Control of corruption

Control of Corruption (estimate)

How it has moved

Loading…

Every country, every year

One row per country and one column per year, sorted by the most recent value. Hover over any cell to see the figure. The country names link to each country’s page.

Loading…

Links that hold up

r is how closely two measures rise and fall together across countries, from −1 (opposite) to +1 (in step), where 0 means no link. These pairs are ranked by the link after adjusting for national income, which means comparing countries with the same GDP per person. Each pair passed three checks. The link held after that adjustment. It was unlikely to be chance (q of 0.05 or less), even allowing for countries in the same region being alike and for the many pairs tested. And the two are not just two measures of the same thing. Very small q values are written in short form: 8.0e-5 means 0.000080.

Links that don’t count as findings

These go together strongly, but they are built from the same numbers or measure the same thing. So they are bound to match, or their match shows that the measures agree, not something about the world.

How much the other measures account for

The 71 most widely reported measures boil down to a few broad patterns. This table shows how much of the differences between countries on this measure those patterns account for, adding one pattern at a time. The patterns are rebuilt without this measure, and without anything built from the same numbers or measuring the same thing. So it never predicts itself under another name. The figures use the 133 countries reporting all of them, and 65 candidate measures.

This measure sits on one of the smaller patterns, not on the one all the measures share.

How it differs by category

This compares the measure across groups of countries, such as those sharing a main religion, instead of against another number. The figure is the share of the differences between countries on this measure that lies between the groups, on a scale from 0 to 1. It is shown before and after adjusting for national income.

The “between groups” figure runs from 0 (the groups are no different) to 1 (the groups explain all the differences between countries). Groups have no up or down, so this says how much the groups differ, never which way. For that, read the group averages. The figure also tends to rise when there are more groups. So to judge whether it could be chance (q), we shuffle the same labels among the same countries many times and see how often a figure this big turns up. Then we allow for the fact that many measures are checked at once, since with enough checks a few big figures turn up by luck alone. Groups with fewer than 12 countries are left out. And countries that share a label share a lot else. Adjusting for income rules out one third factor, but leaves language, region, legal tradition and history untouched.

Every other pair

Everything else this measure was tested against, apart from the links that hold up and the pairs built from the same numbers. Sorted by the link after adjusting. Search by name; every column sorts. A small adjusted figure is a result too. It says the two do not move together once income is set aside.

Source