LIS in Our World In Data (OWID)

Our World in Data (OWID) builds an extensive dataset of inequality and poverty indicators, pulling together multiple sources to provide as comprehensive a view as possible.

To make it easier to navigate this wide range of data, they provide a set of Data Explorers that allow users to explore a very detailed range of indicators on Poverty, Inequality, and Incomes across the distribution. OWID provides three explorers draw from the LIS Databases:

• Poverty Data Explorer
• Inequality Data Explorer
• Incomes Across the Distribution Data Explorer

Information about the definitions and methods behind the data is provided below.

Welfare measure

OWID calculates poverty and inequality indicators for both after-tax and before-tax income. The definitions used align with those used in LIS’ DART data visualization tool and the Key Figures estimates, described here.

As a measure of after-tax income, OWID uses the measure of ‘disposable household income’. This refers to “cash and non-cash income from labor, income from capital, income from pensions (including private and public pensions) and non-pension public social benefits stemming from insurance, universal or assistance schemes (including in-kind social assistance transfers), as well as cash and non-cash private transfers, after deduction of the amount of income taxes and social contributions paid”.

As a measure of before-tax income OWID uses the LIS measure of ‘market income’. This refers to “income received by the households before public redistribution takes place; it includes cash and non-cash income from labor, income from capital, income from private pensions, as well as cash and non-cash private transfers, before deduction of income taxes and social contributions paid”.
The before-tax (‘market’) income is calculated as the sum of income from labor and capital (LIS variable: ‘hifactor’), private cash transfers and in-kind goods and services provided (hiprivate), and private pensions (hi33). The before-tax income is only calculated for surveys in which the required data on tax and contributions are fully captured (including where it has been imputed).

In order to make absolute comparisons of standards of living across countries and over time, the data – measured in local currencies at current prices – is converted into constant international dollars. The LIS data shown in the explorers is all measured in 2017 international dollars.

Accounting for resource sharing within households

The surveys LIS collates are conducted at the household level. The income or consumption reported in the survey data sums across all members of the household.
From the LIS microdata, poverty and inequality indicators are calculated based on two approaches for accounting for resource sharing within households:

• Per capita income: here, each member of the household (both adults and children) is attributed an income equal to total household income divided by the number of household members.
• Equivalized income: on this basis, incomes are adjusted to account for the fact that people in the same household can share costs like rent and heating. We use the ‘square root’ equivalence scale to make this adjustment: each household member (both adults and children) is attributed an income equal to the total household income divided by the square root of the number of household members.

Methods and assumptions applied

LIS provides very detailed documentation of how they process the original survey data on two dedicated metadata platforms: METIS and the Compare.it tool.

In calculating inequality and poverty estimates from the LIS microdata, OWID applies the same ‘top-’ and ‘bottom-coding’ procedure as used by LIS to calculate both the LIS ‘Key Figures’ and the DART interactive visualization tool’ indicators. This is done to remove extreme values from the raw survey data and to make the data across countries more comparable. For a more detailed discussion of why this is done, the methods used, and how it impacts resulting estimates, see here.

A well-known issue with household survey data is that the incomes of the richest are often poorly captured. This can lead to underestimates of inequality. Statistical offices organizing household surveys may adopt various strategies to minimize this, but this varies across countries and over time. In processing this survey data, the Luxembourg Income Study itself takes no steps to try to further correct the problem of missing top incomes. As such, inequality indicators based on this data – particularly those sensitive to the top, such as top income shares – may, in many cases, underestimate inequality.