Detecting which firm-specific characteristics impact market-oriented R&D

See UNCTAD (2005) for details.

The top 80 U.S. corporate R&D spenders deploy approximately 55% of their R&D funds overseas. The top 50 European companies spent 44% of their R&D overseas. The top 43 Japanese firms exported $40.4 billion (56%) of their R&D funds to other countries. See Jaruzelski and Dehoff (2008) for details.

For detailed discussion see Coe and Helpman (1995), Lee (1995), Park (1995), Lichtenberg and De La Potterie (1996) and Diao et al. (1999).

The USA is the most important receiving country. In 2005, the European Union invested USD 19.1 billion in the USA, while US multinationals placed USD 17 billion foreign R&D investment in the European Union. However, Asia has become more and more important. Jaruzelski et al. (2015) indicate that from 2007 to 2015, R&D imports to China grew 79%. India also had a 116% increase in imported R&D.

Similar findings can also be found in Shimizutani and Todo (2008). They found that a high degree of knowledge in the host country about the parent firm’s country attracts basic (applied) research, and market size attracts development (design) activities. Lewin et al. (2009) argued that the purpose behind offshoring innovation is to access qualified and talented personnel. Labor arbitrage is less important.

The literature lists four drivers of global R&D into developing countries: cost (Reddy 1997), production (Amsden and Ted Tschang 2003), market (Gassmann and Han 2004) and technology (Von Zedtwitz 2004) R&D. See summaries in Chen (2008).

For example, Ranjan and Agrawal (2011) found market size to be a significant determinant of FDI inflow into Brazil, the Russian Federation, India, and China. Boasson et al. (2012) found market size to be an important factor in attracting FDI to Nigeria. Kokouma et al. (2013) also found that market size played an important role in attracting Chinese FDI to Guinea.

For example, Chen (2008) found that one of the reasons that multinationals locate their R&D centers in Beijing is the large and dynamic Chinese market.

In a global survey done by UNCTAD in 2000, 85% of the countries, mostly developing countries, offer tax incentives to foreign investment. See UNCTAD (2000) for details.

This requires sufficiently strong intrafirm communication, sufficiently weak product-market competition, and sufficiently strong external spillovers.

Third, R&D offshoring can increase domestic welfare. Fourth, large markets are attractive as R&D hosts because the knowledge generated by the subsidiaries can also be applied to improve competitiveness in those markets.

An increase in R&D expenditure by MNEs in their subsidiaries does not necessarily imply a reduction in R&D at home. The idea of overseas R&D activities in this paper includes both R&D relocation and R&D investment in subsidiaries without a decrease in R&D expenditure at home.

For simplicity, we ignore the discount rate and assume that Northern firms can easily be transformed into multinational firms with no extra research cost.

Because of transportation costs and tariffs, domestic establishments only sell products to domestic markets and other markets where there are no other subsidiaries. Subsidiaries, however, possess the advantage of lower production costs, and they can sell final products back to the parent firm’s country as long as it is profitable. See “reverse imports” in Bayoumi and Lipworth (1998), which discusses the export and import situation in the standardization stage in product cycle theory.

This idea is also supported by Todo and Shimizutani (2008), who found that overseas subsidiaries’ R&D activities can improve the productivity of the parent firm if the subsidiaries engage in “innovative R&D.”.

The technique we use here is similar to the traditional price equation. See Baldwin and Spence Hilton (1984), Leamer (1994), Baldwin and Cain (2000) and Krueger (1997) for details.

Here \(\gamma \), \(\pi _\mathrm{S}\), \(\beta \), \(\hbox {RDE}_\mathrm{S}\), \(\alpha \), and \(\hbox {RDE}_{\mathrm{D}}\) are variables corresponding to P, Q, w, L, r and K in the price regression. We also assume that the subsidiaries’ profit function, in which \(\hbox {RDE}_\mathrm{S}\) and \(\hbox {RDE}_{\mathrm{D}}\), are the inputs, is homogeneous degree one. See appendix for detailed derivation of specification in Eq. (5).

Note that in the quality ladder and product cycle model, the R&D intensity is determined by the firm’s expected profit. An exogenous shock will change the firm’s expected profit and change its R&D intensity, e.g., the impact of international outsourcing on a firm’s innovation activity in Glass and Saggi (2001), and in this paper the impact of firm-specific characteristics on R&D expenditure. When firms determine their R&D intensities before entering the innovation race, they already have expectations for their profit levels as a result of the change brought about by the exogenous shock. Their expected profit will become their profit after they sell their final products. Thus, in our empirical model, we let the profit lag behind R&D expenditure.

In our case, t is 2003 and \(t+\tau \) is 2006. See next section for the detailed information of our dataset.

See Greenaway and Kneller (2008) and Raff and Ryan (2008) for a survey.

Our Eq. (7) corresponds to Eq. (6) on page 916 in Feenstra and Hanson (1999). Our firm-specific characteristics correspond to the structural variables in Feenstra and Hanson (1999). \({\delta }^{\prime }\Delta Z_{i,(t\rightarrow t+\tau )}\) in (7) is similar to the \(\gamma _k \Delta Z_{itk}\) in Feenstra and Hanson (1999). Our (8) corresponds to Eq. (8) on page 916 in Feenstra and Hanson (1999). Our R&D expenditure shares correspond to the primary factor cost shares in Feenstra and Hanson (1999).

The authors gratefully acknowledge the help provided by Professor Dumont.

Our zero-profit condition in Eq. (1) fits better with long-term data (at least 10 years) than with the short-term data we examine, but because of some characteristics of Taiwanese multinational firms, it is less problematic to imply a zero-profit condition in our study than it otherwise might be. First, most Taiwanese MNEs do business in very competitive markets. A question on the ASTM asks Taiwanese MNEs why they have had loss or difficulty in their overseas business in the last year. In 2003–2006, the first reason given was always intense competition from other local and multinational companies (47.20% in 2003, 58.09% in 2004, 58.91% in 2005, 57.95% in 2006). Second, more than half of the Taiwanese MNEs had less than 100 employees in their subsidiaries (54.31% in 2003, 54.30% in 2004, 52.31% in 2005, 52.56% in 2006), and most Taiwanese MNEs had less than 200 employees (67.45% in 2003, 68.22% in 2004, 64.62% in 2005, 64.18% in 2006). Thus, most Taiwanese MNEs are middle-sized or small firms whose long-run period under very competitive markets is relatively short. Third, Taiwanese MNEs’ subsidiaries locate their profit primary in retained earnings (38.46% in 2004, 39.71% in 2005, 39.10% in 2006). The zero-profit condition in our model requires that subsidiaries’ profits be equal to their R&D expenditure. There is no information in the ASTM telling us how subsidiaries finance their R&D expenditure. We believe that market research plays an important role in their retained earnings because strong local demand is the most important reason that Taiwanese MNEs received profit from their overseas subsidiaries (48.46% in 2003, 45.77% in 2004, 45.87% in 2005, 50.71% in 2006).

See proposition 2 in Lommerud et al. (2009).

Note that the sample size in Table 6 is different from the sample size in Table 3. The numbers in Table 3 cannot be computed by using the numbers in Table 6.

The key to understanding the decrease in \(\Delta K/L\_{\mathrm{S}}\) is the change in investment share in subsidiaries (INV_S) and total investment (INV). In Table 6, the total investment increased from 2003 to 2006, but investment share in subsidiaries decreased in 2004. Also, the increase in investment from 2004 to 2006 was less than the change in employment share in subsidiaries. Therefore, we can conclude that the fast increase in employment was the major cause of the decrease in investment per labor.