Does Innovation Co-move with FDI? Evidence from OECD countries

: In this study, the panel co-integration test combined with structural breaks was used to explore the long-term co-movement between FDI and patent and trademark applications ， in accordance with 33 OECD countries from 1999 to 2018. The robust results demonstrate that both innovation variables including patent and trademark co-move with FDI in the OECD sample. Furthermore, this long-term co-movement of FDI and innovation experiences some structural breaks during the period 2003-2010. Finally, there is a long-term co-movement between FDI inflows and innovation activity in OECD countries.

1 Previous literature typically investigates the relationships between FDI and innovation from two directions separately, including uni-directional causality running from FDI to innovation, or the influence of innovation activities on attracting FDI.
However, scant studies pay attention to the bi-directional relationship between FDI and innovation. Moreover, little research has attempted to expand attention onto the dynamic relationships between innovation and FDI. Since structural breaks are normal events in economics, ignoring them may contribute to misunderstanding the relationship between FDI and innovation. This research finds that the bi-directional comovement between innovation and FDI is commonly neglected by previous wisdom. Therefore, the essential contribution of this study is that we examine the long-term dynamic co-movement between innovation and FDI in 33 OECD countries from 1999 to 2018, using a co-integration approach with breaks.
The rest of this paper runs follows. Section 1 is the existing literature review.
Section 2 gives the theoretical background. Section 3 provides a description about the method and data, including the panel stationary test and the panel co-integration model. Section 4 exhibits the empirical results and Section 5 presents our main conclusion.

Literature Review
Plenty of studies have found that innovation activities are positively associated with foreign direct investment (FDI). From the theoretical aspects, the existing literature has explored the mechanism through which FDI affects national innovation performance in the host country (Cheung and Lin, 2004;Awokuse and Yin, 2010;Hsu and Tiao, 2015). Specifically, Cheung and Lin (2004) discuss four channels in greater details. First, local firms can obtain a technology spillover effect from advanced products from international firms. Second, more skilled workers trained by international firms enter local market, which encourage firms to formulate their own innovation program. Third, mature technologies brought by international firms lower the test cost of new technology. Fourth, technology spillover happens through vertical industrial linkages.
As is well recognized in the empirical literature, FDI has a positive effect on innovation activities (Javorcik, 2004;Awokuse and Yin, 2010). For example, using an ordinary least square (OLS) estimation and 138 industry sectors in 2004 as the sample, performance in emerging economies like China. The empirical evidence indicates that international trade and FDI are important determinants of national innovation performance. Additionally, Cheung and Lin (2004) examine the influence of FDI on provincial innovation output in China and note that FDI significantly promotes innovation development. Ascani et al. (2020) utilize data of manufacturing sectors in Italy to explore the impact of FDI on local innovation. They find that inward FDI indeed promotes local innovation, and this effect is stronger in science-based activities.
Numerous studies contrarily demonstrate that innovation exhibits considerable influence on FDI inflow, because international enterprises often bring advanced technologies to host countries. More commonly, Khoury and Peng (2011) indicate that innovation differences affect FDI, especially in developing countries. The authors state that strong intellectual property protection may lead to increased FDI. Aside from this, the other specific reforms are innovation-related, like patent laws that respect foreign intellectual property rights can actually attract more FDI.
Not only is the literature concerned about the determinants of innovation, but also the shocks from innovation to FDI (Wang and Kafouros, 2009;Khoury and Peng, 2011). 1 Studies have shown that intellectual property rights play an important role in attracting FDI. Improvements in intellectual property rights led to a sharp increase in foreign investment in China from 1992 to 2005 (Awokuse and Yin, 2010). Furthermore, the empirical results of Hsu and Tiao(2015) on the impact of property rights protection on FDI in 11 major Asian countries again show that improving patent protection helps to attract more FDI inflows.

Theoretical Background
One of the important reasons for attracting FDI is to gain advanced technology through access to products as well as services and hence improve innovation performance locally (Archigugi et al., 2011). The diffusion spillover effects of FDI are an objective existence. Konstandina et al. (2020) clearly show that FDI is able to promote technology transfer in the host country, as it can bring about a spillover effect of technology, which stimulates independent innovation via technology imports and learning. Independent innovations helps promote feedback in attracting more FDI inflow (Hsu et al., 2015;Awokuse et al., 2010). Acs and Audretsch (2010) find a positive correlation between innovation activity and FDI growth rate in developed countries, but innovation activity and FDI show a negative association in most developing countries. Therefore, FDI may move and grow together with innovation in the long run, especially in OCED countries that exhibit greater economic development levels compared to developing countries.
It is also noted that the relationship between FDI and innovation is unable to remain constant and is likely to experience some structural changes. For instance, Sbia et al. (2014) show that the series of FDI follow a stationary process with some structural shifts. Solarina and Shahbaz (2015) also demonstrate that structural breaks exist in the series of FDI. In fact, FDI series tend to cointegrate with some macroeconomic variables with structural breaks. Umit and Alkan (2016) suggest that FDI has a longterm cointegrated correlation with employment under structural breaks. Therefore, structural breaks may be important in analyzing the long-term co-movement in FDI and innovation correlation activities and hence cannot be ignored in the model.
The need to consider structural breaks could be explained in the following ways.
Structural breaks are a normal case in economics, as many political and economic incidents cause huge impact on FDI attractiveness or innovation performance.
Structural breaks are an unneglectable problem in macroeconomic series since they cannot avoid common exogenous shocks like a global financial crisis (Chang et al., 2011). 2 For example, external factors (i.e., the 2008 global financial crisis) have brought a significant shock to FDI and innovation activities. In this case, there is a sharp decline in global FDI, and enterprises are also less likely to invest in innovation programs in order to survive during a crisis period. As a result, exogenous shocks including common ones and country-specific shifts will impact the long-term relationship between FDI and innovation, and as a result structural changes need to be accounted for in the estimation.

The panel stationary test with structural breaks
Because a structural break may disturb the stationarity of macroeconomic series, structural breaks need to be considered in studies concerning the relationships between FDI and innovation. The literature has demonstrated the importance of structural breakthroughs in examining the existence of unit roots (Chang and Lee, 2015). Zivot and Andrews (1992) develop the minimum test widely-used endogenous test. Lee and Strazicich (2001) as well as Jewell et al. (2003) investigate a model with endogenous structural breaks that brings about more significant result.
In this section we follow the models with the consideration of a structural break developed by Carrion-i-Silvestre et al. (2005), which are designed to test the null hypothesis of panel stationarity as follows: where it y is the variable of interest in country i at time t , the dummy variable The test statistic for the null hypothesis as follow: We note that the limiting distribution of the test statistic is standard normal. To ensure robustness, we calculate the bootstrapped critical values.

Panel co-integration tests with structural breaks
According to the Westerlund (2006) panel co-integration test, the data generation process of dependent variables is as follows: To check for the existence of a long-term cointegrated relationship, the panel Lagrange multiplier (LM) test statistic is constructed as follows: , and ,it e  is the regression residuals obtained using any valid estimate of the co-integration vector, such as traditional dynamic OLS (DOLS) estimates.

Panel dynamic OLS estimations with structural breaks
In order to examine the panel long-term relationships between innovation activities and FDI, we employ the panel dynamic OLS (DOLS) method, including leads and lags of the independent variables, as follows:

The data and variables
We gather the innovation variables, number of patent applications (Patent) and Trademark for the sample, which shows the same trend as Patent. All most countries in our sample exhibit a fluctuation or slumping trend in 2007. This provides preliminary evidence that we should consider structural breaks when discussing the long-term linkage between FDI and innovation variables. Figure 3 and Figure 4 plot the mean scatters of FDI-Patent and FDI-Trademark, respectively. From Figure 3 and Figure 4, we roughly conclude that FDI and innovation have a long-term co-movement trend.

Results of panel co-integration test with multiple structural breaks
Preliminary analysis shows that when analyzing the long-term relationships between FDI and innovation activities, multiple structural breaks need to be considered in the model. We confirm the stationarity of FDI and innovation under the condition of considering structural breaks at first.  Our evidence does suggest that multiple structural changes in the long-term relationships of the panels need to be considered in the co-integration test. In other words, as Grossman and Helpman (1991), Aghion and Howitt (1998)  We also find that break points occur in the number of applications for patents and trademarks around 2008 from Table 2 and Table 3

Panel causality test results
This paper uses the panel DOLS estimation proposed by Kao and Chiang (2001) to determine the relationship between innovation activities and FDI when these variables co-integrate in the long run. Patent has a positive impact on FDI, all effects are positive. However, when the independent variable is Trademark, as shown in Table 4, NULL is found to be rejected at least at the 5% level in 18 of 33 cases. In summary, the country and panel parameters clearly indicate that a favorable innovation environment is conducive to attracting FDI into OECD countries.
[ Table 4 is here] [ Table 5 is here] Through Table 4 and Table 5 we therefore confirm that in OECD countries, there is a long-term bidirectional positive causal relationship between FDI and innovation.  We also account for the financial crisis in the test. As presented in a previous analysis, the 2008 financial crisis did affect FDI and hence innovation. Actually, this global event also has had a direct influence on countries' innovation performance, which may lead to biased results. Hence, we include the financial crisis variable in the model. In particular, the variable financial crisis (FS) is set to 1 when the year is lagged behind 2008 and 0 otherwise. The last column of Table 6 presents the corresponding results. It appears that the test cannot reject the ineffectiveness of long-term movements and argues that FDI and innovation in OECD countries are synchronized.
[ Table 6 is here]

Conclusions
Using annual data from 33 OECD countries from 1999 to 2018, this research examines the long-term relationship between FDI and innovation activity. We aim to address the long-term co-movement between FDI and innovation and account for potential structural breaks. Our study compares studies that focus only on a one-way attracts more foreign direct investment than Trademark, but foreign direct investment contributes more to trademarks than patents.
We believe our empirical results provide more useful information to governments for formulating economic and trade policies. One policy implication in the long run is that innovation must be based on sustained inflows of FDI inflows that can promote innovation activities. From the perspective of macro-economic policy, FDI affects domestic innovation activities and is one of the main influencing factors, 6 host governments must attract greater FDI inflows. Moreover, given that labor-intensive low-tech FDI contributes to less innovation in the host country compared to R&Dintensive high-tech FDI, governments should attach more importance on attracting the latter versus the former. In another sense, innovation can encourage developed countries FDI flows. Therefore, in accordance with our findings, the empirical implication is that governments of OECD countries should encourage the pursuit of more innovation, including original innovation and imitation innovation, in order to increase FDI flows.
We should mention that more specific paths or mechanisms may exist in the longterm co-movement between FDI and innovation. However, due to space limitations, we put such an important issue up to other scholars who are interested in this field.        Notes: t-statistic is in parentheses. The asymptotic distribution of the t-statistic is standard normal as T and N go to infinity. ** denotes statistical significance at the 5% level. All results allow for up to five structural breaks for each country.