“…Consequently, our outcome is not a binary indicator, and we face a multiple treatment problem (Lechner, 2001). We address this issue by estimating a multinomial logit and computing propensity scores for each of the three possible outcomes: not switching (denoted as outcome=0), investing in LDC (outcome=1), 5 This methodology has already been used in international economics to evaluate the effects of exporting and of acquisitions on firms' performance and returns to scale by Wagner (2002), , Girma and Görg (2004), Arnold and Javorcik (2005), Arnold and Hussinger (2005), and Girma et al (2007) and in the context of the relation between FDI and firms' performance, employment and R&D by Barba Navaretti and Castellani (2003), Egger and Pfaffermayr (2003), Debaere et al (2006), and Hijzen et al (2006). 6 Admittedly, this model would include other modes of internationalisation, such as export and contractual modes of international production (such as licensing, outsourcing, and joint ventures), which we will not be able to control for in our empirical analysis.…”