Climate change impacts may be much worse than studies suggest

When cities plan for a hotter future, drawing on studies about how extreme heat drives crime, illness, grid strain, or crop failure, they’re relying on numbers that might be quietly wrong. 

A new study has found hidden measurement errors in the temperature data used in much of this research.

Those errors could make the real impacts of climate change look smaller than they are, understating them by at least 15 percent on average.

The research was co-authored by Wake Forest University assistant professor of economics Alex Yu and distinguished professor of economics Richard Carson from the University of California, San Diego.

Measuring temperature is complicated

Weather stations are scattered miles apart, which means nobody actually has a thermometer reading for most specific neighborhoods, cities, or farm fields. 

Instead, researchers use statistical estimates based on nearby observations to fill in the gaps between weather stations.

It’s a reasonable workaround in principle, and one that’s been used across an enormous body of climate impact research for years. 

The problem, according to this study, is that those estimates carry real errors, and the traditional statistical methods researchers use weren’t built to catch them.

Where the errors hide

The researchers identified several distinct ways this measurement problem shows up, and each one on its own is enough to meaningfully skew results.

In mountainous terrain like Boulder, Colorado, standard weather proxies differed from actual temperatures by an average of nearly 15°F.

That is a substantial gap for a variable researchers often treat as precisely known. 

In flatter, more densely monitored places like Chicago, the same kinds of errors shrank to as little as 1°F. 

The terrain and monitoring density of a place, it turns out, dramatically shape how trustworthy temperature estimates actually are.

This means the accuracy of a climate study can depend as much on geography as on methodology.

Why the problem persists

These errors aren’t random noise that simply cancels out either. They persist over time and vary systematically with local geography. 

When researchers apply these estimated temperatures to real-world outcomes, like crime rates or heat-related illness, the resulting estimates get pushed up or down in a biased, non-random direction rather than averaging out to something close to the truth.

Perhaps most strikingly, the researchers found that for one commonly used weather proxy, temperature measurement errors averaged 3.3°F (1.85°C). 

That’s larger than the entire warming limit set by the Paris Agreement of 1.5°C.

Putting the errors to the test

To see how much this actually matters in practice, the researchers turned to a massive real-world dataset: more than 20 million crime records. 

They first estimated the relationship between extreme heat and violent crime using city-level crime and temperature data as a benchmark.

This gave them a relatively fine-grained benchmark for how heat and violence track each other.

Then they aggregated the same crime and temperature data up to the county level, a coarser geographic scale commonly used in economic research, and ran the same analysis again. 

The estimated effect of extreme heat on crime shrank noticeably at that coarser scale – a clear demonstration of how a seemingly minor methodological choice can quietly understate a real and important effect.

“If climate-impact studies misestimate how severely extreme heat drives crime rates, crop loss, or health emergencies, policymakers may also misestimate the scale of the responses needed to address those impacts,” Yu said.

How to fix the problem

The study doesn’t stop at diagnosing the problem. It also offers researchers practical guidance for doing better going forward.

There’s no single temperature proxy that works best everywhere, which is itself a useful finding for anyone designing future studies. 

In areas with dense weather-station networks, simply using measurements from nearby stations can actually outperform more sophisticated gridded data products. 

In places where monitoring is sparse, though, those same gridded products tend to perform better. 

The right tool depends heavily on the specific location being studied, rather than any one method being universally superior.

Investing in better monitoring

More broadly, the study makes the case that better spatial monitoring of temperature is foundational to getting climate research right in the first place. 

Expanding weather station networks and matching temperature measurements more closely to the places being studied could reduce these errors at the source.

For a field of research that directly shapes how governments prepare for a warmer world, that kind of foundational fix may matter just as much as any single new finding about heat’s effects.

Get the thermometer readings wrong, and everything built on top of them, from crime forecasts to crop models to public health planning, inherits that same blind spot, potentially affecting every policy decision that follows.

The study is published in the Journal of the Association of Environmental and Resource Economists.

—-

Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates.

Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.

——