Real Options: Dos and Don’Ts Real Options in the Pharmaceutical In- Abandons Projects That Are Not Profit- Dustry Became Popular in the Late Nine- Able

Real Options: Dos and Don’ts Real options in the pharmaceutical in- abandons projects that are not profit- dustry became popular in the late nine- able. According to DiMasi2 33% of ties and in the beginning of the new abandonment of clinical projects is due century. At Merck real options have to economic reasons. This means that been in use at least since 1995. “…, within pharmaceutical companies prof- enthusiasts of real options analysis pre- itability or value of a project is an impor- dict their approach will become the pre- tant criterion when deciding on continu- ferred method of investment valuation ing or abandoning a project. In real op- in 10 years.”1 Today most large pharma- tions valuation we therefore consider ceutical companies have a real options various scenarios where some parame- specialist in their valuation teams. But ters, typically the sales expectation, fluc- after about 15 years we can say today tuate. This means that we assume al- that primarily students and professors ready beforehand that at a later point in are enthusiastic about real options time the opinion about the potential of valuation; people that have not seen the a product might change. These changes daily business within the industry. Real can be for the better or for the worse. In options will never replace the good old case of a sufficiently large deterioration risk-adjusted net present value method. management will stop the projects to There are three reasons for our claim: prevent further losses. 1. Real options are still too complex for upper management. 2. Real options have lost a lot of their credibility because of wrong, but widely distributed case studies. It is still taught completely wrong at most universities. 3. Real options require data input that is not easily available. Figure 1: Probability development of a drug development project with real options valuation (calculated Real options: a good idea with ri:val). But let us start at the beginning. Real Real options theory assumes a probabil- options valuation has been developed ity distribution of the critical parameters because of the rigidity of common DCF and models the decisions of the man- methods. People felt uncomfortable giv- agement depending on what value ing sharp assumptions needed for a DCF these assumptions take. In figure 1 we valuation (e.g. sales of “exactly” USD have visualised the probability of the 500 Mio). Everybody struggles estimat- sales along the development path. We ing sales figures far in the future, there- see in the bottom that after each phase fore it seems much more realistic to as- some lower branches are cut out of the sume a certain range. Moreover, a DCF probability tree. These are the scenarios valuation assumes that the project will be continued; in reality management 2 DiMasi, Joseph, “Risks in new drug development: approval success rates for investigational 1 Levinsohn, Alan, „When valuation considers drugs”, Clinical pharmacology and therapeutics, real options“, Strategic Finance, June 2001. 2001.

www.avance.ch where the sales are assumed to be too nies tried to grab some attention with low to justify a continuation of the pro- such examples. ject. A precise description of the real options method can be found in the They tried to convince practitioners to book “Valuation in Life Science. A Prac- use real options because it’s “really sim- tical Guide”. ple”. Take the Black-Scholes formula from quantitative finance (corporate or 1. Real options are complex quantitative finance, what difference There is no doubt that real options are could that make?), fill it with parameters more complex than straight risk- that correspond to your drug develop- adjusted net present value (rNPV). Nev- ment project and you’re done, the for- ertheless real options have a very strong mula returns you the value of the pro- argument in drug development. Real ject. No hassle with Excel sheets, no fi- options claim that managers reconsider nancial modelling. Pretty tempting, project plans and kill projects that do right? The problem was that the values not exhibit sufficient economic poten- did not convince, too bad; nobody could tial. Consequently, it shouldn’t be too see why they should be so much larger difficult to convince upper management now. And because the formula is a of the real option concept as long as the black-box, you couldn’t find out. difference in value to rNPV can be quan- tified and justified. The problem is pri- We have already previously explained marily that the concept is not properly why Black-Scholes is completely unsuit- explained and that people get lost in able for corporate finance3. First, drug technical details. The main hurdle is the development projects have more than probability distribution that is assumed. one embedded option; Black-Scholes We will treat this aspect in more detail can only handle one. Second, the Black- in the third point. Scholes formula does not include the attrition risk, i.e. success rates. Third and 2. Real options have been taught most important, the Black-Scholes for- the wrong way mula is based on replication of the un- When valuing a project with real options derlying and therefore a risk-free portfo- it still remains the same project. It is im- lio. This assumption is complete (!) non- possible that a previously unprofitable sense in any corporate environment and project is all of a sudden worth US$ 100 should immediately ring alarm bells. Mio. Such misevaluations would have Very surprisingly Merck seems to have become obvious over the years; even valued their projects using a Black- without knowledge of real options Scholes approach4. valuation one must have thought that apparently the common valuation tech- niques drastically undervalue projects. But this has never happened. Unfortu- nately academics and would-be special- 3 http://www.avance.ch/downloads/ ists from renowned consulting compa- avance_on_BlackScholes.pdf 4 ardent.mit.edu/real_options/ RO_current_lectures/Realoptions02.pdf

www.avance.ch An article by Shockley et al.5 wanted to Risk-free discounting is a left-over from address the black-box issue by introduc- the original idea that real and financial ing the binomial tree model. Unfortu- options have some similarity. People that nately the article had two major mis- use risk-free discounting focus too much takes. First, they assumed that the attri- on the similarities instead of the differ- tion is inherently modelled by the fluc- ences. Drug development is risky. The tuation of the sales estimate. This is, of risk-free discounting is only justified if course, wrong. It is well possible that the risk can be hedged away as this is the sales potential of a drug increases the case with financial options (at least because, e.g., a competitor fails; but the to a large extent). In drug development trial reveals some severe side effects that it is not possible to hedge the risk, do not allow a continuation of the pro- therefore we have to use a discount rate ject. It is absolutely necessary to include that includes a risk premium. Feinstein the success rates in the tree (a detailed and Lander7 had the correct idea, that description is available in “Valuation in the assumption of adaptive manage- Life Sciences”6). The success rates are ment decisions changes the risk profile valuable information about the risk of a of the project to some extent. The project and it would be careless not to downside risk is reduced, because un- make use of theses statistics. Second, profitable projects are not continued. Shockley et al. use risk free discounting. Feinstein and Lander adapt the discount Absolutely no investor, not even one, rate to this new risk profile. The method would accept a risk-free cost of capital. is not practicable because, again, it as- Why should one invest in a risky asset sumes hedging with something that (the assets remain risky, no matter which doesn’t exist, but the idea that the risk valuation method you apply) if he can profile is slightly different and reduces get the same return with a treasury the discount rate a bit is noteworthy and bond? Although this seems to be a no- correct. brainer, many academics keep teaching risk-free discounting in real options. This Interestingly, all these errors – Black- is probably a side effect of the long iso- Scholes, risk-free discounting, no success lation in the ivory tower. Both errors rates – lead to massively too high values. concern the most value reducing factors, Practitioners looked at the results, attrition risk and discounting; no won- judged them to be completely out of der that Shockley et al. came up with any realistic range, and have not cared wonderful values. Unfortunately, those about real options anymore. An under- were disconnected from reality. standable reaction. Nevertheless, the fact that every pharma company em- ploys a real option specialist shows, that the industry presumes that there is value 5 Shockley, R., Curtis, S., Jafari, J., Tibbs, K., in the managerial flexibility and would 2003, The option value of an early-stage bio- like to quantify it. technology investment. Journal of Applied Cor- porate Finance, 44–55. 6 Bogdan, Boris, and Villiger, Ralph, „Valuation 7 Feinstein, Steven, and Lander, Diane, „A better in Life Sciences“, 2nd edition, 2008, Springer understanding of why NPV undervalues managerial Verlag. flexibility“, The Engineering Economist, 2002.

www.avance.ch is only available to few selected big 3. Data is not readily available for pharma companies. real option valuation An often-cited problem is the volatility How real options should really be as additional input factor. The volatility used is indeed difficult to assess. The parame- For biotech companies the question of ter is a measure of the reliability of the abandoning or continuing is not that sales forecasts. This is only observable important, as voluntary abandonment within large companies that revalue on a corresponds to suicide. But the option regular basis their projects. But to-date to license or not is much more impor- we are not aware of any company con- tant to them. And even some license ducting such an in-house study. Merck contracts contain option clauses that for instance used a volatility based on can be modelled with real options. But comparable biotech stock prices. This is also for pharmaceutical companies the not useful, because a biotech volatility current real option framework of just includes attrition risk, market risk, and abandoning or continuing is too rigid. A operational risk. The real option volatility pharmaceutical company has the op- only refers to market risk. Attrition risk is tions to a) continue, b) abandon, c) put included in the success rates and opera- on hold, or d) out-license a project. And tional risk in the discount rate. And this the continuation option might even be inclusion of these risk factors in the more complex if there is a choice be- wrong parameter, i.e. in the volatility, tween formulations and indications. The leads again to too high valuations. A topic should still be enough food for higher volatility leads to a higher option academics, but please without the men- value. This, of course, is not only coun- tioned errors. terintuitive, but also wrong. If the volatility is higher, then automatically the discount rate is also higher.

But also the success rates must be adapted if we use real options. Usually success rates are a percentage of the projects that were continued in the next phase. The reasons for abandonment can be either safety, efficiency, or profit- ability related. Real option valuation takes care of economic abandonment within the model and therefore has to use success rates that only include safety and efficiency related abandonment, but no economic abandonment8. This data

8 You find a much more detailed description in pharma R&D – a catch-22”, Journal of applied Villiger, Ralph, and Bogdan, Boris, “Valuing corporate finance, 2005.