Statistics for "Sex Doesn't Sell"

There are 917 films in the database, released in the years 2001-2005 in the US. Some of these are listed on IMDb for other years, because their date of first release was an earlier year outside the US. Rereleases from this period are not included. Documentaries are also not included.

When I first started, I started with box office data from www.boxofficereport.com for films with more than $10 million US domestic gross. I then went to www.screenit.com for data on sex/nudity and violence, and downloaded all reviews on screenit.com for the years 2001-2005, including films grossing less than $10 million US domestic. (An earlier version of Table 1 uses this initial data, with box office data for grosses >$10 million only.)

All box office data in this updated tabulation is from IMDb. (Earlier values from boxofficereport.com have been updated to IMDb values.)

% Female is calculated from full cast lists on IMDb. When a name was obviously female, I counted it as female. When a name was obviously male I counted it as male. When I wasn't sure, I followed the link. Since IMDb lists people as actors or actresses, I then had no problem determining sex. It is possible that I got the sex wrong on some of the people I didn't check details on. But with full cast lists for 917 films (49609 speaking parts in total) it's unlikely that even a couple of dozen errors will make any significant difference in the overall results. Once I got into my second year of data the average percent female remained very stable.

Screenit.com rates films for sex/nudity and violence using six categories: none, minor, mild, moderate, heavy, and extreme. MPAA classifications (G, PG, PG-13, R and NC-17) were taken from screenit.com film descriptions for 915 of the 917 films. The classifications for the remaining two films (Y tu mama tambien (2001) and What the #$*! Do We (K)now!? (2004)) were found on IMDb. Correlation coefficients were calculated using Spearman's rank correlation coefficient, in which films are ranked from 1 to 917 for each of box office gross, MPAA classification, sex/nudity, violence and percent female, then the Pearson product moment correlation was calculated on these rankings. Partial coefficients are calculated with the usual formula.

The Christian Film and Television Commission (CFTC) found similar trends in their analysis of box office gross by film content, based on their own ratings of films (newspaper article here). It has been argued elsewhere that films with less sex or nudity do better in large part because they have lower classifications (e.g. not restricted) and therefore can be attended by more viewers. (Restricted films may also appear on fewer screens. Theatre chains don't like them.) Partialling out classification (for my data), however, indicates that sex correlates negatively with box office even when MPAA classification is controlled for. The correlation is smaller, but it is still negative, and it is still statistically significant.

Since sex and violence are also correlated with each other, it makes sense to also partial out violence to see if there is still a relationship between sex/nudity and box office. There is, and it is still negative and statistically significant, though not to the same level of significance as previously.

By this point, the negative relationship between box office and the percentage of women has disappeared. Partialling out sex makes no real difference to this finding.

Therefore there is no excuse for hiring fewer women!

Anemone Cerridwen
updated November 17, 2007