Trade liberalization and firms’ productivity in Vietnam: the role of local business environment

ABSTRACT

This paper examines how frictions in the local business environment where firms operate alter the productivity gains from trade. Using Vietnamese firm-level data from 2006 to 2012, the analysis provides robust evidence of a positive effect of trade opening on firms’ efficiency. However, distortions that operate through certain local market features in the form of unenforceable property rights, an ineffective land-titling system, bureaucratic hurdles and labour market frictions play a crucial role in the transmission of trade liberalization shocks. These findings have important policy implications because they suggest that complementary policies addressing local market constraints need to be implemented alongside trade reforms.

KEYWORDS:

• trade liberalization
• World Trade Organization (WTO) accession
• local business environment
• firms’ productivity
• Vietnamese micro-level data

JEL:

• D22
• F13
• F14
• H73
• P31

ACKNOWLEDGEMENTS

The authors gratefully acknowledge all helpful comments as well as constructive suggestions made by the editor and two anonymous referees. All errors remain the authors’ own.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. A number of firm-level studies have been carried out in the past few years, assessing the role that features of the local business environment play in shaping firms’ economic performance (Agostino et al., 2020; Ganau & Rodríguez-Pose, 2019; Lasagni et al., 2015; Tran, 2019). See section 2 for a review of the literature.

2. Section A1 in Appendix A in the supplemental data online offers an overview of Vietnam’s transition process toward a market-based economy, including the policies promoting international economic integration that have laid the foundation for the country’s accession to the WTO in 2007.

3. Good examples are Pavcnik (2002), Schor (2004), Amiti and Konings (2007), Topalova and Khandelwal (2011) and Brandt et al. (2017) for Chile, Brazil, Indonesia, India and China, respectively.

4. The coefficient is similar in magnitude to the estimates by Topalova and Khandelwal (2011) and Amiti and Konings (2007).

5. In the very demanding specification the coefficients have the expected sign, but they are statistically significant at the 10% level.

6. Baccini et al. (2019) investigate the productivity effect only at the aggregate level and conclude that industry productivity gains from Vietnam’s WTO entry would have been substantially higher without the presence of SOEs.

7. Another strand of the literature, partly related to that emphasizing the role played by the local context, highlights the impact regional interventions have on aggregate productivity and location choices (Baldwin & Okubo, 2006; Okubo et al., 2010; Okubo & Tomiura, 2012). Okubo et al. (2010) show that regional policies for promoting periphery development attract low-productivity firms and adversely affect the productivity gap within a country. The side effects of regional development policies become even more important in the age of globalization. When a country becomes more open to international trade ‘without changing the level of permissible subsidies to firms in remote regions, the subsidies will lead to an increasing distortion of the spatial allocation for industry’ (Baldwin et al., 2003, p. 478).

8. While all these authors acknowledge that ‘institutions matter’, less clear and more controversial is the definition of institutions. ‘Defining institutions is notoriously difficult and the current literature on the topic far from agrees on a common definition’ (Rodríguez-Pose, 2013, p. 1037). In our empirical analysis we do not rely on a single dimension, rather we consider some features of the local business environment that are also related to the well-functioning of institutions.

9. See https://www.aseanbriefing.com/news/land-rights-in-vietnam-what-they-are-and-how-you-can-acquire-land/ASEAN-Briefing for more details.

10. Bonfiglioli et al. (2018) develop a theoretical model where export opportunities induce firms to invest in bigger projects with more spread-out outcomes.

11. In general, the benefit of utilizing a variety of inputs is well known. The seminal work by Ethier (1982) rigorously but simply demonstrates the benefits of input variety arising from a finer division of labour. Ethier models varieties of imported inputs as complementary in production and not perfect substitutes. Firms thus exhibit a ‘love for varieties’ and their efficiency gains result from a greater division of labour. The same approach has been recently adopted by Kasahara and Rodrigue (2008), Kasahara and Lapham (2013) and Bas and Strauss-Kahn (2014), among others.

12. Kugler and Verhoogen (2012) show, both theoretically and empirically, that producing high-quality outputs requires high-quality inputs.

13. Similar positive relationships between trade liberalization and other firm-level performances have been observed. Goldberg et al. (2010) document that tariff cuts in India are associated with an expansion of domestic products. Other micro-level studies link a reduction in trade barriers with export performance (Bas & Strauss-Kahn, 2015).

14. Nunn and Trefler (2014) review the theoretical and empirical literature emphasizing the interdependence between trade and institutions, providing ample evidence of the impact of institutions on trade flows.

15. Anderson and Marcouiller (2002) develop a model of import demand in an insecure world, characterized by difficulties in contract enforcement, where insecure exchanges significantly impede international trade. Using aggregate data they present evidence that poor contract enforcement and low impartiality increase the insecurity of selling to that country, thereby reducing its imports. Similarly, using a gravity equation approach, De Groot et al. (2004) and Álvarez et al. (2018) consider the relevance of the quality of governance in explaining variation in bilateral trade patterns.

16. These frictions alter not only trade activities but also, more generally, reduce firms’ incentives to carry out productive activities involving the use of assets and to undertake investments that either maintain or enhance their value.

17. We thank an anonymous referee for bringing up this point and prompting us to extend our analysis on such an aspect.

18. This survey has already been used by a number of studies, such as Baccini et al. (2019), Bai et al. (2019), Doan and Kiyota (2014), Huang and Yang (2016), Newman et al. (2017), Nguyen (2017, 2019), Nguyen et al. (2018), Le et al. (2019), Ramstetter and Phan (2013) and Tran (2019).

19. There are currently 34 industrial sectors divided into three industrial sections: five in Mining and Quarrying, 24 in Manufacturing, and five in Electricity, Gas and Water Supply.

20. The old version (VSIC 93) was applied to enterprises surveyed from 2000 to 2005, while the new version (VSIC 07) has been applied starting from 2006.

21. We use different deflators to convert nominal values into real values (the base year is 2010). Value-added is deflated by using the producer price indexes (PPI) of each two-digit sector available on the GSO’s website, while capital is deflated by using the gross fixed capital formation deflators (source: the World Bank’s world development indicators). Finally, annual GDP deflators taken from the World Bank’s world development indicators are used to deflate material inputs. Real values in Vietnamese dongs (VND) are then converted into US dollars (US$) using the official annual exchange rate in 2010, that is VND18,612.92 = US$1.

22. As this variable is not available for all years, we follow the procedure suggested by Newman et al. (2017) and applied by Le et al. (2019), and identify, for the years with missing information, an exporter as a firm that either paid export tax during the year or exported in both the year before and the year after.

23. Firms’ ownership has received increased attention, especially in some countries such as Vietnam or China, where the transition from a planned to a market economy has brought several reforms that ease the establishment of private enterprises via the elimination of ownership restrictions, the simplification of regulatory procedures, and the abolishment of related ministries’ and local authorities’ power to grant sub-licences (e.g., Baccini et al., 2019; Wei & Tang, 2019).

24. Section A2 in Appendix A in the supplemental data online provides a detailed description of the methodology implemented to build the PCI indicators, including a detailed description of the sampling procedure. For further information, see https://pcivietnam.vn/en.

25. The 10 sub-indices are: entry costs, land access and security of tenure, transparency and access to information, time costs and regulatory compliance, informal charges, policy bias, proactivity of provincial leadership, business support services, labour and training, and legal institutions.

26. Each component is standardized to a 10-point scale, whereby the worst and best performing provinces are awarded scores of 1 and 10, respectively, and the other provinces are distributed somewhere along the scale between the two scores. Once the components are standardized, an average (either weighted or simple) is taken to create the sub-index. Details on weights are reported in the annually published reports.

27. Table A4 in Appendix A in the supplemental data online reports the other components used to build up the four indexes in 2006.

28. See https://wits.worldbank.org/. TRAINS stands for TRade Analysis and INformation System. The TRAINS database contains data on (applied) MFN tariffs, preferential tariffs, para-tariffs, NTMs and imports by suppliers at HS six-digit level for around 150 countries.

29. Correspondence tables are available at the WITS’s website; https://wits.worldbank.org/.

30. As indicated by Baccini et al. (2019), input–output matrices for Vietnam are available only for some years and at the two-digit sector rather than four-digit industry level. The distinction between input and output tariffs comes at the expense of losing a large number of observations.

31. The average tariff was cut to approximately 18% on the eve of accession compared with around 23% ten years earlier. Before joining the WTO, Vietnam also signed a number of trade agreements that have stepped up its process of international integration.

32. As specified in section 3.3, we use the transformed measure $1+Tarif{f}_{i,t}$ to account for the number of zeros.

33. We choose to run a single model with a dummy variable rather than splitting the sample based on the distinction between a low/high level of local governance quality. A problem of running multiple sub-models is that fewer data are used for each sub-model to find the correct estimators for the confounding variables; their confidence intervals are larger and therewith the individual models are not as good as they could be.

34. When the variable of interest assumes few values fitting a line may not be appropriate and comparing the means of two different groups would be a better approach. Moreover, the use of a dummy variable might be more appropriate when the relationship is suspected not to be linear. Although technically the dummy approach does not use all the information in the data, it answers the research question more directly.

35. To mitigate reverse causality concerns, we also run the empirical model using one-year lag values of the four variables proxying for the level of local governance quality. Results, available from the authors upon request, do not change. However, because PCI information is available only starting from 2006, in the main empirical analyses we prefer to keep the variable at time $t$ to avoid losing observations.

36. Several approaches might be used to explain an interaction of two continuous variables. Figure A3 in Appendix A in the supplemental data online shows the linear predictions of firms’ productivity on tariffs when the provincial index is held constant at different combinations of values from very low to very high.

37. Figure A1 in Appendix A in the supplemental data online shows, for each of the four-indexes used in the empirical analysis, the average level of each index in the South and North provinces across the period 2006–12. As the visual inspection suggests, we observe that provinces in the South have, on average, a better business environment that those in the North.

38. To cope with potential heteroscedasticity and serial autocorrelation, we cluster standard errors at the firm level, but the results are robust to alternative treatments of the error term, such as clustering at the sector–year or sector–province–year level.

39. The dummies for firms’ ownership structure (state owned, private owned and foreign invested) are not included in specifications with firm fixed effects as these variables are time invariant.

40. If we consider the variable tariffs in level rather than in logarithm, we obtain a coefficient of approximately 0.004, also in line with the findings of Baccini et al. (2019) on firms’ profitability.

41. The estimation of firms’ TFP comes with a caveat as the required quantity of physical outputs produced is usually not available in the firm-level dataset (De Loecker, 2011). To partially solve this problem, our study, following the empirical literature, uses industry-deflated sales. The productivity estimated as the residual in the production function is therefore considered as a measure combining real productivity and pricing strategies. The effects of trade liberalization may therefore reflect the response of price–cost mark-ups in addition to actual change in efficiency. However, as discussed by Topalova and Khandelwal (2011) and shown theoretically by Bernard et al. (2003), as long as price–cost mark-ups and true efficiency are correlated, then the revenue-based TFP measure also captures technical efficiency. To show the robustness of our results, we use labour productivity as an alternative proxy for firms’ efficiency. The results, available from the authors upon request, confirm the robustness of our findings.

42. The variable ‘Percentage of firms with LURCs’ might be problematic because many provinces have myriad firms with informal land rights either inherited from previous generations or purchased through informal exchange. Therefore, we show the regression results only for the security of tenure determinant.

43. The first specification also includes a dummy for the North/South of the country and two categorical variables for firms’ ownership. In line with previous results (Nguyen, 2021), we observe that firms located in the South are on average more productive that those in the North. Moreover, in line with the analyses of Le et al. (2019) and Tran (2019), FOEs and SOEs perform better than POEs in terms of their TFP levels.