The Uber Pay Gap

Using data on over one million Uber drivers and millions of trips, Cody Cook, Rebecca Diamond, Jonathan Hall, John A. List, and Paul Oyer show that female Uber drivers earn 7% less than male drivers. What makes this paper new, however, is that UBER’s extensive data lets the authors understand in great detail why the pay gap exists. It’s not discrimination:

Uber uses a gender-blind algorithm and drivers earn according to a transparent formula based on the time and distance of trips. There are no negotiated pay rates or convex returns to long hours worked, factors that have been shown to open a gender earnings gap in other settings. Our research also finds that both average rider ratings of drivers and cancellation rates are roughly equivalent between genders and we find no evidence that outright discrimination, either by the app or by riders, is driving the gender earnings gap.

The authors find that three factors explain the gap; driving speed, experience, and choices about where to drive.

First, driving speed alone can explain nearly half of the gender pay gap. Second, over a third of the gap
can be explained by returns to experience, a factor which is often almost impossible to evaluate
in other contexts that lack high frequency data on pay, labor supply, and output. The remaining
20% of the gender pay gap can be explained by choices over where to drive.

Male Uber drivers, like other males, drive a bit faster than female drivers, about 2.2% faster after controlling for experience and location. Since Uber pays by time as well as by distance the returns to speed are not very high and the difference in speed is small but overall this results in an increase in pay for males of about 50 cents an hour.

Drivers learn by doing and more men than women have driven for Uber for years:

A driver with more than 2,500 lifetime trips completed earns 14% more per hour than a driver who
has completed fewer than 100 trips in her time on the platform, in part because she learn where
to drive, when to drive, and how to strategically cancel and accept trips. Male drivers accumulate
more experience than women by driving more each week and being less likely to stop driving with

Overall, female and male Uber drivers behave remarkably similarly but small differences aggregated over large samples produce a small but systematic gender gap in wages of about 7%. The gap, however, is an artifact, a social construct that has no implications for “social justice,” drivers are treated equally.

The author’s conclude:

Overall, our results suggest that, even in the gender-blind, transactional, flexible environment
of the gig economy, gender-based preferences (especially the value of time not spent at paid work
and, for drivers, preferences for driving speed) can open gender earnings gaps. The preference
differences that contribute to pay differences in professional markets for lawyers and MBA’s also
lead to earnings gaps for drivers on Uber, suggesting they are pervasive across the skill distribution
and whether in the traditional or gig workplace.


I visit Charleston often, and I use Uber when I'm there. It's a college town, but I've yet to be driven by a college student. I don't ask age, but the drivers all seem to be in their thirties. And here's something interesting: none has been from Charleston. That's right, all of my drivers have relocated from some place else, Michigan, Pennsylvania, New York. Charleston is a magnet for people choosing to re-locate, and they seem to be highly motivated, or motivated enough to spend part of their time as Uber drivers. As for gender, I've had more male drivers than female drivers, but I have discerned no difference in speed (either in getting to the pick-up location or the drop-off location). I have discerned some differences in the drivers' familiarity with the area. Some follow the directions on their console even if it's not the shortest route. The fee isn't affected so I don't act like a back-seat driver - I do when I'm riding a cab because the fee is affected.

So I guess I don't understand. I have heard of Uber and assume it is like a taxi. But when I travel I rent a car. Why would I use Uber? A serious question. I am aware there are Uber apps but I scrupulously avoid all apps on my phone and computer so I know nothing about them. But again why Uber? When I visit NY City use public transport or walk. In fact I prefer walking and only use public transport to get to an area where I can walk that I haven't walked before.

Uber is an on-demand taxi service, so it's better for exactly the kind of trips that are advantageous to take by Taxi.

So how does it work? When I call a cab there is a number in the phone book and he shows up and takes me where I need to go. Is there a "number" to call for Uber and are they as dependable as a cab? Is the number the same in New York and LA or do I need 1000 different numbers to call a Uber? Can I find them in the white/yellow pages?

You enter the address you want to go to on your phone, press a button to request a driver, and then within about 15 seconds it matches you with a driver, displays a map of where he is and how he'll get to you, and when he gets there you hop in. Payment is charged automatically at the end of the ride (you put your credit card on file with the app).

Do you know of Google?

OK so it's through a phone app only, right.

Do you know of Google?

The guy in 2018 is asking wtf Uber is. He's obviously either trolling or willfully *extremely* ignorant

Careless: Indeed I am willfully ignorant. I see movie stars and singers in the news or on TV and ask my wife "who is that, have you ever heard of them?" Many of us out here in the world do not follow everything people do. I don't care about social things, never have. I have never watched the oscars or any of the other self congratulatory shows that Hollywood puts on. I don't belong to any social sites and I'm willfully ignorant of what they are and I'm happy about that. Does it mean I don't know the latest songs or high tech do-dads? Yes and it's great. I spend most of my time, now that I'm retired, enjoying life and avoiding the self inflicted drama that the young seem to embrace.. I have recently acquired a smart phone but my single purpose in getting it is so I can "tether" my laptop when I'm traveling. My phone is usually off and stays off for weeks at a time and I like that too. I laugh at the people I see walking along playing on their phones, they miss so much in the world. Just six months ago I was hiking down into the Grand Canyon with it's beautiful views and animal life and another "hiker" was talking on his cellphone (loudly, irritatingly). My thought was what a shame. Did he even see that Desert bighorn (a rare sight in the grand canyon)?

But yes I am sincere in my questions and now understand a little more about Uber and have concluded it is likely of no value to me. Thanks to those who responded..

So I guess I don’t understand. Do you mean to say you carry your phone around with you? Mine is attached to the wall in my home, and when I want to make a call I turn the crank and shout at the operator. I have heard there are new automatic devices that allow one to use a “dial” to place calls manually, but I prefer to have the operator connect mine.

Yes John. Your hilariously funny comment is right on the mark. I am a neanderthal and don't use facebook or most other "in" stuff. I have had my smart phone for just a few months. I changed from my trusty flip phone only because I wanted the ability to "tether" my laptop to it when I am traveling and out of range of wi-fi. However it has kinda backfired in that the smart phone is a pain in the ass to carry around so I often simply leave it at home. Just got back from a week in Florida with my smart phone staying home and it was no problem. I guess since I don't use Uber I don't really need it.

Ironically I am thinking about getting a cheap flip phone too so that I can easily carry a phone with me. I don't really need a phone and I usually keep it turned off but I expect one day I will want to use it to call 911 and when that day comes it would be nice to have a phone with me. Isn't that great! The smart phone is awesome except it is too bulky to carry around so I need a second phone to make up for the smart phones fatal flaw...

New technologies can not only save you the stress of driving, but can also save your household money.

We recently purchased a GSM cellular home security system to replace an older land-line system and save about $55.00 per month.
The cellular service is just $3.00 per month.
The cancelled land-line and long distance carrier services cost $40.00/month and $18.00/month apiece.

I'm surprised you make it back up north this time of year, sort of figured you are a snow bird.

Part time snow bird. But travel year round.

DevOps Dad:
I enjoy driving and do it eagerly. At one time I used to drive for 36 hours straight before sleeping but I'm much older now and can't so that anymore. Twelve hours a day is now my max. Very little stress, we live in the West and I know the back roads and I can often drive for an hour and not see or pass another car. Of course we do also have to drive through big cities too but I don't find that stressful although my wife sometimes does. I get a kick out of watching the fools drive like nuts on the freeway but don't let them bother me.

My home security system is .357. It is extremely loud and makes large holes in anything in front of it. My neighbor is a cop, I have recorded video camera outside and I have nothing much worth stealing except my meager 55" TV ($395 at Walmart). I use the same security system in my motor home or when hiking in the wilderness. It isn't perfect but it is a lot of fun.

This is nonsense. Are you objecting to shortening of the shortening of "application" to "app"? Is there a distinction you are drawing between proper applications and inane apps? Is it a preinstalled vs third party distinction you're making? Or did you manually write out the HTTP request that submitted this comment for you? Maybe you don't have a smartphone, but you absolutely do use apps on your computer, and if you do have a smartphone, you use apps on there too.

Personally, I use Lyft (a competitor to Uber) when I go out to bars and don't want to worry about having a designated driver.

I do understand that essentially all "programming" is now called "apps" and that there are apps on my phone and computer. What I said is I avoid them in I do not download them. Mostly because I am not aware of any that I need and I understand that they are often infected with nefarious coding which will cause me problems.

However you did indeed hit on a use for Uber; i.e. when you drink. However I don't drink. I did consume half a glass of beer in 2010 when I was in Australia so I could try it. Before that perhaps 10-20 years since I drank anything. My big addiction is chocolate and luckily I don't need a designated driver for that.

I did take a cab two years ago but it's pretty easy to get a cab. So if it is easy to get a cab and the other option is Uber I repeat why Uber?

You’ve got to set aside your own (lack of) personal experience here. Uber is in general quicker, cheaper, more reliable, and has a more straightforward pricing scheme than cabs. For travel, if you’re flying to an urban area and staying mostly in that area, it’s often cheaper than a multi-day car rental and avoids parking issues, and as not every American city has New York’s extensive transit system, it beats those options as well.

OneGuy, for you there is no reason to choose Uber, and without the Uber app on your phone there is no way you CAN choose Uber. My own personal experience with Uber has been that it is superior in every way to cabs, buses and every other form of urban transportation, and as Uber, Lyft and similar services have been wildly successful, I must not be alone.

If you are asking why you, personally should use Uber, there is no good answer - it's a personal choice. I can only suggest that you try it and find out for yourself. If you are asking why ANYONE would choose Uber, and maybe implying that Uber is an unnecessary additional choice in the transportation market, I can only say that you have missed the boat.

Clay: I usually put hundreds of miles on my rental car in a week and sometimes drive most of the day. I am open to hear how it is better but be aware I used a cab about 2 years ago and before that perhaps 8 years ago.

Ron H : My question(s) are sincere. I know little to nothing about Uber. As far as I can tell from what has been said here it is more or less like a cab possibly cheaper and possibly faster. I have in the past just walked up to the cab lines at an airport and gotten in the first cab I'm not sure what would be faster.

"So if it is easy to get a cab and the other option is Uber I repeat why Uber?"

The reason to use Uber over a cab is that taxi companies have managed to create a cartel, which limits supply and competition. Taxi drivers typically need a medallion to drive, and the number of medallions is limited. Uber (and Lyft) allow non-medallioned drivers to drive and the increased supply of drivers and competition leads to the lower prices, greater efficiency, and greater reliability that others have mentioned.

Think of Uber and Lyft as market transportation and taxis as cartel/monopoly transportation.

But when I travel I rent a car. Why would I use Uber? A serious question.


I was fortunate enough to attend a small conference where Uber's chief mathematician was one of the keynote speakers. He related the following anecdote about the genesis of Uber:

"We did some research and found that a very high percentage of requests for NYC taxi pickups are never actually completed because the taxi never shows up.

The problem is that fixed taxi rates causes massive swings in supply, and many taxis will just go off duty if it isn't worth their time. Uber made a conscious decision to reverse that equation: allow for a fluctuation in rates, so there is a always enough supply to meet demand for taxi rides."

I very much don't like Uber because of all of the horrible things that it appears they are guilty of, but that nugget of anecdata is still interesting and admirable nonetheless.

Most of my Uber rides originate in the suburbs. Perhaps because of this, most of my Uber drivers tend to be female. Based probably on a suggestion by Tyler, I make an effort to talk to the driver about something other than driving for Uber! I hadn't thought much about it before, when the subject does come up, it seems as if the male drivers have been driving longer than the female drivers.

Discrimination comes into play when female drivers are harassed. This may lead a percentage of them to stop driving, meaning female drivers are less experienced overall resulting in the pay disparity.

As a man, i'm much more likely to sharply criticize a male driver if I don't like their driving, etc. Female drivers i give a free pass.

At last, the discriminatory explanation. The answer, of course, is to employ - or rather not employ, oh dearie me no - more trannies.

The study specifically addresses this. It isn't a factor.

I took a cursory glance at the "Returns to Experience" section. How does the study suggests that Smash's concern is not a factor. They show that women have less experience overall, and then they show that a large chunk of the pay gap goes away when you control for experience. When they look at selection effects, they only include driver fixed effects to examine whether this changes the returns to experience (e.g. if those who drop out were low earners to begin with), but they do not examine if this changes the indicator for male vs. female. Using driver fixed effects does not address Smash's concern. To address Smash's concern you would want to collect data on the distribution of the time and location (i.e. is this a bar where the driver is picking up drunk dudes) of pick-ups and drop-offs and pair that with data on passengers (e.g. did this driver pick up a lot of drunk dudes). Use this to make a variable that could be interpreted situations in which the driver is likely to be harassed (it would be better to survey drivers on their experiences with the passengers (seems relatively easy and would be a more direct measure of being harassed). You would then use that data as independent variables in a regression (probably a logit) to measure the probability of that driver dropping out. You could tackle the issue a few other ways, but I didn't see anywhere in the paper that would even come close to modeling differential drop out rates of men and women, and even if they did, they don't seem to have the data to provide the reasons for any potential differential drop out.

You are suggesting women may choose to quit driving for Uber rather than become selective of their fares? That could be. If so, it didnt show up in ratings. Page 8.

I'm not sure why the customers' ratings of the driver would have much to add to the question of quitting or becoming selective of fares. I wish the authors gave summary data on gender means for rejecting a dispatch (cancelling trips was about equal); this would (in part) get at the issue of selectivity (though that would again be affected by the women who choose to stay, when they choose to work, and what locations they choose).

Cancelling trips was equal, ratings were equal, and gains from experience were equal. I can understand the position that more information is needed to understand if harassment is a factor but there's a lot of proxy data there suggesting it's not a sizeable factor.

I'm still not sure that customer ratings of the driver actually acts as a proxy for harassment. They also don't say that the returns to experience are the say across genders. You would have to include an interaction of gender and experience level to see whether there is a difference in returns to experience. Including only the gender dummy only allows you to compare returns to experience within gender. Again, cancellation doesn't necessarily address the issue of harassment either (you would need to control for the characteristics of the riders, location, time of day) and you would also want to take a look at the rejection rate (again controlling for the aforementioned characteristics).

I'm not sure if harassment leading to dropping out/less experience is a big deal, but they just don't present any analysis that addresses this.

My first thought is that a lot of Uber driving - especially surge driving - involves picking up random drunkards late at night. It would be sensible for women to avoid those time periods and only drive during the less lucrative daytime hours.

Discrimination comes into play when female drivers are harassed.
So what you're saying is that women discriminate more than men do against men who THEY think are jerks? Because that's the flip side of this. Compared to women, men shrug "harassment" off.

It could just be women are harassed more. Maybe it's not, but your conclusion doesn't follow from Smash's comment (or at least, there are several hidden assumptions in your analysis)...

I thought about leaving it there, but your comment is so obtuse as to lead me to be amazed that someone of such a low IQ could actually use a keyboard. Your analysis:
Woman: "Get your hand off my ass. I will no longer let you in my car."
You: "Oh, so you're going to say I'm a jerk because you have a subjective valuation that says me putting my hand on your ass is a bad thing. That's just discrimination."

Your thesis should be easy to test. Uber drivers rate their passengers in the same way passengers rate their drivers. If women are being harassed at a higher rate than men, it should show up in the stats.

The article does not seem to address that question. I think that would be an important data point.

'The gap, however, is an artifact, a social construct that has no implications for “social justice,” drivers are treated equally.'

Really? Considering just how good Uber is at programming and keeping track of multiple variables (think Greyball - ), the fact that speeding drivers are not penalized would actually seem to be a bit of problem.

As this company noticed, though in another context - 'In 1993, in reaction to multi-million dollar settlements arising from car accidents involving its delivery drivers, Domino’s ended its delivery guarantee. While the pizza maker never admitted its drivers drove unsafely in their efforts to beat the 30-minute deadline, at a 1993 news conference Domino’s owner Thomas S. Monaghan said the guarantee was being dropped in an attempt to fight a “public perception of reckless driving and irresponsibility.”

That perception was fueled by a couple of key lawsuits against Domino’s. In 1992, the company agreed to pay $2.8 million to the family of an Indiana woman killed by one of its delivery people. The woman, 41-year-old Susan Noonan Wauchop of Calumet City, Illinois, died when a Domino’s truck struck her van near the Indiana-Michigan border in 1990, an accident that also injured three of her sons and a friend. Domino’s has always maintained that road and weather conditions, not the delivery time requirements, were the primary factors behind that crash.

Yet it was a case ruled upon in 1993 that rang down the curtain on the “30 minutes or it’s free” guarantee. An injury to a 49-year-old St. Louis, Missouri, woman, whose car was struck by a Domino’s

driver in 1989, led to an award of $750,000 in actual damages and $78 million in punitive damages against the pizza company. Jean Kinder, the injured driver, had suffered injuries to her head and back after the driver of the pizza delivery vehicle ran a red light and struck her vehicle.'

Doesn't the driver rating cover this? The ideal driver is likely one that drives as quickly as you are comfortable with, without losing time/money for speeding tickets.

I wouldn't want my driver doing 55 on a 55mph highway and I doubt many others would either. Nor would I want them doing 85. Somewhere between there is ideal and would be reflected in ratings/earnings.

'I wouldn’t want my driver doing 55 on a 55mph highway and I doubt many others would either.'

It is an interesting question - but would you care about your driver doing 30 in a 25mph school zone?

The even more amusing thing is that when autonomous vehicles appear as taxis, they will very likely do a precise 55 on a 55mph highway.

2.2% faster on average means that if the average woman drives 55mph in a 55 zone, the average man is driving 56mph in a 55zone. Not a significant difference on a short trip, but evidently it adds up for the drivers over the course of the day. Personally, I would prefer that the driver stay with the flow of traffic regardless of the posted speed limit. I suspect that autonomous vehicles will do the same.

Speed limits with autonomous cars will be much faster.

What about when there's a mix of autonomous and human-driven?

'I suspect that autonomous vehicles will do the same.'

And I suspect the very first time an autonomous vehicle is involved in a fatal accident where the autonomous vehicle is breaking any traffic law, the manufacturer will be considered liable. And the manufacturer's insurer will happily agree that the manufacturer violated the terms of its insurance policy in such a case.

'Speed limits with autonomous cars will be much faster.'

There is reason to doubt this, but then, I drive on the autobahn basically every day.

Plus, every German driver is taught some simple math which they must use on their driving test - essentially, it has to do with braking distance and speed. The laws of physics will not be rewritten by the introduction of autonomous vehicles - the reason for the 25pmh school zone speed limit has to do with braking distance - 30 feet (this is separate from total stopping distance). An autonomous vehicle doing 60mph in such a zone will require 172 feet to brake.


Physics is mostly the same (rear facing passengers allows for some fun redesign that could shift a lot of weight and alter stopping distance), but reaction time is vastly faster. Something like a quarter of your stopping distance exists for the driver to observe danger, send impulses up and down nerves, and then to hit the break. Likewise, with full 360 degree awareness, self-driving may be able to take much better advantage of lateral space (something humans should NOT be considering in rapid driving calculations) and allow for smaller reaction windows.

'but reaction time is vastly faster'

Well, you did see that I used total braking distance, not total stopping distance, right, thus ignoring reaction time completely? And oddly enough, that is what German drivers are taught to calculate when passing their driving test - reaction time is not considered relevant to the physics of total braking distance.

The numbers provided in my example do not use reaction time as a variable at all.

I'd want the driver to drive at a safe speed. What that speed actually is obviously varies by conditions.

So, yeah, I'd be quite harsh on speeding during school time in a school zone.

I might well accept autonomous vehicles that drive "too slow" sometimes. A 10% increase in journey time would often be tolerable in exchange for the convenience.

The difference in average speed was rather small (page 16):

Speed (mph)
Men: 19.532
Women: 18.760

The leap to "so men break laws" is VERY hard to make from a .8 mph average speed difference, that is (AFAIK as an Aussie) well below the speed limit.

"the fact that speeding drivers are not penalized would actually seem to be a bit of problem."

The paper doesn't say that male drivers are breaking the speed limit. It says they are driving faster. Specifically:

Table 4

Average speed in mph

Men = 19.53

Women = 18.76

Well, there is a commenter above who seems to feel that a rider has the implicit right to ask a driver to speed, so different perspectives are possible in terms of evaluating the data.

Which, admittedly, does not say anything explicit about Uber drivers breaking any speed limits. (And to be cynical - that is the sort of data that a company capable of creating Greyball would likely be interested in not sharing.)

It probably wouldn't physically cause you harm to just admit you are wrong, when someone points it out.

Instead of reflexively trying to deflect by maligning first another poster and then Uber.

'Instead of reflexively trying to deflect by maligning first another poster and then Uber.'

What was maligning? Here is the source quote, after all - 'I wouldn’t want my driver doing 55 on a 55mph highway and I doubt many others would either.' It seems to be accurately summed up with what I wrote - 'Well, there is a commenter above who seems to feel that a rider has the implicit right to ask a driver to speed.'

As for maligning Uber, well, the company does a much better job maligning itself than I ever could, assuming you pay any attention to their actions.

And this was an insufficient admission for you? - 'Which, admittedly, does not say anything explicit about Uber drivers breaking any speed limits.'

Fine - no one, neither you nor I, actually currently have any data concerning speeding on the part of the part of Uber drivers, neither male nor female. Making it impossible with the current available data set to make any empirical statements concerning speeding Uber drivers.

Mea culpa.

Yet it was a case ruled upon in 1993 that rang down the curtain on the “30 minutes or it’s free” guarantee. An injury to a 49-year-old St. Louis, Missouri, woman, whose car was struck by a Domino’s

Former Domino's Driver here. "30 minutes or free" ended in 1986. It was replaced with "30 minutes or $3 off."

If you say so. The Snopes article says this - '... at a 1993 news conference Domino’s owner Thomas S. Monaghan said the guarantee was being dropped in an attempt to fight a “public perception of reckless driving and irresponsibility.”'

"30 minutes or free" ended at different times for different franchisees. IIRC, WOP switched to "...or $3 off" after centralizing phone ordering, because it removed the limit imposed by having only 4-6 phone lines per store.

Before central ordering, if a high school football team had a huge game, only so many people could order from the local Dominos at once ("thank you for calling Dominos can you hold please"), and the load was spread out over multiple pizza chains. After, a phone team large enough to handle the entire region could take orders for that one store, and there was no feedback from the store to the operators to indicate that they were utterly slammed and couldn't handle more orders.

I had stopped working for WOP well before the cutover, but one night when I wandered into my old store, I ended up working the ovens for two hours because they needed everyone either on the make line or on the road.


Interesting research which appears to be competently done. BUT: what a leap of faith and/or reasoning to portray it as a finding relevant to other professions. What is the basis for that conclusion, I mean assumption? Because we can find reasonable factors that explain a 7% gap in wages for Uber drivers, we can infer that there is a 7% wage gap in other professions that is not the result of discrimination? I can think of many reasons that this is faulty reasoning. And, I don't see that the Uber data in any way supports such a conclusion. Couldn't the authors just publish what they found and not speculate beyond what their data shows?

No one made such an inference.

But if one were to make such an inference, the lesson to be learned is "Not all observed gender pay disparities are the result of discrimination."

This is very important because feminists, and especially politicians, treat pay gaps as prima facie evidence of widespread and pervasive discrimination. Studies like this demonstrate that women either need to step up their game (or step on the gas) or realize that their own choices explain their unsatisfactory outcomes.

Who would have guessed that working smarter, harder and longer positively affects pay?

'realize that their own choices explain their unsatisfactory outcomes.' - - > `realize that their own choices explain ***part*** their unsatisfactory outcomes.'

Fixed that for you. It is just a silly to assume 100% of outcomes are controlled by the person as to assume 0% are.

He is talking about women as a group. The default hypothesis is exactly 0%, e.g. "discrimination could favor men on net, or it could favor women on net".

"No one made such an inference." From the conclusion of the paper:
"Overall, our results suggest that, even in the gender-blind, transactional, flexible environment
of the gig economy, gender-based preferences (especially the value of time not spent at paid work
and, for drivers, preferences for driving speed) can open gender earnings gaps. The preference
differences that contribute to pay differences in professional markets for lawyers and MBA’s also
lead to earnings gaps for drivers on Uber, suggesting they are pervasive across the skill distribution
and whether in the traditional or gig workplace."
But let's not quibble about how supportable or not the conclusion is based upon the paper. I find much of the discussion here (particularly the comments that follow) quite tiresome. So much reaction to the fact that many people see discrimination where there may not be any. So many people quick to convict people with different views as being anti-science or anti-facts or anti-research. This is just as bad as the many people out there who might accuse the authors of being anti-women or whatever. What we have is one study with interesting results. From a Bayesian perspective, we don't have much but it is more than nothing. Some people appear to be willing to skip the steps of replication or studying any other labor markets, already sure that the answer will be the same as found in this paper. Uber labor markets may differ in important ways than MBAs and lawyers. Until that is carefully analyzed, I prefer to limit the interpretation of what this one study has found.

Right because the social justice conspiracy theories that permiate every level of society from government to business to academia are data driven and responsible in making reasonable inferences from limited data.

It IS relevant, albeit weakly. Bayesian reasoning - not every piece of evidence has to conclusively resolve the argument, you just gather a bunch of small findings and see which theory has more pointing toward it.

This one in particular is evidence against widespread wage-distorting oppression. It doesn't conclusively prove that there is none, but it points in the direction of less oppression. If you do a bunch of similar trials and most of them end up like this, you start to see a trend. Humans can interpret this intuitively - we don't have to test every single occupation before we can start doubting.

The other thing with Bayesian reason is setting your prior. Based on economic theory, I would think the strong prior is that no discrimination occurs against women.

Against that prior, we have studies demonstrating that, while the pay gap is small, there is some unexplained disparity. That should perhaps slightly weaken the prior. Now, we have this study where we are better able to explain the unexplained. In the context of the underlying theory and the existing literature, I think it is safe to say that wage gap BASED ON DISCRIMINATION is likely immaterial.

Interesting, but unlikely to put social debate to rest. So long as there is any “gap” (and won’t there always be?) people who are looking for unfair discrimination will find it and people who are looking for alternative explanations will find those as well.

This is exactly right. As long as a bottom line gap exists in favor of male workers, in any context, there will be thousands of credentialed women who reason backwards to fault whatever the company does, or allows to happen, as unconscious or structural bias. Look at the comments already. Harassment by customers (as if that could be eliminated by corporate hq) or taking time into account (because when you hire.a car, getting somewhere on time pales in comparison to knowing the driver’s incentives have been altered to produce wage equality — don’t all female executives feel that way?). Whatever a company does or doesn’t do is going to be castigated by professional feminists if the bottom line differs for men vs women.

Why is that bad? Because it kills real thought. It’s such an absolutist position that the reasoning in a given case has no persuasive effect anymore. And it’s unscientific in the extreme to only consider the employer’s practice or non- intervention as a cause while dogmatically excluding the female workers’ preferences as a potential cause.

Maybe, but the Jordan Peterson / Cathy Newman interview is an important stake in the ground here. The unthinking tide has at least been stemmed for now.

Peterson is a pebble in the stream.

Not so sure. Time will tell.

never knowing who to cling to, when the rain sets in.

and I would have liked to know you, but I was just a kid

Jordan Peterson might turn out to be important, but currently he's relatively unknown.

You're implicitly saying that science doesn't matter.

Perhaps that's a good observation, but it bodes poorly for our society.

Eh. We made it from the jungle (or the Rift Valley) to the renaissance without that much formalized science. As long as there are better success and failure mechanisms for the important things we’ll be fine.

>You’re implicitly saying that science doesn’t matter.


It matters a LOT less than many people, especially nice educated liberals, would like to think. Most people just aren't that persuadable by reason compared to naked self interest.

This bitter irony is that the "we love science" crowd are not themselves much persuadable by science if it clashes with the rest of their beliefs; see GMO, Nuclear power, AGW, human biodiversity, IQ, economics, evolutionary psych, and much of the quantitative areas of History and Pol Sci.

Like some of the commenters above, I wonder whether female Uber drivers miss out on some earnings to avoid sexual violence. Anecdotal case in point: a female Uber driver once picked me up at the airport and during the drive told me that she used to pick up bar patrons at night but stopped after being groped by a drunken male passenger. Now she confines herself to airport runs and the like.

Sounds plausible. But then it's a choice of safety over income. And what of a man who chooses an airport run because he doesn't want vomit in his car or being robbed? Can he blame his loss of income on his sex?

Doesn't the driver who is sexually assaulted have legal recourse? Does she not have identifying information on the attacker?

Is a woman who has her tit grabbed worse off than a man who has been punched in the back of the head?

It sounds to me like this argument of second degree effects on earnings is trying to rescue a failed theory.

I think the standard response is that women and men face different sets of risks in general and at night. Both men and women have the same probabilty of drunk riders vomitting but only one has a higher probability of sexual assault or harassment. Wouldn't you only need a small marginal difference for it to have an impact at the aggregate/average?

Whether or not "sexual" harm is a different kind of harm from purely physical harm is a legitimate question. As a guy, I much rather be groped than punched in the head but I also wouldn't have to worry as much that I'd face the risk of both from a perp.


Great point. Women might face the risk of sexual assault in addition to usual problems of all drivers.

On the other hand, women might face less risk of being punched in the head because they are women. Of course the men might get punched more because they're dicks or they confront bad passenger behavior more often.

It does seem to me, though, that there is an endless rabbit hole of feminist/minority explanations for pay disparity. If I show that education explains outcomes, the retort is that there is discrimination in education. If I then explain differences in education on parental involvement or role models, the retort is that those differences have a discriminatory basis.

It's not that I think these explanations are completely without merit, but the burden of proof lies on those making the assertion of discrimination, especially when their proposed remedies include public expenditures or preferences that come at the expense of others outside their group.

We're starting to argue about terminology here. Whether you call it 'sexism', 'discrimination', or whatever, the bottom line is compared to being a man, being a woman is harder. Militant feminists of course go way to far in blaming men, and in their suggested remedies, but it's fair to at least acknowledge the situation, and if you are a man to try to respect it.

"compared to being a man, being a woman is harder."

In some ways, yes. In other ways, no.

Sure, I guess I meant in the working world/employment context.

Men suffer more than 90% of workplace fatalities. They suffer by far more workplace injuries. We die younger from stress, much of it work related. We travel more for work and switch jobs more often. We work longer daily hours. Men vastly predominate women in the dirtiest and most dangerous jobs.

Men have a much higher unemployment rate.

Men wear suits in 100 degree weather while women wear skirts and blouses open to the nipple line.

So tell me again how women have it tougher in the workplace.

Women need to acknowledge and respect that, in the workplace, we are dragging them along for the ride.

Bah. You go through a bunch of fairly rational explanations of why women might get paid less in certain occupations because of differences in preferences that have nothing to do with intelligence or ability, and how that's totally ok, and then you ruin it by concluding with "so men are just better and we're pulling all the weight".

The ending was a bit strong, but he was actually Willitts was responding to the point that women have it harder, which...well who knows? It is far from obvious either way.

I would think in this particular case sexual harassment is highly unlikely. First, there is a paper trail. Second, if you are smart you can record at a cheap rate all rides. So, there isn't much evidnetary risk of being harassed and not able to prove it.

Didn't we have a study on NZ the other day showing women were benefitting from unconscious positive discrimination in the workplace?

I didn't intend to be insulting, but was rather mocking his broad, bold and unsupported assertion with an equally brash statement to the opposite extreme, but one for which actual evidence exists.

To be sure, discrimination might be keeping women out of high paying, dangerous jobs such as fishermen, iron workers, loggers. But I'm quite sure, without evidence to back this up, that few women would want these jobs even absent discrimination. It's merely a lifestyle preference. One may infer that regards women as "weak," but many a blue collar worker regards men in white collar jobs similarly. I make more money than a logger and would never trade my job for his at twice my pay.

Men bear disproportionate physical burdens of police work and the military. We do the "heavy lifting" around the house, in the fields, and in the office. A lot of this has to do with our absolute and comparative advantages in size and strength, in average. It's not any stretch to say men are pulling more weight, figuratively or literally. In mental tasks, such as academic research, men are far more prolific than women. Perhaps there is discrimination in publication, hiring, and education, but at least on the surface it appears men are greatly leading the pursuit of knowledge. It would take a lot of data from careful observation to conclude otherwise.

Another just so story. Just face it, your theory is unfalisible. All data can be contorted into supporting a bizarre, top-down conspiracy social justice theory.

I agree that this is likely the case. However:

1) The magnitude of this effect cannot be large given the data.
2) If true, what's the implication? My inclination is to say that some females (and some males) are making a perfectly legitimate choice to not pick up the likely-worst passengers. The drivers who are willing to pick up unpleasant passengers get a small premium for doing so.

Exactly right, and that's consistent with the entire body of literature of compensating differentials.

The question raised above, though, is whether women face more of these unpleasant situations because they're women. And, if so as a policy matter, is Uber responsible for paying women more because of this?

This is interesting research, but as others have suggested, the findings may have limited (though not zero) applicability to the wider debate about the gender pay gap. I say this as someone who thinks almost all of the pay gap is not due to discrimination. The incentive pay model for the livery profession seems to be quite different from the guaranteed base pay for almost all other jobs (the lack of incentive pay for most professions has been noted on this blog in the past).

Also, this may point to discrimination at the societal level rather than the firm level, as Smash pointed out. If women are afraid to be out and about late on Friday nights (when there are more fares) is that a preference or a reflection of the higher probability that women would face either unpleasant advances from customers or worse?

It's a controlled experiment with a unique data source. It isn't pretending to have wider application. But the inference that can be drawn is that not every pay gap is the result of discrimination. Feminists believe they all are. This is a counterexample.

Also take note that the pay gap, even before data analysis, is only 7 percent - far lower than other professions.

If you look at the data from the paper, average women earn just 68 cents for every 1$ earned by men driving Uber over the same period of time ($397.68 per week for men vs $268.18 for women). But of course much of this is explained by the number of hours worked, which was not accounted for many of the studies showing such large pay gaps in other professions.

Well, as Willitts said; it's an example of a gap created entirely by women choosing the "easier / lower paying " options in the career. Is anyone who is not a raving misandrist surprised by this?

What, women don't want to work alone on Friday nights? Well, I'm sure elderly and disabled people and effeminate men are afraid to be out of Friday nights too. Does this mean they are "discriminated against"? Or is the word "discrimination" so abused as to have no coherent meaning here?

I am of the general opinion that if "discrimination" is measured at most a few % points, then it is not worth expending legal and societal effort to remedy even if it is real. And of course, here it isn't even that.

The legal standard for discrimination is that disparate treatment is because of, not merely in spite of, gender. As you say, if tender men make the same choices as women for the same reason, the cause of the pay gap is from choices about working conditions, not gender.

I do concede that it's possible women suffer more workplace indignities from riders BECAUSE they are women. Perhaps they are seen as sexual targets in addition to just the usual contempt. Perhaps they are perceived to be weaker and thus more easily victimized. But does Uber have a duty to pay women more because of actual discrimination by Ubers customers? Obviously the "answer" is one of public policy and law, but from an objective, analytical, and libertarian perspective, I think Uber cannot be held responsible for conditions beyond its control.

Maybe women with small children should stick their kid in a car seat in the back so they can drive more on the job.
Saves on daycare costs too.

Have experienced this multiple times!

How naive do you have to be to think this "has no implications for social justice?" Everything has implications for social justice. Within a week there will be people screaming about Uber justifying misogyny and demanding they change their algorithms to better compensate women. Then will come the demands that whatever universities employee these unfortunate researchers fire them.

+1. They have found BadTruth and must be punished.

Hahahahaha gender-based "preferences" for time spent not at work.
Off to pick the kids up from school, or as I like to call it: preferences. My husband prefers to work.

You chose to have kids. That's a preference.

Presumably both parents choose to have kids.

Maybe you ought to demand better treatment from your baby daddy.

Or perhaps his opportunity cost of driving the kids is greater than yours, and thus your lower income is consistent with maximizing joint income and NOT the result of societal or individual discrimination.

Well, that would set off a kind of intractable feedback wouldn't it?
The man earns more, so the woman is charged with the task of dropping off and picking up the kids from school/daycare. Which makes it harder to put in long hours, which makes it harder to earn more. Which results in her being charged with the task of picking up and dropping off the kids.

Yes, there are feedback loops like that, Hazel. Doesn't make it discrimination at any stage.

You know the stats; gender earnings match each other until child-rearing age when the women drop out to child-rearing and the pay gap opens. Does anyone believe that women don't, on average (not all of them), prefer taking that route than men?

Does anyone believe that women don’t, on average (not all of them), prefer taking that route than men?

Personally, I don't care about the average pay gap. I care that women who don't prefer it that way have the opportunity to not do it that way.

I think you'd be hard pressed to find many people here who disagree with that, Hazel.

"I care that women who don’t prefer it that way have the opportunity to not do it that way."

So we're back that you chose to have kids, so you didn't have to do it that way.

Great point about the feedback loop.

Beautiful relevant subtle math! Thanks for linking.

To consider the question of social justice, we need to tell equality from equity. Separate stairwells by demographics: inequality. One stairwell for everyone: equality. A stairwell and a ramp: equity.

For example, I almost always prefer smoother, slower rides. I'd pay extra for that option (and no, you can't just ask, the habits are too strong). I perk up when I see that the driver is female, because she's more likely to drive smoother. If Uber designed for that option, women would get more business from people like me. That would be a more equitable design.

What about just allowing you to request a female driver? Would that be even more equitable? Would men be allowed to request female drivers? (Paris has such a system -- a taxi company with female drivers for female passengers...) What if men started using "smooth driver" as a proxy for "female"?

I don't really care about gender. If the style of driving, like "smooth" were a (measurable) option, people would offer that, much like people offer jazz, rock, and other music styles. Requesting gender is not going to lead to that.

You actually can measure driving smoothness by capturing data from the accelerometer in the phone. Some cities have used this to identify sections of road that need repairs. If you’re curious there are lots of apps that will give you a real time read out of all your phone’s sensor data. It’s pretty cool.

I want an option to bring m dog along. I'd gladly pay double the rate which surely would add to the surplus wages the driver receives despite the slightly increased negative scenarios that could arise.

Its an interesting study in that its a very large, rich, real world data set, with automatic data collection. The coming IoT world should make such studies more common.

I'm curious - do studies like this create a benchmark that negatively impact the academic prestige of studies with less extensive data sets? I think particularly of the "we had 75 undergraduates fill out a questionnaire" type studies which seem common in some fields?

Those latter studies always had lower prestige.

But you hit upon a good point. Researchers with better access to data have a huge advantage over those who don't. Aside from the IQ advantage of people who go to Harvard, they have substantial resources for purchasing data. People at lower ranked universities likely could do the same quality research but for the data.

"So you're saying all women are terrible drivers..."

Women are more cautious and less aggressive drivers. This has both costs and benefits.

For that matter, it's worth studying whether this pay disparity evaporates when including the cost of traffic tickets and accidents.

Excellent point! And I suspect women pay less for auto insurance, too...

I was going to make this exact comment. If men drive faster, stands to reason they get in more wrecks (and I think auto insurance data backs this up).

But my (limited) understanding of Uber is that all drivers are covered under the same policy. That means that the safer, slower women are -- albeit subtly -- subsidizing the faster, more dangerous men.

Ur so

Good point about policy premiums, but are the premiums still affected by accidents and tickets notwoth the same policy? And would tickets and accidents incur some risk of losing Uber employment?

One may not be able to quantify the costs of driving faster. I'm merely positing that they exist.

And there could be two equilibria, one that is faster and one slower, with women predominantly in the slower equilibrium, but net pay adjusted by speeding costs roughly equal.

I know lawyers, most of whom are female, who took lower paying government jobs in order to have regular hours, lower stress, and less variable pay. I took a similar route. So pay differences are often accounted for by non gender factors and choices that are highly correlated to gender, but by choice. At a minimum, this makes pay gaps ambiguous and not amenable to a clear policy prescription.

I've often wondered about paid maternity leave. Under what conditions would there be an equitable reason to grant women such paid leave not relying on tribal, self serving concerns? A very low birth rate?

My wife left a successful career to stay at home. to raise our daughters. It was a mutual and conscious choice for us. I had to leave public service for more lucrative options. This obviously widened the wage gap at the margin.

You're in denial as a tool of the Patriarchy; how can we have social justice when your male science-ness hates it?

OBVIOUSLY, all observable differences stem from discrimination all the way down. Everywhere . All the time.:-)

It's rather encouraging that the sex that so often can't reverse, and can't tell left from right, can get within 7% of male pay for driving. I'll bet men - with such inferior fine motor skills - couldn't get anywhere close to a 7% gap at, say, embroidery; 70% less pay for men would be more likely.

I believe the advantage in fine motor skills is a myth taken from observations of employment in comparative advantages. It isn't that women are absolutely better at embroidery, but that the male opportunity cost of embroidery is much higher.

Men dominate women in watch making and diamond cutting, although it's quite possible women are excluded because of discrimination.

Are you aware of studies directly measuring fine motor skills?

And there's a 90% gap between Americans and the rest of the Anglosphere at recognising irony, sarcasm, and satire.

I exemplify that remark!

Did they measure the gap for Asian women? It's probably more like 15%.

Are the accounting for the additional insurance costs paid by men?
Perhaps the pay disparity disappears when you account for the lower auto insurance prices paid by women.

Another good point.

But assuming there is an equilibrium between extra wages and extra insurance, ticket or accident costs, why do women seemingly prefer one equilibrium and men the other?

I'd posit that there are inherent sex differences regarding preferences for risk, patience, excitement. Men take more risks in both chess and poker which leads to much higher representation of men in the upper ranks notwithstanding no physical, mental, or social advantages.

And assuming that's true, is it optimal to nudge women into taking more risks for the cosmetic benefit of pay equality?

> I’d posit that there are inherent sex differences regarding preferences ..

Shhhhhh... the Google is watching...

Why are you presupposing that "riskier" = "better"? IIRC there was a article just last week about how men's high testosterone risk taking reduced their effectiveness as fund managers.

I'm not presupposing riskier is better, especially in an infinite game. If you're playing in a poker or chess tournament, risk taking increases the likelihood of coming out on top. Its the distribution of the maximal order statistic, not the mean, that matters.

I'm actually quite a conservative fund manager. I suppose I buck the trend. As for Uber, I suppose the average matters more for wage gaps unless the lower half of the distribution is censored. In that case, greater risk taking on fares would reflect a wage gap. And if "hustling" for fares is more lucrative, it also requires more "effort." A rational actor might be satisfied with lower income for lower effort.

I don't think what you are saying for chess is quite right. Mark Dvoretsky once observed that women chessplayers have more aggressive playing styles and the games tend to be more hard-fought. If you take the top 10 women chess players today, and compare them to the top 10 men chess player (or 10 random male chess players from the same rating range), I don't think you'd find that the men take more risks or that the women have stodgier playing styles. Nor will you find that if you compare 10 random women expert players and 10 random male expert players.

Men are also more monomaniacal. They devote greater shares of time and resources to singular pursuits.

Interesting point.

I found that when I served in the National Guard, dividing my efforts between two pursuits was detrimental to both, notwithstanding some synergies.

Devoting oneself entirely to a main profession does appear to promote peak efficiency and absolute results.

But now convince me that men have this trait in abundance compared to women. I believe it to be true from my own experience, but I've seen no data.

The Unter side of Uber, interesting article y/day.

It would be interesting to see if the the ratings of male and female drivers are different. Since the study points that male drivers drive a bit fast, that could negatively impact the sense of security, potentially leading to lower ratings.

Or higher, if you get there on time.

I'd guess that driver ratings are only very loosely correlated to driving skills, and are much more likely to correlate with personability, handsomeness, and how nice your car is.

From the paper -- "the average of rider ratings of drivers is statistically indistinguishable between genders. When we regress ratings on gender and the control variables used throughout the paper, we find an economically trivial (and, in most specifications, statistically insignificant) relationship between driver gender and ratings."

It's not that riders prefer male drives (they can't tell which drivers are male), it's that faster driving allows the men to do more rides per unit time.

Yes. This is a main motivation of all cab drivers. Maximize total number of fares. Long one way trips often lead to a fareless return.

The typical fare schedule of city cabs enhances this effect. The rate for the first mile is always higher. But Uber doesn't charge in this fashion, does it?

Interestingly, though, airport runs are usually heavily taxed which makes me wonder why they are more profitable.

Possibly because they avoid the "fareless return" problem discussed above.

Comments for this post are closed