668. Do Taylor Swift and Bad Bunny Have Blood on Their Hands?

Summary of 668. Do Taylor Swift and Bad Bunny Have Blood on Their Hands?

by Freakonomics Radio + Stitcher

53m•March 27, 2026
Listen to Episode

Overview of 668. Do Taylor Swift and Bad Bunny Have Blood on Their Hands?

This Freakonomics Radio episode summarizes a new NBER working paper, "Smartphones, Online Music Streaming and Traffic Fatalities," by Bapu Jena, Vishal Patel, Chris Worsham and Michael Liu. The researchers treat big album-release days as natural experiments to ask whether spikes in smartphone-based music streaming translate into more fatal car crashes. They find a notable short-term increase in traffic deaths on the biggest album-release days, explore robustness checks and mechanisms, discuss limits of available data, and argue the finding reveals a broader research and policy gap around distracted-driving data.

Key findings

  • Dataset: FARS (Fatality Analysis Reporting System), primarily 2017–2022.
  • Treatment: the 10 album-release dates (2017–2022) with the largest first-day Spotify streams (artists include Taylor Swift, Bad Bunny, Drake, Kendrick Lamar).
  • Streaming: first-day streaming volume on those dates rose ≈40%.
  • Fatalities: a ≈15% relative increase in U.S. traffic fatalities on those release days — roughly +18 deaths on each such day (base ≈130 deaths/day).
  • Heterogeneity:
    • Larger effect among younger drivers (who stream more).
    • Larger effect when the driver is alone; smaller when there is a passenger (passenger acts as partial “insurance”).
    • Effect not concentrated only at night and not primarily tied to alcohol-involved crashes.
    • Greater signal in vehicles with Apple CarPlay (hypothesized lowered barrier to interaction).

Methods & data

  • Natural-experiment framing: album-release days produce sudden, exogenous increases in streaming attention.
  • Main comparison: each album-release date vs. the same weekday in the week before and after (controls for weekday patterns, e.g., Friday effects).
  • Falsification and robustness checks:
    • Random placebo dates and random Friday-placebo tests (1,000 iterations).
    • Applying the calendar date of a release to other years without a release (no spike on those years).
    • Stratified analyses by age, alcohol involvement, time of day, and presence of passengers.
  • Data limits: FARS is public but lags (2023 availability arrived later); researchers used only public data (intentional reproducibility).

Robustness, replication and caveats

  • Replication attempts: a producer’s quick re-run with 2023 FARS data initially showed the effect disappearing when 2023 album releases were included.
    • Authors’ explanation: multiple top-release days in 2023 occurred on consecutive Fridays, making treatment and control days similar (i.e., comparing big streaming days to other big streaming days). When restricting to non-overlapping top albums through 2017–2023, the pattern reappears.
  • Limitations and alternative explanations:
    • Cannot directly observe driver smartphone interactions or infotainment logs in FARS — the mechanism (phone fiddling, volume masking cues) remains inferred.
    • Possible unobserved confounders (weather, local events) were not exhaustively ruled out; authors note weather checks could be added.
    • Small absolute impact relative to total annual U.S. traffic deaths — high relative spike on a given day, but limited aggregate contribution.
    • Proprietary telematics and infotainment data (car manufacturers, platform providers) largely unavailable to researchers.

Proposed mechanisms and interpretation

  • Two leading mechanisms:
    1. Active manipulation: drivers touch phones or infotainment (CarPlay/Android Auto may lower action cost and thus increase fiddling).
    2. Sensory masking: louder/novel music reduces auditory cues (horns, sirens) that alert drivers to danger.
  • Passenger presence reduces the effect, consistent with reduced need for the driver to operate the device.
  • Authors see this work as evidence that smartphone-car interactions can be dangerous, but not a call to restrict music releases — rather a flag for better data and vehicle design.

Policy implications & recommendations

  • Authors are not advocating bans on album releases or artist liability.
  • Practical suggestions:
    • Better access to telematics/infotainment data for independent researchers (while balancing privacy concerns) to move beyond inference and identify exact behaviors preceding crashes.
    • Continued vehicle safety improvements and broader adoption of autonomous/semi-autonomous features (authors expect distraction effects to decline as cars get safer).
    • Consideration of design choices in in-car interfaces (e.g., limiting easy access to certain functions while driving).
  • Authors deem the small absolute number of deaths insufficient to justify heavy-handed regulatory action targeted at album releases alone.

Notable quotes and insights

  • Donald Redelmeier: “A moment of inattention can cause a fatal traffic crash…exciting new music can be enchanting to some listeners and indirectly causes some to potentially lower their guard.”
  • Bapu Jena (on a broader solution): “Bring on the frigging autonomous cars already,” reflecting a view that automation may reduce human-driven distraction harms.
  • Vishal Patel (core takeaway): “It’s really all about distraction — understanding how new technologies interact with human behavior on the road is going to be one of the most important traffic safety questions for decades to come.”

Actionable next steps and research agenda

  • Obtain and analyze fine-grained telematics and infotainment logs (vehicle/system-level timestamps of media interaction, phone usage, steering/braking patterns).
  • Partner with automakers, infotainment platforms, and insurers to gain access to proprietary data under privacy-protecting arrangements.
  • Replicate and extend the analysis:
    • Use non-overlapping top-release lists across broader years.
    • Test dose–response (does streaming-volume magnitude correlate with crash magnitude?).
    • Check weather, local events, and spatial patterns (where crashes occur relative to highways, urban streets, theaters).
  • Study specific interface features (CarPlay vs. Android Auto vs. native systems) and the effect of passengers.
  • Monitor trends across 2023–2025 to see if vehicle automation or changing release practices alter the effect.

Bottom line

The study finds a credible short-term association between extremely high-profile album-release days and spikes in traffic fatalities, consistent with the idea that smartphone-infused music consumption can increase dangerous inattention while driving. The result is informative about distraction risks but limited by data access: it points to a larger, addressable problem — researchers need richer telematics/infotainment data to identify exact causal behaviors and to guide more targeted policy or design changes.