Neural network model revealing the impact of gender inequality on women's life expectancy

Our neural network model reveals how gender inequality, along with the growing share of women in full-time employment and the observed positive correlation between life expectancy and socioeconomic status, may be reducing women's chances of living till a very old age (e.g. becoming centenarians).

$ \DeclareMathOperator*{\argmin}{arg\,min} \DeclareMathOperator*{\argmax}{arg\,max} \newcommand{\Kopt}{K_\text{opt}} \DeclareMathOperator{\Var}{Var} $

The gap in life expectancy between men and women in the UK has been closing since the 1980s. While both sexes live increasingly longer, men see a considerably faster improvement rate. The effect is attributed to several socioeconomic and behavioural factors: men benefiting from the industry move from heavy physical labour (manufacturing and mining) to services and adapting healthier lifestyle, while women picking the unripe fruits of emancipation (often juggling a full-time job and housework) and taking up unhealthy habits (smoking and drinking) [1].

Averisera model based on deep learning techniques discovers complex trends and patterns of future mortality rates in the UK. The heatmap shows of the ratio of male to female rate (in logarithmic scale). The lighter spots are women outliving men and the dark patches are men outliving women in particular years and age groups.

From survey data we know that regions 1 and 2 are associated with 4x higher prevalence of deaths from suicides, transport accidents and misuse of alcohol and drugs in young men, and twice higher prevalence of cardiovascular diseases in older men, as compared to their female peers, respectively. Region 3 corresponds to an increased number of women's deaths at old age, mainly due to the Alzheimer disease and dementia [2].

Our forecast for the next 80 years reveals mortality trends transforming those patterns. In dark region 4, mortality rates for women are expected to be higher than for men (although very low in both groups), possibly due to the developing problem of young women adapting unhealthy behaviours. An interesting effect is found in regions 5 and 6, revealing a weaker compression of mortality in men than women. The bright and dark patches correspond to lower and upper tails of the age of death distribution, respectively. The effect may arise from the observed positive correlation between life expectancy and socioeconomic status [3]. Along with the growing share of women in full-time employment, the gender imbalance in the most advantaged groups (more men employed in higher managerial and professional jobs) [4] may be reveal itself as an uplift in the men's age of death and the widening of its distribution. In consequence, the privileged men gain a higher chance of living to a very old age (e.g. becoming centenarians) than their average female peer, but the less lucky ones remain at a higher risk of dying younger.

Our model description and more results can be found under the following links:

Averisera Ltd: Forecasting the UK mortality rates using deep learning

“Forecasting the impact of state pension reforms in post-Brexit England and Wales using microsimulation and deep learning” (paper) and
presentation given at Asia-Pacific Regional Conference of the International Microsimulation Association 2018 in Tokyo and PenCon2018 in Łódź

Python implementation of Averisera RNN mortality model

[1] Office for National Statistics. Health state life expectancies, UK: 2014 to 2016. 2017.
[2] Office for National Statistics. Most affluent man outlives the average woman for the first time. 2015.
[3] Office for National Statistics. 2011 Census: Approximated social grade by sex by age. 2014.
[4] Office for National Statistics. Deaths registered in England and Wales (Series DR): 2015. 2015 and earlier releases.
Agnieszka Werpachowska, 10 April 2018.