The Mechanics of Commercial HMOs’ Gross Profits: Why MLRs Should Remain Stable

Print Friendly, PDF & Email

Richard Evans / Scott Hinds / Ryan Baum


203.901.1631 /.1632 / .1627

richard@ / hinds@ /

twitter.jpg @SecSovHealth

June 14, 2012

The Mechanics of Commercial HMOs’ Gross Profits: Why MLRs Should Remain Stable

  • Estimates and share prices for the commercial HMOs imply rising medical loss ratios (MLRs); stable MLRs are more likely, thus the sub-sector appears under-valued
  • Since 1961, 5 of the 6 MLR peaks and 6 of the 7 MLR troughs were the result of sudden changes in medical costs (price times utilization), rather than changing levels of price competition among insurers
  • Virtually all of these cost surprises were the result of either or a combination of two factors: 1) large accelerations or decelerations in medical price growth; and/or 2) an accelerating rate of change in employment
  • Ignoring large (and random) flu effects, MLR peaks and troughs almost never occur without a corresponding inflection in medical inflation or employment. Medical inflation currently is rapid but stable, and employment is more likely to rise (albeit slowly) than to fall rapidly. Viewed in an historic context, no empirically defensible reason exists to anticipate rising MLRs
  • Rising per-capita utilization at the national level presumably is misinterpreted as a source of rising commercial MLRs. National utilization is rising as a result of employment gains; however employment gains generally reduce per-capita utilization (and thus MLRs) among commercial beneficiaries. Both effects are results of job growth – new employees consume more than the uninsured, but less than ‘legacy’ employees; thus national rates of utilization rise, but commercial rates of utilization fall. New employees consume at about half the rate of ‘legacy’ employees immediately after being hired, and at about 80 percent the ‘legacy’ employee rate toward the end of their first year – but new employees pay the same premiums paid by legacy employees
  • At current valuations, the commercial HMOs imply any or a combination of: a spike in medical inflation, a spike in unemployment, and/or an inordinately severe flu season. Medical inflation is stable, employment is unlikely to fall very far or very rapidly, and flu severity is a random variable. On balance revenue growth (through enrollment gains and current rates of medical inflation) coupled with stable MLRs (eased by the addition of marginal enrollees with lower health consumption, but capped by MLR ‘floors’) is the far more likely outcome, thus our conclusion that the commercial HMOs are undervalued

We believe conventional wisdom misinterprets two crucial elements of the commercial HMO business model, and that these misinterpretations are at the heart of current (we believe too low) valuations. Specifically: 1) the medical loss ratio (MLR) cycle is driven more by costs than by pricing, and these cost effects are unlikely to increase MLRs in the near-term; 2) rising utilization in the context of rising employment is a net benefit, yet commercial HMOs’ current share prices suggest rising utilization is viewed negatively

The MLR cycle: cost-determined, rather than price-determined

The classic service industry theory of gross profit cycles assigns a dominant role to competitive pricing behavior: as pricing rises more capacity comes on-line (and/or the temptation to take share with discounts rises); as price falls capacity goes off-line and/or returns approach costs of capital, extinguishing the temptation to take share by discounting. This classic model of gross profit cycles is essentially ‘price-determined’; i.e. changes to price, not costs, set the cycle in motion

An obvious alternative is a ‘cost-determined’ gross profit cycle, one in which the industry players don’t price compete (or competition is at least a weak influence), and in which underlying costs are sufficiently volatile and difficult to predict that cost trends dominate gross profit changes. To our minds this cost-determined model explains the commercial MLR cycle

We think the history of commercial MLRs proves our point. Looking back, in a price-determined gross profit cycle, we’d expect to see gross profit inflections independent of ‘shocks’ to underlying costs – i.e. evidence that industry players simply chose to raise or lower[1] their gross margins, independent of what was happening on the cost side. In the case of a ‘cost-determined’ gross profit cycle we’d expect the opposite, i.e. changes to gross profit determined by ‘shocks’ to costs. For commercial HMOs, the cost-determined pattern of changes to gross profits explains very nearly all of the MLR cycle since 1961

Across this 50+ year span the MLR cycle has seen 6 peaks, and 7 troughs[2]. We associate 5 of the 6 peaks with ‘cost shocks’[3], and 6 of the 7 troughs[4]. Exhibit 1 shows the relationship; ‘cost shocks’ are represented by a combined measure that reflects the average rate of change in unemployment and medical inflation, and are compared to private insurance MLRs since 1961

With the obvious exception of year-on-year changes in flu severity, most historic cost inflections tie back to accelerating rates of change in either (or in some cases both) medical prices or employment. The effect of sudden changes to medical price on gross profits is straightforward: if medical prices accelerate or decelerate rapidly, costs of care obviously move in the same direction, and if this hasn’t been anticipated in pricing for that contract period, gross profit margins will change. The employment effect is less obvious but still powerful. When overall employment declines rapidly, employees pull discretionary healthcare consumption forward, raising per-capita utilization and MLRs; when employment rises rapidly healthier new employees consume less than (but pay the same premiums as) existing employees, and MLRs tend to fall

Neither source of commercial HMO gross profit pressure is likely to emerge in the near-term, so commercial HMOs’ MLRs should remain stable. Hospital pricing (34% of input costs) has been strong, but acceleration in pricing has been modest (Exhibit 2). Catalysts for hospitals’ commercial pricing (pressure on Medicare payment rates, the need to have brand-name hospitals in-network to maximize Affordable Care Act enrollment gains in 2014) are reasonably predictable and can accordingly be factored in to premiums for upcoming contract periods. Physicians’ (28% of input costs) commercial fees are on a weak and stable trend; as physicians have relatively little negotiating leverage now or in the foreseeable future, we see no reason to expect a sudden spike. Drug pricing (14% of input costs) is rapid on the brand side but more likely to fall than rise; patent losses are outpacing list price gains in the near-term, and rates of list price gain ‘feel’ like they’re maxed (Exhibit 3). Employment is rising slowly off of Great Recession lows (Exhibit 4); since gradual employment gains now appear more likely than very large employment losses, employment seems more likely to aid than impair commercial MLRs

Utilization in the contexts of history and the job cycle

Big picture: utilization v pricing as a source of health cost surprises

To frame utilization as a potential source of unexpected changes to MLR, we need to consider both the relative contribution of utilization to total health spending growth, and the extent to which utilization changes are or are not predictable. Utilization (per-capita, age-adjusted) has much less of an effect on total health cost growth than pricing (Exhibit 5); over the last ten years age-adjusted per-cap utilization has been just less than 20 percent of total growth, and in the last five years just a little more than 14 percent

As regards volatility, we think of utilization in age-adjusted terms simply because the annual aging effect is incredibly predictable – it literally can be estimated a decade (or more) in advance. The standard deviations of age-adjusted per-cap utilization and pricing are very nearly the same, about 1.1% over the last 20 years, so in pure mathematical terms pricing and utilization are comparably ‘noisy’. In more practical terms however, utilization is more predictable than pricing for a very simple reason – the majority of period-to-period ‘noise’ in utilization can be explained by the underlying economic cycle (particularly the employment rate, Exhibit 6), while a large part[5] of health pricing ‘noise’ is unrelated to the economic cycle. Thus HMOs’ ability to forecast changes to per-capita utilization is roughly on par with their (or others’) ability to forecast economic changes, but real changes in medical pricing apparently are random, or very nearly so (Exhibit 7)

Thus changes in utilization are a less potent source (being both smaller and more predictable) of national per-capita health spending ‘surprises’ than changes to medical pricing. And, within commercial underwriting pools the contribution of utilization to health cost ‘surprises’ is smaller still. Much of the change in national utilization is the result of households moving into and out of employment (and employer-sponsored health insurance); this source of rising national utilization does not necessarily translate into rising rates of per-capita utilization within commercial HMOs’ covered populations

Utilization (and commercial HMOs’ gross profits) when employment is rising

Rising employment increases average rates of per-capital utilization at the national level, but tends to improve commercial HMOs’ gross profits

On an age- and health status-adjusted basis, the privately insured consume about 2.7x more healthcare than the uninsured. As employment rises, more households shift from uninsured to insured – nearly tripling their healthcare consumption in the process – so the national average rate of per-capita healthcare utilization rises. This shifting of households from one employment / insurance category to another explains very nearly all of the changes to national rates of healthcare utilization across the current employment cycle[6]

However the shift of households from uninsured to insured has very different effects on utilization and gross profits within a population of commercially insured lives. Average per-capita utilization within the commercial risk pool falls with the addition of each new employee, since new enrollees tend to consume less than existing employees (even though they pay the same premiums). In the first month of coverage average expenditures for new employees are less than half the level of ‘legacy’ employees[7]. New employees’ utilization rises with time, but gradually – employees with coverage obtained in the last 7 to 11 months still consume about 15 percent less than legacy employees (Exhibit 8). Also, rising employment naturally means rising enrollment for the commercial HMOs, thus the direct effect of rising employment is to increase total gross profit – enrollment grows, and marginal enrollees are more profitable. At the net income level profits may improve even more rapidly, as rising membership allows for more efficient amortization of operating costs

For completeness’ sake, we believe there’s an indirect effect of rising employment that increases utilization within commercially-insured populations, but we believe the effect is small. Very simply, as the economy improves consumer confidence tends to improve as well, and households – including those in commercial risk pools that may have been employed throughout the economic cycle – tend to spend more. It follows that per-capita health spending among the ‘long-term’ commercially insured should rise to some degree as the economy improves. However the effect appears very small for two reasons. First, we can show that the movement of households from uninsured to insured status accounts for very nearly all of the observed change in national rates of utilization[8]. Second, the income elasticity of demand for healthcare is quite low – around 0.20 or less[9]

Thus on net, rising employment benefits commercial HMOs – gross profits rise as enrollment grows, especially in light of the fact that marginal enrollees are more profitable than ‘legacy’ enrollees. And, the health utilization effect of rising incomes within ‘legacy’ enrollees’ households appears too small to offset the benefits of enrollment gains

Utilization (and commercial HMOs’ gross profits) when employment is falling

Falling employment reduces enrollment, may lead to near-term spikes in commercial beneficiaries’ utilization, and tends to increase the average age (and health costs) of the remaining beneficiaries

At the national level, declining employment has a large and sudden effect on utilization – households that become uninsured after having had private insurance immediately reduce average healthcare spending by half (Exhibit 9). More narrowly, within commercially insured populations average utilization tends to increase for two reasons: one acute, and one chronic. Households that eventually lose employment presumably see this coming, as these households tend to pull discretionary healthcare demand forward before losing their jobs (and employer-sponsored health insurance, Exhibit 10). The result is a spike in average rates of utilization among commercial beneficiaries. After the rate of job loss stabilizes, utilization rates among the commercially insured should remain at least slightly elevated; the remaining employees tend to be older, and the least healthy of the households (whose health costs exceed premiums) that were recently laid off tend to extend their coverage under COBRA

Summary and investment conclusion

On a capitalization weighted basis, the commercial-predominant HMOs trade at 19, 22, and 33% discounts to the SP500 on FY+1, +2, and +3 year consensus, respectively. Consensus calls for both an acceleration of revenue growth and an increase in MLR’s (Exhibits 11, 12). Ignoring possible mix changes (i.e. growth contributions from non-commercial enrollment growth), revenue should change only as the net result of three drivers: enrollment growth, changes to the generosity of an average contract, and medical inflation. We know that average contract generosity is in fact falling, leaving enrollment and medical inflation as possible drivers of accelerating revenue. If medical inflation is the (presumed) driver, then it can only drive revenue growth if it is anticipated in premium growth – which by extension tends to rule out sudden medical inflation as a driver of rising MLRs. In effect there’s no likely scenario that brings both rising (pre-2014) enrollment and rising MLRs. Instead, assuming (albeit slow) employment gains are more likely than sudden employment losses; and, further assuming that medical inflation is sufficiently predictable to be accounted for in premiums, the more likely outcome for commercial HMOs is rising revenue and stable MLRs

  1. Obviously lowering gross profit could have less to do with competition and more to do with customer pressure, but the historic pattern would be the same – inflections in gross margin independent of costs
  2. We define MLR peaks and troughs as a > 1 s.d. percentage change in MLR
  3. For the record it’s the ’86 / ’87 MLR peak that we can’t blame on accelerating costs
  4. The ’89 MLR trough is unexplained by improving costs
  5. Specifically real medical pricing, which over the last two decades has accounted for roughly 40% of the change in nominal medical prices
  6. Please see Why HMOs are Cheap, Despite Rising Utilization, May 17, 2012, SSR
  7. New employees tend to be both younger, and healthier. Older employees have a priority claim on jobs in the current employment down-cycle, so changes in employment tend to involve younger (and thus healthier) employees. And, even ignoring age, new employees tend to be healthier. Recall that health costs within any sizable population will tend to be driven by a very small percentage of very ill persons. Persons with these levels of illness are less able to get jobs; accordingly the hiring process tends to select for persons with above-average health and thus below-average health costs
  8. Ibid 6
  9. See for example: “The Elasticity of Demand for Health Care …” by Ringel et al. of RAND Health, published in 2005 and available at:
Print Friendly, PDF & Email