Search

We match survey data of Italian firms that includes a repeated experiment in which information about inflation is randomly provided to firms over time with detailed credit data that covers the borrowing decisions of firms. This allows us to study how exogenous variation in inflation expectations causally affects the borrowing decisions of Italian firms. We document a number of new results. Firms with exogenously higher inflation expectations end up paying higher interest rates on average but do not change the overall demand of loans. Instead, we find a significant rebalancing of firms' borrowing decisions away from lower-interest long-term loans and toward higher-interest short-term loans. In anticipation of rising future interest rates linked to higher expected inflation, firms also take on new long-term loans to pay down existing loans, thereby locking in interest rate savings. Firms that are relatively more knowledgeable about financial tools engage in the latter particularly strongly.

As the pandemic spread across the U.S., disagreement among U.S. households about inflation expectations surged along with the mean perceived and expected level of inflation. Simultaneously, the inflation experienced by households became more dispersed. Using matched micro data on spending of households and their macroeconomic expectations, we study the link between the inflation experienced by households in their daily shopping and their perceived and expected levels of inflation both before and during the pandemic. In normal times, realized inflation barely differs across observable dimensions but low income, low education, and Black households experienced a larger increase in realized inflation than other households did. Dispersion in realized and perceived inflation explains a large share of the rise in dispersion in inflation expectations.

An empirically founded and widely established driving force in opinion dynamics is homophily i.e. the tendency of "birds of a feather" to "flock together". The closer our opinions are the more likely it is that we will interact and converge. Models using these assumptions are called bounded confidence models (BCM) as they assume a tolerance threshold after which interaction is unlikely. They are known to produce one or more clusters, depending on the size of the bound, with more than one cluster being possible only in the deterministic case. Introducing noise, as is likely to happen in a stochastic world, causes BCM to produce consensus which leaves us with the open problem of explaining the emergence and sustainance of opinion clusters and polarisation. We investigate the role of heterogeneous priors in opinion formation, introduce the concept of opinion copulas, argue that it is well supported by findings in Social Psychology and use it to show that the stochastic BCM does indeed produce opinion clustering without the need for extra assumptions.