Recent mortality improvement trends in the US

In a previous blog post, we analysed recent mortality improvement trends in the UK. In this post, we look at the US.

Figure 1 below displays the crude mortality rates (on the logarithmic scale) for women in the US for selected ages from the calendar year 1965 to 2016. Overall, this shows a decline of mortality rates at most ages over the time-period considered, albeit some year-to-year fluctuations.

However, the data post-2010 indicates that the speed of decline in the mortality rates observed in earlier years could be slowing down over the most recent past, at least for some ages. Similar comments apply to the crude mortality rates for men (not shown).

Although some early patterns emerge from the crude mortality rates in Figure 1, there are variations from year-to-year. However, these variations are less compared to those in the UK data (mainly due to the larger population size and death counts in the US).

Nonetheless, the variations in Figure 1 can be smoothed out by aggregating data across ages. The result in terms of standardised mortality rates for males and females is shown on the panel of the left-hand side of Figure 2; the resulting mortality improvement rates are shown on the right-hand side.

Both panels confirm the steady fall of mortality rates over time for men and women; in particular, male’s mortality rates have been falling faster than female’s (at least post-1980). Post-2010, the fall of mortality rates seems to have slowed down for both genders. However, we note that the overall pattern of reduction of mortality improvements from 2011 to 2016 for men is broadly similar that observed around 1978 to 1985.

Figure 2: Standardised mortality rates (left) and annualised 5-year mortality improvement rates (right) in the US. These improvement rates were calculated based on the standardised rates.

Table 1 presents a more detailed summary and comparison of average yearly mortality improvements by age-bands over multiple time-periods for both genders. This shows that the average improvement rates vary across time-periods and age bands.

In essentially all the age-bands shown, mortality improvements post-2010 is lower compared to the preceding decade for men and women. In particular, Table 1 shows that average improvement rates post-2010 is as low as -4% in some age-bands (e.g. see men and women aged 30-34).

The exploration carried out so far has been in terms of mortality rates and resulting improvements. Alternative metrics were considered, and the conclusion was consistent regarding the slowdown of mortality improvements in the US post-2010.

For example, Figure 3 displays comparative mortality factors by gender relative to 1965. The smoothed lines were added by fitting one-dimensional P-splines to data up to 2010. The fitted lines were then projected past 2010 using a second order difference penalty.

These projected trends show that mortality experience post-2010 in the US is heavier compared to what one could have anticipated based on 1965-2010 trends.

Further analysis using more advanced mortality forecasting models across multiple countries was also carried out. We shall report the results in upcoming posts.