When Climate Data Don’t Match the Climate Story

With trillions at stake, it pays to double-check the numbers

For decades we’ve been told a simple story: more carbon dioxide means higher global temperatures. That simplicity proved politically convincing and underpins the most expensive set of policies in modern history, covering Net Zero targets, taxes and vast subsidies to decarbonise industry, energy and transport within a generation. The world is effectively betting trillions that the temperature dial can be turned down by squeezing CO₂ emissions.

But what if that core assumption is far less robust than advertised?

My answer came in an analysis published last month in Science of Climate Change that used my background as a finance academic the way a financial analyst would interrogate a market hypothesis. In finance, no matter how elegant a theory looks, you still test it against hard data. If the numbers don’t confirm the model, the model gives way, not the other way around..

The starting point is uncontroversial: since the 19th Century both CO₂ and global temperatures have risen. The correlation is visually potent. But anyone who has worked with time series knows how deceptive such correlations can be. Ice-cream sales and shark attacks rise together in summer; that doesn’t make one the cause of the other. Two trending variables will often appear tightly linked even when the relationship is entirely coincidental.

Econometricians therefore strip out time trends and examine how annual changes relate to each other. Does each year’s increase in CO₂ reliably produce a corresponding nudge in temperature? When we look at the data since reliable measurements began around 1960, the answer is awkward. CO₂ has risen sharply, yet the rate of temperature change has not moved in lockstep. In fact, the annual changes diverge. If CO₂ were the master control knob, you would expect accelerations in CO₂ to march closely with accelerations in temperature. They don’t.

I stress-tested this across multiple temperature and CO₂ datasets. The result was the same: the neat, linear linkage between CO₂ levels and temperature weakens once you look beneath the headline trend. The famous correlation appears, at least in part, to be a statistical mirage.

Then comes causality. For CO₂ to be the principal driver of warming, changes in CO₂ must consistently lead changes in temperature. You can’t have the thermometer moving first. Yet simple regressions show no clear lead-lag pattern in the levels. And when we switch to annual changes, temperature movements often lead those in CO₂, not the other way round. That should raise eyebrows.

In finance we talk about the “joint test problem”: you often can’t measure a key variable directly and end up assuming the very thing you are testing. Climate science runs straight into this. The quantity ‘anthropogenic warming’ cannot be observed; models infer it. But because those same models assume that solar and volcanic effects explain everything that isn’t CO₂ and are tuned to match past temperatures, any conclusions risk circularity. The theory looks right because it was effectively built into the machinery from the beginning.

So what does explain temperature movements in the data? My paper doesn’t claim a full alternative climate model, but it does identify three variables with strong and persistent statistical links to temperature: the Atlantic Multidecadal Oscillation (a long-lasting pattern in North Atlantic sea-surface temperatures), atmospheric humidity and, interestingly, global cereal production. Regressions using both levels and annual changes show these variables have clearer, more stable relationships with temperature than CO₂ does. This doesn’t ‘disprove’ CO₂’s role; it simply shows that the empirical picture is far more complex than the official narrative suggests.

Concerns about spurious correlations and causal ambiguity have been raised before, but they receive little attention because the climate debate has become dominated by model-driven certainty. In most fields, relying almost entirely on internally calibrated models with limited independent empirical challenge would be treated as a recipe for groupthink. In climate policy it has become normal.

The practical implications are huge. Global policy is premised on a near-linear relationship between cumulative CO₂ and temperature. But when the relationship is confronted with historical data it looks surprisingly fragile. Yet governments are pursuing decarbonisation on the assumption that this linearity is scientifically settled and operationally dependable. If it isn’t, then the promised temperature returns on emissions cuts may be much smaller than advertised, despite the economic and social costs.

Finance has a name for this: model risk. When a model drives trillion-dollar exposures, prudent institutions demand independent validation, stress-testing and transparency about uncertainty. Climate policy deserves at least the level of scrutiny applied on Wall Street.

A prudent approach would treat the dominant CO₂ narrative as a leading hypothesis rather than revealed truth. That means diversifying policy: more emphasis on adaptation; more focus on resilient infrastructure; and greater use of ‘no-regrets’ measures that pass cost-benefit tests even if climate sensitivity to CO₂ turns out to be lower or more complicated than today’s models suggest.

It is always tempting for policymakers to dismiss uncomfortable analysis, especially when so much political capital rests on a single storyline. But genuine risk management welcomes challenge. When a finance researcher can take publicly available climate data, apply standard analytical methods and find that the central claim of global policy doesn’t survive basic statistical scrutiny, it is not a sign of scientific heresy. It is a sign that we need more open debate, not less.

Data are stubborn things. They don’t bend to political necessity. And when real-world numbers begin to contradict the narrative, serious people – especially those spending other people’s money – ought to pay attention.

Dr Les Coleman is a research fellow at the University of Melbourne and author of eight books on investment, research, risk management and biography.