FDI, exports and export spillover in Taiwan’s electronics industry

Abstract

This study primarily examines whether and how a firm’s exports are influenced by its own and other firms’ foreign direct investment (FDI) in an industry characterized by high proportions of exporters and FDI. A review of 619 firms in Taiwan’s electronics industry from 2006 to 2011 indicates the presence of optimal FDI in terms of exports. In addition, results suggest a negative (positive) export spillover from FDI for firms with low (high) FDI, signifying that FDI’s congestion/crowding-out effect is greater (less) than the demonstration/imitation effect for such firms. Moreover, a firm’s exports generally decrease with other firms’ exports particularly when its FDI is low, thus suggesting the predominance of competition’s negative effect over its positive effect under such circumstances. Furthermore, export spillover from FDI occurs through the competition effect and technology information effect, where the former (latter) drives a negative (positive) export spillover.

Keywords:

• Foreign direct investment (FDI)
• foreign ownership ratio
• export
• spillover

JEL CLASSIFICATION:

• F10
• F14
• F19

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Please visit the following website for details:

http://unctadstat.unctad.org/ReportFolders/reportFolders.aspx

2 FDI has been shown to positively affect economic growth through a variety of channels such as capital accumulation, technological spillover, spillover of managerial skills, competition effect and demonstration effect (Wu and Buckey 1999).

3 Taiwan was admitted to the WTO in 2002 under the name ‘Separate Customs Territory of Taiwan, Penghu, Kinmen and Matsu (Chinese Taipei)’ with a status of ‘developing member’. In 2018, Taiwan's WTO status was elevated to ‘developed’ member from ‘developing’ member. Please visit the following website for details: https://focustaiwan.tw/business/201810140007

4 Please visit the following website for information on FDI in Taiwan: http://www.moeaic.gov.tw/

5 For details, please visit http://web02.mof.gov.tw/njswww/WebProxy.aspx?sys=100&funid=edefjsptgl

6 Based on the Harmonized Commodity Description and Coding System (HS), Taiwan’s major exporting industries for the period 2006–2011 in descending order of export value include sections 16, 15 and 7. Electronics industry belongs to section 16 of HS classification system (Data source: Trade Statistics of Taiwan’s Ministry of Finance). According to the most recent trade report, the largest industrial export sector in Taiwan is electronics industry (https://www.nordeatrade.com/en/explore-new-market/taiwan/trade-profile).

7 According to Crespo and Fontoura (2007), there are five channels through which FDI has spillover effects in host country, including ‘demonstration/imitation, labor mobility, exports, competition and backward and forward linkages with domestic firms’.

8 Prior studies show that technology improvements contribute to exports (e.g., Montobbio and Rampa 2005; Márquez-Ramos and Martínez-Zarzoso 2009). Huang and Qiao (2005) also document that FDI and international industrial transfer contributed to the growth of China’s electronics industry. In particular, FDI brought not only capital but also technology into China, which fosters the development of Chinese electronics industry.

9 According to Greenaway, Sousa, and Wakelin (2004), fixed cost involves ‘establishment of distribution networks, the creation of transport infrastructures, investment in advertising to gain public exposure, research about the foreign market to gain intelligence on consumers’ tastes, market structure, competitors, regulations and so on’.

10 Domestic exporters are likely to export more to avoid competition with MNEs that produce similar goods in the host country (Girma, Görg, and Pisu 2008).

11 Classification of the electronics industry into eight subindustries is based on Taiwan Stock Exchange.

12 Natural log of exports is used as the dependent variable EXT_i because the original values are too large.

13 According to the Taiwan Economic Journal (TEJ) database, the total foreign capital ratio (FCR) of a firm is the capital contribution portions of all foreign investors and organizations.

14 Following prior research, the study also considers foreign presence in production (FPP) and foreign presence in assets (FPA) in estimation. However, including FPP or FPA results in severe multicollinearity problem. Hence, the study presents only the results based on FPE as a measure of foreign presence.

15 The one-step rather than two-step robust GMM estimation is implemented in this study. The underlying reason is that the standard errors obtained from two-step estimation are downward biased and Windmeijer correction should be applied to fix this problem. However, Windmeijer corrected standard errors have recently proved to be biased when the dynamic panel model is not correctly specified, which is often the case. According to Hwang, Kang, and Lee (2020), a doubly-corrected robust variance estimator should be used when conducting two-step estimation. However, such an estimator remains unavailable in the software package, so we opt to perform one-step estimation and obtain one-step standard errors robust to heteroscedasticity for statistical inference.

16 Sun (2009) used the ratio of exports to total sales to measure export intensity.

17 Variance inflation factor (VIF) is used to measure the severity of multicollinearity. Suppose there are n determinants of some dependent variable. One can regress some determinant on other determinants to get R-squared. VIF is obtained by calculating the reciprocal of one minus such R-squared. If some determinant is well explained by other determinants, meaning that R-squared is high, VIF will be high. In empirics, VIF is considered high and multicolliearity is a concern when its value is greater than 10. VIF is used as a primary measure for detecting multicollinearity because it has been widely used to determine the severity of multicollinearity in prior studies.

18 Given that only EXT_i is log-transformed, the effect of FCR_i on EXT_i is $\left(e^b-1\right)\times 100\%,$ where b is the coefficient on FCR_i. The same goes for FCR_o and EXT_o. The economic magnitude of such estimated effect in other estimations can be obtained based on this approach.

19 According to Chen et al. (2019), in response to appreciation of Taiwan Dollar, labor shortage, and wage increase in 1980s, ‘Expanding and deepening industrial VCs was urgently necessary, as was the promotion of high-tech industries, which forced an outward shift of the ITE industry characterized by labor-intensive and export-processing activities’. ITE here refers to information technology and electronics, which are labor-intensive as mentioned above.

20 Sun (2010) finds that ‘the state and collectively-owned average domestic firms located in coastal or central China appear to benefit from foreign presence’.

21 Some p-values for Hansen’s J statistic are smaller than 0.01 in Tables 4–10 (Hansen and Singleton 1982), resulting in the rejection of the null hypothesis of valid overidentifying restrictions at the 1% level. However, the rejection of the null does not necessarily mean too many instruments and model misspecification. Sato and Söderbom (2017) argue that it could be due to the possibility that the slope coefficients are time-varying and it is recommended to estimate a model where the coefficients vary with time. The p-values for the z statistic of the Arellano-Bond test for the second-order serial correlation in error are all greater than 0.1 in Tables 4–10, indicating no autocorrelation at order two at the 10% level (Arellano and Bond 1991). Hence, the concern about model misspecification can be alleviated (Roodman 2009; Habimana 2017).

22 We also estimate the model that replaces year dummy variables with the dummy variable that captures the effect of the global financial crisis (GFC). The results turn out to be similar to those in Table 4. In addition, the coefficient of GFC is significantly positive, indicating that exports increase after the global financial crisis likely because currency depreciation and reduced interest rate are conducive to increased exports during the post-crisis period. Detailed estimation results are available upon request.

23 If the standard error is not adjusted for heteroscedasticity, the p-values for FCR_o and FCR_o×UPCST_o are both 0.1.

24 We also obtain the results regarding the channels of export spillover from FDI based on sample partitions. However, the results on the interaction variables are insignificant except when examining export spillover from FDI through the channel of industry exports.

Additional information

Notes on contributors

Naiwei Chen

Naiwei Chen is an associate professor in the Department of Banking and Finance at National Chiayi University, Taiwan. His research interests include corporate governance, corporate finance and economic growth.

Ho-Chyuan Chen

Ho-Chyuan Chen is a professor in the Department of Economics at National Chung Cheng University, Taiwan. His research interests include industrial economics, game theory and competition policy.

Rong-Siou Lin

Rong-Siou Lin is a masters student under the guidance of Ho-Chyuan Chen.