Factors Affecting the Support of Industrial Businesses’ Performance in Vietnam’s Digital Economy

Abstract

This study analyzes the factors that affect the technical efficiency (TE) of firms in the supporting industry in the context of Vietnam’s digitalized economy. Stochastic frontier analysis (SFA), Fixed Effect Models, and System-GMM methods are applied to reach the findings that the quality of human resources, capital intensity, and firm size have positive effects on TE. Furthermore, exogenous environmental factors, such as the domestic demand of an industry impacting all upstream businesses, which use inputs that are products of that industry (BSpill-ratio), and the FDI backward effect (BFSpill), also exhibit positive effects. These confirm that the linkage between domestic supporting industry suppliers and FDI assembly enterprises plays an important role in improving TE. Vietnam’s digital transformation since 2020 has also created some interesting changes in the correlation coefficient. Location, sectors, competitiveness, and investment environment are also considered, and the results suggest that they are all determinants to be considered in management policies at both the firm level and the government level. Our contribution in this study is new policies aligned with many major changes in the world economic context, such as the tough tariff policy implemented by recent presidential administrations and a series of reforms of the Vietnamese Government, as well as strong digital transformation in Vietnam. The key findings of this research are important as they confirm which factors are really determinants for the Vietnamese government to implement investment policies for this industry effectively.

1. Introduction

In recent years, Vietnam has become a developing market in Asia that attracts several technological enterprises and prominent global companies, such as Intel (USA), Amata (Thailand), Foxconn (Taiwan), Samsung, Bumjin (Republic of Korea), Toyota, Mitsubishi, Toray, Yazaky (Japan), Wilmar (Singapore), and so on. It can be seen that the structural transformation between industries has changed positively: the proportion of processing, manufacturing, and high-tech industries in Vietnam has increased over the past ten years. The role of the supporting industry has become very important. It is the foundation to help key industries develop and avoid material shortages and supply chain disruptions (Wirjo & Cheok [1]; Le & Diem [2]). Whether a country’s industry develops or not depends on this interdependent causal relationship. That is why the government, as well as researchers, is concerned with understanding how to improve the performance of supporting industry businesses effectively and what factors determine good performance for this kind of business. Moreover, on 31 December 2020, the Ministry of Industry and Trade of Vietnam also approved the Supporting Industry Development Program, with a large budget; the goal is to focus on developing supporting industries. This program requires more in-depth research to determine the advantages of each region, helping to shape the development process of the supporting industry.

However, it is recognized that the supporting industry in Vietnam still mainly relies on the available advantages from natural resources and tends to be labor-intensive. Domestic supporting industry firms mainly produce simple, low-value components and only receive contracts from FDI, which means they participate in the value chain as the second- and third-level suppliers for direct-export FDI enterprises. Therefore, it is difficult for domestic firms to receive technology spillover from MNCs. Regarding previous experimental studies, there are several conflicting results. The majority of studies focus on manufacturing and textiles, garments, and footwear since they are conducted in developing countries. Many research studies focus on the effects of firm-specific factors, such as firm size, age, size and quality-of-labor resources, and so on, on businesses’ technical efficiency. Furthermore, one of the motivations for the authors to conduct this study is that although the Vietnamese government has implemented policies to support the development of supporting industries, they still develop unevenly and ineffectively.

In 2019, the COVID-19 pandemic broke out and quickly spread around the world, depleting many economies. The social distancing policy had a great impact on the production process; many businesses had to suspend production to implement epidemic prevention. The COVID-19 pandemic disrupted the supply chain, causing difficulties for supporting industry enterprises. At the same time, it also promoted the trend of digital transformation in businesses to survive and adapt in the new context of the economy. Businesses are aware of the role of digital transformation in their operational efficiency. Whether digital transformation will have any impact on the development of this industry is still a question that needs to be answered because the COVID-19 pandemic is still showing signs of returning, or other epidemics and political crises can still cause the “shut down the economy” and “social distance” policies.

This study advances our understanding in three significant dimensions. Initially, it establishes a comprehensive framework for the evaluation of technical efficiency (TE) in supporting industries by integrating firm-level characteristics—including human capital quality, capital intensity, and firm size—with external spillover factors, including domestic demand, FDI backward linkages, sectoral competitiveness, and the investment environment. Secondly, it ensures the robustness and reliability of the empirical findings by utilizing sophisticated econometric techniques, such as stochastic frontier analysis (SFA), Fixed Effect Models, and System-GMM estimators, to improve methodological rigor in efficiency analysis. Third, this study contributes new insights into the ways in which digitalization and institutional change influence firm efficiency by situating the analysis within the context of Vietnam’s ongoing digital transformation and policy reforms. Consequently, the discourse on industrial upgrading in emerging economies is expanded.

Because of the contradictions in the results of previous studies, as well as the wish to contribute new policy-making recommendations for the government, we are motivated to perform this study to determine what factors really affect the efficiency of the economy of supporting industry enterprises in the Vietnamese economy. In addition, we aim to answer the question, when the economic context is drastically changed by a new world context and digital transformation, will the impact of those factors on the performance of enterprises in this industry be changed? Furthermore, the world economic context is undergoing many major changes, such as the Ukraine–Russia war, the strict tariff policy implemented by recent presidential administrations with record high import tariffs, and the strong digital transformation and a series of reforms within the Vietnamese economy, so our analysis results are especially useful for Vietnamese economic policy makers as well as foreign investors and company managers in the new context. However, up to now, there has still been no specific indicator to measure the digitalization of the Vietnamese economy, although it has been making efforts to implement many reforms. Our article only begins to examine whether there are any significant differences between the context before the government announced the start of digital transformation and after that by examining sub-samples. In the future, when we have access to data measuring the level of digital transformation, we want to overcome and expand this research with more measurement variables. This research is also expected to be the foundation for further expansion and implementation of more elaborate works.

The research framework is shown below:

2. Materials and Methods

2.1. Determinants of Firm Performance

There are various methods to determine a firm’s performance, such as Fundamental analysis, Warranty analysis, and so on. Firm performance is traditionally evaluated by using the production frontier: Firms located on the frontier are efficient firms, while those below the frontier are considered inefficient. Aigner et al. (1977) [3] and Meeusen and Broeck (1977) [4], when estimating the production frontier, considered the effect of random errors. The production frontier thus formed is the optimal production curve. Therefore, the technical efficiency of each business is a factor considered to reflect the ability to improve a business’s productivity towards the optimal state. In this research, technical efficiency is used to analyze firm performance and explore key factors affecting it to propose sound solutions. Farrell (1957) [5] defines productive efficiency as the ability of producers to efficiently use available resources to produce maximum output at minimum cost. He refers to technical efficiency (TE), which is estimated from the production function, as the performance of the producer to avoid waste of inputs in production.

Recent studies analyzing the technical efficiency of businesses often use two popular methods: SFA (stochastic frontier analysis) and DEA (data envelopment analysis). DEA is a non-parametric estimation that calculates the technical efficiency of businesses according to the distance function. The advantage of this method is that there is no need to define the function format, but the disadvantage is that it is sensitive to outliers, which leads to bias in determining the production frontier and a lack of statistical tests. SFA overcomes the limitations of DEA. SFA is a parametric method for estimating the production frontier, taking into account the random error of the most efficient firms. Many studies apply this approach to assess technical efficiency and the affecting factors, such as Batra and Tan (2003) [6], Ismail and Abidin (2014) [7], and H. T. Nguyễn (2019) [8]. However, we find contradictions in the results of the empirical studies.

There have been many studies on factors affecting the TE of enterprises in general, for all types of enterprises. These factors include firm age, firm size, participation in exporting, human capital, capital intensity, investment environment, location, firm ownership, capital structure, support from the government, competitiveness of domestic enterprises, and so on.

In this study, the authors focus on analyzing some factors that are considered important and have a major impact on the technical efficiency of domestic enterprises in the supporting industry in Vietnam (based on To and Nguyen, 2021 [9]). The authors chose the SFA model to reassess the factors in an inherited manner. This will be presented in the next section to form our research hypotheses.

2.2. Research Hypotheses

Recently, determinants of the development of supporting industry have also interested many researchers (To & Nguyen, 2021 [9], Lien, D.Q2019 [10]), although these studies have been limited within one or some regions in Vietnam, or they only focus on manufacturing firms (Mattsson, P. et al., 2020 [11]; Nguyen, C. et al., 2021 [12], Le, V. et al., 2018 [13]).

For this study, we chose technical efficiency analysis to fit with the features of the supporting industry, which focuses on products in the supply chain for the main industry; this analysis is especially suitable for companies in the processing, manufacturing, and construction industries. Estimating technical efficiency (TE) is the foundation for examining the overall efficiency of firms (Vu, 2016 [14]). By reviewing previous studies (To & Nguyen, 2021 [9], Batra & Tan, 2003 [6]; Ismail & Abidin, 2014 [7]; H.T. Nguyễn, 2019 [8]), we built the foundation for research questions and hypotheses for the determinants of the TE of supporting firms, focused on the six following key factors.

2.2.1. Firm Size

The relationship between firm size and technical efficiency has remained a controversial issue. From both empirical and theoretical perspectives, their relationship is unclear. Some researchers advocate promoting and supporting small businesses based on economic and welfare arguments. For example, it is argued that expanding the small business segment will lead to more efficient resource allocation, less unequal income distribution, and less under-employment because small businesses tend to be more labor-intensive. Employees of small firms may be more motivated by competition-based incentive schemes than by financial schemes, which may make small firms more efficient. Firm size is shown to be a statistically significant determinant of firm technical efficiency and is negatively correlated with the non-technical efficiency effect of the firm (Chapelle and Plane, 2005 [15]).

Lundvall and Battese (2000) [16] found that the relationship between firm size and technical efficiency is not consistent. Yang and Chen (2007) [17] compared the technical efficiency of SMEs with that of large firms and studied the factors affecting technical efficiency in the electronics industry in Taiwan. They found that the average technical efficiency of large firms is higher than that of SMEs, but it is lower when considering the endogenous selection of firm size. Charoenrat et al. (2013) [18] examined determinants of the technical efficiency of small- and medium-sized companies (SMEs) in the manufacturing and assembly industry in Thailand; Kim et al. (2012) [19] studied some Malaysian manufacturing industries in the 2000–2004 period; and other researchers, such as Assefa Admassie and Matambalya (2002) [20], Lundvall and Battese (2000) [16], Le and Harvie (2010) [21], Bhandari and Ray (2012) [22], shared the same point of view when stating that the relationship between firm size and technical efficiency is positive, i.e., the larger the firm size, the higher the technical efficiency. Large-scale factories often have higher technical progress as well as greater technical efficiency. Meanwhile, by using cross-sectional data of 396 companies in the manufacturing and assembly sector in Kenya from 2007 and the SFA method, Cheruiyot (2017) [23] showed that large firm size has a negative impact on technical efficiency. Similarly, Margono and Sharma (2006) [24] found a negative impact of firm size on technical efficiency with a sample of Indonesian textile companies in the 1993–2000 period. These researchers explain that large-scale enterprises may have a complex organizational structure, resulting in a longer and more rigid decision-making process than small businesses.

Amornkitvikai et al. (2014) [25] used stochastic frontier and data envelopment analysis (DEA) to analyze the impact models of technical inefficiency on manufacturing SMEs in Thailand and found that Thai manufacturing SMEs experienced decreasing returns to scale despite their relatively high technical efficiency in production. Their results obtained using both approaches also showed that firm age, medium versus small size, firm location in Bangkok, foreign investment, and government support were significantly and positively associated with technical efficiency. Le and Harvie (2010) [21] assessed technical efficiency and identified the determinants of technical efficiency of non-state-owned domestic SMEs in Vietnam. Their results showed that SMEs in the manufacturing sector in Vietnam have relatively high average technical efficiency, ranging from 84.2% to 92.5%. They explained that small businesses are more efficient due to their flexibility in diversifying and adjusting their business operations in a rapidly changing transition economy.

Alvarez and Crespi (2003) [26], on the other hand, found that larger firms perform better than small firms because small firms may face the following constraints: (i) difficulty in accessing external loans for investment, (ii) lack of efficient resources (e.g., human capital), (iii) lack of economies of scale, and (iv) lack of formal contracts with customers and suppliers. Similarly, Harvie (2002) [27] also mentioned that there are five main constraints that hinder the growth of small and medium enterprises, namely, (i) access to markets, (ii) access to technology, (iii) access to human resources, (iv) access to finance, and (v) access to information.

The research question here is as follows: for firms in Vietnam’s supporting industry, what is the impact of firm size on technical efficiency? Because, previously, there were many contradictory arguments and empirical results, we propose the first research hypothesis, H₁, with a positive impact to test. In the case that hypothesis H₁ is rejected, that can be understood as a negative impact of firm size on technical efficiency.

H₁: 

Firm size has a positive impact on the technical efficiency (TE) of firms in Vietnam’s supporting industry.

2.2.2. Market Size

Not only firm size but also market size is a factor considered to affect the TE of firms, as shown by previous studies (Perelman, 1995 [28]; Gumbau-Albert, 2002 [29]). According to Ohno (2007) [30], market size plays a decisive role in the existence and development of supporting industries because these industries must have a large enough number of orders to be able to participate in the market. In addition, compared to assembly industries, technology in supporting industries is often more capital-intensive. Some supporting industries, such as casting and stamping, require a lot of expensive machinery—production equipment cannot be divided into many parts and, therefore, businesses in the industry must make efforts to reduce unit capital costs by increasing output. Therefore, it is necessary to have a large demand for supporting industry products.

In Vietnam, among the few studies on the same topic, Huyen (2018) [31] demonstrated the positive impact of market size variables—reflected in provincial GDP and enterprise import–export indicators—on the increase in revenue of the electronics supporting industry. Nhan (2019) [32] also confirmed the positive impact of market capacity on the development of supporting industry in Bac Ninh province. In this study, we also consider market size reflected through the BFSpill index, which shows the demand of FDI companies in the upper industry that use products of supporting industry enterprises. Domestic market demand for supporting industry products (BSpill_ratio index) is also included as a factor for analysis and evaluation. These two indices reflect the market demand of the supporting industry, i.e., the market size factor.

We propose the second research hypothesis as follows:

H₂: 

Market size (BFSpill and BSpill_ratio) has a positive impact on the technical efficiency (TE) of firms in Vietnam’s supporting industry.

2.2.3. Quality of Human Resources

Human resources always play an important role in the development of enterprises, in general, and supporting industry firms, in particular (Gumbau-Albert, 2002 [29]; Crespi & Alvarez, 2003 [26]; Charoenrat & Harvie, 2014 [33]). The supporting industry requires skilled workers for the effective use of machinery capacity (Ohno, 2007 [30]). In addition, due to their features as small and medium businesses, the quality of human resources is even more important in improving the efficiency of business operations. Huyen (2018) [31] evaluated the quality of human resources on the revenue of supporting industry firms in Vietnam; however, there is no significant evidence for this.

We propose the third research hypothesis as follows:

H₃: 

The quality of human resources has a positive impact on the technical efficiency (TE) of firms in Vietnam’s supporting industry.

2.2.4. Capital Intensity

Capital and labor are the two most basic factors of production. Technical efficiency refers to the ability to combine inputs, such as labor and capital, at a given state of technology to produce given levels of output. Therefore, capital intensity represents the ability of a business to provide capital to its workers, which is also a factor that researchers, such as Gumbau-Albert (2002) [29], Batra and Tan (2003) [6], and Ismail and Abidin (2014) [7], have confirmed to have an impact on the TE of a firm. In developing countries like Vietnam, capital investment plays a major role in changing the technical efficiency of enterprises. In particular, investment in fixed assets and new assets, replacing old technology with new technology to change production capacity, improves production efficiency (Vu, 2016 [14]; Sinani et al., 2008 [34]). This is especially meaningful in the context of digitalization of the economy, requiring enterprises to invest in new machinery and technology to carry out digital transformation in their own operations.

So, we propose the fourth research hypothesis as follows:

H₄: 

Capital intensity has a positive impact on the technical efficiency (TE) of firms in Vietnam’s supporting industry.

2.2.5. Investment Environment

This factor includes a stable policy environment, preferential policies, and narrowing the gap in awareness and information between domestic suppliers and foreign enterprises. These are factors that ensure the comprehensive development of supporting industries (Ohno, 2007 [30]). Huyen (2018) [31] affirmed that the system of strategies, policies, and information systems are factors that create a favorable business environment for supporting industries in the Vietnamese electronics industry. Information connection and tax policies are also confirmed to have a positive impact on the development of supporting industries in some provinces.

In this study, the authors assess the institutional environment through two indicators, that is, “fair competition” and “the level of corruption”, both of which are extracted from the Provincial Competitiveness Index (PCI) of Vietnam. A good investment environment has highly fair competition and a low level of corruption. Furthermore, the location of a business is also considered through the dummy variable “region”. The location factor contributes to creating differences in the investment environment and competitive advantages (being close to main traffic routes and seaports will create competitive advantages in terms of costs, having a good relationship with the government, being supported by the local government policies, and so on). So, location can have an impact on TE (Nguyen, 2019 [35]; Charoenrat & Charles, 2014 [33]).

We propose the fifth research hypothesis as follows:

H₅: 

A good investment environment has a positive impact on the technical efficiency (TE) of firms in Vietnam’s supporting industry.

2.2.6. Linkage Between Domestic Supporting Industry Suppliers and FDI Assembly Enterprises

This linkage plays an important role, not only by providing a large demand for domestic supporting industry firms but also by contributing to improving the operational efficiency and technological level of domestic companies. One study (Ohno, 2007 [30]) pointed out that narrowing the gap in information and awareness between supporting industry enterprises and FDI assembly enterprises is one of the factors that contribute to promoting the development of the supporting industry. Linkages with FDI enterprises also create spillover effects on the technical efficiency of enterprises in general (Newman et.al, 2015 [36]; Sari et.al, 2016 [37]; Sur & Nandy, 2018 [38]). Through the HFSpill index, we measure the impact of this factor.

We propose the sixth research hypothesis as follows:

H₆: 

A strong linkage between domestic supporting industry suppliers and FDI assembly enterprises (through the HFSpill index) has a positive impact on the technical efficiency (TE) of firms in Vietnam’s supporting industry.

In addition to the above factors, the digital transformation of Vietnam’s economy is also a factor included in this analysis. Therefore, in the next parts, experimental models are tested to identify what determinants really have impacts on Vietnamese supporting industry development.

2.3. Research Functions

Based on theoretical studies, the production function can be described as follows:

$$y=ln{SL}_i={\beta }_0+{\beta }_1{X}_i+v_i-u_i$$

(1)

where ${SL}_i$ is the output of firm i, $X_i$ is the vector of explanatory variables used in the model, $v_t$ is the random error (noise effect), and $u_t$ is the error representing for inefficiency effect.

The assumptions are as follows:

-

$v_i$ is assumed to be independently and identically distributed as N (0, ${\sigma }_v^2$);

-

$u_i$ is assumed to be distributed independently of $v_i$ and to satisfy u_i ≦ 0;

-

$u_i$ is derived from a N (0,${\sigma }_u^2$) distribution truncated above at zero;

-

u and v have no correlation with X.

SFA is a parametric method for estimating the production frontier, considering the random error of the most efficient enterprises. Aigner et al. (1977) [3] and Meeusen and Broeck (1977) [4] independently published this method. The estimation of the stochastic frontier represents a well-established empirical tool, widely employed in many research studies. That is why we use this method in the context of Vietnam.

The production frontier thus formed is the optimal production line; the technical efficiency of each enterprise is, accordingly, the factor considered to assess an enterprise’s ability to improve productivity towards the optimal state. The stochastic production frontier defines the maximum feasible output in an environment characterized by the presence of either favorable or unfavorable events beyond the control of the producers ($v_i$). The error term $u_i$ > 0 is introduced in the model in order to capture the shortfall of Yi, that is, technical inefficiency (Luca & Claudi, 2014 [39]).

After estimating, technical efficiency (TE) is defined as follows:

$${TE}_i={\displaystyle \frac{{SL}_i}{\mathrm{e}\mathrm{x}\mathrm{p}(X_i^{\prime }\beta +v_i)}}={\displaystyle \frac{\mathrm{e}\mathrm{x}\mathrm{p}(X_i^{\prime }\beta +v_i-u_i)}{\mathrm{e}\mathrm{x}\mathrm{p}(X_i^{\prime }\beta +v_i)}}=\mathrm{e}\mathrm{x}\mathrm{p}(-u_i)$$

(2)

SFA is a parametric approach that assumes the production function has the Translog form or Cobb–Douglas form; then, it estimates the coefficient of parameters such as capital, labor, and raw materials (Battese & Coelli, 1995 [40]). SFA is a two-stage estimation method. In the first stage, the technical inefficiency component is decomposed from the error terms. In the second stage, technical inefficiency is the dependent variable, and we run regression models for some independent variables. Maximum likelihood estimation, suggested by Battese and Coelli (1995) [40], is commonly used in the second stage. Some studies adopt different estimation techniques, such as Generalized Methods of Moments (GMM), Fixed Effect Models, and Random Effect Models (Mattsson et al., 2020 [11]; Otsuka & Natsuda [41], 2016; Söderbom & Teal, 2004 [42]). Because of its advantages, we use the SFA method to consider the effects of three groups of factors on the technical performance of supporting industrial businesses: (i) internal factors, such as human resource quality and capital intensity; (ii) external factors, such as the impact of FDI enterprises on the same industry, FDI backward effects, domestic demand of the industry, and the institutional environment, and (iii) firm-specific factors, such as region, firm size, and minor industries.

Either the Cobb–Douglas function or the Translog function is chosen to estimate technical efficiency (TE).

Cobb–Douglas function: $ln{SL}_{it}={\beta }_0+{\beta }_{1}ln{V}_{it}+{\beta }_{2}ln{LD}_{it}+v_{it}-u_{it}$

Translog function:

$$\begin{gathered} ln{SL}_{it}={\beta }_0+{\beta }_{1}ln{V}_{it}+{\beta }_{2}ln{LD}_{it}+{\beta }_3{\left(ln{V}_{it}\right)}^2+{\beta }_4{\left(ln{LD}_{it}\right)}^2+{\beta }_5\left(ln{V}_{it}\right)\left(ln{LD}_{it}\right)+{\beta }_{6}t \\ +{\beta }_7{\left(t\right)}^2+{\beta }_{8}t\left(ln{V}_{it}\right)+{\beta }_{9}t\left(ln{LD}_{it}\right)+v_{it}-u_{it} \end{gathered}$$

2.4. Research Data

The data used in this study is a panel dataset connected from the data of the following enterprise surveys: (i) Economic Census 2014–2022, (ii) Enterprise Census (executed by the General Statistics Office), focusing on supporting businesses; (iii) E-commerce index report data of Vietnam E-commerce Association (VECOM); and (iv) PCI Provincial Competitiveness Index Survey (conducted by VCCI).

The extraction of survey responses from businesses (Economic Census and Enterprise Census) to collect data for variables is the hardest and most difficult stage because the General Statistics Office obtained a large amount of information. We need to extract those answers to record reliable data, which is the contribution of this research.

Table 1. Determinants of the technical efficiency of firms.

Variables	Content
Endogenous factors in a firm:
lnSize	Firm size, measured by the natural logarithm of the number of laborers.
lnHum	Human resources, measured by the natural logarithm of the average cost per employee of a firm.
lnDC	Capital intensity, measured by the natural logarithm of the average asset value per employee of a firm.
State	Dummy variable; = 1 in the case that the firm has relationship with the government, officials, ..., and vice versa.
Region	Dummy variable; takes values from 1 to 6. (1. Red River Delta region; 2. Northern Midlands and Mountains region; 3. North Central Region and Central Coast; 4. Central Highlands region; 5. Southeast region; 6. Mekong Delta region of Vietnam.)
Supind	Dummy variable; takes values from 1 to 6. 1. Textile industry; 2. eather and footwear industry; 3. electronics industry; 4. automobile manufacturing and assembly industry; 5. mechanical manufacturing industry; 6. high-tech supporting industry.
Exogenous environmental factors:
BSpill_ratio	Domestic demand of industry i, which shows the impact of all upstream businesses using inputs that are products of industry j; measured by: $${BSpill\_ratio}_{jt}={\displaystyle \frac{\sum _k{b}_{kl}x{L}_{jt}}{\sum _{i\in j}{L}_{it}}}$$
HFSpill	Impact of FDI enterprises in the same industry, measured by: $${HFSpill}_{jt}={\displaystyle \frac{\sum _{i\in j}{FSh}_{it}x{L}_{it}}{\sum _{i\in j}{L}_{it}}}$$ where ${FSh}_{it}$ is foreign capital ratio of firm i, in year t; j is one of six sub-sectors.
BFSpill	FDI backward effect, measured by: $${BFSpill}_{jt}=\sum _k{b}_{kl}x{HSpill}_{jt}$$ where ${BFSpill}_{jt}$ is the spillover effect of FDI enterprises on supporting industry sub-sector j. Upstream FDI enterprises use inputs from supporting industry firms in industry j. $b_{kl}$ is the coefficient indicating when industry k increases by 1 unit of product, industry l needs to increase by how many units of product; this coefficient is calculated from the Vietnam input–output table.
Competition	To reflect the institutional environment, we use the index of “fair competition”, extracted from the Provincial Competitiveness Index (PCI).
Informal	This variable is an index representing the institutional environment, which reflects the level of corruption, extracted from the Provincial Competitiveness Index (PCI).

Source: Authors.

below provides a summary of the determinants considered in this study. We use the symbol “ln” when naming variables that are measured by the natural logarithm.

Regression models with technical efficiency (TE) as the dependent variable and determinants as independent variables are estimated by four methods (Pooled-OLS, Fixed Effect Models, Random Effect Models, and System-GMM to solve for endogeneity), with a full sample from 2014 to 2022 and two sub-samples (2014–2019, 2020–2022).

Vietnam is still a small economy; it has just started to implement digital transformation and has no indicators to measure the level of implementation. When the COVID-19 pandemic appeared, the Vietnamese government, as well as companies, recognized the necessity of digitalizing the economy; then, it pushed for more digitalization to survive. This transformation can be realized visibly through process digitization, digitized data, investment in developing information technology and communications infrastructure, and so on. The year 2020 recorded many pieces of evidence from reality that Vietnam’s economy is strongly transforming digitally through a lot of new policies from the government. Furthermore, a survey of 400 enterprises investigating “The current status of digital transformation in enterprises in the context of the COVID-19 pandemic”, conducted by the Vietnam Federation of Commerce and Industry (VCCI), shows that since the year 2020, Vietnamese enterprises have begun to recognize and apply digital technologies in stages such as internal management, purchasing, logistics, production, marketing, sales, and payment. In the field of internal management, cloud computing is the most used technical tool by Vietnamese enterprises, at 60.6%, an increase of 19.5% compared to the time before the COVID-19 pandemic; about 30% of businesses have used online work and process management systems, better than in the year 2019. Therefore, the authors chose the year 2020 as the benchmark to divide the data into two sub-samples, namely, before and after the start of the digital transformation in the Vietnamese economy, to determine whether digitalization makes any difference in the impacts of these factors.

Supind is a dummy variable, ranging in values from 1 to 6, that is used to divide and evaluate 7 sub-sectors of the supporting industry. The number of firms falling under each Supind category is as follows:

Supind	Number of Unique Firms
1	767
2	61
3	697
4	494
5	3047
6	247
Total	5313

3. Results

3.1. Test to Choose Appropriate Production Function

Table 2. Estimation results of the technical efficiency of Vietnam supporting industry firms.

Variables	Cobb–Douglas	Translog
Ln(V)	0.9800 ***	−0.0236 ***
	(0.0012)	(0.0075)
Ln(LD)	0.0166 ***	1.5359 ***
	(0.0048)	(0.0168)
T		−0.1208 ***
		(0.0145)
Ln(V)^2		0.0119 ***
		(0.0004)
Ln(LD)^2		−0.0271 ***
		(0.0019)
t^2		0.0227 ***
		(0.0017)
Ln(V) x Ln(LD)		−0.0180 ***
		(0.0012)
t x Ln(V)		0.0715 ***
		(0.0014)
t x Ln(LD)		−0.1282 ***
		(0.0025)
C	1.4785 ***	5.0913 ***
	(0.0732)	(0.0530)
LR Test
Function	Log Likelihood	p-value
Cobb–Douglas	−136,800
Translog	−123,341	2.2 × 10−16 ***

Note *** is statistically significant at the 1% levels. Values in parentheses (...) are standard errors.

shows the results of our test that is used to choose the appropriate production function, either the Cobb–Douglas model or the Translog model, to estimate technical efficiency (TE), as well as to determine whether technical efficiency or inefficiency exists. The stochastic frontier production function and Maximum Likelihood Equation methods are applied for testing.

In the Cobb–Douglas model, the labor, capital, and cost variables all have a positive effect on the output at the 1% significance level. Meanwhile, in the Translog model, the cost variables have a negative effect on output at the 1% significance level. In addition, the Translog model has a higher Log Likelihood index. The very small p-value (p < 2.2 × 10⁻¹⁶) shows that the Translog model has better statistical significance.

Thus, the Translog production function model is chosen as the research model to estimate TE. Our research model is defined as:

$$\begin{array}{l}ln{SL}_{it}={\beta }_0+{\beta }_{1}ln{V}_{it}+{\beta }_{2}ln{LD}_{it}+{\beta }_3{\left(ln{V}_{it}\right)}^2+{\beta }_4{\left(ln{LD}_{it}\right)}^2+\\ {\beta }_5\left(ln{V}_{it}\right)\left(ln{LD}_{it}\right)+{\beta }_{6}t+{\beta }_7{\left(t\right)}^2+{\beta }_{8}t\left(ln{V}_{it}\right)+{\beta }_{9}t\left(ln{LD}_{it}\right)+v_{it}-u_{it}\\ \hspace{1em}{\mathit{u}}_{\mathit{i}\mathit{t}}={\mathit{\delta }}_0+\mathit{\delta }{\mathit{C}\mathit{O}\mathit{N}}_{\mathit{i}\mathit{t}}+\mathrm{}{\mathit{\omega }}_{\mathit{i}\mathit{t}}\end{array}$$

where ${CON}_{it}$ includes variables considered as having an important impact on technical efficiency and ω_it is the error term of the inefficiency model assumed as a truncated normal distribution. ${CON}_{it}$ includes two groups, namely, endogenous and exogenous determinants, as listed in Table 2.

The estimation was controlled by industry and by time to ensure the results. The descriptive statistics are shown in

Table 3. Descriptive statistics of the full sample.

Variables	Min	Max	Mean	Std. Deviation
lnSL	−2.3026	32.0279	11.4886	5.9875
lnV	−2.3026	30.3973	10.2536	5.8733
lnLD	−0.6931	10.0420	2.8725	1.5997
lnSize	0	10.2846	2.8868	1.6110
lnHum	−2.9957	22.2640	6.6776	5.4313
lnDC	−0.2154	27.1610	9.2007	5.5567
Bspill_ratio	0.0526	0.4358	0.1959	0.0572
HFSpill	0	44,776.08	889.0047	4153.076
BFSpill	0	4,118,457.2	444,375.3	564,024.1
informal	2.8100	8.39	5.6501	1.0648
competition	3.1200	8.81	5.1700	1.0977

Source: Authors. Notes: obs = 59,370, lnSL: logarithm of firm output; lnV: logarithm of a firm’s total assets; lnLD: logarithm of a firm’s number of employees.

and

Table 4. Correlation matrix of variables.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
(1) lnSize	1.000
(2) lnHum	0.032 ***	1.000
(3) lnDC	−0.021 ***	0.980 ***	1.000
(4) state	−0.221 ***	0.279 ***	0.280 ***	1.000
(5) Bspill_ratio	−0.244 ***	−0.043 ***	−0.030 ***	0.075 ***	1.000
(6) informal	0.051 ***	0.583 ***	0.577 ***	0.275 ***	−0.021 ***	1.000
(7) competition	0.055 ***	0.406 ***	0.403 ***	0.246 ***	−0.011 ***	0.808 ***	1.000
(8) Hfspill	0.028 ***	−0.095 ***	−0.094 ***	−0.064 ***	−0.259 ***	−0.147 ***	−0.141 ***	1.000
(9) Bfspill	−0.094 ***	−0.343 ***	−0.344 ***	−0.132 ***	0.274 ***	−0.394 ***	−0.370 ***	0.208 ***	1.000
(10) supind	−0.222 ***	−0.059 ***	−0.051 ***	0.017 ***	0.011 ***	−0.073 ***	−0.077 ***	0.063 ***	0.196 ***	1.000
(11) region	−0.013 ***	−0.005	−0.013 ***	0.010 **	0.074 ***	0.069 ***	0.139 ***	−0.006	0.014 ***	−0.125 ***	1.000
(12) TE	0.037 ***	0.073 ***	0.072 ***	−0.019 ***	0.008 *	−0.053 ***	−0.074 ***	0.013 ***	0.071 ***	0.036 ***	0.029 ***	1.000

Significance levels: * p < 0.1, ** p < 0.05, *** p < 0.01. Source: Authors. Based on the above matrix, except for the variable “lnDC”, which may have the problem of multicollinearity with the variable “lnHum” (0.980 ***), resulting in confused conclusions, the other variables show a fairly reasonable relationship with each other, but not quite like the assumptions.

.

3.2. Determinants of TE: Analysis by Pooled OLS, FEM, and REM

The regression models with technical efficiency (TE) as the dependent variable and determinants as the independent variables, estimated by three methods (pooled OLS, Fixed Effect Model, Random Effect Model), are shown in

Table 5. Regression of the full sample (2014–2022).

VARIABLES		Pooled OLS		FEM		REM
		Coefficient	p-Value	Coefficient	p-Value	Coefficient	p-Value
lnSize		0.0042 ***	0.000	0.0050 ***	0.000	0.0022 ***	0.000
lnHum		0.0006	0.156	0.0292 ***	0.000	0.0025 ***	0.000
lnDC		0.0026 ***	0.000	0.0096 ***	0.000	0.0022 ***	0.000
Bspill_ratio		2.00 × 10−1 **	0.024	5.97 × 10−1 ***	0.000	3.50 × 10−2	0.653
HFSpill		0.0000 *	0.056	0.0000 **	0.050	0.0000 ***	0.006
BFSpill		1.61 × 10−8 ***	0.000	7.93 × 10−9 ***	0.000	1.70 × 10−8 ***	0.000
informal		−5.01 × 10−3 ***	0.000	−2.68 × 10−3 ***	0.006	−5.48 × 10−3 ***	0.000
competition		−0.0072 ***	0.000	−0.0008	0.301	−0.0081 ***	0.000
State	1	−0.0031***	0.007	0.0073 ***	0.000	−0.0038 ***	0.000
Region	2	−0.0140 ***	0.000	−0.0075 ***	0.000	−0.0110 ***	0.000
	3	−0.0198 ***	0.000	−0.0109 ***	0.000	−0.0153 ***	0.000
	4	−0.0008	0.859	0.0015	0.734	0.0016	0.684
	5	0.0148 ***	0.000	0.0046 ***	0.000	0.0141 ***	0.000
	6	−0.0011	0.593	0.0012	0.588	0.0031	0.105
Supind	2	−0.0612 ***	0.002	−0.1453 ***	0.000	−0.0252	0.148
	3	0.0270 ***	0.007	0.0704 ***	0.000	0.0088	0.317
	4	0.0325 ***	0.005	0.0799 ***	0.000	0.0110	0.276
	5	0.0144 ***	0.000	0.0107 ***	0.000	0.0124 ***	0.000
	6	0.0478 ***	0.001	0.1048 ***	0.000	0.0209 *	0.089
cons		0.4309	0.000	0.0390	0.061	0.4857	0.000
Prob > F			0.000		0.000		0.000
R-squared			0.0380		0.0564		0.0187

Notes: ***, **, and * show statistically significant results at the levels of 1%, 5%, and 10%. Source: Authors.

for the full sample from 2014 to 2022.

In all three estimations, the correlation coefficients are consistent in sign. The tests conducted to select the most suitable model show that FEM is the most suitable one, as presented in

Table 6. Tests conducted to choose a suitable model.

Hausman test REM FEM Test of H0: Difference in coefficients not systematic      b = Consistent under H0 and Ha; obtained from xtreg.      B = Inconsistent under Ha, efficient under H0; obtained from xtreg.  chi2(17) = (b-B)'[(V_b-V_B)^(-1)](b-B)       = 11,376.42 Prob > chi2 =  0.0000 (V_b-V_B is not positive definite)	Prob > chi2 = 0.00 < α = 0.05; so, reject H0 and conclude that the model chosen in the study is the FEM model.
Breusch and Pagan Lagrange Multiplier Test for OLS REM Breusch and Pagan Lagrangian multiplier test for random effects   te[t,t] = Xb + u[t] + e[t,t]   Estimated results:           |   Var   SD = sqrt(Var)       ---------+-----------------------------          te |  0.0132804   0.1152405           e |   0.0117014   0.1081731           u |      0      0   Test: Var(u) = 0              chibar2(01) =    0.00            Prob > chibar2 =   1.0000	Prob > chibar2 = 1.0000 > significance level 5%, α = 0.05; so, accept H0 and conclude that the selected model is OLS.
F test for OLS FEM F test that all u_i = 0: F(8, 59,342) = 684.38       Prob > F = 0.0000	Prob > F = 0.0000 < 5% significance level, α = 0.05; so, reject H0 and conclude that the selected model is FEM.

Source: Authors.

.

The results of the FEM show the following:

Quality of human resources (lnHum) and capital intensity (lnDC) both have positive effects on TE at a significance level of 1%. Our results are similar to the findings of Yang et al. (2010) [43], Chaffai et al. (2012) [44], Charoenrat and Harvie (2014) [33], Cheruiyot (2017) [23], and Kashiwagi and Iwasaki (2020) [45] on the influence of skilled labor on technical efficiency in the production of firms. Furthermore, this result implies the importance of improving workers’ skills and training human resources, thereby increasing production efficiency.

In addition, firm size (lnSize) has a positive impact on the TE of supporting industry firms in Vietnam at the 1% significance level. This implies that large-scale enterprises can take advantage of organizational strength and technological requirements to achieve larger production levels. At the same time, larger-scale enterprises have higher financial capacity and modern management skills, so they can better handle difficulties, thereby achieving higher technical efficiency.

For the investment environment, we find that the location of a business also affects its technical efficiency. The TE of supporting industry firms in the Southeast region of Vietnam is better than the ones in the Red River Delta region; however, the Red River Delta region shows better performance than at Northerns Midlands and Mountains region (region No.2), as well as the North Central and Central Coast Region (region No.3). This is explained by the fact that, in a comparison to the mountainous and central regions of Vietnam, plains with flat terrain near the sea are more favorable for factories to transport materials and products, saving more costs. Different sectors exhibit different levels of TE, all at the 1% significance level. The TE of the No. (3), (4), (5), and (6) sectors is better than the textile industry’s TE. The leather and footwear industry’s TE is the worst, although the textile and leather and footwear industries were previously the two main export industries of Vietnam. These results also show the shift in the manufacturing industry structure in Vietnam to adapt to the new economic context in the era of the “4.0 Industrial Revolution”. With a negative correlation of the variable “informal”, our results also show that a poor investment environment will inhibit TE, while having relationships with the government can help a firm to perform better (the variable “state” has a positive impact on TE at the 1% significance level).

Furthermore, for exogenous environmental factors, we find that HFSpill has no impact on TE, while BFSpill (FDI backward effect) and BSpill_ratio (domestic demand of an industry impacting all upstream businesses that use inputs that are products of that industry) all show a positive impact on TE at the 1% significance level. This suggests that when domestic demand for supporting industry products increases, supporting industry manufacturers may increase production to meet the demand. This may lead to increased productivity and optimal utilization of production lines, thereby improving technical efficiency. In addition, increased demand for supporting industry products may lead to competition among supporting industry manufacturers, thereby forcing these manufacturers to improve technical efficiency to reduce production costs and compete in the market.

3.3. Determinants of TE by the System-GMM Model

Endogeneity problems are inevitable for data; therefore, to strengthen the consistency of this study’s results, System-GMM models are applied for the research functions. The results are shown in

Table 7. System-GMM model results for the full sample (2014–2022).

	2014–2022
	TE
L.TE	0.436 ***
	(0.058)
lnSize	0.007 ***
	(0.002)
lnHum	−0.013 ***
	(0.002)
lnDC	0.007 ***
	(0.002)
BSpill_ratio	−0.494 ***
	(0.071)
HFSpill	−0.000 ***
	(0.000)
BFSpill	0.000 ***
	(0.000)
informal	−0.012 ***
	(0.004)
competition	0.029 ***
	(0.004)
Region	0.007 ***
	(0.001)
State	−0.003
	(0.006)
Supind	0.014 ***
	(0.002)
Constant	0.144 ***
	(0.034)
AR(1) test	0.000
AR(2) test	0.317
Sargan test	0.256
Observations	3905

Notes: *** shows statistically significant results at the levels of 1%. Source: Authors.

for the full sample, which are similar to the coefficients obtained using the FEM.

3.4. Digital Transformation of the Vietnamese Economy

To consider whether the impact of factors affecting TE is different between before and after digital transformation, we perform regressions with two sub-samples, using the year 2020 as the benchmark to divide the sub-samples. We use 2020 because it is considered as the time Vietnam’s economy made many significant changes after COVID-19 and shows many activities of digital transformation. The results are presented in

Table 8. Regression TE with sub-sample 1 (2014–2019).

VARIABLES		Pooled OLS		FEM		REM
		Coefficient	p-Value	Coefficient	p-Value	Coefficient	p-Value
lnSize		0.0048 ***	0.000	0.0006 ***	0.004	0.0046 ***	0.000
lnHum		0.0432 ***	0.000	0.0432 ***	0.000	0.0444 ***	0.000
lnDC		0.0345 ***	0.000	0.0345 ***	0.000	0.0351 ***	0.000
Bspill_ratio		−1.41 × 100 ***	0.000	−1.41 × 100 ***	0.000	−1.60 × 10−1	0.601
HFSpill		0.0000	0.746	0.0000	0.746	0.0000	0.287
BFSpill		−1.15 × 10−8 ***	0.000	−1.15 × 10−8 ***	0.000	4.12 × 10−9	0.228
informal		2.36 × 10−3	0.135	2.36 × 10−3	0.135	−2.26 × 10−3	0.165
competition		−0.0042 ***	0.003	−0.0042 ***	0.003	-0.0023	0.110
State	1	0.0128 ***	0.000	0.0128 ***	0.000	0.0132 ***	0.000
Region	2	−0.0076 **	0.041	−0.0076 **	0.041	−0.0067 *	0.071
	3	−0.0043	0.130	−0.0043	0.130	−0.0018	0.516
	4	0.0151 **	0.035	0.0151 **	0.035	0.0121 *	0.090
	5	0.0073 ***	0.000	0.0073 ***	0.000	0.0064 ***	0.000
	6	0.0109 ***	0.005	0.0109 ***	0.005	0.0154 ***	0.000
Supind	2	0.2443 ***	0.000	0.2443 ***	0.000	−0.0050	0.938
	3	−0.1863 ***	0.000	−0.1863 ***	0.000	−0.0285	0.452
	4	−0.1815 ***	0.000	−0.1815 ***	0.000	−0.0146	0.717
	5	0.0037	0.187	0.0037	0.187	0.0047 *	0.098
	6	−0.2267 ***	0.000	−0.2267 ***	0.000	−0.0160	0.758
cons		0.3340	0.000	0.3340	0.000	0.0460	0.521
Prob > F			0.000		0.000		0.000
R-squared			0.1027		0.1056		0.1047

Notes: ***, **, and * show statistically significant results at the levels of 1%, 5%, and 10%. Source: Authors.

and

Table 9. Regression TE with sub-sample 2 (2020–2022).

VARIABLES		Pooled OLS		FEM		REM
		Coefficient	p-Value	Coefficient	p-Value	Coefficient	p-Value
lnSize		0.0163 ***	0.000	0.0163 ***	0.000	0.0081 ***	0.000
lnHum		−0.0334 ***	0.000	−0.0334 ***	0.000	0.0421 ***	0.000
lnDC		0.0350 ***	0.000	0.0350 ***	0.000	0.0562 ***	0.000
Bspill_ratio		−4.28 × 10−1 **	0.030	−4.28 × 10−1 **	0.030	−3.30 × 10−1 *	0.078
HFSpill		−0.0002 ***	0.000	−0.0002 ***	0.000	0.0002 *	0.100
BFSpill		−4.11 × 10−8	0.416	−4.11 × 10−8	0.416	1.76 × 10−8	0.836
informal		1.29 × 10−3	0.594	1.29 × 10−3	0.594	−1.89 × 10−5	0.993
competition		0.0060 ***	0.000	0.0060 ***	0.000	−0.0004	0.824
State	1	−0.0174 ***	0.000	−0.0174 ***	0.000	−0.0207 ***	0.000
Region	2	−0.0346 ***	0.000	−0.0346 ***	0.000	−0.0193 ***	0.000
	3	−0.0624 ***	0.000	−0.0624 ***	0.000	−0.0351 ***	0.000
	4	−0.0174	0.101	−0.0174	0.101	0.0014	0.884
	5	0.0153 ***	0.000	0.0153 ***	0.000	0.0010	0.658
	6	−0.0317 ***	0.000	−0.0317 ***	0.000	−0.0104 **	0.020
Supind	2	0.0383	0.396	0.0383	0.395	0.0162	0.704
	3	−0.0448 **	0.024	−0.0448 **	0.024	−0.0385 **	0.037
	4	−0.0540 **	0.032	−0.0540 **	0.032	−0.0494 **	0.038
	5	0.0153 ***	0.000	0.0153 ***	0.000	0.0161 ***	0.000
	6	−0.0356	0.197	−0.0356	0.197	−0.0236	0.367
cons		0.3455	0.000	0.3455	0.000	−0.7811	0.000
Prob > F			0.000		0.000		0.000
R-squared			0.0999		0.2303		0.1009

Notes: ***, **, and * show statistically significant results at the levels of 1%, 5%, and 10%. Source: Authors.

.

Similar tests as those used for the full sample are performed for both sub-samples. Again, the FEM is the most suitable model. We also apply System-GMM for the two sub-samples, and the results are presented in

Table 10. System-GMM model results for the sub-samples.

	2014–2019	2020–2022
	TE	TE
L.TE	0.261 ***	0.592 **
	(0.055)	(0.274)
ln_size	0.004 *	0.024 ***
	(0.002)	(0.007)
ln_hum	0.032 ***	0.010
	(0.005)	(0.015)
ln_dc	0.015 ***	−0.029 **
	(0.002)	(0.012)
BSpill_ratio	−0.198 ***	−0.020
	(0.077)	(0.380)
HFSpill	−0.000	0.002 *
	(0.000)	(0.001)
BFSpill	0.000 ***	0.000 ***
	(0.000)	(0.000)
informal	−0.002	0.021
	(0.005)	(0.044)
competition	0.038 ***	−0.018
	(0.005)	(0.028)
Region	0.003 *	0.000
	(0.002)	(0.011)
State	0.004	0.002
	(0.006)	(0.053)
supind	0.015 ***	0.003
	(0.002)	(0.016)
Constant	−0.148 ***	0.198
	(0.052)	(0.351)
Observations	3648	257

Notes: ***, **, and * show statistically significant results at the levels of 1%, 5%, and 10%. Source: Authors.

.

The results of the system-GMM are similar to the coefficients obtained using the FEM. Our discussions are based on the FEM results, similar to the approaches of To and Nguyen (2021) [9], Nguyen (2019) [35], and Diallo et al. (2019) [46] when they analyzed the TE of firms.

The results of the FEMs for the first sub-sample (without digitalization) are almost similar to the ones for the full sample. However, for the second sub-sample (with digitalization), there are some surprises. Firstly, firm size still has a positive impact on the TE of supporting industry firms in Vietnam at the 1% significance level, but the impact is stronger in the context of digitalization, with a coefficient of 0.0163. Capital intensity still has a stronger fostering impact on TE, while human resources shows a negative relation at the 1% significance level. The correlation coefficients of Bspill_ratio and BFSpill changed from positive to negative in both sub-samples. HFSpill also changed the sign of its correlation coefficient in the second sub-sample.

In brief, this study’s results confirm all research hypotheses from H1 to H5 for the full sample. Firm size, market size, human resources, and capital intensity all have the effect of promoting TE increase, which means promoting the development of companies in Vietnam’s supporting industry. Location, sectors, competitiveness, and investment environment are also considered determinants that should be considered in management policies at both the firm level and the government level. Our results do not confirm hypothesis H6, which requires further detailed discussions of the aspect of policies.

4. Discussion

After confirming what factors are really determinants of the Vietnamese supporting industry from our key research findings, some recommendations should be discussed:

The quality of human resources was proven to be a determinant for the Vietnamese supporting industry. The higher the quality of labor, the more effective the use of existing technology and the adoption of new technology, which, in turn, leads to higher levels of efficiency (Sinani et al., 2008 [34]). This positive relation also implies that increasing capital investment intensity on labor will increase technical efficiency in Vietnamese supporting industry enterprises. This is in line with the digital transformation context of the world economy and Vietnam, requiring workers to grasp and master the use of new technologies. This finding is similar to those of the studies by Nguyen et al. (2019) [35] and Jorge-Moreno and Carrasco (2015) [47] on the role of capital investment intensity per worker on the production activities of enterprises.

A good investment environment, with support from the government and local authorities, will promote better development of firms in Vietnam’s supporting industry. The institutional environment can provide support and incentives for the enterprises to improve productivity and technical efficiency. This may include the provision of resources, investment in research and development, training of high-quality human resources, and policies that encourage investment and industrial development. The institutional environment also provides a legal and regulatory framework to regulate the operations of firms in the supporting industry. Having a clear and stable legal framework can facilitate enterprises in developing and investing in modern technology. In addition, regulations on intellectual property rights and protection of partners can also create incentives for innovation and technical development.

The linkage between domestic supporting industry suppliers and FDI assembly enterprises plays an important role in improving TE, which is shown in the full sample. Technology spillovers from FDI enterprises can generate some benefits for domestic supporting industry enterprises. They can learn and apply advanced technologies, modern production processes, and high-quality management from FDI enterprises. This can improve the labor productivity and technical efficiency of domestic supporting industry enterprises. Advanced technologies and modern manufacturing processes can help improve product quality, increase productivity, and reduce production costs, helping domestic supporting industry enterprises seize market opportunities and compete with domestic and foreign competitors. However, to maximize the benefits from technology spillover from FDI enterprises, supportive policies and solutions should be issued by the government, such as training high-quality workers, facilitating technology transfer, and creating a favorable business environment for enterprises.

Digital transformation has created some interesting changes in the correlation coefficients, causing the government and firm managers to have strong beliefs that policies to increase capital intensity and human resources are still essential at all times. However, the new era makes productivity increase by applying advanced technology, reducing the number of laborers used, and creating an inverse correlation between TE and human resources. So, “investment” here should be understood in broader terms: investing in improving knowledge, the ability to understand and use new technology quickly, long-life learning, and so on. When firms change the traditional management model to a new one by applying new technologies, such as Big Data, the Internet of Things (IoT), cloud computing, etc., it leads to changes in management methods, leadership, work processes, and corporate culture, helping to reduce human resource costs. Digitalization also promotes innovations in factories: when they use more modern technologies with higher automation, workers’ skills and knowledge increase in accordance with the increase in intellectual development in the digitalized society, etc. These are the reasons why enterprises are able to maintain high TE while reducing investment in cost per employee.

The Southeast region continues to have a better performance than the other regions in the development of supporting industry, while the Central Highlands region lost its better performance in comparison with the Red River Delta region. The explanation lies in the fact that some regions receive more government investment, are more accessible, and can more easily implement the digital transformation process.

The variable “informal” has no statistical significance in either sub-sample, while the variable “competition” becomes statistically significant. Before 2020, bitter competition among companies resulted in a negative effect on TE; however, since 2020, the Vietnamese government has implemented particular policies to help businesses in the supporting industry join the global supply chain, and competition has become their motivation to improve themselves, exhibiting a positive effect on TE. Moreover, companies that rely on their relationships with the state are slow to innovate to keep up with the development of digital transformation, so their performance goes down, exhibiting a negative relation between the variable “state” and TE.

One further result we wish to discuss is that the dependent variable TE seems to have a rather low correlation with the other independent variables (the maximum value of this correlation is only 0.0725). This implies that the old policies related to these factors should still be applied. However, the influence of these factors is not large enough for the Vietnamese government to maintain old policies related to these factors. If policies are built based only on familiar factors, Vietnam’s supporting industry will continue to develop slowly; therefore, it will not keep up with other countries and will not meet the needs of the market. Regarding performance, the ability to expand international markets and build brands of supporting industry enterprises is important, as well as the ability to innovate technology and improve the research and development of enterprises, and so on. However, the research models in the literature review mentioned only a few influential factors.

5. Conclusions

In conclusion, this study provides valuable insights into the determinants of technical efficiency (TE) in Vietnam’s supporting industry within the context of a rapidly digitalizing economy. The findings highlight the significant positive roles of human resource quality, capital intensity, and firm size in enhancing TE. Additionally, the influence of external environmental factors, such as the domestic demand spillover effect (BSpill-ratio) and FDI backward linkages (BFSpill), underscores the critical importance of fostering connections between domestic supporting industries and foreign direct investment (FDI) assembly enterprises.

Vietnam’s digital transformation since 2020 has introduced notable shifts in the dynamics of these factors, suggesting that digitalization may amplify or modify their impacts. This study also emphasizes the importance of geographical location, sectoral characteristics, competitiveness, and investment climate as additional determinants that warrant careful consideration in policy formulation. These findings provide actionable implications for both firm-level strategies and government policies, advocating for a holistic approach to enhancing TE in Vietnam’s supporting industries. By addressing these factors, Vietnam can further capitalize on the opportunities presented by its digitalized economy, strengthen its industrial competitiveness, and foster sustainable economic growth. For the government, it is necessary to continue to improve the policy on developing supporting industries, that is, to review, update, and adjust the list of priority products for development in accordance with reality, such as mechanics, automobiles, textiles, footwear, and electronics. At the same time, a strategy to support the export of key industrial products needs to be issued. State departments should support supporting industry firms in developing domestic and global sustainable value chains and issue tax exemption and reduction policies to attract foreign investment in the supporting industry sector. In addition to incentive policies, there must also be commitments applied to foreign investors to ensure effective investment, promoting industrial transfer between FDI enterprises and domestic ones. Enterprises need to make business connections by actively participating in investment promotion programs and fairs, increasing connectivity between Vietnamese enterprises, and developing specialized clusters and groups of enterprises specializing in products to create high competitiveness. Companies themselves need to innovate technology, apply many scientific and technical advances, improve product quality, increase autonomy in input materials, reduce import dependence, strengthen supply chain connectivity, and so on.

Despite all its efforts, this study still has limitations related to the restricted access to data. In our further research, we will try to overcome these by using new methods, including quantitative and qualitative, to identify new factors that previous studies have not mentioned, as well as try to access another source of data.