Research Impact Bonds: A promising instrument for funding scientific research?

Impact bonds help to finance environmental or social projects by transferring their risk of failure to the investors. Recently, it has been suggested to apply this instrument also to scientific research through Research Impact Bonds (RIB). This article highlights the limitations and risks of putting RIBs into practice.

In a recent article, Michael Hill (2023) emphasizes the need to mobilize new financial instruments to fund scientific research. The goal is to improve the current model by attracting private investors in an area which has long been the prerogative of the government; due to potentially high spillovers to society, but high fixed costs, large uncertainty, and low short-term profitability.

The finance industry is highly innovative, and we can see a proliferation of financial engineering (FE) tools to solve such difficult funding problems. Hill argues that impact bonds (IB) can be used to finance research (through research impact bonds or RIB), not only bringing another source of financing, but also improving the current model by promising “strong accountability, no risk to the funding organization and a direct demonstration of impact.” After explaining the rationale behind RIBs, we discuss three fundamental issues of using this model to fund scientific research.

From IB to RIB

An IB is a specific type of bond which return depends in part on the outcome of a given project. The return is larger (resp. lower) if the project succeeds (fails). Because the risk of IBs depends mainly on the specific risk of the project, they are an interesting asset for investors looking to diversify their portfolio. For the issuer of the bond (e.g. the government), the benefit is to share risk by having access to an insurance mechanism in case the project fails. Social or environmental impact bonds (S/EIB) have proven to be useful and effective in various cases such as reducing crime recidivism in the UK; financing infrastructures to reduce stormwater overflow in Washington DC.

Impact bonds work well in these cases because the risk for private investors is limited since it can be suitably hedged (Arrow 1962). This means that the project can be strictly monitored by the investor – in terms of resources, efforts and competences which are allocated – and success criteria are clear and unambiguous. In such cases, investors can commit resources because they can estimate the probability of success, are rather sure that the expected level of effort to make the project successful will be provided, and the success can be assessed against clear and indisputable metrics.

RIBs follow the same principle than S/EIB. However, three fundamental problems arise from the transposition of IBs to scientific research.

Unhedged uncertainty

The fact that researchers only produce knowledge with uncertainty is problematic. This problem could be solved, as in the infrastructure example, if risks can be suitably hedged. However, Arrow (1962) observed that the uncertainty of research is partially related to the fact that one cannot observe the effort of the researcher as well as any further decisions taken as the research project unfolds. Research involves uncertainties which cannot be fully hedged when they depend on unobservable efforts by scientists. Then, it is hard to think of investors ready to finance projects when there is no means to monitor research efforts and decisions.

The ambiguity of success and failure in scientific research

What is a success or a failure in the domain of scientific research, and how can it be measured? Every scientist knows well that a failure can be very useful – generating important information – for the next generation of scientists. Works by Evenson and Kislev (1975), David and Stiglitz (1979) and David, Mowery, and Steinmueller (1992) suggest that, by providing a stronger informational basis for applied research decisions, basic research – even in case of failures – improves the effectiveness with which the resources devoted to applied research can be allocated among competing alternatives. Indeed, any research project that is rigorously implemented eventually produces useful knowledge, whatever its outcome. The lack of a clear definition of “success” and “failure” and the question whether a failure can have some productive effects strongly limit the scope of funding instruments that can be used in the realm of scientific research.

Transaction costs

The reduction of these problems requires the determination of precise criteria to be able to identify the potential outcomes of a given project and their probability of occurring. This is not straightforward – when at all possible – and requires significant work from numerous stakeholders (researchers, public agencies, brokers, investors, experts, lawyers, etc.). As a consequence, the process of implementing RIBs may generate significant transaction costs, contractual complexity, and bureaucracy (Fraser et al. 2018). In large projects such as the implementation of a new social policy or the deployment of green infrastructures involving tenth of millions of dollars, the benefits may significantly outweigh the costs. However, this may be much less the case for smaller, more numerous projects. The risk here is to get less for more, i.e. less research for a larger cost and a larger amount of time spent by researchers dealing with paperwork rather than doing research.

Hill suggests to reduce these problems by using new research management mechanisms such as preregistered research – i.e. defining the research question, methodologies and analysis plan before observing research outcomes (Nosek et. al 2018). This is already happening in research, such as for clinical trials. However, while the increased transparency is useful for RIBs, it does not solve the problem of the definition of what is a success vs failure in research. Moreover, there are several challenges associated to preregistration, such as allowing necessary changes to research procedures during study administration. How to enable these changes in the context of RIB without being exposed to legal issues is unclear.

Conclusion

To summarize, impact bonds are a wonderful mechanism which can help leverage one important mission of financial markets: risk sharing. This generates benefits for multiple stakeholders, including

• governments which can offload part of the risk of failure of their policies to financial markets – an interesting prospect at a time when large public debts increase scrutiny over public spending;
• investors, who have access to an alternative asset with interesting properties – e.g. risk-return profile, diversification benefits;
• and society in general, as projects with potentially important societal impacts can get access to new sources of financing.

However, as we have shown, the conditions for transposing impact bonds to the mission of funding research are not fulfilled, because of unhedged uncertainty, and the absence of a clear determination of what is a success vs a failure in scientific research. The relevance of this instrument in the realm of research seems thus to be limited to a small portion of the whole activity and this will work only at very high transaction costs. This is not a minor issue since it would imply quite some re-engineering of the public research funding agencies (in terms of staffing and professional competences needed to run RIB). The question is then whether the cost is worth it!

But this is not to say that there is no hope for using FE in the area of scientific research. There are many other potential solutions but always fitting a very narrow portion of the scientific activity (such as the mega-fund idea, proposed by Fagnan et al. (2013), to finance cancer research).

To conclude, when all is said and done, it seems necessary to return to the economic fundamentals of scientific research (Nelson, 1959, Arrow, 1962). Scientific research fundamentally aims at producing a public good – knowledge – and in most cases, features of unhedged uncertainty and absence of clear determination of success and failure will dominate. In such cases, it seems vain to try to attract private investors (with a few exceptions). It eventually binds to a choice, which society has to make knowingly: that of bearing all the risks research entails.

Bibliography

Arrow, Kenneth J. 1962. Economic Welfare and the Allocation of Resources of Invention. National Bureau of Economic Research (ed.). The Rate and Direction of Inventive Activity, Princeton University Press, Princeton, NJ: 609–626.

David, Paul, David Mowery, and Edward Steinmueller. 1992. “Analysing the Economic Payoffs from Basic Research.” Economics of Innovation and New Technology 2 (1): 73–90.

David, Paul, and Joseph Stiglitz. 1979. “Analysis of Factors Affecting the R&D Choices of Firms.” Center for Research in Economic Growth, Research Memorandum, no. 232.

Evenson, Robert E., and Yoav Kislev. 1975. “Investment in Agricultural Research and Extension: A Survey of International Data.” Economic Development and Cultural Change 23 (3): 507–21.

Fagnan, David E, Jose Maria Fernandez, Andrew W Lo, and Roger M Stein. 2013. “Can Financial Engineering Cure Cancer?” American Economic Review 103 (3): 406–11.

Fraser, Alec, Stefanie Tan, Mylene Lagarde, and Nicholas Mays. 2018. “Narratives of Promise, Narratives of Caution: A Review of the Literature on Social Impact Bonds.” Social Policy & Administration 52 (1): 4–28.

Hill, Michael. 2023. „World View: How ‘Research Impact Bonds’ Could Transform Funding.“ Nature 618: 887. https://www.nature.com/articles/d41586-023-02070-1

Nelson, Richard. 1959. “The Simple Economics of Basic Scientific Research.” Journal of Political Economy 67: 297–306.

Nosek, B. A. et al. 2018. “The preregistration revolution”. Proceedings of the National Academy of Sciences 115 (11): 2600-2606.