I have written a review of Nicholas Wade’s book, A Troublesome Inheritance: Genes, Race and Human History, for Boston Review. Because there already are a lot of reviews published, I also included discussion of the response to the book. And because I’m not expert in genetics and evolution, I got to do a pile of reading on those subject as well. I hope you’ll have a look: http://www.bostonreview.net/books-ideas/philip-cohen-nicholas-wade-troublesome-inheritance
Category Archives: Research reports
In response to yesterday’s post, “This ‘Supporting Healthy Marriage,’ I do not think it means what you think it means,” Phil and Carolyn Cowan posted a comment, which I thought I should elevate to a new post.
Here is their comment, in full, with my responses.
Since the issue here is one of perspective in reporting, we (Phil Cowan and Carolyn Cowan) need to say that we were two of a group of academic consultants to the Supporting Healthy Marriage Project.
Thank you for acknowledging that. I noticed that Alan Hawkins, in his comment on the new study for Brad Wilcox’s blog, says he has “published widely on the effectiveness of marriage and relationship education programs,” but doesn’t say who paid for that voluminous research (with its oddly consistent positive findings). More about his Hawkins below.
Social scientists who want to inform the public about the results of an important study should actually inform the public about the results, not just give examples that support the author’s point of view.
Naturally, which is why I publicized the study, provided a link to it in full, and provided the examples quoted below.
It’s true as you report that there were no differences in the divorce rate between group participants and controls (we can debate whether affecting the divorce rate would be a good outcome), and that… [quoting from the original post]
“…there were no differences in the divorce rate between group participants and controls and “there were small but sustained improvements in subjectively-measured psychological indicators. How small? For relationship quality, the effect of the program was .13 standard deviations, equivalent to moving 15% of the couples one point on a 7-point scale from “completely unhappy” to “completely happy.” So that’s something. Further, after 30 months, 43% of the program couples thought their marriage was “in trouble” (according to either partner) compared with 47% of the control group. That was an effect size of .09 standard deviations. So that’s something, too. Many other indicators showed no effect. However, I discount even these small effects since it seems plausible that program participants just learned to say better things about their marriages. Without something beyond a purely subjective report — for example, domestic violence reports or kids’ test scores — I wouldn’t be convinced even if these results weren’t so weak.”
1. A slight uptick in marital satisfaction. The program moved 15% of the couples up one point. But more than 50 studies show that without intervention, marital quality, on the average goes down. And, it isn’t simply that 15% of the couples moved up one point. Since this is the mean result, some moved less (or down) but some moved up. Some also moved up from the lower point to relationship tolerability.
It is interesting that, with so many studies showing that marital quality goes down without intervention, this is not one of them. That is important because of what it implies about the sample. Quoting from the report now (p. 32):
At study entry, a fairly high percentage (66 percent) of both program and control group couples said that they had recently thought their marriage was in trouble. This percentage dropped across both research groups over time. This finding is contrary to much of the literature in the area, which generally suggests that marital distress tends to increase and that marital quality tends to decline over time. The decline in marital distress was initially steeper for program group members, and the difference between the program and control groups was sustained over time. This suggests that couples may have entered the program at low points in their relationships.
Back to the Cowans:
While the effects were small (but statistically reliable), they were hardly trivial. For instance, two years after the program, about 42% of SHM couples reported that their marriage had been in trouble recently compared to about 47% of control-group couples. That 5% difference means nearly 150 more SHM couples than control-group couples felt that their marriage was solid.
There are several problems here.
First, this paragraph appears verbatim in Hawkins’ post as well. I’m not going to speculate about how the same paragraph ended up in two places — there are some obvious possibilities — but clearly someone has not communicated the origin of this passage.
Second, this is not the right way to use “for instance.” This “for instance” refers to the only outcome of any substantial size in the entire study. It is not an “instance” of some larger pool of non-trivial results, it is the outlier. (And “solid” is not the same as not saying the marriage is “in trouble.”)
Anyway, third, this phrase is just wrong: “small (but statistically reliable)… hardly trivial.” For most of the positive outcomes they were exactly so small as to be trivial, and exactly not statistically reliable. Quoting from the report again, on coparenting and parenting (p. 39):
Table 9 shows that, of the 10 outcomes examined, only three impacts are statistically significant. The magnitudes of these impact estimates are also very small, with the largest one having an effect size of 0.07. These findings did not remain statistically significant after additional statistical tests were conducted to adjust for the number of outcomes examined. In essence, the findings suggest that there is a greater than 10 percent chance that this pattern of findings could have occurred if SHM had no effect on coparenting and parenting.
And quoting from the report again, on child outcomes (p. 41):
Table 10 shows that the SHM program had statistically significant impacts on two out of four child outcomes, but the impacts are extremely small. SHM improved children’s self-regulatory skills by 0.03 standard deviation, and it reduced children’s externalizing behavior problems by 0.04 standard deviation. … The evidence of impacts on child outcomes is further weakened by the results of subsequent analyses that were conducted to adjust for the number of outcomes examined. These findings suggest that there is a greater than 10 percent chance that this pattern could have occurred if SHM had no effect on child outcomes.
In other words, trivial effects, and not statistically reliable.
2. You say that “Without something beyond a purely subjective report…I wouldn’t be convinced even if these results weren’t so weak.” You were content to focus on two self-report measures. At the 18 month follow-up, program group members reported higher levels of marital happiness, lower levels of marital distress, greater warmth and support, more positive communication skills, and fewer negative behaviors and emotions in their interactions with their spouses, relative to control group members. They also reported less psychological abuse (though not less physical abuse). These effects continued at the 36 month follow-up [should be 30-month -pnc]. Observations of couple interaction (done only at 18 months) indicated that the program couples, on average, showed more positive communication skills and less anger and hostility than the control group. Because the quality of these interactions of the partners, the effects, though small, were coded by observers blind to experimental status of the participants, meaning that not only the self-reports suggest some positive effects but observers could identify some differences between couples in the intervention and control groups that we know are important to couple and child well-being.
I am confused by this. The description of the variables for communication skills and warmth (p. 67) describes them as answers to survey questions, not observations (e.g., “We are good at working out our differences”). I’m looking pretty hard and not seeing what is described here. The word “anger” is not in the report, and the word “hostility” only occurs with regard to parents’ behavior toward children. Someone please point me to the passage that contradicts me, if there is one.
3. When all the children were considered as one group, regardless of age, there were no effects on child outcomes, but there WERE significant effects on younger children (age 2-4), compared with children 5 to 8.5 and children 8.5 to 17. The behaviors of the younger children of group participants were reported to be – and observed to be — more self- regulated, less internalizing (anxious, depressed, withdrawn), and less externalizing (aggressive, non-cooperative, hyperactive). It seems reasonable to us that a 16 week intervention for parents might not be sufficient to reduce negative behavior in older children.
On the younger children, I discounted that because the report said (p. 42): “While the findings for the youngest children are promising, there is some uncertainty because the pattern of results is not strong enough to remain statistically significant once adjustments are made to account for the number of outcomes examined.”
4. For every positive outcome we have cited, you or any critic can find another measure that shows that the intervention had no effect. That’s part of our point here. Rather than yes or no, what we have is a complicated series of findings that lead to a complicated series of decisions about how best to be helpful to families.
That’s just not an accurate description. There are many null findings for each positive finding, and the positive findings themselves are either small, trivially small, or not statistically reliable.
4. Several times you suggest that giving couples the $9,000 per family (the program costs) would do better. Do you have evidence that giving families money increases, or at least maintains, family relationship quality? Is $9,000 a lot? Compared to what? According to the Associated Press, New York city’s annual cost per jail inmate was $167,731 last year. In other words, we are already spending billions to serve families when things go wrong, and some of the small effects of the marital could be thought of as preventive – especially at earlier stages of children’s development.
At the end of your blog, you rightly suggest a study in which giving families money is pitted in a random trial against relationship interventions. That’s a good idea, but that suggests more research. Furthermore, why must we always discuss programs in terms of yes or no, good or bad? What if we gave families $9,000 AND provided help with their relationships – and tested for the effects of a combined relationship and cash assistance.
We have lots of evidence that richer couples are less likely to divorce, of course. I don’t know that giving someone $9,000 would help with relationship quality, but I’m guessing it would at least help pay the rent or pay for some daycare.
It’s important to acknowledge that we’re not talking about research. The marriage promotion program is coming out of the welfare budget, not NIH or NSF. This study is a small part of it. Hundreds of millions of dollars have been spent on this, of which the studies account for a small amount. If this boondoggle continues, and they continue to study it, then they should include the cash-control group.
5. It seems to us that as a social scientist, you would want to ask “what have we learned about helping families from this study and from other research on couple relationship education?” We would suggest that we’ve learned that the earlier Building Strong Families program for unmarried low-income families had low attendance and no positive effects. A closer reading of those reports suggest that many of the unmarried partners were not in long-term relationships and were not doing very well at the outset. Perhaps it was a long-shot to offer some of them relationship help. We’ve also learned that the Strengthening Healthy Marriage program for married low-income families had some small but lasting effects on both self-reported and observed measures of their relationship quality (we think that the researchers learned something from the earlier study). And, notably, we’ve learned that there seemed to be some benefits for younger children when their parents took advantage of relationship strengthening behaviors.
We always learn something. See my comments above for why this is a stretch. I would be happy to see, and even pay for, research on what helps poor families. We already do some of that, through scientific agencies. My objection is not to the research, but to the program that it is studying, which takes money away from things we know are good.
Here is their last word — as good a defense as any for this program.
We know from many correlational studies that when parents are involved in unresolvable high level conflict, or are cold and withdrawn from each other, parenting is likely to be less effective, and their children fare less well in their cognitive, emotional, and social development. It was not some wild government idea that improving couple relationships could have benefits for children. Evidence in many studies and meta-analyses of studies of couple relationship interventions in middle-class families, and more recently for low-income families, have also been shown to produce benefits for the couples themselves — and for their kids. This was not a government program to force marriage on poor families. The participants were already married. It was a program that offered free help because maintaining good relationships is hard for couples at any level, but low-income folks have fewer financial resources to get all kinds of help that every family needs.
We are not suggesting that strengthening family relationships alone is a magic bullet for improving the lot of poor families. But, in our experience over the past many years, it gives the parents some tools for building more productive couple and parent-child relationships, which gives both the parents and their children more confidence and hope.
What we need to learn is how to do family relationship strengthening more effectively, and how to combine that activity with other approaches, now being tried in isolated silos of government, foundations, and private agencies, in order to make life better for parents and their kids.
In our view, trumpeting the failure of Supporting Healthy Marriage by focusing on a few of the negative findings doesn’t help move us toward that goal.
New results are in from the unrelenting efforts to redirect welfare spending to marriage promotion. By my unsophisticated calculations we’re more than $1 billion into this program, without a single, proven healthy marriage yet to show for it.
The latest report is a study of the Supporting Healthy Marriage program, in which half of 6,298 couples were offered an extensive relationship support and education program. Short version: Fail.
Supporting Healthy Marriage is a federal program called “the first large-scale, multisite, multiyear, rigorous test of marriage education programs for low-income married couples.” The program evaluation used eight locations, with married, low- or modest-income parents (or expectant couples) offered a year-long program. Those in the program group had a four- to five-month series of workshops, followed by educational and social events to reinforce the curriculum.
Longer than most marriage education services and based on structured curricula shown to be effective with middle-income couples, the workshops were designed to help couples enhance the quality of their relationships by teaching strategies for managing conflict, communicating effectively, increasing supportive behaviors, and building closeness and friendship. Workshops also wove in strategies for managing stressful circumstances commonly faced by lower-income families (such as job loss, financial stress, or housing instability), and they encouraged couples to build positive support networks in their communities.
This was a good program, with a good quality evaluation. To avoid selection biases, for example, the study included those who did not participate despite being offered the program. But participation rates were good:
According to program information data, on average, 83% of program group couples attended at least one workshop; 66% attended at least one supplemental activity; and 88% attended at least one meeting with their family support workers. Overall, program group couples participated in an average of 27 hours of services across the three components, including an average of 17 hours of curricula, nearly 6 hours of supplemental activities, and 4 hours of in-person family support meetings.
The couples had been together an average of 6 years; 82% had incomes below twice the poverty level. More than half thought their marriage was in trouble when they started.
But the treatment and control groups followed the exact same trajectory. At 12 months, 90% of both groups were still married or in a committed relationship, after 30 months it was 81.5% for both groups.
The study team also broke down the very diverse population, but could not find a race/ethnic or income group that showed noteworthy different results. A complete failure.
But wait. There were some “small but sustained” improvements in subjectively-measured psychological indicators. How small? For relationship quality, the effect of the program was .13 standard deviations, equivalent to moving 15% of the couples one point on a 7-point scale from “completely unhappy” to “completely happy.” So that’s something. Further, after 30 months, 43% of the program couples thought their marriage was “in trouble” (according to either partner) compared with 47% of the control group. That was an effect size of .09 standard deviations. So that’s something, too. Many other indicators showed no effect.
However, I discount even these small effects since it seems plausible that program participants just learned to say better things about their marriages. Without something beyond a purely subjective report — for example, domestic violence reports or kids’ test scores — I wouldn’t be convinced even if these results weren’t so weak.
What did this cost? Round numbers: $9,100 per couple, not including evaluation or start-up costs. That would be $29 million for half the 6,298 couples. The program staff and evaluators should have thanked the poor families that involuntarily gave up that money from the welfare budget in the service of the marriage-promotion agenda. We know that cash would have come in handy – so thanks, welfare!
The mild-mannered researchers, realizing (one can only hope) that their work on this boondoggle is coming to an end, conclude:
It is worthwhile considering whether this amount of money could be spent in ways that bring about more substantial effects on families and children.
For example, giving the poor couples $9,000.
Trail of program evaluation tears
We have seen results this bad before. The Building Strong Families (BSF) program, also thoroughly evaluated, was a complete bust as well:
Some of the people trying to bolster these programs — researchers, it must be said, who are supported by the programs — have produced almost comically bad research, such as this disaster of an analysis I reported on earlier.
Now it’s time to prepare ourselves for the rebuttals of the marriage promoters, who are by now quite used to responding to this kind of news.
- We shouldn’t expect government programs to work. Just look at Head Start. Of course, lots of programs fail. And, specifically, some large studies have failed to show that kids whose parents were offered Head Start programs do better than those whose parents were not. But Head Start is offering a service to parents who want it, that most of them would buy on their own if it were not offered. Head Start might fail at lifting children out of poverty while succeeding at providing a valuable, need-based service to low-income families.
- Rich people get marriage counseling, so why shouldn’t poor people? As you can imagine, I am all for giving poor people all the free goods and services they can carry. Just make it totally voluntary, don’t do it to change their behavior to fit your moral standards, and don’t pay for it by taking cash out of the pockets of single-parent families. I really am all in favor of marriage counseling for people who want it, but this is not the policy platform to get that done.
- These small subjectively-measured benefits are actually very important, and were really the point anyway. No, the point was to promote marriage, from the welfare law itself (described here) to the Healthy Marriage Initiative. If the point was to make poor people happier Congress never would have gone for it.
- We have to keep trying. We need more programs and more research. If you want to promote marriage, here’s a research plan: have a third group in the study — in addition to the program and control group — who get cash equivalent to the cost of the service. See how well the cash group does, because that’s the outcome you need to surpass to prove this policy a success.
Everyone loves marriage these days. But a lot of people like to think of promoting marriage as a way to reduce poverty, and with that they believe poor people are that way because they’re not married. That’s mostly backwards.
I don’t know how I missed this one, from two Valentine’s Days ago…
For an introductory methods course discussion on: when does something cause something else. Question: Are happier couples happier? Some writers think so:
- Washington Post: The Groupon effect: Can date-night deals save your marriage?
- Huffington Post: Date Nights: They Make Your Marriage Work
- Inquisitr: Date Nights Improve Marriage, Better Your Sex Life [Study]
- USA Today: ‘Date night’ can improve marriage, sexual satisfaction
- Cosmopolitan: Why Date Night Is So Important
I can see the study design now: a randomized group of couples were given coupons for date nights, and some time later were compared with a control group without the coupons. Or not. Cosmo summarized:
For their study, researchers from the University of Virginia’s National Marriage Project surveyed 1,600 couples and asked them about everything from relationship satisfaction to sex. They discovered that couples who spend at least one night a week alone together say they’re more committed to their relationship than those who don’t hang out together as much.
Is that it? A simple association between being together and being happy? Almost. First, they say (there are no tables) that they “control for factors such as income, age, education, race, and ethnicity.” Such as? Anyway.
Second, they also claim to have analyzed historical data from the National Survey of Families and Households (1987-1994). They write:
Because we had data from spouses at two time points in the NSFH, we were also able to examine the direction of effects—to determine whether or not couple time reported during the first wave of the survey was associated with marital quality at the second wave. Here, the more couple time individuals reported at the time of the first survey, the more likely they were to be very happy in their marriage at the second survey, five years later. Although the NSFH evidence does not provide us with definitive proof that couple time causes increases in marital quality, the longitudinal character of the data suggests that the relationship may indeed be causal.
So, Wilcox and Dew point #1: If something happened before something else, “the relationship may indeed be causal.” They go on:
It is certainly intuitively true that greater satisfaction with one’s partner should also lead to more time spent in positive, shared activities. Nevertheless, it would be absurd to assume that two partners who intentionally set out to increase positive couple time spent together would typically not benefit from such time with increases in connection and happiness.
So, point #2 is, We already knew the answer before we did the research, because it’s flipping obvious, so who cares about this analysis — it’s almost Valentine’s Day!
There are ways to actually get at “the direction of effects,” like the randomized trial I suggested, or even using longitudinal data and assessing changes in happiness, or controlling for happiness at time 1. Not this.
Anyway, can we think of examples of things that occur before other things without causing them? Here are a few off the top of my head:
- One sibling dies of a genetic disease now, and then the other one dies from the same disease later: Shocking new evidence that genetics works sideways!
- Someone has tennis elbow now, and is playing sports later: The surprising way that getting hurt makes you athletic!
- People who spend more money now have more money later: The more you spend, the more you save!
- And of course, people who have a lot of sex now are good looking later: Sex up your looks!
I’m open to suggestions for better examples.
Note: I guess in some social science neighborhoods it’s common to analyze the effects of extremely similar things on each other, like pleasure being associated with happiness, or strong left arms being associated with strong right legs. Dew and Wilcox actually published a peer-reviewed article, using this survey, on the association between small acts of kindness in marriage and marital satisfaction. And the result? Couples who are nice to each other are happier.
My paper on divorce and the recession has been accepted by the journal Population Research and Policy Review, and Emily Alpert Reyes wrote it up for the L.A. Times today. The paper is online in the Maryland Population Research Center working paper collection.
Married couples promise to stick together for better or worse. But as the economy started to rebound, so did the divorce rate.
Divorces plunged when the recession struck and slowly started to rise as the recovery began, according to a study to be published in Population Research and Policy Review.
From 2009 to 2011, about 150,000 fewer divorces occurred than would otherwise have been expected, University of Maryland sociologist Philip N. Cohen estimated. Across the country, the divorce rate among married women dropped from 2.09% to 1.95% from 2008 to 2009, then crept back up to 1.98% in both 2010 and 2011.
To reach the figure of 150,000 fewer divorces, I estimated a model of divorce odds based on 2008 data (the first year the American Community Survey asked about divorce events). Based on age, education, marital duration, number of times married, race/ethnicity and nativity, I predicted how many divorces there would have been in the subsequent years if only the population composition changed. Then I compared that predicted trend with what the survey actually observed. This comparison showed about 150,000 fewer than expected over the years 2009-2011:
Notice that the divorce rate was expected to decline based only on changes in the population, such as increasing education and age. That means you can’t simply attribute any drop in divorce to the recession — the question is whether the pace of decline changed.
Further, the interpretation that this pattern was driven by the recession is tempered by my analysis of state variations, which showed that states’ unemployment rates were not statistically associated with the odds of divorce when individual factors were controlled. Foreclosure rates were associated with higher divorce rates, but this didn’t hold up with state fixed effects.
So I’m cautious about the attributing the trend to the recession. Unfortunately, this all happened after only one year of ACS divorce data collection, which introduced a totally different method of measuring divorce rates, which is basically not comparable to the divorce statistics compiled by the National Center for Health Statistics from state-reported divorce decrees.
Finally, in a supplemental analysis, I tested whether unemployment and foreclosures were associated with divorce odds differently according to education level. This showed unemployment increasing the education gap in divorce, and foreclosures decreasing it:
Because I didn’t have data on the individuals’ unemployment or foreclosure experience, I didn’t read too much into it, but left it in the paper to spur further research.
Aside: This took me a few years.
It started when I felt compelled to debunk Brad Wilcox’s fatuous and deliberately misleading interpretation of divorce trends — silver lining! – at the start of the recession, which he followed up with an even worse piece of conservative-foundation bait. Unburdened by the desire to know the facts, and the burdens of peer review, he wrote in 2009:
judging by divorce trends, many couples appear to be developing a new appreciation for the economic and social support that marriage can provide in tough times. Thus, one piece of good news emerging from the last two years is that marital stability is up.
That was my introduction to his unique brand of incompetence (he was wrong) and dishonesty (note use of “Thus,” to imply a causal connection where none has been demonstrated), which revealed itself most egregiously during the Regenerus affair (the full catalog is under this tag). Still, people publish his un-reviewed nonsense, and the American Enterprise Institute has named him a visiting scholar. If they know this record, they are unscrupulous; if they don’t, they are oblivious. I keep mentioning it to help differentiate those two mechanisms.
Yesterday I wondered about the treatment of race in the blockbuster Chetty et al. paper on economic mobility trends and variation. Today, graphics and representation.
If you read Brad Wilcox’s triumphalist Slate post, “Family Matters” (as if he needed “an important new Harvard study” to write that), you saw this figure:
David Leonhardt tweeted that figure as “A reminder, via [Wilcox], of how important marriage is for social mobility.” But what does the figure show? Neither said anything more than what is printed on the figure. Of course, the figure is not the analysis. But it is what a lot of people remember about the analysis.
But the analysis on which it is based uses 741 commuting zones (metropolitan or rural areas defined by commuting patterns). So what are those 20 dots lying so perfectly along that line? In fact, that correlation printed on the graph, -.764, is much weaker than what you see plotted on the graph. The relationship you’re looking at is -.93! (thanks Bill Bielby for pointing that out).
In the paper, which presumably few of the people tweeting about it read, the authors explain that these figures are “binned scatter plots.” They broke the commuting zones into equally-sized groups and plotted the means of the x and y variables. They say they did percentiles, which would be 100 dots, but this one only has 20 dots, so let’s call them vigintiles.
In the process of analysis, this might be a reasonable way to eyeball a relationship and look for nonlinearities. But for presentation it’s wrong wrong wrong.* The dots compress the variation, and the line compresses it more. The dots give the misleading impression that you’re displaying the variance around the line. What, are you trying save ink?
Since the data are available, we can look at this for realz. Here is the relationship with all the points, showing a much messier relationship, the actual -.76 (the range of the Chetty et al. figure, which was compressed by the binning, is shown by the blue box):
That’s 709 dots — one for each of the commuting zones for which they had sufficient data. With today’s powerful computers and high resolution screens, there is no excuse for reducing this down to 20 dots for display purposes.
But wait, there’s more. What about population differences? In the 2000 Census, these 709 commuting zones ranged in population in the 2000 Census from 5,000 (Southwest Jackson, Utah) to 16,000,000 (Los Angeles). Do you want to count Southwest Jackson as much as Los Angeles in your analysis of the relationship between these variables? Chetty et al. do in their figure. But if you weight them by population size, so each person in the population contributes equally to the relationship, that correlation that was -.76 — which they displayed as -.93 — is reduced to -.61. Yikes.
Here is what the plot looks like if you scale the commuting zones according to population size (more or less, not quite sure how Stata does this):
Now it’s messier, and the slope is much less steep. And you can see that gargantuan outlier — which turns out to be the New York commuting zone, which has 12 million people and with a lot more upward mobility than you would expect based on its family structure composition.
Finally, while we’re at it, we may as well attend to that nonlinearity that has been apparent since the opening figure. We can increase the variance explained from .38 to .42 by adding a quadratic term, to get this:
I hate to go beyond what the data can really tell. But — what the heck — it does appear that after 33% single-mother families, the effect hits its minimum and turns positive. These single mother figures are pretty old (when Chetty et al.’s sample were kids). Now that the country has surpassed 40% unmarried births, I think it’s safe to say we’re out of the woods. But that’s just speculation.**
*OK, OK: “wrong wrong wrong” is going too far. Absolute rules in data visualization are often wrong wrong wrong. Binning 709 groups down to 20 is extreme. Sometimes you have a zillion points. Sometimes the plot obscures the pattern. Sometimes binning is an inherent part of measurement (we usually measure age in years, for example, not seconds). None of that is an excuse in this case. However, Carter Butts sent along an example that makes the point well:
On the other hand, the Chetty et al. case is more similar to the following extreme example:
If you were interested in the relationship between age and earnings for a sample of 1,400 full-time, year-round women, you might start with this, which is a little frustrating:
The linear relationship is hard to see, but it’s about +$500 per year of age. However, the correlation is only .13, and the variance explained by linear-age alone is only 1.7%. But if you plotted the mean wage over ages, the correlation jumps to .68:
That’s a different question. It’s not, “how does age affect earnings,” it’s, “how does age affect mean earnings.” And if you binned the women into 10-year age intervals (25-34, 35-44, 45-54), and plotted the mean wage for each group, the correlation is .86.
Chetty et al. didn’t report the final correlation, but they showed it, even adding the regression line, so that Wilcox could call it the “bivariate relationship.”
**This paragraph was a joke that several people missed, so I’m clarifying. I would never draw a conclusion like that from the scraggly tale of a loose correlation like this.
What does race have to do with mobility? The words “race,” “black,” or “African American” don’t appear in David Leonhardt’s report on the new Chetty et al. paper on intergenerational mobility that hit the news yesterday. Or in Jim Tankersley’s report in the Washington Post, which is amazing, because it included this figure: That’s not exactly a map of Black America, which the Census Bureau has produced, but it’s not that far off:
But even if you don’t look at the map, what if you read the paper? Describing the series of maps of intergenerational mobility, the authors write:
Perhaps the most obvious pattern from the maps in Figure VI is that intergenerational mobility is lower in areas with larger African-American populations, such as the Southeast. … Figure IXa confirms that areas with larger African-American populations do in fact have substantially lower rates of upward mobility. The correlation between upward mobility and fraction black is -0.585. In areas that have small black populations, children born to parents at the 25th percentile can expect to reach the median of the national income distribution on average (y25;c = 50); in areas with
large African-American populations, y25;c is only 35.
Here is that Figure IXa, which plots Black population composition and mobility levels for groups of commuting zones: Yes, race is an important part of the story. In a nice part of the paper, the authors test whether Black population size is related to upward mobility for Whites (or, people in zip codes that are probably White, since race isn’t in their tax records), and find that it is. It’s not just Blacks driving the effect. I’m thinking about the historical patterns of industrial development, land ownership, the backwardness of racist elites in the South, and so on. But they’re not. For some reason, not explained at all, Chetty et al. offer this pivot:
The main lesson of the analysis in this section is that both blacks and whites living in areas with large African-American populations have lower rates of upward income mobility. One potential mechanism for this pattern is the historical legacy of greater segregation in areas with more blacks. Such segregation could potentially affect both low-income whites and blacks, as racial segregation is often associated with income segregation. We turn to the relationship between segregation and upward mobility in the next section.
And that’s it, they don’t discuss Black population size again, instead only focusing on racial segregation. They don’t pursue this “potential mechanism” in the analysis that follows. Instead, they drop percent Black for racial segregation. I have no idea why, especially considering this Table VII, which shows unadjusted (and normalized) correlations (more or less) between each variable and absolute upward mobility (the variable mapped above):
In these normalized correlations, fraction Black has a stronger relationship to mobility than racial segregation or economic segregation! In fact, it’s just about the strongest relationship on the whole long table (except for single mothers, with which it is of course highly correlated). So why do they not use it in their main models? Maybe someone else can explain this to me. (Full disclosure, my whole dissertation was about this variable.)
This is especially unfortunate because they do an analysis of the association between commuting zone family structure (using macro-level variables) and individual-level mobility, controlling for marital status — but not race — at the individual level. From this they conclude, “Children of married parents also have higher rates of upward mobility if they live in communities with fewer single parents.” I am quite suspicious that this effect is inflated by the omission of race at either level. So they write the following, which goes way beyond what they can find in the data:
Hence, family structure correlates with upward mobility not just at the individual level but also at the community level, perhaps because the stability of the social environment affects children’s outcomes more broadly.
Or maybe, race.
I explored the percent Black versus single mother question in a post a few weeks ago using the Chetty et al. data. I did two very simple OLS regression models using only the 100 largest commuting zones, weighted for population size, the first with just single motherhood, and then a model with proportion Black added: This shows that the association between single motherhood rates and immobility is reduced by two-thirds, and is no longer significant at conventional levels, when percent Black is added to the model. That is: Percent Black statistically explains the relationship between single motherhood and intergenerational immobility across U.S. labor markets. That’s not an analysis, it’s just an argument for keeping percent Black in the more complex models. Substantively, the level of racial segregation is just one part of the complex race story – it measures one kind of inequality in a local area, but not the amount of Black, which matters a lot (I won’t go into it all, but here are three old papers: one, two, three.
The burgeoning elite conversation about economic mobility, poverty, and inequality is good news. It’s avoidance of race is not.