A comparison of stochastic frontier approaches for estimating technical inefficiency and total factor productivity

This article compares standard stochastic frontier models for panel data with a number of recently developed models which attempt to control for unobserved heterogeneity in the inefficiency component. Results are used to construct a generalized Malmquist Total Factor Productivity (TFP) index for the Irish tillage sector. While our application yields similar general TFP trends across models, it is evident that this new class of model leads to fewer theoretical inconsistencies in the production frontier. Furthermore, inefficiency estimates across models are critically compared and the potential benefits of controlling for unobserved heterogeneity are highlighted.     

Using job changes to evaluate the bias of value of a statistical life estimates

This paper presents a new approach to obtaining unbiased estimates of the value of a statistical life (VSL) with labor market data. Investigating job changes, we combine the advantages of recent panel studies, allowing us to control for unobserved heterogeneity of workers, and conventional cross-sectional estimations, which are less affected by measurement error. We find a VSL of 6.1 million euros from pooled cross-sectional estimation, 1.5 million euros from the first-differences panel model and 2.0 million euros from the job-changer specification. Thus, ignoring individual heterogeneity causes overestimates of the VSL, whereas the attenuation bias in panel data models leads to underestimates of the VSL. Our results are less biased than former results and can be used to perform cost–benefit analyses of public projects aimed at reducing fatality risks, e.g., in the domains of environmental, health, or traffic policy.     

Specification tests and tests for overidentifying restrictions in panel data models with selection

Standard tests are generally not applicable in panel data models with selection. The paper shows how the Hausman specification test and the Sargan-Hansen test for overidentifying restrictions can be generalized to panel data models with unobserved heterogeneity and sample selection.     

Efficiency Measurement in Network Industries: Application to the Swiss Railway Companies

The persistence of increasingly high government subsidies in Switzerland’s railroads has led the federal and cantonal authorities to discussing the possibility of high-powered incentive contracts such as those based on cost efficiency benchmarking. Railways are however, characterized by a high degree of unobserved heterogeneity that could bias the efficiency estimates. This paper examines the performance of several panel data models to measure cost efficiency in network industries. The unobserved firm-specific effects and the resulting biases are studied through a comparative study of several stochastic frontier models, applied to a panel of 50 railway companies operating over a 13-year period.* The authors wish to thank Michael Crew and two anonymous referees for their helpful suggestions. Aurelio Fetz provided an excellent assistance, which is gratefully acknowledged. Any remaining errors are the responsibility of the authors.This revised version was published in June 2005 with corrections in the author affiliations.     

COMPARATIVE EFFICIENCY BETWEEN COOPERATIVE,SAVINGS AND COMMERCIAL BANKS IN EUROPE USING THE FRONTIER APPROACH

We test competing hypotheses concerning the comparative behavior of shareholder‐owned commercial banks and stakeholder‐orientated cooperative and savings banks in European banking. One hypothesis is that the risk culture and business models of stakeholder and shareholder‐owned banks have become more alike and so cost efficiency has converged between bank ownership structures. The alternative hypothesis suggests that institutional differences do matter and lead, amongst other things, to variation in network effects and monitoring mechanisms producing differing behaviors and efficiency outcomes. By using a novel panel data set of 521 European banks during 1994–2010, we find: (i) mean inefficiency scores vary by ownership type and are lower for cooperative banks than for commercial and savings banks; (ii) there is a large variation in inefficiency scores among banks within each ownership type but the lower variance for cooperative banks indicates that they are the most homogeneous group; (iii) the inefficiency distribution of savings and commercial banks appear to arise from the same distribution, but this does not hold for cooperative banks. As such our findings are more consistent with the alternative hypothesis. Our first two findings buttress those studies that found significant differences between European banks with differing ownership structures, while our third finding on the significance of the cycle to the distribution of inefficiency is novel.     

Stochastic Cost Frontier and Inefficiency Estimates of Public and Private Universities: Does Government Matter?

Stochastic frontier analysis is used to estimate operating cost inefficiencies of public and private non-profit universities in the U.S. while also accounting for the possible effects arising from differences in the degree of government ownership. Using panel data for four academic years, 2005?C2009, inefficiencies are estimated under two model specifications. Results indicate that public universities are more cost efficient when environmental factors influence cost frontiers but private universities are the cost efficient institutions when those factors are determinants of inefficiency. Increased government funding does matter and increases private sector inefficiency but offers some efficiency improvements among public universities. Following the global financial crisis, there is evidence indicating a considerable slowdown in the inefficiency growth among both public and private universities.     

Benchmarking and firm heterogeneity: a latent class analysis for German electricity distribution companies

In January 2009, the German Federal Network Agency introduced incentive regulation for the electricity distribution sector based on results obtained from econometric and nonparametric benchmarking analysis. One main problem for the regulator in assigning the relative efficiency scores is unobserved firm-specific factors such as network and technological differences. Comparing the efficiency of different firms usually assumes that they operate under the same production technology, thus unobserved factors might be inappropriately understood as inefficiency. To avoid this type of misspecification in regulatory practice, estimation is carried out in two stages: in the first stage observations are classified into two categories according to the size of the network operators. Then separate analyses are conducted for each subgroup. This article shows how to disentangle the heterogeneity from inefficiency in one step, using a latent class model for stochastic frontiers. As the classification is not based on a priori sample separation criteria it delivers more robust, statistically significant, and testable results. Against this background, we analyze the level of technical efficiency of different subgroups from a sample of 200 regional and local German electricity distribution companies for a balanced panel data set (2001–2005). Testing the hypothesis if larger distributors operate under a different technology than smaller ones, we assess if a single step latent class model provides new insights to the use of benchmarking approaches within the incentive regulation schemes.     

Labour Market Inefficiency and Frictional Unemployment in Australia and its States: A Stochastic Frontier Approach

This paper provides estimates of labour market inefficiency and the frictional unemployment rate for Australia and its States over the period January 1978 to December 1997. These estimates are derived from parametric statistical models of employment growth in which technical inefficiencies are accounted for. The mean estimate of the (technically efficient) frictional unemployment rate for Australia over the sample period is S3 per cent of the labour force. Technical inefficiency in the labour market matching process is significant and contributes around 13 per cent to the mean steady-state (naturaľ) unemployment rate. Investigation of the factors explaining the levels of inefficiency suggests that inefficiencies vary countercyclical, are related to which political party is in power and the time of year and that only Western Australia and Queensland have exhibited a significant decline in inefficiency over the period.     

On the choice of functional form in stochastic frontier modeling

This paper examines the effect of functional form specification on the estimation of technical efficiency using a panel data set of 125 olive-growing farms in Greece for the period 1987–93. The generalized quadratic Box-Cox transformation is used to test the relative performance of alternative, widely used, functional forms and to examine the effect of prior choice on final efficiency estimates. Other than the functional specifications nested within the Box-Cox transformation, the comparative analysis includes the minflex Laurent translog and generalized Leontief that possess desirable approximation properties. The results indicate that technical efficiency measures are very sensitive to the choice of functional specification. Perhaps most importantly, the choice of functional form affects the identification of the factors affecting individual performance – the sources of technical inefficiency. The analysis also shows that while specification searches do narrow down the set of feasible alternatives, the identification of the most appropriate functional specification might not always be (statistically) feasible. First version received: November 1999/Final version received: July 2001 RID="*" ID="*"  The authors wish to thank Almas Heshmati, Robert Romain, and an anonymous referee for insightful comments and suggestions. Special thanks go to the associate editor who handled the paper, and whose careful reading and suggestions have improved the paper substantially. The second author wishes to acknowledge the financial support from “President SSHRC” from the University of Saskatchewan. The usual caveats with respect to opinions expressed in the paper apply. Senior authorship is shared. This is University of Nebraska-Lincoln Agricultural Research Division Article No. 13270.     

Economic inefficiency in a revenue setting: the Norwegian whitefish fishery

The main focus in the inefficiency literature is on suboptimal input use and how this causes increased costs, due to technical and allocative inefficiency. Production or cost functions are then typically used to describe the underlying technology of the firm. The possible revenue loss, due to lower than maximum production levels and suboptimal output mix, has received substantially less attention. By using a revenue function to measure inefficiency, the focus, model and estimation technique presented in this article differ from those of previous studies. A shadow revenue model is used to decompose revenue inefficiency into its technical and allocative components, in which the allocative inefficiency is due to a suboptimal output mix. The approach is illustrated using panel data of Norwegian whitefish trawlers. The results reveal large inefficiencies, with respect to output levels as well as output mix, indicating that this can be an important part of the picture when investigating economic inefficiency. To identify the determinants of revenue inefficiency, we conduct a second-step regression, in which technical and allocative inefficiency is regressed upon a set of explanatory variables. The inefficiencies are partly explained by the management system and fleet structure.     

Measurement of technical and allocative efficiencies using a CES cost frontier: a benchmarking study of Japanese transmission-distribution electricity

This paper estimates the technical and allocative inefficiencies of the transmission-distribution sector of Japanese electric utilities using a panel data during the 1981–1998 period. A stochastic production frontier of the CES form is jointly estimated with input demand equations. Taking advantage of the self-duality, we retrieve the cost frontier by which the impacts of technical and allocative inefficiencies on costs and input demands are measured. The estimated elasticity of substitution is significantly different from unity in favor of the CES specification over the Cobb–Douglas. The results show that observed costs are 9 to 48% higher than the efficient level; technical inefficiency raises costs by 1 to 28%, while allocative inefficiency does so by 8 to 30%. Although their impacts on costs are similar, technical inefficiency more fluctuates so the differences in the performance of utilities are mainly due to technical inefficiency. We also find a substantial over-utilization of capital for all utilities.

Is it Beneficial to be Included in a Sustainability Stock Index? A Panel Data Study for European Firms

This paper empirically examines the effect of inclusion in the Dow Jones Sustainability World Index (DJSI World) on corporate financial performance. On the basis of panel data for European firms that were included in the Dow Jones Stoxx 600 Index over time, our micro-econometric analysis with fixed and random effects models implies positive impacts on return on assets for continental European countries, but insignificant effects for Anglo-Saxon European countries (i.e., the United Kingdom and Ireland). Furthermore, the impacts on alternative indicators of corporate financial performance such as Tobin’s Q are generally insignificant. The weak or neutral effect of inclusion in the DJSI World on corporate financial performance can be explained by several mutually confounding factors. Furthermore, the composition of this sustainability stock index is influenced by factors that need not necessarily be directly connected to corporate environmental or social activities, so that potential positive and negative effects of corporate sustainability performance on financial performance can be weakened. Methodologically, this study again supports the strong relevance of unobserved firm heterogeneity since the application of misspecified pooled regression models leads to obviously biased estimation results.     

THE IMPACT OF PUBLICLY PROVIDED EARLY CHILDHOOD EDUCATION PROGRAMS ON DISTRICT‐LEVEL TEST SCORES

While the existing literature largely finds that standardized test scores are positively associated with participation in pre‐kindergarten programs at the student level, there is little research on the policy effect of publicly providing these programs to entire school districts. We attempt to partially fill this void by examining the effect that publicly provided pre‐kindergarten and full‐day kindergarten programs have on fourth grade student achievement in math and reading at the district level. The models utilize panel data from districts in Wisconsin along with fixed effects estimations. We find that once time‐invariant unobserved heterogeneity of a district is accounted for, preschool programs have small but significant impacts on a cohort's future math scores, particularly in less‐educated communities. Additionally, there is no evidence that preschool or full‐day kindergarten has an impact on future reading scores at the district level. (JEL I20, I28)     

Dispersion in wage premiums and firm performance

Using matched employer–employee panel data, we estimate measures of pay dispersion per firm-year that take into account both firm and worker unobserved heterogeneity. Unlike research that controls only for differences in observables, we find that within-firm pay inequality is significantly associated to lower firm performance.     

Spatial stochastic frontier models: controlling spatial global and local heterogeneity

In the last decade special attention has been focused on estimating a firm’s efficiency and productivity; Stochastic Frontier Analysis (SFA) has been one of the most used techniques that allows the separation of inefficiency from stochastic noise, assuming homogeneous technology is available to all producers and that there is independence between observations. However, this second assumption is violated data are spatial auto-correlated, thus biasing statistical inference. Attention has, therefore, shifted to models that allow the controlling of heterogeneity introducing, in the model or in the error term, contextual variables correlated with inefficiency. In our paper we propose viewing the spatial external factors (natural or artificial) in a new way: instead of identifying ex-ante a multitude of determinants, often statistically and economically difficult to detect, we suggested using an original methodology that, following a classical SFA approach, splits efficiency into three components: the first one is linked to the spatial lag, the second one to the DMU’s specificities, and the third to the error term. Finally, we tested our method using simulated data and examined the Italian wine sector, testing the ability to control spatial, global and local heterogeneity.     

Heterogeneity in Consumer Demands and the Income Effect: Evidence from Panel Data

This paper uses unique Spanish panel data on household expenditures to test whether unobservable heterogeneity in household demands (taste, etc.) is correlated with total expenditures (income). The main finding is that tastes are indeed correlated with income for about half of the goods considered, implying that cross‐sectional estimates of income elasticities for these goods are biased. The goods are the following: food eaten outside home, alcohol and tobacco, transportation, and energy. The elasticity of alcohol and tobacco is more than halved when taking unobserved heterogeneity into account. For transportation, the bias is sufficiently large to misclassify the good as a luxury.     

Predicting land use allocation in France: A spatial panel data analysis

Predictions of future land use areas are an important issue as land use patterns significantly impact environmental conditions (biodiversity, water pollution, soil erosion, and climate change) as well as economic and social welfare. In order to improve the prediction accuracy of aggregated land use share models, we propose in this paper a methodological contribution by controlling for both unobserved individual heterogeneity and spatial autocorrelation. Our model is a land use shares model applied to aggregated data in France. Our dataset is a panel which covers both time series observations from 1992 to 2003 and cross-sectional observations by Département (equivalent to NUTS3 regions). We consider four land use classes: (1) agriculture, (2) forest, (3) urban and (4) other use. We investigate the relation between the areas in land in different alternative uses and economic and demographic factors influencing land use decisions. Based on the comparison of prediction accuracy of different model specifications, our findings are threefold: First, controlling for both unobserved individual heterogeneity and spatial autocorrelation outperforms any other specification in which spatial autocorrelation and/or individual heterogeneity are ignored. Second, accounting for cross-equation correlations does not seem to improve the prediction performances and finally, ignoring individual heterogeneity introduces substantial loss in prediction accuracy.     

Bank Ownership and Efficiency in the Aftermath of Financial Crises: Evidence from Indonesia

This paper examines the relationship between types of ownership of banks and their efficiency in the aftermath of a financial crisis using Greene's “true” panel data stochastic frontier model, which takes into account unobserved heterogeneity among banks. The Indonesian banking sector is analyzed using financial data of 144 banks operating in Indonesia over the period of 2000Q4–2005Q2. In the aftermath of the 1997 Asian financial crisis, the cost efficiency of all banks improves over time on average. However, there is some evidence that, as these banks improve their efficiency, state‐owned banks are the least efficient banks while joint‐venture and foreign‐owned banks are the most efficient.     

Efficiency and Equity in School Funding: A Case Study for Kansas

This study measures cost inefficiency of Kansas public school districts and applied both mathematical programming and stochastic frontier approach. The empirical study uses two-stage data envelopment analysis model and the cost inefficiency effects model proposed by Battese and Coelli (Empirical Economics 24:325–332, 1995) and applied to a panel data. The results found mean inefficiencies from these two models are very close. The results indicate that Kansas school districts, on average, exhibit cost inefficiency in their operations, however, there is a tendency for inefficiencies to decline over time. The study does not find any strong evidence for lower efficiency due to lower expenditure per-pupil. Instead, we found inconclusive evidences where lower efficiency for certain school districts could be assigned to unfavorable environmental cost conditions.

Ethnic Networks and US Exports*

This paper provides new estimates of the effects of ethnic networks on US exports. In line with recent research, our dataset is a panel of exports from US states to 29 foreign countries. Our analysis departs from the literature in two ways, both of which show that previous estimates of the ethnic‐network elasticity of trade are sensitive to the restrictions imposed on the estimated models. Our first departure is to control for unobserved heterogeneity with properly specified fixed effects, which we can do because our dataset contains a time dimension absent from previous studies. Our second departure is to remove the restriction that the network effect is the same for all ethnicities. We find that ethnic‐network effects are much larger than has been estimated previously, although they are important only for a subset of countries.