This study investigates the effect of investor attention on stock excess returns in the Tehran Stock Exchange, using weekly data from 52 companies over a five-year period. The Google Search Volume Index (GSVI) is employed as a proxy for investor attention, and its impact across different return levels is analyzed using quantile regression. This study distinguishes itself from similar domestic research both in its focus on stock excess returns as the dependent variable and in its use of quantile regression methodology. By utilizing the quantile regression approach, it is less sensitive to outlier data and avoids statistical issues such as omitted variable bias. The results indicate that investor attention has an asymmetric effect on stock excess returns. Specifically, for stocks whose excess returns lie within the middle quantiles of the excess return distribution, the effect of investor attention on these stocks exhibits short-term persistence. Conversely, for stocks whose excess returns are in the lower quantiles of the distribution, increased investor attention leads to selling pressure and a decrease in their excess returns in subsequent weeks. Furthermore, for stocks of companies whose excess returns are in the upper quantiles of the excess return distribution, an increase in detrended trading volume has a positive and significant effect on stock excess returns. This study shows that investors' behavioral indicators can be used as a complementary tool alongside well-known classical market variables to predict and achieve higher returns for investors.

Introduction and Theoretical Foundations

Traditional financial theories, such as the Efficient Market Hypothesis, assume investors are fully rational. However, behavioral finance challenges this, highlighting the role of psychological factors like "attention" as a limited cognitive resource. In today's digital age, online search behavior provides a measurable proxy for this otherwise difficult-to-quantify concept. This study leverages the Google Search Volume Index (GSVI) as a novel proxy for investor attention. The study investigates a central question: How does investor attention asymmetrically affect stock excess returns in an emerging market like the Tehran Stock Exchange (TSE)? The present research is distinct from prior domestic studies by focusing directly on "excess return" and employing a robust quantile regression approach, moving beyond traditional average-based analyses to capture heterogeneous effects across the entire return distribution.

Data and Methodology

Our empirical analysis utilizes a unique panel dataset of 52 companies listed on the TSE. We collected weekly data over a five-year period (2019-2024), resulting in 11,128 observations. The dependent variable is Stock Excess Return (ER), calculated as the difference between an individual stock's weekly return and the return of the overall market index, adjusted for its beta (systematic risk). The primary independent variable is the change in the standardized Google Search Volume Index (SGSVI) for each company's ticker symbol, which captures the week-over-week shift in investor online attention.
To control for market dynamics, we include detrended trading volume (VLMt) and one-week lags of both the dependent variable (ER1) and the independent variables (SGSVI1, VLMt1). The core of our methodological contribution is the use of Panel Quantile Regression. This method was chosen for three crucial reasons:

It bypasses normality assumptions: A Shapiro-Wilk test strongly rejected the normality of our ER variable (p < 0.001), making standard OLS regression potentially biased.
It is robust to outliers: Financial data is well known to be prone to extreme values, and quantile regression is resistant to their influence.
It reveals heterogeneity: Unlike OLS, which only estimates the average effect, quantile regression allows us to examine how the impact of investor attention differs for stocks with low (e.g., Q1, Q2), medium (e.g., Q5), and high (e.g., Q8, Q9) excess returns. We also conducted Pesaran's Cross-sectional Dependency test and second-generation panel unit root tests (CADF) to ensure the validity of our data.
Key Findings and Results

Our quantile regression results reveal a clear and pronounced asymmetric effect of investor attention on stock excess returns. The coefficients for SGSVI vary significantly across quantiles (confirmed by a Wald test, p < 0.001), which is the central finding of this study. The results can be broken down as follows:
For Low-Performing Stocks (Lower Quantiles, e.g., Q1-Q4): The effect of investor attention (SGSVI) is negative and statistically significant. For example, at Q1 (the 10th percentile), an increase in attention correlates with a -0.0068 decrease in excess return. This suggests that heightened online searches for these "unpopular" or underperforming stocks signal investor concern or panic, leading to selling pressure and a further decline in returns in the following weeks. The lagged attention variable (SGSVI1) shows a similarly negative effect, indicating the persistence of this sell-off pressure.
For Average-Performing Stocks (Middle Quantiles, e.g., Q5): Here, the effect of current attention (SGSVI) is non-significant (p-value 0.805). However, the lagged dependent variable (ER1) is positive and significant, indicating that these stocks exhibit short-term return persistence driven by their own momentum rather than new attention shocks. Investors seem to react more moderately and rationally to this group.
For High-Performing Stocks (Upper Quantiles, e.g., Q7-Q9): The effect dramatically reverses and becomes positive and strongly significant. At Q9 (the 90th percentile), an increase in attention (SGSVI) is associated with a +0.0068 increase in excess return. This finding indicates that for stocks already delivering high returns, increased investor attention acts as a positive reinforcement signal, attracting more buyers, generating buying pressure, and driving prices even higher.
Furthermore, detrended trading volume (VLMt) shows a negative effect at the lowest quantiles and a strong positive, significant effect from the median (Q5) upwards. For example, at Q9, a one-unit increase in VLMt leads to a 0.0211 increase in excess return. This suggests that high trading volume accompanies institutional investor inflows for high-performing stocks and signals divestment for underperforming ones.

Discussion and Conclusion

This study provides robust empirical evidence that investor attention, measured through the GSVI, is a significant and asymmetric driver of stock excess returns in the Tehran Stock Exchange. The central finding is that the effect of investor attention is not inherently positive or negative; it depends entirely on the stock's recent performance context. For high-performing stocks, increased attention reinforces upward return momentum. For underperforming stocks, it accelerates selling pressure and further return decline.
These findings carry practical implications. Investors and analysts can incorporate Google Trends data as a real-time behavioral indicator alongside conventional tools such as P/E ratios and technical charts. More specifically, a sudden spike in search volume for a consistently underperforming stock may serve as an early sell signal, while a steady rise in attention toward a high-momentum stock can reinforce a continuation pattern and suggest a buying opportunity.
This study has two main limitations. First, Google Trends restricts data extraction to a five-year window. Second, our data does not distinguish between search intent (e.g., buying vs. selling research). Future research could incorporate high-frequency intraday search data or classify search terms (e.g., "company news" vs. "company stock price") to refine the measurement of attention and its causal impact.