Home Country Innovation Performance: Moderating the Local Knowledge and Inward Foreign Direct Investment

Introduction

Since the 1990s, economic growth has been rapid and impressive in emerging countries, such as Brazil, Russia, India, China and South Africa (BRICS), contributing to transform the world economy (Fu et al., 2011). In search of growth, productivity and innovation performance, emerging economies facilitate the access of multinational enterprises (MNEs), providing large amounts of FDI in their economies (Blomström & Kokko, 1998; Bruhn & Calegario, 2014; Bruhn et al., 2015; Jin et al., 2019). Inward FDI acquired attention in the extant literature, due to the benefits derived from technology spillover, which became a primary source to promote economic growth enhancing the technology transfers, know-how, exports expansion and product and process innovation (Djulius, 2017; Jawaid et al., 2016; Jin et al., 2019; Mohanty & Sethi, 2019; Wang & Kafouros, 2009).

The literature strongly suggests that, in the past, inward FDI flows has often been a significant contributory factor to the structure and growth of a country’s technological capacity (Dunning & Lundan, 2008). Some noteworthy studies—Liu and Wang (2003), Wang and Kafouros (2009) and Castellacci and Natera (2016), among others—discuss the importance of FDI, considering it not as a mere source of capital but also as a conduit for technology transfer, management and human skills augmentation in host countries. In the specific case of Latin America and the Caribbean (LAC), foreign presence breaks the existing equilibrium, eliminates the monopolistic power of local firms and improves productivity (Castellacci & Natera, 2016).

The post-war era has been a period of significant growth for both world trade and FDI. That is an important factor of innovation for all countries, specially in emerging economies (Bruhn et al., 2015; Fleury & Fleury, 2011; Narula & Wakelin, 1998; Wang & Kafouros, 2009). Forward to the 1950s and 1960s, Brazil was essentially a receiver of FDI (Fleury & Fleury, 2011) when specific governmental programme reforms were created to attract foreign capital as strategy for industrial development through import-substitution industrialization. In the 1970s and 1980s, the inflows of FDI decreased (mainly due to the oil shock crisis) and grew back rapidly after 1990 in countries such as Mexico, Argentina, Brazil and Chile (Benavente et al., 1997; Bruhn & Calegario, 2014). Bruhn and Calegario (2014) and Casanova et al. (2019) argue that during the 1980s and 1990s (a period often referred to as ‘the era of market reform’), most Latin economies experienced changes in comparison to the previous phase of long-run growth. Thus, industrial development resurged under a new philosophy during the 1980s, based on national sovereignty and security doctrines, with a 5-year period of national development plans, national science and technology development programmes (Fleury & Fleury, 2011).

The LAC increased the openness of the economy following the Washington Consensus programme of pro-market reform through trade liberalization, increasing inward FDI (Casanova et al., 2019; Castellacci & Natera, 2016). FDI to LAC rose by 8 per cent to reach US$155 billion in 2017, driven by the region’s economic recovery (UNCTAD, 2018), the first rise in a 6-year period, but the inflows remained well below the 2011 peak during the commodities boom in the 2000s. From Table 1, it is possible to observe that Brazil received the largest amount of FDI in Latin America, over US$67.5 billion in 2017 and US$61.2 billion in 2018, a higher amount than Central American Economies and Transition Economies.

Brazil is considered the largest economy in Latin America, attracting more than 40 per cent of the total flows to the Latin region. Nine of the 10 largest acquisitions by foreign companies in the region were made in Brazil; seven concerned Chinese buyers, which acquisitions involved electricity, oil, infrastructure (gas transmission) and agribusiness companies (UNCTAD, 2019). However, it is possible to observe a decline in FDI when compared to 2017. Following a decline in commodity prices and corruption scandals, market turbulence in Argentina, recession in Brazil and a continued political deterioration in Venezuela, economic growth in the region slowed down between 2016 and 2018 (Casanova et al., 2019). Nonetheless, the authors argued the region grew by 0.6 per cent in 2018 and it is expected to grow by 1.6 per cent in 2019. On the other hand, prospects for foreign investment significantly depends on progress in the new Brazilian administration’s reform programme (UNCTAD, 2019). So far, the confidence of domestic investors reflected in stock markets and has not been matched by foreign direct investors, who appear to be waiting for stronger signals and higher inflows in 2019, on the back of positive economic forecasts, supportive investment policies and the value of announced greenfield projects increasing by more than 50 per cent in 2018 (UNCTAD, 2019).

Concerning empirical evidence, recent studies have highlighted the role FDI has played in the significant transformation in the recipient economy and in improving growth and performance (Bruhn & Calegario, 2014; Guimón et al., 2018; Jawaid et al., 2016; Jin et al., 2019; Mohanty & Sethi, 2019). It is widely observed that for catching-up countries, such as Brazil, the international transfer of technology, through trade, or FDI, is an important source for growth (Goedhuys & Veugelers), which stimulates innovation and productivity in local industries (Castellacci & Natera, 2016; Wang et al., 2010). Considering the role of FDI, this study is based on panel data from Brazilian manufacturing industries, covering the period of 2000–2014. It will be tested empirically how FDI inflows moderates with internal R&D, exports and foreign technology acquisitions. Regarding the manufacturing industry sectors, the segments of high-technology and medium- and low-technology intensity will be analysed, once FDI does not work uniformly in all industry sectors (Mohanty & Sethi, 2019) and differences could appear across industries (Dunning & Lundan, 2008; Javorcik, 2006; Kafouros, 2008).

OPEN IN VIEWER

Table 1.

Inward FDI Inflows by Country in Latin America

FDI Inflows
Region/economy	2013	2014	2015	2016	2017	2018
Latin America and Caribbean	184.392	161.205	155.912	135.405	155.405	146.720
Argentina	9.822	5.065	11.759	3.260	11.517	12.162
Bolivia	1.750	657	555	335	712	255
Brazil	59.089	63.846	49.514	52.751	67.583	61.223
Chile	20.326	23.671	20.011	11.942	6.856	7.160
Colombia	16.209	16.167	11.723	13.850	13.836	11.010
Ecuador	727	772	1.322	767	618	1.401
Guyana	214	255	122	58	212	495
Paraguay	245	412	308	371	456	454
Suriname	188	164	267	309	161	190
Uruguay	3.460	2.328	917	−1.181	−911	−626
Venezuela	2.680	−1.028	769	1.068	−68	956
Central America	58.315	40.903	46.961	42.861	42.887	15.716
Transition Economies	83.764	56.762	36.394	64.654	47.538	34.218

Source: Adapted from UNCTAD (2019).

Note: Values in billions of US dollar.

This article is organized into six sections including ‘Introduction’. The second section discusses the ‘Theoretical Framework’, with the respective grounding literature, followed by the ‘Objective’ and ‘Rationale of the Study’. The third section presents the ‘Methodology’. The fourth section goes on to demonstrate the ‘Discussion and Results’. The fifth section highlights the ‘Conclusions and Managerial Implications’, and finally in the sixth section we present ‘Limitations and Implications for Future Studies’.

Theoretical Framework

Wang and Kafouros (2009) argue that there are three main channels which improve the innovation performance in an economy: (a) internal R&D; (b) exports; and (c) foreign technology acquisitions. The first channel, (a), consists in own internal R&D expenditures, which benefits firms in one country to create new products and processes (Coe & Helpman, 1995; Crespi & Zuniga, 2012; Kafouros, 2008). The second, (b), exports, which can foster growing market size in developing countries (Keller, 2010; Mohanty & Sethi, 2019; Wang et al., 2010). The third, (c), foreign technology acquisitions through imports of machinery and technology equipment, is an important channel for foreign technology spillover (Frank et al., 2016; Fu et al., 2011; Narula & Wakelin, 1998).

Based on these assumptions, we believe that the outlined channels do not improve an industry sector’s innovation performance in an isolated manner, making it necessary to moderate with the country openness, defined by FDI inflows (Castellacci & Natera, 2016). We adopt this idea once the level of foreign presence is low and the possibility of improving innovation performance through inward FDI is expected to be lower (Wang & Kafouros, 2009).

Methodology: Data Sources, Empirical Model and Variables

Empirical Model

Equation (1) shows the models for testing formulated hypotheses based on the link between innovation performance (dependent variable) and independent variables, by both direct and the moderated effects, between inward FDI and R&D, inward FDI and exports and inward FDI and technology acquisition.

γ = α 0 + β 1 FDI i , t + β 2 R&D i , t + β 3 Exports i , t + β 4 Technology acquisition i , t + β 5 ( FDI i , t × R&D i , t ) + β 6 ( FDI i , t ×Exports i , t ) + β 7 ( FDI i , t × Technology acquisition i , t ) + β 8 Labour i , t + β 9 Capital intensity i , t + β 10 Size i , t + β k Year k , t + β j Industry j , t + ε i t

(1)

where γ represents the dependent variable and α the constant. Independent variables are defined by inward foreign direct investments (FDI), internal research and development (R&D), (Exports) and foreign technology acquisition (Technology acquisition) in an industry sector i in year t. The moderated effect is defined by (FDI × R&D), (FDI × Exports) and (FDI × Technology acquisition). The control variables are represented by the number of scientists working with R&D indicated by (Labour), capital (Capital intensity) and industry sector size by (Size). Time (Year) and industry dummy (Industry) variables were adopted to verify any idiosyncrasies across time and industries. Finally, ε denotes the residuals.

Discussion and Results

Table 2 shows the pairwise correlation matrix presenting the means, standard deviations and correlations of the variables. The correlations between ‘exports’ and ‘R&D’ (0.68), ‘capital intensity’ and ‘exports’ (0.69) and ‘capital intensity’ and ‘R&D’ (0.76) are positive moderate and high, hinting at the possibility of multicollinearity. Other coefficients exhibit negative and moderate correlations, such as ‘labour’ and ‘R&D’ (–0.63) and ‘labour’ and ‘exports’ (–0.65). These occur due to the composition of these variables. Considering this fact, we computed the variance inflation factors (VIFs) to check for multicollinearity issues. The highest VIF stands at 4.11 while the lowest value is 1.08. On average, VIFs stands at 2.21. According to Hair Jr et al. (2009), one VIF < 10 is a confirmation that there are no issues with the presence of multicollinearity.

OPEN IN VIEWER

Table 2.

Correlation Matrix: Pairwise Correlation Coefficients Between the Variables

Variables	Mean	Std. Dev.	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
(1) Innovation	0.012	0.045	1.000
(2) FDI	0.000	0.000	–0.078	1.000
(3) R&D	0.037	0.097	0.423*	–0.098	1.000
(4) Exports	0.003	0.007	0.491*	–0.092	0.689*	1.000
(5) Technology acquisition	0.017	0.011	–0.091	–0.155	–0.123	–0.165	1.000
(6) Labour	–0.190	0.468	–0.467*	0.138	–0.635*	–0.655*	–0.100	1.000
(7) Capital	6.597	3.181	0.494*	–0.065	0.761*	0.691*	–0.164	–0.877*	1.000
(8) Industry sector size	17.634	1.220	–0.001	–0.077	0.137	–0.006	–0.461*	–0.041	0.305*	1.000

Source: The authors.

Note: * Indicates significance at the p ≤ 0.01 level of confidence.

OPEN IN VIEWER

Table 3.

Estimates by One-step and Two-step System GMM

Variables	Model 1		Model 2
	One-step System GMM		Two-step System GMM
FDI	0.170***	(0.054)	0.304***	(0.108)
R&D	–0.012ns	(0.155)	–0.142ns	(0.120)
Export	0.397ns	(1.969)	–0.084ns	(1.424)
Technology acquisition	6.201**	(2.783)	13.266***	(4.491)
(FDI × R&D)	–0.072*	(0.040)	–0.069***	(0.023)
(FDI × export)	0.034ns	(0.028)	0.069***	(0.022)
(FDI × technology acquisition)	–0.130**	(0.053)	–0.288***	(0.104)
Labour	0.147ns	(0.086)	0.218***	(0.074)
Capital	–0.034ns	(0.025)	–0.096***	(0.036)
Size	–0.011ns	(0.009)	–0.013ns	(0.014)
Industry dummy	0.125ns	(0.080)	0.205**	(0.084)
Time dummies	Yes		Yes
Prob > F	68.61***
Wald χ2			5028.61***
Number of instruments	22		22
AR (1) (no serial correlation in 1st difference)	p = 0.019		p = 0.468
AR (2) (no serial correlation in 1st difference)	p = 0.077		p = 0.162
Sargan test for instrument validity	p = 0.000		p = 0.000
Hansen (robust) test for instrument validity	p = 0.215		p = 0.282

Source: The authors.

Notes: (1) In model 1, the robust standard errors are in parentheses.

(2) Time dummies are included (based on the survey waves 2000, 2003, 2005, 2008, 2011 and 2014).

(3) Industry dummy variable included: 1, if the industry performs in the high-technology industries sector, and 0, if it performs in the medium- and low-technology industry sectors.

(4) In model 2 the standard errors are included in the parentheses.

(5) ns = Non-significant.

(6) * p ≤ 0.10, ** if p ≤ 0.05; *** if p ≤ 0.01.

Considering the composition of the dependent and independent variables and the endogeneity issues, we estimate model 1 using the one-step system generalized method of moments (GMM) estimation (Arellano & Bond, 1991). The use of internal R&D expenditure can bring endogeneity to the model due to measurement errors of the interest variable (Taveira et al., 2019). In addition, we adopt the two-step system GMM to estimate model 2. The two-step system GMM allows us to control for unobserved heterogeneity and a potential simultaneity between ‘R&D’, ‘exports’, ‘capital intensity’ and ‘labour’.

Table 3 shows the results for models 1 and 2. The two-step makes the coefficients more efficient and robust than the present coefficient in model 1. Considering this, the potentially endogenous variables have the correction recourse considering the external instruments to the model. It is important to mention a short note on the satisfactory diagnostics tests for model 2. The Wald χ² test indicates the joint significance and a good explanatory power of the variables. The first- and second-order serial correlation and the heteroskedasticity-robust Hansen test are shown as adequate. However, the Sargan test shows itself as very sensitive to the increase of the number of instruments (Roodman, 2009). Nonetheless, since our model is based on system GMM, this does not invalidate the instruments considering the Hansen test.

The results indicate for model 2 that ‘FDI’ has a statistically and positive coefficient ( β = 0.304 , p ≤ 0.01 ) . This shows that the FDI presence, generally, is positive to Brazilian manufacturing industries, at least during short periods. Another important result from Model 2 is related to ‘technology acquisition’, which is shown to have a statistically and positive significant coefficient β = 13.266 , p ≤ 0.01 . Our result is in line with that of Calegario et al. (2014) and Mardones and Zapata (2019), at least for import-oriented industries during short periods.

However, this result indicates the high dependence of foreign technology. Frank et al. (2016) argue that technology acquisition is still the main investment category in terms of innovation activities in Brazil. Taveira et al. (2019) assert that companies do not invest heavily in internal technological development such as R&D. This strategy is particularly pertinent in Brazilian companies, since, in most cases, they have a wrong perception about the usefulness of the technology acquisition approach for innovation (Frank et al., 2016), showing interest in upgrading outdated technologies instead of focusing on developing new technologies. In particular, besides investing in R&D and own technology is expensive for a large number of industries; the importing substitution is preferable.

We found evidence of the interaction effect between ‘FDI’ and ‘R&D’ in model 2 with significant coefficient, but with a negative sign β = − 0.069 , p ≤ 0.01 . H₁ predicted a positive direct relationship between inward FDI and R&D and was not supported through the innovation performance output. In general, R&D and innovation processess occur on firms' headquarters located in more developed economies. Castellacci and Natera (2016) identified for LAC that the combination of FDI and R&D does not boost the innovation performance, and Benavente et al. (1997) assert that LAC, including Brazil, presented a decline in technological sophistication. The reason is that enterprises are now more inclined to operate as licensees of foreign ones, assembling products which are part of a wider range of goods and importing other items in the range to be sold domestically (Benavente et al., 1997) or adapting them to the foreign market (Brown & Guzmán, 2014).

Following the analysis, it is possible to observe in model 2 the positive relationship between ‘FDI’ and ‘exports’ β = 0.069 , p ≤ 0.01 . H₂ predicted a positive direct relationship between ‘FDI’ and ‘exports’ and was supported through the innovation performance output. In this case, foreign companies invest in local industries with interest in exports activities, which is in line with Jawaid et al. (2016). Our result suggests the same finding from Brown and Guzmán (2014), which argued that the Mexican manufacturing industries receive large investments of FDI related to exports. However, the interaction between inward FDI and technology acquisition in model 2 does not show a positive coefficient β = − 0.288 , p ≤ 0.01 . H₃ predicted a positive direct relationship between inward FDI and technology acquisition and was not supported through the innovation performance output. This indicates that foreign capital presence is not related to technology acquisitions. We argue that import substitution is preferable.

Concerning the control variables in model 2, we identified a positive and significant effect on ‘labour’ β = 0.218 , p ≤ 0.01 . This is an indication that labour is important for improving innovation performance. However, local industries find it difficult to develop internal R&D activities, since the cost to develop their own technology is higher than that of importing foreign ones. Considering this, it is necessary for labour to have enough qualifications to operate foreign technology. ‘Capital intensity’ is statistically significant with a negative coefficient β = − 0.096 , p ≤ 0.01 . This indicates possible high capital costs in Brazil, compromising the innovation performance. High capital costs are also a common issue in Latin America. Considering this, we argue that the innovation performance could show positive gains after 2 or 3 years after the capital investments. In addition, the industry ‘size’ variable does not show any significance across models 1 and 2.

Aiming to capture a better result across industries, we adopted the industry dummy variable. The result in Model 2 shows a significant and positive coefficient β = 0.205 , p ≤ 0.01 , indicating high-technology industries have a higher innovation performance than medium- and low-technology industries, which is in the same direction as Bruhn and Calegario (2014). The local industries need to develop the internal capacity to explore this foreign knowledge for the long term. To this end, industries, especially medium- and low-technology industries, need government subsidies to develop or adapt their own sources of knowledge to leave the path of buying foreign technology exclusively. Furthermore, more R&D-related FDI is recommended in high-technology industries.

Conclusion and Managerial Implications

The aim of this study was to analyse the main local sources of knowledge that moderate with inward FDI and improve the local industry innovation performance. The empirical results regarding the interaction between several local factors and the FDI inflow suggest a positive impact of FDI on local innovation. In addition, the interaction effect between FDI and exports has a positive impact, but R&D-related FDI does not improve the performance, since this activity occurs in headquarters in developed countries.

Furthermore, inward FDI does not show a positive result moderating with technology acquisition, suggesting the import substitution behavior. This could lead to buying only external technology and not developing their own, a reason why internal R&D investments do not have an impact on Brazilian manufacturing industries. One critical result is related to the high-technology industry sector, which shows higher innovation performance than medium- and low-technology sectors. Our findings infer some managerial and policy implications. In this case, the policymakers should give attention and protection to the potential of the medium- and low-technology sectors. It is necessary to attract more R&D-related FDI, specially in industries from high-technology sectors, to increase import substitution.

Limitations and Implications for Future Studies

The purpose of this research was not to generate a statement but to understand the impacts and possibilities of FDI inflows. Notwithstanding that the data did not represent the firm level, we captured an average effect using an industry survey. Future studies could explore the adoption of firm-level data to improve the results, with other important variables and other types of interactions between variables which were not observed here.