Table 2: Characteristics of the various rate-of-return distributions
Number of |
Rate of return | |||||
Estimated |
Median |
Mean |
Standard |
% obs. < | ||
Developed countries |
(count) 78 |
20.0 |
38.8 |
(percentage) 65.8 |
119.7 |
23.1 |
Developing countries |
123 |
40.0 |
50.0 |
58.9 |
37.9 |
31.7 |
Africa |
25 |
30.0 |
36.1 |
46.4 |
27.2 |
20.0 |
Asia and Pacific |
38 |
45.0 |
56.2 |
77.1 |
51.7 |
21.1 |
Latin America and Caribbean |
56 |
40.0 |
47.9 |
51.9 |
26.7 |
36.4 |
Developed countries, 1985 ≤ |
31 |
20.0 |
41.3 |
79.6 |
152.6 |
22.6 |
Developed countries, >1985 |
47 |
20.0 |
34.0 |
56.7 |
92.6 |
23.4 |
Developing countries, 1985 ≤ |
33 |
40.0 |
47.8 |
55.3 |
32.3 |
39.4 |
Developing countries, >1985 |
90 |
40.0 |
51.4 |
60.3 |
39.9 |
28.9 |
Note: The grouping per time period has been based on the date of publication.
To estimate the underinvestment gap it is not only relevant to know the implicit cutoff rate
and the social cutoff rate, but also the slope coefficient β1. The approach taken here is to estimate β1 by
regressing rate-of-return observations on the right-hand side (those above the cutoff rate) of the ex
post cumulated ranked distribution. Rates of return higher than 100% were excluded in order to
eliminate their distorting effect on the estimation of the slope coefficient. As shown in table 3, this
leads to a substantially better statistical fit of the exponential curve.
Varying the estimated mode or implicit cutoff rate with five percentage points for the
developed countries and 10 percentage points for the developing countries affects the slope coefficient
only marginally. Splitting the dataset into two time periods did not yield radically different slope
coefficients either, nor did it suggest a notable change in the cutoff rate over time. However, the
breakdown of developing countries by region led to a differentiation in implicit cutoff rates as well as
slope coefficients. Given the relative small number of observations, these latter results are statistically
not very robust. Nevertheless, it partially explains why we did not find a very clear mode for the
developing countries as a group.
With estimates for the implicit cutoff rates as well as the slope coefficients and the social
cutoff rate set at 7% for developed countries and 12% for developing countries, the underinvestment
gaps can now be calculated using equation 4. The results of these calculations are presented in table 4
and suggest a considerably bigger underinvestment gap for developing countries (137%) than for
developed countries (40%).
15