Are Uber and Lyft responsible for the decline in drinking & driving in West Hollywood?
In a recent report, we explored the drop in drinking & driving in West Hollywood. We assumed that the decline was largely due to transportation network companies (TNCs) such as Uber and Lyft. Then we started to wonder if the numbers back up that assumption. In this report, we lay out three reasons to believe in an Uber effect on drinking & driving and two issues that leave us uncertain.
Reason #1: What users say
Avoiding drinking & driving is the number one reason users give for picking TNCs over their own cars. Researchers from the University of California, Davis surveyed users in Los Angeles and six other major cities. One-third put drinking & driving at the top, ahead of parking or any other reason.
Reason #2: West Hollywood numbers
Uber’s mass-market service, UberX, arrived in 2013. Drinking & driving has declined in West Hollywood in every year since then. We’re measuring drinking & driving in terms of arrests for driving under the influence (DUI) and alcohol-involved collisions in which DUI was the primary collision factor.
We compared the average numbers for the four years before 2013 to the four years after 2013. The numbers were twice as high before UberX came to town.
Of course, we wouldn’t expect the change to happen all at once. We’d expect drinking & driving to decline over time as TNC ridership grew. We can’t prove that directly, because Uber and Lyft don’t release ridership data. But we have a proxy: the number of taxi drivers, which goes down as TNC ridership goes up. The proxy is 90% correlated with drinking & driving. It suggests that drinking & driving is highly (negatively) correlated with TNC ridership in West Hollywood.
Reason #3: Academic studies
We looked for academic research to corroborate what the West Hollywood numbers seem to show. We found six recent studies that examined the statistical relationship between Uber and drinking & driving. Several of them support the Uber effect in their top-line conclusions:
“We find a significant drop in the rate of fatalities after the introduction of Uber X” — Brad Greenwood, Temple
“We find that Uber’s entry lowers the rate of DUIs and fatal accidents” — Angela Dills, Western Carolina
“Ridesharing services significantly reduce fatal alcohol-related auto accidents and, for a large subset of cities, DUI/DWI arrests” — Frank Martin-Buck, UT Austin
“Estimation of this [Uber] effect implies a…decrease in the alcohol-related collision rate” — Jessica Lynn Peck, CUNY
Issue #1: Correlation is not causation
The West Hollywood numbers show that drinking & driving shrank as TNCs grew. However, that doesn’t necessarily mean that one caused the other. Correlation is not causation. Many other factors — such as enforcement efforts — could be partly or totally responsible.
We noticed two facts that argue against TNCs as the (only) cause of the recent drop in drinking & driving:
- The drop started the year before UberX arrived
- A similar decline happened 10 years earlier, long before Uber existed
The chart below superimposes the earlier decline on top of the recent drop. They look a lot alike. It leaves us uncertain whether TNCs were really responsible for the recent fall.
Issue #2: Mixed academic evidence
That uncertainty makes the academic studies more important. However, after a closer look, we think the statistical evidence is surprisingly mixed.
We give more weight to three of the studies, because they were published in peer-reviewed journals. (“Peer-reviewed” means that the quality of the papers was checked by other experts who weren’t involved in the research.) The three studies came to conflicting conclusions about the Uber effect:
- No: The USC study found no evidence of an Uber effect on alcohol-related traffic fatalities or total traffic fatalities in the top 100 metropolitan areas
- Yes and no: The University of Pennsylvania study detected large drops in alcohol-related collisions in two cities (Portland and San Antonio), but none in a third city (Reno)
- Yes and no: The Temple University study found a small (4% to 6%) reduction in alcohol-related traffic fatalities in California due to Uber, but not on weekends and not in smaller cities like West Hollywood
The three remaining studies concluded the effect is real. We’re more cautious. We view some their detailed statistical results as more mixed or limited.
In the studies that found an Uber effect, the estimated size varied dramatically. The smallest estimate was 4% of drunk-driving fatalities. The biggest were 62% of alcohol-involved crashes and 40% of all traffic fatalities. Some found the effect was immediate, others determined it built up over years.
Overall, we think the evidence is too mixed to prove that TNCs are responsible for the decline in drinking & driving in West Hollywood.
List of academic studies
These are the academic studies we reviewed:
- Noli Brazil (was USC, now UC Davis) and David S. Kirk (Oxford), “Uber and Metropolitan Traffic Fatalities in the United States,” American Jounal of Epidemiology, 2016
- Brad Greenwood and Sunil Wattal (Temple), “Show Me the Way to Go Home: An Empirical Investigation of Ride-Sharing and Alcohol Related Motor Vehicle Fatalities,” MIS Quarterly, 2017
- Christopher Morrison et al (Penn), “Ridesharing and Motor Vehicle Crashes in 4 U.S. Cities: An Interrupted Time-Series,” American Journal of Epidemiology, 2017 (based on press reports in The Verge and other online sources)
- Angela Dills and Sean Mulholland (Western Carolina), “Ride-Sharing, Fatal Crashes, and Crime,” 2017
- Frank Martin-Buck (UT Austin), “Driving Safety: An Empirical Analysis of Ridesharing’s Impact on Drunk Driving and Alcohol-Related Crime,” 2016
- Jessica Lynn Peck (CUNY), “New York City Drunk Driving After Uber,” 2017