Hourly Pay Keeps the Multihomer at Bay

Last week, in an interesting move that is slated to reshape the gig economy, DoorDash announced that it is introducing a new compensation model for its delivery drivers. The popular food delivery service will now offer its drivers the option to receive an hourly minimum wage —a significant deviation from their traditional pay-per-delivery model. DoorDash’s proposed shift may be seen as an attempt to address ongoing criticism regarding the fairness of gig worker compensation. Furthermore, the new model could incentivize drivers to accept smaller orders that were previously disregarded due to their lower pay:

“Drivers will be able to choose whether they earn money for each order — usually a few dollars in base pay plus compensation for miles driven — or receive a flat hourly amount,”

DoorDash stated. The hourly rate, however, only accounts for “active time,” i.e., the period between accepting and dropping off an order, and excludes the waiting time in between orders. Moreover, the company confirmed that tips would be over and above the hourly base pay.

The company’s initiative is similar to Proposition 22, the 2020 California ballot measure which ensured gig workers a minimum wage and limited benefits while circumventing their classification as employees. However, DoorDash pointed out a critical distinction — its drivers are at liberty to switch between hourly and per-delivery pay as they see fit. Interestingly, this new system won’t be available in regions like California, Seattle, or New York, where laws governing minimum pay for drivers have been established.

This pivotal move from DoorDash stands at the intersection of gig economy regulations, labor rights, and innovative business models. It raises numerous questions about the future of gig work and the potential implications for other players in the industry. 

Most of us work with a fixed income, but this is not the norm in the gig economy. Instead, gig workers often earn their income on a per-task basis, with pay rates that fluctuate based on demand, time of day, and the specific platform they're working for. This dynamic pay model provides flexibility but also introduces income uncertainty, which can influence workers’ decisions to multi-home, a.k.a work for multiple platforms simultaneously.

What Drives Multihoming?

This change is very timely since I just finished working on a paper on this exact issue. 

The research paper, co-authored with Maxime Cohen from McGill, Ken Moon from Wharton, and Park Sinchaisri from Berkeley (who just presented it at the MSOM conference in Montreal), is titled Managing Multihoming Workers in the Gig Economy. Together we explore the dynamics of gig economy workers who work for multiple platforms (multihoming) and how they make the decisions of when to work and when to switch.

The paper discusses two major compensation schemes that gig platforms use: pay per work (used by most gig economy firms) and dynamic guaranteed wage rates (used by the firm we collaborated with). The pay-per-work scheme compensates workers only when they perform jobs on the platform, with the pay typically proportional to the duration and effort required to complete jobs. This policy allows the firm to respond quickly to real-time changes in demand and supply. However, it requires workers to constantly check for job availability, while leaving the platform unsure of whether it will be able to attract workers in time for surges in demand.

Under the dynamic guaranteed wage scheme, workers are guaranteed a fixed rate of hourly compensation during their shifts as long as they are actively available for work on the platform. This policy allows the firm to be more certain about its costs while attracting workers who value income certainty and relatively stable work. However, when demand is unexpectedly low, the supply of workers may not adjust despite low utilization. It’s important to note that this policy is slightly different from the one used by DoorDash, where drivers are paid only when they are truly on a delivery task, not while waiting for the delivery. 

The study aims to address several research questions:

1. How do gig economy workers decide among different work opportunities in the presence of multihoming, and how do they respond to the dynamic nature of task availability and incentives across platforms? 

2. Given an understanding of gig workers’ decisions, how can on-demand platforms compete for multihoming workers? What is the optimal incentive scheme that balances responding to demand uncertainty and maintaining control over the supply, how can platforms influence gig workers’ multihoming decisions, and what are the implications for policymakers?

In order to address these questions we use data from a focal firm and data from the Taxi and Limousine Commission of New York, and create a model that estimates how drivers behave at any point in time when they can make a decision (end of a ride, beginning of day):

We then estimate the different costs (driving, waiting, etc.) and the discount rate drivers use to weigh future income vs. current income —any decision drivers make has to do with such a tradeoff.  

Our study found that drivers heavily discount future earnings and respond more strongly to a short-term income shock. A substantial portion of drivers on the focal platform (which offered a fixed per-hour pay) exhibit multihoming behavior, with 66.59% of drivers multihoming more often than not.

Hourly vs. Trip Level Pay

After establishing a model of how drivers behave, we can now run all kinds of “what-if” analyses (i.e., "counterfactual" in academese).

The first result:

"We find that the pay-per-trip wage scheme incurs higher cost compared to the guaranteed pay in order to maintain the same level of service capacity. The pay-per-trip policy equivalent to the guaranteed pay (1×) shortens the number of hours worked and trips completed on average. Such effect is time-dependent. It would require the platform to pay 1.25 to 1.5 times greater than the rate equivalent to the guaranteed pay to mitigate the lower level of supply. This suggests the cost-benefit of the guaranteed wage policy."

In other words, our study concludes that increasing income certainty and compensating workers for continuously working can provide a cost-effective way to maintain a stable workforce. 

What drives this?

"The wage policy further impacts multihoming behavior. Our main result from the estimation of drivers’ discount factor suggests that drivers’ expected value of working is influenced by the timing of the earnings. Although on average the 1× pay-per-trip policy and the guaranteed wage policy would compensate the same driver the same amount, the actual income under the pay-per-trip policy is only received after the trip is completed. Compared to the steady stream of earnings under the guaranteed pay policy, the income earned through the pay-per-trip policy is discounted more heavily and therefore reduces the expected value of working for the focal platform. Subsequently, the expected value of switching to the competing platform appears relatively more appealing, further encouraging drivers to multihome.”

So two different aspects make the per-hour pay better at curbing multihoming: the preference for a steady stream of payments when discounting future payments, and the lower variability of that payment. You may think this discounting is insignificant since it concerns a few minutes or a few hours, but in a world where you make a decision every 20 minutes, the time scale of discounting becomes very fast. 

What are the savings? 

"We find that to retain the loyal drivers as effectively as under guaranteed pay, the pay-per-trip rate has [to] be between 2.5× and 3× of the current hourly offer. In other words, a guaranteed pay scheme could save the platform up to 75% of the cost to maintain the same level of loyal drivers pool. Similarly, to keep the same fraction of drivers who switch most of the time, the pay-per-trip rate has to be between 1.5× and 2× . Finally, to keep the same fraction of drivers who always switch, the scaling factor should be between 1× and 1.25.” 

So why are platforms afraid to use it? One reason is the risk of being tagged as employing these drivers. The second: it sounds riskier and can potentially backfire if the gig worker is not working all that fast. The pay-per-ride or pay-per-delivery should incentivize people to work faster and harder, but there's no evidence to support that in the ride-sharing world. But we must admit that the food delivery business is slightly different: in ride-hailing there's only so much one can do to finish the task faster. In food delivery, drivers have a little more discretion in terms of speed (e.g., finding the fastest route to deliver the goods faster).   

What about the drivers? 

Our study simulated driver earnings under different policies and found that compensating workers per trip at a rate that leads to drivers earning a comparable amount of income to what they would earn under the guaranteed wage policy resulted in slightly higher hourly earnings (30.16% more). However, when aggregated for each day, the pay-per-trip rate generally maintained the level of daily earnings on the platform. So overall, the compensation won't be much different. 

There are, of course, many other benefits to the pay-per-hour scheme. Shifting from task-specific payment models may give workers a sense of independence from market demand fluctuations and portray a sense of secure employment and steady income —a factor highly appreciated by gig economy participants, and particularly those whose main source of income depends on this type of work. Recent studies have indicated that ride-hailing drivers often stick to a regular work pattern and tend to remain loyal to the same platform after prolonged service, despite having the freedom to dictate their own schedules.

The Implications for DoorDash

DoorDash is now offering a version of guaranteed pay that is less forthcoming than the one offered by our collaborator, but I believe that the lower variability of pay will have a similar impact. DoorDash will be paying less with this new model, and will possibly manage to curb multihoming. 

The fact that they will allow Dashers to choose and switch between the two is interesting since it gives considerable power to drivers when deciding what’s best for themselves. But I have two reservations: one is that it will increase uncertainty regarding which option is the best, creating more anxiety for drivers, and the other is that it may raise even more questions on how DoorDash is discriminating among drivers by offering different options to different drivers (assuming this will be played out at the driver level), but it will be easier to compare. In other words, it may backfire due to poor execution. 

In conclusion, our research defines the circumstances under which income stability could potentially reduce the practice of multihoming and offer direction in formulating ideal compensation, incentives, and regulation for the gig economy.  Our study concluded that the dynamic guaranteed pay scheme outperforms the pay-per-work scheme. This is based on estimating the cost of work for individual drivers. The guaranteed pay scheme provides more income certainty to drivers and can help maintain a stable workforce.

Maybe there’s a reason most of us are not paid per task completion.