I used the most recent Annual Social and Economic Supplement (ASEC) to the Current Population Survey (CPS) to produce the following figures, which I think provide unique insights into the state of poverty and health insurance in America.

For the below graphs, each person in the CPS ASEC is assigned an income equal to their family’s income divided by the square root of their family size. I use the family unit definition provided by the Supplemental Poverty Measure (SPM). I define market income as gross cash income from non-governmental sources.¹ I define disposable income using the SPM disposable income concept, which is equal to market income plus cash and cash-like government benefits minus taxes, medical out-of-pocket expenses, child care expenses, work expenses, and child support paid.²

I use the incomes that are recorded in the CPS ASEC because that is the best data source available to me. But we know that these incomes are not wholly accurate as they underreport certain incomes such as food stamps while overreporting other incomes likes the Earned Income Tax Credit and the Child Tax Credit. So just keep that caveat in mind as we proceed below.

The Poor

In order to determine the types of people who are clustered at the bottom of the income distribution (“the poor”), I assigned every individual to one of nine categories. Where an individual is in more than one category — such as a disabled full-time worker — they are assigned to the category that is the lowest in the list.

Most of the categories are self-explanatory, but a few need a bit of elaboration. A full-time worker is someone who was employed the entire year and usually worked 35 or more hours a week. Other workers are people who do not meet the definition of a full-time worker but put in at least one hour of market labor during the year. This therefore means that every other category on the list contains people who did not work at all during the year. They are categorized either by their age or their reason for not working.

Here is what this graph looks like for the market income distribution.

At the very bottom of the market income distribution, there are virtually no workers. Elderly and disabled people, many of whom are on Social Security benefits, make up around 70 percent of the very bottom, with kids making up another 10 percent. As you move up the distribution, the share of people who are working full time marches upward with a slight dip at the very top.

Here is what the same graph looks like for the disposable income distribution.

The main difference between the market income distribution and the disposable income distribution is where the elderly and disabled are. They are overrepresented at the bottom in both distributions, but much less so in the disposable income distribution than the market income distribution. This is a testament to the success of the old-age and disability benefits provided by the Social Security and Supplemental Security Income programs.

The bottom of the disposable income distribution has some workers in it, but still relatively few, and very few full-time workers. When looking at this distribution, it is clear that a program of family benefits — including free child care, a home care allowance, paid leave, and a child allowance — is the best anti-poverty strategy at our current moment as it would provide benefits to the large chunk of children and carers at the bottom as well as to the workers who live with them.

In general, what I like about these two visualizations is that they make it very easy to see why social democratic welfare states are so successful at eliminating poverty, inequality, and income insecurity.

These welfare states levy broad-based taxes on all workers and then use those taxes to fund universal benefits for every category of nonworker. This fights poverty because the poor are almost all nonworkers or people who live with nonworkers. It fights inequality because the share of workers (and therefore taxes paid) increases as you go up the income distribution while the share of nonworkers (and therefore benefits received) decreases as you go up the income distribution. And it fights income insecurity by insuring that even high-earners are cushioned from income shocks when they have children or find themselves unemployed, disabled, or otherwise out of work.

The means-testing alternative to the social democratic welfare state is bad not just because the politics of it are so fraught but because it actually is a substantively worse way to achieve egalitarianism. The universal social democratic approach is not a blunt wasteful approach but is instead a very elegant and effective design once you understand how the population is actually distributed across the income distribution.

The work-focused approach to egalitarianism, which imagines that the bottom of the income distribution is full of low-wage workers and the unemployed, instead of the nonworkers it is actually full of, is of course even more ridiculous than the means-tested welfare state approach.

Health Insurance

Using the same methods above, we can also see what kinds of insurance are present across the income distribution.

Here is how often various types of health insurance are found across the market income distribution.

Given what we know from the graphs above, it is of course not surprising that Medicare and Medicaid (including those who are eligible for both) dominate the bottom of the income distribution. Medicare is primarily for elderly and disabled people while Medicaid is primarily for low-income and disabled people. Despite what some might think, the poorest of the poor defined by market income are actually more likely to be on Medicare than Medicaid, again for the reason that this chunk of the income distribution is dominated by elderly and disabled people.

Here is the same graph for the disposable income distribution.

Once again, we see the huge impact of Social Security on the overall income distribution. Social Security moves a lot of elderly and disabled people up the income distribution and therefore moves Medicare recipients up the income distribution. This smooths out the distribution of Medicare across the disposable income distribution even though it is very concentrated at the bottom in the market income distribution.

Another effect of Social Security is that, in the disposable income distribution, uninsurance is concentrated at the bottom. Under the market income distribution, it is concentrated between the 10th and 30th percentiles because Medicare-receiving elderly and disabled people dominate the bottom 10 percent.

The health insurance graphs help illustrate a few policy facts very vividly.

The first is that the Obamacare exchanges are a miniscule part of the health insurance system. Relative to expectations, they have been a major policy bust.

The second is that Medicaid is actually bigger than Medicare at this point, in terms of how many people are enrolled in it. This only recently became the case due to the Medicaid expansion under Obamacare, which has been a major policy success (at least as far as enrollment is concerned). Going forward, I hope we can do more things like Medicaid expansion and less things like the Obamacare exchanges.

The third is that Medicaid is not exclusively a poor-people program. People misunderstand this because Medicaid eligibility is generally discussed in terms of what your income is as a percent of the poverty line. But Medicaid eligibility is based on short-term tax-unit income, which is a different thing from annual family-unit income. Rich family units can contain multiple tax units, with some tax units poor enough to qualify for Medicaid even though they are not actually poor in terms of family income. Some disabled people are also eligible for Medicaid in a way that does not subject them to the Medicaid income test.

It is important to remember that these graphs are a point-in-time picture of health insurance. It therefore does not reflect the fact that people lose their health insurance constantly and move between different insurance statuses (and between plans within a given insurance status) pretty regularly. Medicaid in particular is subject to an extreme amount of insurance churn, with over 1 in 3 Medicaid participants coming off of Medicaid within a year, often into uninsurance.

When it comes to cleaning up the total mess that is the health care system, the obvious answer continues to be the correct one: create a single national health insurance plan and enroll everyone into that plan. This would solve the problems of underinsurance, uninsurance, and insurance churn while saving hundreds of billions of dollars a year on health spending.


1. This is ASEC variable SPM_TOTVAL minus SS_VAL, SSI_VAL, UC_VAL, VET_VAL, WC_VAL, PAW_VAL. The individual _VAL variables were all aggregated to the SPM unit.

2. This is ASEC variable SPM_RESOURCES.