Olderwork

Future Policy and Information Dissemination: A Natural Language Processing Approach

Which words matter the most in central bank communication? Making use of a rather unique European monetary policy decision setting, we build the first monetary policy dictionary. We train the dictionary on high frequency movements of the stock market around press conferences of the European Central Bank. This allows us to precisely identify which phrases do the market mainly reacts to. We find that phrases such as "Improved economy", "Market development", and "Stability of the euro" are associated with positive returns. On the other hand, phrases such as "Heightened uncertainties" and "Growth of loans" are associated with negative returns.

Food Versus Non-Food Consumption Insurance in Uganda

Households’ income fluctuations in poor countries call for risk smoothing mechanisms, yet insurance is always found to be incomplete. We build a two-goods complete markets model, and confirm this result with the UNPS - a new representative Uganda household-level panel data. The empirical evidence suggests that the degree of consumption insurance differs across consumption goods: Households insure food better than other non-durables.