ECONOMIC DEVELOPMENT: EVIDENCE FROM DIRECTED ACYCLIC GRAPHS* by DAVID A. BESSLER Texas A&M University and NATHAN LOPER{ Purdue University We use directed acyclic graphs to study post-1970 cross-section data from 79 world economies and a subset of 59 less developed economies. Openness to trade, government savings and natural resource exports are direct causes of GDP growth rate. Openness to trade and government savings contribute positively and natural resource exports contributes negatively to growth rate in GDP. An institutional quality index, agricultural productivity, life expectancy, initial GDP and a tropical climate dummy are related to but are not direct causes of GDP growth rate in one or more versions of the models uncovered. Implications on modeling are explored. " Introduction We apply directed acyclic graphs for the construction and interpretation of models of GDP growth based on observational data. The motivation for the paper is recent advances in arti¢cial intelligence. By models of GDP growth we refer to regression analysis on cross-section data in which GDP growth (or a transformation of such) is written as a function of one or more `independent' variables. While such models are not necessarily meant to summarize causal relations (as opposed to associational relations; see Holland (1986) for the distinction), they are used often for just that purpose. The ¢eld of development economics is richly populated with such regressions, as understanding and modeling the causal mech- anism behind GDP growth are fundamental to e¡ecting change. 1 As the data used in such studies are observational and are measured at the same point in time (or change slowly over time), it is oftentimes not clear whether a variable belongs on the left-hand side or the right-hand side of the equals sign  in these equations. There is no experimental manipulation (with random assignment) of one variable (say a measure of agricultural productivity or institutional quality) to observe the sub- sequent response of another variable (say GDP growth rate). It is possible, ß Blackwell Publishers Ltd and The Victoria University of Manchester, 2001. Published by Blackwell Publishers Ltd, 108 Cowley Road, Oxford OX4 1JF, UK, and 350 Main Street, Malden, MA 02148, USA. 457 The Manchester School Vol 69 No.4 September 2001 1463^6786 457^476 * Manuscript received 12.10.99; ¢nal version received 7.9.00. { Two anonymous referees made helpful comments on an earlier draft of this paper. 1 See Barro (1991) or Mellor (1995) for recent examples of such work.