Description

This paper examines the accuracy and other properties of the errors in the short-term GDP and inflation forecasts of the Central Bank of Chile over the past decade, comparing them with the main market surveys. Standard metrics and econometric tests are applied to assess forecast performance according to optimal properties, namely unbiasedness and efficiency. The results show that, in general, the forecasts are “well behaved” and, on average, exhibit smaller errors than those from market surveys. During the COVID-19 pandemic, forecast errors increase, and biases and inefficiencies are found across different variables, consistent with findings for other central banks around the world. Nevertheless, forecast errors remain smaller than those from market surveys during this period of crisis. In recent years, forecast errors have returned to pre-pandemic levels. The study supports the Central Bank staff’s continued efforts to develop predictive models suited to periods of high volatility and crisis.