GREEDS and Stock Returns: Evidence from Global Stock Markets

Abstract

This paper introduces GREEDS as a new measure of optimistic sentiment in the market. We measure the optimistic component of investor sentiment by constructing the Geographically Revealed Economic Expectations disclosed by Search (GREEDS) index from households’ search behavior on Google for a sample of 38 countries. Our results reveal that the GREEDS index positively correlates with global stock returns. We show the asymmetric effect of GREEDS, which is more prevalent in developed countries than emerging markets. Our findings also highlight the role of global sentiment in financial markets through the sentiment commonality effect.

Keywords:

• GREEDS
• Investor sentiment
• Comovement
• Optimism

Acknowledgments

We are thankful to the Editor, Journal of Behavioural Finance, for the support and useful suggestions. We appreciate the reviewers’ effort in reviewing the manuscript and providing valuable comments. The authors would like to thank Prof. Debasish Maitra and Prof. Mehul Raithatha of IIM Indore for their insightful guidance and constructive suggestions. We are grateful to seminar participants at Corvinus University of Budapest, World Finance and Banking Symposium for their helpful comments. All remaining errors are ours.

Notes

1 Detailed description of countries in terms of developed (DEV) or emerging (EMG) along with market share of Google in each country at the end of 2020 are reported in the parenthesis as follows: Australia (DEV, 94%), Austria (DEV, 94.71%), Belgium (DEV, 93.81%), Brazil (EMG, 95%), Canada (DEV, 87%), Chile (EMG, 98.5%), Denmark (DEV, 95.65%), Finland (DEV, 96.19%), France (DEV, 92%), Germany (DEV, 94%), Hong Kong (DEV, 73%), Hungary (EMG, 97.48), India (EMG, 96%), Indonesia (EMG, 96%), Ireland (DEV, 96.08%), Israel (DEV, 97.54%), Italy (DEV, 95%), Japan (DEV, 57%), Malaysia (EMG, 93%), Mexico (EMG, 94%), Netherlands (DEV, 94%), New Zealand (DEV, 94.71%), Norway (DEV, 94.98%), Philippines (EMG, 89%), Poland (EMG, 97%), Portugal (DEV, 96.44%), Russia (EMG, 58%), Singapore (DEV, 92%), South Africa (EMG, 93%), South Korea (EMG, 77%), Spain (DEV, 95%), Sweden (DEV, 94%), Switzerland (DEV, 92.86%), Taiwan (EMG, 92.87%), Thailand (EMG, 98%), Turkey (EMG, 96%), USA (DEV, 72%), and UK (DEV, 90%). We have also considered China as an emerging economy. However, we drop China from our main sample, because of the dominance of Baidu search engine (more than 70%) over Google in China. Similar observations related to low penetration of Google search engine for China have also been reported by Gao, Ren, and Zhang (2020). Thus, our main empirical results are reported excluding China from the sample.

2 The list is downloaded from http://www.wjh.harvard.edu/∼inquirer/spreadsheet_guide.htm.

3 See Kearney and Liu (2014) for a detailed review of literature.

4 On a comparable note, the top 30 negative words listed by Da, Engelberg, and Gao (2015) includes; “jobless,” “inflation,” “recession,” “depression,” “unemployed,” “expense,” “bankruptcy,” “the crisis,” “poverty,” etc.

5 Baker, Bloom, and Davis (2016) provide data for EPU index at https://www.policyuncertainty.com/us_monthly.html.

6 Data for ADS index is obtained from https://www.philadelphiafed.org/surveys-and-data/real-time-data-research/ads.

7 Following Chen (2017) we either use the Bloomberg tickers for country specific benchmark indices or the index name or other equivalent abbreviation for search term to capture market attention. For e.g., “DJIA” for the US and “SENSEX” for India as search term.

8 In all our specifications and results discussion, the significance of Hausman test statistics indicates suitability of fixed effect models over random effects.

9 For instance, in Table 4 of Da, Engelberg, and Gao (2015, 14) the return reversal effect of FEARS on S&P 500 index returns has been documented for subsequent days (t + 1, t + 2).

10 In our unreported results, we estimate our models by including China in our main sample and in the emerging market subsample and our results are qualitatively similar to the Table 4 and Table 5 findings. Hence, excluding China from our main sample due to its low Google search engine penetration does not influence our results.

11 We thank anonymous referee for guiding us in this direction.

12 The standard deviation of GREEDS index for developed countries is 0.065261 and that of emerging countries is 0.065434.

13 The marginal effect of global sentiment is computed as the sum of the GLOBAL_SENT coefficient in the regression models, and the sample means of financial characteristics times the interaction coefficient of GLOBAL_SENT with financial characteristics. The mean values of FI, FM, and FD are 0.685, 0.618, and 0.664, respectively.