4

I am interested in finding out how exactly to calculate the probability of an individual living to a certain age, say 100, based on a life expectancy of X, and knowing his current age, say 30.

I did some reading on the topic but most of the resources I managed to find online regarding the topic, provided only hard numbers or even some online calculators where you could input your variables and get the result (without showing the actual calculations).

What would the formula be for this calculation?

mdewey
  • 16,541
  • 22
  • 30
  • 57
Irina Avram
  • 143
  • 6
  • 2
    Look up 'survival analysis'. Relevant tag here: https://stats.stackexchange.com/questions/tagged/survival – mkt Mar 24 '18 at 18:35
  • 3
    If you have just the information you mentioned, you may not be able to calculate this. Imagine: everyone dies exactly at age 99. Alternatively, deaths are uniform across ages 94 to 104. In either case the life expectancy is 99, but in one case the probability of living to 100 is 0. You need to know more about the distribution. – Björn Mar 25 '18 at 07:55

1 Answers1

2

The life expectancy (the median or average age), which is a summary of an entire distribution (defined by more than just a single parameter), does not capture very well the probability to reach a particular age, which is a particular point in the distribution.

Some example using data (2014) from the Dutch bureau of statistics https://opendata.cbs.nl/statline/#/CBS/nl/dataset/7052_95/table?dl=98D9

death rates, survival rate, and probable age of death - graph 1 shows deathrate for age $i$ $$f_i$$ - graph 2 shows survival rate for age $i$ $$s_i = \prod_{j=0}^{j=i-1} (1-f_j)$$ - graph 3 shows probability for dying at age $i$ $$p_i = s_i f_i$$

These curves based on a single year are of course not accurate (you would need to use estimates for future years since the deathrate of, say, 50 year olds in 2014 might not be the deathrate of 50 year olds in 2064) but are sufficient for demonstration purposes

So for this right figure the mean age of death will be 79.82

But you might imagine that many different types of curves for $f_i$, $s_i$ and $p_i$ may result into this life expectancy number. Thus based on the life expectancy number (which is $\sum i p_i$), a number that combines information from all $p_i$, we can not say what the value is for a single $p_i$ (the probability to die at age $i$).

Sextus Empiricus
  • 43,080
  • 1
  • 72
  • 161