Draft

MEASURING THE IMPACT OF VALUING HEALTH INSURANCE ON LEVELS AND TRENDS IN INEQUALITY AND HOW THE AFFORDABLE

CARE ACT OF 2010 COULD AFFECT THEM

RICHARD V. BURKHAUSER, JEFF LARRIMORE and KOSALI SIMON∗

A substantial part of the U.S. inequality literature focuses on yearly levels and trends in pre-tax, post-transfer cash income and its distribution over time and finds that median income appears to be stagnating, with income growth primarily coming at higher income levels. When we use data from the Current Population Survey for 1995 – 2008 and add the value of employer- and government-provided health insurance coverage, not only does it increase the upward trend in the level of resources controlled by Americans, but also reduces the level of inequality in these resources and its upward trend. We then provide a highly stylized example of this broader income measure’s value in capturing the impact of two key provisions of the Affordable Care Act of 2010 — an expansion in Medicaid and the provision of subsidies to lower-income families for purchasing private coverage on state-run exchanges. Even though these incremental expansions build on existing systems of government-provided health insurance, we find that the vast majority of the benefits would still accrue to the bottom three deciles of the income distribution when we include the value of employer- and government-provided health insurance in our expanded yearly income measure. (JEL D31, H51, I14)

I. INTRODUCTION

The public-use version of the Current Pop- ulation Survey — Annual Social and Economic Supplement (CPS) — is the most commonly used data set to capture yearly levels and trends in U.S. income and its distribution. Each year the Census Bureau reports (U.S. Census Bureau 2009) the previous year’s median pre-tax, in- cash income of households (from both pub- lic and private sources) as well as how the household size-adjusted, pre-tax, post-transfer, in-cash income of persons is distributed. Most

∗This paper was funded in part by the Pew Charitable Trust – Economic Mobility Project. Any opinions expressed in this paper are solely the authors’ and should not necessar- ily be attributed to the Pew Charitable Trust. We thank Scott Winship for valuable advice at all stages of the project. Burkhauser: Professor, Department of Policy Analysis and

Management, Cornell University, Ithaca, NY 14853; Melbourne Institute of Applied Economic and Social Research, University of Melbourne, Victoria, Australia. Phone 607-255-2097, Fax 607-255-4071, E-mail rbv1@ cornell.edu

Larrimore: Economist, Joint Committee on Taxation, United States Congress, SW, Washington, DC 20515. Phone 202-226-7575, Fax 202-225-0833, E-mail jeff.larrimore @mail.house.gov

Simon: Professor, The School of Public and Environmental Affairs (SPEA), Indiana University, Bloomington, IN 47405-1701. Phone 812-856-3850, E-mail simonkos@ indiana.edu

researchers outside of the Census Bureau who use the CPS to capture levels and trends in income and income inequality also focus on this cash measure of income.1 (See Atkinson and Brandolini 2001 and Gottschalk and Smeeding 1997, for reviews of this literature.)

Yet few of these studies consider the impor- tance of excluding non-wage compensation from these measures. Here, we do so, by focusing on the resource value of employers’ contributions

1. In this paper we focus on how conventional measures of median income and income inequality are affected by explicitly including the value of health insurance in these measures. Our work parallels that of Meyer and Sullivan (2003, 2009) showing how pre-tax and post-tax income definitions impact measures of poverty trends in their argument that a shift from a poverty measure based on resources to one based on consumption would more precisely capture poverty trends.

ABBREVIATIONS

ACA: Affordable Care Act CBO: Congressional Budget Office CPS: Current Population Survey — Annual Social and

Economic Supplement ECEC: Employer Cost for Employee Compensation MEPSIC: Medical Expenditure Panel Survey Insur-

ance Component

779 Contemporary Economic Policy (ISSN 1465-7287) Vol. 31, No. 4, October 2013, 779 – 794 Online Early publication September 20, 2012

doi:10.1111/j.1465-7287.2012.00336.x © 2012 Western Economic Association International

780 CONTEMPORARY ECONOMIC POLICY

to health care. It is the most important com- ponent of voluntary non-wage compensation, accounting for 32% of voluntary non-wage com- pensation and 22% of all non-wage compensa- tions (Pierce 2001). Like other types of compen- sations, employer-provided health insurance is a resource that can be “directly” consumed each year.

To be consistent in our efforts to show the importance of health insurance on household resources, we consider the resource value of both employer- and government-provided health insurance, via Medicare and Medicaid, on mea- sured levels of household income and its distri- bution over time.

We do so by the following:

1. Estimating a broader measure of house- hold income that adds employer-provided health insurance contributions and the ex ante value of government-provided health insurance (Med- icaid and Medicare) to the traditional measure of household income (household size-adjusted, pre-tax, post-transfer, in-cash income of per- sons). Importantly, we use the ex ante cost of this insurance to employers and to the govern- ment as our measure of its resource value to the households, not the ex post payments made for health-care services used by these households.

2. Showing the sensitivity of traditional in- cash measures of the level and distribution of income to the addition of the “equivalent income value” of these two health insurance resources. Our focus throughout (except in Figure 1) is on household size-adjusted income of individuals. To examine how the addition of these health insurance values changes trends in inequality by age, we split the population into four age categories: children (aged 0 – 18), young adults (aged 19 – 25), working age (aged 26 – 62), and retirement age (aged 63 and over).

3. Examining how income inequality has changed over the period 1995 – 2008 based on this broader measure of income.

4. Providing an application showing the importance of our expanded measure of income using a stylized example of current policy rele- vance. We examine the change in the level and distribution of total income (our broader mea- sure of income including the value of health insurance) from two key features of the Afford- able Care Act (ACA) of 2010 — the expansion in Medicaid provision to those with incomes below 133% of the poverty line and the provision of publicly funded subsidies and premium tax credits to individuals living in families between

133% and four times the official poverty line for purchasing private coverage on the state-run exchanges. Our aim is not to capture the pre- cise details of how the ACA will affect health insurance coverage or income measures. Rather, it is to illustrate the importance of putting an explicit value on the provision of government- subsidized health insurance coverage in “social success measures” which researchers use to cap- ture public policies’ impact on the resources available to all Americans.

II. RELATED STUDIES

The importance of fringe benefits and non- wage compensation has been considered in the earnings inequality literature (Chung 2003; Levy 2006; Pierce 2001, 2010). However, these papers focus on employer compensation to indi- vidual workers and hence do not demonstrate how their inclusion impacts the overall income distribution. Workers live and share their wages with those living with them (e.g., family or household members), so it is necessary to deter- mine the composition of these larger shar- ing units and then gauge how employer- and government-provided health insurance impacts them.

Fewer researchers have attempted to include the value of employer-provided health insurance and government-provided Medicare or Medi- caid in their measures of household income. Most recently, the Congressional Budget Office (CBO) (2011) did so. Referencing an earlier version of this paper, the CBO analysis also finds similar results to our findings (a reduc- tion in inequality when using the broader income measure) in Appendix C of their report. For their primary estimates the CBO uses the ex ante value of employer-provided health insur- ance, but unlike us they use the fungible insur- ance value of Medicare and Medicaid, which the Census Bureau has estimated and included in the public-use versions of the CPS since 1995 (U.S. Census Bureau 2009).2 Unlike the case for employer-provided health insurance, this means that they only count part of the ex ante insurance value of Medicare and Medicaid for low-income

2. Like us, the Census Bureau also creates a value for employer-provided health insurance in the CPS using a sta- tistical model (U.S. Census Bureau 1993). We were unable to find a recent description of how this measure — which was created using data from 1977 — has been updated since 1993. We were also unable to find a description of the statistical model used to create it in any year.

BURKHAUSER ET AL.: VALUING HEALTH INSURANCE 781

families, depending on whether the individuals are able to afford other basic food and hous- ing needs (see U.S. Census Bureau 1993 for more details). Even using this fungible value, which by definition reduces the value of ben- efits accrued to low-income families, the CBO finds that its inclusion increases the income of low-income households more than high-income households. For the reasons discussed below, unlike Census and the CBO, we use the full ex ante value of Medicare and Medicaid as well as our updated estimates of the ex ante value of employer-provided health insurance. Doing so, we find that including both these ex ante val- ues of health insurance reduces inequality even further.3

Our measure attempts to capture the costs to employers of providing health insurance to their workers and their households as well as the federal and state government costs of providing Medicare and Medicaid to qualified beneficia- ries. This measure is more in keeping with the concept of these programs as insurance against health-related expenses.4 We assign this ex ante value to all those covered in a given year rather than, for instance, assigning a zero value to individuals who are covered by their employer- or government-provided health insurance but

3. It is common in the cross-national literature that focuses on the importance of non-cash benefits on income inequality to use the ex ante value of employer- and government-provided health insurance (Garfinkel, Rainwa- ter, and Smeeding 2006; Garfinkel, Smeeding, and Rain- water 2010; Marical et al. 2008; Paulus, Sutherland, and Tsakloglou 2010; Smeeding et al. 1993). Many such studies, including Garfinkel, Rainwater, and Smeeding (2006) and Garfinkel, Smeeding, and Rainwater (2010), also include the valuation of government-provided education benefits in their comparison and attempt to use a balanced budget framework such that the net budgetary impacts of these benefits are zero rather than deficit financed.

4. To the extent that employees can anticipate their health expenses as in the case of chronic illnesses, those who have higher anticipated expenses may derive greater value (greater consumer surplus) from group insurance that does not account for pre-existing conditions in pricing since it may provide access to otherwise unaffordable care or reduce the costs of these anticipated expenses. This is consistent with Nyman (1999) who suggests that access to unaffordable treatments is important when valuing health care. To the extent that this occurs, assuming that the value of health insurance to the individual is equal to the group insurance value — unadjusted for pre-existing conditions — will overstate the value (consumer surplus) of insurance for those who expect not to use much health care and understate it for those with high health-care usage expectations. However, given the lack of information in our data about health-care usage or the presence of chronic conditions, we do not attempt to model this additional aspect of the value of health insurance and simply measure it at its average cost of provision for the group.

who ex post do not receive any health care in that year.5

III. METHOD AND DATA

We use the 1996 – 2008 CPS to measure lev- els and trends in the household size-adjusted, pre-tax, post-transfer, in-cash income of all Americans as well as their type of employer- and government-provided health insurance cov- erage. (See Burkhauser and Simon 2008, and Burkhauser, Lyons, and Simon 2011, for a fuller discussion of the issues related to using the CPS for this purpose.) For confidentiality reasons, CPS income data have been inconsistently top- coded over this period. If we do not account for these inconsistencies, one will confuse an improvement in the measure of income with a real increase in income, especially at the top of the distribution. Larrimore et al. (2008) have developed methods to provide public users with instructions and additional data needed to cor- rectly account for top-coding in the public-use CPS. Their work produced a consistent set of cell means for all top codes in the public-use CPS that, when used with existing public-use CPS data, provide a comparable data series from 1967 to the present. We use these cell means in our analysis. (See Burkhauser et al. 2012, for a comparison of how the internal data, off which these cell means are based, compare to those of Piketty and Saez 2003, using income tax records.)

While the CPS data provide information on health insurance coverage, we must impute the ex ante value of employer contributions to health insurance and the value of government-provided health insurance from other sources. The value of employer contributions for health insurance comes from the Medical Expenditure Panel Sur- vey Insurance Component (MEPSIC). This sur- vey is conducted by the U.S. Census Bureau and funded by the Agency for Healthcare Research and Quality. It has been conducted

5. Burtless and Svaton (2009) offer an alternative mea- sure of the value of employer- and government-provided health insurance to families. They measure the ex post cost of the health-care actually used by families in a given year rather than the ex ante value of health insurance provided to families in a given year. Burtless and Svaton (2010) expand their previous paper using Census Bureau estimates to mea- sure the ex ante insurance value of employer-provided health insurance and the fungible insurance value of government- provided Medicare and Medicaid as well as their previous ex post cost of health care actually used by families to measure levels and trends in income and consumption.

782 CONTEMPORARY ECONOMIC POLICY

every year since 1996. These are confidential Census Bureau microdata that involve lengthy application periods before access is allowed. Fortunately, the cell means are released pub- licly, and we use them here.6 This includes the employer contribution for single and family plans, by state, by year, and by firm size.7

Because cell-based imputations flatten the distribution of benefits, to the extent that vari- ance within cells is suppressed, we further improve the imputations using relative bene- fits by occupation from the Employer Cost for Employee Compensation (ECEC) index pro- duced by the Bureau of Labor Statistics. In doing so, we assume that the relative compensation by occupation is the same by plan type, state, and firm size, and thus increases or decreases the value of health insurance benefits for each indi- vidual relative to the cell mean by the ratio of their occupation’s mean health benefits in the ECEC to those of all workers in the given year. This increases the number of unique cells valu- ing health insurance benefits in all years. This method results in less flattening of the distribu- tion because workers in low-income households on average work in industries with less gener- ous health benefits. This somewhat reduces the equalizing effect of including health insurance information in our analysis. After imputing the value of insurance, we merge these insurance values to the CPS data using variables common to both data sets. See Appendix S1 in the online version of this article for further details of our imputation procedure.8

We assign values to these non-cash bene- fits which are equal to the private costs paid by employers (this values non-wage compen- sation at its market price as opposed to the value that individuals would pay for it). We then consider the impact of including public sector provision of non-cash transfers in the form of government-provided health insurance through Medicaid and Medicare. We value these pro- grams at their average administrative cost per

6. As with all survey-based data of population samples, even though we use cell means, the MEPSIC data are subject to sampling variability and potential misreporting error to the extent that survey responses do not precisely match actual values.

7. Available at http://www.meps.ahrq.gov/mepsweb/ data_stats/quick_tables_search.jsp?component=2& subcomponent=2

8. In an unreported analysis, we examined national estimates from our database and ensured that they are identical to those released by the Census Bureau, as we have used the same algorithm they use in creating national insurance rates from the CPS.

person. For Medicaid this is the average imputed cost by state and year and age group (adult or child). For Medicare, this is the average imputed cost by state and year among all Medicare par- ticipants. That is, we consistently value Medi- care and Medicaid at their ex ante insurance value, not their “fungible” insurance value. See Appendix S1 in the online version of this article for additional details.

A reasonable question about our imputa- tion method is whether the focus on assigning the value of employer-provided health insur- ance to workers based on the worker’s deci- sion to formally join the plan is appropriate. There are cases, for example, where workers are receiving insurance that to some degree is directly paid by their employers but they cannot afford doctor co-pays and thus do not use their coverage except in extreme emergencies. Such workers are only slightly better off than work- ers whose employers require a small but for- mal employee contribution which makes them decide to not join the insurance plan. In our estimates, we assume that the former workers receive the full subsidized value of the insur- ance, while the later receive none of it. To the degree this occurs we will overstate its value to the former. However, there are also work- ers whose employer-provided group insurance is less costly than what can be purchased through individual non-group programs. For them the employer’s cost of provision of health insur- ance understates its value to them. Therefore, while there may be some bias in our method of valuing this resource, the direction of the bias is not obvious. However, on average, workers must value employer-provided insurance above the zero value currently assumed in pre-tax, post-transfer measures of household well-being because otherwise employees would lobby their employers to increase wages and forego these health insurance schemes.9 The most recent

9. A related question, sometimes conflated with the value of employer- and government-provided health insur- ance, is whether increases in the price of health insurance reflect medical inflation, with individuals receiving the same level of health care at higher costs, or improvements in the quality of health benefits. We abstract from this question, however, based on the observation that rising prices for health care occur for uninsured care and non-group health insurance along with that for employer- and government- provided insurance. Since the cost of that insurance has gone up irrespective of the provider of insurance, the question of why health costs are rising is separate from the ques- tion of the value of having insurance provided by one’s employer or the government rather than having to purchase it independently.

BURKHAUSER ET AL.: VALUING HEALTH INSURANCE 783

evidence from studies using exogenous variation in provision of health insurance finds that work- ers value the newly provided coverage at very close to the cost of the coverage to employers (Kolstad and Kowalski 2012).

Although the focus of this paper is on the impact of including the value of health insur- ance, this is, of course, not the only income source that is currently excluded from the tra- ditional pre-tax, post-cash-transfer income def- inition used by the Census Bureau and in much of the inequality literature. Other ele- ments of income which are not considered in traditional pre-tax, post-cash transfer mea- sures of income include non-cash transfers such as food stamps; tax payments including trans- fers administered through the tax code such as the EITC and child tax credit; and irregularly received income such as capital gains from the sale of stocks, housing, or other assets. Some of these excluded income items, such as food stamps and progressive taxes, will likely fur- ther reduce inequality.10 Others, such as capital gains and other irregularly received income, will likely increase inequality from those described here.11 Recent findings by the CBO (2011), for example, suggest faster inequality growth than that observed here because of different treat- ment of top incomes but also partially because of the inclusion of capital gains in their income measure. These additional income elements are excluded from this paper in order to focus specifically on the impact of health insurance, and their inclusion should not greatly impact the direction or magnitude of the effects of includ- ing health insurance. Nevertheless, these other income elements are important as researchers

10. In the case of taxes, note that the pre-tax, post- transfer income definition potentially double-counts income by including transfer income as a resource without excluding the taxes used to pay for those transfers. Since our paper builds off of the pre-tax, post-transfer income distribution method of measuring income, this concern of double- counting remains. Additionally, since we now include the value of government-provided health insurance, the impact of excluding taxes is further increased. Here we focus solely on the consequences of excluding health insurance in traditional income measure. For a discussion of the impact of taxes on household income trends, see Burkhauser, Larrimore, and Simon (2012).

11. Piketty and Saez (2003) use tax records data to show that inequality increases with the inclusion of capital gains income reported on tax returns. A focus on only taxable capital gains may overstate the inequality increase from the inclusion of all capital gains since gains from housing, which is the most significant capital gain for most middle- class families, is non-taxable. Nevertheless, even with this limitation it is likely that the inclusion of all capital gains would increase inequality levels.

continue to strive for more inclusive income definitions.

While non-cash income and realized capi- tal gains were excluded to maintain the focus on how health insurance changes, others are excluded for more theoretical reasons. Although, even the inclusion of realized capital gains from stocks and bonds is controversial, both because it potentially overstates inequality when realized capital gains from housing and other assets held more broadly by the population are excluded and because unrealized capital gains in general presumably have some effect on the change in resources available to households, but are not considered. In contrast, we do not discuss pen- sion benefits and paid time off here because they are less relevant in the income inequal- ity literature than in the wage inequality liter- ature. Pierce (2010) shows that the inclusion of these two components of non-wage com- pensation further increases wage inequality. But he does not address their inclusion in house- hold income inequality measures. We exclude the accrual value of pension benefits because they are instead included in the CPS as pay- ments during retirement. As a result, the inclu- sion of pension benefits at accrual would result in double-counting of benefits unless such bene- fits are excluded from income during retirement years when the income is actually available to the household for consumption. In the case of paid leave, while the inclusion of such bene- fits is logically consistent for wage inequality studies because paid leave effectively increases the wage per hour for workers over the time those workers actually are working on the job, it makes less sense to do so in traditional measure of household income. In this literature the value of non-work time (leisure) is not included either for part-time workers or non-working household members. So for consistency, unless some value was placed on those “leisure” activities it should not be placed on “vacation” time.

Studies of income inequality usually focus on the entire age distribution and we do so in our main analysis. But because most of the impact of the inclusion of employer-provided health insurance and Medicaid affects the working-age population and their children, while Medicare mostly affects the older population, we sep- arately show their effects on children, young adults, those of working age, and older age pop- ulations. That is, we will look at the household size-adjusted income of working-age people separately from the household size-adjusted

784 CONTEMPORARY ECONOMIC POLICY

FIGURE 1 Trends in Median Income and Total Income by Household

40000

42000

44000

46000

48000

50000

52000

54000

56000

58000

60000

A n

n u

a l

D o

ll a

rs ,

in 2

0 0

8 T

e rm

s

Our median income

Census median income

Our median total income

Notes: All numbers are in real 2008 dollars and use sample weights. Census income of the median household (from reported data, Table H-6) www.census.gov/hhes/www/income/histinc/h06AR.xls. The year is the actual year of data, not the year of the interview in the CPS. Total median income is income plus the ex ante insurance value of employer-provided health insurance and government-provided Medicaid and Medicare of the median household, and was calculated by the authors.

income of older persons, and so on. Note, how- ever, that people of any ages may live in any given household. In all cases we adjust for household size by dividing household income by the square root of the number of house- hold members. This size adjustment technique is commonly used in U.S. and cross-national studies of inequality (Atkinson and Brandolini 2001; Burkhauser et al. 2011; Gottschalk and Smeeding 1997) and closely matches the adjust- ments for household size implied by the official U.S. poverty thresholds (Ruggles 1990). Addi- tionally, all negative values are replaced with zeros, all income values are adjusted to real 2008 dollars using the CPI-U-RS (Stewart and Reed 1999), and the value of medical care is adjusted to real 2008 dollars using the Medical Care CPI.12

12. Since employer- and government-provided insur- ance is not paid directly by the consumer it is not included in the consumer price index. Therefore, as noted by an anony- mous referee, the CPI may understate inflation once we include the resource value of these premiums as income. Deflating these medical costs using the medical CPI rather than the CPI-U-RS reduces the extent to which medical costs slowed the growth in inequality since low-income individu- als receive a higher fraction of their income in this form of medical coverage.

IV. RESULTS

Each year the Census Bureau publishes real median pre-tax, post-transfer, in-cash household income figures for the previous year based on the March CPS data (U.S. Census Bureau 2009). Figure 1 replicates these values for income years 1995 – 2008 above.13 Economic growth in the 1990s propelled median income upward, hit- ting a peak in 2000 and then falling to a 2004 low before once again rising through 2007. Note, however, that by 2007, median pre-tax, post-transfer, in-cash income had not returned to its 2000 business cycle peak high. When we redo these calculations but include our estimated employer- and government-provided health insurance values, not surprisingly, median total income is higher in all years. More surpris- ingly, because these contributions were rising in value, they offset to some degree the fall in other income over this period. While past studies have shown that the prevalence of employer-provided

13. We are unable to exactly match reported CPS figures (http://www.census.gov/prod/2008pubs/p60-235.pdf). We suspect that the difference may occur because we report the median value while the Census Bureau uses a linear interpolation procedure for calculating median incomes to account for clustering of responses at round numbers.

BURKHAUSER ET AL.: VALUING HEALTH INSURANCE 785

health insurance has been falling over these years, Figure 1 shows that the median Ameri- can household’s income, inclusive of the value of employer- and government-provided health insurance, rose sufficiently by 2007 to exceed real 2000 income levels, before falling during the recession year of 2008.

Figure 1 shows how valuing employer- and government-provided health insurance impacts measured levels and trends of the average Amer- ican household’s income. In the next series of tables, we show how valuing health insurance impacts the size-adjusted household income dis- tribution of all Americans.14 Table 1 reports how income was distributed across all Amer- icans in 2008 by assigning persons to deciles based on their household size-adjusted, pre-tax, post-transfer, in-cash income. The bottom row reports mean income values for all Americans. As can be seen in column 1, mean income varies from a highest decile $139,395 to a low- est decile mean of $5,500. The entire popu- lation’s mean income is $44,616. In the next five columns we show by deciles the mean value of employer-provided health insurance and government-provided Medicaid and Medi- care, as well as their sums (total health insur- ance). The last two columns show by deciles the mean value of income plus health insurance (total income) and the share of the total com- ing from health insurance, respectively. While health insurance makes up only 9.93% of all household size-adjusted income in the United States, it is by far a more important share of the income of the lower deciles of the distribution.

As can be seen by comparing columns 2 and 3, the inclusion of industry in our imputa- tion procedure reduces mean employer-provided health insurance benefits to the bottom six deciles while increasing the mean benefits for the top four deciles. However, even with this adjustment, the distribution of employer health insurance is still flatter than the distribution of cash income. For a person in the bot- tom decile, employer-provided health insurance is 4.8% of cash income ($266/$5,500). This rises to a peak of 9.4% for a person in the fifth decile ($2,901/$30,920) and then falls in succeeding deciles to a low of 3.7% for a person

14. In Figure 1, the observation is a household, and all CPS households are included. In the rest of the tables, the observation is an individual in a household (excluding those living in group quarters and those households containing members of the armed forces), and the income measure used is household size-adjusted income.

($5,222/$139,395) in the top one. So employer- provided benefits make up the largest fraction of income for those in the middle of the dis- tribution. This is consistent with health insur- ance being offered more in firms that employ workers living in middle- and upper-income households, but once employed in such firms, benefit packages are the same regardless of salary. This could be the case because of either the non-discrimination clause applied to large, self-insured employers (Carrigan et al. 2002) or the need for small employers to have virtually all eligible employees accept coverage to avoid adverse selection concerns on the part of insur- ers. All further results in this paper use our modified employer health insurance values as reported in column 3 of Table 1.

Not surprisingly, publicly provided health insurance benefits are greater in the bottom deciles, both in their values (columns 4 and 5) and percent of income. For the bottom decile, Medicare and Medicaid combined represent 67% of cash income [($1,591 + $2,085)/$5,500], but by the fifth decile they represent just 5% [($241 + $1,407)/$30,920] and by the top decile they represent less than 0.5% [($30 + 584)/$139,395] of cash income.

Column 6 reports by deciles the sum of mean employer- and government-provided health in- surance. The decile totals are remarkably sim- ilar across the entire income distribution with the heavy targeting of Medicaid to the young in the lower three income deciles and the tar- geting of Medicare primarily to non-working older persons living in the bottom half of the income distribution offsetting the smaller access to employer-provided health insurance for working-age persons in these deciles. The net result is a health-care insurance system that provides benefits very much more equally dis- tributed across the entire income distribution than all other cash income sources. As a result, total income (the sum of mean income and mean health insurance) reported in column 7 is more evenly distributed across deciles than income (column 1), with the share of health insurance (column 8) in the portfolio of lower-income groups, especially the bottom three, much higher than in the rest of the deciles.

To provide some sense of how income has grown across the income distribution in the United States over time and how much including the value of employer- and government-provided health insurance matters in such calculation, Table 2a first reports how income was distributed

786 CONTEMPORARY ECONOMIC POLICY

TABLE 1 How Measured Income in 2008 Changes When the Value of Health Insurance Is Included,

by Decile

Decile Income

Employer Health Insurance Without Industry

Employer Health

Insurance Medicaid Medicare

Total Health

Insurance Total

Income

Share Health

Insurance

1 5,500 312 266 1,591 2,085 3,942 9,437 41.77 2 12,952 927 823 944 3,080 4,847 17,800 27.23 3 18,717 1,619 1,469 621 2,612 4,702 23,419 20.08 4 24,605 2,343 2,206 353 2,059 4,618 29,223 15.80 5 30,920 3,018 2,901 241 1,407 4,549 35,469 12.83 6 37,883 3,579 3,513 149 1,161 4,824 42,707 11.30 7 46,191 3,964 3,990 87 958 5,035 51,227 9.83 8 56,728 4,364 4,463 73 761 5,297 62,025 8.54 9 73,297 4,729 4,991 56 694 5,741 79,038 7.26

10 139,395 4,797 5,222 30 584 5,836 145,231 4.02 Mean 44,616 2,965 2,984 415 1,540 4,939 49,554 9.97

Notes: Income is household size-adjusted, pre-tax, post-transfer in-cash income of persons for everyone in the CPS except those who are in group quarters or in the households of those in the armed forces. All negative values are replaced with zeros, real 2008 dollars, adjusted for household size by dividing by square root of the number of household members, using supplemental sample weights.

Source: Authors’ calculations from the CPS.

across all Americans in the first year of our data, 1995 (column 1), and repeats those val- ues for 2008 (column 2), the most recent year of our data. Like Table 1, this is done by assigning individuals to deciles based on their household size-adjusted, pre-tax, post-transfer, in-cash income and in the bottom row report- ing the mean for all Americans. Column 3 then shows how much mean income has grown by decile. This traditional measure of income growth shows positive growth in all but the bottom deciles. But this positive growth in the nine remaining deciles was uneven with the bot- tom deciles increasing at about two-thirds the rate seen in the next three, and half that of the top three. The next three columns repeat this exercise, but include the value of health insurance as part of total income. The results are quite different. All deciles had positive growth. While the bottom decile substantially lagged behind the others, growth was much closer across the remaining nine deciles. The final two columns show that as a result of sub- stantial increases in the value of employer- and government-provided health insurance, which has been equally distributed across the popula- tion (as seen in column 6 of Table 1), health insurance has grown as a share of the port- folio of total income held by all deciles. But this growth as a percent of total income was

less among the upper parts of the distribu- tion. As shown in Table 2a, when employer- and government-provided health insurance is included in total income, its increasing value since 1995 not only results in greater levels of growth in all deciles but more equal growth across deciles between 1995 and 2008.

Table 2b more formally looks at how income inequality measures and their trends are im- pacted by the inclusion of employer- and government-provided health insurance in their measure. The first three rows focus on 90/10, 90/50, and 50/10 percentile comparisons of income. While these inequality measures rise between 1995 and 2008 when income is mea- sured not including the value of employer- and government-provided health insurance, their increases are substantially lower once the value of health insurance is included in total income and it decreases over 50/10 comparisons. Addi- tionally, while the increase in total income from 1995 to 2008 is statistically significant at the 95% level for all three percentile ratios using cash income, for total income it is only statisti- cally significant for the 90/50 ratio.

While the earnings literature primarily fo- cuses on 90/10 percentile comparisons to mea- sure levels and trends in inequality, Burkhauser, Feng, and Jenkins (2009) argue that once prob- lems associated with top-coding are corrected,

BURKHAUSER ET AL.: VALUING HEALTH INSURANCE 787

TABLE 2 Changes in Measured Income With and Without Health Insurance 1995 – 2008

(a) By Decile

Decile Income

1995 Income

2008

Percent Change

in Income

Total Income

1995

Total Income

2008

Percent Change in

Total Income

Share Health Insurance

1995

Share Health Insurance

2008

1 5,504 5,500 −0.08 8,939 9,437 5.57 38.43 41.77 2 12,228 12,952 5.93 16,001 17,800 11.24 23.58 27.23 3 17,727 18,717 5.59 21,193 23,419 10.50 16.36 20.08 4 23,186 24,605 6.12 26,518 29,223 10.20 12.56 15.80 5 28,725 30,920 7.64 32,057 35,469 10.65 10.39 12.83 6 34,781 37,883 8.92 38,218 42,707 11.75 8.99 11.30 7 41,796 46,191 10.52 45,401 51,227 12.83 7.94 9.83 8 50,909 56,728 11.43 54,838 62,025 13.11 7.16 8.54 9 65,086 73,297 12.62 69,269 79,038 14.10 6.04 7.26

10 123,872 139,395 12.53 128,144 145,231 13.33 3.33 4.02 Mean 40,378 44,616 10.49 44,055 49,554 12.48 8.35 9.97

(b) By Alternative Income Ratios

Inequality Measure

Income 1995

Income 2008

Percent Change Using Income

Total Income 1995

Total Income 2008

Percent Change Using Total Income

90/10 8.248 8.846 7.258 5.966 6.097 2.191 (0.0541) (0.0553) (0.0601) (0.0835)

90/50 2.378 2.491 4.742 2.284 2.340 2.464 (0.0092) (0.0086) (0.0108) (0.0110)

50/10 3.468 3.551 2.402 2.613 2.606 −0.267 (0.0194) (0.0192) (0.0266) (0.0365)

Gini 0.422 0.434 2.723 0.393 0.400 1.826 (0.0012) (0.0010) (0.0012) (0.0009)

Notes: (a) See notes to Table 1. (b) Standard errors in parentheses. For the Gini coefficient, standard errors are calculated based on linearization methods using the svylorenz STATA program (Jenkins 2006). For the 90/10, 90/50, and 50/10 ratios, standard errors are bootstrapped standard errors. To produce the numbers in this table, we create 100 percentile groups, then take the mean values of income, the total income that includes the value of health insurance, and the share that represents insurance, for those who are at the 10th, 50th, and 90th percentile group of total income in that year. The percent change is calculated from these numbers.

Source: Authors’ calculations from the CPS.

it is possible to use Gini coefficients (row 4 of Table 2a) or other scalar measures of the entire distribution to consistently measure income inequality levels and trends. When doing so we find similar result to those found in the top three rows. Income inequality rises slightly when health insurance is not included (2.723%) but more modestly so (1.826%) when it is. Both increases are statistically significant at the 95% level.

In Table 3 we demonstrate how the inclu- sion of health insurance differentially impacts individuals of different age groups. In Table 3a, we provide the value of health insurance in 2008 by deciles for four age groups. In creating the deciles, individuals are classified based on their income rank within their age group rather than their income rank among the entire pop- ulation. The ex ante value of health insurance

is greater for those aged 63 and over than for younger (aged 19 – 25) or middle-aged (aged 26 – 62) workers as well as for children (aged 18 or less). This partially reflects universal Medi- care coverage for those over age 65 but also reflects the higher average health spending of older people.15 Looking at the change in the Gini

15. The higher cost of providing health insurance to the aged reflects their risk of requiring medical care. Provision of this insurance is a real cost to society that is reflected in our measure of its value — the ex ante insurance value of this coverage. And in that sense it provides resources to the aged that exceed those provided to younger persons. So for all those who receive this subsidy it makes them better off than if they had to purchase it themselves. Thus, while the presence of Medicare improves the well-being of the aged relative to a case where they had the same expected medical care requirements and had to pay for them out of their current cash income, it does not mean that the elderly are better off than if they were young and did not require such expensive coverage.

788 CONTEMPORARY ECONOMIC POLICY

T A

B L

E 3

(a )

In co

m e,

H ea

lt h

In su

ra nc

e, an

d T

ot al

In co

m e

by A

ge G

ro up

(2 00

8) ;

(b )

P er

ce nt

C ha

ng es

in A

lt er

na ti

ve In

co m

e M

ea su

re s

W it

h an

d W

it ho

ut H

ea lt

h In

su ra

nc e

by A

ge G

ro up

(1 99

5 –

20 08

), by

D ec

il e

(a )

A ge

d 0

– 18

A ge

d 19

– 25

A ge

d 26

– 62

A ge

d 63

an d

u p

D ec

il e

In co

m e

H ea

lt h

In su

ra n

ce T

ot al

In co

m e

In co

m e

H ea

lt h

In su

ra n

ce T

ot al

In co

m e

In co

m e

H ea

lt h

In su

ra n

ce T

ot al

In co

m e

In co

m e

H ea

lt h

In su

ra n

ce T

ot al

In co

m e

1 3,

88 3

3, 40

0 7,

27 6

4, 78

4 2,

47 4

7, 25

8 6,

67 6

3, 28

7 9,

95 9

6, 64

6 8,

92 2

15 ,5

66 2

10 ,0

34 3,

30 9

13 ,3

43 12

,2 56

2, 97

6 15

,2 33

15 ,6

79 3,

73 6

19 ,4

15 12

,4 88

9, 85

3 22

,3 41

3 15

,1 20

3, 34

9 18

,4 68

18 ,1

12 3,

00 9

21 ,1

21 22

,6 89

3, 84

2 26

,5 31

16 ,2

02 9,

69 6

25 ,8

98 4

20 ,2

48 3,

54 9

23 ,7

97 23

,5 44

3, 21

2 26

,7 56

29 ,4

79 4,

04 7

33 ,5

26 20

,0 77

9, 03

8 29

,1 14

5 25

,9 71

3, 88

6 29

,8 56

29 ,3

69 3,

34 6

32 ,7

16 36

,3 63

4, 44

4 40

,8 07

24 ,6

97 8,

21 8

32 ,9

16 6

32 ,1

50 4,

18 4

36 ,3

34 35

,4 23

3, 76

9 39

,1 92

43 ,8

12 4,

83 3

48 ,6

45 30

,3 04

7, 57

0 37

,8 73

7 39

,3 76

4, 52

6 43

,9 02

43 ,0

46 4,

29 6

47 ,3

42 52

,5 18

5, 04

3 57

,5 61

37 ,6

47 7,

32 4

44 ,9

71 8

48 ,5

70 4,

79 0

53 ,3

60 52

,1 21

4, 56

4 56

,6 85

63 ,5

51 5,

45 2

69 ,0

03 48

,2 42

7, 12

0 55

,3 63

9 63

,0 13

5, 17

9 68

,1 92

65 ,9

46 5,

01 9

70 ,9

65 81

,0 17

5, 77

8 86

,7 95

65 ,3

83 7,

01 6

72 ,3

99 10

12 3,

99 8

5, 31

5 12

9, 31

4 11

6, 82

1 5,

71 1

12 2,

53 2

15 2,

62 0

5, 79

1 15

8, 41

1 12

7, 30

7 7,

04 3

13 4,

35 0

M ea

n 38

,2 27

4, 14

8 42

,3 75

40 ,1

37 3,

83 8

43 ,9

75 50

,4 21

4, 62

5 55

,0 45

38 ,8

91 8,

18 0

47 ,0

71 G

in i

0. 45

1 0.

41 9

0. 41

3 0.

39 2

0. 41

7 0.

39 2

0. 44

1 0.

36 3

(b )

A ll

A ge

d 0

– 18

A ge

d 19

– 25

A ge

d 26

– 62

A ge

d 63

an d

u p

D ec

il e

In co

m e

T ot

al In

co m

e In

co m

e T

ot al

In co

m e

In co

m e

T ot

al In

co m

e In

co m

e T

ot al

In co

m e

In co

m e

T ot

al In

co m

e

1 −0

.0 8

5. 83

−2 .0

5 −0

.5 3

−2 .3

8 −1

.7 1

−3 .0

9 2.

68 −4

.3 3

13 .3

6 2

5. 93

11 .6

9 10

.3 9

9. 80

3. 60

7. 47

1. 78

7. 23

6. 02

15 .9

3 3

5. 59

10 .8

2 8.

03 11

.5 2

4. 72

7. 64

1. 98

5. 87

7. 39

14 .9

6 4

6. 12

10 .2

8 7.

22 11

.3 9

5. 02

7. 38

3. 82

6. 83

8. 19

15 .7

0 5

7. 64

10 .6

4 7.

79 11

.0 4

5. 62

7. 25

5. 57

8. 32

11 .4

6 16

.3 1

6 8.

92 11

.7 3

9. 25

11 .9

9 5.

85 7.

36 7.

11 9.

77 14

.6 5

18 .6

1 7

10 .5

2 12

.7 9

10 .8

1 13

.1 7

7. 41

8. 94

8. 34

10 .3

4 17

.6 3

20 .4

8 8

11 .4

3 13

.0 8

12 .9

9 14

.9 6

7. 62

8. 34

8. 98

10 .9

0 20

.5 7

22 .7

6 9

12 .6

2 14

.0 7

15 .2

6 16

.6 2

8. 97

9. 34

10 .2

6 11

.6 9

22 .5

2 23

. 8 5

10 12

.5 3

13 .3

2 14

.9 8

15 .6

5 10

.4 9

11 .0

4 11

.4 1

12 .1

7 16

.0 0

16 .8

5 M

ea n

10 .4

9 13

.9 1

12 .2

9 13

.8 3

7. 80

8. 85

8. 31

10 .1

9 15

.7 3

18 .6

5 G

in i

2. 72

1. 83

2. 88

2. 78

2. 71

2. 26

3. 72

2. 77

4. 21

2. 10

N o te

s: S

ee no

te s

to T

ab le

1. S

o u

rc e:

A ut

ho rs

’ ca

lc ul

at io

ns fr

om th

e C

P S

.

BURKHAUSER ET AL.: VALUING HEALTH INSURANCE 789

coefficient within each age group, we similarly see that the effect is greatest among older per- sons. While including health insurance reduces inequality within all age groups, it does so the most for those aged 63 and over where health insurance is very evenly distributed across all income deciles and forms a very impor- tant share of the bottom part of their income distribution.

In Table 3b, we show, for each age group, the percentage change between 1995 and 2008 in their mean income and total income (including the value of health insurance) overall and by income deciles as well as their overall income and total income Gini coefficients. In all four age groups, including the value of health insurance increases mean growth as can be seen in the next to last row, and the same is the case within each decile. In general, growth in total income that includes the value of health insurance is more equalizing than growth in income excluding health insurance. But this is especially the case for those aged 63 and over. This same pattern of reduced inequality is more formally seen in the last row which compares income and total income Gini coefficients. Thus, a substantial fraction of the decline in inequality which comes from including the ex ante value of health insurance comes from better representing the resources available to older Americans.

V. APPLICATION TO THE ACA

Having established that the inclusion of health insurance information impacts measured trends in income and income inequality, we now turn our attention to an example of how this matters for evaluating public policy using a very stylized example of two features of the Patient Protection and ACA and the companion Health Care and Education Reconciliation Act of 2010, known collectively as the ACA.16 Our aim here is not to provide a detailed evaluation of the degree to which the ACA will expand health insurance coverage in the United States.17

16. The Patient Protection and ACA of 2010 (P.L.111- 148) can be found at http://frwebgate.access.gpo.gov/cgi- bin/getdoc.cgi?dbname=111_cong_bills&docid=f:h3590 enr.txt.pdf and the Health Care and Education Recon- ciliation Act of 2010 can be found at http://frwebgate. access.gpo.gov/cgi-bin/getdoc.cgi?dbname=111_cong_bills &docid=f:h4872enr.txt.pdf. More detailed reviews of the provisions of this law for employers can be found elsewhere, such as Simon (2010).

17. For instance we do not address other aspects of reform that may impact costs of insurance or medical care

Instead, our ACA-based example is used to make the broader point that income measures used as social success indicators in public pol- icy discussions that do include the value of employer- or government-provided health insur- ance will not, by definition, register the success of public policies that expand health insur- ance coverage. There are 46.3 million uninsured Americans in 2008, according to the Census Bureau estimates (U.S. Census Bureau 2009). But because Census Bureau measures of house- hold size-adjusted, pre-tax, post-transfer, in-cash income do not include the value of health insur- ance, they are not useful for evaluating efforts to improve health-care coverage. Hence, while the ACA is expected to reduce the number of uninsured by almost 30 million (CBO 2010), the resulting changes in resources now available to them can only be tracked by an expanded income measure, such as the one we include here.

There are several analyses of potential im- pacts of the ACA, although none has focused on its distributional consequences. The CBO has estimated the number of Americans likely to be covered by specific provisions of the bill, as well as their cost, over a 10-year horizon. Rele- vant provisions are estimated to increase the rate of insured, non-elderly Americans to between 92% and 94% (CBO 2010) through a combi- nation of “carrot and stick” techniques. Others have criticized these numbers and argued that greater losses in employment or in employee- provided coverage are likely to reduce these projected CBO numbers (Holahan and Garrett 2011; Holtz-Eakin and Smith 2010; Kessler 2011). But there is very little in the peer- reviewed literature thus far simulating impacts of the ACA on overall health insurance cov- erage and none measuring its actual impact because the relevant provisions will not fully be rolled out until 2014. A recent paper by Par- ente et al. (2011) simulates the impact of hav- ing a state versus a national individual health insurance market on the take-up of coverage, and concludes that allowing cross-state sales of insurance would lead to a greater num- ber of individuals with insurance. Abraham and Feldman (2010) simulate the willingness to purchase individual market insurance among those whose employers might drop coverage

for those already insured nor do we address the extent to which the newly covered were receiving free or subsidized care from emergency rooms or hospitals despite not having insurance.

790 CONTEMPORARY ECONOMIC POLICY

as a result of growing premiums and future implementation of health reform. Burkhauser, Lyons, and Simon (2011) show the sensitivity of estimates of the size of the ACA’s crowd-out effects of employer-provided health insurance to whether the affordable coverage definition for employer-provided health insurance is based on a single or family coverage definition for work- ers with families.

Using our newly created measure of total income that includes the value of employer- and government-provided health insurance coverage, we are able to measure the impact of health reform using this extension of the income distri- bution literature. But, given the considerable dis- agreements regarding the specific effects of the ACA, we do so via a highly stylized example in which we assume full take-up of the main insur- ance provisions, ignoring “crowd-out” behavior, changes in the efficiency of the medical care system from the law, the sources of financing for the provisions (e.g., higher Medicare taxes, increases in the federal income tax), and other tax implications. A specific analysis using our measures to fully address the implications of health reform for the distribution of incomes and wealth is beyond the scope of this paper, and is left for future research.

Here we will use our measure to show how two key ACA policies — the expansion of Med- icaid to those with incomes below 133% of the federal poverty line, and the provision of pub- licly funded subsidies for health insurance pur- chased on exchanges by those without afford- able employer-provided health insurance with family incomes between 133% and 400% of the poverty line — would affect the level and distribution of income within a highly styl- ized setting. We first identify currently unin- sured individuals who would be covered by Medicaid expansions (those living in families with incomes under 133% of the official fam- ily poverty line) assuming a take-up rate of 70%.18,19 We assign the average value of current

18. Remler and Glied (2003) place the range of take-up of public insurance by the uninsured (from prior Medicaid expansions for children) at 50 – 70%. We use a 70% value because unlike past expansions, the ACA includes fines for those who do not purchase affordable health insurance, but we also use a 50% take-up rate later.

19. Thus far, our analysis has focused on house- hold sharing units, not families. We will continue to do so following the conventions in the inequality lit- erature, but note here that the ACA refers to family income relative to the federal poverty level when defining subsidies.

Medicaid services to these individuals as is done in the baseline model. Note that this also assumes no behavioral changes (such as moves away from jobs with health insurance to accept Medicaid coverage, which could thus lead to higher wages). We also do not model the extent to which individuals with private coverage who will be under the new Medicaid threshold may take up public coverage instead.

To model the effects of the subsidies for cov- erage, we first identify families whose incomes are under 400% but above 133% of the poverty line. In these families, we identify individuals who are uninsured, and add to them the statu- tory subsidy amounts assuming 70% take-up rates (with calculations also performed under a 50% take-up rate). In terms of the categories we follow, we add these subsidy amounts to the public insurance category. We use the estimates of the subsidy amount provided by the Kaiser Foundation website’s subsidy calculator for the ACA.20 As we treat the subsidy program in a similar fashion to the Medicaid expansion, we add the subsidy estimates we obtain from the Kaiser Foundation to each person in the fam- ily (adults as well as children) as if they pur- chase the exchange coverage as individuals and then aggregate amounts to a family and house- hold basis.

In Table 4 we provide an example of the usefulness of including the value of employer- and government-provided health insurance in a measure of income for policy evaluation. We show the distributional consequences of two of ACA’s expansionary health insurance policies — the extension of Medicaid coverage from 100% to 133% of the poverty line and the availability of subsidized exchange coverage for those living in families that are between 133% and 400% of the poverty line.

20. We use the subsidy amounts for a single person, with no employer coverage available, within a medium-cost area. Although there are some exceptions, in general those with existing employer coverage would not qualify for the subsidies. Since the subsidy calculator shows amounts that are for people of ages 20, 30, 40, 50, and 60 years, we grouped those up to age 24 into the first category, 25 – 34 into the second, 35 – 44 into the third, 45 – 54 into the fourth, and 55 – 64 years into the fifth category. These are meant to be illustrative numbers, rather than precise estimates of what the ACA will do. See the web link provided for addi- tional details about the assumptions made in the calculator. Since these numbers are in 2009 terms, we adjust them to 2008 terms using the July 2008 to July 2009 CPI. See: http://healthreform.kff.org/SubsidyCalculator .aspx.

BURKHAUSER ET AL.: VALUING HEALTH INSURANCE 791

TABLE 4 Distributional Effect of ACA by Decile (2008), 70% Take-Up Rate

Decile

Total Health Insurance

Before Policy

Health Insurance from Medicaid

Expansion

Health Insurance Premium Subsidies

Total Income After Policy

Change in Total Income from Policy

Change in Total Income from Policy:

Percentage

1 9,437 781 4 10,222 785 8.32 2 17,800 581 204 18,584 784 4.41 3 23,419 138 608 24,165 746 3.19 4 29,223 89 372 29,684 461 1.58 5 35,469 69 204 35,742 273 0.77 6 42,707 51 109 42,867 160 0.37 7 51,227 29 58 51,314 87 0.17 8 62,025 20 23 62,068 43 0.07 9 79,038 16 12 79,067 28 0.04

10 145,231 7 9 145,248 17 0.01 Mean 49,554 178 160 49,893 338 0.68 Gini 0.400 0.394 Gini — 50% take-up 0.400 0.395

Notes: See notes to Table 1. Policy changes is a Medicaid expansion from 100% to 133% of the family size-adjusted poverty line and a health insurance premium subsidy for those persons without affordable employer-provided health insurance living in families that are between 133% and 400% of the family size-adjusted poverty line, using a 70% take-up rate for both Medicaid and the subsidies.

Source: Authors’ calculations from the CPS.

Standard Census Bureau measures of in- come — household size-adjusted, pre-tax, post- transfer, in-cash income — will not capture the value of these expansions of health insurance coverage either via subsidies for exchange cov- erage or Medicaid. Using the results we report in Table 1 as a base, we are able to show the con- sequences of these ACA-based changes using our broader income measure which includes the value of health insurance. The first col- umn of Table 4 reports the mean total income, including the ex ante value of all employer- and government-provided health insurance by decile before the policy change. This measure of total income comes from column 7 of Table 1. Columns 2 and 3 of Table 4 report the additional value of health insurance from the Medicaid expansion and the exchange subsidies, respec- tively — assuming that Medicaid coverage and the subsidies are randomly assigned to 70% of those non-elderly who would be eligible if the ACA policies had been implemented in 2008. The Medicaid expansion primarily targets the bottom two deciles while the exchange subsidies target somewhat higher deciles in the distribu- tion. The mean values of the two policies are about the same. We repeat this exercise in an unreported table assuming a 50% take-up rate to examine the sensitivity of our results. Using this lower take-up rate reduces the magnitude

of the change but the qualitative story is similar.

Column 4 of Table 4, the sum of the first three columns, reports the new mean total income amounts by deciles. These increases occur solely because of the increase in the ex ante value of health insurance after the pol- icy. Both these ACA-based expansions of health insurance would primarily increase the mean total income of the bottom three deciles of the distribution. For the lowest decile, total income would change from $9,437 to $10,222, a mean change of $785 (column 5) or 8.3% (column 6). Total income increases by 4.4% in the second decile and 3.2% in the third decile.

The impact of these reforms on overall income inequality measures can be seen in the bottom two rows of Table 4, which contains the Gini coefficients. If enacted in 2008, the reforms would have reduced inequality using our Gini value by about 1.6% under the 70% take-up assumption, from 0.400 to 0.394.

VI. DISCUSSION AND CONCLUSION

Because health insurance in most industrial- ized countries is universally provided by gov- ernment, measures of inequality in those coun- tries can focus on wage or income inequality

792 CONTEMPORARY ECONOMIC POLICY

without greatly distorting trends in their actual level or distribution of yearly resources. But in the United States, where health insurance is the most important component of non-wage compensation but is unequally distributed across employers, and where government-provided health insurance makes up a major and grow- ing component of our social safety net, mea- sures of inequality that focus on cash wages or pre-tax, post-transfer, in-cash income will distort both levels and trends in the resources available to Americans over the period of our analysis and are incapable of measuring the impact of health insurance reform on that distribution.

In this paper, we show the sensitivity of trends in the level and distribution of mea- sured income when the value of employer- provided health insurance and government- provided Medicare and Medicaid are included in our fuller measure of income. We do so by con- structing measures of inequality of income based on the public-use CPS before and after adding the ex ante value of employer- and government- provided health-care insurance (Medicare and Medicaid) to the household size-adjusted, pre- tax, post-transfer, in-cash income of all Ameri- cans for the years 1995 – 2008. Doing so not only increases the income of the average (median) American household, but, because the amount of these ex ante values has been rising over time, pushes median income above its 2000 peak by 2007, the year before our most recent reces- sion. This also reduces income inequality mea- sured either by changes in the mean income of deciles or by Gini coefficients. While includ- ing health insurance reduces inequality within all age groups, it does so most for those aged 63 and over among whom the ex ante value of health insurance is large and evenly dis- tributed. When we use our expanded measure of income, in a stylized example, to illustrate the effects on measured income of two of the ways that the ACA of 2010 will expand health insurance coverage — Medicaid expansion and health insurance exchange subsidies — we find the benefits will primarily go to Americans liv- ing in the lowest-income deciles and will reduce income inequality. This is despite the fact that the two expansions are incremental, operating on top of an existing program of Medicaid coverage for certain low-income groups. While these estimates abstract from other features that should be included in a comprehensive analy- sis of the ACA legislation on the distribution

of income, they illustrate the point that mea- sures such as ours that value employer- and government-provided health insurance are use- ful in capturing the impact of health reforms on the levels and distributions of income.

VII. POSTSCRIPT

In July 2012 a few weeks before we received the proofs for this paper, the CBO issued a report measuring trends in household income (CBO 2012) that for the first time uses the insurance value of Medicare and Medicaid in its estimations of the value of these programs to recipients. In their discussion of their change in methodology from the methodology in the October 2011 CBO report discussed in this paper (CBO 2011), they refer to the arguments in our paper as part of their justification for doing so.

REFERENCES

Abraham, J., and R. Feldman. “What Will Happen If Employers Drop Health Insurance? A Simulation of Employees’ Willingness to Purchase Insurance in the Individual Market.” National Tax Journal, 63, 2010, 191 – 213.

Atkinson, T., and A. Brandolini. “Promise and Pitfalls in the Use of ‘Secondary’ Data-Sets: Income Inequality in OECD Countries as a Case Study.” Journal of Economic Literature, 39(3), 2001, 771 – 99.

Burkhauser, R., and K. Simon. “Who Gets What from Employer Pay or Play Mandates?” Risk Management and Insurance Review, 11(1), 2008, 75 – 102.

Burkhauser, R. V., S. Feng, and S. Jenkins. “Using a P90/P10 Ratio to Measure Inequality Trends with the Public Use Current Population Survey: A View from Inside the Census Bureau Vaults.” The Review of Income and Wealth, 55(1), 2009, 166 – 85.

Burkhauser, R., S. Lyons, and K. Simon. “The Importance of the Meaning and Measurement of ‘Affordable’ in the Affordable Care Act.” NBER Working Paper, No. 17279, 2011.

Burkhauser, R. V., S. Feng, S. Jenkins, and J. Larrimore. “Trends in United States Income Inequality Using the Internal March Current Population Survey: The Importance of Controlling for Censoring.” Journal of Economic Inequality, 9(3), 2011, 393 – 415.

. “Recent Trends in Top Income Shares in the USA: Reconciling Estimates from March CPS and IRS Tax Return Data.” Review of Economics and Statistics, 94(2), 2012, 371 – 88.

Burkhauser, R., J. Larrimore, and K. Simon. “A ‘Second Opinion’ on the Economic Health of the American Middle Class.” National Tax Journal, 65(1), 2012, 7 – 32.

Burtless, G., and P. Svaton. “Health Care, Health Insur- ance, and the Relative Income of the Elderly and Nonelderly.” CRR WP 2009-10, 2009. Accessed August 16, 2012. http://www.brookings.edu/papers/ 2010/0616_health_insurance_ burtless.aspx.

BURKHAUSER ET AL.: VALUING HEALTH INSURANCE 793

. “Health Care, Health Insurance, and the Distribu- tion of American Incomes.” Forum for Health Eco- nomics & Policy, 13(1), 2010. Article 1. Accessed August 16, 2012. http://www.bepress.com/fhep/13/1/1.

Carrigan, W. J., K. McCue, and B. Pierce. “Non Discrimi- nation Rules and the Distribution of Fringe Benefits.” Journal of Labor Economic, 2(20 pt 2), 2002.

Chung, W. “Fringe Benefits and Inequality in the Labor Market.” Economic Inquiry, 41(3), 2003, 517 – 29. Accessed August 16, 2012. http://www3.interscience. wiley.com/cgi-bin/fulltext/120806586/PDFSTART.

Congressional Budget Office. “Letter from CBO Direc- tor Elmendorf to Nancy Pelosi. March 20 2010.” 2010. Accessed August 16, 2012. http://www.cbo.gov/ publication/21351.

. “Trends in the Distribution of Household Income Between 1979 and 2007,” 2011. Accessed August 16, 2012. http://cbo.gov/publication/42729.

. “Trends in Distribution of Household Income and Federal Taxes, 2008 and 2009.” 2011. Accessed August 16, 2012. http://www.cbo.gov/sites/default/ files/cbofiles/attachments/43373-AverageTaxRates_ screen.pdf.

Garfinkel, I., L. Rainwater, and T. Smeeding. “A Re- examination of Welfare States and Inequality in Rich Nations: How In-kind Transfers and Indirect Taxes Change the Story.” Journal of Policy Analysis and Management, 25(4), 2006, 897 – 919.

Garfinkel, I., T. M. Smeeding, and L. Rainwater. The Amer- ican Welfare State: Laggard or Leader. Oxford, UK: Oxford University Press, 2010.

Gottschalk, P., and T. M. Smeeding. “Cross-National Com- parisons of Earnings and Income Inequality.” Journal of Economic Literature, 35(2), 1997, 633 – 87.

Holahan, J., and B. Garrett. “How Will the Affordable Care Act Affect Jobs?” 2011. http://www.urban.org/health_ policy/url.cfm?ID=412319.

Holtz-Eakin, D., and C. Smith. “Labor Markets and Health Care Reform: New Results.” American Action Forum. May 2010. Accessed August 16, 2012. www. americanactionforum.org. http://americanactionforum. org/files/LaborMktsHCRAAF5-27-10.pdf.

Jenkins, S. “SVYLORENZ: Stata Module to Derive Distribution-Free Variance Estimates from Complex Survey Data, of Quantile Group Shares of a Total, Cumulative Quantile Group Shares.” Boston College Department Of Economics Stasitical Software Com- ponents Number S456602, 2006. Accessed August 16, 2012. http://ideas.repec.org/c/boc/bocode/s456602.html.

Kessler, D. “How Health Reform Punishes Work.” Wall Street Journal. 2011. Accessed August 16, 2012. http:// online.wsj.com/article/SB10001424052748704628404 576265692304582936.html.

Kolstad, J. T., and A. E. Kowalski. “Mandate-Based Health Reform and the Labor Market: Evidence from the Massachusetts Reform.” National Bureau of Eco- nomic Research Working Paper 17933, 2012. Accessed August 16, 2012. http://papers.ssrn.com/sol3/papers. cfm?abstract_id=2025309.

Larrimore, J., R. V. Burkhauser, S. Feng, and L. Zayatz. “Consistent Cell Means for Topcoded Incomes in the Public Use March CPS (1976-2007).” Journal of Economic and Social Measurement, 33(2-3), 2008, 75 – 102.

Levy, H. “Health Insurance and the Wage Gap.” NBER Working Paper, No. 11975, 2006. Accessed August 16, 2012. http://www.nber.org/papers/w11975.

Marical, F., M. M. d’Ercole, M. Vaalavuo, and G. Verbist. “Publicly Provided Services and the Distribution of Households’ Economic Resources.” OECD Economic Studies, 44(1), 2008.

Meyer, B. D., and J. X. Sullivan. “Measuring the Well- being of the Poor Using Income and Consump- tion.” Journal of Human Resources, 38(Suppl), 2003, 1180 – 220.

Meyer, B., and J. X. Sullivan. “Five Decades of Consump- tion and Income Poverty.” NBER Working Paper No. w14827. National Bureau of Economic Research, Cambridge, MA, 2009.

Nyman, J. A. “The Value of Health Insurance: The Access Motive.” Journal of Health Economics, 18, 1999, 141 – 52.

Parente, S. T., R. Feldman, J. Abraham, and Y. Xu. “Con- sumer Response to a National Marketplace for Individ- ual Health Insurance.” Journal of Risk and Insurance, 78(2), 2011, 389 – 411.

Paulus, A., H. Sutherland, and P. Tsakloglou. “The Distribu- tional Impact of In-Kind Public Benefits in European Countries.” Journal of Policy Analysis and Manage- ment, 29(2), 2010, 243 – 66.

Pierce, B. “Compensation Inequality.” Quarterly Journal of Economics, 116(4), 2001, 1493 – 525.

. “Recent Trends in Compensation Inequality,” in Labor in the New Economy, edited by K. G. Abraham, J. R. Spletzer, and M. Harper. Chicago: University of Chicago Press, 2010. Accessed August 16, 2012. http://www.nber.org/chapters/c10814.

Piketty, T., and E. Saez. “Income Inequality in the United States, 1913 – 1998.” Quarterly Journal of Economics, 118(1), 2003, 1 – 39.

Remler, D. K., and S. A. Glied. “What Can the Take-Up of Other Programs Teach Us about How to Improve Take- Up of Health Insurance Programs?” American Journal of Public Health, 93, 2003, 67 – 74.

Ruggles, P. Drawing the Line: Alternative Poverty Measures and Their Implication for Public Policy. Urban Institute Press, Washington, DC, 1990.

Simon, K. “Implications of Health Care Reform for Employers: An Analysis of the Patient Protection and Affordable Care Act.” Prepared for the Center for American Progress. 2010. Accessed August 16, 2012. http://www.americanprogress.org/issues/2010/05/pdf/ health_employers.pdf.

Smeeding, T. M., P. Saunders, J. Coder, S. Jenkins, J. Fritzell, A. J. M. Hagenaars, R. Hauser, and M. Wolf- son. “Poverty, Inequality, and Family Living Standards Impacts across Seven Nations: The Effects of Noncash Subsidies for Health, Education, and Housing.” Review of Income and Wealth, 39(3), 1993, 229 – 56.

Stewart, K. J., and S. B. Reed. “CPI Research Series Using Current Methods, 1978-98.” Monthly Labor Review, 122(6), 1999, 29 – 38.

U.S. Census Bureau. Measuring the Effect of Benefits and Taxes on Income and Poverty: 1992. Current Popula- tion Reports, P-60 series. Washington, DC: GPO, 1993. Accessed August 16, 2012. http://www.census.gov/ hhes/www/poverty/prevcps/p60-186rd.pdf.

. Income, Poverty, and Health Insurance Coverage in the United States: 2008. Current Population Reports, P-60 series. Washington, DC: GPO, 2009.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article:

APPENDIX S1. Creation of the Health Insurance Values Database and Details on the Current Population Survey Data Set TABLE S1. Comparisons of Our Estimates of Median Income by Household with Those Reported by Census

794 CONTEMPORARY ECONOMIC POLICY

TABLE S2. Compensation in Health Insurance Policies (Real 2008 Dollars) Means TABLE S3. Health Insurance in the Current Population Survey TABLE S4. Mutually Exclusive Health Insurance Status, Constructed Data Set

TABLE S5. Health Insurance Status for 2008, by Decile of Income

Copyright of Contemporary Economic Policy is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.