AN EXAMINATION OF THE UNINTENDED CONSEQUENCES
IN MOTORCYCLE HELMET LAWS AND ORGAN DONATION POLICY
By
Brian Moore

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Economics
2011

ABSTRACT
AN EXAMINATION OF THE UNINTENDED CONSEQUENCES
IN MOTORCYCLE HELMET LAWS AND ORGAN DONATION POLICY
By
Brian Moore
Chapter 1: "The Effect of Economic Conditions on Organ Donation"
More than 100,000 individuals in the United States are currently on the organ transplant waiting
list, yet surprisingly little is known about the factors that affect organ donation rates. In order to
increase our knowledge about the supply of transplantable organs, this paper assesses the effects
of economic conditions on deceased organ donations. Using data from the Scientific Registry of
Transplant Recipients, I estimate that a 1 percentage-point increase in the unemployment rate is
associated with a 3.5 percent decrease in the supply of organ donors killed in motor vehicle
accidents. Despite evidence of strong cyclical patterns in deaths involving cerebrovascular
disease, suicides, and homicides, no correlation was found between the unemployment rate and
organ donation rates for these causes of death. Understanding cyclical fluctuations in donations
will help policy makers and transplant specialists design policies aimed at increasing the supply
of organs by enhancing their knowledge of expected donors.

Chapter 2: "Not on My Back Roads: Are there Spillover Effects in Motorcycle Helmet
Legislation?"
Previous research on the effectiveness of motorcycle helmet laws has found a roughly 30 percent
increase in fatalities following the repeal of a universal helmet law. This study examines how
motorcycle helmet laws in neighboring states affect fatality rates. Enactment of a universal
helmet law creates incentives for motorcyclists in the state to avoid the helmet mandate by riding

in other states with the more lenient laws. If accidents involving residents of other states lead to
increased expenditure, governments may be justified in eliminating crossover incentives by
imposing helmet laws within their borders. I find that including controls for the helmet laws of
neighboring states has no effect on fatality rates, nor does it change the estimated effectiveness
of a statewide motorcycle helmet mandate. In addition, I find that neither a state's own helmet
law, nor those of its neighbors affect the fraction of motorcyclist fatalities involving nonresidents. Lastly, I find own-state helmet laws are associated with significant reductions in nonresident fatality counts, although the helmet laws of adjacent states remain insignificant. While
cross-state spillover effects could be used to support strategic interaction among state
governments, my examination of fatality patterns shows no evidence of a shift in motorcycle
riding location based on the helmet laws of neighboring states.

Chapter 3: "Donorcycles: Motorcycle Helmet Laws and the Supply of Organ Donors"
(with Stacy Dickert-Conlin and Todd Elder)
Traffic safety mandates are typically designed to reduce the harmful externalities of risky
behaviors. We consider whether motorcycle helmet laws also reduce a beneficial externality by
decreasing the supply of viable organ donors. Our central estimates show that organ donations
resulting from fatal motor vehicle accidents increase by 10 percent when states repeal helmet
laws. Two features of this association suggest that it is causal: first, nearly all of it is
concentrated among men, who account for over 90 percent of all motorcyclist deaths, and
second, helmet mandates are unrelated to the supply of donors who die in circumstances other
than motor vehicle accidents. The estimates imply that every death of a helmetless motorcyclist
prevents or delays as many as 0.33 deaths among individuals on organ transplant waiting lists.

Copyright by
BRIAN MOORE
2011

Dedicated to the memory of my mother, San Juana Moore

v

ACKNOWLEDGEMENTS

I have been very fortunate to work closely with Professor Stacy Dickert-Conlin during
my time at Michigan State University. She has truly been an excellent mentor, providing advice
and support throughout my graduate studies. I thank her for allowing me the opportunity to
serve as her research assistant, which was invaluable for my confidence and professional
development. She always displayed patience and provided encouragement when things were
difficult, especially this past year. Professor Todd Elder has also been an excellent mentor,
always eager to discuss my progress and provide recommendations for improvement. His
enthusiasm for economic research is unparalleled and inspiring, while his ability to “tell it like it
is” made meetings entertaining and unforgettable to say the least. I would also like to express
my gratitude to the other members of my dissertation committee: Professor Jeff Biddle, who has
always been quick to come up with excellent suggestions despite my tendency to provide him
short notice before popping-in to his office, and Professor Mat Reeves, whose sincere interest in
my research and fresh perspective was refreshing. Also, I’d like to thank the faculty members of
Miami University’s Department of Economics, especially Professors Michael Curme and Dennis
Sullivan, who taught me that success at the highest level was within my reach.
I would like to recognize financial support received from the American Society of
Hispanic Economists’ Summer Dissertation Fellowship, the Alliance for Graduate Education and
the Professoriate and the Michigan State University Graduate School. I would also like to
express my appreciation for the friendly attitude and dedication of the office staff both here at
Michigan State and at Indiana University, especially Dean Olsen, Lori Jean Nichols, Jennifer
Carducci and Ruth Cunningham.

vi

I am lucky to have such wonderful family and friends, who have always understood when
I declined invitations to reunions, weddings, and backyard barbeques behind the excuse of this
dissertation. I thank my father for the confidence and visible pride he displayed over the years.
He has always been tremendously supportive, both emotionally and financially. He even went so
far as to research, purchase, and deliver a replacement car when mine (which was technically
also his) was totaled. This generosity made it difficult to be upset when he decided to keep the
first car he found and start the search over again. I would like to thank my brother, Shawn,
whose willingness to discuss sports, or fantasy sports, coupled with an almost complete lack of
interest in the details of my research, were so much more beneficial than they sound. I would
like to thank my many classmates for their contributions, particularly Nicholas Sly, Brian
McNamara, Lizzi Quin and Nicole Funari, whose algorithm has saved me countless hours of data
work. Also, my thanks to Seth Gershenson and Kris Naleszkiewicz for helping me prepare for
prelims, although perhaps I should have hitched my wagon to someone besides the other two
guys who showed up at the wrong location for the first day of math camp.
Lastly, I could not have finished without the sacrifices of my beautiful wife, Katie. I’m
sure my writing and study breaks could have been much more productive for the household,
especially during football season, although she never complained…at least to my face. I could
not ask for a more loving or supportive partner. Although my name will be the only one on the
degree, this accomplishment is ours to share.

vii

TABLE OF CONTENTS

LIST OF TABLES ......................................................................................................................... ix
LIST OF FIGURES ....................................................................................................................... xi
CHAPTER ONE The Effect of Economic Conditions on Organ Donation ...................................1
1.1 Introduction .........................................................................................................................2
1.2 Institutional Details .............................................................................................................3
1.3 Data .....................................................................................................................................6
1.4 Empirical Specifications and Results................................................................................11
1.5 Conclusions .......................................................................................................................19
Appendix 1.1 Tables ................................................................................................................22
Appendix 1.2 Figures ...............................................................................................................28
References ................................................................................................................................37
CHAPTER TWO Not on my Back Roads: Are There Spillover Effects in Motorcycle Helmet
Legislation? ..............................................................................................................................41
2.1 Introduction .......................................................................................................................42
2.2 Background .......................................................................................................................44
2.3 Data and Descriptive Evidence .........................................................................................50
2.4 Methods and Results .........................................................................................................52
2.5 Conclusions .......................................................................................................................64
Appendix 2.1 Tables ................................................................................................................67
Appendix 2.2 Figures ...............................................................................................................76
References ................................................................................................................................83
CHAPTER THREE Donorcycles: Motorcycle Helmet Laws and the Supply of Organ Donors
(with Stacy Dickert-Conlin and Todd Elder) ...........................................................................87
3.1 Introduction .......................................................................................................................88
3.2 Institutional Details of Helmet Laws and Their Link to Organ Donation ........................90
3.3 Data ...................................................................................................................................93
3.4 Do Motorcycle Helmet Laws Reduce Organ Donation Rates? ........................................96
3.5 Conclusions .....................................................................................................................112
Appendix 3.1 Tables ..............................................................................................................115
Appendix 3.2 Figures .............................................................................................................123
References ..............................................................................................................................132

viii

LIST OF TABLES

Table 1.1: Summary Statistics .......................................................................................................23
Table 1.2: Organ Donor Yield Rates by Cause of Death and Age, 1994-2004 .............................24
Table 1.3: The Effect of the Unemployment Rate on Mortality, per Million Persons ..................25
Table 1.4: The Effect of the Unemployment Rate on Organ Donation, per Million Persons .......26
Table 1.5: Average Number of Transplants per Donor .................................................................27
Table 2.1: Summary Statistics .......................................................................................................68
Table 2.2: Estimates of the Effect of Helmet Laws on Motorcycle Fatalities, per Capita ............70
Table 2.3: Estimates of the Effect of Helmet Laws on the Fraction of Motorcycle Fatalities
involving Non-Residents .........................................................................................................72
Table 2.4: Fixed Effects Poisson Estimates of the Effect of Motorcycle Helmet Laws on Total
and Non-Resident Fatalities .....................................................................................................73
Table 2.5: Estimates of the Effect of Helmet Laws on Motorcycle Registrations, per Capita ......74
Table 2.6: Data Sources .................................................................................................................75
Table 3.1: Organ Donors by Year, Gender, and Circumstance, per Million Persons ..................116
Table 3.2: Motor Vehicle Fatalities by Gender, per Million Persons ..........................................117
Table 3.3: Estimates of the Effect of Helmet Laws on Organ Donations ....................................118
Table 3.4: Estimates of the Effect of Helmet Laws on per Capita Motorcycle Fatalities and Organ
Donations, by Gender and Age ..............................................................................................119
ix

Table 3.5: Estimates of the Effects of Helmet Laws on per Capita MVA, NMVA, and
Differenced Organ Donations, by Gender and Age ...............................................................121
Table 3.6: Changes in State Helmet Laws, 1994-2007 ................................................................122

x

LIST OF FIGURES

Figure 1.1: Motor Vehicle Accident Mortality and the Unemployment Rate ..............................29
Figure 1.2: Suicide Mortality and the Unemployment Rate .........................................................30
Figure 1.3: Homicide Mortality and the Unemployment Rate .....................................................31
Figure 1.4: Cerebrovascular Disease Mortality and the Unemployment Rate .............................32
Figure 1.5: Motor Vehicle Accident Donors and the Unemployment Rate .................................33
Figure 1.6: Suicide Donors and the Unemployment Rate ............................................................34
Figure 1.7: Homicide Donors and the Unemployment Rate.........................................................35
Figure 1.8: Natural Cause Donors and the Unemployment Rate..................................................36
Figure 2.1: Arizona Non-Resident Motorcycle Fatality Trends ...................................................77
Figure 2.2: Pennsylvania Non-Resident Motorcycle Fatality Trends ...........................................78
Figure 2.3: Ohio Non-Resident Motorcycle Fatality Trends ........................................................79
Figure 2.4: Oklahoma Non-Resident Motorcycle Fatality Trends ...............................................80
Figure 2.5: Texas Non-Resident Motorcycle Fatality Trends ......................................................81
Figure 2.6: New Mexico Non-Resident Motorcycle Fatality Trends ...........................................82
Figure 3.1: Male Motorcyclist Fatalities per Million Residents .................................................124
Figure 3.2: Male MVA Donors Age 18-34 per Million Residents .............................................125
Figure 3.3: Motorcycle Registrations per Thousand Residents ..................................................126
Figure 3.4: Per Capita Motorcycle Registrations in Three Large States ....................................127
xi

Figure 3.5: Per Capita Motorcyclist Death Rates in Three Large States ....................................128
Figure 3.6: Per Capita Male Motorcyclist Fatalities in Six States ..............................................129
Figure 3.7: Per Capita Male MVA Organ Donors Aged 18-34 in Six States .............................130
Figure 3.8: Per Capita Motorcycle Registrations in Six States ...................................................131

xii

CHAPTER 1

THE EFFECT OF ECONOMIC CONDITIONS ON ORGAN DONATION

1

1.1 Introduction
As of November 2010, more than 109,000 individuals were on the organ transplant
waiting list in the United States. Over the last ten years, an average of 6,730 people died while
1

waiting for an organ annually and another 2,025 became too sick for transplantation. The
shortage of transplantable organs in the United States has attracted considerable attention among
health professionals and policy makers. Understanding the determinants of organ donation rates
will help in designing policies intended to increase the supply of transplantable organs. Because
more than three-quarters of all transplanted organs come from deceased donors, events and
policies that impact mortality rates are also likely to affect organ donation rates.
Using data from the Scientific Registry of Transplant Recipients (SRTR), I estimate the
effect of the state-level unemployment rate on monthly organ donation rates by cause of death of
the donor. I show that an increase in the unemployment rate is associated with decreases in
organ donors due to motor vehicle accidents and cerebrovascular disease (stroke) and increases
in donors from suicide and homicide. Specifically, I find that a 1 percentage-point increase in
the unemployment rate is associated with a 3.5 percent decrease in organ donors from motor
vehicle accidents. For those under the age of 18, the results are even stronger, as a 1 percentagepoint increase in the unemployment rate is associated with a 9.1 percent decrease in organ donors
from motor vehicle accidents.
These results are an extension of existing research finding that mortality rates exhibit a
strong cyclical pattern (Ruhm 2000, Ruhm 2003, Ruhm 2005, and Miller et al. 2009), with
differential effects by cause of death. Furthermore, the causes of death that are most responsive
1

These numbers are taken from the Organ Procurement Transplantation Network (OPTN)
national data available for public use on the OPTN’s website:
http://optn.transplant.hrsa.gov/latestData/step2.asp.
2

to changes in the unemployment rate are the same causes of death most likely to result in organ
donation: those resulting in brain death. Using monthly data, rather than annual data used by
previous authors, I find results consistent with the previous estimates: a 1 percentage-point
increase in the unemployment rate is associated with decreases in motor vehicle accident and
cerebrovascular disease fatalities of 2.5 percent and 1.6 percent, respectively. I also find that a 1
percentage-point increase in the unemployment rate is associated with increases in suicides and
homicides of 1.2 percent and 4.2 percent, respectively. Together with the donation estimates, I
estimate that each additional motor vehicle death caused by economic expansion results in 0.04
additional organ donors and 0.14 additional transplants.
In the following section, I elaborate on the effects of macroeconomic conditions on
mortality and provide a brief overview of organ donation in the United States. Section 3
describes the organ donation data from the SRTR as well as the publicly available mortality data.
Section 4 presents the empirical specifications and results, and Section 5 concludes.

1.2 Institutional Details
Macroeconomic conditions and mortality
Previous work on the effects of the business cycle on mortality rates (Ruhm 2000, Ruhm
2005, Miller et al. 2009) showed a strong procyclical trend in mortality rates. The results were
substantively and statistically significant, as well as robust to changes in specification, estimation
method, and decompositions by age and cause of death. When decomposed by cause of death,
eight out of ten categories examined by Ruhm (2000) were found to be procyclical, with motor
vehicle accident (MVA) deaths exhibiting the strongest correlation with the unemployment rate.
The mechanism behind the procyclicality of mortality rate is not fully understood, especially the

3

strong patterns found among the elderly, who should not be personally affected by changes in the
unemployment rate. The procyclicality in mortality for working-age adults is driven by motor
vehicle accidents, and the estimates are consistent across all age groups. Based on this, Miller et
al. (2009) suggest business-cycle externalities, such as increases in road traffic, rather than
changes in individual behavior, are the major factor for working-age adults.
Patterns seen in the age- and cause-of-death-specific regressions set the foundation for
the connection between unemployment rates and organ donation. Over 90 percent of deceased
organ donors are brain dead at the time of recovery, meaning they have irreversible cessation of
all brain function.

2

Ruhm (2000) and Miller et al. (2009) show mortality rates for the causes of

death most likely to result in brain death, such as accidents and suicides, are very prone to
changes in the unemployment rate. Fatal motor vehicle accidents, in particular, represent over
21 percent of all deceased organ donors since 1994. Suicides, which Ruhm (2000) found to be
significantly countercyclical, represent nearly 8 percent of deceased organ donors. Moreover,
donations due to suicide are of particular interest, as Figueiredo et al. (2007) showed that
families are more likely to agree to donation consent in instances of suicide.

Organ donation in the United States
While roughly 2.4 million people die in the United States each year, less than 1 percent
will be due to brain death. Brain death occurs when injuries lead to as the irreversible cessation
of all brain function. Medical personnel are then able to keep the heart beating with mechanical
ventilation and medication, allowing the internal organs to receive oxygenated blood and remain
viable for transplantation. In contrast, organs deteriorate rapidly following cardiac deaths,
2

Source: Author’s calculation from SRTR data. The remaining cases are donations after cardiac
death. See http://www.organtransplants.org/understanding/death/ for more details.
4

making them less suitable for transplantation. In addition to suffering brain death, potential
donors must be free of contraindications for organ donation. Contraindications are conditions
that prevent the organs from safely being used in transplantation, such as active cancer, HIV, and
3

other infectious diseases. In the absence of these contraindications, Organ Procurement
Organization (OPO) staff will seek consent for organ donation.

4

Organ donation in the United States is an informed-consent system, meaning organs are
5

recovered from a potential donor only if explicit permission has been granted to do so. Consent
for donation must be obtained either from the individual prior to death or from surviving family
members. Although federal law has always maintained that health care professionals only need
the donor’s consent to begin recovery, actual OPO practices have been very different. Until
recently, health care professionals requested consent from the potential donor’s next-of-kin
before beginning organ recovery, regardless of documentation of the donor's own wishes.

6

Starting in 1998, individual OPOs began enforcing the wishes of the deceased regardless of
family opinion, a practice known as first-person consent, in an effort to respect the donor's
wishes and increase the supply of organs. A survey of all OPOs conducted by Wendler and
Dickert (2001) found “widely divergent consent practices” among OPOs, noting that the
3

A complete list of contraindications for organ donation can be found in Table 2 of Guadagnoli
et al. (2003).
4
OPOs are regional members of the OPTN responsible for organ services within a designated
area. There are 58 current OPOs throughout the United States and its territories.
5
This is in contrast to a presumed-consent system, used in many European countries, where all
individuals are presumed to have consented to organ donation unless there is specific evidence
that the individual has chosen to “opt-out” of donation. See Abadie and Gay (2006) for details
about how organ donation rates are impacted by this decision.
6
There are several mechanisms available to individuals wishing to designate oneself as a donor,
including on a driver’s license, signed donor card, state donor registry, living will, or through
durable power of attorney. States differ, to a small degree, on which documents are accepted as
legal consent for donation.
5

differences stemmed from ethical disagreements regarding whose wishes (the deceased or nextof-kin) should determine whether or not organs are recovered. Sheehy et al. (2003) noted that
low consent rates are the major limiting factor to increasing the number of organ donors and
found that the overall family consent rate from 1997-1999 was just 54 percent. Howard et al.
(2007) report that by 2004, some OPOs had reached 60 percent conversion of potential donors.
Once consent has been confirmed or obtained for the potential donor, surgeons initiate the
process of organ recovery.

1.3 Data
Donation and transplantation services in the United States are coordinated and provided
by the Organ Procurement Transplantation Network (OPTN), under directive from the 1984
National Organ Transplant Act. The OPTN is also responsible for collection and management of
data from every donation and transplant occurring in the United States. This study uses data
from the Scientific Registry of Transplant Recipients (SRTR), which includes data on all donors,
wait-listed candidates, and transplant recipients in the United States, submitted by the members
7

st

of the OPTN. Beginning April 1 , 1994, the files also report the donor’s cause of death, with
separate categories for motor vehicle accidents, other accidents such as falls, child abuse,

7

The data reported here have been supplied by the Minneapolis Medical Research Foundation
(MMRF) as the contractor for the Scientific Registry of Transplant Recipients (SRTR). The
interpretation and reporting of these data are the responsibility of the author and in no way
should be seen as an official policy of, or interpretation by, the SRTR or the U.S. Government.
The Health Resources and Services Administration (HRSA), U.S. Department of Health and
Human Services provides oversight to the activities of the OPTN and SRTR contractors. For
more information about the SRTR, please see http://www.srtr.org/.
6

8

suicide, homicide, and, as of October 25th, 1999, natural causes. For the purposes of this study,
the individual-level data were collapsed into state-level counts of deceased donors by month for
the years 1994 through 2004.
Panel A of Table 1.1 displays summary statistics for organ donors by year and cause of
death. Each cell represents the average number of donors per million persons across all 50 states
9

(and D.C.) and 12 months of the year. Before October 25th, 1999, donors dying of natural
causes were coded as "None of the Above." Once the natural cause category was added,
donations in the "None of the Above" category gradually, yet substantially decrease as health
care professionals began incorporating the new category. Natural cause donors increased 22.3
percent from 2000 to 2004.
The mortality data comes from the National Vital Statistics Multiple Cause-of-Death
Mortality Data, which I have compressed into monthly, state-level fatality counts using
International Classification of Diseases (ICD) codes that match the cause of death found in the
organ donor data.

10

In most cases, the match is straightforward; however, death by “natural

causes” does not have a strict medical definition, and therefore does not appear in the ICD codes
underlying the Vital Statistics mortality data. Fatalities coded as cerebrovascular disease, which
include strokes and other instances of intracranial hemorrhage, were used to match with organ
8

The OPTN refers to these categories as the donor's circumstance of death. To avoid confusion,
I will instead refer to them as the donor's cause of death.
9
Due to limitations of the organ donor data, figures for 1994 are from April-December. The
figure for "Natural Causes" in 1999 is from November-December.
10
For the years 1994-1998, I use the following ICD 9th revision codes: MVA (E810-E825),
suicide (E950-E959), homicide (E960-E978), and cerebrovascular disease (430-438). For the
years 1999-2004, I use the following ICD 10th revision codes: MVA (V02-V04, V09.0, V12V14, V19.0-V19.2, V19.4-V19.6, V20-V79, V80.3-V80.5, V81.0-V81.1, V82.0-V82.1, V83V86, V87.0-V87.8, V88.0-V88.8, V89.0, V89.2), suicide (X60-X84, Y87.0), homicide (X85Y09, Y87.1), and cerebrovascular disease (I60-I69).
7

donors in the natural cause category. While 75 percent of natural cause donors died of a stroke
or other intracranial hemorrhage, this classification is admittedly less clean than the other
categories of interest.

11

Panel B of Table 1.1 displays the corresponding average monthly

mortality per million persons by year and cause of death from April 1994 to December 2004.
Cerebrovascular deaths are three times more common than motor vehicle deaths, averaging 50.1
fatalities per month compared to an average 14.9 MVA fatalities per month. Despite this, the
number of natural cause donors is similar to that seen in motor vehicle accidents. In addition to
the relative magnitudes, it is also important to note that Table 1.1 shows a decreasing trend in the
mortality rate for all four cause-of-death categories over this time period, including a 37 percent
drop in homicides. While motor vehicle fatalities decreased over 9 percent from 1994-2004,
MVA donors decreased just 4 percent. Table 1.1 illustrates differences in the probability that a
fatality turns into an organ donation, even among the categories most likely to result in brain
death.
Another simple way to evaluate differences in the probability of becoming an organ
donor is by comparing the ratio of organ donors to mortality counts by cause of death. These
results, which I will refer to as the yield rates, can be seen in Table 1.2. Sample means for
monthly mortality levels are displayed in brackets below each estimate. Row 1 of Table 1.2
shows the yield rates for the pooled-age sample, aggregating over all years. Similar to the
implicit pattern of Table 1.1, higher levels of organ donation can be seen following motor
vehicle accidents, relative to the other causes of death. Frezza et al. (1999) previously noted that
MVA deaths are associated with the highest conversion ratio from potential donor to actual

11

Source: Author’s calculation from OPTN data. Intracranial hemorrhage / Stroke is a category
within the OPTN's mechanism-of-death variable, however, this mechanism also occurs in other
circumstances of death, such as motor vehicle accidents.
8

donor, partly because time available for conversion for MVA deaths is greater than in those
dying from anoxia.

12

Rows 2-6 of Table 1.2 show the yield rates separately by age group,

illustrating that there are higher rates of organ donation when the deceased is younger. Younger
individuals tend to have healthier bodies relatively speaking, with fewer occurrences of
contraindications (Frezza et al. 1999). While the under 18 age group only represents about 13
percent (= 1.9051 / 14.9114) of all MVA deaths, the yield rate is nearly twice as high as any
other age category. The yield rates are highest for those under 34 years old in the natural causes
category, the sample sizes show the 65 and older age group represents over 88 percent (=
43.2185 / 49.0252) of all cerebrovascular deaths. With a large organ shortage, recent attention
focuses on the use of more elderly donors, although disagreement still exists as to what
conclusions can be drawn from the results. Karatzas et al. (2011) concluded that elderly donors
do not affect patient and graft survival in many patients, while Di Cocco et al. (2011) and Cassini
et al. (2010) found the use of elderly kidney donors had a significant negative effect on
transplant survival.
Figures 1.1 through 1.4 show the national time trends of seasonally-adjusted mortality
and the unemployment rate from 1994 through 2004 by cause of death.

13

Seasonally-adjusted

data better illustrate the identification of the regression results to follow, which will include a full
set of binary month indicators. The inclusion of individual month controls is important in
analyzing correlations between the unadjusted unemployment rate and unadjusted mortality rates
due to the large body of evidence that seasonal trends exist in mortality rates. Seasonal trends
12

Marconi et al. (2011) and Singhal et al. (2009) conclude that donor cause of death was not a
strong predictor of graft survival.
13
The mortality data depicted are residuals from a regression of deaths on a full set of month
dummies. While a large increase in homicides occurs in September 2001, the regression results
that follow are not sensitive to inclusion of this month.
9

have been previously documented in cerebrovascular disease or strokes, consistently finding high
mortality rates during winter months (Khan et al. 2005, Myint et al. 2007, Nakaji et al. 2004,
Turin et al. 2009, Wang et al. 2003). Motor vehicle fatalities, on the other hand, have seasonal
peaks in the summer months (Liu et al. 2005, Trudeau 1997). Research has also found seasonal
patterns in suicide (Warren et al. 1983, Bridges et al. 2005) and homicide mortality (Cheatwood
14

1988, Warren et al. 1983, Tennenbaum and Fink 1994).

Despite the use of seasonally-

adjusted data, the mortality counts and the unemployment rate in Figures 1.1 - 1.4 are still
aggregated to the national level, resulting in a noisy representation of the state-level
identification strategy to follow. Nevertheless, suicide mortality appears to have a procyclical
trend while cerebrovascular disease mortality appears to have a countercyclical trend. Cyclical
patterns for motor vehicle accidents and homicides cannot be easily identified in the figures.
Figures 1.5 through 1.8 show the time trend of seasonally-adjusted deceased organ donations
with the national unemployment rate over the same time period.

15

The cyclical patterns in organ

donations are far less clear in this graphical representation than the mortality trends. I next turn
to assessing the strength of the relationship between macroeconomic conditions and the supply
of deceased organ donors (and mortality) in a regression framework.

14

Warren et al. (1983) find suicide counts were high in the spring and low in the winter.
Cheatwood (1988) finds high homicide rates in specific months (December, July, and August),
as opposed to entire seasons. Warren et al. (1983) and Tennenbaum and Fink (1994) find
homicide spikes from July-September as well as December.
15
The donation data depicted are residuals from a regression of donations on a full set of month
dummies.
10

1.4 Empirical Specification and Results
My primary empirical strategy involves estimating state- and month-specific organ
donation rates and mortality rates as a function of the state unemployment rate. The main
estimation equation is:
ysm = αs + µm + γ(unemployment rate)sm + Xsmβ + εsm ,

(1)

where ysm measures the number of deceased organ donors (or fatalities) per million persons by
cause of death, s indexes the state, m indexes the month, and (unemployment rate)sm represents
the unadjusted unemployment rate. All specifications include a full set of state and month
indicators (αs and µm respectively), and the time-varying state-level variables Xsm. The vector
Xsm includes separate measures of the state's population aged 0-17, 18-34, 35-49, 50-64, and 65
and above, indicators for whether the state had an organ donor registry, whether online
registration was available, and whether an OPO in the state enforced a first-person consent
paradigm.

16

I weight each observation by the state’s population in that month using U.S. Census

Bureau estimates.

17

Estimates of γ based on (1) capture the association between within-state

variation over time in organ donation rates (or mortality rates) and economic conditions, proxied
by the unemployment rate.

16

Monthly, unadjusted, state-level unemployment rates were obtained from the U.S. Bureau of
Labor Statistics. Data on first-person consent practices, the existence of state donor registries
and the ability to sign up for those registries online came from the author’s interviews with OPO
employees.
17
State-level population estimates are available annually from the U.S. Census Bureau. These
estimates were expanded to monthly estimates by assuming a constant population growth rate
between years.
11

Thus far, all work on the cyclical nature of mortality has used annual variation. Using a
monthly, state-level aggregation instead would provide another test of the reliability of previous
findings. Annual aggregation absorbs variation in unemployment rates that may be of interest,
such as summer increases in employment due to breaks in the school year. If changes in motor
vehicle congestion are an important part of the cyclical mortality patterns, the monthly
aggregation level would have an advantage. Monthly results that were noticeably different from
previous estimates would be cause for concern that the results could not be interpreted as causal
due to a misspecified model or omitted variable bias. Replicating the previous estimates at the
monthly aggregation level, including binary month controls, strengthens the premise that
identification is coming from changes in the unemployment rate. Another reason for choosing
monthly aggregation in this study is the nature of the organ donation data available. As
previously noted, cause of death was first recorded in April of 1994, and the natural causes
category was added in October of 1999.
Row 1 of Table 1.3 shows results for estimating γ in (1) with cause-of-death-specific,
monthly, state-level mortality rate as the dependent variable.

18

Much like previous studies, the

results show a very strong procyclical relationship for motor vehicle fatalities. A 1 percentagepoint increase in the monthly unemployment rate results in a decrease in motor vehicle fatalities
of 0.3799 per million persons, with a standard error of 0.0530 (all standard errors are robust to
clustering at the state level to allow for arbitrary serial correlation patterns within-state over
time). In addition, a 1 percentage-point increase in the unemployment rate is associated with a

18

The estimates in Table 1.3 and Panel A of Table 1.4 were generated using the command areg
in Stata SE 11.2.
12

19

decrease in cerebrovascular deaths of 0.7772, with a standard error of 0.0931.

Compared to

sample means, these estimates reflect a 2.5 percent (= 0.3799 / 14.9114) decrease in motor
vehicle fatalities and a 1.6 percent (= 0.7772 / 49.0252) decrease in cerebrovascular deaths.
These estimates are similar to Ruhm (2000), who found a 3.02 percent decrease in MVA
fatalities, as well as Miller et al. (2009) who found a 2.94 percent decrease in MVA fatalities.

20

Also similar to the previous literature, a strong countercyclical pattern can be seen in suicides,
with a 1 percentage-point increase in the unemployment rate leading to 1.2 percent (=0.1233 /
10.3454) increase in suicides.

21

Ruhm (2000) and Miller et al. (2009) estimated a 1 percentage-

point increase in unemployment was associated with an increase of 1.27 percent and 1.68 percent
in suicides, respectively.
One interesting departure from the previous studies is the estimated strong
countercyclical relationship for homicides. A 1 percentage-point increase in the unemployment
rate is associated with a 4.2 percent (= 0.2380 / 5.7119) increase in homicides. Ruhm (2000)
showed a procyclical trend in homicides using data from 1972-1991, as did Miller et al. (2009)
after expanding the sample through 2004. They find that a 1 percentage-point increase in the
unemployment rate is associated with decreases in homicides of 1.89 percent and 1.62 percent,
19

The cerebrovascular disease mortality was separately estimated in column 5 of Table 1.3
using a restricted sample of November 1999 through 2004 in order to match the results with
organ donation estimates in column 4 of Panel A in Table 1.4.
20
Miller et al. (2009) also found a 1.66 percent increase in degenerative brain deaths. Although
this may not be a perfect substitute for cerebrovascular disease, it is the closest reported category
found.
21
Hamermesh and Soss (1974) present a utility maximization model in which an increase in the
unemployment rate decreases an individual’s expected future income and utility, leading to an
increase in suicide. Yang (1992) found statistically significant countercyclical variation in
suicide rates, but only for white males. Ruhm (2000, 2003) suggests macroeconomic conditions
have an adverse effect on the mental health of working-age adults, leading to increased suicide
rates during economic downturns.
13

respectively. One reason this discrepancy is expected is because of the drastic decrease in crime
rates throughout the country in the 1990s, as studied before by numerous authors (for example,
see: Blumstein and Rosenfeld 1998, Blumstein and Wallman 2006, Levitt 2004, Rosenfeld 2004,
Zimring 2006). Explanations for this decline in crime rates include levels of police officers,
increased incarceration, capital punishment, racial profiling, concealed-carry laws, the decline in
the crack epidemic, and the legalization of abortion, although there is no widespread support for
any one hypothesis. Whatever the cause, this sharp crime rate decline has a strong impact on my
estimation results for the homicide mortality category, as my sample begins in 1994.
Rows 2-6 of Table 1.3 show the results separately for five different age groups: 0-17, 1834, 35-49, 50-64, and 65 and over. The cyclical estimates for motor vehicle fatalities can be seen
in each age group. Miller et al. (2009) found a similar pattern, leading to their assertion that the
procyclical pattern in motor vehicle fatalities is most likely due to business-cycle externalities
stemming from increased traffic. Similarly, a significant countercyclical pattern in homicides
can be seen in each age category older than 17. The cyclical pattern in the cerebrovascular
disease estimates above is almost completely due to those 65 and older. The fact that the
countercyclical pattern in suicides is concentrated in the 35-49 and 50-64 year old categories,
which are responsible for roughly 50 percent of all suicides, also supports previous findings.
Hamermesh and Soss (1974) hypothesized and showed that changes in unemployment have an
increasing effect on suicide as workers age, explained by the fact that older workers have lower
expected future incomes.

14

Unemployment and organ donation
Panel A of Table 1.4 shows results for estimating γ in (1) with organ donation rates by
cause of death as the dependent variable instead of mortality rates. Row 1 shows the results
when all ages are pooled. Similar to the mortality results, I estimate that the unemployment rate
is negatively correlated with the rate of motor vehicle accident donations, and positively
correlated with the rate of suicide and homicide donations. However, only the motor vehicle
accident donation results are statistically significant. In contrast to the mortality results for
cerebrovascular disease, organ donations from natural causes are countercyclically related to
macroeconomic conditions and highly significant. Data coding may be one explanation for the
unexpected positive estimate. To test this theory, I combine the natural cause donors with those
in the none of the above category, which experienced a sharp decline when the natural cause
category was added. The results of this exercise are reported in Column 5 of Table 1.4 Panel A.
The point estimate in the pooled sample now comes out negative, and although it is not
statistically different from zero, it is consistent with the expected sign, given the fatality results.
When the regressions are run separately by age of donor in Rows 2-6 of Table 1.4 Panel
A, the patterns underlying the pooled-sample estimates become clear. In the case of MVA
donors, a highly significant procyclical correlation for donors under the age of 18 emerges.
Among this group, which represents one third of all MVA donors and 39 percent of all donors
under 18, a 1 percentage-point increase in the unemployment rate is associated with a decrease in
donors of 0.0127 per million persons, with a standard error of 0.0038. This represents a 9.1
percent (=0.0127 / 0.1403) change in organ donors in this age group relative to the sample mean.
This is a dramatic change, especially considering MVA victims under the age of 18 only
experienced a 2.5 percent (= 0.0478 / 1.9051) change in mortality rate. For comparison, the 1

15

percentage-point increase in the unemployment rate would translate to an annual decrease of 8
MVA donors per million (= 0.127 × 12 × 51) and 29 total lives per million (= 0.0478 × 12 × 51)
in the under 18 age group. Additionally, donation decisions for this age group are far more
likely to be made by the donor’s family, suggesting that age and cause of death may play a
significant role in the family's decision to consent. Table 1.4 also shows that the strongest results
in the natural causes category are for those aged 50-64, a group that experienced an insignificant
change in mortality rate due to changes in the unemployment rate. When natural causes are
combined with the none of the above category, the under 18 age group is the only one with
statistically significant results, with a 1 percentage point increase in the unemployment rate
resulting in an 8.9 percent (= 0.0071 / 0.0794) decrease in organ donors. While suicides and
homicides showed strong countercyclical mortality patterns in Table 1.3, the donation trends for
these causes of death do not appear to have a cyclical pattern. One possible explanation for this
is that the types of suicides and homicides that occur may differ throughout the business cycle.
Although rising unemployment is having a significant impact on mortality rates, there may be
differences in the proportion of suicides and homicides that result in potential donors. However,
without individual hospital records for each suicide and homicide, I am unable to formally test
this theory.

Instrumental variables estimates of the effect of fatalities on organ donation rates
The estimates of γ in (1) displayed in Table 1.3 and Panel A of Table 1.4 characterize
reduced-form relationships between macroeconomic conditions and mortality rates or organ
donation rates, respectively. The ratio of the two estimates therefore is numerically equivalent to
an instrumental variables estimate of the effect of mortality rates on organ donation rates. Such

16

an estimate captures the effect of excess deaths due to economic conditions on the supply of
organ donors. The instrumental variables estimates are shown in Panel B of Table 1.4 by cause
of death.

22

In the case of motor vehicle accidents, the estimate of 0.0400 (= -0.0152 / -0.3799)

implies that for each additional motor vehicle death caused by economic expansion, there will be
an additional 0.04 organ donors. The results in the under 18 age group are staggering, with
0.2651 (= -0.0127 / -0.0478) additional organ donors per motor vehicle death. This IV estimate
represents the conversion from death to donor specifically for the marginal individuals who die
in motor vehicle accidents because of the increase in the unemployment rate, and are over three
times larger than the overall yield rate of 0.0795 in Table 1.2.
Instrumental variables estimates of the number of organ donors per fatality for the other
causes of death are not statistically different than zero.

23

This does not mean that additional

deaths in these categories do not lead to organ donors, but that no relationship is found for the
marginal individuals whose death was caused by changes in economic conditions. As seen in
Panel A of Table 1.4, the unemployment rate does not appear to have a strong effect on organ
donation rates in these categories. Within natural causes, this is most likely due to the large
proportion of cerebrovascular deaths in the 65 and older category. This is an age category that is
rarely deemed suitable for organ donation, as demonstrated by the substantially lower yield rates
for those 65 and older in Table 1.2 across all categories of death. While, 58.7 percent of all

22

The estimates in Panel B of Table 1.4 were generated using the command ivreg in Stata SE
11.2. Standard errors for the IV estimates account for the two-stage estimation process.
23
As previously discussed, the estimates for natural causes are plagued by the late addition and
gradual recognition of the category as a cause of donor death in OPTN forms. The negative
estimate implies that increased cerebrovascular disease fatalities are associated with decreases in
donors due to natural causes. The negative estimate goes away in column 5 where the none of the
above category is added to the natural cause donors.
17

fatalities in the data in 2004 occurred in those 65 and older, just 9.7 percent of all deceased organ
donors were 65 or older.

24

The instrumental variables estimates do suggest there are differences in the marginal
benefit of a death resulting from a change in macroeconomic conditions. Those dying in motor
vehicle accidents are more likely to become organ donors. This result is compounded by
differences occurring in the suitability of each fatality for organ donation, as explained in the
next section.

Transplants per donor
The final link between fatalities and the supply of organs centers on the number of
transplantable organs recovered from each donor. According to the OPTN, an overall average of
2.7 organs are transplanted from each deceased organ donor.

25

The information available in the

SRTR data makes it possible to calculate the number of transplants by cause of death of the
donor. Just as certain causes of death lead to more donors than others, the number of transplants
per donor is not uniform either. As seen in Panel A of Table 1.5, motor vehicle accidents,
suicides, and homicides (3.5, 3.5 and 3.6 transplants per donor, respectively) lead to nearly an
entire organ more per donor than the OPTN average. In the previous section we saw that each
motor vehicle accident death leads to 0.04 additional donors. Using the number of transplants
per donor and the assumption that each transplant saves one life, 0.14 (= 0.04 × 3.5) lives are
saved for each motor vehicle accident death caused by economic expansion.

24

Source: Author's calculation from SRTR and U.S. Vital Statistics data.

25

This figure is the author's calculation based on publically available OPTN data for the year
2009, the most recent full year of data available at the time of the calculation.
18

Donors who died of natural causes do not yield as many transplants per donor (2.3),
which suggests that age is an important factor in determining the number of transplants per
donor. When looking at transplants per donor broken down instead by the age of the donor in
Panel B of Table 1.5, that theory is confirmed. Donors under the age of 18 average 3.4
transplants and donors age 18-34 average 3.7 transplants, well over the OPTN average of 2.7.
Therefore, the marginal benefit of a young donor is greater than that of older donors, ceteris
paribus. For those under 18, one in four motor vehicle accident fatalities become an organ
donor, meaning for each motor vehicle accident death caused by economic expansion, an
incredible 0.90 (= 0.2651 × 3.4) other lives are saved via organ transplantation, on average.

1.5 Conclusions
The United States faces a critical organ shortage, as demonstrated by the growing
transplant waiting list. One way to bridge the gap between the supply and demand for organs is
to improve the conversion of potential donors into actual donors. A more complete
understanding of factors that affect mortality rates, especially among the causes of death that
most commonly lead to potential donors, would aid future efforts to increase the number of
deceased organ donors. Cyclical changes in macroeconomic conditions lead to fluctuations in
mortality. This study is the first to examine whether or not corresponding trends are present in
deceased organ donations as well.
Consistent with a larger literature, I find that mortality rates in the United States exhibit a
strong cyclical pattern. Specifically, a 1 percentage-point increase in the monthly unemployment
rate decreases motor vehicle accident and cerebrovascular deaths by 2.5 and 1.5 percent,
respectively. On the other hand, deaths from suicide and homicide follow a countercyclical

19

pattern, with a 1 percentage-point increase in the unemployment rates associated with mortality
increases of 1.2 and 4.2 percent, respectively. While it is expected that similar trends extend to
deceased organ donations for these causes of death, a statistically significant relationship is only
found among motor vehicle accident victims. Among this group, increasing the unemployment
rate by 1 percentage-point decreases monthly organ donation rates by 0.0152 per million
persons, a 3.5 percent change. The results by age of donor are even more telling, as a 1
percentage-point increase in the monthly unemployment rate decreases MVA organ donors
under the age of 18 by 9.1 percent.
Combining the estimates of the effect of unemployment rates on mortality rate with those
on organ donation rates yields an instrumental variables estimate of the impact of mortality on
the supply of organ donors. Using this approach, I find that each motor vehicle accident death
resulting from economic expansion leads to 0.04 additional organ donors and 0.14 additional
life-saving organ transplants. Motor vehicle accident victims under the age of 18 have an even
stronger estimated effect, resulting in 0.27 additional donors and 0.90 additional transplants from
each fatality.
These results show that changes in macroeconomic conditions have a substantial impact
on both mortality rates and organ donation rates. While the additional deaths represent an
unfortunate consequence of improving economic conditions, these additional deaths also
positively contribute to the supply of transplantable organs. In light of these findings, expected
cyclical fluctuations should be incorporated into future organ donation policy analysis.

20

CHAPTER 1 APPENDICIES

21

APPENDIX 1.1

CHAPTER 1 TABLES

22

Table 1.1: Summary Statistics
Panel A: Average Monthly Organ Donors, by Year and Cause of Death, per Million Persons
Non
Motor
Motor
None of
Vehicle
Natural
Vehicle
Child
the
All
Accident
Suicide
Homicide Causes Accident
Abuse
above
Causes
1994
0.460
0.182
0.113
0.110
0.015
0.712
1.591
1995
0.430
0.182
0.090
0.114
0.014
0.791
1.620
1996
0.421
0.164
0.105
0.163
0.012
0.787
1.653
1997
0.404
0.160
0.087
0.172
0.019
0.725
1.574
1998
0.438
0.162
0.080
0.163
0.018
0.841
1.706
1999
0.419
0.146
0.075
0.310
0.178
0.012
0.730
1.713
2000
0.460
0.150
0.073
0.386
0.145
0.012
0.558
1.783
2001
0.415
0.152
0.087
0.433
0.150
0.016
0.492
1.745
2002
0.422
0.158
0.087
0.441
0.131
0.016
0.494
1.749
2003
0.430
0.154
0.090
0.455
0.145
0.019
0.511
1.805
2004
0.441
0.178
0.098
0.472
0.170
0.017
0.619
1.995
Panel B: Average Monthly Mortality, by Year and Cause of Death, per Million Persons
Motor
Vehicle
Cerebrovascular
Accident
Suicide
Homicide
Disease
1994
16.04
10.99
7.56
49.99
1995
15.32
10.90
7.04
50.98
1996
15.10
10.65
6.70
50.87
1997
15.21
10.59
6.21
50.78
1998
15.02
10.43
5.66
49.77
1999
14.64
9.88
5.20
51.78
2000
14.51
9.68
4.96
50.21
2001
14.44
10.10
5.23
48.64
2002
14.86
10.23
5.07
48.20
2003
14.60
10.17
4.93
46.13
2004
14.56
10.35
4.74
43.31
Notes:
1) The OPTN began reporting circumstances of death on April 1, 1994. Natural Causes was added
on October 25, 1999. Averages are taken across 50 states (plus Washington, D.C.) and 12 months of
the year, except where noted.
2) For the years 1994-1998, I use the following ICD 9th revision codes: MVA (E810-E825), suicide
(E950-E959), homicide (E960-E978), and cerebrovascular disease (430-438). For the years 19992004, I use the following ICD 10th revision codes: MVA (V02-V04, V09.0, V12-V14, V19.0-V19.2,
V19.4-V19.6, V20-V79, V80.3-V80.5, V81.0-V81.1, V82.0-V82.1, V83-V86, V87.0-V87.8, V88.0V88.8, V89.0, V89.2), suicide (X60-X84, Y87.0), homicide (X85-Y09, Y87.1), and cerebrovascular
disease (I60-I69).
3) Source: Author's calculation from SRTR and U.S. Vital Statistics data.
23

Table 1.2: Organ Donor Yield Rates by Cause of Death and Age, 1994-2004
Sample means for age and cause-of-death specific mortality (per million persons per month) are
shown in brackets.

MVA

Suicide

Homicide

Natural Causes

all ages

0.0307
[14.9114]

0.0153
[10.3454]

0.0175
[5.7119]

0.0100
[49.0252]

<18

0.0795
[1.9051]

0.0766
[0.4324]

0.0326
[0.6490]

0.2818
[0.1115]

18-34

0.0400
[5.1991]

0.0271
[3.0895]

0.0194
[2.8782]

0.2127
[0.2357]

35-49

0.0229
[3.2858]

0.0114
[3.2194]

0.0124
[1.4153]

0.1023
[1.5428]

50-64

0.0128
[1.9883]

0.0061
[1.8365]

0.0096
[0.4768]

0.0454
[3.9149]

65+

0.0020
[2.5264]

0.0008
[1.7629]

0.0025
[0.2794]

0.0017
[43.2185]

Notes:
1) Calculations are the ratio of organ donors to mortality counts by cause of death from April
1994 to 2004. The natural causes category only covers the period of November 1999 through
2004.
2) The mortality data are calculated using the ICD codes listed in the notes to Table 1.1.
3) Source: Author's calculation from SRTR and U.S. Vital Statistics data.

24

Table 1.3: The Effect of the Unemployment Rate on Mortality, per Million Persons
(robust standard error), [sample mean]; ***, **, and * indicate significance at 1, 5, and 10%
MVA
-0.3799***
(0.0530)
[14.9114]

Suicide
0.1233***
(0.0340)
[10.3454]

Homicide
0.2380***
(0.0437)
[5.7119]

<18

-0.0478***
(0.0152)
[1.9051]

0.0004
(0.0061)
[0.4324]

0.0152*
(0.0088)
[0.6490]

-0.0061*
(0.0033)
[0.1115]

-0.0052
(0.0047)
[0.0846]

18-34

-0.1291***
(0.0269)
[5.1991]

0.0025
(0.0173)
[3.0895]

0.1357***
(0.0245)
[2.8782]

0.0010
(0.0048)
[0.2357]

-0.0030
(0.0072)
[0.2077]

35-49

-0.0613***
(0.0196)
[3.2858]

0.0876***
(0.0182)
[3.2194]

0.0490***
(0.0182)
[1.4153]

0.0069
(0.0130)
[1.5428]

0.0186
(0.0202)
[1.4765]

50-64

-0.0262*
(0.0147)
[1.9883]

0.0291**
(0.0136)
[1.8365]

0.0246***
(0.0093)
[0.4768]

0.0029
(0.0204)
[3.9149]

-0.0116
(0.0323)
[3.8397]

-0.1162***
(0.0178)
[2.5264]

0.0029
(0.0133)
[1.7629]

0.0114**
(0.0056)
[0.2794]

-0.7816***
(0.0883)
[43.2185]

-0.8348***
(0.1372)
[41.8788]

all ages

65+

Cerebrovascular Disease
-0.7772***
-0.8359***
(0.0931)
(0.1447)
[49.0252]
[47.4885]

Years
1994-2004
1994-2004
1994-2004
1994-2004
1999-2004
Notes:
1) Estimation samples for Columns 1-4 consist of 50 states (plus Washington, D.C.) from April
1994 through 2004. Samples for Column 5 cover November 1999 through 2004. The unit of
observation is a state-month, and observations are weighted by state population.
2) The mortality data are calculated using the ICD codes listed in the notes to Table 1.1.
3) All models include indicators for state, month, and year as well as controls for the age
distribution of the state's population, and indicators for whether the state has a donor registry,
whether the state allows online donor registration, and whether organs can be donated without the
consent of family members of the prospective donor.
4) Standard errors, in parenthesis, are robust to clustering within state over time.
5) Source: Author’s calculation from U.S. Vital Statistics data.

25

Table 1.4: The Effect of the Unemployment Rate on Organ Donation, per Million Persons
(robust standard error), [sample mean]; ***, **, and * indicate significance at 1, 5, and 10%
Panel A: The Effect of the Unemployment Rate on Organ Donation, per Million Persons
Natural
Causes +
None of the
Above
-0.0012
(0.0142)
[0.8776]

MVA
-0.0152*
(0.0086)
[0.4301]

Suicide
0.0039
(0.0042)
[0.1621]

Homicide
0.0016
(0.0034)
[0.0888]

Natural
Causes
0.0485***
(0.0169)
[0.4334]

<18

-0.0127***
(0.0038)
[0.1403]

0.0011
(0.0016)
[0.0313]

-0.0015
(0.0013)
[0.0194]

0.0023
(0.0030)
[0.0211]

-0.0071**
(0.0029)
[0.0794]

18-34

-0.0055
(0.0047)
[0.1954]

0.0025
(0.0029)
[0.0819]

0.0033
(0.0024)
[0.0504]

0.0056*
(0.0033)
[0.0440]

-0.0022
(0.0028)
[0.1105]

35-49

0.0022
(0.0032)
[0.0675]

0.0001
(0.0017)
[0.0369]

0.0004
(0.0011)
[0.0150]

0.0073
(0.0075)
[0.1440]

0.0003
(0.0059)
[0.2877]

50-64

0.0001
(0.0014)
[0.0224]

-0.0003
(0.0009)
[0.0105]

-0.001**
(0.0004)
[0.0038]

0.0254***
(0.0094)
[0.1650]

0.0027
(0.0071)
[0.2956]

65+

0.0006
(0.0007)
[0.0045]

0.0005
(0.0004)
[0.0016]

0.0003*
(0.0002)
[0.0003]

0.0078*
(0.0044)
[0.0594]

0.0051
(0.0037)
[0.1044]

all ages

26

Table 1.4 (cont’d)
Panel B: Number of Organ Donors per Fatality

all ages

MVA
0.0400*
(0.0202)

Suicide
0.0317
(0.0323)

Homicide
0.0068
(0.0142)

Natural
Causes
-0.0580**
(0.0271)

Natural
Causes + NA
0.0015
(0.0182)

Notes:
1) Estimation samples for Columns 1, 2, 3, and 5 consist of 50 states (plus Washington, D.C.)
from April 1994 through 2004. Samples for Column 4 are from November 1999 through
2004. The unit of observation is a state-month, and observations are weighted by state
population.
2) All models include indicators for state, month, and year as well as controls for the age
distribution of the state's population, and indicators for whether the state has a donor registry,
whether the state allows online donor registration, and whether organs can be donated without
the consent of family members of the prospective donor.
3) Standard errors, in parenthesis, are robust to clustering within state over time.
4) Source: Author’s calculation from SRTR and U.S. Vital Statistics data.

Table 1.5: Average Number of Transplants per Donor
Panel A: By Cause of Death of Donor
All Causes
MVA
Suicide
Homicide
Natural Causes
Non-MVA
Child Abuse
None of the Above

Panel B: By Age of Donor

2.9
3.5
3.5
3.6
2.3
2.9
2.6
2.5

All Ages
0-17 years old
18-34 years old
35-49 years old
50-64 years old
65+ years old

Notes:
1) Source: Author's calculation from SRTR data.
27

2.9
3.4
3.7
2.9
2.2
1.3

APPENDIX 1.2

CHAPTER 1 FIGURES

28

Figure 1.1: Motor Vehicle Accident Mortality and the Unemployment Rate
400

7.0

300

6.0

200

5.0

100
4.0

0
3.0
-100
2.0

-200

1.0

-400

0.0
199404
199408
199412
199504
199508
199512
199604
199608
199612
199704
199708
199712
199804
199808
199812
199904
199908
199912
200004
200008
200012
200104
200108
200112
200204
200208
200212
200304
200308
200312
200404
200408
200412

-300

seasonally-adjusted MVA deaths

seasonally-adjusted unemployment rate

Note: Author’s calculation from U.S. Vital Statistics and Bureau of Labor Statistics data. The mortality data depicted are residuals
from a regression of deaths on a full set of month dummies. For a specific list of the ICD codes used, see the notes to Table 1.1.
29

Figure 1.2: Suicide Mortality and the Unemployment Rate
300

7.0

6.0

200

5.0
100
4.0

0
3.0
-100
2.0
-200

1.0

0.0
199404
199408
199412
199504
199508
199512
199604
199608
199612
199704
199708
199712
199804
199808
199812
199904
199908
199912
200004
200008
200012
200104
200108
200112
200204
200208
200212
200304
200308
200312
200404
200408
200412

-300

seasonally-adjusted suicide deaths

seasonally-adjusted unemployment rate

Note: Author’s calculation from U.S. Vital Statistics and Bureau of Labor Statistics data. The mortality data depicted are residuals
from a regression of deaths on a full set of month dummies. For a specific list of the ICD codes used, see the notes to Table 1.1.
30

Figure 1.3: Homicide Mortality and the Unemployment Rate
3000

7.0

2500

6.0

2000

5.0

1500
4.0

1000
3.0
500
2.0

0

1.0

-1000

0.0
199404
199408
199412
199504
199508
199512
199604
199608
199612
199704
199708
199712
199804
199808
199812
199904
199908
199912
200004
200008
200012
200104
200108
200112
200204
200208
200212
200304
200308
200312
200404
200408
200412

-500

seasonally-adjusted homicide deaths

seasonally-adjusted unemployment rate

Note: Author’s calculation from U.S. Vital Statistics and Bureau of Labor Statistics data. The mortality data depicted are residuals
from a regression of deaths on a full set of month dummies. For a specific list of the ICD codes used, see the notes to Table 1.1.
31

Figure 1.4: Cerebrovascular Disease Mortality and the Unemployment Rate
2000

7.0

1500

6.0

1000

5.0

500
4.0

0
3.0
-500
2.0

-1000

1.0

-2000

0.0
199404
199408
199412
199504
199508
199512
199604
199608
199612
199704
199708
199712
199804
199808
199812
199904
199908
199912
200004
200008
200012
200104
200108
200112
200204
200208
200212
200304
200308
200312
200404
200408
200412

-1500

seasonally-adjusted cerebrovascular disease deaths

seasonally-adjusted unemployment rate

Note: Author’s calculation from U.S. Vital Statistics and Bureau of Labor Statistics data. The mortality data depicted are residuals
from a regression of deaths on a full set of month dummies. For a specific list of the ICD codes used, see the notes to Table 1.1.
32

Figure 1.5: Motor Vehicle Accident Organ Donors and the Unemployment Rate
7.0

20

6.0

10

5.0

0

4.0

-10

3.0

-20

2.0

-30

1.0

-40

0.0
199404
199408
199412
199504
199508
199512
199604
199608
199612
199704
199708
199712
199804
199808
199812
199904
199908
199912
200004
200008
200012
200104
200108
200112
200204
200208
200212
200304
200308
200312
200404
200408
200412

30

seasonally-adjusted MVA donors

seasonally-adjusted unemployment rate

Note: Author’s calculation from SRTR and Bureau of Labor Statistics data. The organ donation data depicted are residuals from a
regression of deaths on a full set of month dummies. For a specific list of the ICD codes used, see the notes to Table 1.1.
33

Figure 1.6: Suicide Organ Donors and the Unemployment Rate
25

7.0

20

6.0

15

5.0

10
4.0

5
3.0
0
2.0

-5

1.0

-15

0.0
199404
199408
199412
199504
199508
199512
199604
199608
199612
199704
199708
199712
199804
199808
199812
199904
199908
199912
200004
200008
200012
200104
200108
200112
200204
200208
200212
200304
200308
200312
200404
200408
200412

-10

seasonally-adjusted suicide donors

seasonally-adjusted unemployment rate

Note: Author’s calculation from SRTR and Bureau of Labor Statistics data. The organ donation data depicted are residuals from a
regression of deaths on a full set of month dummies. For a specific list of the ICD codes used, see the notes to Table 1.1.
34

Figure 1.7: Homicide Organ Donors and the Unemployment Rate
7.0

15

6.0

10

5.0

5

4.0

0

3.0

-5

2.0

-10

1.0

-15

0.0
199404
199408
199412
199504
199508
199512
199604
199608
199612
199704
199708
199712
199804
199808
199812
199904
199908
199912
200004
200008
200012
200104
200108
200112
200204
200208
200212
200304
200308
200312
200404
200408
200412

20

seasonally-adjusted homicide donors

seasonally-adjusted unemployment rate

Note: Author’s calculation from SRTR and Bureau of Labor Statistics data. The organ donation data depicted are residuals from a
regression of deaths on a full set of month dummies. For a specific list of the ICD codes used, see the notes to Table 1.1.
35

Figure 1.8: Natural Cause Organ Donors and the Unemployment Rate
60

7.0

50
6.0
40
30

5.0

20
4.0

10
0

3.0

-10
2.0

-20
-30

1.0
-40
0.0
199404
199408
199412
199504
199508
199512
199604
199608
199612
199704
199708
199712
199804
199808
199812
199904
199908
199912
200004
200008
200012
200104
200108
200112
200204
200208
200212
200304
200308
200312
200404
200408
200412

-50

seasonally-adjusted natural cause donors

seasonally-adjusted unemployment rate

Note: Author’s calculation from SRTR and Bureau of Labor Statistics data. The organ donation data depicted are residuals from a
regression of deaths on a full set of month dummies. For a specific list of the ICD codes used, see the notes to Table 1.1.
36

CHAPTER 1 REFERENCES

37

CHAPTER 1 REFERENCES

Abadie, A. and S. Gay. 2006. “The Impact of Presumed Consent Legislation on Cadaveric
Organ Donation: A Cross-Country Study.” Journal of Health Economics, 25: 599-620.
Blumstein, A. and R. Rosenfeld. 1998. “Explaining Recent Trends in U.S. Homicide Rates.”
Journal of Criminal Law and Criminology, 88(4): 1175-1216.
Blumstein, A. and J. Wallman (eds). 2006. The Crime Drop in America, New York: Cambridge
University Press, rev. ed.
Bridges, F.S., P.S. Yip, and K.C.T. Yang. 2005. "Seasonal changes in suicide in the United
States, 1971-2000." Perceptual and Motor Skills, 100: 920-924.
Cassini, M.F., A.J. Cologna, S. Tucci Jr., R.B. Reis, A.A. Rodrigues Jr., H.J. Suaid, and A.C.P.
Martins. 2010. “Why Not to Use Kidney Grafts From Elderly Donors.” Transplantation
Proceedings, 42: 417-420.
Cheatwood, D. 1988. “Is There a Season for Homicide?” Criminology, 26(2): 287-306.
Di Cocco, P., G. Orlando, V. Rizza, L. De Luca, M. D’Angelo, K. Clemente, A. Famulari, and F.
Pisani. 2011. “Kidney Transplantation From Older Donors.” Transplantation Proceedings,
43: 1033-1035.
Figueiredo, F.M., F.B. Capaverde, G.G. Londero, M.G. Costa, P.E. Leaes, D.M.S. Oliveria, C.D.
Garcia, and V.D. Garcia. 2007. “Organ Donation in Suicides.” Transplantation Proceedings,
39: 344-345.
Frezza, E.E., L.R.N. Krefski, and C.P. Valenziano. 1999. “Factors influencing the potential
organ donation: a 6-yr experience of the New Jersey Organ and Tissue Sharing Network.”
Clinical Transplantation, 13: 231-240.
Guadagnoli, E., C. Christiansen, and C. Beasley. 2003. “Potential Organ-Donor Supply and
Efficiency of Organ Procurement Organizations.” Health Care Financing Review, 24(4):
101-110.
Hamermesh, D.S. and N.M. Soss. 1974. “An Economic Theory of Suicide.” Journal of Political
Economy, 82(1): 83-98.
Howard, D.H., L.A. Siminoff, V. McBride, and M. Lin. 2007. “Does Quality Improvement
Work? Evaluation of the Organ Donation Breakthrough Collaborative.” Health Services
Research, 42(6): 2160-2173.

38

Karatzas, T., J. Bokos, A. Katsargyris, K. Diles, G. Sotirchos, A. Barlas, E. Theodoropoulou, J.
Boletis, and G. Zavos. 2011. “Advanced Donor Age Alone Is Not a Risk Factor for Graft
Survival in Kidney Transplantation.” Transplantation Proceedings, 43: 1537-1543.
Khan, F.A., G. Engstrom, I. Jerntorp, H. Pessah-Rasmussen, and L. Janzon. 2005. “Seasonal
patterns of incidence and case fatality of stroke in Malmo, Sweden: the STROMA study.”
Neuroepidemiology, 24(1-2): 26-31.
Levitt, S. 2004. “Understanding why crime fell in the 1990s: four factors that explain the decline
and six that do not.” Journal of Economic Perspectives, 18(1): 163-190.
Liu, C., C.L. Chen, and D. Utter. 2005. Trend and Pattern Analysis of Highway Crash Fatality by
Month and Day. DOT HS 809 855, U.S. Department of Transportation, Washington, D.C.
Marconi, L., P. Moreira, B. Parada, C. Bastos, A. Roseiro, and A. Mota. 2011. “Donor Cause of
Brain Death in Renal Transplantation: A Predictive Factor for Graft Function?”
Transplantation Proceedings, 43: 74-76.
Miller, D.L., M.E. Page, A. Huff Stevens, and M. Filipski. 2009. "Why Are Recessions Good for
Your Health?" American Economic Review: Papers & Proceedings, 99(2): 122-127.
Myint, P.K., S.L. Vowler, P.R. Woodhouse, O. Redmayne, and R.A. Fulcher. 2007. “Winter
excess in hospital admissions, in-patient mortality and length of acute hospital stay in stroke:
a hospital database study over six seasonal years in Norfolk, UK.” Neuroepidemiology,
28(2): 79-85.
Nakaji, S., S. Parodi, V. Fontana, T. Umeda, K. Suzuki, J. Sakamoto, S. Fukuda, S. Wada, and
K. Sugawara. 2004. “Seasonal changes in mortality rates from main causes of death in Japan
(1970-1999).” European Journal of Epidemiology, 19(10): 905-913.
Rosenfeld, R. 2004. “The case of the unsolved crime decline.” Scientific American, 290(2): 8289.
Ruhm, C.J. 2000. "Are Recessions Good For Your Health?" Quarterly Journal of Economics,
115(2): 617-650.
Ruhm, C.J. 2003. "Good Times Make You Sick." Journal of Health Economics, 22(4): 637-658.
Ruhm, C.J. 2005. "Mortality Increases During Economic Upturns." International Journal of
Epidemiology, 34(6): 1206-1211.
Sheehy, E., S.L. Conrad, L.E. Brigham, R. Luskin, P. Weber, M. Eakin, L. Schkade, and L.
Hunsicker. 2003. "Estimating the Number of Potential Organ Donors in the United States."
New England Journal of Medicine, 349(7): 667-674.

39

Singhal, A.K., X. Sheng, S.G. Drakos, and J. Stehlik. 2009. “Impact of Donor Cause of Death on
Transplant Outcomes: UNOS Registry Analysis.” Transplantation Proceedings, 41: 35393544.
Tennenbaum, A.N. and E.L. Fink. 1994. "Temporal Regularities in Homicide: Cycles, Seasons,
and Autoregression." Journal of Quantitative Criminology, 10(4): 317-342.
Trudeau, R. 1997. "Monthly and daily patterns of death." Statistics Canada Health Reports,
9(1): 43-50.
Turin, T.C., Y. Kita, N. Rumana, Y. Murakami, M. Ichikawa, H. Sugihara, Y. Morita, N.
Tomioka, A. Okayama, Y. Nakamura, R.D. Abbott, and H. Ueshima. 2009. “Stroke case
fatality shows seasonal variation regardless of risk factor status in a Japanese population: 15year results from the Takashima Stroke Registry.” Neuroepidemiology, 32(1): 53-60.
Wang, Y., C.R. Levi, J.R. Attia, C.A. D’Este, N. Spratt, and J. Fisher. 2003. “Seasonal variation
in stroke in the Hunter Region, Australia: a 5-year hospital-based study, 1995-2000.” Stroke,
34(5): 1144-1150.
Warren, C.W., J.C. Smith, and C.W. Tyler. 1983. “Seasonal variation in suicide and homicide: a
question of consistency.” Journal of Biosocial Science, 15: 349-356.
Wendler, D. and N. Dickert. 2001. "The Consent Process for Cadaveric Organ Procurement:
How Does It Work? How Can It Be Improved?" Journal of the American Medical
Association, 285(3): 329-333.
Yang, B. 1992. "The Economy and Suicide: A Time-Series Study of the U.S.A." American
Journal of Economics and Sociology, 51(1): 87-99.
Zimring, F.E. 2006. The Great American Crime Decline, New York: Oxford University Press.

40

CHAPTER 2

NOT ON MY BACK ROADS:
ARE THERE SPILLOVER EFFECTS IN MOTORCYCLE HELMET LEGISLATION?

41

“Every year, millions of dollars are leaving our state on two wheels. Michigan's outdated
mandatory helmet law is sending motorcyclists racing across our borders and discouraging
motorcyclists in neighboring states from visiting our state.”

-“Opposing points of view: By giving motorcyclists choice on helmets, Michigan would gain
millions”; 30 June 2011, The Detroit Free Press (Rhoades, 2011).

2.1 Introduction
Motorcycle helmet laws remain a fiercely contested issue for public debate across the
country, over fifteen years since the federal government last removed its mandatory coverage
requirement for riders of all ages. Most recently, the legislature of the state of Michigan, which
has had a mandatory helmet law since 1969, proposed a repeal of the law.

26

Supporters of the

bill claim the state loses tourism dollars to states without mandatory laws, both by inducing
Michigan motorcyclists to leave the state and discouraging riders from other states from entering.
“(Tom Casperson, R-Escanaba) said he favors the bill because many Upper Peninsula
motorcyclists vacation in Wisconsin, where they aren't required to wear a helmet,” (Brenzing,
2011). The spillover of motorcycle traffic from a state with a mandatory helmet law to one
without is clearly a topic of concern to lawmakers. Although anecdotal evidence is strong, to
date there is no empirical evidence to support claims of a shift in riding location based on the
helmet laws of neighboring states.
26

See http://www.legislature.mi.gov/documents/2011-2012/billintroduced/Senate/pdf/2011SIB-0291.pdf for more details about Michigan Senate Bill No. 291.

42

Variation in helmet laws across state borders creates incentives for motorcyclists who
prefer helmetless riding to avoid mandatory law states. Tourism revenues, however, are not the
only potential economic impact of spillover traffic. Motorcycle lobbying groups have long
argued that helmets should be a personal choice, and that not wearing a helmet only affects the
individual, thereby negating the need for government action. Accidents involving helmetless
motorcyclists affect more than just the individual, and some courts have ruled that the
government has an interest in “minimizing the resources directly involved” in caring for these
injuries, including emergency response, medical care, and possible disability income transfers
(Jones and Bayer, 2007). If part of this increased expenditure is attributable to residents of other
states, governments may be justified in eliminating crossover incentives by imposing helmet
laws within their borders.
Previous literature has examined state motorcycle helmet laws and their impact on
fatality rates (Dee 2009, Dickert-Conlin, Elder, and Moore 2011, Houston and Richardson 2008,
Sass and Zimmerman 2000), but no attention has been paid to the interaction created by adjacent
states with different helmet laws. A spillover of motorcyclist traffic into states without a helmet
law could partially explain why fatality rates in those states are higher relative to states with
universal helmet laws, resulting in negative externalities due to the high rates of injury or death
of helmetless motorcyclists involved in accidents (GAO 1991).

27

My paper will directly

incorporate this incentive change into the analysis of motorcycle helmet legislation, testing for
evidence of spillover traffic.
The remainder of the paper proceeds as follows: the next section reviews the history and
effectiveness of motorcycle helmet legislation, and describes the conceptual framework behind

43

cross-state spillovers in motorcycle fatalities. Section 3 presents the data used and descriptive
evidence. Section 4 details my empirical specifications and estimation results. Section 5
concludes.

2.2 Background
Motorcycle Helmet Law History
Motorcycle helmet laws in the United States can be grouped into three categories:
universal laws, partial laws and no laws. Universal laws are the most encompassing, mandating
helmet use for riders of all ages. Partial helmet laws only cover part of the population, most
often younger riders.

28

Motorcycle helmet laws are enacted at the state level, although federal

legislation has played a significant role in shaping these state-level regulations.
Universal helmet laws began to appear in response to the 1966 Highway Safety Act,
which authorized the Secretary of Transportation to withhold up to 10 percent of federal highway
construction funds from any state that did not adopt universal helmet laws. By 1976, 48 states
and Washington, D.C. had adopted a helmet law, with California and Illinois being the
exceptions. In 1976, Congress passed an amendment to the Highway Safety Act that eliminated
the universal helmet law mandate and removed the Secretary of Transportation’s authority to
27

Increased accident response expenditures or costs to public health programs are possible
examples of negative externalities incurred due to helmetless riding.
28
The most common partial coverage requires helmets for operators age 17 and younger. Some
states have restrictions for riders under age 14, 18, or 20. A handful of states currently require
helmets for operators with instructional permits, less than one year of riding experience, or less
than $10,000 of personal injury insurance. Passengers riding with operators covered by helmet
laws are also often required to wear helmets (http://www.iihs.org/laws/helmet_history.html).

44

withhold federal funds for noncompliance. This amendment prompted 25 states to reduce their
helmet laws to partial coverage or no laws by 1980.
From 1980 to 1991, eight of the ten helmet law changes moved toward stronger helmet
laws (five states instituted a universal helmet law) amid mounting evidence that the helmet laws
were effective at reducing fatalities. In 1991 the federal government again made a push for
stronger helmet regulations by passing the Intermodal Surface Transportation Efficiency Act of
1991, a two-pronged approach aimed at getting states to adopt laws encouraging public safety.
Under the act, states that enacted both universal helmet laws and seat belt laws would be eligible
to receive federal grant money and states that did not comply would be subject to a 3 percent
reallocation of federal highway funds towards highway safety programs. California and
Maryland each reacted by instituting a universal helmet law in 1992.
In November 1995, Congress removed the threat of reallocating funds from noncompliant states with the passage of the National Highway System Designation Act.

29

Since

then, six states switched from universal to partial-coverage helmet laws and only two states,
Louisiana and Colorado, adopted stronger helmet laws. Today, only 20 states and the District of
Columbia require all riders to wear helmets. 27 states have partial coverage laws, while Illinois,
Iowa, and New Hampshire do not require any riders to wear a helmet.

30

Helmet Use and Fatality Rates
Motorcycle helmet laws matter for behavior changes: affecting people's decision to wear
a helmet and whether or not to ride. These decisions translate to an outcome I can measure:
29

New Hampshire’s partial coverage law was no longer binding without the federal penalties
(http://www.iihs.org/laws/helmet_history.html).

45

fatalities. Numerous studies, mostly observational in design, show that helmet usage in states
with universal helmet laws are very close to 100 percent, while helmet usage following repeal is
approximately 55 percent (Berkowitz 1981, Dare et al. 1978, Gilbert et al. 2008, Kraus et al.
1995, Lund et al. 1991, Preusser et al. 2000, Struckman-Johnson and Ellingstad 1980, Ulmer and
Northrup 2005, and Ulmer and Preusser 2003). Moreover, the change in helmet use is almost
immediate following universal law adoption or repeal (see Gilbert et al. 2008). In addition to
affecting the intensive margin for helmet use within the riding population, helmet mandates also
affect the extensive margin, creating fluxuations in the size of the riding population. Recent
estimates suggest the motorcycle registrations per capita decrease by between 14 percent
(Dickert-Conlin et al. 2011) and 22 percent (Dee 2009) following the enactment of universal
helmet laws. These results, combined with the observational studies on helmet use, suggest that
a non-negligible proportion of motorcyclists prefer not to wear helmets, and alter their behavior
in response to changes in helmet laws.
Recent work on motorcycle helmet laws provide evidence about the effectiveness of
statewide mandates of helmet use. Using a state-year panel data specification from 1976 to
1997, Sass and Zimmerman (2000) find that universal helmet laws decrease per capita
motorcyclist fatalities by 29 to 33 percent. Using a similar specification from 1976 to 2004,
Houston and Richardson (2008) find reductions in fatality rates of 29 percent. Dee (2009) and
Dickert-Conlin et al. (2011) also estimate fatality reductions of 26 to 32 percent and 39 percent,
respectively.
30

31

While these studies testify to the effectiveness of mandatory helmet laws, none

See http://www.iihs.org for more details on the history of motorcycle helmet laws.

31

Dee (2009) covers the years 1988 to 2005 while Dickert-Conlin et al. (2011) uses data from
1994 to 2007.

46

of them considers the impact statewide helmet laws have on the fatality rates of motorcyclists in
adjacent states.
A final area of the helmet literature attempts to estimate a helmet’s ability to reduce a
rider's fatality risk. Specifications exploiting variation in survival and helmet use among withinvehicle panel data provide convincing evidence of the effectiveness of motorcycle helmets at
reducing fatality risk. Using this approach along with data from 1975 to 1986, Evans and Frick
(1988) find that wearing a helmet reduces a rider’s fatality risk by 28 percent. With data from
1993 to 2002, Deutermann (2004) finds that helmet effectiveness for reducing fatalities and brain
injuries to be 37 percent and 69 percent respectively. Most recently, Dee (2009) finds that
helmets reduce fatality risk by 34 percent using data from 1988 to 2005. With an alternative
meta-analysis format, a Cochrane Collaboration found helmets reduce risk of death by 42 percent
and risk of head injury by 69 percent (Liu et al. 2008). These results demonstrate that
motorcyclists crossing a state border in order to remove their helmet would expose themselves to
greater risk of fatality in the event of an accident. The following section will more carefully
develop the conceptual framework of incentives created by state-to-state variation in motorcycle
helmet laws and lay out empirical predictions of the resulting incentive structure.

Incentives and Predictions for Cross-State Spillover
Although to my knowledge no other study of motorcycle fatalities has considered the
effect of variation in helmet laws across states, the phenomenon of crossing borders to evade
laws of one's own state has been tested in other literatures. In a review of the welfare migration
literature, Brueckner (2000) finds evidence for strategic interaction between states in the choice
of welfare benefits, but only mixed evidence for actual migration by welfare recipients.

47

Lovenheim (2008) estimates that between 13 and 25 percent of consumers purchase cigarettes in
border localities to avoid higher tax levels in their own state, and that the relationship between
quantity demanded and home state price is sensitive to the distance to the closest lower-price
border.
Consider two states sharing a border, both with no helmet law (or a partial helmet law
covering only young riders). Then suppose one state enacts a universal helmet law covering all
riders. Let State N designate the state with no helmet law and State H designate the state with a
universal law. In this case, motorcyclists in State N are free to ride with or without a helmet as
they choose. Motorcyclists in State H are faced with the decision to wear a helmet while riding,
ride illegally without a helmet, or discontinue riding in State H. As seen above, only a small
percentage of riders in this circumstance choose to ride illegally without a helmet. Although the
choice to discontinue riding cannot be measured directly, evidence on registration numbers
suggest about 14 to 22 percent of motorcyclists stop riding following the implementation of a
universal helmet law.

32

Whether or not motorcyclists from State H cross into State N depends

on, among other things, the strength of their preferences for helmetless riding and the travel
distance to State N. If cross-state spillover of this nature does occur, given the strong evidence
showing helmetless riding is correlated with fatalities, I expect that some motorcyclists from
State H will be involved in fatal accidents in State N.
Increased fatal accidents in State H are also possible, as any rider hoping to evade the
helmet law in State H must ride to and from the bordering state. Previous work on increased
travel for access to alcohol finds effects of this type. In particular, with respect to minimum
legal drinking age (MLDA) laws, Lovenheim and Slemrod (2010) find that restrictions on

48

drinking age do not reduce youth involvement in fatal accidents for counties within 25 miles of
lower-MLDA state. They also find that fatal accident involvement increases for 18 and 19-yearold drivers in this circumstance. Baughman et al. (2001) examine variation in local access to
alcohol within the state of Texas, finding that the sale of beer and wine within a county actually
decreases alcohol-related accidents by 8 percent. These results are both consistent with people
driving less when alcohol is more accessible, which in turn reduces accidents. Therefore,
enactment of a helmet law that induces people to ride to and from a bordering state to avoid the
helmet mandate may result in increased accidents and fatalities in the state that enacted the law.
Having a larger percentage of neighboring states with a universal helmet law is expected
to lead to increased incentives for spillover traffic and therefore more fatalities, particularly from
non-resident motorcyclists. The rationale for this hypothesis is that more motorcyclists in
neighboring states would be exposed to the situation described above, preferring to ride
helmetless and legally unable to do so in their own state. Some of these motorcyclists are
expected to be within a critical travel distance from a neighboring state with more lenient laws.
On the other hand, having a larger percentage of neighboring states with a universal helmet law
could instead result in reductions to the riding populations in those states, leading to decreased
motorcyclist fatalities. Therefore, the overall effect of neighboring state laws on own state
fatalities is ambiguous.

32

Changes in registrations could also represent a change in the number of motorcycles per
owner. It is also possible that motorcyclists ride without proper registration.

49

2.3 Data and Descriptive Evidence
The fatality data used in this paper come from the National Highway Traffic Safety
Administration (NHTSA) Fatality Analysis Reporting System (FARS), available for public
download at http://www.nhtsa.gov/FARS. The full FARS file contains detailed information on
every accident on public roads that leads to at least one death within 30 days. Among the
variables contained in the full file are both the state of crash occurrence and the vehicle license
state. For this analysis, I aggregate the data to annual counts of fatalities of operators and
passengers on motorcycles for the 48 contiguous states for the years 1976 to 2008 (n = 1,584). A
subset of the analysis will focus on non-resident fatality counts, which is calculated as a
mismatch of license state and state of crash occurrence. Data on statewide motorcycle helmet
law histories come from the Insurance Institute for Highway Safety website
(http://www.iihs.org/laws/default.aspx).
Before testing my predictions with regression models, I illustrate my identification
strategy with a few cases where we might expect cross-state spillover effects. Figures 2.1
through 2.6 show annual trends in non-resident motorcycle fatalities, both in levels and as a
percentage of all motorcycle fatalities, for six different states from 1990 to 2008.

33

Figure 2.1

represents Arizona, which is the only one of California’s neighbors without a universal helmet
law at the time of California’s universal helmet law adoption in 1992. If cross-state spillover
was to occur, Arizona would be expected to experience an increase in the number and percentage
of non-resident motorcycle fatalities beginning in 1992, as California’s riding population looks
to avoid the new helmet mandate. While there is an increase in non-resident fatalities for 1992
33

A non-resident fatality is one in which the state of crash occurrence does not match the state
of residence for the deceased, and is calculated by the author using the detailed FARS data files.
The years 1990 to 2008 were selected for simplification and consistency.
50

and 1993, there were also large decreases in both 1991 and 1994. Figure 2.2 represents
Pennsylvania, which repealed its own helmet law in September 2003. The surrounding states of
Maryland, New Jersey, New York, and West Virginia all have universal helmet laws at this time,
with New Jersey being a particularly interesting border state because of the large number of New
Jersey residents living in the greater Philadelphia region in close proximity to the Pennsylvania
border. Non-resident motorcyclist fatalities in Pennsylvania initially drop in 2004, but display a
large and steady upward trend thereafter, supporting my hypothesis. Figure 2.3 shows data for
Ohio, which has had a partial helmet law, covering riders 17 and under, since 1978. Ohio is of
interest because two of its neighboring states repealed universal helmet laws in this time frame,
each with populous metropolitan areas near the border to increase likely pass-through. In
addition to Pennsylvania mentioned above (September 2003), Kentucky repealed its universal
helmet law in July 1998. Each of these helmet law repeals are expected to decrease non-resident
fatality rates in Ohio, as Pennsylvania and Kentucky residents are no longer required to cross
state borders in order to ride helmetless. Despite a slight initial decrease following Kentucky’s
repeal in 1998 and in the year in 2003 when Pennsylvania repealed, Ohio’s non-resident
motorcyclist fatalities display sharp increases in subsequent years. Figure 2.4 shows nonresident fatality trends for Oklahoma, which also bordered two states that repealed universal
helmet laws since 1990, Arkansas and Texas, both in 1997. Similar to the Ohio case, nonresident fatalities are expected to decrease following repeal in neighboring states. While there
appears to be a large spike in the fraction of motorcyclist fatalities attributable to non-residents in
1998, there is only a small change in non-resident fatality counts. Figure 2.5 represents Texas.
In addition to Texas’ own universal helmet law repeal in 1997, the state shares a border with
Louisiana. Louisiana repealed a universal helmet law in August 1999, which it reinstated in

51

August 2004. As such, non-resident fatalities are expected to decrease in 2000 and increase in
2005. However, Texas’ non-resident fatalities slightly increase following Louisiana’s repeal
before large increases after the reinstatement. Lastly, Figure 2.6 displays data for New Mexico,
which also borders Texas, with likely pass-through from El Paso. New Mexico has had a helmet
law covering only those 17 and younger since 1978. Therefore, decreases in non-resident
fatalities are expected following the Texas helmet law repeal, as Texas residents no longer have
incentives to cross in New Mexico to remove their helmets. The data show no such decrease in
New Mexico, which actually experienced increased non-resident fatalities shortly after the Texas
repeal. With no conclusive evidence of cross-state spillover in Figures 2.1 - 2.6, I turn to
estimating the effect in a regression framework.

2.4 Methods and Results
My main estimating equation for assessing the effect of motorcycle helmet laws on
fatality rates is as follows:

(1)
where yit is the natural log of motorcyclist fatalities per capita in state i and year t, αi and δt
represent a full set of state and year fixed effects, lawit is an indicator for whether or not state i
had a universal helmet law for at least six months in year t, and Xit represents a vector of timevarying, state-level observables expected to influence fatality rates.

34

34

The vector Xit includes

For the purposes of this study, partial coverage helmet laws are assumed to be equivalent to
having no helmet law due to enforcement difficulties and the high proportion of the riding
population facing no restrictions.

52

state population, annual precipitation, heating degree days, gas prices, the state unemployment
rate, an indicator for whether or not the state speed limit is 65 mph or above, and an indicator for
whether or not the state had a primary enforcement seat belt law.

35

Some specifications also

include controls for state motorcycle registration levels per capita. While it may be preferable to
normalize the dependent variable by a measure of motorcycle miles travelled in addition to, or in
place of, normalization by state population levels, estimates of motorcycle miles travelled are not
available.

36

It would also be of interest to include a measure of motorcycle miles travelled in the

set of explanatory variables to control for the amount of riding activity in the state.
The effect of helmet laws in neighboring states is captured by the term,
, where lawjt represents the helmet law in other states j, j ≠ i, and ωij
represents a weighting scheme corresponding to the importance state i attributes to state j’s
helmet law. Due to the difficulty in creating a perfect variable for neighboring state helmet laws,
I will implement four different configurations. One simple formulation is to set lawjt equal to
one for each state with a universal motorcycle helmet law and use ωij equal to 1/ni for each state
j sharing a border with state i, and zero otherwise, where ni represents the number of states that

35

Heating degree days are intended to capture heating requirements for a home or business.
Heating degree days are calculated by taking the average daily temperature and subtracting it
from the base temperature of 65 degrees. The result would be the number of heating degree
days for that day. Average daily temperatures greater than 65 degrees are counted as 0 heating
degree days. The observations from multiple weather recording stations are combined to form a
single statewide heating degree day observation, weighted by state population. Heating degree
days are measured in 1,000s in this paper. Primary enforcement stipulates a driver may be
stopped and issued a citation for not wearing a seat belt. With secondary enforcement, a driver
must be stopped for another violation before a seat belt citation may be issued.
36
Houston and Richardson (2008) normalize fatalities by population, motorcycle registrations,
and vehicle miles travelled with only small changes in results.

53

share a border with state i. This configuration mimics the “contiguity” weighting scheme
utilized in the “race to the bottom” welfare migration literature.

37

The second formulation sets

lawjt equal to the number of shared border miles with an adjacent state containing a universal
helmet law.

38

The weights, ωij, are slightly modified to equal 1/mi for each state j sharing a

border with state i, and zero otherwise, where mi represents the total number of shared border
miles for state i. The third formulation sets lawjt equal to the number of National Highway
System (NHS) roadway crossings from state j with a universal helmet law into state i.

39

The

weights, ωij, are again modified, this time to equal 1/ci for each state j sharing a border with state
i, and zero otherwise, where ci represents the total number of NHS crossings for state i. The
final configuration will be identical to the first with different weights. In this specification, ωij is
equal to si/ni for each state j sharing a border with state i, and zero otherwise, where ni again
represents the number of states that share a border with state i and si is an indicator equal to one
if state i doesn't have a universal helmet law. This weighting scheme has the effect of
controlling for the fraction of neighboring states that have stronger helmet laws than state i, and
37

See Brueckner (2000) for more details.

38

Data on state border lengths came from the website of Dr. Thomas Holmes at the University
of Minnesota, accessed April 1, 2008. 109 borders exist between states, and the data contained
the length (in miles) of each border. Please see
http://www.econ.umn.edu/~holmes/data/BorderData.html for more details.
39
National Highway System crossings in this study include the Eisenhower Interstate Highway
System, other NHS routes and non-interstate STRAHNET routes. This data was obtained from
the U.S. Department of Transportation, Federal Highway Administration website, accessed May
15, 2011. Please see http://www.fhwa.dot.gov/planning/nhs for more details.

54

will test the possibility that the helmet laws of adjacent states will only affect state i’s fatality
rates if state i does not have a universal helmet law of its own.

40

The state fixed effects, αi, account for unobserved differences between states that do not
change over time, such as safety preferences. The full set of year dummies, δt, will control for
time-specific factors that are constant among all states, such as changes in motorcycle engine
technology or the size of other vehicles on the road. Variation within states over time identifies
the model. Of principle interest, estimates of γ capture the effect of a statewide helmet mandate
on motorcyclist fatalities while estimates of λ capture the effect of helmet laws in adjacent states.
Using the natural log of motorcyclist fatalities per capita as the dependent variable allows the
results to be easily compared to previous estimates of the effect of a state's own helmet law,
while observing any additional effect due to the helmet laws of neighboring states. Each of the
configurations for neighboring state helmet laws controls for the expected exposure of state i to
non-resident motorcyclists in a different way. States with long shared borders, such as those in
the western United States, tend to have fewer border crossings than the more highly populated
eastern states. Having a high number of border-crossing roadways serves as a proxy for the
population density near the border and the ease of travel from one state to another. Having a
higher percentage of adjacent states containing universal helmet laws is expected to induce more
non-resident motorcyclists to enter state i, some of which will be involved in fatal accidents as a
result. Therefore, estimates of λ are hypothesized to have a positive sign.
Table 2.1 presents summary statistics for the included variables.

41

Of note, there are 76

annual motorcycle fatalities in an average state and the average fraction of motorcycle fatalities
40

This is because riders forced to wear a helmet following a mandate do not face spillover

55

involving non-residents is 0.122. A universal helmet law is in effect in 46.2 percent of all stateyear observations in the data. States have an average of 4.5 neighboring states, with the
minimum being 1 (Maine) and the maximum being 8 (Missouri and Tennessee). The average
fraction of neighboring states with a universal helmet law is 0.489.

The Effectiveness of Statewide Motorcycle Helmet Laws
Results from estimating (1) are shown in Table 2.2 using the natural log of motorcyclists’
fatalities per capita as the dependent variable.

42

In column (1), the model is estimated without

controlling for the helmet laws of adjacent states to ensure that the estimates are consistent with
those of previous work on the effectiveness of motorcycle helmet laws. As seen in the first row,
adopting a universal helmet law is associated with an approximate 36.4 percent decrease in state
motorcyclist fatalities per capita, which is statistically significant at the 1 percent level,
consistent with previous estimates.

43

There are also highly significant estimates for population,

precipitation, and the state unemployment rate. The estimate of -0.3636 (0.1355) for the
coefficient on the natural log of total state population implies that a 1 percent increase in state
population is associated with a decrease in per capita motorcycle fatalities of 0.36 percent. This
statistically significant estimate suggests there is a difference in fatality patterns between states
with large populations, as the dependent variable has already been normalized by state
population. One explanation for this effect is that states with large urban populations may have

incentives if all neighboring states also have a universal helmet law in place.
41
Data sources for all variables are available in Table 2.6.
42

The estimates in Table 2.2, columns (1) and (2) of Table 2.3, and Table 2.5 were generated
using the command xtreg, fe in Stata SE 11.2.
43
Standard errors shown throughout the paper are robust to clustering at the state level, which
allows for arbitrary correlation in the error terms within state over time.
56

significantly different motorcycle riding patterns, with a smaller proportion of the population “atrisk” for fatal accidents.
There are two obvious ways precipitation levels could affect motorcycle fatality levels.
First, precipitation during a motorcycle ride would make conditions more dangerous, and would
lead to increased accident levels. Secondly, motorcyclists most likely avoid riding during these
spells of inclement weather, reducing the “at risk” riding population and reducing accident
levels. States with large amounts of snowfall would therefore be expected to have shorter riding
seasons, and fewer fatal accidents, ceteris paribus. The precipitation variable used in this study
is an aggregate measure of state-year precipitation levels, and not an indicator for inclement
weather at the time of any specific fatal accident, meaning the latter explanation is probably
more credible. The estimate of -0.1737 (0.0438) for the coefficient on the natural log of
precipitation shows that a 10 percent increase in precipitation would decrease motorcyclist
fatalities by 1.74 percent. Once precipitation levels, which include snowfall, are controlled for,
there is no statistically significant effect of heating degree days on motorcycle fatality rates,
although the estimates corresponding to heating degree days are negative as expected. This
result suggests the variable is capturing a reduction in riding season faced by northern states as
opposed to the decreased safety motorcyclists face in inclement weather.
An increase in the unemployment rate of 1 percentage point is associated with a decrease
in fatalities of 2.94 percent, significant at the 1 percent level. This result is consistent with
previous work on cyclical trends in mortality rates (Miller et al. 2009, Moore 2011, Ruhm 2000),
which explains increased motor vehicle accident mortality rates by the roadway congestion
caused by increases in both business and leisure travel during economic expansions. Higher gas
prices are also expected to decrease motorcycle fatalities rates, as previous research has found

57

higher gas prices decrease the total amount of vehicle miles travelled (Grabowski and Morrisey,
2004). The estimates for the coefficient on the natural log of gas price are negative, but
statistically significant, in all specifications.
Column (2) of Table 2.2 replicates the estimates in column (1), including an additional
control variable measuring the number of motorcycle registrations per 10,000 persons. This is
done to test the possibility that changes in registration rates are responsible for changes in
motorcycle fatality rates, and that the helmet law indicators merely absorb the large changes in
registrations that follow a law change. The results are nearly identical, both in terms of
significance and point estimates, albeit with a slight reduction in the absolute value of the point
estimates. The estimate for the effect of registrations is 0.0965 (0.0667), indicating that the
universal helmet law variables are not inflated due to the omission of registrations from the
model in the other columns, as might be expected if a helmet law enactment resulted in an
immediate change in the number or type of motorcycle riders on the road.

44

Columns (3) through (6) present the results for estimating the full model in equation (1).
Each column corresponds to a different formulation of helmet laws in adjacent states, as
described above. A positive sign for the coefficient representing neighboring state laws would
support the spillover hypothesis. The results are nearly identical throughout. The effect of a
state's own helmet law on fatalities ranges from a 33.7 percent to 36.0 percent decrease,
representing almost no change from columns (1) and (2). The effect of an increase in the
strictness of neighboring state helmet laws is positive in just one of the four formulations,
although insignificantly different from zero in all four cases. The point estimate in column (3)
44

Motorcycle rider type could change in one of two ways following helmet law enactment. The
law could deter more risk-seeking riders, leaving a safer riding population. Alternatively,

58

for the fraction of adjacent states with a universal helmet law implies that an increase of 0.2
45

would decrease motorcyclist fatalities by 2.1 percent (= -0.1067 × 0.2 × 100).

In columns (4)

and (5), increases of 0.2 in the fraction of border length shared with, or road crossings into, a
state with a universal helmet law yield corresponding changes in motorcyclist fatalities of -0.9
percent (= -0.0434 × 0.2 × 100) and -1.9 percent (= -0.0945 × 0.2 × 100), respectively. Finally,
column (6) shows that fatalities increase by 1.3 percent (= 0.0625 × 0.2 × 100) when the fraction
of adjacent states with stronger laws increases by 0.2. Point estimates for the other independent
variables also follow the patterns seen in columns (1) and (2), both in magnitude and
significance. These results suggest the helmet laws of adjacent states do not have a strong effect
on the overall motorcyclist fatality rate, after controlling for a state's own helmet law.

Estimating the Fraction of Motorcycle Fatalities Involving Non-residents
In this section, I specify an alternative model for testing cross-state spillover effects with
respect to motorcycle helmet laws. Rather than attempting to identify changes in the full
motorcycle fatality rate, I will estimate the fraction of motorcycle fatalities involving nonresidents. Over the sample range, an average of 6 non-resident fatalities occurred for each stateyear combination, accounting for 12.2 percent of annual motorcyclist fatalities, on average.
I first estimate a linear fixed effects model identical to equation (1) with yit now equal to
the fraction of motorcyclist fatalities involving non-residents in state i and year t. The estimation
results for this linear model are displayed in column (1) and (2) of Table 2.3. The presence of a

following the Peltzman risk-compensating hypothesis (Peltzman, 1975), riders who stay may
exhibit more risky riding patterns as a result of feeling safer.
45
The mean state has 4.5 neighboring states, meaning a change of 0.2 represents a change in the
helmet law of approximately one neighboring state.

59

universal helmet law, either within the state or among its neighbors, has a positive but
statistically insignificant affect in this model. Population is still highly significant, where a 1
percent increase is associated with a decrease of 0.0675 (0.0227) in the fraction of motorcyclist
fatalities involving non-residents, representing a 55.3 percent (= 0.0675 / 0.122) decrease relative
to the sample mean. While it seems plausible that a change in a state's population, and therefore
potential riding population, is an important determinant of the fraction of non-resident fatalities,
the size of this estimate could indicate a problem in the estimation strategy.
With a dependent variable bounded between 0 and 1, a linear model cannot provide
accurate estimates at all points of the distribution, as literal interpretation of the estimates would
predict levels of yit above 1 or below 0. Following Papke and Wooldridge (2008), I next turn to
estimation that accounts for the fractional nature of the dependent variable, using a fractional
probit model of the form:

(2)
where yit is again the fraction of motorcyclist fatalities involving non-residents. Xit now includes
all state-year observables as before as well as the controls for a state’s own helmet law and the
helmet laws of adjacent states.

represents a set of time-demeaned variables contained in Xit,

and is similar to the state-level fixed effects in the linear regression framework. A full set of
year indicators are also included in the estimation.

46

Columns (3) through (6) of Table 2.3 display the results from estimating equation (2). I
display results for only two of the adjacent state helmet law specifications for brevity. Columns

60

(3) and (4) of Table 2.3 represent the marginal effects using a pooled fractional probit (PFP)
estimator, obtained by maximizing the pooled probit log-likelihood.

47

Columns (5) and (6) use a

generalized estimating equation (GEE) approach with an exchangeable working correlation
48

matrix.

The GEE approach is asymptotically equivalent to a multivariate weighted nonlinear

least squares in this case, and will potentially increase the efficiency of the estimates.

49

The

point estimates follow a very similar pattern, with the exception of those in column (3), where an
increase of 0.2 in the fraction of neighbors with a universal helmet law is estimated to increase
the fraction of non-resident fatalities by 11.4 percent (= (0.0695 × 0.2) / 0.122) relative to the
sample mean. While this result is consistent with the story that an increase in the percentage of
adjacent states with a universal helmet law will induce more non-resident motorcyclists to cross
the border to ride helmetless, it is the only one of six specifications resulting in an estimate
significantly different than zero. The estimates of Table 2.3 do not provide conclusive evidence
about the effect of helmet laws on non-resident motorcyclist fatalities. It is conceivable that the
presence of helmet laws has a significant effect on both the numerator and denominator of the
dependent variable, potentially masking the relationship of interest. In the next section, I test for
this possibility by estimating the number of non-resident motorcyclist fatalities separately with
count data methods that account for the panel nature of my data.

46

The subscript a denotes the use of a scale factor. The results reported in Table 2.3 are average
partial effects that depend on the scale factor, and are comparable to the linear fixed effect
estimates.
47
The estimates in columns (3) and (4) of Table 2.3 were generated using the command glm
with family(binary) and link(probit) in Stata SE 11.2.
48
The estimates in columns (5) and (6) of Table 2.3 were generated using the command xtgee
with family(binary), link(probit) and correlation(exchangeable) in Stata SE 11.2.

61

Fixed Effects Poisson Estimation of Motorcyclist Fatalities
To address the possibility that the numerator and denominator in the above dependent
variable move together, I will separately estimate the number of non-resident motorcyclist
fatalities and the total number of motorcyclist fatalities using count data methods. Count data
methods are necessary here because the nature of the dependent variable requires that the
conditional mean be non-negative for all values of Xit, and a linear model can result in negative
predicted values. Developed by Hausman et al. (1984), I estimate a fixed effects Poisson model
based on the following conditional mean equation:

(3)

)

where Xit includes all explanatory variables from equation (1) as well as the controls for helmet
laws in state i and j≠i, and a full set of year indicators. ci represents unobserved heterogeneity at
the state level, and takes the form ci = exp(αi) in practice. yit will be either the number of
motorcyclist fatalities or the number of motorcyclist fatalities involving non-residents. The fixed
effects Poisson estimator has very strong robustness properties, requiring only proper
specification of the conditional mean for consistency (Wooldridge 1999).
Table 2.4 displays results from estimating equation (3).

51

50

The dependent variable in

columns (1) and (2) is the number of non-resident motorcyclist fatalities, while columns (3) and
(4) correspond to estimates for the total number of motorcyclist fatalities. The estimates for the
coefficient on a state's own helmet law are all negative and statistically significant. This
49

See Papke and Wooldridge (2008) for more details on the fractional probit estimation.

50

The estimation is robust to serial correlation and misspecification of the Poisson distribution
by using the standard errors from Wooldridge (1999).

62

suggests the fractional probit estimates of the previous section failed to capture a significant
effect of helmet laws due to shifts in both the numerator and denominator of the dependent
variable. A similar conclusion can be reached for the effect of the state unemployment rate,
which is associated with statistically significant decreases in both non-resident and total
motorcyclist fatality counts. The pattern of estimates for the natural log of state population show
that states with large populations are expected to have more motorcycle fatalities, but are not
expected to face significant differences in non-resident fatalities. Due to the fact that the
dependent variable is no longer normalized by state population, the positive and significant
estimates in columns (3) and (4) do not contradict the negative and highly significant estimates
seen in Table 2.3 for the population variable, as they capture different effects. Separately
estimating the numerator and denominator did not lead to a significant estimate for the
coefficient on the helmet laws of adjacent states.

The Effect of Helmet Laws on Riding Population
The results in Tables 2.2 though 2.4 provide evidence that a state's own helmet law
substantially affects motorcyclist fatalities, but the helmet laws of adjacent state's do not have a
significant differential effect. As discussed earlier, enacting a universal helmet law affects
motorcycle fatality rates in two distinct manners, by reducing the probability of death conditional
on an accident, and by reducing the size of the riding population. The last possibility to explore
is whether or not universal helmet laws in adjacent states affect fatality rates by reducing the
riding population of those states, and therefore the pool of potential border-crossing
motorcyclists. I next turn to estimating the effect of helmet laws on motorcycle registrations
51

The estimates in Table 2.4 were generated using the command xtpqml, fe in Stata SE 11.2.

63

based on the linear model in equation (1). Table 2.5 shows the results of estimating the equation
(1) with the natural log of motorcycle registrations per 10,000 persons as the dependent variable.
The estimates in column (1) show that universal helmet laws are associated with a 24 percent
reduction in motorcycle registrations, which is consistent with Dee (2009) and Dickert-Conlin et
al. (2011). Columns (2) and (3) add in the controls for helmet laws in adjacent states. Similar to
fatalities, motorcycle registrations do not appear to respond to changes in the helmet laws of
neighboring states.

2.5 Conclusions
Although it has been nearly 50 years since the federal government first mandated
universal motorcycle helmet laws, the merit of such government involvement continues to be
debated. The argument has traditionally pitted civil liberties against the additional costs to
public health programs and emergency response imposed by helmetless motorcyclists, and is
now being updated to include the cost of lost state revenue. Despite the fact that empirical
studies have shown both helmets and helmet laws to be effective at substantially reducing
fatalities, no previous work has incorporated the helmet laws of adjacent states into the analysis.
Neighboring states with different helmet laws facilitate an incentive structure for motorcyclists to
cross state borders to avoid helmet mandates.
I find a state’s own helmet law to be an important factor in determining motorcyclist
fatality rates, as it strongly influences both the helmet status, and therefore safety, of an
individual rider in the event of a crash and the size of the local riding population. Consistent
with a larger literature, my results show a universal law is associated with a decrease in
motorcyclist fatalities of 34.1 percent. After conditioning for a state’s own helmet law, the laws

64

of adjacent states do not appear to have an additional effect on fatality rates. In addition, I find
that neither a state's own helmet law, nor those of its neighbors affect the fraction of motorcyclist
fatalities involving non-residents. Lastly, I find own-state helmet laws are associated with
significant reductions in non-resident fatality counts, although the helmet laws of adjacent states
remain insignificant. While cross-state spillover effects could be used to support strategic
interaction among state governments, either to reduce negative externalities or boost tourism
revenues, my examination of fatality patterns shows no conclusive evidence of a shift in
motorcycle riding location based on the helmet laws of neighboring states. I am also unable to
rule out the possibility that motorcyclists do respond to changes in helmet laws by shifting riding
locations, but are also involved in increasing levels of fatal accidents in their own state as a
result.

65

CHAPTER 2 APPENDICES

66

APPENDIX 2.1

CHAPTER 2 TABLES

67

Table 2.1: Summary Statistics
Variable

Mean

Standard Deviation

Min

Max

Motorcycle fatalities
(annual count)

76.083

98.283

1

879

Motorcycle fatalities
per 100,000 persons

1.537

0.746

0.127

5.156

0.462

0.499

0

1

5,343,621

5,666,977

396,126

36,580,371

Motorcycle registrations per
10,000 persons

235.934

136.762

2.474

1,160.33

Annual precipitation
(in inches)

36.719

15.037

5.37

80.58

Heating degree days
(in 1,000s)

5.238

2.026

0.414

10.758

5.788

1.952

2.3

17.4

1.32

0.57

0.57

3.41

0.616

0.487

0

1

0.171

0.377

0

1

Universal helmet law
Total population

State unemployment rate
Gas price ($ per gallon)
Speed limit ≥ 65 mph
Primary seat belt law

68

Table 2.1 (cont-d)
Variable
Number of adjacent states
Length of borders
(in miles)
NHS road crossings
Fraction of adjacent states with
universal helmet law
Fraction of border miles of
adjacent states with universal
helmet law
Fraction of NHS road crossings
from adjacent states with
universal helmet law
Fraction of adjacent states with
stronger relative helmet law
Number of non-resident
motorcycle fatalities
(annual count)
Fraction of motorcycle fatalities
involving non-residents
Note: Sources available in Table 2.6.

Mean

Standard Deviation

Min

Max

4.5

1.595

1

8

905.013

397.670

102.9

1,563.2

21.083

9.272

4

48

0.489

0.316

0

1

0.486

0.335

0

1

0.490

0.342

0

1

0.197

0.280

0

1

6.044

6.036

0

62

0.122

0.094

0

0.667

69

Table 2.2: Estimates of the Effect of Helmet Laws on Motorcycle Fatalities per Capita

Universal helmet law

(1)
-0.3641***
(0.0428)

ln (motorcycle registrations)

(2)
-0.3408***
(0.0450)

(3)
-0.3549***
(0.0456)

(4)
-0.3603***
(0.0446)

(5)
-0.3542***
(0.0456)

0.0965
(0.0667)

Fraction of neighbors w/ universal law

-0.1067
(0.0751)

Fraction of border miles w/ universal law

-0.0434
(0.0640)

Fraction of National Highway System
road crossings into a state w/ universal law

-0.0945
(0.0650)

Fraction of neighbors w/ stronger
relative helmet law
ln (total population)

ln (heating degree days)

ln (precipitation)

(6)
-0.3369***
(0.0548)

0.0625
(0.0806)
-0.3636**
(0.1355)

-0.3243**
(0.1305)

-0.3304**
(0.1354)

-0.3471**
(0.1420)

-0.3167**
(0.1437)

-0.3754***
(0.1357)

-0.0491
(0.0638)

-0.0465
(0.0620)

-0.0468
(0.0638)

-0.0494
(0.0636)

-0.0494
(0.0639)

-0.0505
(0.0644)

-0.1737***
(0.0438)

-0.1725***
(0.0441)

-0.1763***
(0.0442)

-0.1738***
(0.0440)

-0.1744***
(0.0441)

-0.1708***
(0.0435)

70

Table 2.2 (cont’d)
State unemployment rate

-0.0294***
(0.0087)

-0.0263***
(0.0087)

-0.0290***
(0.0087)

-0.0292***
(0.0087)

-0.0294***
(0.0087)

-0.0288***
(0.0087)

ln (gas price)

-0.1895
(0.2742)

-0.2105
(0.2749)

-0.1444
(0.2691)

-0.1742
(0.2698)

-0.1401
(0.2665)

-0.1996
(0.2785)

Speed limit ≥65 mph

-0.0285
(0.0384)

-0.0269
(0.0386)

-0.0240
(0.0378)

-0.0273
(0.0381)

-0.0256
(0.0380)

-0.0279
(0.0379)

Primary seat belt law

0.0205
(0.0406)

0.0155
(0.0393)

0.0160
(0.0398)

0.0187
(0.0402)

0.0183
(0.0400)

0.0245
(0.0402)

Notes:
1) The dependent variable is the natural log of motorcyclist fatalities per 100,000 persons. All estimation samples consist of 48
states from 1976 to 2008. The unit of observation is a state-year. All models include indicators for years and states.
2) Standard errors, in parentheses, are robust to clustering within state over time. ***, **, and * indicate statistical significance at
1, 5, and 10 percent.

71

Table 2.3: Estimates of the Effect of Helmet Laws on the Fraction of Motorcycle Fatalities involving Non-Residents
Model:
Linear
Fractional Probit
Fractional Probit
Estimation method:
Fixed effects
Pooled QMLE
GEE
(1)
(2)
(3)
(4)
(5)
(6)
Universal helmet law
0.0043
0.0062
0.0006
0.0052
0.0041
0.0054
(0.0069)
(0.0086)
(0.0063)
(0.0073)
(0.0061)
(0.0072)
Fraction of neighbors w/ universal law
0.0148
0.0695***
0.0222
(0.0190)
(0.0176)
(0.0146)
Fraction of neighbors w/ stronger relative law
0.0013
-0.0013
-0.0010
(0.0160)
(0.0137)
(0.0137)
ln (total population)
-0.0675*** -0.0632*** -0.0800*** -0.0579*** -0.0654*** -0.0584***
(0.0227)
(0.0213)
(0.0183)
(0.0165)
(0.0169)
(0.0161)
ln (heating degree days)
0.0079
0.0082
0.0060
0.0066
0.0062
0.0065
(0.0164)
(0.0163)
(0.0131)
(0.0131)
(0.0131)
(0.0131)
ln (precipitation)
0.0270**
0.0267**
0.0254**
0.0241**
0.0246**
0.0242**
(0.0132)
(0.0131)
(0.0111)
(0.0106)
(0.0107)
(0.0105)
State unemployment rate
0.0001
0.0001
-0.0006
-0.0002
-0.0003
-0.0002
(0.0019)
(0.0019)
(0.0017)
(0.0017)
(0.0017)
(0.0017)
ln (gas price)
0.0022
0.0082
-0.0079
0.0181
0.0093
0.0172
(0.0584)
(0.0585)
(0.0517)
(0.0510)
(0.0507)
(0.0509)
Speed limit ≥65 mph
-0.0177**
-0.0171**
-0.0186*** -0.0160*** -0.0167*** -0.0160***
(0.0068)
(0.0068)
(0.0060)
(0.0058)
(0.0058)
(0.0058)
Primary seat belt law
-0.0076
-0.0081
-0.0003
-0.0036
-0.0024
-0.0034
(0.0068)
(0.0067)
(0.0057)
(0.0052)
(0.0053)
(0.0052)
Notes: 1) The dependent variable is the fraction of motorcyclist fatalities involving non-residents (sample mean 0.1222). All
estimation samples consist of 48 states from 1976 to 2008. The unit of observation is a state-year. All models include a full set of
year indicators. The fractional probit estimation includes the time averages of the explanatory variables. Average partial effects are
displayed for the fractional probit estimation.
2) Standard errors, in parentheses, are robust to clustering within state over time. ***, **, and * indicate statistical significance at 1,
5, and 10 percent.

72

Table 2.4: Fixed Effects Poisson Estimates of the Effect of Motorcycle Helmet Laws on
Total and Non-Resident Fatalities
Dependent Variable:
Number of Motorcycle Fatalities
Non-residents Only
Total Fatalities
(1)
(2)
(3)
(4)
Universal helmet law

Fraction of neighbors w/ universal law

-0.2451***
(0.0656)

-0.2620*
(0.1381)

-0.1399
(0.2257)

Fraction of neighbors w/ stronger law

-0.3787***
(0.0358)

-0.3745***
(0.0641)

-0.1104
(0.0989)
-0.0076
(0.1469)

0.0278
(0.0617)

ln (total population)

-0.2056
(0.1942)

-0.2511
(0.1919)

0.7531***
(0.1149)

0.6964***
(0.1065)

ln (heating degree days)

-0.0310
(0.0923)

-0.0387
(0.0926)

-0.0797
(0.0662)

-0.0918
(0.0672)

ln (precipitation)

-0.0946
(0.0682)

-0.0857
(0.0712)

-0.1574***
(0.0334)

-0.1448***
(0.0340)

-0.0381**
(0.0158)

-0.0383**
(0.0155)

-0.0258***
(0.0065)

-0.0251***
(0.0065)

ln (gas price)

0.0316
(0.3728)

-0.0477
(0.3847)

-0.3971**
(0.2005)

-0.4784**
(0.2062)

Speed limit ≥65 mph

-0.0754
(0.0806)

-0.0826
(0.0807)

0.0110
(0.0292)

0.0020
(0.0280)

Primary seat belt law

-0.0436
(0.0733)

-0.0374
(0.0736)

-0.0239
(0.0380)

-0.0195
(0.0381)

State unemployment rate

Notes:
1) All estimation samples consist of 48 states from 1976 to 2008. The unit of observation is a
state-year. All models include a full set of year indicators. The results displayed are coefficient
point estimates and not average partial effects.
2) Standard errors, in parentheses, follow Wooldridge (1999). ***, **, and * indicate statistical
significance at 1, 5, and 10 percent.

73

Table 2.5: Estimates of the Effect of Helmet Laws on Motorcycle Registrations, per Capita
(1)

(3)

-0.2415***
(0.0408)

Universal helmet law

(2)
-0.2517***
(0.0396)

-0.1931***
(0.0570)

Fraction of neighbors w/ universal law

0.1192
(0.0864)

Fraction of neighbors w/ stronger law

0.1114
(0.0767)

ln (total population)

-0.4078**
(0.1966)

-0.4449**
(0.1989)

-0.4288**
(0.2010)

ln (heating degree days)

-0.0274
(0.0310)

-0.0300
(0.0311)

-0.0299
(0.0312)

ln (precipitation)

-0.0127
(0.0743)

-0.0098
(0.0746)

-0.0075
(0.0755)

-0.0325***
(0.0097)

-0.0329***
(0.0097)

-0.0314***
(0.0097)

ln (gas price)

0.2174
(0.2867)

0.1671
(0.2924)

0.1995
(0.2890)

Speed limit ≥65 mph

-0.0172
(0.0639)

-0.0222
(0.0640)

-0.0160
(0.0632)

Primary seat belt law

0.0518
(0.0474)

0.0568
(0.0475)

0.0590
(0.0479)

State unemployment rate

Notes:
1) The dependent variable is the natural log of motorcycle registrations per 10,000 persons.
All estimation samples consist of 48 states from 1976 to 2008. The unit of observation is a
state-year. All models include indicators for years and states.
2) Standard errors, in parentheses, are robust to clustering within state over time. ***, **,
and * indicate statistical significance at 1, 5, and 10 percent.

74

Table 2.6: Data Sources
Variable
Helmet laws,
speed limits, seat
belt laws
Motorcycle fatality
data, state of crash
occurrence, license
state for deceased
riders
Heating degree
days, annual
precipitation
Population
estimates
State
unemployment rate
Motorcycle
registrations
Gas prices

Border crossings
State borders and
border lengths

Source
Insurance Institute for Highway Safety
http://www.iihs.org/laws/default.aspx
National Highway Traffic Safety Administration’s Fatality
Analysis Reporting System (FARS)
http://www.nhtsa.gov/FARS

National Oceanic Atmospheric Administration (NOAA)

U.S. Census Bureau
U.S. Bureau of Labor Statistics
Federal Highway Administration Highway Statistics
U.S. Energy Information Administration State Energy Data
System (SEDS)
http://www.eia.gov/tools/faqs/faq.cfm?id=26&t=10
Federal Highway Administration National Highway System
http://www.fhwa.dot.gov/planning/nhs
Website of Dr. Thomas Holmes (University of Minnesota)
http://www.econ.umn.edu/~holmes/data/BorderData.html

75

APPENDIX 2.2
CHAPTER 2 FIGURES

76

Figure 2.1: Arizona Non-Resident Motorcycle Fatality Trends
30

0.3

California enacts universal
helmet law (January 1992)

25

0.25

20

0.2

15

0.15

10

0.1

5

0.05

0

0

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
# of non-resident motorcycle fatalities
fraction of motorcycle fatalities attributable to non-residents
Notes:
1) Arizona has had a partial coverage helmet law since 1976.
2) Source: Author’s calculation from FARS data.

77

Figure 2.2: Pennsylvania Non-Resident Motorcycle Fatality Trends
30

0.3
Pennsylvania repeals
universal helmet law
(September 2003)

25

0.25

20

0.2

15

0.15

10

0.1

5

0.05

0

0
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

# of non-resident motorcycle fatalities
fraction of motorcycle fatalities attributable to non-residents
Notes:
1) Source: Author’s calculation from FARS data.

78

Figure 2.3: Ohio Non-Resident Motorcycle Fatality Trends
30

Kentucky repeals
universal helmet law
(July 1998)

Pennsylvania repeals
universal helmet law
(September 2003)

0.3

25

0.25

20

0.2

15

0.15

10

0.1

5

0.05

0

0
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
# of non-resident motorcycle fatalities
fraction of motorcycle fatalities attributable to non-residents

Notes:
1) Ohio has had a partial coverage helmet law since 1978.
2) Source: Author’s calculation from FARS data.
79

Figure 2.4: Oklahoma Non-Resident Motorcycle Fatality Trends
30

Arkansas repeals universal helmet law (January 1997)
Texas repeals universal helmet law (September 1997)

0.3

25

0.25

20

0.2

15

0.15

10

0.1

5

0.05

0

0
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
# of non-resident motorcycle fatalities
fraction of motorcycle fatalities attributable to non-residents

Notes:
1) Oklahoma has had a partial coverage helmet law since 1976.
2) Source: Author’s calculation from FARS data.
80

Figure 2.5: Texas Non-Resident Motorcycle Fatality Trends
30

Texas repeals
universal helmet law
(September 1997)

Louisiana repeals
universal helmet law
(August 1999)

Louisiana reinstates
universal helmet law
(August 2004)

0.3

25

0.25

20

0.2

15

0.15

10

0.1

5

0.05

0

0
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
# of non-resident motorcycle fatalities
fraction of motorcycle fatalities attributable to non-residents

Notes:
1) Source: Author’s calculation from FARS data.

81

Figure 2.6: New Mexico Non-Resident Motorcycle Fatality Trends
30

0.3

Texas repeals
universal helmet law
(September 1997)

25

0.25

20

0.2

15

0.15

10

0.1

5

0.05

0

0

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
# of non-resident motorcycle fatalities
fraction of motorcycle fatalities attributable to non-residents
Notes:
1) New Mexico has had a partial coverage helmet law since 1978.
2) Source: Author’s calculation from FARS data.
82

CHAPTER 2 REFERENCES

83

CHAPTER 2 REFERENCES

Baughman, R., M. Conlin, S. Dickert-Conlin, and J. Pepper. 2001. "Slippery when wet: the
effects of local alcohol access laws on highway safety." Journal of Health Economics,
20:1089-1096.
Berkowitz, A. 1981. "The Effect of Motorcycle Helmet Usage on Head Injuries, and the Effect
of Usage Laws on Helmet Wearing Rates." DOT HS 805 851. U.S. Department of
Transportation, Washington, D.C.
Brenzing, B. 2011. "Repeal of helmet law gets new push." The Detroit Free Press, June 8, 2011.
Brueckner, J.K. 2000. “Welfare Reform and the Race to the Bottom: Theory and Evidence.”
Southern Economic Journal, 66(3): 505-525.
Dare, C.E., J.C. Owens, S. Krane. 1978. "Impact of Motorcycle Helmet Usage in Colorado."
DOT HS 803 680. U.S. Department of Transportation, Washington, D.C.
Dee, T. 2009. “Motorcycle Helmets and Traffic Safety.” Journal of Health Economics, 28(2):
398-412.
Deutermann, W. 2004. "Motorcycle Helmet Effectiveness Revisited." DOT HS 809 715. U.S.
Department of Transportation, Washington, D.C.
Dickert-Conlin, S., T.E. Elder, and B.J. Moore. 2011. "Donorcycles: Motorcycle Helmet Laws
and the Supply of Organ Donors." Journal of Law and Economics, forthcoming.
Evans, L. and M. Frick. 1988. "Helmet effectiveness in preventing motorcycle driver and
passenger fatalities." Accident Analysis and Prevention 6: 447-458.
Gilbert, H., N. Chaudhary, M. Solomon, D. Preusser, L. Cosgrove. 2008. “Evaluation of the
Reinstatement of the Universal Motorcycle Helmet Law in Louisiana.” DOT HS 810 956.
U.S. Department of Transportation, Washington, D.C.
Grabowski, D.C. and M.A. Morrisey. 2004. "Gasoline prices and motor vehicle fatalities."
Journal of Policy Analysis and Management, 23(3): 575-593.
Hausman, J.A., B.H. Hall, and Z. Griliches. 1984. "Econometric Models for Count Data with an
Application to the Patents-R&D Relationship." Econometrica, 52: 909-938.
Houston, D.J. and L.E. Richardson Jr. 2008. "Motorcyclist fatality rates and mandatory helmetuse laws." Accident Analysis and Prevention 40: 200-208.

84

Jones, M.M. and R. Bayer. 2007. "Paternalism and its discontents: Motorcycle helmet laws,
libertarian values, and public health." American Journal of Public Health 97(2): 208-217.
Kraus, J.F., C. Peek, D.L. McArthur, A. Williams. 1995. “The Effect of the 1992 California
Motorcycle Helmet Use Law on Motorcycle Crash Fatalities and Injuries.” Journal of the
American Medical Association, 85(1): 1506-1511.
Liu, B.C., R. Ivers, R. Norton, S. Boufous, S. Blows, and S.K. Lo. 2008. "Helmets for
Preventing Injury in Motorcycle Riders." Cochrane Database of Systematic Reviews, Issue
1. Art. No.: CD004333. DOI: 10.1002/14651858.CD004333.pub3.
Lovenheim, M.F. 2008. "How Far to the Border?: The Extent and Impact of Cross-Border
Casual Cigarette Smuggling." National Tax Journal, 61(1): 7-33.
Lovenheim, M.F. and J. Slemrod. 2010. "The fatal toll of driving to drink: The effect of
minimum legal drinking age evasion on traffic fatalities." Journal of Health Economics, 29:
62-77.
Lund, A.K., A.F. Williams, K.N. Womack. 1991. “Motorcycle Helmet Use in Texas.” U.S.
Department of Health and Human Services Public Health Report; 106: 576-578.
Miller, D.L., M.E. Page, A. Huff Stevens, and M. Filipski. 2009. "Why Are Recessions Good for
Your Health?" American Economic Review: Papers & Proceedings, 99(2): 122-127.
Moore, B.J. 2011. “An Examination of the Unintended Consequences in Motorcycle Helmet
Laws and Organ Donation Policy” Ph.D. Dissertation, Michigan State University.
Papke, L.E. and J.M. Wooldridge. 2008. “Panel data methods for fractional response variables
with an application to test pass rates.” Journal of Econometrics, 145: 121-133.
Peltzman, S. 1975. "The economics of automobile safety regulations." Journal of Political
Economy, 83: 677-725.
Preusser, D.F., J.H. Hedlund, R.G. Ulmer. 2000. "Evaluation of Motorcycle Helmet Law Repeal
in Arkansas and Texas." DOT HS 809 131. U.S. Department of Transportation, Washington,
D.C.
Rhoades, J. 2011. “Opposing points of view: By giving motorcyclists choice on helmets,
Michigan would gain millions.” The Detroit Free Press online, June 30, 2011.
http://www.freep.com/article/20110630/OPINION05/106300422/ Opposing-points-view-Bygiving-motorcyclists-choice-helmets-Michigan-would-gain-millions (accessed July 5, 2011).
Ruhm, C.J. 2000. "Are Recessions Good For Your Health?" Quarterly Journal of Economics,
115(2): 617-650.

85

Sass, T.R. and P.R. Zimmerman. 2000. "Motorcycle helmet laws and motorcyclist fatalities."
Journal of Regulatory Economics 18: 195-215.
Struckman-Johnson, C. and V.S. Ellingstad. 1980. "Impact of the Motorcycle Helmet Law
Repeal in South Dakota 1976-79." DOT HS 805 619. U.S. Department of Transportation,
Washington, D.C.
Ulmer, R.G. and V.S. Northrup. 2005. "Evaluation of the Repeal of the All-Rider Motorcycle
Helmet Law in Florida." DOT HS 809 849. U.S. Department of Transportation, Washington,
D.C.
Ulmer, R.G. and D.F. Preusser. 2003. "Evaluation of the Repeal of Motorcycle Helmet Laws in
Kentucky and Louisiana." DOT HS 809 530. U.S. Department of Transportation,
Washington, D.C.
U.S. General Accountability Office GAO. 1991. "Report to Congressional Requesters Highway
Safety: Motorcycle Helmet Laws Save Lives and Reduce Costs to Society." GAO/RCED91-170. Washington, DC.
Wooldridge, J.M. 1999. "Distribution-Free Estimation of Some Nonlinear Panel Data Models."
Journal of Econometrics, 90: 77-97.

86

CHAPTER 3

DONORCYCLES:
MOTORCYCLE HELMET LAWS AND THE SUPPLY OF ORGAN DONORS

With
Stacy Dickert-Conlin and Todd Elder

87

“Motorcycle fatalities are not only our No. 1 source of organs, they are also the highest-quality
source of organs, because donors are usually young, healthy people with no other traumatic
injuries to the body, except to the head… [a mandatory motorcycle helmet law] could put us out
of business – or at least the business of organ transplants.”
- Transplant surgeon quoted in “Brain Dead: Why are There No Mandatory Helmet Laws?”
(Garrett, 2008)

3.1 Introduction
Empirical evidence consistently shows that motorcyclist deaths are disproportionately
concentrated among those riding without a helmet. Based in part on this evidence, helmetless
motorcyclists are perceived as a major source of transplantable organs, particularly within the
medical trauma community.

52

The perceived link between helmet usage and organ donation

even motivated two recently proposed (albeit failed) laws: in 2003, California Assembly Bill
1200 and New Mexico Senate Bill 239 would have made consent for organ donation presumed,
rather than based on explicit written authorization, for all helmetless motorcyclists killed in
traffic accidents.

53

52

For example, Trauma.org, an organization of professionals in trauma and critical care,
published a discussion about helmet laws that included several references to “donorcycles” and
claims by physicians such as “[w]asn't there a study a couple of years ago which showed organ
donations went down by a third when motorcycle helmet laws were strickly (sic) enforced?”
http://www.trauma.org/archive/archives/helmet.html (accessed February 9, 2010).
53
For the specific language of the California and New Mexico bills, see
http://info.sen.ca.gov/pub/03-04/bill/asm/ab_11511200/ab_1200_cfa_20040109_124839_asm_comm.html and
http://www.nmlegis.gov/Sessions/03%20Regular/bills/senate/SB0239.pdf, respectively (accessed
February 10, 2010). Under presumed consent paradigms, motorcyclists can “opt out” of being
potential donors only by signing a form explicitly prohibiting their organs for use in transplants.
88

Despite the perception that helmet usage reduces organ donation rates, no previous
research has investigated whether such a link exists. Estimates of the strength of this relationship
are essential to cost-benefit analyses of government regulation of helmet use. Currently, costbenefit arguments are based on weighing personal freedoms against the negative externalities
associated with helmetless motorcycling, including costs to public health programs resulting
from the relatively high rates of injury and death among those involved in accidents (GAO
1991). This study will quantify an additional, unintended cost of government helmet regulation,
as measured by its effect on the supply of organ donors.
Using state- and year-specific data on organ donations and variation across states and
time in helmet laws, we present evidence that helmet laws reduce organ donation rates.
Statewide helmet mandates are associated with roughly 10 percent reductions in the supply of
organ donors who died in motor vehicle accidents (MVAs). In contrast, helmet laws are
unrelated to the number of organ donors who died due to circumstances other than MVAs, such
as homicide or natural causes. As further support for a causal interpretation of the estimates,
helmet laws affect organ donation rates only among men, who account for more than 90 percent
of annual motorcyclist fatalities, and are also concentrated among those aged 18 to 34. Our
central results suggest that every death prevented by motorcycle helmet laws decreases the
number of organ donors by 0.12. Based on this estimate, along with published estimates of the
number of organs recovered per donor, each death that occurs among helmetless riders saves the
lives of 0.33 persons on the vast organ transplant waiting lists.
In the following section, we review the history of helmet laws and describe the
mechanisms by which helmet laws could influence the supply of organ donors. Section 3
As described by Abadie and Gay (2006), several European nations currently operate under a
presumed consent paradigm.
89

describes the organ donation data from the Organ Procurement and Transplantation Network and
how they relate to publicly available traffic fatality data. Section 4 presents the empirical
specifications and results, and Section 5 concludes.

3.2 Institutional Details of Helmet Laws and Their Link to Organ Donation
Between 1966 and 1995, the federal government twice implemented and retracted acts
that set guidelines requiring all motorcycle riders to wear helmets. Although federal legislation
has never been a strict mandate to states, it has induced substantial swings in state-level
legislation through explicit threats to cut federal highway funding to noncompliant states.

54

No

federal helmet legislation currently exists, but state legislatures continue to debate and modify
their own helmet laws. Currently, 20 states have universal coverage laws, 27 states have partial
coverage laws that typically mandate helmet usage for all riders under age 18, and 3 states do not
require any riders to wear a helmet.

55

Since the most recent federal guidelines for helmet

54

The 1966 Highway Safety Act authorized the Secretary of Transportation to withhold up to 10
percent of federal highway construction funds from any state that did not adopt universal helmet
laws. The Act was amended in 1976, eliminating the helmet law mandate and removing the
authority to withhold funds. Under the 1991 Intermodal Surface Transportation Efficiency Act,
states that enacted both universal helmet laws and seat belt laws would be eligible to receive
federal grant money, while states that did not comply would be subject to a 3 percent reallocation
of federal highway funds towards highway safety programs. The threat of reallocation was
removed with the passage of the National Highway System Designation Act in 1995 (Houston
and Richardson, 1995).
55
For the purposes of this study, partial coverage laws are considered equivalent to having no
helmet law due to enforcement difficulties and the high proportion of the riding population
facing no restrictions. Although the most common partial coverage law requires helmets for
riders under age 18, some states have restrictions for riders under age 15, 19, or 21. A handful of
states currently require helmets for operators with instructional permits, less than one year of
riding experience, or less than $10,000 of personal injury insurance. All helmet statutes specify
maximum punishments for violation; for example, in Georgia, riding without a helmet is
90

mandates were removed in 1995, six states (Arkansas, Texas, Kentucky, Louisiana, Florida, and
Pennsylvania) repealed universal helmet laws, with Louisiana reinstituting their universal law in
2004. Table 3.6 lists the timing of these seven law changes, which play a key role in the
analyses below.
The decline in the prevalence of helmet laws since 1995 is perhaps surprising in light of
strong evidence that these laws increase helmet use and reduce fatalities. Using data from the
2006 and 2007 National Occupant Protection Use Survey (NOPUS), a field study conducted by
the National Highway Traffic Safety Administration to measure the use of motorcycle helmets
and seat belts, we estimate helmet usage rates in states with a universal law to be 97.8 percent,
compared to 54.2 percent in states with partial or no laws. Numerous studies using single-state
data also find that helmet usage decreases from nearly 100 percent to roughly 55 percent
following universal law repeals (see Berkowitz 1981, Dare et al. 1978, Gilbert et al. 2008, Kraus
et al. 1995, Lund et al. 1991, Preusser et al. 2000, Struckman-Johnson and Ellingstad 1980,
Ulmer and Northrup 2005, and Ulmer and Preusser 2003).
Several additional studies measure the effectiveness of helmets in protecting riders in the
event of a crash, with arguably the most convincing approach based on within-vehicle variation
in survival and helmet use among operator-passenger pairs. Using this approach, Dee (2009)
finds that helmets reduce fatality risk by 34 percent. A related literature considers the effects of
helmet laws on state-level fatality rates. Estimates based on within-state variation in helmet laws
over time suggest that universal helmet laws reduce per capita fatalities by more than 20 percent

punishable by a fine of up to $1,000 and one year in jail, but the typical punishment for a first
offense is a fine of $90. See http://www.iihs.org/laws/HelmetUseCurrent.aspx for more
information on current state helmet laws.
91

relative to partial laws and by 27 to 29 percent compared to having no laws at all (Dee 2009,
Houston and Richardson 2008, Sass and Zimmerman 2000).

56

The Logistics of Organ Donation
Although brain death is rare, occurring in less than 1 percent of all deaths in the U.S.,
almost all non-living organ donors are brain dead at the time of organ recovery. The crucial
distinction between brain death, which involves the irreversible cessation of all brain function,
and the more common classification of death (known as “cardiac death”) lies in the fact that the
heart continues to beat after brain death occurs. Current medical technology allows for
essentially indefinite respiration via a ventilator following brain death, so the internal organs
receive oxygenated blood and remain viable for transplantation. In contrast, organs deteriorate
rapidly following cardiac deaths and are therefore unsuitable for transplantation except in
extraordinary circumstances.

57

If the brain dead patient is otherwise healthy and provided

informed consent for donation, either directly or through family members, surgeons instigate the
process of organ recovery.

58

56

A number of additional studies focus on a single state before and after a helmet law change.
For examples, see Auman et al. (2002), Bledsoe et al. (2002), Bledsoe and Li (2005), Eberhardt
et al. (2008), Kraus et al. (1994), Mayrose (2008), Mertz and Weiss (2008), and Muller (2004,
2007).
57
An example of such a circumstance is the growing but controversial practice of “non-heart
beating donation”, in which patients with non-survivable brain injuries (who are not brain dead
because they retain some minimal brain stem function) become donors. Donation in such cases
entails removing the patient from a ventilator, typically in the operating room. Once the patient's
heart stops beating, the physician declares the patient dead and organs are removed. See
http://www.organtransplants.org/understanding/death/ for more details.
58
“Otherwise healthy” individuals are defined as those younger than 70 and lacking
contraindications to organ donation defined by the International Classification of Diseases.
92

The perception that helmetless motorcyclists are prime organ donor candidates is based
on the notion that they can be killed in low-speed, relatively minor collisions which cause brain
death but leave the rest of the body in pristine condition. In contrast, a deceased helmeted cyclist
or automobile occupant is likely to have been in a violent collision that caused widespread
internal damage and cardiac death, both of which are incompatible with organ donation.
Although no previous research has documented the effects of helmet laws on brain deaths in
particular, the existing evidence linking helmet use to motorcyclist death rates implies that the
incidence of brain death is likely to be lower when helmet laws are in place.
The final link between helmet laws and the supply of organs for transplantation involves
the transition from brain death to consenting organ donor. Although federal law has always
maintained that health care professionals only need the potential donor’s consent to recover
organs, a practice known as first-person consent, health care professionals typically also seek
permission from the potential donor’s next-of-kin. Due primarily to low consent rates, donation
rates among all potential donors range from 51 to 60 percent according to Howard et al. (2007).
Despite this apparent inefficiency in organ procurement, policies such as helmet laws that affect
the incidence of brain death will likely affect the supply of organs for transplantation. In the
remainder of this paper, we measure the magnitude of this effect.

3.3 Data
The U.S. first established a unified transplantation network, the Organ Procurement
Transplantation Network (OPTN), under directive from the 1984 National Organ Transplant Act.
Table 2 of Guadagnoli et al. (2003) lists these contraindications, which include cancer, HIV,
hepatitis, and a number of other blood-borne infections.
93

This act provided the authority to divide the United States into mutually exclusive donation
service areas (DSAs), each of which was assigned to an Organ Procurement Organization
(OPO). Each OPO is a local monopoly within its DSA, exclusively responsible for coordinating
and facilitating donation services. One of OPTN’s key initiatives involves the collection and
management of data from every donation and transplant occurring in the United States.

59

The organ donation data used in this paper are DSA-level counts of deceased donors,
available on the OPTN website (http://optn.transplant.hrsa.gov/). There are 57 operational DSAs
in the U.S. that provide data to the OPTN. In most cases, the aggregation from DSA to the state
level is straightforward, but some DSAs cover part or all of multiple states. In these cases, we
assign the donation statistics for the entire DSA to the state where its OPO is headquartered.
Because our identification strategy is based on state variation in helmet laws, this assignment
may present problems if deaths from a county in a given state are designated to another state’s
donor counts. However, as we describe below, our results are insensitive to restricting our
analysis to DSAs that include only one state. Aggregating to the state level results in 38
observations per year because 13 states do not contain an OPO headquarters (D.C. contains its
own OPO); a complete listing of OPO locations can be found at
http://unos.org/members/directory.asp.

59

The National Organ and Transplant Act also outlawed the purchase and sale of organs and
established the OPTN with the responsibilities of creating a system for matching organs to
individuals. It also established OPOs as clearinghouses for acquiring useable organs,
maintaining organ quality standards and allocating donated organs equitably (see
http://optn.transplant.hrsa.gov/SharedContentDocuments/NOTA_as_amended_-_Jan_2008.pdf,
accessed 5/19/09).
94

The OPTN website provides donor counts starting in 1988, along with data on the
demographics, medical history, and certification of consent for each donor.

60

Beginning April

st

1 , 1994, OPTN also reports the donor’s circumstance of death, with separate categories for
MVAs, other accidents such as falls, child abuse, suicide, homicide, and, as of 1999, natural
causes.
Table 3.1 presents the number of organ donors per million persons by circumstance of
death and gender of donor from 1994 to 2007, with all circumstances other than MVAs
aggregated to a single “All Others” category. Because the reporting of circumstance of death
began in April of 1994, all donors from the first three months of 1994 are listed in the “All
Others” category, making recorded MVA donation rates artificially low in 1994. From 1995 to
2007, donations resulting from circumstances other than MVAs rose steadily for both males and
females, increasing by roughly 43 percent from 15.29 to 21.85 per million persons. In contrast,
donations due to MVAs were essentially flat from 1995 to 2007, declining from 5.43 to 5.36 per
million.
Because the MVA category includes donors killed in all motor vehicle accidents, rather
than motorcycle accidents in particular, we also use fatality data from the National Highway
Traffic Safety Administration’s Fatality Analysis Reporting System (FARS) to refine our
analysis.

61

Table 3.2 summarizes motor vehicle fatalities by vehicle type and gender from 1994

to 2007. Overall fatality rates declined 13 percent, from 156.3 per million persons in 1994 to
60

For a sample Deceased Donor Registration Worksheet, see
http://www.unos.org/docs/Deceased_Donor_Registration.pdf.
61

The full FARS file contains detailed information on every person involved in an accident on
public roads that leads to at least one death within 30 days. Our analysis uses fatalities of vehicle
operators and passengers aggregated to the state level.
95

135.3 in 2007, and declined roughly 20 percent (from 147.4 to 118.3 per million) among those in
vehicles other than motorcycles. In contrast, motorcycle fatality rates increased by 90 percent
over the same period, from 8.9 to 17.0 per million persons. Taken together, Tables 3.1 and 3.2
show that the per capita supply of donors killed in MVAs remained roughly constant between
1994 and 2007 in spite of steady decreases in traffic fatalities. Based solely on the yearly
averages, it is unclear whether MVA donation rates were stable because of the dramatic increase
in motorcycle fatalities or because of other factors, such as better technological methods of organ
recovery or higher rates of consent.
Tables 3.1 and 3.2 also show substantial differences across gender in donation and death
rates. In every year, men account for roughly 90 percent of all motorcycle fatalities but only
two-thirds of deaths in other types of vehicles. Because motorcycle fatalities dramatically
increased as a share of all MVA deaths between 1994 and 2007, the share of men in MVA deaths
also increased over time. Returning to Table 3.1, MVA donations among men rose slightly from
1995 to 2007, from 3.55 to 3.74 per million persons, while female MVA donation rates fell from
1.89 to 1.71. These trends are consistent with increased motorcycle deaths being responsible for
holding MVA organ donations roughly constant while non-motorcycle fatalities decreased by 20
percent. Moreover, the dramatic differences across gender contribute to the identification
strategy we pursue below.

3.4 Do Motorcycle Helmet Laws Reduce Organ Donation Rates?
We next turn to assessing whether helmet laws affect the supply of organ donors. As
described above, 38 states headquarter OPOs that report annual data on deceased organ donors.

96

Our primary empirical strategy involves estimating state- and year-specific organ donation rates
as a function of whether the state had a universal mandatory helmet law in place in that year. We
begin by estimating the following model:
(1)

Donorsst = αs + δt + γ(law)st + Χstβ + εst ,

where Donorsst is a measure of the number of deceased organ donors, s indexes the state in
which the OPO is located, t indexes the year, and lawst is an indicator for whether state s had a
universal mandatory helmet law for at least six months in year t. All specifications include a full
set of state and year indicators (αs and δt, respectively), and we indicate below when we also
control for time-varying state-level variables Xst. The vector Xst includes the state’s population,
its square, and separate measures of the fractions of the population aged 18-34, 35-49, 50-64, and
65 and above; state maximum speed limits; separate indicators for whether the state had primaryenforcement and secondary-enforcement mandatory seat belt laws; climate variables correlated
with motorcycle ridership (heating degree days and annual precipitation); and indicators for
whether the state had an organ donor registry, whether online registration was available, and
whether an OPO in the state enforced a first-person consent paradigm.

62

We weight each

observation by the state’s population in that year using U.S. Census Bureau estimates. Estimates
62

Information on helmet and seat belt laws comes from the Insurance Institute for Highway
Safety’s website: http://www.iihs.org/laws/default.aspx. Primary-enforcement seat belt laws
specify that police officers may stop vehicles solely on the suspicion that an occupant is not
wearing a seat belt. Under secondary enforcement laws, officers may cite vehicle occupants for
not wearing seat belts but cannot stop a vehicle solely for this purpose. Data on first-person
consent practices, the existence of state donor registries and the ability to sign up for those
registries online came from interviews with OPO employees. The National Oceanic and
Atmospheric Administration supplied the climate data. Finally, estimates of annual state
population levels and the corresponding within-state age distributions come from U.S. Census
Bureau estimates.
97

of γ based on (1) capture the association between within-state variation over time in mandated
helmet laws and within-state variation in organ donation rates.
Table 3.3 presents estimates of γ from specification (1) separately for donations due to
MVAs and donations due to all other circumstances. The top row presents our central estimates,
in which Donorsst includes only MVA donors. In column (1), we measure Donorsst as the
number of donors per million state residents, so the estimate implies that a universal helmet law
decreases the supply of organ donors by 0.491 per million state residents, with a standard error of
0.157 (all standard errors are robust to within-state clustering over time). As shown in column
(2), inclusion of the time-varying covariates Xst does not markedly change the point estimate. In
both cases, mandatory helmet laws are associated with roughly 10 percent reductions in organ
donor rates relative to the sample average of 5.148 per million persons. In order to assess how
sensitive these results are to the definition of the dependent variable, in columns (3) and (4) we
define Donorsst to be the absolute level of organ donors and its natural logarithm, respectively.
These specifications lead to similar inferences as those in columns (1) and (2); specifically, we
estimate that helmet laws reduce MVA organ donors by roughly 12.7 percent (= -7.696 / 60.595)
when measured in levels and by roughly 9.7 percent when measured in natural logs. Using
donors per capita as the dependent variable, as in columns (1) and (2), allows us to interpret the
estimated coefficients as the effect of helmet laws on donors among the population at risk to
donate, and provides a straightforward interpretation below when we consider the effect of
helmet laws on the excess demand for organs. We therefore focus on specifications using per
capita dependent variables from this point forward.

98

The principal threat to the internal validity of estimates based on specifications such as
(1) stems from differential unmeasured time trends across states in organ donation rates that may
be correlated with the presence of helmet laws. These time trends could reflect hospitals
improving their procedures for organ recovery, education campaigns encouraging informed
consent, or endogenous law changes in response to declining donation rates. If differential
trends drive the negative point estimates in the top row of the table, they would likely also induce
a negative association between helmet laws and the number of donors due to circumstances other
than MVAs. As shown in the “All Others” row of the table, this is not the case – in all four
specifications, the point estimates are both positive and statistically indistinguishable from zero
at conventional significance levels. The remaining rows of the table present estimates for the
individual circumstances that comprise the aggregate “All Others” category. None of these
circumstance-specific organ donation rates is significantly related to helmet laws in any
specification.
As Tables 3.1 and 3.2 showed above, motorcycle fatalities and MVA donation rates differ
substantially across gender, so in Panel A of Table 3.4 we present gender-specific estimates of
the effects of helmet laws. Columns (1), (3), and (5) in the first row show the effect of helmet
laws on motorcycle fatalities, given by estimates of γ based on a specification similar to that in
(1):
(2)

Deathsst = αs + δt + γ(law)st + Χstβ + εst ,

where Deathsst denotes the number of annual motorcycle fatalities per million state residents.
Population-weighted average death rates are shown in brackets. Column (1) shows that universal
helmet laws decrease deaths among male motorcyclists by 3.962 per million persons, a 39

99

percent decline relative to the baseline male death rate of 10.218. Column (3) shows that the
absolute effect is much smaller among women, 0.395 per million persons. This reduction is
roughly 38 percent of the baseline death rate of 1.039, similar to the relative effect among men.
Column (5) presents estimates of γ based on pooled-gender death rates. The 39 percent reduction
(= 4.369 / 11.258) in fatalities is broadly consistent with the findings of Dee (2009), for example,
whose estimates range from 27 to 34 percent in specifications using the logarithm of fatalities as
a dependent variable. In sum, helmet laws decrease motorcyclist deaths among both genders, but
the absolute decrease is much larger among men because men account for over 90 percent of all
motorcyclist deaths.

63

The gender-specific fatality estimates in Table 3.4 suggest an intuitive test for whether
the association between helmet laws and organ donors is causal – because helmet laws can
influence organ donations only through their effect on motorcycle fatalities, the absolute effects
of helmet laws on donation rates should be substantially larger among men than among women.
We therefore present gender-specific estimates of the effect of helmet laws on MVA donation
rates in the even-numbered columns of Table 3.4. The results are striking. Helmet laws have
large, statistically significant effects on the number of male MVA donors, with the estimate of
−0.558 (0.109) being over 16 percent of the baseline male MVA donor rate. In contrast, the
63

The similarity across gender in relative effect sizes may result from a similarity across gender
in the effects of helmet laws on helmet usage. Mayrose (2008) finds that in a sample of fatally
injured motorcycle riders and passengers, the effects of statewide helmet mandates on helmet use
are roughly constant across gender, with 83.8 percent of males and 85.8 percent of females
wearing a helmet in states with a universal helmet law, compared to only 36.4 percent and 34.9
percent in states with a partial law. NOPUS observational data, which are not limited to those
who were involved in a fatal accident, do not differentiate by gender, but these data do allow for
estimates of helmet usage separately for drivers and passengers. In our analyses of these data,
available upon request, the relationship between helmet laws and helmet usage does not
significantly differ between drivers and passengers, suggesting that it also does not differ across
gender.
100

estimated effect among women is positive, although the point estimate of 0.015 (0.061) is both
practically and statistically insignificant. Although helmet laws reduce motorcycle fatalities
among women, the reduction is sufficiently small relative to the number of female MVA deaths
that it does not produce a measurable effect on the supply of female organ donors; as Table 3.2
showed, only 4 percent of all female MVA fatalities in 2007 involved motorcycles, compared to
16 percent among men. We view this pattern as compelling evidence of a causal effect of helmet
laws on organ donations, because the most likely sources of unmeasured confounding trends
would affect organ donation rates among both men and women.
As another check on the plausibility of the results shown above, the remaining rows in
Panel A present estimates of equations (1) and (2) separately by the age categories included in
the OPTN data.

64

We again expect the absolute effects of helmet laws on donation rates to be

largest among the groups most likely to be involved in motorcycle fatalities, and the results
strongly confirm this expectation. Roughly 47 percent (= 4.759 / 10.218) of motorcycle fatalities
among men occur in the 18 to 34 age range, and helmet laws reduce fatalities in this age group
by 2.018, or 44 percent of the baseline rate. Men aged 18 to 34 also experience by far the largest
decline in MVA donors in response to helmet laws: 31 percent of the baseline donation rate
(= -0.526 / 1.716). Among men, the other age groups show much smaller relationships between
helmet laws and fatalities, as well as much smaller effects of helmet laws on MVA organ donors,
providing further support for a causal interpretation of the relationship between helmet laws and
MVA donors among men aged 18 to 34.

64

65

We also estimated these age- and gender-specific

We thank an anonymous referee for suggesting these specifications.

65

We assessed the sensitivity of these central results to four functional form and measurement
issues. First, because states enacted or repealed helmet laws in the middle of calendar years (see
101

models using total organ donors as a dependent variable, rather than MVA donors. The results
of these specifications, available upon request, are broadly similar to those shown in Table 3.4,
albeit less precisely estimated – total donations among men and those aged 18-34 increase in
response to helmet law repeals.
As a final set of specification tests, we next consider the effects of helmet laws on the
supply of donors who were killed in accidents that did not involve motor vehicles. As noted
above, the principal threat to the validity of our central estimates lies in the possibility that statespecific unobserved trends in organ donation rates (or in organ extraction technology) may be
correlated with the presence of helmet laws. It is plausible that these changes similarly affect all
accident victims, regardless of whether the accidents involved motor vehicles. As a result, we
view donors who died in non-motor vehicle accidents (NMVAs) as a more natural “control
group” than the full set of donors who were not killed in MVAs (which includes those who died
via homicides, suicides, and natural causes, in addition to those who died in NMVAs).
Table 3.5 presents three sets of age- and gender-specific estimates that comprise this final
specification test. The first set, in columns (1), (4), and (7), simply reproduces the difference-indifference estimates for MVA donors shown in Table 3.4. The second set, in columns (2), (5),
and (8), shows the analogous estimates based on specifications in which the dependent variable
Table 3.6), we estimated all models using a measure of the fraction of the year in which a state’s
helmet law was in place as the key regressor. Second, we treated the dependent variable as a
count variable, estimating all models by Poisson quasi-maximum likelihood, and alternatively
measured it as the log of per capita death and donor rates. Third, we re-estimate equations (1)
and (2) six times, each time dropping one of the six states that repealed its helmet law, in order to
test the possibility that only one or two states are driving the results. Finally, we excluded from
the analysis all DSAs that cover multiple states (such as the Kansas City-based DSA, which
covers parts of both Kansas and Missouri) because the state in which a death occurred is
ambiguous in these cases. None of these alternative specifications yielded substantively
different results from the central ones reported in the text. All alternative results are available
upon request.
102

is the per capita supply of NMVA donors. Finally, the third set of estimates, in columns (3), (6),
and (9), is the difference between the first two, representing a triple-difference that identifies the
differential effect of helmet mandates on MVA donors relative to NMVA donors. On the whole,
the results provide additional support that the central estimates in Table 3.4 capture real effects
of helmet laws. NMVAs are not strongly associated with helmet laws in either the full sample or
among those most likely to die in motorcycle accidents, such as men and those aged 18-34.
Among men aged 18-34, the NMVA estimate is -0.018, compared to -0.526 for the MVA
estimate, so the triple-difference estimate is -0.508 (= -0.526 + 0.018). The triple-difference
estimates are also negative and statistically significant for males as a whole and in the pooledgender sample of those aged 18-34. In all other age and gender categories, the estimates are
sometimes positive, sometimes negative, but always statistically insignificant. Note that the
triple-difference estimate in the overall pooled-gender sample (shown in the top row of column
(9)) is not statistically significant, partly due to a smaller point estimate than that in column (7)
and partly due to decreased precision because the third difference introduces an additional source
of noise to these estimates. Curiously, helmet laws are significantly negatively associated with
female NMVA donors aged 35-49, but this finding appears to be an anomaly given that none of
the other nine gender- and age-specific estimates are distinguishable from zero.
The estimates in Table 3.5 also shed light on whether motorcycle helmet laws reduce
deaths and organ donations overall or merely shift deaths and donations between circumstances
in a form of “crowd out”. In particular, risk-lovers who die in motorcycle accidents may have
instead pursued other risky activities had their state imposed a universal helmet law.

66

Under

this hypothesis, helmet laws would increase the supply of donors who died in NMVAs, leading
66

We thank Anup Malani for suggesting this possibility to us.
103

to ambiguity in the overall effect of laws on the supply of donors who die in all accidents. As
Table 3.5 shows, there is no evidence to support such a crowding out hypothesis: none of the
NMVA estimates in the table is positive and significant. Among males aged 18-34, who are at
the highest risk of all types of accidents and thus most likely to exhibit crowding out, the
estimated effect of helmet laws on NMVAs is actually negative and insignificantly different
from zero at -0.018 (0.031).

The Dynamic Effects of Helmet Laws
In order to highlight dynamic patterns that the point estimates in Tables 3.3-3.5 do not
show, Figures 3.1-3.3 present graphical evidence of the effects of helmet laws. Figure 3.1 shows
yearly motorcycle fatality rates among men in two groups: those in the six states that repealed
their universal helmet laws from 1994 to 2007, and those in the 32 remaining states and D.C.
For the states in the former group, the X-axis measures the year relative to the state’s law change,
with zero denoting the year of the repeal, 1 denoting the following year, and so on. For each
state without a law change, “year zero” was randomly generated to equal either 1997, 1998,
1999, 2000, or 2003 with equal probability, because these years corresponded to actual law
changes in the other group of states. As the figure shows, death rates in the two groups are
roughly equal in the three years prior to the repeals, but starting in year zero deaths increase
more rapidly in the treatment states than in the control states. By year 4, the death rates in the
treatment states are 14.98 per million persons, compared to 11.88 in the control states.
Figure 3.2 shows MVA organ donation rates among men aged 18 to 34, separately for the
treatment and control states. The divergence between the two groups of states is consistent with
the estimate of γ of -0.526 shown in Table 3.4 – the difference between the series increases from
104

roughly 0.15 in years minus 1 and zero to nearly 0.7 in year 4. Note that the horizontal axis
begins at minus 2 in this figure because the earliest law repeal occurred in 1997, and the OPTN
data do not allow for a clean distinction between MVA and non-MVA donors until 1995.
Perhaps surprisingly, Figures 3.1 and 3.2 imply that helmet law repeals induce gradual
increases in both motorcyclist death rates and organ donation rates, rather than discrete changes
in the year of the repeal. Figure 3.3 illustrates one reason why: per capita motorcycle
registrations increase gradually following helmet law repeals. Registration rates increased by 42
percent over the four years following repeals, from 11.7 to 16.6 per thousand state residents. In
the control states, registrations increased by 15 percent, reflecting increased ridership in the U.S.
in the late 1990s and early 2000s. This pattern implies that at least some of the association
between helmet laws and deaths results from an effect of helmet laws on motorcycle ridership.

67

Although speculative, it is also possible that those induced to ride by the repeals are relatively
risk-loving or accident-prone compared to those who rode while the laws were in place. In order
to assess whether these visual patterns are representative of the trends in each of the six treatment
states, Figures 3.6-3.8 show analogous graphs separately for each state. The individual states
exhibit patterns that are similar (albeit noisier) to those found in the aggregated treatment group.
The lack of an immediate effect of helmet law repeals on motorcyclist deaths and organ
donations may also reflect a gradual effect of repeals on helmet usage. Ulmer and Preusser
67

In agreement with the evidence in Figure 3.1, models analogous to specifications (1) and (2)
with per capita registrations as the dependent variable point to large effects of helmet laws on
registration rates. The coefficient on lawst is -3.317 (0.675), implying that helmet laws decrease
registrations by roughly 14 percent of the baseline registration rate of 23.374 per 1000 persons.
Similarly, the estimated effects of helmet laws on death rates decrease slightly when registrations
per capita is included as a regressor, from -4.369 to -3.887. All data on registrations come from
the U.S. Department of Transportation; see http://www.fhwa.dot.gov/policy/ohpi/hss/index.cfm
for data from recent years.
105

(2003) show that observed helmet usage declined from 96 percent to 76 percent immediately
following the elimination of Kentucky’s universal law in 1998, but that usage continued to
gradually decline over the next three years, falling to 56 percent in 2001. In contrast,
observational studies show that the introduction of helmet laws immediately increases helmet
usage from roughly 55 percent to nearly 100 percent (Kraus et al., 1995).
The asymmetric effects of helmet law repeals and introductions on helmet usage imply
corresponding asymmetries in the effects on death rates and organ donations. To illustrate the
possibility that behavior responds more quickly to the introduction of helmet laws than to their
repeal, Figures 3.4 and 3.5 show motorcyclist registration and death rates for Florida, Texas, and
California, the three most populous states with helmet law changes after 1990.

68

Figure 3.4

shows registration rates in the three states, and Figure 3.5 shows death rates. Each graph
includes two series, one labeled “State with law change” based on data from the given state, and
the other labeled “States with no law change” based on data from the 40 states which have
neither repealed nor enacted helmet laws since 1990. Each series is normed to equal 1 in the first
full year of a new regime, represented by the vertical lines in each graph; for example, Florida
repealed their helmet law in July of 2000, so each series equals 1 in 2001 in the graphs labeled
“Florida”. Registrations and death rates in Florida and Texas were similar to those in the control
states before the helmet law repeals but increased steadily relative to the control states following
the repeals. California, which introduced a universal helmet law in 1992, shows a much different
pattern. Death rates dropped by 38 percent between 1991 and 1992 while registration rates
changed only negligibly, suggesting that the sudden decline in deaths reflected increased helmet
68

California is included as a “control” rather than a “treatment” state in our analyses of organ
donations because its law change occurred in 1992, before the donation data included
circumstance of death.
106

usage rather than a registration effect. Death rates declined an additional 40 percent from 1992
to 1998, mirroring a 40 percent decline in registered motorcyclists over the same period.
The patterns in Figures 3.4 and 3.5 imply that the results shown above, which are based
on an average of seven years of data following law changes, understate the long-run effects of
repeals that gradually influence both registration rates and helmet usage. Moreover, because the
timeframe of OPTN data availability covers six repeals and only one introduction, the estimates
may understate both the short- and long-run impacts of helmet law introductions.
To further investigate the dynamic impacts of helmet law repeals, we estimated variants
of specifications (1) and (2) that include a linear trend in the number of years since a state
repealed its law, measured by the variable “years since repeal”:
(3)

Yst = αs + δt + λ(law)st + φ(years since repeal)st + Χstβ + εst ,

where Yst refers to either deaths or donors per capita.

69

For death rates, the pooled-gender

estimates of λ and φ are -2.05 (with a standard error of 0.60) and 0.77 (0.15), respectively,
implying an effect of 4.75 (0.62) at 3.5 years after a repeal. This number is statistically
indistinguishable from the pooled-gender estimate of 4.369 in Table 3.4, which captures the
effect at roughly 3.5 years following repeals because it is based on an average of seven years of
post-repeal data. Similarly, in models of MVA organ donation rates, the pooled-gender
estimates of λ and φ are -0.27 (0.23) and 0.10 (0.05), implying an effect of 0.63 (0.18) at 3.5
years following a repeal. The estimated effect rises to 0.77 (0.21) after 5 years and to 0.98 (0.27)
after 7 years.

69

We did not attempt to estimate dynamic effects of the introduction of Louisiana’s law in 2004,
primarily because we are hesitant to make inferences based on a single law change.
107

In order to consider whether the treatment and control states had differential trends in
fatalities and organ donations that were not due to variation in helmet laws, we also estimate
variants of equation (3) that include state-specific linear time trends. The inclusion of these
additional variables does not substantially change the point estimates. For death rates, the
pooled-gender estimate of λ declines to -1.66 (0.61) and the estimate of φ increases to 0.85
(0.21), implying an effect of 4.64 (0.72) at 3.5 years after a repeal, similar to (although slightly
noisier) the effect of 4.75 implied by the specification that does not include state-specific trends.
Similarly, the implied effect on MVA donation rates declines from 0.63 to 0.58 (0.24) when
state-specific linear trends are included in (3). Although for brevity’s sake we do not report the
full set of results based on equation (3), these estimates are available upon request.

Instrumental Variables Estimates of the Effect of Motorcyclist Deaths on Organ Donation Rates
The estimates of γ from models (1) and (2) capture the reduced-form relationships
between helmet laws and organ donors and between helmet laws and motorcyclist deaths,
respectively. As a result, the ratio of the two estimates of γ represents a Wald IV estimate of the
effect of motorcyclist death rates on organ donation rates. A causal interpretation of this ratio
essentially requires that helmet laws only influence organ donation rates through their effect on
motorcyclist deaths. Although we cannot formally test this assumption, the informal evidence
presented above suggests that it holds. Specifically, because helmet laws appear unrelated to
non-MVA donation rates and to female MVA donation rates, we interpret the estimated ratio as
measuring a causal effect of motorcyclist deaths on organ donation rates.
The Wald IV estimates, shown in Panel B of Table 3.4, are 0.141 for men (= 0.558 /
3.962), -0.037 for women (= 0.015 / -0.395), and 0.124 for the pooled-gender sample (= 0.543 /
108

4.369). These estimates capture the effect of the excess motorcyclist deaths resulting from state
repeals of helmet laws on the supply of organ donors.

70

Taken literally, the pooled-gender

estimate implies that every motorcyclist death due to the lack of a universal helmet law produces
0.124 additional organ donors. For men aged 18 to 34, this number increases to 0.261 (= 0.526 /
2.018).
Based on the estimates of the sensitivity of organ donation rates to motorcyclist death
rates, we can calculate the size of the positive externality resulting from each helmetless
motorcyclist’s death, as measured by possible lives saved. According to OPTN data, 2.7 organs
are successfully transplanted per deceased donor, on average. Under the upper-bound
assumption that each of these organs saves one life, the pooled-gender estimate of 0.124 implies
that every death of a helmetless motorcyclist saves the lives of 0.33 (= 0.124 × 2.7) individuals
on organ transplant waiting lists.

71

Because this upper-bound estimate is far less than one,

helmetless riding is clearly an inefficient means of preserving life in the absence of a basis for
making normative judgments about the value of one life relative to another. Nonetheless, this
calculation highlights the possibility that helmet laws impact death rates even among those who
do not ride motorcycles.

70

For instrumental variables estimates to be interpreted as local average treatment effects,
Imbens and Angrist (1994) show that a potential instrument must also satisfy the property of
monotonicity. Although monotonicity is untestable in the context of a binary instrument such as
lawst, it is likely to hold in this setting – it implies that the presence of a helmet law does not
increase motorcyclist death rates in any state.
71
According to the 2007 OPTN annual report, one-year survival rates are 94 percent for kidney
recipients, 87 percent for liver recipients, and 88 percent for heart recipients, accounting for the
three most transplanted organs (Table 1.13). The same report shows that multiple organ
transplants are very rare, accounting for fewer than 2 percent of all transplants (Table 1.08).
These numbers, and many others in this section, are taken from OPTN’s rich national data
available for public use on OPTN’s website: http://optn.transplant.hrsa.gov/latestData/step2.asp.
109

As a further gauge of the magnitude of the estimates in Table 3.4, recall that in 2007,
135.3 MVA fatalities resulted in 5.36 organ donors per million persons (see Tables 3.1 and 3.2).
Each MVA death therefore produced 0.040 organ donors, and this overall donor rate, D, is a
weighted average of the donor rates among helmetless motorcyclists (Dhm) and all others
involved in MVAs (Do):
(4)

D = P × Dhm + (1-P) × Do,

where P is the proportion of all MVA fatalities that involve helmetless motorcyclists. The
pooled-gender estimate of 0.124 represents an estimate of Dhm. Based on 2007 FARS national
data on traffic fatalities, P equals 0.050, implying that Do equals 0.035 – while 12.4 percent of
helmetless motorcyclists killed in MVAs eventually become organ donors, the analogous rate
among all other MVA fatalities is only 3.5 percent. This discrepancy presumably results from
higher rates of brain death among helmetless motorcyclists than among others killed in MVAs.
Howard et al. (2007) estimate that 51 percent of viable donors actually become donors, so
roughly 24 percent (= 0.124 / 0.51) of deceased motorcyclists are viable donors, i.e., brain dead
but otherwise healthy. Although we are wary of interpreting this number literally because it is
based on the assumption that consent rates among motorcyclists are equal to those in the rest of
the population, to our knowledge it represents the first estimate of the fraction of deceased
helmetless motorcycle riders who are viable organ donors.

110

Helmet Laws and the Excess Demand for Organs
Finally, we consider the magnitude of the effect of helmet laws in the context of a
broader question: how would eliminating all existing universal helmet laws affect annual organ
donations and transplant waiting lists? According to state-level population estimates from the
U.S. Census Bureau, approximately 155 million people lived in states with universal helmet laws
in 2007. Based on the estimates from Table 3.4, which measure the average effect over the
seven years following a law change, the repeal of all helmet laws would increase the annual
number of deceased motorcyclists by roughly 677 (= 4.369 × 155), or 13.2 percent of the 5,128
deaths in 2007. These deaths would produce 84 (= 0.543 × 155) additional organ donors because
12.4 percent of deceased motorcyclists become donors. Based on OPTN’s estimate of 2.7
transplanted organs per donor, 227 additional organs would be transplanted. Alternatively, the
dynamic estimates from model (3) imply that 332 additional organs would be transplanted five
years after a nationwide repeal. These numbers represent 3.1 and 4.6 percent, respectively, of
the 7,287 individuals who died in 2007 while waiting for a transplant.

72

72

The dramatic shortage of organs has generated a large body of research that evaluates
mechanisms for increasing organ donation rates. Abadie and Gay (2006) find that countries
using presumed consent donation standards have 25 to 30 percent higher donation rates than
observationally similar countries using informed consent regimes. In a special issue of the
Journal of Economic Perspectives addressing the excess demand for organs, Becker and Elias
(2007) focus on financial incentives for increasing living donors, Howard (2007) reviews
policies for increasing consent rates and the pool of potential donors, and Roth (2007) discusses
the compensation of organ donors in light of “repugnance” for the market trading of organs. In a
series of papers, Roth, Sönmez, and Ünver (2004, 2005) design a matching mechanism for organ
recipients and donors that has been implemented in New England.
111

3.5 Conclusions
Motorcycle helmet mandates are effective: consistent with a larger literature, our
estimates indicate that state-level motorcyclist fatalities increase by approximately thirty percent
when universal helmet laws are repealed. Despite the effectiveness of helmets in preventing
deaths, helmet usage rates in states without universal laws suggest that nearly half of all
motorcyclists prefer to ride helmetless. Helmet mandates impose costs on these riders, but these
costs may be justified by a reduction in the negative externalities imposed by those injured or
killed in accidents.
This study is the first to assess the anecdotal belief that helmet laws also decrease the
positive externalities of helmetless riding by reducing the supply of organ donors. Our central
estimates show that organ donations due to motor vehicle accidents (MVAs) increase by 10
percent when states repeal helmet laws. Nearly all of this effect is concentrated among men,
who account for over 90 percent of motorcyclist fatalities, and it is also concentrated among
those aged 18 to 34. Helmet laws are unrelated to the number of organ donors who died in
circumstances other than MVAs, providing further support for a causal interpretation of the
association between helmet laws and MVA donors. Under the upper-bound assumption that
each recovered organ saves one life, the estimates imply that every motorcyclist death due to the
lack of a helmet law saves the lives of 0.33 individuals on organ transplant waiting lists.
Quantifying the unintended consequences of helmet laws allows for more informed
policymaking by providing a more complete picture of the associated costs and benefits. Based
on our preferred estimates, 3.1 to 4.6 percent of those who died while awaiting an organ in 2007
would have instead received a transplant if all helmet laws were repealed. These estimates may
understate the long-run impacts of helmet laws on organ donations, but they show that helmet
112

laws profoundly affect the lives of some potential transplant recipients. This unintended
consequence merits consideration in cost-benefit analyses of helmet laws, which currently focus
on weighing the costs to society of helmetless motorcycling against motorcyclists’ freedom of
choice.

113

CHAPTER 3 APPENDICIES

114

APPENDIX 3.1

CHAPTER 3 TABLES

115

Table 3.1: Organ Donors by Year, Gender, and Circumstance, per Million
Persons

Circumstance: Motor
Vehicle Accident
All
Male
Female

Circumstance:
All Others
All
Male
Female

1994

4.05

2.71

1.34

15.87

9.50

6.37

1995

5.43

3.55

1.89

15.29

8.90

6.39

1996

5.16

3.44

1.73

15.52

8.88

6.64

1997

5.25

3.39

1.87

15.45

8.84

6.60

1998

5.19

3.40

1.80

16.36

9.11

7.25

1999

4.94

3.24

1.70

16.41

9.02

7.39

2000

5.24

3.52

1.72

16.43

9.10

7.33

2001

4.92

3.37

1.55

16.84

9.41

7.43

2002

5.20

3.55

1.65

16.71

9.54

7.17

2003

5.03

3.33

1.70

17.63

9.97

7.66

2004

5.30

3.54

1.77

19.43

10.68

8.75

2005

5.24

3.50

1.73

20.69

11.57

9.12

2006

5.64

3.86

1.78

21.68

12.41

9.26

2007

5.36

3.74

1.61

21.85

12.68

9.15

Note:
OPTN began reporting circumstances of death on April 1, 1994. For the first three
months of 1994, all donors are included in the “All Others” category. Source: OPTN.

116

Table 3.2: Motor Vehicle Fatalities by Gender, per Million Persons

Year
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007

All
156.3
159.0
158.5
156.8
153.4
152.8
148.4
146.5
148.9
147.1
145.8
146.8
140.7
135.3

All MVA
Male
Female
105.3
51.1
107.0
52.0
105.7
52.7
103.9
52.9
102.1
51.3
102.8
50.1
101.0
47.4
100.6
45.9
102.1
46.9
100.7
46.4
100.2
45.6
102.4
44.4
98.5
42.2
95.7
39.6

All Motorcycle Fatalities
All
Male
Female
8.9
8.1
0.8
8.5
7.7
0.8
8.1
7.4
0.7
7.9
7.2
0.7
8.5
7.7
0.8
9.1
8.3
0.8
10.2
9.3
1.0
11.1
10.1
1.0
11.3
10.3
1.1
12.7
11.5
1.3
13.7
12.2
1.5
15.4
13.9
1.5
15.9
14.6
1.4
17.0
15.4
1.5

Note:
Authors’ calculations using Fatality Analysis Reporting System (FARS) data.

117

All Other Vehicles
All
Male
Female
147.4
97.2
50.3
150.5
99.3
51.2
150.3
98.3
52.0
148.9
96.7
52.2
144.9
94.4
50.6
143.7
94.5
49.2
138.1
91.7
46.4
135.4
90.5
44.9
137.6
91.8
45.8
134.4
89.2
45.2
132.1
88.0
44.1
131.3
88.4
42.9
124.8
84.0
40.8
118.3
80.2
38.0

Table 3.3: Estimates of the Effect of Helmet Laws on Organ Donations
Organ Donors Measured in:
Per Capita
(1)
(2)

Levels
(3)

Natural
Logs
(4)

Motor Vehicle Accident

-0.491 -0.543
(0.157) (0.137)

-7.696
(1.975)

-0.097
(0.031)

5.148

60.595

All Others

0.843
0.046
6.215
(1.283) (1.005) (10.545)

0.047
(0.043)

17.668

227.051

Accidents Not Involving
Vehicles

-0.039 -0.234
(0.233) (0.215)

-2.090
(2.410)

-0.106
(0.116)

1.931

23.978

Homicide, Suicide, or
Child Abuse

0.421
0.363
(0.267) (0.251)

5.783
(3.873)

0.094
(0.058)

3.390

43.516

Natural Causes

-0.020 -0.791
-9.762
(1.348) (1.295) (11.534)

0.508
(0.634)

4.086

49.795

None of the Above

0.466
0.546
13.944
(2.229) (2.160) (33.168)

0.258
(0.361)

7.922

105.703

Donor Circumstance

Include State-Year
Controls?

No

Yes

Yes

Sample
Mean
(per Capita)

Sample
Mean
(Levels)

Yes

Notes:
1) All estimation samples consist of 38 states from 1994 to 2007. The unit of observation is a
state-year. All models include indicators for years and states.
2) Models in column (2) add controls for the state maximum speed limit, quadratics in the
total population of the state, the age distribution of the state's population, heating degree
days, annual precipitation, and indicators for whether the state has a donor registry, whether
the state allows online donor registration, whether organs can be donated without consent of
family members of the prospective donor, and whether the state had primary enforcement of
seat belt laws
3) Standard errors, in parentheses, are robust to clustering within state over time.
118

Table 3.4: Estimates of the Effect of Helmet Laws on per Capita Motorcycle
Fatalities and Organ Donations, by Gender and Age
Panel A: The Effect of Helmet Laws on Fatalities and MVA Donations
Males

Females

Motorcycle
Fatalities
(1)

MVA
Organ
Donors
(2)

-3.962
(0.518)
[10.218]

Age Categories:
<18
-0.103
(0.045)
[0.237]

Overall

Pooled

Motorcycle
Fatalities
(3)

MVA
Organ
Donors
(4)

Motorcycle
Fatalities
(5)

MVA
Organ
Donors
(6)

-0.558
(0.109)
[3.445]

-0.395
(0.095)
[1.039]

0.015
(0.061)
[1.702]

-4.369
(0.568)
[11.258]

-0.543
(0.137)
[5.148]

0.047
(0.055)
[0.884]

-0.015
(0.016)
[0.039]

0.092
(0.072)
[0.601]

-0.118
(0.047)
[0.276]

0.139
(0.101)
[1.486]

18-34

-2.018
(0.305)
[4.759]

-0.526
(0.122)
[1.716]

-0.149
(0.050)
[0.352]

-0.041
(0.049)
[0.679]

-2.166
(0.323)
[5.110]

-0.568
(0.136)
[2.395]

35-49

-0.987
(0.170)
[3.297]

-0.017
(0.098)
[0.592]

-0.236
(0.049)
[0.454]

-0.022
(0.049)
[0.286]

-1.222
(0.181)
[3.751]

-0.039
(0.138)
[0.878]

50-64

-0.709
(0.150)
[1.615]

-0.069
(0.027)
[0.219]

-0.018
(0.037)
[0.172]

-0.008
(0.030)
[0.107]

-0.730
(0.163)
[1.787]

-0.077
(0.052)
[0.326]

65+

-0.146
(0.061)
[0.311]

0.008
(0.010)
[0.034]

0.012
(0.016)
[0.022]

-0.007
(0.006)
[0.028]

-0.134
(0.064)
[0.333]

0.001
(0.011)
[0.063]

119

Table 3.4 (cont’d)
Panel B: The Effect of Motorcyclist Deaths on MVA Donation Rates
Males
Wald IV
Estimates:

Females

Pooled

0.141
(0.045)

-0.037
(0.256)

0.124
(0.039)

Notes:
1) All estimation samples consist of 38 states from 1994 to 2007. The unit of observation is a
state-year. All models include indicators for years and states.
2) All models include the state-year controls described in the notes to Table 3.3
3) Standard errors, in parentheses, are robust to clustering within state over time.
4) Sample means for relevant dependent variables are listed in brackets

120

Table 3.5: Estimates of the Effects of Helmet Laws on per Capita MVA, NMVA, and Differenced Organ Donations,
by Gender and Age

MVAs
(1)

Difference
(3)

MVAs
(4)

Females
NMVAs
Difference
(5)
(6)

-0.130
(0.119)

-0.428
(0.113)

0.015
(0.061)

-0.104
(0.109)

0.119
(0.105)

-0.543
(0.137)

-0.234
(0.215)

-0.309
(0.193)

0.047
(0.055)

0.011
(0.067)

.0036
(0.087)

0.092
(0.072)

0.015
(0.031)

0.077
(0.096)

0.139
(0.101)

0.026
(0.086)

0.112
(0.162)

18-34

-0.526
(0.112)

-0.018
(0.031)

-0.508
(0.096)

-0.041
(0.049)

0.002
(0.025)

-0.044
(0.051)

-0.568
(0.136)

-0.016
(0.045)

-0.552
(0.108)

35-49

-0.017
(0.098)

-0.060
(0.051)

0.043
(0.128)

-0.022
(0.049)

-0.094
(0.036)

0.072
(0.060)

-0.039
(0.138)

-0.154
(0.074)

0.115
(0.165)

50-64

-0.069
(0.027)

-0.019
(0.039)

-0.050
(0.045)

-0.008
(0.030)

-0.020
(0.032)

0.012
(0.057)

-0.077
(0.052)

-0.039
(0.044)

-0.038
(0.070)

65+

0.008
(0.010)

-0.043
(0.033)

0.051
(0.036)

-0.007
(0.006)

-0.008
(0.019)

0.001
(0.017)

0.001
(0.011)

-0.051
(0.051)

0.052
(0.050)

Overall
Age Categories:
<18

Males
NMVAs
(2)

-0.558
(0.109)

MVAs
(7)

Pooled
NMVAs
(8)

Notes:
1) All estimation samples consist of 38 states from 1994 to 2007. The unit of observation is a state-year.
2) All models include the state-year controls described in the notes to Table 3.3
3) Standard errors, in parentheses, are robust to clustering within state over time.
121

Difference
(9)

Table 3.6: Changes in State Helmet Laws, 1994-2007
Year None to Partial
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
CO (7)

Universal to Partial

Partial to Universal

Partial to None
NH (9)

AR (8), TX (9)
KY (7)
LA (8)
FL (7)

PA (9)
LA (8)

Note:
The month a law changed is listed in parentheses, where “1” denotes January, “2”
denotes February, and so on. Source: Insurance Institute for Highway Safety:
http://www.iihs.org/laws/default.aspx

122

APPENDIX 3.2

CHAPTER 3 FIGURES

123

Figure 3.1

8

10

12

14

16

Male Motorcyclist Fatalities per Million Residents

-3

-2

-1

0
1
Year Relative to Repeal

States w ith no law change

2

3

4

State w ith law change

Note: Authors’ calculations from FARS, OPTN and Department of Transportation data, 1994-2007.

124

Figure 3.2

1.6

1.8

2

2.2

2.4

2.6

Male MVA Donors Age 18-34 per Million Residents

-2

-1

0

1
2
Year Relative to Repeal

States w ith no law change

3

4

State w ith law change

Note: Authors’ calculations from FARS, OPTN and Department of Transportation data, 1994-2007.

125

Figure 3.3

10

12

14

16

18

Motorcycle Registrations per Thousand Residents

-3

-2

-1

0
1
Year Relative to Repeal

States w ith no law change

2

3

4

State w ith law change

Note: Authors’ calculations from FARS, OPTN and Department of Transportation data, 1994-2007.
126

Figure 3.4

Per Capita Motorcycle Registrations
in Three Large States

.5

1

1.5

2

2.5

Florida

1990

1995

2000

2005

2000

2005

.5

1

1.5

2

2.5

Texas

1990

1995

.5

1

1.5

2

2.5

California

1990

1995

2000

States with no law change

2005
State with law change

Note: Authors’ calculations from FARS and Department of Transportation data. Data
are normed to 1 in the first full year following a law change.
127

Figure 3.5

Per Capita Motorcyclist Death Rates
in Three Large States

.5

1

1.5

2

2.5

Florida

1990

1995

2000

2005

2000

2005

.5

1

1.5

2

2.5

Texas

1990

1995

.5

1

1.5

2

2.5

California

1990

1995

2000

States with no law change

2005
State with law change

Note: Authors’ calculations from FARS and Department of Transportation data. Data
are normed to 1 in the first full year following a law change.
128

Figure 3.6

Per Capita Male Motorcyclist Fatalities in Six States

-3

-2 -1 0 1 2 3
Year Relative to Repeal

4

20
5

10

15

20
15
10
5

5

10

15

20

25

KY

25

FL

25

AR

-3

-2 -1 0 1 2 3
Year Relative to Repeal

-2 -1 0 1 2 3
Year Relative to Repeal

4

25
5

10

15

20

25
20
10
5
-2 -1 0 1 2 3
Year Relative to Repeal

4

TX

15

20
15
10
5
-3

-3

PA

25

LA

4

-3

-2 -1 0 1 2 3
Year Relative to Repeal

States w ith no law change

4

-3

-2 -1 0 1 2 3
Year Relative to Repeal

State w ith law change

Note: Authors’ calculations from FARS data, 1994-2007.
129

4

Figure 3.7

Per Capita Male MVA Organ Donors Aged 18-34 in Six States

-2

-1
0
1
2
3
Year Relative to Repeal

4

3
1

2

3
2
1

1

2

3

4

KY

4

FL

4

AR

-2

-1
0
1
2
3
Year Relative to Repeal

-1
0
1
2
3
Year Relative to Repeal

4

4
1

2

3

4
3
1
-1
0
1
2
3
Year Relative to Repeal

4

TX

2

3
2
1
-2

-2

PA

4

LA

4

-2

-1
0
1
2
3
Year Relative to Repeal

States w ith no law change

4

-2

-1
0
1
2
3
Year Relative to Repeal

State w ith law change

Note: Authors’ calculations from OPTN data, 1994-2007.
130

4

Figure 3.8

Per Capita Motorcycle Registrations in Six States

-3

-2 -1 0 1 2 3
Year Relative to Repeal

4

5

5

5

10 15 20 25 30

KY

10 15 20 25 30

FL

10 15 20 25 30

AR

-3

-2 -1 0 1 2 3
Year Relative to Repeal

4

4

10 15 20 25 30
5

10 15 20 25 30
-2 -1 0 1 2 3
Year Relative to Repeal

-2 -1 0 1 2 3
Year Relative to Repeal

TX

5

5
-3

-3

PA

10 15 20 25 30

LA

4

-3

-2 -1 0 1 2 3
Year Relative to Repeal

States w ith no law change

4

-3

-2 -1 0 1 2 3
Year Relative to Repeal

State w ith law change

Note: Authors’ calculations from Department of Transportation data, 1994-2007.

131

4

CHAPTER 3 REFERENCES

132

CHAPTER 3 REFERENCES

Abadie, A. and S. Gay. 2006. “The Impact of Presumed Consent Legislation on Cadaveric
Organ Donation: A Cross-Country Study.” Journal of Health Economics, 25: 599-620.
Auman, K.M., J.A. Kufera, M.F. Ballesteros, J.E. Smialek, P.C. Dischinger. 2002. “Autopsy
Study of Motorcyclist Fatalities: The Effect of the 1992 Maryland Motorcycle Helmet Use
Law.” American Journal of Public Health, 92(8): 1352-1355.
Becker, G.S. and J.J. Elias. 2007. “Introducing Incentives in the Market for Live and Cadaveric
Organ Donations.” Journal of Economic Perspectives, 21(3): 3-24.
Berkowitz, A. 1981. The Effect of Motorcycle Helmet Usage on Head Injuries, and the Effect
of Usage Laws on Helmet Wearing Rates. DOT HS 805 851. U.S. Department of
Transportation, Washington, D.C.
Bledsoe, G.H. and G. Li. 2005. “Trends in Arkansas Motorcycle Trauma after Helmet Law
Repeal.” Southern Medical Association, 98(4): 436-40.
Bledsoe, G.H., S.M. Schexnayder, M.J. Carey, W.N. Dobbins, W.D. Gibson, J.W. Hindman, T.
Collins, B.H. Wallace, J.B. Cone, T.J. Ferrer. 2002. “The Negative Impact of the Repeal of
the Arkansas Motorcycle Helmet Law.” The Journal of Trauma Injury, Inflection, and
Critical Care, 53(6): 1078-1087.
Dare, C.E., J.C. Owens, S. Krane. 1978. Impact of Motorcycle Helmet Usage in Colorado. DOT
HS 803 680. U.S. Department of Transportation, Washington, D.C.
Dee, T. 2009. “Motorcycle Helmets and Traffic Safety.” Journal of Health Economics, 28(2):
398-412.
Eberhardt, M., E. Goldman, H. Stankewicz, S. Melanson. 2008. “Helmet Use May Affect
Severity of Injuries Sustained in Motorcycle Accidents.” Annals of Emergency Medicine,
52(4): S109.
Garrett, J. 2008, July 7. “Brain Dead: Why Are There No Mandatory Helmet Laws?” New York
Times. Retrieved January 12, 2011 from http://wheels.blogs.nytimes.com/2008/07/07/braindead-why-are-there-no-mandatory-helmet-laws/.
Gilbert, H., N. Chaudhary, M. Solomon, D. Preusser, L. Cosgrove. 2008. “Evaluation of the
Reinstatement of the Universal Motorcycle Helmet Law in Louisiana.” DOT HS 810 956.
U.S. Department of Transportation, Washington, D.C.

133

Guadagnoli, E., C. Christiansen, C. Beasley. 2003. “Potential Organ-Donor Supply and
Efficiency of Organ Procurement Organizations.” Health Care Financing Review, 24(4):
101-10.
Houston, D.J. and L.E. Richardson. 2008. “Motorcyclist fatality rates and mandatory helmetuse laws.” Accident Analysis & Prevention, 40: 200-208.
Howard, D.H., L.A. Siminoff, V. McBride, and M. Lin. 2007. “Does Quality Improvement
Work? Evaluation of the Organ Donation Breakthrough Collaborative.” Health Services
Research, 42(6): 2160-2173.
Howard, D.H. 2007. “Producing Organ Donors.” Journal of Economic Perspectives, 21(3): 2536.
Imbens, G. and J. Angrist. 1994. “Identification and Estimation of Local Average Treatment
Effects.” Econometrica, 62(2): 467-476.
Kraus, J.F., C. Peek, D.L. McArthur, A. Williams. 1995. “The Effect of the 1992 California
Motorcycle Helmet Use Law on Motorcycle Crash Fatalities and Injuries.” Journal of the
American Medical Association, 85(1): 1506-1511.
Kraus, J.F., C. Peek, A. Williams. 1994. “Compliance with the 1992 California Motorcycle
Helmet Use Law.” American Journal of Public Health, 272(19): 96-99.
Lund, A.K., A.F. Williams, K.N. Womack. 1991. “Motorcycle Helmet Use in Texas.” U.S.
Department of Health and Human Services Public Health Report; 106: 576-578.
Mayrose, J. 2008. “The Effects of a Mandatory Motorcycle Helmet Law on Helmet Use and
Injury Patterns among Motorcyclist Fatalities.” Journal of Safety Research, 39: 429-432.
Mertz, KJ. and H.B. Weiss. 2008. “Changes in Motorcycle-Related Head Injury Deaths,
Hospitalizations, and Hospital Charges Following Repeal of Pennsylvania’s Mandatory
Motorcycle Helmet Law.” American Journal of Public Health, 98(8): 1464-1467.
Muller, A. 2004. “Florida’s Motorcycle Helmet Law Repeal and Fatality Rates.” American
Journal of Public Health, 94(4): 556-558.
Muller, A. 2007. “Weakening Pennsylvania's Motorcycle Helmet Law: Preliminary Estimates.”
poster session at 2007 American Public Health Association annual meetings.
Preusser, D.F., J.H. Hedlund, R.G. Ulmer. 2000. Evaluation of Motorcycle Helmet Law Repeal
in Arkansas and Texas. DOT HS 809 131. U.S. Department of Transportation, Washington,
D.C.
Roth, A.E. 2007. “Repugnance as a Constraint on Markets.” Journal of Economic Perspectives,
21(3): 37-58.
134

Roth, A.E., T. Sönmez, and M.U. Ünver. 2004. “Kidney Exchange.” The Quarterly Journal of
Economics, 119(2): 457-488.
Roth, A.E., T. Sönmez, and M.U. Ünver. 2005. “Pairwise Kidney Exchange.” Journal of
Economic Theory, 125(2): 151-188.
Sass, T.R. and P.R. Zimmerman. 2000. “Motorcycle Helmet Laws and Motorcyclist Fatalities.”
Journal of Regulatory Economics, 18(3): 195-215.
Struckman-Johnson, C. and V.S. Ellingstad. 1980. Impact of the Motorcycle Helmet Law
Repeal in South Dakota 1976-79. DOT HS 805 619. U.S. Department of Transportation,
Washington, D.C.
Ulmer, R.G. and V.S. Northrup. 2005. Evaluation of the Repeal of the All-Rider Motorcycle
Helmet Law in Florida. DOT HS 809 849. U.S. Department of Transportation, Washington,
D.C.
Ulmer, R.G. and D.F. Preusser. 2003. Evaluation of the Repeal of Motorcycle Helmet Laws in
Kentucky and Louisiana. DOT HS 809 530. U.S. Department of Transportation, Washington,
D.C.
U.S. Department of Transportation. Various years. “Highway Statistics.” Federal Highway
Administration.
U.S. General Accountability Office GAO. 1991. Report to Congressional Requesters Highway
Safety: Motorcycle Helmet Laws Save Lives and Reduce Costs to Society. GAO/RCED-91170. Washington, DC.

135