THRESHOLDS AND CONTEXT DEPENDENCE IN GROWTH: IntroductionWhat separates winners from losers in the great growth game? A decade of intensive research has identified a range of determinants influencing both rates of factor accumulation and disembodied technological progress. Yet to day no universal growth “recipe” has emerged from this literature: cross-sectional regressions continue to suffer from severe robustness problems [Levine and Renelt ]. At the same time, the set of potential determinants found to be significant in some context or other continues to grow. Indeed, the empirical growth literature now arguably sufFers from an embarrassment of riches, with proposed determinants of growth spanning a wide field including learning by doing [Arrow ], education [Barro and Lee (1993a)], openness [Edwards ], policy distortions [Easterly ], inflation [Bruno and Easterly ], fiscal policy [Barro ], financial sector development [Greenwood and Jovanovic ], income distribution [Perotti ], R&D policies [Grossman and Helpman ], natural resource endowments [Sachs and Warner ], culture [Carroll, Rhee and Rhee ], ethnic divisions [Easterly and Levine ], democracy [Helliwell ], equipment investment [DeLong and Summers ], macro policies [Fischer ], institutions [Knack and Keefer ] and stock markets (Levine ], among others.

Faced with this plethora of causal factors, a number of recent studies have tried to pare the list to a core group of “essential” factors which separate the broad group of “winners” from “losers”.2 By and large, this work has maintained the methodological framework of cross-section regression analysis used in most of the empirical studies cited above. The research strategy has yielded useful insights, yet it has raised the bar quite high: with the exception of the investment ratio and, perhaps, trade openness, few variables appear to be robustly associated with growth.

Upon reflection, it is scarcely surprising that robust causal factors have proven so hard to find for the sixty to more than one hundred countries in the typical cross-country growth regressions. A substantial theoretical literature on nonlinear effects suggests precisely this absence of a strong link valid across a wide variety of countries. Specifically, there are good reasons to suspect that the link between determinants and growth may be quite nonlinear, subject to both threshold effects and complementarities. Such non-linearities of course have a long tradition in the theoretical literature dating back to Young, Rosenstein-Rodan’s “big push”, Gerschenkron’s “backwardness” hypothesis and Schumpeter’s development theory.