6 ms·

Nice UX, but the probability is probably wrong. I am in my 40s, so I'll give the current life expectancy stats about a 50% chance of being valid by the time I

Nice UX, but the probability is probably wrong. I am in my 40s, so I'll give the current life expectancy stats about a 50% chance of being valid by the time I reach the age to die. But somebody who is 10 today, will likely have a life expectancy much longer than the 80 yeas. The entire curve will be pushed out. So for that person, I'd guess the probability of this being correct is closer to 10%.

jtemplin 11y ago
It's very generous to give the current life expectancy stats about a 50% chance of being valid by the time you reach the age to die. If the historical growth rate in life expectancy continues -- or at minimum doesn't go to zero -- there's a 0% chance. Moreover as we witness regularly on HN there are many reasons to believe that the growth rate in longevity will accelerate.
peteretep 11y ago
> But somebody who is 10 today, will likely have a life > expectancy much longer than the 80 years Turns out there's a _highly_ developed and mature prediction market for this called life insurance, and they disagree with your assessment.
- jtemplin 11y ago
  You're saying we have access to the premiums on a 70+ year life insurance policy for a 10-year-old, or at least a reliable proxy?
  - peteretep 11y ago
    If you're asking "Can I insure my newborn for the rest of their life for a predictable premium?" the answer is yes.
    - ghshephard 11y ago
      Indeed (though this one doesn't cover newborns, need to be at least 15 days old): https://www.sbli.com/products/life-insurance/whole-life-for-children https://www.sbli.com/products/life-insurance/whole-life-for-...
      - jsprogrammer 11y ago
        Yes, but what is the implied life expectancy? I'd guess pretty low. Also note that policies like this require ongoing maintenance payments, otherwise you forfeit past payments and benefits. The problem with claiming that the life insurance market is a predictor for death statistics is that firms tend to keep the most important data and calculations used to derive their pricing schemes secret. The firm is only concerned with keeping to projections as far out as they can model. The consumer has no real input into the model (except as a selector of alternatives) and therefore cannot obtain much knowledge about actual death statistics (except perhaps a near meaningless lower-bound).
        ghshephard 11y ago
        There is a One-time payment option.
- reasonattlm 11y ago
  An industry that will be destroyed or bailed out when there is a large upward discontinuity in life expectancy trends thanks to advances in medicine. There is a very big change going on in medical research at the moment, shifting from not trying to treat aging as a medical condition to actually treating aging and its causes. The difference in outcomes will be night and day; some relevant technologies are transferring out of the lab and into startups even now. E.g. Gensight in France, Pentraxin Therapeutics Ltd in the UK, Oison Biotech and Human Rejuvenation Technologies in the US. Anyway, pensions and life insurance will most likely be bailed out when the crash comes, and the expectation of this socialization of losses might explain some of what is observed there in the sense that they are not responding aggressively to the actuarial community's ever more strident warnings on the uncertainty of their projections. Also people are not particularly rational in their expectations about aging - see this for example, a survey of the actuaries who are the same time as warning that life expectancy predictions are increasingly uncertain due to medical advances in the works, still base their future expectations on what happened to their grandparents: https://www.dovepress.com/longevity-expectations-in-the-pension-fund-insurance-and-employee-bene-peer-reviewed-fulltext-article-PRBM https://www.dovepress.com/longevity-expectations-in-the-pens...
  - spenczar5 11y ago
    You undermine the force of your argument by saying "when" and "will be" instead of "if" and "could be". All predictions are probabilistic, after all!
  - AnimalMuppet 11y ago
    I don't think that will destroy the life insurance industry. In fact, I think it will help them. If I buy a life insurance policy for a 15 year old today, I pay them $X in order to receive $Y when the insured dies. If the insured lives for 70 years, the insurance company has 70 years to invest $X, in order to have $Y available when the insured dies. But if the insured lives 80 years, then the insurance company has 10 more years to invest $X. They gain from that situation, rather than lose.
    - ishi 11y ago
- fredkbloggs 11y ago
  But they're probably wrong, too. Insurers have a lot of money riding on the actuarial tables, but they're still just a wisdom-of-crowds best guess. No one will know what the true life expectancy was of everyone born this year until the last of them dies; similarly, no one will know the true remaining life expectancy of everyone who is currently 50 years old until the last of them dies. I agree that actuaries are taking into consideration a lot more data and experience than any HN armchair futurist, but that doesn't mean they're going to be right. Until the 1970s, most of the increase in life expectancy at birth was due to reductions in child mortality, which from an actuarial standpoint is not very interesting because it's rare for infants to have life insurance policies. The life expectancy of a 40-year-old didn't change much from 1850 until then (source: http://www.infoplease.com/ipa/A0005140.html http://www.infoplease.com/ipa/A0005140.html). In essence, nearly everyone now makes it to 40, and what's changed greatly since is what happens after that. The widespread concern is that it's not well understood whether that trend will continue, accelerate rapidly (and if so, whether the mean will increase, skew will increase, or both), or even reverse. Actuaries are professionals paid to make good guesses; they do it well, but they are not oracles.
robzyb 11y ago
Actuarial student here. Mortality improvements are important, but immaterial in this context ;) I think that this should be taken as more of a qualitative illustration of the uncertainty of life, as opposed to a quantitative measurement.
- netcan 11y ago
  Can you elaborate? Are you saying the assumption that the curve will change significantly is questionable or that the effects of the assumed change in the curve change are immaterial? IE, if we accept an assumption that the curve will shift such that average life expectancy at birth is 110 50 years from today, how is this immaterial for a 30 year old today? My common sense understanding is that their chance of kicking it in the 30-35 age bracket is locked in, but once they get to 65, the curve will have shifted enough that their chance of kicking it between 65-70 (a more substantial risk) have meaningfully decreased. Take this extreme scenario.. say we accept Aubrey De Grey's most optimistic (and probably misquoted) prediction that "the first person to live to 1,000 will be born in the next two decades," then that seems to obviously affect people alive today materially, especially if they are young.
  - robzyb 11y ago
    > Are you saying the assumption that the curve will change significantly is questionable or that the effects of the assumed change in the curve change are immaterial? Sorry, that's not quite what I am saying. My point was that "given the purpose of this analysis, mortality factors are immaterial." If the purpose of this analysis was to give some quantitative answers that were going to be used as the basis for decisions - then it's highly likely that you would want to build in mortality improvement. However, in this case the analysis is just a simple illustration to convey a simple idea (actual lifespan vs life expectancy), and I would say that building in mortality improvement would muddy the waters and make it just that little bit harder for the audience to understand.
nxb 11y ago
The top chart looks correctly formed, but you may be right about the bottom chart on that page: "Probabilities For Years Left to Live". If I'm reading correctly, it seems they tried to deduce the bottom chart from the top chart, which is not appropriate for the reason you wrote. The top chart only considers ONE year into the future from TODAY, based on someone's age TODAY. More specifically -- It's a chart for survival analysis (predict rate of failure, based on a certain fixed time period and a certain starting date) not life expectancy (predict the time period over which a certain fixed rate of failure occurs, based on any future date.) The bottom chart though, should be far more influenced by the predicted increase in life expectancy -- since it would have to account for decades of predicted average life expectancy improvement, instead of just increase in average predicted life expectancy over one year, which probably has a negligible effect on the top chart.
adevine 11y ago
I think more important than this is the fact that there are WIDE variations in the stats based on just a few important factors like smoking status, race and income - these can actually have a bigger impact than sex. If you are a non-smoking Asian programmer in Silicon Valley your curve is going to be significantly shifted to the right compared to the average presented here. Overall, great visualization, though. Really like how it highlights the concept of chance and the range of outcomes beyond a single average number.
ChuckMcM 11y ago
Not actually valid analysis, but it is a commonly held belief. I had an interesting conversation with a person who had just passed 50 and they where coming to grips with the realization that science was not going to defeat aging before they needed it to save them. That moment of mortality when you stop thinking about all the things you want to do, and start prioritizing them to the things you have enough time left to finish. There have been a couple of articles discussing this disconnect.
- mlrtime 11y ago
  I'd be interested in reading these articles if you have them.

jtemplin 11y ago

It's very generous to give the current life expectancy stats about a 50% chance of being valid by the time you reach the age to die. If the historical growth rate in life expectancy continues -- or at minimum doesn't go to zero -- there's a 0% chance. Moreover as we witness regularly on HN there are many reasons to believe that the growth rate in longevity will accelerate.

peteretep 11y ago

> But somebody who is 10 today, will likely have a life > expectancy much longer than the 80 years Turns out there's a _highly_ developed and mature prediction market for this called life insurance, and they disagree with your assessment.

jtemplin 11y ago
You're saying we have access to the premiums on a 70+ year life insurance policy for a 10-year-old, or at least a reliable proxy?
- peteretep 11y ago
  If you're asking "Can I insure my newborn for the rest of their life for a predictable premium?" the answer is yes.
  - ghshephard 11y ago
    Indeed (though this one doesn't cover newborns, need to be at least 15 days old): https://www.sbli.com/products/life-insurance/whole-life-for-children https://www.sbli.com/products/life-insurance/whole-life-for-...
    - jsprogrammer 11y ago
      Yes, but what is the implied life expectancy? I'd guess pretty low. Also note that policies like this require ongoing maintenance payments, otherwise you forfeit past payments and benefits. The problem with claiming that the life insurance market is a predictor for death statistics is that firms tend to keep the most important data and calculations used to derive their pricing schemes secret. The firm is only concerned with keeping to projections as far out as they can model. The consumer has no real input into the model (except as a selector of alternatives) and therefore cannot obtain much knowledge about actual death statistics (except perhaps a near meaningless lower-bound).
      - ghshephard 11y ago
        There is a One-time payment option.
reasonattlm 11y ago
An industry that will be destroyed or bailed out when there is a large upward discontinuity in life expectancy trends thanks to advances in medicine. There is a very big change going on in medical research at the moment, shifting from not trying to treat aging as a medical condition to actually treating aging and its causes. The difference in outcomes will be night and day; some relevant technologies are transferring out of the lab and into startups even now. E.g. Gensight in France, Pentraxin Therapeutics Ltd in the UK, Oison Biotech and Human Rejuvenation Technologies in the US. Anyway, pensions and life insurance will most likely be bailed out when the crash comes, and the expectation of this socialization of losses might explain some of what is observed there in the sense that they are not responding aggressively to the actuarial community's ever more strident warnings on the uncertainty of their projections. Also people are not particularly rational in their expectations about aging - see this for example, a survey of the actuaries who are the same time as warning that life expectancy predictions are increasingly uncertain due to medical advances in the works, still base their future expectations on what happened to their grandparents: https://www.dovepress.com/longevity-expectations-in-the-pension-fund-insurance-and-employee-bene-peer-reviewed-fulltext-article-PRBM https://www.dovepress.com/longevity-expectations-in-the-pens...
- spenczar5 11y ago
  You undermine the force of your argument by saying "when" and "will be" instead of "if" and "could be". All predictions are probabilistic, after all!
- AnimalMuppet 11y ago
  I don't think that will destroy the life insurance industry. In fact, I think it will help them. If I buy a life insurance policy for a 15 year old today, I pay them $X in order to receive $Y when the insured dies. If the insured lives for 70 years, the insurance company has 70 years to invest $X, in order to have $Y available when the insured dies. But if the insured lives 80 years, then the insurance company has 10 more years to invest $X. They gain from that situation, rather than lose.
  - ishi 11y ago
fredkbloggs 11y ago
But they're probably wrong, too. Insurers have a lot of money riding on the actuarial tables, but they're still just a wisdom-of-crowds best guess. No one will know what the true life expectancy was of everyone born this year until the last of them dies; similarly, no one will know the true remaining life expectancy of everyone who is currently 50 years old until the last of them dies. I agree that actuaries are taking into consideration a lot more data and experience than any HN armchair futurist, but that doesn't mean they're going to be right. Until the 1970s, most of the increase in life expectancy at birth was due to reductions in child mortality, which from an actuarial standpoint is not very interesting because it's rare for infants to have life insurance policies. The life expectancy of a 40-year-old didn't change much from 1850 until then (source: http://www.infoplease.com/ipa/A0005140.html http://www.infoplease.com/ipa/A0005140.html). In essence, nearly everyone now makes it to 40, and what's changed greatly since is what happens after that. The widespread concern is that it's not well understood whether that trend will continue, accelerate rapidly (and if so, whether the mean will increase, skew will increase, or both), or even reverse. Actuaries are professionals paid to make good guesses; they do it well, but they are not oracles.

robzyb 11y ago

Actuarial student here. Mortality improvements are important, but immaterial in this context ;) I think that this should be taken as more of a qualitative illustration of the uncertainty of life, as opposed to a quantitative measurement.

netcan 11y ago
Can you elaborate? Are you saying the assumption that the curve will change significantly is questionable or that the effects of the assumed change in the curve change are immaterial? IE, if we accept an assumption that the curve will shift such that average life expectancy at birth is 110 50 years from today, how is this immaterial for a 30 year old today? My common sense understanding is that their chance of kicking it in the 30-35 age bracket is locked in, but once they get to 65, the curve will have shifted enough that their chance of kicking it between 65-70 (a more substantial risk) have meaningfully decreased. Take this extreme scenario.. say we accept Aubrey De Grey's most optimistic (and probably misquoted) prediction that "the first person to live to 1,000 will be born in the next two decades," then that seems to obviously affect people alive today materially, especially if they are young.
- robzyb 11y ago
  > Are you saying the assumption that the curve will change significantly is questionable or that the effects of the assumed change in the curve change are immaterial? Sorry, that's not quite what I am saying. My point was that "given the purpose of this analysis, mortality factors are immaterial." If the purpose of this analysis was to give some quantitative answers that were going to be used as the basis for decisions - then it's highly likely that you would want to build in mortality improvement. However, in this case the analysis is just a simple illustration to convey a simple idea (actual lifespan vs life expectancy), and I would say that building in mortality improvement would muddy the waters and make it just that little bit harder for the audience to understand.

nxb 11y ago

The top chart looks correctly formed, but you may be right about the bottom chart on that page: "Probabilities For Years Left to Live". If I'm reading correctly, it seems they tried to deduce the bottom chart from the top chart, which is not appropriate for the reason you wrote. The top chart only considers ONE year into the future from TODAY, based on someone's age TODAY. More specifically -- It's a chart for survival analysis (predict rate of failure, based on a certain fixed time period and a certain starting date) not life expectancy (predict the time period over which a certain fixed rate of failure occurs, based on any future date.) The bottom chart though, should be far more influenced by the predicted increase in life expectancy -- since it would have to account for decades of predicted average life expectancy improvement, instead of just increase in average predicted life expectancy over one year, which probably has a negligible effect on the top chart.

adevine 11y ago

I think more important than this is the fact that there are WIDE variations in the stats based on just a few important factors like smoking status, race and income - these can actually have a bigger impact than sex. If you are a non-smoking Asian programmer in Silicon Valley your curve is going to be significantly shifted to the right compared to the average presented here. Overall, great visualization, though. Really like how it highlights the concept of chance and the range of outcomes beyond a single average number.

ChuckMcM 11y ago

Not actually valid analysis, but it is a commonly held belief. I had an interesting conversation with a person who had just passed 50 and they where coming to grips with the realization that science was not going to defeat aging before they needed it to save them. That moment of mortality when you stop thinking about all the things you want to do, and start prioritizing them to the things you have enough time left to finish. There have been a couple of articles discussing this disconnect.

mlrtime 11y ago
I'd be interested in reading these articles if you have them.