THRESHOLDS AND CONTEXT DEPENDENCE IN GROWTH: DataWe use an annual data set covering all IMF member countries for which data are available over the period 1960-1996. The data are taken from the IMF’s World Economic Outlook and International Finance Statistics databases. The starting sample comprises 2,181 growth observations for 107 countries. We first sort this data set, according to the per capita GDP growth rates, into three equal-sized groups. Observations in the top one-third are classified as “high growth” (corresponding to average per capita growth rates of 6 percent per year), and observations in the bottom third are classified as low growth (corresponding to average per capita growth rates of -2.5 percent per year); the middle one-third observations are dropped, yielding 1,454 observations in the data, set used for the analysis.

Growth can arise from the accumulation of inputs and from disembodied productivity improvements. On the accumulation side, we follow the usual practice of measuring the growth in physical capital by the investment to GDP ratio (I/\r). Human capital accumulation has been proxied by a variety of different variables in the empirical growth literature, including school enrollment rates, average years of primary and secondary education, life expectancy and student/teacher ratios. As these tend to be highly correlated, we use the first principal component of primary and secondary school enrollment rates and life expectancy as a measure of human capital (HK). To simplify interpretation of the variable, results are stated in terms of percentiles (e.g. human capital in the top 10(/l percentile of observations).

To allow for conditional convergence effects across countries, we include the log of the ratio of US per capita income to country j’s per capita income in 1960 ( Yus jY). Fiscal effects are proxied by the ratios (to GDP) of revenues (т/Y), public consumption (G/У), and the fiscal balance (В/У) [Barro ]. As these ratios are endogeneous to current GDP growth, we use the average value of the ratios over the preceding three years. Openness effects are captured by the sample average of the ratio of exports plus imports to GDP, (У) [Edwards ]. The (log) of the black market exchange rate premium (BLK) provides a measure of the overvaluation of the real exchange rate and, in at least some instances, of macroeconomic mismanagement more generally [Barro and Lee (1993b)]. The terms of trade volatility ayj, is used as a measure of the importance of external shocks. Finally, we include indicator variables for two cataclysmic events, drought (DROUGHT) and war fatalities (DEATH).