1a

Estimation of Technical and Allocative Inefficiency in a Translog Cost System:

Mike Tsionas - Subal Kumbhakar

Structural change has dominated the U.S. electric power sector over the past decade. During this time around one third of the 260 or so private operating companies have been involved in mergers. By analysing the data on operating companies for the period 1994-2001 and focussing on the distribution businesses of those companies we seek to analyse the impact of merger on US utilities. We do this using Data Envelopment Analysis. Our analysis allows us to investigate the effect on efficiency of mergers and to test for the determinants of successful mergers. We examine the extent to which the degree of vertical integration, geographic proximity and the sector experience of merger partners influences merger success. Our paper represents the most comprehensive study of the efficiency impacts of merger activity in the US electricity sector to date.

Paper presented at Ninth European Workshop on Efficiency and Productivity Analysis, EWEPA IX, Brussels, 2005

Estimation and Inference in Stochastic Frontier Models:

Paul W. Wilson - Leopold Simar

Stochastic frontier models introduced by Aigner et al (1977) and Meeusen and van den Broeck (1977) specifiy output, cost, etc. in terms of a response function and a composite error term. The composite error term consists of a two-sided error representing random effects and a one-sided term representing inefficiency. Stochastic frontier models are typically estimated by the maximum likelihood (ML) method. Interest typically lies in (i) inferring returns to scale or other features of the response function; (ii) inferring technical efficiency for each firm represented in a sample; or (iii) inferring average technical efficiency within a sample. With regard to (ii) and (iii), many papers have relied only on point estimates, although interval estimates are possible using a ``conventional'''' approach, where one obtains bounds of confidence intervals as percentiles of the estimated distribution of the one-sided error term, conditional on the composite error. However, the intervals estimated by this approach do not reflect the uncertainty surrounding parameter estimates used to construct the interval estimates. This paper develops a bootstrap method for overcoming this limitation. Although asymptotic properties of ML estimators are well-known, their finite sample properties in the context of stochastic frontier models remain unknown. This paper also examines some estimation issues that arise when the ML method is used in practical applications involving frontier models. Monte Carlo experiments indicate the proposed bootstrap method provides better coverages than the existing, conventional approach.

On Estimating the Mixed Effects Model:

Alois Kneip - Wonho Song - Robin C. Sickles

Abstract unavailable.

ENDOGENEITY PROBLEMS IN THE ESTIMATION OF MULTI-OUTPUT TECHNOLOGIES:

CARLOS ARIAS David Roibás

Empirical representations of multi-output technologies use either quantities or ratios of outputs as explanatory variables. These specifications raise questions about the existence of endogeneity problems in the estimation of these technologies. There are two contributions in the present paper. First, by carefully distinguishing between planned and observed output we clarify the problems of endogeneity in the estimation of multioutput technologies. Precisely, we are able to characterize the unlikely conditions under which distance functions can be estimated by plain least squares. In any other circumstances, instrumental variables are required for consistent estimation. Second, we develop a transformation function which incorporates a global efficiency term and we use Monte Carlo simulation in order to compare the relative performance of both primal approaches to the technology: transformation and distance functions. Particularly, we analyze the role played by different degrees of correlation among output disturbances. We find that under different degrees of correlation between outputs the techniques used to correct endogeneity problems are more effective in the transformation function than in the distance function, so the transformation function dominates the distance function on empirical performance

1b

THE COST OF ALLOCATIVE EFFICIENCY FOR MULTIOUTPUT CARGO:

Ana Rodríguez-Álvarez - Beatriz Tovar de la Fé - Lourdes Trujillo

Abstract unavailable.

EFFICIENCY MEASUREMENT IN NETWORK INDUSTRIES: APPLICATION TO THE SWISS RAILWAY COMPANIES:

Mehdi Farsi - Massimo Filippini - William Greene

In 1996, following an alarming growth in government subsidies for railway transport in Switzerland, the federal government introduced a series of regulatory reforms. In particular, the subsidization of regional railway companies that was previously based on a full deficit coverage, has been replaced by an ex-ante fixed payment system. However, given that these subsidies are determined through a long series of negotiations and bargaining between railway companies and the corresponding cantonal governments, companies might use their local monopoly power to maintain high subsidies. Therefore, it is generally believed that the fixed payment system should be supported by a benchmarking practice to induce companies to minimize their costs. Benchmarking methods are mainly based on models of efficiency measurement. However, the reliability of such models in network industries has been often questioned. Network industries are characterized by a high degree of heterogeneity, much of which is network-specific and unobserved. The unaccounted-for heterogeneity can create bias in the inefficiency estimates. This paper examines the performance of several panel data models to measure cost and scale efficiency in network industries. The stochastic frontier models that include additional firm-specific effects, such as the random-constant frontier model proposed by Greene (2005), can control for unobserved network effects that are random but time-invariant. In cases like railway networks the unobserved heterogeneity is potentially correlated with other exogenous, but observed, factors such as network size and density. In such cases the correlation with explanatory variables may bias the coefficients of the cost function in a random-effects specification. However, these correlations can be integrated into the model using Mundlak’s (1978) formulation. The unobserved network effects and the resulting biases are studied through a comparative study of a series of stochastic frontier models. These models are applied to a panel of 50 railway companies operating over a 13-year period in Switzerland. Different specifications are compared regarding the estimation of both cost frontier coefficients and inefficiency scores.

A Comparison of Methods for Efficiency Measurement for Worldwide Railway System:

Ming-Miin Yu - Erwin T. J. Lin

In the recent years, a number of contributions regarding efficiency measurement for rail transport industry can be found in journals. In those papers, one activity data envelopment analysis (DEA) model is used to analysis relative performance of railway companies; it implies that railway companies engage in their services with identical technology. However, railway companies may engage in their services with non-identical technologies and use the shared inputs. Thus, there are at least two shortcomings in the previous studies. First, since a railway company generally provides passenger and freight services, it may not appropriate if one does not take the feature of multi-activity into account. In other words, because the productive technology for passenger service is different from that for freight service, viewing passenger service and freight service as two outputs and measuring efficiency by using conventional DEA model directly may leads biased results. More specifically, when measuring efficiency, the conventional DEA substitutes input factors and outputs into the DEA models and does not take the consideration of the shared inputs. In fact, for a multi-output firm, there is no doubt that shared resources must be imposed on decision-making units (DMUs). A second shortcoming thus arises with respect to how these resources can be allocated in an equitable or optimal way to the various DMUs. The aim of this paper is to compare technical efficiency measures from the multi-activity DEA and conventional DEA and to assess if these differences lead to different policy conclusions. A number of studies about performance evaluation have been presented recently, both from a practical organizational point of view and from a costs research perspective, to deal with the shared input problem. Among them, the multi-activity DEA model, a novel refinement of the conventional DEA approaches, for the joint determination of efficiencies in the DEA context, was proposed by Beasley (1995) and subsequently revised by Mar Molinero (1996) and Tasi and Mar Molinero (1997, 1998, 2002). Specifically, the multi-activity model is used for evaluating efficiencies of organizations that engage in several activities simultaneously. DMUs in this situation may have some inputs and outputs among all the activities. In this paper we intend to apply the multi-activity model to explore the efficiency of individual services within different multi-service railway companies in the world. There are two reasons for this. First, the multi-activity model was designed, in particular, to estimate the efficiency achieved by organizations that face several production functions using shared inputs. Second, to our knowledge, little studies relating to multi-service railway agencies deal with the shared input problem in a proper way. For example, Oum and Yu (1994) applied DEA method to evaluate the efficiency of 19 OECD countries¡¦ rail companies over the period of 1978 to 1989. Chapin and Schmidt (1999) used the DEA approach to measure the efficiency of US Class I railroad companies since deregulation. Cowie (1999) also applied the DEA method to compare the efficiency of Swiss public and private railways by constructing technical and managerial efficiency frontiers and then measured both efficiencies. We argue that for railway companies, in which different individual service efficiency is unlikely the same. However, both the feature of multi-activity and the fact of shared ratio of the resource have been ignored in these studies. It implies that railway companies engage in their services with identical technology. This may not be realistic in the production of a railway company. In this study we selected 40 railways in the year of 2002 as our samples which is drawn from International Statistics published by the International Union of Railways (UIC), all of them provided both passenger services and freight services. Due to the utilization of different production technologies between these two services, they construct different production functions themselves. Moreover, because of some input resources imposed on the multi-service railway companies such as technician and managerial staff are devoted to both types of services, they need to decide how to allocate across different DMUs for the joint determination of the efficiencies of both services, respectively. Considering those shortcomings mentioned above, the multi-activity DEA model is applied to determine the efficiency of individual services within different multi-service railway companies which engage in their services with non-identical technologies and use shared inputs. The results make clear that of 40 railway companies, only BLS (Switzerland), ZSR (Slovak), UZ (Ukraine) and QR (Australia) can be considered as globally efficient. As regard to average efficiency, it is noteworthy that this differs significantly in the two activities, with that of passenger services revealing a higher average rate of efficiency. This may partly be explained by the fact that both loading and unloading in freight service are much time consumption. The principal conclusion of this paper is that, the two different models considered, conventional DEA and multi-activity DEA, provide different information on relative productive performance of the worldwide railway companies considered in this analysis. The multi-activity DEA provides a more adequate measure for contribution of efficiency than conventional DEA models.

1c

Disentangling Risk and Inefficiency Using State-Contingent Frontiers:

Chris O'Donnell

The state-contingent production model of Chambers and Quiggin (2000) is a flexible and realistic model of producer behaviour under uncertainty. O''Donnell and Griffiths (2004) show how a frontier variant of the model can be used to decompose output shortfalls into inefficiency and risk components. In this paper we use Philippine rice data to compare sampling theory and Bayesian estimators of state-contingent production and cost frontiers.

Cost-based Asset Pricing and the Estimation of Stochastic Technologies:

Robert Chambers

Firms facing stochastic technologies and competitive asset markets will optimally remove any arbitrages between the two. Empirically this results in a moment restriction on the joint distributions of stochastic output, returns, capital stock, investment, and prices. This moment restriction can be used to estimate a cost function representation of the firm''s stochastic technology. The method is applied to aggregate US data on output, returns, capital stock, and investment.

Frontier analysis and non-linear pricing for quality in supply-chain relationships:

Angelo Zago - Robert G. Chambers

The payment systems for raw commodities and intermediate products define one of the most critical relationships of many vertically related industries, since they establish how revenues are distributed among growers and processing firms. Intermediate products payment systems also have a pivotal role in setting the incentives that growers and processing firms face: not only do they heavily influence the incentives to improve technical efficiency, they also have far-reaching implications for investment decisions. For these reasons, measuring and evaluating the right attributes in raw materials, commodities, and intermediate products is a common problem in many sectors of the economy (Barkley and Porter, 1996; Buccola and Iizuka, 1997; Ladd and Martin, 1976). This happens to be true in food industries, such as grapes for wine production, milk for cheese, canes for sugar, beans for coffee, but also in other industries, such as for chips in the computer industry, ores in steel production, steel in construction works, crude oil in refined oil production, etc. In this paper we show how to design an optimal payment system for a group of producers using market demand and production data information. Contrary to the New World countries, the wine industry in Europe is very fragmented and appears relatively uninterested by the consolidation processes that are taking place worldwide, especially in Australia and the USA (Economist, 2003; Marsch, FT 2003). In addition, the wine industry in Europe is made of many small firms, which may lack adequate capital for the required investments in new technologies and marketing policies (Saulpic and Tanguy, 2002). A partial solution to the size problem, according to some practitioners, may be the collective organization by farmers through cooperatives. The cooperative movement in the wine sector, however, has been suffering for a reputation for low quality, lack of investment, and often the inability to retain the better members (Touzard et al., 2000; Zago, 2003). One of the critical problems for cooperatives is the remuneration of members'' raw commodities, e.g., grapes. Indeed, in many instances cooperatives have been plagued by excess supply of grapes of low quality which could only be processed to make relatively low quality and cheap wines (Golan and Shalit, 1993). By producing low quality wines, producers face tougher competition, often leading to losses or level of profits not high enough to remunerate investments. Better members, i.e., members with raw commodities of better quality, thus may find more remunerative market outlets by leaving the cooperative, which remains with the worst (quality) members. By changing remuneration schemes, it may be argued, cooperatives and other producer''s groups may improve the quality of the raw commodities delivered by their members, commanding higher prices for processed commodities and ensuring higher profit levels for members (Jarrige and Touzard, 2001). Starting with the paper by Sexton (1986), it has been recognized that it may be better for the stability of a cooperative to use a non-linear pricing scheme. Starting from the private information regarding different members'' technology, Vercammen et al. (1996) take into account asymmetric information and show that a non-linear price could improve over the standard linear pricing even with asymmetric information. Bourgeon and Chambers (1999) show that when the bargaining power of a group of farmers corresponds to its relative importance in the farm population, the quantities produced are the first-best levels. Departures from equal sharing, i.e., redistribution of surplus, appear when the bargaining power of a group does not match its relative importance in the farm population. Most of the contributions in this topic however consider the quantity choice problem and its optimal remuneration. Few contributions deal with quality remuneration in a cooperative setting. Lopez and Spreen (1987) consider the case of sugarcane cooperatives and compare two payment systems, a traditional and a new one, and show that their method may improve efficiency almost twofold. The sugar cane industry is in fact an instance in which the use of different payment systems is relatively well documented. There are indeed a number and different types of payments which may be separated into three main broad groups: fixed cane price systems, fixed revenue sharing systems, and variable revenue sharing systems (LMC, 2002). In the fixed price system, still present in very large sugar industries such as in China, India, and Pakistan, farmers receive a fixed price per tonne of cane, with no premium or discounts paid for cane quality. Its key weakness is the lack of a link with the actual sugar price and thus it represents '.. a lopsided arrangement through which growers and millers do not share price risk..' (LMC, 2002: 2). Under the fixed revenue sharing systems, revenues are shared on the basis of a fixed percentage distribution between growers and millers. In this system, cane prices and mill margins are linked to sugar prices, but the fixed basis can weaken the incentive to improve technical performance and cane quality for both growers and millers. The variable revenue sharing system is the most sophisticated and is based on a formula ensuring that, beyond a benchmark level of cane quality and factory efficiency, growers are the residual claimants for cane quality improvements and millers cash-in the improvements in sucrose recovery at the factory. The system ensures that, at the margin, increased revenues from improvements in cane quality accrue to the grower, while millers capture any gains from milling efficiency (Larson and Borrell, 2001). Touzard et al. (2001) consider the payment systems of the wine cooperatives in South-France and distinguish them into three main groups. The more traditional systems, still used in one sixth of the surveyed cooperatives, is mainly based on sugar content, offering a linear price for sugar content based on the average price for the wine sold by the coop. According to the authors, this first system is easy to manage but it does not seem to recognize the diversity of grapes delivered by the members and thus renders the cooperative a procurer of undifferentiated raw commodities. A more common method, found in around half of the cooperatives interviewed, is used to remunerate varietal grapes such as Chardonnay, Merlot, etc. when they are particularly appreciated in the market. It uses a modified formula of the above mentioned method and thus it applies a quality concept which is a priory based on technical criteria without much consideration for the market effects. Last, a third set of methods is used in one-third of cooperatives, and it differentiates across different plots according to their contribution to the sales of the cooperative. According to Touzard et al. (2001), this set recognizes the efforts made by the member, but it is more difficult to implement since it requires more information, and it leads to a greater inequality among members. In essence, it creates tensions among members to the extent that it introduces market forces into the cooperative. A different strand of the literature considers how to use the results of contract design under asymmetric information with a richer specification of the production technology. Bogetoft (2000), for example, shows how to use DEA estimates of the technology to design an optimal contract between a Principal and an Agent or a group of Agents. In a related series of papers, he exploits this idea under different information settings, that is with moral hazard and adverse selection, and with single and multiple output specification. In this paper we show how to design an optimal payment system for a group of producers using mainly market demand and production data information. We first show how it is possible to implement the first best through higher prices for better quality commodities deriving the optimal pricing schedule from a dual specification of the problem, i.e., with the profit function, taking into account producers'' heterogeneity. Following Sheriff (2004), we then illustrate how technical efficiency interacts with producers'' ability to produce outputs for a given level of inputs and hence affects profitability. After reformulating the pricing scheme in terms of primal measures, we estimate the technology and the technical efficiency of producers via a stochastic output distance function model. We hence use the estimation results to simulate the pricing scheme. This study combines a theoretical model for contract design under symmetric information for a groups of producers with the contributions of the literature on the parametric estimation of production technology using Stochastic Frontier Analysis (SFA). We combine the contributions of these two strands of the literature to design an optimal non-linear pricing scheme for a cooperative using an estimation of the technology obtained with SFA. We use the pricing scheme with a specific dataset for market, weather, and soil quality conditions to show the impact on the choices and payments received by a group of farmers involved in grapes production in Italy, arguing that the model and the methodology are valid for other producer''s groups as well. The plan of the paper is the following. In the next section we explain the relevance of the problem at hand and review some of the related literature. In the following we represent the choices facing producers and we show how the efficiency parameter allow to distinguish among different producers and their choices. We then introduce the game between the members and the cooperative, showing how to derive the optimal pricing scheme with symmetric information. We introduce a model of the behavior of producers and of the cooperative and show what would be the first best pricing scheme. We then formulate the pricing scheme in terms of a primal specification of the technology, i.e., an output distance function, and of market demand information. After introducing the technology, showing how the efficiency parameter enters the primal representation of the technology which may be useful for the empirical implementation, we show how to implement it using stochastic frontier analysis based on a parametrization of the output distance function. We introduce the data used in the empirical application in the following section, where we use market data to estimate an hedonic price function for grapes to have an estimate of the market willingness to pay for quality. In addition, we introduce the experimental data used in the estimation of production technology. We the proceed with the empirical estimation of the optimal pricing rule found in the theoretical section and expressed in terms of the primal parameters. To conclude, we highlight some possible improvements for the methodology and directions for future research.

Contingent price duality in a multi-output state-contingent production model:

Walter Briec - Laurent Cavaignac

In this paper we use the Arrow and Debreu early proof that economic problems under uncertainty can be analyzed exactly the same way that economic problems under certainty usually are, once the appropriate state-space contingent model has been set. Indeed, each state-contingent good in a given state of nature may be considered as a specific good. The main interest in this approach is that it allows to treat the production of a state-contingent good as it usually is done in a standard and certain multi-output production model. Hence, each state-contingent output may be increased independently from the other. This approach fundamentally differs from the well-known ''parametrized distribution approach'' which indexes the production of a given output in each state of nature on a control variable. Consequently, for this kind of model, the output level in every state of nature depends on the control variable level.

The state contingent approach that we adopt is more flexible but inherits a considerable drawback from its flexibility : a single output production model becomes a multi-output model when presented in terms of state-contingent outputs. Therefore, as we focus on a finite number of outputs M and on a finite number of states of nature S we will deal with SxM state-contingent outputs represented by a stochastic production vector in R^M. Following the early demonstration of Peleg and Yaari (1975), Nau (2001) showed that the marginal rate of substitution under uncertainty can be interpreted as a ''risk-neutral'' probability ratio and plays a similar role as do the price ratio in the standard consumer theory. Paralleling this result, we observe that for the producer, the marginal rate of substitution may differ depending on the state contingent output which are substituted. In particular, it is the traditional price ratio if two different outputs are substituted in the same state of nature.

On the other hand, when the prices are deterministic, the marginal rate of substitution becomes Peleg and Yaari''s (1975) ''risk-neutral'' probability ratio if calculated for the same output in different states of nature. In a state-space model and a consumer context, Chambers, Färe and Quiggin (2004) have used the duality between the benefit function and its conjugate, the expected-value function. They showed that if a probability vector associated with the states of nature belongs to the subdifferential of the benefit function for a given consumption bundle, then this same consumption bundle belongs to the subdifferential of the expected-value function (The use of the subdifferential allows a generalization of the analysis to the case of non differentiable preferences). We transpose this approach to the theory of production by defining the convex conjugate of the directional output distance function : the expected and contingent revenue function. Moreover, we generalize their result by proving that the expected value of the optimal prices is equal to minus the subdifferential of the directional output distance function. Following Chambers, Färe and Quiggin''s path (2004), we link the concept of translation homotheticity to the notion of risky technology. Pursuing this goal, we first define a revenue certainty equivalent and its associated risk premium. But we do not use the cost certainty equivalent revenue which is defined in Quiggin and Chambers (2001). Their (cost) certainty equivalent revenue is an indirect certainty equivalent since it is defined as the maximum achievable certain revenue at a given cost. It is directly derived from the indirect cost function. In this paper, we apply a more straightforward transposition of the standard certainty equivalent (see Nau 2001) to define the certainty revenue equivalent as the maximum achievable certain revenue for a given input vector.

Furthermore, applying Karni''s theory of state-dependent preferences (1985) to our model, we consider that if the revenue is not of the same use in every state of nature, the producer might not want to achieve a certain revenue. On the contrary, she might rather try to obtain a higher revenue in a given state of nature than in another. This approach leads her to maximize her revenue in a given proportion. Therefore, we define a revenue equivalent reflecting this proportion which encompasses the case of the revenue certainty equivalent. According to the work of Chambers and Färe (1998), we then show that the associated generalized risk premium does not depend on the input level if and only if the technology is output translation homothetic. Paralleling the consumer theory, such a technology can be considered as constant absolute risky. Next, we focus our attention on efficiency measures and their connection with the stochastic nature of the technology. We first express the Nerlovian overallefficiency index in terms of the equivalent revenue and the risk premium. Eventually, we determine an index which compares the overall efficiency under a deterministic shadow price vector hypothesis to the overall efficiency obtained with stochastic shadow prices.

1d

Multi-Output Technical Efficiency for Hillside Farm Households in Central America Using Stochastic Distance Functions and DEA Methods:

Daniel Solis - Boris Bravo-Ureta - Ricardo Quiroga

INTRODUCTION Productivity analyses in peasant economies are usually undertaken at the farm-level. However, using the farm as a level of analysis to study productivity in developing countries has been criticized during the last years. Singh et al (1986) explain that under competitive markets for commodities and labor, and perfect substitution between family labor and wage labor, farm decisions are separable from other household decisions. Thus, farm-level productivity analysis is appropriate when the separability assumption is fulfilled. However, López and Valdés (2001) indicate that in developing countries, separability between farm and other household decisions does not always hold. Moreover, Chavas et al (2004) argue that performing efficiency studies at the farm-level in an environment with market imperfections may be inappropriate. These authors contend that farm-level analyses neglect possible labor allocation inefficiency between farm and non-farm activities and that off-farm activities are often joint with farm activities. Therefore, Chavas et al (2004) derived formally, from a household production model, a multi-output multi-input model designed to measure efficiency where household outputs include not only farm products but also off-farm income. Several studies have used the Data Envelopment Analysis (DEA) method (e.g., Chavas et al, 2004; Sharma, Leung and Zaleski, 1999; Shafiq and Rehman, 2000) to empirically measure technical efficiency (TE) in a multi-output setting. More recent parametric frontier papers have relied on multi-output multi-input Stochastic Distance Functions (SDF) (e.g., Brümmer, Glauben and Lu, 2003; Coelli and Fleming, 2001; Coelli and Perelman, 2000). Choosing between the stochastic and deterministic frontier methods has generated considerable debate in the literature (Olsen, 1996). Thiam, Bravo-Ureta and Rivas (2001) indicate that although there is a vast literature on agricultural productivity there is not yet a clear picture of the effect that different methodological assumptions might have on measurements of efficiency. In general, empirical studies have found that stochastic frontier models generate higher mean TE estimates than deterministic models. However, there is evidence indicating that this result is usually found under constant return to scale, but contradictory outcomes are attained under variable returns to scale (Wadud and White, 2000; Sharma, Leung and Zaleski, 1999). Consequently, the main objective of this study is to evaluate and compare the TE of hillside farm-households in Central America using deterministic and stochastic specifications for a multi-output technology. DATA AND EMPIRICAL MODELS The data used in this study comes from surveys administered to representative samples of farmers participating in four natural resource management projects in El Salvador and Honduras. The interviews were conducted between May and August of 2002 and for this study a total of 639 observations are available. The data collected in the surveys include three categories of outputs: staples; cash crops; and off-farm income (wage labor earned in off-farm activities including agricultural work). Inputs are classified into the following categories: total area of cultivated land; purchased inputs; family labor; and hired labor. The effect of soil conservation practices and infrastructures on household productivity is also incorporated using two additional variables. Finally, a set of dummies is incorporated to account for a possible project effect. In addition, six socioeconomic variables are included in the stochastic analysis to explain household inefficiency using the Battese and Coelli (1995) approach. PRELIMINARY RESULTS To compare the empirical performance of stochastic and deterministic multi-output/multi-input methods, eleven alternative production frontier specifications are estimated. Five of these models are computed using the DEA approach and six using the SDF. More specifically, input- and output-oriented DEA models are estimated using Constant (CRS), Non-Increasing (NIRS) and Variable Returns to Scale (VRS). Conversely, the input- and output-oriented SDF models are estimated using the Cobb-Douglas functional form, and half-normal and truncated normal distributions for the one-sided error term. In addition, the literature shows that TE measurements could change if inefficiency analysis is included into the stochastic frontier model (Wadul and White, 2000). Consequently, the SDF is also estimated using the Battese and Coelli (1995) inefficiency model. The preliminary results can be summarized as follows. First, the standard tests for the stochastic frontier model reveal the significant presence of technical inefficiency, indicating that this model is more appropriate than the average production function. In addition, the null hypothesis that the one-sided error displays a half-normal distribution is accepted in all stochastic models. Regarding the explanatory variables used in the analysis, 10 out of 12 coefficients in the output-oriented SDF and nine out 12 coefficients in the input-oriented SDF are significant at least at the 10% level. Moreover, the sum of the input coefficients in the output-oriented SDF is less that one suggesting the presence of decreasing returns to scale. On the other hand, the DEA results show that 13.9% of the farms are fully technically efficient under CRS and 20.5% under VRS. Furthermore, the analysis of the scale efficiency index among the sampled farmers reveals that most of the farms are characterized by decreasing returns to scale corroborating the outcome found in the stochastic analysis. The average TE for the input- and output-oriented SDF is 0.75 and 0.74, respectively. This estimate increases to 0.80 and 0.81 when the inefficiency model is included in the analysis. Conversely, the average TE for the DEA models is 0.59. The lower estimates obtained by the DEA models is consistent with a priori expectations. Finally, the Spearman rank correlation coefficient is used to compare the ranking of the efficiency indexes obtained from the different models. This analysis shows a strong positive correlation between the TE estimates of the different DEA models, with values between 0.82 and 0.99. This relationship declines among the SDF models, where the correlation indexes range between 0.55 and 0.83. The correlation between the SDF and DEA models is also statistically significant and positive, but the average value of the Spearman Rank coefficient decreases to 0.41 for the output-oriented models and to 0.35 for the input-oriented specifications.

U.S. Dairy Farm Cost Efficiency:

Loren Tauer - Ashok Mishra

A stochastic cost efficiency equation was estimated for the U.S. dairy industry using national data from the production year 2000. The cost of producing a unit of milk was estimated into separate frontier and efficiency components, with both components estimated as a function of causation variables. Unfortunately, prices of inputs were not collected and thus it was not possible to estimate a standard cost function where cost is a function of input prices. Rather, cost per hundredweight of milk produced was estimated as a function of farm characteristics and practices, resulting in an estimated average cost curve. The average or unit cost curve for a farm is estimated as a function of a covariate set Xi , an error term vi, and an efficiency term u, (1) ci = f(Xi,¦Â) + vi + u(Zi,¦Ä), u(Zi,¦Ä)¡Ý0, where ci is the cost of production per hundredweight of milk on farm i, Xi are the covariates which impact cost, and the vi error term is independent of Xi, Zi and u. The efficiency term, u, is specified as a function of a set of covariates Zi , which may overlap with the covariate set Xi. The ¦Â vector is the coefficients for the frontier cost curve, while the ¦Ä vector is the coefficients for the efficiency cost curve. The error term, v, is modeled as a normal distribution, iid N(0,sigma2), while the efficiency term, u, is modeled as a truncated positive half-normal distribution specified as N+ (g(Zi), sigma2). This allows the error term for an individual farm observation to be either negative or positive, but the efficiency term u, which will be equal to or greater than zero, will shift with covariates Zi. We elect to shift the mean only since shifting the variance also became intractable. Even then, since g(Zi) is the mean of the underlying distribution before truncation, both the mean and variance of the efficiency u are functions of g(Zi) and sigma2. The frontier and efficiency components of equation (1) were estimated jointly using Maximum Likelihood Estimation. The data had been collected using a stratified random sample with an enhanced sample of larger farms since few large farms would have been surveyed with a random sample, so a weighted maximum likelihood model was employed with the weights applied outside the likelihood function. Variables that might influence the cost of production and cost efficiency of an individual dairy farm were entered as impacting the frontier component as well as the efficiency component of the stochastic curve since a prior both components could be impacted. An example is the years of formal education of the farmer. Greater formal education may allow a farmer to select the lowest cost technology defining the frontier cost function, but that education may also allow that farmer to be efficient in the use of that leading edge technology. The factor that has the greatest impact on the cost curve frontier is the number of hours a day the milking facility is used. Using the milking facility more hours per day decreased frontier costs. However, inefficiency increased with increased hours of milking facility use, such that there was no net reduction in costs. Thus farmers can decrease costs with increased utilization of the milking facility, but only if they are efficient in this strategy. Age increased cost of production since older farmers were less efficient. Parlors used for milking as compared to stanchion milking did not decrease frontier costs, but did decrease costs because of increased efficiency, as did the use of a feed nutritionist. Use of rotational grazing decreased frontier costs but also increased fixed cost inefficiency, with a net reduction in cost of production per cwt. of milk sold.

The Effects of Political Violence on Farm Household Efficiency: the Case of Colombia:

Maria A. Gonzalez - Rigoberto Lopez

Many researchers have focused on the impact of adoption of new technologies as the means to increase productivity. However, in many situations, more efficient use of current technology could increase production without requiring additional inputs or introducing new technologies. As importantly, violence and insecurity in the developing world can hinder efficiency and indeed can result in large efficiency loss. Therefore, it is important to quantify the shortfalls in efficiency and the factors underlying these shortfalls, including institutional factors. Political violence is a fact of life in much of the developing world, especially in many countries in Latin America, Sub-Saharan Africa, the Middle East, and Asia. One of the well-established premises in economics and political science is that political violence leads to slower or negative economic growth. Yet nearly all studies of efficiency of farm households ignore the importance of political violence in their analyses (even in countries like Colombia where political violence is rampant) while studies of the economic effects of political violence are conducted at the national level without a micro insight into the effects on farm households, which are particularly vulnerable to political violence. Colombia provides a useful case study for the analysis of the impact of political violence on farm household efficiency. First, Colombia is one of the most politically violent countries in the world. Second, political violence mostly affects rural areas in Colombia. In fact, 93% of the municipalities affected by guerrilla activities are typically rural, with particularly adverse effects on agricultural activities (Bejarano, 1997). Third, there is a wide variation in the incidence of political violence across rural areas, which permits analyzing the impact of political violence on farm household productivity within the same country (Echandia, 2002). The net effect of political violence on farm household productivity is expected to be negative. This paper estimates farm household levels of technical efficiency in Colombia and also identifies the variables that determine the shortfalls in efficiency with special reference to political violence. It explores why farm households often fail to achieve outcomes that can be described as efficient and measures departures from the efficient frontier using a stochastic multi-output, input-oriented distance function, estimated simultaneously with an inefficient effects model, following the approach of Batesse and Coelli (1993). The main data source consists of 822 farms surveyed in 1999 by the Departamento Nacional de Plantación (DNP) in collaboration with the Instituto Interamericano de Cooperación para la Agricultura (IICA) and the World Bank. This survey provides farm-level data on household characteristics and agricultural inputs and outputs. Data from this survey is matched with regional violence indicators, including the number of displaced people provided by the Colombian government in the Sistema Unico de Registro (SUR, unique registration system) and other indicators such as the number of assassinations, kidnappings, and guerrilla attacks, obtained from the University of Los Andes in Santa Fe de Bogotá, Colombia. Note that previous studies on political violence have relied heavily on assassinations. Empirical results indicate that the average level of technical efficiency of farm households in Colombia is approximately 87%. Thus, the results indicate that it is possible for the households in the sample to improve their performance by using the best practice technology and overcoming constraints that might be imposed by factors such as violence. The empirical findings also show that violence has a very influential effect on farm household productivity performance. Simulation results show that if violence is eliminated, average technical efficiency could increase by 6.4 percent, with a particularly strong positive effect on small size farmers.

Structural change in transition: A role for organizational legitimacy? Evidence from Czech Agriculture:

Jarmila Curtiss - Tomas Medonos - Vladislav Valentinov - et al.

The processes of market liberalization and restoration of property rights in post-communist countries of Central and Eastern Europe were expected to result in economic incentives that would stimulate economic restructuring and improved economic performance. Nonetheless, empirical efficiency analyses have shown that ten or more years after the economic system change, a substantial proportion of firms continue to show inefficiencies and weak economic performance. In studies of structural change, this observation has often been ascribed to the delirious effect of an imperfect and incomplete institutional environment as it has implied significant heterogeneity across the economic conditions faced by firms. However, little consideration has received an alternative view on the organizational effect of institutional environment, the neo-institutional approach to organizational theory. It suggests that institutions can influence the firms’ choice of organization form. Firm’s may choose a form that allows efficient management and production of private goods or one which is inefficient in this respect if access to external resources can be secured through social or public approval and support associated with an image of organizational legitimacy. We consider this approach as particularly relevant in a transition setting where many non-economic values and norms are an element of the firms’ environment. We hypothesize that organizational legitimacy is associated with economic gains that provide incentives for firms that find such investment a cheaper path to achieve sustainable economic performance than the path of managing for efficient use of internal resources. In this paper, we consider evidence of legitimacy-seeking behavior, a measure of economic performance, and their relevance to structural change in transition. Furthermore, we analyze characteristics of firms that pursue legitimacy-seeking behavior. We focus on the case of Czech agriculture. Extensive survey and accountancy data for 167 agricultural enterprises for the years 2001 to 2003 are used for the empirical study. As a measure of economic performance we consider revenue efficiency. Our approach is to derive from these data estimates of efficiency based on a parametric stochastic revenue frontier analysis. Consistent with the hypothesis that economic performance is conditioned by organizational legitimacy, we examine using a one-stage approach the relationship between legitimacy indicators and economic performance. Organizational legitimacy can be established through a number of activities, e.g. political and regional activities, provision of public goods, satisfaction of regulations or other activities improving business’s reputation. In the absence of exact measures of organizational legitimacy, to identify indicators we examine co-variation using non-parametric correlation measures and factor analysis on a number of proxies for effort for legitimacy. Four uncorrelated variables are chosen to represent efforts for legitimacy and are used in our examination of the relationship between legitimacy and revenue efficiency. Indicators selected include farm provision of social and cultural activities, the extent of participation in activities organized by associations, payment of higher wages to their employees, and preliminary fulfillment of EU production technical standards. Identified factors associated with legitimacy effort are incorporated in the stochastic frontier revenue model. We specify a model where the first and second moments of the distribution of revenue inefficiency depend on organizational legitimacy effort indicators. This approach allows consideration of heteroscedasticity of the revenue inefficiency error, a condition that is intuitive given that large farms are expected to have greater capacity for providing public goods and other activities as a means of achieving legitimacy. The stochastic frontier analysis results indicate that each of the legitimacy variables except the investment into the EU technological standards have a significant positive effect on revenue efficiency. This indicates that costs of gaining organizational legitimacy through provision of social and cultural activities, and participation in activities organized by associations are exceeded by the resulting economic gains. To bring further evidence to bear on these relationships, we further consider the relationship between legitimacy effort and access to resources and managerial practices using non-parametric Kendall’s tau correlations. These correlations indicate two groups of enterprises. First, we find evidence that farms pursuing social and cultural activities are larger scale enterprises with a significantly higher share of land as owned and with access to higher quality leased land. Second, we find that farms that did not invest in social and cultural activities but invest in EU technological standards are farms that have higher quality worker, better managerial practices, and greater investment in equipment and buildings. We can conclude that evidence that effort to establish organizational legitimacy is associated with significant economic gains, supporting our original hypothesis. We find that gains from legitimacy appear to support larger farms during transition to successfully compete with farms employing efficient managerial practices with greater investment. These findings offer a new perspective on structural change in transition in the Czech agriculture and other similar settings.

On the generation of a regular multiple-inputs multiple-outputs technology using parametric output distance functions:

Sergio Perelman - Daniel Santín

Monte-Carlo analysis is a statistical computational methodology frequently used in efficiency and productivity research for testing different properties of analytical tools. A relatively large number of papers have applied Monte-Carlo experiments in both parametric and non-parametric techniques as well as for comparisons between them. Most of problems in economics have to deal with a multiple-inputs multiple-outputs process. This is especially observed in the public sector where policy makers usually face up to different outputs. However an important drawback in Monte-Carlo experimental design is how to generate a multi-input multi-output production function with noise and inefficiency able to globally fulfill all properties required in the economic theory. In fact, this difficulty motivates that most of Monte-Carlo experiments works only use one output production functions. In this paper our purpose is to show that the theoretical properties of parametric output distance functions provide a useful tool to perform such data generation in order to test efficiency methodologies under different hypothesis. Together with the general requirements for a well-behaved multiple-inputs multiple-outputs production function, we provide a number of sufficient conditions and straight rules to generate data in the simplest case: a two-inputs two outputs production function. In addition to this, the procedure presented in this work can also be used like a rule of thumb for testing regularities in empirically estimated output distance functions. Main conclusion of this research is that the theory of parametric output distance functions can be applied to easily generate a multi-input multi-output production function. Furthermore this result allows enhancing Monte-Carlo experiments conclusions for the empirical work.

Semiparametric Stochastic Frontier Models with Varying Coefficients:A Local Likelihood Approach:

kien tran - mike tsionas - giannis karagiannis

In this paper, we propose a semiparametric stochastic frontier models with varying coefficients as a useful yet flexible specification for studying the measurement of efficiency. Specifically, we consider the following model: (1) where is a vector of unspecified smooth function of . The paper proposes a local maximum likelihood method with a kernel weight function to estimate the varying coefficient function by locally approximating the unknown function in (1) by a polynomial function . The consistency of the estimator and its asymptotic normality are established. A comparison of our proposed method with the Bayesian approach of Tsionas (J. Applied Econometrics, 2002) and parametric approach of Greene (J. Productivity Analysis, 2005) is made in term of real data.

Estimation of Technical and Allocative Inefficiency using a System-Wide Local Maximum Likelihood Approach

Efthymios G. Tsionas - Kien C. Tran - Giannis Karagiannis

Due to the so-called Greene problem, the simultaneous estimation of technical and allocative efficiency is perhaps the most challenging issue in the stochastic frontier literature. In order to obtain separate estimates of technical and allocative efficiency it is necessary to consider a system-of-equations, where the internal consistency of the specification of allocative errors in the objective function (e.g., cost function) and the derivative equations (e.g., share or input demand functions) is the most difficult task. This becomes really complicated when a flexible function form (e.g., translog) is chosen for the objective function, see Kumbhakar (1997) and Kumbhakar and Tsionas (2004a,b) for empirical implementation and detailed discussion of the econometric difficulties and issues involved. Nevertheless, estimation is not straightforward even for simple functional forms, such as the Cobb-Douglas, which by definition is self-dual. Notice that Schmidt and Lovell (1979) have found the Cobb-Douglas dual (cost) system to have a rather complicated error structure and for this reason they have recommended estimating the primal system (i.e., the production function along with the first-order conditions for cost minimization) to measure technical and allocative efficiency. Up to now this has been taken for granted and there is almost no empirical study using the dual (cost) Cobb-Douglas system to estimate technical and allocative efficiency. In an attempt to challenge this assertion, we employ a shadow cost function formulation, similar to that of Kumbhakar (1997), and Balk’s (1997) normalization to show that the error structure of the Cobb-Douglas cost system is manageable and in fact the corresponding likelihood function is available in a closed form. Then we use the local MLE approach to convert the restrictive Cobb-Douglas cost system to a fully non-parametric system, which in principle is an approximation of any cost function system. In that respect, we extent the single-equation local MLE stochastic frontier of Kumbhakar, Park, Simar and Tsionas (2004) into a system-of-equations framework, and we establish the consistency of the estimator and its asymptotic normality. Whereas Kumbhakar, Park, Simar and Tsionas (2004) deliver non-parametric estimates of firm-specific technology and technical inefficiency, in addition we provide non-parametric estimates of allocative inefficiency. For comparison purposes, the proposed estimation procedure is applied to both the primal and the dual Cobb-Douglas system.

A Robust Estimator for Performance Modeling:

Guan Zhengfei - Alfons Oude Lansink

In recent economics literature, Malmquist productivity growth index (Fare et al.) is often used as a measure of total factor productivity. In finding the factors affecting the productivity growth, a two-step approach is often employed where the Malmquist index is computed in the first step and regressed on a set of explanatory variables in the second step. The two-step approach is also frequently used to study factors affecting technical efficiencies. Recent literature in efficiency and productivity studies criticized this two-step approach, arguing that bootstrapping should be used to correct the bias and solve the serial correlation problem which otherwise jeopardizes the validity of econometric procedure. Unfortunately, the bootstrapping method is computationally demanding and may be less appealing to applied researchers. In finding factors affecting firm performance, in many cases variables used in the regression model are likely to be endogenous. For example, investment may be affected by firm performance. Without addressing the endogeneity, analysis may result in inconsistent results. As an additional problem in modeling firm performance, we find that static models are routinely used both in efficiency and productivity studies and in financial studies. However, we notice that firm performance has its specific dynamics. Models not addressing these dynamics are potentially subject to misspecification problem and likely provide biased estimates. The objective of this study is to: 1) develop a robust estimation procedure with GMM which solves the problems in two-step studies, and propose an alternative to the often recommended bootstrapping; 2) address the endogeneity problem in performance modeling; 3) study the dynamics in performance. Empirical application addresses farm businesses in the Netherlands. Results show that the GMM estimator proposed in this study successfully solved the serial correlation problem in two-step studies and provides an alternative to the bootstrap method. The GMM approach addressed the endogeneity problem as well. We conclude that models not addressing the performance dynamics are biased.

Efficiency benchmarking of Railway Infrastructure: a DEA approach:

Bart Roets - Johan Christiaens

This paper presents a benchmarking model, based on DEA, measuring the cost efficiency of railway infrastructure. In this model, determinants of efficiency such as asset age, network complexity, traffic density, etc. are also examined. Nowadays, European railway companies are in the middle of a profound change process. Since 1991, European directives guide the whole railway industry through an important organisational and regulatory transformation. The holistic “Railway System” is vertically separated into national “Infrastructure Managers” (public companies providing railway infrastructure) and several competing “Railway Undertakings” (public or private companies, delivering train services). Infrastructure Managers are not providing infrastructure as such, but rather the infrastructure capacity enabling the train circulations. The infrastructure capacity is an essential constituent of the railway system supply chain. Serving as an input for the Railway Undertakings, it is the principal output for the Infrastructure Managers. As a result of the aforementioned vertical separation, research on productivity and efficiency measurement of railway infrastructure is gradually emerging. Previous research was mainly based on parametric techniques, ranging from a translog cost function to a stochastic input distance function. Benchmark studies on railway infrastructure efficiency remain relatively scarce. No research paper conducting a non-parametric DEA analysis could be identified. Furthermore, previous academic contributions focussed entirely on infrastructure maintenance, thereby overlooking the expensive traffic control operations. And finally, although the Infrastructure Manager provides infrastructure capacity to the Railway Undertakings, previous research still presented the “traffic volume”, a long-established output of the Railway System, as an Infrastructure Manager output. The present paper explores and addresses this research problem, through the construction of a DEA model examining the cost efficiency of infrastructure maintenance and traffic control. Using archival data from accounting and operations databases, the model performs an internal benchmark on the 22 Decision Making Units of Infrabel, the Belgian Infrastructure Manager. Aiming at the development of this cost efficiency benchmarking, the Infrastructure Managers’ multiple outputs are modeled through variables capturing the “capacity”, “availability”, “geometric quality” and “safety” of the infrastructure. Unlike previous research, “traffic volume” is not included in the output specification. Posteriorly to the benchmark, a second stage Tobit analysis examines the regression of environmental variables on the obtained cost efficiency scores. Expected results are based on hypotheses concerning the asset age (average and variance), traffic control technology, network complexity, traffic tonnage and traffic density.

Economies of scope in European railways - an efficiency analysis:

Heike Wetzel - Christian Growitsch

This paper conducts a cross-country analysis to investigate the performance of European railways with particular focus on economies of scope. We test the hypothesis that so called integrated railways – companies owning a network and providing transport services – realise vertical and/or horizontal economies of scope. Our analysis adopts a two step approach. In a first step we analyze the existence of scope economies by estimating technical efficiency of integrated and non-integrated railways using data envelopment analysis (DEA). Following the model introduced by Morita (2002) we compare the efficient frontiers of joint and separate production by constructing a virtual efficient frontier of joint production by separated companies. To determine whether joint or separate production is more efficient we apply a DEA super-efficiency model which relates the efficiency for each type of production to a reference set consisting of the alternative production technology. In a second step, we evaluate the influence of various country-specific and environmental factors of European railway efficiency by using the Tobit regression model approach.

Market Size and Contract Design:What Determines Efficiency and Productivity in Urban Transit Systems?:

Alvaro Costa - ANA FARIA - Ana Faria - et al.

A recent key topic in the study of urban transit systems relates the determinants of their performance. Among the major factors that may explain differences across systems are market organization, contract design, degree and nature of regulatory regime, as well as the characteristics of the network being served (Borger, Kerstens and Costa, 2002). This paper contributes to this literature by providing evidence on the urban public transport sector by using a panel of Portuguese major operators over the period 1996 to 2002. In particular we investigate whether market size and contract design are relevant in explaining firms’ performance. Moreover, our data allows us to discriminate between two performance indicators, namely efficiency and effectiveness. In order to evaluate firms’ performance we construct best practice production frontier and Malmquist indexes over a seven year period using data envelopment analysis. As such, we are able to investigate what have been the main drivers of efficiency and productivity growth in Portuguese urban transit systems.

Technical efficiency between 14 Western European Railways during the period 1977-1999:

CESAR RIVERA-TRUJILLO

The purpose of this paper is to estimate and compare the technical efficiency between 14 Western European Railways during the period 1977-1999. Moreover, this paper analyses the impact of key rail reforms of the 1990s on technical efficiency. It uses a Stochastic Frontier Analysis as this approach has the advantage over the deterministic methods (e.g. Data Envelopment Analysis DEA, Corrected Ordinary Least Squares COLS) of separating inefficiency from the statistical noise. An input distance function approach is used in order to take into account the multi-output characteristic of the rail industry and because this approach does not rely on price or financial data, therefore, it is possible to use the most recent data available. We have given especial attention to the period in which recent rail reforms were applied to determinate their degree of success. Specifically, we investigate the impact of two key rail reforms: separation of infrastructure from operations and third party access. The results suggest that, in average, the separation of infrastructure from operations has affected negatively the technical efficiency but that the introduction of third party access in the provision of transport services has affected the technical performance positively. Moreover, the results showed that the Netherlands, Sweden and Spain are among the most technically efficient railways while Portugal and Greece showed the worst performance.

2c

On the Minimum Extrapolation Principle of Data Envelopment Analysis

Timo Kuosmanen

The minimum extrapolation principle (MEP) is a cornerstone of Data Envelopment Analysis (DEA). This principle defines the DEA production possibility set as “the ''smallest'' set that consistent with the observed data and the postulated properties for the production possibility set” (Banker, Charnes, and Cooper, 1984, Management Science 30(9), p. 1082). Bogetoft (1996, Management Science 42(3)) has earlier noted that the MEP is not applicable with any set of assumptions. This paper presents another paradox: we show that under certain widely used postulates (convexity, free disposability of inputs and good outputs, weak disposability of bad outputs) the DEA production possibility set excludes input-output vectors that are technically feasible by those same postulates. We argue that resolving this type of paradoxes calls for a reformulation of the MEP. We propose a revised MEP that avoids the known paradoxes but preserves the spirit of the original MEP .

Linearly interpolated FDH efficiency score for nonconvex frontiers

Leopold Simar - Seok-Oh Jeong

This paper address the problem of estimating the monotone boundary of a nonconvex set in a full nonparametric and multivariate setup. This is particularly useful in the context of productivity analysis where the efficient frontier is the locus of optimal production scenarios. Then efficiency scores are defined by the distance of a firm from this efficient boundary. In this setup, the Free Disposal Hull (FDH) estimator has been extensively used due to its flexibilty and because it allows nonconvex attainable production sets. However the nonsmoothness and discontinuities of the FDH is a drawback for conducting inference in finite samples. In particular, it is shown that the bootstrap of the FDH has poor performances and so is not useful in practice. Our estimator, the LFDH, is a linearized version of the FDH, obtained by linear interpolation of appropriate FDH-efficient vertices. It offers a continuous, smooth version of the FDH. We provide an algorithm for computing the estimator, and we establish its asymptotic properties. We also provide an easy way to approximate its asymptotic sampling distribution. The latter could offer bias-corrected estimator and confidence intervals of the efficiency scores. In a Monte-Carlo study, we show that these approximations works well even in moderate sample sizes and that our LFDH estimator outperforms, both in bias and in MSE the original FDH estimator.

AXIOMATIC FOUNDATIONS OF EFFICIENCY MEASUREMENT ON FREE-DISPOSAL-HULL TECHNOLOGIES:

William Schworm - Robert Russell

Bol (1986 JET) showed that there does not exist an efficiency index satisfying a set of reasonable axioms proposed by Färe and Lovell (1978 JET) for all (closed) technologies. Dmiturk and Koshevoy (1990 JET) provided a complete characterization of the set of technologies for which there exists an index satisfying the Färe-Lovell axioms. The empirical technologies intrinsic to DEA methods belong to this class. As the Dmitruk-Koshevoy sufficiency proof is constructive, it provides a class of indexes that satisfy the axioms. In our paper presented at the NAPW04 Conference in Toronto, however, we show that the Dmitruk-Koshevoy class of indexes fails to satisfy the more-fundamental axiom of commensurability (independence of units of measurement). We formulated an alternative index that satisfies commensurability as well as the Färe-Lovell axioms on convex polyhedral technologies. In this sequel, we extend those results to the class of free-disposal-hull technologies. In addition, we hope to have results on the empirical implementation of these indexes in time for presentation at EWPE05.

A Shadow Cost Function Model of the Dairy Sector in the United States:

Roberto Mosheim - Knox Lovell

In this study, a new data set based on the 2000 Agricultural Resource Management Survey, the most recent national survey of milk producers, is used to measure and analyze the determinants of technical and allocative efficiency and scale economies for the dairy sector in the United States. A shadow cost function is employed to decompose and analyze economic efficiency. In order to ensure theoretical consistency of the estimated function and the reliance of results, the properties of the shadow cost function are tested. This paper makes a contribution from the standpoint that it applies a sophisticated econometric methodology to a very rich dataset. The study will offer new insights into the dairy sector’s performance , given for example, differences in size, feeding technology and management practices.

A Directional Distance Function Approach to Dynamic Efficiency: An Application to the Dutch Glasshouse Horticulture:

Elvira Silva - Alfons Oude Lansink

Dutch horticulture holds a strong position on world markets of several horticultural products. The contribution of horticultural industry to Dutch agriculture is becoming increasingly important in terms of production value and employment. Meanwhile, the intensive use of some inputs, in particular energy, pesticides and fertilizer, is also posing environmental problems. To design an effective policy it is of interest for policy makers to know the potential of efficiency improvement and identify the factors that cause the inefficiency of individual inputs in Dutch horticultural production. Conventional static measures do not account for dynamic processes at firms such as investments in new capital goods. Input-based dynamic efficiency measures are derived from an adjustment-cost directional distance function approach. A dynamic input directional distance function can be generated from an adjustment-cost technology where the dynamics are explicitly incorporated in the form of the properties of the input requirement sets with respect to the quasi-fixed factors. The properties of the dynamic input directional distance function are inherited from the properties of the technology. The dynamic efficiency analysis is carried out for Dutch vegetable firms over 1991-1995. 345 observations from 123 firms are available in the data set. One output and six inputs are included. The output is aggregated in terms of deflated revenues. The six inputs are energy, materials, service, glasshouse, installations and labour.

Quality production and quality indicators in intermediate products

Angelo Zago

Measuring and evaluating the right attributes of raw materials, commodities, and intermediate products is a common problem in many sectors of the economy. In food industries, for instance, it is well known that the necessary condition for the making of a good wine is the availability of grapes with the right attributes. The same argument can be put forth for the characteristics of milk for cheese production, of fruits for juices, of beets or canes for sugar, of beans for coffee, and many others. In addition, this problem is of interest also in other industries: for example, the quality of chips is important for the computer industry, like that of ores for steel production, of steel for construction works, and of crude oil for refined oil, just to name a few. The problem of taking into account the quality attributes of different goods has a long tradition in economics, and the most well established efforts in this direction are probably those of the hedonic pricing literature in the context of the Consumer Price Index statistics. The question in this case is how to adjust consumer (or industry) prices for increases in the quality of goods, such as computers, cars, and other durable goods (Triplett, 1990). The hedonic pricing literature uses regression techniques to relate the (market) prices of different 'models' or versions of a commodity to differences in their characteristics or 'qualities'. The earliest references of this technique come from agricultural economics, with the early work of Waugh on vegetable prices and Vail on fertilizers (Griliches, 1990). One of the first attempts to incorporate quality attributes in a model of producer behaviour, however, is a paper that views process and quality change as outcomes of a firm''s optimization problem (Fixler and Zieschang, 1992). This contribution shows how a market-determined price-characteristics locus can be used to adjust the Tornquist output- and input-oriented multifactor/multiple output productivity indexes of Caves, Christensen and Diewert (CCD) (1982) for changes in input, output and process characteristics. Using radial distance functions, it shows how the quality adjusted indexes proposed are the product of two indexes, a quality index and a CCD-type Tornqvist productivity index. Extending the work on productivity of CCD, Färe et al. (1992) define an input-oriented Malmquist productivity change index as the geometric mean of two Malmquist indexes as defined by CCD, and develop a nonparametric activity analysis model to compute productivity using linear programming. In a subsequent paper, Färe, Grosskopf and Roos (1995) extends this productivity index by incorporating attributes into the technology. Studying a panel of Swedish pharmacies, they use the attributes together with ratios of distance functions to measure the service quality of each pharmacy. By further imposing a multiplicative separability assumption on the distance functions, they are able to decompose the Malmquist productivity change index into three components, namely quality change, technical change and efficiency change. Another application of the same idea, i.e., of decomposing economic indexes into various components, is the paper by Jaenicke and Lengnick (1999). Merging the soil science literature on soil-quality indexes with the literature on efficiency and total factor productivity indexes, they isolate a theoretically preferred soil-quality index. In addition, using common regression techniques they shed light on the role of individual soil quality properties in a linear approximation of the estimated soil-quality index. A different but somewhat related strand of the literature deals with the environmental impacts in the measurement of efficiency and productivity growth. Färe et al. (1989) indeed started what has become now a relatively vast literature extending efficiency measurement when some outputs are undesirable. The central notion of this paper, and of many that followed (for a recent application and partial survey see Ball et al., 2001), is that of weak disposability of outputs. To credit firms or industries for their effort to cut off on pollutants, technology is modeled so that it can handle the case when the reduction of some (bad) outputs requires the reduction of some of the other outputs and/or the increase of inputs. Besides the concept of output weak disposability, an interesting and useful idea for our setting is the directional distance function, a generalization of the radial distance function introduced to production economics by Chambers, Chung and Färe (1996) who extended and adapted the idea of the translation functions of Kolm (1976) and Blackorby and Donaldson (1980), and of the benefit function introduced in consumer theory by Luenberger (1992, 1994). The directional distance function allows to compare different firms and to measure their distance from the frontier of the technology moving along a preassigned direction. In this fashion it is possible to evaluate the performance of the firms that need to increase the production of the good outputs and decrease that of bad outputs (Chung, Färe and Grosskopf, 1999). The first attempt to use the directional distance function to take into account the quality of outputs in a different context, i.e., health services, is a paper by Dismuke and Sena (2001). They consider the mortality rate as a (bad) quality attribute of the hospital production process and use directional distance functions to calculate a Luenberger-Malmquist productivity index. They are then able to decompose the productivity index into a quality index, plus a technical change and efficiency change components. In this study we propose a methodology to measure the characteristics and composition of intermediate products, i.e., grapes for wine production, and we pursue three objectives. First, we characterize the technology by investigating the relationships among the different quality attributes and the production level. This objective is pursued with a systematic investigation of the disposability properties of the technology, which allows to show that some quality attributes are substitute, while others are complement in production. We also find that many of the disposability properties are not stable across years, presumably because of different weather conditions, and between crop varieties. Second, with the methodological contribution, we address the issue of how to measure quality attributes for intermediate goods using a general representation of the technology. Although there are other instances of this problem in the literature, especially in that dealing with hedonic prices, to the best of our knowledge there are no contributions that address explicitly this topic on the production side. In this paper we model the quality attributes with a multioutput technology, using a general representation of technology based on directional distance functions. These are a generalization of the radial distance functions which, since Shephard''s contributions, have been used to give a single-valued representation of production relations in case of multiple inputs and multiple outputs (Chambers, Chung and Färe, 1996, 1998). Directional distance functions indeed can be seen as an alternative and more general way to represent technology and to compare and measure input, output and productivity aggregates (Chambers, 2002). The quality aggregate measures we propose using directional distance functions may be used to evaluate firms'' output taking into account the whole set of quality attributes. These alternative measures thus can be compared with the standard practice in the industry of using only one attribute, for instance sugar content used to measure the quality of grapes for wine production. In this paper we use the idea of the directional distance function to incorporate quality attributes into the technology, but we depart from the models reviewed above in the construction of an indicator instead of an index. In fact, following Chambers (1998 and 2002), we use the directional distance function to construct an indicator, that is an output aggregator that is expressed in difference forms rather than in ratio forms like in the case of the more traditional Malmquist productivity index. This difference stems from the property of the directional distance functions, which make the Luenberger indicator translation invariant in outputs, to contrast with the property of homogeneity of degree zero in outputs of the Malmquist index coming from the linear homogeneity of the output distance function à la Shephard (1970). We propose an indicator based on directional distance functions for different reasons. First, as explained above, we compare firms based on the distance from the frontier along a preassigned direction which reflects the preference and needs of the buyer or downstream firm with respect to the quality attributes. Second, it may be the case that to be valuable to a downstream firm, the composition of the raw material has to be close to an 'ideal' bundle of attributes preferred by the buyer. In other words, in some instances the composition has to be well balanced and some of the attributes have to be within a certain range. The choice of the direction allows then to take this into account and evaluate the quality attributes produced by a pool of suppliers according to buyers'' needs. The third objective of the paper, more policy-oriented, is to evaluate how quality attributes interact with the quantity level in the production of these intermediate products. The reason for this interest is that in many agricultural markets and food industries, especially in Europe, producer groups are granted the authority to self-regulate the production and trade of many commodities. While in the US the often enforced policy for quality regulation is the use of minimum quality standards, in the European Union a more common policy device is the imposition of ceiling on yields per unit of land. This regulation is common and allowed, for instance, for those producer groups that regulate production and trade of wine with appellation contrôllée; for those that regulate typical products; and for those operating in fruit and vegetables industries. Advocates of this regulation claim that by reducing the production level one can in fact increase quality, and thus quantity ceilings would benefit consumers and producers alike. In other words, output control measures would be justified because they increase economic welfare, and should not be criticized and prosecuted by antitrust authorities (Canali and Boccaletti, 1998). The fact is that the economic analysis on this topic is relatively scarce, one notably exception being a paper by Arnaud, Giraud-Heraud and Mathurin (1999). In a model with vertical differentiation, they are able to show that in some instances output control by a producer group can indeed increase total economic welfare. However, the results of the paper impinge on the assumption of the substitutability between quality and quantity or, put in another way, quality and quantity substitutability should be a necessary condition for the regulation to be welfare-increasing. Although this assumption on the technological relationship may appear reasonable to the reader and to many practitioners, no empirical work has established its nature. In the paper we find evidence of a significant non-linear trade-off between quantity and aggregate quality for the years considered and for both varieties investigated. Moreover, for sugar and total acidity, two major quality components of grapes, for most of the years considered the trade-off with yields occurs at lower production levels in Chardonnay than Merlot. The next section reviews the literature that addresses the issue of how to take into account quality in the production process. Then we introduce the notation, the model and the empirical implementation algorithms we use in the study. In section five we illustrate the data we use, based on production practices and output results of two relatively well known grape varieties, Chardonnay and Merlot. Section six presents and discusses the results. Section seven concludes the paper with the suggestions for further research work.

Do the ban on use of anti-microbial growth promoter impact on technical change and the efficiency of slaughter-pig production:

Lartey Lawson - Lars Otto - Peter Vig Jensen - Mogens Lund.

This study aims at investigating the effects of the ban on the use of anti-microbial growth promoters in the production of “Finishing Pigs” for slaughter. We investigate if the ban on the use of anti-microbial growth promoters has for specialised pig-producers altered the productivity of inputs, technical change and the efficiency of production. This paper complements an earlier paper that investigated the impact of the ban on weaned-pig production. Background: The study is motivated by the fact that antimicrobial growth promoters have been known world wide to protect livestock from bacteria infections, and in effect stimulated the utilization of feedstuff and reduced the mortality rate. However, fears for increasing bacteria resistance with subsequent health hazards for humans and livestock has lead to societal debates about the pros and cons of its use in livestock production. Antibiotic–resistance infectious diseases increase health care costs associated with treatment expenses and possible multi-drug resistance infections. The Danish government and the pig industry in recognition of the negative effects associated with reduced efficacy of antibiotic drugs in the future started a gradual determination of the use of growth promoters in 1995, which resulted in a total ban in 2000. EU as a follow-up to the Danish action also initiated a ban on the use of growth promoters that are known to contribute to human health hazard. In the US the discussion of a ban is an ongoing subject of debate. For US producers, a study by Hayes et al. (2001), suggested that a ban would increase production costs and subsequently lead to an increase in the retail price of pork. In another US study, Wade and Barkley (1992) suggested that with a ban both producers and consumers surplus will increase. However, these studies are based on estimated anticipated results, which may not have very strong desire-to-change impact on the perception and the attitude of producers who are the users of growth promoters. Contributions of the paper: The paper is of an interest because it provides specific knowledge of the impacts of the ban. Therefore from a policy point of view, the paper will provide information for governments and industries contemplating a ban to identify the kind of incentives necessary to ensure a smooth implementation of a ban. For the producers in countries where the ban is still a question of debates the results will provide useful information about the response reaction of Danish slaughter-pig producers. Methods and data: We use the parametric stochastic frontier analysis approach and estimate a production function from an unbalanced panel-data. The data is the national stratified ran-dom sampled of about 75 farms, raising finishing pigs for slaughter, and collected annually for the period 1991 to 2003 by the Danish Agricultural Account Statistic. This sample data allows comparing the input productivities of output for before and after the implementation of the ban. Technical change defined with a time trend and the general index will also be compared. Similarly the over all yearly estimated technical efficiency would be compared.

The role of the policy framework for the effectiveness of benchmarking in regulatory proceedings:

Phil Burns

No abstract.

RESTRUCTURING AND EFFICIENCY IN THE U.S. ELECTRIC POWER SECTOR:

Michael Pollitt - John Kwoka

Structural change has dominated the U.S. electric power sector over the past decade. During this time around one third of the 260 or so private operating companies have been involved in mergers. By analysing the data on operating companies for the period 1994-2001 and focussing on the distribution businesses of those companies we seek to analyse the impact of merger on US utilities. We do this using Data Envelopment Analysis. Our analysis allows us to investigate the effect on efficiency of mergers and to test for the determinants of successful mergers. We examine the extent to which the degree of vertical integration, geographic proximity and the sector experience of merger partners influences merger success. Our paper represents the most comprehensive study of the efficiency impacts of merger activity in the US electricity sector to date.

Model Specification Games in DEA Yardstick Regimes:

Per AGRELL - Peter Bogetoft - Mathias Lorenz

Asymmetric information concerning the underlying technology for a regulated industry is a major obstacle for the development and implemen- tation of yardstick regimes based on production frontiers. Under collective regimes, such as industrywide price-caps, regulated firms have strong incen- tives to collude to withhold information about the true frontier. The regula- tor, fearing stalled investments and collapsing quality standards, implements a minimal extrapolation policy with low rent extraction. The informational rent that results from a too lax regime is then silently split among efficient and inefficient agents in mergers. We present a dynamic framework for cooperative model development and implementation under which the efficient agents con- tribute to model refinement in exchange for a fair reallocation of industry rents. The strategy is derived from a two-period game where the regulator initially launches a conservative model and commits not to lower overall expenditure in the second period. The firms are invited to refine the model for the second period, in which case they will be subject to a DEA yardstick regime. The behavioral di¤erences between efficient and inefficient agents incite the former to reveal the technological information. The resulting second period frontier will be the optimal and the rents for the efficient agents will be the result of industry redistribution.

3b

Absorption of innovations and the relationship between innovations and productivity in Dutch Hospital Industry:

Jos Blank - Bart Van Hulst

The hospital industry is currently the subject of a great public interest because expenditures on hospital care are large, have increased significantly over the past 20 years, and are expected to continue to increase in the future because of the ageing of the population. New technologies, medical as well as organizational, may push up cost further or may downsize cost depending on the nature of the innovation. The main policy question is how we can sustain adequate hospital care at affordable prices in the future. Insights in the dispersion and the consequences of innovations are therefore necessary. In this paper we discuss three important aspects of innovations. : · How do we measure innovations? · Which determinants do affect the dispersion of innovations? · How do innovations affect cost and productivity? We measure innovation status by the number of innovations and an alternative innovation index, in which different innovations are weighted by rarity (see e.g., Baker and Spetz 1999). We distinguish 7 types of innovation: multidisciplinary diagnostics, technical quality, nurse consulting hours, chain care, logistic optimalisation, transferred care and ICT. We distinguish a set of 15 determinants possibly affecting the absorption of innovations. The relationship between determinants and innovation status is investigated by using non-linear estimation techniques based on the Poisson and Negative Binomial distribution. The relationship between productivity and innovation is investigated by estimating a frontier cost function, which also includes the innovation indices. This is the most appropriate approach, since innovations may affect both efficiencies and the frontier itself. Data for this study were obtained from the Ministry of Public Health covering the years 1995-2002. and collected by the National Institute of Public Health. The data are derived from several surveys, such as financial, patient and personnel surveys. Additionally, unique data from a survey on the innovation status of Dutch hospitals were used. This survey contains information on 65 Dutch hospitals about 63 innovations during the period 1995-2004. From the extensive data set we choose 4 quantitative main outputs: the number of discharges of 4 groups of medical specialities, weighted by their case mix. Inputs include nursing personnel, paramedical personnel, other personnel, materials and capital. Capital is assumed to be a fixed input in all models. Input prices are measured as a hedonic price index, based on region and time period. The results show that innovation status is significantly positively influenced by a time trend, hospital concentration, hospital size and presence of a special innovation office. Alleged effects of specialist participation or decentralized management were statistically rejected. The estimates also show that chain care and logistic innovations are positively correlated with productivity. Multidisciplinary diagnostics and transferred care are negatively correlated. Technical quality, nurse consulting hours and ICT have a neutral effect on productivity. Further, it was shown that the innovation-index based on rarity shows a better performance. By using this index, the explanatory power of both the absorption-relationship and the productivity-relationship increased.

Hospitals Affiliations; and Efficiency:

Gerald Granderson - Ila Alam

This paper examines the effects of hospital ownership type, teaching objective, affiliation in an organization (alliance, network, system), and hospital contracts with HMOs and PPOs, on cost efficiency. With ownership type, possible differences in management goals among non-profit and government hospitals may contribute to differences in cost efficiency. We separate non-profit non-government hospitals into church operated versus “other” hospitals (primarily church affiliated and community hospitals) to determine if different types of non-profit hospitals behave dissimilarly. With location, rural hospitals, compared to urban hospitals, tend to deliver more primary than highly specialized care, and may treat a less complex mix and smaller range (distribution) of cases. This study allows for differences in location (urban and rural). With teaching objective, teaching hospitals may handle a more complex and wider distribution of cases compared to non-teaching hospitals. Major teaching hospitals may have different objectives, compared to minor teaching and non-teaching hospitals, with respect to the quality of education provided in teaching students. We separate teaching hospitals into major teaching and minor teaching hospitals to determine if different teaching objective affects cost efficiency. Regarding hospital contracts with HMOs and PPOs, hospitals and HMOs (or PPOs) negotiate the fees hospitals charge patients for providing health care services. Hospitals may offer HMOs and PPOs discounts on fees as incentives to sign the contracts. Given the discounts that are offered and the monitoring of their fees, hospitals that want to maintain a specified level of profits would have an incentive to operate more efficiently. Membership in an alliance, network, or system, may allow hospitals to better coordinate and allocate the treatment of cases across the various hospitals in the organization (alliance, network, or system). Member hospitals could then treat a smaller range of cases, and become more specialized in treating the smaller range of cases. Treating a smaller range of cases could eventually lead to improvements in cost efficiency. We examine if membership in two or more organizations impacts cost efficiency relative to membership in only one organization. The data sample is a panel of 248 general medical and surgical hospitals operating in four Midwestern U.S. states, Illinois, Indiana, Ohio, and Wisconsin, from 1996 to 1999. Hospitals in the Midwest region are more likely to face similar regulations, making it less difficult to compare results across states.

Estimating structural efficiency of the public standard hospital sector in Austria: 1999-2002:

Margit Sommersguter-Reichmann

This paper suggests a framework for assessing the technical efficiency of public, standard acute care hospitals in Austria between 1999 and 2002. Since the hospital service production process is considered to be represented by the use of multiple inputs to produce several outputs the flexible non-parametric approach of Data Envelopment Analysis (DEA) has been chosen to analyse the performance differentials of hospitals. To assess hospital efficiency, we apply an input-oriented radial efficiency measure since hospitals are increasingly confronted with scarce resources which have to be used efficiently to satisfy increasing demand. The prevailing analysis goes beyond previous efficiency studies in the Austrian hospital sector since - in contrast to sample-based analyses – in this case the performance of the population of Austrian public standard hospitals is evaluated. The population of all Austrian public hospitals consists of standard, centre, specialised and central hospitals, mainly differing in size and range of services offered. As preceding sensitivity analyses in terms of different input-output-specifications revealed considerable efficiency differentials for centre, specialised and central hospitals, we concentrated on the population of public standard hospitals as we found (input-output) model-independent efficiency findings for the latter group of public hospitals. The main focus of this research is on identifying the structural efficiency of the hospital sector at the level of the nine Austrian Länder (provinces). For that purpose, we apply the structural efficiency measure used by Førsund and Hjalmarsson (1979) to evaluate the performance of the Länder-specific standard hospital sector. To investigate efficiency differentials we employ the input vector {full-time equivalents of labour and beds} and the output vector {outpatient cases and LDF-points, the latter representing credit points per Austrian-specific DRG} while assuming variable returns-to-scale. As with many other efficiency studies in the hospital sector, the analysis suffers from the lack of quality indicators and has therefore be interpreted cautiously. In order to analyse productivity changes between 1999 and 2002 we use the chain version of the Malmquist index to disentangle efficiency from technology changes. This is considered to be important since Austrian-specific DRGs, used to capture inpatient performance, are redefined almost every year. The findings indicate slight efficiency improvements since 1999 at the federal and also, with a few exceptions, at the Länder level. The results of the sensitivity analyses (peer frequency, super efficiency) reveal genuinely efficient peers and confirm that the shapes of the annual efficiency frontiers can be considered as being robust.

3c

Productivity growth under uncertainty: an application to Spanish dairy farms:

Alan Wall - Luis Orea

In this paper we test for the existence of risk aversion among Spanish dairy farmers. Following the methodology of Kumbhakar (2002) we estimate risk preferences of dairy farmers using a panel data set covering the years 1993-98. We find that farmers are risk averse and find evidence of decreasing absolute risk aversion and increasing relative risk aversion. In the second part of the paper we propose a Malmquist-type index of productivity growth under risk and use the results of the estimation of risk preferences to calculate the index and carry out a decomposition of productivity growth.

ECONOMIC GROWTH WITH INEFFICIENCY:

Luis Orea - Antonio Álvarez

Standard economic theory does not say anything about the behavior of inefficient firms (regions, countries, etc). In fact it implicitly assumes the same behavior regardless of the efficiency level. However, it is reasonable to think that both efficient and inefficient firms would react differently under changes in size, regulation or when an economic policy is implemented. This paper argues that these differences in behavior between efficient and inefficient firms should be taken into account in order to explain changes in total factor productivity and, hence, the in income per capita over time. In particular, we show that overall productivity (economic) growth estimates are biased when they are obtained exploiting information exclusively for efficient firms. To achieve this objective we propose using a model that allows efficient and inefficient firms to behave differently.

Measuring and Improving System Efficiency during Transition Periods:

Warren Vaneman - Kostas Triantis

Transition is an overarching theme in today’s organizations. The body of literature is rich with change frameworks that provide guidance to organizations seeking performance improvement. However, in spite of the exceptional guidance available, the activities that take place during transition periods are often the most disruptive, and contain the most adverse performance and cost consequences, of any period during the system’s life-cycle. One of the reasons for this unfortunate result is the failure of change techniques and methods to identify an efficient path of transition, from the old way of doing business, to the new performance paradigm. We believe that an organization’s ability to master these transition periods is fundamental to achieving steady state operations more efficiently. Our research defines and explores transitional phases within systems and develops new methods for analysis of problems in a complex and dynamic environment. This paper uses a methodological approach that combines system dynamics modeling with the measurement of productive efficiency for the purpose of evaluating systems with respect to efficiency performance and costs during transition periods. The two fundamental premises of this research are that production systems are dynamic in nature, and that dynamic behavior of a system is a consequence of its structure. As systems become more complex and time dependent, alternative methods for evaluating production system performance in a dynamic environment must be explored. The efficiency literature to date has primarily been the interested in system performance measurement rather than causation, because it was thought that: (i) uncovering the pattern of efficient and inefficient practices should be paramount; and (ii) that the comparative advantage is with performance measurement and not determining the causal factors associated with system performance (Färe et. al., 1994). Nevertheless, Färe and Grosskopf (1996) suggest the need to explore what is inside the black box of production technologies to determine how inputs are converted into outputs, so that efficiency performance could be better understood. To this end, they developed a network technology model. In their model, they evaluate how multiple inputs injected into the production process, at multiple time periods, can produce multiple outputs. While this approach is evolutionary, it fails to evaluate the causal relationship that exists within the network. Additionally, we argue while this computational approach studies system performance over time, it only considers system in a steady state of performance, and does not consider non-linear relationships. We believe that system performance is inherent within the system’s structure and policies. Thus, if the system structure (inputs, information, processes, decisions, and outputs) is understood, the sources of good system performance can be replicated for future system design, and the causes of poor system performance can be corrected. Since policies are deep-seated within the system’s structure, determining the causal relationships will provide a window to how system policies affect system performance. This research builds on a system dynamics structure that measures dynamic efficiency performance in a multiple-input output environment and uses a hill-climbing optimization approach that determines the optimal path during the transition period. The objective of this paper is to illustrate the way in which this system dynamics structure can be implemented and how it provides decision-makers with policies to improve dynamic efficiency performance during the transition period. The specific application presented in this paper focuses on how the introduction of a new technology impacts the organization’s capability to provide four fundamental tasks (data preparation, data exploitation, information generation, and exploitation support) as well as the measurement of the organization’s dynamic efficiency performance from when the new technology was introduced.

3d

A Fully Nonparametric Stochastic Frontier Model for Panel Data: A Practical Approach:

Daniel J. Henderson - Léopold Simar

In this paper we estimate the production frontier and technical efficiency fully nonparametrically by exploiting recent advances in kernel regression estimation of categorical data. Specifically, we model firm (unordered) and time (ordered) categorical variables directly into the conditional mean. This approach allows us to smooth the firm and time specific effects, which formally entered the model linearly. Our setup allows for more flexible and accurate estimates of the production frontier and technical efficiency. We apply these techniques to a data set examining labor efficiencies of 17 railway companies over a period of 14 years. Not only are our results for the elasticites more economically intuitive than the parametric and semi-parametric procedures (which are special cases of our procedure), we obtain different rankings in terms of labor efficiencies.

CALCULATION OF SCALE ELASTICITIES IN DEA MODELS:

Finn Førsund - Lennart Hjalmarsson - Vladimir Krivonozhko - et al.

In DEA the qualitative characterisation of returns to scale has occupied many researchers the last decade. However, public policy purposes seem to be better served by numerical information, e.g. how important is the deviation from the competitive market condition that the scale elasticity must be equal or less than one? Following an indirect approach of using efficiency scores and shadow value on the convexity constraint the numerical value of the scale elasticity for radial projections of inefficient points to the frontier can be calculated (provided they are interior points and the facet is fully dimensional). The purpose of the paper is to provide a more powerful and general method using a direct approach to numerically evaluate the scale elasticity at any point on the DEA surface. The approach is based on cutting through the general multidimensional faceted DEA frontier with a two-dimensional plane in any direction from the origin, and calculating the scale elasticities for any point along the intersection of the planes and the frontier. For vertices or points on edges between facets the numerical method gives the scale elasticity based on a right-hand or left-hand derivative in a proportional direction from the origin, corresponding to the basic definition of scale elasticity. A comparison of the indirect and direct approach based on real data is provided.

Conditional Nonparametric Frontier Models for Convex and Non Convex Technologies: a Unifying Approach:

Cinzia Daraio - Léopold Simar

Efficiency and productivity literature primarily focused on the measurement of decision making units (DMUs)'' performance. In recent decades there has been a growing interest for the logical step ahead: the explanation of DMUs productivity differentials. As a matter of fact, the impact of external-environmental factors on the efficiency of producers is a relevant issue related to the explanations of efficiency, the identification of economic conditions that create inefficiency, and finally to the improvement of managerial performance. These factors are neither inputs nor outputs under the control of the producer, but can affect the performance of the production process. In literature, two main approaches have been developed. In the 'one-stage' approach the environmental variables are directly included in the linear programming formulation along with the inputs and outputs. Its main disadvantage is that it requires the classification of environmental factor as an input or an output prior to the analysis. In the 'two-stage' approach the technical efficiency, computed in a standard way, is used as dependent variable in a second-stage regression. The main shortcoming of the two-stage approach, as pointed out in Simar and Wilson (2003), is that the efficiency estimates are serially correlated in a complicated way and that the first stage efficiency scores are biased. Daraio and Simar (2003), hereafter DS, propose a full nonparametric approach which overcomes most of the drawbacks mentioned above. They define conditional (to external-environmental factors) frontiers and conditional order-m frontiers together with their related efficiency scores and the corresponding nonparametric estimators. In particular, order-m frontier estimators (Cazals, Florens and Simar, 2002) are known as being more robust to outliers and extreme values than the full frontier estimates. In this paper we provide a unifying approach to introduce external environmental variables in nonparametric models of production frontiers. Completing the work done in DS we introduce a conditional Data Envelopment Analysis (DEA) estimator, i.e., a DEA estimator of production frontiers conditioned to some external-environmental variables that are neither inputs nor outputs under the control of the producer. In order to control for the influence of extremes or outliers we introduce also a robust version of our conditional DEA estimator, based on the concept of order¡m frontiers. We aim at enriching the toolbox of applied researchers in productivity analysis offering a complete range of conditional measures of efficiency, i.e., measures of performance which take into account the operating environment (or other external factors) in which firms operate in, without imposing their positive or negative impact, but letting the data themselves to tell if and how they affect the performance. The paper is structured as follows. In Section 2 we describe the frontier estima- tion setting and we propose, extending DS, a unifying formalization of the production process based on a probabilistic approach, where the FDH and DEA estimators can be naturally introduced. Section 3 presents the concept of order-m frontiers, and analyzes how convexity can be introduced in these partial frontiers. This leads to define efficiency scores of order-m with respect to convex technologies. Nonparametric estimators are then described and some of their properties are investigated. Section 4 shows how the probabilistic formulation allows to introduce conditional efficiency measures and, extending DS, defines a conditional DEA efficiency score and its robust (order-m) version. In Section 5 we propose a series of indicators of the type of those proposed in Briec, Kerstens and Vanden Eeckaut (2004), extending its application to robust order-m efficiency measures and to conditional and robust measures of performance. Section 6 illustrates the different concepts trough some simulated data sets as well as real data on mutual funds. Section 7 concludes, outlining future development to address. In the Appendix we address some issues about the bandwidth selection procedure necessary for estimating most of the conditional measures. Motivated by the consideration that there exist empirical applications in which convexity could be reasonable we propose in this paper a conditional DEA estimator and a robust version of it based on the concept of order-m frontier. We describe also how these measures can be estimated and we address the problem of their practical computation. These newly introduced measures complete the exploratory tools available for gauging the performance of DMUs when extra information on operating environment are available. We report also some indicators of convexity for several conditional and uncondi- tional, full frontier and robust efficiency measures, extending previous indicators proposed in the literature. Finally, we illustrate all these concepts trough the analysis of some empirical examples: simulated and real data sets. The analysis of the distributions of convexity estimators in the mutual funds example shows that convexity is not clearly established and non-substitutability among the management dimensions (risks, turnover and transaction costs) might be at place in US Aggressive Growth funds. This illustration suggests that the convexity issue should be carefully taken into account in applied works. As a matter of fact, even when convexity could be reasonable from a theoretical point of view, its validity should be empirically checked and verified. Moreover, non-convex estimators are always consistent (even under convexity), whilst convex estimators are consistent only under the convexity assumption. The indicators of convexity presented here, even if useful for descriptive and ex- ploratory purpose, are not able to give a definitive answer about the convexity assumption of the corresponding attainable sets. In fact, the conclusions are drawn in terms of estimated technologies instead of true technologies. A statistical test procedure is requested to make inference with respect to the true technology. In other words, for a particular observation or for the global technology, without a formal testing procedure, it is impossible to determine if the values of the various indicators of convexity less than one are due to non convexity or due to sampling variation. Rigorous statistical procedures for testing convexity should follow the lines of Simar and Wilson (2001, 2002). Their implementation both in the traditional inputs-outputs representation of the production process and in the enlarged inputs-outputs-external factors framework is left for future development of this work.

4a

Using the market at a cost: A note on how the implementation green certificates lead to market inefficiencies:

Thomas Sundqvist and Mats Nilsson

Abstract unavailable.

Mergers in Norwegian Electricity Distribution: An Efficiency Enhancing Exercise?:

Thor Erik Grammeltvedt - Peter Bogetoft

In this paper, we investigate if the mergers of Norwegian electricity distribution companies in the period 1995-2001 can be rationalized as a cost saving exercise – or whether they should be explained by other factors, including the regulatory regime. We use Data Envelopment Analysis to estimate the potential production economic gains from mergers ex ante, and we decompose the gains into potential gains from learning best practice and from better economies of scale and scope. Moreover, we compare with post merger efficiencies to identify extraordinary costs and benefits from the mergers. Our primary focus is on the production economic aspects of mergers, but we supplement with a few observations on strategic implications. We study the regulated revenues and efficiency improvement requirements before and after the mergers to quantify possible gains from “playing the regulation”. We find that mergers lead to a change in input mix which may be associated with potential production economic gains, in particular the possibility to save on labor and service costs. We find some evidence that this change may be associated with a more rapid technological progress or catch up among merged companies, but this result may also be explained by some limitations in the existing analytical instruments. We also find that the “new” companies face somewhat more favorable regulatory conditions, i.e. higher allowed revenues and lower required efficiency improvement requirements. The latter is the consequence of the - possibly flawed - specification of the return to scale in the regulatory model cost model. The regulatory model allows for diseconomies of scale, and the bias in the estimated frontier and hereby in the acceptable revenues may therefore be considerable for the newly formed and relatively larger units. Lastly we find some indications that most of the mergers may also be motivated be gains in related activities, most notably electricity retail sale.

DEA Based Yardstick Regulation: Implementation and New Challenges:

Peter Bogetoft - Per Agrell - Dag Fjeld Edvardsen

The aim of this paper is to discuss the implementation of DEA based yardstick competition. The specific setting is electricity distribution in the Nordic countries and the focus is on some of the new practical and theoretical challenges that arise when theory is put to work. We discuss the transition from CPI-X based regulation to a dynamic yardstick model, how the underlying cost model shall be specified, how to including the yardstick elements in the financial reporting etc. Also, we investigate how to integrate recent advances in the benchmarking methodology, including bias correction based on bootstrapping, and how the new methods relate to the underlying theory of DEA based yardstick competition.

Productivity and deregulation of Norwegian Electricity utilities:

Sverre A.C. Kittelsen - Dag Fjeld Edvardsen - Finn R Førsund - et al.

The deregulation of the Norwegian electricity sector in 1991 separated production and supply on the one hand and transmission and distribution on the other hand. Distribution utilities remain natural monopolies and are subject to a regulatory regime designed to give incentives to productivity improvements. The basic mechanism is based on the RPI-x formula, i.e. costs (or maximal price) are allowed to increase according to the costs of inputs measured by the most relevant index, but not to a full extent; a productivity factor x has to be subtracted. Calibrating the x factor was partly based on an investigation of productivity development of utilities before the change of the regulatory regime. The x-factor was gradually introduced and increased to give the utilities some years to adjust. The full regulation took place in 1999. The purpose of the paper is to investigate the productivity development of individual distribution utilities for the last decade covering both pre- and post periods of the RPI-x regulation to study the impact of the regulation. A Malmquist productivity index approach will be used based on non-parametric frontiers. The decomposition of productivity change into technical shift and catching up will reveal the nature of the utilities’ effort to benefit from the regulation. Bootstrapping techniques will be applied in order to give a better statistical basis for the results.

4b

Assessing the efficiencies of hospitals that provide health services research and teaching:

Marcos Estellita Lins - Maria Stella Lobo - Angela Cristina Moreira da Silva - et al.

This paper presents an empirical illustration of the use of DEA to analyze the efficiency of hospitals that present pdevelop activities in three areas : research, teaching and services. The objective is to identify key factors associated with the efficient use of clinical resources in the provision of hospital services. This study will explore this question by considering a first level where the three aeas abovementioned are taken in separate, and a second level, where an aggregate approach is considered. A weight restricted DEA model is used in each level.

Decomposition of cost changes in Dutch homes for the mentally disabled:

Evelien Eggink - Jedid-Jah Jonker - Michiel Ras

The costs of Dutch homes for the mentally disabled have risen substantially during the last decades. The number of mentally disabled is growing and the care for the mentally disabled changes because of changing ideas on the quality of care. In addition there are many mergers, closures and new start-ups, changing the number of institutions as well as their size and scope rapidly. In this paper the background of the cost increase in the homes for mentally disabled is analysed. For the empirical analysis a panel of Dutch homes for the mentally disabled is used covering the period 1984-1998 and a stochastic parametric frontier cost model is employed. The cost increase in the sector appears to be caused in part by increasing prices and changes in the macro level of production. The contributions of these changes are straightforward to calculate. In addition the costs per product may change. A stochastic parametric frontier cost model is used to assess and decompose the contribution of the costs per product in the cost increase. The impact of changes in the size and composition of production at the micro level are analysed as well as the usual productivity change and its decomposition into technical change and efficiency components. Because of the dynamics in the sector at hand, the panel is unbalanced and the usual techniques to calculate productivity change can not be employed. In addition, the contribution of changes in size and composition of production at the micro level are difficult to calculate. Therefore a circumventing method is constructed to approximate the contribution of each factor to the change in the costs per product. This completes the decomposition of the total cost change in the sector.

Police Reallocation Models:

Mette Asmild - Joseph C. Paradi - Jesus Tadeo Pastor Ciurana

From a central planners point of view, an overall objective of the allocation of resources (inputs) to units can be to make all units efficient. This can be accomplished by reducing inputs for inefficient units and potentially increasing inputs for super-efficient units. However, such reallocations can be done in many different ways. The choice between alternative reallocations depends on the policy objectives of the central management. In this paper we present several different reallocation models that could be relevant for the central management in an empirical case of staff allocation to police units. We consider both reallocations between units but within separate functions, reallocations between functions within units and finally reallocations between both functions and units.

4c

Allocative Efficiency and Decomposition of Productivity Change:

Jiro Nemoto - Mika Goto

This study aims to develop a framework of the productivity decomposition by which the allocative efficiency factor is identifiable as a source of productivity change. Although we have a lot of ways to decompose productivity change, none of them is indicative of allocative efficiency. This is because efficiency has been measured along the fixed ray passing through an observed combination of inputs and outputs regardless of their efficient combination. To make the efficiency measure sensitive to alloccative efficiency, we define the productivity index in the form of the ratio of aggregate inputs and outputs which are implicitly formed from revenues and costs divided by aggregate price indexes of the Konus type. The resulting decomposition analysis is approximately equivalent to that of the Hicks-Moorsteen-Bjurek productivity index measured along the ray of efficient input-output combination. An empirical illustration will be delivered using data on the Japanese prefectures.

Malmquist indexes using a Geometric Distance Function (GDF):

Maria Portela - Emmanuel Thanassoulis

Traditional approaches to calculate total factor productivity change through Malmquist indexes rely on distance functions. In this paper we show that the use of distance functions as a means to calculate total factor productivity change may introduce some bias in the analysis, and therefore we propose a procedure that calculates total factor productivity change through observed values only. Our total factor productivity change is then decomposed into efficiency change, technological change, and a residual effect. This decomposition makes use of a non-oriented measure in order to avoid problems associated with the traditional use of radial oriented measures, especially when variable returns to scale technologies are to be compared.

On the Anatomy of Productivity Growth: A Decomposition of teh Fisher Ideal TFP Index:

Timo Kuosmanen - Timo Sipiläinen

Decompositions of productivity indices contribute to our understanding of what drives the observed productivity changes by providing a detailed picture of their constituents. This paper presents the most comprehensive decomposition of total factor productivity (TFP) to date. Starting from the Fisher ideal TFP index, we systematically isolate the productivity effects of changes in production technology, technical efficiency, scale efficiency, allocative efficiency, and the market strength. The three efficiency components further decompose into input- and output-side effects. The proposed decomposition is illustrated with an empirical application to a sample of 459 Finnish farms over period 1992-2000.

A Decomposition of the Malmquist Productivity Index: Returns to Scale and Technical Progress with Imperfect Competition:

Kevin Fox - W. Erwin Diewert

An extension of the framework of Caves, Christensen and Diewert (1982) is considered. This yields a decomposition of the Malmquist productivity index into technical progress and returns to scale components.

4d

FINDING UNIQUE WEIGHTS FOR EFFICIENT UNITS IN DATA ENVELOPMENT ANALYSIS:

Pekka Korhonen - Rolf Färe

This paper proposes two different principles to find unique weights (multipliers) for efficient units in the Data Envelopment Analysis. In the first approach, an extra single objective function is used to find the weights that discriminate the current efficient unit from the other units as clearly as possible. The min-max rule is applied. The second approach proposes the use of multiple objective linear programming. The multiple objective formulation provides the user with an opportunity to find the most preferred weights for units subject to that the unit remains optimal. The approaches are demonstrated by numerical examples.

Profit efficiency analysis with absolute and uniform shadow prices:

Laurens Cherchye - Mika Kortelainen - Timo Kuosmanen - et al.

We extend the nonparametric toolkit for testing data consistency with profit maximization. Our methodological extensions exploit the dual ‘shadow’ profit efficiency interpretation of the DEA model that builds on assumptions of monotone and convex production possibility sets, and the concomitant use of DEA weight restrictions for constraining the relative shadow prices of the inputs and outputs. Extending earlier work (see Kuosmanen et al., 2004), we then suggest the introduction of uniformity restrictions regarding the shadow prices across the different observational units; the well-established ‘Law of One Price’ (LoOP) provides a solid theoretical basis for such cross-observational shadow price constraints. Within a directional distance function framework, we argue that the selection of a single numeraire good suffices for measuring LoOP-based profit efficiency in absolute price terms. An application to the profit efficiency evaluation of Finnish farms illustrates our methodology. This application shows the potential of the methodology for enriching the conventional shadow price based profit efficiency analysis, including the recovery of LoOP-consistent shadow price information. KUOSMANEN, T., L. CHERCHYE and T. SIPILÄINEN (2004), 'The Law of One Price in Data Envelopment Analysis: Restricting Weight Flexibility Across Firms', European Journal of Operational Research, to appear.

New formulations of DEA and FDH models:

Hervé LELEU

New formulations of Data Envelopment Analysis (DEA) and Free Disposal Hull (FDH) models in a general and unified linear framework are proposed. One of the main objectives of this paper is to derive meaningful economic interpretations of the dual models in the price space. In particular, we introduce a new formulation of the returns to scale assumption with a straightforward economic interpretation. This framework allows for a clear distinction between DEA and FDH models and leads to a natural generalization of both models.

5a

Quality of service technical efficiency and economies of scale - An analysis of European electricity distribution utilities:

Christian Growitsch - Tooraj Jamasb - Michael Pollitt

Although quality of service is of major economic significance in most network industries, it is generally not reflected in technical and scale efficiency analysis. In this paper, we apply stochastic frontier analysis to estimate technical and scale efficiency using a sample of about 500 electricity distribution utilities from eight European countries. We then extent the scope of analysis by incorporating quality of service using multi-output translog input distance function models. We find that introducing the quality dimension into the analysis affects estimated efficiency significantly. In particular, smaller utilities seem to benefit from incorporating quality. We also examine the effect of quality of service on efficient firm size. Our findings emphasise that quality of service should be an integrated part of efficiency analysis and incentive regulation regimes.

Model Development for a Pan-Nordic Incentive Regulation for Electricity Distribution:

Helle Gronli

Abstract unavailable. The productivity of water and wastewater services in Australia:

Tim Coelli

Various government-owned businesses supply water and wastewater services to Australian residents. With the advent of recent competition and regulatory reforms in infrastructure industries in Australia, more and more of these businesses are now facing new types of incentive-based regulatory regimes. This has led to a desire for more information on the performance of these businesses, both relative to each other and over time. In this study we use panel data on the 18 largest Australian water services businesses, observed over an eight-year period from 1995/6 to 2002/3, to measure the relative efficiency and productivity growth of these businesses. Econometric estimates of translog cost and distance functions are used to obtain these measures. The analysis also provides estimates of scale and scope economies, which are of interest when discussing mergers and the long-term structure of this industry.

5b

The long run productivity of primary care in Finland:

Juha Aaltonen - Maija-Liisa Järviö - Kalevi Luoma - et al.

This paper studies the development of productivity and efficiency in primary health care of Finnish health centres from 1988 to 2003, with special emphasis on selection of proper weights for output measures over time. Given that only approximate national wide data are available on unit prices of different outputs, the DEA methods is used to produce optimal weighting structure for each health centre. However, in the long run it is necessary to give bounds within which theoretical marginal rates of substitution between outputs are allowed to vary. We report annual AR-DEA results over period 1988 -2003, with comparable productivity indices. The data consisted annually 195-221 health centres, covering about 70 percent of the Finnish population. We utilised different kinds of visits as output measures for outpatient care: visits to physician, dental visits, visits to other health care personnel and home nursing visits. Short term acute inpatient care was measured in discharges. Long term chronic care and non acute short term care was measured in bed days. Resources were measured by gross operating costs of the health centres. The data are extensively checked for the consistency. Using the panel feature of the data we have followed the productivity improvement of each unit that was at least in one year assigned as fully efficient. If no reasonable explanation could be found for an exceptionally large change in the productivity of the health centre the observation was excluded from the data. Altogether productivity improved by just a couple of percent over the whole period 1988-2003. However, the productivity figures imply a high increase in productivity over 1991-1993, due to decreases in resources allocated to primary care. Thereafter productivity still increased until 1997, but since the primary health care in Finland has experienced downward trend in productivity. Calculations relative to the joint frontier over 1997 – 2003 show, that on average productivity has decreased by about 13 percent. The productivity has fallen rapidly especially since the year 2000, at the time when health centre resources have been increased substantially. We find no evidence that the introduction of new technology within primary care over the 15 year period would have contributed towards significant productivity improvements. On the contrary our results suggest a technological regress in many years. In the long run there seems to be an association between productivity development and resource growth health centres. Productivity falls when health centre resources are increased rapidly. One of the main outputs, visits to physicians, decreased by 13 percent from 1997 to 2002, while the number of physicians in health centres increased by 12 percent at the same period. Differences in efficiency have remained rather stable over the whole period, except the year 1993. The average size of the most efficient health centres is close to the median size of all the units. However, relatively small health centres are on average less efficient than larger ones, but beyond that there is no evidence of economies of scale. Those health centres that were the least efficient in 1988 had on average a more positive productivity development from 1988 to 2003 than the rest of the health centres. Some practical explanations on productivity fall observed since 1997 have been suggested. The inpatient care in the inpatient wards may be on average more demanding than before because the average dependency level of patients in these wards has risen. Community and home care have been substituted for inpatient care more and more for chronically ill elderly patients. However we did not find any support from municipal level analysis for this. A part of the observed productivity decline observed in recent years may be explained by efforts of health centres to raise the quality of the care given in their inpatient wards by raising the number of nursing staff. The explanation models of efficiency differences usually suffer from inappropriate underlying statistical model for the efficiency differences. DEA efficiency scores do not generally have the i.i.d. property required for the statistical inference. However, the inference is usually needed to draw conclusions on population parameters, but our health centre data cover the whole population of health centres in Finland. Also, as the number of fully efficient units is low, 9, we do not need to rely on ML-estimators like Tobit, but OLS-based fixed effect model suffices. Thus, we have used the panel regression models just to explain the variation in productivity. Our set of possible explanatory variables consists of both external and internal variables for the health centre performance over 1997-2003. The external variables can explain about 17 percent of efficiency variations. The most powerful explanatory variables are connected to taxable income within municipality, availability of specialised care (hospitals) and usage of private health care services (reimbursed by health insurance). All these variables are associated with higher economic capacity within the health centre area and have a negative impact on efficiency. The internal variables described the employment shares of different professions in health centres, the role of the acute care and the way patient-physician relation is organised. Their explanatory power was 8 percent of the variation. Health centre productivity was positively associated with the high physician share and negatively associated with high nursing staff share. The results of our analyses suggest that budgetary stringency and financial incentives are significant determinants for differences in productivity and efficiency of health centres. This is consistent with the hypothesis that the grant system reform introduced 1993 led to gains in productive efficiency. As the Finnish health care system appears in international comparisons to have relatively low share of GDP, the tightening of the financial sources does not appear that attractive choice for the policymakers. Therefore in the future the main efficiency gains are most likely to be found in the renewing the internal organisation and operation of providing primary health care. The large and persistent variation in the productivity and efficiency of health centres suggest that there is ample scope for this.

Peer Effects and Productivity: The Role of Competition and Safety Net Hospitals on Performance:

Gary Ferrier - Vivian Valdmanis

Numerous papers have used frontier methods to measure and examine the productivity of firms operating in a host of different industries. To the best of our knowledge, however, there does not exist any research addressing the issue of how a firm’s productivity is related to that of its peers. This paper seeks to address this hole in the literature. Microeconomic theory suggests that competition among firms forces each of them to operate efficiently if they hope to survive in the marketplace. Even in markets that are less than perfectly competitive firms are likely to adjust their performance in light of the performance of their peers. One reason for doing so may be the need to attract capital; another reason may be the “bragging rights” that are associated with superior performance. On the other hand, in less than competitive markets firms may adopt the “quiet life” described by Hicks, operating only as efficiently as required by the pressure placed upon them by the performance of their peers. At any rate, it would seem that evaluating firms’ performances in relation to that of their peers is a worthwhile endeavor. In particular, we examine the relationship among the performance of a sample of hospitals operating in large US urban markets over the period 1993-2002. As a first step, the productivity of the hospitals is calculated using a non-parametric Malmquist index based on DEA models. As a second step, the Malmquist index (and its components--technical change and changes in efficiency) of each hospital is regressed against the mean performance of its peers as well as other factors needed to control for confounding influences. To accomplish the second step, econometric models described in Brock and Durlauf (2001) and Moffitt (2001) will be used. “Peers” will defined in several ways—as those operating in the same urban area, those that are part of the same hospital chain, those of similar ownership type, etc. The expectation is that there will be a strong positive relationship between a given hospital’s performance and that of its peers. In addition to the above analysis, a special examination of the effect of public hospitals on the performance of neighboring hospitals will be performed. In the US, public hospitals serve as “safety nets” for the tens of millions of Americans without health insurance in that they operate with an “open door policy to serve all patients regardless of their ability to pay and provide substantial levels of care to Medicaid, the uninsured, and other vulnerable patients” (Institute of Medicine, 2000). In their role as safety net providers of health care, public hospitals relieve neighboring hospitals of part of the burden of providing care to the un- and under-insured. Therefore the presence of a public hospital is hypothesized to have positive spillover effects onto its geographic peers. This analysis will be aided by the fact that a number of urban areas (e.g., Washington, DC, and Boston) saw their public hospitals close over the sample period, allowing for an assessment of the effects of the closures on peers.

A PARAMETRIC DECOMPOSITION OF THE OUTPUT-ORIENTED MALMQUIST PRODUCTIVITY INDEX FOR PUBLIC HOSPITALS IN GREECE:

Roxani Karagiannis - Michalis Hatziprokopiou

The first objective of this paper is to measure the productivity of a panel of Greek public hospitals during the period 1999-2003. For this purpose a multiple - output multiple - input output distance function is employed to investigate how many patients could be nursed and how many laboratory tests could be accomplished, given the technology and input quantities such as beds, doctors, nurses, administrative personnel and medical and pharmaceutical costs. The adoption of the Malmquist productivity index is necessary because of two reasons: first, it is based only on output and input quantities, render hospital input prices unnecessary. Second, combines technological change and technical efficiency change. The second objective is to combine the measures of technological change, technical efficiency change, scale efficiency change, and output mix into a measure of total factor productivity using the methodology of Balk (2001). The resulting output-oriented Malmquist productivity index is measured using the parameter estimates of the underlying output distance function, which is estimated within a stochastic frontier framework. With the exception of Maniadakis and Thanassoulis (2004), all previous studies for Greek public hospitals have focused on measuring technical efficiency (Giokas 2001; Athanassopoulos and Gounaris 2001; Karagiannis and Hatziprokopiou, 2004). Maniadakis and Thanassoulis (2004) provided estimates of the Malmquist productivity index using however a deterministic approach. The present paper extends their results into a newly developed data set for the period 1999-2003 using a stochastic approach, and employing a more detailed decomposition. Estimation of Information Asymmetries in Housing Markets: A Two-Tier Stochastic Frontier Approach:

Christopher Parmeter - Subal Kumbhakar

In this paper we estimate a hedonic price function for housing characteristics. Since the buyers and sellers do not have full information at the time of the sale, the offer price (maximum that the buyer is willing to pay) will differ from the reservation price (minimum that the seller is willing to accept). If information is freely available to both parties (as is the case if the market is perfectly competitive and centrally located), there will be no difference between the maximum offer price and the minimum acceptance price. If not, there are both positive and negative effects influencing the market equilibrium price due to information asymmetries between buyers and sellers. To measure the impact of these information deficiencies on price we need to estimate the hedonic price function in such a way that both the maximum offer price and the minimum acceptance price functions can be obtained. That is, using the observed data we need to fit a function so that each observation lies between, barring the stochastic noise component, the maximum offer price and the minimum acceptance price functions. This is done using the two-tier stochastic frontier approach, developed by Polachek and Yoon in 1987, composed of one two-sided and two one-sided errors, i.e., the composed error term v+w-u . The price gap represented by u>0 is the seller’s price shortfall revenue due to information deficiency. Similarly, the price gap w>0 represents the buyer’s over payment due to information deficiency. Both the one-sided error components might reflect the costs of obtaining information due to incomplete information. The v term represents random noise in the price determination process and is allowed to take both positive and negative values. To estimate the above model we make distributional assumptions on the error components, viz., u, v, and w. These are: (i)v is distributed normal , (ii) w is distributed exponential, (iii) u is distributed exponential, and (iv) the error components are distributed independently of each other and from the regressors, X. These assumptions will allow us to derive the likelihood function, maximization of which gives the ML estimates of the parameters. Next, we employ the Jondrow et al. (1982) type decomposition to obtain observation-specific estimates of u and w via linear predictors. This will allow us to recover estimates of the offer prices for buyers and reservation prices for sellers. Once these are found estimates of the offer prices and the reservation prices are obtained. An analysis of the valuation of certain characteristics by buyers and sellers is considered and ideas for further research are suggested.

5c

Growth and convergence in the EU: relative contributions of physical and human capital accumulation:

MªMar Salinas-Jiménez - Inmaculada Alvarez-Ayuso - Mª Jesus Delgado-Rodriguez

The aim of this paper is to analyse labour productivity growth and convergence in the EU between 1980 and 1997. Adopting a production frontier approach, labour productivity growth is broken down into components attributable to efficiency change, technological progress and capital accumulation. Non-parametric techniques of linear programming (Data Envelopment Analysis) are used to estimate a common production frontier and TFP is decomposed by means of Malmquist productivity indices. With regard to capital accumulation, we analyse the contribution of private capital to labour productivity growth and, additionally, we consider both a broad measure of physical capital -introducing public capital as an additional input- and human capital accumulation. Focusing on this decomposition, our results show that physical and human capital accumulation appear to be the major source of labour productivity growth in the EU during the 80’s and 90’s. When only physical capital accumulation is considered, labour productivity growth is explained by both capital deepening and TFP growth. However, part of the estimated TFP growth is in fact due to human capital accumulation. Thus, when the human capital variable is introduced in the analysis we observe that the contribution of TFP tend to be negative, reflecting a significant problem in terms of TFP for the European economies. However, one should soften this result by noting the existence of significant differences among the European countries with regard to the factors driving productivity growth. In this sense, we observe that the contribution of TFP growth to labour productivity is positive in most of the European central economies. On the other hand, there are mainly the Mediterranean countries which suffer from problems in TFP, with labour productivity growth being positive as the result of the intense process of capital accumulation undergone by these economies. Once the components of labour productivity growth are analysed, we focus on their relative contributions to convergence. A first approach to analysing this contribution to convergence bases on the commonly used concept of b-convergence. Furthermore, in the spirit of the Quah’s approach, the dynamics of the overall distribution is also analysed, both for the distribution of labour productivity and for each of its components. Regarding the convergence process, a slight tendency toward convergence among the European economies is found during these years. Firstly, we observe a process of technological catch-up (or convergence in efficiency levels) since, on average, the less efficient countries in 1980 underwent greater efficiency gains than the more efficient ones. Nevertheless, efficiency change does not appear to be a source of convergence in terms of labour productivity, even when it appears to be the main factor explaining the overall evolution of the labour productivity distribution. On the other hand, technological progress tends to contribute to divergence, whereas physical and human capital accumulation appears to be the main force driving the process of convergence in labour productivity, with a strong inverse relationship between capital deepening and the initial levels of output per worker. Consequently, given that labour productivity growth and convergence in the EU is largely driven by physical and human capital accumulation, public investment in these types of capital might constitute an appropriate instrument of development and cohesion policy. However, with regard to TFP, our results indicate that some European countries (and especially the Mediterranean ones) are suffering from problems in TFP growth. Therefore, policies aimed at promoting TFP growth in the European countries (i.e. policies promoting macroeconomic stability in order to improve efficiency levels, or R+D activities encouraging technological progress) should also be strongly supported.

Contributions of ICT to Chinese Economic Growth: Almas Heshmati

The view about systematic irrationality of investors and managers in investment with reference to ICT with no effects of productivity growth is called productivity paradox. Research suggest that ICT return in developed nations is significant and positive, but not in developing countries. This paper challenges the above conclusion by examining the contribution of information and communication technology to Chinese economic growth. We will provide some empirical evidence that China has reaped in ICT investment, specifically, the aim is firstly to research on relationship between TFP and ICT capital by using Chinese data. Secondly, provide estimation of the returns or lack of returns to IT investment with sensitiveness analysis, by using different depreciation rate and income share of capital. Thirdly, discuss the policy implications for the Chinese ICT development and investment patterns. Finally, we suggest guidelines for future research in general and also in the China case. The results are of great interest by adding to our understanding of how ICT affects growth in an economic development context.

Productivity technological spillovers and human capital. An analysis on Italian firm data.:

Vania Sena - Sergio Destefanis

The hypothesis that technological knowledge acquired by a firm can spill over to other firms and enhance the latter group''s total factor productivity was first suggested by Arrow (1962) in his work on the effects of learning embodied in new capital equipment. Since then, there has been a considerable theoretical debate on the extent to which a firm can benefit from spillovers and also how much of its productivity growth such spillovers can explain (Romer, 1986; Grossmann and Helpman, 1990; Cohen and Levinthal, 1989). While there exists now considerable evidence that knowledge spillovers exist and do increase total factor productivity by allowing lagging firms to get closer to the technology frontier (see Sena, 2004a), it is not clear yet what are the factors that facilitate and amplify the potential spillovers. In this paper, we test the extent to which human capital can be considered among the factors facilitating spillovers and the diffusion of new technologies. This is highly relevant for the debate on recent trends in the competitiveness of the Italian economy, as the low educational attainment of the latter has often been quoted as one of its main weaknesses. More specifically, extending upon the analysis contained in Sena (2004b) and using stochastic frontier analysis for the main sectors of the Italian manufacturing, we test a) whether technology spillovers help firms to get closer to the technology frontier and b) whether the impact of spillovers on productivity is higher in firms where the stock of human capital is higher. In this process, we control for the endogeneity of the location choice of firms in areas where human capital is higher. We rely on firm data drawn from the various Mediocredito Centrale Surveys and on the regional data on human capital attainment constructed in Destefanis et al. (2004). K. Arrow (1962), The Economic Implications of Learning by Doing, Review of Economic Studies, 29(80), pp. 155-73. W.M. Cohen, D.A. Levinthal (1989), Innovation and Learning: Two Faces of R&D, Economic Journal, 99, pp. 569-596. S. Destefanis, A. Taddeo, M. Tornatore (2004), The Stock of Human Capital in the Italian Regions, WP 3.142, DISES, Università di Salerno. G. Grossman, E. Helpman, E. (1990), Trade, Innovation and Growth, American Economic Review, 80-2, pp. 86-91. P.M. Romer (1986), Increasing Returns and Long-Run Growth, Journal of Public Economics, 94, pp. 1002-1037. V. Sena, (2004a), The return of the Prince of Denmark, Economic Journal, 114, pp. F312-F332. V. Sena, (2004b), The spillover hypothesis of the endogenous growth literature: some new evidence, International Journal of Production Economics, forthcoming.

Linking Investment Bursts and Productivity: An Empirical Investigation in US. Food Manufacturing:

Spiro Stefanou - Pinar Celikkol

Quantifying the importance of factors driving productivity growth such as changes in technology and identifying the relationship between productivity and investment are challenging tasks and have been only partially successful to date. The major complication arises from the causality in the relationship between investment and productivity. Productivity growth implies resource use decisions can impact the quantity of resources available for new production planning, in particular, and activities, in general. Some of these changes may involve doing the same thing more extensively (i.e., extracting scale economies) and some of these changes may involve doing things differently (i.e., introducing new equipment and processes). A detailed empirical analysis at plant-level to gain a better understanding of the relationship between the decision to invest, productivity and plant characteristics can shed some light to this relationship. The main objective of this study is to investigate the causal mechanism between investment and productivity in a structural model using Census Bureau’s Longitudinal Research Database at plant-level from 1972-1999 by focusing on various sub-industries of U.S. Food and Kindred Products Industry. Capacity-improving investment activity is measured by lumpy investment using both absolute and relative measures. Plant investment age, which tracks the time between investment spikes, is calculated as the time elapsed since the plant’s most recent investment spike. To investigate relationship between the decision to invest and productivity growth at plant-level, we estimate a set of least-squares regressions with and without fixed effects where the dependent variable, productivity growth, is regressed to variables such as plant age, plant size, 4-digit industry, year, and lagged plant investment spike variables. For the lagged investment spike variable, we have used various lagged alternatives of investment spikes. We run these models controlling and not controlling for the quartile groups of plants with respect to their TFP growth. The preliminary results in the case of meat products sub-industry finds that the lagged investment spikes for middle quartile ranked plants lead to increase in TFP growth that is attributed to technological change effect dominating the scale effect. When looking at plants that are in only the lowest and highest quartiles together, the lagged investment spike leads to a decrease in TFP growth that is attributed to scale effects dominating the technical change effect. The same result holds when all quartile groups are grouped together. Preliminary results for the case of dairy products sub-industry find that the lagged investment spike leads to increased TFP growth that is attributed to the technological change effect dominating for the plants that are in the lowest and the highest quartile ranked and for plants in all quartile groups together.

5d

Measuring eco-efficiency of production: A frontier approach:

Mika Kortelainen - Timo Kuosmanen

Eco-efficiency of production refers to the capability to produce goods and services by polluting the environment and using natural resources and energy as little as possible. The purpose of this paper is to present a general framework for the measurement of eco-efficiency by drawing insights both from the production theory and eco-efficiency literature. By utilizing methods used in productive efficiency analysis, we construct a general eco-efficiency index that can be applied in different kinds of applications. Although we exploit the existing methods and techniques, our approach differs in some important respects from the usual treatments of firm-level environmental performance analysis (e.g. Färe et al., 1996; Färe et al., 2005). The main difference between our approach and the earlier studies is that we build on the most standard definition of eco-efficiency as the ratio of economic value added to the environmental damage index. This definition gives an equal emphasis on both economic and environmental criteria. By contrast, the efficiency analysis literature typically treats the environment merely as one additional criterion (or constraint) in the technically oriented efficiency assessment. Another difference to be noted is that we focus explicitly on the tradeoffs between the creation of economic value added and its undesirable side-effects to the environment, without direct recourse to physical inputs and outputs. Related to this orientation, we also approach eco-efficiency from a more aggregate perspective. Our general framework is illustrated by means of an empirical application to the evaluation of eco-efficiency of road transportation in Finland.

Nitrate Pollution Control Policy and its impact on Farms'' Performance: A Nonparametric Approach:

Isabelle Piot-Lepetit - Monique Le Moing

In Europe, water quality problems associated with the use of synthetic fertilizers and the disposal of animal wastes have become a major environmental policy issue in many countries. Nitrates in drinking water supplies and eutrophication of inland and coastal waters are especially of concern. High levels of nitrate in water may adversely affect human health as well as the metabolism of livestock.. Increasing concentrations of nitrates in groundwater, the primary source of drinking water in many regions, have been observed, notably in France. Agriculture is not the only source of nitrates in ground and surface waters but it is one of the most concern. There is widespread interest in implementing policies that will be more effective in protecting water quality without causing undue economic harm to agricultural producers. The two farming practices that most concern policymakers are the use of large amounts of fertilizer for crop growth and the disposal of livestock manure. Both materials are sources of nitrogen, which transformed into nitrate once in the soil. Nitrate that is not used by plants or transformed back into atmospheric nitrogen leaches through the soil or runs off into water supplies. Intensive livestock production is an important source of pollution, due to an insufficient area of land available to these farmers on which to apply manure. This is particularly relevant for pig production. The direct impact on the environment of the pig production is in some areas really severe. Along with an expansion of production, there have been significant structural changes in the pig sector. Pig farming has became more intensive with fewer farms producing a larger number of pigs and more specialised with feed obtained from off-farm sources and often with very little land. Developments in production technologies have allowed significant productivity gains, particularly for large-scale producers. Pig farming have became more regionally concentrated. A major factor encouraging the development and uptake of productivity enhancing technologies has been the intense competition in the meat market and the long run decline in real prices received by farmers, which in turn is driven by productivity improvements. In response to high nitrate levels in water supplies, the European Union passed its Nitrate Directive in 1991. Its objective is to limit the amount of nitrogen remaining in the soil as a residual after uptake by crops. The Directive limits the spreading of organic nitrogen per farm to 170 kilograms per hectare in the vulnerable zones. The implementation of this Directive have been organized by each European member states. They have defined a set of constraints relevant for their own country on the use of nitrogen fertilizer, the numbers of livestock, and the storage and disposal of manure. In France, this implementation is effective since 1993. The purpose of this paper is to analyze the impact on farm performance of the Nitrate Directive, mainly the mandatory threshold on the spreading of organic manure. In a first step, we construct a non parametric frontier model that explicitly integrate the individual constraint from the Nitrate Directive on manure spreading. This individual threshold is considered as a productive rights allocated to each farmer. In regards with the activity of each farm some of them are highly constraint while some other are not. The question is to consider how producer will individually adapt their production activity both to fulfil the regulation constraint and to maintain their activity in a good economic performance level. Then, we built a model at a regional level that integrate the same constraint on the spreading of organic nitrogen which allows for trading of productive rights among producers. The objective is how producers from a specific region will fulfil the regulation if they 'collectively' manage their productive rights. The aim is to evaluate the impact of a collective manure management on the individual and regional performance of farms through potential gains that can be made from a trading of productive rights, i.e., of spreading rights. The nonparametric frontier model is constructed that explicitly introduces the constraint of 170 kg/ha of organic nitrogen. The directional distance function is used for the measurement of efficiency. It allows for considering the joint production of good and bad outputs in the measurement of efficiency. The path to the productive frontier is in a direction that increases good outputs and decreases bad outputs, i.e., nitrogen surplus. The empirical application is on a sample of farms with a pig farming activity drawn from the Farm Accountancy Data Network (FADN) for France. The data set covers production years from 1996 to 2000. Results suggest potential gains that can be achieved if productive rights trading or exchange are used instead of individual rights to achieve the same total spreading target. References Brännlund R., Chung Y., Färe R., Grosskopf S. (1998), Emissions Trading and Profitability: The Swedish Pulp and Paper Industry, Environmental and Resource Economics, 12, 345-356. Chung Y.H., Färe R., Grosskopf S. (1997), Productivity and Undesirable Outputs: A Directional Distance Function Approach, Journal of Environmental Management, 51, 229-240. Färe R., Grosskopf S., Li S.K. (1992), Linear Programming Models for Firm and Industry Performance, Scandinavian Journal of Economics, 94, 595-608.

The Cost Implications of Waste Minimisation: Factor Demand Competitiveness and Policy Implications:

Wendy Chapple - Catherine Morrison Paul - Richard Harris

The issue of waste generation and its environmental implications raises important questions about firms’ responsibilities. Waste reduction from either regulations or voluntary action is costly to firms, whether accomplished by “end of pipe” waste management or process solutions involving changes in input use. We evaluate county-level patterns of production costs and waste generation in U.K. manufacturing, using a cost function approach that includes waste as a “bad output.” Using a Leontief quadratic cost function, we estimate overall, input-specific, and marginal production costs of waste reduction, through shadow value, input demand, and marginal production cost elasticities. Of particular interest are the i) substitution effects between inputs and the implications for future Capital/ Labour ratios and ii) cost implications of waste reduction and the implications for competitiveness of UK manufacturing industries. For most counties we find significant costs associated with waste reduction, which arise from increased intermediate materials use but lower labour and capital demand. Therefore we disaggregate the inputs further to analyse whether there has been a substitution effect from raw to more processed materials in the production process, thus leading to these trends in input demand. We find that in certain types of county increasing waste reduction will ultimately lead to a lower Capital/Labour ratio, which could have significant implications for future productivity growth. We also assess the plant level (internal) characteristics and spatial (external) characteristics as drivers of waste minimisation costs.

Productive efficiency models for environmental performance measurement: a literature review:

Ludwig LAUWERS - Hilde Wustenberghs - Lieve De Cock

In order to account for environmental side effects of production processes, several attempts have been made to incorporate them in efficiency and productivity analysis (EPA) models. Based on an extensive literature review, this paper inventories and tries to categorise the existing approaches and models for the environmental adjustment of the conventional productive efficiency measurements. The literature review starts with some “pre- Pittman” papers, the environmentally adjusted multilateral productivity index by Pittman (1983) and the concept of weak disposable bad output, introduced by Färe et al. (1989). The literature review further passes through several applications and further methodological adjustments and finally shows that the various efforts since more than twenty years, yield wide variety of models. Next, this paper tries to categorise these environment-incorporating efforts according to their finality, the way the unwanted side-effects are incorporated, the type of EPA method, and the type of performance measures. The main finality has been, for a long time, the derivation of shadow prices, which then served as proxies for pricing negative externalities in enhanced productivity measures. Alternative finalities are efficiency scores. There are three main ways how unwanted side-effects are incorporated in the EPA models: as a weak disposable undesirable output, as an input or as a balance feature of inputs and outputs. Similar to their conventional counterparts, EPA methods incorporating environment issues fall into three broad categories: on the one hand the adjusted index methods and, on the other hand, the enhanced non-parametric (DEA, Data Envelopment Analysis) and parametric frontier analyses (SFA, Stochastic Frontier Analysis). Finally, performance measures may result endogenously from the models or may be calculated exogenously with the model outcomes. This wide range of models yields various results, which may sometimes allow differentiated, and thus enriching, interpretations (e.g. whether environmental regulations are binding or not), but may also imply nonsense information. In particular, some models do not show coherence with biophysical reality. Färe Rolf, Shawna Grosskopf, C.A.K. Lovell, and Carl Pasurka. (1989). “Multilateral productivity comparisons when some outputs are undesirable: A non-parametric approach.” The Review of Economics and Statistics 71(1), 90-98. Pittman, R.W. (1983). “Multilateral productivity comparisons with undesirable outputs.” Economic Journal 93 (372), 883-891.

6a

Environmental Externalities and Regulation Constrained Cost productivity Growth in the U.S. Electric Utility Industry:

Diego Prior - Gerald Granderson

This paper examines productivity growth and its decomposition into various components for U.S. electric utilities, when accounting for the production of multiple outputs (good and bad), and the regulations the utilities face. The use of fossil fuels in producing electricity contributes to the joint production of good (electricity) and bad (carbon dioxide, nitrogen oxide, and sulfur dioxide) outputs. Reductions in the quantities of the bad outputs can be achieved via reductions in electricity production, employing inputs to reduce the quantities of the bad outputs, or using technologies that contribute to reductions in the bad outputs. Implementing policies to reduce the quantities of the bad outputs, which has benefits for the environment, may lead to higher production costs on electric utilities, and higher prices on consumers. Electric utilities in the U.S. are subject to various state and federal regulations. Regulatory authorities may regulate the rate of return the utilities can earn on their investment. Some state agencies allow the utilities to file for fuel adjustment clauses, which allow the firms to more easily pass fuel price increases onto consumers (compared to undergoing formal rate case hearings). The combinations of such regulations can affect the input choices electric utilities select to produce their desired output levels, in that the input prices of capital and fuel that firms base their choices on may differ from their market prices (Averch-Johnson type effects). Measurements of productivity growth and its decomposition that do not account for the production of bad outputs, and the regulations firms face, may affect productivity growth and its decomposition into various components. In this paper, we follow the work of Ball, Färe, Grosskopf, and Zaim (2004) to develop a rate of return constrained Malmquist Cost Productivity (RMCP) Index. Firms select the quantities of inputs to minimize production cost, subject to quantity and regulation constraints. The RMCP index uses a non-parametric linear programming approach to measure regulated cost efficiency, as defined by Färe and Logan (1992). The standard (unregulated) Malmquist Cost Productivity (SMCP) Index, which Ball, Färe, Grosskopf, and Zaim (2004) developed, can be obtained from the regulated productivity index. A comparison of the regulated and standard indices allows for an examination of the effects of regulation on productivity growth and its decomposition. We can also measure productivity growth and its decomposition when not accounting for the production of the bad output. Derivation of the SMCP index can allow for a determination of whether electric utilities are technically and/or allocatively inefficient outside of regulation. The data sample is a panel of 34 U.S. investor-owned electric utilities from 1992 to 2000.

Measuring the Efficiency and Returns to Scale of Publicly Owned Electricity Distribution Systems in the United States: An INput Distance Function Approach:

David Saal

Over 2000 publicly owned electric power systems operate in the United States. While there are several relatively large public power systems that operate in cities such as Los Angeles, Seattle, and San Antonio, most serve less than 25,000 customers. Given their small scale relative to investor owned utilities, some have argued that the continuing existence of these small public power systems perpetuates an inefficient industry structure. However, as relatively few studies of US electricity distribution have actually included public power systems, there is little evidence to support this assertion. This paper will therefore employ stochastic frontier techniques to estimate an input distance function model of electricity distribution using data available from FERC, EIA, and Platts UDI Electricity datasets. These models will provide evidence not only with regard to the returns to scale which characterize publicly owned distribution systems, but will also allow an assessment of whether smaller systems are less efficient than larger systems. These results will allow meaningful conclusions to be drawn with regard to the appropriate scale of public power systems.

Economies of Scope and Diversification in the Electricity Industry: A Non Parametric Frontier Approach.:

Pablo Arocena

Over the last decade the electricity industry has undergone major liberalising reforms worldwide. Privatisation and substantial regulatory reforms were implemented with the aim of introducing competition in electricity markets. However, in many countries the pro-competitive potential of these reforms has been largely undermined by the high level of market concentration and vertical integration of electric utilities. Further, substantial market power of incumbents often makes difficult new entry. Hence, some regulators seeks to promote competition through policies of market restructuring. Basically, these policies implies the splitting up of existing firms in order to achieve a more competitive market structure, which means the reduction of both horizontal and vertical disintegration. However, such a restructuring of the electric power industry raises questions about the possible sacrifice of scale and vertical economies. Incumbents claims the existence of strong scope and scale economies, because of technological interdependency between stages, the use of common inputs, the need for stage coordination and the market transaction costs. In other words, high integration economies would make inefficient the supply of power by means of smaller and/or specialized firms. The purpose of this paper is to estimate such economies through a non parametric frontier approach. In this respect, our study differs from most previous work on this topic, which traditionally use non frontier techniques to estimate some form of cost function. We use a data set of Spanish electric utilities throughout 1991-1996 period. Our preliminary results show substantial economies of scope and diversification. Particularly, we find that (i) vertical separation between generation and distribution would raise electricity costs about 10 percent on average; (ii) cost savings due to the diversification of power generation accounts for 8 percent on average; (iii) economies of scale are not important provided that scope and product-mix are preserved. In fact, costs could be reduced by 5 percent by dividing each of the existing firms down the middle.

Decomposition of Cost Efficiency and its Application to Japan-US Electric Utilities Comparisons:

Miki Tsutsui - Kaoru Tone

Technology and cost are the wheels that drive modern enterprises; some enterprises have advantages in terms of technology and others in cost. Hence, the management is eager to know how and to what extent their resources are being effectively and efficiently utilized, compared to other similar enterprises in the same or a similar field. This paper presents a new formula for decomposing the actual supply cost into the optimal cost and three types of losses due to input inefficiency to be dependent on technical inefficiency, input price differences, and inefficient cost mix. --Methodology Technical efficiency is measured using the traditional CCR model within the supposed technical production possibility set. The technical efficient input cost, which is obtained by multiplying the CCR optimal input by the input factor price, cannot be reduced further by reducing the amount of input. However, taking into account the differences in input prices under the situation that the unit prices might differ from DMU to DMU, the cost can be reduced by reducing the input factor prices. The traditional cost and allocative efficiencies, which assume the given uniform input prices, suffer from a critical shortcoming if the unit prices of the inputs are not identical among DMUs in the actual economy, as pointed out by Tone (2002). Therefore, using the obtained CCR optimal input value, we construct a cost-based production possibility set and solve the New Tech and New Cost models on this set. This enables us to obtain two efficiency indices, i.e., the price efficiency index and the global allocative efficiency index. The former reflects an efficiency index regarding the differences in input unit prices, while the latter evaluates the efficiency of the input cost mix. Using this methodology, we can decompose the actual cost into three losses and the minimum cost as follows: Actual Cost = Minimum Cost + (Loss due to Technical Inefficiency + Loss due to Input Price Difference + Loss due to Inefficient Cost Mix). --Application to the Electric Power Companies We applied this formula for comparing cost efficiency between Japan and the US electric power companies in order to clarify the main reason for comparatively higher electricity price in Japan. For DMUs, we consider 19 vertical-integrated electric power companies (9 Japanese and 10 US) from 1992 to 1999. According to the results, the Japanese average actual supply cost is nearly twice that of the US average actual supply cost, although it has been declining over the years. It is clearly indicated that the losses due to the differences in the input price levels are much larger than those of the others in Japan as well as the losses in the US. In contrast to the price inefficiency loss, there are no significant differences in losses due to technical and allocative inefficiencies between Japan and the US. This result implies that the comparatively higher electricity price in Japan is caused by higher input factor prices, not technical inefficiency.

6b

Where have all the data gone? Stochastic production frontiers with multiply imputed German establishment data:

Uwe Jensen - Susanne Raessler

In this paper, stochastic production frontier models are estimated with German establishment data. Our data are taken from two waves (2002 and 2003) of the Establishment Panel of the Institute for Employment Research of the Federal Labor Service. We are confronted with 4 % to 15 % of missing values particularly in the most important variables (output, capital and labor), a typical situation in empirical research. Ignoring this would reduce the complete data records available considerably. Whereas information from 18447 observations from the panel waves of 2002 and 2003 is collected in principle, only 13969 observations of them can be used when inference is based only on the complete cases. Ignoring the missing values would certainly lead to lower precision of the estimates. And the question arises whether the remaining data are still representative for the population of interest. If not, the resulting test statistics are no longer valid and the resulting estimates may be biased. Biases can be expected to occur particularly in the establishment''s inefficiency estimates of the stochastic production frontier. Because frontier estimates depend on the extreme efficient establishments in the sample and because the inefficiency estimates are derived from the estimation residuals, the latter are extremely sensitive to any kind of misspecification in the model -- see e.g. Jensen (2005). That is why it is the aim of this paper to demonstrate in an empirical application the dangers of ignoring missing data or the gains of properly imputing them when estimating a stochastic production frontier with establishment data. We decided to use a multiple imputation procedure to fill the missing values. Using a single imputation technique such as mean imputation, hot deck, or regression imputation, in general results in confidence intervals and p-values that ignore the uncertainty due to the missing data, because the imputed data were treated as if they were fixed known values. Thus, basing standard complete data inference on singly imputed data will typically lead to standard error estimates that are too small, t-values that are too significant, and confidence intervals that undercover -- see, e.g., Rässler et al. (2003). Multiple imputation, introduced by Rubin (1987), is a Monte Carlo technique replacing missing values by, e.g., m = 5 simulated versions. Each of the imputed and thus completed data sets is first analyzed by standard methods. Then, the results are pooled to produce estimates and confidence intervals that embed the missing data uncertainty. The theoretical motivation for multiple imputation is Bayesian. Its computation has become tractable with the help of Markov chain Monte Carlo (MCMC) techniques. For the creation of the multiple imputations, we use the stand alone software NORM provided for free by Schafer (1999). One further contribution of our paper is the suggestion of a double imputation of the capital variable. With cross-section data covering two years, we decided to proxy capital by the replacement investment in the current year. Of course, this choice implicitly assumes that capital is replaced uniformly and sufficiently, among others. An alternative would be to approximate capital by the average replacement investment of several years. But since firms are born and die, this approximation would lead to even more missing values or firms. Replacement investment is one the variables suffering from many missing values. This problem is soothed by multiple imputation. But another problem is that a large part (7888 of 18447) of the values on investment in the sample are zero. There is some evidence that many of these firms are simply not able or not willing to provide exact non-zero investment numbers. That is why we decided to multiply impute these zeroes as well. Following Reifschneider and Stevenson (1991), an inefficiency submodel is added to the stochastic production frontier. The production function has translog form in capital and labor. Production function and inefficiency submodel contain many additional variables. Estimation has been performed with LIMDEP 8.0. Using only the observed data is compared with 2 multiple imputation approaches: Filling all missing values by multiple imputation but maintaining the zeroes in the capital variable (MIC0) and filling all missing values and the zeroes in the capital variable. Since we are working with real data and not with simulated data, we are not able to say which results come closer to the truth. Nevertheless, particularly in the inefficiency submodel, working with multiply imputed data seems to reveal some interesting and plausible results which are not available with missing observations. And, summarizing the performance of the two multiple imputation approaches, the MIC0 approach suffers from the serious drawback of counterintuitively producing an insignificant labor parameter in the production function. So, we have a small but significant preference for the results obtained with multiple imputation where the capital zeroes are imputed as well. References: Jensen, U. (2005): Misspecification preferred: The sensitivity of inefficiency rankings, Journal of Productivity Analysis 23/2, to appear. Rässler, S., D.B. Rubin and N. Schenker (2003): Imputation, in: Bryman, A., M. Lewis-Beck and T.F. Liao (eds.), Encyclopedia of social science research methods, Sage. Reifschneider, D. and R.E. Stevenson (1991): Systematic departures from the frontier: A framework for the analysis of firm inefficiency, International Economic Review 32, 715 - 723. Rubin, D.B. (1987): Multiple imputation for nonresponse in surveys, Wiley (New York). Schafer, J.L. (1999): Multiple imputation under a normal model, version 2, Software for Windows 95/98/NT, http://www.stat.psu.edu/~jls/misoftwa.html.

Economics and Olympics: An Efficiency Analysis:

Alexander Rathke - Ulrich Woitek

Applying stochastic frontier analysis, we estimate the technical efficiency of countries in the production of Olympic success since the 1950s. Following Bernard and Busse (2004), population and GDP are used as inputs. We also control for host country advantage and for differences in public support for sports. The results show that the spread of technical efficiency is very wide across countries and across sports. These differences can be seen as caused by different production technologies. Examples for technological differences are training methods, organization, or culture, but also illegal practices like doping. Using the method proposed by Battese and Coelli (1995), we are able to judge the relative importance of these components. Reference Battese, G. E. and Coelli, T. J. (1995), ”A Model for Technical Inefficicency Effects in a Stochastic Frontier Production Function for Panel Data”, Empirical Economics, 20, 325-332. Bernard, A. and Busse, M. (2004), ”Who wins the Olympic Games: Economic Resources and Medal Totals”, Review of Economics and Statistics 86, 413-417.

Efficiency and Convergence in Developing Countries:

CAMILLA MASTROMARCO

Productivity growth is viewed as having three components: technological change, technological catch-up and a scale factor. Using a stochastic frontier model these components are calculated for 57 developing countries over the period 1960–2000. The empirical analysis follows three stages. The first stage tests for the statistical significance of the different hypotheses regarding augmenting factors and neutral technology (Hicks, 1932, Harrod, 1942 and Solow, 1969 neutrality). Technology is found to be non-neutral. Moreover, the specification of the production frontier with human capital affecting the quality of labour force fits the data better. Human capital affects total factor productivity through different channels (efficiency and augmenting labour force). In the second stage, the evolution of the entire distribution of the productivity components is analysed and a formal test for convergence is performed. With respect to regression analysis this approach is likely to be more informative (Quah 1996, 1997). The base of both the test and the visual analysis is the non-parametric kernel density estimator. Following Fan and Ullah (1999), Kumar and Russel (2002) a standard normal kernel is used to derive the test statistic. The first test analyses the importance of TFP and input growth. Both are found to be important for output growth. TFP growth is then further decomposed into technical change, scale effects and the contribution of efficiency. The results from the non-parametric test provide evidence that both innovation and scale effects are important components of TFP growth. Efficiency does not play an important role. In the final stage of the empirical analysis, the channels through which trade affects the components of output growth are analysed. The focus is on foreign direct investment, imports of machinery, import discipline, and export of manufacturing goods. All these channels have a positive effect on efficiency. The result is consistent with the predictions of endogenous growth models with trade (Grossman and Helpman 1991 and Rivera-Batiz, Romer 1991, Young 1991 and Barro and Sala-i-Martin 1995). Explanation: What is new? and Why it is important?: Following Temple, who states that 'a better understanding of what generates economic growth could make a huge contribution to human welfare' (Temple 1999, p.112), this study addresses the determinants of the catching-up effect toward the frontier (efficiency), viz. the forces that drive the growth of developing countries. Given that the focus of this study is the analysis of convergence in developing countries, a major role is expected to be played by technological diffusion, therefore increasing in efficiency (catch-up terms) with respect to technological changes (shifts of frontier). This article uses stochastic frontier methodology to estimate efficiency, technological change and scale effect. Both TFP and input growth are important for output growth. Efficiency does not play the main role in productivity growth of developing countries. It is well known that alternative specifications of the production function lead to ambiguous empirical evidence for competing theories of economic growth (Durlauf and Quah, 1998). Therefore, tests are performed to find the specification in line with the data under analysis. The neoclassical hypothesis concerning the neutrality of technological change is tested against the endogenous one of non-neutral technical progress. Non-neutral technical progress turns out to be the preferred specification. As a result, technical change shifts the frontier and changes the elasticity of substitution between the production factors. The role of human capital in the process of economic growth is not yet unambiguously determined (Islam 1995, p.1154). To address this important issue, the empirical analysis moves to the study of human capital. Evidence indicates that human capital affects growth through multiple channels. Endogenous growth theories (Romer 1992, Grossman and Helpman 1991 and Barro and Sala-i-Martin 1995) argue that open countries absorb technological advance easier than closed ones. Thus, a key role in technological diffusion is played by trade. The effect of trade on the convergence process and therefore on efficiency is investigated. Whether trade openness is beneficial or no to developing countries is a question debated since a long time. Empirical studies that try to determine the impact of openness on growth provide controversial results (Edwards 1992, 1998, Rodrik 1999). The importance is to analyse the transmission mechanisms through which trade affects growth and convergence (Tybout 1998, Wacziarg 2001). Up to now, there is no evidence in the literature on the relative importance of channels through which trade diffuses technology (Tybout 2000). The main contribution of this article is providing for the first time evidence on the macro level.

6c

A Re-examination of the technological catching-up hypothesis across and OECD industries:

Jean-Philippe Boussemart - Walter Briec - Isabelle Cadoret - et al. This study re-examines the catching-up hypothesis at the industry level across the main OECD leading countries, using panel data econometric models involving technological gap indicators calculated with a nonparametric distance function suggested by Färe et al. (1994). The results show that there is statistical evidence of a catching-up process at the industry level. Moreover, both tradables and nontradables sectors exhibit catching-up effects and technology adoption from abroad. This result complements the finding by Bernard and Jones (1996 a,b), Gouyette and Perelman (1997) and Hansson and Henrekson (1997) that there is no (or even a slow) catching-up effect in the manufacturing sector. Moreover, social capability indicators evaluated for each country show that ‘Non Europe’ and ‘Center Europe’ tradables sectors have a rather similar degree of inefficiency while ’North-Europe’ countries are less efficient for both tradables and nontradables. Lastly, both the cross country and the cross sectors dispersions of inefficiency levels are smaller for tradables sectors than for nontradables.

Productivity Analysis of Market and Environmental Regulations in China:

Shunsuke Managi - S. Kaneko

Successful economic and environmental policies can contribute to technological or efficiency improvements by encouraging, rather than inhibiting, technological innovation. Although a large number of studies have been made on the constituents of technological change, little is known about the empirical evaluation of policies that encourage productivity progress and/or regress in China. This paper contributes to the literature on productivity change in several ways. First, we apply a productivity approach to a province-level data set tracked from 1987 to 2001 in China to measure various components of total factor productivity (TFP) within a joint-production model of market and environmental outputs. This contributes to our understanding of the various components of total factor productivity change in China. In addition, this study contributes to better economic and environmental policy design for sustainable development in China by empirically estimating the role of economic and environmental management on market and non-market productivity. Our result for market output is consistent with the literature that there has been considerable TFP growth, while environmental managements have not effectively regulated wastewater, air and solid waste pollutants emissions over our study periods. We found significant negative impacts of pollution abatement and control expenditure (PACE) on market technological progress, although elasticity is small. This PACE, in contrast, positively affects to environmental productivity and technological progress as expected. In addition to the traditional formal regulation, we also find that information, so called informal regulation, affects the TFP and also to reduce the pollution.

The effects of TFP and its components on inequality:

Rogerio Cesar Souza - Fernando Garcia - Jorge Oliveira Pires The paper presents a new and alternative approach within the empirical literature dealing with inequality. It investigates to what extent the dynamics of income inequality is affected by the changes in the Total Factor Productivity (TFP) and its components: technical efficiency, technical progress, allocative efficiency and scale efficiency. The basic idea is that asset allocation and use, as well as production organization, directly influence personal and functional income distribution in countries. Starting with the Stochastic Frontier Analysis, we apply the TFP decomposition suggested by Bauer (1990) and Kumbhakar (2000), for a sample of 38 countries, during the 1970-2000 period. Then, we estimate the direct effects of technical efficiency, allocative efficiency, technical progress, and economies of scale on income inequality of the countries. The results show that allocative, technical and scale efficiencies are positively correlated with Gini. This means that variables associated to the performance of capital channel income to this productivity factor. Technological progress, in turn, works on the entire economies, with direct effects on labor productivity and wages, resulting in a more equitable income distribution. This largely explains why countries such as Brazil, Chile and Mexico still suffer from high levels of income inequality, in spite of economic growth they have enjoyed in the last 30 years.

On the Relative Innovation and Productivity Performance of Domestic and Foreign-Owned Enterprises: A Cross-Country Analysis:

Hans Lööf - Bernd Ebersberger

This paper examines the phenomenon of firms taken over by foreign owners. It has three novelties. First, the it examines not just productivity differences compare to domestically-owned firms, but the efficiency with which it can utilize internal R&D, knowledge flow within the group and scientific-, vertical- and horizontal innovations system to produce patents, new products and labour productivity. Second, in order to incorporate corporate governance in the analyses, the foreign-owned firms are distinguished between three categories of corporate styles (Nordic firms, Anglo-Saxon firms and Continental European and other firms), and the domestically owned firms are separated between multinationals and uninationals (enterprises belonging to a group with only domestic affiliates). Third, the authors are using a uniform set of extensive firm-level data and a uniform econometric approach is applied on observations in four neighbouring countries. The study includes the entirely business sector including both manufacturing and service firms. A main justification for the study is the growing importance of multinational firms and foreign direct investments (FDI) which have grown significantly faster than the trade flow among the most developed countries during recent decades. Between 1990 and 2001 production in enterprises located outside the country of residence of the owners increased from 6 to 11 percent of world output. Export from foreign affiliates of multinational corporations represents more than a third of world trade. However, the literature suggests that the rising trend of foreign direct investments to a large extent reflects increasing and acquisition and mergers in general, rather than a more internationalized economy. The United Nations (2000) reports that the cross-boarder share of total acquisition and mergers has been relatively constant since the late 1980s. Moreover, though evidence clearly suggest a growing significance of overseas R&D activities by MNEs, the firms have not internationalized their innovative activity proportionally to the growth of their production activities (Patel and Vega 1999, Kumar 2001). FDI, which are distinguished from portfolio investment in that it implies a greater degree of foreign control and knowledge transfer (e.g. regarding production, management and marketing), can be divided between acquisitions and investment in new companies or units (greenfield investments). The literature suggests that foreign acquisitions by far exceed the new establishments. U.S. data (Feliciano and Lipsey 2002) show that between 1988 and 1998, outlays for acquisitions accounted for 83% of outlays for acquisitions and new establishments. Swedish Institute for Growth and Policy studies (ITPS) presents more or less identical figures: During 1996 and 2000 acquisition accounted for 77% of the establishment of foreign ownership in Sweden and additional 6% was the result of mergers. This process raises concerns about the role that foreign-takeovers play for R&D, innovation, and productivity in acquired firms. Empirical regularities or stylized facts, which have emerged from a large number of mainly productivity comparison studies on domestic and foreign-owned firms, indicate both positive and negative effects. First, there is robust evidence that within countries, foreign owned firms almost always pay higher wages than domestically owned firms. Second, foreign-owned firms generally have higher productivity then local firms. Third, the evidence for knowledge spillovers from foreign-owned firms to domestic firms is mixed. The literature suggests some alternative explanations for the differences in performance between domestic and foreign-owned firms. Some studies argue that only firms with superior technology or superior productivity are candidates for acquisitions or mergers. Other studies find that FDI investments are oriented toward high productivity sectors. A third finding is that acquisitions and mergers have a positive impact on efficiency of firms per se. Moreover, if the hypothesis on FDI as a strategy to exploit technological activities created within home countries is correct, a higher productivity due to scale economics or other competitive advantages should be expected. It should also be noted that possible performance gap between foreign-owned and domestic firms often cannot be easily examining due to heterogeneity of data sources available. In addition here are also difficult methodological challenges. The basic issue to be explored in this paper is whether firms taken-over by foreign owners perform differently regarding R&D, innovation and productivity than domestic firms ceteris paribus, or every thing else being equal? Since we cannot observe the counterfactual state 'What would the firm behavior and performance have been had the companies not been acquired by foreign owner?' we approximate this state by the two categories of domestically-owned companies which are not foreign owned. The possible differences between domestic and foreign-owned firms are studied with respect to gaps regarding propensity to innovate, R&D and other innovation expenditures, embeddeddness in national innovation systems, innovation output in terms of patent and sales, and finally labor productivity. Among the determinants to firms’ innovation and productivity performance we have different categories of ownership, capital intensity, export and market orientation, whether the firms are focusing on process or product innovations, the firms’ R&D history, human capital and sector classification. The theoretical literature suggests some alternative and complementary hypotheses as to why firms invest in R&D activities abroad. One has to do with opportunities to exploit technological activities created within the home country. A second hypothesis concerns the exploitation of technological advantages of the host country. A third hypothesis emphasizes the increasing complexity and specialization of technology. The different explanations might have varying impact on the behaviour and performance comparison between domestic and foreign-owned firms. The study is based on a large sample of 5 186 firm level observations in a cross-country comparison over the four Nordic countries Denmark, Finland, Norway and Sweden. The proportion of foreign-owned firms has a range from 23% in Finland to 32% in Sweden. The average for the counties is 28%. The econometric analysis account for selection issues - the likelihood that R&D firms constitute a particular group of companies – and idiosyncratic factors such as firm size, market orientation or sector as well. In a sensitivity analysis the study also examine possible simultaneity bias and considers a four-equation model that relates various determinants to research, research to innovation output and innovation output to labor productivity. Our main results give strong support for the hypothesis that domestic multinationals are using the home country for developing technological capacity that is subsequently exploited in affiliates abroad. Correspondingly, the innovation and productivity performance in foreign-owned multinationals are partly returns on activities created in their home countries. This explains the “productivity paradox” that can be observed in Finland, Norway, Sweden and partly also in Denmark.

6d

No abstracts available.

7a

IMPACT OF TRADE AREA ENVIRONMENT ON BANK’S COMPARATIVE ADVANTAGES:

Aude hubrecht - Leleu hervé

This paper analyzes the relationship between the comparative advantages of bank branches and the trade area environment. Bank branches are points of sale whose trade environment influences their activities and performance. Comparative advantages are defined, for each output mix, by the strict dominance of a production technology in a specific trade area over the production technologies of other environments. Using Shephard’s output distance functions on a sample of 728 bank branches, we compare the production technologies for different output mixes and different trade environments. We show that none of the production technologies strictly dominates the others and none of them is strictly dominated. Therefore, each trade area benefits from comparative advantages that we try to highlight. Finally, we evaluate the performance of the central banks regarding their ability to provide the right incentives on output mixes to their bank branches so that the latter may benefit from their comparative advantages.

Allocative Efficiency Measurement Revisited--Do We Really Need Input Prices?:

Oleg Badunenko - Michael Fritsch - Andreas Stephan

The traditional approach to measuring allocative efficiency is based on input prices, which are rarely known at the firm level. This paper puts forward a new approach, which proves that the Farrell output--oriented distance to the frontier in the profit--technical efficiency space indicates allocative efficiency. We apply this new approach to a unique panel of about 35.000 firms from the German Cost Structure Census. The results show that the heterogeneity of German manufacturing firms with regard to allocative efficiency is considerable, and that the potential benefits from improving allocative efficiency are noteworthy. Most strikingly, we find that smaller firms are on average more allocatively efficient than larger ones.

The Importance of Inward and Outward FDI for Productivity Growth in the UK:

Nigel Driffield - Michael Henry - James Love

There is increasing theoretical and empirical evidence that foreign direct investment (FDI) may be motivated not by the desire to exploit some competitive advantage possessed by multinationals, but to access the technology of host economy firms (see Cantwell and Janne, 1999; Serapio and Dalton, 1999; Kogut and Chang, 1991; Fosfuri and Motta, 1999). We extend this developing literature in several ways. First, following van Pottelsberghe de la Potterie and Lichtenberg (2001) we incorporate the importance of outward as well as inward FDI as a source of productivity growth. However, whereas the latter two authors infer motivation ex post from the spillover effects of inward and outward FDI respectively, we infer motivation ex ante. Second, we extend recent work by Driffield and Love (2002, 2003) and Driffield and De Propris (2005) to include, not only the notion of technology sourcing FDI, but also to consider other motives for FDI as an explanation for the heterogeneous effects of FDI on UK productivity. In those papers the authors make the distinction between technology exploiting FDI (TEFDI) and technology sourcing FDI (TSFDI), using R&D intensity differentials between host and source sectors as the means of ascribing motives. This paper builds on those earlier works by also identifying FDI to and from high and low labour cost locations (at the sectoral level) and then testing for productivity effects generated merely through moving low value added activities to low cost locations. The possibility that FDI into high and low cost locations (relative to the host country) generates differential productivity spillover effects has hitherto been ignored in the applied literature. Additionally, the paper seeks to determine whether there are threshold effects in the nexus between inward FDI and domestic productivity mediated by labour quality. That is, whether a threshold level of labour quality exists whereby the productivity spillovers from FDI above and below the critical threshold level differ. Equally, the paper goes on to examine differences in the threshold effects across the different motivations for FDI. We use a panel of FDI flows across OECD countries and manufacturing sectors between 1984 and 1995 to test whether these contrasting motivations, as well as labour quality, influence the effects that FDI has on domestic total factor productivity. We employ both stochastic frontier analysis within a standard fixed effects framework, and a GMM systems approach to the estimation of productivity and spillovers. The preliminary results from our GMM estimations highlight the importance of not treating inward investors in the UK as a homogenous group of firms when seeking to evaluate spillovers, and also appropriately dealing with endogeneity within these models, especially within the externality term where the number of instruments is limited. For example, when we analyse FDI inflows to the UK, we find that only when the motivation for FDI is to exploit some ownership (normally knowledge based) advantage are there positive and significant productivity spillovers from FDI. This outcome, which is consistent with foreign multinationals from sectors with a greater R&D intensity than that of the host country (UK), provides support for the dominant theoretical perspective on the determinants of FDI (see Dunning, 1979). Interestingly, our results indicate that these foreign firms have a positive impact on UK TFP growth notwithstanding the higher unit labour costs (compared to the home country) prevailing in the host country. In contrast, when the motivation for FDI appears to be to source some technological advantage from UK firms, inward FDI has a negative impact on domestic TFP growth. This effect becomes particularly marked in the low cost areas of the UK. The explanation for this is that technology sourcing FDI introduces no new technology to the UK, and hence no technology spillovers, while at the same time increasing competition. This generates the associated crowding out effect that is well understood in studies of productivity externalities following Aitken and Harrison (1999), though seldom analysed in detail. With respect to outward FDI, when this factor is treated as a homogenous variable it is shown not to generate significant productivity spillovers to the UK. However, when we distinguish between the different motivations for FDI, we find that FDI which is motivated by sourcing some technological advantage in the host country generates positive productivity spillovers to the UK. This finding is consistent with that of van Pottelsberghe de la Potterie and Lichtenberg (2001). In an advance on the findings of previous researchers we also uncover evidence which shows that outward FDI to low cost locations also leads to productivity growth at home. This however cannot be seen as ‘technology sourcing’ as it has been ascribed elsewhere, but domestic productivity growth being generated through a ‘batting average’ effect as low value added activities are moved elsewhere. Thus by accounting for the different motivations of FDI, and not simply reporting an aggregated average effect, our findings have gone some way in providing an explanation for the apparently contradictory empirical results reported in the large literature concerning the impact of FDI on productivity growth, as discussed at length by Görg and Strobl, 2001. Key Words: FDI motivation, technology sourcing, productivity spillovers. JEL: F23, O31 (* corresponding author. E-mail m.henry@aston.ac.uk) References: Aitken, B. J. and Harrison, A.E. 1999. ‘Do domestic firms benefit from direct foreign investment? Evidence from Venezuela’, American Economic Review, vol. 89 (3), 605-618. Cantwell J and Janne O (1999) ‘Technological Globalisation and Innovation Centres: The Role of Technological Leadership and Location Hierarchy’, Research Policy, 28, 119-44. Driffield N and Love J H (2002) ''Does the Motivation for Foreign Direct Investment Affect Productivity Spillovers to the Domestic Sector?'' Aston University Working 0202; available at http://research.abs.aston.ac.uk/working_papers/0202.pdf Driffield N and Love J H (2003 ) ‘Foreign direct investment, technology sourcing and reverse spillovers’ The Manchester School, 71, 659-72. Driffield, N. and De Propris, L. ‘FDI, clusters and technology sourcing.’ The Cambridge Journal of Economics, forthcoming Dunning, J.H. (1979) ‘Explaining patterns of international production: In defence of the eclectic theory’. Oxford Bulletin of Economics and Statistics, Vol. 41, 269-95. Fosfuri, A. and Motta, M. (1999) ‘Multinationals without advantages’, Scandinavian Journal of Economics, vol. 101, 617-30. Görg H and Strobl E (2001) ‘Multinational companies and productivity spillovers: a meta-analysis’ Economic Journal, 111, F723-F739. Kogut B and Chang, S J (1991) ‘Technological capabilities and Japanese foreign direct investment in the United States’, Review of Economics and Statistics, 73, 401-13. Serapio MG and Dalton D H (1999) ‘Globalization and Industrial R&D: An Examination of Foreign Direct Investments in R&D in the United States’, Research Policy, 28, 251-70. van Pottelsberghe de la Potterie, B. and Lichtenberg, F. (2001) ‘Does foreign direct investment transfer technology across borders?’ Review of Economics and Statistics, 83, 490-497.

7b

The use of bootstrapped Malmquist indices to reassess productivity change findings: Application to a sample of Polish farms:

Laure Latruffe - Kelvin Balcombe - Sophia Davidova

This paper applies Simar and Wilson’s (1999) bootstrapping procedure for constructing confidence intervals for Malmquist indices calculated with Data Envelopment Analysis (DEA), and correcting the indices for bias. In a second stage, in order to explain the farm-specific change in productivity, both the bias-corrected Malmquist estimates and their standard deviations are employed in a new form of heteroscedastic panel regression, using a maximum likelihood procedure which uses the DEA standard deviations of the Malmquist indices. Polish farm-level data between 1996 and 2000 are used. The final sample consists of 250 farms. A multi-output multi-input DEA model is employed. The three outputs are livestock output, crop output and other output, and the four inputs are land, labour, capital and intermediate consumption. 2000 bootstrap iterations were performed and the 95-percent confidence intervals were constructed. Results show that Polish farm productivity decreased by 2 percent on average between 1996 and 2000. After bias-correction, it appears that the productivity regress is even greater, at 4 percent. As for the confidence intervals, they are relatively large, about 0.50 on average. This indicates that the non-biased-corrected Malmquist indices alone are not sufficient to unambiguously identify farms that have experienced significant progress or regress, that is to say, whose index is significantly different from 1. For example, in both 1996-1997 and 1999-2000, 1 is included in the confidence interval of 82 percent of the farms, indicating insignificant change in productivity for these farms. By contrast, the (non-biased-corrected) Malmquist indices identified about 0 and 1 percent farms with insignificant changes, respectively. In the second-stage consisting in the new form of heteroscedastic panel regression, the share of hired labour and the share of other income in total income are found to be significant sources of productivity progress. The comparison with the results from the standard Ordinary Least Squares estimates reveals differences, in terms of significance (fewer parameters are significant in our new form of regression) and sign of the parameter of the share of other income. By testing for the significance of the heteroscedasticity component in our new model, we find that the integration of the DEA standard errors is significant in explaining a proportion of the variance in the error term.

Relative Efficiency of Service Centres of Sickness Funds: A DEA-Bootstrap Approach:

Matthias Staat

I analyze data on 124 service centres of a German sickness fund. The centres are located across Germany and provide support to customers insured with the sickness fund and at the same time try to canvass new customers. At the time of the analysis, the sickness fund was in the process of restructuring its branches and was interested in identifying efficient branches that could serve as role models for others. The potential reduction of inputs that could be achieved without jeopardizing service quality was also of interest. Therefore, the data were analyzed with an input-oriented Data Envelopment Analysis model The data contained a host of parameters which included numbers of staff as well as number and type of customers, the quality of the location of the branch, its opening hours and every possible aspect of service quality. In addition, there was information on cancellations of policies as well as from customer interviews about parameters such as the customer’s likelihood of a referring somebody to the sickness fund. The premium rates of other local competitors were considered to be indicators for the competitive pressure. Branch management claimed that these parameters, many of them outside its control, did have an impact on the performance of its branch. Since it is not possible to use a huge number of parameters with a limited number of observations, a bootstrap regression of the efficiency scores from a base model on several environmental parameters was carried out. Only the price-performance ratio index from a customer survey did have a significant effect on the score. Therefore, this parameter was included into an extended model. With the final specification, 23 branches were found to be efficient. The average efficiency of the braches was 83%. According to the results obtained, one fifth of the cancellations of policies could have been avoided had all branches serviced their customers efficiently.

A Two-Stage Efficiency Analysis of Portuguese Municipalities:

Paulo Daniel Vieira - Elvira Silva

Portuguese municipalities are responsible to a large extent for the definition and execution of local public policies. Given the contractionary fiscal policy, the limited fiscal autonomy of municipalities and the local government budget law that constraints the borrowing capacity of municipalities, Portuguese local governments face new challenges that consists in finding financial resources necessary to pursue their public services provision. In this context, the efficient allocation of resources becomes a particular relevant issue. The purpose of this paper is to investigate cost efficiency of a sample of 166 municipalities during the year of 2000. The analysis is conducted in two stages. In the first stage, cost efficiency scores are generated using Data Envelopment Analysis. In the second stage, single bootstrapping is used to explain the differences in efficiency scores through a set of variables that characterize the environment within which the activity of municipalities are developed.

7c

The Analysis of Eco-efficiency in an Input-Output Framework:

Mikulas Luptacik - Bernhard Boehm

Eco-efficiency is characterised by production of goods and services with minimal resource use and generation of waste and other emissions of pollutants. Let the production possibility frontier of the economy be determined by the input-output model, primary inputs, pollution generation and abatement, and final demand. The degree by which a netoutput vector, for given primary inputs and environmental standards, could be extended, can be considered as measure of eco-inefficiency. Equivalently this could be also be achieved by a reduction of primary inputs for given environmental standards and given final demand. We have to take into account that desirable outputs are strongly disposable while undesirable ones are only weakly disposable meaning that their reduction can only be achieved by a reduction of desirable outputs or an increase of primary inputs. In the second part of the paper we propose another approach based on the construction of a production possibility frontier. For this purpose a multi-objective optimization problem with maximisation of final demand for each commodity subject to constraints on the primary inputs is formulated. Then, using data envelopment analysis with the production possibility frontier as the standard envelope, the ecoefficiency of an economy will be estimated. As the main result of the paper it can be shown that the eco-efficiency values for both models coincide. Because of the different model structure and variables useful additional insights in performance can be obtained. In view of the sectoral disaggregation of the economy we also propose a new “slack”- based measure of eco-efficiency which goes beyond the traditional proportional approaches to the changes in final demand or primary inputs. Thus, we are able to take into account the changes in the structure of final demand and the composition of primary inputs. A demonstration of the viability of these methods and their relationships will be given using Austrian input-output and NAMEA data.

Environmental Bads and Nonparametric Modelling: Clearing some confusions:

Atakelty Hailu

There has been a growing interest in the measurement of productivity performance in ways that take environmental performance into account. An increasing number of business firms are recognizing the pecuniary and non-pecuniary benefits of improved environmental management and the effective signalling of the same to stakeholders. Socio-economic analysts have also been looking for measures of performance that take into account the environmental effects of economic activity. Both parametric and nonparametric techniques have been used to incorporate undesirable outputs in productivity measurement over the last fifteen years. However, there still remains some confusion regarding the correct manner in which undesirable outputs should be included in DEA models. This paper evaluates the appropriateness of alternative disposability assumptions and clarifies some confusions in the literature including those relating to the normative aspects of pollution abatement.

7d

Research Peformance of Chinese Universities after a Decade of Reform: An Application of DEA Ying Chu NG - Ying Chu NG Research Performance of Chinese Universities after a Decade of Reform: An Application of DEA:

Ying Chu Ng and Sung-ko Li

Department of Economics Hong Kong Baptist University Kowloon Tong Kowloon Hong Kong, CHINA Fax: (852)3411-5580 E-mail: ycng@hkbu.edu.hk Abstract Much work have been done in accessing the research performance of higher education institutions in the US, UK or Australia. Similar study in developing countries has rarely been found. The present study attempts to fill this gap. Using Chinese universities as targets for analysis provides insights into the effectiveness of the education reform. The economic reforms implemented since 1985 facilitated universities in accessing extra resources for development and operation. However, an imbalance growth across regions in China was resulted. Accordingly, the performance of universities is further complicated by this regional effect. With both education reform and economic reforms, an investigation of the efficiency of Chinese universities by region after a decade of reforms is called for. It was found that universities were very inefficient regardless their location. Inefficiency mainly comes from pure technical inefficiency. Across region, universities in the east region were, interestingly, more inefficient over time while there were virtually no changes in inefficiency for universities in the central and west region. In addition to the measures of efficiency, the input-base DEA allows us to examine the possibility of over-utilization in inputs given the amount of outputs. This congestion issue is relevant in the sense that region disparities in accessing resources and the changes in government policies are apparent in the case of Chinese universities. Applying the congestion measure to the data, mild over-utilizing of resources was found for universities in the more developed (east) region. For the universities in the west (least developed) region, over-utilizing of resources got worse between 1994 and 2001 although the overall situation was not too serious. To make the story complete, at the aggregate level the assessment of the overall performance of the higher education sector can be addressed through the estimation of the optimal number of universities by region. With continuous shrinking support from the central government and for the sake of better utilizing social resources, restructuring of universities within each region can be an effective means in managing the education sector.

Efficiency Evaluation of Polytechnic Higher Education Institutions with Considering Output Quality and Institutional Characteristics:

Tsu-Tan Fu - Yun-Shone Lu

The higher education institutions in Taiwan have essentially administrated at the discretion of the government for the past few decades. However, in recent years, the government of Taiwan has gradually lessened its rigid control and allowed higher education institutions more administrative and financial independence. This policy change has exerted pressure on administrators and policy makers to emphasize productivity, efficiency and cost-effective management. In this paper, we evaluate empirically the efficiencies of polytechnic universities and colleges in Taiwan. To evaluate efficiency performance of polytechnic colleges adequately, this paper develops a three-stage performance adjusted DEA evaluation model which can incorporate the quality factors of educational outputs and institutional characteristics of school. A total of 149 schools of Taiwan are empirically evaluated. We also compared efficiencies estimated from the proposed model with those estimated from the traditional unadjusted model. By defined institutional objectives of polytechnic schools, the student enrollments in normal or continued education programs, teacher¡¦s publications, collaborations with industry, professional certificates passed by students are identified as educational outputs for polytechnic colleges. However, to make evaluation at a consistent basis, these outputs are further characterized by the following quality factors for instruction and research services such as student-faculty ratio, research award from the National Science Council, student structure in programs, ratio of faculty with doctoral degree, and ratio of advanced profession certificates passed. The institutional characteristics of polytechnic school evaluated include those organizational factors such as school ownership, type of academic program, school size and field specialization. The empirical results have found substantial efficiency difference between performance adjusted and unadjusted models for those schools with characteristics of public ownership, university type, or commerce field specialization. Thus, the ignorance of those quality and institutional factors could result in measurement bias and inadequate policy implications. The results also indicated that averagely efficiency difference resulted from the output quality adjustment outweighed that from the consideration of institutional characteristics for those public owned, university type, small school size, or engineering field specialized schools.

Substitution effects in multi-output production. Evidence from Italian universities:

Cinzia Daraio - Andrea Bonaccorsi - Leopold Simar

The production of universities is intrinsically a multi-input multi-output activity. This creates a number of challenging theoretical and empirical problems. In this paper we apply recently developed techniques to original datasets, in order to shed light to some controversial issues in economics and policy of science, higher education, and innovation. The problem of relations between various forms of outputs is extremely relevant from the point of view of decision makers and policy makers. First of all, the relation between teaching and research activity may be subject to trade-off effects largely explored in the economics of education and higher education literature (e.g., Graves, Marchand and Thompson, 1982; Cohn, Rhine and Santos, 1989). In principle, substitution and complementarity effects may be equally present. For example, a large number of students may provide universities additional resources that can be employed in research, generating positive effects (e.g., Gautier and Wauthy, 2004). On the other hand, a heavy teaching duty may subtract time and intellectual energy to university professors. So far, most of the evidence has been obtained either in a standard regression framework or using multi-product functions. We will obtain estimates that have a clear interpretation at various points of the interval of size of the outputs or of the total unit of production. Second, more recently the relation between undergraduate and postgraduate teaching has become of interest. Under this respect the institutional differences between Anglosaxon systems, in which doctoral students have formal training and may teach at undergraduate level, and European systems, in which doctoral training is less formalized, have become of interest. A related problem is the relative weight of Master and PhD programmes in postgraduate education. Third, there is increasing interest in the problem of the way in which research activities are affected by so called third mission activities, including creation of spin-off companies, patent application, licensing of intellectual property, applied research with industry, contribution to local and regional growth, consultancy and technical services to the public administration, active promotion of public understanding of science and popular science. It is clear that these activities are becoming increasingly important, as part of the model of entrepreneurial university and of industry-academia-government interactions (often called Triple Helix interactions, see Leydesdorff, 2003). However, several authors have seriously raised the problem of the risk of reducing the quality of academic research, short termism, and loss of tradition of academic independence and neutrality. The problem of the nature of the relation between third mission activities and more traditionally accepted teaching and research activities is therefore on top of the research and policy agenda. In the field of science and education, external-environmental variables may be cause of heterogeneity and may considerably influence the performance of universities. Several studies in the education literature have tried to face this problem by developing and applying one-stage or two or multiple-stage approaches to take into account what they define as socio-economic differences (see Ruggiero, 2004 and the references reported there). The basic idea was to relate efficiency measures to some external or environmental factors which might influence the production process but that are not under the control of the managers. Unfortunately, both one stage and multiple stage approaches are flawed by restrictive prior assumptions on the role of these external factors on the analysed process. On the one hand, as discussed and demonstrated by Simar and Wilson (2003), the multiple-stage approaches suffer of several methodological problems related to the complicated and unknown autocorrelations between first and second stages estimation procedures but also to the inherent bias of the first stage efficiency estimates. On the other hand, in the one stage approach one has to assume the effect of the external factors on the production process, i.e. the analyst should know in advance if the external factors affect positively or negatively the comprehensive performance. Of course, these problems and assumptions are very tight. For this reason, Daraio and Simar (2003, 2005) generalizing the approach of Cazals, Florens and Simar (2002) propose a full nonparametric methodology to explain efficiency differentials by external-environmental factors that overcomes most limitations of previous approaches. In particular, the effect of external factors came as a result of the analysis and is not assumed. In this paper we provide a robust estimation of local and global elasticities of substitution between various outputs of universities. We consider the following outputs: a) Research (number of international publications); b) Undergraduate teaching (number of graduated students, number of enrolled students); c) Postgraduate teaching (number of enrolled students at Master and Doctoral programmes) d) Third mission activities (number of patents by academic inventors, percentage of university budget represented by private contracts). The estimation of both local and global elasticities provides clear indications for analysis and policy making. It avoids several methodological shortcomings of the existing literature, which is often a bad empirical foundation for policy making. We develop several models, taking into account various multiple outputs combinations that are able to jointly capture the effects of teaching and publication activities using the parametrization of robust nonparametric frontiers, as proposed by Florens and Simar (2005). This will allow the estimation of elasticities of substitution between outputs in a robust way, avoiding the drawback of regression-based frontier estimation techniques which capture the shape of the cloud of points (and so elasticities of substitution of the production function) in its middle (due to the averaging of the regression) and not, as it should, at its efficient boundary. These developments will permit to confirm the nature of local effects that were identified in a previous work (Bonaccorsi, Daraio and Simar, 2004).

8a

'Just DEA It!': A New DEA Software and Its Possibilities for Applied Researcher:

Eugene Morgunov - Valentin Zelenyuk

In this paper we review the possibilities of the new version of the software for applied researchers in efficiency analysis—“Just DEA It!”. The software is capable of estimating DEA models under assumptions of constant, no-increasing and variable returns to scale, for input and output orientations, it estimates slacks, Malmquist Productivity indexes, etc. The key novelty of this software is application of statistical bootstrap to obtain bias correction and confidence intervals for the individual and aggregate efficiency scores, as well as to perform some statistical tests. The speed of such estimations in our software is much faster than, for example, in Matlab. Software is enhanced by graphical and user interface, while the data is stored in relational database. Overall, the new software shall be a useful tool for applied researchers, facilitating the use of advanced techniques such as bootstrap.

FEAR 1.0: A Software Package for Frontier Efficiency Analysis with R:

Paul W. Wilson

DEA estimators have been applied in more than 1,800 articles published in more than 490 refereed journals (Gattoufi~{\it et~al.}, 2004). DEA and similar non-parametric estimators offer numerous advantages, the most obvious being that one need not specify a (potentially erroneous) functional relationship between production inputs and outputs. Although much of the nonparametric efficiency literature has ignored statistical issues such as inference, hypothesis testing, {\it etc.,} the statistical properties of DEA estimators have recently been established; see Simar and Wilson (2000) for a survey of these results, and Kneip ~{\it et al.} (2005) for more recent results. Standard software packages (e.g., LIMDEP, STATA, TSP) used by econometricians do not include procudures for DEA or other nonparametric efficiency estimators. Several specialized, commercial software packages are available, as well as a small number of non-commercial, freeware programs. To varying degrees, all of the aforementioned packages are good at what they were designed to do. The existing packages are designed for ease of use (again, with varying degrees of success), but the cost of this is often inflexibility, limiting the user to procedures the authors have explicitly made available. Moreover, none of the existing packages include procedures for statistical inference. Although the asymptotic distribution of DEA estimators is now known (see Kneip et al., 2005, for details) for the general case with $p$ inputs and $q$ outputs, bootstrap methods remain the only useful approach for inference. None of the existing packages incorporate the bootstrap methods proposed by Simar and Wilson (1998, 2001). This paper introduces FEAR 1.0, a software library that can be linked to the general-purpose statistical package R. The routines included in FEAR 1.0 allow the user to compute DEA estimates of technical, allocative, and overall efficiency while assuming either variable, non-increasing, or constant returns to scale. The routines are highly flexible, allowing measurement of efficiency of one group of observations relative to a technology defined by a second, reference group of observations. Consequently, the routines can be used to compute Malmquist indices, scale efficiency measures, super-efficiency scores along the lines of Andersen and Petersen (1993), and other measures that might be of interest. Routines are also included to facilitate implementation of the bootstrap methods described by Simar and Wilson (1998, 2000b). These features can be further used to implement methods of inference for Malmquist indices, statistical tests of constant returns to scale versus non-increasing or varying returns to scale, as well as statistical tests for irrelevant inputs and outputs or aggregation possibilities. A routine for maximum likelihood estimation of a truncated regression model is included for regressing DEA efficiency estimates on environmental variables as described in Simar and Wilson (2005). In addition, FEAR 1.0 includes commands that can be used to compute FDH efficiency estimates along the lines of Deprins et al. (1984), to perform outlier analysis using the methods of Wilson (1993), and to compute the new, robust, root-$n$ consistent order-$m$ efficiency estimators described by Cazals et al. (2002). Most of these features are unavailable in existing software packages.

8b

Rationalising Inefficiency: A Study of Canadian Bank Branches:

Mette Asmild - Peter Bogetoft - Jens Leth Hougaard

Many studies have attempted to explain inefficiency, for instance by bounded rationality, ignorance, lack of incentives or motivation etc. However, the presence of inefficiency remains conflicting with the neo-classical idea of economic rationality. This paper shows how the outcomes of Data Envelopment Analysis-type efficiency models can be rationalised. The allocation of inefficient resources as compared to both economically and technically efficient benchmarks can be understood utilising the theoretical framework suggested in Bogetoft and Hougaard (2003). To illustrate the concepts we consider a data set of Canadian bank branches. The empirical results are encouraging since what appears to be inefficiency in some branches can be argued to be rational choices of reosurce allocation.

Searching for obstacles to further banking integration in Europe:

ANA LOZANO-VIVAS - Jesus T. Pastor

The paper studies the evolution and determinants of productivity banking integration within and across European countries. We study banking productivity considering a worldwide unique constant returns to scale (global) frontier. Resorting to a rencently defined Global Malmquist Index we detect productivity growth and study the influence of its two components, efficiency change and technical change. We further analyze an sho convegence for productivity change and its components to test the pace and tren of European banking integration. We analyze the banking performance of a sample of fifteen European countries over the 1993-2000 period. The paper also presents empirical evidence on how the particular environmental contitions of each country affect the performance of the banking industry and its convergence trend. Banking integration is affected by both, performance and country environmental conditions so it can not be viewed as a uniform and balanced process across all countries.

8c

The assessment of retailing efficiency using Network DEA:

Clara Vaz - Ana Camanho

The objective of this paper is to develop a methodology, based on Data Envelopment Analysis, to assess and improve the performance of retailing organisations with multiple units (shops). The methodology developed is applied to a case-study consisting of 78 supermarkets to ensure its relevance and applicability to real-world assessments. The Decision Making Units (DMUs) of retailing organisations are complex and often include multiple sub-units. For example, in the case of supermarkets, the shops are usually organised in sections that sell different products (e.g., grocery, perishables, textiles, light bazaar and heavy bazaar). This paper proposes assessing the retailing shops considering the DMU at the section level, whose efficiency levels can be obtained from a comparison with similar sections from other shops. This is consistent with the adoption of a commercial perspective for the assessment, such that the outputs are volume of sales, and the inputs measure the resources available at each section (area, stock and number of references available). This analysis is further refined with the inclusion of an additional input reflecting the value of products wasted in the shop floor, which often represents a significant proportion of the operational costs. This analysis enables the identification of benchmarking sections across all retailing stores under analysis, whose best practices should be disseminated across the stores. Considering that the performance improvement targets of retailing stores are focused on volume of sales, this paper suggests using a Network DEA model to aggregate the individual sections in such a way that the targets obtained are consistent with the restrictions at the store level. Therefore, for target setting purposes, the focus of the analysis is changed from sections performance to store performance, whose assessment is made considering that each store is a net of sections. The Network DEA model developed in this paper enables a reallocation of resources between the sections. This model differs from the traditional DEA as it enables both increases and decreases in the level of certain inputs of the sections, subject to the restriction that the input levels at the store are not increased. For example, the assessment reported in this paper considers that the store area is a fixed input, but the area of the sections can be adjusted through the reallocation of floor space to those section that enable a maximization of the volume of sales. This Network DEA model assumes a system-wide perspective, such that the objective is no longer a proportional increase in sales for all sections, but an overall increase in sales at the store. This may involve targets that imply downsizing for some store sections. The applicability of the approach developed in this paper is illustrated in the context of a real-world efficiency assessment of a major Portuguese retailing organisation. The results of the section assessments and the targets obtained for the shops using a Network DEA analysis are discussed. The managerial implications of the results obtained are explored.

Searching for an Optimal Technical Efficiency Measure:

Sung-ko Li

It is well-known that a technical efficiency measure is derived by comparing an observed point with an efficient point on the frontier. Different measures have been proposed in the literature. They identify different efficient points. For the same data set, different methods usually give completely different ranks of the firm. This will cause confusions to managers who want to get useful information from measuring the efficiency of individual units. To search for an optimal efficiency measure, this paper suggests four typical cases and proposes two different approaches. The four typical cases cover all possible considerations of the frontier in efficiency measurement. The two approaches, the shadow-price approach and the directional distance function approach, provide theoretical basis to compare different efficiency measures under the same unified frameworks. To overcome the data problem, the Monte Carlo method will be adopted to generate data that come from multi-output multi-input production technologies with different economic properties. With unified frameworks and data under the control of the researcher, this paper compares the ranks generated by each efficiency measure with the ranks implied by the two unified frameworks. This can identify the best index that can be used in efficiency measurement. It is shown that under some conditions, the radial Farrell input- and output-based technical efficiency measures are the best.

Modifications of the algorithms for visualization of the multidimensional frontier in DEA and efficiency analysis of Russian banks:

Vladimir Krivonozhko - Oleg Utkin - Michail Safin - et al.

In our previous works we have shown that standard parametric optimization algorithms may not be very efficient from a computational point of view. For this reason our group has developed a family of parametric optimization algorithms that allows us to visualize and to analyze production unit behavior in the multidimensional space of inputs and outputs. There are more than one thousand banks in Russia. So, we have to modify our algorithms in order to accelerate the speed of computations and to increase the accuracy of our algorithms. We apply our methods to efficiency analysis of Russian banks.

Constructing Balanced Portfolios of R&D Projects: A DEA Approac:

Boaz Golany - Harel Eilat - Avraham Shtub

We propose and demonstrate a methodology for the construction and analysis of efficient, effective and balanced portfolios of R&D projects with interactions. The methodology is based on an extended Data Envelopment Analysis (DEA) model that quantifies some the qualitative concepts embedded in the Balanced Score Card (BSC) approach. The methodology includes a resource allocation scheme, an evaluation of individual projects, screening of projects based on their relative values and on portfolio requirements, and finally a construction and evaluation of portfolios. The DEA-BSC model is employed in two versions, first to evaluate individual R&D projects, and then to evaluate alternative R&D portfolios. To generate portfolio alternatives, we apply a branch-and-bound algorithm, and use an accumulation function that accounts for possible interactions among projects. The entire methodology is illustrated via an example in the context of a governmental agency charged with selecting technological projects.

8d

ASSESSING THE RELATIVE PERFORMANCE OF UNIVERSITY TECHNOLOGY TRANSFER OFFICES IN THE U.K USING MULITPLE OUTPUT STOCHASTIC DISTANCE FUNCTIONS:

Andy Lockett - Wendy Chapple - Mike Wright - et al.

Universities have been increasingly viewed as engines of economic growth and as creators of intellectual property (IP). IP, through successful commercialization, in theory leads to technological change and productivity growth. Successful IP creation and exploitation can also lead to financial gains for the university. Universities, therefore, are increasingly looking towards maximising the returns from IP through the commercialisation of their research outputs. Traditionally within the technology transfer literature, commercialisation success has been measured by a single measure of licensing activity, either in the form of licensing income, number of licenses, or creation of university spin-out companies (that is companies established as a vehicle to exploit IP). This paper advances previous work to assess the performance of university technology transfer offices, by using stochastic multiple output distance functions. The traditional focus on a single output will arguably only provide a partial analysis of university technology transfer performance. We argue that multiple output measures are more appropriate than the single measures in assessing university efficiency, due to high levels of heterogeneity in the strategies of UK university technology transfer offices. Using a unique UK dataset, collected in collaboration with UNICO (the university new company association) and AURIL (Association of University Industry Liaison), this paper analyses the relative efficiency of UK universities in their ability to turn research inputs into the commercial outputs.

Educational Efficiency in Belgian schools: A Parametric Distance Function Approach at Student Level: - Abstract not available.

Sergio Perelman - Daniel Santín

Measuring public sector performance and productivity:

Philip Stevens - Mary O'Mahony

Explaining differences in countries’ propensities to create wealth is at the heart of economic analysis. This is informed by measures of relative performance of nations and sectors within these nations in terms of output, input and productivity growth. The motivation for this paper is the evaluation of output and productivity performance of sectors where a significant proportion of the output is not sold on the market. This has long been an issue of concern for analysts in the field of productivity measurement since many countries national accounts measure non-marketed outputs by inputs. Hence cross-country comparisons of aggregate productivity growth or levels are contaminated by incorrect measurement in sectors that represent about 20% of aggregate activity. A solution is to exclude non-market sectors when considering aggregate differences and in analysis at the industry level. However this is unsatisfactory given the importance of the sectors excluded. In this paper we consider the measurement of performance in public service provision. We consider the theoretical arguments for employing outcomes in measuring productivity growth and set out a measurement framework. We suggest that the index numbers commonly employed in private sectors can also be applied to non-market sectors, but that the estimation of the information contained in these index numbers proceeds in different ways. We illustrate this using the examples of two of the largest sectors of public service provision: education and health.

Faculty Research Productivity and Promotion: Testing the Behavioral Reinforcement Theory:

Flora F. Tien

This paper aims at exploring how faculty research productivity changes before, during, and after promotion. A mail survey investigating Taiwanese faculty members in nine universities and in fourteen disciplines is conducted. The survey response rate is 52% (N=1017). The measures of faculty research productivity are numbers of published articles and of books. Testing a research hypothesis generated from behavioral reinforcement theory and an alternative claim on the learning effect of familiarity with publishing norms, the study makes several methodological efforts such as collecting career history data and longitudinal records on faculty research productivity to fill literature gaps. The main findings include: (1) overall, behavioral reinforcement theory receives support. Only a few natural sciences & engineering faculty publish articles and a few humanities and social science faculty publish books during the early years before promotion. As faculty approaches the promotion decision, the proportion of publishing the main publication type of their field increases. After promoting to the full professorship, the proportion publishing declines; (2) among different field groups, the research behavior change curve of the natural sciences & engineering faculty best fits the theory; (3) different levels of anticipated promotion rewards have different effects for motivating subjects to publish. The higher the level of the promotion reward, the greater the motivating effect the reward has on faculty research productivity; (4) the results cast doubt upon the argument that only the learning effect of familiarity with publishing norms but not the desire for promotion affects faculty research behavior.

9a

Shadow price of capital and underinvestment in cooperatives and conventional firms

Vania Sena - Ornella Maietta

Cooperatives are usually considered to be allocatively inefficient because they underutilize capital (Mosheim, 2002). This undercapitalisation is attributed to the lack of property rights in the cooperative that often prevents a member from selling an ownership share upon leaving the cooperative: members of cooperatives do not have an incentive to invest in capital equipment as they may not appropriate the increase in value following the investment, in case they decide to leave the cooperative. This means that, comparing the inputs used by conventional firms and by cooperatives, on average, cooperatives use less capital and labour than conventional firms but use more intermediate consumption (see Maietta and Sena, 2004 for the Italian cooperatives). This relative undercapitalisation of coops is quite common and not limited to the Italian cooperatives. The purpose of this paper is to verify whether a different shadow price of capital exists for cooperatives and for conventional firms and whether this can be explained by the existence of finance constraints. The approach followed is to estimate a shadow cost function (Atkinson and Cornwell, 1994, Kumbhakar, 1997) on a panel of conventional and cooperative Italian firms, specialized in the production of wine, over the time 1996-2001 and then relate the differences in the shadow price between the cooperatives and conventional firms to the measures of finance constraints, like debt-to asset ratio and interest coverage ratio. Preliminary results show that credit constraints are important in explaining the differential in the shadow price of capital between the conventional and cooperative firms. Atkinson S. E. and C. Cornwell (1994), Parametric estimation of technical and allocative inefficiency with panel data, International Economic Review 35, 1, 231-243. Kumbhakar S. C., 1997, Modeling allocative inefficiency in a translog cost function and cost share equations: an exact relationship, Journal of Econometrics, 76, 351-356. Maietta O. W., Sena V. (2004), Profit sharing, technical efficiency change and finance contraints, in (eds.), Pérotin V., Robinson A., Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Elsevier, Amsterdam, vol. 8, pp. 149-167. Mosheim, Robert, 2002, Organizational Type and Efficiency in the Costa Rican Coffee Processing Sector, Journal of Comparative Economics 30, 296–316.

Institutional reforms now and benefits tomorrow: how far is tomorrow?:

Pierre-Guillaume Méon - Khalid Sekkat - Laurent Weill

This paper aims to investigate the timing of the impact of changes in the quality of institutions on macroeconomic efficiency. We do so by applying Battese and Coelli (1995)''s stochastic frontier model at the aggregate level. We find that changes in quality of institutions exert two effects on macroeconomic efficiency, a short-term, i.e. after two to four years, and a medium-term effect, around eight years. Robustness checks performed with different measures of the quality of institutions tend to support these results.

Market Fragmentation and Structural Efficiency in China :

Sung-ko Li - Yuk-Shing Cheng

The economic reform in China has been implemented successfully for more than two decades. A key feature of this economic reform is the decentralization of economic power to the local governments. From western point of view, decentralization will lead to a more competitive market. However, some researchers pointed out that in a transition economy like China, decentralizing powers from the central government to local governments would create new distortions. Although the market condition within a local region became more competitive, the competition between regions became less because resources are not allowed to flow freely. This caused a lot of efficiency losses. In this paper, we point out that the shadow price model of ‘structural efficiency’ introduced by Li and Ng (1995) and later extended by Färe and Valentin (2003) is well suited for this issue of regional market fragmentation. To model regional heterogeneity, we use a characteristic variable. Under constant returns to scale, we show that the long-run and short-run structural efficiency are equal. Our modified model is applied to measure the structural efficiency of China from panel data of her provinces. We then decompose the structural efficiency into aggregate inefficiency due to inefficiency within each province and re-allocative inefficiency due to inefficient resource allocation among provinces. Our preliminary results indicate that the productive efficiency within provinces is improving during 1980s and early 1990s. During the same period, resource allocation became more inefficient at the same time period. Our conclusion is that while decentralization should carry on, the central government must put more effort to enhance resource flow across regions.

A Comparison of Productivity in public and private industrial enterprises in India:

Ram Mohan - Subhash Ray

India has been going through a halting and controversial privatization program over the past decade. The program is premised on the hypothesis that India''s public sector enterprises are inefficient and that an improvement in efficiency can come about only by privatizing these. We subject this hypothesis to scrutiny by comparing productivity at public and private enterprises in the industrial sector of India. We use two measures, Malmquist total factor productivity and Tornquist total factor productivity, and compare performance in eight different industries over the period 1991-92 to 1998-99. The Malmquist total factor productivity growth is obtained by applying DEA to the panel data. We find that in six out of eight industrial sectors, there is no difference in Malmquist productivity growth between public and private enterprises. In the remaining two sectors, the public sector does better in one and the private sector in the other. In respect of Tornquist total factor productivity growth,we find that there is no difference in four sectors. In the remaining four sectors, the private sector does better in two (at a 5 per cent confidence level) while the public sector does better in two (at a 10 per cent confidence level). In other words, out of a total of 16 observations, there are no differences in ten cases, the private sector does better in three cases and the public sector in three. Our observations do not support the view that the private sector is inherently efficient in the Indian context. Our finding is of some importance as India is among the few economies with a long history of the prviate and public sectors co-existing over a long period. While we do not speculate on why private ownership has not delivered superior performance in India, we believe our findings accord with the view that the quality of governance mechanisms and legal institutions has an important bearing on whether the private sector can do better than the public sector. These mechanisms and institutions are not of the requisite quality in emerging markets such as India. Two possible policy implications are: first, it is not axiomatic that privatization would lead to improved performance; secondly, privatization must focus on those sectors where the private sector has been seen to do better.

9b

Mean-Variance-Skewness Portfolio Performance Gauging: A General Shortage Function and Dual Approach:

Walter Briec - Kristiaan Kerstens - Octave Jokung

Abstract not available.

Mutual Fund Performance and Persistence in Taiwan:A Non-Parametric Approach:

Chi-Sheng Hsu - Ching-Yao Lia

This paper applies a non-parametric, linear-programming technique, called data envelopment analysis (DEA), to measure performances of Taiwan domestic equity funds during the period of 1999 to 2003. Unlike the traditional Sharpe ratio, the DEA enables us to consider the transaction costs (loads, expense ratios, and turnover ratios) and excess returns simultaneously. The results of Pearson correlation and Spearman''s rank correlation tests show that technical efficiencies (TEs) estimated by DEA are significantly positively correlated with Sharpe ratios. To test the performance persistence, we first divide the sample funds into the low efficiency (0-25th percentile), middle efficiency (26th-75th percentile), and high efficiency (76th-100th percentile) groups according to their ranking on TE value and Sharpe ratio, respectively. The ANOVA analyses and t-tests indicate there are significant differences in mean performances across groups. Then, based on the previous one-, two-, and three-year performances of the sample funds, we employ the iterative seemingly unrelated regressions to analyze the persistence of sample funds'' performances by using the measures of TE value and Sharpe ratio simultaneously. Our results show that there exists a significant 'hot hands' effect in Taiwan domestic equity funds under the TE measure. However, we do not observe the persistence of mutual fund performances under the measure of Sharpe ratios. We conclude that the different results in performance persistence analyses between the above two measures is driven by the DEA taking transaction costs into consideration.

9c

Estimating the Stock of Red Seabream in the Strait of Gibralter: A DEA?Malmquist Analysis:

David Castilla Espino - Harold Fried - Juan Jose Garcia del Hoyo - et al.

The Red Seabream is a valuable fish that is important to port regions of Southern Spain. It is the most expensive fish species in the region, and commercial fishing accounts for more than 10% of employment in some costal villages. This fish is a traditional food fare on Christmas. At one time, Red Seabream was also important to Northern Spain and other Eastern Atlantic regions, but these stocks were virtually depleted due to over fishing ten years ago and have yet to recover. It is critical that the stock in Southern Spain be properly managed to avoid a similar fate. To accomplish this, it is necessary to understand the stock levels and dynamics of the fishery. This paper investigates stock levels over time using DEA/Malmquist techniques. In order to isolate the change in the latent variable fish stocks we exploit the knowledge that over a four year period technology in this fishery was constant. Under this condition the technology change component of the Malmquist productivity index can be interpreted to represent changes in the fish stock rather than changes in technology. In particular, technological regress (progress) suggests a decrease (an expansion) in the stock of the fishery. If technology is constant, we assume that monthly reductions in the Malmquist index are due to reductions in stocks from the beginning of the fishing season. We do not know beginning stocks, but if the reduction of the stock is assumed to be the quantity of fish harvested, then comparing month to month computations of the Malmquist index between any two successive years for 3 consecutive months allows us to solve for beginning stocks in each of the two years and the fishing yield rate from those stocks. We estimate a Cobb-Douglas coefficient for the yield. Daily catch data by vessel are available. These are augmented with fishing vessel characteristics and fishing activity. These are single output day trips to catch Red Seabream. Inputs include vessel length, horsepower and crew size. These data are aggregated into monthly frequencies. One exercise is conducted with monthly data over the period 2000 through 2003 for twelve fishing vessels based in the port of Tarifa. A second exercise utilizes a longer time period, 1998 through 2003, for twenty one vessels. The second exercise is interesting because we have knowledge that technology changed during 1998 and 1999. We exploit the period of constant technology to isolate technological change and stock change over the earlier period.

Estimating Fishing Capacity of Red Seabream Fishery in the Strait of Gibraltar: Parametric and Non-parametric Approaches:

David Castilla Espino

Red seabream (Pagellus bogaraveo, Brünnich, 1768) is a delicious fish, particularly appreciated in Southern Spain. It is one of the most expensive fish sold in the Andalusia fish public exchanges. The high price is due to high demand and to the decline in harvesting levels. Within subdivision IX of ICES area (South west of Spain and Portugal), the Andalusian “voracera” fleet that exploit this fish stock in the Strait of Gibraltar accounted for more than the 50% of red seabream landings, making it one of the most important red seabream fleets in Europe. Moreover, this fleet accounts for a large percentage of full time total employment in the ports of Algeciras and Tarifa. In particular, Tarifa is one of 33 zones in Europe where dependency on fishing is higher than 10% (European Commission 1999). Landings in subdivision IX of ICES area have progressively decreased since 1985, especially sharply in 1998, as a consequence of unsustainable exploitation of the fish stock. Poorly defined property rights combined with high profits have attracted the entry of new vessels. The high mobility and flexibility of vessels allow essentially open access to the fishery. In 1998, it was evident that the fishery was overexploited and two recovery plans aimed at sustainability were implemented in periods 1999-2002 and 2003-2005. The first recovery plan introduced a limited license program to restrict access to the fishery, fishing gear restrictions, a cap on the number of fishing trips, incentives to leave fishing employment, and a size limitation on fish caught. An advisory board was formed, composed of fishermen and other interest groups. The second recovery plan imposed a Total Allowable Catch (TAC) per year. Meanwhile the European Commission (EC) initiated its own TAC for red seabream in subdivision IX of ICES. This paper investigates the effects of the first recovery plan on the fish stock and estimates overcapacity of the “voracera” fleet at the beginning of the second recovery plan. Fleet capacity and capacity utilization are determined by Peak to Peak, Stochastic Frontier Analysis (SFA) and Data Envelopment Analysis (DEA) techniques. The results from the three techniques are compared. The data are monthly red seabream daily trips by vessels located at the port of Tarifa during the period 1998-2001provided by Andalusia regional government. The Peak to Peak technique is based on the identification of peaks (relative maximum outputs) using historical information of the fishery and assuming that capacity is completely used in peaks periods (Ballard and Roberts 1977, Hsu 2003). This technique is simple and requires relatively small amounts of data compared to DEA and SFA. Critical assumptions are constant return to scale and Hicksian aggregation or Leontieff separability of a Coob-Douglas production technology. Peak to Peak is an appropriate technique only when data are limited (Kirkley and Squires 1999b). DEA can handle technologies in multi-species fisheries, it is relatively simple and imposes few theoretical restrictions. DEA does not require the specification of the technology of production, but it does not capture stochastic factors influencing production. Fare et al. (1989) developed a two-stage procedure to calculate production capacity and capacity utilization that was first applied to fisheries by Kirkley and Squires (1999). This procedure imposes full utilization of variable productive inputs in a standard DEA linear program in a first stage in order to determine fishing capacity and observed capacity utilization. A standard DEA linear program is run in a second stage to determine efficient capacity utilization. SFA includes stochastic factors, the specification of the technological characteristics of the production process and permits testing linear restrictions. SFA has difficulties handling multi-output technologies, although this can be accommodated using distance functions (Coelli and Perelman 1999, 2000). Kirkley and Squires (1999) first applied this technique to measuring fishing capacity. This paper employs two procedures that are based upon SFA. The first procedure (Pure SFA) estimates an SFA model without restricting variable inputs in a first stage to determine fishing capacity and observed capacity utilization. A standard SFA model is estimated to obtain efficient capacity utilization in a second stage. The second procedure (Mix DEA-SFA) determines the ratio of utilization of variable inputs using a DEA linear program, where full utilization of variable inputs is imposed. In a second stage, a standard SFA model is estimated and evaluated to determine fishing capacity and observed and efficient capacity utilization using the ratio of utilization of variable inputs calculated from the DEA first stage. The results suggest that the fishing capacity of the “voracera” fleet of the port of Tarifa is 206 T. according Peak to Peak, 313 T. according DEA and between 313 (Mix DEA-SFA) and 322 (Pure SFA) according SFA in 2003. It is interesting that the results for DEA and Mixed DEA-SFA are the same. Extrapolating the results for the whole fishery suggests that overcapacity is between 12% and 74% over the TAC. It would be prudent to base policy on the SFA estimate of overcapacity equal to 74% in excess of the TAC. Capacity Utilization shows a slightly increasing trend after 2000. The mean capacity utilization is 66% according Peak to Peak, 62% according DEA, and 55% according SFA in 2003. The crew rate of utilization (82.4%) exceeds the utilization rate for fishing trips (52.2%) during the whole period. This might be due to the fact that the labour intensive vessels continue to operate in the fishery. Full input utilization for the fleet is 5680 fishing trips and 369 crew members. The mean technical efficiency of vessels is 75.77% according DEA and between 70% and 81% according SFA during 2003. Efficiency rankings using either DEA or SFA are equal. Technical efficiency declines over the period, suggesting regressive technological change due to tighter regulation and the reduction of the fish stock. The management measures applied after 1998 have succeeded in reducing the fishing capacity of the fleet, although not sufficiently to ensure the long run viability of the fishery. Additional regulatory measures are required.

DEA VERSUS ECONOMETRIC ESTIMATION OF EFFICIENCY IN FISHERIES: A REVIEW:

Lamprakis Avdelas - Christos Floros

In this paper we discuss the pros and cons of Data Envelopment analysis (DEA) and econometric methods for estimating efficiency in fisheries. Only limited attempts to estimate DEA, stochastic DEA and stochastic production frontiers (SPF) have been developed and applied to fishing industries (Kirkley, Squires and Strand 1995, 1998; Coglan, Pascoe and Mardle 1998; Pascoe, Anderson and de Wilde 2001). The aim of this paper is threefold. The first intention is to present a review of the theoretical concept underlying the efficiency of fisheries activity. The next intention is to present an overview of quantitative procedures, which can be, used within the framework of DEA and SPF in fisheries under Frontier 4.1, LIMDEP and GAMS software packages. Third, we present a method for deriving a stock index from fisheries data that does not involve any restrictive assumptions regarding the relationship between catch, effort and stock size. In this case, DEA is preferred to SPF. These methods are strongly recommended to fisheries managers dealing with efficiency.

Modelling inefficiency in the product mix:

Frank Asche - Kristin Helene Roll - Ragnar Tveteras

Whether one models inefficiency as output or input oriented depend on what is the endogenous variables for the industry in question. In fisheries, the outputs are often endogenous while the capital (vessel) and other input factors are often treated as fixed. Hence, when investigating efficiency, one has used production functions, requiring that outputs are aggregated to a single output. However, following Squires a large literature indicates that most fisheries should be modelled as multi output technologies. We are in this paper modelling economic inefficiency in a revenue function setting. As it is the outputs that are endogenous, the efficiency measures must be output oriented. Using a revenue function allow us to decompose economic inefficiency into its technical and allocative components. As such, we focus on inefficiency of the output level and the allocative inefficiency caused by an inefficient product mix. In a fishery, efficiency is traditionally modelled through a production function. However by using a production function to describe the underlying technology, it is only possible to get estimates of technical efficiency. Based on a dual approaches all type of inefficiency can be estimated through additional parameters, and firm-specific technical inefficiency as well as output and firm-specific allocative inefficiency, can be identified. This approach was applied to panel data of Norwegian cod trawl fisheries. Panel data allow us to identify and obtain joint parametric estimates of output and firm-specific allocative inefficiency and firm specific technical inefficiency. Our results predict that technical inefficiency effects are significant in explaining the level and variation in vessel revenue. The mean technical efficiency level is estimated to be 78,9%, thus on average the sample could have increased their revenue 21,1% by operating on full technical efficiency. Furthermore, it is found that lager vessels have higher technical efficiency than smaller vessels, and “full-time” fisheries have higher technical efficiency than “part-time”-fisheries. Finally, the results indicates that there are substantial allocative inefficiency in the fleet.

9d

Efficiency and Risk:

Gary Ferrier

A common assumption of economic and finance theory is that firms are risk neutral. While firms may be risk neutral, firms’ managers (i.e., decision makers) may not be risk neutral. The (usually implicit) assumption of risk neutrality is likely to bias measures of efficiency since the assumption results in the omission of risk preferences from the model. A small number of previous efficiency studies using stochastic frontiers have included risk preferences (e.g., Hughes and Mester, RESTAT, 1998). A recent paper by Kumbhakar (AJAE, 2002) has directly specified and estimated risk preferences, production risk, and technical inefficiency for a stochastic frontier model. This paper uses a recently developed technique to simultaneously calculate measures of risk aversion and technical efficiency within a DEA framework. In particular, the method is applied to a sample of U.S. Bank Holding Companies (BHCs) operating from 1994-2001. Results from a model that ignores risk and from the model that simultaneously calculates risk aversion and efficiency are contrasted. The empirical relationship between the calculated risk and efficiency measure is examined as well. Finally, the risk aversion measure calculated by the DEA model is contrasted with standard financial measures of risk to determine whether managers’ risk aversion is reflected in firms’ market performance.

Productivity at the Post:

Emili Grifell-Tatjé - C. A. Knox Lovell

We study economic and financial performance of the United States Postal Services (USPS) after the 1971 reorganization from government (Postal Reorganization Act) using data from 1963 to 2002. The literature on the measuring of productivity and, especially, that which has the identification of the explanatory causes or effects of the productive variation as its goal has grown enormously from the eighties onwards. In contrast, the literature which has the study of the analysis of the distribution of the productive gains achieved by the company as its objective, has tended to disappear. It is essential, for any policymaker, to have information on the levels of efficiency reached as well as of the distribution of the productivity gains (losses) among stakeholders. In this paper we study the USPS analysing both: the variation and the distribution of the productive gains. We propose a methodological approach which has as background Davis (1955) and the Belgium and French literature of the sixties and seventies.

Confident-DEA: A New Methodology Genetic Algorithm Based; For Dealing with Imprecise Data in Data Envelopment Analysis:

Said Gattoufi

This paper proposes a theoretical contribution to the DEA-literature by suggesting an extension we call Confident-DEA approach. The proposed approach involves a bi-level convex optimization model, and hence NP-hard, to which a solution method is suggested. Confident-DEA constitutes a generalization of DEA for dealing with imprecise data and hence allows prediction. Complementing the methodology proposed by Cooper et al (1999) which provides single valued efficiency measures, Confident-DEA provides a range of values for the efficiency measures, e.g. an efficiency confidence interval, reflecting the imprecision in data. For the case of bounded data, a theorem defining the bounds of the efficiency confidence interval is provided. For the general case of imprecise data, that is a mixture of ordinal and cardinal data, a Genetic-Algorithm-based metaheuristic is used to determine the upper and lower bounds defining the efficiency confidence interval. In both cases, a Monte-Carlo type simulation is used to determine the distribution of the efficiency measures, taking into account the distribution of the bounded imprecise data over their corresponding intervals. Previous DEA work dealing with imprecise data implicitly assumed a uniform distribution. Confident-DEA, on the other hand, allows for any type of distribution and hence expands the scope of the analysis. The bounded data used in illustrative examples are assumed to have a truncated normal distribution. The methodology allows for any other distribution for the data.

Minimum Distance and Efficiency Assessment with DEA Models:

Inmaculada Sirvent - Juan Aparicio - Jose Luis Ruiz

In the recent literature, we can find a body of papers that are aimed at determining the minimum distance from an inefficient unit to the weakly efficient frontier. We address here the problem of determining the minimum distance from an inefficient unit to the efficient frontier. This approach makes it possible to identify referents on the efficient frontier that are closer to the assessed unit, which may lead to more sensible efficiency measures, since they would be the result of a more homogeneous comparison. In particular, the second stage of the procedure used for solving the radial DEA models can be adapted to this approach and, what is probably more important, we can define new versions of some existing efficiency measures by means of new DEA formulations following these ideas. Most of the existing approaches to this problem require the specification of all the efficient facets of the frontier, which usually leads to complex algorithms that have to be solved in order to find a solution. In our proposal, the efficiency measures are implemented in several new DEA formulations, which avoids the need for specifying the facets of the frontier, and so, reduces the computational burden.

10a

Management Characteristics Collaboration and Innovation Performance in the UK:

Andy Cosh - Xiaolan Fu - Alan Hughes

This paper explores the impact of management characteristics and patterns of collaboration on a firm¡¯s innovation performance in transforming innovation resources into commercially successful outputs. These questions are investigated using a recent firm level survey database for 465 innovative British small and medium enterprises (SMEs) over the 1998 and 2001 period. Both Data Envelopment Analysis (DEA) and Stochastic Frontier Analysis (SFA) are employed to benchmark a firm¡¯s innovative efficiency against best practice. Quality and variety of innovations are taken into account by combining Principal Component Analysis (PCA) with DEA. We find evidence suggesting that innovative efficiency of SMEs is significantly affected by their management characteristics and collaboration behaviour. Collaboration, organisational flexibility, formality in management systems and incentive schemes are found to contribute significantly to a firm¡¯s innovative efficiency. Managerial share-ownership also shows some positive effect. The importance of these effects, however, varies across different sectors. Innovative efficiency in high-tech SMEs could be significantly enhanced by collaboration, formal management structure and training; and that in medium- and low-tech SMEs is significantly associated with managerial ownership, incentive schemes and organisational flexibility.

Corporate Governance and Firm’s Efficiency: The Case of a Transitional Country Ukraine:

Vitaliy Zheka - Valentin Zelenyuk

"The entrepreneurship structure itself may be critical, with the classic issue of the separation of ownership from control being regarded as one of the earliest and most important sources of X-efficiency” (Button and Weyman-Jones, 1992, American Economic Review). Corporate Governance and Firm’s Efficiency: The Case of a Transitional Country, Ukraine One of the major current areas of research in management theories and practices is the issue of corporate governance as an important determinant of a firm’s performance. This issue has become most pronounced, perhaps, with the corporate scandals at Enron, WorldCom, and other giants, but as a theoretical concept it had been in economists’ minds for at least a half of the century. Among the classical papers is certainly that of Leibenstein (1966), who by conceptualizing his notion of X-efficiency had been talking, essentially, in a language of corporate governance. His concept of X-efficiency has spurred a tremendous amount of literature, pro and con, and also inspired us to look for an empirical confirmation/opposition of his ideas—now explicitly in the light of the corporate governance notion. Whenever there is a division of ownership and management in a firm, it is likely that a well-known principal-agent problem would arise. The owners/investors want to ensure that the professional managers that they hire would run the company in line with the best interests of the owners—working with greatest possible efficiency, maximizing the added value of the firm. On the other hand, the goal of managers or large shareholders might be to maximize their own benefits ignoring the interests of other owners, especially minor shareholders. A similar problem of interest confrontation may arise in the relationship between different levels of authority/ownership at virtually any level and dimension. When interests of managers and workers are not perfectly in line with those of owners, there could be a substantial loss in firm’s efficiency. The core of the problem is the asymmetry of information across levels of authority/ownership—which is also the key issue addressed by the corporate governance theories and practices. The objective of this paper is to investigate the relationship between various indicators of corporate governance quality across firms and the efficiency levels of these firms. We choose Ukraine as our empirical subject not only because we know this country perhaps more than of others, but mostly because a transitional country is a country where changes—including changes in corporate governance practices, ownership structures, etc.—are occurring, for a researcher, at a very attractive pace (although sometimes, it might seem slow to people who live there). In other words, transitional countries are the places where one can often observe changes and consequences with an ‘unequipped eye’—changes that one can hardly see in economies operating close to a steady state. Our main hypothesis is that the higher the quality of corporate governance (which is of course hard to measure) goes together with the efficiency of a firm. Our methodology for testing this hypothesis involves two stages. First, using the data envelopment analysis (DEA) estimator, we estimate efficiency of each firm relative to the so-called best-practice frontier. The resulting estimates serve us as proxies for what Leibenstein called “X-efficiency”. These estimates are then, at the second stage, analyzed by comparing distributions and aggregate efficiencies across different ownership structures, and by regressing efficiencies on the firm specific variables, which we expect (based on theory) to influence the X-efficiency of each firm. For regressions we use the recently proposed approach based on the truncated-regression with bootstrap (Simar and Wilson, 2003), which shown (in Monte Carlo experiments) its good performance. Among the firm-specific variables are the proxies indicating good (or bad) corporate governance practices, shares of state and foreign ownerships, as well as industry and time dummies. Our data comes from the database provided by the main Ukrainian stock exchange, First Securities Trading System (PFTS), and contains information from annual financial statements of all manufacturing companies listed on PFTS in 2000-2001. We use information on total of 158 firms operating in 7 industries. We find empirical support for our main hypothesis that there is a positive relationship between the levels of corporate governance quality across firms and the relative efficiency levels of these firms. This gives another empirical support for Liebenstein’s idea that the major source of X-efficiency is motivation at each level of management/ownership. This is also a support of the argument of Stiglitz (1999), that establishment and enforcement of proper corporate governance principles shall significantly enhance development of individual corporations and economies as a whole—at least due to increasing of efficiency of resource utilization. We also find negative association between the share of the state ownership in a firm and this firm’s efficiency which supports previous findings by other researchers (Brown and Earle, 2000; Andreeva, 2003; and Melnychenko, 2002). Remarkably, we found that the relationship between the share of the foreign ownership at a firm and this firm’s inefficiency level is positive and significant, implying that actually foreign firms are not more but less efficient than local firms, at least in our sample. Together with other findings mentioned above, this may suggest that it does not matter what the origin of the firm is but whether this firm is exercising good corporate governance practices or not. Overall, the empirical evidence we found is consistent with various economic and management theories suggesting that the quality level of corporate governance at an enterprise must be positively correlated with efficiency of such an enterprise. Importantly, this conclusion is found for a transitional country, where the ethics of capitalism is far from being fully inoculated into the culture and mentality—for a country where such issues as transparency have often being viewed as a nuisance rather than as a way to success. We hope that our study will contribute at least a bit to increase in beliefs that good corporate governance is beneficial for enterprise performance in transitional countries as well. More generally, we also hope our results would be provocative for more studies in this area, with better methods and better data, and that altogether these studies will stimulate dissemination of high standards of corporate governance practices.

Population Aging and firm level productivity:

Bernhard Mahlberg - Alexia Fürnkranz-Prskawetz - Vegard Skirbekk

The relationships of worker characteristics and productivity are examined using a matched employer-employee data set from forty branches of the manufacturing and service sectors of Austria. The cross section firm level and plant level data of 2001 are used to for estimating labor productivity according to social demographic characteristics of the staff (e.g. average age, average education, etc.), regional economic variables (e.g. regional GDP per capita, etc.) as well as other firm characteristics. As the indicators of labor productivity we apply the production value per employee and the value added per employee. The results are differentiated by branches as well as per region. On the basis of forecasts about the composition of the Austrian labor force regarding age, projections about the future development of labor productivity are computed. For estimates of productivity we adopt a multilevel regression approach.

"Productivity, Efficiency and Financial Performance of the United States Dairy Sector: A Frontier Approach":

Richard Nehring, Loren Tauer, Erik O’Donoghue, Carmen Sandretto

Abstract not available.

10b

Efficiency and market power in Spanish banking:

Rolf Färe - Shawna Grosskopf - Emili Tortosa-Ausina

Some recent studies have been investigating the existence of market power in the Spanish banking industry. Although results are mixed, the evidence suggests some commercial banks and savings banks benefit from monopoly rents. Some other studies [Berger and Hannan, Review of Economics and Statistics LXXX (1998) 454--465] have also found strong evidence suggesting that banks in more concentrated markets exhibit lower cost efficiency. Our study merges these two groups of findings by exploring how cost and revenue efficiency measures for Spanish banks are related to the so-called return to the dollar (RD), whose advantage over other profitability measures lies on it being a ratio---in contrast to the additive structure associated with profit. This relationship is explored by considering nonparametric and semi-parametric techniques, a set of techniques which is more consistent with those employed to measure efficiency in the first stage of the analysis. Results show that banks'' efficiency is differently related to the return to the dollar according to different circumstances. Specifically, for those firms whose revenues are lower than their costs, the relationship is positive. However, it turns to be negative for those banks with revenues above costs, suggesting that the ``quiet life'''' might be a reality for those banks exerting market power. This finding is more apparent when decomposing efficiency into its different components (cost---or revenue---, technical, and allocative), since the relationship found is more obvious for allocative efficiency, either on the cost or revenue side.

Capital Adequacy Risk Management; Efficiency and Productivity in Taiwan''s Banking Industry:

Cliff J. Huang - Ming-hsiang Huang

Responding to the change of regulation governing financial markets, bank manager tends to adjust their attitudes toward risk accordingly. Numerous studies have examined the effect of the Basel Accord I on the risk taking of banking industry in several countries. Yet, the empirical evidences do not reach consensus conclusion whether the capital requirement encourage or discourage bank risk taking. This might be attributed to the fact that each country has its own unique market structure. The uniqueness of the market structure is indeed significantly related to the dynamic interactions between productivity, efficiency and managers¡¦ attitude toward risk in the underlying banking industry. The objective of this study is to examine the impact of the Basel Accord on the attitudes of bank mangers toward risk taking behavior, and its impact on the productivity and efficiency in the Taiwan banking industry. The panel data of forty-eight sample banks in Taiwan from 1981 to 2003 is utilized to examine the overall impact. A stochastic cash flow cost function is modeled to estimate the bank-specific efficiency, productivity, and its risk premium as a measure of risk behavior. In addition, this study adopts a metric for the risk of insolvency (Z-score) developed by Boyd, et al. (1993) to capture the probability of bank failure. The result of this study would be of interest to the investing publics, bank regulators and bank management.

Bank Concentration and Performance in the Single European Market:

Barbara Casu - Claudia Girardone

The deregulation of financial services in the European Union, together with the establishment of the Economic and Monetary Union, aimed at the creation of a level-playing-field in the provision of banking services across the EU. The plan was to remove entry barriers and to foster competition in national banking markets. However, one of the effects of the regulatory changes was to spur a trend towards consolidation, resulting in the recent wave of mergers and acquisitions. To investigate the impact of increased consolidation on the competitive conditions of EU banking markets, we employ both structural (concentration ratios) and non-structural (Panzar-Rosse statistic) concentration measures. In addition, we examine whether different sizes of banks affect competitive conditions differently. Using bank level balance sheet data for the major EU banking markets, in the decade following the introduction of the Single Banking License (1993- 2003), this paper also develops a model of bank profitability that controls for differences in efficiency estimates, competitive conditions, institutional structures and banks’ overall riskiness, proxied by the volatility of earnings and financial capital.

Generalised Malmquist Productivity Measurement in the Banking Systems of Two Emerging Economies:

Meryem Duygun Fethi - Mohamed Shaaban - Thomas G. Weyman-Jones

Banking systems in emerging and developing economies hold the key to economic growth and productivity change at the macroeconomic level. As financial globalisation proceeds, many emerging economies are reforming their banking systems through the process of liberalisation, privatisation and deregulation, especially where state-owned banks have previously been dominant. The purpose of this paper is to measure productivity change and its decomposition into efficiency change, technical change and scale effects in the banking systems of two emerging economies as liberalisation and privatisation occurred. We specify a generalised parametric Malmquist approach using the distance function, Orea (2002). This can be applied to both output and input distance functions and this permits us to consider a range of appropriate models of the banking production process. Within this approach, we have the ability to explore a number of different stochastic specifications of the inefficiency distribution, including time varying inefficiency, heteroscedastic in efficiency conditional on exogenous determinants, (Wang & Schmidt, 2001). The sample we work with covers two different but related emerging economies, Turkey and Egypt. Both have undergone significant regulatory, ownership and market structure changes in the last ten to fifteen years. Egypt, as an emerging economy, has introduced a wide range of structural economic reforms to create a viable banking system in the past decade. Reforms, mainly aimed to decrease the role of state and increase the role of market forces in the operation of the financial system. Similar reforms had been generated in Turkey, although financial crises had caused some setback to the reform process. We use panel data sets for both economies over this period comprising around 700 observations. In particular, we examine the hypothesis that the initial productivity impact of liberalisation and deregulation appears as a returns to scale effect as banks are able to grow, merge, and alter their market share. The generalised parametric Malmquist is specifically designed to capture the scale effect separately from efficiency change and technical change. Preliminary results on the separate economies suggest scale effects can be important in identifying the initial impact on productivity of financial liberalisation policies.

10c

Water Use Shadow Prices and the Canadian Business Sector Productivity Performance:

Tarek Harchaoui - Kais Dachraoui

This paper develops a production framework that allows for self-supplied water intake, an unpriced ‘natural’ input. The framework is then exploited to estimate the corresponding water shadow prices and to assess the extent to which water impacts on the multifactor productivity performance of the Canadian business sector’s industries. Accounting for water intake leaves the aggregate business sector’s multifactor productivity growth virtually unchanged over the 1981-1996 period, but it increases the productivity performance of the largest water-using industries by 0.7 percentage points on average. The shadow price of water intake amounts to $0.73 cubic metre and varies significantly across industries, thereby reflecting different willingness to pay. While the introduction of water recirculation, a kind of water recycling, does not alter in a significant way most of these results, it reduces the shadow price estimate to $0.55 cubic metre and improves its reliability particularly for the largest water-using industries. Water is found to be a substitute to capital and labour inputs, suggesting that more of these inputs are required to bring about savings in water use.

The Use of DEA in the Measurement of Software Development Efficiency: A Quality Focused Approach: Joseph Paradi - Dwight Schmidt Proper measurement of the software development process is the key to a better understanding of the problems that hinder software development performance. This work examines the relative efficiency of software project development using Data Envelopment Analysis. New DEA models for measuring software development efficiency were introduced using end-user quality experience as the main output instead of traditional size metrics, which have been used extensively in the past. Projects with higher quality were found to be more efficient while larger, riskier, more complex projects had slightly lower efficiency. The relationship between size/risk/complexity and project quality was found to be negligible. DEA based software development productivity measurement offered managerially useful results.

EFFICIENCY FLOODING - BLACK-BOX FRONTIERS AND IMPLICATIONS:

Johannes Sauer - PD Dr. Heinz Hockmann

The availability of efficiency estimation software – freely distributed via the internet and relatively easy to use – recently inflated the number of corresponding applications. The resulting efficiency estimates are used without a critical assessment with respect to the literature on theoretical consistency, flexibility and the choice of the appropriate functional form. The robustness of policy suggestions based on inferences from efficiency measures nevertheless crucially depends on theoretically well-founded estimates. This paper adresses stochastic efficiency measurement by critically reviewing the theoretical consistency of recently published technical efficiency estimates with respect to economic development. The results confirm the need for a posteriori checking the regularity of the estimated frontier by the researcher and, if necessary, the a priori imposition of the theoretical requirements. I) INTRODUCTION In the last 15 years applied production economics experienced a clear shift in its research focus from the analysis of the structure and change of production possibilities to those of technical and allocative efficiency of decision making units. Parametric techniques as the stochastic production frontier model dominate the empirical literature of efficiency measurement (for a detailed review of different measurement techniques see e.g. COELLI ET AL., 1998 or KUMBHAKAR/LOVELL, 2000). The availability of estimation software – freely distributed via the internet and relatively easy to use – recently inflated the number of corresponding applications. The application of the econometric methods provided by these ‚black box’-tools are mostly not accompanied by a thorough theoretical interpretation. The estimation results are further used without a critical assessment with respect to the literature on theoretical consistency, flexibility and the choice of the appropriate functional form. The robustness of policy suggestions based on inferences from efficiency measures nevertheless crucially depends on proper estimates. Most applications, however, do not adequately test for whether the estimated function has the required regularities, and hence run the risk of making improper policy recommendations. This paper shows the importance of testing for the regularities of an estimated efficiency frontier based on flexible functional forms. The basic results of the discussion on theoretical consistency and functional flexibility are therefore reviewed (section 2) and applied to the translog production function (section 3). Subsequently stochastic efficiency measurement is discussed to the background of these findings and essential implications are shown (section 4). Further some stochastic frontier applications with respect to developing countries are exemplary reviewed with respect to theoretical consistency (section 5). It is in particular argued that the economic properties of the estimation results have to be critically assessed, that the interpretation and calculation of efficiency have to be revised and finally that a basic change in the interpretation of the estimated function is required. (…) VII) CONCLUSIONS: THE NEED FOR CONSISTENT AND FLEXIBLE EFFICIENCY MEASUREMENT Existing black box estimation tools foster incorrect and unsound efficiency estimations lacking theoretical consistency and hence lead to inadequate and potentially counterproductive development policy actions. The preceeding discussion hence aims at highlighting the compelling need for a critical assessment of efficiency estimates with respect to the current evidence on theoretical consistency, flexibility as well as the choice of the appropriate functional form. The application of a flexible functional form as the translog specification by the majority of technical efficiency studies is adequate with respect to economic theory. However, most applications do not adequately test for whether the estimated function has the required regularities of monotonicity and quasi-concavity, and hence run the risk of making improper policy recommendations. The researcher has to check a posteriori for the regularity of the estimated frontier which means checking these requirements for each and every data point with respect to the translog specification. If these requirements do not hold they have to be imposed a priori to estimation as briefly outlined in the text. Imposing global regularity nevertheless leads to a significant loss of functional flexibility, local imposition requires a differentiated interpretation: if theoretical consistency holds for a range of observations, this ‘consistency area’ of the estimated frontier should be determined and clearly stated to the reader. Estimated relative efficiency scores hence only hold for observations which are part of this range. Alternatively flexible functional forms – as e.g. the symmetric generalized McFadden – could be used which can be accomodated to global theoretical consistency over the whole range of observations. Furthermore one should always check for a possibility of using dual concepts such as the profit or cost function with respect to the efficiency measurement problem in question. Hence, policy measures based on such efficiency estimates are not subject to possible inadequacy and a waste of scarce resources. Here exemplary applications already exist in the literature. The test for theoretical consistency of an arbitrary selected sample of translog production frontiers published in development relevant literature in recent years revealed the significance of this problem for daily efficiency measurement as well as policy formulation.

A Comparison between Direct and Indirect Measures of Efficiency:

Roxana Ciumara

In this paper we compare direct and indirect measures of efficiency through their properties. We will prove that, in certain conditions, the inverse of the indirect measure of technical efficiency output orientated satisfies the same properties as the measure of cost efficiency. Moreover, the inverse of technical gain - defined as the ratio of indirect to direct measure of technical efficiency output orientated - has the same properties as the measure of allocative efficiency. For a data set from the Romanian Textile Industry, we compute the direct and indirect measures of technical efficiency, the allocative and cost efficiency. The results obtained confirmed our expectations. Then, we derive the indirect measure of technical efficiency output orientated when some inputs are considered fixed and we compare them to the scores found by solving a DEA - type problem in which the same inputs are kept fixed.

10d

Global returns to scale in efficiency analysis:

Victor Podinovski

Global returns to scale (GRS) is a new concept suitable for efficiency analysis with an arbitrary underlying production technology. The types of GRS are indicative of the location of the most productive scale size (MPSS) of the decision making unit (DMU) under consideration. This contrasts with the traditional local classification of RTS, where the latter is indicative of the direction of marginal rescaling that the DMU should undertake in order to improve its productivity. In a nonconvex technology the local and global characterisations of RTS are generally different. The GRS classes are useful in assisting strategic decisions like those involving mergers of units or splitting into smaller firms. It is shown that, in a non-convex technology, the well-known method of testing RTS proposed by Färe et al is in fact testing for GRS and not RTS. Further, while there are three types of RTS: constant, decreasing and increasing, the classification according to GRS includes the fourth type of sub-constant GRS, which describes a DMU able to achieve its MPSS by both reducing and increasing the scale of operations. The notion of GRS is applicable to a wide range of technologies, including the free disposal hull (FDH) and all polyhedral technologies used in data envelopment analysis (DEA)

Estimation Excess Capital Technical and Allocative Inefficiency:

Hung-Jen Wang - Subal Kumbhakar

For many empirical studies using production and cost functions, the input variables are often measured imprecisely. In many cases these measurement errors are one-sided. For example, when idle or obsolete machinery is not properly adjusted in the accounting data, the reported capital stock would over-estimate the actual stock in use. In such a case, reported data on capital stock tend be systematically biased upward. In this paper, we consider a cost minimization model where the measurement of the capital input is systematically biased upward. The setup of the model is somewhat similar to Schmidt and Lovell (1979), except for the fact that we distinguish between measured capital and capital actually used in production. We consider both a single and a two-step estimation methods. The model is applied to the estimation of a US agriculture panel data. The results show that the reported capital overstates the actual capital in use by 2% on average, and that the technical inefficiency reduces output by 12%.

Returns to Scale in DEA with non-discretionary inputs and outputs.:

Ole olesen - niels christian petersen

This paper discusses methods for determining returns to scale in DEA when non-discretionary inputs and/or outputs are present. Returns to scale in multiple input multiple output is a well known problem. Banker and Thrall (1992) derive a measure of returns to scale based upon the standard definition, which considers the ratio between marginal proportionate changes in outputs and inputs. This approach can not be applied when non-discretionary categories are present because radial expansions involving such categories are not allowed. The procedure suggested in this paper involves directional derivatives keeping the level of the non-discretionary categories fixed. The suggested approach can be extended to the case of negative inputs and outputs; this extension will be discussed briefly.

DIRECT AND INDIRECT MEASURES OF CAPACITY UTILIZATION: A NONPARAMETRIC ANALYSIS OF U.S. MANUFACTURING:

Subhash Ray - Kankana Mukherjee - Yanna Wu

The capacity output of a firm can be defined in two alternative ways. The first definition, due to Johansen (1968), known as the physical measure of capacity output, measures the maximum quantity of output that a firm can produce from a given bundle of quasi-fixed inputs even when other (variable) inputs are available without any restriction. The second definition, due to Klein (1960) is the economic measure of capacity output and corresponds to that output where the firm’s short run average cost curve is tangent to the long run average cost curve. Färe, Grosskopf, and Kokkelenberg (1989) provide a nonparametric model using Data Envelopment Analysis for measuring the physical capacity output and the associated rate of capacity utilization in the presence of fixed inputs. In neoclassical economics, a firm, unlike a consumer, does not face a budget constraint. However, there are various reasons related to capital markets why a firm may face budgetary constraints in the short run. In the short run if the firm cannot easily raise capital by issuing new equity then borrowing remains the only plausible way to finance additional expenses. Bur this could adversely affect the debt-to-equity ratio of the firm, which in turn could adversely affect its valuation. Also sometimes borrowing on short notice can only be at unfavorable interest rates. In the short run a quasi-fixed input is held constant due to high adjustment costs. In a comparable manner, the firm might want to keep its total operating expenditure within a budgetary limit to avoid excessive cost of credit and adverse market reaction. Taking this into consideration in this paper we measure the capacity output of a firm as the maximum quantity producible by the firm given a specific quantity of the quasi-fixed input and an overall budget constraint for its choice of variable inputs. Our measures of short run capacity output and the associated rate of capacity utilization are based on a restricted version of the indirect production function introduced by Shephard (1970). The present study extends the nonparametric literature by modeling the restricted indirect production function and deriving a measure of capacity utilization using DEA. We apply this methodology to compute capacity utilization measures in the total manufacturing sector in the US as well as in a number of disaggregated industries within manufacturing for the period 1970-2001 using data from the Bureau of Labor Statistics. Our empirical analysis show considerable variation in capacity utilization rates both across industries and over years within industries. Comparison of the direct measure of capacity output (based on FGK, 1989) and our indirect measure of capacity output shows that in the overall manufacturing sector there is considerable difference between the two measures especially since the 1980s, indicating that the expenditure constraint was binding. In case of individual industries some industries show more variation between the two measures of capacity output (for example primary metals) as compared to other industries (for example electrical and electronic equipment).