A research paper looking at mortality rates in the UK, concludes that there could be errors in the current data provided by the Office for National Statistics (ONS) and suggests that this could lead to basis risk within longevity hedges, swaps or reinsurance transactions.
A discussion paper published by the Cass Business School Pension Institute, authored by Andrew Cairns of Heriot-Watt University, Edinburgh, David Blake of Cass Business School, Kevin Dowd of Durham University & Amy Kessler of Prudential Financial, looks at the problems that unreliable mortality data could cause for insurers and reinsurers.
The researchers suggest, according to the Financial Times which writes here (login may be required), that a statistical blip created by an uneven pattern of birth rate during the Spanish flu epidemic of 1919 in the ONS mortality data has caused the mortality rate to be underestimated.
As you would imagine, underestimating the mortality rate has implications for anyone assuming longevity risks, through insurance, reinsurance or longevity swap transactions. It could also have implications for the annuities market and those investing in these instruments.
David Blake of Cass Business School explained to the FT:
“In 1919, just after the first world war, and at the height of the Spanish flu epidemic, the births at the midpoint of the year were significantly lower than the average for that year.”
“We found that an uneven pattern of births within a given calendar year is a major cause of error in the estimated midyear population.”
“There was also a change in the method used to derive midyear population estimates in the 2001 census. As a result, the number of people born in 1919 was now overstated – introducing ‘phantoms’ into the data.”
The underestimation of mortality rate for this cohort, as a result of the phantoms introduced into the data, has then followed the cohort through time and the error has grown, according to the report. The authors suggest that this inaccuracy in the data impacts the longevity swap and pension risk transfer market, where any increase in the lifespan of covered pensioners results in an increase in payout, or the lifetime cost.
The paper suggests ways to find and remove inaccuracies in the data as well as to improve the accuracy of mortality estimates. In conclusion the paper states:
In particular, we have discovered that an uneven pattern of births within a given calendar year is a major cause of errors in population and exposures data. Different countries or agencies might derive population and exposures in different ways. But, however they do this, the estimates will be subject to potentially significant errors unless they take into account monthly or quarterly births data.
The authors also suggest that other countries, where birth and death data is recorded differently to the UK, may also be at risk from underestimation of mortality rates due to the potential for phantoms to be introduced.
The paper raises issues which interested parties in the longevity insurance, reinsurance and risk transfer market will want to review. With so much riding on the estimates of mortality rates in a longevity risk transfer transaction, having data which is as accurate as possible when you start making your longevity assumptions is vital to avoid ending up with a basis risk gap that was unexpected.
The full paper can be downloaded here in PDF format from the Pensions Institute.