A Simple Formula for Drug (and Biotech and Spec Pharma) Stock Selection

eytan
Print Friendly, PDF & Email

Richard Evans / Scott Hinds

203.901.1631 /.1632

richard@ / hinds@ssrllc.com

twitter.jpg @SecSovHealth

September 6, 2011

A Simple Formula for Drug (and Biotech and Spec Pharma) Stock Selection

  • Therapeutic (drug, biotech and specialty pharma) stocks tend to gradually outperform their peers in the year or so preceding and following major anticipated product approvals. The exception is the period immediately surrounding the scheduled regulatory action, where risks are skewed to the downside
  • We have developed a simple set of portfolio construction rules intended to maximize exposure to share price performance in the months before and after major anticipated regulatory actions, while minimizing exposure to risks in the days surrounding these events. The rules are expressly designed to trigger buy and sell actions using only information that would be available to traders in a real-world setting
  • Backtested from 1996 through 1h2011, a portfolio constructed using these rules outperforms a comparator index of all publicly traded therapeutic stocks by 8% on average each year. The portfolio outperforms the comparator in 11 of 15 annual periods: in years the portfolio outperforms, average relative performance is 13%; in years the portfolio underperforms, average relative performance is -7%. On average, stocks held long in the portfolio represent 46% of the capitalization of stocks in the comparator index, and average annual turnover is 1.44
  • Our systematic, big picture view of therapeutics generally (and large cap pharma especially) is bearish; accordingly we face the challenge of building healthcare portfolios without over-exposing ourselves to these negative systematic risks. Despite our bearish systematic view, we continue to believe that new product flow can create new economic value and thus relative share price outperformance, thus our motive to find a set of portfolio construction rules that develop a portfolio with as much positive idiosyncratic risk (i.e. new product flow), and as little negative risk (broadly, exposure to earnings drivers outside of the immediate scope of new products; narrowly, exposure to asymmetric risk / reward that immediately surrounds regulatory actions), as possible. We believe our rules achieve this goal, and that our approach should offer better risk-adjusted performance than more traditional approaches to therapeutic (drug, biotech, spec pharma) stock selection

Summary

In previous work we established that drug stocks tend to outperform their peers in the two years preceding, and the two years following, anticipated[1] major product approvals. The exception to this general rule was the period immediately prior to the anticipated approval (or PDUFA[2]) date, where risks skewed to the downside as negative (non-approval) news tended to precede PDUFA deadlines

In this note we develop and evaluate a set of specific portfolio construction rules designed to exploit these patterns of share price behavior. We develop and test two rules, one each for the pre- and post-PDUFA periods, and measure the performance of separate and combined portfolios built using these rules. Back-tested from 1996 through 1h2011, the combined portfolio outperformed a cap-weighted index of large cap pharmaceutical, specialty pharmaceutical, and biotechnology stocks by, on average, 8% annually. In this back-test our portfolio outperformed the index in 11 of 15 annual (calendar year) periods; average outperformance during up years was 12.9%, while average underperformance during down years was -7.0%. On average, our portfolio held long positions equivalent to 45.6% of the comparator index, and average annual turnover was 1.44

Portfolio Construction

Candidate stocks include all US-listed[3] pharmaceutical, specialty pharmaceutical, and biotechnology stocks. The current capitalization of this universe is above $1.8T. Collectively these more than 500 stocks constitute the comparator index that we used to measure relative performance of our portfolios, and this comparator index behaves as (approximately) a capitalization weighted composite of the DRG and BTK (Exhibit 1)

Our portfolios are built around therapeutic stocks having thesis-relevant pending US product approvals with established PDUFA dates on (or by) which regulatory action can reasonably be expected. We define thesis-relevant as an NDA or BLA whose regulatory fate is material to the investment case for a given stock. Our dataset extends from 1996 through 1h2011, and includes 275 thesis-relevant regulatory actions involving 130 companies. The pre- and post-PDUFA periods are considered separate and distinct bets, and different rules are established for each period

The Pre-PDUFA Rule: Any therapeutic stock with a thesis-relevant new product (NDA or BLA) submission is included as a long position on the date the regulatory package is submitted. The weighting of any given stock in the portfolio is proportional to its relative capitalization as compared to all other stocks held in the portfolio; however the weighting rule is geared toward relative overweights of smaller capitalization stocks. Stocks having more than one thesis-relevant NDA or BLA submission are held at correspondingly higher weightings, i.e. a stock with two thesis-relevant submissions would be held at a weight equal to twice the weighting it would have been given with only one thesis-relevant submission. After being included in the pre-

PDUFA portfolio on the date a thesis-relevant NDA or BLA submission is filed, each stock is held until a period 75 days prior to the PDUFA date, unless a final regulatory decision is rendered before this period, in which case the stock is sold on the date of the final decision. If the long position remains intact (i.e. no final regulatory decision has been made) into a period beginning 75 days prior to the PDUFA date, that stock is held until the earlier of:

  • a final regulatory decision;
  • a share price gain of greater than a threshold percentage relative to the comparator index (since submission), conditional on a significant portion of the relative performance having occurred within a fixed number of days; or,
  • the PDUFA date

Exhibits 2, 3, and 4 compare this improved pre-PDUFA rule to a simple ‘buy on submission and sell on the PDUFA date’ strategy; the improved rule results in a greater probability of outperformance versus the comparator index (Exhibit 2), a greater magnitude of relative performance (Exhibit 3); and, more predictable performance (Exhibit 4). From 1996 to 2011, a portfolio constructed using the improved pre-PDUFA rule outperformed the comparator index by an annual average of 6.4%. The pre-PDUFA rule outperformed the index 75.7% of the time; in years the pre-PDUFA portfolio outperformed the comparator index average relative performance was 11.7%, in years the portfolio underperformed the comparator index average underperformance was -5.8%. On average the pre-PDUFA portfolio held long positions representing 27.0% of the comparator index market valuation, and average annual turnover was 1.21 (Exhibits 5, 6)

As-built, the pre-PDUFA rule does not incorporate Advisory Committee activities as a potential entry or exit point, simply because we have too few data points to rigorously evaluate trading strategies around these dates. However the general pattern of share price performance around Advisory Committees suggests an asymmetry of risk-reward comparable to that seen around PDUFA dates (Exhibit 7); thus our instinct is that real-world returns would be improved by treating Advisory Committee and PDUFA dates similarly. Specifically, from 90-days pre-Advisory Committee through 60-days pre-Advisory Committee; and 60-days pre-Advisory Committee through 30-days pre-Advisory Committee, stocks were more likely than not to outperform the comparator index. However, in the 30 day periods into- and out-of the Advisory Committee, stocks tended to meaningfully underperform versus the index

The Post-PDUFA Rule: Candidates for inclusion in the post-PDUFA portfolio include all US-listed therapeutic stocks who have received approvals or complete responses on thesis-relevant NDAs or BLAs. As with the pre-PDUFA rule, stocks held in the post-PDUFA portfolio are weighted in proportion to that stock’s capitalization relative to all other stocks held in the portfolio, with smaller capitalization stocks being held at relative overweights. For stocks having more than one thesis-relevant NDA or BLA that qualifies them for inclusion, that stock’s ‘base’ weighting is multiplied by that company’s number of qualifying, thesis-relevant NDAs or BLAs

Our previous work has shown that stocks tend to outperform their peers following both positive and negative final regulatory decisions on major products, but that timing of entry is crucial. In fact the simplest ‘buy on news’ post-PDUFA rule – buying stocks on the date of final regulatory decisions for thesis-relevant NDAs or BLAs – is more likely to produce relative (to the comparator index) losses than gains. We improve on this by requiring both a final regulatory decision and a meaningful share price reaction within 30 days of the final decision. Mathematically we define a meaningful share price reaction in terms of relative (again to the comparator index) volatility over a brief (days) look-back period, and this inclusion criterion is agnostic to the direction of the share price movement. Combining these, the inclusion rule for the post-PDUFA portfolio is a final regulatory decision on a thesis-relevant NDA or BLA, combined with a meaningful share price reaction within 30 days of that decision. Stocks in the post-PDUFA portfolio are held until the earlier of outperforming the comparator index by a threshold percentage, or 360 days following the final regulatory decision

Exhibits 8, 9, and 10 compare this improved post-PDUFA rule to a simple ‘buy on news’ strategy; the improved rule results in a greater probability of outperformance versus the comparator index (Exhibit 8), a greater magnitude of relative performance (Exhibit 9); and, more predictable performance (Exhibit 10). From 1996 through 1h2011, a portfolio constructed using the improved post-PDUFA rule outperformed the comparator index by an annual average of 7.2%. The improved post-PDUFA rule outperformed the index 70.0% of the time; in years the post-PDUFA portfolio outperformed the comparator index average relative performance was 15.1%, in years the portfolio underperformed the comparator index average underperformance was -10.4%. On average the post-PDUFA portfolio held long positions representing 18.6% of the comparator index, and average annual turnover was 1.71 (Exhibits 11, 12)

The Combined Pre- and Post-PDUFA Portfolio: This portfolio is a simple combination of the pre-PDUFA and post-PDUFA rules. From 1996 through 1h2011, a portfolio constructed using the combined pre- and post-PDUFA rules outperformed the comparator index by 8%. The combined portfolio outperformed the index 76.4% of the time; in years the portfolio outperformed  the comparator index average relative performance was 12.9%, in years the portfolio underperformed the comparator index average underperformance was -7.0%. On average the portfolio held long positions representing 45.6% of the comparator index, and average annual turnover was 1.44 (Exhibits 13, 14)

Why (We Believe) These Rules Work

Systematic ‘all boats float’ drivers of revenue and earnings tend not to drive relative performance across a universe of therapeutic stocks; relative performance is instead driven by idiosyncratic, stock-specific fundamentals, the most powerful of which is new product flow. As systematic drivers of performance (in particular, real pricing) are fading, we believe that new product flow is becoming an even more dominant feature of therapeutic companies’ earnings (and thus relative performance) potential

Broadly, we’ve shown previously that therapeutic stocks tend to outperform for long (+/- 2 year) periods before and after anticipated regulatory actions on thesis-relevant new products, but that the period immediately surrounding the anticipated decision offers more downside than upside. We believe that our pre- and post-PDUFA rules both ‘work’ (in terms of producing relative  performance) because they offer quite a lot of access to the relative performance gains before or after the anticipated approval, but mitigate the share price risks immediately around the scheduled approval date

The pre-PDUFA rule buys on submission and holds until 75 days before the PDUFA date, unless a decision is reached before this point, in which case the stock is sold. Within a window beginning 75 days prior to the PDUFA date, the stock is held until the earlier of a decision, a relative performance gain (measured from submission v. the comparator index) of more than a threshold percentage (conditional on much of this gain occurring in the immediate trailing period), or the PDUFA date. Presumably large share price reactions in the immediate period (which we define arbitrarily as 75 days) preceding PDUFA are quite likely to be associated with regulatory expectations having been either met, or missed. And, because we believe the market generally is pricing in its expectation of a largely binary event (approval v. non-approval), we tend also to believe that selling once the market prices in the expectation of a positive event is inherently logical, as that stock is unlikely to offer substantial further relative outperformance before the PDUFA date, but still offers the risk of a regulatory failure. More simply, the discipline of ‘booking’ relative performance gains ahead of the PDUFA date tends to thin the portfolio as the PDUFA date approaches, which is beneficial in that risks appear to shift to the downside in the final days preceding a scheduled regulatory action

The post-PDUFA rule buys only after both a final regulatory decision and a meaningful share price reaction that follows the regulatory decision by no more than 30 days. In effect the rule limits exposure to ‘peri-PDUFA’ share price risks by allowing the stock to have its immediate reaction to the final regulatory decision before the stock is included in the portfolio. Because immediate share price reactions to regulatory news appear to be on average value destroying, the post-PDUFA rule effectively skips these reactions, but exposes the portfolio to the longer-term relative performance gains associated with final regulatory decisions

Real World Utility of the Rules

We believe we have controlled for selection biases that would explain the investment performance of our rules. Most importantly, our rules tend to select smaller capitalization stocks. The primary inclusion criterion is for a company held long to have a ‘thesis-relevant’ NDA or BLA submission, and the odds of a given submission being thesis-relevant rise as the company’s sales base falls – i.e. a $500M peak sales drug is not thesis-relevant to PFE or MRK, but is very thesis-relevant to many smaller capitalization companies that have smaller sales bases. And, because we measure performance relative to a cap-weighted index (even though our weighting rules ‘up-weight’ smaller capitalization stocks) , one could argue that our performance gains are largely due to the fact that we simply tend to over-select and/or overweight smaller capitalization stocks that are ‘more BTK-like’, and therefore our portfolios benefit by tracking more closely to the BTK, which outperformed the DRG over the look-back period. We examine this potential bias in 3 ways. First, we examine the effect of increasing the representation of smaller capitalization stocks in our portfolio by increasing their weights – e.g. where a very large cap name might be weighted in exact proportion to its market cap relative to the cap of other stocks held long, a smaller capitalization stock will be held at some multiple of the ratio of its capitalization to the capitalization of all other stocks held long. We increase the smaller capitalization bias to a point at which the returns in our portfolio more or less mirror the returns of the BTK, and assess whether our portfolio ‘delivers’ this BTK-like performance with less volatility – which it does. Over the 1996 through 1h2011 period, if we fine tune weighting in favor of smaller capitalization stocks, we can effectively match BTK-like share price performance, though our portfolio has a standard deviation of returns of 0.20 versus a standard deviation of returns for the BTK of 0.50 (Exhibit 15). Second, we build a portfolio entirely from ‘BTK-relevant’ stocks, i.e. from stocks whose characteristics are better captured by the BTK than by the DRG. In this case, our rules-based portfolio has substantially higher returns over the look-back period than the BTK (30.0% v. 16.7%), though our portfolio also has marginally greater risk (SD of returns 0.52 v. 0.50, again Exhibit 15). Third, and most comprehensively, we examine the risk-return relationship of our rules-based portfolios as the weighting rule is adjusted from one extreme (simple cap weighting of all stocks) to the other (equal weighting of all stocks). Our rules offer a spectrum of risk / return (back-test) results that are superior to any of the DRG, SPX, our comparator index, or the BTK (Exhibit 16); accordingly we conclude that the rules-based portfolio

generates attractive (back-tested) investment performance, and for reasons other than its tendency to select smaller capitalization names

Turnover in the portfolios is moderate, from 1996 through 1h2011 averaging 1.21, 1.65, and 1.44 turns per year in the pre-PDUFA, post-PDUFA, and combined portfolios, respectively

Individually or combined, the rules tend to hold a fairly substantial percentage of ‘available’ market capitalization: from 1996 – 2011ytd on average the pre-PDUFA, post-PDUFA and combined portfolios held, respectively, stocks representing 27 percent, 19 percent and 46 percent of the comparator therapeutic index

Critically, beyond the largely static probability that smaller capitalization names are more likely to have thesis-relevant regulatory submissions, we see a trend across the back-test period of falling average capitalization of stocks selected by our rules – during 2007-2011ytd, the average shares of total capitalization held in the three rules are less than 45 percent of what they were ten years prior (Exhibit 17). This suggests that the weighting rules derived across the entire back-test period may need to be tempered (i.e. the up-weighting of smaller cap names reduced) going forward, until and unless the relationship between capitalization and thesis-relevant submissions approaches the back-test period average. This also raises the risk that portfolios built using our rules may become so narrow that volatility and/or tracking error reach unacceptable levels, though for the moment we believe the portfolio still has sufficient breadth (# of names) and depth (% of comparator index capitalization)

The question of where in the investment decision our rules apply deserves careful consideration. At one extreme we might use these rules and no other factors to choose names held long. At the other extreme we might use more traditional means of choosing which stocks to hold, and use these rules as a tie-breaker to select from a short-list of candidates. Our bias is closer to the former (wholly rules-based) extreme, for the simple reason that relative returns for individual stocks selected by the rule are somewhat ‘narrow and noisy’[4], which implies that the rules-based ‘bets’ must be made consistently across multiple opportunities in order for the rules to have a reasonable aggregate probability of producing attractive returns

Appendix:

The Current Rules-Based Portfolio

Appendix: Technical Comments

In constructing our dataset of regulatory submissions, we used very strict inclusion criteria which obviously reduced the number of observations at our disposal – the analytical cost of a more accurate, specific and robust dataset. Specifically, we screened for investment thesis-relevant NDA / BLA submissions for which we could obtain submission and PDUFA dates to the calendar day – this screen eliminated a large number of events for which our regulatory data source either was missing observations, or was only able to track events to the month/year of occurrence. Thus the analysis implicitly assumes that – or technically is limited to submissions for which – regulatory events are publicly disclosed as they occur. Further, for the pre-PDUFA rule we excluded any of this subset of events for which the PDUFA date was less than 150 days, or more than 540 days, after submission – presuming an anomalous event or an error in the data. Also note that we have tracked performance associated with each event for just one stock, based only on the listed original applicant and not on any licensees, rights-acquirers, royalty recipients, etc.

  1. Stocks tended to outperform ahead of definitive news related to both regulatory failures and successful product approvals. After definitive (either good or bad) news was priced in, stocks tended to outperform from the post-news baseline
  2. Prescription Drug User Fee Act
  3. And some significant, non-US listed stocks, e.g. RHHBY
  4. We found that returns from following the pre-PDUFA and post-resolution rules were on average in the 58th and 61st percentile respectively. This certainly suggests that a stock within one of the two windows is more likely than not to outperform a randomly chosen alternative innovator – however, we are plainly hesitant to extrapolate to this result, given that (1) choosing between two stocks viewed as comparable save for a proximate regulatory event is not a random draw; and (2) basic probability / portfolio theory give us little confidence that small samples (and samples do not get any smaller than a single stock position) will generate results consistent with characteristics of the population
Print Friendly, PDF & Email