A Threshold Analysis of the Relationship Between Governance and Growth

Abstract

Quality of governance has been found to be significant in economic growth. We investigate, using a threshold technique, whether the quality of governance matters equally across all levels of economic development. We find that the quality of governance is most significant for only a subset of relatively poor countries, while education is most significant for the poorest countries, and region is most significant for the wealthiest.

Acknowledgements

The authors wish to thank Abdelhak Senhadji of the IMF for graciously providing the TFP data and two anonymous referees for helpful comments and suggestions. All remaining errors are solely the authors' responsibility.

Notes

¹This conclusion is similar in spirit to the conclusion reached by Ghosh & Wolf (1998). Using a different technique they find that there is ‘no single recipe for growth.’

²See Krueger (1974), Rose-Ackerman (1978) and Bhagwati (1982) among others.

³As documented by Kaufmann et al. (1999) there are numerous measures of quality of governance and institutions covering various countries and time periods. In general, these measures tend to be highly correlated. Our motivation for using the ICRG data is threefold. First, our methodology extends the methodology used by Olson et al. (2000). And because we wish to compare our results with theirs, we use the ICRG data as our measure of quality of governance. Our second motivation is that the ICRG data are widely used and therefore serve as a benchmark for quality of governance measures in the growth literature (see Barro, 1996, 1997; Clague et al., 1995, 1996; Hall & Jones, 1999; Knack & Keefer, 1995, 1997a, |Knack & Keefer, b; Knack, 2002; and Olson et al., 2000). Finally, as pointed out by Kaufmann et al. (1999) the ICRG measure of quality of governance is a high quality statistic which is subject to peer review to ‘ensure the coherence and comparability across countries.’

⁴We also estimated the production function without imposing the constant returns to scale constraint. The unconstrained coefficient on capital growth is 0.586 and the unconstrained coefficient on labor growth is 0.31. Both coefficients are highly significant and we fail to reject the null hypothesis that they sum to 1. The adjusted R-squared was 0.18%. In addition, the results from our second stage regression were nearly identical when we used the fixed effects estimated from the unconstrained production function.

⁵Data for Cyprus is 1960–1996, Sudan 1996 only, Sierra Leone 1961–1996 and Tunisia 1961–1997.

⁶Government consumption as a percentage of GDP from the Penn World Tables is measured using international rather than local prices. As pointed out by Knowles (2001), using international rather than local prices may artificially inflate government consumption as a percentage of GDP for low-income countries.

⁷GDP data is first available in 1960. For Haiti the GDP data begins in 1967, the Sudan in 1996, Sierra Leone in 1961 and Tunisia in 1961.

⁸Data for China and Egypt are from 1975.

⁹An alternative technique for assessing whether quality of governance matters for all stages of development would be to include an interaction term (quality of governance multiplied by per capita income). The advantage of the threshold technique, however, is that it allows us to determine the optimal break point in the relationship between governance and TFP growth.

¹⁰The threshold technique has been used by previous authors to study aspects of economic growth other than the quality of governance. Earlier studies which used the threshold technique are: Kahn & Senhadji (2000), Ghosh & Wolf (1998), Liu & Stengos (1999), Durlauf & Johnson (1995) and Hansen (2000).

¹¹We chose the model with the breakpoint at observation 17 (Egypt) because the previous bootstrap test indicated that this was a significant breakpoint for all of the measures of quality of governance.

¹²We did estimate our model over the middle sub-sample with 12 observations and the adjusted R-square is negative for all measures of quality of governance. Nonetheless, the coefficients on bureaucratic quality and expropriation risk are significant at the 5% level while the overall average measures as well as corruption and repudiation risk are significant at the 10% level.

¹³An alternative interpretation of our results is that the poor countries in our sample are ‘outliers.’ By including these outliers in our full sample (Table 3) we are mistakenly led to the conclusion that quality of governance matters for growth in TFP. By removing these outliers from our sample (Tables 7 and 9) we find that quality of governance does not matter for TFP growth. Thus, the fact that the sample size for the regressions using the countries for which quality of governance matters (Tables 8) is small does not matter. We are mainly interested in the larger sample results showing that quality of governance does not matter for the vast majority of countries in our sample.

¹⁴See Beck et al. (2001) who compile a new database to determine which institutions are most conducive to development and reform and under what conditions these institutions emerge.