In this chapter, I conduct a series of empirical tests of my argument that changes in patterns of financial globalization are the key determinant of IMF lending policies. Using time-series cross-sectional statistical methods, I analyze a new data set of 197 non-concessional IMF loans to forty-seven countries from 1984 to 2003. These countries are middle-income developing nations that typically borrow on private international markets but periodically and temporarily seek loans from the IMF when facing balance of payments problems. As noted earlier, the IMF also lends on a longer-term, concessional basis to extremely poor countries that rarely borrow on private markets. These loans differ notably from non-concessional Fund programs, however: they are intended to fund long-term development and structural adjustment rather than short-term payments imbalances; they are not linked to specific repayment timetables; they are fully financed from a trust fund entirely separate from the IMF's quota resources; and they are available only to a specified set of extremely poor countries. While many studies of IMF lending pool these two types of loans and countries, doing so is likely to result in biased predictions about the Fund's non-concessional lending behavior.

The original data set analyzed in this chapter has been assembled from official Fund documents available at the IMF archives in Washington, DC. For each loan, data on loan size and conditionality are taken from several categories of documents: the “letter of intent” declaring the borrower country's intent to enter into a Fund program, the attached “memorandum of economic policies” specifying the policy reforms and conditionality a country will implement during the course of the IMF loan, and the “staff report” to the executive board that outlines the draft program and provides further details on conditionality.