OkCupid, a free online dating website (through which I've met two girlfriends myself, incidentally), does some pretty interesting research on the profiles and activity of its users. This particular research blog post presents some fairly on-topic results in "Exhibit C: Looks":
@ArtemKaznatcheev rightly pointed out some range restriction here, so this can't address the validity of an age estimate below 18 years directly, but two things are worth noting:
- The overall slope resembles a simple linear trend that might continue upward past 18 years young.
- The absolute maximum here appears to be around 19, not 18. I doubt the difference is "significant".
It's not clear to me whose ratings these are or how exactly they were collected (as a user of the site, I remember a 1-5 rating index, and chose to rate women only), so it's hard to trust these results, especially in terms of their representativeness of a general population. Artem also pointed out that a better method for controlling social desirability in these ratings would hide direct information on age from respondents; OkCupid's research differs from the kind you're asking about in that regard too. Nonetheless, it's an empirical hint at the more explicit answer to the related, subjective question where respondents do have this info. Social desirability and age info are present in many normal social interactions after all, so removing age info deliberately might reduce naturalistic applicability of results anyway.
This blog post presents many more graphs on related topics that might interest you (e.g., as alternative operationalizations of attractiveness, or as behavioral correlates), but only one seems to address a moderator directly. That moderator is attractiveness itself! A quote follows this graph:
If you separate out the absolute best-looking women, almost all of whom are very young, and also remove the people you won't realistically want to date (the worst-looking women), you find that everyone else's attractiveness doesn't change much with age.
In other words, given that nobody is drop-dead gorgeous or drop-dead hideous, your average 25 year-old is roughly as good-looking as your average 35 year-old. Yes, throwing out the prettiest of the pretty young things is a clumsy handicap to put on an age-by-age comparison. But at the same time, for the vast majority of men, the best-looking women are simply out of reach, so it's actually accurate to exclude them as possibilities. In fact, unrealistic male expectations inspired this article, so we want to do everything we can to correct for them.
Many of you are probably scoffing at the idea that many 35 year-olds are as attractive as many 25 year-olds, but there are social factors at work that you might not consider as you go through life making judgments. Most importantly: nationwide, thirtysomethings are much more likely to be married and therefore much more likely to have stopped optimizing their attractiveness. So the typical 35 year-old woman you see out in the world isn't representative of the single 35 year-olds who are still dating and looking good.
I'm particularly doubtful of this second graph though, because it doesn't seem clear that the top and bottom deciles were identified and excluded separately within each year of age. "If you separate out the absolute best-looking women, almost all of whom are very young" actually sounds more like excluding the top decile while ignoring age, which might invalidate this graph. Removing the most extreme deciles of the sample while ignoring age would necessarily flatten the "everybody else" line somewhat, given any asymmetry in relationships between age and membership in these deciles of attractiveness across all ages.
Edit by what. Delete or edit as you see fit.
Nick, I too see a problem with how the most and least attractive women were deleted from the sample. I add this as an edit, as it does not warrant its own answer, but is too long for a comment.
If attractiveness declines with age, we must assume that it declines equally for all women, until this is disproven. So we must assume that a woman who is among the most beautiful at age 21, while still among the most beautiful at age 35, will have lost in overall attractiveness. The same goes for the average and the unattractive woman. We must assume (because we don't have the data) that over all women of all ages and all levels of attractiveness, we have a parallel decline in attractiveness with age, as shown in the graph below:
Now, if you delete the most and least attractive women from your sample, across all ages, you will, as the graph shows, create a sub-sample with no decline in attractiveness. This is an artifact of you tampering with the data and has no explanatory power whatsoever.
If on the other hand you delete the most and least attractive women for each age group, the general "slant" of the graph and the decrease in attractiveness that it represents does not change at all.
What I believe is that okcupid.com do not want to anger their older female customers, so they pimped or sugarcoated their data in a way that flatters the vanity of those women and allays their fears of being unattractive. In short: this section of their blog is advertising, not research.