How Grant Cycles Set the Price of a Replication

Replication has a price, and grant cycles set it. The cost of checking a published result is usually higher than the cost of producing it, because confirmation work inherits the original's design, adds coordination across many labs, and rarely qualifies for the novelty categories that funders favor. This piece explains how that price gets set, and what a working alternative looks like.

The Price Tag of Reproducibility

Running a replication is not cheaper than running the original. The original study can be small, exploratory, and published on a single lab's budget. A replication must often be larger, preregistered, and multi-site, because the point is to test whether an effect survives outside the conditions that produced it. That adds coordination costs, shared protocols, and independent analysis.

Grant cycles rarely fund confirmation work. Most agencies organize competitions around new questions, and a proposal that says "we will re-run a published study" reads as low-risk and low-novelty to a panel trained to reward discovery. Budget lines follow the same logic. Equipment, staff, and publication costs are easy to justify for a discovery, harder for a check.

Estimates of replication cost range widely. A single small replication in social science might run into the low tens of thousands of dollars, while a coordinated multi-lab effort can reach into the hundreds of thousands. As of the mid-2020s, no standard accounting exists for what a national portfolio of replication would cost, which is itself part of the problem.

The cost structure also depends on what kind of replication is attempted. A direct replication that keeps the original materials and procedure is cheaper than a conceptual replication that tests the same idea under new conditions. But direct replications are often the ones that generate the most controversy, because any deviation from the original protocol invites dispute. That dispute resolution is itself a cost, paid in time and correspondence rather than in grant dollars.

How Funders Shape What Gets Checked

Funders do not merely pay for science; they define its categories. When a call for proposals lists "innovative" and "high-risk, high-reward" as review criteria, replication falls outside the frame. Review panels apply those criteria literally. A careful confirmation study scores well on rigor and poorly on originality, and originality usually wins the tie.

Indirect costs compound the bill. Universities recover a percentage of every grant for facilities and administration, and that percentage applies to replication budgets too. A funder that wants to buy ten replications is also buying ten administrative overheads, which makes a dedicated replication program look expensive next to a single large discovery grant.

This site has argued that Git repositories and lab notebooks diverge on reproducibility, a reminder that the infrastructure of checking is separate from the infrastructure of doing. Funders pay for the doing.

A Worked Example in Social Psychology

The most cited worked example is the large-scale replication of ego depletion, the claim that willpower is a finite resource depleted by use. A coordinated group of laboratories re-ran the basic effect under a shared protocol, with each site contributing its own data collection. The effort was crowdfunded in part, bypassing traditional grant review because no standard category fit.

Per-study costs ran into the tens of thousands of dollars when lab time, participant payments, and coordination were counted. The result challenged the original effect size: the pooled estimate was much smaller than the published literature implied, and close to zero under some specifications. The original authors disputed design choices, which is a normal part of the process.

The funding route matters here. Crowdfunding worked for a high-profile case with a public controversy attached. It does not scale to the thousands of routine findings that no one is arguing about, which is exactly where quiet errors accumulate.

Infrastructure Costs Beyond the Lab

Replication depends on infrastructure that discovery can often avoid. Data-sharing platforms require sustained funding, not a one-time build. Statistical consultants add a line item that small labs rarely carry. Publication fees for null results are real costs, and long-term archiving is almost never budgeted in a grant that ends when the paper is accepted.

A related piece on Software Heritage and legal deposit copies shows what sustained archiving actually requires: an institution, not a grant cycle. Replication has the same dependency. A study that cannot be re-examined in ten years was not really replicated, only re-run.

Statistical review is the quiet cost. A replication that corrects for multiple comparisons, handles missing data transparently, and reports a full multiverse analysis takes analyst time that a simple confirmation does not. That time is skilled, and it is rarely written into a proposal.

Incentives That Keep Replication Scarce

Promotion committees value novel publications, and a replication is hard to sell as a career contribution. Journals historically rejected null findings, which meant a successful replication of a true effect could be unpublishable, and a failed replication could be attacked as a methods complaint. Funders measure success by discovery counts, so a portfolio of replications looks unproductive on an annual report.

Replication remains a career risk for early-career researchers. A postdoc who spends two years checking someone else's finding has two years of work that hiring committees may not count. The objection is not irrational: departments are judged on grant income and citations, and replication generates neither reliably.

The counterargument deserves a hearing. Some funders say that requiring replication would slow the pipeline and divert money from work that only gets done once. That trade-off is real, and it is why blanket mandates tend to fail. Targeted programs work better than universal rules.

What Other Fields Reveal

Psychology is not the only field where replication costs collide with funding structures. In economics, the reproducibility of empirical results has drawn attention, though systematic replication remains rare. A few journals now require data and code for publication, which lowers the cost of checking but does not fund it. The American Economic Association's data availability policy, introduced in the late 2010s, made sharing a condition for publication in its journals. That policy improves transparency, yet it leaves the actual reanalysis unfunded. A researcher who wants to verify a published estimate still needs salary, computing time, and access to proprietary data, none of which the policy provides.

Biomedical research faces a different version of the same problem. Preclinical replication is expensive because animal and cell models require specialized facilities. Some funders have experimented with dedicated reproducibility programs, but they remain small relative to the discovery budget. The result is a patchwork: a handful of high-profile checks, and a vast literature that no one has the budget to revisit.

Why Replication Costs More Than It Should

The gap between the cost of doing research and the cost of checking it widens when the original study was never designed to be replicated. Many published papers omit the raw data, the analysis code, or the exact materials needed to reproduce the procedure. Reconstructing those from a methods section can take weeks of correspondence with authors who have moved on. That reconstruction is unpaid labor, and it falls on the replicator.

There is also the cost of expertise. A replication requires someone who understands the original domain well enough to know which deviations matter and which do not. That person is often the original author, who has little incentive to help. When authors decline to share materials, the replicator must either approximate or abandon the attempt. Both outcomes waste resources.

These hidden costs explain why replication is not simply a matter of will. A funder could double the budget for confirmation work and still see little output if the underlying materials remain inaccessible. The price of replication is set not only by grant cycles but by the everyday practices of data sharing and record-keeping that precede them.

Practical Steps for Reform

Dedicate a fixed percentage of every large grant to replication of the work it builds on. A five to ten percent set-aside, written into the award letter, creates a budget line that panels cannot quietly drop.

Create review tracks for confirmation studies, with their own criteria and their own panels. Reviewers who understand that a well-powered null result is a finding will score it differently from reviewers applying a novelty rubric.

Require data sharing as a condition of funding, with a named repository and a retention period. A grant that funds data collection should fund the archive that keeps it usable.

Recognize replication work in promotion criteria. Departments can list a published replication alongside a discovery publication, and say so in the tenure guidelines before the case is assembled.

Support independent replication centers with multi-year core funding. A center can absorb the coordination, statistical, and archiving costs that individual labs cannot, and it can take on the unglamorous findings that no crowdfunding campaign will touch.