Luis Felipe Céspedes: Carvalho and Bugarin do a good job of providing new estimations on the rationality of inflation forecasts for a group of three emerging economies. The novelty of their analysis stems from the fact that information on inflation expectations based on survey data has recently become available. The topic is important: many central banks around the world use private inflation forecasts as a key element in policymaking. On the one hand, inflation forecasts may be a good proxy of inflationary pressure. On the other, for economies with an explicit inflation target, they could serve as an indicator of the credibility of the inflation target itself. For example, the Central Bank of Chile aggressively reduced its monetary policy interest rate in January 2004 as low inflation rates were feeding into medium- to long-term inflation expectations, pulling them below the 3 percent annual target.
Given the important role of private inflation forecasts as a proxy for inflation expectations, it is vital to determine whether they satisfy some basic rationality conditions. In the first part of the paper, the authors test whether these forecasts are unbiased and efficient.
A first comment on the paper is related to the nature of the information: survey data. These surveys do not necessarily measure informed opinion. The survey used for Brazil is a survey of professional forecasters, while in Chile a mix of academics and professional forecasters are surveyed. In the case of Mexico, the data sources are organizations, so it is less clear who the forecasters are and whether they change frequently. This could shed some light on the authors' empirical results regarding the efficiency of inflation forecasts in Mexico, especially those related to the use of interest rates.
Another important point involves the significance of the unit root analysis component of the paper. The main problem with these unit root tests and with their empirical analysis in general is their robustness. Small sample bias is a serious problem in empirical implementation. The authors only have a few years of observations for each country. Is that enough to be confident about [End Page 139] the results they obtain? The authors acknowledge this problem and proceed by estimating the series without considering its integration order. This strategy may be reasonable for the case of Chile, but it may not be for Mexico. The inflation target was decreasing in Mexico throughout the sample period. Inflation in this case may be stationary around the inflation target, but not stationary in and of itself.
A key issue for the efficiency of forecasts is the information available to the forecasters when the forecast is made. For Chile and Brazil, the authors assume that the output gap is known at the moment the forecast is made. This is not actually the case, however, which reduces the credibility of results based on this information. The forecasters may have other ways of determining current demand conditions before making their forecasts, but if that is the case, the authors should use this information in their analysis.
The second part of the paper investigates the inflation expectations formation process. A key element in the analysis is the role of the inflation target in this process. If the target is fully credible, inflation expectations for longer horizons should be anchored to the target. If the target is not credible, inflation expectations should not matter in the inflation expectation formation rule.
There is a serious problem with the analysis when the inflation target is changing over the sample period. If the economy is converging from high levels of inflation to a lower stable rate, the inflation target itself is very likely to be endogenous. For example, if output is below potential, the authorities are likely to set a less strict inflation target in order to have some room to stimulate the economy. This clearly limits the usefulness of the analysis presented for Mexico and Brazil.
In the case of Chile, the credibility analysis of the twelve-month inflation forecast is...