When national accounts data or microdata are not available, the estimates rely on a regression analysis to impute the missing data. The imputation is based on countries that are similar in terms of key economic and labour market variables. The methodology involves two steps. The first step is to compute the labour income share, adjusted for the labour income of the self-employed, which has been recognized in the economic literature as a crucial element for international comparability.

To achieve this,1 detailed data on status in employment are used to further categorize self-employment into three groups: own-account workers, contributing family workers, and employers. Furthermore, the labour income of each selfemployed group relative to employees is estimated based on a regression analysis of the microdata. The resulting estimate corresponds to the share of total income that accrues to labour: Labour income share = Labour income Gross domestic product .

Illustrative image (internet)

The second step imputes a labour income share for each country and year where a computation based on microdata is not possible. The estimates for 2004-2019 are obtained using regressions involving country fixed effects (where the labour income estimates together with microdata from step 1 are available) or region fixed effects (where the labour income estimates together with microdata from step 1 are not available), along with relevant covariates. A separate cross-validation approach is used to select the model that minimizes prediction error in the year 2020 and then again for the year 2021. For the year 2022, model estimates are calculated using the same approach (using the same models) as for the years up to and including 2019.Finally, for 2023 and 2024, since the current UNSD data is only available up to 2022, macroeconomic data, wages, 2 and other labour indicators, along with country fixed effects are used to estimate the values. Additionally, for a group of countries, OECD data on the unadjusted labour income share up to 2024Q1 is available, which is used as model input.

Disaggregation of labour income by gender From the labour income share estimation procedure, imputed labour income for all workers in the sample is estimated at the micro (individual record) level. Having micro-level imputed labour income enables the production of estimates of labour income of women and men separately, by aggregating all individual records by sex. 1 See ILO 2019 for a complete description of the imputation methodology. 2 From the forthcoming ILO Global Wage Report. When microdata is not available, the estimates rely on regression models to impute the necessary data. The estimates for 2004-2019 are obtained using models with country fixed effects (where at least one microdata-based value is available) or region fixed effects (where no observations are available) along with relevant covariates. A separate cross-validation approach is used to select the model that minimizes prediction error in the year 2020 and then again for the year 2021. For the year 2022, model estimates are computed using the same methodology (using the same models) as those applied for the years up to and including 2019. Finally, for 2023 and 2024, macroeconomic data, wages, 3 and other labour indicators along with country fixed effects are used to estimate the values.