The urge for "fairness" in climate science, and its dangers

This article has been made free-to-view. Like what you read? Then consider buying us a coffee here👇

I think we all recognize that poorer countries closer to the equator will bear the brunt of climate change.  But suppose we limit attention to a big, rich jurisdiction like the US, UK or EU.  Will the outcomes of climate change be “fair” in some sense?  Are we all facing the existential challenge of global warming on an even keel?  Are we all standing shoulder-to-shoulder, staring down a common enemy?  A dust-up over hurricane loss data suggests we are not.

We humans have a deep-seated faith (or at least hope) that the universe is aligned fairly.  But as anyone who has studied the lifecycle of the emerald jewel wasp will tell you, nature can be a very cruel place.  Where the laws of the jungle are allowed to operate – like in our predominantly capitalist economy – some are favored and others face grave disadvantages.  Fairness is a distinctly human construct that always acts in opposition to the forces of nature.  Therefore, fairness must be manufactured through the conscious actions of humans.  

In the context of climate change, our instinct for fairness wills certain things to be true.  For example, we just know that there can’t be any winners from climate change, so if we find some, the data must be faulty.  Everyone – including evil banks, large corporations and billionaire investors – simply must take a hit as global warming takes hold.  Rich people who purchase multi-million dollar properties in coastal locations must be making a huge mistake.  Surely they will lose all of their capital as mother nature bites back in revenge for the harm wrought by decades of noxious emissions?

When someone points out how naive this all is, people get rubbed up the wrong way.  It doesn’t matter that data and theory both suggest sharply unequal outcomes from warming, that the GDP impact of dire climate scenarios might be quite mild in temperate rich countries, that climate change will produce some big winners and many losers, or that banks are indifferent to warming and may even benefit.  They know that climate outcomes will be just and fair, even if we don’t take any action to make them so.

While the instinct for fairness is an endearing and admirable trait in the general population, when scientists and government regulators succumb to it, the outcomes are always rather dire.   Even worse, because policymakers are blindsided by the reality of unfairness, they delay the action required to lean against it.  

I suspect that these are the forces at play in the following scientific and regulatory controversy.

Over the past 30 years or so, climate scientists have been trying to identify a trend in US hurricane-related property damage when observed on a constant dollar basis.   This calculation is difficult because, as Bob Dylan once said, when you got nothing you got nothing to lose.  If a hurricane strikes an empty field, it will do no damage, but if a multimillion dollar block of condos is built on the field, it could be literally ruinous.  Over the past century, Americans – and their wealth – have been migrating southward into the hurricane zone.   A lot of empty fields have thus been replaced by condos, factories, and houses.

The process of making the necessary data adjustments that allow this trend to be measured in a rigorous way is called “normalization”, pioneered by Pielke and Landsea (1998).  Their method has since been extended and enhanced by a number of researchers, and spawned over 70 peer reviewed papers that all show similar results. 

Put simply, their research suggests that hurricane damage does not exhibit an increasing trend once a valid time series has been appropriately normalized and corrected for inflation. In other words, hurricanes have not been getting more destructive –  it’s our increased building in hurricane zones that’s to blame for higher losses. An easily accessible discussion of the methodology is provided on Pielke’s Substack. 

Louie wants you to keep in mind that Pielke himself is not a climate scientist (his degrees are in mathematics, public policy and political science) and that he is currently a nonresident senior fellow at the American Enterprise Institute, a conservative think-tank with a long history of undermining climate science.   Tony wants to further add that he had no idea of Pielke’s affiliation, that his degree in math is more relevant to this issue, and that bad science is bad even if it’s done by your political allies.  If Pielke has supported climate denialism, he’s a bad scientist who doesn't hold himself to his own standards.  However, he’s right on this specific issue.

Pielke’s consensus result was seemingly upended in a paper by Grinsted, Ditlevsen and Christiansen (2019), published in the unfortunately acronymized journal Proceedings of the National Academy of Sciences of the USA (PNAS). They used a novel data set, following a methodology that is a slight variation of that developed by Pielke and Landsea, and found that the trend in hurricane-related damage survives the normalization process.  Specifically, they found that damage increases by 330% per century and that this damage is the result of human actions — migration and property building.

In other words, hurray!  Justice is served. The man who builds his house in a foolish location really does suffer higher losses than the wise man.  The world is a moral and just place after all.

The Grinsted paper was the one referenced by various climate bodies, including the IPCC, without any mention of the multitude of articles that have reached the opposite conclusion.  The new paper obviously met the panel’s requirements (i.e. it aligned with their values) so that’s the one they ran with. 

Scientists, including the aforementioned Prof. Pielke, have called on PNAS to formally retract the Grinsted paper.  It seems the IPCC’s preferred authors used a dataset that was never really intended for use by academic researchers.  It’s a bit like updating the Consumer Price Index using a YouTube survey on inflation expectations.  They both measure the same thing (kind of) but only one is reliable, consistently produced, and ready for primetime on the international stage.

Grinsted allegedly bolted this new source onto the existing, widely-used data that allows the analysis to stretch all the way back to 1900.  Naturally, when analyzing long-term trends, a lengthy sequence is needed to tease out what might be a subtle upward drift in the data. There may be circumstances where economists are forced to splice together datasets, but it’s always a very awkward process.  The data of interest might have been discontinued, and you desperately need to modify it to be comparable to a series that is still being produced.  You’ll then need to jump through a million hoops to ensure the robustness of the data and to finesse your models so they’re able to cope with unavoidable definitional shifts.

It’s dirty work that you would avoid if you could.  If there’s an existing database where others have already expended the effort, you’d only dive into the snakepit if there was a very good reason to do it.  

Not liking the results in the existing literature is a bad reason. 

The net result is that the Grinsted data replicates the older numbers used in the Pielke branch of the literature but records substantially higher numbers when the questionable data kicks in.  A break like this inflates the estimate of the trend or sticks one in where it doesn’t belong.

I’ve reproduced the key chart from Grinsted’s paper below, but I’ve removed a single outlier (Hurricane Harvey for those paying attention.  The remaining outlier is Katrina).  My PhD supervisor – an expert on hypothesis testing – used to say that his grandfather’s eyeball was the most powerful statistical test.  So whaddya reckon?  Trend or no trend?  Remember that this is the controversial dataset, so technically we shouldn’t run the test.  Nonetheless I’m feeling bold.

My eyeball comes back with a flat out negative.  Over a hundred-plus years you might statistically tease out a slight upward tilt, but even if your p-value were <0.05, we’re not talking about a powerful relationship here.  In other words, even accepting the dubious data as valid does not lead to a robust conclusion that hurricane damage is rising over time.  If you think that dropping the outlier was naughty of me, I’ll remind you that one observation does not a trend make.

For so many reasons, I wouldn’t be hanging my hat on this research.  Yet this is the kind of stuff the IPCC runs with.  It sends shivers down your spine that a critical body would apply such loose standards to its work.  There are genuine problems created by climate change, but increasingly consequential property damage in the southern US is apparently not one of them.  The IPCC may feel that there should be a problem, but their feelings are irrelevant.

Those of us who have been around for a while will be aware of the battle that had to be fought to establish the consensus that human-caused climate change was happening.  The behavior of scientists had to be perfect.  Any ethical slip-up would have been grist for the denialists’ mill.  

In the era of consensus we need to maintain our standards.  For all the climate hawks out there, be aware that reality is on your side; you don’t have to fake it.  There can be results that conflict with your worldview yet don’t undermine it.

And remember that the universe is horrendously unfair.  If you’re looking for moral economic and financial outcomes as the world warms, you will be disappointed.