Hacker News new | past | comments | ask | show | jobs | submit login
MIT study that found low pay for Uber drivers to be revisited (reuters.com)
122 points by uptown on Mar 4, 2018 | hide | past | web | favorite | 103 comments



The problem is that the initial, flawed findings fit the biased narrative that the many in the media (both traditional and social) prefer to perpetuate. As a result, the surveys and stories that support their biases get top billing as headlines for days or weeks. People are often so eager to hear claims that support their beliefs that they don't investigate the credibility of the claims.

When the truth comes out, the same media outlets will likely do nothing and at most will publish a small redaction (same goes for individuals who spread this on social media). I have no expectation that the true nature of this study will be widely reported or nearly as promoted as the initial false results.

The world needs more rational skeptics. We have to hold each other accountable by ask probing questions about the nature of "facts" that are presented us. Seek truth.


Someone made a public claim using public data and methodologies. Someone else refuted it. That original someone is going to check again and get back to us.

I agree that the world needs rational skeptics -- and yes, perhaps more -- but looks like the peer review process is working properly here. Don't see a need to blame media bias?


You're confounding two entirely different things: media bias and the scientific process. The whole problem is that the media filters data that fit their agenda, even if its half-baked and incomplete. Then, when more info appears, if it contradicts the media's bias then they won't report it. They turn a blind eye to the facts and keep parttong their talking points.


I don't believe I'm confounding the two; I'm not saying that media bias isn't a problem. I'm pointing out that there's no reason to assume it is the reason for (possible) misinformation in this particular case.


How far off will the refined results be though? The sensational number was half of minimum wage. Is 7/8ths that much different of an issue?


the medium article claims pay was computed as follows: dollars_earned_driving * percentage_of_total_wage_from_driving

assuming this is correct, the less drivers rely on driving as a primary source of income, the more incorrect the $3.37 figure would be. curious to see the updated results.


Here's a pretty good explanation of exactly how the study calculated the number it did, it's pretty embarassing:

https://medium.com/uber-under-the-hood/an-analysis-of-ceeprs...

I'd previously read a paper by an economist using Uber pay data to analyze gender wage gaps. The numbers they had for both men and women were way higher than this study and it's usually best to trust administrative data over self reports.

https://web.stanford.edu/~diamondr/UberPayGap.pdf


That is flat-out embarrassing to report an hourly pre-tax income of $3.37, which implies accuracy down to the cent, when the actual reporting methodology is bucketing within ranges of powers of two. And then on top of that to completely screw up the accounting methodology anyway.

Much more accurate would be to actually get real data from drivers, by giving them instructions to extract desired monthly data from the app itself showing income and hours online (I presume the driver-side apps include this data?).


Thanks for the link. Those questions are worded in a pretty ambiguous way, making this survey pretty much useless. Who knows how the respondents really interpreted those questions?


This has some interesting numbers in it, but I live in a high rent part of the country and these statistics that throw around national averages leave me cold.

In similar conversations you can encounter rhetoric about how we should tax people worth a million dollars. This logic doesn’t consider that a $million for a 40 year old dentist in rural Ohio is very different than for someone who just retired in New York. That’s all they have for the rest of their life.

Financial policy decions need to consider money in terms of cost of living and also age of the people. Especially in a world where pensions are barely a thing now. Going after numbers without context ends up attacking the middle class.

So for instance the $20 an hour would make more sense to Reason if it were restated as a multiple of minimum wage or the poverty line. Because if Ubuntu is concentrated in metropolitan areas that could mean most of their drivers are near the poverty line. Not as incendiary as $3.50/hr but still quite damning.


Hi guys, I made a number of comments on that piece. Please let me know if there are any that you would like to be revised/looked more closely at.

Also you can find me on Twitter where I have over 150 tweets about this from the weekend. Would love to hear your POV.

http://Twitter.com/DaveCraige


Am I the only one that think's it's weird that Uber's Chief Economist is posting on their official Medium account with an article that sources data from a blogger to refute the MIT study? Like this guy has way more information at his fingertips, so I think it's a little suspicious that they aren't using their own internal data to make a public statement.


To be fair to him, the worst move he could make is put an official figure out there.


Wait, Uber could figure out how much their drivers make, then apply standard rates for depreciation and a good local gas estimate to figure out profits? What magic could they use to do that???


i thought they were using the blog data because thats what the MIT study was using.


bingo


Much less scientific but interesting nonetheless: Mr Money Moustache ran some numbers on driving Uber and they are Not Pretty either.

It's not even hard to find the relevant data on amortization and such: relevant government agencies will tell you what they think is reasonable. In Canada, that'll be this page. https://www.canada.ca/en/revenue-agency/services/tax/busines...

> The automobile allowance rates for 2018 are:

> 55¢ per kilometre for the first 5,000 kilometres driven

> 49¢ per kilometre driven after that

The logic goes -- you can be pretty sure to spend at least this much on your car, the taxman won't allow you to deduct more... So using these numbers, you can check the average Uber trip distance length, the time it takes you to drive it and the number of trips a month to figure out how much you are making. UberX in Toronto pays you 81 cents per km and 18 cents a minute. The Toronto minimal wage is currently 14 CAD an hour. If you do quite well and spend 30 minutes driving at a 30km/h average speed then you made 4.8 dollars on the per km part (.81-.49) x 30/2 and 5.4 dollars on the time part .18 x 30. The minimum fare won't save you because they don't deduct the 49¢ per kilometre car cost obviously.

Note https://www.mrmoneymustache.com/2017/11/22/mr-money-mustache... used the rough equivalent of 40 canadian cents/km for the same cost above we used 49c for -- it's not too outlandish, everything in the US vehicle wise is cheaper and there's some guessing involved anyways.


Mr. Money Moustache's cost estimates are wildly unrealistic. I ran a TCO on my own vehicle and it came out to about 35-40 cents per mile, much less kilometer. The MIT study uses 30 cents per mile, and this sounds right if you are being frugal and driving carefully, but apparently even this is being revisited. The idea that the government would not let you over-deduct is, in my opinion, a big mistake -- lawmakers know that people who work out of their cars are generally low-paid and therefore do not contribute much to the tax base in the first place, so it is a higher priority to avoid screwing over the poor than to nickel-and-dime them. Furthermore tax laws in America are influenced by some pro-car, anti-tax elements, who will of course act to make taxes lower for drivers.


If one assumes a fair and rational basis for the standard deductions, it's that people are driving their personal cars for business purposes. Those cars weren't purchased specifically to drive for business and they may not be optimized for economy. And, as you say, many people who put a lot of miles on their personal cars for business are probably average to the lower end of the income scale.

The 30 cents a mile in this study seems reasonable for a modest Prius or Camry assuming that driving empty to pick up passengers isn't a big portion of the total driving. It's probably close to the low-end though for acceptable vehicles.


You’re able to deduct the miles driving while actively seeking a passenger. My Lyft driver was sharing this hack with me while driving me home from the airport yesterday; he turns the app on while driving home from work everyday. If he has a ride, good. If not, he’s gotten to deduct his commute home as a business expense.


Your numbers are way too low as far as I am concerned.

For comparison, you can have a look at France tax deductions. They account for the type of vehicle. https://www.service-public.fr/particuliers/actualites/A12350

The absolute minimum price is 0.286 euro * km. That is if you drive more than 20 000 km a year and have the most economic car of all.


Fuel is way more expensive in France, though, and it wouldn't surprise me if maintenance were as well, considering that CPI is significantly higher. Also see my explanation of why lawmakers would overestimate deductions. Also the minimum deduction in that table is 0.245 euro/km which is for 5000-20000 km although it assumes you drive 5000 km/year at a higher price, which most people will do for personal use anyway.


The minimum deduction is the "d x 0,286". Don't get confused by the French. It's for the "smallest" type of car, I think you actually need a motorbike to reach that rate.

The most economic Prius from this year would fall in the rate d x 0,332.

Yes, there are high taxes on gas but it's not significant when discussing the total costs including maintenance, insurance and depreciation. Keep in mind the exchange rate as well.


> The logic goes -- you can be pretty sure to spend at least this much on your car, the taxman won't allow you to deduct more...

In the US, you are perfectly able to deduct more if you use the actual expenses method.

Source: https://www.irs.gov/newsroom/standard-mileage-rates-for-2018...

"Taxpayers always have the option of calculating the actual costs of using their vehicle rather than using the standard mileage rates."

Here's a Canadian tax advice site that suggests that actual expenses are permitted in Canada as well: https://www.thebalance.com/vehicle-expenses-you-can-claim-on...

Edit: Direct CRA citation that actual expenses are allowed: https://www.canada.ca/content/dam/cra-arc/formspubs/pbg/t777...


Yeah, it's a poor formulation although the intent can be figured out if you focus hard enough but let me make it easier.

> We can use tax allowances instead of actual car expenses because you can be sure the taxman won't allow you to deduct more than what you are spending so the earnings we calculate here will not be more than the actual.

As a formula:

#1: (actual car expenses) <= (tax allowances)

Adding (total uber fees)

#2: (total uber fees) - (actual expenses) <= (total uber fees) - (tax allowances)

and then we go on to prove

#3: (total uber fees) - (tax allowances) < minimal wage

And so:

#4: (total uber fees) - (actual expenses) < minimal wage

Changing <= to = (as you suggest) in #1 and #3 doesn't change the logic of our proof.


I think the government allowance in the USA is very generous. I generally drive reliable, economical cars (e.g. a Honda Civic) and I enjoy driving for work because I'm making a profit. Not much, but money is money.

Maybe if I was driving a lifted F-350 through hilly terrain I'd start to be at the break even point, but using that number as an actual cost is waaay overestimating the actual cost, especially since Uber and Lyft require relatively new (read: efficient) vehicles.


Are you certain your profit numbers include an accurate assessment of the long term maintenance costs and reduction in vehicle resale value you are incurring in connection with the miles driven? If not, as the original paper described, then you are unknowingly taking out a series of small loans on the value of your car.


The original paper actually used a value of $0.30/mile for vehicle cost. This squares pretty well with renting a car from Hertz intended for this purpose (about $200/wk unlimited mileage/insurance) and adding about $0.10/mile for gas.

(Obviously Hertz can buy and maintain cars for less than an individual. On the other hand, they need to make a profit on the rental.)


I did rent a car once for a working trip my SO had. We got out of it with a 40$ profit for a trip of about 600 km (yes gas included). For sure that renter include depreciation cost (I have never seen a car with them that had more than 150 000 km).


Interesting. I wonder if 50c/km includes gas or it is just car depreciation alone. Because here in Moscow, Russia it's very close to what you usually pay when you use Uber, no surge premium.

If that's real cost, it means that all Moscow drivers are engaged in money losing operation, and it's obviously not the case


From what I understand, the number is supposed to include, gas, car depreciation, maintenance and insurance.


> 55¢ per kilometre for the first 5,000 kilometres driven > 49¢ per kilometre driven after that

These are amazing, I love them. My SO and I don't own a car, we rent one when we need. Once she had a working trip to Quebec for a convention, I came with her and we rented a car. We made 40$ of profit from renting that car because of that 50 cents per KM that were refunded (in case it's not clear, it does include the gas price).

I have no idea if the renter we use (Communauto) loses money, I seriously doubt it because they keep adding more and more car into their network.


One might consider that the average drivers wage is highly depressed by a large number of people who are inexperienced or not very serious or not very smart at making money driving. I have a cousin who drives as his primary employment. If you have been driving for awhile, have an aptitude of understanding how to maximize your benefit in complex systems, and have a flexibility on when you work, one can do quite well. Knowing when and where surge is likely to happen. Knowing how to up your odds of getting a nice long trip on a freeway at your max Lux level when the traffic is clear. Having the skills and interest in buying used cars so you can get a good deal on the cheapest car that passes for UberXL or UberSelect. When and where to set your two daily destination options. Keeping up with all the bonus plans and other rules of all the ride sharing platforms in your area. Chatting up other skilled drivers to try to learn more tricks on upping your income.

It does not seem like such a bad idea for society to have some entrepreneurial like jobs that are easy to get but don't pay very well until one becomes good at it.


Driving for Uber is not entrepreneurial. One common feature of entrepreneurship is making money while sleeping. Even ignoring that, entrepreneurship requires a freedom to develop the business and take economic advantage of extraordinary market opportunities should they arise. An Uber driver cannot take advantage of a demand spike by raising prices or reducing less profitable service at will. An Uber driver is limited by Uber's dispatch algorithm at a compensation solely at Uber's discretion.

While a driver can potentially increase the hourly rate at which they work, they can do so only by choosing not to work at certain times. Unlike an entrepreneur, an Uber driver cannot increase their compensation rate by working to develop customers; expand business lines, and importantly by hiring staff. To put it another way, an entrepreneur would have the potential of increasing their profits with the availability of self-driving cars while self driving cars are potential competition for Uber drivers and likely to produce wage suppression.

Not to put too fine a point on it, but the large number of people described as depressing wages in the first sentence of your comment is the large number of people driving for Uber.


I won't even comment on if an uber driver is an entrepreneur, but this is wrong:

> One common feature of entrepreneurship is making money while sleeping.

Do you not consider mom and pop shops entrepreneurs? Local business that close at 5pm-6pm. I'm also not sure small business would usually qualify for this:

> entrepreneurship requires a freedom to develop the business and take economic advantage of extraordinary market opportunities should they arise


Walmart is an example of successful entrepreneurship starting with a mom and pop shop. Sam Walton made money because he had the freedom to hire other people and scale the business beyond his waking hours. An Uber driver can't.

Don't get me wrong, every mom and pop shop isn't run by successful entrepreneurs. But when the local team is doing well, they have the freedom to dedicate shelf space to big foam fingers in the appropriate color and up the price to whatever the traffic will bear. An Uber driver can't set prices when the Super Bowl comes to town.


>It does not seem like such a bad idea for society to have some entrepreneurial like jobs that are easy to get but don't pay very well until one becomes good at it.

Until it gets abused by employers as a cheap source of labor and then sold to the public as "adding jobs". Those entrepreneurial jobs exist already as door to door salesmen and MLM people.

If you can be this smart about Uber driving, you can be this smart about getting a decent salary job, so this doesn't make sense rationally because Uber ultimately has a low earnings ceiling for drivers.


People being able to get transportation at times, places, or prices that were not available before seems like a positive endeavor. People bugging you at your door selling you things that you never would want and MLMs, not so much.


"It does not seem like such a bad idea for society to have some entrepreneurial like jobs that are easy to get but don't pay very well until one becomes good at it."

Because no one ever needs to eat or pay the rent while they are learning their new job?

Sure, pay folks less while learning. That's fine. At least minimum wage, though - that's what minimum wage is there for. Then give them a raise. If employees have to "game the system" to earn minimum wage, what's the point? Besides, such companies do need folks during off hours, just like restaurants need wait staff during non-peak dining hours. Might as well act like it.


You seem to be equating gaming the system with making the system fair to the driver. The problem is that now not only is Uber taking advantage of your coworkers, but you’re indirectly taking advantage of them as well. Because you’re now exploiting the same thing Uber is. This doesn’t fix the system. In fact it probably makes it worse.

Take your advice to the end: nodody should drive unless there’s a surplus of fares. That would cause much of the day to qualify as surge. The premium Uber pays for surge would drop substantially in order to protect their margins. Which means your earning potential would decrease. Now the first quartile of employees are barely making more than the third quartile. That sucks for everybody.


It's easy to complain about the paper, but credit where credit is due: Stephen Zoepf listened to the criticism, considered it, and will be re-running some numbers. All within a few days of the paper being publicized.

That's more than you can say for A LOT of analyses that are published.

This is exactly how public discourse should work, so good job to everybody (so far).


The damage is already done though. Lawmakers and laymen will bring up this $3/h number constantly while making decisions that hurt Uber.


Uber's already a scam to begin with. It's not a bad thing to force them to compensate properly within the law.

1. It is employment, not a contract. They punish you if you don't do X hours a week.

2. They don't pay minimum wage, or other employment things like workers comp, unemployment tax, or others.

3. They actively seek to undermine all laws that protect the taxis and the citizens who ride.

4. They actively encourage lack of awareness with regarding commercial driving and the law. If one does not have commercial insurance, you are driving uninsured. If in a traffic incident, your license will be suspended.

It's just another company trying to socialize the risk and privatize the profit. And they're socializing the risk on the poor and needy by giving them hope and screwing them in ways they don't understand.


Exactly.

At best, a company like Uber could provide a logistics service to taxi drivers and transport companies. Ride-sharing, optimized route planning, easy dispatching etc are all nice features and make the service cheaper by reducing overhead costs.

However this is not what made Uber such a success. It was the unreasonably low prices it offered. How did it manage to do so? By externalizing a number of costs onto workers and society.

And by the way, the article has a very deceptive title. Just because the author agreed to re-run the numbers with Ubers "hopeful" and "optimistic" assumtions doesn't mean the initial numbers where wrong or that the new numbers will be substantially better. They're still employing people for less than minimum wage.


> Ride-sharing, optimized route planning, easy dispatching etc are all nice features and make the service cheaper by reducing overhead costs.

> However this is not what made Uber such a success. It was the unreasonably low prices it offered.

Isn't it? I don't use Uber because it's cheaper I use it because I can have them come straight to my front door hailed from an app leveraging gps with continuous updates to their progress then have my destination programed into the driver's gps ultimately paying automatically with no hassle of cash or fighting over credit cards.

The Uber experience was many light years beyond the existing taxi options (even in cities with good taxi system -- which is not many). That is why they are popular, not because they are cheap.

I also think there's something extremely novel about letting users participate in the system as drivers. There was nothing like that before either.


>Isn't it? I don't use Uber because it's cheaper I use it because I can have them come straight to my front door hailed from an app leveraging gps with continuous updates to their progress then have my destination programed into the driver's gps ultimately paying automatically with no hassle of cash or fighting over credit cards.

This is why I use uber the twice a month I need to get home from a bar or am otherwise unable to drive. And yeah, for twice-a-month rides, I'd be happy to pay full taxi rates; more, really.

The thing is, because uber is so cheap, I've gotten rid of my car and I just use uber all the time. If uber raised their prices, I'd buy a car. Sure, I'd still use them twice a month, sure, but I use them twice a day right now.

The upshot here is that about 1/30th of my uber rides are because the uber system provides a better experience than existing taxis in my area. 29/30ths of my uber rides are because they are so cheap and convenient that I pay only a small premium over owning a car to have someone else drive me all the time.

If uber raises their prices? I'll just go buy myself a Honda. Uber will keep 1/30th or so of my existing business; 29/30ths of the business they get from me will evaporate after they cross that line where I go buy that Honda.


I was surprised the study was from MIT as there were pretty serious flaws. The bigger question is how it passed peer review when they ran the analysis and checked the methods.

The response from the lead author was pretty reasonable and I will camp on his response. I was expecting something defensive rather than “yeah, maybe, let me double check my work.”


The answer is that it has not passed any peer review, it is a "working paper" only available to "associates" of MIT-CEEPR: http://ceepr.mit.edu/publications/working-papers/681

If you read the back-cover of any of these CEEPR working papers, it includes a disclaimer stating that they have not passed any peer review:

CEEPR releases Working Papers written by researchers from MIT and other academic institutions in order to enable timely consideration and reaction to energy and environmental policy research, but does not conduct a selection process or peer review prior to posting


Uber has the data. They know all the trip prices, models of cars (costs can be calculated), and time that drivers are online. This could be figured out, but would you trust what they said?


There's also no upside for Uber to release the numbers if they're not really that high - it'd be discouraging to their future contractor pool.


Costs are going to be a bit squishy but you're right that Uber has a pretty good handle on mileage, time, money paid out, etc. One suspects that if the average driver is making $25+/hr. after reasonable expense assumptions, we'd have heard about it loud and long.

My suspicion with some back of the envelope calculations is that it probably ends up somewhere in the minimum wage range most of the time. Especially at the higher end of that range, it sort of makes sense. Assuming at least a reasonable percentage of drivers have some rational handle on what they make and their real costs, that's about where you'd logically expect pay to need to end up.

To be clear, I'm sure some drivers don't really understand their costs or know that they're effectively borrowing from their future self who will have to pay maintenance bills--and feel they don't have an option. But it's a bit hard believing it's near universal behavior.


>>> My suspicion with some back of the envelope calculations is that it probably ends up somewhere in the minimum wage range most of the time. Especially at the higher end of that range, it sort of makes sense. Assuming at least a reasonable percentage of drivers have some rational handle on what they make and their real costs, that's about where you'd logically expect pay to need to end up.

Your assumptions are incorrect. The world is filled with people who can't estimate costs, it's actually pretty difficult.

I've worked in online gambling before. I think it has a lot of similarities with Uber. It's entirely possible to have the vast majority of your customers lose money.


Seems like the preventative maintenance person would be rewarded. I would also think the car mechanic since they can obtain a car at a lower price.

For budgeting, I’ve tried to just arrive at a cost per month of ownership for a car with expected depreciation (not cost to purchase), ins., gas, and expected maintenance.


No, Uber is shady as hell.


A truly fair study would also research the real income of 'normal' taxi drivers in the same places. Because normal taxis are generally also part of some kind of organization that takes a cut from their earnings. From the drivers point of view there might less difference then implied.


Does anyone have a link to the actual study? The closest I can find is a "research brief" PDF that looks like a university press release; I'm not sure it's even been published at all? And why does Reuters think it's acceptable to not include any links in this sort of reporting?


The actual study link is http://ceepr.mit.edu/publications/working-papers/681 but the full text is "for sponsors only":

> As a benefit to our Associates, the latest Working Papers are embargoed for a period of up to six months before becoming accessible to the public.

The list of associates can be found here: http://ceepr.mit.edu/support/associates


If you google for the paper, then a direct link to the full paper comes up, which works for me: http://ceepr.mit.edu/files/papers/2018-005.pdf


An Analysis of CEEPR’s Paper on “The Economics of Ride-Hailing”

https://news.ycombinator.com/item?id=16513089


Paper updated: Looks like Uber pays 41%-54% of their drivers less than min wage

https://twitter.com/yfreemark/status/970761011895652355


Hopefully these firms haven't yet their shares listed on stock exchanges, otherwise all the ignorant sheeps would have all shorted, only to see their money gone into expert investors pocket a few days later.


Here’s the thing. (1) Uber is not a full time job for a lot of people. They have a couple of hours here or there that fits within their existing schedule. I don’t doubt that it’s helped put food on the table for a lot of families. Unfortunately In a few years driverless cars will take this opportunity away (2) Uber drovers are rational market participants. If they were really averaging less than $4/hour then this wouldn’t be attractive to driver given our current low unemployment rate.


>"Uber is not a full time job for a lot of people"

I don't know this to be true without seeing a citation. In Seattle and Maui (two places I've used a car service recently), it has seemed to me at least the Lyft drivers I've encountered were all full-time drivers.


Could be because, when you use the service, you aren’t sampling drivers uniformly. For example, only 10% of drivers could do it full time, but 90% of all rides could have a full time driver.

(Numbers for illustration only; they aren’t likely to reflect Uber’s population of drivers)


I've heard (not verified) that Lyft's incentive schemes contain cliffs that encourage/select for driving full-time to a much greater degree than Uber.


48% of all Uber driving is done by full time drivers accorinding to their own company data


I talked to one guy who drives for Uber just for his hour long commute. If he was going to drive anyway, why not have someone else pay for it?


How does this work? He can't choose the passenger's destination.


You can tell Uber where you want to end up, and they will route you customers going that location.

Also, I'm astounded at least two people bothered to downvote my factual reporting.


I believe you can, twice a day.


Does anybody else find the Tweet response from Dara Khosrowshahi deeply childish? The in-depth response from Jonathan Hall is, whether you consider it to be biased or not, a well-researched, calm and explained piece of peer-review, but the CEO's tweet is pathetic and kitsch: "MIT = Mathematically Incompetent Theories (at least as it pertains to ride-sharing)"?

Is that the level of communication that we expect from our leaders now? I've always held an image of the character that CEOs of large corporations strive towards as reasoned, rational and "stately", at least in their public role. This sounds like the outburst of a petulant adolescent, and is disturbingly reminiscent of a certain presidential figure.

Is this a consequence of some societal shift towards communicating at the lowest possible denominator, or do I just have an overly conservative view of what is expected from people in those positions?


Well, it does go along with the recent decorum change with respect to "bigger buttons" and the like.

Stately speaking has hit a new low. It's no surprise that CEOs and politicians alike are picking up on what I call Idiocracy-speech.

There was a video a few years ago published on youtube by CCP Grey called "This video will make you angry". It explains how angry memes spread faster, and are more durable than other types. I'd surmise that this is true.


You didn't cite the whole tweet. The real tweet is a bit longer, cites analysis written by Uber, and in context, does not come across as childish:

MIT = Mathematically Incompetent Theories (at least as it pertains to ride-sharing). @techreview report differs markedly from other academic studies and @TheRideshareGuy recent survey. Our analysis: [https://t.co/S2aAqCuDR0]

https://twitter.com/dkhos/status/969805860594581509


I apologise, that's totally true. I did cite only the bit that jumped out at me, and my implication that it was the whole tweet does make it seem worse than it was, but I would disagree that the context redeems the statement.


It’s still childish. It’s the same as writing ‘Micro$oft’. Might be funny if you’re having beers with your mates. Otherwise just refute the claim with better data.


Whatever it is you should accept that it is. If you immediately take the moral high ground and deny that what is IS you contribute to the problem of robbing people of their voice by deeming their thoughts or their way of thinking as wrong. Such judgment immediately closes the potential for dialogue.

Take certain men coming out in women's defense but then being lambasted as hypocritical because he himself had some less serious offenses. This attitude alienates men and causes a defensive reaction. Instead of attacking the real problem which is a culture problem that affects men and women the individual human is attacked. This is no way for change. It marginalizes opinions that fester until they become worse.

Do we dehumanize those who would dehumanize others? No. Because acceptance of the humanity of others and ourselves is the only way to break cycles of hate.


‘“I’m re-running the analysis this weekend using Uber’s more optimistic assumptions and should have new results and a public response acknowledging the discrepancy by Monday,” he wrote.‘

Re-running analysis with more optimistic assumptions is not how science works.


The assumption is with regard to how respondents interpreted a vaguely worded question. Not an assumption of what 2+2 equals.

This study has a huge human component since no single party has all the data needed to calculate the answer.

Uber has a lot of data, but even they don't know total hours spent before or after shifts in prep for work nor expenses with regard to vehicle maintenance.


> Uber has a lot of data, but even they don't know total hours spent before or after shifts in prep for work nor expenses with regard to vehicle maintenance.

I would think Uber has exactly the data they would need to calculate this meaningfully, or at least the points you list aren't the decisive shortcomings. Most jobs require that you get yourself to work and prepared (to some degree, at least, but one that seems to apply fine to Uber drivers) on your own time. Also, in the blog post they state that they agree with the methodology around calculating cost of vehicle use, which regardless is not an area of great contention. There are stable and largely uncontroversial numbers published by tax authorities, and there are rental agencies that provide vehicles at fixed rates.


No, that's precisely how it works. All of science is based on making assumptions.


But should those assumptions really be based on angry tweets from someone with a colossal vested interest?


Uber's complaint was that he assumed that when the questionnaire asked about the hours per week people worked across all on-demand services, their answer only included the hours they worked driving for companies like Uber, but that for the next question about money earned across all on-demand services their answer was their total income and should be multiplied by the percentage of their total income that came from driving given in a later question to get their Uber income. Those assumptions are basically incompatible with each other. See https://medium.com/uber-under-the-hood/an-analysis-of-ceeprs...


It's quite clear that the survey was flawed. However, I'd rather consider all responses tainted and require re-surveying rather than trying to account for it post hoc.


The reply in the uber blog is less "angry tweet" than "good peer review": it points to a very specific methodological weakness which has bearing on the final conclusions.


If somebody with a colossal vested interest makes an angry tweet that also happens to point out the study author made basic errors (that's what this is about: the author of the study likely made invalidating errors; no amount of changing assumptions is going to fix that).


OK, let's imagine he we will re-run the study with optimistic assumptions and the result would be that Uber drivers are compensated lavishly (and if we take more moderate assumptions, they are making adequate money).

And with another set of assumptions result was that they are serverly underpaid.

So, exactly what we have learned here, other then different assumptions yield different results?


Nothing. But learning that different assumptions yield different results is useful in its own right. It tells you that this study is not capable of answering the question it set out to answer, and you will need to do a better study.

Perhaps you have come across interval arithmetic:

https://en.wikipedia.org/wiki/Interval_arithmetic


We learn that the outcome we care about is highly sensitive to the initial assumptions and we should get data on those assumptions.


If the re-run concludes that even more optimistic assumptions still lead to underpaid drivers, though, that’s useful information.

The key aspect here is the study’s author is willing to change his study based on criticism, to try and determine whether it still produces the same conclusions. That is precisely what science is about!


We learn that there are no absolutes when using quantitative methods to measure human actions.


not much, which is a big part of the problem.


I can't tell if this is sarcasm. In case not, the scientific process is about gathering data to test the validity of hypotheses. Your phrasing makes it sound like science is mainly about spewing conjecture.


Assumptions are an essential part of science. Any modelling study (that's what this was) starts with a collection of assumptions (even true for ab initio simulations). Those assumptions are literally coded into the analysis. They should also be clearly stated in the analysis writeup.


Calling this 'a study' is a stretch. This seems more like pure analysis to me. No non-trivial assumptions are necessary to compute the average wage of Uber employees. In this case is was simply a calculation error.


A hypothesis is a conjecture - we hope not a fanciful one, but a conjecture nevertheless.


Its economics its not called the dismal science for nothing


> Re-running analysis with more optimistic assumptions is not how science works.

String theory is nothing more than an assumption.


String theory is still just a speculation by some theoretical physicists, albeit some of them high profile. It's not part of a model accepted my mainstream physics so bad comparison.


And is untestable.


I find that statement a bit disturbing too, but do we need "science" to determine people's average pay now?


I agree that it’s an awkward label, but “good research practices” felt worse.

It was median, not average, as well, which makes sense as potentially being lower than an average wage.


The median is better for things like pay as it removes the extreme outliers - I doubt you'd like it if your negotiation for a pay increase was denied because your lavishly remunerated c level execs skewed the average.


Exactly. Which is why the responses which focus on how the "average" pay is higher than the "median" pay the report mentioned is misleading at best.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: