Re: Ask, And Ye Shall Receive.

1

It may be that the study you are referring to is flawed, but the critique to which you have linked is not devastating; indeed, it generates more heat than light because it mostly points to things that could skew the result, it does not establish that they did. In particular, it says lot that their first complaint is this:

Thus, one major problem with the NBER report is that it contradicts nearly an entire generation of robust and consistent findings that demonstrate the positive effects of quality care on young children. Further, the NBER’s conclusions are also in direct contradiction with those of world-class economists such as the Nobel Prize Winner for Economics, both of whom have spoken out strongly for the economic, labor and family benefits of child care programs.

In other words, the study must be flawed because it tells us something other than what we are used to hearing. Certainly, disagreement with prior studies could indicate a problem, but the authors of the critique cite the disagreement as a problem in and of itself.

horizontal rule
2

I like a lot of what she says, and she's much better than many of the women writing about these issues that I come across in major(ish) media, but the informality of her writing style on TAPPED wierds me out a bit. Particularly as it seems to stand in direct contradiction to the way she would have young women write.

This may be purely a problem of pseudo-ageism, though.

horizontal rule
3

I like her TAPPED posts a lot. Tim's just a pseudo-ageist.

horizontal rule
4

Make that crypto-agism and I think you're on to something.

horizontal rule
5

Er, uh, I mean "crypto-agist"

horizontal rule
6

Damn! foiled again -- OED says I meant "crypto-ageist". "agist" is an entirely separate term, an obsolescent verb having to do with animal husbandry.

horizontal rule
7

In other words, the study must be flawed because it tells us something other than what we are used to hearing.

Something other than what we're used to hearing from better designed studies that, for example, track children in day care separately from children not in day care. When a badly designed study tells you something different from a series of well designed studies, that's fairly persuasive.

horizontal rule
8

I like a lot of what she says, and she's much better than many of the women writing about these issues that I come across in major(ish) media, but the informality of her writing style on TAPPED wierds me out a bit. Particularly as it seems to stand in direct contradiction to the way she would have young women write.

This, I don't get. I saw her exhorting young women to work -- where what the exhortation to write stiffly?

horizontal rule
9

Ignore Tim, he's just a crypto-ageist who (secretly) agists cattle on the side.

horizontal rule
10

I haven't been reading her TAPPED posts, but on the basis of her December article, I have to say I think she's at least sometimes a bad writer. There just shouldn't have been as much confusion as to what the argument of a magazine article actually was, and checking in on the Lawyers, Guns, and Money comments, the comments here, and the 11-D people, it's not as if one or the other position as to the substance of her argument belongs to a small minority. Also, her editors really should have made sure it was clear before it went to press.

horizontal rule
11

Substantive criticisms of Hirshman go in the other thread. This one's for mocking SCMT.

horizontal rule
12

Pseudo-cattle, teofilo. Pseudo-cattle.

horizontal rule
13

Ageist b/c it's off-putting when I see a 60+ write something like, "Omigod." Pseudo b/c it wouldn't strike me as at all odd if mcmanus or Emerson did it (I don't know how old they are). The feeling is akin to the one I had when every woman I know started exhorting other women with, "You go girl!" If it's a part of your normal vocabulary (and, at this point, I'm willing to believe "You go girl" is a part of Emerson's natural vocabulary), you can tell; and when it's not, it's strange. Maybe "Namhsrih."

Wierd attempts to seem "hip" come off badly, and maybe as unprofessional. That's the part that seemed at odds with what I would have taken to be her advice.

It's not exactly criticism so much as something I noticed, that I wanted to mention--I guess I was curious about whether it struck anyone else. Blogging is informal, and you learn the various bits of language by use. But she should learn to use the language under a psued. Maybe Namhsrih.

horizontal rule
14

Also, teofilo has lice. And he caught them from pdf23ds.

horizontal rule
15

At least I don't agist pseud-cattle.

horizontal rule
16

Or pseudo-cattle, even.

horizontal rule
17

Pseud-cattle, of course, being cows who comment on blogs under assumed names and can use all the youthful phrases they want. Tim agists them.

horizontal rule
18

Knock Knock

horizontal rule
19

14-17: You go, girl.

horizontal rule
20

I pseud a cattle once, for breaking my toe.

horizontal rule
21

19: Omigod.

horizontal rule
22

One of the authors (Milligan) has a website with a presentation used to explain the study. If you can make it past the power-point reduction, it's a fair analysis.

It's accepted practice in research of this type to use populations of which only a subset contains the group of interest. For example, research about healthy diets will study large populations, say, mediterannean vs. US to look at differences in diets. Clearly not everybody in those populations eats the same thing.

The main weakness of the Baker, Gruber, and Milligan study is that the correlation they find between an increase in child care and depression in mothers or an increase in negative behaviors in children, may not be caused by the childcare policy changes because other changes may have occurred in Quebec during the years the study covers. However, this isn't a design flaw nor does it call for dismissal of the study, because the authors use two ways to make the study more robust. First, they compare Quebec to other states, thereby eliminating many confounding factors, and, second, they perform a longitudinal study, showing the results were sustained and not an incidental snapshot.

Still, even though the results have numbers that are significant in the sense that they are quite large, there is no measure of how much weight to give this study or of how robust this study really is. If the results are real, or true, then further studies will elucidate those results. So, further studies.

There's one potential problem that would not show up in the results and that is the short-term upheaval that a rapid societal change may bring. If, suddenly, the proportion of mothers in the workforce increases by 50%, it may take more than a few years of adjustment before the benefits in social well-being are realized.

While it would seem unwise to change government policy based on a single, seemingly weak study, continuation of a policy or even implementation should be based on positive findings and not just in economic or equity terms. At some time, somebody had better show the positive effects of social well-being because poor critiques of the present study does not positive evidence make.

(It's paradoxical to me that removing the child of one woman to give her to another woman would result in a net equity gain. If men were running those nurseries there would be some real equity. As it is, the equity gain is forced because of the concentration of a greater number of children with fewer caretakers. Women, and, often, women who have children of their own, take care of multiple children from other women. Does the resultant lower caretaker to child ratio really have a neutral benefit to the child and the mother? The economic gain is readily apparent.

horizontal rule
23

Erm, if there's an obvious economic benefit, and no good evidence of a non-economic harm, that comes out to a net profit, doesn't it?

horizontal rule
24

There is, at least, an indication of real non-economic harm that can't be dismissed. Besides, the two categories are difficult to weigh against each other. E.g. child labor has a great economic benefit while the evidence of harm only becomes apparent with abusive labor practices like long hours.

horizontal rule
25

The other thing you've missed is that this:

It's accepted practice in research of this type to use populations of which only a subset contains the group of interest. For example, research about healthy diets will study large populations, say, mediterannean vs. US to look at differences in diets. Clearly not everybody in those populations eats the same thing.

makes sense because it is very difficult to study what individuals actually eat -- they lie, they misreport, they simply won't put in the effort to keep track. So studies of populations that differ in diet on average are practical where studies of what individuals actually eat aren't. For day care, on the other hand, it's very easy to differentiate subjects in day care from subjects not in day care, and many studies have done so in the past.

That doesn't make this study a right wing plot, or categorically worthless, but it does mean that, given the existence of better targeted studies that contradict it, there's no reason to rush to any judgment on the basis of this study alone.

horizontal rule
26

It isn't true that there exists a large body of data showing net positive outcomes for children in day care. There are some studies, including some that show positive effects, and others that show negative effects, but none are comprehensive or definitive. The authors discuss a literature review in the paper.

There, they point out that other studies often look at the differences between children in child care compared to those not in child care. (what you are demanding). None of those studies address the impact of policy changes directly. This study seeks to address the specific question of what changes the tax policy instituted in Quebec had on labor, child care use and behavioral outcomes.

There may be a better way to design a study (there usually is) to better address the impact of a specific policy change on behavior but it isn't true that such studies exist. There is no easy way to tease out those data.

horizontal rule
27

I checked out that critique briefly when you linked the original paper, and I'd agree with Idealist and Yamamoto that it's weaker than the original paper.

Longitudinal studies of daycare really suffer from self selection - it'd be tough to get randomized assignment of families to stay at home vs. institutional daycare. In particular, assignment to 'high quality' day care - which the critiquers advocate - is particularly fraught because more involved parents fight like crazy to get their kids into better day care, confounding other results. Since this result describes changes in mental health over the entire population when there is greater access to institutional daycare, it suffers from this problem less than other studies.

I'd like to note that I say this as a parent with kids who've attended daycare, sometimes part time sometimes full time. Yamomoto's point that kids are better off with more personalized care is probably true, but it's an artifact of a really unfair labor system in which many women were forced to choose unpaid work at home.

I think it's best to think about this study as an incomplete description of the mental health costs of the institutions we are creating to care for children as our labor markets open to women. I've observed some of these costs in my own kids when they went into daycare - they need new behaviors to meet their emotional needs and some of those might involve anxiety and more conflict with parents. But the mental health picture is obviously incomplete - as I mentioned earlier for the depression finding for mothers - and there are emotional costs/imperfections to most systems to care for kids. (how many of the elite of the 18th and 19th century were actually raised by their parents?)

I do wish the authors wrote their paper with more careful thought and language to avoid this paper sliding into easy fodder for advocates of moving backwards rather than thinking about how we can best structure the way we offer care in a world in which women have good career options and continue to work.

horizontal rule
28

It isn't true that there exists a large body of data showing net positive outcomes for children in day care. There are some studies, including some that show positive effects, and others that show negative effects, but none are comprehensive or definitive.

Moving the goalposts by saying 'net positive', aren't you? No one needs to make the case that daycare is significantly preferable to maternal care, just that it is not harmful, and there is plenty of research showing that to be the case. And 'comprehensive or definitive' is nonsense in this context -- what would it even mean to refer to a study as 'comprehensive or definitive'.

horizontal rule
29

This study seeks to address the specific question of what changes the tax policy instituted in Quebec had on labor, child care use and behavioral outcomes.

Which may make it a source of useful information on the Quebec program. The fact that it doesn't track kids in day care separately from kids not in day care makes it significantly less useful (that is, inferior to a study that does track those groups) as a source of information about the effects on children of being in day care.

horizontal rule
30

24: Are we comparing daycare to child labor now?

horizontal rule
31

I think Y's just saying that non-economic benefits and economic benefits are hard to compare, e.g., if you just look at economic benefits, child labor looks like a fantastic idea.

I just can't see how a study that doesn't isolate kids in daycare can make claims about the effects of daycare on kids. I don't think the diet analogy is apt. It would be more like saying the mediterranean diet is healthier than the American diet because there are more tomatoes, and our proof of this is a study that didn't track how many tomatoes anyone ate.

horizontal rule
32

Well, and I think Y is deliberately trying to stir things up. Nothing wrong with that, of course, but anyone making an analogy of that type is making a calculated effort to piss people off.

It would be more like saying the mediterranean diet is healthier than the American diet because there are more tomatoes, and our proof of this is a study that didn't track how many tomatoes anyone ate.

Exactly. It's also worth noting that a lot of the guesses about what it is about one diet that makes it healthier than another (fiber? low fat?) that come out of the sort of broad study Y is referring to don't pan out when studied directly.

horizontal rule
33

I'm sick of all these studies on day care. It's obvious that there's not a goddamn thing wrong with having kids taken care of by other adults, and that shorter periods of time are better than longer for very young children, and that there are lots of perfectly acceptable options of different levels of formality, and that highly institutionalized care is not optimal but it's what a lot of people end up with because it's cheap. And that the "solution" to this problem is obviously to fucking invest some money in good, flexible options for people. And that what's actually going to happen is just more pointless and stupid oversimplifications about whether mothers "should" or "should not" put their kids in day care.

Bah, I say.

horizontal rule
34

Obviously, in an ideal world, simply investing money (somewhere) would provide good, flexible options for people.

The reality is that any government program is going to be proscriptive and result in changes in the options landscape. That is exactly what happened in Quebec. The tax break policy was instituted via government day care centers with the result that private day care was largely subsumed by institutional care.

Day care options are the easiest thing to provide but they do not present an adequate solution. This study bears that out. Children at young ages don't fare well and parents have a hard time with it as well, particularly mothers. I was surprised that the proportion of mothers who worked outside the home jumped so high. That indicates, to me, some kind of necessity. I don't think it's every mother's preference to be away from her children. For more emotional and time-demanding careers, far more time is spent away from young children, which results in exactly the sort of longer time periods that young children are away from parents that appear to be harmful.


horizontal rule
35

I just can't see how a study that doesn't isolate kids in daycare can make claims about the effects of daycare on kids.

Looking only at children in daycare vs. kids not in daycare is not a framework in which one policy can be compared to no policy or another policy. The reason is that the child care options are completely mixed. So one can look at a whole population before and after a particular policy. It's the sort of thing that is done all the time. Another example is a study that shows whether people are better off under Clinton or under Bush.

More to the point, by the time more than 60% of the state's children are in childcare, as is the case in Quebec, a strong selection has taken place and there is simply no way to study equal random samples from the daycare and the no daycare populations. By that time, there are compelling factors that will skew the results in a far greater way than a serial comparison of whole populations. For example, people with resources to be home with their children, can also provide other services that will contribute to the well being of their children, or, children with special needs who have mothers who can't work because of the undue burden, etc. If one would want to cherry pick matched samples to provide more rigorous controls, one would also, likely, remove the extremes that only become apparent when comparing whole populations. Anyway, that's a way of looking at it.

horizontal rule
36

Are we comparing daycare to child labor now?

No. It wasn't a comparison.
It was used to show lizard breath where her/his reasoning would lead in extreme example.

So many trigger issues here. Why all the hackles?


horizontal rule
37

34 seems completely untrue to me. Why would any govt. program be proscriptive (and do you mean "prescriptive")? There's no reason one can't have X program and other alternatives, e.g., Head Start runs alongside private preschools, childcare subsidies, daycares, etc. Formal institutional day care isn't a one-size-fits-all solution, but it isn't inadequate; licensed family day care, for instance, can be super for young children. How kids do depends on a lot of factors: turnover, child::provider ratios, hours in care, etc. etc. To simply declare it "inadequte" is silly. And it isn't true that more demanding careers necessarily demand more time away from small children; that depends on a lot of factors, like flexibility re. face time, what one's partner does, whether there is on-site daycare, whether one can afford a nanny who can bring the child to your work place for nursing breaks, and so on.

horizontal rule
38

Why all the hackles?

Hormones I bet.

Or guilt at being such terrible terrible mothers.

horizontal rule
39

I blame the raccoons.

horizontal rule
40

Looking only at children in daycare vs. kids not in daycare is not a framework in which one policy can be compared to no policy or another policy.

True. You need a before-policy and after-policy, like we might do with economic policy. Here is Calaville under Bush. Here was Calaville under Clinton. Etc.

But the control group for this study isn't Québécois kids before and Québécois kids after. It's Québéc compared to the rest of Canada over the same time period. And I'm wondering if this introduces too much noise in the signal. Québéc just hasn't been in good economic shape for a while, and I wonder how good a comparison of a financially stressed mom with kids in day care with a financially secure mom in Alberta with kids in day care can be.

It shows that given low-cost daycare, more parents will use it, and that these parents will report being stressed out. What I'm not sure it's doing is accurately catching the link between 'daycare' and 'stressed out mom.' How many Québécois women would have described themselves as stressed before daycare?

Again, this is like comparing the health of the American public under a new FDA rule with the health of the Italians during the same time and concluding something about the FDA rule. It's not a before and after picture of the same population.

horizontal rule
41

I was surprised that the proportion of mothers who worked outside the home jumped so high. That indicates, to me, some kind of necessity

People work because they need the money? All of these offices aren't simply full of people following their bliss? Man, the things you learn online.

Seriously on the study -- did anyone figure out how the 6-10 year old control group worked, and can you explain it to me if it did? I just can't figure out how to meaningfully compare 6-10 year olds to 0-4 year olds on the sort of metrics they were using.

Here, I'm not objecting to the methodology of the study -- I just don't understand it well enough even to object.

horizontal rule
42

They weren't doing that as near as I can tell, LB. They studied newborns through four-year-olds over an eight-year period, and compared little Québéc kids to little other-Canadian kids, concluding that the Québéc kids were more likely to be rowdy, and their parents were more likely to be hostile parents.

I don't think they're tracking 6-10 year olds at all (presumably because they're school age.)

horizontal rule
43

Well, exactly. Simply child care is not the solution, it's childcare plus. And the plus is the difficult part.

The way things are described in 37 (and, yes, prescriptive=much better) everything sounds so simple and easy. It just sounds a bit ludicrous. It just isn't that easy. There is an emotional toll (less for some people than others) that is not taken into this account. There is also the difficulty of providing for all the additional items like less face-time while building a career. That's not the reality of the present situation and throwing money at it will not solve that, either.

I had the good fortune to have access to all kinds of wonderful child care options for my children. It opened the gateway to a career but it was damned hard because of career demands. (Weekends, travel meetings, deadlines on grants.) In what fantasy world do mothers have it easy?

horizontal rule
44

I don't think they're tracking 6-10 year olds at all (presumably because they're school age.)

No, they definitely are -- I just checked my link and realised that I'd accidentally linked to a press release rather than the full study, which is here. 6-11 year olds are used as a control group (because they're in school regardless of the policy, so the policy doesn't affect them) somehow, but I'm not getting how.

horizontal rule
45

In what fantasy world do mothers have it easy?

Dunno. Has anyone alluded to a world like that? Comment 37 certainly didn't.

horizontal rule
46

"There is an emotional toll." I love sweeping, subject-free statements like this.

horizontal rule
47

(less for some people than others)

I think that's where some people get upset. There sometimes appears to be the implication that if the toll isn't large enough, you don't love your kids enough. That's sort of crazy, but I don't think it's any less influential a view for all that.

horizontal rule
48

They use the 6-10 year olds as a control for confounding variables. They are not used as a comparison group for the 0-4 year olds.

But the control group for this study isn't Québécois kids before and Québécois kids after. It's Québéc compared to the rest of Canada over the same time period.

Actually, it's both.
I agree the data are noisy.

horizontal rule
49

Yeah, an interesting facet of this study is that it measures maternal evaluation of the children's emotional states, rather than a blind evaluation by a third party. You think after enough discussions like this, women with their kids in day care might be less confident that their kids are all right, and more likely to evaluate their mental health status less positively? I don't see how this study could be controlled for a possible effect of increased maternal concern.

horizontal rule
50

They use the 6-10 year olds as a control for confounding variables.

Spell this out for me, using short words, if you would?

horizontal rule
51

There sometimes appears to be the implication that if the toll isn't large enough, you don't love your kids enough.

I was, rather, thinking of the implication that if the physical arrangements of daycare are provided, there is no toll.

I think there are many women who are very good at being on fully on task at work and fully on task at home. I don't think that says anything about love.

horizontal rule
52

Okay, here's what seems to be going on with the 6-11 year olds:

Second, where possible, we use 6-11 year old children as a control group. Since these children were in the same economic environment, we can use them to check for false positives.

I'm not quite sure how the math on all this works, but the idea seems to be that to insulate the study from other non-child care related occurrences (like the fact that Québéc's economy sucked, the referendum, potential terrorist attack, what have you) that could affect the children's behavior.

Whatever affects the 6-11 year olds (economy falls, kids become 4% more rowdy due to stressed out parents) will be assumed to be having the same effect on 0-4 year olds. So if there was a spike in rowdy behavior in four year olds, they'd look and see if there was a spike in rowdy behavior in the 6-11 year olds. If there is, discount the spike as probably not a result of the daycare policies.

It's not perfect, but I don't think it's bad methodology.

horizontal rule
53

That just seems like a weird assumption -- that anything having emotional effects on 0-4 year old children would be reliably likely to have the same emotional effects on 6-11 year old children. 3 year old 'rowdiness', e.g., doesn't seem to me to be that obviously related to 8 year old rowdiness.

horizontal rule
54

"There is an emotional toll." I love sweeping, subject-free statements like this.

Let me frame it this way. Many of the mothers who join the workforce are doing it out of necessity. IOW, there is an emotional converse to that.

horizontal rule
55

Well, here's the thing. It is a weird assumption, but it's an assumption that will work against the author's hypothesis.

Without the control group, we'd falsely conclude that every rowdyness increase was due to the new daycare laws. Well, that seems obviously bad, so we'll just assume that anything that affects older kids will affect younger kids, and we'll eliminate the resulting rowdyness. This will probably be overcorrecting, but it's overcorrecting in a way that works against your preferred conclusion. Biases introduced by this assumption will work against you.

Basically, you're making the case harder for yourself when you include the 6-11 year old control group. It's like giving your opponent the strongest form of their argument and then attacking that.

horizontal rule
56

Many of the mothers who join the workforce are doing it out of necessity.

Yeah, just like many of the dads. So?

horizontal rule
57

Why is "rowdyness" bad?

horizontal rule
58

Why is "rowdyness" bad?

Is Wolfson dead?

horizontal rule
59

re: 56
All things being equal the patriarchy would have ensured long ago that men were the ones to stay home with the kids.

horizontal rule
60

This will probably be overcorrecting, but it's overcorrecting in a way that works against your preferred conclusion.

How is it a conservative assumption? I can imagine confounding variables that might affect 0-4 year olds and not 6-11 year olds -- to pick a silly example, a Quebec formula manufacturer might change its formula to something with behavioral effects. Or a local TV station could change the time Sesame Street was on. I don't think assuming that any non-daycare effect is more likely to effect 6-11 year olds than 0-4 year olds is necessarily true.

horizontal rule
61

The study's looking at increased aggression and hyperactivity in kids, which I referred to here as 'rowdyness'. I assume that the authors of the study are assuming that aggression, hyperactivity and anxiety are bad.

'Obviously bad', though, in the sentence I wrote, refers to the methodology (i.e., not using a control) not the rowdyness.

horizontal rule
62

I assume that the authors of the study are assuming that aggression, hyperactivity and anxiety are bad.

For aggression, at least, that too is an odd assumption. I'd think of aggression as a normal personality trait that would be problematic if either too high or too low -- a slight change in the mean, if not tied to an increased incidence of individuals with problems relating to aggression, seems not to be much to worry about.

horizontal rule
63

And that you for the explanation -- that bit of the study just puzzled me.

horizontal rule
64

LB, I think the comparison that's worthwhile here is between

A) Drawing conclusions about the 0-4 year olds without controlling for anything within Québéc.

vs.

B) Drawing conclusions about the 0-4 year olds with controlling for economic stressors by looking at 6-11 year olds.

B) may indeed be underestimating the stressors, but it's clearly better than A), which ignores those stressors entirely.

horizontal rule
65

Cala's got it. The association they're finding is that between the time period before universal day care and after universal day care, these measures became worse. If there's any other change that simultaneously happened in Quebec over that period and affected kids and parents, then they could have a spurious association. By using this control group and finding their results hold, they rule out anything that affected kids evenly. So you can't just tell an alternative story like Cala's - that the worsening economy would affect kids - you'd have to explain why it affected 0-4 year olds significantly more than other kids. Not impossible, but it starts to feel strained.

horizontal rule
66

All things being equal the patriarchy would have ensured long ago that men were the ones to stay home with the kids.

OK, yamatrollo, you're not doing much to shore up your statistical arguments here.

horizontal rule
67

Eh, as trolls go, she's gotten the conversation started, and I'm not coming up with anything of interest to blog about today.

horizontal rule
68

All things being equal the patriarchy would have ensured long ago that men were the ones to stay home with the kids.

I must be out to lunch; I took that to be an offhand tribute to the pleasure of staying home with the kids.

horizontal rule
69

a Quebec formula manufacturer might change its formula to something with behavioral effects. Or a local TV station could change the time Sesame Street was on. I don't think assuming that any non-daycare effect is more likely to effect 6-11 year olds than 0-4 year olds is necessarily true.

I think the argument is not more likely but as likely. And see what I mean about strained arguments?

I think I like the tack better of saying that these are clearly a partial description of the mental traits of mothers and kids after the policy change, and we need a fuller picture over time - that includes cognitive & school performance, adaptation to diverse social settings (in which homecare kids are probably weaker), and mothers' overall welfare.

I'd also be curious what the data for single mothers & their kids show for the policy changes, even though it was a largest basket of reforms for single parents.

horizontal rule
70

All that said, it does seem like the study would have benefitted from looking at kids in institutionalized day care vs. kids not in day care vs. kids in private day care, especially since I think the authors are trying to argue that Québéc should be giving the parents tax breaks or vouchers ather than subsidizing daycare.

horizontal rule
71

Re aggression being a good thing: from memory, the authors do talk about a couple of possibilities, including that increased aggression in the 0-4 yo Quebec kids is just early socialisation to 'school' (ie the kids in other provinces catch up once they start grade 1), and they are pretty clear this is about short term effects vs long term effects, so they're definitely not saying the kids are going to turn into aggressive adults. Not that there'd be anything wrong with that!

horizontal rule
72

I think the argument is not more likely but as likely. And see what I mean about strained arguments?

I think I was unclear. Using the 6-11 year-olds as a control group would be very conservative only if you were certain that any confounding variable would either affect both groups in the same way, or would affect the older group only or more intensely than it affected the younger group. If there is an unaccounted for factor that affects the younger group only, or more intensely than the older group, the older group is useless as a control w/r/t that factor.

(You want a less strained possibility? Younger children spend more time with adult caregivers than older children, whether family or paid. Economic downturns make those caregivers emotionally stressed, and that stress affects younger but not older children. Now, this is a just-so story, I have no reason to think that it's true -- but I have no reason to think that it's false, either, and there may be a dozen other possibilities that I haven't thought of. That's why you want well matched control groups -- a poorly matched group just doesn't tell you much.)

If it is possible for some factor other than day care to affect

horizontal rule
73

The last half-sentence is an editing error, ignore it.

horizontal rule
74

If there is an unaccounted for factor that affects the younger group only, or more intensely than the older group, the older group is useless as a control w/r/t that factor.

Not useless, just not ideal or complete. The comparison is between a) not controlling at all, and taking all changes as evidence for your hypothesis or b) controlling for some of those changes. It doesn't mean that you can't postulate other reasons, like a Sesame Street shortage, but it does shift the burden of proof. I can't claim X% effect, I can only claim X-Y% at most. If I want to argue for a mitigating Z%, I can still do that.

It's better to have a more similar control group, but an imperfect control group is better than none at all.

Take a silly example. It's October of 2001 in NYC, and I"m doing a study on how 5-year-olds adjusted to kindergarten after some new regulation. So I survey their moms to see if their kids are stressed. And man, those are some stressed kids in the fall of 2001. Much more stressed than kids in California in 2001 or NYC kids in 2000. So I conclude that the NYC schools have become much more stressful in a year, and I blame it on Guiliani.

Wait, you say sensibly. What about September 11? I don't have unbiased evidence on 5-year-olds, but I can say that six-year-olds were reported to be stressed, too, etc, and I should maybe consider that when thinking about the changes to the kindergarten program. It's probably not perfect, but it's better than ignoring 9/11 entirely.

horizontal rule
75

Mmm. I'd agree with your 74. I was disagreeing with this from your 55: This will probably be overcorrecting, but it's overcorrecting in a way that works against your preferred conclusion. Biases introduced by this assumption will work against you.

You don't have any good reason to think that the use of the older kid control group will result in over, rather than under correcting. If an unaccounted for factor affects older kids more intensely, you'll overcorrect. If it affects older kids less intensely, you'll undercorrect. If it affects them in a precisely similar manner, you'll correct the right amount. I don't see any a priori means of guessing which is more likely.

horizontal rule
76

Fair enough. I wasn't being very careful in 55. I meant that controlling for economic changes, whether it over- or under- corrects, isn't going to trump up the evidence in favor of your hypothesis unfairly. Which is what we're worried about -- if by controlling it we overcorrect, we haven't presumed in our own favor. If by controlling it we undercorrect, we're still closer to 'accurate' than we would be if we didn't control at all.

horizontal rule
77

Except in that you've bolstered the credibility of your research by purporting to have controlled for unaccounted-for factors when you may not have effectively done so. E.g., if you do a study of injury rates in non-unionized meat-packing plants, and compare the injury rates of your subjects to those in a control group of Australian shepherds, you haven't really controlled for anything useful at all.

This isn't, of course, anything like that bad, but it's some distance in that direction.

horizontal rule
78

The difficulties in this study show just how hard it is to do this kind of research. It's really difficult to establish social effects that result from day care vs. no day care. Finding the specific effects that day care has on the development of young children is probably a near intractable question in the general way that it can be researched. It's not just because of the variability in care and some of the other criticisms that are reasonable, but also the inability to establish how a particular population would have fared otherwise. There are no true tests or real controls.

Breastfeading makes a useful analogy to daycare. There aren't any studies that show how a generation of formula fed babies in developed nations turned out to become disfunctional adults. Even specific problems like eye development are difficult to causally attribute to lack of breastmilk. Bottlefed babies do fine. But, we know of specific benefits that breastmilk offers infants and that breastfeading offers to the mother-child bond. To the point where breastfeading is now strongly recommended.

horizontal rule
79

That doesn't make any sense, Y. We know of specific benefits from breastfeeding because of well-controlled studies -- matched cohorts of breast and bottle-fed babies. It's not difficult at all to do such studies. What you seem to be saying (which is true) is that the benefits are small enough that for any given child you can't attribute any particular outcome to breast v. bottle, which is certainly true, and why everyone should stop giving people a hard time about it. (But, you know, large enough to make encouraging breast feeding a worthwhile public health move.)

horizontal rule
80

Further, this:

The difficulties in this study show just how hard it is to do this kind of research. It's really difficult to establish social effects that result from day care vs. no day care.

is absolutely true. And what it means is that whatever the effects of day care may be, they can't be all that large or they would be easier to study. No one worries all that much about how to design a study to show that smoking has bad health effects -- the effects are so great that they leap out at you, regardless of other factors.

horizontal rule
81

77: Oh come on. It's not that badly designed. All they claim is they're trying to reduce some false positives, and their big assumption is '0-4 year olds will be affected in the same direction, if not the same magnitude, by unpleasant events as 6-10 year olds.'

It could turn out to be wrong, of course -- maybe when mommy & daddy are out of work they're nice to the babies who thrive because their parents are home more but the older kids resent it.

And sometimes, assumptions like that are wrong. Medical studies often use young white men because they're what hang around campuses when the universities need volunteers, and there's been a few recent articles that this leads to drug recommendations that don't work as well on women and/or minorities.

But it's still not a bad guess, and it's certainly better than throwing up your hands and not using a control group. But that's why nothing ever turns on one study; those assumptions merit investigation, too.

horizontal rule
82

Certainly.

horizontal rule
83

Comity!

horizontal rule
84

I'm not sure that all the benefits Y refers to in 78 have been established by matched cohort studies. I can conceive of many that would simply be assumed good: the transmission of antibodies, for instance. Certainly that one about the mother-child bond could be established by other means. If a process is known to occur, then I can think of proceding on the assumption it was beneficial without double blind or other comparative studies..

horizontal rule
85

Sure, there's nothing wrong with doing uncontrolled studies if that's all that's practical, or as a preliminary step toward identifying areas for future research. But breastfeeding isn't particularly difficult to do a reasonably well controlled study of, (double-blind? No.) and day-care isn't impossible.

horizontal rule
86

There are no true tests or real controls.

This is just not true. This study goes much farther than others in looking at the effect of a shift in day care availability. And if you believe this is important enough, then we should consider randomized assignment to daycare groups.

I think that might be reasonable not for the daycare/no daycare choice - that really should be driven by the parents' desires and job options - but for the different forms of support for childcare that we could provide as a society.

horizontal rule
87

Oops, 86 is me.

horizontal rule