LTN bonds was 17.7%, the Selic rate expected for the end of October 2004 was 14,8% and
LFT bonds were issued at a 0.4% discount. The expected return differential, TPt,canthus
be set at 2.5%.
To estimate the expected cost differential between 1-year fixed-rate bonds and dollar
denominated bonds, FPt, the 1-year yield on LTN bonds must be compared to the expected
return in Reais on US$ Global bonds. At the end of October Global bonds with a 5-year
maturity have been issued at a rate of 9.45% but the yield on bonds with a 1-year to
maturity appears much lower; the yield curve shown on the Treasury website points to a
4% 1-year yield (see Tesouro Nacional 2/12/2003). On the other hand, in the same period,
the expected depreciation from the daily survey was 9.4%. With an interest rate of 17.7%
on LTNs, the expected return differential, FPt, can thus be estimated at around 4.3%.
The premium on price-linked bonds (NTN-C and NTN-B) over 1-year fixed-rate bonds,
IPt, is the sum of an inflation-risk premium and, eventually, a “credibility spread” due
the higher inflation expected by the market than by the government. The inflation risk
premium can be estimated as the difference between the interest rate on LTN bonds and
the (real) yield at issue of 1-year price-linked NTN-C bonds augmented by the expected
IPG-M inflation. At the end of October NTN-C bonds with a 3-year maturity were issued
at 9.32% while, according to the daily survey, the expected 12-month ahead IPG-M inflation
was around 6.5%. This implies an inflation risk premium of 1.9% in October. As the real
yield on 1-year bonds might be lower than the yield on 3-year bonds, the cost advantage
of 1-year NTN-C bonds could even be greater than 1.9%. We do not add to this estimate
the difference between the inflation expected by the market and by the government, i.e. the
“credibility spread”, since there is no official target for IPG-M inflation. It is worth noting,
however, that expected IPCA inflation from the daily survey was 6.2% slightly higher than
the inflation rate implicit in the projection of the 90th COPOM meeting of November.
4.3 Uncertainty of debt returns
The conditional variance of debt returns and their covariances with output growth and
inflation can be estimated from one-year ahead forecast errors of the Selic rate, inflation,
the exchange rate, inflation and output growth. Ideally, one would like to run forecasting
regressions on yearly data for such variables. Then, the residuals of the regressions could
be taken as the estimates of the one-year ahead unanticipated components of the Selic rate,
the exchange rate, inflation and output growth. Unfortunately, this procedure is precluded
in the case of Brazil both because time series at yearly frequency are not sufficiently long
and, more importantly, because of the frequent regime shifts experienced over the last two
decades.
To circumvent this problem we consider the following three alternatives. The first ap-
proach exploits the daily survey of expectations of GDP growth, inflation, the exchange
rate and the Selic rate. The unexpected components of these variables can be obtained as
the difference between the realization of the relevant variables and their expectations one
year earlier. The conditional covariances can then be computed as the mean of their cross
products.
The second method focuses on the most recent period of inflation targeting, starting
in mid 1999, and relies on a structural backward-looking model of the Brazilian economy
estimated with monthly data. The model, which is presented in the Appendix, is consistent
10