Paul Johnson, Steven Durlauf and Jonathan Temple 1155is that the errors may then be serially correlated, and the standard errors unreli-
able. Using simulations, Bertrand, Duflo and Mullainathan (2004) showed that
differences-in-differences estimators are potentially highly vulnerable to this prob-
lem, to the extent that “placebo” interventions are often found to have effects
that are statistically significant. This could well be a major concern for the cor-
responding growth studies carried out to date. In samples with the cross-section
dimensions associated with cross-country data, some of the available solutions for
calculating panel-robust standard errors may also face problems.
24.6 Endogeneity and instrumental variables
In this section, we consider the use of instrumental variables in cross-section and
time-series contexts. One obvious criticism of growth regressions is that they do
little to establish directions of causation. As well as reverse causality, there is the
standard problem that two variables may be correlated but jointly determined by
a third. Variables such as growth and political stability could be seen as jointly
determined equilibrium outcomes associated with, say, a particular set of institu-
tions. The usual response has been the use of instrumental variables, for reasons
discussed in section 24.6.1, but there are some grounds for caution in their appli-
cation. These include the difficulty of establishing credible exclusion restrictions
(section 24.6.2) and the problems raised by heterogeneous effects in small samples
(section 24.6.3).
24.6.1 Concepts of endogeneity
There are many instances in growth research when explanatory variables are clearly
endogenously determined in an economic sense. The most familiar example would
be a regression that relates growth to the share of investment in GDP. This may
tell us that the investment share and growth are associated, but stops short of
identifying a causal effect, or explaining why investment varies; presumably it is
endogenous to a range of economic variables. When variables are endogenously
determined in the economic sense, there is also a strong chance that they will
be endogenous in the statistical or technical sense, namely correlated with the
disturbances in the structural equation for growth. To give an example, consider
what happens if political instability lowers growth, but slower economic growth
feeds back into political instability. The OLS estimator will conflate these two effects
and yield an inconsistent estimate of the causal effect of instability.^23
Views on the importance of these considerations differ greatly. One position is
that the whole growth research project effectively capsizes before it has even begun.
Mankiw (1995) and Wacziarg (2002) have suggested an alternative and more posi-
tive view. According to them, one should accept that reliable causal statements are
almost impossible to make, but use the partial correlations of the growth literature
to rule out some possible hypotheses about the world. Wacziarg uses the example
of the negative partial correlation between corruption and growth found by Mauro
(1995). Even if shown to be robust, this correlation does not establish that some-
how reducing corruption will be followed by higher growth rates. But it does make