since the shorter the period of time since the adoption occurred the stronger the
signalling e∏ect is compared to others possible reasons for the rescheduling. In this
model timing is crucial. In the ...rst place, the indebted country could either receive
or does not receive the IMF loan (and accept the IMF conditionality that goes
with it). Then, creditors decides whether or not to grant the debt rescheduling
to the country.
The bivariate probit speci...cation is the following:
I*=Xb+u
C*=Zg+ Id+v
I = 1 iff I * > 0; 0 otherwise (4.1)
C = 1 iff C* > 0; 0 otherwise (4.2)
The disturbances are assumed to be bivariate normally distributed.
μ fl
u
v
1½
fl
0;½ 1
Equation (4.1) ofthe bivariate speci.cation describes the IMF adoption. The
latent variable for the IMF adoption, I *, is a linear function of the countries’
macroeconomic characteristics (vector X) which a∏ect the probability to adopt
an IMF programme (they will be speci.ed more carefully in Section 4.1.3). Since,
after the adoption of the IMF programme, these macro-variables would be a∏ected
by the implementation of the programme itself (and thus they would become
endogenous), we take their values two year before the programme is adopted, in
order to make sure they are predetermined.
I * occurs both in the observed di chotomous form in equation (4.2) and in
the latent-variable form in equation (4.1). The sign of the coe^ient of the di-
chotomous variable I , in equati on (4.2), will measure the role of the IMF in
debt concessions schemes and our prior expectation is that it will be signi.cantly
greater than zero.
Equation (4.2) describes the “concession” of a debt rescheduling. The latent
variable for the debt rescheduling, C*, is a linear function of the countries’ macro
18