The Congressional Budget Office, Jonathan Gruber and others attempting to forecast the effect of the Affordable Care Act (ACA) on the propensity of employers to offer health insurance have gotten the microeconomics wrong.
Background: Beginning in 2014, employers who offer affordable health insurance will thereby render their employees ineligible for health insurance tax credits that can be large as $15,000 per family per year. All analysts agree that there are SOME employers who will react to this situation by dropping their coverage, and that their employees will benefit by obtaining the subsidies. The debate is about the size of this effect and whether it would large enough to offset other factors that might be encouraging employers to offer coverage.
In one way or another, the answer to the question is ultimately found through empirical analysis. One could wait until say, 2015, and measure what happened. For those of us who want a forecast before then, we must somehow relate historical episodes to what the ACA will do in 2014.
The approach of CBO, Professor Gruber (see 27:02 in this video), and others has been to (a) think of a "demand" for employer-sponsored insurance (hereafter, ESI) and (b) look at the historical sensitivity of that demand to the price of ESI, especially variation associated with the ESI subsidy implicit in the exclusion of ESI premiums from payroll taxes and employee income taxes. Item (b) is acceptable enough -- I agree that subsidies impact prices -- but item (a) is fundamentally flawed because the wrong demand curve has been identified.
Historically, there has not been a viable non-group insurance market, so that employers dropping their coverage are in effect asking a significant fraction of their employees to go uninsured or to apply for Medicaid. The demand curves traced out by such episodes are telling us about the distribution of employee preferences for being uninsured (or on Medicaid) and the distribution of ESI administration costs (broadly defined to include insurance loadings and other factors). Economic theory does not tell us the precise shape of these distributions, so it's nice to have examined the historical episodes, which suggest that the elasticity of the propensity to offer ESI with respect to the price of ESI is roughly -0.5. In other words, historical increases in ESI prices in the amount of 10% have caused roughly 5% of employers to drop coverage.
But, under the ACA, dropping ESI does not mean leaving the employees uninsured. The propensity of employers to drop ESI under the ACA therefore has little to do with the distribution of employee preferences for being uninsured or for participating in Medicaid. Thus, to a first approximation, the -0.5 elasticity cited above is irrelevant. (More precisely, the magnitude of the historical elasticity is probably a conservative lower bound on the magnitude of the elasticity relevant for ACA forecasts, because going uninsured is a feasible but unlikely choice for employees who lose ESI).
What we really need to know is the distribution of employee preferences to participate in the ACA's "exchange" plans, and the distribution of ESI administration costs as compared to exchange administration costs (again, broadly interpreted to include insurance loadings, etc.). Health economists have not studied that yet.
You might think that the distribution of ESI administration costs would be relevant for the ESI-exchange margin, but the exchanges will have administrative costs too. The application for exchange plans is 21 pages long: employers who drop ESI may well replace it by helping their employees with application and other administrative tasks, much like employers help employees with immigration paperwork. From this perspective, ESI and exchange participation look like very close substitutes and the 0.5 price elasticity magnitude looks like a wild underestimate.
Although the exchange plans will not be called "Medicaid," you might think that the distribution of preferences for Medicaid participation could be important because people will attach some of the same stigma to exchange participation as they do to Medicaid. I would agree with you if the purpose was to quantify the effect of Massachusetts' health reform on ESI coverage, because premium support was made available in Massachusetts primarily through its "Commonwealth Care" program which was a collection of 4 Medicaid managed plans (one plan has been added since). In contrast, recipients of federal ACA subsidies will be receiving cash they can spend on a plan of their choice, one or two of which will be the plans in which their state's U.S. Senators are enrolled: that’s a lot closer substitute for ESI than Medicaid is.
Because they have helped uncover distributions of administration costs and preferences for being uninsured, the historical studies are better suited to predict the number of people who will change from uninsured to insured as a consequence of the ACA, as RAND has done. But that's very different from predicting how many people move from ESI to non-group policies. Even without considering labor-market equilibrium effects or general equilibrium effects of the ACA on ESI coverage, we can already see that demand analysis has been misapplied in ACA forecasting and that the propensity of the ACA to reduce ESI coverage has been underestimated.