%0 Report
        %A Schmieder, Johannes F.
        %T fevc: Leave-Out Bias-Corrected Variance Decompositions in Stata
        %D 2026
        %8 2026 Oct
        %I Institute of Labor Economics (IZA)
        %C Bonn
        %7 IZA Discussion Paper
        %N 19053
        %U https://www.iza.org/publications/dp19053
        %X Variance decompositions with worker and firm fixed effects measure heterogeneity and sorting, but estimation noise biases plug-in estimates. This article introduces fevc, a Stata command implementing the leave-out bias correction of Kline, Saggio, and Sølvsten (2020). It reports corrected worker and firm variances, their covariance, and the variance of their sum, with choices over deletion units, controls, and population weights. It also reports standard errors for projections of fixed effects on observed characteristics. A source-only Mata backend provides portability; a parallel Rust backend handles large datasets. Randomized calculations report Monte Carlo standard errors. In public Italian social-security data, the correction halves estimated firm-effect variance and reverses the worker–firm covariance’s sign. In synthetic benchmarks with 6.4 million observations and 28 threads, Rust is 1.3–3.5 times as fast as the KSS Matlab package; under match deletion, measured peak memory is about 30 percent of Matlab’s.
        %K fevc
%K two-way fixed effects
%K variance decomposition
%K leave-out estimation
%K limited-mobility bias
%K Mata
%K Rust
%K randomized linear algebra