Principal investigator: Eduarda Miller Figueiredo
Original title: The Value of Flexible Work: Evidence from Uber Drivers
Authors: M. Keith Chen, Judith A. Chevalier, Peter E. Rossi, and Emily Oehlsen
Location of the Intervention: United States
Sample Size: 197.517 drivers
Sector: Labor market
Primary Variable of Interest: Salary
Type of Intervention: Uber (flexible work)
Methodology: Multivariate Probit
Summary
Companies have launched business models to meet the demand for services using independent providers at unconventional hours. Although these companies do not offer many of the traditional employment benefits, they offer an opportunity for workers to receive compensation on a flexible schedule. This article seeks to analyze behavioral patterns of Uber drivers and provide preliminary evidence on the types of flexibility sought by these drivers. Using a Multivariate Probit model, they found that the sample drivers experience large, inconsistent salary changes from week to week and therefore place a high value on flexible work arrangements.
- Policy Problem
In recent years, several companies have launched business models that seek to meet the demand for services with independent service providers who work intermittent or non-standard hours. While these companies do not offer many of the traditional employment benefits, they provide an opportunity for service providers to receive compensation on a flexible schedule.
The literature has examined flexible workplace practices. The American Council of Consultants[1] (2010) reports that 81% of employers would allow some employees to periodically change their start and end times within a set range of hours, while 27% of employers would do the same for most or all employees. However, only 41% would allow some employees to change their start and end times daily. Thus, employers typically seem to have preferences for certain working hours for employees. Furthermore, the literature also suggests that lower-paid workers have less flexibility than higher-paid workers (Bond and Galinsky, 2010).
Therefore, this article aims to analyze behavioral patterns of Uber drivers and provide preliminary evidence on the types of flexibility sought by these labor providers.
- Implementation and Evaluation Context
Uber is a platform where drivers, once approved, can use their own cars – or rented ones – to offer rides whenever they want. There are no minimum hour requirements and few restrictions on maximum hours. The fares paid by passengers are set at the city level and adjust dynamically (increase) when demand is high relative to the supply of drivers in a given area.
Since drivers can work whenever they want, a driver's earnings at any given time are effectively determined by the driver's marginal willingness to work. However, wages are uncertain and compensation can be quite low. In other words, the main advantage is also a significant disadvantage: drivers can work whenever they want.
Hall and Krueger (2016) examine research evidence and administrative data from Uber. They document that drivers cite flexibility as a reason for working for Uber and that for many it is a part-time activity, secondary to more traditional employment (Hall and Krueger, 2016; Campbell, 2018).
Given that Uber drivers largely work part-time, it is not surprising that their work hour patterns do not resemble those of people with conventional jobs. Figure 1 compares the work habits of drivers with the work habits of employed men over 20 years old in 2014 (ATUS).[2]).
Figure 1: Comparison of Uber driver activity with ATUS workers.
The graph in Figure 1 shows the proportion of these drivers working each of the 168 hours of the week. It is evident that ATUS work occurs mainly between 9 am and 17 pm, while Uber drivers are more likely to work at 18 pm or 19 pm than at 14 pm or 15 pm. While ATUS men are about half as likely to work on Saturday afternoons, Uber drivers are more likely to work on Saturday afternoons and evenings.
- Policy/Program Details
The authors used data provided by Uber, which contains all the hours of drivers on the platform in the United States from September 2015 to April 2016, totaling 197.517 drivers, with 881.826.744 hourly observations.[3]The study focused on the UberX service, as it accounts for the majority of Uber trips. Specifically, the data consists of an anonymous driver identifier and a record of time spent actively using the system, driving time, city, and payments.
The authors divided time into discrete hours as the unit of observation, 168 hours per week. A worker is considered "active" if they are active (transporting a passenger or picking up a passenger) for at least 10 minutes within that hour. The salary is calculated as the driver's total salary for that hour, divided by the minutes worked, multiplied by 60. On the Uber platform, drivers are expected to pay for both the capital costs of their vehicle and all operating costs of the vehicle. Thus, these costs were incorporated into the driver's reserve salary.
- Assessment Method
The authors used a multivariate Probit model, with a latent regression for each time period. Furthermore, the labor supply model has two important points: (i) the censoring point varies according to the observation, as observed wages vary between observations; and (ii) the labor supply wage reservation model imposes an exact restriction on the coefficient that achieves identification – that is, the restriction of wages on the latent variable of the model is equal to -1.
Figure 2 shows a graph of the variations in expected wages and the logarithm of expected wages. It is possible to see that even within the city, week, and day of the week, there is a large variation, with a standard deviation of more than $3 per hour, which corresponds to a variation in wages of at least 10%.
Figure 2: Variation in Expected Wages.
A: Expected Salary. B: Log of Expected Salary.
- Main results
The authors' results suggest that Uber drivers do not have homogeneous preferences regarding the time of day and day of the week. Preferences for the rush hour From Monday to Friday, they are very heterogeneous.
It was also found that the drivers in the sample experience large changes in pay that are not consistent from week to week, and therefore they may place a high value on the flexible work arrangement. For example, by randomly selecting 100 drivers in Philadelphia who worked Monday through Thursday nights, the authors found that while most average city wages hover around $20, there is one week when the actual pay is very high.
The authors find that the average driver earns about $21,67 per hour, with a surplus of about $10 per hour, suggesting a reserve wage of $11,67 per hour. The reserve wage includes the cost of the driver's time as well as driving costs.
The ability to work split shifts and unconventional hours at Uber is valuable for drivers. Adapting to both positive and negative wage shocks is possible under the Uber-style work arrangement. A conventional work arrangement rarely allows workers to opt to work more if they unexpectedly run out of money.
In other words, the authors documented an important value analysis of flexible work arrangements: the ability to adapt working hours to reservation wages that vary over time.
- Lessons in Public Policy
The expectation is that technology will allow for the growth of more work modalities in the style of the Uber platform. Although such arrangements may have significant disadvantages compared to traditional careers with their labor rights, flexibility is an important source of value in such arrangements.
References
Bond, James T., and Ellen Galinsky. 2011. “Workplace Flexibility and Low-Wage Employees.” https://familiesandwork.org/downloads/WorkFlexandLowWageEmployees.pdf .
Campbell, Harry. 2018. “The Rideshare Guy 2018 Reader Survey.” https://docs.google.com/document/d/1g8pz00OnCb2mFj_97548nJAj4HfluExUEgVb45HwDrE/edit .
Council of Economic Advisors. 2010. “Work-Life Balance and the Economics of Workplace Flexibility.” https://digitalcommons.ilr.cornell.edu/key_workplace/714 .
Hall, Jonathan V., and Alan B. Krueger. 2016. “An Analysis of the Labor Market for Uber Driver-Partners in the United States.” Working Paper no. 22843 (November), NBER, Cambridge, MA.
[1] Council of Economic Advisors (2010).
[2] American Time Use Survey.
[3] These numbers were found through a series of cleanups of the original database.
