The Skeptic’s Guide to Drug Stock Selection
September 22, 2010
The Skeptic’s Guide to Drug Stock Selection
- Despite our enduring large-cap pharma bear hypothesis, we recognize this sub-sector dominates healthcare capitalization, and so cannot be entirely avoided in most healthcare portfolios
- Our pharma bear thesis anchors to the mis-match between companies’ structures and secular/systematic risks; we seek to reduce exposure to these big-picture risks (all of which we view as unfavorable) by loading up on favorable idiosyncratic risks: generally speaking this means: concentrated positions, smaller companies, younger product portfolios and a preference for biologics over small molecules
- Plainly our approach rejects intra-sector mean reversion as a basis for drug stock selection (we’ve shown elsewhere this tends not to work), but embraces the notion that companies with improving product mix can be bought cheaply over long-time periods, rather than only ahead of major product-related news such as regulatory filings / approvals
- We examine relative (to peers) share price performance of companies for two years preceding and two years following scheduled major product approvals; with the exception of asymmetric risks and rewards in the 30 days before a scheduled regulatory action, stocks tend to outperform for the entire period – both before and after a major product approval. This supports the notion that stocks with improving product mix can be bought cheaply over long time periods
- To complete the view of relative performance around mix shifts we examine patent expiries as well, and find that stocks show no clear relative performance pattern before or after the loss of major products to generic competition – all of which suggests the market is reasonably efficient at pricing anticipated product losses
- LLY, BMY, ABT, & Roche are favored for smaller size, younger product portfolios, relative adequacy of near-term pipeline (filed products) and/or a focus on biologics; for their standings on these same metrics our thesis argues against PFE, NVS, JNJ and GSK
We continue to hold our long-standing bearish view of the large-cap drug industry. Structurally, we see an industry whose comparatively large and inflexible cost base is unsuitable to the slower growing and more volatile revenue patterns in the industry’s near future. Moreover we see new product flow generally failing to offset product losses in the near- to mid-term, as the visible components of future product mix (late-stage and filed products on the one hand, and expiring patents on the other) plainly point toward net erosion. Prospects for mix in the longer-term are challenging as well, particularly as returns on R&D spending appear to be well below costs of capital. Environmentally, we expect continued pricing pressure in the export markets near-term, know that current very high real rates of US pricing almost certainly cannot go higher, and ultimately cannot be sustained; and, are convinced that the Independent Payment Advisory Board, which begins operating in 2015, is very likely to result in government setting of prices for products used by Medicare beneficiaries. Even in the face of these pressures, we still believe that the industry has effective options for improving its position (smaller more flexible cost structures and simple optimization of developmental portfolio risks lay at the heart of these), though recent opportunities to adapt structurally (PFE’s integration of WYE, MRK’s integration of SGP) have been missed, which hints that the sub-sector’s structural shortcomings will persist, along with its relatively poor earnings performance.
Despite our bearish view, we recognize that the sector represents more than half of total healthcare capitalization, making it practically impossible for many investors to avoid the sub-sector entirely – thus the purpose of this note is to explore the best option(s) for integrating pharmaceuticals exposure into broader healthcare portfolios.
We begin by recognizing that our bear hypothesis is grounded entirely in the mis-match between industry structure and secular / systematic risks, but that individual companies can still largely (though temporarily) escape this trap by simple virtue of improving their product mix. In very general terms this argues for:
- buying smaller rather than larger companies — companies with very large market shares by definition tend to be relatively over-exposed to secular / systematic risks — and, the impact of new products on mix is diluted less at small companies;
- buying companies with younger product portfolios wherein products are more likely to be on the upswing of a lifecycle than at or beyond maturity; and,
- buying companies with a relatively large percent of total sales coming from biologics, as biologics lifecycles are far less likely than small molecule lifecycles to be ended by generic competition.
These rules notably ignore mean reversion as a basis for intra-sector stock selection, for the simple reason that mean reversion appears not to work. Quantitatively, within the sub-sector buying lower valuations and/or selling higher valuations does not produce reliable returns; all we’ve ever been able to show is that the combination of extreme (high or low) relative valuation AND high relative estimate dispersion signals a subsequent large relative share price move – though there is no advanced signal as to whether this will be a gain or loss. More qualitatively, we’ve elsewhere shown that ALL evaluable instances of substantial (>= 20%) relative performance within the sub-sector (since 1975) can be explained by clearly superior fundamentals, positive relative earnings revisions, a proximate major product approval, or some combination of these. In short, cheap drug stocks don’t become expensive – or vice versa – by chance; substantive changes in relative valuation within the group consistently tie to idiosyncratic, stock-level fundamentals. It follows that pharmaceuticals stock selection should bias fundamentals over valuation.
Still, for our general ‘rules’ to work, we have to show that changes in mix are associated with corresponding changes in relative performance, and that relative performance unfolds gradually rather than immediately – so that opportunities to buy or sell remain ‘open’ long enough to be truly relevant to portfolio construction. Aside from certain highly unpredictable events (e.g. safety-related product withdrawals or manufacturing violations), the most impactful product mix-related events arguably are the beginning and end of a major product’s lifecycle. To better understand how valuations perform around these events, we examined patterns of relative drug stock performance at both ends of the product lifecycle – product approvals (i.e. PDUFA dates), and the first entry of a major generic competitor.
Starting with PDUFA dates – using data looking back twenty years, we analyzed the relationship between anticipated or recent final US regulatory action on a major new molecular entity (NME), and the performance of that company’s stock relative to its peers. Our dataset includes 198 product approvals (positive events), and 82 ‘negative events’ (NDA withdrawal, receipt of a non-approvable letter, or receipt of an approvable letter when approval would have been expected). In each instance, we analyzed the relevant drug company’s relative share price performance either over a four-year period, beginning two years prior to the anticipated approval date and ending two years after; or, for as long as possible before and after the scheduled action date if four years of data were not available.
n the period preceding anticipated approval, drug companies whose major products were in fact approved outperformed their peers at each of the points analyzed (Exhibit 1). (Please bear in mind that all of the relative performance values in this call are cumulative, rather than annualized, and that accordingly, dates further away from the event tend to have much larger relative performance values. On an annualized basis, at each point preceding the PDUFA, relative performance values are roughly the same.) Also, at each point pre-approval the odds of the successful company’s stock outperforming its peers were strongly positive, though these odds fell as the approval date approached. And, after the approval date, shares of companies receiving approval of major new drugs continued to outperform at a substantial rate, though outperformance was slightly weaker than before approval.
Drug companies whose major products were not approved as expected also outperformed their peers pre-approval, with the obvious exception of the 30 day period immediately preceding the scheduled action (Exhibit 2). Even omitting the negative relative performance just before expected approval, pre-decision outperformance was roughly half as strong (on an annualized basis) as for companies whose products received positive decisions. Odds of outperforming were favorable at 730 and 545 days preceding the decision (61 and 56 percent respectively), but unlike companies that received positive decisions, fell below 50 percent in each subsequent period preceding the decision. The share price performance consequences of a negative decision were substantial, but brief. On average these stocks underperformed peers by 3.3 percent between 30 days of the scheduled decision and the date itself; which reflects the tendency of bad news to come out just before the scheduled action date. Note that the performance penalty associated with a negative decision tends to become smaller as we move past the action date; in effect this implies the stocks are in fact outperforming modestly from a post-decision low.
Obviously, in the period preceding the regulatory action date, it’s impossible to know with certainty the outcome of the pending decision – so the risks and returns to investors in the pre-decision period are of course a blend of the relative likelihoods of positive or negative decisions. Averaging the positive and negative pre-decision returns is complicated by the fact that we can’t definitively set the relative odds of a decision being good or bad. Approval rates vary from year to year, but generally speaking we know that slightly less than half of NME submissions are approved in their first review cycle. If we average returns during the pre-return period assuming – in line with historic averages of first round approvals – that odds of a positive or negative decision are 47% and 53% respectively, then we see the expected returns, and odds of positive returns, shown in Exhibit 3. This almost certainly under weights the odds of a positive event, as companies tend to signal, and investors price in, greater odds of an extended review for many of the products facing higher levels of regulatory risk. Accordingly, we expect that the actual approval odds priced in would lie somewhere between the odds of first round approval and the odds of ‘eventual’ approval (roughly 80% positive / 20% negative). For arguments’ sake, we solved for the approval odds that would produce a zero expected return in the 30 days preceding the regulatory action date – these odds are 70% probability of a positive decision, and 30% probability of a negative decision (Exhibit 4). Using these odds, companies with major product decisions pending were more likely than not to outperform as the decision approached, though the odds of outperformance fell as the scheduled action date drew closer. Ironically, in all cases – whether or not the product was approved – companies tended to outperform their peers immediately following a regulatory action.
These data support several conclusions; for our immediate purposes the most important is that major product approvals have a substantive and long-lasting positive impact on relative performance. This is consistent with our recommendation to buy stocks with improving mix, and argues against the competing notion that the impact of major product approvals is ‘priced in’ at a very early stage of the product lifecycle, as we’re still seeing substantial outperformance two years after product approval. On a more tactical level, we note that risks and rewards are asymmetric in the period immediately preceding a scheduled regulatory action date – the (annualized) relative performance of positive events is not significantly greater in this period than in earlier pre-approval periods, but the odds of a negative event are concentrated in this period, and the average under-performance tied to negative events is at least triple the average out-performance potential of a positive event. Also on a tactical level, and in contrast to the share price consequences of product approval, we note that the negative consequences of a product failure tend to be priced in excessively and immediately – consistent with the traditional notion that immediate share price reactions to product failures tend to be overdone. Relative performance penalties for product failure fade quickly, which is to say that buying these shares on the negative news has generally led to outperformance in the following periods.
Turning to the far end of the product lifecycle – patent expiry – we examined the relative share price performance of companies before and after the loss of a major product (or products) to generic competition; our dataset includes 121 events over a 20-year look-back period. As we did with pending regulatory actions, we examined relative performance at discrete time points surrounding the event of interest, from two years before the entry of generic competition until two years after.
Somewhat counter-intuitively, companies facing major product losses tended to outperform their peers in the period preceding generic entry, though the relative performance gap was very small (Exhibit 5). In the 180 days following patent expiry, whether the relevant stock outperformed appears to be random, and the relative (under)performance gap also was quite small. Further out, stocks that had suffered major product losses on average outperformed, though again by a small margin.
We struggle to find a clear pattern in share price performance around LOE, with the exception of the high probability of under-performing (though by a very small margin) in the 30 days immediately after generic entry. In effect, this argues that the market ‘prices’ generic entries reasonably well, on average. This stands to reason, in that LOE’s tend to be far more ‘concrete’ events than PDUFA dates (with very few exceptions we know for certain a generic launch will occur); and, the earnings consequences are much easier to forecast (we’re subtracting a product of known sales value, and reasonably estimable earnings value, rather than adding a product of unknown sales and earnings value). We recognize this conclusion appears to run up against an argument we’ve recently made – that consensus appears to be under-estimating the gross margin consequences of pending major product losses. We’d revise that to say that sell-side estimates are probably too generous with respect to post-patent gross margin, but that history suggests the buy-side sees and prices in a more ‘complete’ erosion of gross margin – and, by extension, that we have no empiric basis for arguing that the shares of companies facing major LOE’s should be avoided simply because of the LOE.
Summarizing, we believe that pharmaceuticals stock selection should seek to maximize favorable idiosyncratic risks, and minimize secular / systematic risks – all of which we view as unfavorable. This argues in favor of concentrated positions in smaller companies, rather than large numbers of stocks and/or a bias to larger stocks, either of which raises exposure to secular/systematic risks. Using share price performance around major product approvals, we’ve shown that relative share price gains are sufficiently large — and occur over sufficiently long time periods — to be exploitable in portfolio construction. The considerable benefits of new product approval have in the past, on average, remained ‘on-sale’ for an extended time.
Beyond the specifics of share price performance around PDUFA dates, we see other more general reasons for favoring companies with improving product mix. The sector faces immediate (Europe) and mid-term (US Independent Payment Advisory Board) threats to real pricing power; and, as we argued earlier, current high real rates of US pricing are unlikely to go higher, and eventually must fall. It follows that we should prepare for falling real pricing power, and we believe that companies with improving product mix are far more capable of growing revenues in the face of pricing pressure than companies with either static or declining mix. Returning to where we started, we also believe that company’s structures are too large and inflexible in light of future (slower and more volatile) revenue patterns. Companies that can grow revenues despite pricing pressure (companies with mix gains) delay the earnings consequences of relative large and inflexible cost structures; conversely, as real pricing power fades, companies with mature to declining product mix face these consequences much more quickly.
Exhibit 6 presents a ‘rough’ screen of the large-cap pharmaceuticals universe according to four dimensions: size (we favor smaller), biologics exposure (we favor more), revenue weighted portfolio age (we favor younger), and the number of new molecular entities (NME’s) filed for approval (of course we favor more). LLY, BMY, ABT, & Roche are favored for smaller size, younger product portfolios, relative adequacy of near-term pipeline (filed products) and/or a focus on biologics; for their standings on these same metrics our thesis argues against PFE, NVS, JNJ and GSK.
- Prescription Drug User Fee Act
- “Independent Evaluation of FDA’s First Cycle Review Performance – Retrospective Analysis Final Report” Booz Allen Hamilton, January 2006; and “Performance Report to the President and the Congress for the Prescription Drug User Fee Act” FY 2008, www.fda.gov
- “Gross Margin Expectations Too High for Pharma …” Sector & Sovereign Research LLC, January 12, 2010