Firm performance and backward and forward linkages: the case of the garment sector in Myanmar

ABSTRACT

This study examines how vertical spillovers through backward and forward linkages with multinationals relate to productivity and inputs allocation for garment firms in Myanmar. The results confirm that while these linkages promote spillover effects on productivity growth with capital accumulation, they fail to accelerate employment growth. The Myanmar government has emphasized employment growth by attracting foreign investment. Although linkages with multinationals lead to productivity improvement, such positive effects may be realized without clear evidence of increased job opportunities.

KEYWORDS:

• Foreign direct investment
• vertical spillovers
• backward and forward linkages
• Myanmar

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. See, e.g. Blomstrom and Kokko (1998), Blomstrom, Globerman, and Kokko (2001), Crespo and Fontoura (2007), Gorg and Greenaway (2004), and Lipsey (2004).

2. The critical issue in previous studies is that the measures of backward and forward linkages rely fully on the input-output coefficients of the IO tables (Barrios, Görg, and Strobl 2011; Giroud, Jindra, and Marek 2012). Following Javorcik (2004), most studies use the IO tables to construct the measure of backward linkages by calculating the share of total revenue downstream sectors explained by foreign-owned firms and that of forward linkages by calculating the share of total revenue in upstream sectors explained by foreign-owned firms. In this specification, all firms in the same sector share identical measures of backward and forward linkages without the consideration of heterogeneous features of firms’ structures. Since certain firms have international linkages and others do not even within the same sector, treating all firms as identical may fail to capture precise features of backward and forward linkages. To solve this issue, this study constructs the individual firm-level measures of backward and forward linkages by conducting business surveys of individual firms in the garment sector. Exceptions may include the work of Vacek (2010) on the case of Czech Republic, which shows positive spillovers through backward linkages but insignificant spillovers through forward linkages.

3. Our firm-level data covers years 2014 and 2016. Since the exuberant general election was conducted in November, 2015, we believe that it would be difficult to collect reliable data of firm characteristics in 2015.

4. One possible problem is on the common trend assumption. When treatment effect relative to controls being evaluated, treatment and control groups should follow the common trend, which means that the outcome variables should reveal the similar trend between the two groups before the treatment. Because our data have only two years of data, we simply conduct the balance check using the level of variables in the first year. However, this does not guarantee the common trend. Another possible problem is related to the argument that matching approaches can control only for observable selection biases, but they may generate unreliable results if unobservable biases exist. Although we admit such shortcomings, our analysis would be a first step to examine backward and forward linkages of vertical spillovers in Myanmar.

5. Many studies have also examined horizontal spillovers theoretically and empirically, although their empirical analysis has shown the mixed results (Abraham, Konings, and Slootmaekers 2006; Aitken and Harrison 1999; Blalock and Gertler 2008; Blomstrom and Persson 1983; Castellani and Zanfei 2003; Caves 1974; Djankov and Hoekman 2000; Du, Harrison, and Jefferson 2012; Fons-Rosen et al. 2013; Globerman 1979; Gorg and Strobl 2003; Gorodnichenko, Svejnar, and Terrell 2014; Haddad and Harrison 1993; Haskel, Pereira, and Slaughter 2007; Javorcik 2004; Javorcik et al. 2011; Keller and Yeaple 2003, 2009; Kokko, Tansini, and Zejan 1996; Konings 2001; Kosova 2010; Lin, Liu, and Zhang 2009; Reganati and Sica 2007; Sembenelli and Siotis 2005; Teece 1977; Yudaeva et al. 2003).

6. Backward and forward linkages are potential channels for productivity spillovers, which refer to contacts between multinationals (foreign-invested firms) and their domestic supplier and those between multinationals (foreign-invested firms) and their domestic buyers, respectively (Javorcik 2004; Vacek 2010). Scott-Kennel (2007) also notes that vertical linkages may be backward (with suppliers and subcontractors), forward (with customers and agents), and contractual (with domestic franchisees and licensees). In contrast to horizontal linkages closely related to competition effects, vertical linkages occur through collaborative activities with local partners (Girma, Görg, and Pisu 2008).

7. Moran (2001) applies various case studies to argue that technology transfers are common with foreign firms as foreign firms often provide technical assistance and management experience to their domestic suppliers. For positive effects, backward spillovers may arise when domestic firms supply their inputs to foreign firms which have an incentive to provide assistance to their suppliers to ensure high quality and on-time delivery of their production inputs (Newman et al. 2015). Javorcik (2004) also suggests the role of indirect channels of domestic firms’ productivity gains since intense competition with foreign customers and greater demand for domestically-produced intermediate goods encourage domestic suppliers to supply high quality inputs and improve their efficiency. Conversely, domestic firms may experience negative backward spillover effects through direct linkages with downstream foreign firms. The profit of domestic firms decreases when foreign firms take advantage of the bargaining power of domestic firms during contract negotiations, resulting in a decline in productivity (Girma, Görg, and Pisu 2008). Thus, domestic suppliers should produce various inputs to supply the inputs required by foreign firms (Rodriguez-Clare 1996). In addition, if foreign firms import intermediate goods, domestic firms may also experience negative impacts on productivity through a competition channel, leading to a loss in customers and results in lower profits for domestic suppliers.

8. Grossman and Helpman (1991) argue that domestic firms may improve productivity through direct linkages when upstream foreign firms supply inputs with higher technologies to domestic customers, such that domestic firms can learn advanced technologies, resulting in positive productivity gains for domestic producers. Javorcik (2004) also shows that forward spillovers may appear in the form of externalities when foreign firms supply inputs of embodied services or other forms of support, which helps improve the productivity gains of domestic users. In addition, Newman et al. (2015) suggest the possibility of positive forward spillover effects through indirect linkages. Conversely, negative forward spillover effects are also possible when the entry of foreign firms into upstream sectors may be anti-competitive, and each foreign firm possesses a larger market share (Newman et al. 2015; Jordaan 2011). In addition, foreign firms generally perform well in various aspects including efficiency, technical know-how, and managerial skill; thus, upstream domestic firms cannot compete with such foreign firms, resulting in a negative impact on downstream domestic firms due partly to the payment of higher prices for their inputs (Liang 2017; Aitken and Harrison 1999).

9. See, e.g. Lin, Liu, and Zhang (2009) for China, Du, Harrison, and Jefferson (2012) for China, Javorcik (2004) for Lithuania, Kubny and Voss (2014) for Vietnam, Reganati and Sica (2007) for Italy, and Barrios, Görg, and Strobl (2011) for Ireland.

10. See, e.g. Schoors and van der Tol (2002) for Hungary, Javorcik and Spatareanu (2005) for the Czech Republic and Romania, and Stancik (2007) for the Czech Republic.

11. See, e.g. Chang, Chung, and Xu (2007) for China, Lin, Liu, and Zhang (2009) for China, Kubny and Voss (2014) for Vietnam, and Schoors and van der Tol (2002) for Hungary.

12. To check the robustness of our empirical results, we also use the different critical values of the shares of sales (purchases) differentiating backward (forward) linkages. Table A1 in Appendix 1 shows the results based on alternative critical values by replacing 50 percent in the baseline with 30 percent. The estimated results are qualitatively similar to those in the baseline case.

13. See Van Beveren (2012) for an extensive review on the TFP.

14. The identifying assumption of the Olley-Pakes approach is that, conditional on capital, investment is monotonically increasing with respect to the shock. Since capital responds to the shock with time lags through contemporaneous investment, the return to the other inputs can be obtained by non-parametrically inverting investment and capital as a proxy for the unobserved shock (Lin, Liu, and Zhang 2009). The Olley-Pakes approach consists of a two-step estimation method, where semi-parametric methods are used to estimate the coefficients on the variable inputs in the first step; then, the parameters on capital inputs can be identified under assumptions on the dynamics of the productivity process in the second step.

15. The monotonicity condition of the Olley-Pakes estimation that investment is strictly increasing in productivity requires that only observations with positive investment can be used in the empirical model. If firms involve zero investment due to substantial adjustment costs with capital stock, these observations cast doubt on the validity of the strict monotonicity condition. The Levinsohn-Pertin estimation mitigates this problem by using intermediate inputs instead of investment as a proxy. Firms typically report positive use of intermediate inputs; therefore, the analysis retains most observations with the strict monotonicity condition more likely satisfied.

16. The approach of Ackerberg, Caves, and Frazer (2006) allows for the possibility that a firm’s private knowledge of its productivity may affect the input decisions and for firm-specific productivity differences that exhibit idiosyncratic changes over time to mitigate the simultaneity bias between productivity shocks and input choices (Javorcik et al. 2011). The Olley-Pakes and Levinsohn-Pertin approaches assume that firms can instantly adjust certain inputs at no cost when they are subject to productivity shocks. However, Ackerberg, Caves, and Frazer (2006, 2015) suggest that the labor coefficient can be consistently estimated in the first stage only if the free variables show variability independently from the proxy variable. If this is not the case, the coefficients would be perfectly collinear in the first-stage estimation and hence would not be identifiable.

17. The advantages of the Wooldridge method include that it overcomes potential identification issues in the first step estimation, emphasized by Ackerberg, Caves, and Frazer (2006, 2015) and that it obtains robust standard errors, easily accounting for both serial correlation and heteroscedasticity.

18. Many empirical studies estimate the ATT by applying various matching methods, like propensity score matching (PSM), which could reduce the selection bias by creating comparable counterfactual outcomes for treated units. Once the treated units are matched, the methods assume no systematic differences in unobservable characteristics between treated and untreated units, given the estimated propensity scores under some assumptions. The first is conditional independence assumption (CIA) or confoundedness; that is, after controlling for observed covariates, the potential outcomes are independent of the treatment assignment. The second is the independent and identically distributed observations assumption, which requires that the potential outcomes and treatment status of each individual are independent of the potential outcomes and treatment status of all other individuals in the sample. The third assumption is the common support or overlap condition, which suggests that every observation has a positive probability of being both treated and control (Heinrich, Maffioli, and Vazquez 2010).

19. In the second step regression, we include all covariates used in the first step to improve efficiency, as in the works of Neuenkirch and Neumeier (2016).

20. Matching approaches may be an appropriate method to overcome the selection bias caused by observables and to estimate the average treatment effect in observational studies. However, it should be noted that these matching approaches can control only for observable selection biases and may generate unreliable results if unobservable biases exist, i.e. systematic differences between treatment and control groups.

21. This study also discusses the roles of two channel classifications of backward and forward linkages, (i) the external-domestic channels and (ii) the indirect-direct channels, in Appendix 2. First, concerning the external-domestic channels, the external channel means that domestic firms trade directly with foreign-located firms operating abroad (direct exports or imports), while the domestic channel means that domestic firms trade with multinationals or foreign affiliates operating in the domestic country. The trade literature generally emphasizes the roles of external trade (imports and exports) in inducing spillover effects without the consideration of the domestic channels, and the existing studies on FDI do not consider different features between the external and domestic channels. Second, for the contexts of the indirect-direct channels, we evaluate the prevalence of domestic firms’ use of brokers or intermediaries, often called ‘middlemen.’ Such middlemen often play a crucial role as an intermediary or distributor in a transaction or process chain in facilitating interaction between buyers and sellers (Li, He, and Sousa 2017). In our study, the indirect channel of backward and forward linkages means that domestic firms trade with foreign middlemen operating in the country, while the direct channel means that domestic firms trade with foreign firms which are not middlemen. The examination of the external-domestic and indirect-direct channels would help us understand the source of vertical spillover effects in developing economies.

22. See Lin, Liu, and Zhang (2009), Du, Harrison, and Jefferson (2012), Merlevede, Schoors, and Spatareanu (2014), Vacek (2010), Javorcik (2004), Blalock and Gertler (2008), Kubny and Voss (2014), Reganati and Sica (2007), and Barrios, Görg, and Strobl (2011) for positive effects of backward linkages and Chang, Chung, and Xu (2007), Lin, Liu, and Zhang (2009), Kubny and Voss (2014), Schoors and van der Tol (2002), Du, Harrison, and Jefferson (2012), and Xu and Sheng (2012) for positive effects of forward linkages.