Who Likes What Movies

In The New Yorker, Tad Friend on what movie marketers say about who likes what:

Young males like explosions, blood, cars flying through the air, pratfalls, poop jokes, "you're so gay" banter, and sex – but not romance. Young women like friendship, pop music, fashion, sarcasm, sensitive boys who think with their hearts, and romance – but not sex (though they like to hear the naughty girl telling her friends about it). They go to horror films as much as young men, but they hate gore; you lure them by having the ingénue take her time walking down the dark hall.

Older women like feel-good films and Nicholas Sparks-style weepies: they are the core audience for stories of doomed love and triumphs of the human spirit. They enjoy seeing an older woman having her pick of men; they hate seeing a child in danger. Particularly once they reach thirty, these women are the most "review-sensitive": a chorus of critical praise for a movie aimed at older women can increase the opening weekend’s gross by five million dollars. In other words, older women are discriminating, which is why so few films are made for them.

Older men like darker films, classic genres such as Westerns and war movies, men protecting their homes, and men behaving like idiots. Older men are easy to please, particularly if a film stars Clint Eastwood and is about guys just like them, but they’re hard to motivate. "Guys only get off their couches twice a year, to go to `Wild Hogs' or `3:10 to Yuma'."

This seems a nice set of "stylized facts" to explain. Must we invoke age and gender specific random cultural drift to explain these, or can we find more systematic and functional explanations?

I can roughly understand young men liking action, violence, and sex while young women like fashion, gossip, and romance.   But why do old men like darkness and idiocy while old women like critical praise, "doomed love and triumphs of the human spirit"?   Can you, for example, tell a plausible story of how this helps them learn about something useful, or helps them signal a valued characteristic?

GD Star Rating
loading...
Tagged as:
Trackback URL:
  • http://profile.typekey.com/aroneus/ Aron

    A woman could learn a lot about a man based on his feelings after seeing ‘3:10 to Yuma’. I’d almost recommend women make that a pre-marriage date night film.

  • Russell Wallace

    As you say, a lot of the patterns are easily explained. But:

    old men like darkness and idiocy

    How reliable is this conclusion? I find it surprising not least because my own liking for these things is much less now than when I was a teenager.

  • http://t-a-w.blogspot.com/ Tomasz Wegrzanowski

    IMDb has way more clue about demographic preferences than critics. Just look at these lists: http://www.imdb.com/chart/male http://www.imdb.com/chart/female
    There’s very very very little difference between them, male and female taste in movies is virtually identical.

    As an independent confirmation of that, from what Netflix Prize people say, predictions don’t get even a tiny bit better with demographic information included.

  • Ambi

    may be as we age our reaction to excitement,adventure decreases/suppressed/matures & in contrast to youngsters [except clinically depressed], older people are relatively more unhappy/depressed/pessimistic.
    may be this pessimism pervades to everything they view, to reflect on their experiences & hope less change but more order.

  • http://blog.efnx.com Schell

    Perhaps these marketing agencies are reinforcing the behavior they would most like us to display, so they can more easily cater to the public. I’d say that for the individual these assertions are false, they definitely don’t hold for me, in fact they seem like the generic stereotypes of a 17 year old boy. How convenient it would be if this were true? At least for Hollywood.

  • http://www.mccaughan.org.uk/g/ g

    Tomasz, I don’t think it’s very strong confirmation; demographic information could easily be useless for the Netflix Prize task even if men and women have very different tastes in movies. Imagine, for instance, that there are four kinds of person: men of type A and B, women of type C and D. And that A, B, C, D have *perfectly disjoint* tastes in movies; and that all people of each type have identical tastes. Then as soon as you know even one movie someone has liked, you know everything, and demographic information won’t help you a bit; but men and women in this imaginary world have very different tastes indeed.

    (Of course, if a substantial fraction of the Netflix Prize dataset is made up of people for whom you have little or no prior-choice data, then a failure of demographic data to help *is* evidence. Anyone know whether that’s the case?)

  • http://profile.typekey.com/aroneus/ Aron

    “..predictions don’t get even a tiny bit better with demographic information included…”

    This is a misinterpretation. Demographics are predictive. However, you do not have primary demographic data in the Netflix Prize contest. They do not give you that information. You could estimate demographics using various external data sources, but you aren’t adding any usable information in that process.

  • Douglas Knight

    I am shocked by the IMDB top movies by sex, that the gap between the sexes’ ratings is so small compared to the range over movies.

    There are serious selection problems here. The ratio of males to females rating a given movie varies widely. It looks to me to be strongly correlated with the gap between male and female rating and with stereotypes. The important question, to Tad Friend and for revealed preference, is how many people see the movie, not what they rate it.

  • Johnicholas

    This quote makes it seem to me that Robin Hanson believes he is cutting with Occam’s razor:

    “Must we invoke age and gender specific random cultural drift to explain these, or can we find more systematic and functional explanations?”

    However, I think this is a misapplication.

    There certainly is age and gender-specific random cultural drift. There is also certainly some impact on people’s minds from different biology (Judith Butler notwithstanding). Even if we believe the marketer is entirely correct about who likes what, we need at least one other data point from a substantially different time or culture.

    Otherwise, the two effects are compounded.

    Also, I would trust the Netflix Prize competitors over a professional marketer. Marketers probably divide their attention among many concerns, only one of which is predicting who will like what. The competitors have a lot more data, and their attention is not divided.

  • http://t-a-w.blogspot.com/ Tomasz Wegrzanowski

    g: Netflix Prize dataset is mostly people with few hundred ratings + 1-9 predictions per person to make. It’s plausible that with less data demographic information would be useful.

    Douglas Knight: I don’t have IMDb Pro account, but if I understand it having it allows you to do all kinds of advanced searches in their database, maybe you can find it out and tell us.

  • Douglas Knight

    Tomasz Wegrzanowski,
    how did you reach that belief about IMDB Pro? That’s certainly not how it’s advertised, so I doubt that if I sign up it will become obvious how to do such searches.

    The pro version offers a top 500 list of movies. Before I worry about that, I should compute the correlation for the public list. But I’d rather not look at the top-rated movies because I suspect that they have even worse selection effects. eg, who watches Citizen Kane?

  • Upload2045

    When I’m uploaded, I won’t watch movies. I’ll act in them.

    It generally saddens me to see these are the things people like and want to see in movies. But, then again, humanity generally saddens me. The entertainment business has some real content issues.

  • Upload2045

    When I’m uploaded, I won’t watch movies. I’ll act in them.

    It generally saddens me to see these are the things people like and want to see in movies. But, then again, humanity generally saddens me. The entertainment business has some real content issues.

  • http://t-a-w.blogspot.com/ Tomasz Wegrzanowski

    Douglas Knight: I might be wrong, I vaguely recall that some rankings and search features were supposed to be IMDb Pro-only, but that was ages ago, and might have even never been true.

  • ad

    The idea is that men fight their way out of dangerous situations, while women also try to befriend their way out of them.

    It seems to me that if you are trying to fight your way out of trouble, friends ands allies are something you should want rather badly.

    So why should there be a significant difference between the sexes?

    Depending on the nature of the threat, women might be more likely to prefer opposite sex allies.

    They might also see less to gain from having been seen to overcome a threat.

    And less to lose from having been seen to be afraid to see a horror film alone…

  • ad

    There’s very very very little difference between them, male and female taste in movies is virtually identical.

    The lists for worst films were entirely different. They only agreed about the best films…

  • Douglas Knight

    The IMDb top 50 by sex has 28 overlaps. Males score those movies higher by .2 +- .2 (avg +- stdev). The ratio of male raters to female is 6 +- 1.3. The correlation is .6. It’s only .5 if I take logs of ratios. I’m surprised it’s that high, since a scatterplot makes it look uncorrelated, except for 2 outliers, Godfather 1 & 2, which are the only movies with a rating of .6 and the only ones with ratios in excess of 8. Also, the first two Star Wars films are the only films with a rating of .4 and they have large ratios. But the two films with the lowest ratios have male excesses of .3, above average (It’s a Wonderful Life, Sunset Blvd).

    I suppose it is not surprising that old movies have more extreme ratios: there’s more choice involved in whether to watch the film. As I indicated before, I’d rather avoid that selection issue.