Page 141 - Contributed Paper Session (CPS) - Volume 6

CPS1850 Paula Alves de Almeida et al.
Graph 3 Percentage distribution of Brazilian feature films directed by women by sex
of key roles, Brazil, 1996-2016

For the modelling procedure we considered as response variable “the
director is a woman” and as auxiliary variables the following characteristics of
films: sex of scriptwriter, sex of cinematographer, sex of producer, sex of
protagonist, and film genre. The variables “sex of cinematographer” and “film
genre” were not statistically significant in the logistic regression and were
excluded from the model. The “sex of protagonist” was significant, but the
model with this variable has a smaller value of the pseudo R2 and lower
predictive power than the chosen model.
Table 1 indicates that when the scriptwriter is a woman, the odds ratio of
the film’s being directed by a woman is 115 times that for films with male
scriptwriters. In addition, the odds ratio for a film’s being directed by a woman
is roughly 6.5 times when the producer is a woman then when the producer is
a man.
Table 1 Estimated coefficients, corresponding standard errors and odds ratios to
“sex of director”

The chosen model presents the best value for goodness-of-fit statistics
and predictive power among the fitted models. According to the value of the
pseudo R , the set of variables included in the model may explain 71% of the
2
variability of the response variable. The model also correctly predicts about
86% of the outcomes when the film is directed by a woman and 96% of the

130 | I S I W S C 2 0 1 9