We currently use a Gini of 31.1 for Ethiopia (from the World Bank, 2021). It's a real and well-sourced number, but it measures the wrong thing for our calibration.
- The World Bank's Ethiopia number (31.1) is built using the spending concept.
- how unevenly people earn. The World Inequality Database's Ethiopia number (about 54.5 for the same year) is built using the earnings concept.
These tell very different stories about the same country because richer households save a big share of what they earn, while poorer households spend everything (and sometimes borrow). Spending always looks more even than earning — even for the same population. That's why 31.1 and 54.5 are so far apart.
Since the model is about earnings, 31.1 makes Ethiopia look more equal than it really is.
There's a second, related issue. The same step uses a US Gini of 41.5 as a reference point — but the World Bank's US number is earnings-based (the US collects that data well), while their Ethiopia number is spending-based. So we're already comparing apples to oranges, before we even get to which Ethiopia value to use.
Suggested fix
Switch both countries over to the World Inequality Database. It uses an earnings-based definition for every country, so the comparison stays consistent:
- Ethiopia: about 54.5
- US: about 57–58
Small change in ogeth/income.py — two numbers and the comments around them. The longer-term improvement would be to build a Gini directly from labor surveys, but this swap is the right next step.
Ethiopia profile (World Bank PIP): https://pip.worldbank.org/country-profiles/ETH
Ethiopia (World Inequality Database): https://wid.world/country/ethiopia/
cc: @jdebacker @rickecon
We currently use a Gini of 31.1 for Ethiopia (from the World Bank, 2021). It's a real and well-sourced number, but it measures the wrong thing for our calibration.
These tell very different stories about the same country because richer households save a big share of what they earn, while poorer households spend everything (and sometimes borrow). Spending always looks more even than earning — even for the same population. That's why 31.1 and 54.5 are so far apart.
Since the model is about earnings, 31.1 makes Ethiopia look more equal than it really is.
There's a second, related issue. The same step uses a US Gini of 41.5 as a reference point — but the World Bank's US number is earnings-based (the US collects that data well), while their Ethiopia number is spending-based. So we're already comparing apples to oranges, before we even get to which Ethiopia value to use.
Suggested fix
Switch both countries over to the World Inequality Database. It uses an earnings-based definition for every country, so the comparison stays consistent:
Small change in
ogeth/income.py— two numbers and the comments around them. The longer-term improvement would be to build a Gini directly from labor surveys, but this swap is the right next step.Ethiopia profile (World Bank PIP): https://pip.worldbank.org/country-profiles/ETH
Ethiopia (World Inequality Database): https://wid.world/country/ethiopia/
cc: @jdebacker @rickecon