Arnold Kling  

I Trust My Statistical Intuition

PRINT
Climate Change, Capital Mobili... I Heart Kevin Lang...

Bryan writes,


Sincere question: Have you personally reviewed the evidence on smoking, Arnold? I haven't. I believe that smoking causes cancer based on scientific consensus.

...Still, Arnold seems to be saying that you should base all your beliefs on direct examination of the evidence, and ignore expert consensus. I just can't buy that - and I wonder if Arnold really buys it either. Arnold?


The reason I have not reviewed the evidence on smoking and lung cancer is that I trust my statistical intuition, which tells me that this is an easy situation in which to obtain reliable data. So if the people who have looked at the data come to strong conclusions, I believe those conclusions are robust.

On global warming, I see a chaotic system with a ratio of potential explanatory variables to data points that far exceeds one. My intuition is that the climate models are highly unreliable.

Another way of saying this is that I believe that the scientific consensus on smoking is robust. I believe that the scientific consensus on global warming is fragile.

I believe that it is silly to trust all consensus estimates in all scientific fields equally. I imagine that with a bit of introspection, Bryan would agree.


Comments and Sharing






COMMENTS (13 to date)
Buzzcut writes:

Yes, statistical intuition. Epidemiologically speaking, figuring out if smoking causes cancer should be relatively easy. It's just a matter of gathering data (and the data should be relatively easy to gather and robust) and crunching numbers.

Contrast that with climate science. Data gathering is extremely difficult (how do you go back in time to gather the needed data?), and the amount of data needed to be crunched is astronomical. As a result, climate models are not statistical, they're more of the finite element variety. The computational power needed to run them as the model gets more accurate is staggering.

aaron writes:

Lungs and cells are also much more stuctured and not nearly as dynamic.

Brad Hutchings writes:

So what about second hand smoke? Cancer menace or fraud? And what percentage of smokers will contract lung cancer? Just off the top of my head, I can't believe it's 100%. Wouldn't everyone in France die of lung cancer?

ryan writes:

If the data on smoking are easier to gather, and if the scientific consensus is so much more robust, shouldn't we wonder how much we should trust Fred Singer's judgment on climate change? Larry Summers wants to compare those who don't believe in global warming to those who don't believe second-hand smoke causes cancer -- well, there's at least one example where he is perfectly, absolutely correct.

aaron writes:

2nd smoke. Fraud (though I'm not sure about asthma effects in smoking households). However, the aesthetic benefits of having non-smoking resaurants are far better than the mythical health benefits. Might be worth it if it caused a little cancer.

Arnold Kling writes:

just for clarification, both Larry Summers and I were referring to smokers and cancer, not second-hand smoke and cancer. Second-hand smoke and cancer would be a perfect example where I would want to probe the analysis before believing any purported consensus that there is a significant relationship.

Mike writes:

The scientific consensus on smoking and cancer is also older than the so-called consensus on [anthropogenic] global warming. People make extravagant claims -- "As certain as gravity," says Al Gore -- about a branch of science, paleoclimatology, that is only 25 years old (the first textbook on paleoclimate came out in 1980). A better analogy would be between the so-called consensus on second-hand smoke and the so-called consensus on [anthropogenic] global warming.

And the evidence of consensus is also shaky. Gore quotes a survey of the literature done by Oreskes. But Oreskes' methods omitted not just some, but most of the relevant literature. Oreskes says that almost none of the papers in her sample contradict the anthropogenic theory. Well, neither does my phone book nor my grocery list. A closer examination of the literature demonstrates that a clear and compelling majority of the papers are silent on the issue of anthropogenic forcing.

Furthermore, any "consensus" that does exist is suspect. The Mann team behind the hockey stick graph would not disclose their sources and methods until a congressional investigation forced the issue. The paper was published without the normal steps of peer review. Unlike the consensus that underpins the link between smoking and lung cancer, the "consensus" behind anthropogenic global warming is based on pathological science, not normal science.

Bill Conerly writes:

I've spent a lot of time on macroeconomic analysis, trying to discern trend from noise. I've seen many economists screw up their analysis because they estimated a relationship over one business cycle. This has made me very wary of conclusions based on noisy data.

I'm not saying that the consensus isn't true, but that it would be easy to get something like this wrong.

Jim writes:

If you think the science of climate change is so uncertain, then why are you not worried that global warming could be much greater than the consensus predictions? Why do you conclude - as far as I can see - that it will probably be much less? Apart from that simply being a more comforting thought, of course.

Jim writes:

Scratch that, have just seen your earlier post where you address the possibility of under-predicting global warming.

Patrick writes:

Oh, I see Arnold. So we're not required to use direct examination of the data & analysis - instead, you would encourage us to trust our "statistical intuition". Sorry, but I don't trust most people's "statistical intuition". I get your point, but think you are over-estimating the ability of most people.

aaron writes:

I thought this post, at econopundit, was amusing:

According to my blurred recollection of undergraduate European economic history, Greenland's depopulation, England's deforestation and its subsequent "early" industrial revolution (see Herbert Heaton, Rise of the British Coal Industry) -- not to mention Englang's increased international quest for resources like New World codfish, sugar cane, etc. -- were all consequences of Medieval warming's end and the coming of the "little ice age."

In other words: we adapted (wir passten uns an).

Nigel Kearney writes:

Also, lung cancer does exist. Warming of the earth outside the range of temperatures that have occurred in the relatively recent past has not (yet) happened.

In 1920, the scientific consensus was there was no reason to link smoking and lung cancer. In fact they didn't expect lung cancer to be very common at all.

Explanations of events that have occurred many times are inherently more likely to be correct than predictions of a single future event that has never happened before.

A related point is that lung cancer researchers get paid because lung cancer exists - regardless of how they attribute the cause. If global warming researchers attibute the cause to something other than human activity they may cause their research funding to dry up.

Comments for this entry have been closed
Return to top