Debunking the EPI on the Cadillac tax

The Economic Policy Institute report titled "Employer Health Care Costs Do Not Drive Wage Trends" has become the definitive study for those who oppose a "Cadillac (excise) tax" on expensive group (employer-provided) health insurance plans.  The EPI report has been cited by prominent bloggers on websites like Daily Kos and MyDD to "debunk" the supposed "myth" that health care costs are not a factor in wage growth or decline.  Well, it turns out that EPI gets it wrong on several fronts on the data and interpretation of the data alone.  Their conclusions, to say the least are at best simply mistaken.

The EPI's report is based on three fundamental (and as I will demonstrate, mistaken) claims.

Claim One: "Health care costs are not large enough to substantially move wages."
It is easy to understand that health care cost trends have not been a significant driver of wage trends when one examines the scale of employer expenditures on health care. Health care costs were just 7.6% of total compensation and 9.4% of total wages (all wages paid, including premium pay, paid leave, and so on) in 2007. [emphasis mine]
Reading this, my first thought was a bit of disbelief.  Really?  Health care costs are just 7.6% of total compensation?  If that's true and the cost of health care is that small a portion of total compensation, why would employers sweat the extra tax on a few of their employees' policies - or even all of them, if they offer Cadillac plans to all employees and they all pick it?  Second question: if you're an employee whose employer provides health care and you know how much it costs, does it sound like less than 10% of your wages (at least the employer portion)?  This would mean that your annual wages are over $230,000 if you have a family plan eligible for the excise tax, or over $85,000 if you have an individual plan that qualifies for the excise tax, if your employer plays the entire premium, slightly less if you pay part of it but still substantial.

But here is the bigger question.  Is this true?  The answer: of course not.  You see, the EPI uses data from the Bureau of Labor Statistics (BLS).  So far so good.  And the data says that at the end of 2007, the employer cost for total compensation per hour per employee was $28.11 and total health insurance cost per hour per employee was $2.21.  Turns out that's actually 7.9%, but whatever.  By September 2009 (latest BLS data available), the number climbs to 8.1%.  But here's the real problem: this is the cost for all employers put together and then averaged.  But as we know, not all employers provide health coverage.  So there were tons of employees who made wages but no health benefits as part of their compensation and they are averaged within that 8.1%.  In fact, the Kaiser Foundation's most recent survey found that only 59% of the workers were covered by employer-sponsored health insurance in 2009.  So that 8.1% expenditure is spread among 59% of workers, not 100%.

Adjusting for this factor, for the employees who do have employer-provided coverage, the cost of health insurance per employee per hour climbs to 14% of total compensation in 2009.  And as a portion of wages only, it was 20% in 2009.  And that's for your employer's side of the cost alone.  It doesn't even count the employee contributions.  That's also just the average.  I would expect this percentage to be higher for employers who do provide high-premium plans.  Moreover, under the status quo, these trends are heading up.  Now does it look like health care costs are, as the EPI put it, "large enough" to move your wages?  By the way, that figure of the average employer plan costing 20% of your wages?  It matches up quite nicely if you consider that in 2008, the average family health plan cost $12,298, or about 18% of the median income for a family of 4 that year, $67,019.  Yes, I know I'm doing a little rough justice by using the mean for one number and the median for the other (sorry for all the wonk talk), but I can't find the median in 2008 for a family plan, so I'm having to assume the mean is the median there.

[Note: these numbers can be further adjusted if someone can find the wages and benefits specific to employees with health benefits rather than all workers. However, I don't think it would change the calculations much, and it is entirely unclear in which direction.]


All of their numbers have to be similarly adjusted for the report to make any proper numerical sense.  You cannot fail to factor in the severe problem of declining portion of Americans being covered by employer-based insurance.

EPI also claims that total benefits as a percentage of total are constant and that employers just move money around:
Further, overall benefits’ (health care plus all other fringe benefits) share of total compensation has actually been stable for the last 20 years or so—as health costs expanded, pension and payroll tax shares diminished.
This is true (well it's true for the part that we have data for among those 20 years, which is about 5, anyway) , according to the Bureau of Labor Statistics, which is where EPI's end-notes indicate they obtained the data from.  However, what are these benefits, aside from health insurance?  Apart from the legally required ones such as paying half your payroll taxes (not income taxes), here is a full list, with September 2009 percentage of total compensation in the parentheses:
  • Paid leave: vacation, holidays, sick and personal days off (6.9%)
  • Supplemental pay: overtime, shift differentials and bonuses (2.9%)
  • Retirement benefits [stop laughing!] (4.4%)
  • Other insurance: life, short and long-term disability (0.5%)
  • Other (0.1%)
It should jump out at anyone that the the top two items on this list are actually cash benefits!  These are things you receive in your paycheck.  You can either take a vacation or cash in your vacation time.  You get a check for either!  Overtime, shift differentials and bonuses are obviously cash.  If your health insurance cost can be kept in check, and as EPI suggests, your health insurance money is redistributed to your "other benefits", it is more than likely that you will get more cash, more take-home pay.  That is essentially a wage increase if your employer offers you more overtime or more time off.

Retirement benefits are also essentially cash.  They are deferred wages. 

So out of the non-health insurance, non-legally mandated benefits, 95% are cash or deferred wage benefits.  66% are direct cash benefits (paid leave and supplemental pay) that put money in your paycheck.

Also notice that last part: "as health costs expanded, pension and payroll tax shares diminished."  When do payroll taxes (i.e. social security, medicare, etc.) diminish?  When the payroll - i.e. your total income (wages, bonuses, paid leave, overtime, etc. - anything your employer has to pay a payroll tax on) diminishes.  So even if your 'wages' technically do not diminish, EPI is admitting that monetary compensation (deferred, i.e. pension, or current, i.e. paid leave, bonuses etc.) do in fact dip downwards when health costs expand.  Note that this is not indicative of lay-offs, since these numbers are per employee per hour.

In addition once again, there is the data problem.  There are a significant amount of employers who fall into 3 categories:
  1. Employers that offer health insurance and other benefits such as a retirement plan, long term care insurance, etc.
  2. Employers whose only major benefit is health insurance
  3. Employer who offer no or next to no benefits at all
And a breakdown of how much employers in each of those categories spend on total benefits vs. wages would be interesting to see.  However, EPI does not give us any such thing.  As a result, we see the crowding out effect - firms which have no voluntary benefits being averaged with firms that have some or a lot of voluntary benefits.  It would stand to reason that firms that provide a lot of additional benefits beside health care would have to spend more than the 29-30% average noted in the BLS report, in order to balance out the firms that have none. 

EPI's report's second pillar: "Examination of actual wage and benefit trends confirms that changes in the trajectory of health care costs did not materially affect wage trends over the last 20 years."  To make this point, EPI uses this chart (partial chart below) to demonstrate how the costs of wages, benefits (specifically health benefits) have changed through three periods:


Once again, the EPI uses market-wide changes in wages and benefits without any regards to the fact that not every employee has benefits, and without consequently adjusting the numbers.  There is no way for me to redo these numbers without having the originals from which the EPI built this table.  The original shares of workers who had employer based coverage in those years is absolutely crucial, and that data is missing.  But based on my analysis above, I would not expect these numbers to be very reliable, except for the trend (i.e. when it went up and when it went up more).  The EPI is adding wage data, which every employee has, and pitting them together against health insurance premium data, which not every employee has.  Without accounting for this, the calculations cannot meet the test of reliability.

But, more importantly, this table showing only the changes in three periods over the span of 1989 to 2006 does not tell you anything about whether or not your total cash take-home pay would be more if health insurance costs were controlled.  At best, it tells you that cost of health insurance has been a drag on employers (and your take-home cash compensation) for a really, really long time.  That shouldn't come as a surprise to anyone, since in the current debate, some have suggested that Democrats should have made a deal with Nixon on health care back in the 1970s.  The health care situation that is a gigantic crisis threatening to blow a hole through our economic fabric did not start in the last couple of decades.  It has been accumulating over a very long time.

Granted, not the entirety of wage increases or decreases or stagnation can be accounted for by health insurance costs to one's employer, and the effects can be overstated if one is not careful.  But there is no way to avoid facing up to the idea that it is in fact a major factor.  Actually, EPI kind of demonstrates this for us:
But the lessons of the 2000s are also instructive: despite the faster productivity growth there has been no real wage growth recently,
Hmm, so in the 90s, productivity growth drove wage growth (and it did, but note that health care costs were also held to smaller increases), but no real wage growth in the 2000s despite faster productivity growth.  I wonder what might have contributed to that.  Oh look, between 2000 and 2006, of the three periods in the above table, health care costs rise the fastest.

EPI's third claim:
The wage behavior described—accelerating in the late 1990s and more slowly thereafter— actually best characterizes wage growth for low-wage workers who have minimal access to employer-based health care. Conversely, this pattern of wage-growth over time is least pronounced for higher paid workers with the most health coverage.
They also note that in the 2000s, it's the poor and the middle class whose wages were stagnated or reduced, not that of the high income earners.  According to the report, only 27% of workers in the bottom fifth of income earners and 64% of workers in the middle fifth had employer provided health insurance, whereas 80% of the top fifth of income earners had health insurance.  The data certainly bears this out.  During the boom of the 1990s, the poor and middle class had the most to gain.  In the 2000s, they had the most to lose.  But the conclusion EPI draws from it, namely that this means that wages are not significantly influenced by health care costs of the employer, is not obvious, nor, in my judgment, reasonable.  Just because higher income earners have the most coverage does not mean their wages are not suffering because of the costs.  I will demonstrate:

It should be noticed that we are looking at data about wages and what percent of what kind of income earners get employer provided coverage over an 18-year period, but we are not told whether coverage proportions in each income category held steady, rose, or declined. EPI does not reveal that for the middle-class, the availability of employer provided coverage has been on a solid downhill curve.  Their coverage is being dropped, or their portion of the premiums are growing, and small businesses are struggling to keep their employees on the payroll and provide them insurance.  Think I'm making it up?  An article by EPI's own Elise Gould in 2005 made this point clear as daylight:
Middle-income Americans between the ages of 25 and 54 were 26.7% more likely to be uninsured in 2004 than in 2000.
Indeed, the data included in that article shows an overall 2.9% decline in employer provided coverage between 2000 and 2004 alone, with the bottom fifth of workers' coverage declining by 2.9%, the second fifths' coverage declining by 4.9% and the middle fifth's coverage declining by 2.3%.  Even the top two fifths lost coverage in that period to the tune of 2.4% and 3%, respectively.  Also note from the first table in that report that blue collar workers lost health coverage at over 1.5 times the rate of white collar workers.  Let me point out that those declines were much bigger over the 18 year period EPI was discussing in its recent paper.  So, I do not really see how one can claim that health insurance premiums have no significant effect on wages while we are staring at data showing us that coverage has been declining among the high income earners and low income earners alike.  Yes, higher income earners remain the most likely to receive health benefits, it is undeniable that even for them, coverage has been on the decline.  If 20 years from now, only 50% of high income earners were covered and 10% of low income earners, I don't think that we can say then "hey, look, high income earners have the most coverage anyway, so health care costs are not affecting wages."

Another thing popped in my head when this argument came to my attention.  I will quote it once again for your convenience:
The wage behavior described—accelerating in the late 1990s and more slowly thereafter—actually best characterizes wage growth for low-wage workers who have minimal access to employer-based health care. Conversely, this pattern of wage-growth over time is least pronounced for higher paid workers with the most health coverage.
I stared at the highlighted parts for a while.  If this is true, does it not follow that EPI is fighting to protect the wealthier income earners from the effects of a tax that might be put on their insurance plans?  If it's the highest paid workers who have the most coverage, why the outrage about this affecting the middle class?  More to the point, isn't this trying to have it both ways -- indignantly protesting the excise tax on the claim that it will "hurt the middle class" and then turning around and saying "well, middle class wages aren't affected anyway because it's the high income earners who have the most coverage?"  I don't get it.  Shouldn't the Economic Policy Institute be fighting for more poor and the middle class people to get coverage and care instead of trying to protect those with the most coverage -- the high income earners, according to their own paper -- from a tax on their insurance company (or passed down, their employer) that might apply on some of their health plans over a certain threshold to their insurance companies?  Their mission sure seems to be to fight for the poor and the middle class:
The Economic Policy Institute, a nonprofit Washington D.C. think tank, was created in 1986 to broaden the discussion about economic policy to include the interests of low- and middle-income workers. Today, with global competition expanding, wage inequality rising, and the methods and nature of work changing in fundamental ways, it is as crucial as ever that people who work for a living have a voice in the economic discourse.

EPI was the first — and remains the premier — organization to focus on the economic condition of low- and middle-income Americans and their families.

EPI also felt it important to disclose their funding sources on their "About Us" page:
EPI is a 501(c)(3) corporation. In 2005 through 2007, a majority of its funding (about 53%) was in the form of foundation grants, while another 29% came from labor unions. EPI also receives support from individuals, corporations, and other organizations.
I must stress that my argument and this piece is based the substance, not the funding source of the issue brief I discuss here.  But I felt it important to mention given the recent funding fallouts that have contributed to a lot of attacks on the objectivity of MIT Prof. Jon Gruber.

So in the end, the EPI bases its campaign to oppose the excise tax based on three claims, each of which is flawed and/or deeply questionable, and given that, none of those lead to the conclusions the EPI claims to.  These certainly do nothing to undo the validity of the support of the excise tax by three Nobel winning economists (Paul Krugman among them), 21 other prominent economists, the CBO, the JCT, the Center for Budget and Policy Priorities, and economic policy columnists of several newspapers including the Washington Post's Ezra Klein who finds a stronger correlation between wage growth and the growth-rate of the cost of employer-provided coverage than between wage growth and economic expansion.  The EPI brief does not negate their judgment that there is a tradeoff between wages and the cost of health insurance benefits.  While the cost of health insurance isn't the only factor in determining and driving wages, it is one of the major factors, and the EPI paper fails negate it.

Technical Note: Emphases may be mine in quotes provided.