How a Future Review Can Improve Funding Decisions Today

A funder wants resources to match a project’s quality. Its manager wants more resources, whatever the project deserves. Asking the manager how promising the project is therefore creates an obvious problem: why not exaggerate?

In Resource Allocation with Midterm Review, Yijun Liu examines how a promised future evaluation can help address that conflict. The arXiv preprint, version 2 revised on November 5, 2026, develops a theoretical mechanism rather than testing actual grant, corporate research or startup funding decisions.

Its central result is counterintuitive: under the optimal funding rules, the funder’s expected payoff is strictly higher before the review than afterward, despite having more public information later.

A review that changes the initial bargain

The model has two periods and one project. The recipient privately knows its fixed quality. The funder wants each period’s allocation to equal that quality, with a quadratic penalty for deviations: larger mismatches become disproportionately costly. The recipient simply values the total resources received across both periods.

Between periods, a public review produces either a good or bad signal. Higher-quality projects are more likely to receive good signals, but the review need not reveal quality conclusively.

Crucially, the funder commits at the outset to initial funding and to later allocations for each possible review outcome, all contingent on the recipient’s initial report. No separate payment rewards honest reporting. Instead, the allocation schedule must make truthful reporting at least as attractive as any alternative report.

The future review therefore matters before it happens. Claiming higher quality can secure more initial resources, but it also changes the funding consequences of the subsequent evaluation.

Better matching now, more variation later

Liu calls the main result “frontloading information value.” This does not mean spending more money initially. In the quadratic-loss model, each project’s initial allocation equals its expected second-period allocation under truthful reporting.

The distinction is between expected amounts and variability. Initial funding is fixed once the report is made. Later funding fluctuates with the review. Because its conditional mean equals the initial allocation, this extra variation adds to the funder’s expected squared mismatch with quality. The paper establishes that this variation is nonzero overall, making expected second-period loss strictly greater.

That payoff is a mathematical measure of allocation accuracy—not an estimated financial return, scientific output or social benefit.

The comparison also does not show that reviews are harmful. Liu contrasts the optimal mechanism with a benchmark that ignores reports and allocates according to expected quality: the prior mean initially and the relevant posterior mean after the review. The optimal mechanism uses anticipated review outcomes to support more informative initial allocations, rather than using the review only after it arrives.

Higher claims bring tougher continuation terms

In the optimal mechanism, initial funding does not decrease with quality. After a bad review, however, continuation funding does not increase with quality. Higher-quality projects can consequently end up with less funding than lower-quality projects receiving the same bad signal.

The allocation after a good review is more complicated. It cannot be increasing throughout the quality range. Under additional conditions—for example, when even the highest-quality project can receive a bad review—it strictly decreases near the top.

For each quality, the initial allocation lies between the bad-review and good-review allocations. Thus good reviews weakly raise funding and bad reviews weakly lower it; equality is possible.

The paper’s simplest example makes the reversal concrete. Quality is uniformly distributed between zero and one, and the probability of a good review equals quality. A project with quality 0.25 receives 0.45 initially, then 0.75 after a good review or 0.35 after a bad one. A project with quality 0.75 receives 0.55 initially, then 0.65 or 0.25. These are model units, not observed funding amounts.

There is an important subtlety: in this example, every report gives a recipient the same expected total resources. Truthfulness is optimal, but not uniquely optimal. The mechanism does not make lying strictly worse.

What the result leaves open

The conclusions depend on the model’s structure. The funder can commit to future allocations; quality stays fixed; and review probabilities depend on quality, not initial funding or hidden effort. Allocations can also be negative. For some distributions, the optimal bad-review allocation involves a funding clawback.

Liu discusses funding-dependent reviews and alternative information structures as extensions, not established results. The paper’s contribution is therefore a theoretical insight: a review can shape the incentives behind today’s information, as well as guide tomorrow’s allocation. Whether those rules work in practice requires evidence beyond this preprint.