Jazz to EDM

Does music help your mental health?

736 listeners answered a survey about their habits and their scores. Here is what the data says.

Find your genre

Music helps, they said.

74.5%

of 728 people who answered said music improves their mental health. 169 noticed no effect and 17 said it makes things worse.

Heavy listeners hurt more.

Add up anxiety, depression, insomnia and OCD, each self-scored from 0 to 10. The total climbs with every band of daily listening. The best model, age plus log of hours, says the effect fades as hours grow, and it can't say which way it runs: people who struggle may simply reach for music more.

14.8

Out of 40, for 148 people who listen 0–1 hours a day

19.6

Out of 40, for 87 people who listen 7 or more hours a day

0–1 h 14.8

2–3 h 16.6

4–6 h 18.2

7+ h 19.6

Engraving of an old man tuning a radio while the clock on the wall reads two in the morning
“Perpetual entertainment” at 2 a.m.: a 1925 Paragon radio advertisement by Phil Cook.

Sleep goes with it.

Insomnia scores rise from 3.2 among people who listen an hour a day or less to 4.6 among those who listen seven or more. Depression follows the same climb, from 4.2 to 5.5.

The survey

An open Google Form shared on Reddit, Discord and in print, with no limit on age or place. Every answer, and every score, is self-reported.

736
responses, 33 questions each
Aug–Nov 2022
when the answers came in
16 genres
to pick a favourite from
Ages 10–89
median age 21
Bar chart of respondents' favourite genre: Rock 188, Pop 114, Metal 88, then Classical, video game music and EDM
From the notebook: favourite genre, all 736 respondents.

Pick your genre.

See how the people who share your favourite answered. Averages, not diagnoses, and small groups swing.

How we tested it.

Four symptom scores and sixteen genre frequencies are too many moving parts for one regression, so we compressed both, then fit simple and multivariate models with errors that tolerate uneven variance.

Prepare

Split the answers 30 / 70219 · 517

Keep the study variables8

Compress

PCA on the four conditionsPC1 50.0%

MCA, then cluster the genresk = 5

Post-hoc joint MCA, silhouettek = 7

Model

OLS, robust HC3 errorsBreusch-Pagan

Best fit: age + log(hours)adj. R² 0.056

Hours and age matter; genre and tempo don'tp < 0.0001

From the notebook

Scree plot: the first principal component explains most of the shared variance in the four mental health scores
One component carries half the variance in the four scores.
Dendrogram grouping the sixteen genres into seven clusters
Seven genre families, by how often people listen to each.
Genre clusters plotted in the first two MCA dimensions, overlapping heavily
The families overlap, which is why genre predicts so little.
Every respondent placed in MCA space, coloured by genre group and sized by daily listening hours
Every respondent, sized by hours a day.

Who asked

Hope Tadsen, Kala Fejzo, Emmanuel Damas and Shola Able-Thomas ran this study for MIDS 203 at UC Berkeley's School of Information in April 2026, using the public MxMH survey. It finds an association, not a cause: the answers are self-reported, the sample chose itself online, and one snapshot can't show whether listening changes mood or mood changes listening. Genre and tempo explained almost nothing once age and hours were in the model. For streaming platforms the useful lesson is care, not claims: heavy listening may be a sign someone is coping, and recommendations could respond to that without promising a cure. Read the full report.