3 ms·

This is bogus. "Life expectancy" is the number you obtain after you run a statistical model. More precisely the Lee-Carter model [1]. The Lee-Carter model is b

This is bogus. "Life expectancy" is the number you obtain after you run a statistical model. More precisely the Lee-Carter model [1]. The Lee-Carter model is based on the idea that over more than one hundred years the average annual mortality rate (or rather its logarithm) decreases linearly for all ages and other buckets (sex or race for example). So, you fit a one factor linear model using SVD. Now, after Covid-19, we've had an outlier event. Any linear regression would be thrown out of wack after such an event. Or, to be more technical, any L2 linear regression. What is the predictive power of the Covid event for the future mortality? One can claim that people affected by long Covid will show increased mortality rates in the years and decades to come. Maybe. In any case, the Covid event holds zero predictive power to newborns. And life expectancy is by definition the life expectancy of newborns. Bottom line: this is just a statistical artifact. Life expectancy now is not lower than before Covid. If anything, Covid accelerated some advances in medicine, so life expectancy is higher than the projected increase over 2 years. [1] https://en.wikipedia.org/wiki/Lee%E2%80%93Carter_model https://en.wikipedia.org/wiki/Lee%E2%80%93Carter_model

pmoriarty 4y ago
"the Covid event holds zero predictive power to newborns" If their parents, siblings, or grandparents died or suffered long-term health consequences then newborns could be affected in all sorts of ways, up to and including shortening their lifespan. The economic impact of COVID-19 will also affect people's mortality, as fewer opportunities, less earning potential, and economic recession will affect people's health by resulting in poorer health care, unhealthier living conditions and food, and quite possibly depression -- all of which are known to affect lifespan. So I wouldn't be so quick as to rule out any affect on newborns. It's just more complicated than merely saying that X number of newborns died directly due to COVID-19.
- credit_guy 4y ago
  You are right, but all those things can't be predicted by running the Lee-Carter model, which looks at the realized mortality in the past. So, one might say, "There's a possibility future mortality could potentially be higher", this is not the statement made by Scientific American. The statement is "Our evidence tells us the mortality will be higher; it will be about as high as if we turned back time by 26 years". That is absolutely not the case. It's misinterpreting the results of a model that is applied mechanically to data that should not be applied to.
TomSwirly 4y ago

pmoriarty 4y ago

"the Covid event holds zero predictive power to newborns" If their parents, siblings, or grandparents died or suffered long-term health consequences then newborns could be affected in all sorts of ways, up to and including shortening their lifespan. The economic impact of COVID-19 will also affect people's mortality, as fewer opportunities, less earning potential, and economic recession will affect people's health by resulting in poorer health care, unhealthier living conditions and food, and quite possibly depression -- all of which are known to affect lifespan. So I wouldn't be so quick as to rule out any affect on newborns. It's just more complicated than merely saying that X number of newborns died directly due to COVID-19.

credit_guy 4y ago
You are right, but all those things can't be predicted by running the Lee-Carter model, which looks at the realized mortality in the past. So, one might say, "There's a possibility future mortality could potentially be higher", this is not the statement made by Scientific American. The statement is "Our evidence tells us the mortality will be higher; it will be about as high as if we turned back time by 26 years". That is absolutely not the case. It's misinterpreting the results of a model that is applied mechanically to data that should not be applied to.

TomSwirly 4y ago