DUI rates decline in Uber cities | Hacker News Reader

DUI rates decline in Uber cities

blog.uber.com

100 pointsjmduke51 comments

If this is true, this is a big deal. However study is inherently flawed because it uses only two cities for it's data. There are a number of factors that could have called this result to be observed, for example increase in SF's DUI due to more rigorous police patrols. I also did not see any seasonal controls. DUI rates might be different depending on the time of year.

Here is what would provide for a way stronger claim:

1. Aggregate DUI levels change since UBER started operating across all cities with no UBER service.

2. Aggregate same data for cities with UBER.

If UBER causes decline in DUI rates hypothesis is true, we can expect two trends:

1. On average cities with UBER will observe relative number of DUI's going down when compared to cities without UBER.

2. Amount of UBER rides given will be positively correlated with decrease of DUI rates.

Personally I think it would make sense that UBER decreased amount of DUIs and I would really like a data released that can prove it.

It is probably true. Cheaper taxi services would cause ppl who drove drunk for financial reasons to hire an uber car.

Tbh, I am surprised no city ties liquor licenses to taxi services. (E.g. sell an open container, have a contract with a taxi company to drive people home)

>Cheaper taxi services

don't know about cheaper. "Reliable" would do the trick for me. Remember waiting for a taxi for an hour for a 5 mi ride. I'd walk it easy if i knew that it'd take an hour to wait.

Reliable, convenient, timely, and available anywhere. Those are all pretty significant since bars are not always in downtown cores, and taxis rarely patrol around house parties.

This. When I leave I want to get home so I can lay down and make the world quit spinning as quickly as possible.

I've never really had any trouble getting a taxi to arrive within 15 min of when they said they would. So maybe we just deal with different ones. I'm not in the SF or NY areas.

It actually doesn't even provide data for two cities, just one. It talks about anecdotal data that is the equivalent of hearsay: "we consistently hear from our riders in Chicago and elsewhere."

Did you read the article? It gave the sources for the data for both cities.

Honest question: is this not a classic example of the fallacy of claiming causation in an observational study, in the presence of a potentially infinite number of confounding variables? I'm not a statistics expert by any means, but this is one of the first things taught in an intro stats class: observational studies can establish relationships (correlation), but only well designed experiments can establish causation. This is quite obviously an observational study, and a tiny one at that.

Furthermore, wouldn't the fact that marijuana was legalized in Washington violate the parallel trend assumption that has to hold to use difference in differences when comparing with San Francisco?

I understand the powerful incentive for Uber to provide this narrative, and I think anything that encourages people to not drive while intoxicated is a good thing. I just find it disheartening that companies feel the need to start with these powerful marketing narratives (which these types of 'studies' really are, not honest efforts to uncover useful knowledge), and use dubious analysis to give those narratives credibility.

It's not that I'm disputing there is a relationship - they've obviously identified that. I have a problem with statements like these: "We estimate that the entrance of Uber in Seattle caused the number of arrests for DUI to decrease by more than 10%." And, "We can reasonably extrapolate these results and apply them to what we consistently hear from our riders in Chicago and elsewhere." It is quite a stretch to claim that Uber caused anything based on an observational study with a minuscule amount of data, and it is even further of a stretch to claim that you can combine this claim with anecdotal evidence and somehow generalize it into principle.

The author probably has no background in experimental design, so I would spare him of statistical scrutiny. This includes the author's furtive attempt to sneak in causation into a simple regression.

Or, if I were to give the author of the benefit of the doubt, Uber realized that this potentially inaccurate statistical claim ("Uber is inducing a drop in DUI") can help them win people's goodwill in areas where their operations have been thwarted/stunted by bureaucracy.

If so, they are really shrewd (for most laypeople don't understand anything about statistical inference), but I am not sure if I would trust a company that resorts to such a disingenuous campaign.

I actually want Uber to be a success. I love their service and envision it to be a real-time bidding platform for all sorts of (offline) services. But at the same time, I would be concerned if a company of such great promise is already acting intellectually dishonestly.

Hence, I really, really hope the author is just an idiot.

> Hence, I really, really hope the author is just an idiot.

Not knowing good experimental design doesn't make someone an idiot.

I agree. Not knowing anything does not make anyone an idiot. Not realizing that you do not know something, however, kind of does.

If "unknown unknowns"[1] make people idiots then everyone is an idiot and the term is neither an insult nor meaningful. Likely there's some other reason you have for wanting to use a mean and angry word that you haven't articulated well. What you describe in your original comment is probably just the Dunning–Kruger[2] effect.

[1] https://en.wikipedia.org/wiki/There_are_known_knowns [2] https://en.wikipedia.org/wiki/Dunning%E2%80%93Kruger_effect

It might be a case of the Dunning-Kruger effect. Either way, I am willing to concede that I misused the term (perhaps I am the one being an idiot here). I am still surprised that the blog post got published on Uber's blog though.

Oh god, that chart [1] took me a long time to understand. Since when was it a good idea to use floating, negative bars in a bar chart to indicate change over time.

[1] https://lh6.googleusercontent.com/3VC2E1ok_hRV5MPXS3J9s5u9xe...

I think that is a waterfall chart, badly drawn. The gain bars are typically shown in green, and loss in red.

This is a pretty big deal. As short-sighted and destined to fail as the current backlash against Uber is (see the cease-and-desist order issued in Virginia this week), it becomes a different narrative altogether if you can show that Uber saves lives. Any correlation with a decrease in DUI's is going to be hard momentum to counteract. If Uber can show definitively that highway safety increases in Uber cities, it's very compelling evidence that they should be in every major city regardless of existing taxi laws.

Although lobbying for increased highway safety has gotten us some atrocious national policy in the past (the 21 drinking age) - it would be nice to see it be used for a good cause here and give more cities modern transportation options.

> If Uber can show definitively that highway safety increases in Uber cities, it's very compelling evidence that they should be in every major city regardless of existing taxi laws.

No, it's evidence that cities should increase the availability of taxis or taxi-like services in general, not Uber specifically. In fact, due to more stable pricing, I'd expect Lyft to do better than Uber, and normal taxis to do even better.

Why would normal taxis do better? (Currently in SF) they're more expensive, the cars are less-well-maintained, and the drivers are often... I won't say "rude", but... less than pleasant to deal with. Not to mention the dirty looks you get when you want to use a credit card, and just the hassle of payment and tipping. Their dispatch services suck (they say 20 minutes and most of the time never show up), and hailing a taxi is a pain.

Simply increasing the availability of taxis won't fix those problems.

If operating a taxi was $250k cheaper (the price of a medallion in SF) you get Uber's results.

Most of uber's "advantages" stem from the fact taxi medallions get leased for about $1k/month that uber's contracted drivers don't pay.

And not exploiting their drivers, and having a gay-friendly employment force, and having rated drivers, and having trip itinerary transparency, and not having to cruise as much, an having visibility over where other drivers are to improve coverage...

Some of that is good and true, but trip itinerary transparency and rated drivers don't necessarily make for a good ride...

4 of 5 of my Uber rides in the last two weeks had 4.8-4.9 star drivers that didn't know the intersection I wanted to go to in SF. They all searched directions to/from: Sutter/Stockton, 20th/Valencia, a block from Market/Church, and 4th/Howard - all of those being non-trivial streets in one form or another (especially Market St, really, you're asking me for directions to Market St, that's like asking me where the Strip is in Vegas). Not to mention I'm not sure why they keep asking for directions too, since it's not like a cab I'll never see again overcharging me through my own fault - if I have a problem I will contact Uber and they will fix it (also a good thing!). One guy even asked for instructions, disregarded them, got stuck in traffic without ending the ride early, and made me and my husband late for an appointment that we would have been 5-10m early for if he didn't space out - helloooo 1 star rating. Oh, and half the time, I will get drivers with so much cologne on that I feel sick.

Say what you will about the terrible state of cabs in SF because it's all true and then some, but I think both cabs and Uber have a lot of _different_ problems to work out. A 5 star driver is not someone who has to enter in "Market and Church" in a GPS, it's not someone that poured half a bottle of cologne on themselves that morning, and it's most definitely not a person that fucks around driving in the direction they were explicitly told not to go in.

Some days I am jealous of NYC and London and other cities where I've had excellent experiences with cabs and Uber just adds a small layer of luxury/convenience on top of that if I choose. SF used to be that way with Uber a few years ago but not so much anymore - the price decrease and choices are nice but I wish I could get quality too - both Uber Black and UberX are just meh now.

The rating inflation could be lyft's fault. The lyft system is specifically inflated: 1 & 2 are really bad rides, 3 is "bad enough that you never want to be matched up again", 4 is "needs work" and 5 is "good". If you have an excellent ride, it's 5 + tip. However, deactivation is ALSO inflated, Lyft wants only 4.5 + drivers (after a certain number of rides). Passengers crossing between Lyft and Uber could be skewing the statistics of driver ratings.

As a lyft (and uber, from time to time) driver with 9 years of living experience in San Diego, there are times when I know exactly where the destination is but I am not sure of the "best" way to get there, or I am worried that there are multiple ways to get there and having the GPS decide gives it an "objectivity" that CYA's me for 'dispute resolution' purposes.

And both lyft and uber have been expanding like crazy recently, so having a 4.8 or 4.9 could be the law of averages in small numbers working against your experience.

This is interesting, but it would be more compelling if it mentioned the time scales involved. I really don't know if we're looking at 6 months' worth of data, several years' worth, or what. I'm assuming it's corrected for seasonal fluctuations, but with no longitudinal scale it feels a little handwavey.

I would like to see a line graph, not some really weird bar graph (I've never seen a graph like that, so it's unfamiliar and hard to read). Show me the trend, preferably with different lines for marijuana in order to show it's impact separately.

I don't really buy it. What if DUI rates fell across all cities nationwide, regardless of Uber being on the market?

Doesn't really matter for winning the narrative.

The US enacted a 21-year old drinking age in 1984. It has been an absolute train wreck. One of the most backwards and out of touch with reality national policies that we have. The only reason it's still around is because of an idea that it somehow "saves lives". The stat most often cited to "prove" this, is the number of alcohol related traffic fatalities that have occured in the US since 1984. They have declined, significantly. Does it matter that this has nothing to do with the drinking age? Does it matter that number of highway deaths have declined across the entire world, and at a much faster rate than the United States? Does it matter that highway deaths declined at the same time as a rapid increase in seat belt usage? No. None of it matters. In terms of the narrative MADD won, and half the country believes the drinking age saves lives, even though it's utter baloney.

If Uber can play up the idea that having Uber around saves lives, it's a trump card. It doesn't matter how sound the statistics are, they'll win the narrative in every city. As a politician do you really want to be against saving lives?

It does do some good.

Countries with lower drinking ages have much higher rates of binge drinking.

The WHO Europe region has a 70% higher incidence of teen (15-19) binge drinking than the Americas region: http://i.imgur.com/oAbeg7x.png?1

Canada, with its 18 and 19 year old drinking laws has a far higher rate of teen binge drinking than the US, as do the European countries: http://i.imgur.com/iaJbDLd.png?1

Finally, we find that the US has a overall rate of binge drinking in the middle of Europe, lower than France and the UK but higher than Germany and Spain: http://i.imgur.com/h0BkKGe.png?1

You also have to give the value of 20-something binge drinking in both continents for the picture to be complete. Us could have a perfectly good case of postponed binged drinking.

This is possible although I don't have that data. It's still preferable as the amount of brain damage would be significantly less than during teen years, and they would hopefully be at least somewhat more responsible.

The lack of error bars when tracking illegal activity is vary suspect. However assuming there numbers are accurate heavy drinking in and of it's self is a vary minor issue compared to say DUI related deaths which are lower in Europe. Sure it's complex but only looking at negatives tends to create bad pollicy as you end up with stuff like "Taboo till 21 may actually promote drinking in the over 21 crowd."

Lower DUI in Europe might have more to do with availability of public transport then with anything else. Drinking group does not need non-drinking member nor pay for taxi to get home. They just take relatively cheap bus.

Completely anectodal, but I've been a big time binge drinker in my teens in two different EU countries, and I never really felt the urge to drink starting in my 20s. I've gotten that out of my system, no longer as cool as you'd think. I don't even drink when I go out to a bar and club (you won't know that though, it looks like I am) and I go wilder than most drunkards.

I'm having trouble understanding their statistical work. How is the "differences in differences" framework different from having two experimental groups and no control?

I'm also not sure that extrapolating to Chicago makes sense. I live there. Since anecdotal observations seem to matter, it seems that the people who would order an Uber car would call a cab if Uber wasn't available. It's hard for me to imagine Uber providing such an addition in service that it would change DUI rates.

Hello Uber, a better study would be the number of ER visits cuases by suspected drunk driving. Hire me. There are data scientists and there are creative data scientists.

Uber can't hire you if they can't email you. So put one in your profile. :)

"cuases"

Creative data scientists, but maybe not great spellers. :-)

Bugs me when an analysis doesn't start with a visual inspection of the data. How about a plot of DUIs vs Time? is it linear? Next does it show a discontinuity at some pint in time? what effect does changing the bin size (time) have?

Uber has definitely cut down on the drinking and driving of my friends, no question.

Hypothetical plot twist: The actual number of people driving under the influence has not declined, but as police resources are diverted to Uber-related incidents — e.g. drivers (allegedly) assaulting [1] or kidnapping [2] passengers, or the generic disputes you would expect to arise from a taxi service — their capacity to arrest people for DUI has declined.

[1] http://www.latimes.com/local/lanow/la-me-ln-uber-driver-arre...

[2] http://www.hollywoodreporter.com/news/la-uber-driver-arreste...

Uber should give people 10% discounts if their blood-alcohol level is above the legal limit.

You might want to consider all the incentives you'd be creating there....

Hey, Uber is buying the last round :)

Smart Analysis. Governments miss it all the time.

Many people would argue that the cost savings of Uber/taxi are already substantially greater than a DUI.

The results don't even looks statistically significant at the 5% alpha level.

I'm not really following the analysis too well (I'd like to just see the raw data) but when they did their differences in differences seattleTRUE:uberTRUE had a p-value of over 0.05

I, too, will comment on this article without reading it!

Yes, I read the article and I applied an idea I took from one of the first articles on Nate Silver's 538 'news' site [0] of using Bayesian statistics to decode news as you read it and in my opinion the data presented is very underwhelming and there isn't any firm ground they are standing on here.

Here's my 'Bayesian logic' applied to it. My 'prior believe' in the claim being made is that I don't believe Uber has broad penetration across all demographics. Yes, people who are in the tech world and under 30 know of them, but not much outside of that. I was the first one of all my friends to use Uber (I would know because I have gotten over $60 in credits for signing them up). So outside of SF, my initial personal belief is that Uber doesn't have wide usage. So either the largest demographic of DUI violators are nerds who are younger than 30 or Uber is magically being used by people who would prefer using a more expensive Uber service than driver over a cheap taxi ride.

So let's look at the evidence they provide to see if it changes my prior significantly: one city was explicitly given data about without any hint as to what time-frame they are talking about. The other cities were literally the equivalent of hearsay arguments so because I'm the judge of this court, that evidence doesn't count. When you're talking about month-over-month and even year-over-year trends, a drop of 10% is just as easily explained by a random process modeled as a poisson [1] over the entrance of Uber. I would know having dealt with lots of time series data.

One finally comment, my apologies to the author of the bar chart, but that is just terrible. It's hard to read and doesn't make much sense from looking at it first. How many bar-chart graphs are out there that look like that? It could just be me speaking aesthetically because I'm weirded out by the random floating bars in the middle, but I feel Edward Tufte would have something to say about that.

jusitizin, I know you weren't speaking to me directly, but I read the article. Hope this comment proves it :)

[0]

http://fivethirtyeight.com/features/a-formula-for-decoding-h...

[1]

http://www.wired.com/2012/12/what-does-randomness-look-like/

This has about as much weight as pirates and global warming

http://www.forbes.com/sites/erikaandersen/2012/03/23/true-fa...

No, it has more weight because there is a plausible mechanism for causation. It isn't proof, but it's an interesting potential theory.

Reply on news.ycombinator.com