Countries · structure
Most of this data is one pattern
Every measure here was chosen to be about something different. Most of them are not. Take the 71 measures that 190 or more countries report, and look at how they rise and fall together. A single pattern accounts for 45% of all the differences between countries. That pattern lines up with national income at r = +0.94 (where 1 would mean a perfect match). That is as close to being income as a thing can get without being it.
So the useful question is not what the patterns are. It is which measures are just that one pattern again, and which are not. The ones that are not can tell you something that income (GDP per person) cannot. Both lists are below, in percentages.
The shared pattern45%of all differences between countries
What it is+0.94how closely it matches income (r)
Everything else24%split across 5 smaller patterns
Countries133reporting all 71
The pattern itself
Countries at one end of it have long lives, few young people, working roads, power and water, and high income. Countries at the other end have the reverse. It is one sliding scale from poor to rich. These are the measures that move along it most strongly.
Highest at the rich end
HDI · Infrastructure score · Life expectancy · Health coverage index · Median age · Education score
Highest at the poor end
Population 0-14 · Birth rate · Infant mortality · Fertility rate
Which measures are that pattern, and which are not
Each figure is how much of that measure the shared pattern accounts for, across the countries that report it. For each score, the pattern is rebuilt without that measure. It is also rebuilt without any measure built from the same numbers or measuring the same idea. So no measure here is being used to predict itself. A measure near the top is national income under another name. A measure near the bottom is something income does not reach. That is where this data is most useful. It is also where a link between two measures is least likely to be the rich-poor pattern in disguise.
Mostly the shared pattern
| HDI | 93% | 133 |
| Inequality-adjusted HDI | 92% | 125 |
| Gender Inequality Index | 89% | 129 |
| Infrastructure score | 88% | 133 |
| Informal employment | 86% | 104 |
| Life expectancy | 85% | 133 |
| Population 0-14 | 85% | 133 |
| Birth rate | 84% | 133 |
| Health coverage index | 84% | 133 |
| Median age | 80% | 133 |
| Fertility rate | 79% | 133 |
| Infant mortality | 78% | 133 |
Almost none of it
| Births outside marriage | <1% | 41 |
| FDI inflows | <1% | 133 |
| Retirement age, women | <1% | 47 |
| Conflict deaths | <1% | 133 |
| Divorce rate (UN) | <1% | 68 |
| Single-parent households | <1% | 112 |
| Male labor participation | <1% | 131 |
| Christian share | <1% | 133 |
| Inflation | <1% | 133 |
| Military spending | <1% | 123 |
| Primary enrollment | <1% | 136 |
| Forest area | <1% | 133 |
The last column is how many countries the figure rests on. All 152measures are scored this way. Every measure’s own page shows its figure.
Is there anything else in here?
Yes, but not much, and none of it is large. Once the shared pattern is taken out, what is left splits into 5 smaller ones that add up to 24%. They are listed to be complete, not as findings. At this size, this page can’t tell a real pattern apart from an accident of which measures happen to be included. None of them is income again, which is the one thing worth checking.
| What separates countries along it | Share | Match with income (r) |
|---|---|---|
| Muslim share · Christian share · Civil liberties (FH) · Political rights (FH) | 8.0% | +0.14 |
| Migrant share of population · Age at childbearing · GDP per capita (market) · Population growth | 5.7% | +0.11 |
| Health score · Adult obesity · Women in a union · Primary enrollment | 4.0% | -0.09 |
| Refugees originating · Conflict deaths · Education spending · Armed conflict present | 3.5% | +0.11 |
| Forest area · Protected land · Agricultural land · Environment score | 3.1% | +0.06 |
How this was computed, and what it is not
A pattern is not a cause. This splits the differences between countries into patterns of measures that move together. It says how much of the data is one thing and which measures sit on it. It says nothing about what produces that. Two measures on one pattern need not be connected in any way.
It describes 133 countries. Only they report all 71 measures. They are richer and better documented than the ones that do not. Patterns found among the countries that collect the best statistics describe those countries.
Counts are left out. Total population and net migration say how big a country is, not what it is like. They would pull out a pattern that is only about size.
It describes 133 countries. Only they report all 71 measures. They are richer and better documented than the ones that do not. Patterns found among the countries that collect the best statistics describe those countries.
Counts are left out. Total population and net migration say how big a country is, not what it is like. They would pull out a pattern that is only about size.
Using more measures leaves fewer countries that report all of them. And splitting up more measures than there are countries gives shaky results. So here we keep measures that 190 or more countries report. The main figure barely moves across every cut-off worth using, so it does not rest on that choice.
| Countries required to report | Measures | Countries with all of them | Shared pattern |
|---|---|---|---|
| 160 | 114 | 55 | 41% |
| 170 | 105 | 79 | 42% |
| 180 | 88 | 94 | 39% |
| 190 · used here | 71 | 133 | 45% |
| 200 | 36 | 178 | 47% |