Remembering David Kendrick (1937-2024)

It was with great sadness that we and many other colleagues, students, and friends learned about the passing of David Kendrick, Emeritus Professor of Economics at the University of Texas at Austin. He died on April 7, at an age of 86.

We had the privilege of interacting with him as a colleague, professor and friend, Alex most intensely during the early days of the development of GAMS at the World Bank and Hans as a graduate student at the University of Texas. David joined its Department of Economics in 1970, after having earned his Ph.D. from MIT in 1966 and taught as Assistant Professor at Harvard’s Economics Department 1966-1970. David was invariably exemplary, both as a distinguished economist, dedicated teacher, and caring mentor, opening his home to visitors and students alike. Throughout his career, his contributions to the field were of high quality, not driven by professional fads but by a desire to enhance and communicate research and tools in areas that he found promising. We will here take the opportunity to describe the parts of his legacy that are most directly related to GAMS community, in the process evoking some of the early intellectual history of GAMS and the people involved.¹

Already as a graduate student, David entered the field of computational economics. His Ph.D. dissertation at MIT, supervised by Richard S. Eckaus, developed linear and mixed-integer programming (LP and MIP) models for investment planning in process industries with geographically distributed and interdependent production units with economies of scale (Kendrick 1967). At the time, this was a research area that engaged several of the brightest minds of the field; among others, David’s dissertation draws on Alan Manne (Markowitz and Manne 1957; Manne and Markowitz, eds 1963). David’s application of his model to the steel industry in Brazil pointed to the payoffs from further advances in technology, both hardware and software – he reports that, when applying a mixed-integer model that by today’s standards was small (23 integer variables, 433 other variables, and 122 constraints), he was forced to end computations due to limitations on computer time after having found a solution that he considered “good” without necessarily being optimal. Beyond computer capacity, the steps needed to translate his algebraic model and its database into a computer-readable format and receive the results were cumbersome and error-prone.

During his years at Harvard, David was part of the Project for Quantitative Research in Economic Development led by Hollis Chenery, at the time Professor of Economics at Harvard. Among other things, the project produced a volume on development planning (Chenery 1971) that includes a chapter based on a non-linear economywide planning model that David coauthored with Lance Taylor, at the time one of his students, one of the CGE pioneers and later a leading heterodox macroeconomist (Kendrick and Taylor 1971); as indicated in the reference list, that model and several others referred to in this blog are represented in the GAMS library. Other traces of David’s years at Harvard include the volume “Notes and Problems in Microeconomic Theory,” the second edition of which is coauthored with Samuel Bowles and Peter Dixon, one of the students in the course that the first edition was based on (Dixon et al. 1980). As many in the GAMS community know, Peter has moved on to become a luminary in the field of CGE modeling, albeit as part of the GEMPACK community.

In 1972, Hollis Chenery became the World Bank’s Chief Economist, a position he held until 1982. He brought with him personal connections and a commitment to draw on quantitative structuralist economics in this institution’s research and operations. At about the same time, David started a 16-year spell of part-time consulting with the World Bank while Alex landed a staff position a few years after having left his native Austria. Alex had also been working on LP models, most recently for General Electric. The task of turning LP and other mathematical programming models into tools that could be built and applied with high efficiency – at a low cost in time and with built-in error checks to maintain high quality – led to GAMS, including the crucial early step of developing the first GAMS compiler, which Alex recalls was thanks to David’s networking with computer scientists. In coming years, their joint work on GAMS frequently brought Alex to Austin, on one occasion appearing in a graduate economics class that Hans attended in the mid-1980s.

Among the first public signs of GAMS’ emergence is a technical note with familiar-looking GAMS code that Alex coauthored with Johannes Bisschop in 1979, and a 1980 World Bank monograph by Choksi, Meeraus, and Stoutjesdijk on the fertilizer industry in Egypt (Bisschop and Meeraus 1979; and Choksi et al. 1980). The model by Choksi et al. applies the methodology documented in Kendrick and Stoutjesdijk (1978), which is close to what David developed for his dissertation: linear and mixed-integer programming models with linearization of non-linearities to permit economies of scale. During the next few years, this led to a series of sectoral studies, including oil refining, steel, and fertilizers in a multi-country setting (Kendrick et al. 1981 and 1984; Mennes and Stoutjesdijk 1985). The different pieces reflect David’s concern with style in modeling: models should be presented so that the reader with minimum effort can understand their structure and data, among other things putting brief explanations in words under each part of complex equations, something that he labeled “Manne notation”; not surprisingly good style leads to model presentations that are similar to a well-structured GAMS statement (Kendrick 1984). This paper also includes what may be the first application of a CGE model in GAMS – a miniature ORANI model, linearized in the Johansen tradition and shared by Peter Dixon (1979). A few years later, Condon et al. (1987) were the first to apply GAMS to a non-linear CGE model, drawing on the seminal work by Dervis et al. (1982) and benefitting from improved non-linear solvers. In parallel with the industrial sector work, GAMS was applied to agriculture, including spatial agricultural sector models that simulate equilibria in multiple markets, drawing on a formulation due to Paul Samuelson (1952) that Hans, for his dissertation under David’s supervision, adapted for a regional Egyptian context (Lofgren 1993). Kendrick (1996) surveys the broader area of sectoral economics, referring to much of the research described above, including several other dissertations that, under his supervision, applied GAMS (Adib 1985, Letson 1992, and Linden 1992), benefitting from the fact that, starting from the mid-1980s, he made GAMS available to department students on a mainframe computer and on floppy disks for use on PCs.²

David’s most active period of work on GAMS concluded with the publication of the first edition of the GAMS User’s Guide (Brooke et al. 1988). Up to his retirement in 2014, his courses in computational economics covered the sectoral and economywide model types referred to above, in later years using as their main text the volume “Computational Economics” which he coauthored with Ruben Mercado (a student of his) and Hans M. Amman, for many years his main collaborator (Kendrick et al. 2006).

David is remembered with fondness and appreciation by the two of us and many others who were fortunate to cross his path. He advised his students to stand on the shoulders of the previous generations and their models. Today’s computational economists are well advised to draw on and be inspired by his work in their future endeavors.

References

Adib, P. Manouchehri. 1985. An investment planning model of the world petrochemical industry. PhD dissertation, University of Texas, Austin, TX.

Bisschop, Johannes, and Alexander Meeraus. 1979. Selected Aspects of a General Algebraic Modeling Language. Technical Note. Development Research Center, World Bank.

Brooke, Anthony, Alexander Meeraus, and David Kendrick. 1988. GAMS: A User’s Guide.

Chenery, Hollis B. ed. 1971. Studies in Development Planning. Harvard University Press.

Choksi, Armeane M., Alexander Meeraus, and Ardy J. Stoutjesdijk. 1980. The Planning of Investment Programs in the Fertilizer Industry. The Johns Hopkins University Press.

Condon, Timothy, Henrik Dahl, and Shantayanan Devarajan. 1987. Implementing a Computable General Equilibrium Model on GAMS: The Cameroon Model. Development Research Department Discussion Paper DRD 290, World Bank. [GAMS library: cammcp.gms, camcns.gms, camcge.cms, cammge.gms]

Dervis, Kemal, Jaime de Melo, and Sherman Robinson. 1982. General Equilibrium Models for Development Policy. Cambridge University Press.

Dixon, Peter B. 1979. A skeletal version of Orani 78: Theory, data, computations, and results. Preliminary Working Paper No. OP-24 , Impact Research Center, Industrial Assistance Commission, 608 St. Kilda Road, Melbourne, Victoria, 3004, Australia.

Dixon, Peter B., Samuel Bowles, and David Kendrick. 1980. Notes and Problems in Microeconomic Theory. North-Holland.

Kendrick, David A. 1967. Programming Investment in the Process Industries, The MIT Press.

Kendrick, David A. 1984. Style In Multisectoral Modelling. Chapter 15 in Andrew J. Hughes Hallett, ed. Applied Decision Analysis and Economic Behaviour. Martinus Nijhoff Publishers.

Kendrick, David A. 1996. Sectoral Economics, in eds. Hans M. Amman, Hans M., David A Kendrick, and John Rust. Handbook of Computational Economics, Volume 1. Elsevier Science.

Kendrick, David A., Alexander Meeraus, and Jaime Alatorre. 1984. The planning of investment programs in the Steel Industry. The Johns Hopkins University Press. [GAMS library: mexss.gms].

Kendrick, David A. Alexander Meeraus, and Jung Sun Suh. 1981. Oil refinery modeling with the GAMS language. Research Report No. 14. Center for Energy Studies. University of Texas at Austin. [GAMS library: macro.gms].

Kendrick, David A., P. Ruben Mercado, and Hans M. Amman. 2006. Computational Economics, Princeton University Press, Princeton, NJ.

Kendrick, David A., and Ardy J. Stoutjesdijk. 1978. The Planning of Industrial Investment Programs A Methodology. The Johns Hopkins University Press.

Kendrick, David A., and Lance Taylor, ”Numerical methods and Nonlinear Optimizing models for Economic Planning,” in Chenery, ed. 1971. [GAMS library: chakra.gms]

Letson, David. 1992. “Investment decisions and transferable discharge permits: An empirical study of water quality management under policy uncertainty”, Environmental and Resource Economics, Vol. 2, pp. 441-458.

Linden, Gary. 1992. An integrated approach to energy investment: A project level model for Colombia’, PhD dissertation, University of Texas, Austin, TX.

Lofgren, Hans. 1993. “‘Liberalizing Egypt’s agriculture: A quadratic programming analysis”, Journal of African Economies, Vol. 2, No. 2, pp. 238-261.

Manne, Alan S., and Harry M. Markowitz. eds. 1963. Studies in Process Analysis. Cowles Foundation Monograph No. 18, John Wiley & Sons, Inc.

Markowitz, Harry M., and Alan S. Manne. 1957. “On the Solution of Discrete Programming Problems,” Econometrica, Vol. 25, No. 1.

Mennes, Loet B. M., and Ardy J. Stoutjesdijk, 1985. Multicountry Investment Analysis. The Johns Hopkins University Press.

Samuelson, Paul. A. 1952. “Spatial Price Equilibrium and Linear Programming.” American Economic Review, vol. 42, pp. 283-303.

Thompson, Gerald L., and Sten Thore. 1992. Computational Economics: Economic Modeling with Optimization Software. San Francisco, CA: The Scientific Press.

Thore, Sten. 1991. Economic Logistics: The Optimization of Spatial and Sectoral Resource, Production, and Distribution Systems. Praeger.