I tried to post the following as a comment at http://rogerpielkejr.blogspot.de/2012/10/top-10-damaging-hurricanes-within-50.html but it did not show up (spam filter?). I repost the comment here for reference.
I have got some questions about the different incarnations of your normalised Hurricane loss data, which you seem to be using in your table. When visually comparing the data sets from 2001, 2005 and 2009 I noticed some non-trivial differences between them. For a better comparison I have digitized the graphs from your following publications:
In the following I will refer to them as P1, P5, P9, as the use 2001, 2005 and 2009 dollars, respectively.
Here is a plot of all three data sets, each of them normalized to its 1900-2000 average.
It is readily apparent that the values vary considerably from one another (extreme differences indicated by arrows) and I can not discern a systematic pattern.
I have also calculated the ratios of P9/P5, P9/P1 and P5/P1, leaving out value pairs where the denominator is “small” in an attempt to minimize the influence of digitization errors.
The data sets P1, P5 and P9 are given in 2001, 2005 and 2009 USD, respectively, so it is no surprise that the ratio is not equal to 1. What does come as a surprise to me is the huge scatter of the ratios, as well as the huge ratio of some outliers like 1933. Has there been extensive reanalysis of past losses? There is also a large change in the ratio of P5/P1 and P9/P1 with time, when looking at the periods before and after 1950, indicating that a substantial change in the damage assessment has taken place sometime between P1 and P5. Furthermore the ratio of P5/P1 is about 1.5 in the recent years and about 1.7 on average. This factor can not be explained by inflation from 2001 dollars to 2005 dollars over the four years between the data sets, as then you would have to assume an annual inflation rate of 10%.
Could you please comment on the following points regarding the ratio plot:
a) What is the reason for the strong scatter
b) What is the reason for outliers like 1933.
c) The change in ratios pre-/post-1950.
d) The magnitude of loss value changes from 2001 to 2005.
Especially c) is of interest, as I would expect this kind of change to influence trends. I did not notice a mention of change of methology in the 2006 Oceonagraphy article, but I may have missed it.
Thanks in advance.