Recession Pop
An analysis into whether the music people reach for changes with economic vitality
My Role:
Data Analyst
Timeframe:
2 months
Tools:
RStudio
Category:
Machine Learning
Overview
Testing whether "recession pop" is a real signal or just a media story.
Pop media says hard economic times produce high-energy, feel-good hits. I tested that claim with machine learning on 630 Billboard top-10 songs from 1962 to 2024, matching their Spotify audio features against U.S. economic data.
Problem
"Recession pop" rests on anecdote: a few upbeat hits from 2009 and 2020. Similar pop-culture indicators like the lipstick and hemline indexes faded once more data came in. No one had checked whether music tracks the economy, or predicts it.
Solution
I used lagged regression to test whether music and the economy predict each other. Then I used PCA with logistic regression and SVMs to see whether audio could flag hard years. A formal recession flag failed: the model either guessed "no recession" every time or performed no better than chance. The misery index, which is inflation plus unemployment, separated years far more cleanly, with high-misery years clustering at lower musical positivity and energy. Music is a reflection of how people feel and genuinely responds to economic conditions.
