Paper: arXiv 2308.10039
Authors: HyeonJun Kim
Abstract
This study explores the potential of internet search volume data, specifically Google Trends, as an indicator for cross-sectional stock returns. Unlike previous studies, our research specifically investigates the search volume of the topic ‘happiness’ and its impact on stock returns in the aspect of risk pricing rather than as sentiment measurement. Empirical results indicate that this ‘happiness’ search exposure (HSE) can explain future returns, particularly for big and value firms. This suggests that HSE might be a reflection of a firm’s ability to produce goods or services that meet societal utility needs. Our findings have significant implications for institutional investors seeking to leverage HSE-based strategies for outperformance. Additionally, our research suggests that, when selected judiciously, some search topics on Google Trends can be related to risks that impact stock prices.
Complexity vs Empirical Score
- Math Complexity: 4.0/10
- Empirical Rigor: 6.0/10 (capped from 7.5: transaction costs not modelled; no out-of-sample test; see penalty flags)
- Quadrant: Street Traders — empirical and applied, lower math complexity (effective empirical rigor capped from 7.5)
Why this score: The paper applies relatively simple linear regression and portfolio sorting methods (low math complexity) but uses real financial data (Korean stock market via DataGuide, Google Trends), performs rigorous backtesting with Fama-MacBeth regressions, double sorts, and provides detailed statistical results (t-values, significant returns), indicating strong empirical implementation.
Research Flowchart
flowchart TD A["Research Goal<br>Impact of 'Happiness' Search Volume<br>on Stock Returns"] --> B["Data Collection"] B --> C["Data Preparation"] C --> D["Computational Modeling"] D --> E["Key Findings"] B --> B1["Google Trends Data<br>Happiness Search Exposure HSE"] B --> B2["Korean Stock Market Data<br>Returns & Firm Characteristics"] C --> C1["Align Temporal Data<br>Monthly Frequency"] C --> C2["Define Variables<br>Portfolios by Size & Value"] D --> D1["Time-Series Regressions<br>Fama-MacBeth Approach"] D --> D2["Factor Analysis<br>Risk Pricing Evaluation"] E --> E1["HSE Significantly Predicts Returns<br>Big & Value Firms"] E --> E2["HSE Reflects Firm Utility Production<br>Risk Pricing Mechanism"] E --> E3["Implications for Investors<br>HSE-based Strategies"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes that happiness search exposure (HSE) from Google Trends can be used as a risk factor to explain cross‑sectional stock returns, especially for large and value firms.
Source sentence
Empirical results indicate that this ’happiness’ search exposure (HSE) can explain future returns, particularly for big and value firms.
Assumes 6
- Happiness Sentiment Exposure is systematic but noisy. model
Source sentence
First, as we assumed that Happiness Sentiment Exposure is systematical but noisy, we had a concern that Happiness Sentiment Exposure will be misestimated due to market proxy
- Market proxy is omitted to avoid contaminating the HSE factor loading. model
Source sentence
Second, if the market proxy is added to the analysis, then the factor loading for Happiness Sentiment will measure the relationship with Happiness Sentiment with abnormal returns, which is not this paper’s topic.
- Returns are winsorized at the 1st and 99th percentiles. statistical
Source sentence
The returns are winsorized by 99percentile and 1percentile thr…
- BE/ME is calculated by dividing equity minus preferred stock market cap of December of year t. data
Source sentence
BE/ME at year t was calculated by dividing equity minus preferred stock last trading day market cap of December of year t
Show 2 more
- Portfolios are formed in June using HSE data from July t‑4 to June t and rebalanced in July t+1. execution
Source sentence
The portfolio is formed in June of year t using the HSE data from July of year t −4 to June of year t(with at least 24 observation) and reblanced that July of year t + 1.
- Daily return and market‑cap data are converted to monthly frequency. data
Source sentence
Return data is changed from daily to monthly frequency, and BE/ME at year t was calculated...
Data
- Google Trends happiness search volume alternative public
Source sentence
search volume of the topic ’happiness’
- Daily return including dividends and market cap data market-prices commercial
Source sentence
Daily return including dividends, daily market cap of common stocks, equity, preferred stock data was collected from DataGuide.
Costs
Not modeled
Finds 5
- HSE has explanatory power for future returns even after controlling for size and BE/ME, with stronger predictive power for big and value firms. predictive-accuracy
Source sentence
Empirical analysis show that stock return’s exposure to the search volume of topic ’happiness’(or HSE) have explanatory power to future return even when BE/ME and size is controlled.
- Average excess return increases monotonically with HSE; the 10th decile (Happy) outperforms the 1st decile (Unhappy) by over 17% annualized. performance
Source sentence
After 8th decile, annualized expected return is over 9.3%, and the annualized excess return difference between 10th decile(Happy) and 1st decile(Unhappy) is over 17%.
- Non‑linear relationship: both low and high HSE stocks are small‑cap and low‑value, while middle HSE stocks have the highest BE/ME and market cap. statistical
Source sentence
low HSE stocks and high HSE stocks are both small cap and low value firms, while middle HSE
- Fama‑MacBeth regressions on subsamples confirm that HSE explains cross‑sectional returns better for big and value firms than for small and growth firms. pricing-calibration
Source sentence
Big and Small firm Subsample and Value and Growth firm Subsample Fama-Macbeth regression support the latter hypothesis.
- The equally weighted high‑decile portfolio yields a robust and statistically significant excess return, though the monotonic pattern weakens due to size and value effects. performance
Source sentence
for the equally weighted portfolio, the excess return for high decile portfolio is robust and statistically significant, but the monotonic increase of return in higher decile became less apparent.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 30 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.