Access to EU-SILC microdata is restricted for scientific purposes only. See https://www.gesis.org/en/missy/metadata/EU-SILC/.
The analytical dataset used for the regression models is built from the original EU-SILC 2004-2019 microdata by combining information from the personal (P) and household (H) files, and by linking each female respondent to her co-resident partner within the same household and year. We first collected respondent- and partner-level characteristics (including labour-market variables and annual labour incomes) from the personal file, then merge in household-level material deprivation information from the household file, and finally generate a panel-ready dataset.
The resulting file is structured as a yearly individual panel (one record per respondent-year), with a stable individual identifier and a household identifier to allow clustering at the household level. Standard data-cleaning steps are applied to ensure that (i) the material deprivation index is based on a consistent set of items over time, (ii) income concepts are handled consistently within individuals, and (iii) key variables used in the regressions are observed (e.g., change in deprivation, earnings-loss indicator, family composition, and female employment status). The final output is directly used by the regression do-file after declaring the panel structure.
| Variable in regressions | EU-SILC source (main inputs) | Simplified construction / recoding |
|---|---|---|
deltamd6 |
Household deprivation items (see md6) |
Year-to-year change: deltamd6 = md6 - L.md6 (after xtset pid year). |
md6 (and L.md6) |
hs060, hs040, arrears items (hs010/hs011, hs020/hs021, hs030/hs031), hs050, hh050, hs110 (household file) |
6-item count of enforced lack. Arrears are harmonised across questionnaire changes (pre-/post-2008). Observations with >1 missing item among the six are dropped. |
wloss2 |
Partner annual labour income components: py010g/py010n and py050g/py050n (personal file) |
Partner income is aligned to the analysis year and built from employee + self-employment components. Earnings loss: wloss=1 if income drops by ≥20% vs previous year; wloss2=1 if loss occurred in t-1 -> t or t-2 -> t-1. |
family |
Child age information available in the prepared sample file (derived from EU-SILC household roster/children ages) | Indicator for having at least one child aged 0–5 in the previous wave (lagged). |
f_empA |
Respondent current labour status rb210 (personal file) |
Three-category status based on rb210 and L.rb210: not employed at t; employed at t-1 and t; entry into employment t-1 -> t. |
country |
Country identifier pb020 |
Encoded to numeric and used as factor i.country. |
cluster (stratified models) |
Derived grouping of countries into clusters (Nordic, Continental, Southern, Anglosaxon, Central-Eastern, Baltic). | |
pid (panel id) |
Individual identifier upid |
Numeric panel identifier created from upid. |
uhid |
Household identifier uhid |