Hacker News Reader

Comment by spangry | Hacker News Reader

As I understand the data in this article, midlife mortality rates for those who hold college degrees has declined from 1992 to 2019, whereas the rate has remained largely stable for non-college degree holders.

I wonder if this trend is due, in part, to college degree holders becoming disproportionately female over time, and women having lower midlife mortality rates? https://www.aei.org/wp-content/uploads/2018/05/degrees-1.png

View on HN

It's exactly these kind of issues with statistics that cause us all kinds of problems. I'm glad you pointed this one out.

It reminds me of a YT video I was watching with similar issues about cancer mortality rates. We've been doing all these treatments, and cancer survival rates have been going up. So everybody cheers about how good the treatments are. But when you control for the fact that earlier detection puts more people into the 'cancer' category earlier, causing 'cancer' people to live statistically longer from diagnosis, then the benefits of the treatments mostly go away (for many but not all types of cancer).

And these kinds of misleading issues are all throughout statistics. See Simpson's paradox, etc.

This seems like an extremely broad brush. There are cancers that were literally untreatable and guaranteed death within years, that with treatment now can see patients living 5+ years. Lung cancer specifically, but others as well.

You're not rebutting his point.

Let's assume fate has decreed that patient X will die of lung cancer at 70. Detect it at 68, dies in 2 years. Detect it at 64, dies in 6 years. Your early detection "increased" survival by 200%.

And I think there's a lot to his point. Fundamentally, cancer can be divided into three groups:

1) Slow growth. Leave it alone and it probably never harms the patient. Many prostate cancers fall into this category.

2) Fast growth. These are the ones where the oncologists hitting it hard can make a real difference.

3) Fast growth/fast spread. The oncologists don't have a chance. Some tumors can be slowed.

Unfortunately, our ability to figure these out (other than in hindsight) is limited. Both of my parents died of stuff that spread rapidly, in both cases treatment was a negative. (Although there was some palliative stuff for my father.)

Given that Americans with and without college degrees are split pretty much 50/50, then we’d expect there to be an equal increase in non-college degrees holder mortality rates if this was caused by changing who got degrees.

I suppose it’s possible that the gender ratio change is the cause of half of the mortality decrease, and the other half is a broad decrease in mortality rates. That would cause it to cancel out in non-college degrees holder mortality holders and double in college degree holders.

Women have had higher life expectancy and lower mortality rate in a number of age brackets, including the one studied in the paper, for much of recorded history, at least in the US. So if more women graduate from college you would expect to see some decrease in average mortality rates (and eventually increase in average life expectancy) among college graduates simply because of that.

Kinda like Bill Gates walking into a bar causes bar patron's average net worth jump up a few million. Funny thing, statistics.

Not just more female, but also a broader proportion of the population in total. You'd need to control for these effects to draw any conclusions at all.

> During the same period, mortality inequality across counties rose 30 percent, accompanied by an increasing rural health penalty

Surely women aren't a 30% greater percentage of rural county dwellers than urban county dwellers, though?

Reply on news.ycombinator.com