Comments on Alfaro and Chen Ann Harrison Wharton, University of Pennsylvania and NBER May 2012 Goal of this paper Does incoming FDI raise productivity in a country because: 1. 2. 3. Incoming MNCS are more productive? Incoming MNCS weed out inefficient domestic firms? Incoming MNCs transfer knowledge (“spillovers”) to domestic enterprises? Answer: all three (MNCs, spillovers, and reallocation play a role) Truly impressive theory and empirical paper • Alfaro and Chen use heterogeneous trade theory to identify and distinguish between different sources of productivity changes due to MNC entry • Million observations across many countries, years (2002-2007) • Detailed TFP estimation using OP and a number of robustness tests Some possible concerns 1. Endogeneity: can the tail wag the dog? 2. Are there knowledge spillovers or not? 3. Why don’t policies to encourage FDI help? 4. Minor issues (magnitude of effect, direction of spillovers) Usual question about causality: • 1 million observations; 36,000 MNCs, implying 3.6 percent of data • Countries with more MNCs had higher TFP growth • But can the tail wag the dog? Aren’t MNCs just attracted to more dynamic, growing countries and industries? Clever IV Approach: first stage • Instrument for multinational entry in the host country using productivity of the multinational firm at headquarters. • Other determinants of location of subsidiary include distance (-ive), share a border (positive) and share a language (positive) But…. First stage R-square is low in Table 1, suggesting instrument is “weak”. Can we see F-statistics? Can we see over-identification tests for the validity of the instrument? Can we see the OLS results without IVs? Are there knowledge spillovers from foreign investment? Table 5: Table 10: yes no NOT for developing countries YES for developed but only vertical not horizontal So what are we to conclude? Why don’t policies to encourage FDI help? Table 11 suggests that tax incentives and other pro-FDI policies reduce foreign entry Clearly reflects the fact that incentives tend to be introduced when a country is less attractive to foreign investors My suggestion: cut this part out and devote a future paper to it once you resolve the reverse causality problem (that less attractive locations offer bigger incentives). Authors estimate impact of 100 percent increase in probability of MNC entry: Direction of spillovers • Assumed to move from MNCs to host country firms • Increasingly this assumption is violated, especially for emerging market MNCs, who travel in order to learn from their host country counterparts: CEMEX learns from Federal Express Haier learns from US competitors Emerging market MNCs in Silicon Valley New evidence for China China (Du, Harrison, and Jefferson) Strong vertical (backward) linkages or spillovers, disputing Table 10 Linkages increase post-WTO entry Perhaps you used the US input-output tables, and developing country inputoutput coefficients are different? More significant knowledge spillovers in subsidized sectors, disputing Table 11, and suggesting Chinese did it right. New evidence for India: FDI promotes learning but not reallocation Source: Harrison, Martin, and Nataraj, forthcoming More hard evidence on IP (China) IP instruments (tax incentives, tariffs, subsidies) work better when there is competition. (joint work with Aghion) IP instruments work better when targeted at exporting firms (joint work with (Justin Lin) BUT actual government targeting limited so much scope for improvement. Concluding Comments Important work—on both theory and estimation--disentangling the impact of FDI on host countries. Alfaro and Chen show us how to disentangle impacts on productivity and procompetitive effects of MNCs that weed out weak firms. Summarizing suggestions: More tests of validity of instruments Are there spillovers ? Do FDI policies help or hurt (cut Table 11) Direction of spillovers and magnitudes 谢谢你，再见！ Thank you, and goodbye!