Patent rights protection and Japanese foreign direct investment

Research Policy 42 (2013) 738–748

Contents lists available at SciVerse ScienceDirect

Research Policy journal homepage: www.elsevier.com/locate/respol

Patent rights protection and Japanese foreign direct investment Tatsuo Ushijima ∗ Aoyama Gakuin University, Graduate School of International Management, Shibuya 4-4-25, Shibuya-ku, Tokyo 150-8366, Japan

a r t i c l e

i n f o

Article history: Received 3 June 2011 Received in revised form 20 August 2012 Accepted 30 September 2012 Available online 31 October 2012 Keywords: Foreign direct investment Patent laws Technology spillover Japanese firm

a b s t r a c t This paper estimates the link between Japanese foreign direct investment (FDI) and the host country’s patent rights protection (PRP) in 1985–2004. Regressions performed on data that are aggregated in a variety of ways identify a positive and significant link that is concentrated in countries with a high innovative (imitative) ability and technology-intensive industries. Firm-level logistic regressions show that the link is stronger for firms that depend more on patents to protect innovations than their industry peers. These patterns lend strong support to the argument that PRP and FDI are correlated across countries because the strengthening of PRP ameliorates investors’ concerns about the spillover of proprietary technology. © 2012 Elsevier B.V. All rights reserved.

1. Introduction The relationship between a country’s patent rights protection (PRP) and its inflow of foreign direct investment (FDI) has considerable implications for national development policies and the location strategies of multinational corporations (MNCs). If PRP increases FDI inflows, as suggested by many authors and institutions [e.g., The World Bank (2002)], then policymakers designing national patent systems should consider this effect in addition to the traditional trade-off between domestic innovation and technology diffusion. However, if PRP is unimportant to the location decision of MNCs, then the merit of the TRIPS (Trade-Related Aspects of Intellectual Property Rights) agreement (and other international initiatives to strengthen global PRP) might be exaggerated, especially for countries with low innovative ability. A growing body of research has estimated the PRP–FDI link across countries, using mostly U.S. data. Consistent with the view that PRP stimulates FDI inflow by protecting investors from unintended spillovers of proprietary technology, many authors have identified a positive and statistically significant link between a country’s PRP and FDI inflow from the U.S. (e.g., Lee and Mansfield, 1996; Maskus, 1998; Smith, 2001). However, at least two issues remain unclear in this literature. The first is the role of foreign PRP in the geographical distribution of non-U.S. FDI. The second is whether the PRP–FDI link is caused by firms that depend largely on patents to protect innovations. If the link is not driven by the investment behavior of these firms, then the conventional wisdom becomes

∗ Tel.: +81 3 3409 8544; fax: +81 3 3409 4167. E-mail address: [email protected] 0048-7333/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.respol.2012.09.010

dubious because the link could merely reflect the collective effect of other policy and institutional factors coevolving with patent laws that remain uncontrolled for in these estimations (Primo Braga and Fink, 1998; Lall, 2003). In this paper, I help resolve these ambiguities by using microeconomic data on Japanese FDI over two decades. Japanese FDI provides an interesting case because, as Mansfield’s (2000) international survey of managers suggests, the location decision of Japanese MNCs is just as sensitive to foreign PRP as that of their U.S. counterparts. Nevertheless, no previous studies have estimated the effects of foreign PRP on the geographical distribution of Japanese FDI. Another unique feature of this study is that it takes a more detailed look at the cross-sectional variations of this PRP–FDI link. If PRP is correlated with FDI because of the direct effects of patent laws on the reduction of technology spillover, then the correlation should be particularly strong for the FDI of firms that largely depend on patents to appropriate innovations. On this critical issue, research has supplied only limited and mixed evidence at the industry level while supplying no evidence at the firm level. In this paper, I provide unusually detailed evidence on the heterogeneity of the PRP–FDI link across industries and firms. In particular, I estimate how a firm’s sensitivity to foreign PRP varies with its innovative activities (especially patenting) by matching FDI and patent data at the firm level. My empirical focus is on FDI undertaken in 1985–2004 by industrial firms listed on the Tokyo Stock Exchange (TSE). I estimate factors influencing these firms’ FDI location in two steps. First, data are aggregated for cross-country regressions to estimate the country-level determinants of FDI flows. After controlling for geographical distance and a variety of development variables, I find a statistically and economically significant link between Japanese FDI

T. Ushijima / Research Policy 42 (2013) 738–748

and the host country’s PRP. Consistent with Smith’s (2001) findings for U.S. MNCs, this link concentrates in host countries with a high ability to imitate foreign technology. In addition, disaggregating FDI by industry reveals that the link is specific to technology-intensive industries such as chemicals and electric machinery. These results are consistent with the view that the strengthening of PRP alleviates foreign investors’ concerns about the spillover of proprietary technology and thereby increases the FDI inflow into a country. In the second step, I perform logistic regressions based on firmlevel data to estimate a firm’s likelihood of investing in a country. I find that after controlling for various country- and firm-specific factors, a country’s PRP significantly increases the probability of a firm’s investing in the country. In addition, the estimated pattern of inter-firm variation in PRP–FDI links is highly consistent with the direct effect of technology protection because the link is stronger for firms that depend heavily on patents as compared to their industry peers. An interesting result is that a firm’s sensitivity to foreign PRP increases with its patent intensity, even in industries for which aggregate estimations fail to identify a positive PRP–FDI link. The PRP-sensitivity of FDI is therefore a function of the investing firm’s technology strategy as well as industry characteristics. Regressions also reveal that PRP increases FDI inflow into a country mainly by affecting the initial investment decision of new investors rather than the expansion decision of incumbent firms. Taken together, my results are consistent with the notion that strengthening PRP to improve the appropriability of innovation increases the inflow of FDI. The finding that this effect is mainly driven by technology-intensive firms implies that PRP affects the quality as well as the quantity of incoming FDI. It is important to recognize that the effect of PRP in attracting FDI concentrates in countries with relatively advanced technological ability. Governments must carefully weigh the costs and benefits of strengthening PRP in light of their country’s technological needs and capabilities. The organization of this paper is as follows. Section 2 reviews the background of this research. Section 3 introduces data. Section 4 performs cross-country regressions. Section 5 performs logistic regressions to estimate the PRP–FDI link at the firm-level. The final section concludes. 2. Background 2.1. Literature The relationship between a country’s PRP and FDI inflow is a highly empirical question. On the one hand, stronger PRP increases foreign firms’ ability to reap profit from proprietary technology transferred via FDI. This effect would make the relationship positive as PRP increases a country’s attractiveness to foreign firms, especially technology-intensive ones. On the other hand, PRP decreases the transaction costs of arms-length technology transfers such as licensing. This effect can make the relationship negative by inducing firms to substitute licensing for FDI. In the OLI framework of Dunning (1993), therefore, PRP increases a country’s location (L) advantage to attract investment by firms whose ownership (O) advantage is in technology, but simultaneously decreases these firms’ internalization (I) advantage.1 Empirical studies have shown that the cross-country correlation of PRP and inward FDI is generally positive. Lee and Mansfield (1996) make an early attempt to estimate the link. They find that

1 For this reason, researchers should ideally look at alternative transaction modes simultaneously, such as exporting, licensing, and FDI, to estimate the total effect of PRP on international technology flows. Maskus (1998) and Smith (2001) are examples of such an attempt while using aggregate data. Data on alternative transaction modes are difficult to obtain at the firm level.

739

the strength of PRP, as perceived by managers, is positively correlated with U.S. direct investment in a sample of fourteen developing countries. Ensuing studies use more objective PRP measures to examine a wider cross-section of countries. Maskus (1998) identifies a positive correlation between PRP and U.S. FDI stock in a panel of 46 countries. Smith (2001), studying a cross-section of 50 countries, finds that PRP and the foreign affiliate sales of U.S. MNCs are correlated positively and significantly. Unlike studies based on aggregate data, Smarzynska (2004) estimates the PRP–FDI link based on microeconomic data. She finds that a firm’s probability of investing in a transition economy in Europe and the former Soviet Union increases with the strength of PRP in the focal economy. Although evidence for the correlation of a country’s PRP and FDI inflow abounds, the implications of this link are not unambiguous. As Maskus (2001) notes, PRP is likely to be only one element of the “cocktail” of policies and institutions considered by firms in deciding where to invest. PRP could be correlated with FDI merely because it picks up the collective effect of many coevolving factors. The PRP–FDI link could also arise because of a signaling effect if investors regard countries with stronger PRP as providing more favorable climates for private business and investment (Lall, 2003). These confounding effects can be serious after the mid-1990s because many countries that joined the World Trade Organization (WTO) strengthened PRP to comply with the TRIPS agreement while simultaneously liberalizing policies in many areas. Studying the cross-sectional variation of PRP–FDI link can help resolve this ambiguity. If the link appears because of the direct effect of technology protection, it should be strong where the spillover of proprietary technology is a serious threat to foreign investors. For instance, firms investing in technologically sophisticated countries will face a large risk of losing technological advantage if they are not protected by strong patent laws because indigenous firms in such countries have a high ability to imitate advanced foreign technology. Smith (1999) finds that U.S. exporters are sensitive to foreign PRP when trading with technologically advanced countries. Likewise, Smith (2001) finds that sensitivity to foreign PRP of U.S. affiliate sales and licensing is significantly positive only if a country with high imitative ability is involved. Nunnenkamp and Spatz (2004) report that the cross-country correlation of PRP and U.S. FDI stock is positive only among countries with large endowments of human capital. The PRP-sensitivity of FDI would also vary across firms because, as the innovation literature documents, industries and firms differ considerably in the extent to which they use patents to appropriate innovations (Levin et al., 1987; Hall et al., 2005). Unfortunately, research has generated only limited and mixed evidence on the inter-industry heterogeneity of PRP–FDI links and has generated no evidence on the intra-industry heterogeneity among firms. Lee and Mansfield’s (1996) survey shows that managers in the chemical industry (including pharmaceuticals) are the most concerned about foreign PRP, which is consistent with the fact that patents are critical assets in this sector (Levin et al., 1987). Fink (2005) compares four technology-intensive industries, but fails to identify a significant link between the foreign affiliate sales of U.S. firms and the host country’s PRP in any of the selected industries. By contrast, Smarzynska (2004) finds that PRP and FDI are only correlated positively and significantly in patent-sensitive industries.2 Smarzynska (2004) does not examine whether the sensitivity to foreign PRP varies across firms because of firm-specific factors even though her estimations are based on firm-level data.

2 These industries include drugs, cosmetics, health care products, chemicals, machinery and equipment, and electrical equipment. Smarzynska (2004) groups these industries into a single category.

740

T. Ushijima / Research Policy 42 (2013) 738–748

Table 1 The distribution of sample FDI by region. Asia

(China)

Europe

North America

Oceania

Africa

Middle East

Total

1985–1989 1990–1994 1995–1999 2000–2004

490 756 1125 957

(35) (298) (458) (635)

303 318 182 122

380 199 189 111

Latin America 25 37 44 20

27 30 20 12

2 0 12 5

2 3 0 0

1229 1343 1572 1227

Total

3328

(1426)

925

879

126

89

19

5

5371

Note: Asia includes China that is reported in parentheses.

Overall, the literature suggests two avenues for further research in this area. The first is to examine the effect of PRP on the geographical distribution of non-U.S. FDI. The second is to study the heterogeneity of PRP–FDI links across industries and firms. This paper pursues both by estimating the determinants of Japanese FDI location based on microeconomic data. The use of disaggregated data enables me to provide the first evidence on the heterogeneity of PRP–FDI links as a result of firm-specific factors. Specifically, I estimate how a firm’s sensitivity to foreign PRP varies with its technology intensity by matching FDI and patent data at the firm level. 2.2. Technology and Japanese FDI Research has documented that technology is a critical driver of Japanese outward FDI. Industry-level studies such as Kogut and Chang (1991) and Pugel et al. (1996) show that Japanese FDI tends to emanate from R&D-intensive industries. Firm-level studies focusing on either selected industries (Kogut and Chang, 1996; Belderbos and Sleuwaegen, 1996) or a wide range of industries (Berry and Sakakibara, 2008) both report that a firm’s propensity to engage in FDI increases with its R&D intensity. To my knowledge, however, no studies have estimated the effect of PRP on the distribution of Japanese FDI across countries. An important fact highlighted by research on Japanese FDI is that technology can be a “pull factor” of international firm growth. In other words, firms invest abroad not only for the traditional motive of exploiting their own technology but also to obtain technology developed by foreign firms (Branstetter, 2006). Distinguishing these alternative FDI motives is important because strong PRP can repel investors who seek access to foreign technology. Evidence suggests that technology sourcing is a relatively important motive in the acquisition of foreign firms (Blonigen, 1997) and in investment in R&D subsidiaries (Odagiri and Yasuda, 1996). To obtain an accurate estimate of the effects of PRP on FDI that transfers the investing firm’s own technology to a foreign country, my estimations focus on green-field FDI in non-R&D subsidiaries. 3. Data In this paper, I study the geographical distribution of FDI made in 1985–2004 by TSE-listed Japanese industrial firms. I obtain FDI data from Toyo Keizai Shimposha’s Kaigai shinshutsu kigyo soran (Japanese Overseas Investment), a comprehensive directory of foreign subsidiaries operated by Japanese firms. Based on this directory’s annual CD-ROM editions, all subsidiaries satisfying the following conditions are identified: (1) operates in a country for which the regression variables introduced below are fully available; (2) is operated by a manufacturing firm listed on the TSE’s first section; (3) was established in 1985–2004 by green-field investment; (4) has a Japanese parent with an equity stake of no less than 10%; (5) is engaged in the production or distribution of manufactured goods. I focus on TSE-listed firms to obtain most firm-specific variables from financial statements. The third condition excludes acquisitions. The fourth condition excludes portfolio investment.

The last condition excludes service, finance, and R&D subsidiaries. I exclude service and finance subsidiaries because technology transfer from the parent is unlikely to be essential for these operations. R&D subsidiaries are excluded because the acquisition of foreign technology can be the main motive of overseas R&D (Odagiri and Yasuda, 1996). Ultimately, 5378 subsidiaries operating in 58 countries satisfy the above conditions. In this study, I measure FDI discretely as the creation of a new subsidiary. My sample therefore consists of 5378 FDI events. Ideally, one would like to know the size (value) of these FDI events. However, firms rarely disclose the value of individual FDI projects. Because of this constraint, researchers of firm-level FDI normally employ a discrete investment measure as I do in this paper. Two potential biases exist in my estimations. First, to the extent that initial FDI to set up subsidiaries responds more sensitively to the host country’s PRP than do follow-on investments to upscale existing subsidiaries, my estimations focusing on the former can overstate the effect of PRP on total FDI.3 Second, if larger investments involve more intensive technology transfer and are therefore more sensitive to foreign PRP, then my estimations will underestimate the effect of PRP by treating all investments with equal weights. Hopefully, these potential biases cancel out one another, but the net effect is hard to predict. Of the above 5378 subsidiaries, 3652 (68%) are production subsidiaries (which may also perform distribution functions) and 1726 (32%) are distribution subsidiaries that specialize in the sales of manufactured goods. Of these distribution-only subsidiaries, 1212 (70%) are likely to sell only imported products because they belong to firms without a production base in the host country.4 A total of 2955 subsidiaries (55%) are owned 95% or more by the Japanese parent. Table 1 describes the distribution of sample FDI events across regions. It is evident that Asia steadily gained importance as the destination of Japanese FDI. The share of FDI going to Asia increased from 40% in the late 1980s to 78% in the early 2000s. This increase is largely attributable to the emergence of China as the largest recipient country. In 2000–2004, China alone accounts for half of the sample FDI events. Another noteworthy fact is the low ability of developing countries outside Asia to attract Japanese FDI. According to Toyo Keizai, many low-income countries in Africa and Middle East received no investment by TSE-listed firms during the sample period. Estimating the link between a country’s PRP and FDI inflow requires a quantitative measure of PRP. This study employs the PRP index of Ginarte and Park (1997), which is widely used in the academic literature because of its comprehensive coverage of major aspects of national patent laws and its availability for a large number of countries. This index measures the strength of patent laws in five areas: (1) extent of coverage, (2) membership in international patent agreements, (3) provisions for loss of protection,

3 Smarzynska (2004) advances that firms are most sensitive to PRP when making the initial investment decision. 4 Although less likely in practice, another possibility is that these subsidiaries outsource production to local firms.

T. Ushijima / Research Policy 42 (2013) 738–748

(4) enforcement mechanisms, and (5) duration of protection. Each area has subcategories whose score has a sum from zero to one. The Ginarte–Park index, which measures the overall strength of patent laws, is the sum of the scores in these five areas, and ranges from zero (no protection) to five (full protection). Because my sample period of 1985–2004 goes beyond the coverage of original Ginarte–Park index (which goes up to 1990), I employ the updated index described in Park (2008). Within the sample period, the updated Ginarte–Park index is available every five years from 1985 to 2000. Given this data structure, my dependent variable is FDI during the five-year interval [t, t+4] in which t = 1985, 1990, 1995, 2000. 4. Cross-country regressions 4.1. Specification and variables In this paper, I estimate the link between Japanese FDI and foreign PRP with two alternative methods. The first is a cross-country regression based on aggregated data. The second is a logistic regression based on firm-level data. This section documents the analysis based on the former method. The cross-country regression I estimate is similar to the gravity model widely used in international economics to identify the determinants of international trade and investment flows. Although the pure gravity model consists of variables such as country size, income level, and distance, researchers normally incorporate additional variables to estimate factors causing friction in international commodity and capital flows [see Bergstrand and Egger (2011) for a review of the literature]. Because PRP can cause friction in FDI flows, the cross-country regression is a natural empirical framework for the present study. As noted in Section 2, earlier studies based on aggregated data have employed a similar approach to estimate the PRP–FDI link. Recent applications of the gravity model to FDI include Mutti and Grubert (2004), di Giovanni (2005), Bergstrand and Egger (2007), and Eichengreen and Tong (2007) among others. Although the gravity model is normally “strictly descriptive” (Leamer and Levinsohn, 1995), some recent cross-country studies of FDI are underpinned by theory such as the knowledge-capital model, which posits that FDI is a vehicle of MNCs to utilize technology and other valuable knowledge in foreign countries (Markusen, 2002; Carr et al., 2001; Blonigen et al., 2003). Unlike the standard gravity model, my model is specified in the negative binominal (NEGBIN) regression framework and is therefore highly non-linear. This specification is appropriate because my dependent variable, the number of subsidiaries newly established in a country, takes only non-negative integer values. Guimaraes et al. (2003) have shown that an investment count regression model can be derived as a transformed version of the conditional logit model, which is another important method to estimate the country level determinants of FDI location (Coughlin et al., 1991; Head et al., 1995). My specification is therefore closely related to the two most common methods in the FDI location literature. The model assumes that the logged expected count () of FDI country j receives in the five-year period beginning in t is a linear function of the country’s PRP index and control variables (z) in year t and period-specific fixed effects ()5 :

741

institutions, human capital, and the stock of past Japanese FDI, all of which take the value for year t. The first three control variables are standard in the gravity models. The effect of population is expected to be positive as it represents the host country’s market potential. The effect of GDP per capita depends on the dominant motive of FDI. It will be positive if “horizontal FDI” (FDI intended to produce identical goods in multiple locations for local sale) is dominant, and negative if “vertical FDI” (FDI intended to perform vertically linked activities in different locations to save costs) is dominant (Carr et al., 2001). GDP per capita and population measures are obtained from Penn World Table 6.2. The effect of distance is theoretically ambiguous. Empirically, distance tends to be negatively correlated with FDI (Bergstrand and Egger, 2011), perhaps because remote foreign operations are more difficult to manage than proximate ones. Distance can be positively correlated with FDI if firms substitute local production for exporting to save distance-related trade costs. I use the distance between Tokyo and the host country’s capital obtained from the CEPII database. The market orientation of the host country is one of my key control variables and is measured with the economic freedom index published by the Fraser Institute. I include this variable in regressions to control for policy and institutional factors coevolving with PRP as comprehensively as possible. If these factors are uncontrolled, they can seriously confound the direct effect of technology protection, as noted by Maskus (2001) and Lall (2003) among others. The economic freedom index ranges from zero (minimum freedom) to ten (maximum freedom) and measures a country’s market orientation in five major areas: (1) government size; (2) legal structure and security of private ownership; (3) access to sound money; (4) freedom to trade internationally; and (5) regulation of credit, labor, and business. Eaton and Walker (1997) have found that this index is positively and strongly correlated with a country’s steady-state income level. The effect of human capital depends on the type of labor firms seek to employ in foreign countries. It can be negative if foreign subsidiaries mostly perform activities that intensively use unskilled labor, which is relatively scarce in a high-income country such as Japan. I measure human capital with the average years of schooling for the population aged over 15, as estimated by Barro and Lee (2000). The Japanese FDI stock is included to capture various factors, including the agglomeration economies stressed by Head et al. (1995) and unobserved country characteristics. It is the cumulative number of subsidiaries formed by Japanese manufacturers before t. Table 2 reports the descriptive statistics of the regression variables. Unsurprisingly, some independent variables are strongly correlated with one another.6 The correlations are particularly strong among four development-related variables: PRP, economic freedom, GDP per capita, and human capital. Among these variables, the correlation coefficient ranges from 0.65 (economic freedom and human capital) to 0.79 (GDP per capita and human capital). PRP is correlated most strongly with GDP per capita (0.74). Given these correlations, I carefully check the sensitivity of coefficient estimates to model specifications.

ln(jt ) = t + ˇ · PRPjt +  · zjt .

4.2. Estimation results

The control variables include population, GDP per capita, distance from Japan, the market orientation of government policies and

Table 3 reports estimation results based on the sample of 58 countries, all of which had received at least one Japanese FDI by

5 For the sake of simplicity, the subscript t is used to denote both variables in year t and those measured over [t, t+4].

6

The correlation matrix is available from the author upon request.

742

T. Ushijima / Research Policy 42 (2013) 738–748

Table 2 Definition and descriptive statistics of cross-country regression variables. Variable

Definition

n

Mean

Min

Median

All subsidiaries Production subsidiaries Distribution subsidiaries Wholly-owned subsidiaries

# Subsidiaries formed in five years from t # Production subsidiaries formed in five years from t # Distribution subsidiaries formed in five years from t # Wholly-owned (equity stake ≥ 95%) subsidiaries formed in five years from t # Joint ventures (equity stake < 95%) formed in five years from t Patent rights protection index of Ginarte and Park (1997) and Park (2008) Economic freedom index of the Fraser Institute Real GDP per capita (1000 USD, PPP) Average years of schooling of the population aged over 15 Population (million persons) Distance between the capital city and Tokyo (1000 km) Cumulative count of FDI made by Japanese manufacturers before t

227 227 227 227

23.69 16.09 7.604 13.04

65.92 52.85 16.78 35.48

0 0 0 0

3 1 1 1

635 496 139 384

227

10.66

36.14

0

1

347

Joint ventures PRP Economic freedom GDP per capita Human capital Population Distance Japanese FDI stock

SD

Max

227

2.862

1.166

0.588

2.867

4.875

227 227 227 227 227 227

6.364 12.09 7.118 71.10 9.811 68.03

1.292 8.747 2.494 191.2 3.851 146.7

3.023 0.486 2.059 0.241 1.157 0

8.960 5.728 6.960 17.50 9.52 9

9.088 34.37 12.05 1262 18.55 1128

Note: Reported descriptive statistics are for the main sample of 58 countries. All variables are for year t unless otherwise noted.

2004, according to Toyo Keizai.7 Column (1) estimates a specification omitting economic freedom. The coefficient for PRP is positive and highly significant. The estimated coefficient implies that a unit increase of the PRP index, which is approximately 85% of the standard deviation, is on average associated with a 46% increase of investment count. The link between Japanese FDI and foreign PRP is therefore statistically and economically significant. Column (2) incorporates the economic freedom index. The coefficient for this variable is also positive and highly significant, indicating that countries that are more favorably inclined to private business attract more FDI. Note that the coefficient for PRP is virtually unaffected by the introduction of the economic freedom index. This stability suggests that these two variables mostly capture distinct aspects of a country’s policies and institutions when used jointly in regressions. The coefficients for other variables mostly have expected signs in columns (1) and (2). The effect of Japanese FDI stock is positive and highly significant. Consistent with earlier studies, the effects of population and distance are significantly positive and negative, respectively. The coefficient for GDP per capita is negative, which is seemingly at odds with a stylized fact that FDI tends to flow between high-income countries (Shatz and Venables, 2000). Alternative estimations show that this tendency does exist in the present data because the effect of GDP per capita becomes significantly positive when other development-related variables such as human capital and economic freedom are omitted.8 The coefficient for GDP per capita is negative in the reported specifications possibly because it picks up the effect of labor costs when various aspects of economic development are separately controlled by multiple variables. In column (3), estimation is performed on an enlarged sample, which includes an additional 32 countries. These countries, which are predominantly small, less-developed countries, are excluded from the main sample because they appear to be permanently outside the choice set of Japanese firms investing abroad. According to Toyo Keizai, no Japanese firms had established a subsidiary in

these countries before or during the sample period. Because the dependent variable for these countries is constant at zero, these countries can cause an over-dispersion problem because of excess zeros.9 In fact, the dispersion parameter in column (3) increased substantially as a result of the inclusion of these countries, but the coefficient for PRP is significant and is similar to the one based on the main sample. The PRP–FDI link is therefore robust to the choice of estimation sample. Another noteworthy result is that the coefficient for human capital, which is insignificant in columns (1) and (2), is positive and highly significant in column (3). This change suggests that investment in human capital is critically important if a country is to be considered as a potential FDI location by MNCs. Needless to say, a very important event for global intellectual property rights protection during my sample period is the TRIPS agreement, which took effect in 1995 and established minimum protection levels of for all WTO members. After the agreement, the level of PRP increased considerably around the world. The average PRP index of 53 countries, which are the original WTO members in my sample, increased from 2.47 in 1990 to 3.23 in 1995, and then to 3.70 in 2000.10 Given the positive PRP–FDI link identified above, this increase suggests that the TRIPS agreement stimulated Japanese FDI in many parts of the world. However, the effects of the agreement could be greater than predicted by the PRP index because the WTO provided developing countries with a transition period of five or ten years.11 Firms investing in these countries might have based their location decisions on the anticipated level of PRP rather than the current one. To examine this scenario, column (4) focuses on the WTO original member countries and estimates a specification in which the effect of PRP after 1995 is allowed to be different between developed and developing countries.12 Specifically, a dummy variable, with a value of one for developing countries only for the post-TRIPS period (t = 1995 and 2000), is created and interacted with the PRP index. The estimation result suggests that the international location choice by Japanese firms was mostly based on the realized level of PRP because the coefficient for the interaction term is insignificant.

7 These countries include China, which is by far the largest recipient country of Japanese FDI, as reported in the last section. Estimations excluding China generate substantially the same results. 8 To examine the sensitivity of estimation results to the correlations among PRP, economic freedom, GDP per capita, and human capital, I estimated specifications that include only one of these variables in addition to population, distance, and Japanese FDI stock. The estimated coefficient (associated standard error) is 0.393 (0.098) for PRP, 0.476 (0.087) for economic freedom, 0.028 (0.012) for GDP per capita, and 0.143 (0.045) for human capital.

9 The ratio of observations whose dependent variable is zero is 32% in the main sample of 58 countries and 56% in the enlarged sample of 90 countries. 10 Maskus (2001) notes that a country’s PRP level measured by the Ginarte–Park index must be at least 3.0 to be consistent with the TRIPS agreement. 11 This transition period for the full compliance with the TRIPS agreement was provided only to the original WTO members and initially limited to five years for developing countries and ten years for the least-developed countries. 12 Developing countries are as defined by the WTO. Separating developing and the least developed countries do not change the conclusion below.

Table 3 Cross-country NEGBIN regressions of Japanese FDI. All subsidiaries (1)

All subsidiaries (2)

All subsidiaries (3)

All subsidiaries (4)

Production subsidiaries (5)

Distribution subsidiaries (6)

Wholly-owned subsidiaries (V)

Joint ventures (8)

Patent rights protection (PRP) PRP × developing (t ≥ 1995) PRP × high imitative ability PRP × low imitative ability Economic freedom (EF)

0.464*** (0.136)

0.425*** (0.140)

0.379** (0.166)

0.420*** (0.142) −0.100 (0.096)

0.488*** (0.147)

0.351** (0.144)

0.499*** (0.144)

0.338** (0.146)

EF × high imitative ability EF × low imitative ability GDP per capita Human capital Population Distance Japanese FDI stock Dispersion parameter (˛) Log likelihood # Observations # Countries

0.408*** (0.120)

0.421*** (0.126)

0.372*** (0.119)

0.441*** (0.128)

0.316** (0.143)

0.431*** (0.129)

All subsidiaries (9)

All subsidiaries (10)

0.642*** (0.153) 0.215 (0.196)

0.445* (0.242) −0.161 (0.173)

0.520*** (0.149) 0.591*** (0.125) −0.094*** (0.018) 0.022 (0.068) 0.002** (0.001) −0.104*** (0.032) 0.010*** (0.002)

0.400*** (0.126)

−0.031* (0.017) 0.070 (0.057) 0.002*** (0.001) −0.104*** (0.032) 0.010*** (0.002)

−0.056*** (0.019) 0.049 (0.056) 0.003*** (0.001) −0.071** (0.030) 0.009*** (0.002)

−0.038* (0.022) 0.240*** (0.054) 0.005*** (0.001) −0.060** (0.030) 0.013*** (0.002)

−0.072*** (0.025) 0.074 (0.063) 0.003*** (0.001) −0.073*** (0.026) 0.010*** (0.001)

−0.092*** (0.022) 0.030 (0.062) 0.003*** (0.001) −0.108*** (0.034) 0.009*** (0.002)

−0.002 (0.020) 0.038 (0.063) 0.001*** (0.000) −0.024 (0.033) 0.009*** (0.002)

−0.054** (0.021) 0.068 (0.061) 0.002*** (0.001) −0.038 (0.033) 0.010*** (0.002)

−0.069*** (0.021) −0.007 (0.059) 0.003*** (0.000) −0.136*** (0.033) 0.007*** (0.002)

0.328*** (0.119) 0.533*** (0.145) −0.058** (0.023) 0.065 (0.056) 0.003*** (0.001) −0.063** (0.030) 0.009*** (0.002)

1.558

1.446

2.281

1.366

1.470

1.282

1.492

1.235

1.409

1.288

−673.4 227 58

−667.8 227 58

−726.9 348 90

−621.2 210 53

−553.0 227 58

−502.1 227 58

−560.2 227 58

−502.0 227 58

−665.8 227 58

−656.0 227 58

T. Ushijima / Research Policy 42 (2013) 738–748

Dependent variable

Note: In parentheses are robust standard errors. All regressions include period dummies. * Significant at the 0.10 level. ** Significant at the 0.05 level. *** Significant at the 0.01 level.

743

744

T. Ushijima / Research Policy 42 (2013) 738–748

The next four columns disaggregate the dependent variable by the type of subsidiary created by FDI. Columns (5) and (6) focus on investments in production and distribution subsidiaries, respectively.13 Not surprisingly, FDI for local production responds sensitively to the host country’s PRP. The effect of PRP on distribution FDI is smaller but also positive and significant. Mansfield (1994) finds that setting up a distribution subsidiary normally involves few technology transfers from the parent. Distribution FDI can nevertheless be sensitive to PRP because of the vertical link of production and distribution activities and reverse engineering, which can cause technology spillovers even from imported products.14 The role of reverse engineering is likely to be large because a majority of distribution subsidiaries in my sample only sell imported goods, as I noted earlier.15 Columns (7) and (8) separate wholly owned subsidiaries in which the Japanese parent holds an equity stake of at least 95% from joint ventures (equity stake < 95%). The effect of PRP is smaller for joint ventures but positive and significant for the both types of FDI. Overall, the estimation results in columns (5)–(8) suggest that PRP is an important determinant of the location of various types of FDI. Given this finding, I focus on total FDI in the rest of this paper. Columns (9) and (10) examine the heterogeneity of the PRP–FDI link across countries, which is of interest because weak PRP should be a major barrier to FDI only if the host country is able to imitate advanced foreign technology. Consistent with this view, Smith (2001) finds that PRP affects foreign affiliate sales of US firms only in countries with a high imitative ability. Column (9) performs a similar test by estimating a specification in which the PRP and economic freedom indices can take different coefficients for countries with high and low imitative abilities. A country’s imitative ability is defined as high if the per capita number of scientific and engineering journal articles published in year t is above the median and is defined as low otherwise.16 I find that the effect of PRP is positive and significant only for countries with high imitative ability. Qualitatively the same result is obtained in column (10), in which countries are divided based on the number of scientific and engineering articles not normalized by population. The present data therefore robustly shows that the PRP–FDI link arises only where weak PRP substantially increases the risk of unintended technology spillovers. Note also that, unlike PRP, economic freedom exerts qualitatively the same effect on Japanese FDI regardless of the host country’s technological ability.

report the coefficients for the PRP and economic freedom indices and the associated standard errors to conserve space. In columns (1) and (2), the mean R&D and patent intensities of firms in each industry are reported.18 Columns (3) and (4) report the estimation results of the baseline model, assuming constant coefficients on the PRP and economic freedom indices. The effect of economic freedom is positive and significant in all industries. By contrast, the effect of PRP is positive and significant in only nine industries, which are relatively R&D- and patent-intensive, as shown in columns (1) and (2).19 FDI is most sensitive to PRP in the chemical industry (including pharmaceuticals). Columns (5)–(8) report the estimation results of the specification, assuming separate coefficients on the PRP and economic freedom indices for high- and low-imitative-ability countries, which are segmented at the median per capita number of scientific and engineering journal articles. I find that in eight industries with a positive and significant PRP–FDI link, the link is specific to high-imitative-ability countries in which investors will likely face a high risk of technology spillovers if they are not protected by strong patent laws. The effect of economic freedom, which is positive and significant in all industries, does not exhibit such asymmetry. If anything, my results suggest that the ability of economic freedom to attract FDI is stronger for countries with low imitative ability. Taken together, aggregate estimations performed in this section suggest that a strong PRP–FDI link arises when (1) technology is an important asset of investing firms and (2) firms in the host country are technologically sophisticated enough to imitate advanced foreign technology transferred via FDI. 5. Firm-level regressions 5.1. Specification In this section, I provide firm-level evidence for the link between Japanese FDI and foreign PRP with particular attention to the role of the investing firm’s innovative activities. The dependent variable is the dummy variable dijt , which is coded as one if firm i invests in country j by forming a subsidiary in the five-year period beginning in t (t = 1985, 1990, 1995, 2000) and is coded as zero otherwise. The baseline model assumes that this variable has a value of one with the probability given by the following logistic distribution function:

4.3. Industry-level regressions P(dijt = 1) = This subsection examines the variation of the PRP–FDI link across 2-digit industries by performing fifteen separate regressions on the number of newly formed subsidiaries in each industry.17 The NEGBIN estimation results reported in Table 4 are based on the main sample of 58 countries. Although the model includes the full set of control variables employed in the last subsection, I only

13 As noted in the last section, production subsidiaries include subsidiaries that perform distribution as well as production activities. These dual-function subsidiaries are not included in distribution subsidiaries. 14 There is substantial evidence that bilateral trade flows are sensitive to the importing country’s PRP (Maskus and Penubarti, 1995; Smith, 1999). 15 I examined whether the PRP-sensitivity of distribution FDI differs between firms operating a production subsidiary in the host country and firms without a local production base. Unreported regressions show that the sensitivity is comparable between the two sets of firms. Specifically, the estimated coefficient for PRP (standard error) is 0.392 (0.237) for firms with a production subsidiary and 0.377 (0.148) for firms without a production subsidiary. 16 Data are taken from US National Science Foundation’s Science and Engineering Indicators. 17 The creation of an industry-specific FDI count is based on Toyo Keizai, which assigns each subsidiary to a 2-digit industry.

exp(t + h + ˇ · PRPjt +  · zjt + ı · xit ) 1 + exp(t + h + ˇ · PRPjt +  · zjt + ı · xit )

,

in which  is the period-specific fixed effect,  is the industryspecific fixed effect (2-digit), z is a vector of country-level variables other than PRP, and x is a vector of firm characteristics. The firmspecific variables include R&D and patent intensities, firm size (total assets), profitability (operating income/total assets), leverage (liabilities/total assets), and previous investment (the cumulative number of subsidiaries the firm had formed in the focal country by year t). It is important to note that the above model is not a disaggregated version of the NEGBIN model estimated in the last section. Transforming an FDI count model by disaggregating the dependent variable into FDI projects results in the conditional logit model mentioned earlier (Guimaraes et al., 2003). I do not employ the

18 The average is taken over the entire sample period. See the next section for the definition of R&D and patent intensities. 19 The R&D and patent intensities are significantly greater for firms in the nine PRP-sensitive industries than for firms in the six PRP-insensitive industries at the 0.05 or higher level.

T. Ushijima / Research Policy 42 (2013) 738–748

745

Table 4 NEGBIN regressions of Japanese FDI by industry. R&D intensity (%) Patent intensity Patent rights protection

Economic freedom

Patent rights protection

Economic freedom

(1)

(2)

(3)

(4)

High imitative ability (5)

Low imitative ability (6)

High imitative ability (7)

Low imitative ability (8)

Foods

0.9

0.6

Textile products

1.1

1.5

Paper and pulp

1.1

1.1

Chemicals

3.4

2.5

Petroleum products

0.6

0.5

Rubber products

2.5

2.5

Ceramic products

1.9

2.1

Iron and steel

1.3

1.0

Non-ferrous metal

1.3

1.1

Metal products

2.0

2.5

Machineries

2.1

2.4

Electric machineries

2.9

3.3

Transportation equipment 2.5

1.8

Precision instruments

3.0

4.2

Other manufacturing

1.9

1.4

0.077 (0.212) 0.057 (0.349) −0.063 (0.347) 0.805*** (0.190) 0.336 (0.362) 0.530** (0.265) 0.676** (0.300) 0.709** (0.331) 0.724** (0.320) 0.110 (0.278) 0.389** (0.173) 0.318** (0.137) 0.326** (0.158) 0.334* (0.200) −0.299 (0.247)

0.693*** (0.182) 0.390* (0.232) 1.250*** (0.238) 0.691*** (0.184) 1.599*** (0.474) 0.951*** (0.238) 0.926*** (0.195) 1.241*** (0.336) 1.367*** (0.300) 0.749*** (0.165) 0.254* (0.154) 0.536*** (0.141) 0.275* (0.152) 0.471*** (0.152) 0.547** (0.229)

0.386 (0.241) 0.472 (0.420) −0.212 (0.393) 1.129*** (0.209) 0.668 (0.449) 0.863*** (0.303) 1.167*** (0.326) 0.698* (0.381) 1.146*** (0.356) 0.458 (0.294) 0.612*** (0.190) 0.614*** (0.161) 0.602*** (0.186) 0.335 (0.263) −0.113 (0.288)

−0.534* (0.300) −0.649* (0.390) 0.309 (0.437) 0.371 (0.251) −0.384 (0.485) −0.017 (0.319) −0.076 (0.373) 0.700 (0.528) 0.198 (0.392) −0.369 (0.355) 0.047 (0.220) −0.221 (0.179) 0.085 (0.235) 0.357* (0.214) −0.584 (0.389)

0.597*** (0.203) 0.284 (0.243) 1.285*** (0.241) 0.628*** (0.183) 1.448*** (0.453) 1.062*** (0.225) 0.927*** (0.189) 1.259*** (0.314) 1.226*** (0.313) 0.685*** (0.182) 0.183 (0.163) 0.465*** (0.150) 0.174 (0.159) 0.446*** (0.164) 0.466* (0.255)

0.881*** (0.201) 0.634*** (0.234) 1.097*** (0.267) 0.937*** (0.221) 1.736*** (0.508) 1.160*** (0.269) 1.249*** (0.203) 1.230*** (0.361) 1.618*** (0.339) 0.930*** (0.196) 0.399** (0.176) 0.745*** (0.171) 0.405** (0.176) 0.473*** (0.182) 0.657** (0.270)

Note: R&D and patent intensities are the average of all firms in the industry over 1985–2004. In parentheses are robust standard errors. * Significant at the 0.10 level. ** Significant at the 0.05 level. *** Significant at the 0.01 level.

conditional logit specification because it is unable to incorporate firm-specific determinants of FDI location, a critical weakness to the present study. The above ordinary logit specification is more ad hoc, but overcomes this limitation. A cost of using the above specification that takes firms as the observation unit is that the correspondence between the dependent and independent variables can be weak for firms investing outside their main industry (e.g., a chemical firm investing in a textile mill to produce synthetic fibers).20 I speculate, however, that the bias as a result of this measurement error is not serious, if any exists at all because the share of FDI outside the investing firm’s main industry is less than 10% in the present data. R&D intensity is defined as R&D expense normalized by total assets and averaged over five years beginning in t. Patent intensity is patent stock in year t normalized by total assets in the same year. To create this variable, I retrieved data of all patents granted to sample firms by Japan Patent Office from the Institute of Intellectual Property’s database.21 Based on the perpetual inventory method, a firm’s patent stock was then estimated as the capitalized number of claim-weighted patents granted to the firm per year (depreciation rate = 15%). I use the claim-weighted patent count because the number of claims per patent increased considerably in Japan

20 I appreciate an anonymous referee for bringing my attention to this issue. Taking segment (firm-industry) as the observation unit mitigates this problem. Unfortunately, the availability of segment-level data is very limited for Japanese firms during my sample period. 21 See Goto and Motohashi (2007) for a detailed description of this database.

after patent reforms in 1988 (Sakakibara and Branstetter, 2001). The weight of a patent is unlikely to be constant over the sample period. Although the above baseline specification assumes that the sensitivity to foreign PRP (ˇ) is identical for all firms, my main specification sheds light on the heterogeneity of ˇ both between and within industries by sorting firms in each 2-digit industry into quartiles based on the R&D and patent intensities in each period. Using dummy variables for these quartiles, the specification describes the variations of the PRP–FDI link across firms as follows:

ˇit = ˇh +

4  k=2

k ˇrq rqkit +

4 

k ˇpq pqkit ,

k=2

in which ˇh is the PRP-sensitivity of the least R&D- and patentintensive firms in industry h. {rq} and {pq} are a set of dummy variables for firms whose R&D and patent intensities are in the kth quartile of the industry (k = 2, 3, 4). This specification captures the variation of PRP-sensitivity across firms in the same industry due to differences in innovation input (R&D expense) and output (patent). Table 5 provides the descriptive statistics of regression variables. Although I do not report the correlation matrix, the correlations of firm-level variables are moderate. Among the most strongly correlated variables are firm size and previous investment (correlation coefficient = 0.20) and R&D and patent intensities (0.20). The estimation results reported below are robust to the inclusion/exclusion of these variables.

746

T. Ushijima / Research Policy 42 (2013) 738–748

Table 5 Definition and descriptive statistics of logistic regression var. Variable

Definition

FDI dummy

1 if investing in the focal country during in five years from t, 0 otherwise See Table 2 See Table 2 See Table 2 See Table 2 See Table 2 See Table 2 See Table 2 R&D expense/total assets averaged over five years from t Patent stock/total assets (billion yen) in t Total assets (trillion yen) in t Operating income before depreciation/sales averaged over five years from t Liabilities/assets in t Number of subsidiaries a firm has formed in the focal country before t

PRP Economic freedom GDP per capita Human capital Population Distance Japanese FDI stock R&D intensity Patent intensity Firm size Profitability Leverage Previous investment

n

Mean

SD

Min

Median

Max

151,519

0.027

0.162

151,519 151,519 151,519 151,519 151,519 151,519 151,519 151,519 151,519 151,519 151,519

2.856 6.391 12.33 7.198 70.64 9.791 116.4 0.024 2.407 0.227 0.056

1.172 1.264 8.603 2.439 191.0 3.837 266.5 0.018 3.556 0.00 0.032

0.588 3.023 1.014 2.059 0.241 1.157 0 0 0 0.001 −0.164

2.892 6.508 9.526 7.047 16.02 9.518 18 0.020 1.256 0.077 0.051

4.875 9.088 34.37 12.05 1262 18.55 2432 0.194 47.13 7.775 0.222

151,519 151,519

0.570 0.096

0.242 0.469

0.045 0

0.573 0

8.353 27

0

0

1

Note: All variables are for year t unless otherwise noted. Firm-specific variables are obtained from Nikkei NEEDS database except for patent stock obtained from the Institute of Intellectual Property’s patent database and previous investment obtained from Toyo Keizai database.

5.2. Estimation results Table 6 reports estimation results. The estimation sample has 151,519 observations based on 698 firms having engaged in at least one FDI in 1985–2004 and 58 countries having received at least one Japanese FDI by the end of sample period.22 Column (1) reports the estimation result of baseline specification. The coefficients for PRP and other country-level variables are qualitatively unchanged from those obtained in aggregate estimations. This similarity is unsurprising because the present model describes the FDI behavior of the same firms in the same countries during the same period studied by the NEGBIN models in the last section. All of the firm-level variables are significant. The coefficients for R&D and patent intensities are both positive, which is consistent with earlier evidence that technology is a critical driver of Japanese outward FDI (Kogut and Chang, 1996; Berry and Sakakibara, 2008). The specification in column (2) separates the effects of PRP for high- and low-imitative-ability countries segmented at the median per capita publication of scientific and engineering journal articles. Unlike in earlier estimations based on aggregate data, the effect of PRP is positive and significant (although only marginally, with a p-value of 0.096), even for countries with low imitative ability. As in earlier estimations, however, the effect of PRP is considerably larger for countries with high imitative ability. The hypothesis that the effect of PRP is identical for high- and low-imitative-ability countries is strongly rejected by a Wald test (p = 0.010). Column (3) sheds light on the heterogeneity of the PRP–FDI link across firms based on the main specification. The fifteen industryspecific estimates of ˇh , the PRP-sensitivity of the least R&D- and patent-intensive firms, are not reported to conserve space. Instead, I report the simple average and standard deviation (in brackets) of these coefficients.23 A Wald test rejects the hypothesis that the averaged ˇh is zero (p = 0.014), suggesting that even the least technology-intensive firms in many industries normally consider foreign PRP when deciding where to locate a new subsidiary. A Wald test also rejects the hypothesis that ˇh is identical for all industries (p = 0.001). The main result in column (3) is that the sensitivity to foreign PRP varies not only between but also within industries. Specifically,

22 The sample size is smaller than the theoretical maximum (161,936 = 698 firms × 58 countries × 4 periods) because of firms and countries entering and exiting the sample. 23 The individual estimates of ˇh are available from the author upon request.

the interaction effects of PRP and patent intensity quartile dummies are all positive and significant. Moreover, the size of these effects is larger for higher quartiles. We therefore have evidence that firms depending more on patents relative to industry peers to secure monopolies of innovations respond more sensitively to foreign PRP in deciding where to invest. By contrast, all of the interaction effects of PRP and R&D intensity dummies differ insignificantly from zero. This may be because patent stock is a more direct measure of technology protected by patent laws and/or because the R&D expenses reported by Japanese firms was a noisy measure of their innovative activities until accounting standards reforms in the late 1990s.24 Columns (4) and (5) separate firms based on whether industrylevel estimations in the last section have identified a positive and significant PRP–FDI link for their industry. The industry-specific estimates of ˇh are again averaged across industries for reporting. As reported in column (4), the estimation results for firms in the nine PRP-sensitive industries are similar to the full-sample estimation results. Column (5) focuses on firms in the six PRPinsensitive industries. The averaged PRP-sensitivity of the least R&D- and patent-intensive firms is insignificant in these industries. However, the estimated interaction effects of the PRP index and patent intensity dummies broadly suggest that a firm’s sensitivity to foreign PRP increases with its patent intensity even in these industries. The PRP-sensitivity of FDI is therefore a function of the investing firm’s technology strategy as well as industry characteristics. Columns (6) and (7) perform another set of subsample estimations. Smarzynska (2004) advances the argument that the effect of current PRP is mitigated for sequential FDI, which is stimulated by the firm’s previous investments made in response to the host country’s earlier policies and institutions. Because my regressions show that a firm’s likelihood to invest in a country increases with its stock of past investments in the same country, it is of interest whether PRP exerts differential effects on the initial and sequential FDI of Japanese firms. In columns (6) and (7), I separate firms into initial investors (no previous investment in the focal country) and

24 For instance, some prominent firms clearly investing in R&D, such as Hitachi and Toyota, reported no R&D expense. Because these firms’ true R&D expense before the late 1990s is unknowable, my data assume that it is zero, as reported in their financial statements. Removing zero R&D firms from regressions hardly affects the estimation results because their share in total observations is only 1%. Griliches and Mairesse (1990) provide a detailed discussion of the problems of Japanese R&D reporting before the accounting standards reforms.

T. Ushijima / Research Policy 42 (2013) 738–748

747

Table 6 Logistic estimations of a firm’s likelihood to invest in a country. Estimation sample

PRP

Full

Full

Full

(1)

(2)

0.445*** (0.141)

PRP × high imitative ability

PRP × Q2 (R&D intensity)

R&D intensity Patent intensity Firm size Profitability Leverage Previous investment Mean dependent variable Log likelihood # Observations

0.194 [0.196]

0.271** [0.196]

0.160 [0.270]

−0.004 (0.023) 0.022 (0.022) 0.015 (0.028) 0.027 (0.020) 0.086*** (0.022) 0.092*** (0.026) 0.689*** (0.138) −0.081*** (0.027) −0.036 (0.057) 0.002*** (0.000) −0.194*** (0.028) 0.001*** (0.000) 1.119 (1.687) −0.019 (0.011) 0.468*** (0.029) 2.806*** (0.726) * 0.261*** (0.106) 0.255*** (0.035)

0.085* (0.046) 0.072* (0.042) 0.046 (0.066) 0.088** (0.039) 0.083* (0.044) 0.171*** (0.046) 0.759*** (0.160) −0.075*** (0.021) −0.008 (0.070) 0.002*** (0.000) −0.216*** (0.034) 0.001*** (0.000) 2.707 (5.806) 0.052*** (0.011) 1.042*** (0.110) −0.306 (1.278) −0.396 (0.257) 0.298*** (0.093)

−0.001 (0.021) 0.017 (0.023) 0.008 (0.029) 0.052** (0.039) 0.101*** (0.021) 0.122*** (0.046) 0.674*** (0.119) −0.069*** (0.019) −0.020 (0.056) 0.002*** (0.000) −0.216*** (0.028) 0.002*** (0.000) 2.381 (1.846) −0.015 (0.015) 0.545*** (0.033) 2.266*** (0.836) −0.525*** (0.112)

0.013 (0.038) 0.037 (0.039) 0.060 (0.051) −0.002 (0.024) 0.052 (0.032) 0.063* (0.035) 0.238* (0.143) −0.055*** (0.021) −0.071 (0.072) 0.002*** (0.000) −0.194*** (0.027) 0.001*** (0.000) −2.031 (3.058) 0.004 (0.015) 0.287*** (0.036) 2.814*** (1.487) 0.261*** (0.097) 0.255*** (0.035)

0.029 −11,263.2 115,879

0.021 −2568.7 35,640

0.019 −10,198.0 141,523

0.142 −3348.4 9996

0.027 −13,855.6 151,519

0.027 −13,867.5 151,519

PRP × Q4 (patent intensity)

Japanese FDI stock

0.398*** [0.093]

0.027 −13,911.8 151,519

PRP × Q3 (patent intensity)

Distance

0.358** [0.140]

0.787*** (0.119) −0.130*** (0.027) −0.034 (0.056) 0.002*** (0.000) −0.216*** (0.028) 0.002*** (0.000) 2.498* (1.385) 0.014** (0.007) 0.477*** (0.029) 2.185*** (0.615) −0.213** (0.090) 0.286*** (0.042)

PRP × Q2 (patent intensity)

Population

Sequential investment (1)

0.698*** (0.139) −0.079*** (0.021) −0.031 (0.058) 0.002*** (0.000) −0.193*** (0.028) 0.001*** (0.000) 2.508* (1.388) 0.015** (0.007) 0.475*** (0.028) 2197*** (0.617) −0.213** (0.090) 0.288*** (0.043)

PRP × Q4 (R&D intensity)

Human capital

Initial investment (6)

0.010 (0.021) 0.030 (0.022) 0.029 (0.027) 0.033* (0.018) 0.083*** (0.021) 0.106*** (0.025) 0.700*** (0.139) −0.080*** (0.021) −0.031 (0.058) 0.002*** (0.000) −0.194*** (0.028) 0.001*** (0.000) 0.869 (1.779) −0.010 (0.009) 0.481*** (0.029) 1.989*** (0.627) −0.273*** (0.097) 0.286*** (0.041)

PRP × Q3 (R&D intensity)

GDP per capita

PRP-insensitive industries (5)

0.561*** (0.134) 0.266* (0.160)

PRP × low imitative ability

Economic freedom

(3)

PRP-sensitive industries (4)

Note: In parentheses are standard errors clustered by country and period. In columns (3)–(7), ˇh that is averaged over industries is reported in the place of PRP coefficient. In brackets are the standard deviation of ˇh . Significance level of the averaged ˇh is based on the Wald test. All regressions include period and industry dummies. * Significant at the 0.10 level. ** Significant at the 0.05 level. *** Significant at the 0.01 level.

sequential investors (previous investment > 0).25 Consistent with Smarzynska (2004), I find that PRP mostly affects initial FDI because the effect of PRP is significantly positive only in column (6) which focuses on firms with no previous investment in the focal country. This result suggests that the strengthening of PRP increases FDI inflow into a country mainly by expanding the extensive margin of investing firms rather than stimulating additional investment on the part of incumbent firms. Overall, the above results suggest that the PRP–FDI link arises because patent laws block or slow the competitive diffusion of the investing firm’s proprietary technology transferred via FDI. The

25 Previous investment is omitted from the estimation of initial investment decision because the variable is constant at zero for all observations. The size of the sequential investment subsample is much smaller than that of the initial investment subsample because firms typically have no earlier investment in most countries.

finding that a firm’s sensitivity to foreign PRP increases with its patent intensity is particularly important as it is the first direct evidence on the role of patents in the formation of PRP–FDI links. 6. Conclusion This paper provides detailed evidence for the effects of PRP on FDI inflows, based on Japanese data over two decades. I find an economically and statistically significant link between the Japanese firm’s FDI and foreign PRP. Moreover, I find substantial evidence that the link is heterogeneous across countries, industries, and firms. My major findings are summarized as follows: (1) The positive PRP–FDI link concentrates in countries with a high ability to imitate foreign technology. In technologically lesssophisticated countries, the link disappears.

748

T. Ushijima / Research Policy 42 (2013) 738–748

(2) The link with foreign PRP is positive and significant only for FDI in technology-intensive industries, such as chemical and electric machinery. (3) The sensitivity of a firm’s FDI to foreign PRP increases with its patent intensity relative to industry peers. The effect of PRP diminishes substantially when a firm has previous investment experience in the same country. These patterns suggest that the strengthening of PRP increases FDI inflow into a country as it mitigates investors’ concern about unintended spillovers of proprietary technology. They also suggest that PRP affects the quality as well as the quantity of FDI inflow because PRP mainly affects investment by innovative firms in technology-intensive industries. It is important to recognize that the effect of PRP to attract FDI varies considerably across countries. The strengthening of PRP can increase neither FDI inflow nor innovations by domestic firms if a country’s innovative (imitative) ability is low. Governments must carefully weigh the cost and benefit of PRP in light of their country’s technological capability and needs and the nature of foreign investment they wish to attract. Acknowledgements I am grateful to two anonymous referees, Ryuhei Wakasugi, Eiichi Tomiura, Hiroshi Ohashi, Banri Ito, and Takeo Hoshi for valuable comments. Walter Park and the Research Institute of Economy, Trade, and Industry (RIETI) generously provided data used in this research. The usual disclaimer applies. References Barro, R.J., Lee, J.W., 2000. International data on educational attainment: updates and implications. Center for International Development Working Paper 42. Belderbos, R., Sleuwaegen, L., 1996. Japanese firms and the decision to invest abroad: business groups and regional core networks. Review of Economics and Statistics 78, 214–220. Bergstrand, J.H., Egger, P., 2007. A knowledge-and-physical-capital model of international trade flows, foreign direct investment, and multinational enterprises. Journal of International Economics 73, 278–308. Bergstrand, J.H., Egger, P., 2011. Gravity equations and economic frictions in the world economy. In: Bernhofen, D., Falvey, R., Kreickemeier, U. (Eds.), Palgrave Handbook of International Trade. Palgrave Macmillan, New York. Berry, H., Sakakibara, M., 2008. Resource accumulation and overseas expansion by Japanese multinationals. Journal of Economic Behavior & Organization 65, 277–302. Blonigen, B.A., 1997. Firm-specific assets and the link between exchange rates and foreign direct investment. American Economic Review 87, 447–465. Blonigen, B.A., Davies, R.B., Head, K., 2003. Estimating the knowledge-capital model of the multinational enterprise: comment. American Economic Review 93, 980–994. Branstetter, L., 2006. Is foreign direct investment a channel of knowledge spillovers? Evidence from Japan’s FDI in the United States. Journal of International Economics 68, 325–344. Carr, D.L., Markusen, J.R., Maskus, K.E., 2001. Estimating the knowledge-capital model of the multinational enterprise. American Economic Review 91, 693–708. Coughlin, C.C., Terza, J.V., Arromdee, V., 1991. State characteristics and the location of foreign direct investment within the United States. Review of Economics and Statistics 73, 675–683. di Giovanni, J., 2005. What drives capital flows? The case of cross-border M&A activity and financial deepening. Journal of International Economics 65, 127–149. Dunning, J.H., 1993. Multinational Enterprises and the Global Economy. AddisonWesley Publishers, Harlow, England. Eaton, S.T., Walker, M.A., 1997. Income, growth, and economic freedom. American Economic Review 87, 328–332.

Eichengreen, B., Tong, H., 2007. Is China’s FDI coming at the expense of other countries? Journal of the Japanese and International Economies 21, 153–172. Fink, C., 2005. Intellectual property rights and U.S. and German international transactions in manufacturing industries. In: Fink, C., Maskus, K.E. (Eds.), Intellectual Property and Development: Lessons from Recent Economic Research. Oxford University Press, Oxford. Ginarte, J.C., Park, W.G., 1997. Determinants of patent rights: a cross-national study. Research Policy 26, 283–301. Goto, A., Motohashi, K., 2007. Construction of a Japanese Patent Database and a first look at Japanese patenting activities. Research Policy 36, 1431–1442. Griliches, Z., Mairesse, J., 1990. R&D and productivity growth: comparing Japanese and U.S. manufacturing firms. In: Hulten, C.R. (Ed.), Productivity Growth in Japan and the United States. University of Chicago Press, Chicago, IL. Guimaraes, P., Figueirdo, O., Woodward, D., 2003. A tractable approach to the firm location decision problem. Review of Economics and Statistics 85, 201–204. Hall, B., Jaffe, A., Trajtenberg, M., 2005. Market value and patent citations. Rand Journal of Economics 36, 16–38. Head, K., Ries, J., Swenson, D., 1995. Agglomeration benefits and location choice: evidence from Japanese manufacturing investments in the United States. Journal of International Economics 38, 223–247. Kogut, B., Chang, S.J., 1991. Technological capabilities and Japanese foreign direct investment in the United States. Review of Economics and Statistics 73, 401–413. Kogut, B., Chang, S.J., 1996. Platform investments and volatile exchange rates: direct investment in the U.S. by Japanese electronic companies. Review of Economics and Statistics 78, 221–231. Lall, S., 2003. Indicators of the relative importance of IPRs in developing countries. Research Policy 32, 1657–1680. Leamer, E.E., Levinsohn, J., 1995. International trade theory: the evidence. In: Grossman, G., Rogoff, W. (Eds.), Handbook of International Economics, vol. III. Amsterdam, North-Holland. Lee, J.Y., Mansfield, E., 1996. Intellectual property protection and U.S. foreign direct investment. Review of Economics and Statistics 78, 181–186. Levin, R.C., Klevorick, A.K., Nelson, R.R., Winter, S.G., 1987. Appropriating returns from industrial research and development. Brookings Papers on Economic Activity 3, 783–820. Mansfield, E., 1994. Intellectual property protection, foreign direct investment and technology transfer. International Finance Corporation Discussion Paper. Mansfield, E., 2000. Intellectual property protection, direct investment and technology transfer: Germany, Japan, and the USA. International Journal of Technology Management 19, 3–21. Markusen, J.R., 2002. Multinational Firms and the Theory of International Trade. MIT Press, Cambridge, MA. Maskus, K.E., 1998. The international regulation of intellectual property. Weltwirtschaftliches Archiv 134, 186–208. Maskus, K.E., 2001. IPRs and FDI. In: Bora, B. (Ed.), Foreign Direct Investment: Research Issues. Rutledge, London. Maskus, K.E., Penubarti, M., 1995. How trade-related are intellectual property rights? Journal of International Economics 39, 227–248. Mutti, J., Grubert, H., 2004. Empirical asymmetries in foreign direct investment and taxation. Journal of International Economics 62, 337–358. Nunnenkamp, P., Spatz, J., 2004. Intellectual property rights and foreign direct investment: a disaggregated analysis. Weltwirtschaftliches Archiv 140, 393–414. Odagiri, H., Yasuda, H., 1996. The determinants of overseas R&D by Japanese firms: an empirical study at the industry and company levels. Research Policy 25, 1059–1079. Park, W.G., 2008. International patent protection: 1960–2005. Research Policy 37, 761–766. Primo Braga, C., Fink, C., 1998. The relationship between intellectual property rights and foreign direct investment. Duke Journal of Comparative and International Law 9, 163–187. Pugel, T.A., Kragas, E.S., Kimura, Y., 1996. Further evidence on Japanese direct investment in U.S. manufacturing. Review of Economics and Statistics 78, 208–213. Sakakibara, M., Branstetter, L., 2001. Do stronger patents induce more innovation? Evidence from the 1988 Japanese patent law reforms. RAND Journal of Economics 32, 77–100. Shatz, H.J., Venables, A.J., 2000. The geography of international investment. World Bank Policy Research Working Paper Series 2338. Smith, P.J., 1999. Are weak patent rights a barrier to U.S. exports? Journal of International Economics 48, 151–177. Smith, P.J., 2001. How do foreign patent rights affect U.S. exports, affiliate sales and licenses? Journal of International Economics 55, 411–439. Smarzynska, B., 2004. The composition of foreign direct investment and protection of intellectual property rights: evidence from transition economies. European Economic Review 48, 39–62. The World Bank, 2002. Global Economic Prospects and the Developing Countries. The World Bank, Washington, DC.