ArticlePDF Available

Do Traffic Tickets Reduce Motor Vehicle Accidents? Evidence from a Natural Experiment

Authors:

Abstract and Figures

This paper analyzes the effect of traffic tickets on motor vehicle accidents. OLS estimates may be upward-biased because police officers tend to focus on areas and periods with heavy traffic and thus higher rates of accidents. This paper exploits the dramatic increase in tickets during the Click-it-or-Ticket campaign to identify the causal impact of tickets on accidents using data from Massachusetts. I find that tickets significantly reduce accidents and non-fatal injuries. However, there is limited evidence that tickets lead to fewer fatalities. I provide suggestive evidence that tickets have a larger impact at night and on female drivers.
Content may be subject to copyright.
1
Do Traffic Tickets Reduce Motor Vehicle Accidents?
Evidence from a Natural Experiment
Dara N. Lee §
University of Missouri-Columbia
January 2012
Abstract
This paper analyzes the effect of traffic tickets on motor vehicle accidents.
OLS estimates may be upward-biased because police officers tend to focus
on areas where and periods when there is heavy traffic and thus higher rates
of accidents. This paper exploits the dramatic increase in tickets during the
Click-it-or-Ticket campaign to identify the causal impact of tickets on
accidents using data from Massachusetts. I find that tickets significantly
reduce accidents and non-fatal injuries. However, there is limited evidence
that tickets lead to fewer fatalities. I provide suggestive evidence that tickets
have a larger impact at night and on female drivers.
JEL Codes: K32, K42, R41
_____________________
§ Email: leedn@missouri.edu. Address: Department of Economics, University of Missouri-
Columbia, 226 Professional Building, 909 University Avenue, Columbia, MO 65211.
2
I. Introduction
Reducing motor vehicle accidents is a key concern for health policy makers. Motor
vehicle accidents cause more than 40,000 deaths and several million injuries each year, and are
also the leading cause of death among children in the United States (Center for Disease Control
2009). While a large body of literature examines the impact of regulations and technological
innovations, such as seat belts, airbags, and child safety seats (Braver et al., 1997; Carpenter and
Stehr 2008; Levitt 2008), there has been considerably less work on the effect of traffic law
enforcement. However, addressing the question is complicated by the issue of reverse causality
more police officers are stationed at areas and during periods with higher rates of traffic
accidents, in which case OLS estimates of accidents on tickets may be upward biased. Figure 1,
which depicts a scatter point graph of the daily average number of motor vehicle accidents
against tickets using data from Massachusetts, reinforces this concern at first glance, there is no
discernible relationship between the two variables.
Further, while the ostensible goal of traffic tickets is to deter reckless driving, recent
literature demonstrates that traffic tickets are often used as a tool to generate revenue for local
municipality budgets (Makowsky and Stratmann 2008; Garrett and Wagner 2009).
1
There is also
compelling evidence that police officers are influenced by personal preferences when giving out
tickets (Anbarci and Lee 2008; Antonovics and Knight 2009).
2
It is thus unclear whether tickets
would fulfill their intended purpose of improving road safety.
1
Makowsky and Stratmann (2008) show that police officers in towns which are more budget-strapped are more
likely to issue a ticket than a warning. Garrett and Wagner (2009) find that significantly more tickets are issued in
counties the year following a decline in revenue.
2
Anbarci and Lee (2008) find that minority officers, particularly African-Americans, are harsher on all motorists but
even harsher on minority motorists. Antonivics and Knight (2009) find evidence for preference-based discrimination
3
The limited literature addressing the impact of traffic law enforcement on road safety has
produced mixed results. Carr, Schnelle and Kirchner (1980) use the Nashville traffic police strike
in 1978 as a natural experiment and report no significant deterioration in road safety during the
strike. Conversely, Beenstock et al. (1999) use panel data on road sections in Israel and find
some evidence that large-scale enforcement reduces road accidents. Babcock et al. (2008) show a
negative correlation between police and accidents in Kansas. Makowsky and Stratmann (2011)
use the financial health of town as an instrument for the number of tickets issued to demonstrate
that more tickets lead to fewer accidents. However, most of the existing studies rely on monthly
data of the number of policemen as the measure of traffic law enforcement, which may not be
accurate because police have other duties besides patrolling traffic. Further, monthly data may
obscure the sensitivity of the behavior of drivers to traffic tickets.
This paper exploits exogenous variation in the number of tickets issued to identify the
causal impact of traffic tickets on motor vehicle accidents using daily municipality-level data
from Massachusetts. In the fall of 2002, Massachusetts participated for the first time in the Click-
it-or-Ticket (CIOT) program, a federal program that was initiated and funded by the National
Highway Traffic Safety Administration (NHTSA). The program is carried out throughout the
federal year through six “mobilizations” of one to two week periods, during which designated
police officers specifically and aggressively focus on traffic law enforcement. Although the
program focuses on seat belt use, Massachusetts had a secondary seat belt law until 2007, which
meant that while drivers and passengers were required to wear seat belts, police cannot pull them
over solely for failing to wear a seat belt. The motorist would have to be committing another
among police officers. Their results demonstrate that officers are more likely to search if officer race and driver race
differ.
4
traffic offense, such as speeding, in order for the police to have sufficient grounds for pulling a
motorist over. Thus, I argue that the main impact of the CIOT campaign is to increase the degree
of traffic law enforcement overall, and serves as a natural experiment to examine the impact of
tickets on motor vehicle accidents.
There were two state-wide mobilizations that took place in Massachusetts during
November and December of 2002. The November mobilization was held the two weeks
surrounding Thanksgiving and the December mobilization was carried out the week before
Christmas. However, these periods were presumably chosen because there are higher traffic
volumes (and therefore potentially higher rates of accidents) near holiday periods. To address the
concern of endogeneity, I control for the periods that would presumably have been chosen for
mobilization in 2001 had the campaign taken place as comparison. After controlling for these
periods in 2001, time and municipality fixed effects, as well as a host of other variables, the
increase in tickets during the actual mobilization periods in 2002 is then arguably exogenous to
accidents. Both results from reduced form regressions and using the mobilizations as an
instrument for tickets suggest that there is a negative and significant relationship between tickets
and accidents. The estimated accident elasticity with respect to tickets is approximately -0.28 a
1 percent increase in tickets issued leads to a 0.28 percent decline in motor vehicle accidents.
The elasticity of non-fatal injuries with respect to tickets is smaller at -0.17; there is no
discernible impact on fatalities. Further, I show the reduction in accidents is higher in
municipalities that issued more tickets during the mobilizations, which provides evidence tickets
per se are reducing accidents rather than other concurrent factors. The main results are robust to
a number of different specification checks, which I discuss in Section V. Finally, I explore when
tickets are more effective and who tickets affect most in Section VI, which could help inform
5
policy makers on how to allocate enforcement in order to achieve the highest impact. I provide
suggestive evidence that tickets have a larger impact at night, and a larger effect on female
drivers. However, tickets do not appear to differentially affect age groups.
The rest of the paper is organized as follows. Section II describes the CIOT program and
using data. Section III presents the identification strategy. Section IV reports and discusses the
main results. Section V offers several robustness checks. Section VI explores heterogeneous
effects of tickets, and Section VII concludes.
II. Background and Data
II.A. The Click-it-or-Ticket (CIOT) Program
The CIOT program was first conceived in North Carolina in 1993, and by 2001 the
program had spread to a number of other states in the region. (The program is still in existence
today.) The initial program combined 3,000 enforcement checkpoints and paid advertising to
build public awareness about seat belt use. In 2002, eighteen states participated in a national
CIOT campaign pilot program, including Massachusetts (Tison et al. 2006).
In Massachusetts, the CIOT campaign is organized as following. Local police agencies
are given grants according to their population size, and police officers apply for overtime to work
during the campaign mobilizations. During the mobilizations, police officers focus exclusively
on traffic law enforcement and did not have to respond to any other calls. Further, they are
expected to give out a certain number of tickets per shift, although they are not penalized if they
do not attain the number. Since the campaign focuses on safety belt use, officers target drivers
who were not using seat belts. However, Massachusetts had a secondary seat belt law until 2007,
meaning drivers and passengers were required to wear seat belts but police cannot pull them over
solely for failing to wear a seat belt. Officers must have another reason to pull over a car, such as
6
speeding, and then may issue a $25 citation to the driver and passengers who are not wearing
seat belts. Therefore, it is not just tickets issued for seat belt violations that rose in number during
the mobilization periods, but tickets of all types of offenses (as described in more detail in
Section II.C). In addition to increased traffic law enforcement, Massachusetts engaged in a state-
wide media campaign to publicize the CIOT program via radio and television advertisements.
The CIOT program, combined with the fact that Massachusetts still had a secondary seatbelt law
in 2002, provides the unique setting to examine the impact of traffic citations on motor vehicle
accidents.
II.B. Data Sources and Description
The data used in this paper are drawn from two main sources. The tickets data include all
traffic tickets issued in the state from April 1, 2001, through January 31, 2003, covering all 350
municipalities. The database has detailed information on the offense type, ticket amount, location
and time, demographic data about the driver, as well as the model and make of the car. The
particular beginning and end dates of the sample period are exogenous to this study the tickets
data are kindly shared by Bill Dedman and Francie Latour, who first collected the extensive data
from the Massachusetts Registry of Motor Vehicles for a series of Boston Globe articles
examining ticketing behavior and racial profiling in Massachusetts. The authors were not aware
of the CIOT mobilizations.
The motor vehicle accident data are from the Massachusetts Highway Safety Division
and includes all accidents that were reported to the Registry of Motor Vehicles for the same
period. Each crash includes information on date, time, location, number of vehicles involved, and
crash severity. I also obtain additional demographic data from the Highway Safety Division
regarding the age and gender of the drivers involved in each accident. The demographic data are
7
available only from 2002 onwards. I then link the two datasets by municipality and date and
combine them to create a panel dataset describing the number of tickets issued and accidents for
each day in the sample period, for each municipality. The unit of observation is thus at the
date/municipality level.
In addition to the tickets and accidents data, I collect daily weather information and gas
data, which could potentially affect both traffic volume and accidents. The weather data includes
daily precipitation, snowfall, snow depth on the ground, and mean temperatures for each
municipality based on the closest weather station. There are altogether 57 weather stations which
collected daily weather data for the relevant data period. I link the weather of each municipality
to the closest weather station. Gas data consists of retail prices for regular unleaded gas for each
municipality for each day in the sample period.
Finally, I include year-varying municipality level data that are intended to capture
characteristics that could be correlated with both traffic volume and accidents, such as the
municipality unemployment rate, population, number of registered motor vehicles per capita,
luxury cars per capita, trucks per capita, motorcycles per capita, and average car age.
II.C. Summary Statistics
Panel A of Table 1 reports the overall summary statistics. On average, there are 4.8
tickets issued and 0.9 motor vehicle accidents per 100 miles of public road length. There are
0.381 nonfatal injuries per 100 road miles and there are close to zero fatalities due to motor
vehicle crashes. The measures of tickets and accidents are adjusted per length of public road in
each municipality to account for the different sizes (and also populations) of municipalities.
Figure 2 depicts the number of tickets and accidents by week for the sample period. There is a
8
large and distinct spike in tickets for both the highlighted areas in 2002 relative to the same
periods in 2001, and relative to all other periods.
Panel B of Table 1 presents the summary statistics of the main dependent variables by
different groups time of day, gender, and age group. Unsurprisingly, there are more tickets and
accidents in the day because of higher traffic volumes. Males are involved in more accidents than
females, which could reflect both that there are fewer female drivers and gender differences in
driving behavior. There are also more tickets and accidents among the 25-64 age group.
3
III. Identification Strategy
The identification strategy of this paper uses the sharp increase of tickets during the
CIOT mobilizations to examine the impact on accidents. It also rests on the assumption that the
increase in tickets is exogenous, an assumption that I argue in this section is reasonable
controlling for other factors. The first CIOT mobilization occurred from November 18, 2002
until December 1, 2002, and Thanksgiving Day was on November 28. The concern is that these
periods were chosen because there are high traffic volumes surrounding these holidays, in which
case any changes in tickets or accidents could simply be an artifact of increased traffic volume.
To address this issue, I control for the comparable periods in 2001. More specifically, if the
CIOT campaign took place in 2001, the mobilization period would presumably have been chosen
in the same way the week of Thanksgiving and the week before. Since Thanksgiving was on
November 22 in 2001, I define the comparable period to be November 12-25, 2001. The second
mobilization in 2002 occurred from December 16-22. I similarly define the comparison period in
2001 to be December 17- December 23. The comparison periods are defined to begin and end on
3
Adjusted for population, however, the accident rate of the 15-24 age group is twice as large as that of the 25-64 age
group.
9
the same day of week, in order to prevent any bias from having unequal number of weekdays and
weekends. For lack of a better term, I call the following weeks Treatment Periods: November
12-25, 2001, December 17-23, 2001, November 18-December 1, 2002, and December 16-22,
2002. Treatment Periods x Year2002 is then the equivalent of the actual CIOT mobilizations.
The control periods are all other days in the sample period.
The identification strategy would also be violated if other factors not controlled for
systematically correlated with the mobilization periods. For example, the mobilizations in 2002
could be systematically correlated with inclement weather or shocks to gas prices, which could
affect traffic volume and in turn both the number of tickets and crashes. I therefore control for
weather conditions and retail unleaded gas prices as well.
Another issue to consider is that using the CIOT mobilizations to capture the overall
impact of tickets may not be appropriate if only tickets for seat belt violations for given out.
However, as discussed earlier, Massachusetts did not have a primary seat belt law until 2007,
which means that before 2007, law enforcement officers may issue a ticket for not wearing a seat
belt only when there is another citable traffic infraction. Figure 3 presents the overall ten most
frequent traffic violations over the control and mobilization periods. In the control sample,
speeding tickets are the most common, followed by seat belt violations and failure to stop.
During the CIOT mobilization periods, seat belt violations surpass speeding tickets as the most
common type of ticket, but there is also a uniform rise of almost all types of offenses. The types
of offenses that tickets are issued for are also the same for eight of the ten most frequent offenses
in the control periods. It therefore seems reasonable to interpret the mobilization periods led to
overall stricter traffic law enforcement as measured by tickets.
10
IV. Results
IV.A. OLS Results
According to the basic model in Becker’s seminal paper (1968) on crime and punishment,
an individual makes the decision of whether to commit a crime or not by weighing the expected
costs against the expected benefits. In the context of this paper, the degree of traffic law
enforcement increases dramatically during the CIOT mobilizations, which can be translated into
a higher probability of getting caught and thus a higher expected cost of committing a traffic
violation. We should therefore expect to see fewer traffic violations and potentially fewer
accidents, i.e., we would expect to carry a negative sign if we model the relationship between
accidents and tickets as such:
      
where and are municipality and calendar fixed effects, which include day-of-week, month-
by-year, and public holidays (and the contiguous weekend if the holiday falls on a Friday or
Monday) fixed effects.  represents a vector of controls that may influence both traffic volume
and accidents, which include the amount of snowfall, snow depth on ground, rainfall,
temperature, retail gas prices, population, unemployment rate, registered cars per 100 miles,
trucks per 100 miles, trailers per 100 miles, and motorcycles per 100 miles. The weather
conditions and gas prices variables vary at the day-municipality level; the rest vary at the
calendar year-municipality or fiscal year-municipality level.
However, OLS estimates of equation (1) would lead to biased estimates if there is a
correlation between  and . This is probable because police officers tend to focus on
areas where and periods when accidents are prone to happen. If so, estimates of will be
upward-biased. Table 2 reports the OLS estimates from regressing accidents on tickets directly.
11
The results in Table 2 confirm what we have already seen in Figure 1 the coefficients on tickets
are close to zero and in fact slightly positive, suggesting that the estimates are indeed upward-
biased.
IV.B. First Stage Results
The identification strategy proposed in this paper is to use the CIOT periods as a single
instrument for tickets. To be valid instrument, the CIOT periods should be able to demonstrate
strong explanatory power of tickets (Bound et al., 1995). The first stage regression representing
the relationship between tickets issued in town i on date t and the CIOT periods is modeled as
follows:
       
where and are again municipality and calendar fixed effects.  is a
dummy representing the treatment weeks defined in Section III.  indicates when the CIOT
mobilizations took place in 2002, and is also equivalent to 
. is the same vector of controls as before. thus captures the rise in tickets
during the CIOT mobilizations in 2002.
Column 1 of Table 3 shows that the CIOT mobilizations increased on average the number
of tickets issued per 100 miles by 1.9 per day/municipality. Given the mean of 4.8, the change
represents approximately a 40% increase. The F-statistic is approximately 100. The first-stage
relationship between the CIOT periods and tickets is clearly strong.
IV.C. Reduced Form Evidence
Table 3 (Columns 2-4) reports the results from estimating the reduced-form relationship
between the number of accidents and the CIOT mobilization program:
        
12
Using accidents as the dependent variable (Columns 2 4), the results suggest that the
reduced form impact of the campaign on motor vehicle accidents was around -0.10, which would
translate into a 11 percent reduction. It appears that the number of accidents fell significantly
during the mobilization periods, relative to both the comparable periods in 2001 and the control
periods. Injuries went down during the CIOT periods as well, but the reduction in reduced form
is not statistically different from zero at conventional levels. The estimate on fatalities is very
imprecisely estimated. This is perhaps unsurprising given the low incidence of fatalities overall.
IV.D. IV Results
The main results of the paper, which are the estimates from the IV regressions, are
documented in Table 4. The coefficient on tickets in Column 1 implies that a unit increase in
tickets decreases accidents by 0.05. Since there are on average 4.8 tickets issued and 0.88
accidents in a municipality, this implies the elasticity of accidents with respect to tickets is
approximately -0.28 a 1 percent increase in tickets leads to a 0.28 percent decrease in motor
vehicle accidents.
4
The estimates shown in Table 4 are those produced when using the full set of
controls, but the estimates are stable across different specifications.
Next, I examine whether tickets have any impacts on nonfatal injuries and fatalities
caused by motor vehicle accidents. Column 2 documents the result from using the number of
nonfatal injuries as the dependent variable. For injuries, the IV estimate is significant at the 10
percent level. If we take the estimate at face value, the elasticity of injuries with respect to tickets
would be -0.17. Column 3 presents the coefficient on tickets for fatalities. Neither the reduced
form nor IV estimates support that tickets reduce fatalities.
4
All results using the level numbers of tickets and accidents (instead of rates per 100 miles) yield similar results and
can be obtained by request.
13
IV.E. Mechanism: Tickets versus Information
There are two main mechanisms that could lead to the observed result of fewer accidents
during the CIOT mobilizations. The first is a deterrence effect, both for drivers who receive a
ticket and those who observe other drivers receiving a ticket. For example, giving a ticket to a
driver who was going 80 mph in a 55 mph road could deter the recipient from further speeding,
and thus be less likely to be involved in an accident. There could also be a visual deterrence
effect, for example, drivers observe another car being pulled over for speeding and therefore
slow down and drive more carefully.
Second, it is possible that tickets per se are not driving the reduction in accidents. As
noted in Section IIA, there was a concurrent state-wide media campaign during the CIOT
periods.
5
While the media campaign would only be credible if there were an actual increase in
ticketing, the effect of tickets found in Section IV would be overstated if the media campaign
was the main mechanism through which accidents were reduced.
I argue that the main mechanism was indeed through tickets and not through the spread
of information through the media campaign (or other related channels, such as through the social
networks of ticketed drivers). If the spread of information about the CIOT mobilizations was the
main mechanism, the reduction in accidents should not vary systematically by how many tickets
were written in a particular town. In other words, if ticketing mattered, then it should have had a
larger effect in the towns that issued more tickets. In Panel A of Table 5, I include an interaction
term of CIOT and Active Ticketing, a binary variable which I define to be 1 if the average
number of tickets issued daily during the CIOT periods exceed the average number of tickets in
5
It is to my understanding from discussions with the Massachusetts Highway Safety Division that the main
component of the CIOT program, particularly in its inception year, was the increase in traffic law enforcement.
14
the control periods. This seems to be a reasonable proxy for active traffic enforcement during the
mobilizations. By construction, the coefficient on the interaction term CIOT*Active Ticketing for
Tickets is much larger than and statistically different from the coefficient on CIOT. Of more
interest are the coefficients on the interaction term of CIOT*Active Ticketing for Accidents and
Injuries (Columns 2 and 3) they are both statistically different from zero and from the CIOT
term. In municipalities where the main exposure to the CIOT program is through the media
campaign and not through increased ticketing, we do not observe a significant decrease in
accidents or injuries, whereas the opposite is true for municipalities that actively ticketed. This
helps rule out the hypothesis that the media campaign is driving the results. IV regressions
(Table 5 Panel B) using only the participating municipalities yield similar but even more
precisely estimated coefficients to using the entire sample.
IV.F. Comparison of Results to Existing Literature
How do these results compare with other mechanisms that aim to improve road safety?
Carpenter and Stehr (2008) look at the effect of mandatory seatbelt laws, and find that they
significantly reduced traffic fatalities and serious injuries resulting from fatal crashes by 8 and 9
percent, respectively. Levitt (2006) finds that child safety seats are no better than seat belts at
reducing fatalities among children aged 2-6. Dee et al. (2005) demonstrate that graduated driver
licensing restrictions for teens reduced traffic fatalities among 1517-year-olds by at least 5.6%.
However, these interventions focus more on traffic regulations rather than traffic law
enforcement. By contrast, there has been little direct investigation of the effect on road safety of
traffic law enforcement. This paper contributes to the understanding of how traffic law
enforcement, as measured by tickets, affects road safety by offering a novel identification
strategy.
15
To my knowledge, Makowsky and Stratmann (2010) is the only other study that
examines the impact of traffic tickets on road safety. They use the financial health of town -
whether a town asks voters to approve a property tax override referendum as the instrument for
tickets. By putting an override referendum in front of voters, the town board indicates that the
town is in fiscal distress and that they would like to raise additional revenue.
6
They find
comparable results. Their IV estimates indicate that a unit increase in tickets leads to 0.12
(s.e.=0.034) fewer motor vehicle accidents, 0.044 (s.e.=0.022) injuries and 0.002 (s.e.=0.001)
fatalities. Their unit of observation is at the month-municipality level and the mean of tickets,
accidents, injuries are at 82.68, 36.93, and 15.83, respectively. Their results thus translate into an
elasticity of accidents with respect to tickets of -0.27, which is very close to the elasticity found
in this paper. They find a somewhat higher elasticity of injuries with respect to tickets of -0.23.
They also present some but inconclusive evidence that tickets reduce fatalities.
V. Robustness Checks
V.A. Using Only November and December Months
As there were only 3 weeks of mobilizations in 2002, a potential concern would be there
are factors in other months that are not controlled for that could confound the estimation. For
example, there is a spike in tickets in May 2002 relative to May 2001 (Figure 2). The reason for
this is because a number of the larger municipalities in Western MA participated in a trial CIOT
mobilization the week surrounding Memorial Day Weekend (Solomon et al. 2002). However, I
am unable to obtain data on which specific municipalities participated in the mobilization as the
6
Their instrument rests on the caveat that the only impact on fiscal distress on motor vehicle accidents is through the
number of tickets issued. However, one could imagine that the fiscal distress would affect accidents through many
other unobservables such as unemployment (and therefore less traffic), poorer road maintenance and a host of other
unobservable variables. The authors attempt to address these issues by including various controls and by using
different specifications.
16
Department of Highway Safety did not keep records of the particulars of the trial mobilization.
As a robustness check, I restrict the analysis to only the November and December months of
2001 and 2002.
The results are qualitatively similar to those using the entire sample period. Table 6 Panel
A summarizes the reduced form estimates of the CIOT mobilizations on tickets and accidents
using the full set of controls. The reduction in the number of injuries due to motor vehicle
accidents associated with the mobilization periods is now statistically significant at the 5 percent
level. Table 6 Panel B presents the IV estimates. The estimate on injuries implies that a one-unit
increase in tickets decreases accidents by 0.07 injuries. As before, there is no discernible impact
of tickets on fatalities. The results from using the restricted sample tell a consistent story: tickets
appear to reduce the overall number of accidents and injuries, but there is no support for tickets
being an effective tool in reducing fatalities.
V.B. Using Averages
Although the number of tickets varies across days and municipalities, the instrument
the CIOT mobilizations was a state-wide program and supposedly took place across the entire
state of Massachusetts, i.e., the variation in CIOT participation was by day (although there are
varying degrees of participation). While I group standard errors by municipality in all regressions,
one may be concerned that failing to account for intra-day correlation in errors could generate
standard errors that are biased downwards. I argue that it is more important to account for intra-
municipality correlation than intra-day correlation since Massachusetts has 350 municipalities
and hence distinct variation in weather and traffic conditions across the state. Nonetheless, as a
robustness check, I collapse the data by day into a simple time series to see if the results still hold.
Panel A of Table 7 contains the reduced form estimates and Panel B the IV results. The results
17
are consistent with those produced when using the full sample. Columns 4-8 of Table 7 Panel B
give the elasticity using logarithmic forms of accidents and tickets. The estimated elasticity is
around -0.35, which is also close to the elasticity (-0.28) calculated using the full sample.
VI. Heterogeneous Effects
When are tickets most effective in reducing accidents? What kind of driver do tickets
affect most? From a policy perspective, it is important to understand these issues in order to
efficiently allocate enforcement efforts. To my knowledge, this is the first paper to attempt to
explore heterogeneous effects of traffic law enforcement.
VI.A. By Time of Day
First, I examine the impact of tickets by time of day (Table 8 Panel A). The independent
variable of interest is the number of tickets issued during the day and the number of tickets
issued at night. The first stage relationships between the mobilizations and tickets issued during
daytime and nighttime are both strong (not shown), and the separate estimates during the day and
night are statistically different at the 5 percent level. The implied elasticity for accidents with
respect to tickets is close to -1 at night and is only around -0.16 during the day. This finding that
tickets have a larger impact at night seems to make sense despite 60 percent less traffic on the
roads, close to half of all fatal car accidents occur at night (Elliot 2009). Driving at night could
be potentially more dangerous because of poorer light conditions and more dangerous drivers
(for example, drunk or fatigued drivers) on the road.
VI.B. By Gender
The Massachusetts Highway Safety Department did not start collecting demographic data
of the drivers involved in motor vehicle accidents until the beginning of 2002, so unfortunately
the comparable treatment periods in 2001 cannot be included in the analyses by gender and age
18
groups. To ensure the results would not be completely discredited without using the control
periods in 2001, I run the basic IV regression of accidents on tickets using only data from 2002
onwards. The coefficient on tickets is slightly smaller at -0.369 (s.e. = 0.008), but not statistically
different from using the entire sample.
Since it is probably difficult for a driver to ascertain whether a fellow driver on the road
being pulled over is male or female, the main independent variable of interest is the overall
number of tickets (instead of the number of tickets issued to male and female drivers). Table 8
Panel B presents the coefficient on tickets using the number of male and female drivers involved
in motor vehicle accidents per 100 miles. While the actual coefficients for males and females are
close, the implied elasticities are very different. This is because females are involved in much
fewer accidents overall (presumably because there are fewer female drivers and/or they are more
careful drivers). The accident elasticity with respect to tickets for females is -0.72 and -0.24 for
males, a striking three-fold difference. Given the large literature within psychology and
sociology that show women tend to be more risk-adverse than men (see Croson and Gneezy
(2009) for a review), it is plausible that women are more deterred by traffic law enforcement
than men.
VI.C By Age Group
Finally, I investigate whether there are heterogeneous impacts of tickets by age groups in
Panel C of Table 6. This is of particular interest to policy makers because young drivers under 25
tend to be at most risk on the road (Center for Disease Control 2009). Again, because it would be
difficult for the driver to determine the age of another driver being pulled over, I use the overall
number of tickets as the main independent variable. The results imply that the elasticities with
respect to tickets are 0.76 and 0.62, for 15-24 year olds and 25-64 year olds, respectively, and
19
these are not statistically differentiable from one another. There does not seem to be marked
differences in how tickets affect the two age groups.
VII. Conclusion
This paper examines whether traffic tickets affect road safety as measured by motor
vehicle accidents. A naïve OLS regression of accidents on tickets suggests that there is no impact
of tickets on accidents. However, an analysis using exogenous variation in the number of tickets
issued to identify the causal effect of tickets on road safety gives rise to distinctly different
results tickets in fact lead to fewer motor vehicle accidents. Further, tickets help to reduce non-
fatal injuries stemming from motor vehicle accidents. In addition, the heterogeneous impact of
tickets suggests that there is scope for intervention, for example, by allocating more resources
towards traffic enforcement at night since tickets have a larger impact during nighttime. Also,
females appear to be more deterred by traffic law enforcement than men. However, there do not
appear to be differences in the impact of tickets on different age groups. Overall, the findings of
this paper suggest that as unpopular as traffic tickets are among drivers, motorist behavior does
respond to tickets.
20
References
Anbarci, Nejat, and Jungmin Lee. 2008. “Speed Discounting and Racial Disparities:
Evidence from Speeding Tickets in Boston.” IZA Discussion Paper No. 3903.
Antonovics, Kate, and Brian Knight. 2009. “A New Look at Racial Profiling: Evidence
from the Boston Police Department.The Review of Economics and Statistics, 91(1): 163-177
Babcock, Michael, Thomas Zlatoper, and Andrew Welki. 2008. “Determinants of Motor
Vehicle Fatalities: A Kansas Case Study.” Journal of the Transportation Research Forum, 47(1):
89-106.
Becker, Gary. 1968. “Crime and Punishment: An Economic Approach.” Journal of
Political Economy, 76(2): 169-217.
Beenstock, Michael, Dalit Gafni, and Ephraim Goldin. 2001. “The Effect of Traffic
Policing on Road Safety in Israel.Accident Analysis and Prevention, 33(1): 73-80.
Bound, John, David Jaeger, and Regina Baker. 1995. Problems with Instrumental
Variables Estimation when the Correlation between the Instruments and the Endogenous
Variable is Weak. Journal of the American Statistical Association, 90(430): 443-450.
Braver, Elisa, Susan Ferguson, Michael Greene, and Adrian Lund. 1997. “Reductions in
Deaths in Frontal Crashes among Right Front Passengers in Vehicles Equipped with Passenger
Air Bags.The Journal of the American Medical Association, 278(17):1437-1439.
Carpenter, Chris and Mark Stehr. 2008. “The Effects of Mandatory Seat belt Laws on
Seat belt Use, Motor Vehicle Fatalities, and Crash-Related Injuries among Youths.” Journal of
Health Economics, 27(3): 642-662.
Croson, Rachel and Uri Gneezy. 2009. “Gender Differences in Preferences.” Journal of
Economic Literature, 47(2): 448-474.
Dedman, Bill & Francis Latour. 2003. “Race, Sex, and Age Drive Ticketing. The Boston
Globe.
Elliott, Hannah. 2009. “Most Dangerous Times to Drive.” Forbes.com, January 21,
http://www.forbes.com/2009/01/21/car-accident-times-forbeslife-cx_he_0121driving.html
Levitt, Steven. 2008. “Evidence that Seat Belts Are as Effective as Child Safety Seats in
Preventing Death for Children Aged Two and Up.” The Review of Economics and Statistics,
90(1): 158-163.
Makowsky, Michael, and Thomas Stratmann. 2009. “Political Economy at Any Speed:
What Determines Traffic Citations?” American Economic Review, 99(1): 509-527.
21
Makowsky, Michael, and Thomas Stratmann. Forthcoming. “More Tickets, Fewer
Accidents: How Cash-Strapped Towns Make for Safer Roads.Journal of Law and Economics.
Soloman, Mark, Robert Ulmer, and David Preusser. 2002. “Evaluation of Click It or
Ticket Model Programs.” Department of Transportation National Highway Traffic Safety
Administration Technical Summary.
Tison J., Solomon, M., Nichols, J., Gilbert. S, Siegler J., Cosgrove. L 2006. “May 2006
Click It or Ticket Seat Belt Mobilization Evaluation: Final Report.” National Highway Traffic
Safety Administration.
22
Figures and Tables
Figure 1 Scatter Point Graph of Accidents against Tickets
.5 11.5 22.5
Daily Average of Accidents
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Daily Average of Tickets
Fitted values
23
Figure 2 Tickets and Accidents by Week, April 1, 2001 January 31, 2003
0 1 2 3 4 5 6
Mean number of accidents per week/municipality
2 3 4 5 6 7
Mean number of tickets per week/municipality
2001w14
2001w29
2001w44
2002w7
2002w22
2002w37
2002w52
week
Tickets Accidents
Treatment_Periods
X Year 2001
Treatment_Periods
X Year 2002 = CIOT
24
Figure 3 Most Common Traffic Violations by Period
0
100
200
300
400
500
600
Non-CIOT periods
CIOT periods
25
Table 1A. Summary Statistics
Variable Mean Std. Dev.
Tickets per 100 miles 4.78 6.88
Accidents per 100 miles 0.88 1.44
Injuries per 100 miles 0.38 1.03
Fatalities per 10,000 miles 0.30 6.11
Temperature (F) 51.01 17.45
Snowfall (inches) 0.11 0.69
Snow Depth (inches) 0.81 3.16
Rainfall (inches) 0.11 0.30
Population 18,307 37,357
Retail gas price 1.42 0.17
Unemployment rate 3.77 1.61
Registered cars per 100 miles 10,326 15,155
Trucks per 100 miles 5,126 6,158
Trailers per 100 miles 723.21 774.56
Motorcycles per 100 miles 411.47 484.61
Unit is at the day/municipality level.
The number of observations is 234,850
The sample period is from 4/1/2001 to 1/31/2003
Table 1B. Differences in Means of Selected Variables
Panel A: By Time of Day Day Night Difference
Tickets 2.96 1.82 1.14 ***
Accidents 0.67 0.21 0.45 ***
Injuries 0.28 0.10 0.18 ***
Fatalities 0.15 0.15 0.00
Panel B: By Gender Female Male Difference
Tickets 1.25 3.07 -1.82 ***
Accidents 0.57 0.81 -0.24 ***
Injuries 0.11 0.11 0.00 **
Fatalities 0.05 0.14 -0.08 ***
Panel C: By Age Group 15-24 25-64 Difference
Tickets 1.41 2.88 -1.469 ***
Accidents 0.36 0.90 -0.535 ***
Injuries 0.06 0.14 -0.079 ***
Fatalities 0.04 0.12 -0.073 ***
* significant at 10% ** significant at 5% *** significant at 1%
Unit is at the day/municipality level.
Panel A
Day is defined from 6 a.m. to 6 p.m; night from 6 p.m. to 6 a.m.
The sample period is from 4/1/2001 to 1/31/2003
The number of observations is 234,850
Panels B & C
The number of observations is 138,204
The sample period is from 1/1/2002 to 1/31/2003
26
(1) (2) (3)
Tickets 0.0082*** 0.0042*** 0.0002
(0.0021) (0.0013) (0.0033)
– Controls include daily municipality-specific measures of average snowfall,
snow depth, rainfall, temperature, retail gas prices (plus squared terms),
annual municipality-specific measures of population, unemployment rate,
number of registered automobiles, trucks, motorcycles, and trailers per 100
miles.
Table 2. OLS Results: The Impact of Tickets on Motor Vehicle Accidents,
Injuries, and Fatalities
Accidents
Injuries
Fatalities
– The sample size is 235,521. Standard errors clustered by municipality are
presented in parentheses.
– Tickets, Accidents, and Injuries are measured per 100 miles of public road.
Fatalities are measured per 10,000 miles of public road.
– All regressions include month-by-year, day of week, and holiday weekends
dummies, a dummy for Treatment_Periods , and a full set of controls.
(1) (2) (3) (4)
CIOT 1.904*** -0.100*** -0.0260 0.0341
(0.176) (0.0238) (0.0171) (0.1045)
– All regressions include month-by-year, day of week, and holiday weekends dummies, a
dummy for Treatment_Periods , and a full set of controls.
– Controls include daily municipality-specific measures of average snowfall, snow depth,
rainfall, temperature, retail gas prices (plus squared terms), annual municipality-specific
measures of population, unemployment rate, number of registered automobiles, trucks,
motorcycles, and trailers per 100 miles.
Table 3. Reduced Form Evidence: The Impact of CIOT Mobilizations on Motor Vehicle
Accidents, Injuries, and Fatalities
Tickets
Accidents
Injuries
Fatalities
– The sample size is 235,521. Standard errors clustered by municipality are presented in
– Tickets, Accidents, and Injuries are measured per 100 public road mileage. Fatalities are
measured per 10,000 public road mileage.
27
(1) (2) (3)
Tickets -0.0519*** -0.0137* 0.0179
(0.0125) (0.0084) (0.0548)
– Controls include daily municipality-specific measures of average snowfall,
snow depth, rainfall, temperature, retail gas prices (plus squared terms),
annual municipality-specific measures of population, unemployment rate,
number of registered automobiles, trucks, motorcycles, and trailers per 100
miles.
Table 4. IV Results: The Impact of Tickets on Motor Vehicle Accidents,
Injuries, and Fatalities
Accidents
Injuries
Fatalities
– The sample size is 235,521. Standard errors clustered by municipality are
presented in parentheses.
– Tickets, Accidents, and Injuries are measured per 100 public road
mileage. Fatalities are measured per 10,000 public road mileage.
– All regressions include month-by-year, day of week, and holiday
weekends dummies, a dummy for Treatment_Periods, and a full set of
controls.
28
Table 5. Comparing Mechanisms: Tickets versus Information
Panel A. Reduced Form Evidence: The Impact of the CIOT Campaign by Active Ticketing
Tickets Accidents Injuries Fatalities
(1) (2) (3) (4)
CIOT * Active_Ticketing 4.0517*** -0.1183*** -0.0873*** 0.163
(0.3284) (0.0359) (0.0310) (0.136)
CIOT -0.6039 -0.0407 0.0159 -0.0480
(0.1194) (0.0325) (0.0198) (0.134)
Panel B. IV Results: The Impact of Tickets for only Municipalities that Actively Ticketed
Accidents Injuries Fatalities
(1) (2) (3)
Tickets -0.0456*** -0.0210*** 0.0033
(0.0100) (0.0074) (0.0528)
* significant at 10% ** significant at 5% *** significant at 1%
– A municipality is defined as participating if the daily average of tickets issued during the CIOT
mobilizations is greater than the rest of the sample period.
– The number of observations in Panels A and B are 234,850 and 121,451 respectively. Standard errors
clustered by municipality are presented in parentheses.
– All regressions include month-by-year, day of week, and holiday weekends dummies, a dummy for
Treatment_Periods, an interaction term between Active_Ticketing * Treatment_Periods, and a full set of
controls.
– Controls include daily municipality-specific measures of average snowfall, snow depth, rainfall,
temperature, retail gas prices (plus squared terms), annual municipality-specific measures of population,
unemployment rate, number of registered automobiles, trucks, motorcycles, and trailers per 100 miles.
– Tickets, Accidents, and Injuries are measured per 100 miles of public road. Fatalities are measured per
10,000 miles of public road.
29
Panel A: Reduced Form Evidence
Tickets Accidents Injuries Fatalities
(1) (2) (3) (4)
CIOT 1.777*** -0.129*** -0.0423** -0.0032
(0.171) (0.0242) (0.0182) (0.1003)
Panel B: IV Results
Accidents Injuries Fatalities
(1) (2) (3)
Tickets -0.0729*** -0.0238** -0.0018
(0.0144) (0.0108) (0.0697)
* significant at 10% ** significant at 5% *** significant at 1%
Table 6. Robustness Check: Using November and December Months only
– The number of observations is 42,700. Standard errors clustered by municipality
are presented in parentheses.
– All regressions include month-by-year, day of week, and holiday weekends
dummies, a dummy for Treatment_Periods, and a full set of controls.
– Tickets, Accidents, and Injuries are measured per 100 miles of public road.
Fatalities are measured per 10,000 miles of public road.
– Controls include daily municipality-specific measures of average snowfall, snow
depth, rainfall, temperature, retail gas prices (plus squared terms), annual
municipality-specific measures of population, unemployment rate, number of
registered automobiles, trucks, motorcycles, and trailers per 100 miles.
30
Table 7. Robustness Check: Using Time Series Data from April 1, 2001 to January 31, 2003
Panel A: Reduced Form Evidence - The Impact of the CIOT Campaign on Tickets, Accidents, Injuries, and Fatalities
Tickets Accidents Injuries Fatalities Tickets Accidents Injuries Fatalities
(1) (2) (3) (4) (5) (6) (7) (8)
CIOT 1.9411*** -0.1155* -0.0297 0.0505 0.4685*** -0.1625** -0.1301* -0.2428
(0.2945) (0.0665) (0.0290) (0.1251) (0.0802) (0.0686) (0.0750) (0.3789)
Observations 671 671 671 671 671 671 671 479
Panel B: IV Results - The Impact of Tickets on Accidents, Injuries, and Fatalities
Accidents Injuries Fatalities Accidents Injuries Fatalities
Logs
(1) (2) (3) (4) (5) (6)
Tickets -0.0595* -0.0153 0.0261 -0.3468** -0.2776* -0.6322
(0.0329) (0.0123) (0.0623) (0.1507) (0.1583) (0.9552)
Observations 671 671 671 671 671 479
* significant at 10% ** significant at 5% *** significant at 1%
Levels
Levels
Logs
– All regressions include month-by-year and day of week dummies, and a dummy for Treatment_P eriods. Controls
include daily averages of snowfall, snow depth, rainfall, temperature, retail gas prices (plus squared terms), annual
averages of population, unemployment rate, number of registered automobiles, trucks, motorcycles, and trailers per 100
miles.
– Tickets, Accidents, and Injuries are measured per 100 miles of public road. Fatalities are measured per 10,000 miles of
public road.
31
Table 8. Heterogeneous Effects by Time of Day, Gender, and Age Group
Panel A - By Daytime or Nighttime
Time of Day Day Night
Tickets by Day/Night -0.036*** -0.117***
(0.014) (0.032)
% change in accidents -5.425 -54.50
% change from 1-unit increase in tickets by day/night 33.80 54.97
Elasticity w.r.t tickets by Day/Night -0.161 -0.991
Panel B - By Gender
Gender Female Male
Tickets -0.029*** -0.027***
(0.008) (0.010)
% change in accidents -15.05 -5.056
% change in 1-unit increase in tickets 20.93 20.93
Accident elasticity w.r.t tickets -0.719 -0.242
Panel C - By Age Group
Age Group 15-24 25-64
Tickets -0.016*** -0.032***
(0.006) (0.012)
% change in accidents -15.99 -13.14
% change in 1-unit increase in tickets 20.93 20.93
Accident elasticity w.r.t tickets -0.764 -0.628
* significant at 10% ** significant at 5% *** significant at 1%
– Each cell represents the IV estimate from a different regression.
Accidents
– Number of observations in Panel A is 234,850. Number of observations in Panels B and C is
138,204.
– Tickets, Accidents, and Injuries are measured per 100 miles of public road. Fatalities are
measured per 10,000 miles of public road.
– All regressions include month-by-year, day of week, holiday weekends dummies, and a full set
of controls.
– Controls include daily municipality-specific measures of average snowfall, snow depth, rainfall,
temperature, retail gas prices (plus squared terms), annual municipality-specific measures of
population, unemployment rate, number of registered automobiles, trucks, motorcycles, and
trailers per 100 miles.
Accidents
Accidents
... Carpenter and Stehr [6] showed that high risk drivers and heavy drinkers are more responsive to seatbelt laws. Luca [7] showed that "Click-It or Ticket" campaigns are more successful at night, suggesting that nighttime drivers would have more reason to avoid police, perhaps being under the influence of drugs or alcohol. Adams et al. [8] tested whether drivers operating vehicles under the influence of alcohol self-report that they are more likely to wear their seatbelts. ...
... Occupants are in fact heavily influenced by additional costs and benefits imposed upon them from other sources, specifically stemming from potential interaction with law enforcement. In this sense, our paper fits well within the very recent literature that has assessed other external considerations on individual seatbelt use [7,8,12]. ...
Article
Full-text available
One way to avoid detection of law enforcement officials if you are engaging in illegal activities is to wear a seatbelt. Therefore, an unintended consequence of laws allowing people to possess marijuana for medical purposes is that seatbelt use may decline among groups whose possession of marijuana is now legal. We find a decrease in seatbelt use among middle-aged males, providing evidence that drivers use seatbelts as a means to avoid police interaction. We find no such reduction in seatbelt use among those less likely to possess medical marijuana cards. Our evidence supports the contention that drivers use seatbelts more if they fear interaction with law enforcement officials, which is consistent with evidence of heightened seatbelt use among drunk drivers. These findings are important in understanding how to best design traffic safety laws and enforce them.
... During this period the number of MVCs decreased by 20%, MVC fatalities decreased twofold, and speed-related fatalities decreased threefold. Lee (2012) used citation data that included all traffic tickets issued in Massachusetts from 2001 to 2003, and MVC data from the Massachusetts highway safety division. The Click-it-or-Ticket program was used as a treatment period. ...
Article
Full-text available
Motor vehicle crashes (MVCs) have a huge cost to society in terms of death, injury and property damage. The cost of fatal MVCs alone is estimated at US $44 billion per year. Among many confounding factors, traffic citations as an element that may reduce MVC frequency are not well understood, and most research carried out to date has evaluated the effects of the total number of citations on the number of MVCs. However, certain types of citations may be more likely to reduce the number of MVCs, whereas other types are not very effective. This research was set out to examine the impact of different types of traffic citations on MVCs on two hazardous main US highways in Wyoming US-30 and US-26. A negative binomial modeling technique was implemented by exploiting 4 years of crash and citations data to identify the causal impacts of traffic citations on crash frequency by incorporating traffic and geometric features. The modeling results showed that higher numbers of speeding and seat belt citations reduce the number of crashes significantly. These findings are the results of law enforcement efforts along the highways. Traffic count and the number of horizontal curves were found to significantly increase the number of MVCs.
... During this period the number of MVCs decreased by 20%, MVC fatalities decreased twofold, and speed-related fatalities decreased threefold. Lee (2012) used citation data that included all traffic tickets issued in Massachusetts from 2001 to 2003, and MVC data from the Massachusetts highway safety division. The Click-it-or-Ticket program was used as a treatment period. ...
Article
Traffic crashes result in a huge cost to society in terms of death, injury and property damage. The cost of fatal crashes alone is estimated to be 44 billion dollars a year. Among many factors that can reduce the crash frequency, traffic citations is not well understood and most of the researches carried out so far mostly focused on the number of citations not the specific types of citations. However, certain types of citations may be more likely to reduce the number of crashes and some types are not as effective. This research was set forward to examine the impact of different types of traffic citations on traffic crashes in two main hazardous US highways in Wyoming: US-30 and US-26. The Negative binomial modeling technique was implemented by analyzing four years of citation data and incorporating traffic and geometric features to identify the causal impacts of traffic citations on crash frequency. The modeling results uncover that the higher number of speeding and seat belt citations can reduce the traffic crashes significantly. Traffic counts and number of horizontal curves were also found to be significant in increasing the traffic crashes.
... On the other hand, higher fines were followed by a reduction in monthly accident and injury rates of 0.5 % on average. Using a similar design, Lee (2012) found that dramatically increased tickets had an effect on accidents and traffic injuries in Massachusetts, particularly during night time hours and by female drivers. Finally, a randomized controlled trial has been conducted in Tsingtao (China) by Lu et al. (2012). ...
Article
Full-text available
Assess the impact of a 100-per cent increase of fines on compliance with speed-limits (50 km/h) and parking regulations in urban areas. Norm violations were measured through standardized observations in all five major cities in Switzerland before and after fines had been raised by 100 % or more at two pre- and two post-intervention periods. The speed of 20,000 cars was measured, and 10,000 parked cars were controlled. Levels of police controls remained stable over the entire period, but enforcement activity varied considerably across cities. Violations of parking regulations decreased by 32 %. Speed-limit violations remained unaffected in the longer run by higher fines. When cities without relevant enforcement activity were excluded, a reduction by 17 % was observed. No causal inferences can be drawn given the absence of a control area. However, the data suggest that higher fines deter traffic violations, provided enforcement activity remains at relevant levels.
... The two major studies on the frequency of ticketing and budget and public safety concerns, (Lee, 2012) One might also use the amount of revenue raised on fines in state comprehensive annual financial reports (CAFR). There are two problems with this data. ...
Article
Full-text available
States and municipalities across the country are struggling to match revenues with expenditures. Sometimes these governments use traffic fines and fees to help balance the budget more so at the city level than at the state level. This article explores the rationale for the issuance of traffic tickets and provides a state-level analysis on the occurrence of tickets and its relation with budget or public safety factors. Utilizing a cross-sectional multiple regression with lags, it was found that public safety concerns as evident in fatal crashes data has a significant and larger negative effect on the issuance of traffic tickets than budget concerns as measured by state credit ratings, unemployment rates, and housing prices.
... Unlike Sass and Zimmerman (2000) and Sen and Mizzen (2007), who use the number of police officers as a proxy to capture police enforcement, we followed the operationalization proposed by Lee (2012) in using traffic tickets. Indeed, police officers have duties other than patrolling traffic, such that an increase in the number of police officers cannot be assumed to automatically lead to a commensurate increase in road safety. ...
Article
Full-text available
Objectives: The objective of the current study is to determine the contribution of Chile's 2005 traffic law reform, police enforcement, and road investment infrastructure to the reduction of traffic fatalities and severe injuries from 2000 to 2012. Methods: Analyses based on structural equation models were carried out using a unique database merging aggregate administrative data from several Chilean public institutions. The sample was balanced (13 regions, over 13 years; N=169). Dependent variables were rates of traffic fatality (total, drivers, passengers, and pedestrians), severe injuries, and total number of crashes per vehicle fleet. Independent variables were (1) traffic law reform, (2) police enforcement, and (3) road infrastructure investment. Oil prices, alcohol consumption, proportion of male population 15-24 years old, unemployment, years' effects and regions' effects, and lagged dependent variables were entered as control variables. Results: Empirical estimates from the structural equation models suggest that the enactment of the traffic law reform is significantly associated with a 7% reduction of pedestrian fatalities. This association is entirely mediated by the positive association the law had with increasing police enforcement and reducing alcohol consumption. In turn, police enforcement is significantly associated with a direct decrease in total fatalities, driver fatalities, passenger fatalities, and pedestrian fatalities by 17%, 18%, 8%, and 60%, respectively. Finally, road infrastructure investment is significantly associated with a direct reduction of 11% in pedestrian fatalities, and the number of total crashes significantly mediates the effect of road infrastructure investment on the reduction of severe injuries. Tests of sensitivity indicate these effects and their statistical significance did not vary substantively with alternative model specifications. Conclusions: Results suggest that traffic law reform, police enforcement, and road infrastructure investment have complex interwoven effects that can reduce both traffic fatalities and severe injuries. Though traffic reforms are ultimately designed to change road user behaviors at large, it is also important to acknowledge that legislative changes may require institutional changes--that is, intensification of police enforcement--and be supported by road infrastructure investment, in order to effectively decrease traffic fatalities and injuries. Furthermore, depending on how road safety measures are designed, coordinated, and implemented, their effects on different types of road users vary. The case of Chile illustrates how the diffusion of road safety practices globally promoted by the World Health Organization and World Bank, particularly in 2004, can be an important influence to enhance national road safety practices.
Article
Full-text available
Anecdotal evidence suggests that local governments may have a revenue motive for traffic fines, beyond public safety concerns. Using California's county‐level data over a 12‐year period, this article shows that counties increased per capita traffic fines by 40 to 42 cents immediately after a 10 percentage point tax revenue loss in the previous year; however, these counties did not reduce traffic fines if they experienced a tax revenue increase in the previous year. This finding indicates that county governments probably view traffic fines as a revenue source to offset tax revenue loss, but not as a revenue stabilizer to manage revenue fluctuation. This article also finds that low‐income and Hispanic‐majority counties raised more traffic fines. Counties that generated more revenue from the hotel tax—a tax typically paid by travelers and visitors—raised more traffic fines, indicating a possible tax‐exporting behavior by shifting the traffic fine burden to nonlocal drivers.
Article
The objective of the current study is to determine to what extent the reduction of Chile's traffic fatalities and injuries during 2000-2012 was related to the police traffic enforcement increment registered after the introduction of its 2005 traffic law reform. A unique dataset with assembled information from public institutions and analyses based on ordinary least square and robust random effects models was carried out. Dependent variables were traffic fatality and severe injury rates per population and vehicle fleet. Independent variables were: (1) presence of new national traffic law; (2) police officers per population; (3) number of traffic tickets per police officer; and (4) interaction effect of number of traffic tickets per police officer with traffic law reform. Oil prices, alcohol consumption, proportion of male population 15-24 years old, unemployment, road infrastructure investment, years' effects and regions' effects represented control variables. Empirical estimates from instrumental variables suggest that the enactment of the traffic law reform in interaction with number of traffic tickets per police officer is significantly associated with a decrease of 8% in traffic fatalities and 7% in severe injuries. Piecewise regression model results for the 2007-2012 period suggest that police traffic enforcement reduced traffic fatalities by 59% and severe injuries by 37%. Findings suggest that traffic law reforms in order to have an effect on both traffic fatality and injury rates reduction require changes in police enforcement practices. Last, this case also illustrates how the diffusion of successful road safety practices globally promoted by WHO and World Bank can be an important influence for enhancing national road safety practices. Published by the BMJ Publishing Group Limited. For permission to use (where not already granted under a licence) please go to http://group.bmj.com/group/rights-licensing/permissions.
Article
Full-text available
This paper analyzes the determinants of motor vehicle fatalities in Kansas. The objectives of the study include the following: $ Conduct a literature review of motor vehicle safety studies. $ Formulate a theoretical model of the determinants of motor vehicle fatalities. $ Estimate the statistical significance of the various determinants. $ Compare the empirical results of the study to other recent state studies of motor vehicle fatalities. The unemployment rate had the expected inverse relationship to fatalities but was statistically significant in only two of seven estimated models. Alcohol consumption and the speed limit on rural interstate highways had the expected positive relationship to fatalities. Other variables that were statistically significant included Kansas seat belt law dummy, the proportion of young and old drivers to the rest of the Kansas population, and the ratio of rural to urban driving. The study measured the effect of three measures of highway safety law enforcement on fatalities: Kansas police per 10,000 population, Kansas police per 100 miles of road, and Kansas per capita expenditure for police protection. All three had the expected negative sign and were highly significant.
Article
Full-text available
We draw attention to two problems associated with the use of instrumental variables (IV), the importance of which for empirical work has not been fully appreciated. First, the use of instruments that explain little of the variation in the endogenous explanatory variables can lead to large inconsistencies in the IV estimates even if only a weak relationship exists between the instruments and the error in the structural equation. Second, in finite samples, IV estimates are biased in the same direction as ordinary least squares (OLS) estimates. The magnitude of the bias of IV estimates approaches that of OLS estimates as the R(2) between the instruments and the endogenous explanatory variable approaches 0. To illustrate these problems, we reexamine the results of a recent paper by Angrist and Krueger, who used large samples from the U.S. Census to estimate wage equations in which quarter of birth is used as an instrument for educational attainment. We find evidence that, despite huge sample sizes, their IV estimates may suffer from finite-sample bias and may be inconsistent as well. These findings suggest that valid instruments may be more difficult to find than previously imagined. They also indicate that the use of large data sets does not necessarily insulate researchers from quantitatively important finite-sample biases. We suggest that the partial R(2) and the F statistic of the identifying instruments in the first-stage estimation are useful indicators of the quality of the IV estimates and should be routinely reported.d
Article
Full-text available
Law enforcement officers are allowed to exercise a significant amount of street-level discretion in a variety of ways. In this paper, we focus on a particular prominent kind of discretionary behavior by traffic officers when issuing speeding tickets, speed discounting. Officers partially forgive motorists by writing a lower speed level than the speed that officers observe. Verifying the level of speed discounting by different groups of officers and motorists and ascertaining the presence of racial disparities in this lenient policing are the main objectives of this paper. We find that minority officers, particularly African-Americans, are harsher on all motorists but even harsher on minority motorists regarding speed discounting. The minority-on-minority disparity appears to be stronger in situations involving Hispanic officers, infrequently ticketing officers, male motorists, those driving old vehicles, and minority neighborhoods.
Article
Traffic accidents are one of the leading causes of injury and death in the U.S. The role of traffic law enforcement in the reduction of accidents has been studied by relatively few papers and with mixed results that may be due to a simultaneity problem. Traffic law enforcement may reduce accidents, but police are also likely to be stricter in accident-prone areas. We use municipal budgetary shortfalls as an instrumental variable to identify the effect of traffic citations on traffic safety and show that budgetary shortfalls lead to more frequent issuance of tickets to drivers. Using a panel of municipalities in Massachusetts, we show that increases in the number of tickets written reduce motor vehicle accidents and accident related injuries. The findings show that failure to control for endogeneity results in a significant underestimation of the positive impact of law enforcement on traffic safety.
Article
This paper reviews the literature on gender differences in economic experiments. In the three main sections, we identify robust differences in risk preferences, social (other-regarding) preferences, and competitive preferences. We also speculate on the source of these differences, as well as on their implications. Our hope is that this article will serve as a resource for those seeking to understand gender differences and to use as a starting point to illuminate the debate on gender-specific outcomes in the labor and goods markets.
Article
Speeding tickets are determined not only by the speed of the offender, but also by incentives faced by police officers and their vote-maximizing principals. We hypothesize that police officers issue fines more frequently when drivers have a higher opportunity cost of contesting a ticket, and when drivers are not residents of the local municipality. We also predict that local officers are more likely to issue a ticket to out-of-town drivers when fiscal conditions are tight and legal limits prevent increases in property taxes. Using data from traffic stops in Massachusetts, we find support for our hypotheses. (JEL H76, R41)
Article
I. Introduction Since the turn of the century, legislation in Western countries has expanded rapidly to reverse the brief dominance of laissez faire during the nineteenth century. The state no longer merely protects against violations of person and property through murder, rape, or burglary but also restricts "dis­ crimination" against certain minorities, collusive business arrangements, "jaywalking," travel, the materials used in construction, and thousands of other activities. The activities restricted not only are numerous but also range widely, affecting persons in very different pursuits and of diverse social backgrounds, education levels, ages, races, etc. Moreover, the likeli­ hood that an offender will be discovered and convicted and the nature and extent of punishments differ greatly from person to person and activity to activity. Yet, in spite of such diversity, some common properties are shared by practically all legislation, and these properties form the subject matter of this essay. In the first place, obedience to law is not taken for granted, and public and private resources are generally spent in order both to prevent offenses and to apprehend offenders. In the second place, conviction is not generally considered sufficient punishment in itself; additional and sometimes severe punishments are meted out to those convicted. What determines the amount and type of resources and punishments used to enforce a piece of legislation? In particular, why does enforcement differ so greatly among different kinds of legislation?
Article
Over the past thirty years, the use of child safety seats in motor vehicles has increased dramatically. There is, however, relatively little empirical evidence regarding the efficacy of child safety seats relative to the much cheaper alternative of traditional seat belts. Using data on all fatal crashes in the United States from 1975 to 2003, I find that child safety seats, in actual practice, do not provide any discernible improvement over adult lap and shoulder belts in reducing fatalities among children aged two to six. Lap-only belts are somewhat less effective, but still far superior to riding unrestrained. Copyright by the President and Fellows of Harvard College and the Massachusetts Institute of Technology.
Article
This paper provides new evidence on racial profiling using information on the race of both motorists and officers. Extending the model of Knowles, Persico, and Todd (2001), we develop a new test for distinguishing between preference-based and statistical discrimination. Our test is based on the notion that if search decisions are driven purely by statistical discrimination, then they should be independent of officer race. Our results, by contrast, demonstrate that officers are more likely to search if officer race and driver race differ. We then investigate and rule out two alternative explanations for our findings. Copyright by the President and Fellows of Harvard College and the Massachusetts Institute of Technology.