Laura Gee, Tufts University
When ice cream sales rise, so does crime. Of course, ice cream doesn’t cause crime—hot weather drives up crime rates, and it’s also a great excuse for mint chocolate-chip. Because economists have a large impact on policy makers, it’s important that they identify causation and not correlation in behavioral studies. We wouldn’t want anyone outlawing ice cream, after all.
Photo by Annie Shelmerdine.
Over the last 40 years, the field of economics has increasingly focused on distinguishing causation from correlation in order to better predict the decision-making of individuals.
Before the so-called "credibility revolution," economists tended to use a combination of very simple theoretical models and non-experimental observational data. These methods provided very clear predictions of behavior—but sometimes at the cost of accurately predicting real-world choices.
For a new generation of economic researchers—especially in behavioral economics—a new approach has gained popularity, both in modeling preferences for decision-making and in validating causal effects.
Much has been written about behavioral research disputing the "rational economic agent" model assumed by certain schools of economics. Newer studies are adopting more nuanced models of preference (known in both contemporary and historical contexts as utility functions).
Secondly, economists have increasingly turned to experimental methods such as randomized controlled trials, like you might see in medical research, to better identify causal factors.
To illustrate both effects—more nuanced models and more rigorous validation—let’s turn to a recent study on the effects of "feeling pivotal" in charitable giving.
Understanding Charitable Giving
Charitable giving provides a clear example of why theoretical models matter. A simple model of the decision about whether to donate to a charity might use this equation:
Utility from action = (Dollars you have) - (Dollars you give to charity)
An economist using this model would predict that no one will ever donate to charity—the end result of donation always leaves you with less money, after all. Yet that clearly isn't what happens in the real world: in 2018 alone individuals in the USA donated 286 billion dollars.
So we should tweak the decision model.
Something that has been observed many times is that when people feel pivotal to reaching a goal, they are more likely to take an action. Making a difference seems to give people a pleasurable “warm glow.”
We could alter our model slightly to reflect this:
Utility from action = (Dollars you have) - (Dollars you give to charity) + (Happiness from reaching the goal) + (Warm glow from taking the action)
With our more complex model we now predict that people will donate if they get enough happiness from some combination of making the donation and seeing the goal met.
Photo by Sharon McCutcheon
Our model is now a lot better at matching the real-world facts that people do actually donate money, and that people are more likely to give when a charity is just about to reach its goal. But this could just mean that charities that are close to their goals are also more deserving, and feeling pivotal might not really be causing the donation. To disentangle causation from correlation and to test our new theoretical model, we ran some experiments.
Testing the Effect of Feeling Pivotal
To test the model we ran an experiment where we made some people feel more pivotal to reaching the goal, and others less so. We did this by partnering with a charity and sending out three different messages to their potential donors, which we paraphrase below:
Control Message: Please give.
Less Pivotal Message: Please give. You're in a group that needs to help the charity reach a goal, but you're not very likely to be pivotal.
More Pivotal Message: Please give. You're in a group that needs to help the charity reach a goal, and you're very likely to be pivotal.
The results were clear. While only 1.6% of the control group donated, 2.0% of people donated who got the less pivotal message, and 3.7% donated when they got the more pivotal message. That might not seem like much, but it is a 230% overall increase in donations.
Because we randomly assigned the messages, we are confident that the message caused the increase in donations.
Looking beyond charitable donations
This analysis of decision-making is relevant beyond just the realm of charitable giving. One readily available analogy to our "feeling pivotal" model arises in the domain of politics, specifically in voting behavior.
To apply here, the model could be generalized:
Utility from action = (Happiness you have) - (Loss of happiness from taking an action) + (Happiness from reaching the goal) + (Warm glow from taking the action)
This more general model could then be used to explain why people vote, for example, even though it takes time and effort. (Indeed about 60% of the population voted in the 2016 election.)
Photo by Parker Johnson.
We know that voter participation rates are higher when an election is close. Do more people vote because they think, if they perceive an election to be close, that they might be pivotal to their candidate prevailing? Perhaps, but close elections also have more resources put into them, some of which drive voter participation.
We could design an experiment where we show people different polls that vary how close the election is, and measure if those who are made to feel more pivotal are more likely to vote.
If both components of the “credibility revolution”—more nuanced theoretical models and more rigorous experimental validation—can be adapted and applied to broad areas of the social sciences, we may have an increasingly better understanding of what causes (rather than what is merely correlated with) a diverse array of outcomes. This understanding of causes should, in turn, lead to better recommendations for government policy makers, business leaders, and individuals.
Laura Gee
is an Associate Professor of Economics at Tufts University. Her research is in behavioral economics, with a particular focus on how individual decision-making is influenced by group dynamics. Her studies rely on both lab and field experiments, as well as observational data.