Manufacturing Productivity and Firm Ownership in a Transition Economy: Analytical Issues and Evidence from Vietnam
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Manufacturing Productivity and Firm Ownership in a Transition Economy: Analytical Issues and Evidence from Vietnam Prema-chandra Athukorala and Hai Thanh Nguyen Arndt-Corden Department of Economics Crawford School of Public Policy Australian National University Prema-chnadra.athukorala@anu.edu.au haithanh.nguyen@anu.edu.au Abstract The paper contributes to the fledgling literature on firm ownership and industrial productivity in transition economies drawing on the experience of Vietnam. The ownership structure of Vietnamese manufacturing has undergone a dramatic transformation thanks to significant opening up of the economy to foreign direct investment and relaxing restriction on domestic private enterprises, notwithstanding the government’s commitment to preserving the role of state-owned enterprises in the economy. The results of our analysis undertaken using a new establishment-level panel dataset over the period 206-2017 indicate that transformation of the ownership structure has contributed to significantly improving productivity of the manufacturing sector, with both fully owned subsidiaries of multinational enterprises (MNEs) and MNE joint ventures in the domestic private sector paying a key role. Productivity of fully owned MNE subsidiaries is significantly higher than that of MNE joint ventures, supporting the view that relaxing ownership restrictions on foreign direct investment has been instrumental in improving manufacturing productivity. Both, state owned enterprises (SOEs) and MNE-SOE joint ventures are at the bottom of the productivity ranking by ownership mode. This comparison suggest that the choice between the state and private entrepreneurs is important in determining productivity implications of joint venture operation of MNEs: the state sector joint ventures are not immune from various productivity-retarding factors affecting SOEs. Key words: transition economies, Vietnam, productivity, ownership. JEL codes: F23, O14, P24, P23
1 Manufacturing Productivity and Firm Ownership in a Transition Economy: Analytical Issues and Evidence from Vietnam Abstract The paper to the fledgling literature on firm ownership and industrial productivity in transition economies drawing on the experience of Vietnam. The ownership structure of Vietnamese manufacturing has undergone a dramatic transformation thanks to significant opening up of the economy to foreign direct investment and relaxing restriction on domestic private enterprises, notwithstanding the government’s commitment to preserving the role of state owned enterprises in the economy. The results of our analysis undertaken using a new establishment- level panel dataset over the period 206-2017 indicate that transformation of the ownership structure has contributed to significantly improving productivity of the manufacturing sector, with both fully owned subsidiaries of multinational enterprises (MNEs) and MNE joint ventures in the domestic private sector paying a key role. Productivity of fully owned MNE subsidiaries is significantly higher than that of MNE joint ventures, supporting the view that relaxing ownership restrictions on foreign direct investment has been instrumental in improving manufacturing productivity. Both, state owned enterprises (SOEs) and MNE-SOE joint ventures are at the bottom of the productivity ranking by ownership mode. This comparison suggest that the choice between the state and private entrepreneurs is important in determining productivity implications of joint venture operation of MNEs: the state sector joint ventures are not immune from various productivity-retarding factors affecting SOEs. 1. Introduction The role of firm ownership in industrial performance is a subject of interest in many areas of economics and business studies. The subject has gained added impetus in recent years in the context of the process of economic transition in the former socialist countries. Unshackling the private sector from the state dominance is central to the move from ‘plan to market’. However, reflecting socioeconomic resistance and strong vested interested, not only the timing and sequencing of these reforms but also the choice of ownership modes in the divesture of state- owned enterprises has varied significantly among these countries. In most transition economies SOEs and domestic firms operates side by side with fully foreign owned firms and joint ventured of foreign owned forms and SOES and domestic firms with different degree of township among them. Differences in performance among diverse ownership groups is vital
2 for analysing overall performance of the manufacturing sector and informing the debate on further reforms. The early literature on ownership – productivity nexus has mostly focussed on the difference between public and private ownership and/or subsidiaries of multinational enterprises (MNEs) relating to competing domestic firms.1 The advent of transition economies calls for extending the analysis to account for hybrid nature of ownership patterns. However, the lack of good -quality firm and plant level data has been a serious impediment for researchers to meet the quest for knowledge.2 The purpose of this paper is to contribute to the fledgling literature on firm ownership and industrial productivity in transition economies drawing on the experience of Vietnam. Gradual transition from plan to market commenced with the announcement of the renovation reforms (doi moi) announced in 1986 have dramatically transformed the ownership structure of Vietnamese manducating over the past three decades. The notable ownership transition, coupled with the availability of establishment-levels data from a comprehensive annual manufacturing survey covering a period of sufficient length, make Vietnam an ideal case study of this subject. There are two previous studies on on the Vietnamese experience with ownership transition and industrial productivity: Ramstetter and Ngoc (2012) (covering the period, 2001- 2006 and Le, Pieri and Znomotto (2019) (for the period 2001-2011). The novelty of this paper lies in the following aspects. Firstly, the analysis is based on a newly constructed data set covering more recent years (2006-2017). This helps us to better capture structural changes resulting from the new phase of ownership resourcing took place in the second half of 2010s. Second, specific emphasis is placed on capturing the hybrid nature of the ownerships structure. We distinguish between fully state owned firms and joint ventures of state-owned firm, and fully owned subsidiaries of multinational enterprises and MNE joint ventures with local firms. Thirdly, there are several methodological improvements. For calculating real value added (output) we use the double deflator method, which take into account differences in price movements of final goods and intermediate inputs. There is evidence that reforms have significantly altered the relative prices gross manufacturing output and material inputs. Therefore, it is important to use separate price indices to deflate output and material inputs instead of directly converting nominal valued added into real term based on the mistaken assumption of fixed relative prices. We use generalized moments methods of Ackerberg, Caves and Frazer (2015) to estimate total factor productivity (TFP). This method has the advantage of addressing identification issues involved in the methodology commonly used in previous 1 See Ehrlich et al (1994), Caves 2007, Gorg and Greenaway 2004, Griffith (1999), Griliches and Regev (1999) and the works listed therein. 2 The previous studies of the productivity – ownership nexus in transition economies include Konnigs (1997), Jeffersob, Rawaski, Li and Yuxin (2000); Asaftei, Kunbhakar and Mantescu (2008), Jindra, Giroud and Scott- Kennel (2009), Ramstetter and Ngoc (2013), Basic and Slyexevic (2018), Chang, Chun and Moon (2013) and Le, Pieri and Znomotto (2019).
3 studies. Finally, deviating from the standard practice of estimating firm-level productivity based on a production function estimated by pooling all firms, we estimate the production function at the at the two-digit level of the standard industry classification to allow for differences in technologies. To preview the key findings, there is strong empire evidence that transformation of the ownership structure brought about by remorse over the past three decades has contributed significantly improving productivity of Vietnamese manufacturing, with both fully owned MNE subsidiaries and joint ventures with domestic private sector paying the key role. However, productivity of fully owned MNE significantly higher their private sector joint ventures. This results support the hypothesis that relaxing ownership restrictions on foreign direct investment has been instrumental in improving manufacturing productivity. Both, state owned enterprises (SOEs) and MNE-SOE joint ventures MNEs have recorded are at the bottom of the productivity ranking by ownership mode. This comparison clearly suggest that the choice between the state and private entrepreneurs important in determining productivity implications of joint venture operation of MNEs: state sector joint ventures are not immune from various productivity-retarding factors affecting SOEs. It further indicates that partial privatisation of fully state-owned enterprises would produce small marginal improvement in productivity growth that complete privatisation would. Firms that shift their ownership from SOE to domestic private enterprises perform better compared to firms that remain state ownership. On the other hand, domestic firms purchased by or in cooperation with a fully owned foreign firms to become. These results are consistent with our analytical priors and there are several indications that ownership serve as a predetermined variable in detraining inter firm in productivity. Firm-specific characteristics, such as size, age, ownership, location and capital intensity, and sector-specific characteristics, such as export exposure, import dependence and competition, are the important sources of productivity gains over and above the ownership modes for Vietnamese manufacturing firms. The paper proceeds as follows: Section 2 succinctly discussed the analytics underpinning the ensuing enquiry into productivity differentials by ownership. Section 2 provides information on policy reforms and transformation of the ownership structure of Vietnamese manufacturing. Section 4 analyses the patterns and trends of productivity performance of Vietnamese manufacturing enterprises. An econometric analysis of the
4 determinants of TFP growth of manufacturing enterprises in undertaken in Section 5. The concluding section summarise the key findings and make suggestions for further research. 2. Analytical context What could explain differences in productivity performance of firms belonging to different ownership categories? To the extent that enterprises of different ownership face the same production possibilities and have access to similar markets, systematic differences in productivity growth among ownership types must be interpreted through some firm-specific factors/assets that are endogenous to the performance of each ownership mode. In other words, persistent productivity differences among ownership type should prevail after controlling for variables commonly applicable to the conduct and performance of all firms. (Hart and Moore 1990, Ehrlich et al 1994, Shleifer and Vishny 1994). The dominant model of the effect of public ownership on firm performance is the public choice or property right model (Boardman and Vining 1989, Toninelli 2000). It postulates that property rights are more attenuated in state-owned enterprises than in private sector firms. A crucial difference between private and publically owned firms is that ownership in the latter not effectively transferable. Limits to transferability of ownership inhibits capitalisation of the firm’s future consequences into its current property rights and thus reduces the owner’s incentives to monitor managerial behaviour. Obligatory links with the state rooted in nonmarket objectives create ‘agency problem’ hinder profitable operation of state owed enterprises. Partial divesture of the state-owned industries in the process of economic transition could help perpetuate the ‘agency problem’ (Roland 2002, Schaffer 1998, Kornai, Maskin and Roland (2003). Economic reforms have not been synonymous with elimination of state intervention in enterprise decision making which is a sine quo nonfor productivity improvement. At the same time the ‘soft bouget constraint’ (Kronai 1986)3 resulting from the availability government subsidies and other budgetary supports and various extra-market privileges such as access to easy finance from government-owned financial institutions cushion these firms against competitive pressure created by natural maker forces. Under the soft budget constraint, enterprises can not only survive but also even engage in anti-competitive behaviour. Performance of SOEs and their joint-venture operations is also subject to ‘state capture’: the capacity of private agents to shape government policy impacting on these firms through illicit and non-transparent methods (Hellman and Shankerman 2000). 3 The ‘softening’ of the budget constraint appears when the strict relationship between the expenditure and the earnings of an economic unit has been relaxed, because excess expenditure will be paid by some other institution, typically be the paternalistic State.
5 There is a well-developed literature on the decision of multinational enterprises (MNEs) set up overseas subsidiaries rather than export directly or license their products or technology and the unique features of the operations of their foreign subsidiaries compared to the local firms in the host countries.4 MNEs possess relatively large amounts of intangible assets that facilitate efficiency (e.g., patents and other fruits of R&D, management knowhow, marketing resources). Thus, in the process of global reach, MNEs not only bring new investment to the host countries but also these intangible proprietary assets. Thus it is oftener asserted that multinational subsidiaries’ (foreign invested enterprises, FIEs) tend to be relatively more efficient other firms in a given host country: better management practices, specialized knowledge about production and access to better technology give them a cost advantage over domestic firms. Subsidiaries of manufacturing MNEs probably has the potential to avoid inefficiencies of small scale operations more often than their rivals because ‘parents’ supply of internally generated funds frees offspring of the bondage of capital rationing’ (Caves 1974). Technological secrecy and desire not to share proprietary assets with local entrepreneurs could make joint ventures less attractive compared to fully owned subsidiaries. Entering into host country industry by setting up a joint venture is subject to considerable risk of the joint venture partner and related parties free riding on the MNE’s intangible proprietary assets,, which determine its competitive edge in global operation. If the joint venture is the only ownership mode permitted under foreign investment law in a given host county, the MNE could resort to other means to restraint spillover of firm-specific assets and managerial expertise. Unlike in the case of a fully-owned subsidiary, performance of a joint venture also depend on the congruence of the partners’ goals, which is not an issue in the case of a fully owned subsidiary, So joint venture subsidiaries could be less productive compared to wholly owned subsidiaries, ceteris paribus (Brouthers and Hennat 2003). However, a joint venture could be a preferable entry mode for MNEs when a local partner has the potential to offer complementary knowledge, such as deep understanding of the local market and business environment, ability to circumvent bureaucratic read pater when meeting domestic administrative requirements, or access to distribution channels and domestic resources (Hennart 2009). These considerations are presumably more relevant for MMEs operating in diffused-technology industries for which protecting specific proprietary assets is not a major concern. Compared to MNE joint venture with domestic private entrepreneurs, performance of MNE-SOE joint venture could be adversely affected by the productivity-reading factors common to pure state-owned enterprises such as the agency problem, soft budget constraint 4 For surveys of this literature see Caves (2007), Dunning (1998), Gorg and Greenaway (2004), and Helpman (2006).
6 and the state capture. This could make MNE-SOE joint ventures less productive compared to their private sector counterparts. 3. Reforms and the ownership transition 3.1 Reforms Vietnam, unlike the transition economies in the Central and Eastern Europe and somewhat similar to the Chinese reforms, adopted a gradualist approach to unshackling the economy (Riedel and Comer 1997, Leong 2015, World Bank 2009, Vu-Thanh 2007 & 2019). The reform process, which was introduced under the slogan of doi moi (renovation) in 1986 and implemented gradually during ensuing three decades, set the stage for the development of a market economy characterised by a mix of state and private ownership. The first law on FDI passed in December 1987 specified three modes of foreign investor participation: (i) business cooperation contracts (BCCs), (ii) joint-ventures and (iii) fully foreign-owned ventures. Foreign participation in the fields of oil exploration and communication was strictly limited to BCCs. In some key sectors such as transportation, port construction, airport terminals, forestry plantation, tourism, and cultural activities, joint venture with domestic state-owned enterprises (SOEs) was specified as the sole mode of foreign entry. Fully foreign-owned ventures in other sectors were permitted only under special considerations according to policy priorities of domestic industrial development. The duration of foreign ownership of approved projects was limited to a maximum of 20 years, unless under exceptional circumstances. The Private Enterprises Law promulgated in the same legally recognized sole proprietorships and a legal basis for opening up limited liability and joint-stock companies. However, during the next decades or so notwithstanding some simplifications of legal procedure , retaining the SOE dominance continued to the preferred option of the reform process. The initial significant opening of the economy to foreign investment led to a growing resentment against FDI within certain circles of the Communist Party, resulting in adoption of a number of restrictive policy measures in1995-96. These included establishing Communist Party cells in FIEs, doubling commercial and residential rents for foreign enterprises and expatriate staff, imposing a maximum time limit of three years on work permits issued to foreigners employed in FDI projects, restricting foreign capital participation
7 in labour-intensive industries, and imposition a domestic-content and export-performance requirements on FIEs in a number of key industries. This policy backsliding raised serious concerns in the international investment community. Policy reforms gained momentum following the economic downturn during 1997-99, supporting the adage that ‘bad times lead to good reforms’. A new Enterprise Law that came into force in 2000 introduced a simplified procedures for setting up new enterprises, and permitted conversion of joint-venture FIEs (including joint ventures formed under BCCs) into fully-owned subsidiaries of parent companies. In December 2005, a new unified Investment Law was promulgated in place of the Law on Foreign Investment and the Law on Domestic Investment Promotion. The key features of this landmark legislation included treating foreign and domestic investors equally with regard to investment approval and incentives, providing investors with complete freedom in the choice of the mode of business entry, abolishing local-content and export-performance requirements, and introducing a decentralized three-tier system of investment approval. There have been further revisions to foreign investment and enterprise law in 2014 and 2016 to make it more consistent with international standards and practices. The noteworthy future of ownership reforms over the past three decades dramatic has been its dualistic nature. Significant peening up of the economy to foreign direct investment and relaxing restriction on domestic private enterprises took place side by side with the government commitment to preserving the role of SOEs in the economy (Vu-Thanh 2017 and 2017). Interestingly, following the country’s WTO accession in 2016, restructuring the SOE sector restricting SOE sector in order to make a leading force in the leading force in global economic integration because declared government policy. As part of this new policy emphasis some of the largest GCs were reorganized into large and diversified state-owned conglomerates, called “State Economic Groups” (SEGs). 3.2 Manufacturing performance and ownership structure Manufacturing sector has played a major role in Vietnamese economy since the ‘renovation reforms’ initiated in the late 1980s, and in particular, following significant trade liberalization and enterprise reforms in the early 2000s. Real manufacturing output has grown at an average annual rate of nearly 10 per cent during 1990-2017, with the exception of 2010 when there was an overall contraction in economic growth in the country. The direct contribution of manufacturing to GDP growth increased from 12.3 per cent in 1990 to around 20 per cent
8 during 2000-2009. After significant contraction in 2010, the contribution has gradually increased to 15.3 per cent in 2017. The manufacturing sector has directly generated 4.3 million jobs between 2005 and 2017. The share of manufacturing in total formal sector employment in the economy increased continuously from 11.8 per cent in 2005 to 17.3 per cent in 2017.5 Manufacturing growth has been underpinned by a notable shift in the ownership structure. The share of State-owned enterprises (SOEs) in manufacturing output (value added) declined from above 21.9 per cent in 2006 to 3.3 per cent in 2017. The share of SOEs in manufacturing employment dropped from 15.8 to 2.8% between these two years ( ). Output and employment shares foreign invested enterprises (FIEs) stood at 67.5% and 59.6% respectively in 2017, up from 47.6% and 38.9% in 2006. The output and employment shares of private domestic enterprises (PDE) declined marginally between these two years reflecting the growing importance of FIEs in manufacturing, but still accounted for much larger shares (29.2% and 37.6%) compared to that of the SOEs. Within FIEs, output and employment shares of joint venture (JV) decreased gradually over the period due to relaxation of ownership restrictions that permitted MNEs to form fully owned subsidiaries. The shares of fully-owned 60.6% output and 56.7% of employment in 2017. The comparable figures for MNE joint ventures had declined sharply to 6.9% and 2.9% respectively. The decrease in output and employment contribution of JV was mainly attributed to the decrease in that of joint-ventures with state-owned enterprises (JV-SOE). In sum, there has been a notable shift in the ownership structure of Vietnamese manufacturing from public ownership to private ownership, with foreign owned enterprises, in particular fully-foreign owned firms accounting playing a pivotal role in this structural change. Table 1: Ownership structure of manufacturing output and employment in Vietnam, 2006-171 2000-01 2006-07 200-11 2016-17 % Output (value added) State-owned enterprises (SOEs) 56.5 19.4 11.0 4.2 Private domestic enterprises (PDEs) 23.3 33.3 37.5 29.3 Foreign invested enterprises (FIEs) 20.2 47.4 51.5 66.6 Fully foreign owned (FOF) 14.5 32.6 41.8 58.7 Joint-venture (JV) 5.7 14.8 9.7 7.9 JV-SOE --- 12.6 7.2 5.5 JV-PDE --- 2.2 2.5 2.5 % Employment State-owned enterprises (SOEs) 55.5 14.1 6.8 3.1 Private domestic enterprises (PDEs) 32 46.2 46.6 38.8 5 Data used in this paper, unless otherwise stated, are from The Statistical Yearbook, Hanoi: General Statistical Office (GSO) (various issues).
9 Foreign invested enterprises (FIEs) 12.5 39.8 46.7 58.2 Fully foreign owned (FOF) --- 35.2 43.1 55.1 Joint-venture (JV) --- 4.7 3.7 3.1 JV-SOE --- 2.4 1.7 0.8 JV-PDE --- 2.3 2.0 2.4 Note: (1) Two-year averages --- Data not available Source: Data for 2000-01 are from Ramstetter and Ngoc (2012) ; for other years compiled from the Vietnam Enterprise Survey database. Manufacturing enterprises of all ownership types are heavily concentrated in the Red River delta in the North and the Southeast in the South (Table 2). These regions have better infrastructure, especially better roads and proximity to airports and seaports. Within these regions, there has been a heavy concentration of firms in Hanoi and Hochiminh city, the two largest metropolises of the country. The degree of concentration of fully owned MNE subsidiaries and MNE joint ventures in in and joint venture in red river delta and south East regions is much larger compared to SOEs and domestic private firms.
10 Table 2: Performance of manufacturing enterprises by region and ownership, 2017 Ownership Northeast and Red River Delta North South Central South-east Mekong Total Mountainous (Hanoi)1 Central and Highland (Hochiminh city)2 Delta region region % of plants SOE 8.1 36.1 (20.3) 9.55 7.8 30.8 (17.3) 7.8 100 PDE 4.8 33.8 (17.5) 4.35 8.1 39.6 (24.1) 9.5 100 FOF 5.7 28.2 (4.4) 0.99 2.6 54.8 (10.4) 7.8 100 JV-SOE 3.9 39.7 (14.1) 5.13 3.9 38.5 (20.5) 9.0 100 JV-PDE 5.1 28.9 (10.7) 3.65 7.0 42.7 (14.3) 12.6 100 Total 5.0 33.1 (15.8) 3.96 7.3 41.5 (22.2) 9.3 100 % Output (value added) SOE 8.6 29.6 (18.8) 3.86 10.7 40.7 (30.0) 6.6 100 PDE 4.5 27.0 (9.8) 2.29 8.0 45.5 (26.3) 12.8 100 FOF 15.8 34.7 (5.3) 1.22 1.6 41.6 (10.6) 5.2 100 JV-SOE 0.1 78.2 (18.4) 1.78 2.1 16.6 (8.1) 1.6 100 JV-PDE 2.8 42.9 (23.2) 1.70 3.4 33.6 (12.1) 15.6 100 Total 11.2 34.5 (8.1) 1.66 3.8 41.4 (15.8) 7.5 100 % Employment SOE 10.4 30.1 (17.4) 7.40 17.1 30.2 (21.6) 4.9 100 PDE 5.5 31.0 (10.8) 3.33 8.9 37.6 (18.9) 13.6 100 FOF 8.0 28.3 (3.9) 3.71 2.8 48.6 (11.1) 8.6 100 JV-SOE 3.5 66.0 (28.1) 2.32 7.0 18.1 (10.0) 3.3 100 JV-PDE 4.0 27.2 (12.4) 1.81 3.2 29.2 (12.1) 34.7 100 Total 7.0 29.6 (8.1) 3.62 5.5 43.3 (14.3) 10.9 100 Note: (1) Figures for Hanoi are given in brackets. (2) Figures for Hochiminh City are given in brackets Source: Data compiled from the VES 2006-2017. 4 Total Factor productivity: measurement and patterns 4.1 Measurement of productivity Labour productivity is the most commonly used measure of productivity. However, it is not a comprehensive measure of productivity. As a partial measure of only the efficiency of labour but also the impact of other inputs, in particular capital, which with the workers work. For instance, two firms with the same quality of labour may have different labour productivity if one of them have more capital (and or technology) for the workers to work with. Total factor productivity (TFP), the measure of productivity used in this paper, is the change in output over and above all factors of production used in the production process. Unlike labour productivity, Let us begin with the standard production function:
11 = 0 + + + (1) where is the log of output (value added), is the log of capital input, and is the log of labour input, all of which are observed, for firm i at period t. Hence, TFP is ̂0 + ̂ , where the first term can be interpreted as the mean efficiency level across firms and the second term is the deviation from that mean for a given firm. Estimation of (1) raises econometric issues because input choices are likely to be correlated with the firms’ productivity : more efficient producers are, all else equal, likely to use more inputs. There is also potential selection bias when a panel data are used, because less efficient producers – those with low are more likely to exit from the sample coverage. Therefor, is assumed to be additively separable in a transmitted component and an i.i.d component . = 0 + + + + (2) The term represents “productivity” shocks that are potentially observed or predictable by firms when they make input decisions. Intuitively, might represent variables such as innate technology or managerial ability of a firm, expected production disruption due to machine breakdown, expected defect rates in a manufacturing process, soil quality or the expected rainfall at a particular farm’s location. In contrast, the term represent shocks to production or productivity that are not observable (or predictable) by firms before making their input decisions at time t. might represent deviations from expected breakdown, defect, or rainfall amounts in a given year. Inputs and are chosen by firms. This means that if the econometrically unobservable is observed at least partially by the firm prior to choosing and , then these choices likely depend on , creating correlation between ( , ) and , and rendering OLS estimates of and inconsistent. To control for the correlation between unobservable productivity shocks and input levels, Olley & Pakes (OP) (1996) suggest using a firm's investment as a proxy variable for a firm's productivity. Since investment may not fully respond to changes in productivity, Levinsohn & Petrin (LP) (2003) propose using intermediate inputs rather than investments in the control function. According to OP and LP, inputs are assumed to be of two types: variable inputs such as labour and intermediate inputs, the choice of which in the current period does not have an impact on their cost of use in the future periods and ‘state’ variables whose choice have an impact on the future cost of input. Thus, equation (2) is rewritten as: = 0 + + + + (3)
12 Ackerberg, Caves and Frazer (ACF) (2015) argue that the approaches of OP and LP suffer from an identification issue because of treating labour as a state variable: choice of labour potentially have an impact on he future cost of inputs given significant hiring or firing costs. ACF, therefore, allow labour input to have dynamic effects and include it in the demand function of the proxy variable in (4), which is = −1 ( , , ) = ℎ ( , , ). Thus, labour coefficient, along with all other input coefficients, are identified in the second stage of their procedure instead of in the first stage as proposed by OP and LP, particularly: Stage 1: Insert the inverse of the demand function for the proxy variable: = 0 + + + ℎ ( , , ) + (1) To obtain the estimate of the composite term: ( , , ) = 0 + + + ℎ ( , , ) Stage 2: With this estimate, ̂ ( , , ), estimate all parameters using the equation: ̂ = 0 + + + ̃( −1 − 0 − , −1 − , −1 ) + + (2) After obtaining consistent estimators for and , the level of productivity is estimated as (7): ̂ = ( ̂ − ̂ − ̂ ) (3) In this paper we use the ACF method. Estimation is done using the Stata (acfest) subroutine developed by Manjón & Mañez (2016), which provides for estimate the productivity with the Wald test to test joint significance of the explanatory variables, and the Sargan-Hansen J test for over identification. 4. 2 Data Data were compiled for the period 2006-2017 from the unpublished returns to the Vietnamese Enterprise Surveys (VES) conducted annually by the General Statistical Organization of Vietnam (GSO). This is one of the most disaggregated establishment-level data sources available for transition economies. The annual returns of establishments have been linked across years so that an annual data panel can be constructed for the relevant variables. We extracted data for establishments operating in manufacturing industries VSIC 10 to VSIC 31 excluding petroleum and gas (VSIC 19). Petroleum and gas industry is excluded for two reasons: most of the production is exported in semi-processed forms, and, unlike other products, petroleum is subject to world market price fluctuations. The survey data enabled us to groups establishments under five ownership categories: State-owned enterprises (SOEs), Private domestic enterprises (PDEs), fully owned foreign invested enterprises (FOF), joint
13 ventures of SOEs and foreign invested enterprises (JV-SOE), and joint ventures of PDEs and FOFs (JV-PDE). The VES has been conducted annually since 2000. However, we decided to limit the time coverage of f our analysis to 2007-2017 because precise lining of firms and the relevant performance variable is not possible for the entire period because of changes in industry code with effect from 2007. In any case, we believe that focussing on this period rather than the entire period from 2000 better suit our purpose given the significant structural break in ownership patterns resulted from the policy reforms undertaken during 2005-06. The VES data presumably has an inherent “large firm bias” because the ‘size’ cut-off point used in selecting firms (which varies across years), and probable low response by relatively smaller firms. However, we believe that this is does not pose a major problem in inter-industry analysis of productivity patterns. By inspecting firm (establishment)-level data from the 2006 manufacturing census, we found that firms with employment of more than 10 workers account for over 96 per cent of gross manufacturing output and 89 per cent of manufacturing employment. Output (value added) and capital stock are measured in real terms (at 2010 prices). Real output is computed using the ‘double deflator’ method with separate deflators for gross output and intermediate inputs constructed from data provided by the GSO. Wholesale price indices at the two-digit level of VNSIC are used as gross output deflators. Deflators of intermediate inputs are constructed by combining wholesale price indices with data on the input stature of production from the 2012 Input-Output constructed by the GSO. Data series for real capital stock are derived using implicit price deflators derived from data on current and real (2010 =100) gross fixed capital formation (at the 2-digit level).6 The number of workers employed is used to measure labour input. This is presumably a good proxy for the number of labour hours involved in manifesting production because there has not been any changes in the length of the working week during the period under study. Unfortunately, the available data does not permit us to adjust the employment data for change in labour quality. Previous plant/firm level studies have mostly estimated firm-level TFP based on a production function estimated for the entire manufacturing sector based on the implicit assumption of homogeneity in technology among all firms. In order to minimise possible estimation bias resulting from using this restrictive assumption, we estimate plat-level productivity based on production functions estimated for each 2-digit manufacturing 6 Price indices used for deflating current gross output, inputs and capital stocks together with a methodological note are available from the authors on request.
14 industries.7 Tobacco industry (VNSIC 12) is excluded from estimation because it has too few firms. 4.3 Trend and pattern of Total factor productivity TFP estimates for total manufacturing and the five ownership groups are plotted in Figure 5. To facilitate comparison data series are plotted as indices (2016 = 100). The index of TFP in total manufacturing increased from 100 to 135 during 2016-2017, recording an average annual growth rate of 2.0%. Within manufacturing, all ownership groups record increase of productivity during 2006-2017. However, both SOEs and SOE-joint ventures with MNEs have recorded slower TFP growth (1.1% and 2.1%, respectively) compared to the other three ownership groups (FOF: 3.2%, PDE: 4.3, JV-PDE: 4.2). Overalls, this first look at data show an interesting patterns of domestic private firms closely following (or even performing slightly better than) the productivity patterns of fully- owned MNE subsidiaries and their joint ventures with private firms. This has occurred in a context of significant increase in the role of foreign firms (both forgoing ventures and fully owned MNE subsidies) in Vietnamese manufacturing (Table 1). At first blush, these patterns seems to suggest ‘superior technology or productivity imported by the subsidiaries progressively spills into their domestic rivals’ (Caves 2007, p. 214). If there were no efficiency levels were exogenous, large foreign shares in output should be correlated with low productivity levels for domestic competitors. Figure 2 Manufacturing TFP growth by ownership 7 In this we follow Newman, Rown, Talot and Tarp (2015 ).
15 160 150 140 TFP (2006 = 100) 130 120 110 100 90 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 SOE PDE FOF JV-SOE JV-PDE Total Note: (1) Average annual TFP growth rates: state-owned enterprises (SOEs): 1.1% ; private domestic enterprises (PDEs): 4.3)%; fully owned MNE subsidiaries (FOF): 3.1% ; SOE-MNE Joint-ventures (JV-SOE): 2.1%; PDE-MNE joint ventures (JV-PDE): 4.2%; total manufacturing: 2.0%) Source: Data compiled from the VES 2006-2017. TFP estimates for two-digit industries by ownership mode are summarised in Table 2. All industries, except printing (VNISC 18) and pharmaceuticals (VNISC 21) show productivity growth, though at varying degree. The data shows relatively higher productivity growth in industries in which Vietnamese manufacturing has been rapidly integrating within global production networks (eg. Electronics (NVISC 26), electrical goods (VNISC 27), Motor vehicles (VNISC 29)). These are also the industries with heavy concentration of the presence of foreign-invested enterprises.
16 Table 2: Vietnamese manufacturing: Output (value added) share and TFP growth by ownership type (2006-17) (%) VSIC code Industry Average output (value added) share 2006-2017 TFP average growth 2006-2017 Total SOE PDE FOF JV-SOE JV-PDE Total SOE PDE FOF JV-SOE JV-PDE 10 Food products 14.3 6.5 55.3 26.5 6.7 4.9 1.0 -1.3 6.4 1.2 4.4 2.6 11 Beverage 3.3 35.0 18.5 16.7 29.1 0.8 0.5 -1.0 8.0 1.1 -7.4 8.2 13 Textile products 3.8 11.5 28.7 54.6 3.6 1.7 0.9 5.4 2.2 2.3 -11.1 4.7 14 Wearing apparel 11.6 5.7 40.1 51.7 0.7 1.8 0.5 0.3 1.9 0.8 -0.4 1.5 15 Leather and related products 9.6 1.4 20.3 73.9 0.3 4.2 0.2 1.8 0.9 0.4 -2.4 -0.7 16 Processed wood/bamboo (excluding furniture) 1.4 7.7 72.3 15.1 2.4 2.5 0.7 2.7 2.4 0.1 -0.2 2.2 17 Paper and paper products 1.9 8.2 52.1 37.2 0.6 1.9 0.4 -0.7 0.1 0.7 -11.9 4.7 18 Printing and publishing 1.1 37.3 43.8 18.3 --- --- -0.5 -2.7 -0.7 -0.4 --- --- 20 Chemical 5.5 33.3 20.4 39.8 4.9 1.6 0.6 1.4 2.3 2.8 -5.3 0.7 21 Pharmaceuticals and chemical products 1.7 8.9 70.1 14.5 3.2 3.5 0.0 2.5 0.5 -1.2 -8.1 -0.3 22 Rubber, plastic products 4.5 9.2 39.9 47.5 1.0 2.3 0.9 7.4 1.2 2.6 29.9 6.7 23 Non-metallic mineral products 6.1 22.0 50.8 12.3 13.6 1.4 1.8 15.3 5.8 3.7 21.3 8.9 24 Metals 2.3 19.0 52.0 20.0 6.9 2.1 1.2 3.3 9.4 0.1 --- 7.4 25 Fabricated metal products 4.8 6.7 43.7 41.9 5.0 2.7 1.4 8.5 4.9 2.1 14.0 7.2 26 General machinery and equipment and parts 1 10.8 3.0 3.1 86.5 6.8 0.7 3.1 10.4 2.1 7.3 0.6 5.9 27 Electrical machinery and apparatus 3.6 8.6 24.1 62.1 2.8 2.4 0.9 4.0 2.8 1.7 19.1 4.3 28 Machinery and equipment, other 1.8 3.9 34.7 59.8 --- 1.1 0.8 0.4 2.0 1.3 --- 3.9 29 Trailers and motor vehicles 3.0 10.8 17.9 48.1 21.8 1.4 4.4 18.0 2.0 10.0 --- --- 30 Other transport means 4.7 11.4 7.1 30.5 46.9 4.1 2.0 5.9 6.4 1.5 21.1 4.7 31 Furniture 4.3 2.4 44.4 50.8 0.3 2.2 2.0 4.7 1.8 4.3 -- 3.5 100.0 12.6 37.0 40.4 7.9 2.2 1.1 4.3 3.1 2.1 4.2 4.2 Note: (1) Includes electronics --- In this industry, there were no firms of this ownership type in operation during the entire period or in any year/s within this period. Source: Data compiled from the Vietnam Enterprise Survey database.
17 5. Ownership and productivity In this section, we empirically analyse the effects of ownership types on TFP. For this purpose, the estimation equation is specified as follows: = + 1 + 2 + + µ + + (1) Where TFP is total factor productivity, = 1, 2, … , is the firm, = 1, 2, … , is the 2-digit sector and = 1, 2, … , is the time unit in years. : Firms’ fixed effects; µ: 2-digit sector dummy variables; : Year dummy variables; : Error term assumed to be independent of explanatory variables. The main variable of interest, ownership, is captured by ownership dummies: state-owned enterprises (SOE), private domestic enterprises (PDE), fully-owned foreign enterprises (FOF), joint-ventures with state-owned enterprises (JV-SOE) and joint-ventures with private domestic enterprises (JV-PDE), private domestic firms (PDE) as the reference group (base dummy). The Control variables (Xij) are listed below with the expected sign of the regression coefficients in brackets (+ or -) Firm-size dummy variable based on the number of employees: micro: 0-9 employees, small: 10-49 employees (base dummy), medium: 50-299 employees and large: more than 300 employees.8 (+ or -) Number of years of operation based on the year of entry (+ or -) Geographical location dummies: Northeast and Mountainous region as the base dummy, Red River Delta, North Central, South Central and Highland, Southeast and Mekong Delta (+) Export orientations (export-output ratio at 2-digit industry level) (+) Import dependence (import-output ratio at 2-digit industry level) (+ or -) Herfindahl Index of industry concentration measured at 2-digit level Among the control variables, firm size ( ) is included to capture the impact of scale on firms’ productivity. Large firms are generally expected to be more productive compared to micro and small firms because large firms tend to lower production costs and absorb more resources (Van Biesebroeck, 2005). In contrast, small firms may have higher productivity due to their lean organizational structure (Utterback 1994 and Williamson 1967). Firm age ( ) is included to control for the possibility that observed productivity- level difference across different types of ownership could be a function of the age distribution 8 In experimental runs, we used firm size measured by the number of variable as a continuous variables, but it was not possible to retain it because of high multicellularity.
18 of the firms being compared. Age can capture the vintage capital effect and learning-by-doing. Younger firms produce with greater efficiency and better technology than older plants. In particular, when an innovation is introduced, it is generally adopted by newer enterprises, while others tend to delay as it may be costly to scrap products of not so recent-vintage (Caves and Barton, 1990, Gittleman, ten Raa, & Wolff, 2006). However, the longer a firm survives in an industry the more productive thanks to learning-by-doing (Jovanovic & Nyarko, 1996; Kraft, 1991) or the purging of unproductive firms over the years (Hopenhayn, 1992). Location dummies variables (GLDs) are included in the model by dividing the firms by six main regions of Vietnam: Northeast and Mountainous (base dummy), Red River Delta, North Central, South Central and Highland, Southeast and Mekong Delta. Given the large historical, institutional, and economic differences persist over time between these areas, it is expected that firms’ productivity is different across these regions. Of these six regions, South Vietnam was under central planning for only a much shorter period compared to those in the North. Presumably, this can have a significant impact on firm-level productivity differences. The trade-intensity of the sector may affect the productivity performance of firms. There is evidence that only more productive firms can withstand exposure to trade and the least efficient firms are forced to exit as they can no longer compete (Melitz, 2003). Similarly, in import competing sectors, firms have to remain efficient in order to survive (Pavcnik, 2002). The main impact of trade liberalization is thus to induce a reallocation of resources across firms forcing the least productive to exit and the most productive to expand. Industry concentration ( ) is included to capture the effects of market power on firms’ productivity. Greater competition will pressure firms into adopting new technologies and operating more efficiently (Nickell, 2002). Also endogenous growth theory postulates that models that the monopoly rents coming from low level of competition are invested in R&D which in turn leads to innovation and improvements in TFP (Dixit & Stiglitz, 1977; Grossman & Helpman, 1991; Romer, 1990). It has also been shown that, under some conditions, increased competition can lower the expected income of managers and therefore their effort, which in turn reduces firm efficiency levels (Hermalin, 1992). For these reasons, industry concentration is included as an additional explanatory variable in the regression. To study the industry concentration, we use the Herfindahl index9 (Hart, 1975). This index is a measure of the size of firms in relation to the industry and an indicator of the amount of competition among them. Decreases in the Herfindahl index indicate an increase in competition, whereas increases imply the opposite. 9 The standard Herfindahl Hirschman Index (HHI) constructed by squaring the market share of each firm competing in a market and then summing the resulting numbers: as = ∑ =1 2 , where 2 is the output share of firm in sector at time .
19 Data for all variables other than the two trade exposure variables (MOR and EOR) are compiled from the VER database. Data series for MOR and EOR are constructed at VNSIC two-digit level by combining trade data from the UN COMTRADE database and gross manufacturing output data from VER databases. In estimating the TFP equation, , , , , and are measured in natural logarithms. 5.2 Estimation method We use the correlated random effects (CRE) method for estimating the TFP equation ( to the static TFP model (Schunck & Perales 2017, Wooldridge 2019). Fixed effects and random effects estimators are the common methods applied to panel data. Unlike the standards OLS estimators, both these estimators take account of unobservable individual heterogeneity. The fixed-effects (FE) estimator offers consistent estimators but does not allow us to estimate time- invariant variables since it is based on the within operator. On the other hand, the random- effects estimator (RE) increases the efficiency of estimations but imposes a strong assumption that individual effects are not correlated with explanatory variables. However, it can yield biased and inconsistent coefficient estimates if one or more explanatory variables are endogenous (i.e., if they are jointly determined together. The CRE approach is a mid-way house between the FE and RE estimators. It provides within estimates, analogous to FE by subtracting the cluster mean of time‐variant variables in a random effects model and offer the offers advantages over FE in terms of capability to estimate the effects of time-invariant variables. It is relaxes the restrictive assumptions of un-correlation between the unobserved heterogeneity term and other explanatory variables. Moreover, it allows for the statistical disentangling of effects of time-variant within-firm determinant factors and time-variant between-firm determinant factors, taking the panel data design into account (Schunck and Perales 2017). In our case, this feature of CRE is particularly important for capturing productivity level differences due to an ownership change within a firm or switch from state to private ownership. 5.3 Results Table 3 presents CRE estimates of the productivity equation. Alterative RE and FE estimates are given in the Appendix Table A-1 for comparison. The correlation metres of all variables is given in Table A-2 to facilitate interpretation of the results.
20 Table 3: Total Factor Productivity in Vietnamese Manufacturing: CRE estimates (Dependent variable = Log TFP) (1) (2) (3) Ownership (with PDE as the base dummy) L.SOE 0.091** 0.038 0.0135 (0.028) (0.028) (0.0333) L.FOF 0.179*** 0.156** 0.140** (0.048) (0.048) (0.0483) L.JV-SOE 0.126* 0.068 0.0802 (0.062) (0.061) (0.0614) L.JV-PDE 0.154** 0.124** 0.127* (0.050) (0.047) (0.0499) Log AGE 0.182*** -0.179*** (0.012) (0.00849) Plant-size dummies (SIZE) Micro -0.179*** 0.0936*** (0.008) (0.00846) Medium 0.093*** 0.102*** (0.008) (0.0155) Large 0.101*** 0.182*** (0.015) (0.0119) Regional dummies (RGD) Red River Delta 0.155*** 0.142*** (0.019) (0.0188) North Central -0.058* -0.0632* (0.026) (0.0255) South Central and Highland 0.028 0.0142 (0.022) (0.0225) Southeast 0.406*** 0.387*** (0.019) (0.0187) Mekong Delta 0.154*** 0.147*** (0.022) (0.0214) Export orientation (EOR) 0.352*** 0.360*** (0.012) (0.0123) Import dependence (MDR) 0.176*** 0.175*** (0.013) (0.0134) Log Herfindahl Index (HPI) -0.035*** -0.0315*** (0.004) (0.00485) FOF*EOR -0.0553** (0.0177) SOE*AGE 0.0136 (0.0115) Constant 4.840*** 5.888*** 5.889*** (0.0210) (0.077) (0.0757)
21 Observations (N) 154317 154317 154317 AIC2 360395 346597 345639 BIC2 361081 347472 346554 Notes (1) Heteroscedasticiy robust standard errors are are given in parentheses with the statistical significance of the coefficients denoted as * p
22 The coefficient of SEO*AGE is statically insignificant. This result suggests our inference relating to the relative productivity perforce of SOEs is not susceptible to the vintage effect. The coefficient of FOF*EOR is statistically sufficient with the negative sign. The results suggest that the level of productivity in export-orated FOF amount to abbot two-thirds of the average level of FOF. The results could possibly reflect lower productivity gains at the formative stage of engagement in simple assembly processes within global production networks. This is an interesting issue, which deserves further investigation. To comment briefly on the results for the control variables, coefficient estimates of the size variables support the hypothesis that large firms are generally more productive. In particular, the coefficients of the micro dummy variable are significantly negative. The coefficient of AGE is positive and statistically significant suggesting that firms in older cohorts operate more productively. There is evidence from the regional dummies that on average productivity of firms located in the Southeast and Mekong Delta are much higher than the national average. This means, lack of domination of large firms in the market structure may stimulate competition and increase firm productivity. Both export orientation and import penetration seems to promote productivity, with that former having a much greater effect competed to the latter. Increase in industrial concentration seems to have a mild negative effect on manufacturing productivity. 5. 4 Robustness check In the productivity - ownership nexus, arguably causality may not necessarily go from later to the former. For instance, private firms with poor productivity may be more likely to come under state ownership for political reasons. Alternatively, the more efficient enterprises are privatised first to make the privatisation strategy more active to the private sectors. Any unobserved time- invariant firm-specific characteristics, which can increase or decrease productivity of one enterprise, may encourage its owners to decide to merge with or acquire other enterprises, change their ownership types. For example, a good management level of domestic enterprises may allow them to merge with FIEs, leading to changes in ownership. There is a tendency for R&D-oriented and human-capital intensive MNEsas well as domestic firms to their operations in regions with high TFP in an attempt to benefit from local technology slipovers or other agglomerations advantages.
23 It is therefore important to tackle the problem of indigeneity of firm’s ownership in the econometric analysis. However, unfortunately, suitable external instruments that meet the criterion of highly correlated with the ownership variables but not correlated with productivity, are hard to come by in this type of an econometric analysis. We therefore attempted to check the robustness of the results in two ways. First, we estimated the model limiting the sample only to the firms whose ownership did not change during the period under study. This is a suitable test of the sensitivity of results to productivity proposed ownership changes (Ehrlich, Gallais-Hamonno, Liu and Lutter 1994). Second, we estimated the model using Hausman- Taylor instrumental variable estimator (HT) (Hausman & Taylor, 1981) which aims to minimise indigeneity by using instruments derived within the model. Third, we conducted the standard Hauman test of the potential endogeneity of the ownership variables for btoh CRE and and HT. Result are rematablely rescilients to these alternative test suggesting that ownership serves as a pretermined variables in our regessions. In addtiton, we reestimated the model alfer exclluding micro firms from the data set and also after truncating the time coverage to 2010-16. As already discussed there has been significant chnages. From titme to time, in coverage of these farins from time in the VES. There are also gneral concerns about the quality of data reported by these firms. The time coverage was truncated to 2010-16 allows for production disruptions during th global financial crisis (2008-09). We tested five-firm concentration ration as an alternative to HPI, square of AGE in addtiton to AGE (totest for possible non-linerity of the vintage effect), used non-log forms of HPI, EOR, MPR is alternative regression runs. The results were robust to all these test. 6. Conclusions We have examined the role of ownership in manufacturing productivity in Vietnam in the context of policy reforms undertaken over the past three decades. The ownership structure of Vietnamese manufacturing has undergone a dramatic transformation thanks to significant opening up of the economy to foreign direct investment and relaxing restriction on domestic private enterprises, notwithstanding the government’s commitment to preserving the role of SOEs in the economy.
24 The results of our analysis undertaken using a new establishment-level panel dataset over the period 206-2017 indicate that transformation of the ownership structure has contributed to significantly improving productivity of the manufacturing sector, with both fully owned MNE subsidiaries and MNE joint ventures in the domestic private sector paying a key role. However, productivity of fully owned MNE significantly higher their private sector joint ventures. This results support the hypothesis that relaxing ownership restrictions on foreign direct investment has been instrumental in improving manufacturing productivity. Both, SOEs and SOE joint ventures with MNEs hare at the bottom of the productivity ranking by ownership mode. This comparison clearly suggest that the choice between the state and private entrepreneurs important in determining productivity implications of joint venture operation of MNEs: state sector joint ventures are not immune from various productivity-retarding factors affecting SOEs. Superior productivity performance of MNE subsidiaries naturally directly reflects in the measured productivity performance of domestic manufacturing. However, in assessing national gains from MNE presence in the economy, a crucial issue is the extent to which the quasi rent generated by MNEs from their proprietary assets spill over to the domestic private sector. This is an important subject for further research. References Ackerberg, D. A., Caves, K., & Frazer, G. (2015), Identification Properties of Recent Production Function Estimators. Econometrica, 83(6), 2411–2451. Asaftei, Gabriel, Subal C. Kumbhakar, and Dorin Mantescu (2008), ‘Ownership, business environment and productivity change.’, Journal of Comparative Economics 36 (3), 498- 509. Aghion, Philippe, Richard Blundell, Rachel Griffith, Peter Howitt, and Susanne Prantl (2009), ‘The effects of entry on incumbent innovation and productivity’, The Review of Economics and Statistics 91(1), 20-32. Athukorala, Pprema-chandra and Tien Kien Nguyen (2009), Transition to a market economy, foreign direct investment and export performance in Vietnam, in J. London (ed.), Routledge Handbook of Contemporary Vietnam, London: Routledge (forthcoming) Bačić, Katarina, Ivana Rašić Bakarić, and Sunčana Slijepčević (2018), ‘Sources of productivity differentials in manufacturing in post-transition urban South-East Europe.’, Post-Communist Economies 30 (4), 526-548. Benfratello, Luigi, and Alessandro Sembenelli (2006). ‘Foreign ownership and productivity: Is the direction of causality so obvious?.’, International Journal of Industrial Organization 24(4 ), 733-751.
You can also read