• How to really predict employer offers of health insurance

    The McKinsey “study” that loudly predicted a huge decline in employer-sponsored health insurance in response to the Affordable Care Act is back in the news today (e.g., NY Times, NPR, LA Times) because McKinsey finally released their methodology.  It turns out they did a market research survey of executives.  This approach has been widely criticized as inaccurate, so perhaps it’s a good time to ask how these kinds of predictions can be made more credibly.

    Start by asking why employers sponsor health insurance for their employees at all.  The answer is that employer-sponsored health insurance is not taxed, so a dollar contributed to health insurance premiums buys a dollar of insurance while a dollar devoted to wages translates to less than a dollar of take-home pay.  As an employer, if I devote a portion of my compensation budget to health insurance and my competitor doesn’t, the dollar value of total compensation at my company will be greater than at my competitor’s.  I’ll attract the best workers.  So employers sponsor health insurance because the labor market is competitive.  They might wish they could cut these costs or drop health benefits entirely, just like they’d like to cut wages, but they have to consider the realities of the labor market or they won’t be able to hire.

    So how does the ACA change the labor market?  It does several big things: 1) It expands Medicaid, making more low-income people eligible, 2) It creates insurance exchanges through which individuals can obtain (subsidized) insurance if they don’t have a qualifying employer offer, and 3) It gradually reduces the tax exclusion for employer-sponsored insurance, starting with the most expensive policies.  It does lots of other things too, but I think these will have the biggest effects on the labor market.

    It turns out that there are well established empirical methods for predicting the effects of Medicaid expansion and changes in tax rates (see for example, Gruber and Simon (2008) and Bernard and Selden (2002)).  Predicting the effects of the exchanges is harder, but the fact that access will be limited to those without employer offers simplifies things somewhat.  Most workers will not be eligible for subsidies if they had access to exchanges and the tax benefit is a major factor, especially for higher income workers, so a firm that drops coverage will be cutting compensation significantly for most of its workers.  It’s not likely that many firms will be able to do this unilaterally.

    All of these issues are discussed in more detail and accompanied by a forecast that shows very modest changes in employer behavior in “The Effect of Health Reform on Public and Private Insurance in the Long Run,” by Pizer, Frakt and Iezzoni , March 2011 (ungated working paper available).

    References

    Gruber J, Simon K. Crowd-out 10 years later: have recent public insurance expansions crowded out private health insurance? Journal of Health Economics 2008; 27 (2): 201-17.

    Bernard D, Selden TM. Employer Offers, Private Coverage, and the Tax Subsidy for Health Insurance: 1987 and 1996. International Journal of Health Care Finance and Economics 2002; 2 (4): 297-318.

     

    Share
    Comments closed
     
    • So on the one hand we have a prediction model (based on “well-established empirical methods”) and on the other hand we have a survey that asked a sample of people who will actually be making the decision to say what they would do. The model and survey results don’t agree (by a wide margin). Model v. survey–it takes real academics to poo-poo the results of the survey.

      While one can argue about potential bias of the questions, or whether the HR people surveyed would actually decide the way they say the would, ignoring the fact that the people who actually will make the decisions are not even close to the model’s prediction indicates that there is likely a factor (or factors) the model is missing.

      Forcing data to fit a model is not science.

      • Not necessarily. Two of the questions in the survey asked takers to estimate how much the company spent on health benefits per employee and what percentage of the total premiums the company paid for their employees. Nearly three out of every five people surveyed said they didn’t know the cost of the health benefits (including more than one of every two primary decision makers), and a little over one out of every three said they didn’t know the percentage of premium covered by the company. If they don’t know what their company is paying, then how are they supposed to make even semi-plausible judgments on whether ending the employer-sponsored insurance plan will save them money?

        • Generally, executives will float along with the status quo in areas that seem to be relatively stable while spending most of their time putting out operational fires. So I am not surprised that most of them did not have an exact understanding of health benefit costs. When the environment changes, such as with the ACA, more attention is placed on the specific issues and costs.

          What is of concern is that when decision-makers were presented with some facts about the ACA (perhaps somewhat biased but not excessively so), the number who reported they might drop coverage was much higher than the models predicted.

          Of course, after a deeper dive most of these executives might stick with the status quo, there is, after all, a strong bias against change. But when a significant number are entertaining change, there is also the possibility that a wave of change might sweep through businesses.

          Bottom line is that the survey results raise the level of uncertainty, and hand-waving is not going to reduce it. Perhaps some academics could get out and talk with the decision makers?