Wednesday, 6 November 2019

NOAA accounting leads to disaster

Reblog of Roger Pielke Jr twitter thread.

Let's take a quick look at @NOAA's very popular and very misleading "billion-dollar" disaster dataset.

It's bad economics, bad science, but apparently good marketing . . .

NOAA counts up the number of disasters that cause more than a bilion $ in damage, and then presents that simple count as a time series.

That count has shown an increase since 1980, when NOAA begins its tally.

Must be climate change, right? ...

When NOAA first presented the billion dollar count, they ignored inflation, whoops.

So when this was pointed out to them, the adjusting of past losses for inflation led to NOAA quietly adding 19 new events to their dataset

But it's not just inflation

We have more people, more wealth, more buildings today than in 1980 or 1992 etc.

So a 1980 event that caused, say $950m in inflation-adjusted losses, would certainly cause > $1B today

But it is not included

Here's 9 events from 1980 not included by NOAA

Obviously a simple count of losses above a threshold is problematic because loss potentials change over time.

So why not just present total economic losses, with no threshold?

Brilliant idea . . .

Here then is total US "Hazard Losses" (from Sheldus/ASU) presented as a percentage of US GDP (via OMB) for 1980-2016 (bounded by NOAA start, Sheldus end).

The lack of any upward trend is notable

The down trend over this time period was due to a drought of big US hurricanes 2006-2016 (since ended)

Compare to NOAA over same period

So when we raised these issues back in the day, NOAA held a workshop & published a paper

They concluded:

"It is difficult to attribute any part of the trends in losses to climate variations or change, especially in the case of billion-dollar disasters"

Despite all this, the billion dollar disaster time series lives on as zombie science.

It is simple, seems to fit a narrative, and utterly misleading.

So to help protect you from this particular zombie, I have prepared the attached graph to deploy when it attacks!

No comments:

Post a comment