A colleague writes:
Thought you might be interested in this front page data journalism take down of an article. I don’t know the article but this amounts to a journalist talking with someone who didn’t like the piece and ripping it based on a measurement detail. How bad though is this measurement detail? Are you better off having data on parental effort for 2 days on 30 parents or for 2 parents on 30 days? It’s not a cluster question since outcomes are at the parent level not the day level.
The New York Times article in question, by Justin Wolfers, says:
The latest salvo in the mommy wars is that all that time you spend parenting just doesn’t matter. But it’s a claim that, despite the enthusiastic and widespread coverage by news media . . . does not hold water. . . .
The claim that parenting time doesn’t matter . . . largely reflects the failure of the authors to accurately measure parental input. In particular, the study does not measure how much time parents typically spend with their children. Instead, it measures how much time each parent spends with children on only two particular days — one a weekday and the other a weekend day.
The result is that whether you are categorized as an intensive or a distant parent depends largely on which days of the week you happened to be surveyed. . . . Trying to get a sense of the time you spend parenting from a single day’s diary is a bit like trying to measure your income from a single day. If yesterday was payday, you look rich, but if it’s not, you would be reported as dead broke. You get a clearer picture only by looking at your income — or your parenting time — over a more meaningful period.
I took a quick look at the paper [an article by Melissa Milkie, Kei Nomaguchi, and Kathleen Denny, in the Journal of Marriage and Family] and I’m concerned that they controlled for “mother’s work hours.” This doesn’t seem quite right, because if the mother works fewer hours she can have more time to spend with kid, right? Anyway, the news article did seem a bit breathless.
My colleague replied:
I spent time puzzling over this today (without reading the original article yet). On the one hand it seems a little odd—like it’s sort of controlling for the treatment. But if (perhaps a big if) you think that working is pretreatment with respect to caring for children then you probably do want to control for this. It’s like staying at home is the encouragement, spending time with kids is a treatment, and the outcome is the outcome. Assume away selection issues and say that there was random assignment to staying at home and also random assignment to spending time with kids conditional on staying at home — ie with different propensities for those that stay at home or not. Then you would want to take the average of the spending time effect in each of the stay at home categories. You might worry that staing at home is hugely predictive of spending times and so when you condition on staying at home you have no variation in spending time with kids to use to estimate an effect. But that’s another way of saying that you cannot tell if the effect is due to staying at home or to spending time with kids. If there is an independent effect of staying home with kids it should be observed in both groups. If it is different in both groups then a linear control is not correct (given the probably different assignment propensities to each group); but that just means you have to be careful how you control not that you shouldn’t control.
As I’ve said before, one of the benefits of blogging is that it’s acceptable to retain some level of uncertainty. In a scientific paper or, in a different way, in a news article, there’s pressure to make a strong conclusion. Here we can discuss and leave some questions open.
The paper link does not work for me—here is an ungated pdf: http://www.soc.iastate.edu/sapp/soc302Afamily1.pdf
and here is abstract: http://onlinelibrary.wiley.com/doi/10.1111/jomf.12170/abstract
“I don’t know the article but this amounts to a journalist talking with someone who didn’t like the piece and ripping it based on a measurement detail.”
Wolfers is an economics professor, not just some journalist.
“Are you better off having data on parental effort for 2 days on 30 parents or for 2 parents on 30 days?”
This seems to be making an excuse for the researchers that maybe they used their resources as well as they could. I don’t think that’s what’s at issue. The question is whether their work provides strong evidence for the conclusions drawn from it. Particularly given that it was a null finding, a noisy measurement of treatment would suggest that the work provides weak evidence (even if it’s the best evidence the researchers could produce on their budget).
On the evidence: it seems this is not really a story of small N (though it is that as well), but of biased effect estimates. If they have random error in the classification of treatment/non-treatment, the estimated effects will be steered towards zero (if there is an effect to talk about).
This is not an experimental study, there is no manipulated treatment and like most social science is making weak or no causal claims, which is always the best you can do. With 2300 parents and and careful, well designed national sample (Panel Study of Income Dynamics) and a measure that makes sense, this is a pretty well designed study. Of course you can criticize the work but the idea that you are choosing between 2 parents to study and 30 parents is ridiculous and really kind of insulting to the authors. Somehow the idea that “what you did yesterday” would be a worse measure than “what you did last year” is ridiculous; sure would you like diary data every day for a year? Why not, but the value added over taking a sample of days is not likely to be much. It’s not what you did on that one day that makes a difference, whether that day was measured with error or not (the idea of not is strange obviously), it’s that the one day is a sample of all of the days because it is all of the days of a child’s life that potentially this matters for, not any one day. I also like the way the controlled for father’s time, given that the article was about the impact of mother’s available and engaged time not total parental time. It’s funny that Wolfers says “you” when he really means “you women” throughout.
Also, re the Disneyland story in the Wolfers article, they actually have a control for the subjective rating of how typical the recorded day was, so presumably the Disney day would be recorded as atypical.
The article link is broken so I don’t know exactly what they said, but the NYT post sounds plausible:
This non-finding is just an artifact of their study design
(“We didn’t measure X very well, so we just can’t tell whether X predicts Y”)
and not necessarily a feature of the real world
(“We’re sure that X has no relationship with Y”).
What you’ve said about the original article’s approach (measure time spent with kids on just 2 days per parent) would be perfectly fine if they only wanted to estimate the average amount of time-spent-with-kids, over all parents.
But if they wanted to estimate the *relationship* between average-time-spent-with-kids and a child-success outcome, they would need to estimate each parent’s time-spent-with-kids much more precisely. Here, it sounds like they basically regressed the child-success outcome on a predictor with tons of measurement error. It’s no wonder they found no significant relationship.
Without the article itself, I don’t know if the authors actually made this claim (“X is unrelated to Y!”), or if they realized there’s a measurement error and it was just journalists who misread the article.
“Does your time as a parent make a difference?” Nope, It does not. Moreover, spending time with your children could be detrimental to their well-being. Actually, the further you are from them the better.
Joking aside, this is really, really bad science (Vhttps://www.youtube.com/watch?v=IaO69CF5mbY ). I do not think that the authors checked their claims severely enough before they sent them out for review, and I do not think the reviewers cared. As long as there is a controversial claim that will elicit some response, publish it, especially in sociology. I hope that no parent takes seriously the idiotic claim that parents who keep away from their children are more likely to have happier and more successful offspring.
The article says nothing of the sort. I would really suggest reading the article before discussing what claims it makes or does not make.
Of course it does. Read the last paragraph. It argues for mothers to “ease up” and for social services and financial resources to help with upbringing, because parenting time is not important for children’s well-being. Of course the authors use less loaded way to describe it but the message remains.
The article says nothing like “keep away from” your children. It absolutely does not say that if you “keep away” from your children they will be better off. It says that spending it makes no difference how much time you spend with children once you control for all of the most important predictors e.g. income and education of the parents. And it actually says that in adolescence time can be positive. It says that mothers do not have to feel that by sending their child to preschool or by letting them participate in sports or the school newspaper after school they are going to have a more unhappy, more likely to be criminally involved child. It’s also okay to go to a PTA meeting if dad is okay with watching the kids, you don’t have to have dad do the PTA because of that extra special mother effect.
Perhaps they did overstate the claim, but it would not be inappropriate to claim that there is a lack of evidence for the opposite claim of a strong relationship between Time and Outcome which is being made with equal strength. This is one of those studies that makes me shake my head and say — how would anyone think that Time in and of itself — independent of quality, of learning, of nurturing, etc., etc., etc. — would be strongly related to an outcome such as this. These types of hypotheses say a lot about the social science literature — but almost nothing positive about it.
George,
I assume you were being ironic when you said, ‘Joking aside’. Claiming an absence of an effect is not equivalent to claiming an effect in the opposite direction.
How do you think parents would act as a consequence of accepting the article’ s claims? Spending more time with their children,the same amount, or less?
I don’t think either is proposed as an intervention.
Your inferences are going well beyond the data and beyond the author’s claims.
To answer directly. If they understood the article, they would conclude that Time — independent of potential interactive effects — is unrelated to the outcome. Thus, one cannot logically argue for any of the three options you present given the evidence.
To clarify: Given the evidence presented in the article.
The article’s one speculation at the end is that mother’s of young children do not need to exhaust themselves with spending every waking hour actively engaged with their child, which is contrary to lots of parenting advice that many middle class mothers of children get. So actually if someone was for some reason going to use a regression analysis as the basis for their parenting, the one thing might be that they would decide that it is okay not to home school. Also possibly that it is okay for your child to attend pre-K or K.
Point taken.
James,
The authors argue that mothers spent less time with their children (from the discussion – last paragraph): ” it suggests that mothers ease up on practicing more intensive mothering during childhood, especially given that it may end up exhausting them.”
I did not make any inferences. I wrote what is a parent to do with the evidence of the paper. Nothing is not a choice.
The authors also suggest to invest on economic resources and social services, since they (authors) see no value on investing on parenting time. Who’s inferences are going beyond the data??
They do not say that time on parenting is not important, they say “intensive mothering” — meaning NOT the kind of mothering that says “go out and play, come back for lunch” but the practice of mothering that says never let your child be on their own — is neither beneficial nor harmful to the child and if it is exhausting the mother to practice she can stop being exhausted and let the kid play on his or her own or be with a grand parent for that matter. And yes absolutely they are saying that investing in programs for kids, like high quality pre k and after school programs, particularly those that are focused on kids with the highest risks of problems, those whose parents did not finish high school or have only a high school degree and mothers who are in emotional distress, would be focusing on the variables that are most predictive of future problems . This is hardly earth shattering news.
“Questions as to how the amount of time mothers spend with their children matters for their off- spring are fraught with tension. As part of polit- ical ideology and the Mommy Wars, how much time mothers spend with children and should spend with them is hotly contested terrain. The ideology of intensive mothering, which has been prominent in U.S. culture over recent decades, underscores the idea that mothers are unique and their time irreplaceable for children. ” I would say it’s more questionable that the “ideology of intensive mothering” has been central for any group besides upper middle class women (like those who are anti-vaxxers) and the fundamentalist home schooler world. Everyone else lets their 3-11 year olds (the age groups studied) go to school for most of the day and then the variation they are measuring is mostly about how much time is spent in after school activities, i.e. between 3 and bedtime and breakfast.
Also they point out “there is an important case to consider, however, in our non-findings. We cannot measure the possibility that mothers who spend more time with children could be negatively selected into being with them more often. For example, unobserved factors of mothers may keep them from engaging in work or social activities and thus make them more available for spending time with their children; these same factors may exert a less-than-positive influence on offspring (e.g., Chase-Lansdale et al., 2003; McLoyd, Jayaratne, Ceballo, & Borquez, 1994), which may be driving the observed null results.” This is not that different than the point about it being helpful to focus on the needs of children of low education moms and dads and also on the needs of children whose mothers are in distress. If someone is going to insist on reading some kind of policy implications from this, I would say that they say that if you have low education and/or depressed mothers telling them to spend more time with their children is not going to have a beneficial effect.
Elin,
How do you think parents shoild act as a consequence of accepting the article’ s claims? Spending more time with their children,the same amount, or less? If none of the above then the title of the article is misleading. It’s central focus is not about parenting time and children outcomes, but about something else.
Note, I think various people are using “the article” to mean different things: the scientific article vs the popular version. Elin seems to be focused more on the content of the scientific article, and many others are focused more on the popular journalism. It seems like there’s a “talking past” each other going on here.
If you are talking about the article published in Marriage and Family, no, it’s not an article (nor would they publish an article) that is intended to guide change individual level behavior — it does not do anything like look at what happens if you deliberately change individual level behavior. It only looks at humans in their more or less natural state except that they are completing time diaries. The title of the article is “Does the Amount of Time Mothers Spend With Children or Adolescents Matter?” and the answer they give is “no” so there is absolutely no advice to change anyone’s behavior. the range of reported available hours goes up to 77 in the child sample. They are saying at the end, if you are exhausted or overwhelmed and you are practicing the kind of mothering that is up at that end of the spectrum, it’s probably not going to hurt your kids for you to cut back to 70 hours. But they don’t make a definitive claim there because obviously they haven’t done (and no one ethically could do) an experiment on that, but they do point out that there is some question about whether those 70+ hour a week mothers are distorting the results because they are selected out of activities involving contact with adults. The whole last section of the articles is caveats and possible implications and areas for future research, as they say
“Thus, this is not the definitive
study that trumpets “Mothers Do Not Matter for
Children”: It is clear that mothers’ practices matter
in myriad ways. But given the findings here,
it is incumbent upon other researchers to show
how and why the amount of mother time does
matter for children.”
As to what advice a mother of non adolescent children would take from this article, I think it is “do what makes sense for you and your family, on average across a representative sample it didn’t make a difference, so if you want to do more great, if you want to do less that’s fine, stay the same, that’s fine as well.” One the other hand, if you have an adolescent you might want to pay attention to how much actual engaged time you have, an hour a day is about average nationally. If I were writing a parenting article I’d focus on that as the take away.
George, by the way the article most clearly does not address “parents” it only addresses mothers. Specifically the variable they are using is time spent with the child not in the presence of the father. So this article, if you really want to use it for parenting advice, could be taken to mean it’s okay that some of mothers time engaged with child happens when the father is also engaged with the child. Like if you are all playing a game together. They don’t seem to have really dealt with the paternal time with child at all since they are interesting in taking on the intensive mothering concept, but part of what this says is that it’s fine for dad to be playing with the kids or helping with homework, the idea that maternal time is especially and uniquely effective is not supported.
George,
You are making inferences beyond the data and you are mischaracterizing the arguments presented by the authors. What they can logically claim is that the evidence reviewed in the paper does not support the arguments made by those who claim ‘Time’ in and of itself is the deciding factor on the outcome. From that conclusion, the authors go on to discuss policy issues implicated by the unwarranted claims of those who have argued the opposite. They then go on to give their ‘opinion’ in the discussion section of the paper based on a hypothesized belief that the proscribed ‘intensive mothering’ practices may have detrimental effects on the mother and that their review of the literature and their own data provides no evidence that ‘intensive mothering’ the claimed positive effect.
So .. “common sense” would say “mother’s work hours” would be inversely proportional to time spent with kids. But reality is that actually, lots of moms and dads both do schedule shifting of their work so that they can be available at times when kids are awake and at home and in two parent families shift around so that, for example, one parent is available for breakfast, while another parent is available for after school or bed time. So we see lots of female cleaners who work the overnight. One of the women who comes on duty late afternoon where I work sometimes sets her kids up in the student lounge. Figure a third of nurses work the overnight shift and the moms and dads among them get home in time to take their kids to school in the morning. Same for many other occupations. Myself, I schedule all of my teaching so that if I need to I can do pick up and drop off. And I can shift grading and writing until after bedtime. There are also tons of k-12 teachers who are motivated for their occupation by having a yearly schedule that matches that of their school age children. It’s important to remember that most people in the workforce do not have college degrees or any college for that matter.
what do you mean by “breathless”?
I think Wolfer’s point, and it’s a good one, is that this two day measurement does not have a high correlation with how much time is spent with the child over a longer period of time.
That’s an empirical measurement claim, and I don’t recall seeing this directly evaluated (either by Wolfers or anyone else connected with this study). I’d be happy to find out I’m wrong.
The more measurement error you have on the input (time spent, quality of time spent) and the output (how well did those kids turn out?) the worse your correlation is, just to state the obvious.
This is, of course, similar to other work that looks at exercise, diet, purchases etc over a short period of time and tries to make long term conclusions. I’ve generally found these tenuous if not contemporaneous. If I have a two week food diary, I would be prepared to make group comparisons, e.g. correlations between eating beets and income, ethnicity, age, etc. but any connection between those two weeks of beet eating and health 20 years later I’d not trust.
“All analyses were cross-sectional. We examined
longitudinal associations between maternal
time at W1 and developmental outcomes at
W2 but do not present results here. First,
there were no statistically significant associations
between time with mothers in childhood
(W1) and adolescent outcomes (W2). Second,
theoretically, maternal time should be
important for the concurrent experiences and
well-being of children. Existing research also
indicates a lack of long-reaching influence (e.g.,
Hsin, 2009).”
Really I find it is amusing that this topic is so fraught that no one is mentioning the use of p values or that they should have been using a multilevel model. But that’s the way it goes around here on some topics :).
I think one of the reasons to control for mothers’ time at work is that time is a limited resource. Even if the best interest of a child is the only driving factor, it is arguable that it is better for a child to grow up in a more affluent family (achieved by a working mother) than in a more caring one (achieved by stay-at-home mom). If we want to know precisely what parental involvement achieves as opposed to how much it stacks up against making more dough, the control make sense.
This thread would never have happened if this article had an appropriate title.
This is totally misleading: “Does the Amount of Time Mothers Spend With Children or Adolescents Matter?”
This is much better: “Does the Marginal Hour of Maternal Time Spent with Children in the United States Generate Measurable Effects on Several Dimensions of Human Development that the Authors Consider ‘Good'”
I agree with Andrew that the problem is mostly about the refusal of scientific discourse (in its current state) to embrace uncertainty. But in this case it is also about terrible generalizations from specific circumstances (a marginal hour of maternal time in the United States in the latter half of the 20th Century) to general claims about humanity (maternal time with children doesn’t matter).
But, if academics wrote such realistic titles, how would they ever get tenure? I suggest a study:
“The effect of the marginal minute of navel gazing on tenure decisions in top ranked research institutions in the latter half of the 2010’s segregated by gender identity and physical sex, and the effect of fertility cycle on the color of navel lint”
Also, how much data would you have to collect to shift the prior on whether “maternal time makes (some or no) difference?” How many babies abandoned at infancy wind up in the upper ranks of society? It really is obvious that starting from zero, the marginal effect of time spent is large, and starting from 100% of 24x7x365 the marginal effect of one less hour per year is small and we know this from the start. The only hope of meaningful information to come out of this kind of study is to somehow estimate a curve G(T) where G is goodness of outcome (which is highly fraught) and T is time spent per day, and even with good quality measurements of G and T anyone should expect such a curve to have a LOT of noise. Furthermore most parents are probably already putting T at a point where they perceive that the marginal effect is small, and this choice is going to be associated with all kinds of things like their socio-economic status, religious background, cultural background, family size, etc etc
This is really right except that they set the maximum number of hours as 112/week because they only care about child’s waking hours and they are looking at time change within a week .. the thing is that I think most people are thinking about children under 3. The mean age of the child sample was 7. These are children who are in school, almost all of them for full days. In the case of the adolescents we are talking about a mean of under 8 hours a week. And then you have people at 51 hours! With an American adolescent! And keep in mind that is “exclusive mother time” it doesn’t count times when two parents are together with the child. The variances are really big too.
I think there is quite a bit to criticize in this article, starting with the overstatement of the influence of the intensive mothering ideology on the population of families and on scholarly research. but people’s emotional reactions to the idea that there is not something magical about exclusive mother time is keeping that from happening.
I assume that the far end of the spectrum is dominated by home-schooling, so the real message may be “home schooling neither helps nor hurts school age children on average”.
My next project was going to be called “The Costs of Discrimination”, but now that I’m thinking about it I’m going to call it “Medical Expenditure Increases and Game Quality Decreases Caused by Adult Women Playing Soccer on Artificial Turf Instead of Grass: an IV Strategy Exploiting Differential Gender Valuations by FIFA”.
Oh wait – I got distracted didn’t I?
#USA