OECD economies shown in Missale (2001). In the latter countries, a strong negative co-
variance between short-term interest rates and output growth is observed over the period
1970-1998 while the covariance between short-term rates and inflation is small and not
significant. Only Greece, Portugal and Sweden display a correlation between short-term
rates and inflation as strong as in Brazil. This fact can be explained by the specific shocks
experienced by the Brazilian economy during the short period considered. However, these
correlations could also reflect structural features of the economy and/or the need for a more
flexible approach to inflation targeting in emerging economies exposed to large shocks. In
particular, the low correlation of output with the policy rate may reflect a lower elasticity
of the output gap to such rate or the case for a smoother convergence of the inflation rate
to the target. Quoting the Open Letter sent by Banco Central do Brasil’s Governor to the
Minister of Finance (January 2003): “It is a standard practice among Central Banks when
facing supply shocks of great magnitude to postpone the convergence of current inflation
towards the targets over a longer period, avoiding unnecessary costs to the economy. This
was the case faced by Brazil in the last year.”
As the positive correlation of the Selic rate and the exchange rate with inflation domi-
nates their negative but small correlation with output growth, bonds indexed to the Selic-
rate, inflation and the exchange rate all provide some insurance against variations in the
primary surplus and the debt ratio due to unexpected changes in nominal output growth.
This is shown in the first Column of Table 3, which reports for each type of debt the opti-
mal share for risk minimization in the case we abstract from hedging against variations in
the interest payments of the other instruments. All shares are positive reflecting the same
distribution of the returns of variable-rate instruments.
Column 2 shows the debt composition that allows to minimize both the risk of variations
in the primary surplus and in the interest payments. As Selic-indexed, price-indexed and
dollar denominated bonds are close substitutes in the government portfolio, variations in
their interest payments should be hedged by holding a long position in Selic-indexed bonds
(for example by means of foreign currency swaps).
When cost considerations are introduced into the analysis, the composition of the debt
clearly moves in favor of price-indexed bonds. Column 3 shows that the government should
issue price-indexed bonds in amount far exceeding the total debt and hedge this position by
holding assets denominated in dollars, along with Selic-indexed bonds. This result may look
surprising given the cost advantage, 2.3%, of dollar denominated bonds over price-indexed
bonds, but it is worth recalling that expected return differentials must be normalized by the
conditional variance or returns and the standard deviation of exchange-rate depreciation
has been 3.6 times that of inflation. Since for practical reasons a structure of assets and
liabilities as shown in Column 3 is clearly unfeasible, in Column 4 the share of price-indexed
bonds is estimated in the case the government cannot hold Selic-indexed bonds. The case
for price indexed bonds is again strong.
Evidence from the daily survey of expectations thus suggests price indexation as the
optimal strategy for debt management, thus supporting the policy indications by Bevilaqua
and Garcia (2000). As bonds indexed to the price level currently represent less than 15%
of the domestic marketable federal debt (in the hands of the public) this would imply that
funding in the next few years will have to rely on price indexation. It is however important to
realize the risk of a strategy that increases the exposure of the government budget to unex-
12