Why and how FDI stocks are a biased measure of MNE affiliate activity Sjoerd Beugelsdijk 1 , Jean-Franc ¸ois Hennart 2 , Arjen Slangen 3 and Roger Smeets 4 1 Department of International Business & Management, University of Groningen, The Netherlands; 2 CentER and Department of Organization & Strategy, Tilburg University, The Netherlands; 3 International Strategy & Marketing Section, University of Amsterdam, The Netherlands; 4 CPB Netherlands Bureau for Economic Policy Analysis & Department of International Economics and Business, University of Groningen, The Netherlands Correspondence: S Beugelsdijk, Department of International Business & Management, University of Groningen, PO Box 800, 9700 AV Groningen, The Netherlands. Tel: þ 31 (0) 50 363 9095; Fax: þ 31 (0) 50 363 3458; E-mail: s.beugelsdijk@rug.nl Received: 12 March 2009 Revised: 23 April 2010 Accepted: 25 April 2010 Online publication date: 1 July 2010 Abstract Many international business (IB) studies have used foreign direct investment (FDI) stocks to measure the aggregate value-adding activity of multinational enterprises (MNE) affiliates in host countries. We argue that FDI stocks are a biased measure of that activity, because the degree to which they overestimate or underestimate affiliate activity varies systematically with host-country character- istics. First, most FDI into countries that serve as tax havens generate no actual productive activity; thus FDI stocks in such countries overestimate affiliate activity. Second, FDI stocks do not include locally raised external funds, funds widely used in countries with well-developed financial markets or volatile exchange rates, resulting in an underestimation of affiliate activity in such countries. Finally, the extent to which FDI translates into affiliate activity increases with affiliate labor productivity, so in countries where labor is more productive, FDI stocks also result in an underestimation of affiliate activity. We test these hypotheses by first regressing affiliate value-added and affiliate sales on FDI stocks to calculate a country-specific mismatch, and then by regressing this mismatch on a host country’s tax haven status, level of financial market development, exchange rate volatility, and affiliate labor productivity. All hypotheses are supported, implying that FDI stocks are a biased measure of MNE affiliate activity, and hence that the results of FDI-data-based studies of such activity need to be reconsidered. Journal of International Business Studies (2010) 41,1444–1459. doi:10.1057/jibs.2010.29 Keywords: foreign direct investment; affiliate value-added; affiliate sales; MNE affiliate activity; measurement bias INTRODUCTION Multinational enterprises (MNEs) are firms that own value-adding activities in more than one country, and are hence central to international business (IB). While many IB studies have examined the location and geographic distribution of such activities at the level of individual MNEs (e.g., Davidson, 1980; Henisz & Delios, 2001), a significant number of them have taken the value-adding activity performed by all MNE affiliates in a country as their level of analysis. Between them these studies have investigated: (1) how the foreign value-adding activities of MNEs from a given home country are distributed across host countries (e.g., Dunning, 1980, 1993; Dunning, Fujita, & Yakova, 2007); (2) how such cross-country distribution is affected by home and host-country characteristics (e.g., Buckley, Clegg, Cross, Liu, Voss, & Zheng, 2007; Grosse & Trevino, 1996; Habib & Zurawicki, 2002; Sethi, Guisinger, Phelan, & Berg, 2003); and Journal of International Business Studies (2010) 41, 1444–1459 & 2010 Academy of International Business All rights reserved 0047-2506 www.jibs.net