Income prediction bias in the gig economy

A TFI Research project by Ray Charles “Chuck” Howard, David J. Hardisty, and Dale W. Griffin
Posted on May 20, 2021

People who work in the gig economy (e.g. driving for Uber or delivering food for Just Eat) normally have both variable working hours and variable incomes. Thus, it can be difficult for them to plan ahead financially. In this TFI research report, Ray Charles “Chuck” Howard, David J. Hardisty, and Dale W. Griffin examine whether people who work in the gig economy tend to over-predict their income or not, and how they can be prompted to better predict their income and improve their financial decision-making

Download the report or read about the summary below to find out about the results

More and more people are turning to the “gig economy” to earn money; for example, driving via platforms like Uber, delivering food through intermediaries like Deliveroo, and completing human intelligence tasks on platforms like Amazon Mechanical Turk. The speed and magnitude of the shift toward gig economy employment has been remarkable. From 2005 to 2015, the number of Americans “gigging” increased by nearly 50%, and 94% of net employment growth in the U.S. economy occurred in gig economy work arrangements (Katz and Kruger, 2016).

An important and previously understudied aspect of working in the gig economy is that this type of income is highly variable, which makes it difficult to predict. In this research report, we introduce and test the hypothesis that gig economy workers display an income prediction bias in which they over-predict their gig income, and we test two “nudges” designed to improve income prediction accuracy.

Income prediction bias: fact or fiction?

Whether gig workers accurately or inaccurately predict their future income is currently an open question. On the one hand, there is research suggesting that prediction accuracy may be the norm. For example, gig workers may engage in “income targeting” and simply work for as long as it takes to earn the amount they predicted they would. On the other hand, research also shows that financial predictions tend to be optimistic (e.g., Peetz and Buehler 2009). Assuming that optimism in the context of income means expecting to earn more money rather than less, this implies that overprediction is the norm.

Do gig workers display an income prediction bias? To test our hypothesis, we began by running three studies with participants from three different types of gigs: driving for Uber, completing online human intelligence tasks for Amazon Mechanical Turk, and delivering food for apps like Deliveroo. In each of these studies we had participants predict their income at the start of the week, then report how much they had actually earned at the end of the week. We also asked participants to predict and report how many hours they would work, allowing us to calculate their expected hourly wage.

We found that participants in Studies 1–3 significantly overpredicted their gig income. In the third study, where the bias was largest, the magnitude of overprediction was $63.5 or 19.9% for a single week. We also found that participants overpredicted the number of hours they would work at their gig. However, our participants’ expected hourly wage was quite accurate. Taken together, these results indicate that gig workers do display an income prediction bias, and that the bias is associated with overpredicting how many hours one will be able to work, rather than overpredicting how much money one will earn per hour.

Improving income prediction accuracy

One reason people tend to make optimistic predictions is simply that it is difficult for the human mind to think of atypical outcomes (e.g., Howard et al. 2021). Uber drivers, for instance, may overpredict their future income because it is easy for them to think of a typical day during which everything goes smoothly, but it is difficult for them to consider atypical events like a car repair that will inevitably interrupt their ability to work.

One solution to this problem is nudging people to take an “outside view” and base their prediction on relevant past experience (e.g., Buehler et al. 1994), which includes atypical outcomes. We reasoned that an effective way to do this in our research was to have people report their total gig income over the past four weeks (e.g., $1,000), calculate their average weekly earnings for them (e.g., $250), and suggest they base their income prediction for the next week on their average past income.

A second nudge that helps people consider atypical outcomes when making predictions is to have them explicitly consider reasons why the future might be different than usual before they make their prediction (Howard et al. 2021). We reasoned that asking gig workers to list two reasons why the number of hours they work at their gig in the next week might be different than usual would reduce their tendency to overpredict. We call this our “atypical” nudge.

To test the “outside view” and “atypical” nudges we ran a fourth study with 662 gig workers who deliver food for apps like Deliveroo. Participants were randomly assigned to predict their gig income for the next week in either a control condition with no special instructions, or in one of the nudge conditions described above. One week later, participants reported how much money they had actually earned from their gig.

The results of our fourth study are illustrated in Figure 1. Participants in the control condition overpredicted their gig income by $62.4 or 23.2%, which replicates the results we observed in Studies 1–3. Participants in the atypical condition overpredicted by $67.4 or 25.7%, indicating that our atypical nudge did not work as planned. However, participants in the outside view condition overpredicted by only $33.1 or 13.4%, meaning the outside view nudge reduced the income prediction bias by $29.3 or 53.0% versus control. Follow-up analyses confirmed that taking an outside view nudged participants to make lower predictions than participants in the control and typical condition, but that participants in all three conditions ended up earning roughly the same amount of money.

The bottom line

Our findings have two clear implications for consumers, financial advisors and policy-makers. First and foremost, gig workers over-predict their gig income. Second, taking an outside view improves prediction accuracy. A third result with practical implications is that income prediction bias is associated with over-predicting the number of hours one will be able to work, rather than over-predicting one’s hourly wage. In other words, gig workers do not appear to be misled about how much they will earn from their gig when they work; rather, they seem to mislead themselves about how many hours they will be able to work. This finding is also important for theorists, because it suggests that optimism in the context of income prediction applies to how hard one will be able to work, but not necessarily how well one can perform their job.

The rapid expansion of the gig economy shows no signs of abating. This makes it important to identify the unique challenges of gig work, and to provide solutions for these challenges. We have done this in our research by documenting an income prediction bias, and showing how it can be reduced. Our hope is that giving gig workers a clearer view of their financial future will help improve their long-term financial well-being.

References

  • Buehler, Roger, Dale Griffin, and Michael Ross (1994), "Exploring the ‘planning fallacy’: Why people underestimate their task completion times," Journal of Personality and Social Psychology, 67 (3), 366.
  • Howard, C., Hardisty, D., Sussman, A., and Lukas, M. (2021). Understanding and Neutralizing the Expense Prediction Bias. Working paper. Available upon request from the first author of this report.
  • Katz, L. F., & Krueger, A. B. (2016). The rise and nature of alternative work arrangements in the United States, 1995-2015 (No. w22667). National Bureau of Economic Research.
  • Peetz, Johanna and Roger Buehler (2009), “Is there a budget fallacy? The role of savings goals in the prediction of personal spending,” Personality and Social Psychology Bulletin, 35 (12), 1579-1591.