Understanding innovation · Part III · Capital and value

10 · Investing in innovation: the statistics of returns

13 min read

In brief

The returns on investment in innovation do not follow a normal distribution but a power law: a few exceptional outcomes generate most of the aggregate return, the majority of transactions lose money or yield little, and the average does not describe the typical case.
Everything else follows from this statistical property: the single transaction is a bet, the portfolio is a strategy; without a sufficient number of positions, the probability of missing the outcomes that pay for all the rest is high by construction.
Phased investment is a real option: each round buys, as well as a shareholding, the information needed to decide whether to exercise the next one. Much of the value of the model lies in the right not to continue.
Illiquidity is part of the contract, not a defect: the capital invested has no market until a liquidity event, and the horizon is measured in years. The required return incorporates it.
The empirical evidence on angel investing indicates that positive aggregate returns exist but are concentrated, and that two behaviours correlate with results: the time devoted to due diligence and post-investment involvement.

Section 1

The shape of the distribution, and why it changes everything

Anyone who comes to investment in innovation from traditional markets brings with them a statistical intuition that does not work here: that of the normal distribution, in which outcomes cluster around the average and the extremes are rare and symmetrical. The returns of venture investing have another shape, documented convergently by academic research (Kaplan and Schoar, 2005, on the persistence and dispersion of fund returns; Korteweg and Sorensen, 2010, on the dynamics of the returns of individual financed companies; Kerr, Nanda and Rhodes-Kropf, 2014, on the experimental nature of the activity) and by the operational data discussed in the sector literature: a power law, strongly asymmetrical. The majority of transactions return less than the capital or lose it entirely; a minority return a few multiples; a small fraction return very high multiples, and this fraction generates most of the aggregate return of a portfolio or of an entire vintage of the sector.

Three consequences overturn ordinary intuitions. The first: the average describes nothing typical. A portfolio can have an excellent average return with 80 per cent of the positions in loss, because one position has returned fifty times its capital; "how does it do on average" and "how does it usually do" are questions with opposite answers, and they must be kept separate. The second: the error that costs most is not investing in a failure, it is missing the exception. In a normal distribution, errors weigh symmetrically; under a power law, the maximum loss on a position is the capital invested, while the gain forgone on the exception has no ceiling. Selection that is too prudent, discarding everything that might fail, thereby discards everything that might pay for the portfolio as well. The third: the dispersion between investors is enormous and persistent (this is Kaplan and Schoar's result): the difference between the top and bottom quartile of funds is much wider than in traditional asset classes, and those who have done well tend to do well again. Access to the best transactions and the capacity to select count more than elsewhere, and "the return on venture capital" as a single number is an abstraction that hides this dispersion.

Section 2

From the single deal to the portfolio

If the value lives in the tails, the rational strategy is the one that gives access to the tails, and access is a question of numbers. An investor with two or three positions is not doing venture investing: they are making concentrated bets on a distribution in which the mode is loss. The probability of having at least one exception in the portfolio grows with the number of independent positions, and the operational literature converges on portfolios of dozens of positions as the threshold below which the variance of outcomes remains dominated by luck; the logic is the same as the innovation tournaments of chapter 03 and the project portfolio of chapter 08, applied to capital: volume at the entrance, disciplined selection, and acceptance that individual outcomes remain unpredictable while the aggregate becomes manageable.

Hence the two construction rules that prudent practice recognises. The first is partial allocation: only a fraction of total wealth is assigned to an illiquid, high-variance asset, sized so that the worst outcome (the loss of the whole allocation, which the distribution makes possible) does not compromise the solidity of the investor; venture is a component of a wealth portfolio, not a substitute for one. The second is diversification within the allocation: more positions, as far as possible independent of each other (sectors, technologies, investment vintages), because concentration in a few transactions puts luck back in command. The two documented errors that violate these rules have names of their own: under-diversification (a few large positions instead of many small ones) and home bias, concentration on what is near and familiar, geographically or by sector, which reduces the independence of the positions precisely where it appears to increase control.

A third error deserves a paragraph because it is the most insidious for those who come from the technical side: confusing the quality of the technology with the quality of the investment. They are different variables: an investment is the combination of an asset, a price and a structure, and an extraordinary technology bought at a valuation that already discounts its success is a mediocre investment, while a good technology on terms that incorporate the risk can be an excellent one. Technological due diligence (chapters 06 and 07 give the instruments for it) answers the question "does it work and will it mature?"; the investment decision answers "at this price, with these rights, in this structure, is the risk-return profile attractive?". Chapter 11 will add the third question, "who will capture the value?", which completes the trio.

Section 3

Phased investment as a real option

Chapter 09 described phased financing from the project's side; from the investor's side, the structure has a precise name in financial theory: it is a compound real option. Each round buys two things: a shareholding, and the right (without the obligation) to take part in the next round with the information that the project will have produced in the meantime. Kerr, Nanda and Rhodes-Kropf (2014) describe the whole sector in these terms: investment in innovation is experimentation, and rounds are experiments financed in sequence, each of which is worth something for the information it generates as well.

The accounting consequence is counter-intuitive and important: much of the value of the model lies in the right to stop. Investing all the necessary capital in a single initial commitment means giving up the value of future information; investing in phases means paying for uncertainty only in tranches, abandoning the positions in which the experiment has given a negative result and concentrating capital where it has given a positive one. Options theory (Dixit and Pindyck for the formal treatment of investment under uncertainty) gives this flexibility a quantifiable value, which grows with uncertainty itself: the more uncertain the outcome, the more the possibility of deciding after having seen is worth. This is why experienced capital does not regard phased financing as an inconvenience imposed by scarcity, but as the correct structure of the contract; and it is also the right reading of milestones: from the investor's side, a milestone is the point at which the option is exercised or abandoned, and its quality is measured by how far the evidence that defines it is capable of changing that decision (the criterion of the decision window of chapter 07, seen in the mirror).

The reverse side of the model is the horizon: the option is exercised over years, and between one exercise and the next the capital has no market. Illiquidity is not an operational defect to be corrected but a property of the contract: a shareholding in an unlisted company converts into money only at a liquidity event (sale, secondary, listing, dividends), which may be many years away from the entry and has no guaranteed date. The return required by risk capital incorporates a premium for this renunciation, and the practical rule that follows is one of financial hygiene: capital assigned to venture should be capital that you do not expect to need over the horizon of the investment, because the need for early liquidity turns illiquidity from a known property into a realised loss.

Section 4

The evidence on angel investing, and the behaviours that correlate with results

For the individual investor, the reference study remains that of Wiltbank and Boeker (2007) on the returns of more than a thousand North American angel investors across more than three thousand exits: the aggregate return turned out to be positive and competitive (of the order of 2.6 times capital in about three and a half years across the distribution of observed exits), but with the distributional shape of section 1: more than half the exits below the capital invested, and the bulk of the return concentrated in a small fraction of transactions. The study also identified two behavioural correlations that later research has substantially confirmed: transactions preceded by more hours of due diligence showed better outcomes than those decided quickly, and the investor's post-investment involvement (competence and experience made available, not interference) correlated with results. The correct reading is a cautious one: these are correlations on self-reported data, not laws; but they indicate that in early stage investment the investor's work (before and after signing) is part of the return, and that the model of "a cheque written quickly and forgotten" is the configuration with the worst evidence.

For the investor who cannot or does not want to do this work themselves, the consequence is structural, not moral: it is preferable to invest through vehicles or operators that do it professionally (funds, syndicates, operators that select and accompany), accepting the cost of intermediation as the price of selection and of the portfolio. The choice between direct and intermediated investment is therefore a choice about who performs the functions that the evidence indicates as decisive: selection with real due diligence, portfolio construction, accompaniment.

In the Volcano method

This chapter gives the statistical frame within which the model's materials addressed to investors should be read. The "mapa del riesgo y del premio", the essays "Invertir en innovación" and the declared golden rule ("only a part of available capital is assigned to innovation") are the partial allocation of section 2; the model's communication is explicit that the high multiples describe what happens when a project arrives, not an average, and that the majority do not arrive: it is the distinction between average and typical case of section 1, declared instead of hidden. The segmentation of investor figures is the translation of the division of risk: the tax investor has a return that materialises by the tax route independently of the outcome plus a possible shareholding (a hybrid position that stands outside the power law for the first component and inside it for the second); the private investor in a convertible note buys full exposure to the distribution, with the shareholding fixed on day one that remunerates early entry; the pure lender stands outside the upside by contract. The pre-entry de-risking (public and tax funds that absorb the most uncertain phase before private capital comes in) modifies the point of the distribution at which the private investor enters, not the shape of the distribution: it is a reduction of risk, not a promise of return, and that is how the model declares it. Illiquidity too is declared as part of the contract ("this is risk capital, and illiquidity is part of the agreement"), with the conversion and exit events defined in advance; and the choice by project (you invest in the specific project, not in a basket) puts the construction of their own portfolio in the investor's hands, with the responsibility for diversification that section 2 describes.

Readings

Frequently asked questions

If the majority of transactions lose money, is investing in innovation irrational?

The single transaction taken on its own resembles a bet; a portfolio built with discipline does not. The rationality of investment in innovation lives at the aggregate level: enough independent positions, real selection, partial allocation of wealth, a horizon consistent with illiquidity. Under the wrong conditions (two transactions, concentrated wealth, a short horizon) the irrationality lies not in the asset class but in the construction.

Is it not better to wait until the risk is lower and pay more?

It is a legitimate choice of position on the curve, not a free lunch: entering later buys less risk at a price that discounts it, entering earlier buys more risk on terms that remunerate it. The error is not preferring one point or the other, it is demanding the terms of one point while standing at the other. The phased structure exists precisely to make this choice explicit: every round is a different point on the curve, with its own price.

Does a certain tax return eliminate the risk of the investment?

No: it distributes it differently. Where there is a component of return by the tax route, that component does not depend on the outcome of the project, and its risk is regulatory and one of tax execution, not entrepreneurial; any shareholding, by contrast, remains exposed to the distribution of returns like any other equity. Assessing a hybrid transaction requires assessing the two components separately, each with its own risks, and not using the certainty of the first as a reason to stop analysing the second.

How much of one's wealth makes sense to assign to this asset class?

There is no universally correct percentage, and it is the question to put to a qualified adviser on the specific case; there is, however, the criterion with which the answer must be constructed: the fraction assigned must be able to be lost entirely without compromising financial solidity and liquidity needs over the horizon of the investment, because the distribution makes that outcome possible and illiquidity prevents it from being corrected along the way. The golden rule (only a part of the available capital) is the perimeter within which every specific answer has to sit.

Disclaimer. This chapter has a training purpose: it describes the statistical and structural properties of investment in innovation as documented by the literature. It does not constitute investment advice or a recommendation of any transaction; investment decisions should be taken by assessing the specific case, your own financial situation and, where appropriate, with the assistance of qualified professionals.

← Previous chapter: 09 · The capital of innovation Next chapter: 11 · Who captures the value →

Back to the Introduction.

We choose to go to the Moon in this decade and do the other things, not because they are easy, but because they are hard.
John F. Kennedy, 1962

Tax close before 31 December Turn your tax burden into a return