Why it’s hard to predict gig income

by Chuck Howard, David Hardisty, & Dale Griffin
Posted on October 08, 2020

As unemployment rates spike due to Covid-19, an increasing number of people are turning to the gig economy for work. This increase in the number of gig workers has driven their wages down and made it even more difficult to predict their own income over time.1 The impetus to help gig workers plan their financial future has only grown since our previous blogpost.

To understand the impact that the pandemic may have had on the accuracy with which gig workers can predict how much they'll earn, we recently collected additional data from Uber drivers and people who gig for food delivery apps like DoorDash and Skip the Dishes in the US.2 In August and September 2020 we asked our participants to:

  • Predict their gig income and hours of gig work for the next week.
  • Report their perceived average income for the previous eight weeks (so we could examine the relationship between perceived average income and predicted future income).
  • Report how long they had been working at their gig (so we would measure the extent to which prediction accuracy is associated with experience).
  • Report their actual earnings and hours worked at the end of the target week (so we could examine their prediction accuracy).

Expectations vs reality

As shown in the figure below, we find that Uber drivers over-predicted their income for the next week by $64 or 18%. On average, they predicted their income at the start of the week to be $415 but reported their real income at the end of the week to be $351. They also over-predicted the number of hours they would work by almost 6 hours or 31%. On average, they predicted they would work 24 hours at the start of the week but ended up working 18.3 hours. Interestingly, more experienced drivers made somewhat more accurate predictions than less experienced drivers, but even the experienced drivers tended to over-predict.

Food delivery drivers show similar results: they over-predicted their income for the next week by $64 or 20%. They also over-predicted the number of hours they would work by 4 hours or 21%. Among the delivery workers, more experienced drivers were just as inaccurate as less experience drivers, and full-time drivers were as inaccurate as part-time drivers.



Why are these gig workers over-predicting their future income? The data provide a compelling clue for this optimistic bias: predicted income in both samples was remarkably similar to drivers’ perceived average income, which can be influenced by a few high-earning weeks.3 This, along with our previous findings, implies that what drivers perceive to be their average income may be higher than what they earn in a typical week.

Thinking things through

Our results suggest that we can help drivers improve their prediction accuracy by prompting them to consider reasons why the future may not be “average”. We are currently testing this hypothesis is a longitudinal experiment with 600 delivery US drivers, and we look forward to sharing the results early next year.

A final insight derived from our new data that warrants mention is that although a strong majority of drivers over-predict their future income, a small number do end up significantly under-predicting. Why? Our qualitative data suggests this is because drivers can occasionally “hit the jackpot” and earn a very generous tip. This is undoubtedly beneficial to these drivers in the short-term, but it also underlines the fact that gig economy income can be highly variable and therefore highly unpredictable.

Dale W. Griffin is Professor of Marketing and Behavioural Science at UBC Sauder School of Business

David J. Hardisty is Associate Professor of Marketing and Behavioural Science at UBC Sauder School of Business

Chuck Howard is Assistant Professor of Marketing at Texas A&M’s Mays Business School