How Sensitive is the Analysis of Firm and Employment Dynamics to Longitudinal Linkage Problems?

Empirical measures of firm and employment dynamics based on administrative datasets are biased due to missing links in the longitudinal observation of firms. This paper quantifies the bias in a set of widely used measures and evaluates the performance of two linkage methods in reducing bias. We find that a linkage method that builds on the continuity of the firm's workforce is more effective for producing reliable estimates of these measures than a traditional record linking method commonly applied by statistical agencies.Using improved linkages considerably modifies results regarding the contribution of different classes of firms to employment growth. It almost completely shifts the mass of employment at entry and exit towards the smallest firms. It reduces job creation by entrants and job destruction due to exit by half and reveals a greater importance of established firms for employment growth