01 · The Question
Can doing another study actually leave us worse off?
Research is usually discussed as though producing additional evidence must be beneficial. A new study adds observations, narrows an estimate, tests another population, or contributes another result to the literature. Surely knowing more is better than knowing less.
But research is not free. It consumes funding, researcher time, participant effort, institutional capacity, and opportunities to investigate other questions. Research can also delay decisions while evidence is collected.
If a study is expected to produce information with little capacity to improve knowledge or consequential decisions, those costs may exceed its benefits. In that sense, research can have negative net value.
03 · What You Need to Know
More evidence is not automatically more valuable evidence
Research value should be judged against what the study consumes
A study can produce legitimate new information and still be a poor use of resources. The relevant question is not simply whether researchers will learn something. It is whether what they are expected to learn is sufficiently valuable relative to what must be given up to learn it.
This is an opportunity-cost problem. Resources committed to one study cannot simultaneously fund another study, provide a service, support infrastructure, or be used for some other valuable purpose.
Research prioritization therefore involves choosing among competing uses of scarce resources, not deciding whether knowledge is desirable in the abstract. Formal value-of-information methods make this logic explicit by comparing the value of reducing decision uncertainty with the resources required to reduce it.
Negative value is better understood as negative net value
The phrase “negative value of research” can be misleading if interpreted to mean that the knowledge produced by a study is intrinsically harmful. The more precise concept is usually negative net value.
Information or knowledge value
The benefit produced by what the study allows researchers or decision-makers to learn.
Net research value
The value of those benefits after accounting for the resources and other consequences required to produce them.
A study can therefore generate some useful information while still having negative net value if producing that information costs more than it is expected to be worth.
Value-of-information analysis provides a formal version of this comparison
In decision-oriented research, the expected value of sample information (EVSI) estimates the expected improvement in decisions resulting from the information that a particular study could generate.
The expected net benefit of sampling (ENBS) then compares that expected information value with the expected cost of the proposed research. ISPOR's value-of-information guidance identifies EVSI and ENBS as study-level measures that can inform whether further research is worthwhile and how research might be designed efficiently.
This framework has been developed particularly extensively in health economics, where decision uncertainty and opportunity costs can be represented in formal models. The broader principle is more general: scarce research resources should not be consumed merely because additional information can be produced.
Research may add information without changing a consequential decision
Suppose current evidence already makes one option clearly preferable across the plausible range of uncertainty. Another study could make the effect estimate more precise, but all credible outcomes of the proposed research would leave the same option preferred.
The study may still have scientific value. It could contribute to estimation, replication, theory, prediction, or future evidence synthesis. But its immediate decision value may be small.
If those other benefits are also modest, an expensive additional study becomes difficult to justify. This is why greater precision should not automatically be treated as sufficient research value.
Research can have low value because it addresses the wrong uncertainty
A decision problem may contain many uncertainties, but only some meaningfully affect what should be done. A study can be methodologically sound yet concentrate on a parameter that contributes little to decision uncertainty.
For example, researchers might collect another large sample to estimate a short-term effect more precisely when the decision actually depends on uncertainty about long-term persistence. The study adds data, but it leaves the consequential uncertainty largely intact.
Research value therefore depends partly on whether the study targets the uncertainty that is actually worth reducing.
Redundant research can have diminishing returns
Replication and cumulative evidence are essential to science, so duplication should not automatically be classified as waste. A replication can reveal instability, test generalizability, identify bias, or strengthen confidence in an important finding.
Yet the incremental value of another very similar study can decline as evidence accumulates. If existing evidence already supports a robust conclusion and another study is unlikely to resolve a consequential remaining uncertainty, the expected benefit of further replication may become small.
The relevant question is not whether the topic has been studied before. It is what the proposed study adds that matters.
Poorly informative research can consume resources without resolving uncertainty
A study does not need to be redundant to have low expected value. It may simply be incapable of producing sufficiently informative evidence.
A sample may be too small to meaningfully reduce uncertainty. Follow-up may be too short for the outcome driving the decision. Measurement may be too noisy. The comparator may not reflect the actual choice faced by decision-makers. The study population may not correspond to the population in which the decision must be made.
In such cases, the problem is not merely that the study could produce an inconvenient result. The design itself may have low expected capacity to improve the relevant knowledge.
Research has opportunity costs even when someone else pays for it
A common mistake is to treat funded research as though its cost disappears once a grant has been awarded. From a broader resource-allocation perspective, the resources remain scarce.
Funding, reviewer attention, research staff, laboratory or computing capacity, participant pools, institutional support, and researchers' time could have been used elsewhere. Research opportunity cost therefore exists even when an individual research team does not directly bear the financial cost.
This is why research costs need to be compared with the consequences of remaining uncertain rather than considered only from the project team's budget perspective.
Participant burden can matter independently of financial cost
Research participants may contribute time, disclose information, undergo interventions, accept randomization, experience inconvenience, or face risks. These burdens are ethically relevant even when they are difficult to translate into monetary units.
A study expected to provide little useful knowledge is harder to justify when participation imposes substantial burdens. Ethical review involves considerations beyond value-of-information analysis, but both perspectives reinforce the importance of asking whether a study is capable of producing worthwhile knowledge.
Research can impose costs by delaying action
Another study may be valuable in isolation yet have negative consequences if decision-makers insist on waiting for it when existing evidence already supports action.
During the delay, beneficial interventions may remain unavailable, inefficient practices may continue, and affected populations may lose opportunities that cannot later be recovered.
Thus, research timing can matter as much as research content. When additional research delays a decision that already has enough evidence, the forgone benefits of waiting should enter the comparison.
A null or confirmatory finding does not make a study negative-value research
Research decisions are made before results are known. A study can be worthwhile prospectively even if it eventually finds no effect, confirms the current estimate, or leaves the preferred decision unchanged.
Before the study began, several results were possible. If those possible results could have materially improved consequential decisions, the study may have had substantial expected value even though the particular result observed did not change the action.
Conversely, a study that unexpectedly produces a dramatic finding was not necessarily a good research investment if, prospectively, its design had little expected capacity to provide useful information relative to its cost.
Watch Out
Do not classify research as valuable or wasteful simply from the result it eventually produces. Research prioritization is a prospective problem: what information could reasonably be generated, how valuable could that information be, and what must be consumed to obtain it?
Not all research value can be reduced to immediate decision change
Decision-oriented value-of-information analysis does not capture every legitimate contribution of research. Basic research can develop theory, identify previously unknown phenomena, create methods, build datasets, train researchers, or open research programs whose downstream benefits are difficult to predict.
These benefits should not be assigned a value of zero merely because they cannot be represented in an immediate decision model.
The stronger conclusion is narrower. When a study is justified primarily because it will reduce uncertainty for a particular decision, the expected improvement in that decision should be large enough to justify the resources required. When the study serves another scientific purpose, that purpose should be stated and evaluated on its own terms.
