Abstract
This contribution focuses on testing the empirical impact of the convexity assumption in estimating costs using nonparametric specifications of technology and cost functions. Apart from reviewing the scant available evidence, the empirical results based on two publicly available data sets reveal the effect of the convexity axiom on cost function estimates: cost estimates based on convex technologies turn out to be on average between 21% and 38% lower than those computed on nonconvex technologies. These differences are statistically significant when comparing kernel densities and can be illustrated using sections of the cost function estimates along some output dimension. Finally, also the characterization of returns to scale and economies of scale using production and cost functions for individual units yields conflicting results for between 19% and 31% of individual observations. The theoretical known potential impact as well as these empirical results should make us reconsider convexity in empirical production analysis: clearly, convexity is not harmless.
Similar content being viewed by others
Notes
The same reasoning applies to the revenue function and all variations of the profit function, except of course the long run profit function where the effect of convexity or not cannot logically be distinguished.
Budish et al. (2014) claim that the current defects in market microstructure are linked to a continuous-time market design based on a continuous limit order book. Instead, they propose a discrete-time frequent batch auction.
Bogetoft et al. (2000) is the first article maintaining convexity on input and output space, but relaxing convexity in the graph of technology without restrictions on the dimensionality of input and output spaces. Li (2019) adds variable returns to scale to this quasiconcavity of production, but only manages to treat the single output case. This framework is justified by appeal to, for instance, the law of diminishing marginal rates of substitution in inputs space, or to the idea of diminishing marginal rates of transformation in outputs space. But, one may question whether there really is a law of diminishing marginal rates of substitution in inputs space. For example, Brokken (1977) summarises three studies revealing that there are increasing marginal rates of substitution of grain for roughage in beef production: thus, establishing nonconvexities in input space.
Just as a nonconvex or a convex technology yields the same profit function. This implies that convexity cannot logically be tested using the profit function (even though any variation on the long-run profit function would allow testing for convexity).
If one is willing to accept CRS, then the plausibility of additivity and divisibility separately is at stake. First, perfect divisibility of inputs and/or outputs is probably the most debatable assumption. Many if not most operations management problems in industry and distribution involve some form of indivisibilities and input fixities resulting in complex integer and possibly nonlinear optimization problems. In general, all production processes seem to have some lower limit below which a process cannot possibly be scaled down realistically. Therefore, divisibility is highly questionable (see Scarf (1994) or Winter (2008) for more detailed criticisms). Second, while additivity is essential to define free entry and may seem plausible at first, it is not beyond criticism since it presupposes spatial separation and noninteraction (see Winter (2008)). Since additivity relates to the aggregation of results of activities occurring in geographically distinct places, transportation costs must be small in order to be safely ignored. When activities are close for transportation costs to be negligible, then the risk of production externalities looms when activities get “too close” to create interactions. Furthermore, location is important in the definition of quite some outputs (e.g., Italian and Californian lemons are considered different).
The Hasenkamp (1976) results are at least partially attributable to the strong assumptions on input/output separability, among others.
While some studies offer explicit comparisons, some results in Table 1 are based on own computations. In particular, some studies report cost efficiency ratios based on cost frontier estimates under convexity and nonconvexity. Taking a ratio of these efficiency ratios nets out the observed cost and reveals the difference in cost estimates under convexity and nonconvexity.
These authors also report results for six size classes each time for solvent and insolvent firms separately: the difference varies between a minimum of 9.38% and a maximum of 29.63%.
Nonconvex results are presented graphically. Our numbers are based on information provided by the author.
For the dynamic profit frontiers we report the difference between convex and nonconvex results relative to nonconvex results. We thank F. Ang for providing these results in personal communication.
Nonparametric deterministic frontier estimators that minimally extrapolate the data subject to some maintained axioms are labeled Data Envelopment Analysis (DEA) models in contrast to different (e.g., parametric) methodologies. The same moniker is also used to distinguish these convex models from the nonconvex ones in (5) that are sometimes labeled Free Disposal Hull (FDH) technologies. This use of monikers to denote both general methodologies and specific models is confusing.
Substituting \(t_k = \delta z_k\) in (5), one rewrites the sum constraint on the activity vector. Realizing that the constraints on the scaling factor are integrated into the latter sum constraint, the LP appears.
Badunenko et al. (2012) show that the reliability of efficiency scores of nonparametric, deterministic convex frontier estimation remains excellent when the ratio of the variation in efficiency to the variation in noise is low. Assuming this result would also be confirmed for nonconvex similar estimators and would also hold for the unexplored case of the cost function, this reinforces our argument to opt for a simple deterministic nonparametric frontier framework to test for convexity.
To enhance comparability we use a common Sheather and Jones plug-in bandwidth for the convex and nonconvex data series to be compared.
Indeed, in the single output case it may be intuitively clear that convex and nonconvex cost functions trace the same function at the extremes of the empirical range, but that in between the approximation depends on the precise structure of the sample. This intuition for the single output case is developed in the figures in Appendix C in Electronic Supplementary material by computing the models in Sect. 3.1 on a numerical example with two inputs and a single output. However, the situation is less clear in the multi-output case: convex and nonconvex cost functions need not have very much in common except at some of the extremes of the empirical range (see Appendix D in Electronic Supplementary material for figures computed on a numerical example with one input and two outputs).
As pointed out by a referee, there is a multitude of approaches to compute the difference between two functions and thus to assess global fit.
As pointed out by the same referee, to the best of our knowledge it remains an open question to which extent this better fit of the nonconvex cost function may lead to overfitting when predicting.
Results in, e.g., Ray and Kim (1995) learn that accounting for nonregressive technical change increases technical efficiency, thereby improving the empirical fit. However, their panel data set covers a long time period, while ours have a short time dimension.
References
Afriat, S. (1972). Efficiency estimation of production functions. International Economic Review, 13(3), 568–598.
Alam, I., & Sickles, R. (2000). Time series analysis of deregulatory dynamics and technical efficiency: the case of the US airline industry. International Economic Review, 41(1), 203–218.
Allahverdi, A., Ng, C., Cheng, T., & Kovalyov, M. (2008). A survey of scheduling problems with setup times or costs. European Journal of Operational Research, 187(3), 985–1032.
Ang, F., Mortimer, S., Areal, F., & Tiffin, R. (2018). On the opportunity cost of crop diversification. Journal of Agricultural Economics, 69(3), 794–814.
Atkinson, S., & Dorfman, J. (2009). Feasible estimation of firm-specific allocative inefficiency through Bayesian numerical methods. Journal of Applied Econometrics, 24(4), 675–697.
Badunenko, O., Henderson, D., & Kumbhakar, S. (2012). When, where and how to perform efficiency estimation. Journal of the Royal Statistical Society, 175A(4), 863–892.
Balaguer-Coll, M., Prior, D., & Tortosa-Ausina, E. (2007). On the determinants of local government performance: a two-stage nonparametric approach. European Economic Review, 51(2), 425–451.
Barnett, W. (2002). Tastes and technology: curvature is not sufficient for regularity. Journal of Econometrics, 108(1), 199–202.
Bjørndal, M., & Jörnsten, K. (2008). Equilibrium prices supported by dual price functions in markets with non-convexities. European Journal of Operational Research, 190(3), 768–789.
Bogetoft, P., Tama, J., & Tind, J. (2000). Convex input and output projections of nonconvex production possibility sets. Management Science, 46(6), 858–869.
Briec, W., Kerstens, K., & Vanden Eeckaut, P. (2004). Non-convex technologies and cost functions: definitions, duality and nonparametric tests of convexity. Journal of Economics, 81(2), 155–192.
Brokken, R. (1977). The case of a queer isoquant: increasing marginal rates of substitution of grain for roughage in cattle finishing. Western Journal of Agricultural Economics, 1(1), 221–224.
Budish, E., Cramton, P., & Shim, J. (2014). Implementation details for frequent batch auctions: slowing down markets to the blink of an eye. American Economic Review, 104(5), 418–424.
Cesaroni, G., & Giovannola, D. (2015). Average-cost efficiency and optimal scale sizes in non-parametric analysis. European Journal of Operational Research, 242(1), 121–133.
Cesaroni, G., Kerstens, K., & Van de Woestyne, I. (2017a). Estimating scale economies in non-convex production models. Journal of the Operational Research Society, 68(11), 1442–1451.
Cesaroni, G., Kerstens, K., & Van de Woestyne, I. (2017b). Global and local scale characteristics in convex and nonconvex nonparametric technologies: a first empirical exploration. European Journal of Operational Research, 259(2), 576–586.
Chavas, J.-P., & Cox, T. (1999). A generalized distance function and the analysis of production efficiency. Southern Economic Journal, 66(2), 294–318.
Copeland, A., & Hall, G. (2011). The response of prices, sales, and output to temporary changes in demand. Journal of Applied Econometrics, 26(2), 232–269.
Coviello, D., Ichino, A., & Persico, N. (2014). Time allocation and task juggling. American Economic Review, 104(2), 609–623.
Cummins, D., & Zi, H. (1998). Comparison of frontier efficiency methods: an application to the U.S. life insurance industry. Journal of Productivity Analysis, 10(2), 131–152.
De Borger, B., & Kerstens, K. (1996). Cost efficiency of Belgian local governments: a comparative analysis of FDH, DEA, and econometric approaches. Regional Science and Urban Economics, 26(2), 145–170.
Diewert, W., & Parkan, C. (1983). Linear programming test of regularity conditions for production functions. In W. Eichhorn, K. Neumann, & R. Shephard (Eds.), Quantitative Studies on Production and Prices (pp. 131–158). Würzburg: Physica-Verlag.
Eaton, B., & Lipsey, R. (1997). On the foundations of monopolistic competition and economic geography: the selected essays of B. Curtis Eaton and Richard G. Lipsey. Cheltenham: Edward Elgar.
Estache, A., Rossi, M., & Ruzzier, C. (2004). The case for international coordination of electricity regulation: evidence from the measurement of efficiency in South America. Journal of Regulatory Economics, 25(3), 271–295.
Fan, Y., & Ullah, A. (1999). On goodness-of-fit tests for weakly dependent processes using kernel method. Journal of Nonparametric Statistics, 11(1), 337–360.
Färe, R. (1988). Fundamentals of production theory. Berlin: Springer.
Farrell, M. (1959). The convexity assumption in the theory of competitive markets. Journal of Political Economy, 67(4), 377–391.
Fuss, M., McFadden, D., & Mundlak, Y. (1978). A survey of functional forms in the economic analysis of production. In M. Fuss & D. McFadden (Eds.), Production economics: a dual approach to theory and applications (Vol. 1, pp. 219–268). Amsterdam: North-Holland.
Garbaccio, R., Hermalin, B., & Wallace, N. (1994). A comparison of nonparametric methods to measure efficiency in the savings and loan industry. Real Estate Economics, 22(1), 169–193.
Grifell-Tatjé, E., & Kerstens, K. (2008). Incentive regulation and the role of convexity in benchmarking electricity distribution: economists versus engineers. Annals of Public and Cooperative Economics, 79(2), 227–248.
Griffin, J. (1979). Statistical cost analysis revisited. Quarterly Journal of Economics, 93(1), 107–129.
Hackman, S. (2008). Production economics: integrating the microeconomic and engineering perspectives. Berlin: Springer.
Hackman, S., Passy, U., & Platzman, L. (1994). Explicit representation of the two-dimensional section of a production possibility set. Journal of Productivity Analysis, 5(2), 161–170.
Hasenkamp, G. (1976). A study of multiple-output production functions. Journal of Econometrics, 4(3), 253–262.
Hung, N., Le Van, C., & Michel, P. (2009). Non-convex aggregate technology and optimal economic growth. Economic Theory, 40(3), 457–471.
Ivaldi, M., Ladoux, N., Ossard, H., & Simioni, M. (1996). Comparing Fourier and translog specifications of multiproduct technology: evidence from an incomplete panel of french farmers. Journal of Applied Econometrics, 11(6), 649–668.
Izadi, H., Johnes, G., Oskrochi, R., & Crouchley, R. (2002). Stochastic frontier estimation of a CES cost function: the case of higher education in Britain. Economics of Education Review, 21(1), 63–71.
Jacobsen, S. (1970). Production correspondences. Econometrica, 38(5), 754–771.
Kerstens, K., & Managi, S. (2012). Total factor productivity growth and convergence in the petroleum industry: empirical analysis testing for convexity. International Journal of Production Economics, 139(1), 196–206.
Kerstens, K., Sadeghi, J., & Van de Woestyne, I. (2019). Convex and nonconvex input-oriented technical and economic capacity measures: an empirical comparison. European Journal of Operational Research, 276(2), 699–709.
Kerstens, K., & Vanden Eeckaut, P. (1999). Estimating returns to scale using nonparametric deterministic technologies: a new method based on goodness-of-fit. European Journal of Operational Research, 113(1), 206–214.
Klein, L. (1960). Some theoretical issues in the measurement of capacity. Econometrica, 28(2), 272–286.
Krivonozhko, V., Utkin, O., Volodin, A., Sablin, I., & Patrin, M. (2004). Constructions of economic functions and calculations of marginal rates in DEA using parametric optimization methods. Journal of the Operational Research Society, 55(10), 1049–1058.
Kumbhakar, S., & Lovell, C. (2000). Stochastic frontier analysis: an econometric approach. Cambridge: Cambridge University Press.
Li, Q. (1996). Nonparametric testing of closeness between two unknown distribution functions. Econometric Reviews, 15(1), 261–274.
Li, S.-K. (2019). A nonparametric test of quasiconcave production function with variable returns to scale. Journal of Mathematical Economics, 82, 160–170.
Maddala, G., & Roberts, R. (1981). Statistical cost analysis re-revisited: comment. Quarterly Journal of Economics, 96(1), 177–182.
Mas-Colell, A., Whinston, A., & Green, J. (1995). Microeconomic theory. Oxford: Oxford University Press.
Narbón-Perpiñá, I., Balaguer-Coll, M., Petrović, M., & Tortosa-Ausina, E. (2020). Which estimator to measure local governments’ cost efficiency? the case of Spanish municipalities. SERIEs, 11(1), 51–82.
Narbón-Perpiñá, I., Balaguer-Coll, M., & Tortosa-Ausina, E. (2019). Evaluating local government performance in times of crisis. Local Government Studies, 45(1), 64–100.
Olesen, O., & Petersen, N. (2016). Stochastic data envelopment analysis-a review. European Journal of Operational Research, 251(1), 2–21.
Oude Lansink, A., Stefanou, S., & Kapelko, M. (2015). The impact of inefficiency on diversification. Journal of Productivity Analysis, 44(2), 189–198.
Podinovski, V. (2004a). Efficiency and global scale characteristics on the no free lunch assumption only. Journal of Productivity Analysis, 22(3), 227–257.
Podinovski, V. (2004b). Local and global returns to scale in performance measurement. Journal of the Operational Research Society, 55(2), 170–178.
Preckel, P., & Hertel, T. (1988). Approximating linear programs with summary functions: pseudodata with an infinite sample. American Journal of Agricultural Economics, 70(2), 397–402.
Ray, S. (1997). Weak axiom of cost dominance: a nonparametric test of cost efficiency without input quantity data. Journal of Productivity Analysis, 8(2), 151–165.
Ray, S. (2004). Data envelopment analysis: theory and techniques for economics and operations research. Cambridge: Cambridge University Press.
Ray, S., & Kim, H. (1995). Cost efficiency in the US steel industry: a nonparametric analysis using data envelopment analysis. European Journal of Operational Research, 80(3), 654–671.
Ray, S., & Mukherjee, K. (1995). Comparing parametric and nonparametric measures of efficiency: a reexamination of the Christensen-Greene data. Journal of Quantitative Economics, 11(1), 155–168.
Resti, A. (1997). Evaluating the cost-efficiency of the Italian banking system: What can be learned from the joint application of parametric and non-parametric techniques. Journal of Banking & Finance, 21(2), 221–250.
Romer, P. (1990). Are nonconvexities important for understanding growth? American Economic Review, 80(2), 97–103.
Sauer, J. (2006). Economic theory and econometric practice: parametric efficiency analysis. Empirical Economics, 31(4), 1061–1087.
Savani, R. (2012). High-frequency trading: the faster, the better? IEEE Intelligent Systems, 27(4), 70–74.
Scarf, H. (1986). Testing for optimality in the absence of convexity. In W. Heller, R. Starr, & S. Starrett (Eds.), Social choice and public decision making: Essays in honor of Kenneth J. Arrow (Vol. I, pp. 117–134). Cambridge: Cambridge University Press.
Scarf, H. (1994). The allocation of resources in the presence of indivisibilities. Journal of Economic Perspectives, 8(4), 111–128.
Shephard, R. (1970). Theory of cost and production functions. Princeton: Princeton University Press.
Simar, L., & Wilson, P. (2000). Statistical inference in nonparametric frontier models: the state of the art. Journal of Productivity Analysis, 13(1), 49–78.
Varian, H. (1984). The nonparametric approach to production analysis. Econometrica, 52(3), 579–597.
Viton, P. (2007). Cost efficiency in US air carrier operations, 1970–1984: a comparative study. International Journal of Transport Economics, 34(3), 369–401.
Wibe, S. (1984). Engineering production functions: a survey. Economica, 51(204), 401–411.
Winter, S. (2008). Scaling heuristics shape technology! should economic theory take notice? Industrial and Corporate Change, 17(3), 513–531.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We acknowledge the most helpful comments of G. Cesaroni, P. Kerstens, F. Maier-Rigaud and M. Vardanyan on earlier versions. We thank conference and seminar participants in Auckland, Berlin, Bordeaux, Edinburgh, Harbin, Helsinki, Houston, Karaj, Palmerston North, Rennes, Shanghai, Taipei and Tianjin for useful comments on earlier drafts. We are most grateful to two referees of this journal. The usual disclaimer applies. Note that this contribution has circulated under different titles.
Supplementary Information
Below is the link to the electronic supplementary material.
Rights and permissions
About this article
Cite this article
Kerstens, K., Van de Woestyne, I. Cost functions are nonconvex in the outputs when the technology is nonconvex: convexification is not harmless. Ann Oper Res 305, 81–106 (2021). https://doi.org/10.1007/s10479-021-04069-1
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10479-021-04069-1