The Role of FDI on Exports Performance: Evidence from the Mexican States

Abstract

This article examines manufacturing export determinants across Mexican states and regions from 2007 to 2015. Paying particular attention to the role of FDI, the analysis considers internal and external determinants of manufacturing exports under static and dynamic panel data methods. Several interesting results were obtained. First, the ratio of manufacturing to total GDP is the most consistent determinant of exports performance, regardless of the estimation method or specification employed. Second, static panel data estimations under GMM techniques suggest different sensitivity to FDI across regions, with the Mexico-U.S. border region observing the most substantial short-term effect of FDI on manufacturing exports. Finally, using dynamic panel data methods, we found significant persistence and similar long-term effects of FDI across most of the regions.

Keywords

exports foreign direct investment Mexico panel data

Introduction

Over the last three decades, the Mexican economy has undertaken significant structural changes in terms of its relationship with the rest of the world. The country shifted its economic development strategy from a protectionist policy, based mostly on import-substitution industrialization and an oil-dependent economy, to a liberal policy of economic openness with an export orientation, especially with respect to manufactured goods (Williamson 1990; Ten Kate 1992). Following its accession to the World Trade Organization (formerly known as the General Agreement on Tariffs and Trade) in 1986, and the enactment of the North American Free Trade Agreement (NAFTA) in 1994, Mexico’s trade and capital flows rose significantly. Moreover, since then Mexico has strategically promoted free trade by signing twelve free trade agreements with forty-six countries and twenty-eight agreements for the promotion and reciprocal protection of investments.

Today, exports and FDI flows are two crucial engines for the Mexican economy—especially those associated with the manufacturing sector. Figure 1 presents the evolution of total Mexican exports and FDI inflows between 1990 and 2016. From the period before the start of NAFTA to the most recent years, exports and FDI have experienced remarkable increases of about nine-fold and six-fold in value, respectively. Although the rise in exports and FDI has been remarkable, its effect has not been felt homogeneously across all Mexican states and regions. While manufacturing activity and its corresponding exports have become a central element for the economies of some states, others hardly benefit, being largely absent from export-related businesses—despite having similar access to foreign markets.

Figure 1.

Mexico’s total exports and FDI flows (millions of U.S. dollars). Source: Prepared with data from the World Bank, World Development Indicators.

Figure 2 shows the ratio of manufacturing exports to GDP for the different states and the four regions of Mexico.¹ The figure illustrates a very dissimilar pattern of foreign trade across the country, with a significant concentration of exports along the (northern) Border region, where every state performs well above the national average (22.7%) and some states possess exports that exceed the size of its GDP (e.g., Chihuahua with 113%). Contrastingly, in the South region none of the states exceeds the national average and for some of them manufacturing exports represent less than 1% of GDP (e.g., Campeche, Quintana Roo, and Guerrero). Although FDI is more volatile in nature, Figure 3 shows a geographical distribution like that of exports.²

Figure 2.

Average State Exports to GDP Ratio 2007–2015 (%). Source: Own estimations using data from INEGI.

Figure 3.

Average Manufacturing Exports and FDI Flows, 2007–2015 (real pesos of 2008). Source: Own estimations with data from INEGI and Secretaría de Hacienda.

Considering the above patterns of trade and FDI, this paper studies the determinants of manufacturing exports across Mexican states, while paying special attention to the impact of foreign capital flows. A number of papers have studied the determinants of exports in industrial and emerging economies. A first strand of literature mainly examines the causal relationship between exports and FDI. Overall, studies analyzing causality report mixed results. For instance, Boubacar (2016) employs annual data on U.S. FDI to twenty-five OECD countries between 1999 and 2009. He uses spatial econometrics panel data techniques and finds a complex bidirectional causality between FDI and exports. Goswami and Saikia (2012) also analyze causality making use of aggregate data for India’s exports, FDI, GDP, and gross fixed capital formation. Estimating a vector error-correction model, they report the presence of bidirectional causality between exports and FDI. Ahmed, Cheng, and Messinis (2011) analyze causality for Ghana, Kenya, Nigeria, South Africa, and Zambia, employing an error-correction model to test for Granger causality. Their findings show bidirectional causality between exports and FDI in Ghana and Kenya, Granger causality from FDI to exports in South Africa and from exports to FDI in Zambia. Similarly, Hsiao and Hsiao (2006) analyze causality in China, Korea, Hong Kong, Singapore, Malaysia, Philippines, and Thailand, using time series for 1986–2004. Estimating panel data Granger causality test between GDP, exports and FDI, they report individual direct causality from exports to FDI only in China, but from FDI to exports in the cases of Taiwan, Singapore, and Thailand. For the eight countries in the sample analyzed together, they only observe direct causality from FDI to exports.³

In a second strand found in the literature, several other studies have followed a multivariate approach that not only looks at causality between exports and FDI, but also at other relevant determinants of exports. Many of those studies have made use of industry- or firm-level data. For instance, Franco (2013) employed data pertaining to U.S. FDI on sixteen OECD countries from 1990 to 2001, separating asset-seeking from asset-exploiting FDI. Employing panel data techniques, she addresses endogeneity problems caused by FDI and exports, and observes that market-seeking FDI influences export intensity more than other forms of FDI. Also, Rahmaddi and Ichihashi (2013) analyze Indonesia’s manufacturing exports by industry from 1990 to 2008 using fixed effects panel data methods. They find that higher levels of FDI enhance the performance of manufacturing exports and that FDI effects on exports vary across manufacturing industries with capital-intensive, human-capital-intensive and technology-intensive exporting industries gaining the most from FDI inflows. Finally, Karpaty and Kneller (2011) analyze manufacturing firms in Sweden with at least fifty employees during the years 1990–2001. Using the two-stage probit procedure proposed by Heckman (1979), they find that FDI has positive effects on Swedish exports.⁴

In a third strand of literature, some studies have examined the effects of FDI on exports at either the subnational or regional level. Perhaps due to the absence of data on exports for other countries, the existing evidence studying the regional influence of FDI on exports seems to be concentrated on Chinese regions. For instance, Zhang (2015) employs data for thirty-one manufacturing sectors and thirty-one regions of China over 2005–2011. Using panel data fixed effects and instrumental variables techniques, he observed that FDI has exerted a significant influence on China’s export success and that its absorptive capacity is reinforced through human capital availability. Similarly, Zhang and Song (2000) used data from twenty-four Chinese provinces for 1986–1997 and employed ordinary and generalized least squares techniques. Their paper provides evidence on the role of FDI in promoting Chinese exports and reports that a 1% increase in the level of FDI in the previous year is associated with a 0.29% increase in exports in the following one. Finally, Sun and Parikh (2001) analyze a panel of twenty-nine provinces across three regions of China for a period of eleven years (from 1985 to 1995). They find that the strength of the impact of exports on GDP varies significantly across regions. Their results also imply that the relationship between exports (FDI) and economic growth depends on regional, economic and social factors.

Evidence on export determinants for Mexico is less abundant and mostly focuses on the causality between exports and FDI while employing aggregate data (see, for instance, Vasquez-Galán and Oladipo 2009; De la Cruz and Núñez Mora 2006; Pacheco-López 2005; Cuadros, Orts, and Alguacil 2004; Alguacil, Cuadros, and Orts 2002, among others). A paper that uses a different approach to that of simple causality analysis is Aitken, Hanson, and Harrison (1997). They studied 2,104 Mexican firms for 1986–1990 employing a Probit specification to analyze a firm’s probability of exporting. They found that foreign firms are a catalyst for domestic firms and a firm’s probability of exporting is positively correlated with its proximity to multinational firms.

In this paper, we take a regional approach to look at internal and external factors that affect manufacturing exports in Mexico. We pay special attention to agglomeration economics resulting from the presence of local manufacturing activity and the stock of foreign capital. We think that this is an important link in the FDI-exports relationship that has been overlooked in some of the previous existing empirical literature. We develop first a partial equilibrium model that disentangles the FDI-export link and lays the foundations for our empirical analysis. Afterward, we rely on different static and dynamic panel data model specifications and various econometric techniques that allow us to control for potential endogeneity problems and identify short- and long-term effects of FDI on manufacturing exports.

Several interesting findings are obtained in this paper. First, regardless of the method or specification employed, we observe that the most consistent determinant of exports is the ratio of manufacturing to total GDP. This result is consistent with the idea that agglomeration economies are necessary for the existence of a robust exporting platform in each state and region. Second, using GMM estimation techniques to control for endogeneity, two important results were obtained. On the one hand, estimating a dynamic panel specification, we observed significant export persistence, but, most importantly, similar long-term effects coming from FDI across most regions—with only slightly less sensitivity to FDI in the Center region. The intuition for this result is that, once we consider long-term export dynamics, there seems to be little difference in how regions respond to FDI variations. On the other hand, under our static specification, the results suggest that, in the short-term, states show different sensitivities to FDI across regions, with the Border region experiencing the strongest effect of FDI on manufacturing exports, followed by the North, Center and South regions. This result is consistent with the idea that, as we move further away from the U.S.-Mexican border, the short-term sensitivity of exports to FDI is less significant due to the lower presence of agglomeration economies in the form of foreign capital.

The rest of this paper is organized as follows. The second presents a brief literature review on the connection between FDI, agglomeration economics and exports. The third section describes the data used in the analysis and presents some descriptive statistics. The fourth section develops a theoretical model that lays the groundwork for our empirical assessment. It also defines the static and dynamic empirical models that are employed to study manufacturing exports determinants. The fifth section presents the results of the empirical estimations. Finally, the last section concludes.

Literature Review on FDI, Agglomeration Economies and Exports

In line with vertical foreign direct investment theories (Markusen and Maskus 2002), a low-wage location, good transport infrastructure, and a fragmented chain of production, where goods at different stages may cross borders many times at no or low cost, are all factors that are responsible for Mexico’s significant increase in FDI inflows in recent years (Cuevas, Messmacher, and Werner 2005). Indeed, the country has transformed into an export-platform, to a large extent, because of FDI inflows in the manufacturing sector. Waldkirch (2003) analyzes the effect of the North American Free Trade Agreement (NAFTA) on FDI in Mexico and found that it had a significantly positive impact.⁵ Similarly, South (2016) also points out the crucial role that NAFTA played in the evolution of Mexico’s manufacturing exports, which are no longer primarily from the maquiladora industry, but are now dominated by higher value-added manufacturing products.

Undeniably, regional integration has been fundamental to attract FDI flows to Mexico and promote its exporting sector. According to Blomström and Kokko (1997), Mexico’s accession to NAFTA led to significant policy changes, which explain the growing interest of foreign multinationals in the country. Agglomeration economies have also been a fundamental factor to attract industrial activity, primarily through the so-called cumulative causation process (Ottaviano and Puga 1998). Agglomeration has helped firms obtain productivity gains by concentrating in geographical areas, which results in lower transportation costs, access to a specialized labor pooling, and potential technological spillover effects (Combes and Gobillon 2015).

Gains from agglomeration can be traced back to Marshall (1890) and more recently to Duranton and Puga (2004), who have proposed other agglomeration mechanisms based on the concepts of sharing, matching, and learning. In the case of Mexico, Jordaan (2012) observes that FDI causes positive externalities among Mexican firms and that agglomeration economies enhance these spillovers. Moreover, he suggests that agglomerations of foreign and local manufacturing firms have the largest positive effect on the probability of attracting new FDI flows.

The presence of agglomeration economies also influences firms’ export behavior in different ways. Malmberg, Malmberg, and Lundequist (2000) examine the impact of various types of agglomeration economies on the export performance of Swedish exporting firms. They find that export performance is related to an increasing scale of operations, the size of the company being the most important factor, which is associated with urban agglomerations and corporate groups. Also, Greenaway and Kneller (2008) study the exporting behavior of manufacturing firms in the United Kingdom and conclude that regional and industry agglomeration are relevant to the chances of successful entry of new exporting firms. Similarly, Koenig (2009) analyzes the export decisions of French manufacturing firms and shows evidence of export-agglomeration economies; that is, proximity to local exporters positively influences the probability of starting to export to a given country.

Rodríguez-Pose et al. (2013) report that agglomeration effects, education, and transport infrastructure endowment play a particularly relevant role in the export propensity of Indonesian manufacturing firms, while export spillovers increase export intensity. Likewise, Ito, Xu, and Yashiro (2015) find that the agglomeration of incumbent Chinese manufacturing exporters contributes significantly to export participation. Interestingly, they estimate that the spillover from the agglomeration of exporters exists in both coastal and inland areas and is more relevant among inland firms. Finally, Ramos and Moral-Benito (2017) examine Spanish manufacturing export firms selling to specific foreign markets and document the existence of destination-specific agglomeration economies. They report that firms exporting to countries with worse institutions, a different language and different currencies are significantly more agglomerated. These results suggest that the value provided by agglomeration is higher for those destinations where entry could be more difficult.

Following this brief literature review, we explore the impact of agglomeration economies—in the form of the FDI stock and the proportion of manufacturing economic activity—on Mexico’s exporting sector.

Data

Our sample comprises all thirty-two Mexican states. For the purpose of our analysis, the country is divided into four large regions, following the regionalization proposed by Banco de México (see Figure 2). The period of analysis is determined by the availability of information on manufacturing exports and extends from 2007 to 2015. Our data come from various sources. Exports and manufacturing GDP come from Mexico’s National Institute of Statistics INEGI (Instituto Nacional de Estadística y Geografía). Foreign Direct Investment flows were obtained from Mexico’s Ministry of the Economy (Secretaría de Economía). The real exchange rate is from Mexico’s central bank (Banco de México) and the U.S. manufacturing productivity index came from the U.S. Federal Reserve Economic Data.

Since FDI flows are highly volatile, we build an FDI stock using the perpetual inventory method.⁶ Ever since the seminal works by Mankiw, Romer, and Weil (1992) and Islam (1998), who study the neoclassical determinants of economic growth, the empirical literature has intensely used the perpetual inventory method to calculate capital stocks. In this paper, we also follow Wacker’s (2013) recommendation, who concludes that FDI stock data is generally recommendable to measure the economic activity of multinational firms, and Mollick and Cabral (2009) who use the perpetual inventory method to estimate capital stocks for the Mexican manufacturing industry.

In calculating FDI stocks, we take advantage of the fact that FDI data at the state level is available from 1999. In Table 1, we present some descriptive statistics for the full sample and each of our four regions. As can be observed, average state exports are considerably more substantial in the Border region. States in the Center region are the second most important average exporters, followed by the states in the North and South regions. Looking at the FDI stock at the end of the sample period, in 2015, we observe that there is very little difference between the Border and Center regions (38.8% and 38.0% of the total, respectively), with the latter surpassing the former just marginally.⁷ The North region FDI stock is less than half that of the Border region (16.7%), and the South region possesses only a small fraction of total stock (6.4%). Figure 3 provides a picture of the geographical location of exports and FDI across states. It is clear from this picture that there is a close relationship between the distribution of exports and FDI, with a significant geographical concentration in the Border and Center regions.

Table 1.

Average Regional Indicators by State (Millions of 2008 Pesos).

Region	Average Total Exports	FDI Stock in 2015	Total GDP	Manufacturing GDP	Manufacturing to Total GDP (%)
Border	358,447	147,146	569,847	133,368	23.1
North	47,335	37,960	272,833	46,672	14.5
Center	85,979	108,071	735,549	130,616	24.0
South	13,992	18,267	411,576	36,822	9.0
Total	106,994	71,037	478,888	81,451	17.1

Source: Own calculations with data from INEGI and Secretaría de Economía.

In Table 2, we review the correlation between the main variables of our model. The first column shows the correlation between exports and every one of its possible determinants. As expected, we observed a positive correlation between exports and FDI, state GDP, the U.S. manufacturing activity index, the real exchange rate, and the ratio of manufacturing to total GDP within each state. A potential problem of multicollinearity is only observed for the correlation between the FDI stock and state GDP (0.84). To assess this potential problem in more detail, we calculate the variance inflation factors (VIF) for the set of variables in Table 2. Jointly assessed, all variables present a mean VIF of 1.94 and individually they are all smaller than 4, which suggests that our model is not beleaguered by multicollinearity problems.^8,9

Table 2.

Correlation Matrix.

Variables	Average Total Exports	FDI Stock	State GDP	U.S. Index of Manufacturing Production	Real Exchange Rate	Ratio of Manufacturing to Total GDP
Average total exports	1
FDI stock	0.6842	1
State GDP	0.5597	0.8361	1
U.S. Index of manufacturing production	0.0516	0.0349	0.0503	1
Real exchange rate	0.0967	0.1063	0.0278	−0.1880	1
Ratio of manufacturing to total GDP	0.4612	0.2917	0.5026	0.0286	−0.0057	1

The Model

Theoretical Framework

Our empirical model is based on the partial equilibrium models proposed by Cabral and Mollick (2011) and Barrell and Pain (1997, 1999). In this model, the production of exports, X, by state i and time t is given by the following Cobb-Douglas production function with constant returns to scale:

X_{i t} = A_{i t} K_{i t}^{α} L_{i t}^{1 - α}

(1)

where, as usual, A, K and L stand for technology, capital, and labor, respectively. Taking (natural) logs and using circumflexes to represent them, the expression above is transformed into

{\hat{x}}_{i t} = {\hat{a}}_{i t} + α {\hat{k}}_{i t} + (1 - α) {\hat{l}}_{i t}

(2)

In equilibrium, the labor required to satisfy the production of exports of each state depends on the real exchange rate, $z$ (a demand factor), and the relative importance of the manufacturing sector in the state, ${\hat{y}}^{m}, (a supply factor),$ such that:

{\hat{l}}_{i t} = l ({\hat{z}}_{t}, {\hat{y}}_{i t}^{m}) = δ_{1} {\hat{z}}_{t} + δ_{2} {\hat{y}}_{i t}^{m}

(3)

where $δ_{1}$ and $δ_{2}$ are the sensitivity of labor to the real exchange rate and the proportion of manufacturing activity in the economy. The real exchange rate is state invariant. In equilibrium, a depreciation of the real exchange rate makes the cost of hiring labor lower, and thus labor demand cheaper, leading to a higher demand of labor in equilibrium. Meanwhile, as the proportion of manufacturing activity to total production in the economy increases, the pooling of labor supply available for the exporting sector rises. When this happens, entity i becomes a more competitive place for the production of exports and more firms locate next to each other, as they benefit from positive spillovers that arise from agglomeration.

The stock of foreign direct investment, F, is a fraction $λ_{i}$ of the total stock of capital K in state i:

F_{i t} = λ_{i} K_{i t}

(4)

taking logs and solving for $\hat{k}$ , this expression is simply

{\hat{k}}_{i t} = - {\hat{λ}}_{i} + {\hat{f}}_{i t}

(5)

The technology in equation (1), $A_{i t},$ depends on the relative influence of domestic (Y) and foreign output (Y*), as well as on a state-specific fixed factor (effect), $γ_{i}$ , such that:

{\hat{a}}_{i t} = γ_{1} {\hat{y}}_{i t} + γ_{2} {\hat{y}}_{t}^{*} + γ_{i}

(6)

This partial equilibrium model framework gives rise to the empirical equations that we define next and provides the foundations for the empirical model we test in the following section.

Empirical Model

Substituting (3), (5), and (6) into (2), adding a constant and an error term, we obtain the following empirical equation:

{\hat{x}}_{i t} = β_{o} + β_{i} + β_{1} {\hat{f}}_{i t} + β_{2} {\hat{z}}_{t} + β_{3} {\hat{y}}_{i t}^{m} + β_{4} {\hat{y}}_{i t} + β_{5} {\hat{y}}_{t}^{*} + u_{i t}

(7)

where $β_{o}$ is the constant term, $β_{i} = γ_{i} - {α \hat{λ}}_{i}$ , $β_{1} = α,$ $β_{2} = (1 - α) δ_{1}$ , $β_{3} = (1 - α) δ_{2}$ , $β_{4} = γ_{1}$ , $β_{5}$ = $γ_{2}$ and $u_{i t}$ is the error term.¹⁰

In our empirical model, ${\hat{x}}_{i t}$ represents the log of total Mexican manufacturing exports by state i at time t. Exports depend on external and domestic factors. Across the external factors, ${\hat{f}}_{i t}$ is the log of the FDI stock, ${\hat{z}}_{t}$ is the log of the real exchange rate, and ${\hat{y}}_{t}^{*}$ is the log of the US industrial production index. The two domestic factors that influence exports are state GDP, ${\hat{y}}_{i t}$ , and the ratio of manufacturing to total GDP, ${\hat{y}}_{i t}^{m}$ . We expect each of our foreign and domestic control variables to exert a positive effect on manufacturing exports.

Two of our control variables, the FDI stock and the ratio of manufacturing to total GDP, capture agglomeration economies that emerge from the presence of foreign capital and manufacturing activity across states. Following the works by Kinoshita and Campos (2003), the FDI stock captures the agglomeration effect arising from the location of foreign capital. In the case of manufacturing activity, we can clearly see in $β_{3} = (1 - α) δ_{2}$ that agglomeration effects emanate from the interaction between the intensity in the use of labor $(1 - α)$ and the relative importance of the manufacturing activity in total production of state i ${(δ}_{2})$ . Intuitively, by locating in the same state as other manufacturing firms, exporters benefit from positive externalities that arise from knowledge spillovers, available specialized labor, and intermediate inputs (Kinoshita and Campos 2003).

In order to include exports inertia in the model, we augment the equation in (7) by adding lagged exports, which captures exports persistence. Our dynamic specification is thus defined by:

{\hat{x}}_{i t} = β_{o} + ρ {\hat{x}}_{i t - 1} + β_{i} + β_{1} {\hat{f}}_{i t} + β_{2} {\hat{z}}_{t} + β_{3} {\hat{y}}_{i t}^{m} + + β_{4} {\hat{y}}_{i t} + β_{5} {\hat{y}}_{t}^{*} + u_{i t}

(8)

where $ρ$ captures exports persistence or inertia. In a dynamic specification like equation (8), the lagged dependent variable on the right-hand side would be correlated with the error term, invalidating the results obtained through traditional OLS panel estimations. In addition, there are also some potential endogenous variables in our model (e.g., FDI stocks, state GDP, manufacturing to total GDP), which might bias the estimation of equation (7). To deal with these problems, we adopt two different approaches. The first consists of estimating an equation similar to (8), but disregarding the persistence of exports (ρ = 0), which biases the estimation of the model using OLS, and fitting the first lag of all the potential endogenous variables in the model to avoid reverse and simultaneous causation:

{\hat{x}}_{i t} = β_{o} + β_{i} + β_{1} {\hat{f}}_{i t - 1} + β_{2} {\hat{z}}_{t - 1} + β_{3} {\hat{y}}_{i t}^{m} + + β_{4} {\hat{y}}_{i t - 1} + β_{5} {\hat{y}}_{t}^{*} + u_{i t}

(9)

This specification allows us to avoid the use of potentially invalid or weak instrumental variables (Clements et al. 2012). The estimations then rely on the Hausman test to establish whether fixed or random effects methods are more appropriate. Moreover, to revise how these determinants change from one region to another, we partition our sample of thirty-two states in the four regions, Border, North, Center, and South, described above.

Recently, Bellemare, Masaki, and Pepinsky (2017) and Reed (2015) have criticized the use of lagged regressors to control for endogeneity. Henceforth, the second approach we follow to deal with endogeneity is the use of system generalized method of the moments (SGMM) techniques to solve the consistency problem of OLS in (8), as well as potential problems of reverse and simultaneous causation. Arellano and Bond (1991) and Blundell and Bond (1998) propose a model in which lagged differences are employed in addition to the lags of the endogenous variables, producing more robust estimations when the autoregressive processes become persistent. SGMM estimators are said to be consistent if there is no second order autocorrelation in the residuals, according to the Arellano-Bond test of second-order serial correlation (AB (2)), and if the instruments employed are valid according to the Hansen-J test. To avoid overidentification problems, the instrument set is constrained to its minimum by employing the collapse procedure proposed by Roodman (2009), which restricts our specification to one instrument for each lag distance and instrumenting variable.

The partition of our full sample of thirty-two states into regions would provide us with sub-samples in which the number of time periods (years) is greater than the number of units of analysis (states). Under this scenario, SGMM tends to suffer from problems of overidentification due to the proliferation of instruments. Therefore, rather than splitting our sample, we rely on the interaction between regional dummies and FDI stock to revisit the role of capital flows in the dynamics of manufacturing exports. Consequently, the model to be estimated is:

\begin{array}{l} {\hat{x}}_{i t} = β_{o} + ρ {\hat{x}}_{i t - 1} + β_{i} + β_{1} {\hat{f}}_{i t} + β_{2} {\hat{z}}_{t} + β_{3} {\hat{y}}_{i t}^{m} + + β_{4} {\hat{y}}_{i t} + β_{5} y_{t}^{*} + \\ \sum_{i = 1}^{4} δ_{i} ({\hat{f}}_{i t} * r e g i o n_{i}) + μ_{t} + u_{i t} \end{array}

(10)

where $r e g i o n_{i}$ is a set of dummy variables comprising the four regions defined. We expect each of these interactions to present a positive sign ${(δ}_{i} > 0)$ and $μ_{t}$ to appear in the model only when state-invariant regressors are not included in the model.

Empirical Results

Static Panel Data Estimations

Table 3 reports the estimations of equation (9) using random effects. According to the Hausman test (χ² = 9.15, p-value = 0.1031), random effects are preferred over fixed effects for the estimations of our static model specification. The first column reports the results of the estimations for all thirty-two states, while the rest of the columns describe the results for each of the four regions. Looking at the whole sample of thirty-two states, we observe first that the FDI stock does not appear to be a reliable determinant of exports.¹¹ Meanwhile, state GDP, the real exchange rate, and the ratio of manufacturing to total GDP yield the expected positive sign and are statistically significant at the 1% level.

Table 3.

Static Model Estimations Employing Random Effects.

Variables	National	Border	North	Center	South
FDI stock lagged $({\hat{f}}_{i t - 1})$	0.108	0.420**	0.114	0.407⁺	−0.376
	(0.140)	(0.190)	(0.164)	(0.254)	(0.464)
State GDP lagged $({\hat{y}}_{i t - 1})$	0.939***	0.116	1.001*	1.012**	0.523
	(0.267)	(0.282)	(0.592)	(0.494)	(0.479)
U.S. index of manufacturing production ( ${\hat{y}}_{t}^{*})$	0.159	1.002***	0.423	0.294	0.140
	(0.318)	(0.121)	(0.699)	(0.352)	(0.574)
Real exchange rate lagged ( ${\hat{z}}_{t - 1})$	1.047***	0.672*	1.419***	1.353***	0.736
	(0.238)	(0.382)	(0.488)	(0.378)	(0.587)
Manufacturing to total GDP lagged (( ${\hat{y}}_{i t - 1}^{m})$	0.091***	0.019**	0.090***	0.088***	0.124***
	(0.018)	(0.008)	(0.022)	(0.025)	(0.039)
Constant	−9.863***	−1.690	−13.544*	−16.065***	−0.101
	(3.536)	(2.339)	(8.228)	(5.215)	(9.382)

Number of observations	288	54	90	72	72
R² within	0.115	0.655	0.254	0.723	0.022
R² between	0.719	0.382	0.843	0.514	0.661
R² overall	0.703	0.434	0.826	0.521	0.613

Note: Standard errors robust to heteroskedasticity are reported in parenthesis. The symbols ⁺, *, **, and *** refer to levels of significance of 12%, 10%, 5%, and 1%, respectively.

Moving on to the estimation results for the regions, we observe that the ratio of manufacturing to total GDP is statistically significant for each of the four regions, with coefficients ranging from 0.02 in the Border region to 0.12 in the South region. Considering the ratio of manufacturing to total GDP as a proxy of economies of agglomeration, it makes sense for this variable to be a relevant determinant which increases its magnitude as we move away from Mexico’s northern border with the U.S., where perhaps other factors—such as transportation costs and economic integration with the U.S. economy—could potentially be more relevant. The real exchange rate is statistically significant for each region, except the South. A possible explanation for this finding is that, since manufacturing exports in the South region are not substantial, those states tend to benefit less from the competitive gains that a depreciation of the real exchange rate can bring to the rest of the economy. The state GDP is statistically significant only for the North and Center regions, but not for the Border or South. Finally, we observe that FDI stock is only statistically relevant for the Border (coefficient of 0.42, significant at the 5% level) and Center (coefficient of 0.41, statistically significant at the 12% level) regions. We conjecture that this result responds to the fact that these two regions comprise the largest shares of manufacturing exports in the country: 63% and 20%, respectively, but also have the highest shares of FDI: 39% and 38% of the total stock.

Dynamic Panel Data Estimations

There are several advantages to estimating the model in (10) using SGMM. The first is that, by introducing lagged exports on the right-hand side, we can control for the inertia or persistence of manufacturing exports over time and for more complete dynamics. The second advantage relates to the first and has to do with the fact that since lagged dependent variables perpetuate their effect into the infinite future, we could interpret the estimated coefficients and their significance as long- rather than short-term effects.¹² Assuming the economy is in a steady state, and thus that all variables growth at the same rate, the long-term coefficients for the FDI effects on exports are obtained as: $β_{LR} = β / (1 - ρ)$ . These long-term coefficients, however, merit a word of caution, given that our period of analysis (nine years) is relatively short. Third, by using SGMM we do not need to lag all our potentially endogenous variables by one period. Instead, we can instrument these variables using lags and lagged differences of the ones we consider to be potentially endogenous.

Table 4 presents estimations of the dynamics specifications in equation (10) by employing panel system-GMM techniques. In columns (1), (3) and (5) we consider only the FDI stock, real exchange rate and lagged exports as endogenous, while in columns (2), (4) and (6) the state GDP and the ratio of manufacturing to total GDP are also regarded as endogenous. The results of the model estimations considering all the control variables, analogous to the specification in Table 3, appear in columns (1) and (2). Notice that, to avoid perfect collinearity, one of the interactions, Border × FDIS, is dropped from the regression. With this modification, the coefficient of FDI stock corresponds to the effect of the Border region, and is taken as the reference region. To calculate the effect of FDI on manufacturing exports of, for instance, the North region, the coefficient for the total FDI stock (corresponding to our reference region) must be added to that of the interaction for the FDI stock and the North region (FDIS × North), whenever such coefficients end up being statistically significant.

Table 4.

Dynamic Model Estimations Employing SGMM.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
Lagged exports ( ${\hat{x}}_{i t - 1})$	0.656***	0.655***	0.639***	0.638***	0.633***	0.638***
	(0.104)	(0.117)	(0.093)	(0.103)	(0.093)	(0.098)
FDI stock ( ${\hat{f}}_{i t})$	0.267**	0.217*	0.242***	0.221***	0.261***	0.218***
	(0.106)	(0.125)	(0.082)	(0.078)	(0.091)	(0.082)
State GDP $({\hat{y}}_{i t})$	−0.106	0.120			0.076	0.169
	(0.594)	(0.727)			(0.352)	(0.355)
U.S. manufacturing production index ( ${\hat{y}}_{t}^{*})$	0.308	0.053
	(0.868)	(0.750)
Real exchange rate ( ${\hat{z}}_{t})$	−0.206	−0.157
	(0.524)	(0.355)
Manufacturing to total GDP ratio ( ${\hat{y}}_{i t}^{m})$	0.046***	0.064**	0.050***	0.058***	0.050***	0.061***
	(0.016)	(0.026)	(0.016)	(0.022)	(0.016)	(0.017)
FDI stock × north region	−0.028	−0.023	−0.026	−0.040	−0.016	−0.021
	(0.073)	(0.087)	(0.026)	(0.034)	(0.034)	(0.045)
FDI stock × center region	−0.042	−0.060*	−0.045**	−0.059**	−0.043**	−0.056*
	(0.026)	(0.035)	(0.021)	(0.029)	(0.021)	(0.029)
FDI stock × south region	−0.031	−0.018	−0.036	−0.035	−0.029	−0.026
	(0.056)	(0.050)	(0.038)	(0.034)	(0.034)	(0.032)
Constant	1.168	−0.549	0.679	0.834	−0.490	−1.412
	(5.596)	(7.179)	(0.763)	(0.885)	(3.946)	(4.045)
Number of observations	256	256	256	256	256	256
States	32	32	32	32	32	32
Number of instruments	31	31	26	26	29	29
Second order test of serial correlation	−0.442	−0.512	−0.455	−0.470	−0.419	−0.492
p-value	[0.659]	[0.609]	[0.649]	[0.638]	[0.675]	[0.622]
Hansen test	24.033	24.033	25.764	25.764	25.745	25.745
p-value	[0.291]	[0.291]	[0.137]	[0.137]	[0.216]	[0.216]

Note: Standard errors robust to heteroskedasticity are reported in parenthesis. The symbols *, **, and *** refer to levels of significance of 10%, 5%, and 1%, respectively. The Hansen test reports that under the null the overidentified restrictions are valid. Second order test of serial correlation corresponds to the Arellano-Bond test for serial correlation, under the null of no autocorrelation.

In Table 4, column (1), as expected, lagged exports are statistically significant at the 1% level. With respect to the effect of FDI stock, the coefficient is only statistically significant for the reference region, implying that FDI stock has the same impact on every region. In other words, in the long run, the effect of FDI stock on exports is the same across the country. This result, however, changes when we consider state GDP and the ratio of manufacturing to total GDP in column (2) as endogenous. Here, the coefficient for the Center region is negative and statistically significant at the 10% level. Given this significant interaction, we would interpret that, in the long term, the effect of the Center region (coefficient of 0.455 = (0.217 – 0.060)/(1 − 0.655)) is smaller than for the Border region (coefficient of 0.628).

A problem that we encounter in the regressions of columns (1) and (2) is that, due to the presence of the lagged dependent variable on the right-hand side, there are several variables which are not statistically significant and could be deemed redundant. This drawback is particularly problematic when we are instrumenting some of those irrelevant regressors. To deal with this issue, we follow two different approaches. The first is to estimate the model in columns (3) and (4), omitting those variables that were not statistically significant. The second is to eliminate from the model, in columns (5) and (6), the time-invariant regressors (i.e., real exchange rate and the U.S. manufacturing productivity index) and include time effects instead. The results from following those strategies are consistent with those described before in column (2). Overall, except for the Center region, all others observe a slightly larger impact from the FDI stock on exports in the long run. For the Center region, the coefficient ranges from 0.447 to 0.594, while all other regions range from 0.602 to 0.711. A possible explanation why the Center experiences less sensitivity in its exports to FDI than the rest of the country in the long term is that states in this region have traditionally attracted FDI that is primarily oriented toward serving the domestic rather than export market.

Robustness Checks

The advantage of the model estimated in (8) is that we are able to control for the persistence of FDI and instrument potentially endogenous variables. Nevertheless, if we are merely interested in short-term effects and take advantage of GMM, we can just omit from (8) the lag of manufacturing exports as a regressor and continue to tackle endogeneity issues, as in Table 4. Given that we are now omitting the lagged dependent variable, in addition to the Hansen test, testing for first order serial correlation becomes necessary. In Table 5 we replicate the estimation of Table 4 by employing the static specification in (1) using GMM and interpret the results as short-term effects.

Table 5.

Static Model Employing GMM Estimations.

Variables	(1)	(2)	(3)	(4)
FDI stock ( ${\hat{f}}_{i t})$	0.618***	0.594***	0.533***	0.593***
	(0.162)	(0.146)	(0.154)	(0.110)
State GDP $({\hat{y}}_{i t})$	0.986	0.668*	0.855*	0.892***
	(0.737)	(0.351)	(0.483)	(0.197)
U.S. manufacturing production index ( ${\hat{y}}_{t}^{*})$	−0.595	0.051
	(0.958)	(0.626)
Real exchange rate ( ${\hat{z}}_{t})$	−0.377	0.100
	(0.592)	(0.374)
Manufacturing to total GDP ratio ( ${\hat{y}}_{i t}^{m})$	0.061**	0.054**	0.061**	0.050**
	(0.030)	(0.026)	(0.025)	(0.022)
FDI stock × north region	−0.064	−0.087*	−0.096*	−0.088***
	(0.090)	(0.052)	(0.054)	(0.034)
FDI stock × center region	−0.108***	−0.081*	−0.116***	−0.099**
	(0.035)	(0.045)	(0.033)	(0.049)
FDI stock × south region	−0.169**	−0.150**	−0.200**	−0.172***
	(0.083)	(0.061)	(0.095)	(0.056)
Constant	−4.376	−4.987*	−6.115	−7.027***
	(5.797)	(2.923)	(5.781)	(1.914)
Number of observations	288	288	288	288
States	32	32	32	32
Number of instruments	30	30	28	28
First order test of serial correlation	0.603	0.733	0.661	0.860
p-value	[0.547]	[0.464]	[0.508]	[0.390]
Second order test of serial correlation	−0.921	−0.202	−1.003	−0.421
p-value	[0.357]	[0.840]	[0.316]	[0.674]
Hansen test	17.140	17.140	18.178	18.178
p-value	[0.703]	[0.703]	[0.638]	[0.638]

Note: Standard errors robust to heteroskedasticity are reported in parenthesis. The symbols *, **, and *** refer to levels of significance of 10%, 5%, and 1%, respectively. The Hansen test reports that under the null the overidentified restrictions are valid. First and second order test of serial correlation corresponds to the Arellano-Bond test for serial correlation, under the null of no autocorrelation.

Columns (1) and (2) in Table 5 present the results for the model that includes all the original regressors. Except for column (1), where the interaction between the FDI stock and the North region is not statistically significant, we observe evidence suggesting that each region’s exports are affected differently by the FDI stock. As before, there are again irrelevant variables that might best be omitted from the model. In columns (3) and (4), we excluded the state-invariant, irrelevant regressors that were not statistically significant in (1) and (2) and instead included time effects. This time the state GDP is positive and statistically significant along with the ratio of manufacturing to total exports. Moreover, the results suggest that the Border region is the most sensitive to variations in the FDI stock (coefficients of 0.533 and 0.593), followed by the North (0.437 and 0.505), Center (0.417 and 0.494) and South (0.333 and 0.421) regions. Once again, we interpret the latter result as evidence that as we move away from Mexico’s northern border with the U.S., the impact of FDI on manufacturing exports is less relevant in the short run. We think this evidence is consistent with the findings of Aitken, Hanson, and Harrison (1997), since the presence of multinational firms and FDI is less important as one moves south—at least in the short run. The relevance of this factor for manufactured goods exports also diminishes as one moves away from Mexico’s northern border.

Conclusion

In this paper, we examined the determinants of manufacturing exports across Mexican states and regions. Developing first a theoretical model that lays the foundations of the empirical assessment, our analysis takes into consideration internal and external determinants of manufacturing exports and pays special attention to the role of FDI. We make first use of traditional static fixed-effect estimations, followed by dynamic and static panel techniques employing SGMM. Regardless of the method or specification employed in the estimations, the most reliable determinant of manufacturing exports is the ratio of manufacturing to total GDP. This result is consistent with the idea that agglomeration economies are necessary for the existence of a robust exporting platform in each state and region. This result is also consistent with the evidence reported by Jordaan (2012), who finds that new multinational firms have concentrated in a selected group of states in Mexico—mainly in the north and central regions—due to the regional presence of the agglomeration of manufacturing firms that provide knowledge spillovers and other externality-based productivity advantages.

Using GMM estimation techniques to control for endogeneity, we also obtain two important results. First, by employing a dynamic panel specification, we observed significant export persistence but, most importantly, similar long-term effects of FDI across most of the regions—with only slightly less sensitivity to FDI in the Center region. The intuition for this result is that, once we take into account the long-term dynamics of manufacturing exports, there seems to be little difference in how responsive regions are to FDI variations. This fact has important policy implications, especially when considering promoting the less-developed South region and facilitating its economic integration into the rest of the country. Indeed, policymakers should bear in mind that foreign capital could a crucial element for developing an export platform in that region. Second, under our static specification, the results suggest that, in the short run, there are dissimilar responses to FDI variations across Mexican states, with the Border region observing the strongest effect of FDI on manufacturing exports, followed by the North, Center and South regions. This result is consistent with the finding of Aitken, Hanson, and Harrison (1997), which states that as we move further away from the U.S. Mexican border, and hence toward the south, the sensitivity of exports to FDI is less relevant given that foreign capital is also less abundant.

We acknowledge that there might be differences in the analysis of total FDI, as employed in this paper, and the individual analysis of its three components: reinvestments, balances to subsidiaries and new investments. As pointed out by Lundan (2006), the three FDI components might respond to different corporate business strategies. For instance, once a firm has positive earnings, it might choose to reinvest or to repatriate a proportion of those earnings depending on corporate factors, including agency costs. Indeed, we would expect that, out of the three categories of FDI listed above, new investment and reinvestment made by international firms would have the most substantial impact on manufacturing exports. The available data from Mexico’s Ministry of Economy distinguishes between the three FDI components. Hence, we leave this issue as an exciting line for future research.

Footnotes

Authors’ Note

An earlier version of this paper was presented at the 87th Annual Conference of the Southern Economic Association in Tampa, FL and at the VIII Congreso de Investigación Financiera FIMEF in Monterrey, México. The usual disclaimer applies.

Acknowledgments

The authors wish to thank Leonardo Torre and Miroslava Quiroga for their very detailed comments that helped improve this paper. We also acknowledge two anonymous referees of this journal for helpful comments that improved exposition and readability. We thank Eva González for her assistance in preparing the political maps of Mexican regions.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

René Cabral

Jorge Alberto Alvarado

Notes

References

Afonso

Gonzalez Alegre

. 2011. “Economic Growth and Budgetary Components: A Panel Assessment for the EU.” Empirical Economics 41 (3): 703–23.

Ahmed

A. D.

Cheng

Messinis

. 2011. “The Role of Exports, FDI and Imports in Development: Evidence from Sub-Saharan African Countries.” Applied Economics 43 (26): 3719–31.

Aitken

Hanson

G. H.

Harrison

A. E.

. 1997. “Spillovers, Foreign Investment, and Export Behavior.” Journal of International Economics 43 (1): 103–32.

Alguacil

M. T.

Cuadros

Orts

. 2002. “Foreign Direct Investment, Exports and Domestic Performance in Mexico: A Causality Analysis.” Economics Letters 77 (3): 371–76.

Arellano

Bond

. 1991. “Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations.” Review of Economic Studies 58 (2): 227–97.

Baliamoune-Lutz

M. N.

2004. “Does FDI Contribute to Economic Growth?” Business Economics 39 (2): 49–56.

Banco

de Mexico

. 2011. Reporte sobre las Economías Regionales, Enero-Marzo. Accessed December 2, 2020. https://www.banxico.org.mx/publicaciones-y-prensa/reportes-sobre-las-economias-regionales/%7BB22A7137-361C-E776-4BE9-7A6C29C26401%7D.pdf.

Barrell

Pain

. 1997. “Foreign Direct Investment, Technological Change, and Economic Growth within Europe.” The Economic Journal 107 (445): 1770–86.

Barrell

Pain

. 1999. “Domestic Institutions, Agglomerations and Foreign Direct Investment in Europe.” European Economic Review 43 (4–6): 925–34.

10.

Bellemare

M. F.

Masaki

Pepinsky

T. B.

. 2017. “Lagged Explanatory Variables and the Estimation of Causal Effect.” The Journal of Politics 79 (3): 949–63.

11.

Blomström

Kokko

. 1997. “Regional Integration and Foreign Direct Investment: A Conceptual Framework and Three Cases.” Policy Research Paper No. 1750, World Bank, Washington, DC.

12.

Blonigen

B. A.

2005. “A Review of the Empirical Literature on FDI Determinants.” Atlantic Economic Journal 33 (4): 383–403.

13.

Blundell

Bond

. 1998. “Initial Conditions and Moment Restrictions in Dynamic Panel Data Methods.” Journal of Econometrics 87 (1): 115–43.

14.

Boubacar

2016. “Spatial Determinants of U.S. FDI and Exports in OECD Countries.” Economic Systems 40 (1): 135–44.

15.

Cabral

Mollick

A. V.

. 2011. “Intra-industry Trade Effects on Mexican Manufacturing Productivity before and after NAFTA.” The Journal of International Trade and Economic Development 20 (1): 87–112.

16.

Calderón

Loayza

Servén

. 2004. “Greenfield Foreign Direct Investment and M&A: Feedback and Macroeconomic Effects.” Policy Research Working Paper No. 3192, World Bank, Washington, DC.

17.

Chowdhury

Mavrotas

. 2006. “FDI and Growth: What Causes What?” The World Economy 29 (1): 9–19.

18.

Clemens

M. A.

Radelet

Bhavnani

R. R.

Bazzi

. 2012. “Counting Chickens When They Hatch: Timing and the Effects of Aid on Growth.” The Economic Journal 122 (561): 590–617.

19.

Combes

Gobillon

. 2015. “The Empirics of Agglomeration Economies.” Handbook of Regional and Urban Economics 5A:247–348.

20.

Conconi

Sapir

Zanardi

. 2016. “The Internationalization Process of Firms: From Exports to FDI.” Journal of International Economics 99:16–30.

21.

Crespo

Fontoura

M. P.

. 2007. “Determinant Factors of FDI Spillovers—What Do We Really Know?” World Development 35 (3): 410–25.

22.

Cuadros

Orts

Alguacil

. 2004. “Openness and Growth: Re-examining Foreign Direct Investment, Trade and Output Linkages in Latin America.” Journal of Development Studies 40 (4): 167–92.

23.

Cuevas

Messmacher

Werner

. 2005. “Foreign Direct Investment in Mexico Since the Approval of NAFTA.” The World Bank Economic Review 19 (3): 473–88.

24.

Davies

Desbordes

Ray

. 2018. “Greenfield Versus Merger and Acquisition FDI: Same Wine, Different Bottles?” Canadian Journal of Economics 51 (4): 1151–90.

25.

De la Cruz

J. L.

Nuñez-Mora

J. A.

. 2006. “International Trade, Economic Growth, and Foreign Direct Investment: Causality Evidence in Mexico.” Revista de Economía Mundial 15:181–202.

26.

Dritsaki

Adamopoulos

. 2004. “A Causal Relationship between Trade, Foreign Direct Investment and Economic Growth for Greece.” American Journal of Applied Sciences 1 (3): 230–35.

27.

Duranton

Puga

. 2004. “Micro-Foundations of Urban Agglomeration Economies.” Handbook of Regional and Urban Economics 4:2063–2117.

28.

Eicher

T. S.

Helfman

Lenkoski

. 2012. “Robust FDI Determinants: Bayesian Model Averaging in the Presence of Selection Bias.” Journal of Macroeconomics 34 (3): 637–51.

29.

Franco

2013. “Exports and UFDI Motivations: Empirical Evidence from .S. Foreign Subsidiaries.” International Business Review 22 (1): 47–62.

30.

Girma

Gorg

Pisu

. 2008. “Exporting, Linkages and Productivity Spillovers from Foreign Direct Investment.” Canadian Journal of Economics 41 (1): 320–40.

31.

Goswami

Saikia

K. K.

. 2012. “FDI and Its Relation with Exports in India, Status and Prospect in North East Region.” Procedia-Social and Behavioral Sciences 37:123–32.

32.

Greenaway

Kneller

. 2008. “Exporting, Productivity and Agglomeration.” European Economic Review 52 (5): 919–39.

33.

Greenaway

Sousa

Wakelin

. 2004. “Do Domestic Firms Learn to Export from Multinationals?” European Journal of Political Economy 20:1027–43.

34.

Harms

Méon

. 2018. “Good and Useless FDI: The Growth Effects of Greenfield Investment and Mergers and Acquisitions.” Review of International Economics 26 (1): 37–59.

35.

Heckman

1979. “Sample Selection Bias as a Specification Error.” Econometrica 47:153–61.

36.

Hsiao

F. S.

Hsiao

M. C. W.

. 2006. “FDI, Exports, and GDP in East and Southeast Asia—Panel Data Versus Time-series Causality Analyses.” Journal of Asian Economics 17 (6): 1082–106.

37.

Islam

1998. “Growth Empirics: A Panel Data Approach—A Reply.” The Quarterly Journal of Economics 113 (1): 325–29.

38.

Ito

Yashiro

. 2015. “Does Agglomeration Promote Internationalization of Chinese Firms?” China Economic Review 34:109–21.

39.

Jordaan

J. A.

2012. “Agglomeration and the Location Choice of Foreign Direct Investment: New Evidence from Manufacturing FDI in Mexico.” Estudios Económicos 27 (1): 61–97.

40.

Karpaty

Kneller

. 2011. “Demonstration or Congestion? Export Spillovers in Sweden.” Review of World Economics 147 (1): 109–30.

41.

Kinoshita

Campos

N. F.

. 2003. “Why Does FDI Go Where It Goes? New Evidence from the Transition Economies.” International Monetary Fund Working Paper No. 3-228.

42.

Kneller

Pisu

. 2007. “Industrial Linkages and Export Spillovers from FDI.” The World Economy 30 (1): 105–34.

43.

Kok

Ersoy

B. A.

. 2009. “Analyses of FDI Determinants in Developing Countries.” International Journal of Social Economics 6 (1/2): 105–23.

44.

Koenig

2009. “Agglomeration and the Export Decisions of French Firms.” Journal of Urban Economics 66 (3): 186–95.

45.

Lundan

S. M.

2006. “Reinvested Earnings as a Component of FDI: An Analytical Review of the Determinants of Reinvestment.” Transnational Corporations 15 (3): 33–64.

46.

Malmberg

Lundequist

. 2000. “Agglomeration and Firm Performance: Economies of Scale, Localisation, and Urbanisation among Swedish Export Firms.” Environment and Planning 32:305–21.

47.

Mankiw

N. G.

Romer

Weil

D. N.

. 1992. “A Contribution to the Empirics of Economic Growth.” The Quarterly Journal of Economics 107 (2): 407–37.

48.

Markusen

J. R.

Maskus

K. E.

. 2002. “Discriminating among Alternative Theories of the Multinational Enterprise.” Review of International Economics 10 (4): 694–707.

49.

Marshall

1890. Principles of Economics, 8th ed. London, UK: Macmillan and Co.

50.

Mollick

A. V.

Cabral

. 2009. “Productivity Effects on Mexican Manufacturing Employment.” The North American Journal of Economics and Finance 20 (1): 66–81.

51.

Ottaviano

Puga

. 1998. “Agglomeration in the Global Economy: A Survey of the ‘New Economic Geography’.” The World Economy 21 (6): 707–31.

52.

Pacheco-López

2005. “Foreign Direct Investment, Exports and Imports in Mexico.” The World Economy 28 (8): 1157–72.

53.

Rahmaddi

Ichihashi

. 2013. “The Role of Foreign Direct Investment in Indonesia’s Manufacturing Exports.” Bulletin of Indonesian Economic Studies 49 (3): 329–54.

54.

Ramos

Moral-Benito

. 2017. “Agglomeration by Export Destination: Evidence from Spain.” Journal of Economic Geography 18 (3): 599–625.

55.

Reed

W. R.

2015. “On the Practice of Lagging Variables to Avoid Simultaneity.” Oxford Bulletin of Economics and Statistics 77 (6): 897–905.

56.

Rodríguez-Pose

Tselios

Winkler

Farole

. 2013. “Geography and the Determinants of Firm Exports in Indonesia.” World Development 44:225–40.

57.

Roodman

2009. “A Note on the Theme of Too Many Instruments.” Oxford Bulletin of Economics and Statistics 71 (1): 135–58.

58.

South

2016. “The Flying Geese Metaphor: Export-oriented Manufactures in Mexico.” Regional Studies 50 (9): 1483–95.

59.

Sun

Parikh

. 2001. “Exports, Inward Foreign Direct Investment (FDI) and Regional Economic Growth in China.” Regional Studies 35 (3): 187–96.

60.

Tekin-Koru

Waldkirch

. 2010. “North-South Integration and the Location of Foreign Direct Investment.” Review of International Economics 18 (4): 696–713.

61.

Ten Kate

1992. “El ajuste estructural de México: Dos historias diferentes.” Comercio Exterior 42 (6): 519–28.

62.

Vasquez-Galan

Oladipo

O. S.

. 2009. “Have Liberalization and NAFTA had a Positive Impact on Mexico’s Output Growth?” Journal of Applied Economics 12 (1): 159–80.

63.

Wacker

2013. “On the Measurement of Foreign Direct Investment and its Relationship to Activities of Multinational Corporations.” Working Paper Series No. 1614. European Central Bank, Frankfurt, Germany.

64.

Waldkirch

2003. “The ‘New Regionalism’ and Foreign Direct Investment: The Case of Mexico.” Journal of International Trade and Economic Development 12 (2): 151–84.

65.

Williamson

1990. What Washington Means by Policy Reform. Latin American Adjustment: How Much Has Happened? Washington, DC: The Peterson Institute for International Economics.

66.

Zhang

K. H.

2015. “What Drives Export Competitiveness? The Role of FDI in Chinese Manufacturing.” Contemporary Economic Policy 33 (3): 499–512.

67.

Zhang

K. H.

Song

. 2000. “Promoting Exports: The Role of Inward FDI in China.” China Economic Review 11:385–96.

68.

Zhang

Felmingham

. 2001. “The Relationship between Inward Direct Foreign Investment and China’s Provincial Export Trade.” China Economic Review 12:82–99.