On October 20, the Wall Street Journal published an interview with LLY’s CEO, addressing the question of whether LLY’s relatively high R&D spending versus peers (Exhibit 1) is warranted. With no intended criticism to the author, the Journal’s general framing of the question – whether past R&D spending is warranted in light of prospects for the current crop of late phase projects – reflects the prevailing standard test for R&D productivity, which we believe is far too subjective, and far too narrow. Hard data can and should be used to measure R&D productivity; and, because R&D is an inherently long-cycled investment, productivity measures should include metrics that capture both investment and output over time – rather than measuring output in terms of the current late phase pipeline. With such an approach we show empirically, across long time frames, that LLY’s R&D productivity has been substantially below that of peers, and that the company’s relative R&D performance is worsening
As noted in the article LLY’s share of peer group R&D spending has been rising; however at the same time the company’s share of quality adjusted innovation has been declining. Exhibit 2 compares LLY’s share of peer group R&D spending (green lines) to its share of peer group innovation (gray lines). In this comparison, innovation is measured as the quality-adjusted number of patents granted to the company in any given year, and is expressed as a percentage of quality-adjusted patents granted to the company’s peer group in that same year. Early (’94 – ’95) and late (’11-’13) years in the analysis are shaded, because measures of innovation are less reliable in these most distant and most recent periods[1]
Exhibit 2 has two simple takeaways: 1) LLY is spending substantially more per ‘unit’ of innovative output than its peers; and, 2) LLY’s relative over-spending per unit of innovation appears to be worsening as compared to peers. Exhibit 3 expresses these same inputs differently, in terms of relative costs per unit of innovation. Specifically, Exhibit 3 shows R&D dollar cost per standard, quality-adjusted unit of innovation, divided by the peer group’s R&D dollar cost per standard quality-adjusted unit of innovation. Exhibit 3 shows that on average, from 1994 to 2013, LLY spent $3 of R&D to produce the same amount of innovation that the peer group produced for $1
Exhibits 2 and 3 use data for granted patents only, and ignore the companies’ pending patent applications. If we expand the analysis to include applications, the findings suggest LLY’s core problem – a rising share of R&D spending and falling share of R&D output – will worsen. Controlling for currently pending patent applications (LLY’s and the peer group’s), we expect LLY to have a falling share of patents granted over time (Exhibit 4). We would expect to see LLY receive about 2.25% of peer group grants in 2014 (despite spending more than 5% of total peer group R&D), falling to 1% or less of peer group grants by 2019. Combining these anticipated patent grants with LLY’s current portfolio of granted patents, we project that LLY’s share of total, quality-adjusted innovation in the peer group ‘Hidden Pipeline’ (i.e. projects in phase II or earlier stages of development) will fall from just more than 2% today, to roughly 1.6% over the next decade (Exhibit 5)
Despite outspending its peers on R&D, and producing only one-third as much innovation (per R&D dollar), LLY has produced a patent state of lower average quality than peers, and has achieved leadership positions in very few of the therapeutic areas it has entered. Exhibit 6 shows the average quality of LLY’s patents by grant year, as compared to the peer group average. Quality measures in this exhibit are heavily dependent on the rate at which a company’s patents accumulate citations from patents that are filed afterward (so-called ‘forward cites’). From ’96 to ’09, the average quality score of a LLY patent was just less than 0.75x the peer group average. Exhibit 7 lists the 20 research areas (by World Health Organization ‘WHO’ Anatomic Therapeutic Chemical ‘ATC’ category) that comprise the bulk of innovation in LLY’s pre-phase III (i.e. ‘hidden’) pipeline. In these 20 research areas, representing more than 80 percent of the total innovation in LLY’s mid- to early pipeline, LLY has a top 3 share of innovation in only 2 categories (Other Sex Hormones and Modulators, and Antimycotics for Systemic Use). LLY’s average competitive rank in its own top 20 research areas is tenth; in the company’s top 10 research areas its average competitive rank is twelfth
Patents aside, the argument can be made that a given company’s quality-adjusted patent output might not predict the actual economic value of its research, since a single discovery might produce sufficient earnings to salvage the relationship between the company’s R&D spending and its profits. For completeness’ sake, we show that this argument does not apply to LLY. A proxy of economic returns to R&D spending (essentially yr1 R&D spending / yr10 operating income) confirms that LLY produces lower economic returns on its R&D spending (Exhibit 8)
Regardless of whether LLY’s current late-phase pipeline meets or even exceeds expectations (an outcome affected by a large degree of chance), over the long-time frame and across the broader scope that R&D productivity can and should be measured, LLY is very clearly less able than its peers to turn R&D spending into innovation
What should be done about LLY’s high relative R&D spending is a far deeper question than whether or not spending should be cut. At current levels of R&D productivity, the answer is simple: other than completing its very late phase projects, LLY should not be conducting R&D at all. The ‘add or cut’ R&D spending framework falsely implies that LLY’s current poor level of R&D productivity determines its productivity going forward. LLY can and should identify the underlying causes and correct these, putting itself in a position to conduct R&D profitably
As convinced as we are that R&D productivity in the industry generally — and at LLY specifically — is negative, we believe just as strongly that the underlying causes are identifiable, and can be addressed. From a distance, LLY appears to be committing two of the more common R&D productivity sins – it appears to set its R&D budget top down (what can we afford to spend?) rather than bottoms-up (what projects are available to us that offer a positive return?); and, the company appears more likely to put its own discoveries into development than to put the best available discoveries into development. Both of these sins are common in the peer group, and have the effect of creating an R&D environment in which a large number of dollars is chasing an artificially narrowed (not enough attention given to outside discoveries) opportunity set – with the inevitable result being a ratio of R&D spending to innovative output that is far too high
[1] ^ Early years are compromised by the fact that patents in this period are beginning to expire; and, since our dataset focuses on active patents only, patent counts in these early years can in some cases be too small to be meaningful. Because our method of quality adjustment relies heavily on citation patterns, and because recently issued patents generally are too young to have developed citation patterns that are indicative of the relative quality ‘status’ these patents will attain across their lifetimes, the quality-adjusted innovation measure is unreliable until the underlying patents are at least 3 years old
