OPI report on Medicaid growth projections under Obamacare less than enlightening

October 26, 2011

This morning, the liberal Oklahoma Policy Institute (OPI) released a report challenging the findings of a May 2011 OCPA study that projects growth in the state’s share of Medicaid expenditures under Obamacare. Earlier this year, OCPA collaborated with Cato Institute senior economist Jagadeesh Gokhale, Ph.D., to undertake a comprehensive, ground-up study of the likely impact that the expansion of Medicaid under Obamacare will have on Oklahoma’s budget.

While I certainly appreciate and welcome OPI’s willingness to venture into this important conversation, their report is nothing more than reprint of absurd cost estimates by the Oklahoma Health Care Authority and the Congressional Budget Office and a summary of various studies by left-leaning organizations – the Kaiser Family Foundation, the Robert Wood Johnson Foundation and the ultra-liberal Urban Institute.

By contrast, OCPA and Cato did not pull from any other studies; instead, Gokhale started from scratch, looking at all the relevant factors that lead to Medicaid expenditure growth, analyzing data from apolitical databases (the U.S. Census Bureau, the Medicaid Statistical Information System State Datamart, the Oklahoma Health Care Authority, the Bureau of Labor Statistics, etc.) and projecting historical growth trends into the future.

The OPI report claims that OCPA’s projections – and remember, these are only projections – are based on “mistaken assumptions and methodologies.” I would start by countering that Mr. Gokhale is a Ph.D. scholar, a member of the Social Security Advisory Board and one of the most highly regarded research economists in the nation. In short, the man and his abilities are bona fide. That doesn’t mean his assumptions and methodologies couldn’t be mistaken, but as we’ll show below, they’re not.

OPI’s criticisms of our study warrant a response. Below, Gokhale offers a point-by-point response to the OPI report (quotes from the OPI report are in italics, while Gokhale’s responses are in bold).

“A May 2011 report released by the Oklahoma Council of Public Affairs (OCPA) and co-authored by Jagadeesh Gokhale and Angela Erickson of the Cato Institute departs wildly from these national and state-specific studies in contending that the Affordable Care Act will impose enormous new spending obligations on the state budget. The OCPA/Cato study concludes that “implementation of [the ACA] would increase Oklahoma's Medicaid spending by $11.4 billion during the law's first ten years (2014-2023), which would represent a 35 percent increase over estimated Medicaid spending” in the absence of the law. As a result, they contend that total state Medicaid expenditures in the first decade under the law would exceed $40 billion, which would clearly place huge and unmanageable strains on the state budget.”

“The OCPA/Cato Institute's cost estimates cover a longer time period – from 2014 to 2023 – than do either the CBO and Holahan-Heady studies (2014-2019) or the Oklahoma Health Care Authority (2014-2020).”

Correct. We are interested in the full decade's cost, beginning from program inception.

“The additional three or four years covered by OCPA/Cato includes a period where the federal share of costs for newly-eligible Medicaid enrollees de-clines to 90 percent. Still, even accounting for the inclusion of the additional years at a slightly higher state share, their conclusion that state Medicaid costs will rise by $11 billion and 35 percent due to the ACA is far out of line with the other studies that all project additional state costs of less than $1 billion and an increase of no more than 6 percent (under the Holahan-Heady “enhanced outreach scenario”).”

Based on my responses below, who is "out of line" is debatable.

“Despite the inclusion of a two-page methodological appendix, one cannot determine exactly how the OCPA/Cato Institute study arrives at its estimates based on the information in its report.”

I have authored other studies (methodology index begins on page 34) where the methodology used in the OCPA study is more detailed and illustrated with charts on historical trends that the OCPA study extrapolates. Indeed, the studies done by the Urban Institute, CBO and others base participation and cost estimates not on long-term historical trends (the correct approach), but on single year (2007) estimates. It is inappropriate to make projections on a limited subset of data, which by construction does not average across years of high and low participation and cost growth.

“Participation Assumptions: The OCPA/Cato report assumes an additional 321,000 Medicaid recipients in 2014, which is in between the number of new Medicaid recipients assumed by Holahan-Heady (357,000 under the standard participation scenario and 470,000 under enhanced outreach) and by OHCA (137,000 standard, 180,000 enhanced). The significant difference is that OCPA/Cato assumes that a very high proportion of those who will sign up for Medicaid will be “old eligibles‟, who will be reimbursed at the regular federal match rate (approx. 65 percent) rather than the enhanced rate (90 to 100 percent). OCPA/Cato projects that by 2014, there will be 104,000 “old eligibles‟ enrolled in Medicaid, rising to 142,000 by 2023. By 2023, “old eligibles” will constitute 40 percent of their total increased Medicaid population of 359,000.”

Our estimates are based on the Current Population Survey data for Oklahoma, derived by applying current Medicaid eligibility rules by detailed demographic and eligibility categories. In contrast, the studies by the Urban Institute and others base their projections on behavioral "simulations," of which the methodology is sparsely described. Furthermore, these simulations were developed in 2009, well before Obamacare was enacted, and were validated only for 2007 – these simulations did not account for trends in eligibility, enrollments, benefit recipiency, etc. during the past decade ─ which is important because estimating spending growth from a level difference versus a trend difference (with and without Obamacare) would produce very different estimates. That is, matching some features from a given historical year does not imply that the rates of change incorporated in the model’s equations would produce projections consistent with observed past trends.

From the language of these studies, the researchers looked at the “new equilibrium,” not the transition process, which must occur on the path to “equilibrium.” That can lead to various problems – the new equilibrium may not be achieved for a long time and the transitional process may involve high costs. This approach seems inappropriate.

In a simulation model (unlike a strictly data-based projection model, such as the OCPA study) ancillary behavioral specifications could affect the outcomes; without understanding how the model is constructed in detail, it’s difficult to say how that “black box” simulation works. Finally, although they build in reactions to things such as non-compliance penalties for individuals and employers to the individual mandate, it’s not clear how those reactions are calibrated. There’s no information in the past that would provide a means to implement it.

Indeed, there are several other questionable assumptions made in the Urban/simulation model (compromised because of data constraints, limited information about the type of health insurance plans and their features such as premiums, cost sharing. etc.) – far too many to have a reasonable degree of confidence on the simulation model’s operation.

“Their paper states: Oklahoma's Medicaid spending will surge because the federal government will provide only the state’s regular match rate for those who were Medicaid eligible but not enrolled under pre-PPACA laws, also known as the “old eligibles”.

“But given that Oklahoma's Medicaid program currently covers only adults who are parents of dependent children with incomes below 37 percent of the federal poverty level, how many “old eligibles‟ are there in the state who are not en-rolled?

According to OHCA, there are only 61,000 uninsured individuals who are currently eligible but un-enrolled in Medicaid – 44,000 children and 17,000 adults. This total “old eligibles” population is less than half of the 142,000 “old eligibles‟ on which OCPA/Cato bases its projections. The surge in enrollment among the “old eligibles” assumed by OCPA/Cato goes well beyond any “woodwork effect” due to enhanced outreach efforts and the individual mandate that all other studies take into account in calculating state costs, and appears to be more of a pure “thin air” effect.”

Again, our estimates of current enrollees by detailed eligibility categories are taken from the Medicaid Statistical Information System State Datamart (MSIS), a non-partisan administrative database. The estimates of non-enrolled “old eligible” is based on the Current Population Survey data for Oklahoma. These data are weighted to provide a representative sample of the state's population, with weights provided by the U.S. Census Bureau. In short, the figure of 61,000 uninsured individuals who are currently eligible for Medicaid but un-enrolled is inaccurate.

“Expenditure Assumptions – The OCPA/Cato brief provides very limited data to explain how it calculates Medicaid costs under the ACA. However, it starts by asserting: Currently, 23 percent of the Oklahoma population is en-rolled in Medicaid (828,000 enrollees) at an average cost of $4,195 per person. A non-disabled adult on Medicaid in Oklahoma costs nearly $4,500 annually. Both the population figure of 828,000 and the per person costs of $4,500 are overstated. Currently as of August 2011, there are just over 733,000 total enrollees in Medicaid; for the last complete fiscal year, OHCA reported an average of 707,453 members enrolled each month. The OCPA-Cato number thus overstates enrollment by some 13 to 17 per-cent, which affects their projections of the program's cost from 2014 to 2023.”

Again, a critique based on a single year's data point is simply not valid. Our projections are extrapolations of historical trends – the only sensible way to make estimates of future enrollments and benefit receipt. Our enrollments are based on CPS data, which is a representative sample of the Oklahoma population. It's debatable whether our enrollment estimate is an overstatement or the other studies understate their enrollment estimates. The same remarks apply to cost estimates.

The authors of the critique claim that we base our per person costs on the average for non-disabled adults. That's incorrect. A careful reading of the methodology we provide would clarify that the total cost estimates are based on aggregating the product of benefit recipients and average costs, both estimated separately for each eligibility category and separately for each demographic group – by age and gender.

“Even more significantly, their average annual cost for a nondisabled adult – the population that will account for the great bulk of the growth in Medicaid enrollment under the ACA – is fundamentally flawed. It appears from the report that their baseline average cost of nearly $4,500 per adult recipient reflects the average cost per enrollee who receives Medicaid-paid health services rather than the average cost of all enrollees in the program, including the significant percentage of enrollees who receive no services.”

This claim is incorrect. It's true that our tables show "new enrollees from among “old eligibles" resulting from Obamacare. But the average cost per benefit recipient (not per enrollee) – by detailed demographic and eligibility category – is applied to our estimate of new beneficiaries (not enrollees) who would qualify under pre-Obamacare eligibility rules. In fact, not all new enrollees from "old-eligibles" are assumed to be beneficiaries. Based on estimates from our data sources on alternative insurance coverage, we estimate and exclude from our new enrollment estimates of those individuals who would not receive Medicaid benefits because they are healthy or would obtain benefits from alternative (employer sponsored/private) insurance sources.

“The average cost of all Medicaid enrollees, including those who are enrolled but do not receive services in a given year, is much lower: for FY 2007, the average cost per adult enrollee was $2,716, or 40 percent less than the OCPA-Cato figure, according to the Kaiser Com-mission on Medicaid and the Uninsured. However, it appears that OCPA-Cato carries forward their substantially higher per-care-recipient cost across the entire population of enrollees in projecting Medicaid expenditures from 2014-2023. This substantially inflates their projections of the total cost of Medicaid expansion.”

Again, current average costs are not appropriate to apply to cost estimates for 2014-23. Those must be adjusted for expected health care cost growth – again, extrapolating from historical trends rather than basing projections on a single year's cost estimate.

“It is worth noting that the average adult cost for current recipients in both the OCPA/Cato and Kaiser Commission data are based on a population composed largely of pregnant women, who will have relatively high per person expenditures than other adults. A good case can be made that the expansion population will be less expensive than the cur-rent population because it will include a large number of young and healthy adults.”

Our cost estimates are, again, distinguished by eligibility groups.

“In its calculations, OHCA projects the annual per person costs of those newly enrolled in Medicaid at $2,529 in FY 2014, rising to $2,916 by FY 2010. This is also some 40 percent less than OCPA/Cato’s base-line figure of $4,500 per person in 2009.”

See above.

“Compounded together, the faulty assumptions in the OCPA-Cato Institute report that exaggerate the share of new Medicaid recipients who are “old eligible,” the number of current Medicaid recipients, and the expenditure per recipient produce highly inflated estimates of the future cost of the Medicaid budget.”

“Finally, we should note that the OCPA/Cato Institute report looks only at state Medicaid costs and fails to consider any anticipated savings or offsetting revenue associated with implementation of the ACA. However, several studies at both the national and state levels that have considered a broader range of cost, savings, and revenue factors associated with the law have concluded that the cost of the health care law will be less for states and could even yield net savings. A July 2011 report from the Robert Wood Johnson Foundation (RWJF) states that, “State governments will collectively save between $92 billion and $129 billion from 2014 to 2019 because of provisions in the Affordable Care Act that are designed to reduce the uninsured population and provide federal funding for functions that, in the past, have been financed by states and localities.””

Our estimates subtract hospital cost savings form extending coverage to the uninsured. It is not appropriate to say that "state costs" would be low because of "revenues factors." Those revenues would pay for increased costs of providing for new beneficiaries and would reduce state budget deficits. But the costs would, nevertheless, be higher. Our study was about estimating new costs from the introduction of Obamacare.

In summary:

1. The studies cited by the Oklahoma Policy Institute provide very sparse methodological detail to have any confidence in their estimates.

2. When it's available, the methodology from these studies appears to be based on ad-hoc assumptions within a “black-box” simulation model. In contrast, OCPA’s study takes a much simpler approach, is completely data based, and makes projections based on longer-term historical trends in eligibility, enrollments, benefit receipt, and average benefits distinguished by detailed demographic and eligibility categories. They are not based on single-year averages of these variables.

3. A claim that our study is "wildly out of line" is simply not credible. Administrative records and micro-survey historical data in our report are based on representative samples of the state's population. OCPA and Cato would counter with the opposite claim: other studies, particularly those simulation model studies such as done by the Urban Institute – wildly under-represent the likely costs of Medicaid expansion under Obamacare for Oklahoma.