What Next for the MLR Cycle?

eytan
Print Friendly, PDF & Email

Richard Evans / Scott Hinds

203.901.1631 /.1632

richard@ / hinds@ssrllc.com

June 10, 2011

What Next for the MLR Cycle?

  • We’re convinced MLRs remain stable through 2012 and potentially 2013, though we see significant risks of price competition in 2014. Consensus calls for rising MLRs into 2012 and 2013; we think this is premature, so we expect further outperformance from health insurers relative to both the SP500 and the rest of healthcare
  • The MLR cycle is far more cooperative than competitive; underwriting profits change more as a result of trend reversals in per beneficiary claims cost than as a result of price competition among underwriters
  • 2nd-derivative changes to employment or medical prices are the main causes of MLR peaks and troughs; in the immediate term employment is more likely to improve than worsen, and medical prices generally are decelerating – both of which augur an improving MLR trend
  • MLR floors are in effect a predictable underwriting cost. Under the theory of cooperative pricing insurers should consistently price to a modest expected rebate, which so far is consistent with results. MLR floors encourage higher (pre-rebate) premiums and a smoother MLR cycle, since insurers must more carefully avoid losses
  • As 2014 approaches, hospitals’ pricing power grows, as the marginal enrollee in 2014 is far more likely to recognize a hospital brand name than an insurer brand name, suggesting enrollment has much to do with the hospitals a plan includes in-network. We expect this hospital pricing effect to hit in 2014, but acknowledge it could reach back to 2013, and this is the primary risk to our call that MLRs remain stable in both 2012 and 2013
  • We’re less concerned with Blues’ reserves, believing the plans need large reserves ahead of an expected enrollment expansion in 2014. Factoring in potential enrollment gains, Non-profit Blues’ ‘true’ reserves are about 8 percent lower than they appear. And, when we factor in the much less predictable nature of the marginal enrollee – which compels a higher reserve – Blues’ capital levels seem even less remarkable
  • To our minds the key risk to the underwriting cycle comes in 2014; the one-time nature of this large enrollment expansion presumably will lead insurers to prioritize enrollment gains over short-term MLRs, though we admit this type of price competition has not been seen in other one-time expansions (managed Medicare in 1994 or Medicare Part D in 2006)
  • Beyond 2014, we see spiraling per beneficiary costs on the exchanges as a result of adverse selection. Premiums should be able to stay ahead of the costs, though the exponential nature of the cost trend amplifies actuaries’ forecasting errors, which will tend to raise MLRs. Most important, adverse selection is unsustainable, forcing a change in the rules of the Health Insurance Exchanges (HIEs). What these rules might be and how they might affect insurers is practically impossible to predict; presumably risks are to the downside

The Backdrop: Is Health Insurance Pricing Cooperative or Competitive?

Classic models of service industry pricing behavior center on price / capacity links; i.e. over-capacity leads to price competition, which leads to capacity exiting the industry, which leads to less competition and higher pricing. Capacity then re-enters the industry to exploit the high prices, and the cycle repeats. Contrary to what we perceive is conventional wisdom, this classic model fails to describe the medical loss ratio (MLR) cycle

We see the MLR cycle as more cooperative[1] than competitive. Pricing levels and capacity are very poor predictors of peaks and troughs in the MLR cycle, which argues against the classic competitive model. Conversely, large disruptions to input costs per beneficiary are very good predictors of peaks and troughs. Specifically, if we assume that insurers do not compete on the mark-ups they earn on claims paid, and further assume that the (percentage) level of mark-up the insurers seek is roughly constant, then this implies that the MLR cycle will peak or trough primarily when insurers’ predictions of input costs per beneficiary are substantially wrong. Even more specifically, we assume that insurers simply apply a non-competitive mark-up to the input costs per beneficiary they project for the upcoming contract period, and that their projections for the upcoming contract period rely heavily on claims experience from the trailing period(s). Thus when claims costs (per beneficiary) in the upcoming period are substantially different from those in the trailing period(s), the industry’s MLR level rises or falls

This cooperative model is highly predictive of peaks and troughs in the MLR cycle. And, we consistently find that peaks and troughs are a consequence of trend breaks in underlying per beneficiary costs as a result of 2ndderivative (acceleration or deceleration) changes in either medical prices or employment. Clearly changes to medical prices lead to changes in input costs per episode of care, and changes in employment levels lead to changes in the quality of the risk pool (in downturns employees over-consume as discretionary consumption is pulled forward; in upturns new employees’ claims costs are lower than incumbent employees’ which temporarily reduces average claims costs per beneficiary)

Exhibit 1 is a line graph which overlays our MLR model with actual (private payor) MLRs since 1970; Exhibit 2 is a table showing our model’s accuracy in predicting MLR cycle peaks and troughs. The model describes historic peaks and troughs within 2 years 100% of the time

We conclude that health insurers’ pricing behavior is more cooperative than competitive, and that directional expectations for the MLR cycle should be largely a function of changes to directional expectations for medical pricing and employment

Stage 1: The Probable Effect of MLR Limits on Pricing

The Affordable Care Act (ACA) limits insurers’ gross profits to no more than 20% for individuals and small groups, and to no more than 15% for large groups – i.e. the ACA sets MLR ‘floors’ of 80% and 85%

The expected cost of MLR limits to an insurer is a combination of two values: 1) within any given contract period, the MLR limits (which are applied at the combined plan / state level) eliminate the ability of very profitable plans within an insurer’s national portfolio to offset the cost of plans with losses. Thus all else held equal, an insurer’s consolidated MLR will rise as a result of the MLR limits. We term this the ‘portfolio effect of MLR floors.’ And, 2) at the national level and across contract periods, now that limits are in effect, if insurers price below trend they must accept the resulting losses, but if they price above trend they cannot keep the resulting gains. Before the limits, insurers had (presumably) equal probabilities of being above or below trend, thus the expected value of realized gross profit was exactly equal to premiums less expected claims. With MLR limits in force, insurers have zero probability of an MLR below the limit and (again, presumably) a 50% probability of MLRs above the limit (assuming all insurers price to an expected MLR of exactly the maximum). Thus with no change in pricing behavior, with MLR limits in force insurers’ expected gross profit falls to: (premiums – expected claims) – (0.5 x claims per beneficiary forecasting error); we term this added cost the ‘forecasting error effect of MLR floors’

We estimate that the portfolio effect of MLR floors is roughly 2.9 percent. MLR limits are applied at the combined plan / state level, and ultimately will be applied to three-year rolling average MLRs. Using 2010 actuals for entity-level MLRs, and longer-term data for the persistency of entity-level MLRs, we measured the effect of applying MLR limits to a national portfolio of distinct plans (Exhibits 3 and 4; please see Appendix 1 for an expanded version of Exhibit 3). Limiting individual plan MLRs to 80% raises the MLR of a national portfolio of these plans by 12.2%; limiting small group MLRs to 80% raises the MLR of a national small group portfolio by 3.9%; and, limiting large group MLRs to 85% raises the MLR of a national large group portfolio by 2.4%. Individual plans commonly have lower MLRs for reasons that ultimately (2014) are obviated by the ACA, thus we believe the effect of MLR limits on small and large group portfolios (average increase of 2.9%) is a more meaningful estimate of the ongoing cost of these limits to insurers

The forecasting error effect of MLR floors is difficult to estimate, though we believe this effect is smaller than the ~2.9% ‘portfolio effect.’ With the exception of large 2nd-derivative changes in medical price and employment trends, annual per beneficiary claims cost trends are quite smooth (Exhibit 5). As a crude prediction of forecast error, we can describe the annual per beneficiary cost trend logarithmically, then measure the average annual tendency of the actual trend to diverge from this prediction. If we assume that insurers forecast per beneficiary costs for the upcoming contract period purely as a function of the long-term logarithmic trend (and of course further assuming the model continues to be descriptive), we would expect an average annual forecast error for per beneficiary costs of roughly 5.3% (Exhibit 6).

Clearly insurers can and will do better than the longer-term model in almost all years. Deviations from the long-term logarithmic trend are serially correlated; if plans simply assume that the gap between the long-term logarithmic trend and actual costs is constant from year to year, the average annual error in estimated per beneficiary costs falls by more than half, to 2.4 percent. In reality, because inputs into the national trend (pricing, employment) offer significant early visibility, we believe true error in annual per beneficiary claims cost prediction is well below two percent. Recall that the forecasting error effect of MLR floors is one-half of the average annual forecasting error, making the incremental cost a somewhat less than 1 percent increase in the expected value of insurers’ MLRs, assuming no change in pricing behavior

Thus all in, MLR limits are an estimable cost that would be expected to raise MLRs by roughly 3 to 5 percent if insurers made no changes. Ignoring other steps that improve reported MLRs versus the benchmark (e.g. reducing and/or re-categorizing broker commissions; carving out a larger percentage of benefits to other managers), insurers can eliminate the effect of the floors on MLR by increasing price between roughly 4.3 percent (large groups) and 6.7 percent (small groups). Or, insurers can eliminate the effect of the MLR floors on absolute dollars of gross profit (but accept a marginally higher MLR) by increasing prices roughly 2.9 percent (large group) to 5.0 percent (small group) (Exhibit 7)

Returning to the premise that health insurance pricing is more cooperative than competitive, we would expect insurers to raise prices by an amount sufficient to at least maintain the absolute value of gross profit dollars, after taking into consideration the margin benefit of ‘structural’ steps such as reducing and re-categorizing broker commissions. This is consistent with 1Q11 evidence of larger insurers consistently setting prices at levels where each anticipates paying a small rebate. We believe insurers will continue pricing to a small expected rebate, and that the magnitude of the expected rebate will be roughly equal to the size of the pricing action necessary to offset the effect of MLR floors on absolute margins (2.9 – 5.0 percent)[2]

Stage 2: Near-Term MLR Prospects – the Impact of Employment and Medical Prices

Unemployment is presently at 8.7%, well above the long-term average of 5.7%. In fact ‘true’ unemployment may well be much higher, given that a large number of persons who have been out of work for prolonged periods may no longer be counted as actively seeking employment, and thus no longer counted as part of the labor force. Given our present low point in the employment cycle, it seems far more likely that employment will substantially improve than substantially worsen, thus employment represents more upside than downside to the MLR cycle

Hospitals (inpatient and outpatient), physicians and prescription drugs dominate input costs, thus our consideration of whether pricing is likely to accelerate or decelerate focuses on these three broad groups of inputs. Unlike hospitals and prescription drugs physicians are ‘price-takers’ and have been for some time; accordingly we see little likelihood of physicians’ pricing (~36% of total input costs) disrupting the underwriting cycle in the near- to mid-term. Prescription drug (~17% of input costs) average prices are likely to decelerate in the near-term (4Q11 thru 1H12) for two reasons: major patent losses create a negative mix effect beginning in 4Q11, and brand drug list pricing (presently running at 6.9%) is likely to decelerate significantly (we estimate to +/- 4%) in early 2012. As we’ve shown previously, brand drug list pricing tends to decelerate going into presidential primaries (which begin 1Q12), especially when prices were growing in real terms in the year preceding the election (Exhibits 8, 9)

Hospital pricing (~39% of total costs), like physician and pharmaceutical pricing, recently has been decelerating (Exhibit 10); however we see prospects for acceleration in both the near- and middle-term. Near-term, we expect states to cut hospitals’ Medicaid re-imbursement rates more than is generally expected, as Medicaid subsidies tied to the American Recovery and Reinvestment Act (ARRA) expire at the end of this month, and as states may no longer be able to rely on other sources of Medicaid cost-savings. Hospitals traditionally seek to re-capture lost Medicare / Medicaid pricing from commercial payors, thus accelerating Medicaid re-imbursement cuts may force hospitals’ commercial prices higher. Nevertheless even if this occurs, we believe these effects would be in plain view well before insurers’ write the bulk of their 2012 commitments, thus we see little immediate risk to the underwriting cycle

Stage 3: Insurance Pricing Effects in the Run-Up to Health Insurance Exchanges (HIEs)

In the middle- to longer-term, we believe hospitals that have positive name recognition in their communities also have substantial pricing power, especially during the first few years of expanding enrollment on the health insurance exchanges. The reasoning is straightforward: consumers who are new to purchasing health coverage are far more likely to recognize hospital names than health plan names, and so are likely to choose health plans on the basis of which hospitals they can or cannot have access to. We believe the roll-out of Massachusetts’ health reforms serves as a real-world example of this effect: in the two years after reforms came about in 2006, Massachusetts hospital pricing accelerated to the point that hospital costs accounted for over 55% of the state’s total growth in healthcare expenditures in that period (Exhibits 11, 12). Massachusetts hospitals do not appear capacity constrained (Exhibit 13), which would tend to rule out this cause of rapid price increases. Charge-per-day growth does correlate with hospital size (Exhibit 14), which tends to support the notion that recognizable Massachusetts hospitals enjoy considerable pricing leverage over Massachusetts health plans. The tendency of hospitals to gain pricing power as 2014 approaches should re-accelerate hospital pricing in the middle-term, but we believe insurers are likely to see the trend early enough to incorporate higher hospital charges into premiums

A more traditional source of MLR concern in the run-up to the HIEs is the Blues plans; reserves are relatively high, and Blue Shield of California recently pledged to limit its net income to 2 percent of premiums, down from 3 percent

We examined the non-profit Blues’ reserves on a per-member per-month (PMPM) basis, adjusting all years to 2010 health spending levels to eliminate the effect of medical inflation on the trend (Exhibit 15). Plainly reserves are very high relative to history; however we stress the need for these plans to have more reserves than normal as 2014 approaches, given that they and their regulators will expect a large enrollment expansion in that year. If we adjust reserve levels to reflect the percent change (implied by CBO estimates of non-Medicaid enrollment gains) in an average plan’s enrollment in 2014, then the ‘true’ reserve level falls by 8 percent. And, we note that enrollment expansions typically bring a great deal of uncertainty with respect to the marginal enrollee, thus reserves per marginal enrollee arguably should be higher, which would serve to further normalize the current reserve levels – particularly in the eyes of state regulators who are charged with having a diversity of well-capitalized plans available on their state-level HIEs in 2014. In short, we expect regulators’ bias is to let the non-profits build reserves ahead of 2014, which lessens our concern that present high reserve levels will lead the Blues to raise their loss ratios

Against this backdrop, we do not read California Blue Cross’ pledge to reduce its net income to 2 percent as material. Politically, insurance premiums are a particularly charged issue in California; the state is one of 11 in which insurance commissioners are elected rather than appointed, and California Blue Cross recently disclosed their CEO’s 2010 compensation of $4.6M under pressure from Consumer Watchdog. Mathematically, California Blue Cross is reducing its net income by a single percent, which – assuming all of the delta is devoted to claims cost – would only raise its MLR by a single percent

Stage 4: Roll-Out of HIEs

Despite our conviction that health insurance pricing is more cooperative than competitive, we can envision scenarios in which the reverse might be true. The total market for health insurance – specifically in terms of the number of persons enrolled – typically grows at a glacial pace. On the commercial side eligible populations grow roughly at the rate of change in the number of persons employed, net of gradual changes in the rates of health insurance offer and acceptance between employers and employees. Against this backdrop, consider the competitive relevance of suddenly adding several million new potential enrollees, all at once. Using Congressional Budget Office (CBO) estimates – which for the record we think are too high[3] – 2014 brings roughly 14 million new commercial enrollees. In an industry with an annual demographic tailwind of roughly 1.5%, 2014 effectively packs over four years of enrollment growth into a single contract cycle

The one-time nature of this enormous potential enrollment gain should, theoretically, have a very large impact on 2014 pricing behavior. Far better to sacrifice 2014 margins for the sake of enrollment, then raise prices to profitable levels over time, than to lose 2014 enrollment share. This seems especially true given the presence of loss-sharing provisions in the ACA, which offset insurers’ underwriting losses during the roll-out of the HIEs – in effect, the temporary loss-sharing provisions function as enrollment subsidies, and should encourage lower pricing

In purely behavioral terms we find the argument that 2014 brings price competition among insurers compelling. However empirically, we have to accept that history offers two natural tests of this hypothesis – and in neither instance did insurers lower price to take advantage of large one-time enrollment gains. We’ve twice seen large gains in Medicare enrollment: first for the growth of managed forms of Medicare in 1994, and again in 2006 with the roll-out of Medicare Part D. In both of these cases, we would have expected to see large potential enrollment gains correlate with fairly sharp increases in MLR (i.e. lowering of prices), followed by a return to more typical MLRs over time. However in neither case were large enrollment gains clearly associated with price competition (Exhibit 16). Nevertheless on balance, we still see substantial risk of price competition reducing MLRs as eligible populations expand in 2014

Stage 5: The Potential for Runaway Adverse Selection on the HIEs

We’ve shown elsewhere that rates of un-insurance are likely to remain high even after the roll-out of the HIEs; this is mainly due to the relatively modest perceived value of health insurance to potential buyers, coupled with the very large difference between the low cost of penalties (for not being insured) and the fairly high cost of coverage for most households. This is a classic setting for adverse selection, wherein policies are more often purchased by persons who are ill or soon expect to be, driving average claims costs higher, which drives premiums higher, which further reduces the value of plans to those who are less ill, and so forth as the cycle repeats

We’ve also shown that because the exchanges offer a much broader range of coverage ‘generosity’[4] than is typically available elsewhere, that further potential for adverse selection exists even among those choosing to be insured. We believe most households in fair or good health that choose to be insured will select the cheapest available option; whereas households in poor health will choose the most expensive option. Thus even though some healthy households may choose to be insured, each contributes fewer premium dollars to the care of less healthy households than typically occurs in health insurance markets (especially employer-sponsored insurance or ESI) where beneficiaries have a more narrow choice of coverage levels – all of which further feeds the HIEs’ tendency toward adverse selection

We’re convinced that the HIEs, as structured, will suffer substantial adverse selection, which is by definition unsustainable. This argues that rules of the HIE ‘game’ must also change within a very few years after they begin operating, and it’s practically impossible for us to predict how these will change. We do believe that insurers will be able to anticipate the effect of adverse selection on average per beneficiary claims cost, and so will be able to raise premiums ahead of the accelerating cost trend; however the accelerating cost trend adds considerable forecasting (and thus MLR) risk

Summary and Conclusion

We conclude that underwriting margins should remain stable through 2013, with a risk of compression in 2014 as a by-product of a one-time fight for large enrollment gains. Profits on the HIEs presumably will be substantial as underlying per beneficiary costs spiral due to adverse selection, further amplifying the gap between gross margin growth and growth in operating costs, to the benefit of overall profitability. Any such out-sized gains likely are very short-lived, as adverse selection compels further reforms

Typically we value insurers on long-term earnings power relative to the SP500, though we increasingly struggle with this framework due to the tremendous uncertainties in and beyond 2014. We’re defaulting toward a shorter-term and more subjective framework of expectations v. reality; and, believing that consensus expectations for an upturn in the MLR cycle (Exhibit 17) are misplaced, we see potential for health insurers to continue outperforming both the SP500 and the broader healthcare industry

  1. We use the word ‘cooperative’ for no other reason than to adhere to standard economics / game theory terminology – we do not believe, nor are we suggesting, that de facto collusion exists
  2. Expected rebates should be smaller in the first year of MLR floors, as the additional pricing needed to maintain margins is reduced by the margin gains from various structural changes, e.g. to broker commissions and carve-outs. In subsequent years, new structural offsets to MLR floors presumably will not be available; accordingly we expect pricing to be consistent with an expected rebate that is roughly equal to the all-in ‘cost’ of MLR floors
  3. Post-2014 Reform-Related Volume Gains are Modest” March 2, 2011, Sector & Sovereign Research, LLC
  4. The HIEs will offer actuarial values (AVs, the % of eligible claims cost paid by the plan) of 60, 70, 80, and 90; whereas most beneficiaries of government or employer-sponsored plans are offered a much more narrow range of AVs
Print Friendly, PDF & Email