US Healthcare Demand Slow for Cyclical Reasons; Volume-Sensitive Names are Undervalued

Print Friendly, PDF & Email

Richard Evans / Scott Hinds / Ryan Baum

203.901.1631 /.1632 / .1627

richard@ / hinds@ /

twitter.jpg @SecSovHealth

January 12, 2012

US Healthcare Demand Slow for Cyclical Reasons; Volume-Sensitive Names are Undervalued

  • From 1960 to 2007 average per-capita ‘units’ and ‘mix’ of care (which we follow using an aggregate measure termed ‘elasticity/mix’) grew 2.2% (real, per-capita) annually, then fell 40bp each year from 2008 through 2011, reaching a 50-year low in 2010
  • Changes in per-capita elasticity/mix are starkly cyclical, and account for the entirety of the post-2007 fall in US healthcare demand. The fall in per-capita ‘units’ of care (e.g. physician visits & acute care admissions) appears to have been more important than any change in the ‘mix’ of care
  • Secular changes to per-capita demand (e.g. HDHP enrollment) are comparatively (and absolutely) small, explaining no more than one-eighth of the drop in per-capita elasticity/mix, and in all likelihood far less
  • As an almost entirely cyclical phenomenon, the (now trailing) trough in US healthcare demand presumably foreshadows a comparable recovery. Healthcare companies whose earnings are more sensitive to volume (cyclical) than either pricing power or new product flow (neither being cyclical) should see more substantial earnings acceleration in a recovery
  • Fundamental expectations and valuations for volume-sensitive healthcare names generally imply that a 50-year demand trough (2010) extends well into the future (+/- 2013). In effect these valuations signal a zero likelihood of improvement in the broader economy, a belief that the slowdown in healthcare demand is a secular ‘new normal’ or some combination thereof. Believing the healthcare slowdown is cyclical, and that broader economic conditions are as likely to improve as to deteriorate further, we see the volume-sensitive names as generally undervalued
  • Manufacturers / distributors of non-prescription consumables are most favored (e.g. BAX, BDX, COV), followed by commercial HMOs (e.g. UNH, WLP) and acute-care facilities (e.g. HCA, UHS, THC, HMA, LPNT, CYH). Because dental demand (both mix and volume) appears even more elastic than traditional medical demand, we see upside to manufacturers / distributors of dental consumables (e.g. PDCO, HSIC, XRAY, SIRO), though premium valuations make these names relatively less attractive

US healthcare demand is relatively weak in a long-term historic context (Exhibit 1); and, as compared to consensus we believe demand also has been surprisingly weak in the near-term (e.g. 2009 – 2011)

This note examines the causes of slow demand growth, whether these are cyclical or secular, and whether subgroups of healthcare stocks may be misvalued as a consequence of any confusion regarding the demand trend

Our conclusion is that the causes of slow US healthcare demand are predominantly cyclical, but valuations appear to reflect, at least in part, a more permanent secular slowdown. Accordingly volume-sensitive subgroups of healthcare stocks – in particular manufacturers and distributors of non-pharmaceutical consumables (e.g. BAX, BDX, COV) – appear undervalued. Acute care hospitals (e.g. HCA, UHS, THC, HMA, LPNT, CYH) also appear undervalued; beyond the obvious relevance of the volume trend to these names we believe they retain substantially more net pricing power than is generally understood. Commercial HMOs (e.g. UNH, WLP) continue to appear undervalued despite recent share price outperformance; we see stable gross profit percentages on revenues that grow faster than operating costs in a near-worst case (e.g. even at stable enrollment medical inflation exceeds underlying inflation in operating costs); in an upturn expanding enrollment magnifies operating leverage, and offsets the gross-margin effects of rising per-capita demand by improving the risk pool

What’s Changed?

We frame healthcare demand growth as the simple product of: change in number of persons, change in ‘units of care’ per person, and change in the average cost of a unit of care. Because population growth and the effect of population aging (or more simply ‘demographic effect’) move glacially, and have in total changed very little over a half-century, for the purposes of de-constructing the current slowdown their effect can be ignored. This moves measurement of the demand trend to an ‘age-adjusted per-capita’ basis, and leaves pricing, innovation, and elasticity-related changes in unit demand as candidate causes of slow growth

Backing out CPI leaves real age-adjusted per-capita demand growth, which can be separated further into two values: real medical pricing, and ‘elasticity/mix’. Exhibit 2 shows the 50-year trend in these values. Current real medical pricing (2.3% from ’09 to ‘11e) is on par with its 50 year average (1.8%); however elasticity/mix (-0.4% from ’09 to ‘11e) is dramatically below its 50 year average of 2.0%, and is in fact coming off of a 50-year low (2010). We conclude that slow healthcare demand ties directly to an outright fall in elasticity/mix

Despite its two-part name, elasticity/mix consists of three reasonably distinct components: 1) changes in per-capita unit demand attributable to changes in elasticity; 2) changes in per-capita unit demand attributable to the availability of treatment options in year X than were not available in year X-1; and, 3) changes in the average cost of a ‘unit’ of care attributable to the displacement of less expensive treatment options by more expensive treatment options

We believe the decline in elasticity/mix has more to do with falling unit demand than with a change in the mix of care. The definitive datasets for longer-term analyses of unit demand and mix are only current through 2008, so we’re unable to take a detailed view of demand and mix through the 2009 – present slowdown. However current prescription per-capita data are available, and these correlate reasonably well with physician visits per-capita, which in turn correlate reasonably well with the per-capita rate of health system interactions (or, alternatively, per-capita ‘units’ of care). The prescriptions per-capita trend is generally consistent with the recent (and longer-term) trend in elasticity/mix, which by extension is at least consistent with our presumptive view that elasticity/mix has fallen simply because fewer patients are showing up for care (Exhibit 3)[1]

More broadly, and to defend further our view that the slowdown is more about units than mix, we would point to the conventional wisdom – which we accept – that health care elasticity generally manifests more as a fall in the number of interactions between patients and the health system than by changes in the mix of care consumed once an episode of care begins. Empirically, this pattern is evident in studies of elasticity reaching as far back as the RAND Health Insurance Experiment (HIE). More subjectively, the pattern fits general observations: patients are more able to influence whether or not they interact with the healthcare system than what care is received once they show up; and, providers tend to offer a single state-of-the-art treatment option rather than a spectrum of treatment options of varied costs and/or levels technical sophistication. As a final point, building on the tendency of providers not to offer options ‘below’ the state-of-the-art, we would emphasize that the elasticity/mix trend currently is negative; changes in mix could only produce a negative value for elasticity/mix in aggregate if patients were reverting to substantially cheaper care options[2] en masse, and this seems incredibly unlikely

Cyclical, or Secular?

We believe slowing demand is far more cyclical than secular, and that we can show this irrespective of whether one accepts our preceding argument that falling unit demand is having a far greater impact than changes in the mix of care

At the risk of being pedantic, demand for healthcare is (at least somewhat) elastic; and, demand elasticity is a consequence of both income and health insurance, both of which are affected by changes in employment. Comparing national employment levels and changes to healthcare elasticity/mix across the 51 years of available data, it appears that elasticity/mix generally mirrors the same – cyclical – trend as the broad labor market (Exhibit 4). We believe the implied cause and effect is in fact true: i.e. that falling incomes and falling employer-sponsored insurance enrollment have reduced demand for care, which explains the fall in elasticity/mix. Correspondingly we believe that an economic up-cycle that includes rising employment ultimately will drive the elasticity/mix trend higher

The relationship between consumers’ healthcare and non-healthcare spending similarly argues that healthcare demand has slowed for predominantly cyclical reasons. Consumers obviously moderate their overall spending according to both objective realities (e.g. real income) and subjective expectations (e.g. confidence); it follows that changes to consumers’ non-healthcare spending is a useful benchmark for analyzing changes in healthcare spending. We analyzed the relationship between consumers’ healthcare and non-healthcare spending across the 1984-2008 time period, and used the result to estimate 2008-2010 healthcare spending (which we presume reflects cyclical factors) as a function of consumers’ non-healthcare spending patterns (which are known to reflect cyclical factors). Against the backdrop of changes to consumers’ non-healthcare spending, changes in healthcare spending look more or less normal (Exhibit 5). In fact if anything, based only on consumers’ non-healthcare spending we would have expected 2008 – 2010 healthcare spending to have been lower than it actually was, despite the fact that the drop in healthcare demand ‘felt’ surprisingly large

We find the evidence supporting cyclical drivers as a cause of slow healthcare demand compelling; but recognize that a secular change in the nature of insurance also is occurring (in particular the shift to high deductible health plans, or HDHPs), and that this has to be considered carefully as at least a contributing cause, or even as an alternative explanation. The trend toward HDHPs, and the related transfer of health cost risk to consumers, is fundamentally important[3], but has very little to do with current low levels of demand

Despite rapid enrollment growth, HDHPs’ current-market demand effects are mitigated by two key considerations: the amount of total health spending that falls above even HDHP deductibles; and, the still modest total penetration of HDHP plans into the total (commercial & public) insurance base. Healthcare spending is highly skewed on a per-capita basis; very few persons spend the very large majority of total health dollars, and these persons and dollars are rarely influenced by even very high deductibles.

Using single policies as an example, average HDHP deductibles are just less than $2,000, and average deductibles for all other forms of employer-sponsored coverage are just less than $1,000. Only thirteen percent of persons have total annual health spending that falls between these deductibles; the total spending that falls between the $1,000 and $2,000 annual levels is only 5 percent of total spending, and fully 91 percent of total spending exceeds the higher ($2,000) HDHP deductible (Exhibit 6)

Numerous studies have examined the effects of HDHP enrollment on healthcare demand; no clear consensus exists on whether HDHPs lower costs or by how much, though where HDHPs have been shown to lower costs, much of the effect appears to be temporary. Using the results of a recent paper[4] that did attribute a reduction in healthcare demand (14 percent) to first-time HDHP enrollment, and making no provision for the likelihood that HDHP enrollees’ demand would normalize over time, we developed a conservative (i.e. presumably over-) estimate of the effect of HDHP enrollment patterns on demand during the downturn (Exhibit 7). On this basis we estimate an average 19bp headwind to real annual per-capita healthcare spending growth since 2007. This level of headwind is non-trivial across long time periods but is far too small to explain the present low level of demand. Exhibit 8 shows the components of US healthcare demand growth by decade since 1960, with a focus on the 2009 – 2011 downturn (bold). The current (2009 – 2011) elasticity/mix trend is 1.6 percent below the 2000-2008 average, and fully 2.4 percent below the long-term average; this implies the conservatively estimated 19bp demand headwind from HDHP enrollment explains no more than one-eighth of the current slowdown

What do Valuations Imply?

The valuations backdrop poses an immediate dilemma: we have no direct way of knowing whether low expectations and valuations for healthcare names signal a general expectation of continuing weakness (or even further weakening) in the broader economy, or instead a more narrow belief that current weak healthcare demand is a secular ‘new normal’

Of the 156 US-listed healthcare companies with market capitalizations ≥ $1B that make up our universe, we view 38 of these as being more sensitive to volume trends than their average peer (Exhibit 9). These companies fall into several broad categories, including: manufacturers / distributors of non-prescription consumables, acute care facilities, commercial HMOs, drug retailers, drug wholesalers, diagnostic laboratories, and manufacturers / distributors of dental consumables.

As we’ve established, the immediate healthcare demand trend is dominated by cyclical influences. Across the healthcare universe earnings tend to be driven by some combination of volume, real pricing power, and new product flow – with volume being very sensitive to the economic cycle and pricing power / new product flow being very insensitive. It follows that healthcare names whose earnings power is more levered to volume than to pricing power or new product flow should underperform their peers as the economy weakens and demand falls, and outperform as the economy strengthens and demand improves

Volume-sensitive healthcare names generally have both revenue and earnings forecasts reflecting slower growth than our broader healthcare universe; and, with the exception of commercial HMOs these companies have tended to underperform their healthcare peers since the beginning of the economic downturn. This implies the market has either of two views – the broader economy will weaken further; or, current economic weakness will continue for the foreseeable future. And, valuations further imply that either of these two outcomes are more likely than the alternative scenario of (even modest) economic improvement

Bearing in mind that the cyclical component (elasticity/mix) of healthcare demand is coming off a 50-year low in 2010, we believe the scenario of at least modest demand improvement is at least equally likely vis-a-vis the alternatives of further economic weakening, or continued economic weakness sufficient to hold elasticity\mix at or near a 50-year low. This in turn argues that volume-sensitive names should trade very nearly on par with their healthcare peers, instead of the present discount

Of the more volume-sensitive subgroups, we believe that manufacturers / distributors of non-prescription consumables, and in particular the larger names in this category (BAX, BDX, COV) are the most attractive. Consensus revenue estimates for these names call for roughly 4 percent nominal revenue growth through 2013. Beyond the point that this reflects a multi-year extension of the lowest rate of real healthcare demand in measured US history, we would point out that buying into a 4 percent nominal top-line growth in turn implies (at 1% CPI) a long-term per-capita elasticity/mix trend of zero[5]. Further, implied productivity gains for these names are on par with the broader healthcare index, which runs counter to our belief that these volume-sensitive companies should have greater relative operating leverage as volumes (eventually) improve. Valuations (PE forwards through 2013) relative to healthcare and the broader SP500 are among the lowest in healthcare, with the exception of acute care hospitals and HMOs. The risk to our cyclical bet on these names is their greater relative exposure to ROW and in particular EU markets, which appear somewhat more likely than the US to see further economic weakening. To our minds this source of earnings risk is mitigated by the tendency of EU healthcare demand to weaken cyclically more in terms of price than volume, and by the tendency of pricing pressure to fall disproportionately on more innovative sellers whose products carry larger, more subjective, and thus more negotiable gross margins[6]

Commercial HMOs generally imply both slower top- and bottom-line growth than both the broader SP500 and our >=$1B market-cap healthcare universe. Even in the context of a continued slow economy, if per-capita healthcare demand stayed at its current 50-year low, HMOs would see roughly 2 percent real per-capita revenue growth, simply because of ongoing medical inflation. Gross margins have become far more stable – in large part owing to rebate requirements – this implies that as long as enrollment levels don’t fall, HMOs have a reasonable expectation of making a stable gross profit percentage on a revenue line that consistently grows faster (at the rate of per-capita medical inflation) than operating costs (growing at CPI). And this is an approximation of the worst case. As the economy improves, national per-capita demand also grows as the elasticity/mix trend recovers, but per-capita demand in commercial underwriting pools on average falls as (marginally more healthy) new hires drive smaller claims against the prevailing premium. We recognize that the gross margin effect of healthy new hires is capped by limits on medical loss ratios; however current estimates and valuations appear not to accept that the gross profit effects of hiring tend to more than offset the gross profit effects of a secular rise in the per-capita demand trend, and that HMOs see additional operating leverage as membership-driven revenue growth translates into earnings leverage as operating costs are more efficiently amortized. Despite having reasonable earnings prospects in a continued defensive scenario and growth potential in the context of an economic recovery, commercial HMO valuations (PE forwards) are 15 to 20 percent below broader healthcare and SP500 levels. We favor names with more commercial than Medicaid exposure (Exhibit 10) for the simple reason that an improving economy grows commercial but shrinks Medicaid enrollment. Of the larger commercial names we favor UNH and WLP over CI and AET largely because UNH and WLP tend to have large shares of their local healthcare markets where CI and AET do not, and we believe this is crucial to longer-term success under the Affordable Care Act (ACA)

Hospitals are perhaps the most sensitive to underlying volumes, though expectations here are weaker and valuations lower than anywhere else in the broader healthcare universe. And, concerns regarding hospitals’ middle-term pricing power appear to be compounding bearish sentiment, though we believe these concerns are largely misplaced. We recognize that the failure of the Deficit Reduction Act’s super-committee triggers provisions that should reduce Medicare reimbursement rates by roughly 2 percent in 2013, that hospitals are levered to Medicare reimbursement (Exhibit 11), and finally that hospitals cannot directly affect the Medicare pricing trend. However hospitals’ long-term offset to Medicare pricing pressures – namely commercial pricing – is very much available as an offset to pending Medicare rate cuts, and may in fact be more ‘available’ than normal. Commercial revenues are a sufficiently large percentage of total revenues (Exhibit 11, again) for revenue weighted commercial pricing changes to truly matter; and, the relationship between commercial and Medicare pricing levels is not so extreme (Exhibit 12) as to preclude a further growth in commercial prices. Most important of all we believe hospitals’ commercial pricing power grows as 2014 approaches, assuming the ACA remains largely intact. 2014 is a rare opportunity for commercial plans to gain large chunks of enrollment, but most candidate enrollees are not familiar with plans’ brand names. Candidate enrollees are however very familiar with names of local hospitals, and should (as appears to have happened in Massachusetts) tend to choose plans largely on the basis of which hospitals are or are not in a given plan’s network. In effect, local hospitals are the brand equity that drives enrollment gains in the 2014 land-grab, and any plan hoping to participate in 2014 enrollment gains cannot exclude desirable hospitals from its network(s) as 2014 approaches. In Massachusetts hospital pricing accelerated into the enrollment expansion associated with that state’s health reform, and we believe the same will be true nationally as we approach 2014[7] – thus all in, we believe hospitals have more than sufficient commercial pricing power to offset any Medicare pricing pressures that hit in 2013. Finally, and independent of the preceding commercial pricing arguments, we would point out an improving economy tends to improve not only hospitals’ volumes but their mix-adjusted pricing, as better-paying commercial payors displace lower-paying Medicaid

Manufacturers / distributors of dental consumables (e.g. PDCO, HSIC, XRAY, SIRO) may be more beneficially levered to cyclical changes in demand than healthcare companies broadly; our thesis being that both volume and mix in dentistry should be relatively more elastic than for non-dental medical care. Visits to dentists show more apparent cyclicality than visits to physicians (Exhibit 13); and, because dentists have a greater tendency than physicians to offer a range of treatment options, we suspect that average sales per dental visit also shows more cyclicality than for medical visits. Expectations for sales and earnings growth among the dental companies are slightly below those for the broader healthcare universe; however valuations are at a slight premium, thus the dental companies are only marginally attractive

Diagnostic laboratories (LH, DGX) are plainly levered to underlying volume demand. Revenue and earnings growth expectations are in-line (LH) to well below (DGX) broader healthcare, though valuations for both names are at a slight premium. Despite their volume leverage we’re neutral on the labs; valuations are not as attractive when compared to other volume-sensitive subgroups, and we have longer term concerns with respect to pricing[8]

Prescription volumes are plainly cyclical, which suggests the drug trades (drug wholesalers, drug retailers, and PBMs) as preferred subsectors given any expectation of improving demand. We’re nevertheless generally negative on the drug trades, particularly PBMs, in large part because we believe generic dispensing margins ultimately will fall, and that consensus expectations of expanding margins ultimately won’t be met

  1. To be clear, the prescription per-capita trend is not directly affected by the shift from brands to generics, i.e. a brand prescription in period 1 that becomes a generic prescription in period 2 simply shows up as a prescription in both periods. We recognize that promotional intensity is a function of companies having patent-protected brands to sell, and that promotional intensity has fallen as patents have expired. Having come from the branded drug industry, our experience is that patterns of care driven by promotional intensity (e.g. the penetration of statins into the population) are relatively ‘sticky’, such that the post-patent fall off in promotional intensity has a limited effect on total unit demand. On net, we see later lifecycle promotional intensity as more of a zero-sum contest between alternative therapies than a driver of overall unit demand growth
  2. For the record, patent losses have not produced a falling mix trend – even in pharmaceuticals narrowly, much less healthcare broadly. The average retail price of a prescription has risen consistently each year for as far back as we can measure, and certainly has risen throughout the 2009 – present slowdown in healthcare demand
  3. For more on this trend, please see: “Consumers Add Risk; Insurers Add Power … “ Sector & Sovereign Research LLC, July 29th, 2010
  4. Bunten, Haviland, et al “Healthcare Spending and Preventive Care in High-Deductible and Consumer-Directed Health Plans” Am J Manag Care. 2011;17(3):222-230).
  5. Taking 1 percent population growth and 40 bp aging effects as given, we would anticipate roughly 3.8 percent nominal growth in US healthcare demand if CPI were roughly 1% (v. a 2% steady-state), medical inflation held constant (v. 2000-2008) at 1.3%, and elasticity/mix held at zero (v. the 2000-2008 trend of 1.2%, and the 50-year trend of 2.0%). See Exhibit 8 for further context
  6. For more details, please see “Single-Payor (European) Governments and your Healthcare Portfolio” Sector & Sovereign Research LLC, August 24, 2010; and, our 11/15/2011 blog post “Dismal EU Growth Forecasts Mean Pricing Pressure on (Innovative) Healthcare Exporters”
  7. Details of hospitals’ role in the Massachusetts reform roll-out are in our December 12, 2009 call “The Biggest Reform Worry ….”; we’ll post these details on our healthcare blog in the coming days
  8. Commercial contract wins carry the spill-over benefit of Medicare volumes, since physicians using a given lab because of commercial insurance requirements will tend to send most or all of their practice’s tests – including Medicare patients’ tests – to that same lab. The result is that commercial pricing is in many cases much lower than Medicare pricing, and we feel this is inherently unstable in the longer-term
Print Friendly, PDF & Email