Mobilised volume is the wrong test of a development finance institution

A development finance institution should be judged on additionality, and mobilised volume should stop being the figure at the front of its annual report. That is my position, and I hold it knowing that the measure I want promoted is harder to audit than the one I want demoted.

The two standards count different things. Mobilised volume is a gross number: the private capital that closed alongside a public commitment, summed across a portfolio. Additionality is a counterfactual: the part of that capital that would not have moved with the institution out of the structure. The first is arithmetic on signed documents. The second is a claim about a world that did not happen.

What a volume number can and cannot carry

Nothing on this site measures either quantity for any development bank, so take a domestic series where the same problem is in plain view and the numbers are published. Construction spending on manufacturing plant in the United States, as the Census Bureau records it, ran at 75 billion dollars a year in 2020, 82 in 2021, 125 in 2022, 202 in 2023, 244 in 2024 and 215 in 2025. Those points are annual rates: each is already a yearly figure, not one month's spending. The rise from 2021 to 2024 is 162 billion dollars a year, and the largest single-year step is 77 billion dollars a year, between 2022 and 2023.

Manufacturing output over the same years, on the Federal Reserve Board's index with 2017 set at 100, reads 93, 97.6, 98, 97.3, 96.4 and 97.3. The 2024 reading is 1.2 index points below where the series stood in 2021. Across the four observations from 2022 onward it has sat between 96.4 and 98, a spread of 1.6 index points, and it has not reached its own base year in any observation of the window. In 2025 plant spending fell 29 billion dollars a year and the output index rose 0.9 points.

That pair does not show the spending was wasted, and I am not saying it does. Plant takes years to build and longer to run, so output in any of these years reflects capacity commissioned earlier, and I have not estimated the lag. The point is narrower and harder to get around. A very large, precisely dated, fully auditable volume number sat in public view for six years, and at no point in those six years did it tell anybody whether the capacity behind it was additional to what would have been built regardless.

Why the weaker measure keeps winning

Availability is the whole explanation. A mobilisation figure exists on the day of financial close, comes off signed documents, reconciles to a ledger and survives an audit. An additionality judgement is not available on that day and may never be available. Institutions report what can be reported by the deadline, and shareholders ask for what can be tabled at the meeting.

That has a portfolio consequence, and it runs the wrong way. An institution judged on volume has every reason to take the transactions that were closest to closing without it, because those close on time and at size. The deals where the institution is genuinely the reason anything happened are smaller, slower and likelier to fail. Under a volume standard, being decisive is expensive and being ornamental is cheap.

The objection with real force

The strongest reply is that additionality is unfalsifiable and therefore corruptible. An institution asked to report on a counterfactual will report a flattering one. There is no ledger for a world that did not happen, no auditor can sign it, and the measure decays into a narrative written by the party being measured. Volume cannot be talked up in that way. It is in the documents or it is not.

I accept most of that. The objection lands cleanly on estimating the counterfactual after the fact, which is how additionality is usually attempted and why it usually fails. So do it before the fact instead. At approval the institution states in writing what it believes will not happen without it, in terms specific enough to be wrong: which tranche no commercial lender would hold, which tenor the market would not extend, which currency nobody would price. The statement is dated and it is published. Judgement then runs against a record rather than a mood, and the institution can be shown to have been mistaken.

Refusals belong in the same record. An institution that declines a transaction because the private money was already committed has produced additionality by declining, and under the present standard that decision is invisible and reads as underperformance. Publishing the declines is the cheapest evidence available that the standard is real.

What moving the standard costs

Four things change, and none of them is comfortable. The headline number falls, because volume counted for standing nearby stops being counted. The ratio of private dollars to public dollars stops working as a target, since a high ratio is as consistent with arriving late as with crowding capital in. The portfolio shifts towards first-of-kind transactions, which take longer to close and fail more often, so the loss rate rises and has to be defended as evidence rather than explained away. And the case for fresh capital has to be made on a contested measure instead of an audited one, to shareholders who find the audited one easier to sign.

Those costs are the argument for the change, not against it. A measure that falls when an institution does the easy thing is working. The present measure rises in that case, which is what makes it the wrong test however sound its arithmetic.

One admission, since this page argues from a series about American factories rather than about any development bank. The absence is part of the case. The standard that governs these institutions is the standard their data happens to support, and convenience is a poor reason to keep a test that rewards turning up to deals that were already done.

One question to put to a mobilisation figure

When a mobilisation total is quoted, ask what was written down at approval about the capital that would not otherwise have arrived. If something was written down, the figure has a standard behind it and can be tested against the outcome. If nothing was, the total is a count of dollars that were present. That is a real fact about the portfolio and no fact at all about the institution.