Global warming policy has become the world’s most expensive bet. Governments have committed trillions of dollars on the assumption that carbon dioxide (CO2) from human activity is the principal driver of rising temperatures.
The story is simple: more CO2 in the air means higher global temperature. That simplicity proved politically convincing and underpins net‑zero targets, carbon pricing, and vast subsidies to decarbonize industry, energy, and transport within a generation.
But what if that core relationship is statistically less solid than advertised?
My new paper, published in the journal Science of Climate Change, probes that possibility with a disarmingly basic question: “Could CO2 be the principal cause of global warming?”
Instead of turning to climate models, I used my financial research experience to approach the problem the way economic analysts examine a market hypothesis: by testing how well data supports the assumed cause and effect.
This approach offers promise because climate and financial markets have a lot in common. Both are complex global systems with many feedbacks, incomplete data, and multiple plausible drivers. Both rely heavily on time‑series data, where establishing causality is notoriously difficult.
In finance, skeptical regulators and risk managers insist that models be stress‑tested against hard numbers. My analysis applied that toolkit to the CO2-temperature link to provide what Nobel laureate Daniel Kahneman recommended as an outside view.
The starting point is familiar. Since the 19th century, atmospheric CO2 and global average temperature have both trended upward. This co-movement is widely taken as empirical support for a mechanistic link from CO2 to temperature.
But as economists have known for decades, two variables that rise together can look tightly correlated even if they are not directly related at all. Although ice cream sales and shark attacks both increase in summer, that doesn’t mean one causes the other.
To guard against this trap, econometricians test whether correlations persist once time trends are removed. This is done by looking at correlations between year‑to‑year changes as well as levels: does each annual uptick in CO2 produce a corresponding nudge in the global temperature?
Since reliable data became available around 1960, annual increases in atmospheric CO2 have accelerated markedly, while the annual rate of warming has stayed roughly constant.
If CO2 were the main driver and had a linear influence, the pace of warming should have sped up in step. It has not.
On this basis, the headline correlation between CO2 levels and temperature is likely spurious, driven largely by the fact that both happen to trend upward over time.
Causality is the next hurdle. For CO2 to be the principal driver of warming, changes in CO2 should consistently lead temperature changes, not the other way round.
Simple regressions confirm that levels of CO2 and temperature move together, but without a clear lead‑lag pattern. Shifting to changes, temperature does not reliably follow earlier movements in CO2, and actually leads subsequent changes in CO2.
Because robustness is crucial, the exercise is repeated using multiple temperature datasets (HadCRUT5, NASA GISS) and CO2 records (Mauna Loa, Barrow, Cape Grim).
The pattern persists: CO2 and temperature levels co‑rise, but once you look at annual changes and timing, the simple story of CO2 driving temperature falls apart.
If CO2 is not significantly contributing to global warming, then what is? While no alternative climate model was presented, three variables were identified as having a strong statistical correlation with temperature: the Atlantic Multidecadal Oscillation (a long-lived pattern in North Atlantic sea surface temperatures), global cereal production, and atmospheric humidity.
Regression of temperature against these variables, using both levels and year-to-year changes, reveals strong linear relationships that withstand the standard statistical tests. Humidity stands out as a particularly powerful factor, accounting for a significant portion of temperature variation and consistently leading the data.
This analysis is not the first to raise these statistical concerns and builds on earlier work in climate econometrics. What is striking is how little weight such work has had in policy debates compared with physics‑based models that are calibrated on the very data they seek to explain.
In other fields, relying almost entirely on internally tuned models, with limited independent empirical challenge, would be seen as a recipe for groupthink.
The critical conclusion of this analysis is that the central claim of climate science – that global warming is driven in a near‑linear relationship by cumulative CO2 – does not survive basic statistical scrutiny using historical data.
This matters for climate policy because current strategies are effectively an all-or-nothing bet on that relationship being right.
Finance recognizes this as classic model risk. When a model underpins trillion-dollar exposures, good governance demands independent validation, stress testing, and transparency about uncertainties. Climate policy deserves no less.
A prudent approach would treat the dominant CO2 story as a leading theory, not a closed case, and design policy around that uncertainty. It would put more weight on adaptation, energy resilience, and practical measures that still pay off even if CO2 proves less influential than advertised.
It is politically tempting to dismiss uncomfortable analyses as heresy. But genuine risk management values challenge.
When a finance researcher can take public climate data, apply standard tools, and cast doubt on the central plank of global climate policy, decision-makers should not look away.
They should be asking the question posed by this article, and demand that climate models and policies alike be built on relationships that have been tested and retested against the real world.
Dr. Les Coleman is a research fellow at the University of Melbourne and author of eight books on investment, research, risk management, and biography.
