Drug & Biotech Companies with Undervalued Pipelines: Updated (and Expanded) List

Print Friendly, PDF & Email

Richard Evans / Scott Hinds / Ryan Baum


203.901.1631 /.1632 / .1627

https://twitter.com/images/resources/twitter-bird-blue-on-white.png richard@ / hinds@ / baum@ssrllc.com


July 6, 2015

Drug & Biotech Companies with Undervalued Pipelines: Updated (and Expanded) List

  • Phase 2 and earlier pipelines account for more than a third of larger biopharma companies’ market values; however because the market knows much less about these ‘hidden’ pipelines than about companies’ ‘non-hidden’ operations (e.g. existing, filed, or phase 3 projects), we believe hidden pipelines drive substantially more than a third of relative share price performance
  • Using patent data, we can better determine what’s actually in companies’ hidden pipelines, and by extension whether share prices fairly account for hidden pipelines’ contents. Since 2Q13, stocks identified as undervalued using this method have outperformed the peer group by 1.6x – and have in fact outperformed the peer group in every quarter
  • Our patent-based method is very good at establishing the relative amounts of innovation in companies’ pipelines, but does little to establish where companies stand on the timeline of bringing these innovations to market
  • To fill this gap, we’ve begun using a systematic analysis of clinicaltrials.gov data to identify products that are moving along in clinical development, but not yet captured by consensus or reflected in share prices. On a back-tested basis, stocks selected using our clinicaltrials.gov-based method outperformed peers by 2.9x over the last 8 quarters, as compared to the 1.6x prospective outperformance achieved with the patent-based method
  • The patent-based and clinicaltrials.gov-based methods tend to choose different stocks; in effect each method tends to choose stocks the other method would likely miss. Accordingly, we intend to publish portfolios of both methods separately, as well as a combined portfolio of all stocks selected by either method
  • Stocks presently screening as undervalued using the patent-based method are BMY, SNY and VRTX; stocks presently screening as undervalued using the clinicaltrials.gov-based method are AMGN, CELG, and GILD

Where we’re BULLISH: Biopharma companies with undervalued pipelines (e.g. VRTX, BMY, SNY, ROCHE, AMGN, CELG, GILD); Biopharma companies with pending major product approvals (e.g. ALIOF, ALKS, AMGN, BDSI, ENDP, HLUYY, HSP, ICPT, JAZZ, NVS, PTCT, RLYP, RPRX, TSRO, UCBJY, VRTX); ABBV and ENTA on sales prospects in Hep C; SNY on undervalued basal insulin franchise and sales potential for Praluent (alirocumab), in addition to its undervalued pipeline; AZN and LLY on the likelihood that excess SG&A/R&D spending must be reined in, in addition to pending major product approvals; CFN, BCR, CNMD and TFX on rising hospital patient volumes; XRAY and PDCO on rising dental patient volumes and rising average dollar values of dental products and services consumed per visit; CNC, MOH and WCG on bullish prospects for Medicaid HMOs; and, DVA and FMS for the likely gross margin effects of generic forms of Epogen; RAD as an acquisition target as WBA and CVS seek to defend against narrowing retail networks

Where we’re BEARISH: Biopharma companies with overvalued pipelines (e.g. GILD, ALXN, SHPG, REGN, CELG, NVO, BIIB); PBMs facing loss of generic dispensing margin as the AWP pricing benchmark is replaced (e.g. ESRX, CTRX); Drug Retail as dispensing margins are pressured by narrowing retail networks and replacement of AWP (e.g. WBA, CVS, RAD); Research Tools & Services companies as growth expectations and valuations are too high in an environment of falling biopharma R&D spend (e.g. CRL, Q, ICLR); and, suppliers of capital equipment to hospitals on the likelihood hospitals over-invested in capital equipment before the roll-out of the Affordable Care Act (e.g. ISRG, EKTAY, HAE)

Premise and rationale

Share prices of drug and biotech stocks can be broken down into two major components: the capital markets’ estimates of the value of assets that are well characterized (i.e. on-market, filed, and late-stage developmental projects), plus the markets’ estimates of the value of assets about which relatively little is known (i.e. pre-phase III projects). For simplicity, we refer to early, pre-phase III developmental projects collectively as the ‘hidden pipeline’

To estimate the market capitalizations of hidden pipelines, we subtract from each company’s enterprise value (EV) an estimate of the present value of everything else – marketed products, products filed for regulatory approval, products in phase III development, and all non-pharma lines of business. On average, hidden pipelines account for about 38 percent of companies’ EVs, and the percentage of EV explained by hidden pipelines ranges from a low of 6, to a high of 74 percent (Exhibit 1)

Even though hidden pipelines are less than half of companies’ market values, we believe they account for a majority of the mid- to longer-term change in these companies’ share prices. Because the markets have substantially more information regarding non-hidden operations than they have regarding hidden pipelines, the valuations of hidden pipelines are much rougher guesses. By extension, this implies that hidden pipeline valuations are far more volatile than the valuations of non-hidden operations, and by further extension that hidden pipelines explain substantially more than one-third of relative performance among these stocks. It follows that anchoring biopharma stock selection to sound estimates of ‘true’ hidden pipeline values is likely to generate outperformance

How we establish relative values of hidden pipelines

Having solved for the market capitalization of a company’s hidden pipeline, the question becomes whether or not the assigned market value closely reflects the value of what’s actually in the hidden pipeline

Since 2Q13, we’ve relied exclusively on patent data to estimate the relative value of hidden pipelines. This method has worked extremely well; in every quarter since 2Q13 stocks identified as undervalued using this method have outperformed stocks with fairly (or over-) valued pipelines by a cumulative 1.6x (Exhibit 2, top rows)

Despite the proven usefulness of relying solely on patent data, we’ve taken steps to further improve our hidden pipeline valuation process, for the following reason: Patents offer a very thorough inventory of companies’ research activities, but offer virtually no insight into stages of development. For example a company might have a very substantial lead over its peers in terms of the number of quality-adjusted patents it holds in therapeutic category X, but could still be well behind its peers in terms of category X products in advanced stages of development. Because sell-side analysts tend to publish forecasts for meaningful projects in mid- to late-stage clinical development, and because companies must disclose all clinical trials (on clinicaltrials.gov) for all products for which they intend to seek US regulatory approval, we felt reasonably comfortable assuming that share prices would faithfully reflect full(ish) information about companies’ mid- to late-stage projects, thus making patent analysis a fairly comprehensive means of explaining everything that wasn’t yet factored into share prices

As it turns out, consensus estimates fail to capture very large portions (about 88%) of the clinical development activity that companies disclose (Exhibit 3). As it also turns out, by making use of the clinical development activity that consensus ignores, we can reliably identify undervalued biopharma companies – even if we ignore patent-based estimates of pipeline value (Exhibit 2, middle rows)

Exhibit 2 shows that stocks selected using clinicaltrials.gov data performed even better (2.9x peers) than stocks selected using only the patent data (1.6x peers), though at a substantially higher s.d. of returns. N.B. the performance comparison is not apples : apples – patent-selected stocks’ performance was achieved prospectively by an algorithm we set in motion in 2Q13; in contrast the clinicaltrials.gov selected stocks’ performance is a back-test of an algorithm we’ve developed over the last several months. Thus despite the performance differences shown in Exhibit 2, it’s important to realize that in a very real sense the performance record of the patent-selected stocks is more firmly established

Exhibit 2 (bottom rows) also shows that a combined portfolio, holding long all stocks identified by either method, yields performance (2.7x peers) nearly on par with the stocks selected using only clinicaltrials.gov data (2.9x peers) though at a substantially lower s.d. of returns (0.05 combined v. 0.09 for clinicaltrials.gov alone). Indexed performance data are shown graphically in Exhibit 4

N.B. the two strategies generally choose different stocks. Over the last 8 quarters, only 10% of the stocks chosen by the patent-based strategy were also chosen by the clinicaltrials.gov strategy (Exhibit 5). Arguably this is a good thing – because either strategy alone only identifies only about one-fifth of the stocks that outperform in any given quarter, combining the relatively non-overlapping strategies results in a combined strategy that identifies a substantially larger percentage of outperformers (Exhibit 6)

For these reasons, we believe the patent-based and clinicaltrials.gov-based methods are complimentary, and our plan is to use both methods going forward. We’ll continue to publish the patent-based portfolio on its own, but will also publish the clinicaltrials.gov-based portfolio, as well as a portfolio combining the two methods

Hidden Pipeline valuation using quality-adjusted patents

We assign all active[1] patents to their corresponding parent companies, then take three key additional steps: 1) each patent is quality weighted[2]; 2) patents are segregated into those that correspond to either known (e.g. marketed, filed, or late stage projects) or hidden (pre-phase III) products / projects; and 3) each ‘hidden’ patent is assigned to disease, mechanism, and/or physical (chemical / biochemical characterization) categories[3]. The result is a company-by-company database that characterizes the sizes (numbers of quality-adjusted patents) and relative values (quality-adjusted patents, weighted further by the relative economic values of the disease areas in which the company conducts research) of the analyzed companies’ hidden pipelines

To determine which companies’ pipelines appear relatively over- or undervalued, we simply compare the market capitalization (price) of each company’s hidden pipeline to the apparent (relative) economic value of that company’s hidden pipeline

Exhibits 7 and 8 summarize the results. In Exhibit 7, column (b) provides our estimate of the market capitalization of each company’s hidden pipeline[4], and for reference column (c) shows hidden pipeline capitalization as a percent of total enterprise value. Column (d) gives each company’s hidden pipeline value as a percent of the total hidden pipeline value for all companies. Using BMY as an example, BMY’s $49B hidden pipeline capitalization is about 4.5 percent of the $1,090B combined capitalization for all 22 companies’ hidden pipelines. In column (e) we express the quality- and sales-weighted amount of innovation in each company’s hidden pipeline, as a percent of the total quality- and sales-weighted innovation in all 22 companies’ hidden pipelines (BMY has 10.4 percent of total innovation across the 22 companies, but only 4.5 percent of the market value). Column (g) is the ratio of columns (e) and (d); i.e. column (g) is the share of peer group innovation in a given company’s hidden pipeline, divided by that company’s share of total peer group hidden pipeline market value. Companies with larger shares of innovation than of market value (e.g. BMY) have hidden pipelines that are apparently undervalued, and vice versa[5]. Column (f) sales weights the shares of hidden pipelines depicted in column (e); and column (h) calculates the ratio of columns (f) and (d); i.e. the share of peer group sales weighted innovation in a given company’s hidden pipeline, divided by that company’s share of total peer group hidden pipeline market value. Exhibit 8 depicts the data in Exhibit 7 graphically, comparing each company’s share of the peer group’s quality-adjusted hidden pipeline (y-axis), to its share of the total peer group’s hidden pipeline capitalization (x-axis). Companies that depart significantly from the 45 degree ‘normal’ have hidden pipelines that are apparently under- (above the normal line) or overvalued (below the normal line)

AZN, BAYER, BMY, GSK, JNJ, MKGAY, ROCHE, SNY, and VRTX have hidden pipelines that appear undervalued[6]; ALXN, BIIB, CELG, GILD, LLY, NVO, REGN, and SHPG have hidden pipelines that appear overvalued[7]

Implied performance, and actual performance to date

Implied performance is simply the percentage change in relative share price that would bring a given company’s hidden pipeline valuation to par. Fundamentally, the idea is that a given hidden pipeline, adjusted for the sales potential reflected in its therapeutic area mix and the quality-adjusted volume of total innovation, should have the same or nearly the same economic value[8] as an ‘average’ pipeline with the same therapeutic area mix, and quality-adjusted volume of innovation

Results are provided in Exhibit 9, both with and without sales weighting of hidden pipeline values. Again using BMY as an example, based only on the amount and quality of innovation in BMY’s hidden pipeline and assuming only that this amount and quality of innovation ultimately is valued at par with other hidden pipelines of comparable amounts and quality of innovation, we would expect BMY to outperform its peers by roughly 47 percent. If in addition, we adjust our estimate of the value of BMY’s hidden pipeline to account for the mix of therapeutic areas BMY is pursuing, we would expect BMY to outperform its peers by roughly 55 percent. We make no assertion that our valuation method identifies single-digit percentage mis-valuations. Instead, we simply argue that companies with apparently undervalued pipelines are more likely to outperform peers

Exhibit 10 provides the actual relative performance of a drug / biotech portfolio whose stock selection is based entirely on hidden pipeline valuation (all names with ≥ 20 percent implied share price gain are held long). Performance figures are provided for the entire period since we first published the method (Nov 2012) and for each interval between updates. Since inception, returns to stocks chosen based on hidden pipeline valuations are roughly 1.6x peer group[9] returns (1.69x equal-weighted; 1.56x cap weighted). The equal-weighted portfolio has beaten the eligible stocks not held during every interval between updates since initiation. The odds of matching the performance of the Hidden Pipeline portfolio in every period since inception by randomly selecting peer group companies are roughly 67,000:1[10]


Changes in pipeline values since our last update (August 2014)

Exhibit 11 tracks the changes in each company’s hidden pipeline since November 2014; and breaks those changes down into their component parts. ALXN, ABBV, NVO, and REGN had the largest increases in hidden pipeline apparent values; conversely AZN, BIIB, SHPG and PFE had the largest declines

We break the total hidden pipeline change into three critical drivers—positively, (1) quality improvements in the existing hidden pipeline; and (2) new additions to the hidden pipeline, whether through patent grants, purchases, new assignments or a first citation on an existing patent; and, negatively, (3) patent expirations, sales, reassignments, etc.[11]

ALXN, REGN, NVO and LLY led the peer group in quality improvement over the past three months. ABBV and NVO had the most new grant / assignment activity. Conversely, AZN, PFE, SHPG and BIIB hidden pipelines lost more ground than any other company’s due to expiration, attrition and reassignment

Hidden Pipeline valuation using clinicaltrials.gov

Companies developing prescription drugs for US approval must disclose their clinical trials on clinicaltrials.gov. We downloaded all active trials for the 22 companies analyzed, and compared active trials as disclosed on clinicaltrials.gov with pipeline products having consensus forecasts – and found that fully 88% of the active projects disclosed on clinicaltrials.gov are not captured by reliable[12] consensus (Exhibit 3, again)

We recognize that consensus rarely forecasts very early stage products, especially for the larger companies analyzed here, and that this normal pattern contributes to the high percentage of clinicaltrials.gov products that fail to find their way into consensus forecasts. However we also recognize that the moment a significant project moves into consensus is a point at which mid- to longer-term estimates are likely to rise. As such, systematically identifying active projects before they find their way into consensus (or into share prices) is potentially worthwhile

As a first step, we back implied growth (7-year horizon) out of current share prices using a dividend-discount model (DDM) based approach; however because dividend policies vary across the companies of interest (and in the case of faster growing and/or biotech companies dividends may not exist) we use free cash flows as a substitute for dividends

As a second step, we estimate the 7-year growth of companies by essentially tacking clinicaltrials.gov data onto the ends of consensus forecasts. Where reliable consensus estimates exist for pipeline products we use these; however where clinicaltrials.gov data identify pipeline products that are not in consensus forecasts, we add these products to the underlying consensus forecasts

More specifically, we use standard estimates of time by phase of development and cumulative attrition across phases of development to project the probable number and timing of product launches, by company. We make the simplifying assumption that all launches have peak sales and lifetime sales trajectories identical to the average historic peak sales and average sales trajectories for products launched over the last twenty years. We assume large and small molecule products achieve average peak sales and average trajectories for large and small molecules respectively, but (as of yet) we make no further forecast refinements to reflect product characteristics such as therapeutic area, presence or absence of competition, and so forth

As a third step, we compare growth implied by share prices with the growth we estimate using a combination of consensus estimates and clinicaltrials.gov data. Share-price implied growth aligns reasonably well with fundamentally-derived growth (R-squared=0.71, Exhibit 12). However in some cases fundamentally-derived growth will be substantially higher than share-price implied growth, which indicates these companies’ mid- to longer-term product flow (aka their Hidden Pipelines) may be undervalued

In each of the last 8 quarters, the 3 companies (of the 22 analyzed) whose fundamentally-derived growth estimates most exceeded share-price implied growth have outperformed their peers (the remaining 19 companies) by a cumulative 2.9x (Exhibit 2, again.

  1. We work with US patents only. At least since the Patent Cooperation Treaty (PERCENT), patent filings for larger, multi-national research-based organizations have become global – thus we believe adding ex-US patents to our datasets would add little or no useful signal
  2. Our method of quality weighting relies on several variables, including in particular citation patterns (e.g. number of citations and rate of citation accumulation) and patent ‘vintage’. Raw (unweighted) patent counts have very limited relevance
  3. Specifically, we categorize patents according to the World Health Organization’s ATC (Anatomic, Therapeutic, Chemical) classification scheme
  4. Essentially enterprise value less the present value of all products / projects / lines of business other than projects in phase II and earlier development
  5. We recognize there are two limits of our method which emerge when analyzing smaller companies. First, the most fundamental premise of the ‘hidden pipeline’ is that it is in fact hidden. In the case of very large companies, the assumption that pre-phase III pipelines are in fact hidden is robust; however the quality of this assumption deteriorates as companies become smaller. Smaller companies will have fewer pre-phase III projects, making it more practically feasible for investors to identify and assess these – making them more comparable to the assets we define as ‘tangible’. Second, our method relies on an assumption of central tendency – i.e. that a large portfolio of projects in a given set of therapeutic areas and with a given overall quality-adjusted patent count will have a similar economic value to a comparable large portfolio of the same therapeutic area mix and overall quality-adjusted count. This assumption is robust with large portfolios – especially since the larger companies have such tremendous overlap in terms of diseases (and disease mechanisms) pursued. However the assumption is weaker with small portfolios. Smaller companies A and B may be in similar therapeutic areas with similar quality-adjusted patent counts, but could be working on highly distinct approaches
  6. Based on arbitrarily defining overvalued as having a ratio of innovation : value ≥1.5
  7. Based on arbitrarily defining undervalued as having a ratio of innovation: value ≤ 0.67
  8. In reality we’re very aware that ‘hidden pipelines’ with similar therapeutic area mix and quality-adjusted ‘weight’ of total innovation will have different economic values. However, we argue that when these pipelines are ‘hidden’ that the market has no factual basis (other than quality-adjusted patents and therapeutic area sales weighting – which we don’t believe the market uses) on which to assign the pipelines different values – and that the market’s assignment of different economic values reflects nothing other than the market’s inability to efficiently value intangibles. Thus over time, the most reasonable guess as to the relative values of two pipelines of identical therapeutic area mix and quality-adjusted volume of innovation is that the two pipelines have identical values
  9. We define the peer group as all of the companies in the screen that are not held long
  10. Assumes a random choice of three peer group stocks at each update, except for initiation when the peer group had only 10 companies and a random two stock portfolio is chosen. Random portfolios are equal-weighted.
  11. The fourth, “other” category captures any uncategorized mathematical remainder due largely to algorithmic calculations.
  12. Separate forecasts from 3 or more analysts for a given product in any year.


Print Friendly, PDF & Email