When Scientists Slip From Evidence Into Advocacy, Climate Policy Pays the Price

Somewhere near the end of a perfectly good climate paper, the tone shifts. The modelling has been careful, the statistics rigorous, the caveats noted. Then comes the policy recommendations section, and suddenly the prose stops sounding like science. Obsolete, polluting technologies “must be forbidden”, one study declares. Citizens should be weaned off fossil fuels through unprecedented reforms, urges another. The numbers have left the building, and opinion has taken their seat.

That slippage, repeated across hundreds of papers, is the quarry that a team led by the University of Cambridge set out to track. What they found is not reassuring for anyone hoping that good climate science reliably turns into good climate policy.

The researchers trawled four databases for studies on green energy and transport published since 2019, the prominent corners of the net zero push: wind power, hydrogen, road vehicles. More than 3000 papers came up. After two rounds of screening, just 23 met the criteria of actually offering policy recommendations tied to cutting emissions. Then they graded those recommendations the way you might grade the science itself, using a new appraisal tool, and the marks were poor across the board. Every single one of the 23 studies ended up rated as carrying a high risk of bias in how it communicated its advice to policymakers.

“Our first step in making research more accessible to policy makers was to understand how researchers in engineering and climate science make policy recommendations now,” says lead author Vangelis Danopoulos at Cambridge’s Statistical Laboratory. The science, he found, was generally robust. The recommendations, the bit where science is meant to become action, were too often treated like an afterthought.

Three Ways to Go Wrong

The failures sorted themselves into a few recognisable patterns. Sometimes the recommendations barely connected to the study’s own findings, floating free as a sort of wish list of things everyone in the field already agrees with. Sometimes the language tipped over into advocacy, the “must” and “should” of a campaigner rather than the measured register of a referee. And almost always, uncertainty went undisclosed.

That last one was the worst. On the rule the team cared most about, disclosing how confident you actually are, 96 per cent of the studies were marked high risk. Plenty of papers were happy to flag uncertainty in their findings; almost none carried it through to the recommendations, where it arguably matters more. Of all 23 studies, exactly one properly addressed the uncertainty wrapped up in its own policy advice.

The tool doing the grading is itself part of the story. Danopoulos and colleagues built it, christened it the Evidence Communication Rules for Policy, or ECR-P, and this review was its first proper outing. It rests on five plain-spoken rules drawn from the Winton Centre for Risk and Evidence Communication: inform rather than persuade, offer balance but not false balance, disclose uncertainties, state how good your evidence is, and head off likely misunderstandings before they take root. Run a paper’s recommendations through those five filters and the cracks show. Most papers communicated their findings well enough; it was the leap from finding to recommendation where rigour drained away, as if the authors had switched out of their lab coats and into something more comfortable.

Why does that switch happen? Part of it is simple unfamiliarity. A climate scientist can be a world authority on turbine wake effects and still have no real grasp of how a government actually drafts an energy policy.

“Being clear about the uncertainties, being clear about the trade-offs is so important if we’re going to bring our science to the people, industries and governments who will be charged with turning that science into decisions,” says Danopoulos. The people writing policy tend to have broad remits and thin time; if they cannot see where the risks and unknowns lie, whatever they build sits on shaky ground. “Highlighting what we don’t know is just as important as highlighting what we do know,” he adds.

The Pull of Whatever Is Topical

There is something almost human in the wish-list problem, which might be why it recurs. A researcher finishes a tight piece of analysis on, say, fiscal incentives, then feels obliged to gesture at the bigger picture, and the bigger picture is whatever the climate conversation happens to be fixated on that month. “A lot of the recommendations that fell into the ‘wish list’ camp seemed to be written from a point of view of whatever was topical at that point in time,” says Danopoulos. Emotive language showed up in roughly a quarter of the studies. Tempting, perhaps, to reach for stronger words when you are trying to move a minister, but it costs the work some of its credibility. “As researchers, we have to be so careful about how we express our findings and how we make our recommendations.”

Not everyone fell short. One paper on rolling out electric vehicles across the UK treated its recommendations as a primary goal from the outset rather than an optional flourish, and it came out the best of the lot, the only study to grapple honestly with the uncertainty in its own advice. The lesson is faintly obvious in hindsight: papers that take recommendations seriously tend to make good ones. There was also a stranger absence. Among all those thousands of records, the team could not find a single engineering or experimental study that bothered to offer climate policy recommendations at all. For fields built to solve practical problems, that is a curious silence.

The fix, the team argues, is not to scold. It is partly structural: training that scientists have never been offered, and funders who treat policy reporting as part of the job rather than a courtesy. A formal reporting guideline for recommendations, of the kind that already disciplines clinical trials and systematic reviews, would help. “This work wasn’t intended as a way to scold scientists, but rather to highlight that actually turning good science into good policy is a difficult and complex process, and it’s one that many scientists simply haven’t had training in,” says Danopoulos.

The stakes are not abstract. Trust in institutions is low, scientists are under pressure to conjure net zero solutions at speed, and every recommendation built on undisclosed uncertainty is a small hostage to fortune, ready-made ammunition for anyone looking to cast doubt on the whole enterprise. Get the handover right and robust science actually shapes the decisions it was meant to inform. “Anything we can do to support scientists in that process will ultimately mean better science, and better policy to support the transition to clean energy,” says Danopoulos. The harder question, lurking under all of it, is whether a research culture that rewards the confident answer can learn to value the honest admission of doubt.

Frequently Asked Questions