High frequency data, volatility, and disaster risk in macroeconomic modeling

This dissertation consists of three essays on high frequency data, volatility, and disaster risk in macroeconomic modeling. In Chapter I, I verify a credit card transaction dataset for Austria and Germany. I use this dataset in Chapter II to investigate the weekly impact of financial market uncertainty on the real economy in a structural vector autoregressive model. Chapter III investigates the distinct features and economic effects of stochastic volatility and rare disaster risk in the Real Business Cycle model, which is a workhorse model in macroeconomics. Chapter I examines expenditures exploiting three key aspects of a novel credit card expenditure dataset: information about detailed spending categories, the location of the merchant and the credit card user, and the time of the transaction at high frequency. The aggregated credit card expenditures closely track official data when credit card holders' spending behavior resembles that of residents and the two groups overlap significantly. Additionally, geographic variations in the data replicate established patterns, such as higher spending in the proximity of the card user's residence. Finally, I show that the improvement in nowcast accuracy through high frequency transaction data hinges on the model and training period. Chapter II uses weekly transaction data to analyze the impact of financial market uncertainty on credit card expenditure. Financial uncertainty shocks lead to a significant decrease in credit card expenditure growth rates. A novel identification method challenges the standard Cholesky identification scheme. Expenditure on durables emerges as the main driver of the negative effect on consumption, indicating a precautionary motive among households. The non-Gaussian shocks identified in the model motivate the final chapter of my dissertation. In Chapter III, I depart from the standard Gaussian shock assumption and examine the effect of non-Gaussian shocks in the Real Business Cycle model. By introducing disaster risk instead of stochastic volatility, I address two puzzles related to uncertainty shocks in RBC models with CRRA preferences that have been emphasized in the literature. The results demonstrate that the RBC model with rare disaster risk significantly outperforms the RBC model with stochastic volatility in matching the empirical third-moments of crucial macroeconomic variables by at least an order of magnitude.