Learning by doing – the notion that workers can get better at a task by tinkering and refining their technique – is a central concept in economics. The authors of a new study appearing in the American Economic Journal: Applied Economics look for signs of learning by doing in a sample of microentrepreneurs: 6,300 New York City taxi drivers who provided tens of millions of rides in the city during 2009.

First they look to see how much experienced drivers earned in sixty-minute windows after making dropoffs in different neighborhoods at different times of the day and week. This allows them to distinguish the most lucrative locations and times (lower Manhattan and the areas near JFK airport, especially during rush hour) from other locations and times that require more effective searching to turn up fares. In the figure below, neighborhoods shaded more darkly are more profitable.

Figure 1 from Haggag et al. (2017)

The authors find that the newest drivers do tend to earn significantly less after making dropoffs in more difficult neighborhoods or during less profitable times of day, presumably because they have not yet mastered techniques for scrounging up fares in less dense areas or neighborhoods they don’t know well. By contrast, the novices do basically just as well as experienced drivers in the best neighborhoods and times when it is very easy to find the next fare.

Ultimately, though, there is relatively little learning in this setting. The average novice starts out only earning only about 5% less per hour than an experienced driver. He typically catches up within about 100-120 shifts, and only loses out on about $350 total over that span.