Mortality improvement is a topic of interest to many in the life insurance industry, and it will surprise none of our readers that we at SCOR are no exception! Assumptions for future mortality improvement can be derived in many different ways but one thing all methods have in common is that they rely to some extent on historical mortality improvement rates – the actual, measurable mortality improvement rates in past periods. (For more information, see “US Mortality Improvement Analysis Methodology” in the December 2015 issue of SCORviews.)
At SCOR we recently compared historical mortality improvement rates among several different countries, including the US. Before discussing these comparisons, let us state some definitions.
Definition of annual mortality improvement:
So if the mortality rate does not change, annual mortality improvement is zero. If the mortality rate changes, for example, from 10 per 1000 to 9.9 per 1000, the annual mortality improvement rate is a positive number, 1.0%. Likewise, if the mortality rate increases year over year then the annual mortality improvement rate is a negative number.
For purposes of this article, we begin with sex and attained age specific mortality rates by calendar year. We consider males and females separately but instead of considering individual attained ages we look at a weighted average mortality rate across the population. For this purpose we used a set of weights more representative of the insured population than of the general population. We also used the same set of weights regardless of country, removing the weight vector as a possible source of variation
Finally, our source data come from the Human Mortality Database (HMD) as of January 2016. We used the CDC multiple causes of death file to compute the mortality rates in the population for 2014 – one year beyond the current range of the HMD. This method has been back-tested and closely approximates the HMD data for years where they overlap.
With these preliminaries out of the way, let’s look at some results.
Annual mortality improvement rates for males (blue) and females (red) is closely correlated in the US.
Figure 1 shows the annual mortality improvement for men and women in the US. Several features of this graph stand out. One is the close correlation of the female and male results. Another is the negative auto-correlation with frequent up and down movements. But the most overriding feature is the trendline which is apparent: each succeeding ‘good’ year is less good and the last few years have had small improvement rates, or even negative ones (females 2011 and males 2013). Improvement seems to have slowed, perhaps even to have reversed.
Although our research was driven originally by interest in US mortality improvement trends, we decided to broaden our research to compare trends in the US with those in other developed economies. For comparative purposes we limited our review to markets with sufficient population and up-to-date HMD records. All figures are reported using the same scale (-4%-+8%) and the same time frame where possible (2001-2014).
Mortality improvement for females in Austria slowed down following 2011. The relative mortality improvement from year to year appears to be the most volatile of this sample. We believe that this volatility is partly attributable to the small population: it is the smallest nation among our five nations studied.
Results for France show slowdowns for both males and females after 2011. This decline is more pronounced among females.
Similar to the data from Austria, mortality improvement for females in Sweden was consistent over the observed period with the exception of 2012. Male mortality improvement similarly was fairly stable over the period, and in fact increased slightly after 2011.
UK data illustrate a significant slow-down in mortality improvement after 2011. Annualized improvement from 2000-2011 was about 2.0%, but fell to 0.5% over 2011-2013. Males exhibited a similar drop in annual improvement. The Continuous Mortality Investigation (CMI) released a report in October 2015 indicating that this trend continued in 2014 and 2015.
Of the five nations we observed during this study, females in each country experienced a slowdown in mortality improvement following 2011 when compared to the decade before. Males in three of the countries exhibited similar trends.
Changes in annual mortality improvement can have many different causes and interactions. Examining the causes of these changes was beyond the scope of the original study. However, given the nature of the findings, more research may be warranted for specific countries and/or periods of time.
To the extent that historical mortality improvement rates are used to project future mortality improvement rates, it will be interesting to see if post-2011 trends persist as the “new normal.”