Regional innovation and the retention of foreign direct investment: a place-based approach

ABSTRACT

The relationship between regional innovation and foreign direct investment (FDI) has been extensively examined, but the impact of regional innovation on FDI retention has received scant attention. Drawing on the logic of place-based approaches, the paper investigates this impact by conducting survival analyses on data collected across 31 provinces of China. We find that regional technician supply (RTS) and regional intellectual property flexibility (RIPF) help regions retain foreign ventures. However, RTS is less likely to retain the foreign investments of multinational enterprises with large research and development expenditure, while RIPF is more likely to retain them. Theoretical and policy implications are discussed.

KEYWORDS:

• regional innovation
• foreign direct investment retention
• firm innovation
• survival analysis

JEL:

• C41
• F23
• O32
• R11

INTRODUCTION

Economic geography (EG) has found strong relationships between regional innovation and foreign direct investment (FDI), showing that regional innovation determines regional FDI (Castellani et al., 2016; Wei et al., 1999) and that, in turn, regional FDI shapes regional innovation (Smith & Thomas, 2017; Wang et al., 2016). Despite the important implications for the regional innovation–FDI relationship, prior research has focused on either the attraction of FDI via regional innovation or the leveraging of FDI to boost regional innovation (e.g., Castellani et al., 2016; Smith & Thomas, 2017). It has been largely silent on regional innovation for FDI retention in a region. Place-based approaches contend that multinational enterprises (MNEs) and their investments are crucial for regional value creation and capture (Bailey et al., 2020; Barca, 2009). The loss of foreign ventures or the retention of those that do not create region-desired value may impede economic growth. In this sense, it is not only crucial for regions to have FDI inflow, but also to retain foreign ventures, especially those that build local productive capacity.

International business research has examined FDI retention with a focus on MNEs (Tang et al., 2021b). A growing corpus of work has shown that firm heterogeneity in innovation profiles (Xu & Lu, 2007), technological capital (Shaver & Flyer, 2000), and technology adoption (Zaheer & Mosakowski, 1997) may affect the divestment decisions of MNEs in a host country. This research stream argues that whether MNEs retain their ventures in a host country depends on their technological losses or gains in that nation (Shaver & Flyer, 2000). Incorporating international business research, prior EG research has presented findings on FDI retention, with Kim et al. (2010) finding Japanese ventures were less likely to exit China if they were geographically close to other Japanese ventures in China. Cui et al. (2020) observed regional competitive profiles (e.g., factor and demand conditions) that could affect the backflow of FDI inflow in China. These findings have added to our knowledge of FDI retention in a country. However, MNEs may have multiple investments in different regions in any one nation, and not all regions may see the maintenance of FDI. This study focuses on FDI retention at the subnational level and contributes new findings to the regional innovation−FDI research.

We build on the place-based approach and develop our arguments by identifying regional capabilities for value creation (Barca, 2011) and by considering ‘the specifics of the local and wider regional context’ (Barca et al., 2012, p. 140). Specifically, we investigate regional innovation’s direct impacts on regional FDI retention and the variation of these impacts among foreign ventures held by different MNEs. In so doing, we demonstrate how region-specific innovation advantages help a region retain FDI by creating values with MNEs (i.e., parent firms of foreign ventures) and capturing values co-created with foreign ventures. These findings advance our understanding of regional development by integrating regional innovation with regional stakeholders (e.g., MNEs). Our study, therefore, helps to shift place-based approaches from a policymaking tool to a coherent framework for understanding how FDI can advance the development of regions.

Empirically, we conduct a set of survival analyses of foreign ventures in a host country’s subnational regions. These ventures were held by MNEs operating in the host country even after divesting one venture in that nation. This empirical strategy enables us to tease out FDI retention in a region and distinguish such retention from national-level FDI activities. We examine the survival of 927 foreign ventures held by 236 MNEs in 31 provinces of China between 2007 and 2016. The analysis of China allows for significant diversity in economic and institutional conditions across regions in a country. China is a large market with many investment opportunities (United Nations Conference on Trade and Development (UNCTAD), 2020) and an emerging economy in the process of marketization, suggesting both economic and institutional variations across subnational regions (Tang et al., 2021a; Wang et al., 2019).

The rest of this paper is organized as follows. The next section provides the theoretical background and develops the main hypotheses. The third section introduces the empirical framework. The fourth section presents the empirical findings. The paper concludes by discussing the theoretical and policy implications of our core key findings.

THEORY AND HYPOTHESES DEVELOPMENT

The place-based approach and regional FDI retention

Capital investments and technological innovation are driving forces behind regional economic growth (Grossman & Helpman, 1991; Romer, 1986, 1990), and FDI is an important conduit for these forces (Bajo-Rubio et al., 2010; Borensztein et al., 1998; Crespo & Fontoura, 2007). Besides physical investments, inward FDI creates technological spillovers that may benefit lagging regions, especially those in emerging economies (Fu et al., 2010). Prior research has focussed on either FDI-led innovation (e.g., Smith & Thomas, 2017; Wang et al., 2016) or the impact of regional innovation on FDI (e.g., Fallon & Cook, 2010; Halvorsen, 2012), assuming, albeit implicitly, that FDI is a one-way path and will stay in a region after inflow. This assumption, however, does not always hold true.

Place-based perspectives on regional development suggest natural and societal characteristics vary across regions and, consequently, governments should seek to leverage regional advantages while reinforcing areas of weakness (Bailey et al., 2018; Barca, 2009; Beer et al., 2020). In the context of inward FDI, regions develop policies for creating and capturing values according to specific trajectories (Borensztein et al., 1998; Hildreth & Bailey, 2013). By contrast, MNEs invest in a region to create and capture values for their own benefit (Brandenburger & Nalebuff, 1995), aiming to profit from investment in the region (Dunning & Lundan, 2008; Nielsen et al., 2017). Although it is possible for a region to co-create value with MNEs investing in the region, the alignment of these inconsistent objectives may not endure (Aitken & Harrison, 1999; Clarysse et al., 2014), opening up the possibility of divestment, as well as the potential mismatch between regional policies and the retained FDI.

Despite recognizing the regional innovation–FDI relationship, the EG literature is largely silent about how regional innovation may help a region retain foreign ventures and, particularly, retain those that fit with the region’s innovation environment. In a large emerging market such as China, political and economic environments vary significantly across subnational regions due to substantial differences in natural endowments and institutional development (Wei et al., 1999). Such variations shape two pillars of the regional innovation environment: technician supply and intellectual property (IP) flexibility. For a region, the former reflects a regional innovation capability (Barbero et al., 2021; Wang et al., 2019) and defines the extent to which foreign ventures may create value for their parent firms (i.e., MNEs) by developing innovation in the region (Wang et al., 2016). The latter represents a type of regional institutions for innovation (Wojan, 2019), and determines the degree to which foreign ventures may capture value derived from regional innovation (Mieszkowski & Barbero, 2021). For MNEs holding ventures in a region, their research and development (R&D) expenditure denotes innovation competence (Aarstad & Kvitastein, 2020). Such competence determines whether and how MNEs may benefit from their ventures in a region’s innovation environment.

Regional innovation and FDI retention

The place-based approach emphasizes regional characteristics, which underpin regional unpredictability, heterogeneity, and uncertainty and, consequently, affect a region’s long-run economic development (Bailey et al., 2018; Barca et al., 2012). Yet, while regional innovation is expected to be efficient in filling the innovation-based economic gap between developed and lagging regions (Han et al., 2016), regional differences in natural endowments and institutional conditions may result in imbalances in regional innovation. In China, for example, coastal regions have adequate resources and liberalized institutions for creating knowledge and undertaking innovation activities, but other regions do not (Wang et al., 2019). These innovation imbalances are derived from the variation of regional technician supply (RTS) and regional intellectual property flexibility (RIPF), both of which constitute a region’s innovation environment.

A region’s RTS has a co-integration relationship with regional and external knowledge stocks (Sanso-Navarro & Vera-Cabello, 2018). Published research has accepted RTS as an innovation input and shown its positive impact on regional innovation performance (e.g., Guan & Yam, 2015), and this is largely due to the fact that technicians translate tacit knowledge into codified knowledge and manifest it in technologies. These technologies are, in turn, externalized in products and services (Nonaka et al., 2000). Moreover, researchers note that RTS is directly consumed in the R&D process (Guan & Chen, 2010) and related to financial input of innovation (e.g., technician salaries; Wang & Huang, 2007), thereby reflecting a region’s R&D capability (e.g., Barbero et al., 2021; Wang et al., 2019).

RTS is crucial for regional FDI in that a region’s R&D capability determines how the region may absorb the knowledge of foreign investors (Smith & Thomas, 2017) and affects how MNEs may create value via investments in the region (Cano-Kollmann et al., 2016). When investing in a region, MNEs seek complementary assets for local production (Dunning & Lundan, 2008). These assets include physical assets and technological capabilities derived from technicians (e.g., competence in high-tech manufacturing). Such capabilities are found to affect foreign ventures’ survival because technicians are crucial for understanding the knowledge base of MNEs, and for transferring technologies into the region (Ott et al., 2014). A larger number of technicians also strengthens knowledge connections between MNEs and the region (Tsai & Ghoshal, 1998). Such connections increase the irreversibility of foreign ventures held by these MNEs (Song & Lee, 2017), thereby retaining FDI in a region. Accordingly, we propose:

Hypothesis 1: A region’s technician supply has a positive (negative) impact on the retention (divestment) of foreign ventures in the region.

Furthermore, RIPF – the implementation of effective innovation regulations in a region – is another pillar of the regional innovation environment, representing an important element of regional innovation-related institutions. Previous studies have shown the importance of the number of patents as an indicator of a region’s innovation achievement (Han et al., 2016), innovation activity (Cantner et al., 2019), or innovation capacity (Mieszkowski & Barbero, 2021). In addition, flexibility in granting patents is important for regional innovation. Griliches (1990) argued that the ratio of granted patents to patent applications reflected the flexibility (or stringency) of patent examination in a region. Although IP regulations are mostly consistent within a country, the implementation of these regulations varies. Stringent examination of patent applications results in the strict implementation of IP regulations, which in turn results in a small granting ratio (Griliches, 1990) which may be unattractive to some MNEs.

Previous research has recognized the importance of institutional quality for regional development. For example, Rodríguez-Pose (2013) discussed the importance of institutions for regional development. D’Ingiullo and Evangelista (2020) found that regional institutions determined a region’s innovation performance as they affected knowledge transfer and technological externalization in the region (Crescenzi & Rodríguez-Pose, 2009).

As ‘the rules of the game’, institutions determine ‘humanly devised constraints that structure human interaction’ in a region (North, 1990, p. 360). In this sense, MNEs may divest their foreign ventures when being affected by unfavourable institutions (e.g., Belderbos & Zou, 2009; Getachew & Beamish, 2017). Flexible innovation institutions favour MNEs and help them retain ventures in a region because RIPF enables MNEs to access IP protection more efficiently (Griliches, 1990), facilitating knowledge transfer. Smoothly transferred knowledge strengthens the MNE−region connection and enhances the likelihood of foreign ventures being retained (Song & Lee, 2017). Thus, we posit:

Hypothesis 2: A region’s IP flexibility has a positive (negative) impact on the retention (divestment) of foreign ventures in the region.

Firm R&D expenditure (FRNDE)

The interaction between regional innovation and other stakeholders such as people, local assets, place-based communities, institutions, and geography is crucial for regional value creation and capture (Beer et al., 2020). It is not only regional characteristics per se that matter, but also the integration of these characteristics with other stakeholders (e.g., MNEs). Prior research has shown that innovation was both location- and firm-specific (Turkina et al., 2019). Advocates of place-based policies contend that a region needs to be well-positioned with respect to competence-supporting policies and activities (Bailey et al., 2020). Such place-based positioning helps a region develop locational advantages (e.g., an attractive innovation environment) to retain FDI. However, policies and activities based on regional positioning may not benefit all firms equally in the region. As such, an attractive innovation environment may not retain all foreign ventures, whose parent firms (i.e., MNEs) engage in R&D to different degrees. Such differences are likely to alter the relationships suggested in Hypotheses 1 and 2.

Previous studies have examined firm innovation as a function of R&D expenditure (Atkeson & Burstein, 2010), since this expenditure reflects a firm’s absorptive capacity (Cohen & Levinthal, 1990) and innovation resources (Brown et al., 2012). Turkina et al. (2019) argued that a firm’s absorptive capacity affected the extent to which the firm could search and integrate regional knowledge into its internal innovation process. Sarkar et al. (2021) found R&D expenditure was a vital resource for firms seeking to respond to market volatility and foster regional competitiveness, showing a place-based policy (e.g., Smart Specialisation Strategy) did not influence the innovation of all firms in the same way.

FRNDE may affect the impact of RTS on regional FDI retention for multiple reasons. First, MNEs with a small R&D expenditure are likely to stay in a high-RTS region for an extended period in order to benefit from regional innovation capabilities, but MNEs with a large R&D expenditure may not rely on local technicians to develop innovation capabilities in the long run. When an MNE has acquired local knowledge and cannot obtain additional value in a region, the MNE is likely to leave (Inkpen & Beamish, 1997). Second, because sufficient innovation resources facilitate knowledge transfer (Wei & Liu, 2006), a high level of RTS may enable MNEs with a large R&D expenditure to access regional knowledge and integrate it quickly into their internal innovation process (Cohen & Levinthal, 1990), suggesting a rapid process of knowledge transfer and a short period of local engagement. Therefore, we predict:

Hypothesis 3a: FRNDE weakens the RTS−FDI retention relationship, such that a large RTS is less likely to retain foreign ventures held by MNEs with a large R&D expenditure.

Further, R&D expenditure may alter the relationship between RIPF and regional FDI retention in several ways. First, MNEs with a large R&D expenditure are more likely to seek flexible knowledge protection in a region because MNEs with sufficient innovation resources have valuable knowledge to transfer (Cano-Kollmann et al., 2016; Wei & Liu, 2006), and the outcome of such transfer relies on accessible IP protection (Berry, 2017). Since a flexible IP-protecting environment enables MNEs with large R&D expenditure to access innovation protection in a region, these MNEs are less likely to leave the region. Second, flexible innovation institutions encourage MNEs with large R&D expenditure to develop local connections such as research collaboration with local communities because of the accessible protection on external R&D outcomes derived from large R&D expenditure (Oxley, 1999; Zhao, 2006). As such, we postulate:

Hypothesis 3b: FRNDE strengthens the RIPF−FDI retention relationship, such that high RIPF is more likely to retain foreign ventures held by MNEs with a large R&D expenditure.

Figure 1 summarizes the above-proposed relationships in a theoretical framework.

EMPIRICAL FRAMEWORK

Research data and sample

To test the hypotheses, this study relies on two databases: the Orbis database maintained by Bureau van Dijk and the Statistical Yearbook of provinces in China (SYPC) compiled by the Provincial Bureau of Statistics of 31 subnational regions in China. The Orbis database covers publicly listed firms worldwide and has been widely used in research (e.g., Rasel et al., 2020; Turkina et al., 2019). The data contains extensive firm-level information such as ventures (i.e., subsidiaries, branches) in home and foreign countries, R&D expenses, the year of incorporation, the number of employees, financial performance, etc. This study focuses on non-Chinese firms that held at least one venture in China after divesting another venture between 2007 and 2016, thereby ensuring that a divestment event occurs in a subnational region rather than at the national level. The SYPC data has been widely used in previous EG studies (e.g., Cui et al., 2020; Wang et al., 2016). It covers regional information about economic development, technology and science, and regional FDI amount. For a geographical distribution of the research, see Appendix A in the supplemental data online.

Figure 1. Theoretical framework.

Estimation approach and the dependent variable

This study aims to understand how long a foreign venture would survive (i.e., when it would be divested) in a region, suggesting an event history analysis of its survival in the region. Compared with other regression analyses, an event history analysis (i.e., a survival analysis) allows for the inclusion of events (e.g., divestment of foreign ventures) at different moments in time. It takes the temporal bias of observations into consideration (Hosmer et al., 2008). In the analysis, a dummy variable is used as the dependent variable and inserted in a hazard function, where 1 indicates the event’s occurrence (i.e., the divestment of an MNE) and 0 indicates that a foreign venture retains. Prior EG research has applied this method to answer various research questions, including the performance of US counties during the Great Recession (Lewin et al., 2018) and the migration of university students after graduation (Teichert et al., 2020).

We model the retention of foreign ventures by estimating the hazard rate of divestment events. Specifically, the hazard rate $h_i\left(t\right)$ is the hazard (i.e., odds, likelihood or probability) that a foreign venture i has been in a region for a period of t before it is divested or censored, suggesting: (1) $h_i\left(t\right)=Pr\left(T_i<t+1|T_i\ge 1\right)$(1) where $T_i$ is the length of a retention spell (duration) of a foreign venture in a region. In the analysis, retention is identified as an observation that the foreign venture is not divested. We investigate the effects of regional and firm factors on retention by assuming the baseline hazard to be: (2) $h_0\left(t\right)=p{t}^{p-1}\mathrm{e}\mathrm{x}\mathrm{p}\left({\text{β}}_0\right)$(2) where $p$ is an ancillary shape parameter. It indicates the distribution shape, showing an exponential increase or decrease of hazard rates with time. Given a vector of determinants $x_i$ including RTS, RIPF, FRNDE and control variables discussed in the next subsection, the hazard function of FDI retention is estimated in a parametric model (i.e., Weibull regression): (3) $h\left(t,x_i\right)=p{t}^{p-1}\mathrm{e}\mathrm{x}\mathrm{p}\left({\text{β}}_0+x_i{\text{β}}_x\right)$(3)

In model estimation, we take firm–region crossed effects into consideration by estimating robust standard errors clustered by firm–region to the data (Rabe-Hesketh & Skrondal, 2012).

Explanatory variables

• Regional technician supply (RTS). A region’s supply of technicians is measured as the number of full-time-equivalent R&D personnel per 10,000 population. Theoretical justification and empirical findings of previous studies have shown the relationship between RTS and regional innovation (e.g., Wang et al., 2019) and the association between regional innovation and FDI (e.g., Huang et al., 2012).

• Regional intellectual property flexibility (RIPF). This variable measures the extent to which a region is flexible about patent applications by computing the ratio of a region’s number of granted patents to the number of applications. Prior research has conceptualized this ratio as the flexible (or stringent) implementation of IP protection (Griliches, 1990).

• Firm R&D expenditure (FRNDE). The R&D expenditure of an MNE that held a venture in a region is measured as the firm’s annual expenditure (US$ millions) on R&D. This measurement is based on the theoretical justification and empirical evidence in previous studies (e.g., Sarkar et al., 2021), arguing that R&D expenditure indicates a firm’s innovation resources and absorptive capacity for knowledge.

The retention of foreign ventures in a region is not only affected by regional innovation and firm R&D, but also by other regional and firm factors, suggesting a set of control variables in the analysis. Thus, this study controls for regional FDI volume, regional FDI countries, regional marketization, regional economic growth, a focal firm’s return on assets (ROA), firm profit margin, firm age, firm FDI in China and firm size. Appendix B in the supplemental data online presents the rationale of these control variables, their measurement, and data sources. To minimize the effect of potential endogeneity, all time-varying explanatory variables are lagged by one year. Lastly, we include region, industry, and year dummies to control the unobserved regional, industrial, and temporal effects.

RESULTS

Table 1 presents the survival function of foreign ventures observed in the sample time window. It shows that among 927 foreign ventures that started at time zero (i.e., when a foreign investment entered a region), 192 were divested (i.e., not retention, or ‘fail’) in time period one. This creates a survival function of 79%, implying that 79% of foreign ventures had been retained in a region for one time unit (i.e., year). Likewise, 65% of foreign ventures had been retained for two time units. By the end of the time window, 42 foreign ventures were retained and consequently censored.

Table 1. Survival function table of divestment.

Time	Foreign ventures	Fail	Net lost	Survival function	Standard error	95% Confidence interval (CI)
1	927	192	59	0.793	0.013	0.765	0.818
2	676	119	47	0.653	0.016	0.621	0.684
3	510	47	40	0.593	0.017	0.560	0.625
4	423	34	74	0.545	0.017	0.511	0.579
5	315	35	39	0.485	0.018	0.449	0.520
6	241	15	51	0.455	0.019	0.418	0.491
7	175	7	45	0.437	0.019	0.399	0.474
8	123	3	42	0.426	0.020	0.387	0.464
9	78	1	35	0.420	0.020	0.381	0.459
10	42	0	42	0.420	0.020	0.381	0.459

The pattern of survival functions (Table 1) is visualized in Figure 2, showing a decreasing step function with each step (i.e., 1 unit of time) across the horizontal axis. The vertical axis indicates the probability of a foreign investment being retained in a given period.

Figure 2. Kaplan–Meier survival estimates.

Table 2 reports descriptive statistics and the correlation matrix of key variables in the empirical tests. Given several large correlations, we conducted a variance inflation factor (VIF) test and found that all VIF values were < 2.0, suggesting little reason for concern about multicollinearity in estimation (Ryan, 2008).

Table 2. Descriptive statistics and correlation coefficients of key variables.^a

		1	2	3	4	5	6	7	8	9	10	11	12	13
1	Divestment (the event)	1
2	Regional technician supply	–0.26	1
3	Regional IP flexibility	−0.19	0.19	1
4	Firm R&D expenditure	−0.06	0.00	0.01	1
5	Regional FDI volume	−0.05	0.33	−0.06	0.02	1
6	Regional FDI countries	0.07	−0.27	−0.07	−0.06	−0.18	1
7	Regional marketization	−0.06	0.19	0.32	0.04	0.09	−0.11	1
8	Regional economic growth	0.64	−0.29	−0.12	−0.03	−0.10	0.19	−0.02	1
9	Return on assets	0.01	−0.01	0.01	0.01	−0.01	0.06	0.06	−0.02	1
10	Firm profit margin	0.02	−0.06	0.00	−0.01	0.02	0.02	0.04	−0.02	0.02	1
11	Firm age	−0.14	0.00	0.00	0.02	0.02	−0.02	0.02	−0.14	−0.01	0.04	1
12	Firm FDI	0.02	−0.29	−0.10	0.02	−0.04	0.09	−0.25	0.12	0.01	−0.04	0.00	1
13	Firm size	−0.07	−0.02	0.00	0.47	−0.01	−0.02	−0.04	−0.07	0.01	−0.04	0.04	0.24	1
	Mean	0.49	47.17	56.30	0.50	242.75	35.70	7.80	9.44	−0.01	8.98	73.60	21.39	62.62
	SD	0.50	35.02	11.97	1.07	291.63	37.85	1.21	6.24	1.32	12.84	46.21	27.69	122.52
	Minimum	0.00	3.38	22.80	0.00	0.00	3.00	1.23	−3.69	−24.07	−93.50	2.00	1.00	0.05
	Maximum	1.00	113.81	78.05	5.75	1800	159.00	11.01	27.61	17.92	70.71	323.00	165.00	2100
	VIF	–	1.42	1.17	1.32	1.16	1.12	1.22	1.16	1.01	1.03	1.41	1.02	1.24

Note: ^aCorrelation coefficients with absolute values > 0.05 are statistically significant at p < 0.05. N = 927. VIF, variance inflation factor.

Table 3 displays the results of a set of parametric analyses (Hosmer et al., 2008), showing the hazard ratio that represents the probability of a foreign venture being divested (i.e., not retained). A hazard ratio < 1 (> 1) means that an explanatory variable lengthens (shortens) the time to the event (i.e., divestment), implying an increase in the probability of retention (divestment). A hazard ratio of 0.939 means that a 1 unit increase in an explanatory variable leads to a decrease in the hazard ratio by 6.10% (i.e., 1 – 0.939), implying an increased probability of retention and a decreased probability of divestment. In Table 3, model 1 is a baseline model. It includes control variables only. Models 2 and 3 add the explanatory variables to test Hypotheses 1 and 2, respectively, and model 4 includes both explanatory variables to confirm the tests. Models 5 and 6 include interaction terms of regional innovation and FRNDE to test Hypotheses 3a and 3b, respectively. Model 7 takes all variables and interaction terms into the analysis. Changes in the log-likelihood of models 1–7 demonstrate the improved goodness of fit of models by including the explanatory variables and their interaction terms. The value of parameter p confirms the shape of the Weibull distribution, where a parameter p > 1 means a significant increase in the divestment probability following the increase in time.

Table 3. Hazard ratios of foreign direct investment (FDI) retention.^a

_t (time to event)	Model 1		Model 2		Model 3		Model 4		Model 5		Model 6		Model 7
H1: Regional technician supply (RTS)			0.939	***			0.942	***	0.940	***	0.942	***	0.940	***
			(0.007)				(0.007)		(0.007)		(0.007)		(0.007)
H2: Regional intellectual property flexibility (RIPF)					0.961	***	0.968	**	0.967	**	0.972	**	0.972	**
					(0.009)		(0.009)		(0.010)		(0.010)		(0.010)
H3a: RTS × FRNDE									1.004	^†			1.005	**
									(0.002)				(0.002)
H3b: RIPF × FRNDE											0.992	*	0.988	**
											(0.004)		(0.004)
Firm R&D expenditure (FRNDE)	0.914		0.893		0.916		0.898		0.776	*	1.365		1.369
	(0.064)		(0.065)		(0.066)		(0.067)		(0.079)		(0.277)		(0.266)
Regional FDI volume	1.000		1.000		1.000		1.000		1.000		1.000		1.000
	(0.001)		(0.000)		(0.001)		(0.000)		(0.000)		(0.000)		(0.000)
Regional FDI countries	1.006		1.022		0.998		1.018		1.017		1.017		1.016
	(0.006)		(0.018)		(0.007)		(0.020)		(0.020)		(0.019)		(0.019)
Regional marketization	2.104	***	1.265		1.887	***	1.200		1.206		1.207		1.210
	(0.240)		(0.192)		(0.220)		(0.172)		(0.171)		(0.174)		(0.167)
Regional economic growth	1.189	***	1.116	***	1.186	***	1.117	***	1.116	***	1.120	***	1.121	***
	(0.017)		(0.021)		(0.018)		(0.022)		(0.022)		(0.021)		(0.021)
ROA	1.041		1.040		1.029		1.031		1.029		1.033		1.030
	(0.049)		(0.041)		(0.049)		(0.041)		(0.042)		(0.041)		(0.041)
Firm profit margin	0.993		0.992	^†	0.995		0.994		0.993		0.994		0.993
	(0.005)		(0.005)		(0.005)		(0.005)		(0.005)		(0.005)		(0.005)
Firm age	1.002		1.002		1.002		1.002		1.002		1.003	^†	1.003	^†
	(0.002)		(0.001)		(0.002)		(0.001)		(0.001)		(0.001)		(0.001)
Firm FDI	0.999		0.997		0.999		0.997		0.998		0.998		0.998
	(0.002)		(0.002)		(0.002)		(0.002)		(0.002)		(0.002)		(0.002)
Firm size	1.001		1.001		1.001		1.001		1.000		1.001		1.000
	(0.001)		(0.001)		(0.001)		(0.001)		(0.001)		(0.001)		(0.001)
Constant	0.000	***	0.029	*	0.001	***	0.153		0.147		0.115		0.102
	(0.000)		(0.048)		(0.001)		(0.234)		(0.220)		(0.175)		(0.149)
Parameter p	1.87		1.99		1.90		2.02		2.03		2.02		2.04
Log-likelihood	−611.16		−563.51		−603.40		−558.15		−556.12		−556.06		−552.22
ΔLog-likelihood			47.64	***	7.75	***	53.01	***	2.03	*	2.09	*	5.93	**
(compared with model number)			1		1		1		4		4		4

Note: ^aRobust standard errors are shown in parentheses. Region, industry and year dummies are included. The number of observations = 927; and the number of firm-region clusters = 251. Variables in interaction terms are mean centred.

†p < 0.10; *p < 0.05; **p < 0.01; ***p < 0.001.

Hypothesis 1 proposes that RTS has a positive (negative) effect on the retention (divestment) of foreign ventures. The results of model 2 support this hypothesis, with a negative and significant effect for RTS (hr = 0.939, p < 0.001), suggesting that 1 unit of increase in RTS results in a decreased probability of divestment by 6.10%. RTS’s negative and significant effect is consistently observed in models 4–7, where other explanatory variables and interaction terms are added. Likewise, Hypothesis 2 gains support in model 3. Specifically, this hypothesis predicts a positive (negative) effect of RIPF on the retention (divestment) of foreign investments. This prediction is consistent with results in model 3, showing a negative and significant effect for RIPF (hr = 0.963, p < 0.001). Support for this hypothesis is also confirmed in models 4–7. Figure 3 visualizes these effects with Kaplan–Meier survival estimates for each quintile of RTS and RIFP, respectively, in panels A and B. These figures show that the probability of retention (divestment) is higher (lower) for regions with higher RTS and RIPF levels (e.g., quintiles 4 and 5 in both panels).

Figure 3. Kaplan–Meier survival estimates for each quintile of the regional technician supply (RTS) and regional intellectual property flexibility (RIPF).

Hypothesis 3a argues that FRNDE weakens the relationship between RTS and regional FDI retention. An interaction term between RTS and FRNDE is added in model 5, which shows a positive and significant effect for RTS × FRNDE (hr = 1.004, p < 0.10). This effect remains in model 7, demonstrating the support to Hypothesis 3a. Moreover, Hypothesis 3b posits that FRNDE strengthens the relationship between RIPF and regional FDI retention, which is consistent with the results of model 6. In the model, an interaction term of RIPF and FRNDE is negative and significant (hr = 0.992, p < 0.05), providing support to Hypothesis 3b. These effects are confirmed in model 7, in which all variables and interaction terms are included. We visualize the effects of FRNDE by presenting marginal effects of RTS and RIPF, respectively, in panels A and B of Figure 4. In panel A, we note that RTS’s effect on the predicted time to event decreases when FRNDE increases. For example, when FRNDE increases from 0 to 1, RTS’s marginal effect decreases from 10.60 to 9.70. Conversely, panel B shows that the effects of RIPF become larger when FRNDE increases. These findings provide additional support for Hypotheses 3a and 3b.

Figure 4. Marginal effects (with 95% confidence intervals – CI) of the regional technician supply (RTS) and regional intellectual property flexibility (RIPF) at different levels of firm research and development (R&D) expenditure.

Besides the main variables, a control variable – regional economic growth – shows interesting findings. It has a positive and significant effect across models 1–7, suggesting that fast economic growth does not retain FDI but encourages divestment. This result is not consistent with previous studies arguing the positive effect of economic growth on FDI (e.g., Castellani et al., 2016). However, it is consistent with theoretical arguments on MNEs in emerging markets (Hoskisson et al., 2000). These firms enter an emerging market mainly for low-cost advantages. When a region’s economy grows fast, the cost advantage may not persist due to growing costs of key production factors (e.g., labour), and the firm may leave the region. Moreover, we conduct a set of post hoc tests to check the robustness of empirical results in the supplemental data online.

DISCUSSION AND CONCLUSIONS

The extant literature has shown extensive evidence on the relationship between regional innovation and regional FDI inflow but lacks insight into the impact of regional innovation on local FDI retention. Such insights are important because both innovation and FDI are essential for regional economic growth. Nevertheless, MNEs may withdraw their investments from a region, and regions need to keep foreign ventures that can create region-desired value. Thus, this study aims at answering two research questions: how may regional innovation help regions retain foreign ventures, and which foreign ventures are more likely retained?

Answers to these questions advance the place-based approach by extending its theoretical logic into post-FDI research. Previous studies found the significant influence of regional innovation on FDI inflow in a region and FDI distribution among regions (Fallon & Cook, 2010; Halvorsen, 2012), but did not consider FDI exit from a region. In this paper, we find that RTS and RIPF help regions retain foreign ventures because a region’s technician supply and the implementation of IP regulations determine how foreign ventures may create values in a region and how the region may capture the value created by MNEs – an important group of regional stakeholders (Beer et al., 2020). As two pillars of a regional innovation environment, RTS and RIPF may facilitate the value creation of foreign ventures in a region, thereby helping regions retain FDI.

Despite the consistently positive effects of RTS and RIPF on regional FDI retention, their implications for theory and policy vary. Large RTS implies a high level of regional innovation capabilities (Barbero et al., 2021; Wang et al., 2019), which help regions retain FDI and facilitate the development of regional innovation (Guan & Yam, 2015). Therefore, regional policies focussed on technicians (e.g., working conditions of technicians, income, and other incentives for technicians) increase both FDI and the innovative capacity of a region. They enable the region to retain FDI as well as develop innovation, both of which are driving forces of regional economic growth (Grossman & Helpman, 1991; Romer, 1986).

In contrast, while RIPF has a positive impact on regional FDI retention, a high level of RIPF may not always facilitate innovation in a region of an emerging economy. Prior research found the innovation of firms in lagging regions was vulnerable to foreign entrants (Fu et al., 2010), because MNEs have better-developed capabilities for innovation and can produce a larger number of innovation outcomes (e.g., patents) than local firms. When a region has flexible institutions to protect innovation, foreign entrants are likely to obtain more protection than local firms. Such imbalanced protection may discourage regional firms’ innovation and, consequently, harm a region’s long-term economic growth. In this sense, regional policymakers need to consider the flexible implementation of IP regulations carefully.

The different policy implications of RTS and RIPF effects are also demonstrated in the variation of retained foreign ventures. This study finds that large RTS is less likely to retain foreign ventures whose parent firms invest more in R&D, but a high level of RIPF is more likely to retain these foreign ventures in a region. These findings support place-based policies and shed new light on their pathways to impact, as regional policymakers need to leverage regional advantages and complement regional weaknesses with place-based policies (Beer et al., 2020). Our findings show how regional innovation may retain which foreign ventures and can inform policies to sustain FDI in a region. While RTS helps the retention of foreign ventures held by MNEs with a low level of R&D expenditure, RIPF retains FDI by high R&D-expenditure MNEs. These insights provide a better understanding of a region’s value creation and value capture with different MNEs.

In addition, our findings shed light on a dilemma confronting many regions as they seek to navigate a path between reliance on internal innovation (i.e., increasing the number of technicians in the region) and dependency on external innovation (i.e., leveraging favourable institutions to retain foreign innovators in the region). While a large RTS facilitates regional innovation (Guan & Yam, 2015), it may not help regions benefit from foreign innovators (i.e., MNEs with large R&D expenditure) over an extended period. If a region of an emerging economy retains these foreign innovators via a high level of RIPF, the region may not develop its innovation capability effectively due to imbalanced IP protection between foreign and local innovators. Future research and policy formulation needs to examine this conundrum and seek solutions that provide the best possible solutions for different types of regions.

Furthermore, the current study has limitations that suggest future research to produce fruitful output. For example, while a focus on the innovation-related factors allows us to develop a theoretically consistent framework, the cost is at a lack of attention to incentive policies on FDI retention. A region’s proactive policies or administrative tools may help FDI stay in the region. Moreover, although we have undertaken multiple sets of tests and analyses to examine the hypothesized relationships, we acknowledge that it is hard to rule out completely the potential endogeneity in a regression-based research design. Future research is consequently encouraged to conduct studies with experimental or quasi-experimental designs (e.g., difference-in-difference estimates, Bertrand et al., 2004; regression discontinuity designs, Lee & Lemieux, 2010; behavioural experiments, Tang, 2017), which could (in)validate the relationships investigated in this study.

Overall, our study provides support for place-based programs of economic development through the investigation of the relationship between regional innovation and FDI retention across 31 subnational regions of China. Analyses of the regional innovation−FDI retention relationship extend our understanding by showing what types of innovation policy may create and capture value with which MNEs in a region. We find that both RTS and RIPF have the potential for retaining foreign investments, but in different ways. While RTS does not help a region keep ventures held by MNEs with large R&D expenditure, RIPF does. Knowing this, policymakers will be able to fine-tune their interventions in regional economies to achieve better outcomes.

ACKNOWLEDGEMENT

The authors are grateful to Ben Derudder and three anonymous reviewers for their invaluable and constructive comments on the earlier version of this work. Any errors are the authors’ alone.

DISCLOSURE STATEMENT

No financial interest or benefit has arisen from the direct applications of this research.