1. FDI spillovers and economic growth

Existing literature provides a number of different channels through which positive FDI spillovers occur.Footnote ¹ These channels include a learning from foreign companies (Wang and Blomström, Reference Wang and Blomström1992; Blomström and Kokko, Reference Blomström and Kokko1998),Footnote ² diffusion of knowledge through the domestic employees of multinational corporations (Fosfuri et al., Reference Fosfuri, Motta and Rønde2001), positive backward linkages (Spencer, Reference Spencer2008), and domestic competitions with foreign companies (Wang and Blomström, Reference Wang and Blomström1992).Footnote ³ Interestingly enough, although these channels for positive FDI spillovers may well contribute to host countries' economic growth, empirical evidence is contested (e.g., Haddad and Harrison, Reference Haddad and Harrison1993; Aitken and Harrison, Reference Aitken and Harrison1999; De Mello, Reference De Mello1999; Javorcik, Reference Javorcik2004; Carkovic and Levine, Reference Carkovic, Levine, Graham, Moran and Blomström2005; Havranek and Irsova, Reference Havranek and Irsova2011; Iršová and Havránek, Reference Iršová and Havránek2013).

One possible reason for these mixed empirical findings lies in the heterogeneity of host economies.Footnote ⁴ Some countries may be better equipped with absorb technology and know-how from MNCs than others. For example, FDI spillovers are likely to occur only in countries that have sufficient human capital (Borensztein et al., Reference Borensztein, De Gregorio and Lee1998; Xu, Reference Xu2000; Bengoa and Sanchez-Robles, Reference Bengoa and Sanchez-Robles2003). Hermes and Lensink (Reference Hermes and Lensink2003) and Azman-Saini et al. (Reference Azman-Saini, Law and Ahmad2010) show that the development of a sound financial system is a precondition for FDI's positive effect on growth. Similarly, Alfaro et al. (Reference Alfaro, Chanda, Kalemli-Ozcan and Sayek2004) contend that FDI has a positive effect on growth in countries with well-developed financial markets. Balasubramanyam et al. (Reference Balasubramanyam, Salisu and Sapsford1996) demonstrate that the positive effect of FDI on growth is significant for host countries with export promotion policies. Damijan et al. (Reference Damijan, Rojec, Majcen and Knell2013) find that positive spillovers appear in medium or high productivity firms with higher absorptive capacities.

In sum, these studies disentangle the FDI-growth nexus by investigating how heterogeneous economic preconditions in host countries affect FDI spillovers. However, to understand the microfoundation of the relation between FDI and growth, the following question needs to be answered: what makes certain host countries more likely to adopt growth-friendly institutions and policies in the first place? The aforementioned preconditions for FDI-led growth are not exogenously dictated, but created and promoted (or demoted) by host governments. More importantly, the literature lacks discussion on the role of property rights protection in moderating the relationship between FDI and growth. In the next section, we attempt to demonstrate how ATHs influence host countries' property rights protection, in turn affecting FDI-led growth.

2. Autocratic time horizons, property rights, and FDI-led growth

FDI is not a magic pill for host countries' economic growth, but can be a catalyst for the development of local businesses and entrepreneurship. For FDI's positive spillovers to occur, domestic economic actors should not only adapt themselves to new technologies and business practices brought by FDI, but must grow to be self-sustaining. Unless domestic businesses do away with significant reliance on foreign capital and productivity, negative spillovers may have an adverse impact on the host country's economic growth. For example, foreign firms may dominate a host country's domestic market, crowd out potential domestic competitors (Aitken and Harrison, Reference Aitken and Harrison1999), and hurt infant domestic industries by increasing labor costs (Zhang et al., Reference Zhang, Li, Li and Zhou2010). Thus, FDI-led growth is likely to occur (1) when foreign firms are willing to transfer technologies and know-hows, and (2) when local businesses are willing to make investments to accommodate new operating systems, train local workers, and become competitive in both domestic and global markets. We contend that these two general conditions are likely to be fulfilled in autocratic countries with long-time horizons, particularly through the protection of property rights.

ATH is an autocrat's expectation of job security in the future. As Olson (Reference Olson1993) posits,Footnote ⁵ self-interested autocrats with long-time horizons are likely to adopt optimal (or revenue-maximizing) taxation, promote domestic economic productivity, and build market-friendly domestic institutions, thus protecting property rights strongly and providing public goods to domestic economic actors. That is because they expect their long-term benefits through economic growth to outweigh short-term predatory rents. To the contrary, autocrats with short time horizons resemble roving bandits who tend to be highly myopic and predatory. Their utmost interest is in maximizing their short-term private gains at the expense of domestic economic productivity. Thus they are likely to build and strengthen rent-extraction mechanisms that require significant government control over domestic assets, economic activities, and markets, by imposing excessive taxation and confiscating capital goods. Since Olson's pioneering research, scholars have uncovered empirical evidence for this conjecture. Wright (Reference Wright2008), for example, argues that autocrats with long-time horizons are more likely to use foreign aid effectively for economic growth. Clague et al. (Reference Clague, Keefer, Knack and Olson1996) find that they are more likely to protect property and contract rights. Moon (Reference Moon2015) demonstrates that autocratic regimes with long-time horizons provide better property rights institutions, thus attracting more FDI. Scholars also find that those far-sighted autocrats are less likely to confiscate private properties and expropriate foreign investments (Li, Reference Li2009a), and less prone to corruption (Chang and Golden, Reference Chang and Golden2007). Similarly, it is also argued that the long-time horizons give governments incentive to maintain greater room to manipulate policy when it comes to designing bilateral investment treaties (Blake, Reference Blake2013).

This market-friendly preference structure leads autocratic countries with long-time horizons to provide conditions under which FDI-led economic growth is likely to occur. We argue in particular that such countries are liable to benefit from FDI spillover through secure property rights institutions,Footnote ⁶ via property right institutions. Since autocratic governments with long-time horizons are generally interested in economic growth, underlying preference could be realized through a development of infrastructure, human capital, or other market friendly economic policies. We will discuss more on this in the empirical analysis section. First, foreign firms' willingness to transfer technologies and know-how depends on there being sound rule of law under which they are likely to believe that their property rights will not be infringed by the host government and local businesses. The growth of infant domestic industries, often expedited by FDI, starts from imitating advanced technologies and business operations. However, foreign firms would be reluctant to transfer their technology and skills if the host government and local businesses are willing and able to take them without proper payment. Thus, foreign investors' confidence in the enforcement of property rights over their intellectual properties or physical assets is essential. Conversely, the absence of well-enforced property rights should hinder such positive spillovers by making foreign firms more risk-averse and myopic. For example, country case studies find evidence that firms are unwilling to transfer technologies when the developing host countries have a weak system of intellectual property rights protection (Mansfield, Reference Mansfield1994; Maskus, Reference Maskus1994). Olofsdotter (Reference Olofsdotter1998) also shows evidence that sound property rights institutions and efficient bureaucracies facilitate FDI spillovers.

Property rights protection also reassures domestic firms that their investments will be protected from illegal expropriations. For local businesses to increase their capacity to absorb advanced technologies and efficient management skills, they should invest in training programs for local labor and reforming business operations, which are costly. Similarly, unless local firms have confidence in the legal protection of their investment and subsequent economic outcomes, they would be extremely risk-averse for fear of government exploitation and corrupt (and/or inefficient) bureaucracies. Even when FDI's positive spillovers are available, the host country's economic growth is not guaranteed if local companies remain to be uncompetitive subsidiaries of foreign firms. On the other hand, once domestic economic actors have confidence in the government's credible promise to protect their property rights, they will be more willing to make investments to increase their absorptive capacities and imitate management know-hows.

In sum, we hypothesize that FDI has positive effect on growth when an autocratic host country's leadership time horizon is sufficiently long. In the empirical analysis section, we further test whether this conditional relationship is through property right institutions, as Figure 1 denotes.Footnote ⁷

Figure 1. Moderating effect of ATH FDI-led growth.

3. Research design

We estimate the impact of FDI on economic growth, moderated by autocratic leadership time horizon, in a panel dataset of 64 countries for 36 years (1970–2005). The sample of autocratic countries is taken from the regime-type dataset created by Geddes et al. (Reference Geddes, Wright and Frantz2014).Footnote ⁸ In years of leadership turnover, we consider only a single leader who held the leadership position on the first day of each year (Goemans et al., Reference Goemans, Gleditsch and Chiozza2009). As Figure 1 shows, our theoretical mechanism is more than just showing the moderating effect of ATH. After demonstrating empirical evidence for our hypothesis, we further investigate whether the mechanism itself is empirically supported.

3.1 Estimator

Our methodological approach is incremental starting from a simple ordinary least squares (OLS) framework to more sophisticated models. First, we estimate a growth model using OLS controlling for country-specific fixed effects and serial autocorrelation.Footnote ⁹ However, in estimating these models for the dynamic panel dataset, we encounter multiple econometric problems that may not be solved using a simple OLS framework.

First, we run fixed-effects models because country-specific characteristics are likely to be correlated with the independent variables as well as with the error term that contains unobserved country-specific confounders. Yet, a fixed-effects model poses a problem (so-called Nickell bias) in our dynamic panel data with a small number of time periods compared with a larger number of countries; i.e., Nickell (Reference Nickell1981) demonstrates that the standard fixed-effects estimator generates inconsistent estimates in dynamic panel data with small T and large N.Footnote ¹⁰

Second, the explanatory variables are likely to be endogenous to economic growth. For example, FDI may well promote economic growth by invigorating and expanding a host country's market, in turn inviting further FDI.Footnote ¹¹ In addition, ATHs may be endogenous to growth from a theoretical standpoint; that is, an autocratic leader may be more likely to expect to survive in office when the country's economy is vibrant and growing, which makes the leader better equipped with fund government projects and provide patronage goods to elites in the winning coalition (Bak and Moon, Reference Bak and Moon2016). Given multiple endogeneity problems, using a conventional two-stage least squares estimator also poses a daunting methodological challenge; that is, finding highly relevant and exogenous instruments for multiple endogenous variables is challenging.

Given these econometric concerns, we use the Arellano–Bond generalized method of moments (GMM) estimator (Arellano and Bond, Reference Arellano and Bond1991), originally proposed by Holtz-Eakin et al. (Reference Holtz-Eakin, Newey and Rosen1988).Footnote ¹² By taking first differences in the variables of interest, we eliminate country-specific effects. In addition, the estimator makes endogenous regressors pre-determined by using their lagged values, which also allows us to exploit the in-sample information as much as possible. In particular, Carkovic and Levine (Reference Carkovic, Levine, Graham, Moran and Blomström2005), who examines the growth-accelerating effect of FDI, provide a useful starting point for the econometric framework.

Consider the following growth equation first:

(1)$$\eqalign{{\rm \Delta GD}{\rm P}_{i,t} =\; &{\rm GD}{\rm P}_{i,t}-{\rm GD}{\rm P}_{i,t-1} \cr =\; &\lpar {\alpha -1} \rpar {\rm GD}{\rm P}_{i,t-1} + \beta x_{i,t} + \gamma {\rm CONTRO}{\rm L}_{i,t-1} + \tau {\rm YEA}{\rm R}_t + \eta \lambda _i + \epsilon _{i,t}} $$

where the dependent variable, GROWTH, is real GDP per capita growth. x refers to a set of endogenous variables that include FDI inflows (FDI), ATH, and the multiplicative interaction term (FDI × ATH). CONTROL is a set of control variables, YEAR refers to year-specific effects (year dummies), λ represents country-specific fixed effects, and ε shows the error term.

Rearranging the above equation gives

(2)$${\rm GD}{\rm P}_{i,t} = \alpha {\rm GD}{\rm P}_{i,t-1} + \beta x_{i,t} + \gamma {\rm CONTRO}{\rm L}_{i,t-1} + \tau {\rm YEA}{\rm R}_t + \eta \lambda _i + \epsilon _{i,t}$$

To swipe away the country-specific effects, we take first differences of the above, and get the following dynamic panel equation:

(3)$${\rm GD}{\rm P}_{i,t}-{\rm GD}{\rm P}_{i,t-1} = \alpha \lpar {{\rm GD}{\rm P}_{i,t-1}-{\rm GD}{\rm P}_{i,t-2}} \rpar + \beta \lpar {x_{i,t}-x_{i,t-1}} \rpar + \gamma \lpar {{\rm CONTRO}{\rm L}_{i,t}-{\rm CONTRO}{\rm L}_{i,t-1}} \rpar + \tau {\rm YEA}{\rm R}_t + \lpar {\epsilon_{i,t}-\epsilon_{i,t-1}} \rpar $$

This differential equation is required to meet moment conditions to ensure the weak exogeneity of the instruments; e.g., E[x _{i,t−m(m > 1)}(ε _i,t − ε _i,t−1)] = 0, and E[GDP_{i,t−m(m > 1)}(ε _i,t − ε _i,t−1)] = 0. Differencing may make the lagged levels of the endogenous regressors weak instruments for differences, thus making the model inefficient and biased (Blundell and Bond, Reference Blundell and Bond1998). This concern looms large ‘when the explanatory variables are persistent over time’ (Carkovic and Levine, Reference Carkovic, Levine, Graham, Moran and Blomström2005, 200), such as FDI and ATH in our models. Thus, as a robustness check, we also use the system GMM estimator that incorporates the level equations using lagged differences of explanatory variables as additional instruments (Blundell and Bond, Reference Blundell and Bond1998). To ensure the validity of the instruments, we check the Hansen-J over-identification test.Footnote ¹³ Roodman (Reference Roodman2009) warns that too many instruments can weaken the Hansen test, which is signified by the P-values of 1.000. Roodman (Reference Roodman2009) suggests that ‘Researchers should also test for robustness to severely reducing the instrument count’ (129). To address this issue, we report the number of instruments, and perform several robustness checks. First, we determine the optimal number of lags such that the root mean squared error is minimized (reported in the appendix). In addition, we reduce the number of instruments significantly using a single instrument for each variable and lag distance rather than using all lagged values as instruments, and replacing the GMM instruments with their principal components. We also check the potential second-order autocorrelation in first differences.Footnote ¹⁴ Finally, we use robust standard errors that are consistent against heteroskedasticity and autocorrelation within panels.

5. Tests of theoretical mechanism

The theoretical mechanism indicates that FDI induces economic growth when an autocratic country strongly protects property rights. We examine this mechanism empirically in two steps.

First, we replicate an empirical model in Clague et al. (Reference Clague, Keefer, Knack and Olson1996) in which the authors examine the impact of leadership time horizons on the level of property rights protection. Using our measure of ATH and their model specifications, we find similar results. Using models 1 and 2 in Table 3, we examine how ATH affects the level of contract-intensive money (CIM). For a robustness check, although the temporal dimension is limited, we use the property rights index from the International Country Risk Guide (ICRG) (PRS, 2007) as the dependent variable in models 3 and 4. The estimator of models 1 and 3 is OLS with a lagged dependent variable and country fixed effects, while models 2 and 4 use Driscoll and Kraays' (1998) covariance matrix estimator (D–K) with fixed effects and AR(1) autocorrelation error structure. In all models, we find that as ATH decreases or Pr.(Leadership Failure) increases, the level of CIM significantly decreases, which is consistent with our theoretical argument as well as the finding of Clague et al. (Reference Clague, Keefer, Knack and Olson1996).

Table 3. Effect of ATH on property rights protection

Note: D–K stands for Driscoll and Kraays (1998) covariance matrix estimator with fixed effects. All models include country-fixed effects. D–K models 2 and 4 use the first-order autocorrelation structure of error. Robust standard errors are presented in parentheses. *P < 0.10; **P < 0.05; ***P < 0.01 (two-tailed tests).

Next, we examine whether FDI has positive effect on growth only when CIM is sufficiently high, using an interaction term between FDI and CIM. Using the same dynamic panel GMM, excluding information about ATH, the results in model 2 of Table 4 show that property rights is a significant factor that moderates the relationship between FDI and economic growth. Interestingly enough, such an empirical pattern does not hold in democracies. This finding implies that our theoretical mechanism is likely to be unique to autocratic countries. Additional evidence for our mechanism is found in model 3, in which we include ATH and FDI × ATH, in order to see how the conditioning effect of CIM changes when we include the political variable (ATH) that is expected to determine the level of property rights protection. The results show that the moderating effect of CIM is absorbed by ATH, while the moderating effect ATH remains significant.

Table 4. Effect of FDI on growth conditioned by property rights protection

Note: Pr(Fail) indicates the probability of leadership failure. All models use a single instrument for each variable and lag distance instead of using all lagged values as instruments, replacing the GMM instruments with their principal components. All variables are lagged by 1 year. All columns include year dummies (not reported). Robust standard errors are presented in parentheses. *P < 0.10; **P < 0.05; ***P < 0.01 (two-tailed tests).