Date of Award
Union College Only
Inflation, forecasting, Neural Network, Google Trends
This paper incorporates Google Trends in addition to other macroeconomic predictors to nowcast the regional-level Consumer Price Index in the United States by using the Artificial Neural Network (ANN). In order to assess the contribution of adding Google Trends, we first consider a benchmark forecasting model to regional inflation. The benchmark model solely utilizes panel data of macroeconomic and financial variables such as employment, money supply M2, gold price, and bond yield. And then we extend the model by incorporating regional Google Trends into the macroeconomic panel. Google trends data is not available at the regional level. Therefore, I first construct regional Google Trends data by aggregating state-level Google Trends in which each state receives a weight proportional to its GDP. The results show that the ANN model with Google trends data nowcasts U.S regional level CPI more accurately based on the MAE and MSE measures on a 12-month horizon testing period ranging from Feb 2022 to Jan 2023. Finally, I present regional-level CPI nowcasts for the period Feb 2023-April 2023 with the ANN model using Google trends data and a panel data of macro-financial data.
Li, Shizhe, "A Neural Network Approach to Inflation Nowcasting Using Google Trends" (2023). Honors Theses. 2720.