Private businesses are usually financed by more than one investor, each of which may want to attribute some share of what those businesses achieve to themselves. For example, if two investors equally share the financing of a new factory that employs 500 workers, they may each report 250 jobs created.

That sounds simple enough, but it’s easier said than done. Attribution also tends to get muddled up with other claims of impact. One might even say attributed outcomes are not claims of impact at all. I’m going to try to clear that up, and discuss some of the complications that arise when trying to do this in practice.

Before getting started, it’s probably worth recognising that the credit (or blame) for the development outcomes created by private enterprises lies with their founders and employees at all levels, and numerous external influences. Whilst financing is necessary, and investors have some influence, when an investor reports that they created 250 jobs, that should not be interpreted as taking sole credit.

That said, statements such as “we helped financed the construction of a factory that now employs 500 people” are an indispensable part of investors’ impact reporting, and the remainder of this article is concerned only with attribution among investors.

Unattributed impact data (e.g. the total number of workers employed by portfolio companies) is unproblematic when interpreted correctly as portfolio descriptive statistics but it can be liable to misinterpretation as one investor taking credit for outcomes that other investors also contributed towards, and these numbers can verge on the ridiculous when a portfolio contains very small stakes in very large companies.

The word impact means different things in different contexts.[1] If many banks would be willing to finance the construction of a factory, the bank which actually supplies the loan can claim its money was used to build a factory but cannot claim the factory would not have existed without it. The correct way for a bank to describe its impact depends on the question being answered.

Sometimes the question is: what difference did my intervention make? The answer should consist of the difference between what happened after the intervention, compared to what would have happened without it.[2] In the context of investment, if one investor does not invest then another might. Development finance institutions (DFIs) exist to do things that private investors are not, so the possibility that a private investor might undertake a project is especially important. Money allocated by governments to DFIs cannot be counted as Official Development Assistance unless investments are delivering something that would not have happened otherwise. This is often referred to as additionality, but impact investors use the term contribution to combine the questions of whether they are doing anything others would not have (additionality) with the question of what difference that makes to development outcomes.

Sometimes the question is: what resulted from the investment? Attributed development outcomes answer this question. A bank can report what its money was used for, without making any claim about what would have happened under an alternative scenario. That entails taking responsibility for having financed a factory and everything that follows from that, including the good (such as jobs created) and bad (such as greenhouse gases emitted). We would not accept an investor claiming zero responsibility for carbon emissions because another investor would have made the same investment in its place.[3]

The answers to these two questions (“what difference did our intervention make?” and “what resulted from the investment?”) only coincide when nothing would have happened without your investment.

Attributed outcomes are not claims about the impact of an intervention. Investors usually cannot report what difference their intervention as an investor makes, because whilst the companies they invest in can report how many workers they employ, for example, they cannot report how many they would have employed under an alternative history. The causal impact of interventions can sometimes be estimated from a dedicated research effort, but it is not always possible to distinguish causation from correlation.[4]

The most natural basis for attributing development outcomes is the proportion that an investment represents in the total amount of capital being raised, or the total capital base of an enterprise. If you only provided 10% of the money needed to build a factory, your attributed job creation should be 10% of the total.

The methodology proposed for by the GIIN is to attribute an enterprise’s development outcomes to individual investors by multiplying them by:[5]

Investment amount outstanding / enterprise value.

PCAF also specifies that a financial institution’s share of emissions is proportional to the size of its exposure to the borrower’s or investee’s total (company or project) value. The method varies somewhat according to asset class and whether the instrument of investment is listed or private.[6]

There are flaws with these approaches, but the absence of a perfect solution should not prevent the adoption of a reasonable solution. Especially if the method is easily understood and its limitations are visible.

There are arguments that more risk-bearing capital should take proportionately more credit (and blame) for enterprise outcomes. This principle is evident in the agreed MDB/DFI methodology for recognising mobilization, for example, where mobilization is based on which party is taking the commercial risk, not on which party is disbursing funds (liquidity).[7] If an MDB guarantees 50% of a loan provided by a commercial bank, it is considered to have provided 50% of the finance itself and to have mobilized only the other 50%. Enterprise financing involves a spectrum of risk exposures from equity through to the most senior of loans. An attribution method that weights development outcomes in proportion to risk exposure would be complicated, and difficult to agree between investors.

Methods based on enterprise value priced to market suffer from the problem that when market valuations rise and fall, the valuation of equity changes while the quantity of debt does not. Most people would probably not accept that the owners of a business (equity investors) bear less responsibility for a company’s carbon emissions as its share price falls.

The value of total assets (or total liabilities), as recorded in the enterprise’s balance sheet, could also be used at the denominator in an attribution calculation, in which case the value of each investment in the attribution formula must also be at book value not market value. This method also suffers from the problem that the book value of equity relative to debt moves around as accountants revalue assets.

In principle attribution could be based on the historical value of each investor’s contribution to primary fund raisings, which is not affected by subsequent changes in valuations either by the market or by accountants, but the historical information required to do that will often not be available and the method breaks down for investors that have acquired equity at market prices in secondary transactions.

Methods based on total enterprise value or total assets are also better suited to attributing the level of a development outcome among investors, not changes. Suppose a business employs 1000 people and has an enterprise value of £5m. An investor that holds 10% of the EV might report an attributed “jobs footprint” of 100. But now suppose that the business raises £2m from two new investors and creates 200 new jobs and the enterprise value rises to £7m. The level of employment is now 1200 and the change is 200. Each of the two new investors could reasonably claim to have financed the creation of 100 jobs. Their share of enterprise value is 1/7 each. Applying that to the change in employment would attribute 28 jobs created to each (and applying that to the new level of employment would attribute a jobs footprint of 171 to each).

When new capital is being raised to finance the expansion of a firm, the obvious solution to the problem of attributing changes (job creation), is to consider each investor’s share of the sum raised against the incremental outcomes associated with it, and not apply the new investors’ share of the total enterprise value to the incremental outcomes. That is how to get the correct answer of two investors having 100 jobs created attributed to each of them, in the example above.

Firms do many things at once and raise money from different sources at different times so enterprise-level outcomes such as employment growth often cannot be neatly separated into distinct increments associated with separate capital injections. Firms can create jobs without raising fresh capital — how should that attributed be to investors?

Even in relatively simple cases, without multiple rounds of capital raising, attributing changes in development outcomes often requires some modelling. If an investment is made in a business that currently employs 1000 people and the investor assumes that it would employ 1100 people in a year’s time in the absence of their investment, then when it reports 1500 employees in a year’s time, after receiving investment, the investor can infer it created 400 jobs. That is an estimate based on comparing reported data against a modelled baseline trend. If more than one investor participated, then they can attribute those 400 jobs in proportion to the share of the capital raised.

Such exercises are only possible for a subset of investments and often require subjective judgements. This makes hard to use for annual impact reporting at the portfolio level (although understanding incremental impact is of course essential for informing investment decisions and setting monitoring plans).

As things stand, DFIs are starting to adopt the PCAF methodology for carbon accounting, but in the absence of consensus about how to attribute other outcomes, most report unattributed portfolio impact data. An easy step forward would be to agree a rough and ready method for attributing the levels of enterprise outcomes based on investments as a share of portfolio companies’ capital, as the GIIN propose. Standardising an approach to reporting estimated incremental impact is going to be much harder.

One of the unattributed development outcomes that DFIs sometimes report is the number of jobs supported by their portfolio companies. This term is liable to be misunderstood.

Jobs supported is a static concept, best understood as the job “footprint” or “reach” of a company, which describes something about a portfolio of investments. It should not be confused with job creation. Even if the total quantity of employment in an economy does not change, who supports those jobs will change as firms grow (and others shrink), assets change hands, and suppliers and customers move between businesses. It is also a concept that inherently double counts. For example, the jobs at a factory are “supported” three times, by its investors, its suppliers, and its customers.

Despite these conceptual shortcomings, there are reasonable demands on investors to describe the economic significance of their portfolios. Models, such as the Joint Impact Model, can estimate the number of jobs that operating companies support in their supply chains and the wages they pay; that electricity suppliers support; and that bank loans support. These numbers should be interpreted as saying something about the reach of these types of business, keeping in mind that each job is reached in many ways.

Footnotes

[1] For more detailed definitions see Vosmer and de Bruijn (2017) Attribution in Results Measurement: Rationale and Hurdles for Impact Investors. DCED.

[2] In the language of empirical economics, what would have happened is called the “unobserved counterfactual”.

[3] To complicated matters, investors may sometimes estimate a counterfactual to compute more meaningful attributed development outcomes. This is discussed below in the section “incremental impact”. The crucial point about such estimates is that the counterfactual is “no investment takes place” and not “we do not invest, but somebody else might”.

[4] More or less the entire field of empirical development economics is concerned with the problem of identifying the causal effects of interventions or policies. Randomised control trials offer a neat way of comparing what happens when something is done to what happens when it is not done, but they are only feasible in certain circumstances, and very rarely in the context of development finance. Carter, Van de Sijpe and Callel (2021) “The Elusive Quest for Additionality” World Development, explains the difficulties of estimating additionality.

[5] Global Impact Investing Network (2020). Methodology for standardizing and comparing impact performance.

[6] PCAF (2020). The Global GHG Accounting and Reporting Standard for the Financial Industry. First edition.

[7] MDB methodology for private investment mobilization: reference guide