The Effect of State Workers' Compensation Program Changes On The Use of Federal Social Security Disability Insurance

The Effect of State Workers Compensation
Program Changes on the Use of Federal Social

Security Disability Insurance
MELISSA MCINERNEY and KOSALI SIMON*
In addition to traditional forms of private and public medical insurance, two other
large public programs help pay for costs associated with ill health. In 2008,
Workers Compensation (WC) insurance provided $57.6 billion in medical care
and cash benefits to employees who are injured at work or contract a work-related
illness, and Social Security Disability Insurance (DI) provided $106 billion to
individuals who suffer from permanent disabilities and are unable to engage in
substantial gainful activity. During the 1990s, real DI outlays increased nearly 70
percent, whereas real WC cash benefit spending fell by 12 percent. There has
been concern that part of this relationship between two of the nations largest
social insurance programs may be due to individuals substituting toward DI as
state WC policies tightened. We first show that this negative correlation between
the national series does not hold over time within states, the level at which a cau-
sal relationship should operate. We then test for a causal effect of changes in WC
enrollment on DI applications and new DI cases within states over time, using
state policy (the maximum WC benefit) as an instrument for WC enrollment.
Despite a strong first stage fit, we find no statistically significant evidence that
WC tightening caused DI rolls to increase, although the standard errors are large
enough that we cannot reject effects of substantial magnitude. We conclude it is
unlikely that state WC changes were a meaningful factor in explaining the rise in
DI during our study period of 19862001, although further study using individual
level data is warranted.
* The authors affiliations are, respectively, Department of Economics, College of William and Mary,
E-mail: mpmcinerney@wm.edu; School of Public and Environmental Affairs, Indiana University, Blooming-
ton, IN, E-mail: simonkos@indiana.edu. Simon acknowledges support from the National Institute on Disabil-
ity and Rehabilitation Research (NIDRR) through the Employment Policy RRTC funded to Cornell
University (Grant No. H133B040013). We thank Richard Burkhauser, John Burton, Xuguang Guo, Andrew
Houtenville, and David Stapleton for sharing data that made this project possible. We are grateful to Richard
Burkhauser, Susanne Bruyere, John Burton, Mark Duggan, Xuguang Guo, Judith Hellerstein, Andrew
Houtenville, and David Stapleton for advice and encouragement, and two anonymous reviewers. We thank
Anna Choi, James Hunsberger, and Aparna Lhila for outstanding research assistance.
INDUSTRIAL RELATIONS, Vol. 51, No. 1 (January 2012). 2012 Regents of the University of California
Published by Wiley Periodicals, Inc., 350 Main Street, Malden, MA 02148, USA, and 9600 Garsington
Road, Oxford, OX4 2DQ, UK.
57
58 / MELISSA MCINERNEY AND KOSALI SIMON
Introduction
IN ADDITION TO TRADITIONAL FORMS OF PRIVATE AND PUBLIC MEDICAL INSURANCE,
two other large public programs help pay for costs associated with ill health.
In 2008, Workers Compensation (WC) insurance provided $57.6 billion in
medical care and cash benefits to employees who are injured at work or con-
tract a work-related illness, and Social Security Disability Insurance (DI) pro-
vided $106 billion to individuals who suffer from permanent disabilities and
are unable to engage in substantial gainful activity. During the 1990s, the
employment of working-aged people with disabilities fell dramatically (e.g.,
Stapleton and Burkhauser 2003), while total annual spending on DI and the
number of individuals receiving DI rose. There have been many factors
found to be partially responsible for explaining these trends. For example,
Duggan and Imberman (2008) attribute much of the rise in the DI rolls to
the characteristics of the individuals insured by the DI program, the state of
the economy, and the generosity of program benefits, while other researchers
have considered the effect of federal policy changes such as the Americans
with Disabilities Act of 1990 in explaining changes in employment rates
among the disabled community (e.g., Acemoglu and Angrist 2001; DeLeire
2000a, 2000b; Houtenville and Burkhauser 2003). However, there is a sub-
stantial amount of this trend left to explain. As policymakers seek to better
integrate working-aged people with disabilities into the workforce, it is
important to understand the factors leading to this rise in DI spending and
the concurrent fall in employment rates among individuals with disabilities.
This is an important policy question because while employment of working-
aged people with disabilities was falling, the poverty rate for adults with
disabilities also deteriorated relative to their nondisabled peers (Burkhauser,
Houtenville, and Rovba 2005). There may also be an unexplored role played
by state policies to explain changes in the DI program and related outcomes.
In this paper, we consider the impact that changes in one specific program,
Workers Compensation (WC), may have on entry into the DI program (mea-
sured as DI applications and new DI cases) through their effect on reduced
WC enrollment.
As motivation for considering the relationship between WC and DI, we note
that total real dollars spent on WC fell 12% during the 1990s when many
states enacted changes to make WC more difficult for workers to claim. This
has led to the question of whether there is any substitution between the two
programs that may help explain the observed trend in DI, above and beyond
the factors researchers already have found account for some of this trendthat
is, do individuals who are at the margin and potentially eligible for both pro-
grams substitute toward DI when WC rules tighten? Although this has been
State WC Program Changes and Federal DI Use / 59
hypothesized in the literature (see, e.g., Burton and Speiler 2001; Sengupta,
Reno, and Burton 2009; Thomason, Burton, and Schmidle 2001), there has
been no empirical testing of this conjecture other than contemporaneous work
by Guo and Burton, an earlier version of which is in LERA Proceedings
(2008).
The expected impact of WC policy changes on the employment of individu-
als with disabilities is ambiguous. Consider a broad definition of disability that
includes anyone who suffers a workplace injury who would have received any
WC. Some, but not all, have injuries that are severe enough to be eligible for
federal disability benefits. For the severely injured workers, a restriction of
WC may lead to a decline in the employment rate for those persons with dis-
abilities if they earlier received WC but now would move onto DI.1 For the
less severely injured workers, any decline in receipt of WC that is not made
up for by an increase in receipt of DI may cause the individual to return to
work soonerthat is, being without either program causes a stronger attach-
ment to the labor force than being on WC by itself. Thus, the theoretical effect
of WC policy changes on the employment of people with disabilities is ambig-
uous. Under a restricted definition of disability that only includes workers who
would be eligible for DI were they to apply, tightening WC benefits is pre-
dicted to unambiguously reduce the employment of people with disabilities.
Of course, one assumption inherent in this study of whether WC tightening
leads to DI program growth is that there are no concurrent changes in DI pro-
gram generosity that explain the trend. If DI became more generous, the
observed national trend may simply reflect the fact that individuals eligible for
both programs may choose the more generous of the two, leading to a growth
in total DI spending and a fall in total WC spending. Autor and Duggan
(2003) document an increase in the DI earnings replacement rate during the
1980s and 1990s that especially favored lower earners.2 We argue that the
analysis we conduct is able to identify how WC generosity impacts claiming
in the two programs, above and beyond any shifting toward DI that would
occur because of DIs increasing generosity. First, we quantify WC generosity
by the states maximum weekly benefit amount. Injured workers potentially
1
DI programs are generally available only to those not employed, whereas once permanent partial WC
benefits are awarded, injured workers in many states may return to work while receiving benefits. For exam-
ple, in California permanent partial disability benefits are awarded based on an injured workers previous
earnings and a doctors determination of the extent of permanent disability (adjusted for the workers age
and occupation). Permanent partial disability benefits are not reduced if the injured worker is employed
while receiving such benefits (California Division of Workers Compensation 2009).
2
Autor and Duggan (2003) show that due to rising earnings inequality, the bend points used in the com-
putation of DI benefits increased by more than average earnings of low earners. This made DI relatively
more generous for low earners during this time. There were no similar, concurrent changes to WC generos-
ity for low earners whose weekly benefit is generally two-thirds of pre-injury earnings.
eligible for maximum WC benefits are less likely to benefit from the increased
DI generosity for low earners that Autor and Duggan (2003) identify. Second,
WC benefit generosity is determined at the state level, whereas changes in DI
generosity impacted all workers in the United States. We exploit state-by-year
level changes in WC generosity to isolate the role of WC changes in any shift-
ing away from WC to DI, above and beyond national trends in increased DI
generosity.
In the remainder of the paper, we first describe the two programs to consider
the scope for a causal effect of WC reforms on DI applications. Next, we
describe the prior literature relevant to this question and lay out our empirical
tests and data. We first present national descriptive time trends in DI and WC
program statistics and then conduct both visual and statistical inspection of
time trends in the two programs within states and over time. Our causal test of
the effect of declining WC generosity on DI outcomes uses policy variation
within states and over time in the real maximum weekly WC benefit for severe
cases as an instrument for state WC enrollment of severe cases.3 In the case of
several state-years, the variation in the maximum weekly benefit occurs partly
because the state average weekly wage, to which it is pegged, changes. While
this relationship is set by policy, it also involves changes in worker outcomes
in the state; we conduct two types of analyses to ensure that our effects are
robust to variation that occurs only from policy changes.
As a synopsis, the first observation from our data inspection documents that
although there is a concurrent rise in DI and fall in WC at the national level,
the rises in DI are not occurring in the same states as the falls in WC. We then
examine the causal effect of decreasing WC participation on DI cases and
applications, using WC benefit policy parameters as the instrumental variable.
Despite a strong first stage, we find no evidence of a second-stage effect of
WC case declines on DI caseloads or applications. However, the standard
errors do not allow us to reject sizable effects (e.g., large, negative effects con-
sistent with substitution between the two programs on the order of one new DI
case for every five injured workers who do not receive WC, as well as large
positive effects consistent with the programs being complements on the order
of two new DI applicants for every five workers who receive WC). We con-
clude based on the patterns seen in the visual inspection of the data and the
3
Data on DI applications and new cases were generously provided by Richard Burkhauser, Andrew
Houtenville and David Stapleton. Data for 19801992 were created by Lewin Associates and used in Bur-
khauser, Butler, and Weathers (2002), and Burkhauser, Butler, and Gumus (2004). Data for 19932001 were
created for and used in Houtenville and Burkhauser (2003). We thank John Burton and Steve Guo for shar-
ing data on permanent WC receipt, which originate from the National Council on Compensation Insurance
(NCCI). We obtained the permanent partial disability (PPD) maximum weekly benefit amounts from the
annual U.S. Department of Labor publication, State Workers Compensation Laws.
IV results that it is unlikely that WC policy changes have contributed to the
rise of DI applications, new cases, and by extension, to the decline in employ-
ment among individuals with disabilities.
Background
Workers Compensation Benefits. Workers Compensation insurance pro-
vides medical care and cash benefits to employees who are injured at work or
contract a work-related illness and is administered at the state level. Workers
injured on the job receive medical care immediately but must be away from
work between 3 and 7 days before receiving cash benefits to replace lost
wages. WC is one of the largest social insurance programs in the United
States. In 2008, individuals received a total of $57.6 billion in cash and medi-
cal payments for WC$28.6 billion in cash benefits and $29.1 billion in med-
ical care (Sengupta, Reno, and Burton 2010). In 2008, spending on cash
benefits for WC was larger than federal spending on Temporary Assistance for
Needy Families ($15.5 billion), comparable to federal spending on the Supple-
mental Nutrition Assistance Program ($34.61 billion), and about two-thirds the
size of Unemployment Insurance benefits ($40.7 billion) and the federal
Earned Income Tax Credit ($42.4 billion).4 The injured workers weekly bene-
fits are a function of their weekly earnings and generally replace two-thirds of
pre-injury weekly earnings, subject to a maximum that varies across states. In
2008, the state maximum ranged from $399 per week in Mississippi to $1311
per week in Iowa.
Typically, injured workers receiving cash benefits first receive temporary
total disability (TTD) benefits. Injured workers receive these benefits until
one of three things happen: (1) they are able to return to work, at which
point they are no longer considered totally disabled and cash benefits cease;
(2) the physician has determined that the injured worker has reached Max-
imum Medical Improvement and the injured worker is evaluated for per-
manent disability benefits; or (3) the injured worker has reached the state
time limit for temporary benefits and is evaluated for permanent disability
benefits.
4
Data on federal Unemployment Insurance spending from ET Financial Data Handbook 394 Report
(http://www.workforcesecurity.doleta.gov/unemploy/hb394.asp, viewed December 5, 2011), Supplemental
Nutrition Assistance Program from SNAP Annual Summary (http://fns.usda.gov/pd/SNAPsummary.html,
viewed January 11, 2011), Temporary Assistance for Needy Families from FY 2008 TANF Financial Data
(http://www.acf.hhs.gov/programs/ofs/data/2008/overview.html, viewed January 21, 2011), and Earned
Income Tax Credit estimates from Estimates of Federal Tax Expenditures for Fiscal Years 20082012 (JCS-
2-08 http://www.jct.gov/publications.html?func=startdown&id=1192, viewed January 21, 2011).
There are two types of permanent disability benefits in WCpermanent par-
tial disability (PPD) benefits and permanent total disability (PTD) benefits.
The most common is PPD benefits, comprising 66 percent of all WC cash
benefits in 2004. If any injured workers are substituting DI for WC because of
a policy change in WC, it is likely to be workers potentially eligible for PPD
who are driving the aggregate trends in DI outcomes and the employment of
working-aged people with disabilities. Those on PTD constitute less than 1
percent of all WC cases and would have been more likely to be on DIand
unable to workregardless of the tightening of WC policies as the nature and
duration of their injuries are so severe.5
A weekly benefit amount is computed for individuals awarded PPD bene-
fits. The weekly benefit is generally two-thirds of a workers pre-injury
weekly earnings, subject to a maximum that varies by state.6 There are some
important distinctions in how the duration of PPD benefit receipt is deter-
mined. As by definition the injury is only partial in nature, states dictate the
number of weeks the injured worker may receive the corresponding weekly
benefit, using one or more of the following methods. First, some injuries (such
as the loss of an arm or a thumb) are compensated based on a fixed schedule
that dictates the number of weeks an injured worker may receive benefits.7
Second, PPD claims for injuries that are not on the states schedule,
unscheduled injuries, tend to be provided as an increasing function of the
fraction of the body injured. Either this benefit tends to be delivered as a lump
sum, or the injured worker will receive benefits for a fixed number of weeks
(Barth 2005). When benefits are delivered as a lump sum, the injured worker
receives a one-time payment while agreeing to waive the right to receive fur-
ther cash or medical benefits for the claim (Barth 2005). Third, in thirteen
states, unscheduled PPD benefits are related to the loss of earnings in the
workplace (Barth 2005).
Social Security Disability Insurance. Disability Insurance is a federal pro-

gram administered by the Social Security Administration. In 2008, DI paid
$106 billion to 7.8 million disabled workers and their families. To receive
5
We report results using PPD as our main category of WC data, but have also estimated models with
PTD cases added to PPD and found the results to be almost identical.
6
Alaska and Washington are two exceptions. Instead of computing benefits weekly, PPD benefits are
paid as a one-time lump sum.
7
For example, in Ohio, injured workers who lose a thumb will receive scheduled PPD benefits for
60 weeks, whether or not they return to work. This is in contrast to DI benefits, which are reduced or cease
when an individual returns to work. Ohio Revised Code 4123.57.
DI benefits, disabled workers must have worked in Social Securitycovered
employment in the past and show that they have a disability that is expected
to last at least 12 months and that they are unable to do the jobs they most
recently left or engage in any substantial gainful activity. Neither medical
nor cash benefits are available at the time of the disabling event. At least
5 months must pass between the disabling event and the time period for
which the first benefit check is received, and there is a mandatory 29-month
waiting period between the disabling event and the onset of Medicare bene-
fits. Furthermore, the application process can be long and costly; it can take
between three and 5 months for a decision to be made, during which time
the applicants must not engage in substantial gainful employment.8 State
boards make the initial disability determination and play a key role in
whether a DI award is granted; rejected applicants can appeal the decision.
According to the 2009 Annual Statistical Report on the Social Security Dis-
ability Insurance Program, of those applying for DI in 2008, only 38 percent
of applicants were awarded DI at the state board level. This means that over
60 percent of applicants either appealed the decision or were never awarded
DI benefits. Thus, many applicants face an initial rejection, and those choos-
ing to appeal this decision must remain out of work and without DI benefits,
for well over 5 months. In fact, disabled individuals appealing an initial deci-
sion often remain out of work much longer than the requisite 5 months.
There is a 24-month waiting period before Medicare coverage begins for DI
recipients, and Livermore, Stapleton, and Claypool (2009) find that the aver-
age DI recipient is already 9 months into the 24-month Medicare waiting
period when awarded benefits.
If awarded DI, the proportion of covered wages replaced declines as covered
wages increasethe Social Security Administration computes benefits follow-
ing the same benefit schedule used to compute Social Security retirement ben-
efits. Recipients receive the full award for their life, and the average after-tax
replacement rate is approximately 60 percent (Bound and Burkhauser 1999).9
After a 2-year waiting period, recipients are eligible to receive Medicare cover-
age. There are work incentives to encourage recipients to return to work, but
those policies are fairly recent.
8
Substantial employment is defined as earning more than $980 a month pre-tax in 2009. While an
application is being considered, the person may be working (and even after it has been approved) as long as
the work does not exceed $700 per month. This amounts to <2.5 h of work a day at around $15 h. If the
person earns more than that, they will be denied on technical grounds (Social Security Administration
2009).
9
Individuals awarded DI benefits receive DI benefits until they reach full retirement age, at which point
the benefit becomes an Old Age benefit from the Social Security Administration. However, the payments do
not change, so the individual essentially receives benefits for life (Social Security Administration, 2009b).
Relationship between Workers Compensation and Social Security Disability
Insurance. Recent research has found that when generosity is reduced in
one program, there have been substantial increases in enrollment in a more
generous program. For example, Duggan, Singleton, and Song (2007) docu-
ment an increase in DI receipt after increases to the full retirement age for
Social Security retirement benefits, effectively reducing generosity. Schmidt
and Sevak (2004) find increases in enrollment for the Supplemental Security
Income (SSI) program following welfare reform. Several other papers show
a substitution effect between SSI and welfare. Generosity effectively
increased in SSI in 1990 following the Sullivan v. Zebley decision, and
Duggan and Kearney (2007), Garrett and Glied (2000), and Kubik (2003)
document substitution toward SSI and away from welfare following this
decision.
The potential for spillovers to the DI program from tightening WC state
programs has been suggested by researchers familiar with the two programs,
but the empirical economics literature has just begun to investigate the rela-
tionship. In 2004, the National Academy of Social Insurance (NASI) held a
conference on the relationship between these two programs.10 In some of the
earliest discussions of this phenomenon, Mont, Burton, and Reno (2000) and
Thomason, Burton, and Schmidle (2001) point to the growth rate of WC
benefits falling dramatically at the national level between 1992 and 1998.
They show that after many years of double-digit growth rates, the annual
percentage increases in WC costs for medical and cash benefits grew (or
contracted) by minuscule amounts between 1992 and 1998, the latest year of
their data. During this same period, national DI benefit spending increased
by 33 percent. The possibility of a link between the two programs has also
been suggested in the annual publication, Workers Compensation Benefits,
Coverage, and Costs, published by NASI. This publication presents the two
national series, DI benefits per $100 of covered wages (or the earnings of
those insured under WC) and WC benefits per $100 of covered wages for
the years 19802007. Between 1980 and 1990, WC benefits per $100 in
covered wages were rising, while DI benefits per $100 in covered wages
were falling. These trends reversed from 1990 onward, suggesting a negative
relationship between the two programs when benefits are measured relative
to covered wages.
10
Proceedings from this conference were published in the Social Security Bulletin, 65(4), released May
2005. In it, James B. Lockhart, III, Deputy Commissioner of Social Security, stressed how important the
relationship between the two programs is to his agency. He noted that the Social Security Administration
devotes over 1000 full time equivalent employees to handling the offset between the two programs (Lock-
hart 2005).
TABLE 1
RECIPIENTS OF SOCIAL SECURITY DISABILITY INSURANCE AND WORKERS COMPENSATION, 19922001
Workers Compensation
(WC) Recipients DI Recipients
Age 40.19 (11.27) 51.05 (13.35)
Female 0.37 0.55
Single 0.36 0.46
Black 0.12 0.17
Hispanic or other 0.07 0.05
Educational Attainment
Less than HSD 0.19 0.32
High school degree 0.43 0.38
Some college 0.29 0.20
College or more 0.09 0.10
Receive Medicaid 0.07 0.26
Currently working 0.11 0.03
Wage and salary income (2000 $) 17,980 (20,095) 3544 (11,972)
Share receiving DI 0.08 1.0
Share receiving WC 1.0 0.02
N 10,085 43,937
SOURCE: Authors calculations from March CPS (weighted by population weights).
One relevant question for examining the causal connection between the two
programs is the extent to which workers are potentially eligible for both.
Although there are no data we know of that would allow us to measure how
many of those receiving permanent WC benefits are also potentially eligible
for DI, we can examine characteristics of recipients of the two programs.11
Table 1 presents characteristics of WC and DI recipients in the March Current
Population Survey (CPS) from 1992 to 2001. DI recipients are older than WC
recipients, and DI recipients are also more likely to be women. DI recipients
have lower earned income than WC recipients, which is not surprising given
the work restrictions in the DI program. Eight percent of those on WC also
report receiving DI, while two percent of those on DI also report receiving
WC.
Another way to judge whether there are WC recipients who may be poten-
tially eligible for DI and vice versa is to look at data on the extent to which
the injuries of DI recipients are work related. Reville and Schoeni (2005) use
11
Roberts (1994) obtained data from the Michigan Bureau of Workers Disability Compensation for
claims occurring in 1984 and 1985. Beginning in 1981, WC benefits were coordinated with other types of
employer-paid benefits. Employers were required to report the types of additional benefits injured workers
received. Even this dataset does not contain information about DI receipt because at the time of this study,
WC benefits were not reduced by DI benefits in Michigan. (Approximately one percent of WC recipients
concurrently received some other employer-provided long-term disability benefit.)
the 1992 Health and Retirement Study that is reflective of those 5161 years
of age and show that among those receiving DI, 37 percent became disabled
because of a work-related injury. They also show that the fraction on DI
because of a workplace injury differs by type of disability, with high fractions
in musculoskeletal causes. Similarly, Sengupta, Reno, and Burton (2009) find
that in 2008 approximately 15.5 percent of DI beneficiaries had some connec-
tion to WC (or other public disability benefits, such as civil service disability
benefits, military disability benefits, or state temporary disability benefits).
Policymakers have recognized from the start of the DI program that a cer-
tain group of workers could be eligible for both programs. Starting in 1965, a
Social Security Amendment specified that workers who receive DI benefits
will see reductions in the DI payments when they receive WC payments so
that the combined value of DI and WC payments cannot exceed 80 percent of
a workers average pre-injury earnings. In fifteen states, there is a reverse off-
set, and WC payments, not DI payments, are reduced. In states with the
reverse offset, employers and insurers (and the state) have a clear incentive to
see that more of the WC cases are on DI concurrently; in all states, there is an
incentive to see fewer workers on WC and more on DI to keep employer WC
costs low and a state attractive to business.
From the injured workers perspective, WC and DI benefits complement one
another. WC cash and medical benefits are available immediately after the
workplace injury in contrast to DI cash benefits (and Medicare) that are subject
to lengthy waiting periods. WC benefits are not taxed, so after-tax replacement
rates can be close to one for WC (Meyer 2002). However, DI benefits are
taxed and have after-tax replacement rates of approximately 60 percent. There
are also features of the DI program that are attractive to injured workers, even
in the absence of WC program changes and despite the offset rules described
earlier. While many WC benefits are time limited, DI benefits are not time lim-
ited, which is especially attractive to WC PPD beneficiaries who often receive
benefits for a fixed number of weeks. Furthermore, DI benefits are adjusted
over time to reflect changes in the cost of living, whereas most state WC bene-
fits are not adjusted in a similar way. DI also has a third advantage over WC;
when Medicare coverage begins for DI beneficiaries, it provides health cover-
age for all conditions, not only for the workplace injury.12 Access to Medicare
is especially valuable for injured workers who do not return to work and lose
employer provided health insurance. In fact, qualitative evidence from surveys
of health workers suggests that fear of losing benefits, such as comprehensive
employer health insurance, is one of the reasons injured workers choose not to
apply for WC benefits (Galizzi et al., 2010). Finally, as documented in Autor
12
Individuals who receive a lump sum WC benefit also have every incentive to pursue DI benefits.
and Duggan (2003), the DI program was becoming more generous during the
time period we examine, especially to low-wage workers. This may be another
reason why injured workers might seek DI benefits. In this paper, we examine
how WC program changes impact the incentive for injured workers to claim
DI benefits above and beyond concurrent claiming that may arise because the
two programs complement one another. We focus on changes to maximum
benefit levels to isolate the effect among workers least likely to be affected by
increased generosity of DI. In the next section, we examine the motives to
workers, employers, and states to shift injured workers from WC to DI.
Hypothesis
Our hypothesis is that as the expected benefits of applying for WC decline,
certain marginal workers with permanent injuries are more likely to apply for
DI and more likely to become DI cases. Employers (particularly those who
self-insure for WC), insurers, and the state also have incentives to shift work-
ers from WC to DI, and they could play a role in making WC harder to obtain
or in assisting the workers to apply for DI instead. To the extent that there are
enough such marginal workers to affect the aggregate statistics, the causal rela-
tionship should be observable in state-level data.
The motivation behind our hypothesis is similar to that used in Autor and
Duggan (2003), who document the decline in unemployment in the 1990s and
the concurrent increase in the receipt of DI. They argue that as DI became rel-
atively more generous over time, the number of workers who are conditional
applicants increased and more of them applied for DI. They classify potential
DI applicants into three groups: always apply, never apply, and apply if unem-
ployed (the conditional applicants). The group of conditional applicants will
grow if the probability a DI claim is accepted grows or the benefit grows. Our
hypothesis is quite similar. If the cost of applying to DI remained the same for
these injured workers, but the generosity of WC decreased because of WC
reform, then more workers will apply for DI, and fewer workers will apply for
WC (although those who apply for DI might continue to apply for WC as well
if they are the more severely injured workers who will not be denied WC
because of WC reforms).13 As in Autor and Duggan (2003), we use policy
13
Another factor that would determine whether the application was successful would be the climate for
reviewing applications. We do not have data on this factor, nor on any policy variables that would be poten-
tial instruments for it. We acknowledge that although it would have been desirable to fully capture in our
model all factors that affect the number of individuals on WC in a state, it is sufficient for our main purpose
to capture sufficient exogenous variation that allows us to test the hypothesis and quantify the impact of
increased WC generosity.
changes to characterize the generosity of a program. We do so in the case of
WC benefits; as they become less generous, injured workers would be more
likely to substitute toward DI.
Workers Compensation Policy Changes. In testing the causal effect of fall-

ing WC enrollment on DI applications and cases, we use an instrument for
WC enrollment as the two series may otherwise be affected by unobserved
omitted factors at the state level over time. Our instrument for WC enrollment
is changes in WC policy that affects the programs generosity. We focus on
changes in WC weekly benefit generosity driven by regulations and statutes as
our instrument, and only for the type of cases that are most likely to have DI
as an outside option. We use variation in the state regulation-set maximum
weekly benefits for permanent partial disability as our instrument for changes
in WC enrollment when studying the effect of declines in WC enrollment on
DI applications and new cases.
Recall that because the WC benefit award is a function of a workers previ-
ous earnings, subject to this state-set maximum, our policy parameter affects
WC benefits more for injured workers with high weekly earnings. This raises
the concern that the treatment effect we are identifying may not apply broadly
but only to a small subset. However, using March Current Population Survey
data for the years 1986 through 2001, we determined that 28 percent of work-
ers would be eligible to receive the state-set weekly maximum benefit. The
share of workers impacted by changes in the generosity of the maximum is
not small, but our results are nevertheless to be interpreted as the effect of the
treatment on the treated.
The mean maximum benefit rises slightly from $375 in 1983 to $480 in
2006 (in 2000 $). Although this aggregate trend suggests that WC was becom-
ing more generous over this time period, it obscures changes within a state. A
state will rarely decrease the nominal value of the maximum weekly benefit,
but in many states and years, the nominal maximum benefit level stays con-
stant and signifies a decline in real terms. In many states that do increase the
nominal maximum in a given year, the increases to the maximum do not keep
up with inflation. In fact, 40 percent of our observations (state-years) experi-
ence a decline in the real value of the maximum weekly benefit.
Relevant Prior Work. The main contribution of our paper is to quantify

whether there is substitution between WC and DI because state WC program
changes resulted in DI becoming relatively more generous to disabled individ-
uals over the time period. To our knowledge, the only other papers to address
this question result from concurrent work in progress by Guo and
Burtonwhich updates Guo and Burton (2008)and conclude that WC
FIGURE 1
EMPLOYMENT RATE OF WORKING-AGE PEOPLE WITH DISABILITIES
35
30
25
20
15
10
0
1980 1985 1990 1995 2000 2005
SOURCE: Bjelland, Erickson, and Lee (2008).
tightening increased DI applications. One way our approach differs from theirs
is that we examine how WC enrollment changes are affected by policy-
prescribed changes in state maximum weekly benefits and how that then
affects DI applications and cases. In contrast, Guo and Burtons work almost
exclusively examines measures they construct of expected WC permanent
generosity as right-hand-side variables in equations explaining DI outcomes.
The main difference between our two papers, however, is our inclusion of
year fixed effects. We include year effects to capture nonlinear year-to-year
changes happening nationwide that would impact all those potentially eligible
for either program. We are interested in how state WC generosity impacts
DI applications and cases, so we wish to control for these national changes
in estimating the relationship between state WC generosity and DI claiming.
In our first-stage regressions, shown below, we find a positive and statisti-
cally significant relationship between WC generosity and WC receipt. This
suggests that enough variation remains within states to identify any possible
substitution between the two programs that might arise as state WC generos-
ity changes.
Descriptive Analysis
National Trends in Workers Compensation and Social Security Disability
Insurance. As background, we first describe trends in national data series
related to WC and DI. Figure 1 shows the well-known dramatic decline in
employment of people with disabilities that occurred since 1990. Researchers
FIGURE 2
SOCIAL SECURITY DISABILITY INSURANCE AND WORKERS COMPENSATION PROGRAM OUTCOMES,
19862001
1,400,000
1,200,000
1,000,000
800,000
Applications to DI
600,000 New DI Cases
WC PPD Cases
400,000
200,000
0
1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
SOURCE: Data on new DI cases and applications to DI provided by Burkhauser and Stapleton. Permanent Partial Disability
(PPD) claim data from Burton and Guo.
have found many possible reasons for this pattern, including changes in the
types of individuals who are of working age and with disabilities (Houtenville
and Daly 2003); changes in the types of work required and the relative com-
petitiveness of those with disabilities (Stapleton, Goodman, and Houtenville
2003); the diminished value of employment because of rising healthcare costs
for the disabled (Hill, Livermore, and Houtenville 2003); changing health pro-
file among the disabled (Kaye 2003); the Americans with Disabilities Act
(DeLeire 2003); the expansion of federal disability programs in terms of eligi-
bility and benefits (Goodman and Waidmann 2003); as well as the state of the
economy (Duggan and Imberman 2008).
Reflecting the trend in Figure 1, Figure 2 shows that the DI rolls have risen
during the same period. We plot the number of initial DI applications and initial
DI allowances, what we will refer to as new DI cases. In Figure 2, we also
illustrate changes in PPD receipt. The trends in PPD receipt seem to mirror
trends in DI applications, at least through the mid-1990s. When DI applications
(and new cases) fell in the late 1980s, PPD receipt rose. DI applications (and
new cases) rose in the early 1990s, a period when PPD receipt was falling.14
Figure 2 provides a sense of the magnitude of the total number of DI and
WC recipients at the national level, but does not describe the situation at the
14
Note that the negative relationship between WC and DI is not as clear as in the two series published
in NASI publications. When we do not divide benefits by covered wages, we only see a negative relation-
ship between the two programs during the 1990s. In contrast, NASI data, presented per $100 in covered
wages, suggest a negative relationship from 1980 to 2007. We do not present results by covered wages
because data on covered wages are not available to us at the state level, nor are these data available at the
national level prior to 1989.
TABLE 2
STATE-LEVEL MEANS OF WORKERS COMPENSATION (WC) OUTCOMES, DI OUTCOMES, AND WC
PROGRAM PARAMETERS (2000 $)
Mean (SD)
WC Claim Rate (per 100,000 workers)
Permanent partial disability cases 650.90 (377.95)
DI Claim Rate (per 100,000 workers)
Applications to DI 1122 (319)
New DI cases 409 (96)
PPD Program Parameters
PPD maximum weekly benefit 410.57 (156.76)
Unemployment rate 5.52 (1.56)
Share of population ages 4554 22.36
Share of population ages 5564 16.42
Industrial composition
Agriculture, forestry, fisheries 2.47
Mining 0.61
Construction 6.74
Manufacturing-durable goods 11.35
Manufacturing-nondurable goods 7.84
Transportation, communication, public utilities 8.17
Wholesale trade 4.39
Retail trade 13.87
Finance, insurance, and real estate 7.46
Business and repair services 6.48
Personal services 3.39
Entertainment and recreation 1.35
Professional and related services 25.90
N 724
Data set consists of one observation for a state-year, for 1986 through 2001. Data on permanent partial disability (PPD)
WC receipt are missing in select states and years because not all states report claims information to the National Council
on Compensation Insurance (NV [missing only 19861995], ND, OH, TX [missing 87 and 89 only], WA, WV, and
WY). The average state has two million workers. There are only 708 observations for real maximum PPD benefits
because benefits are determined by lump sum in Alaska (and Washington).
state level. In Table 2, we present summary statistics for our measures of WC

and DI receipt at the state level. So that we can make comparisons across
states of different sizes, throughout the rest of the paper we present all mea-
sures of WC and DI receipt per 100,000 workers in a state and year. We see,
on average, 651 PPD cases (per 100,000 workers) at the state year level. DI
applications are much more common than the severe WC claims. There are
more than 1000 applications to DI per 100,000 workers each year, resulting in
over 400 new DI cases per 100,000 workers.15
15
Note that our measures of WC receipt reflect both the stock of existing severe cases and the flow of
new cases. In contrast, our measure of DI receipt only reflects new cases. When we consider both the stock
and flow of DI cases, we find on average roughly 4000 DI cases (per 100,000 workers).
The Relationship Between Social Security Disability Insurance Outcomes
and Workers Compensation OutcomesTrends Within States Over
Time. Using our state by year data set, we next look within states to determine
whether the trend seen in Figure 2 is upheld at the state level. To present this
most concisely in a visual manner, Figure 3a,b are two scatterplots depicting
within-state year-to-year changes in measures of DI and WC receipt. Changes
in DI are presented on the vertical axis, and changes in WC are on the hori-
zontal axis. These scatterplots are constructed from state-by-year data, and
each point represents a states 1-year change in DI receipt and 1-year change
in permanent partial WC receipt in a given year. In Table 3, we show the
regression analysis version of these scatterplots, with and without additional
controls. The first specification contains no covariates and then adds state and
year fixed effects individually, and then in combination, as additional controls
for time invariant state characteristics as well as national time trends that
would affect both programs, such as the aging of the population.
DIs;t b0 b1 WCs;t ls st e1;s;t 1
Disability Insurance (DI) and WC stand for the different measures of DI

and WC receipt illustrated in Figure 2 and Table 2.
In (2), we also control for macroeconomic conditions that change within a
state over time and may affect DI applications, such as the unemployment rate.
Workplace injury ratesand the type of workplace injuriesvary across
industries, so we also control for the share of a states workforce in the follow-
ing industries: agriculture; mining; construction; manufacturing-durable goods;
manufacturing-nondurable goods; transportation; wholesale trade; retail trade;
finance, insurance, and real estate; business and repair services; personal ser-
vices; and professional services. Finally, we also control for the states demo-
graphic composition. An older population may be more at risk for DI and
permanent WC benefits, so we include the share of the state population ages
4554 and ages 5564.16
DIs;t c0 c1 WCs;t c2 unems;t c3 share45 to 54s;t
c4 share55 to 64s;t Ws;t K ls st e2;s;t 2
where Ws,t is a vector of the share of workers employed in different industries,

and other variables are as indicated by labels. Although this richer specifica-
tion controls for more of the factors that may influence WC and DI receipt,
16
In results not shown, we test sensitivity to the fact that workers are most prone to workplace injuries
when beginning a new job, so an increase in the states population may suggest an influx of workers. There-
fore, we also include the number of workers ages 2564 (in millions). Results are not sensitive to the inclu-
sion of a control for population.
FIGURE 3
(A, B)WITHIN-STATE YEAR-TO-YEAR CHANGES IN DI AND WORKERS COMPENSATION (WC)
RECEIPT. (A) WITHIN-STATE ANNUAL CHANGES IN NEW DI CASES AND WC PERMANENT PARTIAL
DISABILITY (PPD) CASES. THE CORRESPONDING CORRELATION COEFFICIENT IS )0.007 AND IS NOT
STATISTICALLY SIGNIFICANT (P = 0.85). (B) WITHIN-STATE ANNUAL CHANGES IN DI APPLICATIONS
AND WC PPD RECEIPT. THE CORRESPONDING CORRELATION COEFFICIENT IS )0.1313, AND IT IS
STATISTICALLY SIGNIFICANT (P < 0.001)

A:
200
Within-state 1-year change in new DI cases
-100 0-200 100
-1000 -500 0 500

Within-state 1-year change in WC PPD claims
B:
600
Within-state 1-year change in DI apps.
-200 0 200
-400 400
-1000 -500 0 500

Within-state 1-year change in WC PPD claims
SOURCE: Data are at the state-by-year level. Each point represents a states 1-year change in DI receipt and 1-year change in
permanent WC receipt in a given year. As an example, DI applications (per 100,000 workers) rose from 1108 in 1992 to
1303 in 1990 in CA. So the value of the within-state change for CA in 1993 is 194.4.
TABLE 3
THE RELATIONSHIP BETWEEN SOCIAL SECURITY DISABILITY INSURANCE OUTCOMES AND WORKERS
COMPENSATION OUTCOMES
DI applications per New DI claims per

100,000 workers 100,000 Workers
Panel A: no covariates
Permanent partial cases per 100,000 workers )0.126 (0.031)** )0.023 (0.009)**
Panel B: state fixed effects only
Permanent partial cases per 100,000 workers 0.038 (0.033) 0.072 (0.012)**
Panel C: year fixed effects only
Permanent partial cases per 100,000 workers )0.166 (0.030)** )0.051 (0.009)**
Panel D: state and year fixed effects only
Permanent partial cases per 100,000 workers )0.036 (0.028) 0.010 (0.012)
Panel E: All covariates
Permanent partial cases per 100,000 workers 0.026 (0.025) 0.025 (0.012)**
Mean of dependent variable 1122 (319) 409 (96)
Each regression contains 724 state-by-year observations between 1986 and 2001. Data on permanent Workers
Compensation receipt are missing in select states and years because not all states report claims information to the
National Council on Compensation Insurance (NV [missing only 19861995], ND, OH, TX [missing 87 and 89 only],
WA, WV, and WY). Each cell contains a coefficient estimate and standard error for the independent variable of interest
from a separate regression where the dependent variable is the column heading, and the independent variable of interest
is the row heading. Each regression includes an intercept. All dollars are in year 2000 terms. The full set of covariates in
Panel E includes the state unemployment rate; the share of the state population aged 4554 and the share of the state
population aged 5564; the share of workers in the following industries (agriculture, mining, construction,
manufacturing-durable goods, manufacturing-nondurable goods, transportation, wholesale, retail trade, finance, insurance,
and real estate, business and repair services, personal services, and professional services are the left out category); as
well as state and year fixed effects. Regressions are weighted by the states population between the ages 25 and 64. R2
for the regressions in Panels D and E with DI applications as the dependent variable ranges from 0.92 to 0.94. R2 for the
regressions in Panels D and E with new DI claims as the dependent variable ranges from 0.82 to 0.85.
we can only interpret these results as the conditional correlation between WC

and DI receipt because we still are unable to capture all of the relevant vari-
ables that would allow us to infer causality. The omitted variables likely bias
our results upward because any omitted variable that would increase WC
receipt, such as increased propensity to claim social insurance benefits, likely
also increases DI receipt. Similarly, omitted variables that are expected to
decrease WC receipt, such as improvements in workplace safety, are also
expected to decrease DI applications and cases.
Figure 3a, which shows the relationship between PPD receipt and new DI
claims, is such that the scatterplot depicts a mass of points (with outliers) and
no clear trend. Although the corresponding correlation coefficient is negative
()0.007), it is very close to zero and not statistically significant (p = 0.848).
In Figure 3b, we also examine the relationship between PPD receipt and applica-
tions to DI. Again, we observe a mass of points with no clear negative trend. The
corresponding correlation coefficient is negative but statistically significant at
conventional levels (p < 0.01). Furthermore, examining the outliers suggests that
those state-years that see large changes in DI do not have corresponding changes
in WC receipt, and vice versa. Visual inspection of the scatterplots suggests that
changes over time in DI and WC are not happening within the same states. The
OLS regression version of these scatterplots includes state and year fixed effects
and shows that the relationship is not statistically significant. In Table 3, Panel
A shows that without the use of any covariates, there is a statistically significant
negative relationship between WC and DI. Adding state fixed effects in Panel B
yields results that are positive in sign, although including only year fixed effects
in Panel C leaves results that are qualitatively similar to Panel A results. Results
in Panel D show that when we control for state effects and year effects, none of
the estimated relationships is statistically significant. There are many reasons
why state time invariant confounders, and national time period effects, need to
be accounted for in this model, but we presented results in Panels AC to show
that the changes in results from including state and year fixed effects are not sim-
ply because of an increase in standard errors making results statistically insignifi-
cant. The coefficients themselves change substantially, while the standard errors
stay fairly similar in all panels. In Panel E of Table 3, we include our full set of
covariates and find that the estimated relationship between WC and DI is quite
sensitive to the inclusion of covariates. The coefficients on WC receipt describ-
ing DI applications are now positive, and in the regressions describing new DI
cases, they are positive and statistically significanta result inconsistent with a
story of substitution between the two programs.
In summary, although there is a concurrent rise in DI and fall in WC at the
national level, the scatterplots show that rises in DI and falls in WC do not come
from the same states. This suggests it is unlikely that the changes observed
reflect individuals causally substituting DI for WC when WC tightened. OLS
results also do not show a substitution pattern, but as mentioned earlier, these
coefficient estimates may be biased upward. For example, we cannot perfectly
control for deteriorating worker health, and this omission exerts an upward bias
on our coefficient of interest. For these reasons, we now turn to changes in DI
and WC receipt induced by changes in the maximum weekly benefit parameter.
The Effect of Permanent WC Receipt on DI Applications and Receipt. To

determine whether there is a causal relationship between the two programs, we
construct instrumental variables (IV) estimates of the effect of declining WC
enrollment on DI receipt using within-state changes to WC program generos-
ity, as measured by the real maximum weekly PPD benefit.17 The first- and
second-stage equations take the forms (3) and (4), respectively, shown below:
17
Many other papers quantify WC generosity using measures of weekly WC benefits (see, e.g., Bronch-
etti and McInerney forthcoming; Guo and Burton 2010; Butler and Worrall 1983; Hirsch, Macpherson, and
Dumond 1997; and Krueger 1990).
FIGURE 4
MAXIMUM WEEKLY PERMANENT PARTIAL DISABILITY (PPD) BENEFITS OVER TIME (2000 $)
1000
900
800
700
600
Mean
500
Iowa
400
New York
300
200 Tennessee
100
0
SOURCE: U.S. Department of Labor, Employment Standards Administration, Office of Workers Compensation Programs.
State Workers Compensation Laws.
WCs;t d0 d1 WCgenerositys;t d2 unems;t d3 share45 to 54s;t

d4 share55 to 64s;t d5 pop25 to 64s;t Ws;t D ls st e3;s;t 3
DIs;t h0 h1 WCs;t h2 unems;t h3 share45 to 54s;t

h4 share55 to 64s;t h5 pop25 to 64s;t Ws;t H ls st e4;s;t 4
where s denotes state and t denotes year. In (3), WC generosity is the real
maximum weekly PPD benefit for state s in year t, and variation in it occurs
at the unit of observation (state by year). We weight our regressions by the
state working-age population.
Within-state variation in WC generosity arises year to year as the maximum
benefit changes.18 There are two ways states legislate changes from year to
year in the maximum weekly benefits. Most states (31) legislate that the maxi-
mum benefit each year be indexed to a fixed percentage of the state average
weekly wage, as computed each year by the state Department of Labor. For
example, maximum PPD benefits in Iowa are pegged to 185% of the state
average weekly wage. In 2005, the weekly maximum benefit was $1042, and
this increased automatically to $1079 in 2006. In Figure 4, the maximum ben-
efit in Iowa, one of the more generous states, is shown with the dotted line.
Although nominal benefits increase each year in Iowa, there are periods where
annual changes in real maximum benefits decline. For example, real maximum
benefits fall between 1986 and 1991. A few other states (4) make periodic
18
We also considered quantifying changes to generosity with changes to the maximum duration of PPD
receipt. However, we document fewer instances of states changing maximum duration.
adjustments to the percentage of the state average weekly wage that comprises
the maximum. For example, in Tennessee, the drop in real weekly maximum
benefits in 1993 corresponds to the state beginning to change the percentage
of the state average weekly wage to which maximum weekly benefits are
pegged. Between January 7, 1994 and June 30, 1995, maximum PPD benefits
in Tennessee were set at 86.8% of the state average weekly wage. This per-
centage of the state average weekly wage increased to 91.2% on January 1,
1996. Most (8) of the remaining states legislate the dollar value of the maxi-
mum benefit directly. States that employ this approach increase the maximum
less frequently, leading to longer stretches of maximum benefits remaining the
same in nominal termsand greater erosion in real benefits. This is illustrated
in Figure 4 for New York, one of the least generous states that set the dollar
value of the maximum in state statutes. We observe five increases to nominal
maximum weekly PPD benefits over this time period, and the maximum bene-
fit remains fixed at $400 per week from 1993 to 2006. In real terms, maximum
benefits declined from $477 per week in 1993 to $342 per week in 2006 (in
2000 $).19 Finally, we exclude two states (Alaska and Washington) from our
analysis as they do not compensate injured workers based on weekly benefits,
awarding PPD beneficiaries one-time lump-sum payments instead. In total, six
states change the method of computing the maximum over time.
Results from the first stage (equation 3) are shown in Table 4. In each case,
we find a positive and statistically significant relationship between maximum
weekly PPD benefits and PPD WC receipt, as we would expect. We find a
$100 increase in the maximum weekly benefit yields approximately fifty new
PPD recipients (per 100,000 workers) (correspondingly, a $100 decrease in
maximum weekly benefits results in fifty fewer injured workers claiming WC).
This positive, statistically significant relationship is quite robust, and the partial
F-statistic is 24. Workers more likely to claim WC when maximum weekly
WC benefits are more generous are those workers with relatively higher
19
In addition to these changes in benefit payment amounts, several states enacted other specific reforms
to their WC systems in the 1990s in efforts to curb rising WC costs, including allowing employers, not
employees, to select the doctor who first treats an injured worker. Doctors serve as gatekeepers to the WC
system, so this reform enables employers to choose physicians who are more conservative about what they
consider a valid workplace injury. Several states began to require medical evidence to prove the existence of
a workplace injury, making it more difficult for injured workers to receive WC for injuries such as back pain
or carpal tunnel syndrome. Finally, several states began to require that the workplace injury be the major
or predominant cause of the disability. This type of reform especially impacts older workers who may have
pre-existing conditions aggravated at work. Finally, twenty-two states enacted stiff penalties for employees
found filing fraudulent claims. The existing evidence on the impact of these reforms on the WC program is
mixed (for example, Boden and Ruser [2003] found no effect). In results not shown, we explored the use of
these reforms as instruments, but found them to have weak first stage effects on WC receipt.
TABLE 4
78 /
THE EFFECT OF WORKERS COMPENSATION PROGRAM PARAMETERS ON WORKERS COMPENSATION AND SOCIAL SECURITY DISABILITY INSURANCE
OUTCOMES
First stage Reduced form Second-stage instrumental variables results

Permanent
partial cases Applications to DI New DI cases Applications to New DI cases
Dependent variable per 100,000 workers (per 100,000 workers) (per 100,000 workers) DI (per 100,000 workers) (per 100,000 workers)
(1) (2) (3) (4) (5)
MELISSA MCINERNEY
Maximum weekly PPD 0.52 (0.10)** 0.076 (0.067) 0.005 (0.033) na na

benefit (2000 $)
AND
Permanent partial na na na 0.15 (0.12) )0.06 (0.08)

cases (per 100,000
workers)
Unemployment rate )43.76 (6.64)** 51.16 (4.25)** 13.80 (2.07)** 57.63 (6.48)** 12.05 (2.92)**
Share of population Aged 27.88 (13.16)** 1.12 (8.43) 19.11 (4.10)** )3.01 (8.53) 10.04 (3.97)**
4554
KOSALI SIMON
Share of population Aged )69.80 (13.47)** )24.85 (8.64)** )8.44 (4.20)** )14.52 (11.78) )11.19 (4.04)**
5564
State industry shares
Agriculture, forestry, )9.88 (8.33) 6.60 (5.34) 0.57 (2.60) 8.06 (5.27) 4.45 (2.26)**
fisheries
Mining 33.59 (13.84)** )6.56 (8.87) 1.89 (4.32) )11.54 (9.72) )0.80 (5.06)
Construction 8.74 (6.34) )12.18 (4.06)** )3.94 (1.98)** )13.48 (4.04)** )8.21 (1.74)**
Manufacturing-durable 5.13 (4.97) )6.24 (4.06)** )1.76 (1.55) )7.00 (3.17)** )4.40 (1.38)**
goods
Manufacturing-nondurable )7.39 (5.36) )16.29 (3.44)** )6.02 (1.67)** )15.19 (3.44)** )6.13 (1.57)**
goods
Transportation, 6.48 (6.37) 2.20 (4.09) )0.75 (1.99) 1.24 (4.05) )0.96 (1.80)
communication, and other
public utilities
TABLE 4 (Cont.)

Permanent
partial cases Applications to DI New DI cases Applications to New DI cases
per 100,000 workers (per 100,000 workers) (per 100,000 workers) DI (per 100,000 workers) (per 100,000 workers)
(1) (2) (3) (4) (5)
Wholesale trade 6.09 (8.12) )8.44 (5.21) )5.02 (2.53)** )9.34 (5.08)* )4.09 (2.51)*
Retail trade )5.15 (4.97) )1.76 (3.18) )1.83 (1.55) )1.00 (3.14) )3.99 (1.40)**
Finance, insurance, and real estate 20.48 (6.62)** )4.45 (4.24) )0.64 (2.06) )7.48 (4.79) )2.74 (2.23)
Business and repair services )0.94 (6.42) )1.48 (4.11) )0.49 (2.00) )1.34 (3.98) )0.46 (1.73)
Personal services )17.52 (8.98)** )5.14 (5.76) 0.03 (2.80) )2.55 (6.04) )0.59 (2.66)
Professional services 13.31 (13.77) )16.58 (8.83)* 8.25 (4.30)* )18.55 (8.65)* 3.71 (3.94)
N 708 708 708 708 708
R2 0.904 0.944 0.853 0.942 0.855
F 24.46
Mean of dependent variable 577.93 (312.53) 1172 (397) 430 (111) 1172 (397) 430 (111)
See notes to Table 3. Alaska and Washington are excluded from the analysis because permanent partial disability (PPD) benefits were delivered in lump sum, not based on maximum
weekly benefits. Mean maximum weekly PPD benefits are $410.
pre-injury weekly earnings. Recall that WC benefits generally replace two-
thirds of pre-injury weekly wages, up to a maximum that varies across states
and over time. As we calculated that about 28% of workers are eligible for the
maximum, this effect is driven by individuals above the bottom two-thirds of
the earnings distribution. However, there is great variation across states in the
maximum weekly benefit and the fraction of workers affected by the maxi-
mum. For example, in California, we calculate that 69 percent of all workers
would be eligible for the PPD WC maximum.
Results from the reduced form are presented in columns (2) and (3), and we
now present results from our second-stage IV regression in columns (4) and
(5) to allow for easier interpretation of the coefficient estimates. No IV coeffi-
cient on permanent WC receipt is statistically significant, and the coefficient
estimates in column 4, where the dependent variable is applications to DI, are
actually positive. For example, the coefficient estimate in column 4 is 0.15
with a standard error of 0.13. This point estimate corresponds to fifteen new
DI applications for every 100 new PPD cases. From a mean of 1172 applica-
tions to DI, this corresponds to an increase of one-ten-thousandth of a percent.
However, the standard error is so large that we cannot rule out a wide range
of effects (including negative effects). In this case, effect sizes of )0.1 and 0.4
are both in the 95% confidence interval.
In column 5, we present estimates where the dependent variable is new DI
cases (per 100,000 workers). The coefficient is negative but not statistically
significant. For example, in column (5), the coefficient estimate is )0.05, sug-
gesting that for every 100 workers not receiving WC because benefits became
less generous, there are five additional people going onto DI. From a mean of
430 new DI cases, this corresponds to an increase in DI claiming of one-ten-
thousandth of a percent. As in columns 2 through 4, the standard errors are so
large that we cannot rule out much larger effectsboth )0.21 and 0.11 are in
the 95 percent confidence interval.
We interpret our findings above as the impact of changes to generosity in
WC maximum benefits on those eligible for the maximum weekly benefits.
Although these findings cannot be extrapolated to the entire working popula-
tion, we argue that the sizable (almost one-third) population eligible for the
maximum is an interesting population. Furthermore, in this paper, we seek to
identify the impact of changes in WC generosity within states on DI applica-
tions and cases. During the time period we study, changes to the DI program,
as documented by Autor and Duggan (2003), made DI relatively more gener-
ous than it had been to low earners. Such nationwide changes in program gen-
erosity would confound our results as they would show how even absent WC
changes there may be a substitution between the two programs as DI genero-
sity increased. This would be evidence of a different mechanism than the one
we set out to test in this paper. By focusing on changes to generosity that
would impact the top third of workers, we are able to avoid this possibility.
The identifying assumption that will allow us to infer a correct causal inter-
pretation is that state-set maximum weekly PPD generosity impacts WC
receipt but does not directly impact DI receipt, conditional on the observables
in our second-stage regression. This assumption would be violated if there was
a reason DI receipt was sensitive to the factors that drove WC generosity. One
concern might be that within a state, year-to-year changes in average weekly
wages may affect WC generosity and may also affect DI applications in some
unobservable way (for example, if average wages have dropped in the state
because the worker stock is sicker and thus may be more likely to apply for
DI). In many states, maximum weekly benefits formulas are pegged to the
state real average weekly wage that introduces variation that is not entirely dri-
ven by changes in state policy. This is a somewhat similar situation to what
Autor and Duggan (2003) face in quantifying changes in DI generosity, as
their variation comes from changes in expected DI replacement rates for work-
ers at the 75th percentile of that state years earnings distribution. To under-
stand whether there is cause for concern, we present results limiting the
sample to states whose PPD maximum benefits are not automatically pegged
to state average weekly wages, and also results from a model that includes all
states but controls directly for the state average weekly wage in all the equa-
tions.
Specifically, we first estimate models based just on the states that changed
the percentage of the state average weekly wage used in computing the max-
imum (e.g., Tennessee), or that did not have any automatic indexing of the
maximum weekly benefit to the state average weekly wage (e.g., New York).
Results for the key parameters of interest are reported in Panel A of Table 5.
Among this subset of states, our qualitative conclusions are the same. We
find a strong, statistically significant positive relationship between WC gener-
osity and WC receipt in our first stage. In our second stage, we find a small,
not statistically significant relationship between WC and DI receipt with large
standard errors that do not allow us to rule out large negative or positive
effects.
We can also be sure that our coefficient estimate on maximum weekly PPD
benefits is not merely reflecting changes to average weekly wages if we
include the state average weekly wage directly as an additional regressor in
equations (3) and (4). This ensures that whatever variation used to identify
effects of the maximum weekly benefit has been already conditioned on any
changes in the state average weekly wage. Again, our main qualitative conclu-
sions are upheld. We document a strong, statistically significant relationship
between WC generosity and WC receipt in the first stage. We then find a
TABLE 5
82 /
ISOLATING THE EFFECT OF WORKERS COMPENSATION PROGRAM PARAMETERS ON WORKERS COMPENSATION AND SOCIAL SECURITY DISABILITY
INSURANCE OUTCOMES FROM CHANGES IN STATE AVERAGE WEEKLY EARNINGS

Permanent
Partial Cases per Applications to New DI Cases Applications to New DI Cases
Dependent Variable 100,000 Workers DI (per 100,000 workers) (per 100,000 workers) DI (per 100,000 workers) (per 100,000 workers)
(1) (2) (3) (4) (5)
MELISSA MCINERNEY
Panel A: states that experience changes to the maximum weekly benefit level above and beyond automatic updates from growth in the state average weekly
earnings
AND
Maximum weekly PPD benefit 0.877 (0.186)** 0.134 (0.104) 0.059 (0.056)
(2000 $)
Permanent partial cases 0.153 (0.110) 0.013 (0.091)
(per 100,000 workers)
N 192 192 192 192 192
R2 0.934 0.942 0.834 0.934 0.809
F 22.23
KOSALI SIMON
Panel B: all states, include control for state average weekly earnings
Maximum weekly PPD benefit 0.461 (0.105)** 0.059 (0.068) )0.003 (0.003)
(2000 $)
Average weekly earnings 0.800 (0.279)** 0.254 (0.180) 0.104 (0.087) 0.151 (0.222) )0.026 (0.091)
(2000 $)
Permanent partial cases 0.128 (0.141) )0.057 (0.085)
(per 100,000 workers)
N 708 708 708 708 708
R2 0.905 0.945 0.854 0.943 0.855
F 19.2
See notes to Table 3. Alaska and Washington are excluded from the analysis because permanent partial disability (PPD) benefits were delivered in lump sum, not based on maximum
weekly benefits. Panel A only includes those states that either change the percentage of the state average weekly wage used to compute maximum benefits (Arkansas, Missouri,
Rhode Island, and Tennessee) or legislate the dollar value of the maximum benefit (Alabama, Arizona, California, Georgia, Indiana, New York, Oregon, and Wisconsin). The
average weekly wage included in Panel B is computed for each state and year from March Current Population Survey data.
small, not statistically significant relationship between WC and DI receipt in
the second stage. However, the standard errors are large enough that we can-
not rule out large negative effectsconsistent with substitution between the
two programsand large positive effectsconsistent with the two programs
being complements. We take the fact that these two sets of results are qualita-
tively similar to our base results to indicate that the way many states set maxi-
mum weekly benefits tied to state average weekly wages does not introduce
notable concerns for our method of identification.
Discussion and Conclusion

Although trends in national data suggest a negative relationship between
DI and WC, we examine the relationship between the two programs at the
state-year level. Scatterplots are particularly useful in examining this relation-
ship because they show that the large increases in DI were not happening in
the same state-years that experienced falls in WC. To study this more rigor-
ously, we used an IV strategy exploiting plausibly exogenous variation in
the state real maximum weekly PPD benefit. We found that WC receipt is
quite responsive to the state-set maximum benefit parameter (and produces
variation that applies only to the top third or so of the worker population for
whom this maximum is relevant). Although we find no second-stage evi-
dence of substitution between WC and DI at the state by year level, our esti-
mates do not allow us to rule out large positive or negative effects. For
example, both one new DI case for every five injured workers who do not
receive WC and one new DI case for every ten injured workers who receive
WC are in the 95 percent confidence interval of our coefficient estimate.
Although it is possible that a causal relationship between the two programs
exists that we cannot detect with our data (i.e., that at the individual level,
there could be some for whom WC tightening leads to DI applications), we
argue that our approach answers the policy question that motivated this paper:
that the decline in WC outcomes during the 1990s is not a significant factor in
the increase in DI outcomes during the same period. Even if individuals at the
margin are increasingly claiming DI more than WC, if these changes are too
small to detect in our state-by-year level data, then this behavior is unlikely to
be driving the aggregate trends that prompted the initial concern about WC
tightening leading to DI growth. Understanding whether there is a program
substitution effect at the individual level is an unexplored area of research; the
Health and Retirement Study, with its confidential-use state identifiers, is one
possible data set for this future line of work.
REFERENCES
Acemoglu, Daron, and Joshua D. Angrist. 2001. Consequences of Employment Protection? The Case of
the Americans with Disabilities Act. The Journal of Political Economy 109(5): 91557.
Autor, David, and Mark Duggan. 2003. The Rise in the Disability Rolls and the Decline in Unemploy-
ment. Quarterly Journal of Economics 118(1): 157206.
Barth, Peter. 2005. Compensating Workers for Permanent Partial Disabilities. Social Security Bulletin
65(4): 1623.
Bjelland, M. J., W. A. Erickson, and C. G. Lee 2008, November 8. Disability Statistics from the Current
Population Survey (CPS). Ithaca, NY: Cornell University Rehabilitation Research and Training Center
on Disability Demographics and Statistics (StatsRRTC). Retrieved August 31, 2009 from
http: www.disabilitystatistics.org.
Boden, Leslie I., and John W. Ruser. 2003. Workers Compensation Reforms, Choice of Medical Care
Provider, and Reported Workplace Injuries. The Review of Economics and Statistics 85(4): 92329.
Bound, John, and Richard V. Burkhauser. 1999. Economic Analysis of Transfer Programs Targeted on
People with Disabilities. In Handbook of Labor Economics, Vol. 3C, edited by Orley C. Ashenfel-
ter, and David Card, pp. 3417528. Amsterdam: Elsevier Science.
Bronchetti, Erin T., and Melissa P. McInerney. Forthcoming. Revisiting Incentive Effects in Workers
Compensation: Do Higher Benefits Really Induce More Claims? Industrial and Labor Relations
Review.
Burkhauser, Richard V., J. S. Butler, and Gulcin Gumus. 2004. A Dynamic Programming Model of Social
Security Disability Insurance Application. Journal of Applied Econometrics 19(6): 67185.
, J. S. Butler, and Robert R. Weathers II. 2002. How Policy Variables Influence the Timing of
Social Security Disability Insurance Applications. Social Security Bulletin 64(1): 5283.
, Andrew J. Houtenville, and Ludmila Rovba. 2005. Rising Poverty in the Midst of Plenty: The Case
of Working-Age People with Disabilities. Ithaca, NY: Cornell University, Rehabilitation Research and
Training Center for Economic Research on Employment Policy for People with Disabilities. Available
at http://digitalcommons.ilr.cornell.edu/edicollect/162.
Burton, John F. Jr., and Emily Speiler. 2001. Workers Compensation and Older Workers. Health and
Income Security for an Aging Workforce, Brief No. 3. Washington, DC: National Academy of Social
Insurance.
Butler, Richard F., and John D. Worrall. 1983. Workers Compensation: Benefit and Injury Claims Rates
in the Seventies. The Review of Economics and Statistics 65(4): 58089.
California Division of Workers Compensation. 2009. California Division of Workers Compensation fact
sheet D: Answers to your questions about permanent disability benefits. Available at http://
www.dir.ca.gov/DWC/iwguides.html on September 7, 2009.
DeLeire, Thomas. 2000a. The Unintended Consequences of the Americans with Disabilities Act. Regula-
tion 23(1): 2124.
. 2000b. The Wage and Employment Effects of the Americans with Disabilities Act. Journal of
Human Resources 35(4): 693715.
. 2003. The Americans with Disabilities Act and the Employment of People with Disabilities. In
The Decline in Employment of People with Disabilities: A Policy Puzzle, edited by David C. Staple-
ton, and Richard V. Burkhauser, pp. 25977. Kalamazoo, MI: W.E. Upjohn Institute.
Duggan, Mark, and Scott Imberman. 2008. Why are the DI Rolls Skyrocketing? The Contribution of Popu-
lation Characteristics, Program Changes, and Economic Conditions. In Health in Older Ages: The
Causes and Consequences of Declining Disability Among the Elderly, edited by David Cutler, and
David Wise, pp. 33779. Chicago: University of Chicago Press.
, and Melissa S. Kearney. 2007. The Impact of Child SSI Enrollment on Household Outcomes.
Journal of Policy Analysis and Management 26(4): 86186.
, Perry Singleton, and Jae Song. 2007. Aching to Retire? The Rise in the Full Retirement Age and
Its Impact on the Disability Rolls. Journal of Public Economics 91: 7.
Galizzi, Monica, Petra Miesmaa, Laura Punnett, Craig Slatin, and The Phase In Healthcare Research Team.
2010. Injured Workers Underreporting in the Health Care Industry: An Analysis Using Quantita-
tive, Qualitative, and Observational Data. Industrial Relations 49(1): 2243.
Garrett, Bowen, and Sherry Glied. 2000. Does State AFDC Generosity Affect Child SSI Participation?
Journal of Policy Analysis and Management 19(2): 27995.
Goodman, N., and T. Waidmann. 2003. Social Security Disability Insurance and the Recent Decline in the
Employment Rate of People with Disabilities. In The Decline in Employment of People with Dis-
abilities: A Policy Puzzle, edited by David C. Stapleton, and Richard V. Burkhauser, pp. 33968.
Kalamazoo, MI: W.E. Upjohn Institute.
Guo, Xuguang (Steve), and John F. Burton Jr. 2008. The Relationship Between Workers Compensation
and Disability Insurance. In Proceedings of the 60th Annual Meeting, edited by Adrienne E. Eaton,
pp. 2537. Champaign, IL: Labor and Employment Relations Association.
Guo, Xuguang (Steve), and John F. Burton Jr. 2010. Workers Compensation: Recent Developments in
Moral Hazard and Benefit Payments. Industrial and Labor Relations Review 63(2): 34055.
Hill, Steven C., Gina Livermore, and Andrew Houtenville. 2003. Rising Health Care Expenditures and the
Employment of People with High-Cost Chronic Conditions. In The Decline in Employment of Peo-
ple with Disabilities: A Policy Puzzle, edited by D. Stapleton, and R. Burkhauser, pp. 181215. Kala-
mazoo, MI: W.E. Upjohn Institute for Employment Research.
Hirsch, Barry T., David A. Macpherson, and J. Michael Dumond. 1997. Workers Compensation Recipien-
cy in Union and Nonunion Workplaces. Industrial and Labor Relations Review 50(2): 21336.
Houtenville, Andrew, and Richard B. Burkhauser. 2003. Did the Employment of People with Disabilities
Decline in the 1990s, and Was the ADA Responsible? A Replication and Robustness Check of Acemo-
glu and Angrist (2001). Research Brief. Ithaca, NY: Cornell University Employment and Disability
Institute.
, and Mary C. Daly. 2003. Employment Declines Among People with Disabilities. In The Decline
in Employment of People with Disabilities: A Policy Puzzle, edited by David Stapleton, and Richard
Burkhauser, pp. 87123. Kalamazoo, MI: Upjohn Press.
Kaye, H. Stephen. 2003. Improved Employment Opportunities for People with Disabilities. National
Institute on Disability and Rehabilitation Research. Available at http://www.dsc.ucsf.edu/pdf/
report17.pdf.
Krueger, Alan B. 1990. The Incentive Effects of Workers Compensation. Journal of Public Economics
41: 7399.
Kubik, Jeffrey. 2003. Fiscal Federalism and Welfare Policy: The Role of States in the Growth of Child
SSI. National Tax Journal 56(1): 6179.
Livermore, Gina, David Stapleton, and Henry Claypool. 2009. Health Insurance and Health Care Access
Before and After SSDI Entry: Final Report. Princeton, NJ: Mathematica Policy Research.
Lockhart James B. III. 2005. Reassessing the Relationship Between Disability Insurance and Workers
Compensation. Social Security Bulletin 65(4): 12.
Meyer, Bruce D. 2002. Unemployment and Workers Compensation Programs: Rationale, Design, Labor
Supply and Income Support. Fiscal Studies 23(1): 149.
Mont, Daniel, John F. Burton, and Virginia Reno. 2000. Workers Compensation: Benefits, Coverage, and
Costs, 19971998, New Estimates. Washington, DC: National Academy of Social Insurance.
Reville, Robert T., and Robert F. Schoeni. 2005. The Fraction of Disability Caused at Work. Social Secu-
rity Bulletin 65(4): 3137.
Roberts, Karen. 1994. The Role of Supplemental Disability Benefits in Income Security for Injured Work-
ers. Industrial Relations 33(3): 36576.
Schmidt, Lucie, and Purvi Sevak. 2004. AFDC, SSI, and Welfare Reform Aggressiveness: Caseload
Reductions vs. Caseload Shifting. Journal of Human Resources 39(3): 792812.
Sengupta, Ishita, Virginia Reno, and John F. Burton Jr. 2009. Workers Compensation: Benefits, Coverage
and Costs, 2007. Washington, DC: National Academy of Social Insurance.
Sengupta, Ishita, Virginia Reno, and John F. Burton Jr. 2010. Workers Compensation: Benefits, Coverage
and Costs, 2008. Washington, DC: National Academy of Social Insurance.
Social Security Administration. 2008. Chart 11: Final outcome of disabled worker applications, 1998
2007. Annual Statistical Report on the Social Security Disability Insurance Program, 2008. Social
Security Administration. Available at http://www.ssa.gov/policy/docs/statcomps/di_asr/2008/
sect04.html. Viewed September 7, 2009.
. 2009a. Questions? How much can I earn and still receive Disability benefits? Available at http://
ssa-custhelp.ssa.gov/cgi-bin/ssa.cfg/php/enduser/std_adp.php?p_faqid=317. Viewed September 7,
2009.
. 2009b. 511. When does your period of disability end? Social Security Handbook. Available at
http: www.ssa.gov OP_Home handbook.05 handbook-toc05.html. Viewed January 21, 2010.
Stapleton, David C., and Richard V. Burkhauser (eds). 2003. The Decline in Employment of People with
Disabilities: A Policy Puzzle. Kalamazoo, MI: W.E. Upjohn Institute for Employment Research.
Stapleton, David C., Nanette Goodman, and Andrew J. Houtenville. 2003. Have Changes in the Nature of
Work or the Labor Market Reduced Employment Prospects of Workers with Disabilities? In The
Decline in Employment of People with Disabilities: A Policy Puzzle, edited by David C. Stapleton,
and Richard V. Burkhauser, pp. 12579. Kalamazoo, MI: W.E. Upjohn Institute for Employment
Research.
Thomason, Terry, John F. Burton Jr., and Timothy P. Schmidle. 2001. Workers Compensation: Benefits,
Costs, and Safety under Alternative Insurance Arrangements. Kalamazoo, MI: W.E. Upjohn Institute
for Employment Research.
Appendix 1
Data on Workers Compensation Outcomes

Permanent Partial Disability Cases per 100,000 Workers. Data on the fre-
quency of WC receipt (PPD claims per 100,000 workers) were coded from documents gener-
ously provided by John Burton and Steve Guo, based on NCCI data.20 Data for the years 2003
and 2004 can be found in the publication, Workers Compensation Policy Review.21 Data on
these two series are available for all states which have premiums rated by the NCCI for the
years 19852004. The data series also include information for a few states and years not rated
by the NCCI (i.e., Delaware, Pennsylvania, and West Virginia).22 Note that with the exception
of West Virginia (an exclusive state fund), these data series only represent the frequency of WC
claims files with competitive, private insurers. Claims filed with self-insuring employers or state
funds are not included in these totals.
Data on DI Outcomes
Applications to DI per 100,000 Workers and New DI Cases per 100,000
Workers. State totals on the number of new applications to DI and the number of new DI
cases by state and year were generously provided by Richard Burkhauser and Dave Stapleton.
Data on applications and new cases are available from 1980 to 2001 for all states. We obtain
the total number of applications by summing the number of initial applications to the DI pro-
gram (initDItotal) and the number of initial applications to the DI and SSI programs (initconto-
tal). The total number of new cases is the sum the number of initial DI only allowances
(initDIallow) and the number of initial DI and SSI allowances (initconallow). To obtain these
series per 100,000 workers, we use total employment constructed from County Business Pat-
terns (CBP).
DI Applications or New Cases per 100; 000 workers

Total Application or Cases
100; 000
CBP total Employment
20
Data primarily are derived from NCCIs Annual Statistical Bulletin.
21
Data for 2003 are available in the September October 2007 edition of the Workers Compensation
Policy Review (Vol. 7, Issue 5), and data for 2004 are available in the July August 2008 edition of the
Workers Compensation Policy Review (Vol. 8, Issue 4).
22
Data are missing for Nevada prior to 1996 and for Delaware, Pennsylvania, Texas, and West Virginia
in certain years.
State DI Benefits (000s) per 100,000 Workers. Data on total DI benefits paid in
each state, in each year, are published in the Social Security Bulletins Annual Statistical Sup-
plement. Data are available in each state for the years 19802006 and presented in millions.23
To obtain this series per 100,000 workers, we use total employment constructed from County
Business Patterns (CBP).
Total DIBenfits 1000

DI Benefits Paid per 100k Workers 000s 100;000:
CBP Total Employment
Data on WC Program Parameters

PPD Program Parameters. Data on PPD program parameters were entered from hard
copy and microfiche editions of the BLS annual publication, State Workers Compensation
Laws. Annual data are available in all states from 1983 to 2006. In this paper, we examine max-
imum weekly PPD benefits.
23
We obtained hard copies of the publication for the years 1980 through 2001. Data from 2002 through
2006 are available online at http://www.socialsecurity.gov/policy/docs/statcomps/supplement/2008/index.html
(last viewed September 9, 2009).

The Effect of State Workers' Compensation Program Changes On The Use of Federal Social Security Disability Insurance

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Effect of State Workers' Compensation Program Changes On The Use of Federal Social Security Disability Insurance

Uploaded by

Copyright:

Available Formats

The Effect of State Workers Compensation

Program Changes on the Use of Federal Social

MELISSA MCINERNEY and KOSALI SIMON*

Social Security Disability Insurance. Disability Insurance is a federal pro-

Workers Compensation Policy Changes. In testing the causal effect of fall-

Relevant Prior Work. The main contribution of our paper is to quantify

SOURCE: Bjelland, Erickson, and Lee (2008).

state level. In Table 2, we present summary statistics for our measures of WC

Disability Insurance (DI) and WC stand for the different measures of DI

where Ws,t is a vector of the share of workers employed in different industries,

STATISTICALLY SIGNIFICANT (P < 0.001)

-1000 -500 0 500

-1000 -500 0 500

DI applications per New DI claims per

we can only interpret these results as the conditional correlation between WC

The Effect of Permanent WC Receipt on DI Applications and Receipt. To

WCs;t d0 d1 WCgenerositys;t d2 unems;t d3 share45 to 54s;t

DIs;t h0 h1 WCs;t h2 unems;t h3 share45 to 54s;t

First stage Reduced form Second-stage instrumental variables results

Maximum weekly PPD 0.52 (0.10)** 0.076 (0.067) 0.005 (0.033) na na

Permanent partial na na na 0.15 (0.12) )0.06 (0.08)

First stage Reduced form Second-stage instrumental variables results

First stage Reduced form Second-stage instrumental variables results

Discussion and Conclusion

Data on Workers Compensation Outcomes

DI Applications or New Cases per 100; 000 workers

Total DIBenfits 1000

Data on WC Program Parameters

You might also like