Some investigators submit that SPI disrupts production and increases uncertainty in the economy. By doing so, it undermines the incentives for accumulating physical capital and reduces the rate of economic growth. By contrast, others argue that economic growth leads to either higher SPI because growth entails substantial structural changes that undo political coalitions and induces painful readjustments in the balance of power among different interest groups, or to lower SPI because it reduces social and political tensions. Even though the existence of a negative relationship between SPI and economic growth has been elevated to "stylized fact" status, the empirical studies on which this purported relationship is based have been heavily criticized for their ad hoc selection of explanatory or control variables, their excessively narrow definitions of SPI, their insufficient sensitivity analysis, and their failure to investigate the direction of causality. Even though we do not agree with all these criticisms, we do believe that this negative relationship should not be elevated to stylized fact status without demonstrating that causality exists, and that it runs from SPI to growth rather than in the opposite direction.

The literature seems to contain two rather different understandings of SPI, one stressing regular and irregular government transfers, the other focusing on much harsher aspects of SPI. The fact that these overlap does little to diminish the different intensities that each attaches to "instability." While the former interpretation constrains it to relatively tame phenomena, the latter places it closer to social chaos. To recognize both views, we construct two measures of SPI: one captures the more severe forms of SPI, while the other captures the less severe forms of SPI. While we could have used many other variants, our justification is that the ones we use can be considered the bounds of the realistic range of such measures, permitting a more complete depiction of the causality between SPI and growth.

Our measure of severe or upper-bound SPI follows the existing literature in using the following three indicators: the numbers of political assassinations per million people, revolutions, and coups d’etat. The first of these is particularly important, because it captures a dimension of magnitude that is largely missing from the other measures.

For the measure of moderate or lower-bound SPI we use indicators that include the regulation of political participation and the openness of executive recruitment. Because political actors and processes are subject to systematic regulation, this set of indicators can capture the extent of even subtle changes in both legal and actual practice. The less regulated actors and processes are, the greater the potential for social and political change, and the higher the value of this SPI index.

As both indexes measure SPI, but capture quite different aspects of it, one would expect them to be positively, but not highly, correlated. In general, this expectation is fulfilled except for the Middle East and North Africa. For all other regions the correlation between the respective pairs of SPI indexes is positive and statistically significant.

Next we turn to the rate of real GDP growth and to the time periods chosen. We collected measures of GDP growth for nonoverlapping five-year periods from 1960 through 1995 for an unbalanced panel of 98 developing countries: 14 countries in Asia, 21 in Latin America, 17 in the Middle East and North Africa, and 46 in Sub-Saharan Africa. We found that a significant negative relationship exists both for the pure cross-section relating to growth over the whole period as well as for the individual five-year intervals.

We find that the evidence supporting the hypothesis that high levels of SPI cause lower rates of economic growth is much weaker than generally believed, and we find no traces of a long-run causal relationship. How can we reconcile these conclusions with the differing results of other similar studies? Our sensitivity analysis shows that the Sub-Saharan Africa sample constitutes a large part of the explanation. Not only is the Sub-Saharan sample much larger than those for other regions, but also its SPI seems to take a more structural form. Our findings support the explanation that, once one controls for institutional development or for the terms of trade, the causality results vanish. Hence we suspect that if other studies were to exclude African countries from their samples, the existing results of a negative relationship between SPI and growth would disappear.

Given the prominence recent macroeconomic research has attached to SPI, we offer a number of suggestions for further research. First, in light of the inconsistency between the existing understanding that a negative relationship exists between SPI and economic growth and our findings of the lack of a causal negative relationship between SPI and growth, one should ask at what frequencies and lag lengths the relationship changes from noncausal to causal. A second direction for future research would be to investigate whether a causal negative relationship emerges between economic growth and other important sources of instability, for instance, policy variability. Third, investigators should readily be able to identify additional omitted variables, especially those of an institutional nature, that might be related to both SPI and growth. Numerous institutional variables may be relevant, like the fairness and effectiveness of the judicial system and the stability of property rights. Fourth, given the difficulties in constructing a lower-bound measure of SPI, exploratory research of this sort with other SPI measures should be encouraged. Finally, in light of the wide variety of other consequences that investigators have ascribed to SPI, an examination of causal relationships between SPI and these other variables should be seriously considered. In particular, seeing whether the Sub-Saharan Africa sample would again play such a determinant role would be interesting.