Funding bias, also known as sponsorship bias, funding outcome bias, funding publication bias, and funding effect, refers to the tendency of a scientific study to support the interests of the study's financial sponsor. This phenomenon is recognized sufficiently that researchers undertake studies to examine bias in past published studies. Funding bias has been associated, in particular, with research into chemical toxicity, tobacco, and pharmaceutical drugs. It is an instance of experimenter's bias.
The psychology text Influence: Science and Practice describes the act of reciprocity as a trait in which a person feels obliged to return favors. This trait is embodied in all human cultures. Human nature may influence even the most ethical researchers to be affected by their sponsors, although they may genuinely deny it.
Scientific malpractice involving shoddy research or data manipulation does occur in rare instances. Often, however, the quality of manufacturers' studies are at least as good as studies that were not funded by a special interest. Therefore, bias usually occurs for other reasons.
Research results can be selected or discarded to support a predetermined conclusion. The tobacco industry, for example, would publish their own internal research that invariably found minimal adverse health effects of passive smoking.
A company that hires researchers to perform a study may require the researchers to sign a nondisclosure agreement before they are funded, by which researchers waive their right to release any results independently and release them only to the sponsor. The sponsor may fund several studies at the same time, suppressing results found contrary to their business interests while publicizing the results that support their interests. Indeed, a review of pharmaceutical studies revealed that research funded by drug companies was less likely to be published, but the drug-company-funded research that was published was more likely to report outcomes favorable to the sponsor.
A double-blind study with only objective measures is less likely to be biased to support a given conclusion. However, the researchers or the sponsors still have opportunities to skew the results by discarding or ignoring undesirable data, qualitatively characterizing the results, and ultimately deciding whether to publish at all. Also, not all studies are possible to conduct double-blind.
Scientist researcher Anders Sandberg writes that funding bias may be a form of publication bias. Because it is easier to publish positive results than inconclusive or no results, positive results may be correlated with being positive for the sponsor.
Outcome reporting bias is related to publication bias and selection bias, in which multiple outcomes are measured but only the significant outcomes are reported, while insignificant or unfavorable outcomes are ignored.
Selection bias may result in a non-representative population of test subjects in spite of best efforts to obtain a representative sample. Even a double-blind study may be subject to biased selection of dependent variables, population (via inclusion and exclusion criteria), sample size, statistical methods, or inappropriate comparators, any of which can bias the outcome of a study to favor a particular conclusion.