Biased FDI spillovers in incomplete datasets: An empirical examination

Abstract

We examine biases in foreign direct investment (FDI) productivity spillovers that can arise when using incomplete datasets, by comparing estimates for Indonesia from the World Bank Enterprise Survey (WBES)—an incomplete dataset example—with estimates from the Indonesian Manufacturing Survey (MS). Furthermore, we conduct estimations on samples drawn from MS, following the sampling methodology of WBES. We find that estimates with this sampling framework are inaccurate, due to measurement error in industry-level horizontal and vertical FDI, strong presence of small firms, and small sample size. Relaxing the WBES sampling criteria and using FDI variables from MS produces substantially more reliable findings.