effort of one of the main shareholders. We capture this eventuality by normalizing IFI
disbursements by the borrowing region’s GDP.
We first assess the impact of financial crises on disbursements. We estimate two separate
regressions for the World Bank and for the IMF by seemingly unrelated methods. We
allow, therefore, for the presence of common shocks to the lending operations of the two
institutions. This set-up offers one additional advantage. By estimating two separate sets
of regional fixed effects, respectively for the Bank and for the Fund, we are able to infer
the relative role of each institution in a given region. In Table 3 we report the results of a
regression relating annual regional disbursements by the World Bank and the Fund to a
set of regional fixed effects and to our crisis indicator. Two facts stand out. First, the
World Bank typically plays a quantitatively large role than the IMF, as shown by a
comparison of the regional effects. This is true in Africa, Asia, Latin America, and the
Middle East. Only for Europe is the amount of IMF lending in normal, i.e. non crisis,
conditions, at par with that of the World Bank. Second, both institutions respond
significantly to the emergence of a financial crisis. However, and not unexpectedly, the
response of the IMF is substantially larger than that of the World Bank. Following a
balance of payments crisis, the IMF is shown to increase its lending by almost half a
percentage point of the borrowing region’s GDP. For the Bank the increase is less than
one decimal point (always in relation to the region’s GDP).
Summing up, this set of findings support the notion that the International Monetary Fund
plays a key role in cushioning the impact of financial crises on developing countries, with
the World Bank having at best a supportive one. However, in normal times, World
Bank’s lending is substantially larger than that of the IMF. These results conform to well
established notions of an IMF lending massively only in time of crises, and of a World
Bank Group intermediating, instead, capital to member countries in a normal way.
We then assess how the pattern of response to crises changed between the past two
decades. Econometric results are reported in the second column of Table 3. The first
striking finding is in the large increase over time in the size of the coefficient of IMF
response to crisis. In the 1980s its value was 0,271. In the 1990s, it is seen to increase by
0,268, thus virtually doubling in size. IMF disbursements reached almost 0.5% of the
GDP of recipients. Clearly, this reflects the sizeable increase in IMF intervention during
the 1990s when balance of payments crises increasingly involved the capital account of
several large countries, rather than simply their current account, and required official
intervention on a much larger scale than in the past. Interestingly enough, the size of
World Bank response declined in the 1990s compared to the eighties and became not
significantly different from zero. Most likely, this finding reflects the fact that while the
World Bank Group was quick to react to the debt crisis in the 1980s through a massive
increase in the amount of structural adjustment lending, in the next decade shrunk due to
overall lending constraints and exposure considerations.
We also test whether the pattern of regional lending by the two institutions changed
significantly over the two decades in question. To this purpose, we check whether the
pattern of regional fixed effects shows a structural break between them. The results,
14