Global Socio-Economic Analysis of Suicide Rates

Business problem

Public-health teams need to know where risk is concentrated before deciding where to focus attention. This project asks which groups show the highest recorded rates, how rates changed around major economic events, and whether national wealth explains the differences.

Dataset

A public table of suicide counts and rates by country, year, age group and sex, with population and GDP fields: 27,820 rows and 12 features.

Limitations:

• Data is aggregated by country, year and group. It cannot say anything about individuals.
• Recorded rates depend on how each country reports, and underreporting is likely in some places.
• It has no clinical variables such as depression or treatment history.

Analytical workflow

1. Cleaning. The HDI column was removed because about 70% of its values were missing, and GDP formats were standardised.
2. Exploratory analysis by age group, sex and country.
3. Time-series trends across 1985 to 2016.
4. Correlation analysis between GDP per capita and suicide rate.
5. Segmentation by demographic group.

Business insights

• Recorded rates are highest in men aged 75 and over, and male rates are several times higher than female rates in the data.
• Male rates moved far more than female rates during periods of economic disruption, notably around 1995 in several Eastern European countries, while female rates stayed comparatively steady.
• GDP per capita shows essentially no correlation with suicide rate in the pooled data (reported r of about 0.00).

Methodology and technical decisions

• Dropping the HDI column was preferred over imputing 70% missing values, which would have introduced mostly invented data.
• Rates per 100,000 are used rather than raw counts so groups and countries of different sizes can be compared.
• Interpretation is kept at group level. The repository goes further and proposes screening rules and funding triggers. This page keeps to what the data supports and treats those as ideas for public-health professionals to evaluate.

Results and limitations

• A near-zero pooled correlation does not mean wealth is irrelevant. Pooling countries and years can hide relationships within countries, and unmeasured factors may offset each other.
• The data cannot explain why rates are high in a group. Terms such as isolation are interpretations, not findings from this dataset.
• Spikes around 1995 line up in time with economic upheaval, but timing alone does not establish cause.
• No intervention was designed, tested or evaluated here.