Decompose aggregate average-compensation growth into within, between, cross, entry, and exit effects (Foster-Haltiwanger-Krizan style)

Takes the per-group output of compute_growth_decomposition() and answers a different question than that function does: not "why did group i's own wagebill change" but "why did average compensation across ALL groups combined change" – specifically, how much is pay growth within groups versus a shift in headcount share toward higher- or lower-paid groups.

Usage

compute_wage_decomposition(growth_decomp, group_cols = NULL)

Arguments

growth_decomp

Output of compute_growth_decomposition()

group_cols

Optional character vector of columns identifying a higher-level unit (e.g. "country_code") within which shares/composition are computed separately. NULL pools all rows into one global composition per ref_date.

Value

A data.table, one row per ref_date (or per ref_date x group_cols), with avg_compensation, avg_compensation_lag, within_effect, between_effect, cross_effect, entry_effect, exit_effect, and total_effect (which equals avg_compensation - avg_compensation_lag by construction, except at the panel's first period, which is NA – no lagged baseline).