Is "some men might exaggerate about how many women they've screwed" really a controversial idea?
People tend to be honest on written questionaires though...
Actually we have strong evidence that people do not tend to be honest on written questionaires. For example, the Conflict Tactics Scale which is used in domestic violence estimates asks "did you do X" and "was X done to you". When asked to couples, which should match up there's a very poor correlation (it's at or below random chance) between the two members of the same couple, but the overall number of responses (abusing and being abused) are about the same size. This super-low correlation hold for male sufferers with female abusers, and female sufferers with male abusers.
The bizarre thing is that both abusers and victims PROPORTIONS match, but, the wrong people say yes or no: e,g. 10 men say they're an abuser, and 10 women say they're victims, but they don't even match. The 10 male abusers had wives who didn't report being victims, and the 10 female victims had husbands who didn't report being abusers. Gender reverse this, and they find the same thing. So, it literally isn't simply a case of "abusers just pretend it didn't happen", there's something more complex going on.
This is seen as a weakness of the "conflict tactics scale" but in reality it just means ALL similar questionnaires are fundamentally suspect, and most similar surveys "get around" the problem by eliminating one side's answers beforehand (don't ask both people about an event, and you never have to deal with discrepancies in answers).
So at best we can just hope people are lying or confused in a consistent manner and that the overall averages are right.
http://en.wikipedia.org/wiki/Conflict_tactics_scale#CriticismAnother methodological problem is that interobserver reliability (the likelihood that the two members of the measured dyad respond similarly) is near zero for tested husband and wife couples. That is, the chances of a given couple reporting similar answers about events they both experienced is no greater than chance. On the most severe CTS items, husband-wife agreement is actually below chance: "On the item "beat up," concordance was nil: although there were respondents of both sexes who claimed to have administered beatings and respondents of both sexes who claimed to have been on the receiving end, there was not a single couple in which one party claimed to have administered and the other to have received such a beating."
Many domestic violence measures are derived from the Conflict Tactics Scale.