Lifted from Crooked Timber comments

At http://crookedtimber.org/2012/04/01/autism-a-public-discourse-blaming-teachers-and-parents/#more-23859

Well, yes, of course autism is overdiagnosed: what the hell did people think that DSM was for?

The purpose of diagnostic manuals is to ensure that nobody who enters a psychiatrist's office will lack a diagnosis that will enable them to claim the cost back on their medical insurance.

This is perhaps seen most clearly in the diagnosis of Oppositional Defiant Disorder in children, where the criteria are
A pattern of negativistic, hostile, and defiant behavior lasting at least 6 months, during which four (or more) of the following are present:
(1) often loses temper
(2) often argues with adults
(3) often actively defies or refuses to comply with adults' requests or rules
(4) often deliberately annoys people
(5) often blames others for his or her mistakes or misbehavior
(6) is often touchy or easily annoyed by others
(7) is often angry and resentful
(8) is often spiteful or vindictive
Note: Consider a criterion met only if the behavior occurs more frequently than is typically observed in individuals of comparable age and developmental level.

(The note itself might conceivably be helpful if it offered some information on what frequency of these behaviours was typically observed, which of course it doesn't.)

Autism in DSM is a Chinese restaurant menu diagnostic system - two from column one, three from column two, one from column three - that mathematically allow something like twelve thousand different ways to be autistic.

Though, come to think of it, the 'Chinese restaurant menu' idiom is American; here in Australia Chinese restaurants have never had it.  We don't do fortune cookies, either.

Auties

In a critique of a paper by McDonald and Paul(2010) examining autism incidence
(http://www.all.org/pdf/McDonaldPaul2010.pdf) over at The Misbehaviour of Behaviourists (http://www.quicktopic.com/27/H/vJvhV4fDnBgw7) Michelle Dawson queries the article's assumption that "criteria for the specific diagnosis of autism have stayed about the same since 1978".
I fully appreciate that that claim's a bit dodgy, but I don't think that differences in diagnosis are an additional problem over and above the problem that the DSM-whatever criteria are so vague and ambiguous that if taken literally they vastly overselect. I mean, you might even accept that in Venn diagram terms the circle of "People who have autism" is entirely within the circle "People who fit DSM criteria for autism", but the difficulty is finding a way to check whether the expansion or contraction of either is particularly related to expansions or contractions in the other. Unless, of course, you take the view that autism actually is "whatever fits the criteria", without any further interpretive strategies.
Which then brings one up against the problems in assessing (say) "preoccupation with .. restricted patterns of interest that is abnormal..." when there are no measures attached, and if there were measures there would still be no documentation of how much 'preoccupation' falls within the normal range for three-year-olds.