Tuesday, April 26, 2016

The Curious Case of The Dropping Life Expectancy At Birth For Non-Hispanic White Women. A Lesson in Interpretations.

I've been reading about the recent changes in life expectancy at birth* for various US demographic groups.  The main findings are summarized here:

Between 2013 and 2014, overall life expectancy at birth for the total US population held steady at 78.8 years, according to the latest data from the Centers for Disease Control and Prevention's (CDC's) National Center for Health Statistics (NCHS).
For men, life expectancy was 76.4 years, and for women, it was 81.2 years, Elizabeth Arias, PhD, from the NCHS, Division of Vital Statistics, notes in a new NCHA data brief, published online April 20.
Life expectancy increased by 0.4 years for non-Hispanic black men (going from 71.8 to 72.2 years) and by 0.1 years for Hispanic men (going from 79.1 to 79.2 years). Life expectancy remained unchanged for non-Hispanic white men (76.5 years).
Among women, life expectancy increased by 0.2 years for Hispanic women (going from 83.8 to 84.0 years), remained unchanged for non-Hispanic black women (78.1 years), and declined by 0.1 years for non-Hispanic white women (going from 81.2 to 81.1 years).

I have bolded the sentences which describe changes in life expectancy at birth. Neither non-Hispanic black women nor non-Hispanic white men experienced a change in the life expectancy measures, but the other groups mentioned in that quote did. Life expectancy at birth rose for non-Hispanic black men and for both Hispanic men and women, but it fell for non-Hispanic white women.

When I started looking for the reasons behind that drop** (because this is the first recorded drop in a more general life expectancy measure in the US during a time period without major wars or epidemics)  I came across this, in the context of an interview with various experts:

WILLIAMS: It's important to realize that, although the life expectancy is declining for white women, that white women, nonetheless, are still living almost 10 years longer than African-American men. So there still is a large racial gap in health.
KODJAK: And that gap is even wider with Hispanic men and women. The CDC report shows they live longer than everyone else. Alison Kodjak, NPR News.

What these statements refer to is the following:  Arrange the life expectancy at birth figures for the six demographic groups mentioned in the above quote in an order from the longest life expectancy to the shortest and you get this list:

1.  Hispanic women (84)
2.  Non-Hispanic white women (81.1)
3.  Hispanic men (79.2)
4.  Non-Hispanic black women (78.1)
5.  Non-Hispanic white men (76.5)
6.  Non-Hispanic black men (72.2)

Williams' point is that non-Hispanic white women, as a group, still have a higher life expectancy at birth than non-Hispanic black men and most other groups mentioned in the quote.

Williams makes a mistake, however,  when he attributes the whole difference between the life expectancy figures of non-Hispanic white women and non-Hispanic black men to race, given that one group consists of women and the other group of men.  A more correct comparison would be to compare non-Hispanic white and black women to each other or non-Hispanic white and black men to each other.***

But never mind.  That's not the point I wish to make.

It's something more complicated, a reverse of the usual kind of statistical misunderstanding where people try to refute statistical information about the average, the most frequent or the modal value of some data by noting that it doesn't jive with their own personal experience or the personal experiences of those they know.  In short, personal anecdotes are used to argue against data about averages derived from large populations of people.

In this case even experts appear to interpret data about average life expectancy in some group as if it applied equally to all its sub-groups.  Thus, we get quotes like this:

Public health experts say the rising white death rate reflects a broader health crisis, one that has made the United States the least healthy affluent nation in the world over the past 20 years. The reason these early deaths are so conspicuous among white women, these experts say, is that in the past the members of this comparatively privileged group have been unlikely to die prematurely.
Well, if we go far back into the past we are going to find a lot of premature deaths among all women due to maternal deaths.  But it is indeed true that the life expectancy at birth in the US, on average, is high for  non-Hispanic white women, higher than for all other groups mentioned here except for Hispanic women.****

But not all non-Hispanic white women are equally "privileged" in having good physical and mental health.  Some are poor, lack education, are trapped (or see themselves trapped) in poor rural areas or small towns with few jobs.

It is that sub-group among the much larger category of non-Hispanic white women which is experiencing a rapidly shrinking life expectancy at birth.

The focus on the overall group of non-Hispanic white women in this example is like not seeing that some trees are stunted because much of the forest looks so healthy.

But it would be equally incorrect to use those stunted trees to demand more resources for the whole forest.  Some experts have expressed that fear:

Others have questioned the sudden focus on whites, pointing out that African Americans continue to have shorter life spans and face severe health challenges exacerbated by racial segregation and discrimination. Why, they ask, give so much attention to a group that remains statistically advantaged?
“The truth is that white death rates are still much, much lower than they are for African Americans,” said Bridget Catlin, senior scientist at the University of Wisconsin. “My concern is that people will think, ‘Oh, it’s whites that need to be helped.’ ”

It's the stunted trees which deserve attention.  Some forests have many more of them than others.  But they all deserve help.


*  Life expectancy at birth in year x in a particular place refers to the number of years a child born in year x at that place can expect to live if current mortality patterns of year x prevail in the future.  When the measure is calculated for a demographic sub-group, the current mortality patterns that are assumed to prevail are those applying to that demographic sub-group in year x.

The bolded words are important.

Because of the way the measure is calculated it is more sensitive to deaths at younger ages than at older ages.

** Several hypotheses have been proposed to explain this change.  The more medical ones include increasing obesity and diabetes, increased misuse of opioids and heroin and the resulting overdoses, cigarette smoking among the poorer whites,  increased rates of suicide by middle-aged white women, increased rates of alcoholism by the same group and a rise in apparent accidents and accidental poisoning.

More sociological explanations suggest that economic expectations have changed for poorer whites, with more despair about the future.  Poor minorities are more accustomed not to expect a better economic future and so may not experience the same kind of loss of hope, having had little in the first place.

Some have proposed greater rates of divorce and single-parenthood among the poorer whites as one possible reason. Even the possibility that traditional gender roles, when combined with the new economic reality that most women need to work with money, might be at the root of the extra pressure poor white women experience in terms of the double shift.  On the other hand, one study found that employment seemed to have a protective effect on poorer white women's health.

Note that the crucial question all such explanations have to be able to answer is why they would explain a drop in white women's life expectancy but not in the life expectancy of otherwise identical (i.e., equally poor, equally likely not to have finished high school) white men or in the life expectancy of otherwise identical Hispanic women and men or of otherwise identical black men and women.  Some do better in that respect than others.

But what the data mostly agrees on is that this drop has its roots in the experience of one sub-group among white women:  Poor white women with little education who live in poor rural areas and small towns.  Poor white men in similar circumstances are also losing years of life expectancy, though not quite as many.  That the overall life expectancy for non-Hispanic white men has not declined could be because other improvements in men's longevity are large enough to have canceled out this particular effect.

All this is tentative.  I hope that future studies can tell us something more definitive not only about the life expectancy changes for poor white women but for all groups of the poor in the United States.

***  This article summarizes some of the issues which affect the life expectancy measures of black women and men in the US, compared to white women and men.  Some have a larger effect than others:  Neonatal deaths because of the years of life lost and the greater rates of violent deaths at early ages.  But note the general need for better access to medical care which the ACA might help to fix.

I couldn't find similar comparisons covering Hispanic men and women.

****  It is probably even higher for Asian-American women.  The available data doesn't address that question.