Correlation Equals Causality


The indispensable Roger Pielke highlights, in a recent post, a risible piece of analysis presented in the detailed work product used to support the most recent UN IPCC Summary for Policymakers.


The key analytical exhibit purports to demonstrate that global warming has caused escalating economic losses using a “gee whiz” chart that shows a rising trend in global temperature from 1970 to 2005 moving in rough lockstep with a rising trend in global economic losses from weather-related catastrophes over the same period. 

I’ll resist the temptation to take the low road in illustrating the obvious point that correlation does not equal causality.


OK, I won’t. What the authors don’t mention is that, obviously, ANYTHING showing an increasing trend over the past 35 years will show a great fit with the losses from weather-related catastrophes line. The variable “Jim Manzi’s Age” is statistically correlated with estimated losses at the 5% significance level – so apparently I am personally responsible for many billions of dollars of disaster costs.


The much more significant point is that it’s equally obvious, even to a non-expert like me, that the first thing you would have to consider in evaluating the chart presented in this report is the growth in global wealth between 1970 and 2005, the proportion of this represented by physical assets and the spatial distribution of this wealth vs. disaster-prone areas in order to start to normalize the trend in weather-related losses.


Not surprisingly, competent analysts have considered these issues in detail. For example, see this, this and this. Not to mention this and this.


What these papers show is that once you normalize for population, wealth and inflation at the national level, there is a weak upward trend in normalized weather-related disaster losses only if you include the 2004 / 2005 hurricane season. The authors are explicit that that US losses dominate these numbers and that a shift in US population into more vulnerable areas in Florida probably accounts for any trend.


It took me 15 minutes on Google to find the relevant research, and maybe two hours to assimilate it. Apparently this was too much work for 2,500 scientists.


What’s so shocking is not that IPCC report doesn’t agree with these conclusions, but that it doesn’t even address them – it’s blind to the fundamentals of the analysis in this area.


Sign up for free NRO e-mails today:

Subscribe to National Review