Female representation in Bollywood hits: A crude analysis

From Wikipedia, I collected data about the top 10 Bollywood hits each year from 1947-2022. I used OpenAI's models to classify each movie as 'female-led' or not, and to assign genders to the cast and the movie director. All these help us paint the picture of female representation over the years, specifically among Bollywood hits. Read more about this data and how it was collected on the Kaggle dataset page that I created.

Measures of female representation in Bollywood

I am using three measures, all imperfect, to understand the representation of women in Bollywood hits.

1. Whether or not a movie can generally be said to 'female-led', and is binary (1 if female-led, 0 if not). This is a fuzzy concept that is hard to define. To give an example, a movie where "boy falls in love with girl" is not 'female-led'. A movie where "girl goes on a journey to discover her true self" is female-led. Capturing this in the wide diversity of Indian movies using imperfect data is tough and inevitably this is a subjective, but not useless measure.
2. 'female cast importance' (normalized to lie between 0 and 1), and uses the scraped cast information. This measure is proportional to two things: the number of female actors listed in the cast, and also the order in which they're listed (listed earlier leads to a higher score). The exact definition can be seen in the code below.
3. female director, i.e. whether the director of the movie is female (0 if male, 1 if female).

Qualitative findings

While each of these measures gives an incomplete picture, the combination of all three is more reliable and tells an interesting story.

1. In the last ~15 years, there has been a mild uptick in all 3 measures, especially 'female director'. Depressingly, there's only one Bollywood movie which scores high on all 3 measures (i.e. has a 'female-led' plot, has a female director, and has a female cast importance > 0.6) - "Panga").
2. Shockingly, in the entire dataset of 740 movies, only 10 of them have been directed by women. Not until Farah Khan's 2004 hit "Main Hoon Na" was there a female directed Bollywood hit.
3. Surprisingly, the first 5 years 1947-'52, score pretty well on both 'female-led' and 'female cast importance' metrics. An abundance of female stars - Suraiya, Nutan, Madhubala et. al. - who were likely just as popular as their male equivalents was possibly a factor here.

Caveats and Limitations

1. Possibly the biggest limitation here is that I'm only looking at upto 10 hits from each year. Many critically successful or even reasonably popular movies do not make this list. For example, Kangana Ranaut's "Queen") was reasonably successful, yet does not make the top 10 hits of the year. Another recent example is "Pagglait", which is also not in this dataset. My guess is that this limitation generally has the effect of making the situation look worse in terms of female representation, since female-led movies may not appeal to as broad an audience as the regular Bollywood fare. At the same time, limiting to just the hits perhaps gives us a better indication.
   1. Potential improvement: Consider all movies, but weight them by their performance at the box-office. This requires updating the data collection scripts.
2. Classifying a movie as 'female-led' in a consistent way, over 75 years of movies, is extremely hard. I have no idea whether we are correctly classifying the really old movies (e.g. "Mirza Sahiban") ), and I mostly did not override the AI's classifications in these cases. Among recent examples, the two Akshay Kumar starrers "Toilet" and "Pad-Man" are arguably not female-led, but I erred on the side of generosity here and let the AI's classification that they are female-led remain.

Imports

import json
import pandas as pd
import matplotlib.pyplot as plt
from IPython.display import display, HTML
import math

Load the data and manually override false positives

with open('hindi_hits_from_1947_to_2022_ai_enhanced.json', 'r') as f:
    lines = set(f.readlines())
movies = json.loads("[" + ",".join(lines) + "]")
print("Loaded {} movies".format(len(movies)))

Loaded 740 movies

# a given gender's score is proportional to two factors:
# 1) how many actors of that gender there are
# 2) how high up in the cast list they are
def female_cast_importance(cast):
    score = {'male': 0, 'female': 0}
    
    for i, actor in enumerate(cast):
        gender = actor['gender'].lower()
        if gender in ['male', 'female']:
            score[gender] += 1 / math.log(i+2)  # Higher weight for actors earlier in the list
        
    return score['female'] / (score['male'] + score['female'])

ex = female_cast_importance([{'gender': 'male'}, {'gender': 'male'}, {'gender': 'male'}, {'gender': 'female'}])
print("female-cast-importance if the cast is 'male, male, male, female' is {}".format(ex))

female-cast-importance if the cast is 'male, male, male, female' is 0.16812753627111746

rows = []
for movie in movies:
    if movie['title'] == "Roti":
        # this one's broken
        continue
    cast_string = ", ".join([x['name'] for x in movie['cast'][:4]])
    # put all of these into a pandas dataframe
    rows.append({
        'year': movie['year'],
        'title_link': '<a href="' + movie['movie_url'] + '">' + movie['title'] + '</a>',
        'title': movie['title'],
        'female-led': movie['female-led'],
        'female_cast_importance': female_cast_importance(movie['cast']),
        'female_director': movie['director']['gender'].lower() == "female",
        'director_name': movie['director']['name'],
        'actors': cast_string
    })
df = pd.DataFrame(rows)

# Manually override some movies that were incorrectly classified
female_led_false_positives = [
    "Mujhse Dosti Karoge!",
    "Pardes",
    "Anari",
    "Amrit",
    "Judaai",
    "Tum Haseen Main Jawan",
    "Sharafat",
    "Aya Sawan Jhoom Ke",
    "An Evening in Paris",
    "Phool Aur Patthar",
    "Guide",
    "Gehra Daag",
    "Dil Apna Aur Preet Parayi",
    "Phagun",
    "Phir Subah Hogi",
    "Jagriti",
    "Sangdil",
    "Jadoo",
    "Arzoo",
    "Saathiya",
    "Ready",
    "Goliyon Ki Raasleela Ram-Leela",
    "Badrinath Ki Dulhania"
]
df.loc[df['title'].isin(female_led_false_positives), 'female-led'] = False

pd.set_option('display.float_format', '{:.2f}'.format)

List of movies considered 'female-led'

# female led, sort by year
subset = df[df["female-led"]==True].sort_values(by="year")
display(HTML(subset.to_html(index=False, escape=False)))

year	title_link	title	female-led	female_cast_importance	female_director	director_name	actors
1947	Neel Kamal	Neel Kamal	True	0.70	False	Kidar Sharma	Begum Para, Raj Kapoor, Madhubala
1947	Mirza Sahiban	Mirza Sahiban	True	0.61	False	K. Amarnath	Nurjehan, Trilok Kapoor
1949	Bari Behen	Bari Behen	True	0.49	False	D. D. Kashyap	Suraiya, Rehman, Ullhas, Geeta Bali
1949	Singaar	Singaar	True	0.77	False	J. K. Nanda	Suraiya, Madhubala, Jairaj
1950	Beqasoor	Beqasoor	True	0.39	False	K. Amarnath	Madhubala, Ajit, Yakub, Gope
1951	Hum Log	Hum Log	True	0.75	False	Zia Sarhadi	Nutan, Balraj Sahni, Shyama, Durga Khote
1952	Anhonee	Anhonee	True	0.39	False	K. A. Abbas	Raj Kapoor, Nargis
1953	Parineeta	Parineeta	True	0.39	False	Bimal Roy	Ashok Kumar, Meena Kumari
1957	Mother India	Mother India	True	0.39	False	Mehboob Khan	Nargis, Sunil Dutt, Rajendra Kumar, Raaj Kumar
1958	Sadhna	Sadhna	True	0.31	False	B.R. Chopra	Sunil Dutt, Vyjayanthimala, Leela Chitnis, Radhakrishan
1959	Sujata	Sujata	True	0.70	False	Bimal Roy	Nutan, Sunil Dutt, Shashikala
1962	Anpadh	Anpadh	True	0.45	False	Mohan Kumar	Balraj Sahni, Dharmendra, Mala Sinha, Shashikala
1963	Bandini	Bandini	True	0.47	False	Bimal Roy	Nutan, Ashok Kumar, Dharmendra
1967	Raat Aur Din	Raat Aur Din	True	0.26	False	Satyen Bose	Pradeep Kumar, Nargis, Feroz Khan, K N Singh
1968	Neel Kamal	Neel Kamal	True	0.47	False	Ram Maheshwari	Waheeda Rehman, Manoj Kumar, Raaj Kumar
1969	Aradhana	Aradhana	True	0.43	False	Shakti Samanta	Sharmila Tagore, Rajesh Khanna, Sujit Kumar, Farida Jalal
1970	Tum Haseen Main Jawaan	Tum Haseen Main Jawaan	True	0.36	False	Bhappi Sonie	Dharmendra, Hema Malini, Pran, Helen
1986	Nagina	Nagina	True	0.57	False	Harmesh Malhotra	Sridevi, Rishi Kapoor, Komal Mahuvakar, Amrish Puri
1986	Chameli Ki Shaadi	Chameli Ki Shaadi	True	0.39	False	Basu Chatterjee	Anil Kapoor, Amrita Singh
1988	Khoon Bhari Maang	Khoon Bhari Maang	True	0.51	False	Rakesh Roshan	Rekha, Kabir Bedi, Sonu Walia, Shatrughan Sinha
1991	Lamhe	Lamhe	True	0.51	False	Yash Chopra	Sridevi, Anil Kapoor, Waheeda Rehman, Anupam Kher
1993	Damini	Damini	True	0.39	False	Rajkumar Santoshi	Meenakshi Sheshadri, Sunny Deol, Rishi Kapoor, Amrish Puri
1994	Laadla	Laadla	True	0.48	False	Raj Kanwar	Sridevi, Anil Kapoor, Raveena Tandon, Anupam Kher
1996	Bandit Queen	Bandit Queen	True	0.61	False	Shekhar Kapur	Seema Biswas, Nirmal Pandey
1999	Taal	Taal	True	0.34	False	Subhash Ghai	Aishwarya Rai, Akshaye Khanna, Anil Kapoor, Amrish Puri
2001	Lajja	Lajja	True	0.70	False	Rajkumar Santoshi	Manisha Koirala, Rekha, Madhuri Dixit, Mahima Chaudhry
2016	Dangal	Dangal	True	0.59	False	Nitesh Tiwari	Aamir Khan, Sakshi Tanwar, Fatima Sana Shaikh, Zaira Wasim
2017	Secret Superstar	Secret Superstar	True	0.59	False	Advait Chandan	Zaira Wasim, Aamir Khan, Meher Vij, Raj Arjun
2017	Toilet: Ek Prem Katha	Toilet: Ek Prem Katha	True	0.39	False	Shree Narayan Singh	Akshay Kumar, Bhumi Pednekar
2018	Hichki	Hichki	True	1.00	False	Siddharth P. Malhotra	Rani Mukerji
2018	Pad Man	Pad Man	True	0.39	False	R. Balki	Akshay Kumar, Radhika Apte
2020	Chhapaak	Chhapaak	True	0.61	True	Meghna Gulzar	Deepika Padukone, Vikrant Massey
2020	Panga	Panga	True	0.62	True	Ashwiny Iyer Tiwari	Kangana Ranaut, Jassi Gill, Yagya Bhasin, Richa Chadda
2020	Thappad	Thappad	True	1.00	False	Anubhav Sinha	Taapsee Pannu
2022	Gangubai Kathiawadi	Gangubai Kathiawadi	True	0.59	False	Sanjay Leela Bhansali	Alia Bhatt, Shantanu Maheshwari, Seema Pahwa, Jim Sarbh
2022	Jugjugg Jeeyo	Jugjugg Jeeyo	True	0.54	False	Raj Mehta	Neetu Kapoor, Anil Kapoor, Varun Dhawan, Kiara Advani

List of movies with female directors

subset = df[df["female_director"]][["year", "title_link", "director_name", "actors", "female-led", "female_cast_importance"]]
display(HTML(subset.to_html(escape=False)))

	year	title_link	director_name	actors	female-led	female_cast_importance
180	2010	Tees Maar Khan	Farah Khan	Akshay Kumar, Akshaye Khanna, Katrina Kaif	False	0.23
225	2007	Om Shanti Om	Farah Khan	Shah Rukh Khan, Deepika Padukone, Shreyas Talpade, Kirron Kher	False	0.36
267	2014	Happy New Year	Farah Khan	Deepika Padukone, Shah Rukh Khan, Abhishek Bachchan, Sonu Sood	False	0.27
289	2019	Gully Boy	Zoya Akhtar	Ranveer Singh, Alia Bhatt, Siddhant Chaturvedi	False	0.30
320	2011	Zindagi Na Milegi Dobara	Zoya Akhtar	Hrithik Roshan, Abhay Deol, Farhan Akhtar, Katrina Kaif	False	0.28
347	2020	Chhapaak	Meghna Gulzar	Deepika Padukone, Vikrant Massey	True	0.61
483	2004	Main Hoon Na	Farah Khan	Shah Rukh Khan, Sushmita Sen, Sunil Shetty, Amrita Rao	False	0.36
552	2020	Panga	Ashwiny Iyer Tiwari	Kangana Ranaut, Jassi Gill, Yagya Bhasin, Richa Chadda	True	0.62
596	2015	Dil Dhadakne Do	Zoya Akhtar	Anil Kapoor, Shefali Shah, Priyanka Chopra, Ranveer Singh	False	0.46
641	2012	Talaash: The Answer Lies Within	Reema Kagti	Aamir Khan, Kareena Kapoor, Rani Mukerji, Nawazuddin Siddiqui	False	0.45

List of movies with female_cast_importance > 0.6

subset = df[df["female_cast_importance"] >= 0.6]
# Select specific columns
subset = subset[["year", "title_link", "director_name", "actors", "female_cast_importance"]]
subset = subset.sort_values(by="female_cast_importance", ascending=False)
display(HTML(subset.to_html(escape=False)))

	year	title_link	director_name	actors	female_cast_importance
737	2020	Thappad	Anubhav Sinha	Taapsee Pannu	1.00
659	2018	Hichki	Siddharth P. Malhotra	Rani Mukerji	1.00
704	1947	Dard	Abdul Rashid Kardar	Munawwar Sultana, Suraiya, Nusrat, Husn Banu	0.80
575	1949	Singaar	J. K. Nanda	Suraiya, Madhubala, Jairaj	0.77
735	1951	Hum Log	Zia Sarhadi	Nutan, Balraj Sahni, Shyama, Durga Khote	0.75
317	2001	Lajja	Rajkumar Santoshi	Manisha Koirala, Rekha, Madhuri Dixit, Mahima Chaudhry	0.70
220	2000	Fiza	Khalid Mohammed	Karisma Kapoor, Hrithik Roshan, Jaya Bachchan	0.70
671	1947	Neel Kamal	Kidar Sharma	Begum Para, Raj Kapoor, Madhubala	0.70
528	2022	Bhool Bhulaiyaa 2	Anees Bazmee	Tabu, Kartik Aaryan, Kiara Advani	0.70
625	1960	Barsaat Ki Raat	P. L. Santoshi	Madhubala, Bharat Bhushan, Shyama	0.70
385	1959	Sujata	Bimal Roy	Nutan, Sunil Dutt, Shashikala	0.70
405	1955	Mr. & Mrs. '55	Guru Dutt	Madhubala, Guru Dutt, Lalita Pawar	0.70
648	1978	Main Tulsi Tere Aangan Ki	Raj Khosla	Nutan, Vinod Khanna, Asha Parekh, Trilok Kapoor	0.64
694	2007	Chak De! India	Shimit Amin	Shah Rukh Khan, Vidya Malvade, Shilpa Shukla, Sagarika Ghatge	0.63
654	1986	Love 86	Esmayeel Shroff	Tanuja, Rohan Kapoor, Govinda, Farha Naaz	0.62
552	2020	Panga	Ashwiny Iyer Tiwari	Kangana Ranaut, Jassi Gill, Yagya Bhasin, Richa Chadda	0.62
639	1985	Ram Teri Ganga Maili	Raj Kapoor	Mandakini, Rajiv Kapoor	0.61
34	1996	Bandit Queen	Shekhar Kapur	Seema Biswas, Nirmal Pandey	0.61
359	2013	Goliyon Ki Raasleela Ram-Leela	Sanjay Leela Bhansali	Deepika Padukone, Ranveer Singh	0.61
48	2013	Chennai Express	Rohit Shetty	Deepika Padukone, Shah Rukh Khan	0.61
139	1951	Baadal	Amiya Chakravarty	Madhubala, Prem Nath	0.61
141	1994	Hum Aapke Hain Koun..!	Sooraj Barjatya	Madhuri Dixit, Salman Khan	0.61
304	1958	Phagun	Bibhuti Mitra	Madhubala, Bharat Bhushan	0.61
334	1948	11 O'Clock	Homi Wadia	Fearless Nadia, John Cawas	0.61
347	2020	Chhapaak	Meghna Gulzar	Deepika Padukone, Vikrant Massey	0.61
480	2002	Saathiya	Shaad Ali	Rani Mukerji, Vivek Oberoi	0.61
373	1947	Mirza Sahiban	K. Amarnath	Nurjehan, Trilok Kapoor	0.61
418	1949	Dillagi	A. R. Kardar	Suraiya, Shyam	0.61
372	1960	Kohinoor	S. U. Sunny	Dilip Kumar, Meena Kumari, Leela Chitnis, Kumkum	0.61
630	1961	Hum Dono	Amarjeet	Dev Anand, Sadhana, Nanda, Leela Chitnis	0.61
315	1998	Kuch Kuch Hota Hai	Karan Johar	Shah Rukh Khan, Kajol, Rani Mukerji, Sana Saeed	0.61
553	1983	Justice Chaudhury	K. Raghavendra Rao	Jeetendra, Sridevi, Hema Malini, Moushumi Chatterjee	0.61
721	2008	Bachna Ae Haseeno	Siddharth Anand	Ranbir Kapoor, Bipasha Basu, Deepika Padukone, Minissha Lamba	0.61
455	1984	Naya Kadam	K. Raghavendra Rao	Rajesh Khanna, Jaya Prada, Sridevi, Padmini Kolhapure	0.61
490	1987	Satyamev Jayate	Raj N. Sippy	Vinod Khanna, Meenakshi Sheshadri, Madhavi, Anita Raj	0.61

Plotting all three measures together, over time.

#import os
#os.environ["PATH"] += os.pathsep + '/Library/TeX/texbin'
#plt.style.use(['science'])
#plt.style.use('default')

cast_importance = df.groupby('year')['female_cast_importance'].mean().rolling(window=5).mean()
female_led = df.groupby('year')['female-led'].mean().rolling(window=5).mean()
female_director = df.groupby('year')['female_director'].mean().rolling(window=5).mean()

desired_dpi = 160
fig_width, fig_height = 4, 3

plt.figure(figsize=(fig_width, fig_height), dpi=desired_dpi)
# Plot each series
plt.plot(cast_importance, label='Female Cast Importance')
plt.plot(female_led, label='Female-led Movies')
plt.plot(female_director, label='Female Director Movies')

# Set plot title and labels
#plt.title('Evolution of Female Representation in Movies')
plt.xlabel('Year')
plt.ylabel('5-year rolling averages')

# Show the legend
plt.legend()

# Display the plot
plt.show()