We use indicators for central bank independence based on the turnover rate of central bank governors.
Apart from the TOR as provided by Cukierman et al. (1992), we employ a new data set.5 Based on
information gathered from central banks and the IMF's International Financial Statistics (IFS) we have
constructed the turnover rate of the central bank governors in more than 80 developing countries for
various sample periods. To enable comparison with the data of Cukierman (1992), we calculated the
turnover rates for two periods: 1980-1989 and 1990-1998. 6 For the first period Cukierman (1992) also
provides turnover rates, albeit for a substantially smaller sample of countries. Detailed information is
shown in Table A1 in the Appendix.
3. Central bank independence and inflation in developing countries
We focus on the relationship between CBI and inflation in developing countries for two periods: 1980-
1989 and 1990-1998. The first period is chosen on the basis of the availability of the data of Cukierman
(1992). We not only estimate simple bivariate regressions, but following Campillo and Miron (1997) also
estimate models that include various control variables (see Section 2). Unfortunately, this reduces the
number of observations.
We assume that the causality runs from our CBI proxy to inflation, although, in principle, the
causality could be in the other direction (see Cukierman, 1992). For our purpose, this does not make any
difference, because the aim of the aim of the paper is to analyze whether the relationship between CBI and
inflation as reported in the literature holds in general or is driven by a few influential observations. We
start with the simple bivariate model using our new data set for the TOR. Row 1 of table 1 shows the
OLS estimation results for the period 1980-89. The coefficient of the TOR is highly significant. The same
results show up when we employ Cukierman's TOR (row 2 of table 1). Also for the period 1990-98 the
coefficient of our CBI proxy differs significantly from zero. These results are very much in line with
almost all previous studies, which found that there is a clearly significant positive relationship between
the TOR and inflation in developing countries.
[Insert Table 1 about here]
We checked whether the relationship between inflation and CBI is linear using the RESET (Regression
Error Specification Test). This test means regressing the in-sample residuals on the same regressors as in
the original linear regression and on powers of the in-sample forecasts and testing for the joint
significance of the coefficients performing an F-test (see e.g. Granger and Terasvita, 1993).7 These tests
clearly lead to the conclusion that the relationship is linear.8
Next, we checked for the influence of high inflation countries. A very simple way to visualise the
role of these countries - as suggested by Temple (1998) - is to order the countries according to their level
of inflation and then add countries one by one to the sample. Figure 1 shows the estimated coefficients