Inflation, a central concern in macroeconomics, is challenging to explain and forecast accurately. While traditional models utilize economic and financial factors like interest rates and oil prices, their predictive power is often limited. This paper introduces a novel factor—investor attention—to address this gap. Investor attention, originating from behavioral finance, reflects the information acquisition and asset pricing behaviors of investors. Theoretically, it's linked to inflation because it influences asset pricing, which directly reflects price level changes. Furthermore, investor attention is connected to inflation expectations—a crucial determinant of inflation according to the New Keynesian Phillips Curve (NKPC). This study, therefore, aims to empirically explore the relationship between investor attention and inflation in the United States, testing three hypotheses: 1) investor attention can explain inflation; 2) investor attention can forecast inflation; and 3) investor attention influences inflation through inflation expectations. The originality lies in connecting investor attention, a micro-level factor, to macro-level inflation dynamics, potentially enriching our understanding of both behavioral finance and macroeconomics.

Existing literature extensively addresses inflation explanation and forecasting. Studies have attributed inflation variations to monetary policy, the Phillips curve, and other factors like industrial production, unemployment, and nominal wages. Various models, including VAR, dynamic factor analysis, neural networks, and Bayesian methods, have been employed for inflation forecasting. Recent research has explored the role of novel factors like commodity prices, online price indices, and even climate variables in improving forecast accuracy. Behavioral finance, particularly investor sentiment, has also been considered, but with mixed results. This study uniquely focuses on investor attention, an under-researched aspect of behavioral finance, to assess its impact on inflation.

This study uses the Google Search Volume Index (GSVI) for the keyword "inflation" from Google Trends as a measure of investor attention. Monthly data from January 2004 to July 2020 is utilized. The seasonally adjusted monthly CPI inflation rate for all urban consumers from the Federal Reserve Economic Data serves as the inflation measure. Both series are logarithmically differenced. Before analysis, the stationarity of both time series is tested using ADF, KPSS, and PP tests. A VAR(1) model is estimated to explore the dynamic relationship between investor attention and inflation, with a Granger causality test applied to determine the causal relationship. Impulse response analysis is performed to examine the effect of shocks to investor attention on inflation. Linear regression models, controlling for lagged oil market returns (Brent oil futures returns), are then employed to explain and forecast inflation. Equations (1) and (2) represent the VAR model, while equation (3) incorporates oil market returns and interaction terms with investor attention. Equations (4) and (5) provide the out-of-sample forecasting models. The rolling window method is used for out-of-sample forecasts, and forecast accuracy is assessed using out-of-sample R-squared, Mean Squared Forecast Error (MSFE), and MSFE-adjusted statistics, comparing against AR(1) and Random Walk (RW) benchmark models. Robustness checks update investor attention using the keyword "monetary policy" and modify model specifications by adding the real interest rate, including higher moments of investor attention, and changing to an interactive model. Finally, the relationship between investor attention and inflation expectations (using University of Michigan Survey data) is investigated using a model with twelve lags to account for information processing time (equation 14). A VIF test checks for multicollinearity.

The VAR(1) model analysis demonstrates a significant negative relationship between investor attention and inflation. The Granger causality test confirms that investor attention Granger-causes inflation. Impulse response analysis reveals that the impact of a shock to investor attention on inflation lasts around seven months. After controlling for oil market factors, the negative impact of investor attention on inflation remains significant. Out-of-sample forecasting reveals that models incorporating investor attention significantly outperform AR(1) and RW benchmark models for both short and long horizons. Robustness checks using the keyword "monetary policy" and different model specifications support the primary findings. Specifically, when both "inflation" and "monetary policy" are used in GSVI, the results are even stronger than those using only "inflation." Using the keyword "monetary policy" alone yields weaker results. The analysis of inflation expectations shows that investor attention significantly influences these expectations, with lagged effects indicating a time lag in information processing.

The findings strongly support the hypotheses. The negative relationship between investor attention and inflation suggests that increased attention to inflation might lead to decreased future consumption and thus lower prices. This aligns with the investor recognition hypothesis and the theory of limited attention. The significant forecasting ability of models incorporating investor attention underlines its importance in inflation prediction, adding to the existing economic and financial factors. The influence of investor attention on inflation expectations further reinforces the mechanism through which this effect operates, bridging the gap between micro-level investor behavior and macro-level inflation.

This study highlights the previously overlooked role of investor attention in inflation dynamics. The findings demonstrate investor attention's explanatory and predictive power for inflation, offering valuable insights for policymakers and economic participants. Future research could explore the use of more sophisticated models to better capture the complex relationships between investor attention and inflation, including potential nonlinear effects. The use of alternative models and the investigation of cumulative investor attention also warrant further investigation.

The study relies on GSVI as a proxy for investor attention, which may not fully capture the complexity of investor sentiment. The linear models used may not fully capture the potential nonlinear interactions between investor attention and inflation. The study focuses solely on the US market; further research could explore the generalizability of these findings to other countries.