The Minimum Wage in an Online Job Market

John Horton has written a novel paper that uses an experiment and a policy change in an online job market to understand the effects of the minimum wage. The job market in question is something like the Upwork platform where firms can post jobs and workers from anywhere in the world can post offers to work at an hourly wage to do tasks such as computer programming, data entry, design and transcription. Typically workers are hired for a week or two.

Horton was able to implement a minimum wage by simply not allowing a worker to offer to work at less than the minimum wage for a randomly chosen set of jobs.

During the experimental period, firms posting an hourly job opening were immediately assigned to an experimental cell. The experiment consisted of four experimental cells: a control group with the platform status quo of no minimum wage, which received 75% of the sample (n = 121, 704), and three active treatment cells, which split the remaining 25% of the sample. A total of 159,656 job openings were assigned. Neither employers nor workers were told they were in an experiment. The active treatments had minimum wages of $2/hour in MW2 (n = 12, 442), $3/hour in MW3 (n = 12, 705), and $4/hour in MW4 (n = 12, 805).

Horton found that the minimum wage did reduce hiring, especially in low-wage job categories when the minimum wage was high relative to the median wage. The hiring reduction was measurable but, consistent with previous research, not large. All the work on the platform, however, is logged through the software so Horton also has very good data on hours worked and here the story is quite different. The minimum wage substantially reduced hours worked.

A higher minimum wage likely causes firms to scale back projects but that seems somewhat inconsistent with the small effect on hiring (fixed costs of hiring would suggest fewer hires and fewer hours but perhaps more hours per hire.) Horton finds another factor explains the reduction in hours worked. At a higher minimum wage, firms are careful to hire more productive workers. He finds that about half of the decline in hours can be explained by substitution towards higher productivity workers. Previous studies have found suggestive effects along these lines. For example, Giuliano 2013 found that the higher minimum wages could shift teenage employment to teenagers from more affluent regions who were likely more skilled and less likely to quit. Horton finds similar demographic effects as hiring shifts away from Bangladeshi workers and towards US workers but since his data on productivity is much cleaner than in previous studies there is less need to rely on demographic correlates of productivity.

In part (it seems) due to the experiment, the job-platform later instituted a $3 per hour minimum wage for all jobs. Horton is thus able to supplement his experimental results with analysis of a policy change in the same environment. Consistent with the experimental result, the imposition of the minimum wage across the board caused substantial declines in hours worked with little effect on hiring overall but a big effect on the lowest-wage workers who found that their probability of being hired dropped substantially after the minimum wage was imposed.