Is there too much certainty when measuring uncertainty

This paper criticises the econometric inflation uncertainty proxies found in the literature, which show an overly optimistic picture about our real ability to forecast, and highlights the sharp contrast between the evidence portrayed by that literature and the evidence conveyed by the literature on surveys of inflation expectations. While the latter shows that actual forecasts are usually biased and systematic forecast errors are pervasive the former shows a much more optimistic picture, in accordance with the rational expectations paradigm. Also, both literatures have historically shown conflicting evidence on the inflation level – inflation uncertainty link. Next, the performance of inflation forecasts from both the Central Bank of Brazil Inflation Report and the Focus Survey are analysed. The paper then pinpoints some simple measures that could be taken to improve the reliability of econometric inflation uncertainty proxies, and carries out a (pseudo) real-time forecasting simulation exercise to derive a set of such proxies for Brazil. The features of those forecasts are shown to be very similar to those found in surveys.