with that proposed by Favero and Giavazzi (2003) under the hypothesis of “Ricardian fiscal
policy”. As we use monthly data, the one-year ahead unanticipated components of the
Selic rate, the exchange rate, inflation and output growth are estimated as the 12-month
cumulated impulse responses of these variables to shocks of inflation, the output gap and
the EMBI spread.
The third approach approximates the one-year ahead unanticipated components of the
relevant variables using the residuals of forecasting equations estimated on quarterly data
for the period Q3 1995 to Q1 2003.
This method requires the extension of the sample period to include the fixed exchange
rate period and the currency crisis of 1999. On the other hand, the estimated stochastic
structure is independent of the modeling strategy.
5.1 Estimating the debt composition from the daily survey of expectations
Table 1 decomposes GDP growth, IPCA inflation, exchange-rate depreciation (relative
to the US dollar), and the Selic rate between its expected and unexpected component for
the years 2000, 2001 and 2002 for which expectations can be obtained from the daily survey.
Except for the first year when output growth was higher than expected, the Brazilian
economy performed much worse than expected. Output growth was substantially lower in
2001, while inflation and exchange-rate depreciation exceeded expectations in both 2001 and
2002. The Selic rate also turned out much higher than expected. Had the government issued
fixed-rate conventional bonds instead of Selic-rate indexed bonds and dollar denominated
debt, debt sustainability would not be a problem for Brazil. Hence, prima facie evidence
appears to make a strong case for fixed-rate long-term debt. This depends however on
the specific —short— period considered. If times of unexpected deflation, falling short-term
interest rates and unexpected appreciation, as those experienced in 2003, are as likely as the
events of the period 2000-2002, then issuing fixed-rate bonds paying a high term premium
would be a poor strategy.
To correctly address the issue of the optimal debt composition we must look at the
covariances of debt returns with output and inflation. Table 1 clearly points to a negative
correlation between all types of indexation and unexpected output growth but also shows
that unexpected inflation has been positively associated with higher returns on dollar-
indexed bonds and Selic-rate bonds. Unexpected inflation has also led to higher returns on
price-indexed bonds. This suggests a role for price-indexation (and, to a lesser extent, for the
other types of indexation) in hedging against unexpected deflation. This requires, however,
that the observed comovements between inflation and debt returns were a systematic feature
of the Brazilian economy and not just an episode confined to the period under consideration.
The qualification makes it clear that policy indications are not robust when the available
evidence is limited to a short period of time as in the present case.
The conditional covariances of debt returns with output and inflation (relative to the
conditional variance of returns) are presented in Table 2. The covariances of output growth
are negative but small with all types of indexed debt, while inflation displays a strong
positive correlation with the Selic-rate and a mild correlation with the exchange rate. Hence,
all types of indexation are useful hedges against inflation, although they introduce additional
risk when negative output shocks already impair debt sustainability.
Importantly, the magnitude of these effects is in sharp contrast with evidence from
11