A data feature on the film industry

Which Genres Actually Pay Off?

By Jiselle Resendiz

72,775 movies, 19 genres, 75 years of releases. This report follows the money from production budget to box office to find out which genres are the smartest bets โ€” and which just look expensive.

0.00x median return on every dollar budgeted for โ€” โ€” the highest of any genre in the dataset
Scroll โ†“

Now showing in the data

๐ŸŽฌ
0
Movies analyzed
๐ŸŽญ
0
Genres covered
๐Ÿ“…
โ€”
Years of releases
๐Ÿ’ฐ
โ€”
Highest median yield
01

Horror delivers the highest return on every dollar spent

Among the โ€” movies in the dataset with a reported budget and revenue, โ€” has the highest median yield of any genre at โ€”x revenue per dollar budgeted โ€” despite an average budget of only โ€”, one of the smallest of the 19 genres. At the other end, โ€” returns a median of just โ€”x.

Median (revenue รท budget) by genre, genres with at least 30 financially-reported titles. Hover a bar for the genre's defining title.

02

Bigger budgets don't buy better returns

Adventure and Animation carry the largest average production budgets of any genre โ€” โ€” and โ€” respectively โ€” and generate enormous absolute box office. But per dollar spent, their median yields (โ€”x and โ€”x) both trail the top genre's โ€”x.

Each bubble is a genre: position is average budget vs. average revenue, size is number of financially-reported titles.

03

The best-reviewed genres aren't the best-paying ones

Horror is TMDB voters' lowest-rated genre of the 19, averaging โ€”/10 โ€” yet it tops the yield leaderboard. โ€” is rated highest at โ€”/10, but doesn't come close to Horror's โ€”x median return.

Average TMDB vote by genre, all 19 genres.

04

Output boomed for six decades, then pulled back

Release volume grew from โ€” films in the 1950s to โ€” in the 2010s โ€” a โ€”x increase โ€” before falling to โ€” in the 2020s so far.

Number of movies released per decade.

Intermission

Back to back, that's โ€” years of movies

Total combined runtime of all 72,775 films in this dataset: โ€” hours.

05

Drama and Comedy dominate production, not the yield chart

Drama is the largest genre with โ€” films (โ€”%), Comedy follows with โ€” (โ€”%). Yet their median yields โ€” โ€”x and โ€”x โ€” both sit mid-pack.

Number of movies by genre, all 19 genres, sorted by volume.

06

Foreign-language films rate better but stay far less visible

Non-English-language films average a higher critic rating โ€” โ€” vs. โ€” out of 10 โ€” but score barely more than half the popularity, โ€” vs. โ€”. English is the original language for โ€”% of the dataset.

Average TMDB vote by original language, top 10 languages by title count.

07

The cost of making a movie varies by an order of magnitude

Average production budgets range from โ€” for โ€” to โ€” for โ€” โ€” a โ€”x gap between the cheapest and most expensive genre to produce.

Average production budget by genre, sorted highest to lowest.

08

Runtime is a genre signature, not a coincidence

โ€” runs shortest at โ€” minutes on average, while โ€” runs longest at โ€” minutes โ€” a โ€”-minute spread across the 19 genres.

Average runtime (minutes) by genre, sorted longest to shortest.

About this data

Source
Full TMDB Movies Dataset (asaniczka, Kaggle), downloaded 2026-09-27. Poster art in the marquee above is served live from TMDB's public image CDN.
What one row is
One row is one movie with a "Released" status on TMDB, identified by its release date and its primary genre.
Rows dropped and why
From ~1.5 million raw rows: kept only status == "Released"; dropped rows with an unparseable release date or a release year outside 1950โ€“2026; dropped rows with no listed genre (took the first listed genre as each movie's primary_genre); dropped rows with fewer than 10 TMDB votes. Result: 72,775 rows. See scripts/clean_data.py for the exact code.
How rates and averages are computed
Yield = revenue ÷ budget, computed only for movies with both budget > 0 and revenue > 0 reported (8,645 of 72,775 movies). Genre-level yield uses the median to avoid outliers dominating, and is only reported for genres with at least 30 financially-reported titles. All other averages are means over the full 72,775-row dataset unless stated otherwise.