Here is a strange pattern. Countries where more of the managers are women tend to have more homicide. It is not about wealth, and it is not chance. But it is also not what it first seems. This article shows how a link across the whole world can come mostly from one part of it.
The pattern
Each dot is a country. Across the page is the share of managers who are women. Up the page is the homicide rate, killings per 100,000 people. Across 131 countries the link is +0.42. Comparing only countries with similar income makes it slightly stronger, +0.46.
best-fit line line after adjusting
Every figure here is r: how closely two measures rise and fall together, from −1 (one goes up as the other goes down) to +1 (they rise together). 0 means no link.
Look region by region
Now split the countries by region and look inside each one, still comparing countries of similar income. In Latin America & Caribbean the link is strong, +0.60. Elsewhere it is weak or runs the other way: from −0.38 to +0.26.
Europe & Central Asia (43 countries)+0.22
Latin America & Caribbean (24 countries)+0.60
Sub-Saharan Africa (24 countries)+0.01
East Asia & Pacific (21 countries)+0.26
Middle East, North Africa, Afghanistan & Pakistan (12 countries)−0.38
In the middle Latin American or Caribbean country, 42% of managers are women, against 35% in the middle country worldwide. Its homicide rate is 17.7 per 100,000, against 2.4 worldwide. The countries there with the most homicide are Jamaica (64% of managers are women), Ecuador (45% of managers are women), Trinidad and Tobago (53% of managers are women), Bahamas, The (60% of managers are women) and Honduras (46% of managers are women). So the region sits in the top corner of the chart, and pulls the worldwide line up with it.
What is left once region is allowed for
Comparing countries only with others in their own region, the link drops from +0.46 to +0.26. Allowing for income inequality too leaves +0.24. So most of the worldwide link is a difference between regions, and only a small part is a pattern found inside them. Even that small part may be chance: within countries over the years there is no link at all (+0.01 across 74 countries).
All countries, before adjusting (131 places)+0.42
Comparing countries with similar income (131 places)+0.46
…and also region (comparing countries only within their own region) (131 places)+0.26
…and also adjusting for income inequality (the Gini index) (122 places)+0.24
The same trap catches a famous link
Income inequality and homicide are often said to go together, and across countries they do: +0.49 after allowing for income. But compare countries only within their own region, and it shrinks to +0.21. Much of that link is also about which region a country is in. See inequality vs homicide in full.
Why this matters
When a link spans the whole world, it is worth asking whether it holds inside each region. Regions share history, culture, laws and trade, so countries in the same region tend to be alike in many ways at once. A link that comes from one region may be about that region, not about the two things being measured. Every pair page on PlaceStats shows the link region by region for this reason.
What this cannot tell you
It describes countries, not people. It says nothing about women managers and crime.
Homicide counts are less complete in some countries, and women-in-management figures come from labour surveys of different quality.
A link is not proof that one thing changes the other.