Loading today’s stories
Science1 distinct publisher3 min readPublished
A UNSW-led study builds its climate exposure measure from US temperature records alone, then finds the banks carrying more of that exposure wrote about 10 percent fewer small-business loans in median-risk counties before any losses landed.
The Scientist · Science desk
Compiled by The ScientistSomething wrong?How this is made
Start with the arithmetic, because the two magnitudes reported side by side can be read against each other. Five fewer loans is the average effect [7] against a decline of about 10 percent [5], which puts the baseline near 50 small-business loans per bank-county pair each year [1]. The US$130,000 is the same effect measured as a 5 percent fall in value [5][7], implying roughly $2.6m of lending in that same cell [2]. That works out to about $52,000 for the average loan in the portfolio and about $26,000 for the average loan that disappears [3][4]. The credit withdrawn sits at the small end, around half the size of the loans the bank keeps writing. The division assumes the median-exposure percentages and the average-pair dollar figures describe the same unit, which is how the source presents them [5][7].
Then the design, which is the reason to take the lending result seriously and also the reason it cannot answer the question people most want answered. The researchers compared different banks lending into the same US county in the same year [9]. That absorbs everything local: borrower demand, property values, and whatever insurers are charging or refusing in that county [9]. What survives the comparison is a bank's own portfolio exposure, which is the point. But insurance variation lives inside the part of the data this method deliberately discards, and the study did not examine premiums directly [9]. Cortes's reading is that insurance applies to everyone in a county equally, and that a bank is weighing the risk of repayment rather than the insurability of a particular property [10]. So the evidence here does not establish that credit repriced before insurance did. It establishes that credit moved before losses did.
On causation, the measure was validated forward rather than backward: counties with larger deviations from historical temperature norms went on to have more disasters and greater damage [3]. That is what licenses the word anticipation [15], and it is a stronger claim than a post-disaster study can make, since most of this literature begins after the storm [16]. The lending finding itself remains an association [5], and it is reported per one-unit move in a bank-level index whose scale the account does not state, so it will not convert into an effect per degree of warming [5].
What follows is a channel that operates with no loss event in it. Exposure was constructed without hurricane, flood, fire or loss records [2], the banks holding more of it also held more capital [4], and the reallocation was selective rather than a general pullback [6]. Small businesses are the riskier borrower class to begin with [13], the withdrawn loan averages $26,000 [4], and the places losing that lending are the places that need capital to adapt [14]. The effect grew larger once climate shocks actually arrived [8], which is the part that looks like the older literature; the interesting part is everything that happened before that.
Ranked by verification strength, evidence, and original report placement.
Kristle Romero Cortes is an associate professor at the UNSW Sydney Business School and senior deputy director of the UNSW Institute for Climate Risk & Response, and is a co-author of the study.
The study's measure of systematic physical climate risk was created from US temperature data alone, not from records of hurricanes, floods, fires or the financial losses they caused.
Counties more exposed to broad temperature shifts subsequently experienced more disasters and greater damage, according to the study.
Cortes says banks with greater exposure to physical climate risk held more capital and altered their lending behaviour.
For a county with the median level of climate exposure in the study, a one-unit increase in a bank's physical climate risk measure was associated with about 10% fewer small-business loans and a 5% fall in the value of lending.
Cortes says that rather than cutting credit everywhere, banks shifted it away from places that were more exposed to the same climate risks.
Distinct publishers with included, body-backed reporting in this cluster.
phys.org
1 article · September 4, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
science
What you expect from your own old age shows up a decade later in who you still see1 distinct publisher
science
NASA's two Sun papers ask paleoclimate to carry an astrophysical term2 distinct publishers
science
Researchers trace a third of Central Valley groundwater loss to 122 named producers1 distinct publisher
science
The sharp crystal seed was never there: 8,000 nuclei imaged atom by atom, all gradients1 distinct publisher
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
One interview, no paper
The 10 percent, the five loans and the US$130,000 all come from the same conversation with one of the study's authors: no journal named, no co-authors, no link to a working paper, and no scale for the "one-unit increase" the whole effect is measured in. The validation step that makes the temperature-only measure credible — hotter-deviating counties later suffering more disasters — is reported as a result, never shown. The design description is coherent and the numbers are internally consistent, which is why this is not a floor score.
Nothing to count yet
This is a research result, and our coverage records no supervisory guidance, bank disclosure, replication or dataset release that an outsider could verify. The lending shifts described happened inside the study's own sample, which is a finding about past behaviour rather than evidence that anyone has taken the measure up — so we score nothing here rather than dress a finding as uptake.
Framing outruns the disclosure
"Priced in ahead of time" and the jump to communities being starved of adaptation finance are doing more work than the reported numbers can carry: the community-level consequence is asserted, not measured, and the headline effect is quoted in a unit whose size the reader never learns. The concrete part — banks holding more capital and writing fewer small-business loans in exposed counties — is stated plainly and without embellishment, so the overshoot is in the interpretation rather than the arithmetic.
Research communication from an interested institute
The piece reaches readers through the channel universities use to publicise their own work, and the sole voice in it is not just a co-author but the senior deputy director of the UNSW institute whose remit is climate risk and response — an institutional stake in the finding mattering. That is ordinary academic promotion rather than a commercial position, and nothing here suggests funding or client interests, so the pull is real but bounded.
Plausible mechanism, unverifiable numbers
We are reasonably sure what the study says and reasonably unsure whether it holds: the identification strategy is the standard one for this question and the result is consistent with itself, but a single interview with an interested author, no paper to check and no second voice leave little room for confidence in the magnitudes. Our own arithmetic on the implied baseline is solid only if the two quoted figures describe the same sample.