4. Concluding remarks
Using forecasts of the inflation rate in South Africa, we study the rationality of forecasts and the shape of forecasters’ loss functions. We find strong micro-level evidence of forecast rationality. Evidence of the asymmetry of forecasters’ loss functions is mixed, indicating the presence of at least some extent of heterogeneity with respect to the shape of forecasters’ loss functions. For a large proportion of forecasters, we cannot reject the hypothesis of forecast rationality. An estimated asymmetric loss function in some cases brings forecasts closer in line with forecast rationality, but the results depend on the model being studied. In contrast, we largely reject the hypothesis of forecast rationality when we study aggregated forecasts. We observe violations of forecast rationality for both the pooled Bloomberg data (which is a comparable survey focusing only on financial analysts) and the sectoral BER data. A comparison with our results for individual forecasts, however, shows that the deviations from rationality in the case of pooled forecasts and sectoral forecasts should be interpreted with caution, as forecaster heterogeneity at the microeconomic level of individual forecasters can lead to deviations from forecast rationality at the aggregated level of pooled and sectoral data. Hence,the main message to take home from our empirical analysis is that, given that the concept of forecast rationality is of key importance for monetary policy, it is advantageous to assess the rationality of South African inflation expectations, and the functional form of forecasters’ loss function, using micro-level data.
