Richard Murphey, 11/12/2018
In the first post in this series, we discussed the basics of forecasting a P&L for an unapproved drug. In this post, we will learn about frameworks used to value biopharma companies.
Drugs have short and explosive life cycles, with new products growing from nothing to billions of dollars of high-margin revenue in just a few years, and then going to zero overnight when patents expire. Newly launched competing products can erode what once seemed like lasting franchises. Even $100B+ companies can lose 40% of their market cap almost overnight when one study of an unapproved product doesn’t go as planned (look at Bristol Myers in summer 2016).
The drug industry is characterized by high growth, high profits, binary risk, and volatility. In many cases, significant portions of a company’s valuation don’t show up in financial metrics – late-stage unapproved drugs can be valued at tens of billions of dollars before they even enter the market.
To value biopharma companies, you need to do more than dig into the financials -- you need to dig into the products. And that means you need to get into the science.
But valuing products is a subject for another post. This post will be a "stepping stone" from the traditional finance techniques you are familiar with to more science-oriented concepts you'll need to become familiar with in order to evaluate products. Specifically, in this post I’ll provide an overview of the valuation methodologies commonly used in biopharma, and highlight highlight a few biopharma-specific modeling considerations.
DCFs are tricky in any industry, as they are very sensitive to assumptions, and it can be hard to nail down key assumptions.
In biopharma, DCFs are even more difficult, as they layer on even more assumptions than DCFs for companies in other industries. Despite this lack of accuracy, they play a larger role in biopharma valuation than many other sectors, and are often the main method for determining a company’s valuation, especially for earlier stage companies.
For many (if not most) pharma companies, significant value lies in the pipeline of unapproved drugs, or drugs at the earliest stages of commercialization. It is tough to use multiples for these products, as they don’t generate revenues or profits. You can use forward multiples to try to capture the value of early-stage products, but that can get complicated quickly: do you use revenue or P/E multiples? What year's revenue or earnings do you apply the multiple to? Should you use an NTM multiple, or 2-year multiple, or discounted 5 year multiple? These answers will be different depending on the company.
You’ll still use multiples based approaches, but rely a bit more on the DCF to give you more granularity. In some cases, you'll do a sum-of-the-parts DCF valuation, sometimes you'll do a SOTP and use different multiples for different products. If you are evaluating a $50B+ company, you can probably just use LTM and NTM P/E, but the earlier stage the company, the more you’ll need to explore using revenue multiples, discounted forward multiples, 2+ year forward multiples, or some combination of those.
In biopharma, P/E is used more often than EV/EBITDA. In most cases for larger, commercial-stage companies, LTM and NTM P/E are used, but often people will use a discounted forward P/E for earlier-stage companies. The rationale for this is that a company may have valuable products that are either not launched or are in early stages of launch, so the products will generate little or no earnings for years, so using NTM P/E would not accurately capture the value of the products. So you pick a year in the future where these products become more mature and can be valued using P/E, then put a multiple on those forward earnings, then discount it back.
For example, if you have a product that you think will be a $1B revenue product in five years with 60% contribution margins, but that product is only expected to generate $10M in revenue next year, you might use earnings in year 5 for your P/E. Then you would look at comps, figure out the mean and median NTM P/E for those comps, multiply that by your year 5 earnings, and then discount that back four years (because you are using NTM P/E on year 5 earnings, so you are effectively using year 4 as your valuation date).
If that sounds confusing, it’s probably because it is a somewhat contrived approach. It makes theoretical sense, but in practice it is pretty squishy. What year’s earnings do you apply the multiple to? In the above example it’s year 5, but could it be year 4, or year 6? Let’s say you want to look at year-5 earnings; should you use an NTM P/E on year 5 earnings and discount it back four years, or a 2-year forward P/E on year 5 earnings and discount it back three years? Or use an LTM P/E on year five earnings and discount it back 6 years? What comp set should you use?
Each of these could yield significantly different valuations, and there isn’t really a “standard” methodology as far as I can tell. Your VP may end up asking you for 15 different permutations, all of which are semi-justifiable but none of which are that great.
Like all valuations, you’ll use a combination of different methodologies, but these complexities make it a bit trickier to figure out “reasonable” multiples to use for a biopharma company.
Using revenue multiples can get around some of those issues and can be a “cleaner” way to use multiples to capture the growth potential of early-stage products. Typically a product will generate a meaningful amount of revenue before it generates a meaningful amount of profit, so you can put revenue multiples on earlier years than you can for P/E multiples (ie, if a product generates $100M revenue but only 20% contribution margins in year two, and “mature” margins are expected to be 50% starting in year 5, you may need to use year 5 for a P/E multiple but could potentially get away with year 2 for revenue multiple). However, if a company has high-growth products as well as material earnings from “mature” products, it’s probably better to use P/E, or a sum-of-the-parts approach with different multiples for different products.
Of course it is still a very imperfect method. As discussed in the prior post, forecasting the sales trajectory of a new drug launch is very difficult. It’s also not straightforward to pick which year’s revenue to use – do you use 2-year forward revenue, 3-year, 4-year?
One way you can get around this is to use a “peak sales” multiple. Pick some comps that were bought at a similar stage to your company, figure out the peak sales of their main product(s) (assuming there are only a few major products), then divide equity value (“tech” value, assuming a company has no debt) by peak sales and there’s your multiple. This way, you don’t have to guess about the sales trajectory, and you don’t have to worry about picking a year for your forward multiple. Of course the downside is that you may not have a great comp set, and this doesn’t really account for how quickly / slowly your product will ramp.
In the current market, many small or mid-cap companies are valued more on likelihood of getting acquired by big pharma than on conservative estimates of expected value of their cash flows. In the last five years pharma companies have shown extraordinary appetite to buy companies at very high premiums that can only be justified by very optimistic projections. In some cases, a pharma company will pay whatever it takes to get the deal done for assets that their science team deems highly strategic. So there has existed an “M&A thesis” that many investors follow, where they try to identify the next M&A candidate and then value the companies based on a probability-adjusted discount to M&A comps.
You probably shouldn’t rely too heavily on this technique as it is difficult to justify with traditional valuation methodologies, but it is helpful to be aware of this. Many valuations are difficult to explain using traditional techniques, but that doesn’t necessarily mean that they are wrong – just that the investors are betting on pharma’s willingness to pay up for strategic assets.
If you look at models in equity research reports that use DCF or multiples methods to value companies, you will often see that there are a few pretty aggressive assumptions hidden in there. Without these aggressive assumptions, a reasonable DCF in many cases would get you a valuation below the current market price. In these cases, it’s likely that the market views the company as a prime takeout candidate.
VCs use a framework that is sort of similar to this when valuing early-stage companies. The idea is that you focus on the few assumptions that matter and "outsource" to the market assumptions that don't move the needle. You're giving up some precision in return for better accuracy. For example, probability of technical success at a given stage (preclinical development, Phase 1, etc) is probably the biggest determinant of value for early-stage companies, assuming there is a large enough market. Getting 50% smarter on probability of success will inform your valuation much more than getting 50% smarter on steady-state contribution margin. Consequently, startup valuation becomes more about diligencing technical risk than doing complex financial analysis (which is why most VCs are PhDs, not bankers). More on this in another post.
Beyond differences in valuation frameworks, there are a few finance and accounting idiosyncracies unique to biopharma. Many of these are very complex and technical, and you won’t be expected to master them – you’ll just need to know how to make reasonable assumptions and, in live deal or bakeoff situations, consult specialized lawyers, accountants or consultants for their take.
Most drugs have a terminal value of zero: when key patents expire, generics flood the market, and revenue drops 90%+ basically overnight. For these products, it is incorrect to model much if any terminal value. Rather, you can often model out sales every year until key patent expiry, then assume the drug is worth nothing.
For small molecule drugs, you should almost always assume that once patents expire, the drug is worthless. FDA has well-established regulations for enabling fast approvals of “generic” small molecule drugs.
For large molecules, however, “genericization” is less black and white. Due to the more complex nature of these molecules, it is harder to prove a large molecule is “biosimilar” to another large molecule. FDA only recently established a pathway for approval of “biosimilar” large molecules that are substitutable for pioneer large molecules, and this pathway is harder and more expensive than the pathway for small molecules. I’m not super current on the biosimilar world, but basically large molecules have longer and larger “tails” than small molecules when patents expire, so it is not always appropriate to forecast significant revenue erosion upon patent expiry. You need to look into the details of the particular drug you are forecasting.
So while you generally won’t include a terminal value for products, sometimes investors place a terminal value on a platform (though you really need to exercise caution when doing this). The idea here is that a company has some fundamental scientific competency that makes it possible for them to discover and develop lots of new drugs, or a BD competency that gives them an advantage in identifying and licensing promising assets. So you could include a terminal value to account for the value of products the company has not invented / acquired yet, but is likely to invent / acquire in the future.
If you do this, you should have a good reason for it, and this should not be a big part of the value. The value of a drug increases exponentially as it advances in development. At the earliest stages, a drug candidate has a 1/10,000 chance or less of getting approved, thus very little value. Most of the value of a platform is determined by one or two products, and the value of the rest of the platform is typically almost meaningless.
The deal terms in the press releases around biopharma M&A or licensing deals are often pretty complex. You’ll often see an “upfront” payment of some smallish amount, maybe combined with an equity investment, and then some development, regulatory or commercial milestones, and then royalty rates. You might also see that the licensor has agreed to pay x% of all development expenses for a particular product, or has an option to license rights to a few products in certain geographies, or even an option to acquire the whole company.
These contingent payments are referred to as “biobucks” (although often royalties are excluded from this). Nearly all partnering deals in biopharma have some sort of “biobucks”. Biobucks are important because they enable companies to structure around risk. Sellers want to get paid if their products work, but buyers don’t want to pay up for risky drugs that will most likely end up worthless. Biobucks enable companies to structure deals that bridge this gap.
However, this can make valuation more complex. For one, it adds another dimension of complexity to M&A comps. Often you’ll have a column in your comp set for “upfront” value, and a column for “total” value. These numbers can be very far apart.
It also makes modeling more difficult in some cases. In addition to forecasting and valuing a P&L, you need to account for any deal-specific terms that alter the P&L or balance sheet. A few common terms to look out for:
Modeling all of this can get pretty complicated, but it is important not to gloss over these terms. In some cases I’ve seen investors get this wrong and value a company as if they had nearly all the economics for a key asset, when in reality the company owns only a fraction of these rights. Don’t do that :).
Discount rates in pharma are tricky, which is problematic because DCFs play such a large role in valuation. Traditional methods of calculating discount rates like CAPM don’t really yield usable numbers in biopharma. The risks associated with biopharma companies are often unique to the company itself, and this can make it hard to compare risk and volatility across companies. Volatility for biotech stocks, especially pre-revenue companies, can be crazy and beta values are often useless.
In practice, investors will typically bucket companies into different groups based on development stage and size, and then apply progressively lower discount rates to larger companies. Figuring out the right discount rate when a big company buys a small company can be tricky – do you use the small company’s discount rate, or the big company’s discount rate? Should they be that different in the first place if the risks are diversifiable?
This is a very complex area and for live deals or bakeoffs you may need to consult with lawyers and accountants. I’ll just highlight a few pharma-specific issues, and then let you do further research to figure out how to model these factors.
The first issue is NOLs. Many pharma companies accrue a lot of net operating losses when they are developing their drugs, and these losses can be valuable tax shields. In M&A, you’ll want to pay close attention to how NOLs are valued. This is an area where you’ll want to figure out how your group typically handles this accounting, and get feedback from lawyers and accountants on important deals. For the purposes of this post, I’ll just flag it as something to look out for.
The next major tax issue is understanding how various corporate structures and IP domiciles influence tax rate. Many pharma companies have historically domiciled much of their core IP in low-tax countries like Ireland, so profits from drugs using this IP is taxed at a very low rate. These laws are complex and rapidly changing, and when you need to get sharp on this, usually for a live deal, consult with lawyers.
IP domiciling and tax can be a really interesting and important area, and while it isn’t expected you know much about this, if it piques your interest and you learn about it then it can be a nice way to differentiate your skill set.
This is another very complex topic that is also very important, and one you aren’t expected to know much about as an analyst (or probably even as an MD). Being wrong about IP is often a costly mistake, so if you are in a position where you need to make assumptions about IP in a live deal context, talk to a lawyer or consultant. It is also a highly technical intellectual rabbit hole, and if you find that you like this area, it can be a valuable way to differentiate your skillset.
When you are modeling drug revenues, it can be important to have good assumptions about patent expiry, and you probably can’t get away with spending a couple grand on a lawyer for your pitchbook model. In these cases it’s probably best to read equity research reports, company presentations or SEC filings to figure out where the market is bounding patent risk, and just run sensitivities on that.
One major takeaway from all of this is that “garbage in, garbage out” applies even more in biopharma than for many other industries. You don’t have historical financial information to anchor estimates for a substantial part of most businesses. There are a lot of assumptions, and they stack up very quickly. You need to be very diligent about your assumptions, and you will often have to make assumptions that you don’t feel 100% confident about, because you just don’t have data.
In addition to doing the usual sensitivities, spend extra time thinking through 1) which assumptions move the needle the most, 2) your “confidence intervals” for each assumption and 3) where you differ most from the market and your conviction level around that (you should do all of this for any valuation work you do, but it is especially important for biopharma). This is probably less important for banking than for the buyside, but it's never a bad idea to be extra diligent about your assumptions.
Especially for pre-revenue companies, you won’t select comps based on sector, size and other financial metrics. You determine your comp set based on factors like disease being treated (the more specific the better), stage of development, type of drug (what is the molecular target of the drug, mechanism of action, how good is the data), and commercial factors like price, number of patients who are candidates for the drug, and sales channel / reimbursement (are you selling to hospitals, primary care physicians, dermatologists, etc). CapIQ and FactSet aren’t great for this, although you can use tools like Evaluate Pharma, or just do a ton of googling.
If you can find a fairness opinion for a relevant biopharma deal, that can be a helpful resource. Equity research reports can also be useful as a framework, although you should obviously take a critical look at their models. A helpful exercise for learning is looking at a bunch of ER models, seeing where they have divergent assumptions, and then developing your own view on the fundamentals and appropriate valuation methodologies.
Putting your money where your mouth is can also be a great incentive to learn quickly. Treat it like bitcoin -- only invest as much as you can afford to lose, and expect a wild ride. And beware of shorting – it can work, but many stocks are hard to borrow, and it is not unheard of for companies to get acquired for 300-600%+ premiums, so you may lose more than you bargained for.
Eventually, you’ll start asking questions that are critical to your valuation but that you don’t have answers to. Will FDA approve this drug? How likely is it that the upcoming Phase 2 studies are positive? Will a generic company invalidate a key patent next year? Is this drug better than a competing drug? Answering these questions is at the heart of the work you will do on the buyside, and requires learning enough science to have intelligent discussions with physicians and scientists. You don’t need a PhD or MD to do this, but there is a fairly steep learning curve. I’m not aware of any good articles / books on these topics, so I’ll try to write another post on how to learn these things in the future.
As a jumping off point to learning about the science of drug discovery and development, you may want to read this post providing a high-level overview of the drug development process.