Inefficiency in survey exchange rates forecasts

We use survey data on five bilateral exchange rates to provide empirical evidence of the fact
that professional forecasters of foreign exchange rates behave irrationally, in the specific sense that
they respond inaccurately to available information in the market when forming their predictions.
In particular, we find systematic biases in the forecasts resulting in the overreaction of analysts
to past information contained in the exchange rate dynamics: forecasters change their prediction
more than would be rational on the basis of past realized changes. In addition, forecasters are
heterogeneous in their irrationality: low performers in previous periods show a more pronounced
overreaction effect. This can be read as an indication of perpetration of past errors and continued
inability to learn from the past. In the second part of the paper, we exploit the novel structure
of our dataset, which consists of survey data extracted from the Bloomberg platform and readily
available to anyone. This feature allows us to consider own and others’ past forecasts as part of the
information set that analysts use in making their predictions. By using past forecasts as proxies for
relevant macroeconomic variables, we find evidence that analysts fail to correctly process not only
the information contained in the spot rate past dynamics, but also the information in this broader
set. We see this as confirmation of the existence of inefficiency and heterogeneity between low and
high performers also when full information is available.