These advantages of the single-country modelling approach are, however, gained at a cost which goes beyond the money and time required to build the model. In the first place, the 'without programme aid' situation is no longer an observable construct, such as a group of nonprogramme countries, as was the case in Chapter 6, but rather the imaginary situation in which the country under analysis received no adjustment finance from the Bank and made no associated policy changes. But this situation is not easy to define with absolute accuracy, since some adjustment to economic crisis is invariably made by recipient countries whether or not the Bank intervenes, and the question of what kind of policy change would have been carried out in the absence of a deal with the Bank cannot be resolved with certainty. In addition, the choice between models is a subjective matter. Even a model which has predicted well in the past may be rejected by the user on the grounds that its structure does
not square with his intuitive belief about how the economy works. 1 The Philippines desk of the World Bank, for example, commissioned a
couple of modelling exercises in 1986 as a preparation for the programme of new lending to the Aquino government (discussed in Volume 2, Chapter 12). One of these was an econometric model designed by a member of the Bank's Economic Analysis and Projections Department (Hwa, 1986). It was rejected, not on account of its ability to forecast the Philippines economy (which was quite good) but rather because: (a) it was unstable (i.e. any external disturbance drove it further and
further away from macroeconomic equilibrium)2 and thus mathematically inconvenient;
(b) it exhibited certain counter-intuitive features. One was a negative response of the demand for money to the interest rate, which accounted for the instability above described; the other was a price elasticity of demand for demand for exports well below one. Hence, as an internal Bank commentary on the model put it, 'an export-led growth strategy is incompatible with balance of payments equilibrium. Clearly then, this elasticity is another key parameter on which more work must be done before simulation results can be accepted' (Kharas, 1986: 12).