MRK: R&D Productivity, as Compared to Peers

Print Friendly, PDF & Email

Richard Evans / Scott Hinds / Ryan Baum


203.901.1631 /.1632 / .1627 richard@ / hinds@ /


May 5, 2014

MRK: R&D Productivity, as Compared to Peers

  • Please see the SSR Health YouTube channel for podcasts of recent research
  • MRK’s economic returns to R&D spending (the relationship between Yr1 R&D spending and Yr10 adjusted earnings) have been slightly below the (falling) industry average since 2009. For reference, industry returns to current R&D spending appear to be below cost of capital
  • MRK’s lower relative returns to R&D spending appear to be driven by at least two underlying characteristics of its R&D operation: MRK spends 23 percent more in R&D for each quality-adjusted ‘unit’ of innovation produced as compared to peers; and, the average quality of a ‘unit’ of MRK innovation is below the peer benchmark
  • Across the peer group, we believe an assortment of unforced management errors result in weak R&D productivity. For MRK, two in particular stand out:
    • Across the 24 research areas that account for 80 percent of MRK’s pre-phase 3 innovation, MRK has an average competitive rank of less than 5th, and ranks among the top 3 in just 11 of these 24 areas. Unlike its best-in-class peers (BMY and Roche have average ranks of 1.9 and 2.0, respectively) MRK has not actively managed its discovery efforts to ensure that all of its teams are competitive
    • MRK generates 1.8% of global biomedical innovation, but two-thirds of the products in MRK’s human testing originated in the Company’s own laboratories. Given its relatively small share of biomedical innovation (true even of peers with the largest R&D budgets), and its relatively low average rank in targeted research areas, MRK’s development bias in favor of its own discoveries drives unacceptably high opportunity costs
  • MRK’s legacy of breakthrough innovation arguably has been achieved at a too-high dollar cost; high spending per idea and lower than average quality per idea were overpowered by very large levels of absolute R&D spending. As a result, despite its legacy of innovation MRK’s economic returns to R&D spending now lag the peer group’s. MRK can make substantial productivity gains by exiting discovery areas in which it is not world class and re-deploying these resources to discovery areas in which it can compete; and, by more actively and honestly assessing whether externally available molecules are better options than each internally-sourced lead it plans to place into human development
  • These and other R&D productivity metrics covering the 22 largest publicly-traded companies (by R&D spending) are available in a comprehensive benchmarking report at

Returns on R&D Spending are Low v. Peers

Merck’s economic returns to R&D spending (an enterprise-wide measure of R&D productivity that compares yr1 R&D spending to yr10 adjusted earnings)[1] exceeded those of the peer group in almost every year from 1988 to 2005; since then Merck’s returns have tracked or slightly lagged the peer average (Exhibit MRK1)

R&D Spending per Unit of Innovation is High v. Peers

The rate at which Merck generates innovation per R&D dollar invested (a narrow measure of productivity within R&D operations)[2] followed a similar pattern, exceeding the peer average until 2003, and tracking or slightly lagging the peer group thereafter (Exhibit MRK2). Merck ranks 16th among 22 peers in terms of R&D spending per quality-adjusted patent, spending an average of $26.5M across the 1993-2012 timeframe, which is 23 percent more than the peer average of $21.6M (Exhibit MRK3)

The Average Quality of a ‘Unit’ of Innovation is Low v. Peers

The average quality of Merck’s innovation has lagged the peer group across the entire period of analysis (beginning 1988), and appears to have worsened relative to peers in recent years (Exhibit MRK4). Across the 1988 to 2013 period, Merck ranked 15th among 22 peers for average quality of innovation

MRK’s Average Rank in Key Research Areas is Just Less than Fifth; Above the Peer Average but Substantially Worse than Best-In-Class

24 research areas account for 80 percent of the innovation in Merck’s phase II and earlier (aka ‘hidden’) pipeline. Merck ranks among the top 3 companies in 11 of these areas, and has an average rank of 5.4 (Exhibit MRK5). By way of comparison, the average peer company has an average rank of 8.1 in the research areas that represent 80 percent of its early- to mid-stage research activity, thus Merck has been more successful than its average peer at achieving leadership positions in the research areas it currently targets. Nevertheless MRK has been substantially less successful at achieving leadership in target research areas than large cap peers such as BMY, Roche, and PFE, who have average ranks in their key research areas of 1.9, 2.0, and 2.5, respectively

MRK Favors its Own Discoveries, to the Exclusion of Potentially Better Leads Available Externally

63.8 percent of the projects in Merck’s clinical trials are the Company’s own discoveries, even though Merck discoveries represent just 6.4 percent of peer group discoveries, and 1.8 percent of all-source biomedical discoveries (Exhibit MRK6). Thus Merck’s ‘internal bias index’ is 35.1 (63.8 percent of clinical projects are own discoveries / 1.8 percent of biomedical discoveries are Merck’s), which is somewhat less (better) than the average peer group index of 54.0. Note that the odds of clinical development projects being externally sourced rises with later stages of development; this indicates MRK’s in-licensing tends to occur late in development after its own internally sourced leads have failed at earlier stages of development.

  1. Conceptually, this metric reflects the firms’ Discover → Develop → Make → Sell value chain, relating earlier R&D investment to later earnings. Research and development dollars are invested in the ‘Discover’ and ‘Develop’ steps, initiating value creation; and, further necessary investments (‘Make’, ‘Sell’) are made in order to commercialize the resulting products. The economic return on the earlier R&D spend is thus the subsequent (years later) earnings stream

    More specifically, we compare year 1 R&D spend to year 10 adjusted operating income; for example the return on 1978 R&D spending is defined as 1987 adjusted operating income. Obviously the 1978 R&D spend results in products that should generate earnings across several years, so this metric cannot be read narrowly as a specific measure of productivity in a single year. Rather, by comparing year 1 R&D spend and year 10 adjusted operating income over multiple decades, we’re able to estimate two things: 1) representative economic return levels for a given ‘era’ of spend; and 2) the trajectory of economic returns to R&D spending over time

    The ten year gap between investment (R&D spend) and return (adjusted operating income) is chosen because it approximates the distance between the midpoint of a single product’s R&D costs (even though spending takes place across more than a decade, the greatest spending occurs in development phases that precede launch), and that same product’s peak earnings (which occur at a midpoint in the on-market lifecycle, after launch costs are covered, but before competitive positioning erodes). This obviously is a rough approximation of the investment / return timeframe; however because our estimate of economic returns to R&D spending is fairly insensitive to the assumed gap between R&D spend and associated earnings, a rough approximation is defensible

    The key benefit of this metric is that it is an enterprise-wide measure of R&D productivity, because it takes into account all associated costs, not just the direct costs of R&D. The key weakness of this metric is that its most recent values reflect returns on R&D investments made 10 years earlier, i.e. the metric is silent on economic returns to R&D spending within the last 10 years

  2. Where the preceding metric (economic returns) is an enterprise-wide measure of returns to R&D spending, innovation yield to R&D spending is a narrower measure of productivity within the confines of the R&D operation. Conceptually, this metric reflects the amount of innovation produced per R&D dollar spent

    The patenting behavior of biopharma companies across the early- to mid-phases of discovery and development appears to be both comprehensive (i.e. all potentially promising ideas are patented) and consistent (the decisions of when to patent and what to patent are similar across companies). As a result patents – and in particular these patents’ citation patterns – offer a reasonable basis for comparing firms’ innovative output

    We define the innovation yield of R&D spending as the amount of innovation produced in a given year, divided by the amount of R&D spent in that same year. More specifically, we define the year in which innovation is ‘produced’ as the year in which the patent describing that innovation is granted

    The first limitation of this metric is that discovery precedes the patent filing, and that the filing precedes the grant – often by several years. As such, it’s important for readers to bear in mind that patents granted in a given year – and thus counted by this metric as being innovation ‘produced’ in a given year – in truth will have been made as many as several years earlier

    The second limitation of this metric is that R&D spending in any given year will be heavily weighted to ‘D’, i.e. to development spending, which generally has more to do with commercializing ideas than with generating ideas. If we had access to ‘clean’ measures of research spending – sans development – we obviously would prefer a narrower metric reflecting the innovation yield to ‘R’ (i.e. research) spending only. Nevertheless the combined R&D spending measure is all we have, so we’re forced to work within its limits

    The third and final limitation of the metric is that it is only accurate for grant years at least 5 years in the past. Because our measure of innovation output relies heavily on patents’ citation patterns, and because the citation patterns of patents less than 5 years old may change significantly across these patents’ lifetimes, our estimate of innovative output can be very noisy until approximately 5 years following the underlying patents’ grant dates. To remind readers that these more recent readings are less reliable, our graphics show the most recent 5 years’ observations in a shaded area

    The benefits of the ‘innovation yield to R&D spending’ metric are that it allows us to estimate productivity within the confines of the R&D organization (it is roughly equivalent to the number of quality-adjusted ideas produced per dollar of R&D spend); and, despite its own time lag, it is a metric that can be used for more recent years (5 year lag) than the economic returns metric (10 year lag)

Print Friendly, PDF & Email