On the ten biggest album releases measured by Spotify streams, US traffic fatalities averaged 139.1. On the surrounding comparison days, the average dropped to 120.9. That 18.2-death gap sits at the center of a music streaming traffic fatalities study built on an event-study comparison of release days against nearby control dates in an NBER working paper.
Patel and coauthors first published the paper through the National Bureau of Economic Research in February, tying that death toll to something else spiking on the same dates: music streaming. The researchers treat streaming as a proxy for smartphone use, and the study's results put the release-day increase at 43%.
The Harvard Gazette reported the same 43% figure in March. It's one of the more provocative pieces of distracted-driving research to surface this year, and reactions have split about how you'd expect.
Some have read it as proof that a chart-topping album can get people killed on the road. What the data actually supports is narrower: streaming activity and traffic deaths rose together on ten unusually large cultural days. Whether one caused the other is a separate question, and it's the one this piece spends the rest of its time on.
What the music streaming traffic fatalities study found
The paper, titled "Smartphones, Online Music Streaming, and Traffic Fatalities," built its primary analysis around the ten biggest album releases by Spotify streaming volume from 2017 through 2022.
For each release date, the researchers compared US traffic fatalities against a control window covering the ten days before and after. Economists call this an event study: it treats the surrounding days as a stand-in for what fatalities would have looked like on the release date itself, had no album come out.
That's a design assumption, not a separately reported finding, and it's worth keeping in mind before the numbers start doing the talking.
The release-day estimate came out to 139.1 adjusted fatalities, with a 95% confidence interval running from 126.8 to 151.4. The control-day estimate was 120.9, with a much narrower interval of 119.6 to 122.2.
The control estimate draws on many more days, which helps explain its tighter interval. It doesn't mean every control day was typical, only that the estimate is built from considerably more observations than the ten release days.
The 18.2-fatality gap between the two carries its own confidence interval of 4.8 to 31.7, with a p-value of 0.01, which translates to a relative increase of 15.1%. Summed across the ten release dates, that gap totals 182 additional fatalities relative to the study's control comparison.
That total is an arithmetic sum of the reported gap, not 182 deaths individually traced back to a specific stream or a specific driver.
A p-value of 0.01 means a gap this size would be unlikely to appear if the true underlying difference were zero, given the model's assumptions. It doesn't rule out other explanations, and it doesn't identify streaming as the cause.
Ten release events are still a small, deliberately selected sample of unusually large cultural moments, chosen because they generated outsized streaming numbers. The published study does, however, include sensitivity analyses expanding the cutoff to the top 5, 15, 20, and 25 albums, with the excess-fatality estimate remaining positive and statistically significant at each cutoff.
Testing whether the pattern holds beyond the study period still means running the comparison against newer or otherwise independent releases, rather than relying only on the original 2017-to-2022 sample.
One more precision note matters here. A discussion of the study on the Statistical Modeling blog describes the headline claim as a rise in fatal car crashes. The study itself measures traffic fatalities, a related but not identical count, since a single crash can kill more than one person.
This piece sticks with the study's own term, fatalities, throughout.
Smartphone music streaming and distracted driving
The study's central move is treating streaming activity as a stand-in for smartphone use, on the reasoning that streaming commonly occurs on smartphones. A proxy works like a thermometer for weather: useful for inferring a condition without directly measuring what you actually care about.
Streaming isn't cast as dangerous in itself. It's a stand-in for something the study doesn't observe directly, which is how often people interact with smartphones while driving.
Streaming rose by 43% in the study's analysis. Either way, streaming activity and traffic fatalities climbed together on the same ten dates, which is the core pattern behind this album release days traffic fatalities finding.
The proposed chain runs like this: a major release coincides with a streaming surge, the surge is used as a marker of phone activity, and phone activity behind the wheel raises crash risk. Each link is plausible on its own.
But the study measures aggregate streaming volume and aggregate fatality counts by date, not what any individual driver did behind the wheel. It does not observe whether drivers were using phones, or what they were doing on them, when crashes occurred.
That's a key unobserved link in the argument, and naming it precisely matters more than letting a strong correlation quietly imply a mechanism the data can't confirm.
Does music streaming increase traffic fatalities, or just coincide with it?
The care in how the researchers phrase their result is worth sitting with. The authors describe major album releases and increased streaming as being associated with higher traffic fatalities, language deliberately more cautious than a direct causal claim.
The published model does account for several obvious alternative explanations. It adjusts for federal holidays, day of the week, week of the year, and calendar year, and the ten-day window includes the same weekday one week before and after each release.
The researchers also ran falsification tests using random Fridays and random dates, a same-date comparison in other study years, and a Friday-only sensitivity analysis. Those checks strengthen the reported association, but they still don't directly establish smartphone use behind the wheel as the mechanism.
The analysis began as an NBER working paper, but it was published in JAMA Network Open on August 17, 2026. That removes the earlier working-paper and peer-review caveat, but it doesn't turn an observational event study into direct evidence of what individual drivers were doing.
The Statistical Modeling discussion also highlights an independent replication and extension by Gaurav Sood. That analysis reproduced an elevated effect in the original period but found no comparable out-of-sample effect across seven major 2023 and 2024 releases, along with no positive relationship between how heavily an album streamed and the size of its fatality estimate.
That analysis isn't a peer-reviewed refutation of the published study. It is, however, another reason to distinguish a replicated statistical pattern within the original sample from a settled causal explanation.
Coincide with and cause describe two different claims, and the evidence reviewed here backs the first far more solidly than the second.
What this means for drivers and streaming platforms
If the association survives further scrutiny, it reframes an album release, fundamentally a marketing and cultural event, as a narrow road-safety question rooted in smartphone engagement rather than in the music itself.
That's a distinction platforms would need to act on directly. The problem, if it holds up, wouldn't be songs; it would be how often people interact with a phone while driving on those particular days.
The size of the reported gap, 18.2 additional fatalities averaged across the ten biggest release days studied, is large enough to warrant attention even before the causal question is settled.
Streaming platforms and safety advocates could explore stronger driving-mode defaults, playback that doesn't require touching a screen, or delaying push notifications during commute hours on release dates.
None of that has been tested by the research reviewed here. These are proposals, not conclusions, and nothing in the current data measures whether any of them would actually reduce fatalities.
What exists is an association between release dates and higher fatality counts, not evidence about what fixes it.
What would test the mechanism
The study's reported finding is narrower than the headline number suggests: fatalities were higher, in a statistically distinguishable way, on major album-release days measured in the study.
Turning that into a stronger case for the proposed mechanism, rather than a repeated headline, requires evidence beyond the aggregate release-day pattern.
A few designs would move the question forward:
A larger out-of-sample set of release dates, particularly releases after 2022, would show whether the pattern holds beyond the period used for the primary study.
Crash-level records noting phone use at the time of a collision would connect aggregate streaming data to actual driver behavior, the link the current design can't establish on its own.
A breakdown of fatalities by time of day would show whether deaths cluster around release-day listening peaks or spread evenly across the day, pointing toward different explanations either way.
Further independent replication, ideally preregistered or peer-reviewed, would show whether the 15.1% figure remains stable when other researchers build the comparison from scratch.
Until more evidence like this arrives, the fair description of what happened on those release days is narrower than some headlines suggest: streaming and traffic deaths rose in tandem, with a statistically significant release-day gap in the published analysis.
Turning that pattern into a verdict on smartphones and driving requires stronger evidence about the mechanism itself.
Comments
Be the first, drop a comment!