and the band of plus and minus two times the standard error for the regressions reported in rows 1-3 of
table 1. The figures clearly show that only after the high inflation countries are included in the sample, the
coefficient of the TOR index becomes significant. In other words, the conclusion that CBI matters for
inflation is driven by a limited number of observations.
[Insert Figure 1 about here]
Next, we turn to the multivariate models. The lower part of table 1 presents the regressions. As we are
mainly interested in the effect of CBI on inflation, we only report the estimated coefficients for the TOR.9
It follows that in the regression for the period 1980-89 the coefficient of our proxy for CBI is not
significantly different from zero (row 4 of table 1). When we use Cukierman's TOR instead, its coefficient
is significantly different from zero, although only at the 10 per cent level (row 5). For the 1990s we do not
find a significant effect of our TOR on inflation for the multivariate model (row 6). The variables that turn
out to be significant in all regressions are openness, the exchange rate regime (except in the regression
with Cukierman’s TOR) and the debt ratio (see table A2 in the Appendix).
Figure 2 shows the estimated coefficients and the band of plus and minus two times the standard
error for the regressions reported in rows 4-6 of table 1. Again, it clearly follows that Cukierman's TOR
only becomes significant once high inflation countries are added to the sample. Although the coefficients
of our proxy for CBI never become significant, the same important role of high inflation observations can
be discerned.
[Insert Figure 2 about here]
Finally, we have ordered the countries according to their turnover rate and redid the recursive regressions.
The results are reported in figures 3 and 4, respectively. The results confirm our conclusion. The
coefficient of the TOR variable generally only becomes significant at high turnover rates.
[Insert Figures 3-4 about here]
4. Conclusions
In this paper we have re-examined the relationship between central bank independence and inflation in
developing countries. We extend the existing literature in three ways. Firstly, we present a new data
set for the turnover rate of central bank governors in a very large sample of countries, which also
covers the 1990s. Secondly, we employ turnover rates in a multivariate model. Previous studies which
used control variables (notably Campillo and Miron, 1997 and Temple, 1998) employed legal