24 January 2010

What a Tangled Web We Weave

[UPDATE BELOW!]

There is another important story in involving the Muir-Wood et al. 2006 paper that was misrepresented by the IPCC as showing a linkage between increasing temperatures and rising damages from extreme weather events. The Stern Review Report of the UK government also relied on that paper as the sole basis for its projections of increasing damage from extreme events. In fact as much as 40% of the Stern Reivew projections for the global costs of unmitigated climate change derive from its misuse of the Muir-Wood et al. paper.

I documented this in a peer reviewed paper published in 2007, which you can see here in PDF. In that paper I wrote:
Furthermore, the Stern Review uses the Muir-Wood et al. (2006) as the sole basis for projecting future global losses from extreme events (see Table 5.2, p. 138). This means that the Stern Review’s conclusions on the costs of future extreme events under conditions of climate change are based almost entirely on projections of future hurricane losses, which Stern projects somewhat mysteriously will increase to 1.3% of global GDP or higher. Its reliance on estimate of tropical cyclones losses is both direct and indirect. Its summary Table 5.2 on p. 138 indicates that increasing losses from hurricanes are one or two orders of magnitude larger than other losses that it has examined. . . inexplicably, the Stern Review concludes that US tropical cyclone losses will increase from 0.6% of GDP today to 1.3% of GDP under 2[degrees] of warming (Table 5.2). Yet, on page 130 the Stern Review cites Nordhaus (2007) to suggest that 2–3[degrees] of warming could double tropical cyclone losses from 0.06% of GDP (2005 losses) to 0.13% (future losses). There is no justification provided for increasing the Nordhaus (2007) values by a factor of 10. This apparent error (simply a typo?) is consistent with the Stern Review’s overstatement of future economic losses from extreme weather events more generally.
As I was preparing this post, I accessed the Stern Review Report on the archive site of the UK government to capture an image of Table 5.2. Much to my surprise I learned that since the publication of my paper, Table 5.2 has mysteriously changed! Have a look at the figures below.

The figure immediately below shows Table 5.2 as it was originally published in the Stern Review (from a web archive in PDF), and I have circled in red the order-of-magnitude error in hurricane damage that I document in my paper (the values should instead be 10 times less).

Now, have a look at the figure below which shows Table 5.2 from the Stern Review Report as it now appears on the UK government archive (PDF), look carefully at the numbers circled in red:

There is no note, no acknowledgment, nothing indicating that the estimated damage for hurricanes was modified after publication by an order of magnitude. The report was quietly changed to make the error go away. Of course, even with the Table corrected, now the Stern Review math does not add up, as the total GDP impact from USA, UK and Europe does not come anywhere close to the 1% global total for developed country impacts (based on Muir-Wood), much less the higher values suggested as possible in the report's text, underscoring a key point of my 2007 paper.

Consequently, anyone wanting to understand or replicate my analysis from the original source would no doubt be confused because evidence of the error in Table 5.2 was quietly changed after the publication of my paper. Had they noted the error it would have obviously led to questions about the implications, and ultimately the bottom line estimates of the costs of unmitigated climate change. [SEE UPDATE BELOW. THERE WAS ANOTHER POSSIBILITY I DID NOT CONSIDER.] Rather than rewrite the report, apparently, it was decided instead to rewrite history. Fixing facts to fit a policy conclusion is not a good idea for any government, but to do so with the quiet participation of leading academic advisors is doubly bad. Once again, not good.
UPDATE:

In the comments Phil Clarke points to an FAQ page on the UK Treasury site that has this interesting admission at number 20:

20. You state the cost of US hurricanes at temperatures of 3°C above pre-industrial levels as 0.13% and 1.3% of US GDP in different places in the report. Which is correct?

The correct figure is 0.13%. There is an error in Chapter 5, pg. 139, which cites the cost as 1.3%. An Errata page will be published to cover this and any other typographical errors.

The FAQ page is now in error, as Chapter 5 no longer cites the cost as 1.3%, because, as documented above, the error has been whitewashed away. I am unaware of any Errata page (readers?), however, it is now unnecessary as the report itself has been revised post-publication. Someone coming to the report would have no reason to suspect any problem. In the comments boballab writes:
The disturbing aspect is that someone checking the points in your paper would almost certainly checked the online copy of the Stern report. With the quite change with no attribution it would cast doubt over the conclusions in your paper and make it look like you did shoddy and/or dishonest work, thus damaging your credibility with the scientific community, policymakers and the public at large.
The issue is much deeper than a typo -- you can seen in my excerpt from my paper above that I had already assumed that it was a typo. The problem is that once the typo is corrected it then reveals that the numbers presented by Stern just do not add up.