For academics: research design to study Individual investors’ preferences for corporate social responsibility
This page explains how academics could use our research design to study individual investors’ preferences for corporate social responsibility in financial markets. With the SimTrade market simulations used in the experiments, you can get relevant data about the behavior of individual market participants such as trading behavior, learning behavior, investor profile, etc.
In our blog post The Three Rs of Science: Relevance, Rigor, and Reach published on ESSEC Pedagolab, we oultine a balanced framework for academic research that aspires to be both intellectually robust and socially impactful. The 3R approach advocates for Relevance, by addressing meaningful and timely societal and managerial issues; Rigor , through the use of sound and transparent scientific methods; and Reach, by ensuring that research findings extend beyond academic circles to inform practice, policy, and public discourse. Rather than viewing these dimensions as trade-offs, we encourage researchers to integrate them thoughtfully, aiming to produce knowledge that is methodologically credible, contextually significant, and widely accessible.
Research design
To explore individual investors’ perceptions of corporate social responsibility (CSR), we design a simulation-based lab experiment where participants trade their portfolio (cash and stocks) on a company and react to the news flow, which is a series of events about the company that may unfold during the day. Among these events, one of them is a CSR-related event: the announcement of the corporate donation after a natural incident (a tsunami that ravaged the Atlantic coast). Participants expect such a corporate donation, but they ignore beforehand whether that donation will be high, moderate or low (compared to the company’s profit). By randomizing the donation level for each individual participant, we can identify the participants’ trading reactions according to different levels of corporate donations. The participants’ trading decision to buy, hold or sell stocks signals their perceptions regarding the company’s CSR.
We adopt an experimental approach as it enables us to identify a causal relationship between corporate social responsibility (CSR) and investor behavior. Traditional empirical methods often suffer from limitations such as omitted variable bias and reverse causality, which obscure whether CSR drives investment decisions or merely responds to investor expectations. In addition, survey-based studies are susceptible to social desirability bias, whereby investors may claim to value CSR but fail to reflect this in their actual trading behavior. Our experimental design addresses these limitations by focusing on observed trading actions rather than self-reported preferences.
Experimental design
We use an experiment because it allows us to observe trading decisions at the individual level, mapping information about participants’ profile and trading behavior to the CSR-related event. In the experiment we manipulate only the amount of corporate donation (small, medium, and high amounts) while keeping all other factors constant; this ensures high internal validity.
Choice of experimental design
There are two different methods to design an experiment: the between-subject and the within-subject design. Both design methods have their advantages and drawbacks. The main point to consider when choosing the design method is the importance of the “demand effect”, which is a spurious effect reflecting the attempt of participants to behave to satisfy their perception of the experimenter’s expectations. By changing the value of a parameter from one experiment to another, the participant understands what the experiment is about. In our case, by donation level between the two consecutive simulations, the participant would understand that the experiment is about corporate donations. For our design experiment, to avoid a “demand effect”, we then decided to adopt a between-subject design.
Realistic simulation tool
To carry out our research, we use an on-line trading simulation platform called SimTrade. In terms of experimental design, compared to traditional out-of-context experiments (such as lotteries used to measure preferences), the added value of SimTrade is contextualization: similar to traders in an investment bank, participants in the simulation can buy and sell stocks of a company in reaction to stock market price fluctuations and economic news. The experiment conducted by using the SimTrade simulation platform then provides a realistic environment to study the decisions taken by the participants of the experiment.
The experiment was carried out online and in the ESSEC experimental research lab designed for conducting experiments in a controlled environment. The experiment was presented to students as an opportunity to contribute to a research project studying how economic agents make their financial decisions. Following common practice, the CSR aspect of the research project was not revealed to participants to avoid disclosing our research subject.
At the beginning of the experiment, the participants are required to read the description of the scenario, which includes a description of the firm, a brief history of the stock price, the events expected to occur during the trading session (the appointment of the next CEO is one of the events presented) and the market consensus by financial analysts. The participants are presented the sequence of events that may unfold during the simulation. This sequence includes several major events: a CSR-related —the announcement of a corporate donation —and several non-CSR-related — such as the announcement of financial results, the news disclosure of a macro-economic indicator, and the result of a nation-wide tender offer. Those non-CSR-related events serve as placebo check to study the robustness of our results. Prior to the announcements of these events, the participants are uninformed about their final outcome. Regarding the CSR-event, while they know that the company is going to announce a corporate donation, they do not know its level (i.e., high, moderate, or low). Regarding the announcement of the company’s financial results, they do not know whether the company will achieve the analyst’s consensus or not. Regarding the macro-economic indicator, they do not know if it will be the market consensus or not. Finally, and regarding the tender offer, they do not know whether the company will win it or not. Participants are asked to act as investors whose objective is to maximize their gains out of the trading day, something of which they are clearly reminded at the start of the simulation. As an incentive, participants receive compensation in the form of a course grade bonus derived as a function of their realized gains during the simulation. The participants, who have followed the core finance course, are familiar with the functioning of financial markets, but this is the first time that they launch the simulation used for our experiment.
Balancing internal and external validity in experimental research
Our laboratory experiment, conducted via the simulation-based SimTrade platform, enables precise measurement of individual participant behavior. By collecting granular data on trading decisions alongside demographic information—such as age, gender, and educational background—we enhance the internal validity of our findings. This controlled environment allows us to systematically manipulate key variables, including the level of corporate donations, as well as the dissemination of information to participants.
Improving external validity with our research design
While this methodological rigor strengthens internal validity, it may limit the generalizability—or external validity—of our results. This trade-off is a common challenge in experimental research. To address this, we offer our experimental setup for replication and adaptation by the academic community. Implementing the experiment across diverse contexts would be a valuable step toward enhancing the external validity and broader applicability of our conclusions.
