ec499 capstone paper
TRANSCRIPT
Michigan State University
Impact of Chapter 58 on Insurance
An Analysis of the Migration of Health Care Coverage
4/27/2012
By James Regan and Grant Cleberg
Abstract: In spring of 2006, Massachusetts enacted a health care insurance bill with the intent to expand coverage to nearly all residents. Six years after the reform significant data has been collected, which show how the reform affected Massachusetts. The increases in coverage show that the legislation is in accord with its intent and the migration of coverage post reform shows that the problems of crowding-out are minimal. Parallels between Massachusetts’ health reform and the Federal reform exist and could shed light on how the PPACA could affect the nation.
Introduction
In early 2006, Massachusetts enacted a bill known as Chapter 58. The purpose of the bill
was to alleviate disparities in coverage and make health care insurance more available to all
income groups. Chapter 58 was drafted in response to a threat of losing federal funding, if the
number of uninsured residents using the states’ Free Care Pool was not reduced. The legislation
included a mandate for individuals to obtain a minimum level of health care insurance,
subsidies for people below 300% of the Federal Poverty Line (FPL) and market reforms in the
way insurance is provided. Having legislation designed with these features puts an emphasis on
individual responsibility, which is thought to be effective in the current health care delivery
system.
With the enactment of Chapter 58, new incentives have changed how people obtain
their coverage. These new incentives brought up questions about how Chapter 58 would alter
the market for insurance in Massachusetts. Six years after the reform, our research has been
influenced by these questions, but more specifically, “How have people changed their health
insurance coverage in Massachusetts since the enactment of Chapter 58?”
In answering this question, we will be able to explain how the incentives influenced by
Massachusetts' health reform affected the demand to obtain coverage. Three incentives that
we expect to have the largest impacts on coverage are, the individual mandates, the expansion
of Medicaid and the use of government subsidies. We believe the individual mandates will
pressure residents to obtain coverage by removing some of the opportunity cost of not
purchasing insurance through a tax inclusion, hence the demand for coverage will increase. The
expansion of Medicaid will provide lower income groups with free health insurance, so we have
concern about how much the government will crowd-out insurance in the private sector. The
subsidies for income groups that do not qualify for Medicaid will ease the cost of purchasing
insurance, making the market for insurance more enticing. We will be using these incentives to
help explain procurement and migration of coverage.
Background
There have been a number of papers written to estimate and interpret the effects of
Chapter 58. The Long 2008 paper studied the pre-post comparisons directly after the reform
using data from survey by International Communications Research or ICR. The study found that
the uninsured rate dropped from 13 percent to 7 percent. The Long 2008 paper had several
limitations such as small sample sizes and not controlling for factors unrelated to the health
reform. In addition, the estimates were from 2007 and the reform was not fully implemented
until fall of 2007 (Long, 2008).
The Long et. al. 2009 paper expanded on the 2008 paper by using new data from the
Current Population Survey’s (CPS) Annual Social and Economic Supplement (ASEC) and a
stronger research design (Long et. al. 2009). The study used difference-in-difference (DD) and
multivariate regression methods for the years 2004 to 2007 to estimate the impacts of the
reform. The findings of the pre-post and DD estimates were fairly consistent to the Long 2008
findings (Long et. al., 2009). Similar limitations existed in the 2009 paper, but the regressions
helped dissect the effects of the reform in Massachusetts.
Our research is an expansion of the two Long papers, where we further estimate the
impact on coverage from the health reform in Massachusetts. We used additional data sets
from the years 2004 to 2007, excluding 2006, to better estimate the long term effects from the
reform. Since the subsidies are provided based on what income group you fit into, we divided
people into three income groups, up to 150 percent of FPL, 150 to 300 percent of FPL, and
greater than 300 percent of FPL. The Long papers divided people only into two income groups,
up to 300 percent of FPL and greater than 300 percent of FPL. People in the lowest income
group are eligible for free health insurance provided by Commonwealth Care if they are
ineligible for Mass Health (Medicaid), people in the 150 to 300 percent income group are
eligible for subsidies on a sliding scale relative to their income and the high income group does
not receive any additional subsidies from Chapter 58.
We also examined more types of coverage than the Long papers. We have included
Medicaid, directly purchased insurance and employer sponsored insurance. We further
explained the migration in both directly purchased and employer sponsored coverage by
separating them by policy owners and dependents. Since we took dependent coverage into
consideration, we included minors in our analysis to interpret reform impacts on their
coverage.
Study Design
For our research, we utilized the difference-in-difference experiment design to compare
health insurance coverage in Massachusetts before and after the state enacted Chapter 58. In
our experiment, Massachusetts is the treatment group and the states Pennsylvania and New
York are the control groups. Unlike Massachusetts, these control states did not implement any
significant health reform policy. From the experiment, we created a DD model to analyze the
data collected from the surveys.
The reasoning behind using this DD model is that there may be other effects that are
unrelated to the Chapter 58 health reform that may have affected health insurance coverage in
Massachusetts and the control states. For example, cost of living may have increased or
decreased in these states, which could have potentially transferred into decreased or increased
health coverage. Likewise, a state economic boom or recession could have contributed to
similar effects. The DD model helps control for these other factors by incorporating an
interaction term into the model; in our case it was Massachusetts, a dummy variable, and the
post reform year, another dummy variable. Since this interaction term takes on zero if it is
either “not” Massachusetts, the post reform year, or both, and takes on one if it “is”
Massachusetts and post reform year, its coefficient captures the effect of Chapter 58 post
reform in Massachusetts relative to the control states.
With this DD experiment, we estimated the effects on health coverage for children (ages
0-18) and non elderly adults (ages 19-65) within different income groups. The three income
groups correspond to the different levels of health insurance subsidies provided to individuals
and families.
Data
We used data sets from the 2005-2011 CPS ASEC. These surveys are better known as the
March Supplement because they are conducted in March. They collect “detailed information on
income and health insurance coverage” of the previous year (Long et. al., 2009). Thus, the data
represents the years 2004-2010 excluding 2006. Our choice of years is based on pre and post
reform years. Chapter 58 was enacted in April 2006 and the amendments to the original bill
spanned till November 2007. Our data from years 2004, 2005 capture the economic
environment directly prior to the enactment of Chapter 58. The data from years 2007, 2008,
2009 and 2010 depict the effects of the reform.
The states we used for controls are New York and Pennsylvania since they are in the
same region as Massachusetts. With these control states we are able to show the change in
health insurance with the reform in Massachusetts and adjust for the economic environmental
changes that affected the New England region.
We have used age to separate minors and non-elderly adults into two groups. The first
group being ages 0 to 18 and the second group being ages 19 to 65. The ages for non-elderly
adults are 19 to 65 because the data collected in the supplements are from the previous year.
Our data is divided between three different income levels based on percentages of the
federal poverty line(FPL). To calculate the percentage of the FPL we used the reported
individuals’ income level and divided it by the corresponding FPL in relation to their family size;
the three income groups, less than 150 percent of FPL, 150 to 300 percent of FPL, and greater
than 300 percent of FPL depict the differences in subsidies.
We estimate coverage in four different groups, any type of insurance(total insurance),
Medicaid, direct purchased, and employer sponsored(ESI) coverage. We further divided ESI and
direct purchased coverage into, direct purchased policy owner and dependent, employer
sponsored policy owner and dependent. With these variables we are able to see the migration
in coverage after the reform.
Method
We demonstrate the effects of the Massachusetts reform by using linear regression
models based on a difference-in-difference model. To estimate the impact of Chapter 58, we
ran 148 DD regressions. In the regression we used the dependent variables that covered
Insurance (any coverage), Medicaid, employer sponsored (total, policy owner & policy
dependent), and directly purchased (total, policy owner & policy dependent). The independent
variables in the regression were Massachusetts, and three different post reform variables. The
Massachusetts variable shows the cross sectional differences between Massachusetts and the
control states. We used three different post reform variables to illustrate how the reform
affected coverage over time. Using multiple post variables helps show how stable the effects of
the reform were; post1 helps us understand the direct affect of the reform and post2 shows if
the effects of the reform were consistent over time. These post variables were defined as post,
post1 and post2. Post was the four years after the reform (2007-2010), post1 was the years
2007 and 2008 and post2 was years 2009 and 2010. The post reform coefficients show the cross
sectional effects between pre- and post reform coverage. We used the interaction between
Massachusetts and the post reform variables to show the effects of Chapter 58 in
Massachusetts, while controlling for other factors that could have affected the New England
region; the focus of our research has been on the interaction coefficient. An example of our
regression model looks like this:
Yinsurance=β0 + β1MA + β2post + β3(MA*post) + ε
where Yinsurance, MA and post are dummy variables and (MA*post) is the interaction term that
conveys the DD estimate1. Since the dependent variables are dummies, this is a linear
probability regression model, where each coefficient estimates the probability of having a type
of insurance based on the variable.
Findings
Table 1 shows the DD estimates of the changes in insurance coverage rates from 2004-
2010, four years since the implementation of Chapter 582. Total insurance coverage increased
by 2.07 percent for minors and by 12.95 percent for non-elderly adults who were in the low
income group. The mid income group had increases of a similar fashion by 4.14 percent for
minors and 16.29 percent for non-elderly adults in total insurance coverage. The increases for
the high income group were not quite as large as the groups below them; they had increases of
only 1.08 percent for minors and 4.29 percent for non-elderly adults in total insurance. These
estimates were all statistically significant at the one percent level for non-elderly adults across
the three income levels, but only the mid income level estimate was statistically significant for
minors3. The increases in total insurance were mainly due to increases in public coverage
(Medicaid) and employer sponsored coverage (ESI). Changes in direct purchased played a
minimal role in the increase of insurance coverage; there were many decreases in this type of
coverage and the increases that were present were very small compared to the increases from
the other types of coverage. Likewise, the direct purchased estimates were all statistically
insignificant. This result shows that the majority of people migrated from direct purchased
1 See Appendix C for proof of what β3 conveys.2 Since Chapter 58 was enacted in 2006, it was excluded from our regressions.3 See Appendix A for a complete table of estimates from the regression.
coverage to Medicaid and ESI coverage. As expected, Medicaid played a significant role in the
increase of total insurance coverage in the low income group with an increase of 10.21 percent
for non-elderly adults and 2.93 percent for minors. ESI was the larger contributor to increases
in health coverage for the mid and high income groups. As expected, the increases for minors
from ESI were tied to its dependent component and the policy owner’s component for non-
elderly adults. All of the estimates for ESI and Medicaid were statistically significant at the one
percent level for non-elderly adults, however only one estimate in the mid income group of ESI
was statistically significant. Our results were consistent with the results of the two Long
papers4.
This array of results suggest that Chapter 58 had a larger impact on the non-elderly
adult population; the minor groups did see increases, but most of these estimates were not
statistically significant and were only a fraction of the estimates for non elderly adults. It makes
sense that minors were not as affected by the reform because many of them already had
insurance prior to the reform; approximately two/thirds5. Likewise, the results illustrate that
the reform had more impact on the low and mid income level groups. This makes sense
because many of the high income level individuals already had insurance prior to the reform
and do not qualify for the expansion of Medicaid or any insurance premium subsidies. The pre-
and post means for Massachusetts illustrate the same picture as the DD estimates that the
increases in health insurance coverage were a result of large increases in both Medicaid and ESI
coverage. They show that Medicaid was the large contributor to increases in health coverage
for the low income group and ESI was most prominent in the increases for the mid and high
4 See Appendix D for a side-by-side comparison of our results with the two Long papers results.5 See MassHealth Waiver: 2009-2011....and Beyond for the chart that depicts this result.
income groups. The means for direct purchased coverage in Massachusetts were relatively
small compared to the other means and showed moderate decreases for minors of the low and
mid income groups and very small increases for high income minors and non-elderly adults6.
We also split the four post years into two different post periods and ran two other sets
of regressions. Again, the reason for doing this is to look at Chapter 58’s stability; to see if its
effects were more prominent in the first two post years or in the second two post years. Table 2
shows the DD estimates of the change in insurance coverage for the first post period, from
2004-2008 and table 3 shows the estimates for the second post period, 2004-2005 and 2009-
20107. In the first post period there was a larger impact on insurance coverage for the minor
population than in the second post period, which was most contributed by the increase in
Medicaid for the low income group and the dependent component of ESI for the mid and high
income groups. These estimates were statistically significant at the one percent level. This
result is consistent with the fact that many children already had some type of health insurance
prior to the reform, so the changes in the second post period should be less than in first post
period. There were many decreases in direct purchased coverage in the first post period for the
minor population and only small increases in the low and mid income levels for non-elderly
adults. This reflects the recession of 2008, where many people could not afford to have health
insurance. There was a larger impact on coverage for non-elderly adults in the second post
period than in the first post period, which was contributed by increases in Medicaid for the low
income group and the policy owner component of ESI coverage. Again, these estimates were
statistically significant at the one percent level. We saw some increases in direct purchased for
6 See Appendix B for complete table of means.7 See Appendix A for complete tables of estimates for first and second post period regressions.
non-elderly adults, which were tied to the policy owner component of direct purchased. This is
what we should expect since the second post period is after the recession, so people were
becoming able to afford health insurance once again. However, these increases were relatively
small compared to the increases from Medicaid and ESI coverage.
In sum, these two post period regressions illustrate the big picture that was seen from
the original post regression. They show that since the reform was passed, the majority of the
impact on insurance coverage was due to increases in Medicaid and ESI coverage and that the
minority of the impact was due to direct purchased coverage. The two post period regressions
and the original regression suggest that many people migrated from direct purchased to either
Medicaid or ESI coverage after the reform was passed, depicting that there may have been
some crowding out into Medicaid from direct purchased†. On the other hand, these regressions
also suggest that since the reform was passed, there has been more of a "crowd-in" effect from
the large increases in ESI coverage(Gruber, 2011). It is important to note that the estimates on
ESI coverage for Massachusetts are intensified by decreases in ESI coverage in the control
states. That is, the decreases in ESI coverage in the control states increased the gap between
Massachusetts' and the control state's coverage, which translated into higher estimates for
Massachusetts. Hence, the increases in insurance coverage that are a result from ESI coverage
may not be as large as they appear to be. The decreases in ESI coverage in the control states
was most likely the result of increasing costs for employers to offer insurance, which translated
into less employers offering insurance or employees that could not afford the inflated price that
the employer was asking for health insurance. Perhaps if we used different New England
control states, our estimates would not have been so large, but nonetheless ESI was still a
major contributor to increases in insurance coverage in Massachusetts.
Discussion
This analysis was an extension of the Long et. al., 2009 paper that estimated the impact
of Chapter 58 pre- and post reform using data from years 2004-2007 (Long et. al., 2009),
however we used more years of data and went into more detail. To see how stable the effects
of Chapter 58 were in the post years, we split our post period into three different post periods
and ran regressions using each of these periods in our model. Our findings indicate that since
Chapter 58 was implemented, there has been an increase in the insured rate in Massachusetts
for both the minor and non-elderly populations. There were increases across all income groups;
however the larger effects of the reform were prominent in the low and mid income groups.
This is the result that we should expect from Chapter 58 since people of the high income group
are not eligible for any of the subsidies that are outreached in the legislature and because there
is a good chance that these people already had insurance prior to the reform. The main factors
that drove the increases in insurance coverage in Massachusetts were the increases from
Medicaid and ESI coverage; the increases in direct purchased coverage were very minimal in
the non-elderly population and there were decreases in direct purchased coverage for the
minor population across all income groups. The increases in Medicaid were strongest in the low
income group and increases in ESI coverage were larger in the mid and high income groups, so
perhaps there was some crowding out; at least for the low income group. This is not surprising
because from the expansion of Medicaid up to 150 percent of the federal poverty line, you
have people who were just outside eligibility for Medicaid now qualifying for it. The more
interesting result was from the large increases in ESI coverage in the non-elderly adults. In a
paper written by Jonathan Gruber, there were similar results of large increases in ESI coverage.
Gruber wrote that this "crowd-in" to ESI coverage can be explained by increases in demand for
group insurance by employees of a firm to meet the requirements of the mandate and
employers meeting this increased demand(Gruber, 2011).
Limitations of the Study
There are limitations in our study stemming from our data. Since the CPS is a self
reported survey, inaccuracies may exist. Under or over-reporting can be caused by errors in
memory, social norms and worst, deception, which a Cato Institute publication elaborated on
(Yelowitz et. al. 2010). The individual mandate in Chapter 58 can influence the way people react
to a governmental surveyor asking if they own insurance; there exists the possibility that
respondents would not want to report that they are in violation of the mandate(Yelowitz et. al.,
2010). Some individuals do not answer the specific questions in the CPS. When this happens the
Census Bureau will impute an answer based on other respondents' answers (Davern et.
al.,2007), which can further misrepresent the data collected from the surveys.
Unlike the Long et. al. 2009 paper, we did not adjust for “age, race/ethnicity, sex,
citizenship, educational attainment, marital status, health status, employment, [or] residence in
an urban area” (Long et. al., 2009). We also did not use propensity score weights included in
the CPS. This would change our results to better represent the population.
The health reform for our entire nation, the Affordable Care Act(ACA), was recently
passed in 2010. The ACA is modeled after Massachusetts' health reform, so parallels can be
drawn from Chapter 58 to predict what will be the impact on insurance coverage from the ACA.
However, the analysis of the Massachusetts experience from its health reform is very limited
when trying to estimate what will be the impact of the ACA on the nation's insurance coverage
rates because the two reforms are not exactly the same. For example, Chapter 58 expands
Medicaid coverage to those earning up to 150 percent of the federal poverty line(FPL) and the
ACA only expands Medicaid coverage to those earning up to 133 percent of the FPL. Chapter 58
partially subsidizes health care insurance for those earning up to 300 percent of the FPL and the
ACA provides these type of subsidies for those earning up to 400 percent of the FPL and these
subsidies are only available for people who are not offered insurance by their employer. You
could possibly control for the expansion differences by designating your income groups in
accordance to the ACA expansions, but there is still the issue of what the employers in each
state are going to do; they may decide to drop coverage and pay the small penalty for not
offering insurance to their employees or they may decide to offer more health coverage. There
have been economists like Jonathan Gruber, who have estimated the impact of the ACA on the
nation. However, the models that he used were very complicated and only a simulation of what
the impact of ACA could be(Gruber, 2011)8. We need to remember that the ACA has not been
fully implemented and most of the expansions have not taken effect. In 2014, the individual
mandate and the expansions of Medicaid and subsidies take effect. Perhaps in a few years,
when these expansions take place, the effects of the ACA will become more apparent and data
is available, economists can more accurately predict the impact of the ACA.
References
8See Gruber,Jonathan.2011. "The Impacts of the Affordable Care Act: How Reasonable Are The Projections?" for an explanation of the simulation models used.
Anthony, S., Seifert, R. W., & Sullivan, J. C. (2009). The MassHealth waiver: 2009-2011..and
beyond. Center for Health Law and Economics, University of Massachusetts Medical
School,
Davern, M., Rodin, H., Blewett, L. A., & Call, K. T. (2007). Are the Current Population Survey
uninsurance estimates too high? An examination of the imputation process. Health
Services Research, 42(5), 2038–2055.
Gruber, J. (2011). The impacts of the affordable care act: How reasonable are the
projections?.National Bureau of Economic Research Working Paper Series, (17168),
Retrieved from http://www.nber.org/papers/w17168
Long, S. K. (2008). On the road to universal coverage: Impacts of reform in Massachusetts at
one year. Health Affairs, 27(4).
Long, S. K., Stockley, K., & Yemane, A. (2009). Another look at the impacts of health reform in
Massachusetts: Evidence using new data and a stronger model. American Economic
Review: Papers & Proceedings, 99(2), 508–511.
Yelowitz, A., & Cannon, M. F. (2010). The Massachusetts health plan much pain, little gain. Cato
Institute's POLICY ANALYSIS SERIES, 653,
Appendix ABelow are the complete tables of regressions corresponding to the three differently defined post years.
Table 1: Coverage Rate of Change in MA (2004, 2005, 2007, 2008, 2009, 2010)Percent of FPL <=150% 150%< x <=300% >300%Insurance Age 0-18 2.07 4.14*** 1.08
(n= 38297) (0.02) (1.55) (0.66)19-65 12.95*** 16.29*** 4.29***
(n= 76686) (2.20) (1.90) (0.65)Medicaid 0-18 2.93 -0.81 -1.32
(n= 38297) (2.96) (2.97) (0.93)19-65 10.21*** 5.96*** 1.46***
(n= 76686) (2.62) (1.84) (0.53)Direct Purchased0-18 -2.76 -2.72 -0.06
(n= 38297) (2.00) (2.05) (0.92)19-65 1.94 1.63 -0.24
(n= 76686) (1.51) (1.27) (0.53)Employer Sponsored0-18 3.30 9.29*** 1.96
(n= 38297) (2.55) (3.15) (1.26)19-65 4.48** 11.53*** 3.77***
(n= 76686) (2.18) (2.42) (0.92)Direct Purchased Policy Owner0-18 -0.02 0.28 0.04
(n= 38297) (0.08) (0.21) (0.08)19-65 2.13** 2.29*** -0.61
(n= 76686) (0.97) (0.87) (0.40)Direct Purchased Dependent0-18 -2.74 -3.00 -0.11
(n= 38297) (2.00) (2.04) (0.92)19-65 -0.19 -0.66 0.37
(n= 76686) (1.20) (0.96) (0.37)Employer Sponsored Policy Owner0-18 0.29 0.37 -0.07
(n= 38297) (0.30) (0.49) (0.26)19-65 2.93 9.62*** 2.12
(n= 76686) (1.91) (2.40) (1.36)Employer Sponsored Dependent0-18 3.01 8.93*** 2.04
(n= 38297) (2.54) (3.16) (1.28)19-65 1.55 1.91 1.65
(n= 76686) (1.28) (1.86) (1.27)*** Significant at the 1 percent level ** Significant at the 5 percent level * Significant at the 10 percent level
Table 2: Coverage Rate of Change in MA (2004, 2005, 2007, 2008)
Percent of FPL<=150% 150%< x <=300% >300%
Insurance Age 0-18 3.22* 4.93*** 0.65
(n= 26836) (1.96) (1.64) (0.77)19-65 11.13*** 15.43*** 3.54***
(n= 52529) (2.50) (2.11) (0.73)Medicaid 0-18 7.43** -0.93 -2.48**
(n= 26836) (3.34) (3.48) (1.04)19-65 9.73*** 6.34*** 0.89
(n= 52529) (3.09) (2.23) (0.61)Direct Purchased0-18 -2.61 -4.34* -1.50
(n= 26836) (2.30) (2.27) (1.01)19-65 1.08 1.22 -0.72
(n= 52529) (1.78) (1.49) (0.60)Employer Sponsored0-18 -2.55 11.11*** 3.77***
(n= 26836) (2.80) (3.64) (1.41)19-65 1.69 10.88*** 3.60***
(n= 52529) (2.55) (2.83) (1.04)Direct Purchased Policy Owner0-18 -0.11 0.27 0.06
(n= 26836) (0.11) (0.29) (0.13)19-65 0.80 1.92* -0.60
(n= 52529) (1.13) (1.06) (0.45)Direct Purchased Dependent0-18 -2.50 -4.62** -1.56
(n= 26836) (2.30) (2.25) (1.01)19-65 0.28 -0.71 -0.12
(n= 52529) (1.43) (1.09) (0.41)Employer Sponsored Policy Owner0-18 0.31 0.21 -0.14
(n= 26836) (0.31) (0.62) (0.29)19-65 1.60 10.50*** 1.85
(n= 52529) (2.26) (2.84) (1.56)Employer Sponsored Dependent0-18 -2.86 10.90*** 3.91***
(n= 26836) (2.79) (3.66) (1.44)19-65 0.09 0.38 1.75
(n= 52529) (1.44) (2.16) (1.46)*** Significant at the 1 percent level ** Significant at the 5 percent level * Significant at the 10 percent level
Table 3: Coverage Rate of Change in MA (2004, 2005, 2009, 2010)
Percent of FPL<=150% 150%< x <=300% >300%
Insurance
Age
0-18 0.91 3.43* 1.55**
(n= 25962) (1.95) (1.79) (0.75)
19-65 14.60*** 17.19*** 5.09***
(n= 52092) (2.42) (2.11) (0.74)
Medicaid
0-18 -1.56 -1.18 -0.13
(n= 25962) (3.34) (3.53) (1.18)
19-65 10.54*** 5.53** 2.10***
(n= 52092) (2.96) (2.20) (0.67)
Direct Purchased
0-18 -2.81 -1.25 1.45
(n= 25962) (2.18) (2.43) (1.12)
19-65 2.70 2.01 0.28
(n= 52092) (1.72) (1.49) (0.64)
Employer Sponsored
0-18 8.88*** 8.12** 0.06
(n= 25962) (2.96) (3.71) (1.52)
19-65 6.96*** 12.30*** 3.93***
(n= 52092) (2.47) (2.80) (1.09)Direct Purchased Policy Owner
0-18 0.07 0.29 0.03
(n= 25962) (0.07) (0.28) (0.07)
19-65 3.27*** 2.66*** -0.63
(n= 52092) (1.17) (1.03) (0.46)Direct Purchased Dependent
0-18 -2.88 -1.54 1.42
(n= 25962) (2.18) (2.42) (1.12)
19-65 -0.57 -0.64 0.91**
(n= 52092) (1.32) (1.11) (0.46)Employer Sponsored Policy Owner
0-18 0.27 0.55 0.00
(n= 25962) (0.31) (0.58) (0.31)
19-65 4.18* 8.97*** 2.41
(n= 52092) (2.14) (2.76) (1.59)Employer Sponsored Dependent
0-18 8.61*** 7.57** 0.07
(n= 25962) (2.95) (3.72) (1.55)
19-65 2.78* 3.33 1.52
(n= 52092) (1.49) (2.17) (1.49)
*** Significant at the 1 percent level
** Significant at the 5 percent level * Significant at the 10 percent level
Appendix BBelow are the tables of means from the regressions corresponding to the different post years.
Table 4: Average Coverage in Massachusetts (2004, 2005, 2007, 2008, 2009, 2010)Type of Coverage Insurance
Age 0-18 19-65(Sample) (n= 38297) (n= 76686)
Pre
ReformPost
ReformPre
ReformPost
ReformLocation MA Control MA Control MA Control MA ControlPercent of FPL <=150% 0.9206 0.8684 0.9414 0.8685 .7472 0.6957 0.8678 0.6868150%< x <=300% 0.9256 0.9097 0.9704 0.9132 0.7651 0.7932 0.9027 0.7678>300% 0.9724 0.9685 0.9800 0.9653 0.9316 0.9238 0.9644 0.9136 Direct Purchased 0-18 19-65 (n= 38297 ) (n= 76686)
Pre
ReformPost
ReformPre
ReformPost
Reform MA Control MA Control MA Control MA Control <=150% 0.1266 0.0879 0.0869 0.0758 0.0848 0.0910 0.0963 0.0831150%< x <=300% 0.1335 0.0882 0.0981 0.0800 0.0696 0.0796 0.0765 0.0703>300% 0.0553 0.0507 0.0506 0.0467 0.0399 0.0387 0.0405 0.0417 MedicaidAge 0-18 19-65(Sample) (n= 38297 ) (n= 76686)
Pre
ReformPost
ReformPre
ReformPost
Reform MA Control MA Control MA Control MA Control <=150% 0.6625 0.6087 0.7600 0.6769 0.4608 0.3721 0.5971 0.4063150%< x <=300% 0.2954 0.2509 0.3414 0.3050 0.1496 0.1076 0.2415 0.1398>300% 0.0486 0.0593 0.0555 0.0795 0.0336 0.0306 0.0543 0.0366 Employer Sponsored 0-18 19-65 (n= 38297) (n= 76686)
Pre
ReformPost
ReformPre
ReformPost
Reform MA Control MA Control MA Control MA Control <=150% 0.2035 0.2371 0.1781 0.1787 0.2144 0.2440 0.2110 0.1958150%< x <=300% 0.5711 0.6385 0.6317 0.6061 0.5459 0.6215 0.6077 0.5680>300% 0.8945 0.8890 0.9036 0.8786 0.8660 0.8607 0.8841 0.8411
Table 5: Average Coverage in Massachusetts (2004, 2005, 2007, 2008)Type of Coverage Insurance Age 0-18 19-65(Sample) (n= 26836) (n= 52529)
Pre
ReformPost
ReformPre
ReformPost
ReformLocation MA Control MA Control MA Control MA Control
Percent of FPL <=150% 0.9206 0.8684 0.9437 0.8593 0.7472 0.6957 0.8649 0.7020
150%< x <=300% 0.9256 0.9097 0.9806 0.9154 0.7651 0.7932 0.9067 0.7805
>300% 0.9724 0.9685 0.9772 0.9668 0.9316 0.9238 0.9627 0.9195 Direct Purchased 0-18 19-65 (n= 26836) (n= 52529)
Pre
ReformPost
ReformPre
ReformPost
Reform MA Control MA Control MA Control MA Control <=150% 0.1266 0.0879 0.0968 0.0843 0.0848 0.0910 0.0930 0.0883
150%< x <=300% 0.1335 0.0882 0.0776 0.0758 0.0696 0.0796 0.0731 0.0710
>300% 0.0553 0.0507 0.0390 0.0494 0.0399 0.0387 0.0365 0.0425 MedicaidAge 0-18 19-65(Sample) (n= 26836) (n= 52529)
Pre
ReformPost
ReformPre
ReformPost
Reform MA Control MA Control MA Control MA Control <=150% 0.6625 0.6087 0.7748 0.6467 0.4608 0.3721 0.5860 0.3999
150%< x <=300% 0.2954 0.2509 0.3102 0.2750 0.1496 0.1076 0.2348 0.1294
>300% 0.0486 0.0593 0.0409 0.0765 0.0336 0.0306 0.0473 0.0354 Employer Sponsored 0-18 19-65 (n= 26836) (n= 52529)
Pre
ReformPost
ReformPre
ReformPost
Reform MA Control MA Control MA Control MA Control <=150% 0.2035 0.2371 0.1329 0.1920 0.2144 0.2440 0.2000 0.2127
150%< x <=300% 0.5711 0.6385 0.6842 0.6404 0.5459 0.6215 0.6221 0.5888
>300% 0.8945 0.8890 0.9201 0.8769 0.8660 0.8607 0.8890 0.8477
Table 6: Average Coverage in Massachusetts (2004, 2005, 2009, 2010)Type of Coverage Insurance Age 0-18 19-65(Sample) (n= 25962) (n= 52092)
Pre
ReformPost
ReformPre
ReformPost
ReformLocation MA Control MA Control MA Control MA ControlPercent of FPL
<=150% 0.9206 0.8684 0.9437 0.8779 0.7472 0.6957 0.8649 0.6728
150%< x <=300% 0.9256 0.9097 0.9806 0.9106 0.7651 0.7932 0.9067 0.7552
>300% 0.9724 0.9685 0.9772 0.9637 0.9316 0.9238 0.9627 0.9074 Direct Purchased 0-18 19-65 (n= 25962) (n= 52092)
Pre
ReformPost
ReformPre
ReformPost
Reform MA Control MA Control MA Control MA Control <=150% 0.1266 0.0879 0.0968 0.0670 0.0848 0.0910 0.0930 0.0783
150%< x <=300% 0.1335 0.0882 0.0776 0.0848 0.0696 0.0796 0.0731 0.0696
>300% 0.0553 0.0507 0.0390 0.0436 0.0399 0.0387 0.0365 0.0409 MedicaidAge 0-18 19-65(Sample) (n= 25962) (n= 52092)
Pre
ReformPost
ReformPre
ReformPost
Reform MA Control MA Control MA Control MA Control <=150% 0.6625 0.6087 0.7748 0.7081 0.4608 0.3721 0.5860 0.4123
150%< x <=300% 0.2954 0.2509 0.3102 0.3381 0.1496 0.1076 0.2348 0.1502
>300% 0.0486 0.0593 0.0409 0.0827 0.0336 0.0306 0.0473 0.0380 Employer Sponsored 0-18 19-65 (n= 25962) (n= 52092)
Pre
ReformPost
ReformPre
ReformPost
Reform MA Control MA Control MA Control MA Control <=150% 0.2035 0.2371 0.1329 0.1649 0.2144 0.2440 0.2000 0.1801
150%< x <=300% 0.5711 0.6385 0.6842 0.5684 0.5459 0.6215 0.6221 0.5472
>300% 0.8945 0.8890 0.9201 0.8804 0.8660 0.8607 0.8890 0.8340
Appendix C
This is the proof of the information that is conveyed from the coefficient of the interaction term. In our difference-in-difference(DD) model, we are trying to estimate if there were changes in coverage in Massachusetts due to the enactment of the reform, while controlling for other factors that could have contributed to the effects on coverage. Our model was,
Yinsurance=β0+β1MA+β2Post+β3(MA*Post)+ε,
where Yinsurance is a dummy variable taking on one if the person had insurance, MA is a dummy taking a value of one if in Massachusetts and zero for control states, and Post is another dummy variable taking a value of one if in post and zero in the pre reform period. Let's look at the different combinations of values for the variables.
MA=0, Post=0, MA*Post=0 → β0 (control states, pre reform)
MA=1, Post=0, MA*Post=0 → β0 + β1 (Massachusetts, pre reform)
MA=0, Post=1, MA*Post=0 → β0 + β2 (control states, post reform)
MA=1, Post=1, MA*Post=1 → β0 + β1 + β2 + β3 (Massachusetts, post reform)
The DD estimate is derived from subtracting the Massachusetts means from the control states means:
[(β0 + β1 + β2 + β3)-( β0 + β1)] - [( β0 + β2)- β0]=(β2+β3) - β2 = β3
From this you can see that (β3) is the DD estimate for Massachusetts. This concludes the proof.
Appendix D
This table is a side-by-side comparison of our results with the Long papers.
Table 7: Estimates of the Impacts of Health Reform on Insurance Coverage, by Income
Long et. al. 2008
Pre-post adjusted estimates
Long et. al. 2009
Pre-post adjusted estimates
Cleberg Regan
Pre-post estimates
Long et. al. 2009
DD adjusted estimates
Cleberg Regan
DD estimatesAll adults
Uninsured -5.6 -6.5 -6.6 −6.6*** -8.4***ESI coverage 2.9 1.7 1.8 3.1* 6.2***
Higher-income adults
Uninsured -1.8 -0.2 -3.2 −0.2 -4.29***ESI coverage 0.9 0.8 1.8 0.4 3.77***
Lower-income adults
Uninsured -10.5 -17.1 -12.9 −17.3*** -14.8*** ESI coverage 4.9 3.9 2.0 5.6* 7.8****** Significant at the 1 percent level ** Significant at the 5 percent level * Significant at the 10 percent level