Post-2014 Reform-Related Volume Gains are Modest
Richard Evans / Scott Hinds
richard@ / email@example.com
March 2, 2011
Post-2014 Reform-Related Volume Gains are Modest
- The common notion of 30 million newly insured in or around 2014 (13% gain) likely overstates reform-related increases in demand for health care
- Persons without a health insurance offer from a current employer choose coverage on the health insurance exchanges (HIE’s) less frequently than the 30 million estimate assumes; and,
- Any transition of employees from employer-sponsored (ESI) to HIE’s opens the door for previously insured employees to newly choose to be un-insured. The total number of employed un-insured actually grows if more than roughly half of subsidy-eligible employees transition to HIE’s
- On net, we see best-case gains of 19 million insured (9% gain, assumes no transition from ESI to HIE); worst case, we see zero gain in the insured (complete transition from ESI to HIE’s)
- Moreover, the average value of insurance purchased by the employed-insured is likely to fall. Under ESI, employees have little opportunity to choose very limited coverage, but can make this choice on HIE’s. And on the HIE’s, enrollees’ pay the entire marginal cost of any insurance more generous than the cheapest option. As a result, most healthy enrollees buy the cheapest plan available; only enrollees who expect very high health costs buy generous plans. The result is an average AV on the HIE’s of roughly 0.65; this compares to average AV’s under ESI of roughly 0.82
- Thus an ESI to HIE transition can both raise the number of employed un-insured and lower the value of insurance purchased– in fact, if slightly more than one-quarter of ESI beneficiaries are switched to HIE’s, we would expect to see zero reform-related volume gains in and around 2014
- HMO’s clearly are pressured by any tendency of previously insured ESI beneficiaries to forego insurance on the HIE’s, and for those on HIE’s to buy less generous coverage. Assuming 6% medical inflation and no changes in rates of insurance or AV (i.e. zero ESI to HIE transition), we would expect roughly 7.5% gross margin growth for insurers (best-case). If all current ESI beneficiaries transitioned to HIE (worst-case), gross margin growth would fall to roughly 3.4% from 2014 to 2020. The most likely case is the median; if half of ESI beneficiaries transition to HIE, gross margin growth would be roughly 5.6%. Current HMO valuations imply the worst-case; we still see upside to valuations, though upside is more modest in light of recent outperformance
- Hospitals in particular are negatively affected; net gains in the insured of +/- 30 million almost certainly will not be realized, nor will corresponding declines in costs of uncompensated care, which may in fact rise as AV’s fall, since falling AV’s simply shift cost responsibility from insurers (98% hospital collection rate) to the insured (+/- 30% collection rate). The net shift in insurer mix from commercial to Medicaid in and around 2014 is a potent source of incremental price pressure
- The tendency of HIE enrollees to choose the cheapest option – unless they’re ill – compounds the already tremendous adverse selection pressures on the HIE’s. And, the fact that healthy enrollees on cheap plans are more profitable than less health enrollees on expensive plans creates an economic incentive for insurers to exclude beneficiary-preferred providers of chronic care (e.g. diabetes, oncology) from networks. These findings increase our conviction that PPACA will have to be modified soon after HIE’s begin operating
We have written extensively about the relationships between health insurance, demand for health care, and the inflationary tendency in health pricing. The insurance / demand link is simple: insurance coverage accelerates consumption. We estimate that an insured adult (19-44) in good health consumes more than three times the healthcare of a comparable person who is uninsured
Insurance also breeds price inflation, which itself is a predictable feature of an economic system having substantial inelasticity – if prices rise and unit consumption remains strong, producers have every reason to raise price further. Since aggregate healthcare demand is remarkably inelastic – elasticity estimates range between roughly 0.17 and 0.50 (presumably because only a fraction of direct health costs are borne, or often even seen by, insured patients) — the health system is geared to inflationary price growth
Thus an influx of newly insured patients when (if) key provisions of the PPACA begin to take effect in 2014 could easily be viewed as broadly bullish for providers of health care goods and services, especially those goods and services that are insulated from regulatory or commodity pricing pressures. We believe that this (apparently consensus) view is a logical conclusion when “coverage” is framed simply in terms of “insured v. uninsured.” However, we also believe this framing (e.g. “30 million newly insured”) is overly simplistic. Instead, we believe that the rollout of health insurance exchanges (HIE’s) may lead to increased rates of un-insurance among the employed, and reduced quality of coverage (i.e. actuarial value or AV) among the employed who remain (or become) insured. Most concerning, we believe that the tendency of HIE enrollees to purchase more or less generous coverage depending on their perceived health status may lead to a classic adverse selection spiral
Health insurance exchanges (HIE’s) are likely to increase the percent (and number) of uninsured full-time workers
In prior calls, we have argued that when HIEs begin operating in 2014, many employers will stop offering employer sponsored insurance (ESI) and shift their (particularly, subsidy-eligible) workers into the HIE’s. To the extent this shift occurs, it will increase the uninsurance rate among workers, and on average will decrease the actuarial value of coverage purchased
The shift from ESI to HIE changes the nature of the health insurance enrollment decision for the individual in very important ways. Behavioral economics show that individual decision making is subject to a significant “status quo bias” – individuals disproportionately choose to stick to a current condition, particularly when changing that position requires an action and the status quo requires no action. The phenomenon has been observed experimentally, and proven empirically in a range of settings – including portfolio allocation, retirement savings and health plan selection. The ESI to HIE transition upsets the ESI status quo, and meaningfully changes the nature of the enrollment decision
Choosing no insurance
We would argue that one of the most important (and perhaps overlooked) aspects of ESI is that it sets the insurance status quo to “insured” – roughly speaking, in the presence of an ESI offer, there is very little marginal work required to become insured. The enrollment decision set is generally limited to very few plans, administrative costs are borne by the employer, and premiums are either (1) paid directly by the employer; or (2) paid by the employee with funds withheld from payroll by the employer (i.e., the employee doesn’t have to write a check to their insurer each month to remain insured). By contrast, an employee that is moved into the exchange market has a new status quo – i.e., her default status is “uninsured” until she undertakes a series of (potentially complicated and costly) steps in order to obtain coverage. Her net premium costs are observed, and she must pay them directly, removing her own funds from her bank account every month. Subjectively, we are convinced that this restructuring of the health insurance “choice architecture” implies nontrivial loss of enrollment — i.e. people who did not opt out of ESI will choose not to opt in to HIE
To very roughly quantify the potential loss of enrollment, we looked at individuals’ self-reported preferences for insurance, and the rates at which they were offered (and accepted) employer coverage. Monheit and Vistnes (2004) find that those with a strong preference for insurance are more likely to have a job that offers ESI than those with weak/uncertain preference (Exhibit 1). However, when an offer is made, those with a strong preference are only slightly more likely (91% v 86%) to accept ESI. Extrapolating from their numbers, we estimate that among those who do not receive an ESI offer, individuals with a strong preference for insurance are far more likely to obtain coverage on their own (70% v 48%). These findings support our broad hypothesis that the default option impacts the choices that individuals make; and that ESI’s setting the default option to “insured” inflates insurance enrollment. The directional implication then is obvious – removing the ESI offer reduces the likelihood that any individual obtains coverage, especially in the case of an individual who doesn’t exhibit a preference for insurance to begin with (but who accepts what may be effectively a near costless offer of ESI)
Critically, these findings suggest that HIE’s will increase the uninsurance rate among the full-time employed. We modeled a hypothetical basket of 100 workers, utilizing known distributions of income, insurance preference, ESI offer rates, etc (Exhibit 2). For simplicity, we assume that none of these workers are Medicaid eligible, by expanded PPACA Medicaid eligibility criteria. In this baseline scenario, about 68 of the workers have ESI coverage; another 16 have coverage from another source, and 16 are uninsured
Next, we modeled a scenario in which all subsidy-eligible workers’ ESI offers are rescinded once the HIE’s begin operating, placing the burden of finding, obtaining and paying for coverage on the employee (Exhibit 3). Again, for simplicity and to err on the side of understating the uninsurance rate, we assume that all non subsidy-eligible workers retain their employer sponsored coverage. The net effect was a 6-worker increase in the number of uninsured (i.e., a decline in the insured, from 84 to 80) – the 6 uninsured workers who obtained coverage after introduction of the exchanges were more than offset by the 10 ESI workers who had accepted the ESI offer (i.e. chose not to opt out) but then did not seek coverage on the exchanges (i.e. chose not to opt in)
Exhibit 4 shows the net change in number of insured workers as a function of the share of subsidy-eligible workers who lose their ESI offer. We estimate that if more than 53% of subsidy eligible workers lose ESI, then aggregate uninsurance (again, among workers) goes up when exchanges begin operation
We recognize that this analysis simplifies a number of moving parts in the ESI to HIE transition – most notably the likelihood that the existence of an HIE (and the legal mandate to obtain coverage) will yield an improvement to the market for individual coverage which would increase the enrollment rate among those not offered ESI. Nevertheless we also believe that the market forces and behavioral realities we’ve outlined make it impossible to conclude that HIEs will absorb all of the newly uninsured when employers begin exiting the insurance market
Choosing less insurance
Putting aside the binary decision of whether or not to obtain insurance after having been shifted from ESI to HIE, we believe that the HIE structure nudges enrollees into the least generous (i.e. cheapest or “bronze”) plans, effectively shrinking the average actuarial value (AV) of coverage purchased
There are three key features of health costs and the exchanges that lead us to this conclusion: 1) the distribution of health costs is extremely skewed; 2) HIE’s offer a broader array of coverage options than ESI; and, 3) unlike ESI, HIE enrollees’ subsidies are fixed-dollar, such that the marginal cost of buying more than the subsidized level of AV is borne 100 percent by the enrollee
Beginning with the distribution of health costs; recall that the bottom half of the population accounts for just 3 percent of health spending, while the top 1 percent accounts for more than 21 percent of spending. Given this, any coverage at all is a losing bet for the vast majority of individuals (i.e., for about 87 percent of the population, paying 100 percent of their health costs out-of-pocket is cheaper than their most cost-effective option on the HIE)
Second, most employers that offer coverage offer a limited set of choices, so employees are relatively unable to lower premiums by choosing less generous coverage. On HIE’s, ignoring ‘young invincible’ policies, workers have a much broader range of plans to choose from, with actuarial values ranging from 0.60 (“bronze”) to 0.90 (“platinum”)
Third, under ESI the marginal cost of purchasing more generous coverage is subsidized by the employer, and also by tax effects. This is not true on the HIE’s, where the marginal costs of buying more generous coverage are neither subsidized nor tax advantaged. HIE premium subsidies are set such that the cost of the second-cheapest “silver” (70% AV) plan does not exceed the caps established in the PPACA (Exhibit 5). Importantly, the HIE subsidies are fixed in dollar terms, not percentage terms, and do not vary based on the actual plan purchased. This means that the same subsidy applies whether an enrollee chooses a cheaper “bronze” plan (with the caveat that individuals can’t claim subsidies in excess of premiums paid) or a more expensive “gold” (80% actuarial value) or “platinum” (90%) plan. Thus every dollar by which the chosen plan’s premiums exceeds the federalsubsidy is paid entirely by the enrollee with no offset.
Our assumption that employees switch plans (when given the option) to reduce premium costs is supported by empirical evidence. Samuelson and Zeckhauser, using Harvard University employee health insurance enrollment data from the 1980s, unsurprisingly demonstrate that enrollees were extremely unlikely to switch their insurance plan (among options with relatively similar costs) after their initial enrollment (<3% switch each year). However, in 1985 – when Harvard added a “low cost” version of an existing BCBS plan (with higher deductibles and lower premiums) to the menu of options, there was a spike in switching as “current enrollees in the standard BCBS coverage transferred in significant numbers to the BCBS low option.” The low cost BCBS option was the only plan to have comparable enrollment among new and old enrollees, which shows that there was no status quo bias – all of which is consistent with a willingness to trade actuarial value for premiums when the option is available. Under the HIE’s, this tendency to ‘trade down’ should be more pronounced; the Harvard employees would have at least enjoyed tax offsets against the marginal cost of insurance, where HIE participants will not
We modeled coverage level selection for individuals assuming that (1) they have perfect foresight into their health costs for the year; (2) they could choose from a bronze, silver, gold, or “near-platinum” (i.e., a plan equivalent to the Federal Employee Health Plan actuarial value of 87 percent) plan; and (3) that they choose the plan with the lowest total cost (premiums + cost share) for their level of health care consumption
Based on premium and actuarial value data from a sample of health plans from LA County in 2006, we very roughly estimate that a 1 percentage point increase in actuarial value raises premiums by 1.25%. Combining these premium estimates with estimated subsidies yields the out-of-pocket costs depicted in Exhibit 5 for bronze, silver, and gold plans, as well as a plan comparable to the Federal Employee Health Benefit Program BCBS plan (87% actuarial value). These marginal costs – above the cost of a qualifying bronze plan – are substantial at income levels below 400% FPL. To contextualize, an individual’s marginal out-of-pocket costs, in order to purchase a gold plan on the HIE, are roughly comparable to annual spending on such household necessities as electricity, telephone services, gasoline, and restaurant dining (Exhibit 6). These costs are substantial, and we are very skeptical that a significant number of individuals between 150% and 400% of FPL will choose to make such marginal outlays — unless they expect to be quite ill
All in, we estimate an average actuarial value of coverage purchased on HIE’s of less than 65 percent. If we relax our assumption of perfect foresight by allowing individuals to (fairly dramatically) overestimate their chances of a catastrophic event, average actuarial value increases to only 66 percent. Current AV’s for ESI are roughly 82 percent; this implies that HIE’s could reduce the average AV of insurance purchased by subsidy-eligible workers by as much as 16 percent
Implications for adverse selection
The net effect of the preceding is that on the HIE’s, enrollees naturally sort themselves into only the lowest and highest actuarial value plans available. In a broadened choice set where the marginal costs of buying more than the most basic level of insurance are fully borne by the enrollee, only the very small percentage of persons that drive a very large percentage of total spending choose anything other than the most basic coverage, and they all choose the most generous option available. Conversely, the very large percentage of persons who expect non-catastrophic health costs choose the cheapest option
The dilutive effects of the ESI to HIE transition potentially are profound. First, we see a familiar source of risk pool dilution in the tendency of more workers to choose uninsurance – without healthy persons paying more into the risk pool than they take out, less healthy persons are forced to pay premiums that more closely reflect their actual costs of care. Second, we see further dilution in the form of healthy workers choosing ‘cheaper’ (i.e. lower AV) plans – i.e. the excess of healthy workers’ premiums over health costs under the HIE’s tomorrow is smaller than it is under ESI today, so even when healthy workers choose to be insured, they each contribute less to the risk pool under HIE than under ESI
To be clear, since the structure of the HIE’s first came to light we’ve felt that the design ultimately was unworkable because of adverse selection of the ‘first type,’ i.e. because of healthy persons choosing not to be insured. The ‘second type’ of adverse selection, i.e. the tendency of more health insured to buy much cheaper plans on the HIE’s, simply amplifies the magnitude of adverse selection effects on the HIE’s, and accelerates the related consequences
Implications for changes in demand
We began by expressing skepticism regarding the apparent consensus that ’30 million newly insured’ (an 13 percent increase in the number of insured) would produce a corresponding +/- 13% boost in unit demand, and we conclude that this framing in fact overstates demand growth in or around 2014 for two reasons – net gains in numbers of insured should be much smaller; and, this figure makes no provision for falling generosity of coverage among the (working) insured
We’ve shown that as employers shift workers from ESI to HIE, rates of uninsurance among workers increase, and the generosity of insurance purchased on average declines. Thus the more extensive the shift from ESI to HIE, the less growth we are likely to see in unit demand in or around 2014. We calculated likely numbers of un-insured in 2014 as a function of the percent shift from ESI to HIE: if no workers are shifted from ESI to HIE, we estimate 31 million uninsured; conversely, if all workers are shifted from ESI to HIE, we estimate 50 million uninsured (Exhibit 7). We estimate roughly 50 million uninsured today, thus we estimate the net effect of PPACA is to reduce the number of uninsured by anywhere between 19 million (best case, no ESI to HIE shift, a 9% gain in total insured) and zero (worst case, complete ESI to HIE shift). Surely the answer lies somewhere in between; if we simply assume it lies in the middle, we’d argue that 2014 brings a net gain in the insured population of roughly 10 million persons (a 4% increase in the number of insured) – far less than the apparent consensus notion of 30 million newly insured
We’ve also shown that average AV’s are likely to fall under the HIE’s, which raises the question of whether any gains in numbers of insured actually brings a corresponding gain in net demand. Thinking in terms of total national demand for care, we expect newly insured persons to go from an average AV of zero to an average AV of 0.65, newly un-insured persons to go from an average AV of 0.82 (the current ESI average) to 0.65, and persons transitioned from ESI to HIE’s to go from an average AV of 0.82 to an average of 0.65. Against this backdrop of assumptions, we can calculate the percent of current ESI beneficiaries that would have to be shifted into HIE in order for the whole-nation weighted average ‘generosity’ of insurance to actually fall as a result of PPACA – and the result is 26 percent. More simply, if more than a quarter of current ESI beneficiaries are shifted into HIE’s, then the insurance losses among these workers (the sum of no- and less-insurance) are sufficient to offset the gains from expansion of Medicaid, and enrollment onto HIE’s from persons previously employed but un-insured
On net, we conclude that 2014 gains in unit demand are far less than one would expect under the simple framing of ’30 million newly insured’, and could in fact be negative. We expect total change in the un-insured to be much smaller – anywhere from zero to 19 million. And, we believe that the generosity of coverage purchased will fall, and could easily fall far enough to offset any demand gains from growing numbers of insured persons
Healthcare in aggregate trades at sub-SP500 multiples, thus share prices appear not to reflect an expectation of a large 2014 step-change in demand. This is consistent with our finding that such a demand step-change is an unlikely event; accordingly we have no basis for making a broad claim that healthcare in aggregate is overvalued
More narrowly, we recognize that falling rates of insurance among the employed, and reduced generosity of coverage among the employed-insured both work against HMO’s. We modeled HMO gross margin growth on an assumption of 6% medical inflation, constant rates of insurance among the employed, constant AV among the employed, and a very slow (2020) return to full (6.5%) employment. In this baseline scenario gross profit growth is approximately 7.4%; roughly the product of medical inflation, population growth, and re-employment. More realistically we see rates of insurance and average AV’s falling among the employed. Holding all else constant v. the baseline scenario, we modeled worst-case and median-case scenario’s; in the worst-case we assume a complete transition to HIE’s from ESI, and in the median case assume a 50% transition to HIE’s from ESI. The median case leads to gross profit growth of roughly 5.6% from 2014 to 2020, after which gross profit growth would once again become roughly equal to the product of medical inflation, population growth, and re-employment (Exhibit 8). The worst-case leads to gross profit growth of roughly 3.4% from 2014 to 2020, with subsequent growth normalizing in-line with the other scenarios. HMO’s trade at a roughly 20% (fPE) discount to healthcare in aggregate, which itself trades at a roughly 10% (again, fPE) discount to the SP500, and these valuations are reasonably consistent with a pending reform-related deceleration of HMO fundamentals (following an ’11 – ’13 employment related improvement in fundamentals), though valuations seem to imply the worst-case. We recognize that the transition from large group underwriting to smaller group and individual underwriting in and beyond 2014 implies a tightening of operating margins, which further compounds the earnings pressure of decelerating gross margin growth. Normal rates of gross margin growth eventually return after an ESI to HIE transition is complete, as does productivity growth on a remodeled base of operating expenses, but this may not happen until as late as 2020. In light of recent relative outperformance, HMO’s now imply decelerating earnings, where they once implied dis-intermediation, though we still believe the implied deceleration is worst-case, and that valuations can still improve, though upside is now more limited
Hospitals in particular are negatively affected by un-realized expectations for reduced rates of uninsurance, declining AV’s, and a net shift in insurer mix away from employers and toward Medicaid. Net gains in the insured of +/- 30 million almost certainly will not be realized, nor will corresponding declines in costs of uncompensated care. In fact, costs of uncompensated care should rise as AV’s fall, since falling AV’s simply shift cost responsibility from insurers (98% hospital collection rate) to the insured (+/- 30% collection rate). The shift in insurer mix from commercial to Medicaid corresponds to a substantial reduction in hospitals’ blended effective pricing
Returning once again to broader arguments, we note that the falling AV’s among the employed insured compounds already considerable adverse selection risks of the HIE’s, meaning we’re more convinced that further legislation will be required in order to keep the HIE’s from collapsing under adverse selection, and that such changes will be needed very early in the ‘life’ of the HIE’s. And, we note that the tendency of the ‘well’ to choose the lowest AV and the ‘ill’ to choose the highest AV creates a perverse incentive for plan designers / administrators
Under ESI, insurers work to bring preferred providers into networks in order to win the business of the employer sponsor. Under HIE’s, networks offering particularly good access to preferred providers – especially preferred providers for high cost chronic illnesses – are inviting adverse selection. If we’re correct that enrollees ‘clump’ into either the cheapest or most expensive plan based on their expected health needs; and, that those in the cheapest plans pay premiums above their actual health needs and those in the most expensive pay premiums below their health needs, then plans are motivated to recruit the former and discourage the latter. All else equal, plans with easy access to the area’s best sports specialists but poor access to specialists in cancer and diabetes are more profitable than plans with essentially the inverse design. In effect, the HIE’s create an economic incentive to avoid providers of quality chronic care. This, coupled with compound tendencies toward adverse selection, convince us even further that PPACA will require considerable modification very soon after the HIE’s begin operating
- We found similarly large relative demand differences attributable to insurance status across age, income, and health status subpopulations.
- See, among others, Health and Capital; and “The Political Economics and Investment Relevance of American Health Reform,” August 18, 2009.
- For the purposes of this call, we assume that the key provisions of the PPACA (e.g. health insurance exchanges; expansion of Medicaid eligibility; mandatory coverage) will take effect as currently legislated. Defunding efforts by the new Republican House (see “What a Republican House Means for Health Reform”, October 4, 2010) and an inevitable — and uncertain — Supreme Court ruling on the law’s constitutionality make ultimate implementation far from certain.
- See especially “Why the Average Employer Will Drop Health Insurance in 2014,” May 26, 2010.
- Richard Thaler and Cass Sunstein’s “Nudge” contains an excellent nontechnical treatment of the status quo bias, choice architecture and default options. For an academic view, see Samuelson and Zeckhauser, “Status Quo Bias in Decision Making,” Journal of Risk and Uncertainty 1: 7-59 (1988).
- Monheit and Vistnes, “Health Insurance Enrollment Decisions: Understanding the Role of Preferences for Coverage” ERIU Working Paper 31, 2004.
- More precisely, we assume that employers rationally drop ESI coverage for subsidy-eligible workers and provide cash sufficient to go purchase their own coverage on an exchange.
- Of course, this is no way to advocate forgoing health coverage as an optimal choice.
- Set in terms of percent of income spent on premiums
- Based on our reading of the literature on persistence of health costs through time, we believe that this is not a particularly strong assumption.
- These estimates compare very reasonably to CBO’s estimate of annual bronze premiums of $4,500 to $5,000, relative to the reference silver plan premiums of $5,200.
- Peterson, Chris, “Setting and Valuing Health Insurance Benefits” CRS R40491, April 6, 2009.
- Specifically, we modeled expected health expenditures such that individuals assigned a 95 percent chance of realizing their actual health costs, and a 5 percent chance of realizing a catastrophic event that would put their health costs in the top 1 percent of the country. This causes a structural overestimate of approximately $2,000 per person for 99 percent of the population.
- This 50 million figure is CBO’s spring 2010 estimate of uninsured, which is surely inflated by higher than normal levels of unemployment. If we normalize the employment effect, the estimate of uninsured would be roughly 48 million. Either way, the broad conclusions remain unchanged
- We recognize that AV is only the proportion of covered benefits paid for by the plan, and says nothing about the scope of benefits covered. For this analysis we assume all sources of coverage (e.g. Medicaid and HIE’s) offer the same scope of benefits. Because Medicaid is less generous than HIE’s this assumption tends to overstate the gains in healthcare demand associated with expansion of coverage in 2014, which is the conservative way to test our thesis that demand gains are smaller than expected
- As we emphasized in our last call, because we expect medical inflation to exceed CPI for some time, once the ESI to HIE gross margin effects play out, insurers should once again enjoy the associated leverage of having absolute gross profits grow faster than underlying costs