Monday, August 16, 2010

Disaster Losses Over Time

The New York Times had a piece over the weekend, noting that there's been a lot of weather related disasters recently, and relating these to climate change:
The floods battered New England, then Nashville, then Arkansas, then Oklahoma — and were followed by a deluge in Pakistan that has upended the lives of 20 million people.

The summer’s heat waves baked the eastern United States, parts of Africa and eastern Asia, and above all Russia, which lost millions of acres of wheat and thousands of lives in a drought worse than any other in the historical record.

Seemingly disconnected, these far-flung disasters are reviving the question of whether global warming is causing more weather extremes.

The collective answer of the scientific community can be boiled down to a single word: probably.

“The climate is changing,” said Jay Lawrimore, chief of climate analysis at the National Climatic Data Center in Asheville, N.C. “Extreme events are occurring with greater frequency, and in many cases with greater intensity.”
Given the generally poor state of climate change reporting in the US media, this is welcome.

However, it motivated me to want to know more about the quantitative aspect of natural disasters - how significant are natural disasters to the global economy?  What is the trend?  Etc.

The best data I have found so far are from a 2009 report, Global Assessment Report on Disaster Risk Reduction, by the United Nations International Strategy for Disaster Reduction Secretariat (UNISDR).

Firstly, here are data on just the total economic losses due to natural disasters, in nominal (ie not inflation adjusted) dollars.

The growth here is, of course, pretty much meaningless - we have to take account of inflation, and furthermore we have to take account of economic growth - as the economy gets more valuable, we would expect natural disasters to cause larger economic losses even after adjusting for inflation, even if there were no trend at all in the disasters themselves.

So they also do the obvious thing and provide a graph of losses as a ratio to GDP:

Here "100" denotes the average value over 1975-2007.  Given that in 2005, the total losses were about $220 billion (US), and that global GDP that year was $45.5 trillion according to the IMF (estimated at market exchange rates), we are talking about natural disaster losses of about 0.5% of GDP that year, and the "100" average in the above graph corresponds to about 0.17% of global GDP.

These percentages sound small, but probably the way to think about them is as compared to economic growth.  If the losses due to natural disasters were to become of the same size as current year-to-year economic growth, presumably there would not be much growth (we would just be replacing the stuff destroyed in disasters).  Since global growth has generally been 4-5%/year, the worst case for natural disasters has been about a tenth of that.  Not completely negligible - though certainly not a harbinger of imminent collapse.

Of the above graph, the report itself says "Figure 2.37 shows that when these losses are adjusted for inflation and expressed as a percentage of global GDP, the trend is far less pronounced and statistically insignificant."

My eyeball glance leaves me unsure that trend is statistically insignificant. Most of the big values are on the right hand side of the graph. However, analyzing it is not completely straightforward, since the data are certainly not normally distributed and a simple linear regression will not be reliable. I hope to present a non-parametric analysis tomorrow.

However, even if there was a residual trend, I think we would still have work to do to separate any exogenous climate signal from endogenous economic issues. In particular, it's very likely that there's been an overall shift in asset prices relative to GDP during the 1975 to 2007 period of this graph, and we would expect disaster losses to be proportional to asset prices (because they will be dominated by destruction of real estate). That too will have to wait for further analysis.


Dean said...

Actually, there have been peer-reviewed published papers on this subject and you are correct, there is no statistically significant increase in disater losses.

I don't have the direct links, but the author's blog is:

Greg said...

"Reported loss" is a very narrow basis for assessment, but even so, it should be decreasing over time as a proportion of GDP. Let's review some of the other losses:

In microeconomic terms, there is the lost opportunity for further investment building on the destroyed capital (lost option value). There is unreported -- because unknown -- loss, such as reduced lifespan of apparently unharmed assets (e.g. reservoirs silting up, weakened road pavements or sub-bases). There is increased susceptibility of remaining assets to future disasters (increased risk). (For example, one part of a hillside slipping away increases the risk of neighbouring parts also slipping.) And replacement of lost assets "crowds out" other uses for finance, both directly and in increased insurance premiums (opportunity cost).

Macroeconomically, over the long term there is a social bias against capital investment in countries with disaster-prone areas, a bias greater than strict risk analysis suggests is correct.

Political repercussions can make things worse too. One scenario: The floods in Pakistan drive the country back to military rule, and thence to more clashes with India. More money would be spent on military hardware, so, with the debt servicing, there's less available for investment.

These additional risks should be powerful drivers towards loss prevention. As the economy grows and becomes more capital intensive, it makes sense to spend more money on protecting assets. (Slope stabilization, erosion control, flood control, structure strengthening, etc.) Therefore, fractional reported loss should be declining in the null hypothesis.

Anonymous said...

The GDP measurement itself could be increased by the catastrophe as recovery is capital intensive and makes jobs so it could balance out the destruction(which might reduce economic activity) which is not recorded in GDP as it is only theoretical (GDP is not a profit/ loss accounting but takes everything together not valuing one actvity over another.) So if working capital (Russian/Pakistani farmland) is destroyed GDP does not measure that as it is only an account of cash making activities.

Alexander Ac said...

"as the economy gets more valuable, we would expect natural disasters to cause larger economic losses even after adjusting for inflation"

- yes, BUT the counter-trend is much better weather forecasting (i.e. avoided economic losses). Thus graph showing losses in developed (better forecasting) vs. developing (worse forecasting) would be maybe more telling...?

Borepatch said...

I'm with Dean - Pielke has been all over this. Most recently about very questionable reports from Reinsurance companies. The data simply don't back the report's conclusions.