Comments and Discussion

Aart Kraay

In a previous paper, Pattillo, Poirson, and Ricci (PPR) argue that the effect of debt on growth is nonlinear. ¹ At low levels of debt (below 18 percent of GDP or 65 percent of exports), the effect of debt on growth is positive, while at high levels of debt the effect is negative. In this paper PPR provide evidence that this nonlinearity is also reflected in two sources of growth: factor accumulation and TFP growth.

Understanding the links between debt, growth, and the sources of growth is undoubtedly important. As the authors correctly note, this line of research has significant implications for the construction of empirical indicators of debt sustainability. Research along these lines will certainly be important to inform the policies of creditor countries and multilateral lenders toward debt relief. Therefore more research along these lines should be high on the agenda. The authors should be commended for their work in this area.

I would like to speculate about the underlying factors that may be driving the observed positive correlation between debt and growth at low debt levels, and the negative correlation at high debt levels. In particular I suggest that the positive correlation between debt and growth may well be driven by unmeasured differences in returns to investment. In contrast, the negative correlation between debt and growth may be driven by unmeasured differences in institutional quality. To the extent that the latter observation is correct, it may temper PPR's policy implication that reductions in debt can contribute to faster growth rates.

Consider first the positive effect of debt on growth. As PPR note in their paper, some countries may have domestic investment opportunities with higher returns than other countries. Some countries may be poor and have lower capital stocks per worker. If diminishing returns at the country level are important, these poor countries will enjoy higher returns. Some countries may simply have better productivity levels or growth rates that raise the return to investment. To the extent that such cross-country differences in returns provide a motivation for international capital flows, one should expect to see higher borrowing in countries with high returns. To the extent that return differences are persistent over time, one should also see higher debt stocks in countries with higher returns. ² And finally, to the extent that higher returns lead to faster capital accumulation and growth, one should expect to see a positive correlation between debt stocks and growth.

This simple view of debt, capital accumulation, and growth all responding to favorable returns has sharp implications for the regressions that PPR estimate. PPR first estimate a standard, cross-country growth regression of real per capita GDP growth on a list of the usual suspects from this literature: initial income; flows of schooling and investment in physical capital; policies; and shocks. PPR augment this with debt as a fraction of GDP and allow it to enter nonlinearly as a spline function with a predetermined breakpoint at 18 percent of GDP obtained in their previous paper. They then also estimate three other regressions—with the growth rates of physical and human capital as well as TFP growth as dependent variables—using the same set of explanatory variables.

Suppose that differences in returns are entirely due to differences in initial capital stocks. Then debt will be correlated with growth only to the extent that it finances faster capital accumulation in countries with low initial capital stocks. But since the authors already control for capital accumulation (by including investment rates) and initial income levels, one should not expect debt to enter significantly in the growth regression. However, in table 2a PPR find that the direct effect of debt on growth is positive, significantly so in two of the five specifications they consider. Similarly, debt should not enter significantly in the equation for growth in physical capital, since the regression already controls for investment and debt matters only to the extent that it finances investment...