Studying Gig Economy Workers’ Decisions – Wharton Prof. Gad Allon at Global Forum London

Studying Gig Economy Workers’ Decisions – Wharton Prof. Gad Allon at Global Forum London


>>There’s a whole new economy out there called
the gig economy, driven primarily by the fact that all of us have a mobile device, by
excess amount of data that all firms have, and driven primarily by new platforms that allow
for a perfect match between supply and demand. And when I say supply, it means supply
for everything from point A to point B, food – we saw today, Deliveroo. If you want to develop a
website, you can do it on Upwork. If you want someone to do handyman
work, you can get that as well. So, on one side, we see that
kind of notion of opportunity with having these very asset-like type
of platforms that enable perfect match, but at the same time, there are
significant challenges with these platforms. The main challenge is that while on one
side you need to compete on the customer, you also need to compete on these
the same gig economy workers because the same workers are working
for Uber and Via and handymen. What we are going to try to do in this work
primarily is really understand how this gig economy worker Whether it’s a driver, a handyman, or
bike rider working for Deliveroo, how do they make labor decisions. How do they make decisions
that both motivates them? What are the main levers that make them work. Getting a little bit numbers, so I’ll throw a
few numbers at you, in the interest of time, 44 million people in the U.S. took some work. Thirty-four percent of the
overall economy tried that. Numbers from today from The Guardian,
9.6 percent of the labor in the UK did at least once a week gig
economy work over the last year. We are talking about by boosting the economy
by 2.7 trillion by 2027, and with the emergence of more and more automation, we expect to
see that work that as we see it now is going to be more fragmented, and more and more
jobs are going to look like gig economy work. So, with that probably, you ask yourself,
how is that different from freelancing? Well, not all that different. Gig economy is not new. People always have worked gigs, but today we
have, when most people refer to gig economy, they specifically talk about new
technology-enabled kind of work. What technology did here is it reduced
transaction costs, reduced friction. If there is someone willing
to do the work around you, it’s very easy to match between
that and that person. So, who are these people? Seventy percent are doing that by choice, and
what I mean by choice, they start initially just as an attempt to try to see what it is, for 44
percent it became their primary income. Fifty percent of them are millennials. What’s really the main difference? If many of us, when we make a labor decision,
we make it once a year, maybe twice a year, maybe once in seven years, for example. Here, we talk about people that make decisions
by the minutes, and when they make decisions by the minute, they make
decisions on where to work, how long to work, and also which firm to work. Many of them, and we’re going to
talk about later about later about multi-homing, many of them are switching
between multiple firms. You problem took an Uber drive
where the same driver was driving for Uber and for Lyft, for example. So, while at the same time these
drivers have many more opportunities, the level of complexity increases as well. When I say complexity, firms have the ability to
continuously tailor the wages for each driver, to tailor the incentive they offer
for each driver. So, for example, in our data,
we had 8000 drivers. We had a thousand different
wages in every point in time. So, the complexity is increasing
significantly, and with that, also the challenges we are going to address. We are going to try to address
the very simple question. How do gig economy workers make labor decisions? Do we need a new type of labor economy? How can the platform influence their decision. Let me go first and try to talk
about what are they doing now? So, what they do now is search pricing,
basically moving their prices up and down, as a way to incentivize more
employees, pre-announced bonuses. So, they announced the bonuses in advance
saying during this time that’s what you’re going to get. We have an article from the New York Times
saying how Uber uses psychological tricks. And so let me stop here for a second and say I think that’s a little
bit like a cheap shot at Uber. That was a time where Uber was down,
and it was easy to take a shot at them. We all use psychological tricks. All of you, we are at alumni events, we know that all of you were students, you all remember the psychological
trick we played on you called grades. [laughter] and so they also pay
psychological tricks on them. Anyway, continuing, giving
more data is another thing. There are also significant
implications for policy. As we know, all of you are here in London very
close to the EU or not far yet from the EU, we understand that, I mean there are
new regulations about what are the type of relationship we want to
have with these workers? Are they, are there any requirements
for minimum wage? Are they allowed to multi-home
and things like that. We won’t be able to answer these questions well. So, let’s talk about our study. Our study was done with a firm called Via. Via in London, you see them as ViaVan. We did a project primarily in New York. Via is different than Uber in the following way. Uber will pick you from where you are and
will take you to where you want to go. Via will ask you to walk two
blocks and pick you up from there and will drop you two blocks
from where you want to go. It’s the Israeli version of Uber. As an Israeli, I can say that. Anyway, you have, what we have is data 358
days, 5.5 million observations for each and every driver, for each
and every point in time, we know exactly what wage was offered to them. We know exactly what they did. We know when they accepted and when they decide
not to accept, and we take all this information, and we ask our ourself, why shouldn’t
I just run it [inaudible] system. Why shouldn’t I just reuse data. So, let me show you the first challenge. The first challenge is the
following, once it’s going to click. The first challenge is the following. I’m going to draw on the x axis
the average offer that was made, and on the y axis, the likelihood
they accepted it. How do you expect it to be? You expect more people as they pay
you more, what are you expecting? You’re going to expect people
to accept more jobs, right. Only that it reverses later on. Why? Is it the case that drivers don’t want to
work and get paid more, or the is it case may be that the decision to offer
them more was already knowing that they’re actually not going
to accept it – what we call simultaneity. So, what we have to think
about is how to solve that. I’m not going to go into a statistics class now,
but believe me, we are trying to solving that. We have issues around, I’m not going to talk
about that, we have issues around the fact that we see only those that drove, and we
cannot really treat them as a random sample. Not going to talk about that, but I’m
going to mention we are addressing that. All of these slides were just to say
everything we are doing is kosher. So, if you ask yourself why we are doing, why
it’s complicated, it’s exactly because of that. We’re going to look at two decisions. One decision, shall I work or not. Then the decision how long to work. Because that’s really the sequence
by which they make decisions. So, remember that, and we’re going
to look at really three main factors. What’s the impact of wage? As they offer you more, are you working more? Second one, as you earn more until now,
are you going to work more or less. As you worked more hours until now,
are you going to work more or less? So, our wage, hours so far, and income so far. That’s going to be. Wage is rational. You want people to work more as they earn more. Hours so far and income so
far and not so rational. So, let’s see what we get. Drum roll, let’s talk about the results. So, the first result is the following. Work or not. What you see here, green means positive impact. Yellow means negative impact. No color means not significant. The first thing you see is the more we
pay people, the more they pay people, the more likely they work,
the more hours they work– sorry, the more income they had until now,
the less likely they are going to work. The more hours they worked until now, the
more likely they’re going to continue to work. Maybe that’s just a coincidence. Let’s [inaudible] to see more. In fact, we see even the same where it
comes to the number of hours worked. So, the more they pay them, the more likely they
continue to work, the more hours they worked. Unexpected, but really the surprising part
is the last one, which we call inertia. It’s the notion that the more hours they worked
until now, the more likely they continued to work, and the more hours they worked. And you don’t see as shocked as you
should, but it has significant implications for what it means for the
amount of hours they worked. As we see, in more and more cities, New York,
London as well, getting into that trying to reduce the amount of these employees working, what we see here is significant
inertia in their behavior. We see that’s the same, for any shift
we do, we see the same behavior. I will not get into that. We see similar behavior when it comes to across
days, and so overall, what they are going to take from this talk is three things. The interesting thing about
these employees is that while on one side they’re looking
for more flexibility. Most of them are working gig economy
work because they want more flexibility. So, we see on one side, yes,
they are very rational. They are reacting to incentives
the way you expect them to. But the interesting thing is, they
have significant income targeting, which means the more you pay them,
the more they’re [inaudible], the less likely they’re going to continue. But also significant inertia. Significant inertia, which means
the more hours they work for you, the more likely they’re going to continue. It has implications for how
you personalize payments. So, for example, one experiment we do with Via now is how do you actually convince
drivers not to work rather than to work. How do you convince them to go off the
road when they should go off the road. We can talk about also the implication for
firms that are doing, allowing multi-homing. As we know, one of the main issues
of many of these firms is the fact that customers can actually,
which between these, which has significant implication
for Uber’s share price. Uber’s share price is under the belief
that over time they can actually reap more of that market we don’t actually see that. So, what do you want to take? Significant inertia, these are rational people
but have significant income targeting effects, so overall much more complex than
just running one more dataset. Any question afterwards,
I’m happy to take offline. Thank you.

Author:

Leave a Reply

Your email address will not be published. Required fields are marked *