The gig economy offers tremendous flexibility for workers and companies, but it also comes with a host of unknown factors for both parties. A new study co-authored by Wharton professor of operations, information and decisions Gad Allon looks at how drivers for a ride-hailing firm make labor decisions – when to work, and for how long – and aims to improve predictions about labor supply and to shed light on more effective financial incentives. The paper, titled “The Impact of Behavioral and Economic Drivers on Gig Economy Workers,” is co-authored by New York University Stern School of Business professor Maxime Cohen and Wharton doctoral candidate Park Sinchaisri. Allon recently visited the Knowledge@Wharton radio show on SiriusXM to discuss the implications of the study and what the future of work could look like.
An edited transcript of the conversation follows.
Knowledge@Wharton: Why is this an important topic?
Gad Allon: There is a whole new economy out there that people call “the gig economy.” Some people refer to it as the sharing economy. It’s basically the idea that the old economy was one of freelancing. People had full-time jobs but did freelancing once in a while. There are more and more opportunities now through new platforms for people to engage in freelancer work at every type of scale. They can do it once a day, once a week. They can do it almost as a full-time job, continuously shifting between, let’s say, working for a few hours for Uber, then switching to TaskRabbit, then doing maybe a few hours for GrubHub or Postmates.
That economy is going to grow to around $2.7 trillion by 2025. But do we know how these employees actually behave? There is a lot more flexibility on the firm side. I think we always want to have a situation where the firm can match exactly supply and demand, and they indeed can do that. The only issue now is that the same workers can choose continuously where they want to work. The key question that we wanted to answer is to fundamentally understand how they make decisions, when the decisions now are [based on] seconds, minutes, hours or days rather than, “I need to decide where I’ll work next year.”
Knowledge@Wharton: Companies like having that flexibility because it allows them to have more workers working fewer hours, and it protects them from having to provide health insurance or higher rates of pay. That was the original motivation for a lot of these gig economy companies coming on board, correct?
Allon: Exactly. It allows Uber to be five times bigger than any car firm. People compare it to BMW without having a single car. You have no asset on your balance sheet, but you operate at a valuation of $120 billion.
Knowledge@Wharton: When you look at these two sides — workers and companies — which side has the greatest level of angst in this whole scenario?
Allon: I would say both sides. I think the stakes are just vastly different. If you look at a firm like Uber or Lyft, this is a situation where, because drivers can switch continuously between firms, there is a steep competition for what we call a land grab. You see Uber leaving China when they realize that DiDi, which is the main player in China, is going to outplay them. You see the same in Indonesia and Malaysia, and you see it in the U.S. In some cities, Lyft will be competitive, and some cities not so much. There is a lot of angst in that [situation] primarily because there is a lot of VC money poured into it.
“[The gig] economy is going to grow to around $2.7 trillion by 2025. But do we know how these employees actually behave?”
Now, let’s talk about the drivers. For many of them, that is their livelihood. They are trying to make the decision knowing that 15, 20 years forward, there will be self-driving cars. Many of these drivers know that driving is maybe going to end in 20 years, so there is a lot of uncertainty for them. I tend to speak with drivers often, asking them how long they have been working, what they have been doing. You will find many Uber drivers saying, “I’ve been spending three hours at the airport.”
There is a lot of uncertainty over earnings, because [drivers] don’t have security anymore. You will hear a lot about the impact on families — what is the impact when you don’t have the overall social safety net that you had before? I think there will be a lot of uncertainty in the next few years because of that.
Knowledge@Wharton: That brings up a question about what will these companies look like, especially if autonomous vehicles become as important as a lot of people believe they will be? These companies are already looking at partnerships with the Big Three automakers, so all these workers in the gig economy may or may not be working in these jobs down the road.
Allon: In my opinion, autonomous cars with no one at the wheel are probably many years out. We will need someone there, but the driver might be very different than what we have now, and the set of skills they will need will be vastly different. Imagine getting into a taxicab, and rather than having a driver whose main skill is the ability to get you to where you want to go, he is [instead] a concierge who knows the city well, who can advise you where to go, who can give you comments on how to prepare for your next meeting. The set of skills will change.
Knowledge@Wharton: In your research, you found that there are drivers who believe they want to get a certain level of compensation, hit that plateau, and then decide they are done for the day or the week or the month. Can you talk about that?
Allon: Let me talk a little bit about our method — we spoke with drivers to be able to develop the hypothesis, but ultimately [the study] was data driven. We followed some 8,000 drivers. We saw every decision they made for the last year. We saw how they made decisions when they were offered more money, when they were offered less money, when they [worked] in the morning, or in the afternoon. We saw interesting behavior. First of all, the more money you offer them, the more likely they are going to work, which might be surprising to some people but essentially is not that surprising.
But the interesting phenomena is exactly what you are referring to, which is we saw a very strong income-targeting effect. When drivers get closer to a certain income, we see a surprising outcome, which is that you pay them more and they are less likely to come and work. If they start working a shift, they are going to work fewer hours. This is something that was observed before, but we did manage to reconcile it here. We have seen it across drivers, across types, and in a very robust way.
Knowledge@Wharton: If you have guaranteed money, you are not willing to work as hard as maybe you would if you know that longer hours would get you more money. Is that correct?
Knowledge@Wharton High School
Allon: Right. We saw one more behavior that was interesting. Generally speaking, the more you pay people, the more they want to work and the longer they want to work. But we did see another effect, which was what we call inertia: The longer people worked, the more likely they are to continue to work and the longer they worked.
I will go back to the income effect. The reality is that many of these drivers are trying to balance leisure, they are trying to balance family life. It is true that in the past you paid them a lump sum, and they worked and gave their best effort. We know already that there is quite a bit of research showing that for gig economy workers, their family lives deteriorate very quickly. What we [find] is that there are internal mechanisms, so to speak, that are trying to prevent them from doing that. This notion of whether it is within a day, within a week or within a month — once [drivers] reach a certain [income] level, the strong financial incentives become weaker.
Knowledge@Wharton: What are some of the other things that drivers are considering besides time and money?
Allon: One thing that seems to be fairly consistent is that people want to be working at times when they can actually drive rather than stand in traffic. People want to be feeling that they are moving forward, their work is being rewarded. They hate waiting.
I want to go back to the inertia that we see, which is the fact that the more they work for a specific firm, the more likely they are going to continue to work for that firm. The way I explain that, at least for now, is [that they are] trying to find more consistent revenue streams. If you think about being a gig economy worker, it means that you are fairly entrepreneurial. You need to drive where you think you are going to find more work. We know people do what we call multi-homing, they switch between platforms. But at the same time, we see an attempt to try to streamline and create a more consistent vision of where things are going. So, even though there are many drivers and they behave in very different ways, we see that as very consistent behavior.
“We saw a very strong income-targeting effect.”
Knowledge@Wharton: In the research, you estimated that most of the workforce will be in the gig economy by 2025. What will that do to our overall economy?
Allon: These are not our projections, but those that people tend to agree on. As we see more and more automation, we are going to see more and more people moving to places where what we call the last mile is going to be important – [reaching] the customer. We are going to see fewer and fewer jobs that are going to be 9-to-5 or [located] at the plants. We are going to see more and more jobs that [will involve] working a shift, and then you have a shift to do whatever you want.
It was interesting for us to learn that a significant number of people who work for the gig economy did not start because of necessity; they started to supplement their income. They are doing one more shift, trying to pay for their vacation, and over time realizing that they actually enjoy it and enjoy the flexibility that comes with it. So, we are going to see more fragmentation moving forward.
Knowledge@Wharton: Your research was with a ride-sharing company, but is the decision-making the same for people working in other aspects of the gig economy?
Allon: Yes. Most of the people [in our study] work around 75% of their time for that specific ride-sharing firm, but the rest of the time they work for other ride-sharing firms or other delivery firms. That is a good question because we modeled how an employee is making a decision, and we believe that there was nothing in that modeling that was specific to that firm. Ikea bought TaskRabbit to help them build their furniture. Amazon is doing more work with the gig economy. If you buy furniture that needs assembly, Amazon can bring you someone to assemble that — and that is going to be a gig-economy worker, not an Amazon employee.
The same person might be driving in the morning for a ride-sharing firm, and in his lull time he may go out to assemble a table tennis table. We see a lot of that. Again, flexibility is great, but it has many challenges with it.
Knowledge@Wharton: How is this going to affect the bottom line for companies that are fully vested in the gig economy, like Uber and Lyft, or retailers that are tied to some component of the gig economy?
Allon: I think the main issue for all of these [firms] is that there is very tough competition for employees. The fact that employees can switch very quickly from one firm to another creates a challenge where you need to lure customers on one side, but then you need now to [attract] the employees. It is a double-sided competition. In fact, most of the competition now is not so much for the customer, it is much more for the employee. The reality is that the more employees you bring, the more drivers you have on the road, the faster your response time is between customers. The more customers you bring, the more employees you bring. It is a chicken-and-egg kind of problem, but we know that the main challenge is bringing more drivers or more employees.
Knowledge@Wharton: Gig economy companies benefit from the fact that workers are freelance. There may be times when there are more drivers in a particular city than you really need, or fewer drivers than you need. As you said, if drivers aren’t having a successful night with Uber, they may switch and earn their money through Lyft.
“Most of the competition now is not so much for the customer; it is much more for the employee.”
Allon: That is exactly one of the main challenges. It is not just having enough employees, it is having the right number of employees. If [firms] have too many drivers waiting and doing nothing, they actually penalize themselves — or at least that is what they feel they are doing. [Many] times they have to pay drivers through subsidies, and long term [that means] burning cash. At the same time, if they don’t have enough drivers, they lose money by virtue of having customers wait. They prefer having customers wait rather than having drivers wait. Drivers waiting, generally speaking, hurts their bottom line more.
Knowledge@Wharton: You have many more unknowns in the gig economy than you do in traditional business, correct?
Allon: Exactly. You have many more unknowns. Second, you have many more new business models continually thrown at these firms. For example, when Lime, the scooter firm in San Francisco, was introduced, Uber saw a dip in demand. Everybody is competing on the last mile. Every new business model is going to try to take and streamline that and try to take a cut from that. That is why there is a lot of uncertainty in these models.
Knowledge@Wharton: What incentives did you find really worked so that the company and the employees both felt good about the overall operation?
Allon: We tested one type of incentive — the financial one. One of the advantages of not only the gig economy but also the new era of having so much data on every driver is realizing which drivers are the ones you want to keep and which are the drivers that you prefer stay home for this specific shift. What we managed to show is that if you can identify the ones who are already near their income target, you have to offer a significant amount of money to be able to get them to drive. If you identify and categorize drivers properly, you can bring the right number of drivers with 30% less incentive, for example. Or if you use the same incentive, you can target them with much higher precision. But it really comes down to understanding what the financial incentives are for each and every driver.
Knowledge@Wharton: How much are these companies looking at every one of their drivers so that they can incentivize the good ones to keep working?
Allon: Continuously. These are very highly data-driven firms. Think about machine learning and what we call deep learning. I will give you an example. When we look at the 7,000 drivers [in this study], we had probably 1,000 different incentive schemes that were tried at different times using different mechanisms. These firms are very savvy in terms of their use of data. They analyze it well. They think about it well. What we showed, however, is that if you want to project outside of what you see, you need to have a fundamental model. But they continuously try to randomize and experiment in finding better incentives.
By the way, some of the incentives were if you work for an hour, if you work for two more hours, if you show four days in a week, you are going to get a raise. A lot of it was about creating more consistency for a single driver.
Knowledge@Wharton: Is that model of financial incentives accepted in the industry?
Allon: They are also bound a bit by the type of contractual agreement that they [have with] the drivers. The bigger risk they face is of regulation. Uber, for example, is trying to ensure that drivers are not considered to be employees. For that to happen, you have to make sure that there is a very simple agreement between you and them, which is usually financial and very transactional.
The agreement in the industry overall is that you can randomize, you can try different things, you can guarantee a little bit of a wage, but you cannot do anything much more creative than that. Anything more creative than that will bind you, in which case they become an employee and you have to provide for their health benefits and other things.
“If you buy furniture that needs assembly, Amazon may bring you someone to assemble that — and that will be a gig-economy worker, not an Amazon employee.”
Knowledge@Wharton: By taking this freelance approach, are companies able to manage their costs better using the data that they collect and analyze?
Allon: Right, but at the same time it is about the bottom line. To get a driver to drive on the road, Uber needs to ensure this driver gets $17 an hour. Through driving, they can get $14, which means that Uber is subsidizing each and every driver for $3 coming from VCs, which means it is coming essentially from our pension funds. So, it is flowing directly to the bottom line. True, they can predict that. Everything is much more predictable, but $3 is a lot of money.
Knowledge@Wharton: What’s next for this research?
Allon: There are two areas that we mentioned here. One of them is going back to this notion of inertia. We really want to understand what drives this inertia because it may have significant impact on other areas. We think that there is something unique about the firms that we work with that is creating that. And it is creating opportunities for firms to engage with their customers, with their drivers. We think there is a fundamental question of what drives that.
Second, we think that it may create opportunities for not just financial agreements but also mechanisms around reducing friction that may allow firms to create more interesting models of how to ensure that they have consistent service. One thing that is very fundamental here and that we have to be clear about is the customer. We are generally winning, right? We have better service than ever. If you want to get a cab from [one place to another] within a few minutes, someone is there. Overall, we are better off. The only question is, is that sustainable? To make it more sustainable, we need to find more novel models to keep these drivers driving, but also do it more responsibly.
Knowledge@Wharton: The consumer understands that the final mile is important. Certainly, consumers feel like they can pick and choose. If I want to go get a TV from Best Buy and they don’t give me the right service, I can take it back and go somewhere else.
Allon: And that is the challenge. In most of the new economy, whether it is Uber, Amazon, Facebook or Google, we get better service cheaper than ever. The questions are about the long-term implications, and that is what we are trying to understand here. What are the long-term implications of that on the welfare of employees?
Jeff Bezos says that the customer always takes the current service level as a given and is never content with whatever you offer them. So, if you currently can have a driver in five minutes, next year it is going to be less than a minute. After that, I want it before I even think about it. That is going to drive even more innovation, more cost to try to satisfy that and more competition in trying to find interesting models around it.