From a Post article while I was traveling:
Wal-Mart and Toys R Us … will stop selling plastic baby bottles, food containers and other products that contain [BPA]. … One of the eyebrow-raising statistics about the BPA studies is the stark divergence in results, depending on who funded them. More than 90 percent of the 100-plus government-funded studies performed by independent scientists found health effects from low doses of BPA, while none of the fewer than two dozen chemical-industry-funded studies did. This striking difference in studies isn’t unique to BPA. When a scientist is hired by a firm with a financial interest in the outcome, the likelihood that the result of that study will be favorable to that firm is dramatically increased. …
Within the scientific community, there is little debate about the existence of the funding effect, but the mechanism through which it plays out has been a surprise. At first, it was widely assumed that the misleading results … came from shoddy studies done by researchers who manipulated methods and data. … But close examination of the manufacturers’ studies showed that their quality was usually at least as good as, and often better than, studies that were not funded by drug companies. …
"Tricks of the trade" … include testing your drug against a treatment that either does not work or does not work very well; testing your drug against too low or too high a dose of the comparison drug because this will make your drug appear more effective or less toxic; publishing the results of a single trial many times in different forms to make it appear that multiple studies reached the same conclusions; and publishing only those studies, or even parts of studies, that are favorable to your drug, and burying the rest. … Decisions about which articles to include in a meta-analysis and how heavily to weight them have an enormous impact on the conclusions. …
The answer is de-linking sponsorship and research. One model is the Health Effects Institute … [with] an independent governing structure. … HEI conducts studies paid for by corporations, but its researchers are sufficiently insulated from the sponsors that their results are credible.
Alas this solution is mostly wishful thinking. "Government funded" does not mean "unbiased" – it just means a different mix of biases. Instead we need evaluation institutions, such as prediction markets, which can better resist funding biases of all sorts.