Death by Pokémon GO: The Economic and Human Cost of Using Apps While Driving

• Author: Mara Faccio,John J. McConnell
• Date: 01 September 2020
• DOI: http://doi.org/10.1111/jori.12301
• Published date: 01 September 2020

© 2019 American Risk and Insurance Association. Vol. 87, No. 3, 815–849 (2020).

DOI: 10.1111/jori.12301

DEATH BY POKÉMON GO: THE ECONOMIC AND HUMAN

COST OF USING APPS WHILE DRIVING

Mara Faccio

John J. McConnell*

ABSTRACT

We investigate the link between smartphone usage by drivers and increases

in insurance premiums by using police accident reports and exploiting the

introduction of the augmented reality game Pokémon GO as a natural

experiment. We document a disproportionate increase in crashes, vehicular

damage, injuries, and fatalities in the vicinity of locations where users can

play the game while driving. Ignoring incremental lives lost, the lower

bound estimate of the cost of users playing the game while driving trans-

lates to an increase in insurance premiums of 2.47 percent.

INTRODUCTION

In the recent decade, smartphones and their applications or “apps,”as they are

popularly known, have become ubiquitous as have concerns with their usage.

The concern that we investigate in this study is the allegation that the use of

smartphone apps while driving has given rise to an increase in the cost of vehicular

*Mara Faccio and John J. McConnell are at the Krannert School of Management, Purdue

University. Faccio is also affiliated with the Asian Bureau of Finance and Economics Research

(ABFER); the European Corporate Governance Institute (ECGI); and the National Bureau of

Economic Research (NBER). McConnell can be contacted via e‐mail: mcconnj@purdue.edu.

We thank Chief Jason Dombkowski of the West Lafayette Police Department, Captain Brad

Bishop of the Lafayette Police Department, Chief John Cox of the Purdue Police, and Sheriff

Barry Richard of the Tippecanoe County Sheriff’s Department for permitting access to police

accident reports. We are especially thankful to Lafayette Police Department crime analyst

Steve Hawthorne who provided the police accident reports data and patiently answered our

questions. We thank Stuart Graham of CyanSub for providing the latitude and longitude of

all PokéStops and Gyms in Tippecanoe County. We thank Joan Schmit (the editor), two

anonymous reviewers, John Barrios, Maria Boutchkova, Michael Brennan, Jillian Carr,

Shrijata Chattopadhyay, Logan Emery, Susan Lu, Maria‐Teresa Marchica, Steve Martin,

David Parsley, Donald Teder, Joel Waldfogel, and seminar participants at the 5th Annual

Magnolia Finance Conference at Mississippi State University, at the 2018 Dawn or Doom

Conference at Purdue University, and at the 2018 NBER’s Summer Institute IT and Digiti-

zation Workshop for helpful comments and suggestions. We also thank Adam Lawson and

Nate Hitze of Purdue University for developing the parsing program used to extract street

intersections data used in the Online Appendix.

815

insurance due to increases in vehicular crashes and related vehicular damage,

injuries, and fatalities.

As reported in the Wall Street Journal, insurers cite a possible connection between

smartphone usage and vehicular crashes as one explanation for the 16 percent in-

crease in U.S. automobile insurance premiums between 2011 and 2016 (Scism and

Friedman, 2017). Certain circumstantial evidence supports such a connection. To

wit, following a steady, though not uninterrupted, 25‐year decline, vehicular fatality

crashes in the U.S. reversed course in 2011. In 1988, fatality crashes totaled 42,130. In

2011, that number reached a low of 29,867. By 2016, the total had increased to 34,439

(NHTSA, various years). That reversal has been widely reported and commented

upon. Less well reported is that total vehicular crashes followed a similar course

with 6.887 million reported in 1988 falling to a low of 5.338 million in 2011, and

reversing course to reach 7.277 million in 2016.

The supporting circumstantial evidence is that diffusion of smartphone apps has

increased in parallel with the incidence of vehicular fatalities and crashes. Ac-

cording to Wikipedia, in 2008, Apple’s App Store had available 800 smartphone

apps with 10 million downloads. By 2011, those numbers were 500,000 apps and 18

billion downloads, and by 2017, there were 2.2 million apps and 130 billion

downloads.

1

Of course, attributing any increase in crashes and fatalities to smart-

phone usage and app availability is extraordinarily difficult given that many other

factors also changed over the years in which both increased. The same WSJ article

cited above notes that insurers said “[t]he rise in traffic deaths is the result of many

factors. Low gas prices and a U.S. economic recovery combined to put more drivers

on the road, giving drivers less time to react if they are distracted by their smart-

phones…” Other factors cited include “reaching for a dropped item, attending to

children in a back seat or eating.”

In this study, we circumvent the difficulty of isolating the effect of app usage by

examining the introduction of a specific app that can be associated with specific

geographic locations. The app is the highly popular Pokémon GO augmented reality

game. The game was introduced on July 6, 2016. Within 1 month, worldwide, the

game was downloaded more than 100 million times. For our purposes, the virtue of

the app is that the stockpile of a participant’s“weapons”used to play the game can

be replenished in the vicinity of specific well‐identified “PokéStops”many of which

are located near public thoroughfares. This unique feature of the app enables us to

investigate the link between playing while driving and crashes in a well‐identified

setting. If the game is played while a player is driving and if playing the game while

driving increases the likelihood of crashes occurring, locations near PokéStops

should experience a disproportionate increase in crashes following introduction of

the game.

2

1

https://en.wikipedia.org/wiki/App_Store_(iOS). A similar trend is documented for the

other large digital distribution platform, Google Play (https://en.wikipedia.org/wiki/

Google_Play).

2

Anecdotally, a link between the introduction of Pokémon GO and crashes has been reported

in media outlets including the Wall Street Journal,“Pokémon Go‐Related Car Crash Kills

816 JOURNAL OF RISK AND INSURANCE

To conduct our analysis, we examine nearly 12,000 police accident reports for

Tippecanoe County, Indiana, for the period of March 1, 2015, through November 30,

2016. We find a disproportionate increase in crashes near PokéStops from before to

after July 6, 2016. In the aggregate, these crashes are associated with increases in the

dollar amount of vehicular damage, the number of personal injuries, and the

number of fatalities.

Specifically, a difference‐in‐differences analysis indicates that 136 incremental cra-

shes that occurred at locations in the proximity of PokéStops over the 148 days

following the introduction of the game can be attributed to the introduction of

Pokémon GO. This incremental increase in crashes accounts for 48 percent of the

increase in the total number of county‐wide crashes. Based on the assessments of

damage in the police reports, this increase in crashes in the vicinity of PokéStops

results in lower bound estimate of the incremental vehicular damage of $334,145

over the 148 days following the introduction of the game.

The 136 incremental crashes give rise to 33 incremental personal injuries in the

vicinity of PokéStops. These account for 27 percent of the aggregate increase in the

number of personal injuries experienced county‐wide over the same time period.

Based on data from the Insurance Information Institute and the Centers for Disease

Control and Prevention (CDC), a lower bound estimate of the incremental claims for

bodily injuries and the loss of lifetime income imply a county‐wide total incremental

cost of $892,155.

On a sadder note, our analyses point to two incremental vehicular fatalities in the

vicinity of PokéStops following the introduction of the game, resulting in a lower

bound estimate of the insurance payout for fatalities of $482,772.

These numbers result in a lower bound estimate of the implications for insurance

premiums due to users playing Pokémon GO in the vicinity of PokéStops of $1.71

million over the 148 days following introduction of the game. A nontrivial fraction

of this total is the value of lives lost. Regardless of whether that number is included,

the incremental payouts associated with users playing Pokémon GO while driving

are significant. Indeed, even ignoring the value of lives lost, the lower bound esti-

mate of the incremental payouts associated with users playing the game while

driving imply a 2.47 percent increase in auto‐insurance premiums.

Our interpretation of regression results may give rise to two broad concerns. First, it

could be that the observed increase in crashes is not due to the introduction of

Pokémon GO but to another shock that correlates with the introduction of the game

and vehicular crashes. In that regard, one concern specific to our analysis is that

Tippecanoe County is the home of Purdue University, a 40,000‐student body uni-

versity in a county with a “fulltime”population of 188,000 (i.e., excluding students).

Woman in Japan”, August 25, 2016; USA Today,“Pokémon Go player crashes his car into a

tree”, July 14, 2016; Fox News,“Death by Pokemon? Public safety fears mount as Pokemon

GO craze continues”;The Guardian,“Pokémon Go player crashes car into university while

playing game,”July 18, 2016. A study by Ayers et al. (2016)reports that a nontrivial number

of users are characterized in tweets as playing the game while driving.

DEATH BYPOKÉMON GO817