Tuesday, April 17, 2007

S 59 Debate: The Basic Problem with the Empirical Anti-smacking Literature

There's a lot of empirical literature that at least vaguely correlates kids who are smacked with some bad attributes and life outcomes for them later. The negative outcomes appear to be small for cases involving true smacking (i.e., so that injurious and close to injurious cases are separated out), but they don't disappear. And no researchers that I'm aware of report finding that smacked kids ever do better later in measurable ways than the non-smacked.
At best, then, smacking starts to look like a relatively harmless minor vice. And yet parents, who overwhelmingly love and want the best for their kids, nonetheless overwhelmingly support preserving smacking for correction as an option. Why might that be? What might parents be sensitive to that the researchers aren't catching?

Consider what a lot of parents report (e.g., this one here): that their kids are quite different. That with some "time out" etc. works great, but with others it just doesn't. Some respond to bribes but some do not, etc.. Some kids are a day at the beach and some are a complete handful. For these parents, smacking often gets into the picture as a specific response to their child's non-responsiveness to other options etc. and to their child's specific nature more generally.
Very roughly such parents seems to be saying – and I take this to be close to common sense – that some kids are naughtier (or more transgressive or more chaotic or....) than others, and that they are the ones whom they end up smacking.

Let's suppose that the parents are onto something when they say this. Then the sort of question one needs to ask when looking at later negative outcomes is not
  • How do the smacked kids fare in comparison to the non-smacked?
  • How do smacked kids fare in comparison to similarly-naughty non-smacked kids? or
  • How do the smacked kids fare in comparison to the non-smacked when we control for their intrinsic naughtiness levels?
Consider the following ultra-simple model: 10 sets of parents (the smackers) who smack their naughty kids, and 10 sets of parents (the pushovers) who never smack any of their kids. Suppose each family has 2 kids, so that there are 40 kids in play total, and suppose that 80% of kids in both camps are nice and 20% are naughty. Suppose further that all of the nice kids (none of whom gets smacked - the smacking parents only smack their naughty kids remember) end up scoring a 10 on whatever later success-measure you like, the pushovers' naughty kids all score a 7, and the smackers' naughty kids end up scoring 8's. In this scenario, the smackers' naughty kids do better than the pushovers' naughty kids. Smacking has a real benefit it seems.
But now suppose one just compares the smacked kids to the non-smacked. The average smacked kid scores an 8. The average non-smacked kid scores (16*10+16*10+4*7)/36 = 348/36 = 9.667. Now it erroneously look like smacking is a complete disaster!

This is a very simple model... but it captures the flaw in most current "smacking is bad" studies, I believe.
In our example smacking's benefits
would show up even without keeping track of or knowing who's naughty or nice: you just have to be careful to make sure that you look at the smacking parents vs. the pushovers and note that the average kid of the former does better, i.e., (160+32)/20 = 9.6 > 9.4 =(160+28)/20. But I've not found a single study that keeps track of both smacking and non-smacking parents and also of the children they smack vs. those they don't. And of course in the real world where a much more complicated model is needed, you need much much more data than just these two basic statistics to have a hope of finding out the truth. (Still, with a little trickery and shrewd variable selection and reanalysis, I think I should be able to get smacking to come out fully positive in some existing studies. That's gonna take some serious library and analytical work to make happen, but stay tuned.)

Moreover the idea that most parents take for granted - that kids are differentially smackworthy: some naughty, some nice - is coded for in a causal sense in relatively little of the literature (I plan to spend some time in the library soon reading the material that claims to do some of this.), and is given systematically short-shrift whenever it's mentioned elsewhere. Here are some lines (with my commentary in italics) from Elizabeth Gershoff's widely cited review paper [Note added April 30 2007: Cindy Kiro interviewed on TV3 tonight leaned on this paper heavily and absolutely mischaracterized its results - albeit in ways that Gershoff herself has tended to do in informal media presentations!] that give a general idea of how primitive and loaded the state of the art is:
  • The stronger association between corporal punishment and the aggression composite for boys may also be accounted for by child effects; because boys tend to exhibit aggression more than girls, they may also elicit more corporal punishment from parents than do girls. However, as I discuss below, boys in general tend to receive more corporal punishment than girls. Taken together, these findings constitute a chicken-and-egg problem: Are boys spanked more because they are aggressive, or are they more aggressive because they are spanked more? Longitudinal studies beginning in infancy are needed to resolve these questions. {Isn't it bizarre that there's any serious debate that boys might be smacked more than girls because of their additional stroppiness? Surely there might be influences back in the other direction too, but what sort of nutty social constructionist do you have to be to entertain for a second that "child effects" generally and sexed/gendered child effects in particular aren't going to be very important?}
  • Because these meta-analyses are based primarily on correlational studies, parental corporal punishment cannot be identified definitively as the cause of these child behaviors and experiences, with the exception of immediate compliance. For some of the child behaviors considered in the meta-analyses, it is conceivable that the causal direction is reversed from what might be expected, such that children are driving the associations (e.g., aggressive children tend to elicit more corporal punishment from their parents). {Conceivable? Reversed from what might be expected? Only if your conceptions and expectation were governed by a nutty social theory could you doubt that the children themselves drive a lot of what goes on.}
  • The parent–child relationship is complex, and the mere fact that parents use corporal punishment is unlikely to be entirely responsible for how a child develops and behaves. {Mastermind subject: the bleeding obvious. What sort of field is it that needs to say this to itself?}
A final note about Gershoff's paper. Essentially my main complaint was registered in a commentary by Baumrind et al..(I'm less sympathetic to the rest of Baumrind et al.'s criticisms) Here's how Gershoff replied:
I must disagree [with Baumrind et al.]that assuming a primarily parent-to-child direction of effect is completely arbitrary. The main rationale parents have for using corporal punishment is that it will have an effect on their children; although some parents may spank their children out of frustration with their aggressive behavior, even then they do so to stop the behavior. It is no shock to anyone that indeed parents’ behaviors do affect their children and that disciplinary behaviors may be primed to have more of an effect than other behaviors because they typically involve arousal of both child and parent (Hoffman, 1983).This is not to say that children do not elicit behaviors from their parents; indeed, certain child behaviors may make some parents more likely to react with corporal punishment. However, the large majority of corporal punishment is enacted purposefully with the goal of correcting child behavior and bringing it in line with the norms and expectations of parents and of society. It is facile to suggest that parents are powerless to resist using corporal punishment in the face of their children’s noxious behaviors. As caregivers responsible for teaching their children how to behave, parents can choose how to respond to their child’s behavior; they can respond by spanking but also by sending the child to his or her room, taking away privileges, or ignoring the behavior.
Ho-lee crap. Gershoff suggests that someone who says they smack their kid because he's naughty etc. and to correct them/set a boundary for them etc. must be saying of themselves that they are a behavioristic black box that's had a brute involuntary response elicited! And that if they can't say that then the child's behaviors etc. can't be thought of as causally prior, as the thing to which their punishment/discipline is a measured response.
But that's just not so at all. A rational response to something can and normally does have the thing it's a response to as a cause. It makes perfect sense to say that your rational response is a response precisely to your child's defiance or cruelty or whatever it is that you think you just can't let go unchallenged/uncorrected. The child's behavior is smack-worthy by your lights. That's your judgment: it deserves/warrants a smack. You now have to act in response. No behaviorist reflexes need apply! Gershoff's so nutty and misdirecting and unforgivably behaviorist, and distorting of her legitimate critics here that it really does give one pause... I'm not impressed with her at all.

There's a serious problem here. Gershoff's supposedly authoritative view is the exact equivalent of taking seriously that crime is primarily or overwhelmingly an enforcement issue rather than than a people behaving badly issue. While what laws and enforcement regime you have does matter, to refuse to take seriously that crime/bad behavior is perfectly real and that its incidence is a prime driver of where you do and should deploy police, what laws you do and should have, and so on, is absolutely nuts. (Hence not even the most completely whacked out thinkers about crime say things as daft as "Police, crime it's a chicken-and-egg problem"!) I'd guess that the behavioral incidence variable isn't quite as strong in the child naughtiness, etc. case as in the crime case, but of course it also doesn't need to be.

Once upon a time people wondered whether smoking caused cancer. Quite a lot of serious thought was put into the idea that there might a "smoking gene" that was a common cause of both smoking behaviors and of lung cancer. Ultimately, however, the direct causal mechanism was compelling, and the idea of a gene causing a specific high-level behavior and also a specific cellular malfunction was way too much to swallow. It was a purely speculative, arguably very suspicious and implausible move to introduce it in the first place, so it was easy to just let that hypothesis die.
The reverse is the case with smacking and possible bad later outcomes. We've already
got a plausible basic model to work with, and it's what parents are reasoning with intuitively every day. According to that model, smacking overall is child-driven to a significant extent, and in the larger scheme of things, the thing that makes a kid more smackworthy is itself a risk factor for later bad behaviors and life outcomes. Surely there's all sorts of circular causation going on later – the world's a complex place – and there's always the outside chance that the highly intuitive model may be deeply misguided or not explain anything much at all. But to assume it's wrong and doesn't have a large role to play in any explanation is unbelievably obtuse. Sadly, this is the sort of dice-loading, pseudo-empirical social science that every ten years or so gets taken very seriously (e.g., "Sex/gender is all a social construction so if your kid has a circumcision disaster, just do a little external genitalia surgery and raise him as a girl - being a female/girl is primarily an effect of adult/cultural causes... go for it!" was Orthodoxy around 1990) with predictably disastrous results. (Sensible parents looked on in horror at how such smart people could have talked themselves into believing and recommending the transparently idiotic.)

I salute NZ's parents for intuitively grasping the fundamental reality of their situations and for not being impressed by almost certainly deceptively negative correlational data. It is a credit to the good sense of the New Zealand people that they have not allowed themselves to be bamboozled on this front. Various tiers of academia and of the social services establishment on the other hand, systematically oversell their data, and, in my view, live in iterated forms of social constructionist dreamland and denial (as they do on other issues when it suits them to do so).

I began this post by asking "What might parents be sensitive to that the researchers aren't catching?" And I might have added "...and that anti-smacking crusaders and politicians aren't catching?" Our answer, ultimately, is that Parents tend to be sensitive to the differences in their kids (even if they themselves choose not to let those differences make a difference on how they behave) whereas, for various reasons, the researchers and the crusading politicians alike are committed to treating all kids as the same or "equally", as something like a point of principle (although different principles apply in different cases). It's the later sort of commitment, which has both ideological and moral dimensions and versions, that blinds them and leads them to systematically either degrade or ignore the working, realistic and in my view, probably basically correct assumptions of most parents. In doing so they rightly earn the incredulity shading into scorn ("Call us with your results when you don't idealize away everything we actually think is important to our own decision-making about raising the children we love more than you ever will") of the prudent middle of society.

No comments: