Beliefs About the Stock Market and Investment Choices: Evidence from a Field Experiment

How to cite

Wohlfart, Johannes; Laudenbach, Christine; Weber, Annika; Weber, Ruediger. 2021. "Beliefs About the Stock Market and Investment Choices: Evidence from a Field Experiment." AEA RCT Registry. AEARCTR-0008462. https://doi.org/10.1257/rct.8462-1.1

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Johannes Wohlfart* , Christine Laudenbach , Annika Weber , Ruediger Weber

* Primary investigator

Registry ID
AEARCTR-0008462
DOI
https://doi.org/10.1257/rct.8462-1.1
First registered
Last updated
Status
Completed
Trial dates
2019-09-01 – 2019-12-31
Countries
Germany ()
JEL codes
D14, D83, D84, D91
Registry record
http://www.socialscienceregistry.org/trials/8462

Abstract

We study households' beliefs about the autocorrelation of annual returns of the aggregate stock market and the role of these beliefs in investment decisions. Based on a survey of retail investors at a German online bank, we document substantial heterogeneity in individuals' perceived autocorrelation, with a majority believing in mean reversion of returns. These beliefs predict how respondents adjust their portfolios in response to market movements, as measured in administrative account data. We provide a random half of our respondents with historical information on the low predictive power of realized returns for year-ahead returns. The intervention persistently reduces respondents' perceived autocorrelation, and shifts their expected year-ahead return accordingly. We find only minor adjustments of investment decisions in the short term. However, among those believing in mean reversion before the intervention, treated respondents display a significantly smaller increase in equity purchases in response to the COVID-19 stock market crash four to five months after the treatment. Our results suggest that heterogeneity in perceived models is a causal driver of trade in asset markets, and may more generally lead to differences in economic decisions across individuals.

Intervention

Between September and November 2019, we conducted a survey of about 2,000 stockholders that are clients of a major German online bank. In our survey, we elicit respondents' beliefs about the historical autocorrelation of stock returns based on a new, individual-level measure. Specifically, we first ask respondents to think of six intervals of historical annual return realizations of the German stock market index (DAX) during the last 50 years. For each return interval, respondents are asked to provide an estimate of the average return of the DAX over the subsequent 12 months if the return over the previous 12 months fell into that interval. Subsequently, a random half of the respondents are informed about the actual historical conditional mean return over the following year for each of the six scenarios of returns over the previous year. Actual conditional mean returns in the six scenarios vary only narrowly around the unconditional historical average return of the DAX of 8.5 percent, illustrating the historical low degree of predictive power of recent returns for future returns at the annual frequency. In both the main survey and a follow-up survey conducted four weeks later, we then measure our respondents' posterior beliefs about the autocorrelation of stock returns and the return over the 12 months after the survey.

In July and August 2022, we conducted four additional surveys:
- an additional experimental investor survey using an analogous approach as pursued in our main study to measure and shift beliefs about return predictability based on the price-dividend ratio.
- an additional descriptive investor survey to study the sources of investors' beliefs about the autocorrelation of returns.
- an additional general population survey from Germany measuring respondents' perceived autocorrelation of returns.
- an additional general population survey from the US measuring respondents' perceived autocorrelation of returns.

Experimental design

Our setup generates exogenous variation in individuals' beliefs about the autocorrelation of aggregate stock returns, which in turn allows us to examine whether subjective models of the stock market are a causal driver of people's return expectations. Moreover, by linking the treatment variation with administrative account data on trading decisions before and after the intervention, we can obtain causal evidence on the role of individuals' perceived models of the stock market in their investment decisions.

Design details

Randomization method
By computer.
Randomization unit
individual
Clusters
Main study: 2,083 individuals Additional experimental investor survey: 772 individuals Additional descriptive investor survey: 227 individuals Additional general population survey from Germany: 504 individuals Additional general population survey from the US: 508 individuals
Observations
Main study: 2,083 individuals Additional experimental investor survey: 772 individuals Additional descriptive investor survey: 227 individuals Additional general population survey from Germany: 504 individuals Additional general population survey from the US: 508 individuals
Arms
Main study: 1,058 individuals in the control group; 1,025 individuals in the treatment group. Additional experimental investor survey: 336 individuals in the control group; 357 individuals in the treatment group.
Primary outcomes
Beliefs about the autocorrelation of stock returns, stock return expectations, trading behavior

Cited by

No citation data available for this registration yet.

Keywords