The Vietnam Provincial Competitiveness: An Efficiency-based Approach

Abstract

The provincial competitiveness index (PCI) serves as an important measure for the efforts of localities in Vietnam in improving the business environment. However, the methodology of PCI excludes the influence of the traditional conditions on business growth, only measuring the “net” competitiveness at the provincial level. This study is conducted (1) to supplement the approach to measure the relation between the improvement of the business environment and the changes in the business results of the enterprises; at the same time and (2) to pay attention to factors that have been excluded in the PCI methodology. Measuring technical efficiency (TE) by data envelopment analysis, fixed effect model, and random effect model are employed to achieve these research objectives. The findings show that the impact of PCI to business growth has a lag of approximately 1 year. Furthermore, in the TE approach, the business growth indicators are affected not only by the improvement of the business environment created by the local government but also by the basic conditions of that locality. This approach seems to be “fair” across localities, especially those with low starting points. This theoretical approach, however, also needs to be further complemented, especially regarding data, to overcome its potential limitations.

Keywords

DEA efficiency enterprises provincial competitiveness

Introduction

The provincial competitiveness index (PCI) of Vietnam was first introduced by the Vietnam Chamber of Commerce and Industry (VCCI) and US Agency for International Development (USAID), started in 2005 with 42 provinces and cities evaluated. PCI is considered as an index to measure and evaluate the actual quality of economic governance of provinces and cities in Vietnam, thereby promoting the development of the private economic sector.

Over the years, the PCI index has been gradually updated and adjusted to suit the realities of the business environment in Vietnam, including important changes in the sub-indices and their corresponding weights; at the same time, update localities participating in the assessment. From 2013 up to now, the PCI and its sub-indices have been maintained stably with 10 sub-indices and deployed in all 63 provinces and cities nationwide.

From the initial debates on PCI, localities have quickly transformed to respond and actively participate through solutions to improve the competitiveness index, in a variety of ways (Minh, 2013; Tuan, 2017). Improving the quality of economic management, creating a favorable business environment and administrative procedures will bring benefits first of all to businesses and investors. In the long run, these efforts contribute to the growth of localities, even overcoming limitations and disadvantages in terms of geographical location or infrastructure conditions.

So far, the effects of PCI as assessed by VCCI and USAID (2018) include four broad areas:

PCI changes the governance thinking of localities, from focusing on geographical advantages, the development of infrastructure as a prerequisite, to affirming the enhancement of regulatory quality as a driving force for the development of the business sector and investment attraction.

PCI creates a tool for citizens and businesses to effectively monitor the administrative efforts of local governments.

PCI is the driving force for change, reflected in the efforts and initiatives that have been implemented by the provinces.

PCI promotes cooperation and sharing among localities, extending to district and departmental competitiveness assessment initiatives in many localities.

Besides the advantages, the PCI index also always raises issues for discussion. As is often the case, no single metric can satisfy all evaluation purposes. Similarly, for PCI, one of the possible concerns is: Is it appropriate for all localities to be placed in the same ranking, using the uniform scale to measure the efforts of localities that have different starting points?

To answer this question, first of all, it is necessary to start with the PCI methodology that was confirmed from the very beginning by the research team of VCCI and USAID. Among the basic features of the PCI approach, it should be noted that the PCI is constructed by excluding the influence of traditional conditions on the economic growth of a province (geographical location, infrastructure, market size, or human resources) (VCCI & USAID, 2018, p. 13). This approach examines in isolation the impact of business climate practices at the provincial level on growth in Vietnam; hence, focuses on solutions to encourage provincial governments to improve the quality of governance. As such, PCI uses a very clear intent approach.

However, the implementation process is always paralleled by higher requirements. One of the first questions is how does the business environment, as reflected by PCI, contribute to local business growth? In addition, if we consider PCI components such as policies that localities strive to implement, then, in turn, their impact on practice often has a certain lag. On the other hand, efforts to improve the business environment are often associated with a certain authority, with specific “political tenure.” Is the growth of enterprises in period t the result of local government efforts in period (t−n)? Finally, but not all and again, do localities with high socio-economic development backgrounds have certain corresponding advantages?

The issues raised above are the premise of this study to try to link the improvement of the business environment and the performance of enterprises through different indicators, taking into account the lags of the PCI impacts and paying attention to differences in basic conditions of localities. All of these are measured through the efficiency between inputs (including PCI and basic conditions of localities) and outputs (business growth indicators) as an additional approach to the PCI index.

In recent decades, the terms, business environment and competitiveness, are mentioned frequently to reflect the ability of countries to attract domestic and foreign investment. Business environment, as defined by Cherunilam (2015, p. 112), consists of
both the internal and external environments, the term business environment is generally used to refer to the external environment, etc. The external factors, on the other hand, are, by and large, beyond the control of a company. The external or environmental factors such as the economic factors, socio-cultural factors, government and legal factors, demographic factors, geo-physical factors.

These factors, then in turn, affect the competitiveness of a country or region. On the other hand, the term of competitiveness has also been widely used at both the national and economic region levels to reflect a country’s capacity in various aspects, including attracting investment and contributing to its economy growth. In theory and practice, to measure the competitiveness of a country or a locality, there are a number of common and widely used approaches that can be mentioned as follows.

Global competitiveness index (GCI) is as an index to measure the competitiveness of countries. According to this approach, “competitiveness as the set of institutions, policies, and factors that determine the level of productivity of a country” (Schwab, 2017). The World Economic Forum grouped the components of GCI into 12 categories, the pillars of competitiveness including institutions, infrastructure, macroeconomic environment, health and primary education, higher education and training, goods market efficiency, labor market efficiency, financial market development, technological readiness, market size, business sophistication, and innovation. Despite the popularity of the GCI, this index still has certain limitations, including measurement methods and data. Benítez-Márquez et al. (2022) proposed an objective global competitiveness index using the data provided by the World Economic Forum but quantifying to what degree the resulting rankings are associated with those corresponding to the GCI.

Also at the country level, the World Competitiveness Ranking was introduced by the IMD World Competitiveness Yearbook and first published in 1989. This ranking aims to analyze and rank countries according to how they manage their competencies to achieve long-term value creation. World Competitiveness Ranking is based on 334 competitiveness criteria selected as a result of comprehensive research using economic literature, international, national, and regional sources and feedback from the business community, government agencies, and academics (IMD World Competitiveness Center, 2022).

In another approach, Djankov et al. (2004) presented the ease of doing business score and the ease of doing business ranking which is based on the ease of doing business score. According to this approach, the ease of doing business ranking compares economies with one another while the ease of doing business scores benchmark economies with respect to regulatory best practice, showing the proximity to the best regulatory performance on each indicator. The ease of doing business score measures an economy’s performance with respect to a measure of regulatory best practice across the entire sample of 41 indicators for 10 doing business topics.

At the regional level, the competitiveness is also measured by alternative approaches such as Global Urban Competitiveness Report focusing on sustainable urban competitiveness, urban land and urban finance (Kamiya & Pengfei, 2020) or regional competitiveness index measuring the major factors of competitiveness across the regions of European Union (Annoni & Dijkstra, 2020).

In general, in both theory and practice, measuring national and local competitiveness usually generates debates from the subjects being measured and from academia. Then, a number of alternative approaches have been introduced to provide a multi-dimensional perspective on the approach to measuring competitiveness (Benítez-Márquez et al., 2022; Kozyr et al., 2018; Porter, 2004; Rizzi et al., 2015; Berumen, 2005). Therefore, the measurement of competitiveness of localities in each particular country is increasingly required to be associated with national and local characteristics to meet the requirements of national competitive policies.

Methods and Data

Overview of Measuring Methods of Provincial Competitiveness Index in Vietnam

In Vietnam, the provincial competitiveness index has been stably deployed over the years. PCI is a useful index to assess the efforts of localities in improving the competitiveness of provinces, cities and specific sectors (Hau, 2021). The components of PCI and calculation methods have changed and adjusted over some stages. Specifically, in the period 2005–2012, the survey data were preliminarily calculated at the provincial level before adjusting and building the component indexes (VCCI & USAID, 2013). Starting in 2013, the authors switched to calculating and adjusting each indicator value, component indexes, and PCI score for each enterprise that responded to the scoring survey. At the same time, the PCI index was added a new component index—“Fair competition” to reflect the requirements of the business community for a fair business environment (VCCI & USAID, 2014).

From 2013 and stable up to now, the PCI index is composed of 10 sub-indices for all 63 provinces and cities in the country. Accordingly, a locality is considered to have good governance when it has: (1) low market entry costs; (2) easy access to land and stable land use; (3) transparent business environment and public business information; (4) low informal costs; (5) time to inspect, examine and implement regulations and administrative procedures quickly; (6) fair competitive environment; (7) the provincial government is dynamic and creative in solving problems for businesses; (8) high quality, business development support services; (9) good labor training policy, and (10) fair and efficient dispute resolution procedures and maintained security. Survey data are normalized and weighted on a 10-point scale, composites PCI scores, and ranks by provinces and cities.

Methods to Measure the Efficiency Improvement of the Provincial Competitiveness Index

The purpose of improving the business environment is to facilitate business operations and promote business growth. Hence, one of the questions raised above is how does the business environment, as reflected in PCI, actually contribute to the growth of local businesses? Is a locality with a high ranking in the PCI rankings associated with high growth for businesses in that locality?

Some studies of VCCI have determined: “Under the assumption that other factors remain constant, statistical analysis in the 2008 PCI report has shown that: One point increase in the unweighted PCI in a given province is expected to result in: i) An increase of 6.9% in the number of active enterprises; ii) New investment per capita increased by 2.6 million Vietnam dong and iii) increased by 1.6% of GDP per capita in the next year” (VCCI & USAID, 2018, p. 31) or “one point improved in the PCI will increase the rate of new business start-ups by 2.7%” (VOV, 2017) or “The statistical model used in PCI 2016 can also account for this long-term effect. That is, one point increase in PCI will increase new business start-ups by 3% over the next 10 years” (VOV, 2017).

The above impact assessment approaches use econometric methods, based on establishing regression models with each dependent variable describing the corresponding business growth indicator. The above calculations lead to two important conclusions: (1) PCI has a positive impact on business growth which can vary from period to period, and (2) the impact of PCI has lagged over time.

Using a different approach, Tung (2014) used the efficiency measurement method by data envelopment analysis (DEA) and panel data from 2009 to 2011 to calculate the technical efficiency (TE) of each locality based on comparing inputs and outputs.

This method uses mathematical tools instead of econometrics. The outstanding advantage of the DEA method is that it allows combining multiple inputs and outputs for efficient computation. Accordingly, the input sets include nine component indexes of PCI and the output includes eight indicators reflecting the growth of the business.

Compared with VCCI’s approach, the approach of Tung (2014) allows considering the relationship between inputs (components of PCI) creating outputs (business growth indicators) in the same index. However, due to data limitations, the study of Tung (2014) has the following limitations: (1) the inputs of the model only include components of PCI, not including the basic advantages of each locality (such as population, labor). Therefore, the study cannot answer the question of how different regions with basic advantages affect firm growth; (2) PCI data for a specific year are considered as inputs to the model for the corresponding year, without considering policy lag, and (3) the current PCI data have changed compared to the data of 2009–2011 period in terms of both data and how the index was calculated.

From the advantages and limitations of the existing studies, this study uses a combination approach of econometric models and mathematical models to calculate the efficiency index of localities in the following two steps:
Estimating the econometric model to determine the lag impact of PCI on business growth. The purpose of this step is to select PCI data that are suitable for the expected year of influence, to overcome the second limitation of Tung (2014);

PCI data, once defined, will be used as inputs. At the same time, those will be combined with some other local inputs to form a set of inputs, overcoming the first limitation of Tung (2014). In addition, the panel data of PCI are selected from 2013 (when the index set was stable with 10 components) to 2019, updating new changes of PCI and other indicators.

Specific steps include as follows.

Step 1. Determine the Impact Lag of PCI on Business Growth

Given the characteristics of PCI data over the years and based on other accessible data, this study develops a panel data regression model, which takes advantage of the data observed on both regions (63 provinces and cities) and time (7 years, from 2013 to 2019). The panel data are superior based on a combination of cross-sectional data (observations at the same time) or time-series data (few observations at multiple points in time). To consider the effects of variables, this study proposes to compare and choose between fixed effect model (FEM) and random effect model (REM) as presented briefly by the following Equations (1) and (2):

FEM:
$Y_{i t} = β_{j} X_{j i t} + μ_{i t}$
(1)

REM:
$Y_{i t} = β_{j} X_{j i t} + v_{i} + ε_{i t}$
(2)

where Y_it is the dependent variable reflecting the business growth (number of businesses) in locality i in year t (from 2013 to 2019). The variable of number of businesses is selected as the dependent variable because it is the most visible indicator among business growth indicators such as revenue or profit, less dependent on the various characteristics of industries and localities. Then, it is easy to compare across localities of the country.

X _jit is the value of the variable j (including four variables: population, labor, ${P C I}_{t}, {P C I}_{t - 1}$ ) of locality i in year t; $β_{j}$ is coefficients of variable j; $μ_{i t}$ is the error term of locality i in year t. Variable ${P C I}_{t - 1}$ is added to consider the 1-year lag of ${P C I}_{t}$ .

For the random effect model, the classical error is divided into two components: The component $v_{i}$ represents locality-specific random effects which vary between localities but do not change over time; the component $ε_{i t}$ represents random effects in each locality.

The purpose of determining both models (1) and (2) at the same time is to select a model that is suitable for the above assumptions. Accordingly, the Hausman test will be used to select a suitable estimation method between the two methods of estimating fixed effects and random effects (Gujarati, 2004, p. 652). Based on the selected model, the results will be reviewed and analyzed to consider the impact of PCI components on business growth.

Step 2. Measure the Efficiency of Provincial Competitiveness Improvement Based on Business Growth Indicators

Measuring Efficiency

The concept of efficiency was first stated and measured by Farrell (1957), developed from the work of Debreu (1951) and Koopmans (1951). Efficiency can be measured in two approaches: input orientation (minimizing inputs without changing outputs) or output orientation (maximizing output without additional inputs) (Coelli et al., 1998). Accordingly, TE reflects the ability of an enterprise (or organization/locality) to achieve maximum output using certain inputs (or minimizing inputs without changing outputs). Meanwhile, allocative efficiency (AE) measures the capacity to optimize the use of inputs, including their cost, with the corresponding production technique. As a result, economic efficiency (EE) or overall efficiency (OE) is the combination of TE and AE (EE = TE × AE). In our study, we choose the output oriented-efficiency to match the goal of PCI (as inputs) toward maximizing the business growth (as outputs).

To illustrate the efficiency of output-oriented efficiency, Farrell (1957) used an example of a firm with one input (x) and two outputs (y1, y2) under the assumption of constant return to scale. The production possibility curve ZZ′ is shown in Figure 1. Point A is below the production possibility frontier, thus corresponding to the point of technical inefficiency. The AB gap represents the technical inefficiency or potential outputs that can be increased without requiring additional inputs. Then, the output-oriented efficiency is measured by the ratio: TE = 0A/0B.

Figure 1.
Source: Coelli et al. (1998).

In cases of the data including price information, then the iso-revenue line DD′ can be added to measure AE. Then, AE is measured by the ratio: AE = 0B/0C. Finally, EE is calculated by combining TE and AE: EE = TE × AE = (0A/0B) × (0B/0C) = 0A/0C.

Measuring Efficiency by Data Envelopment Analysis

There are several methods to measure efficiency, however, they can be divided into two groups of approaches: parametric and non-parametric. The efficiency measurement by DEA is based on a non-parametric approach with the outstanding advantage that it does not need to assume the production function and can handle many inputs and outputs. For DEA, there are different approaches based on assumptions of return to scale. In this study, the output-oriented efficiency based on the variable return to scale assumption is described as follows:
${m a x}_{ϕ, λ} ϕ,$

constrained by
$- ϕ y_{i} + Y λ \geq 0$

$x_{i} - X λ \geq 0$

$N 1^{'} λ = 1$

$λ \geq 0$

where $ϕ$ represents the level of efficiency, $1 \leq ϕ \leq \infty$ . $ϕ - 1$ is the incrementable outputs with constant inputs. In this approach, $1 / ϕ$ defines TE ( $0 \leq T E \leq 1,$ with a value of 1 being the most efficient); X and Y are inputs and outputs, respectively; $i$ is decision-making units (in this case, localities); $λ$ is a vector N × 1; $N 1^{'} λ = 1$ is the constraint of the variable return to scale (VRS) assumption.

Measuring Productivity Change by Malmquist Method

Besides the DEA, panel data characterization can also allow the calculation of the Malmquist TFP index to measure changes in productivity change, decomposed into technological change and TE change.

$m = t f p c h = t e c h c h \times e f f c h$

The value of $m > 1$ indicates an improvement in total factor productivity. The value of $m < 1$ represents a decrease in the total factor productivity and m = 1 shows that the total factor productivity does not change.

Data

To meet the above analysis requirements, data are expected to be organized as balanced panel data. Accordingly, the inputs and outputs of efficiency measurement are described including the variables presented in Table 1, respectively.

Table 1.
Variables Used in the Models.

No. Variable Explanation

Efficiency measurement

Inputs

1 Pop Population (thousand people)

2 Labor Labor force aged 15 years and over (thousand people)

3 Pci The total PCI

3.1–3.10 pci1–pci10 PCI component index 1–10

Outputs

1 business Number of active businesses as of December 31 of each year

2 fdinum Number of licensed FDI projects in the year

3 fdicap Total registered FDI capital (million US dollars)

4 buscap Average investment capital of active enterprises (billion VND)

5 buslabor Total number of employees in active enterprises as of December 31 of each year (person)

6 asset Value of fixed assets and long-term financial investment of active enterprises as of December 31 every year (billion VND)

7 revenue Net revenue of active enterprises (billion VND)

8 profit Profit before tax of active enterprises (billion VND)

9 profitrate Profit rate of active enterprises (%)

Regression

Dependent variable

1 business Number of active businesses as of December 31 of each year

Independent variables

2 Pci The total PCI

3 d1_pci First difference of PCI

4 d2_pci second difference of PCI

5 Pop Population (thousand people)

6 Labor Labor force aged 15 years and over (thousand people)

Source: Compiled from PCI report (VCCI & USAID, 2021b), statistical yearbook over the respective years and Business White Paper in 2021 (GSO, 2021).

Panel data starting from 2013 to 2019 to ensure 10 stable component indexes of PCI for 63 provinces and cities are collected from PCI website (VCCI & USAID, 2021b). The PCI 2020 has been published; however, other data including business statistics corresponding to the output variables from 1 to 9 in Table 1 are incomplete. Thus, PCI 2020 is excluded for all variables. It should be noted that the fully published PCI data for the years mentioned above are weighted. The remaining variables are aggregated from statistical yearbooks over the years. Some business indicators in 2019, which have not been updated in the corresponding yearbook, are supplemented from the Business White Paper 2021.

Efficiency by DEA technique and Malmquist index were analyzed and bootstrapped 1,000 times using the Benchmarking package 0.29 and Productivity 1.1.0 on the R platform, respectively.

Research Results

The Relationship Between Provincial Competitiveness Index and Business Growth

The purpose of this analysis is to determine the relationship between the PCI and the business growth indicators. At the same time, it determines the lag of the PCI impacts, if any. As analyzed above, two fixed effects and random effects models were analyzed and selected by the Hausman test, respectively. In which, the dependent variable is the number of businesses (business). Independent variables include PCI (pci), the first difference of PCI (d1_pci), the second difference of PCI (d2_pci), population (pop), and the number of labor (labor) of the province or city. After calculating the first and second difference values and removing the missing data due to taking the difference, the remaining panel data consist of 315 observations.

As the result of the Hausman test in Table 2, the value of Prob>chi2 = .000 allows to conclude that the effects of independent variables adequately modeled by a REM is resoundingly rejected. In other words, the fit model chosen is the FEM. According to the FEM model, the population and labor of localities have statistically significant influences on the number of businesses in the province or city. The overall R-squared is approximately 0.7925. Regarding the influence of the PCI, the first difference of PCI has an impact on the number of businesses at the statistically significant level. On the other hand, the influence of the second difference of PCI is reversed, inconsistent with the assumptions.

Table 2.
Regression Results of Fixed Effect Model (FEM) and Random Effect Model (REM).

Variable FEM REM

Coef. p > |t| Coef. p > |t|

PCI (pci) −75.94 .454 212.70 .076

First difference of PCI (d1_pci) 592.01 .019 345.53 .263

Second difference of PCI (d2_pci) −223.75 .073 −192.51 .211

Population (pop) 36.15 .000 51.38 .000

Labor 76.86 .000 −52.95 .000

_cons −107,241.50 .000 −34,160.99 .000

sigma_u 79,687.55 10,559.98

sigma_e 3,390.57 3,390.57

Rho 0.98 0.91

R ²

Within 0.6727 0.4730

Between 0.8089 0.8637

Overall 0.7925 0.8499

Prob > F, Prob > Chi2 .000 .000

Source: Author’s calculation based on data from 2013 to 2019 compiled from PCI reports (VCCI & USAID, 2021b), statistical yearbook over the respective years and Business White Paper in 2021 (GSO, 2021).

As a result, the first difference of PCI (instead of the original PCI), population, and labor of the localities and other inputs in Table 1 will be used as input sets for the efficiency measurement in step 2. Based on the characteristics of the first-difference variable of PCI, hence, we will call it the competitiveness improvement instead of competitiveness.

Efficiency of Provincial Competitiveness Improvement Based on the Business Growth Indicators

The results of calculation of TE by DEA method with 1,000 iterations by bootstrap, using a set of 13 inputs and 9 outputs, are presented in Tables 3 and 4.

Table 3.
Descriptive Statistics of Technical Efficiency over the Years.

Variable Number of Observations Mean Standard Deviation Minimum Maximum

TE 2014 63 0.7846 0.1346 0.3794 0.9021

TE 2015 63 0.9321 0.0720 0.5802 0.9788

TE 2016 63 0.9054 0.0808 0.6205 0.9411

TE 2017 63 0.9034 0.0772 0.5929 0.9596

TE 2018 63 0.9125 0.0948 0.3383 0.9664

TE 2019 63 0.9664 0.0395 0.7615 0.9788

Source: Author’s calculation based on data from 2013 to 2019 compiled from PCI reports (VCCI & USAID, 2021b), statistical yearbook over the respective years and Business White Paper in 2021 (GSO, 2021).

Table 4.
Technical Efficiency of Localities in 2019.

Locality TE 2019 Rank 2019 Locality TE 2019 Rank 2019

Lao Cai 0.97881 1 Hau Giang 0.97602 33

Ha Tinh 0.97850 2 Ha Giang 0.97602 34

Bac Lieu 0.97813 3 Ca Mau 0.97598 35

Kien Giang 0.97805 4 Hoa Binh 0.97595 36

Nghe An 0.97805 5 Long An 0.97591 37

Quang Ngai 0.97791 6 Tien Giang 0.97574 38

Dien Bien 0.97777 7 Phu Tho 0.97570 39

Dong Nai 0.97774 8 Thai Binh 0.97564 40

Bac Giang 0.97769 9 Thanh Hoa 0.97563 41

Hung Yen 0.97760 10 Binh Dinh 0.97556 42

Thua Thien – Hue 0.97745 11 Khanh Hoa 0.97554 43

Da Nang 0.97735 12 Ninh Binh 0.97548 44

Soc Trang 0.97727 13 Ha Nam 0.97536 45

Hai Duong 0.97723 14 Ben Tre 0.97534 46

Tuyen Quang 0.97721 15 Ba Ria-Vung Tau 0.97530 47

Ho Chi Minh city 0.97719 16 An Giang 0.97528 48

Binh Duong 0.97712 17 Quang Tri 0.97518 49

Kon Tum 0.97682 18 Đak Lak 0.97512 50

Lam Dong 0.97680 19 Thai Nguyen 0.97512 51

Lang Son 0.97672 20 Cao Bang 0.97487 52

Tay Ninh 0.97661 21 Bac Kan 0.97484 53

Nam Dinh 0.97660 22 Binh Phuoc 0.97462 54

Tra Vinh 0.97660 23 Vinh Phuc 0.97459 55

Dak Nong 0.97644 24 Son La 0.97459 56

Bac Ninh 0.97642 25 Phu Yen 0.97457 57

Binh Thuan 0.97639 26 Yen Bai 0.97414 58

Gia Lai 0.97631 27 Lai Chau 0.97353 59

Ha Noi 0.97618 28 Hai Phong 0.86141 60

Dong Thap 0.97615 29 Vinh Long 0.84432 61

Ninh Thuan 0.97612 30 Can Tho 0.81404 62

Quang Binh 0.97611 31 Quang Nam 0.76145 63

Quang Ninh 0.97610 32

Source: Author’s calculation based on data 2019 compiled from PCI reports (VCCI & USAID, 2021b), statistical yearbook over the respective years and Business White Paper in 2021 (GSO, 2021)

Table 3 presents descriptive statistics on the average TE over the years from 2014 to 2019. As mentioned at the end of step 1, the first difference of PCI instead of the original PCI will be used in efficiency measurement. Due to the first difference of PCI to measure the differences of variables among 2 years, data change between 2013 and 2014 used to calculate TE 2014.

In general, the average TE of the whole country increases gradually over the years. The lowest average TE reached 0.7846 in 2014 and the highest one reached 0.9664 in 2019. In addition, the TE in 2004 also reflect relatively large inter-provincial variation. The standard deviation is approximately 0.1346, while the locality has the lowest TE value at 0.3794 and the highest at 0.9021. Over the years, the data between localities are relatively more uniform. They also reflect the fact that the interest of localities in improving PCI is increasing. According to VCCI and USAID (2021a), the trend of PCI scores among provinces after 16 years of PCI implementation seems changing. The gap between the top and bottom provinces in terms of both original PCI and weighted PCI scores is narrowing. Positive improvement trend maintained. However, while the improvements of the bottom group of provinces are considered a good sign, it seems that the reform achievements of the top provinces of the PCI seem to be limited to areas that are easy to reform.

In Table 4, provinces and cities are ranked based on TE scores, respectively. It should be reminded that this ranking is not for comparison with the PCI rankings due to the difference in methodology. While the PCI looks at businesses’ perceptions of the business environment, TE examines the relationship between inputs (PCI improvement with one year lagged), basic conditions of provinces and cities (population, labor), and outputs (business growth indicators). A locality with a high TE score implies that the locality makes more efficient use of combinations of inputs to maximize outputs. Therefore, although the localities at the top of the PCI 2019 rankings such as Quang Ninh, Dong Thap, Bac Ninh, and Da Nang are still in the upper half of the TE rankings, most of the remaining provinces have very different position between the two rankings.

Changes in the Efficiency of Provincial Competitiveness Improvement Based on Business Growth Indicators

Besides measuring TE as above, panel data also allow measuring changes in total factor productivity (tfpch), technology change (techch), and TE change (effch). The analysis results for the Malmquist index are shown in Table 5.

Table 5.
Changes of Indicators in the Period 2014–2019.

Variable Number of Observations Mean Standard Deviation Minimum Maximum

Total factor productivity change 63 1.1086 0.4444 0.1436 5.3574

Technology change 63 1.0068 0.2296 0.5029 2.7951

Technical efficiency change 63 1.1034 0.3735 0.2197 5.3574

Source: Author’s calculation based on data from 2013 to 2019 compiled from PCI reports (VCCI & USAID, 2021b), statistical yearbook over the respective years and Business White Paper in 2021 (GSO, 2021).

The results of measuring changes in the indexes show that the average values of total productivity change, technology change, and TE change are all greater than 1, implying an improvement over the years. However, the values are quite close to 1 and these changes are relatively small over the whole period.

Measuring technical performance by DEA has limitations regarding data. In this study, we use 2019 data to calculate TE. This time point allows us to use the data in a relatively up-to-date manner and have the most data available for analysis. However, the measurement at one point in time does not allow the change over the years to be observed. As the results presented above, the analysis is based on panel data using the Malmquist index is useful to describe the changes of TE over the years. Hence, it shows the need for further studies, with possible data, to assess the change and better reflect the efforts of localities over the years.

Conclusion and Policy Implications

Going back to the issues raised at the beginning of this study, within the available data, some conclusions can be drawn including:

First, the business environment, as reflected by PCI, contributes to the growth of local businesses. However, these effects of PCI do not necessarily reflect at the same time as business growth indicators. From the results of this study, it can be seen that the impact of PCI improvement has a lag of approximately one year. In other words, efforts to improve PCI this year can create favorable conditions that positively affect the business growth in the next year. Compared with the conclusion of the VCCI research team that “one-point increase in PCI will increase new business start-ups by 3% over the next 10 years” (VOV, 2017), the results show the similarity of the trend, but different in degree of impact (different methods and data of the studies). The finding shows that the goal of improving the business environment of localities must be a continuous process and has a lasting impact on business growth.

Second, if PCI considers the business environment of localities based on the assumption that other factors remain constant, TE-based-approach will lead to different results. In the TE approach, the business growth indicators are affected not only by the improvement of the business environment created by the local government but also by the basic conditions of that locality. This approach seems to be “fair” across localities. In a locality with a low starting point (or low inputs), although its output may be lower than other localities, that locality can still be more efficient, in terms of the input–output correlation.

It is necessary to notice that the approaches between two methods of measuring TE (between input and output) and PCI (based on component indicators to assess competitive capacity) and purposes are different. The differences between the two approaches, TE or PCI, do not portray which one is better. The PCI approach considers the efforts of the local governments to improve the business environment. Meanwhile, the TE approach allows for a link between these efforts, combined with the local characteristics that lead to business growth. In other words, TE and PCI should be seen as complementary. Therefore, comparing the results of ranking the provinces between the two methods is not necessary. Instead of criticizing PCI’s practices, through this study, we only want to contribute to PCI and emphasize the relationship between PCI and business growth outcomes, including emphasizing basic characteristics of localities. In addition, we intended to compare this ranking with previous studies. However, the methods and variables used in the model are different. Such rankings may only be meaningful when compared at the same time. Compared with Tung (2014), our data are more than 10 years later. Therefore, we did not make such a comparison would be disproportionate and result in the readers’ confusion.

However, this approach is also along with its limitations. First, the attempt to include basic characteristics of localities in the calculation is often accompanied by variable shortcomings. For example, in this study, only the population and labor variables are considered. Meanwhile, production theory includes two basic types of inputs, capital, and labor. From the perspective of the economy, capital can be understood as the size of the economy, which can be expressed through GRDP, the total social investment capital of each locality. Even, the quality of transport infrastructure can be the factor to consider. However, within the scope of this study, those are inaccessible data. We believe that our efforts are incomplete due to limitations in data access. However, this approach is expected to make sense and should be considered when evaluating the efforts of localities. Hence, we believe that future studies, with better access to data, need to combine more inputs and outputs to reflect the richness of practice and the diverse characteristics of business environment.

Last but not least, the PCI data used in this study published are not original, but weighted data. On the other studies of the VCCI team, the analytical data are mostly based on unweighted data. Because of limitations in data access, differences in results are inevitable.

Appendix.
Descriptive Statistics for Panel Data.

Variable Mean SD Min Max Observations

Year Overall 2,016 2 2,013 2,019 N = 441

Between 0 2,016 2,016 n = 63

Within 2 2,013 2,019 T = 7

Idprovince Overall 32 18.20 1 63 N = 441

Between 18.33 1 63 n = 63

Within 0.00 32 32 T = 7

Pop Overall 1,473.29 1,310.05 305.20 9,038.60 N = 441

Between 1,316.73 315.73 8,332.51 n = 63

Within 78.16 960.77 2,188.33 T = 7

Labor Overall 865.57 715.89 208.70 4,826.00 N = 441

Between 719.47 213.66 4,473.54 n = 63

Within 43.70 514.33 1,218.03 T = 7

Business Overall 8,731.46 26,913.22 399.00 239,623.00 N = 441

Between 26,282.79 521.14 178,562.10 n = 63

Within 6,554.08 −49,106.68 69,792.32 T = 7

Fdinum Overall 40.85 132.23 0.00 1,365.00 N = 441

Between 124.02 0.00 827.86 n = 63

Within 48.10 −330.01 577.99 T = 7

Fdicap Overall 470.61 1,069.49 −7.40 8,669.70 N = 441

Between 886.16 0.00 4,938.64 n = 63

Within 607.64 −2,484.93 5,387.33 T = 7

Buscap Overall 349,608.60 1,154,116.00 7,330.00 9,614,692.00 N = 441

Between 1,117,492.00 9,054.00 6,550,992.00 n = 63

Within 316,584.80 −1,746,360.00 3,716,170.00 T = 7

Buslabor Overall 209,075.70 455,691.90 7,156.00 2,958,127.00 N = 441

Between 456,661.50 7,606.00 2,781,411.00 n = 63

Within 44,261.89 −171,825.00 416,418.10 T = 7

Asset Overall 151,286.70 458,784.00 2,352.00 3,820,547.00 N = 441

Between 444,861.10 2,981.29 2,787,345.00 n = 63

Within 123,612.20 −885,958.30 1,184,489.00 T = 7

Revenue Overall 276,015.90 696,490.80 4,047.00 5,993,159.00 N = 441

Between 675,960.90 4,630.14 4,228,986.00 n = 63

Within 185,491.60 −1,047,053.00 2,040,189.00 T = 7

Profit Overall 9,743.66 26,789.07 −23,498.00 190,636.00 N = 441

Between 25,829.74 −4,108.29 149564.30 n = 63

Within 7,718.58 −45,915.63 61,602.09 T = 7

Profitrate Overall 2.14 2.69 −8.40 16.06 N = 441

Between 2.16 −2.10 9.91 n = 63

Within 1.63 −4.85 11.60 T = 7

pci Overall 60.63 4.14 48.96 73.40 N = 441

Between 2.68 54.99 68.51 n = 63

Within 3.17 51.32 68.45 T = 7

pci1 Overall 7.89 0.69 5.86 9.47 N = 441

Between 0.27 7.40 8.60 n = 63

Within 0.63 6.21 9.24 T = 7

pci2 Overall 6.29 0.75 4.12 8.68 N = 441

Between 0.49 4.95 7.33 n = 63

Within 0.58 4.85 8.41 T = 7

pci3 Overall 6.19 0.52 4.45 7.63 N = 441

Between 0.31 5.43 7.01 n = 63

Within 0.41 4.85 7.22 T = 7

pci4 Overall 6.63 0.80 4.83 8.90 N = 441

Between 0.66 5.49 8.35 n = 63

Within 0.46 5.04 7.77 T = 7

pci5 Overall 5.66 0.98 2.81 8.94 N = 441

Between 0.69 4.32 7.08 n = 63

Within 0.71 4.02 8.53 T = 7

pci6 Overall 5.43 0.99 3.12 8.19 N = 441

Between 0.61 3.90 6.85 n = 63

Within 0.79 3.41 8.06 T = 7

pci7 Overall 5.33 0.96 3.08 8.37 N = 441

Between 0.58 4.11 7.06 n = 63

Within 0.77 3.58 7.31 T = 7

pci8 Overall 5.87 0.75 3.68 7.82 N = 441

Between 0.46 4.73 7.28 n = 63

Within 0.59 3.92 7.46 T = 7

pci9 Overall 6.06 0.84 4.10 8.24 N = 441

Between 0.67 4.79 7.66 n = 63

Within 0.50 4.68 7.41 T = 7

pci10 Overall 5.91 0.81 3.68 7.99 N = 441

Between 0.52 4.83 7.17 n = 63

Within 0.63 3.77 8.13 T = 7

Source: Compiled from PCI report (VCCI & USAID, 2021b), statistical yearbook over the respective years and Business White Paper in 2021 (GSO, 2021).

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Diep Thanh Tung

References

Annoni

, & Dijkstra

(2020). The EU regional competitiveness index 2019 .

Benítez-Márquez

M.-D.

, Sánchez-Teba

E. M.

, & Coronado-Maldonado

(2022). An alternative index to the global competitiveness index. PLoS ONE , 17(3), e0265045. https://doi.org/10.1371/journal.pone.0265045

Berumen

S. A.

(2005). An approach to local and regional competitiveness. Cuadernos de Administración , 18(29), 13–32.

Cherunilam

(2015). Business policy and strategic management . Himalaya Publishing House.

Coelli

, Prasada Rao

D. S.

, & Battese

G.E.

(1998). An introduction to efficiency and productivity analysis . Kluwer Academic Publishers.

Debreu

(1951). The coefficient of resource utilization. Econometrica , 19(3), 273–292. https://doi.org/10.2307/1906814

Djankov

, McLiesh

, & Klein

(2004). Doing business in 2004: Understanding regulation . World Bank.

Farrell

M. J.

(1957). The measurement of productive efficiency. Journal of the Royal Statistical Society: Series A (General) , 120(3), 253–281. https://doi.org/10.2307/2343100

GSO. (2021). Statistical data . https://www.gso.gov.vn/so-lieu-thong-ke/

10.

Gujarati

D. N.

(2004). Basic econometrics . Tata McGraw-Hill.

11.

Hau

H. Q.

(2021). Improvement of the competitiveness of Vietnam tourism. Journal of Economics and Development , 283, 59–69.

12.

IMD World Competitiveness Center. (2022). World competitiveness ranking . https://www.imd.org/centers/world-competitiveness-center/rankings/world-competitiveness/

13.

Kamiya

, & Pengfei

(2020). Global urban competitiveness report (2019–2020) .

14.

Koopmans

T. C.

(1951). An analysis of production as an efficient combination of activities .

15.

Kozyr

N. S.

, Petrovskaya

N. E.

, & Zazimko

V. L.

(2018). Modern approaches assessing global competitiveness. European Research Studies Journal , XXI(2), 1034–1041. https://doi.org/10.35808/ersj/1355

16.

Minh

(2013). After PCI, the first reactions of provincial leaders . https://vneconomy.vn/hau-pci-nhung-phan-ung-dau-tien-cua-lanh-dao-tinh-thanh.htm

17.

Porter

M. E.

(2004). Building the microeconomic foundations of prosperity: Findings from the business competitiveness index . Oxford University Press.

18.

Rizzi

, Graziano

, & Dallara

(2015). The regional competitiveness: An alternative approach. Rivista Internazionale Di Scienze Sociali , 123(3), 307–335.

19.

Schwab

(2017). The Global Competitiveness Report 2017–2018 . https://www3.weforum.org/docs/GCR2017-2018/05FullReport/TheGlobalCompetitivenessReport2017%E2%80%932018.pdf

20.

Tuan

D. A.

(2017). Choose how to deal with PCI . https://nhandan.vn/dien-dan-nhan-dan-cuoi-tuan/chon-cach-ung-xu-voi-pci-287900/

21.

Tung

D. T.

(2014). Additional approaches to assess the Vietnam provincial competitiveness index (PCI). International Business Research , 7(3), 1–12. http://doi.org/10.5539/ibr.v7n3p1

22.

VCCI & USAID. (2014). Vietnam provincial competitiveness index 2013 . https://pcivietnam.vn/en/publications/pci-full-report

23.

VCCI & USAID. (2018). Manual for the use of the results of the provincial competitiveness index (PCI) survey . https://pcivietnam.vn/en/about/pci-methodology.html

24.

VCCI & USAID. (2021a). Vietnam provincial competitiveness index 2020 . https://pcivietnam.vn/en/publications/pci-full-report

25.

VCCI & USAID. (2013). PCI methodology . https://pcivietnam.vn/gioi-thieu/phuong-phap.html

26.

VCCI & USAID. (2021b). PCI data . https://pcivietnam.vn/du-lieu-pci

27.

VOV. (2017). What effect does PCI have on local economic efficiency ? https://vov.vn/kinh-te/pci-co-tac-dung-gi-toi-hieu-qua-kinh-te-dia-phuong-602781.vov

No.	Variable	Explanation
Efficiency measurement
Inputs
1	Pop	Population (thousand people)
2	Labor	Labor force aged 15 years and over (thousand people)
3	Pci	The total PCI
3.1–3.10	pci1–pci10	PCI component index 1–10
Outputs
1	business	Number of active businesses as of December 31 of each year
2	fdinum	Number of licensed FDI projects in the year
3	fdicap	Total registered FDI capital (million US dollars)
4	buscap	Average investment capital of active enterprises (billion VND)
5	buslabor	Total number of employees in active enterprises as of December 31 of each year (person)
6	asset	Value of fixed assets and long-term financial investment of active enterprises as of December 31 every year (billion VND)
7	revenue	Net revenue of active enterprises (billion VND)
8	profit	Profit before tax of active enterprises (billion VND)
9	profitrate	Profit rate of active enterprises (%)
Regression
Dependent variable
1	business	Number of active businesses as of December 31 of each year
Independent variables
2	Pci	The total PCI
3	d1_pci	First difference of PCI
4	d2_pci	second difference of PCI
5	Pop	Population (thousand people)
6	Labor	Labor force aged 15 years and over (thousand people)

Variable	FEM		REM
Variable	Coef.	p > \|t\|	Coef.	p > \|t\|
PCI (pci)	−75.94	.454	212.70	.076
First difference of PCI (d1_pci)	592.01	.019	345.53	.263
Second difference of PCI (d2_pci)	−223.75	.073	−192.51	.211
Population (pop)	36.15	.000	51.38	.000
Labor	76.86	.000	−52.95	.000
_cons	−107,241.50	.000	−34,160.99	.000
sigma_u	79,687.55		10,559.98
sigma_e	3,390.57		3,390.57
Rho	0.98		0.91
R ²
Within	0.6727		0.4730
Between	0.8089		0.8637
Overall	0.7925		0.8499
Prob > F, Prob > Chi2	.000		.000

Variable	Number of Observations	Mean	Standard Deviation	Minimum	Maximum
TE 2014	63	0.7846	0.1346	0.3794	0.9021
TE 2015	63	0.9321	0.0720	0.5802	0.9788
TE 2016	63	0.9054	0.0808	0.6205	0.9411
TE 2017	63	0.9034	0.0772	0.5929	0.9596
TE 2018	63	0.9125	0.0948	0.3383	0.9664
TE 2019	63	0.9664	0.0395	0.7615	0.9788

Locality	TE 2019	Rank 2019	Locality	TE 2019	Rank 2019
Lao Cai	0.97881	1	Hau Giang	0.97602	33
Ha Tinh	0.97850	2	Ha Giang	0.97602	34
Bac Lieu	0.97813	3	Ca Mau	0.97598	35
Kien Giang	0.97805	4	Hoa Binh	0.97595	36
Nghe An	0.97805	5	Long An	0.97591	37
Quang Ngai	0.97791	6	Tien Giang	0.97574	38
Dien Bien	0.97777	7	Phu Tho	0.97570	39
Dong Nai	0.97774	8	Thai Binh	0.97564	40
Bac Giang	0.97769	9	Thanh Hoa	0.97563	41
Hung Yen	0.97760	10	Binh Dinh	0.97556	42
Thua Thien – Hue	0.97745	11	Khanh Hoa	0.97554	43
Da Nang	0.97735	12	Ninh Binh	0.97548	44
Soc Trang	0.97727	13	Ha Nam	0.97536	45
Hai Duong	0.97723	14	Ben Tre	0.97534	46
Tuyen Quang	0.97721	15	Ba Ria-Vung Tau	0.97530	47
Ho Chi Minh city	0.97719	16	An Giang	0.97528	48
Binh Duong	0.97712	17	Quang Tri	0.97518	49
Kon Tum	0.97682	18	Đak Lak	0.97512	50
Lam Dong	0.97680	19	Thai Nguyen	0.97512	51
Lang Son	0.97672	20	Cao Bang	0.97487	52
Tay Ninh	0.97661	21	Bac Kan	0.97484	53
Nam Dinh	0.97660	22	Binh Phuoc	0.97462	54
Tra Vinh	0.97660	23	Vinh Phuc	0.97459	55
Dak Nong	0.97644	24	Son La	0.97459	56
Bac Ninh	0.97642	25	Phu Yen	0.97457	57
Binh Thuan	0.97639	26	Yen Bai	0.97414	58
Gia Lai	0.97631	27	Lai Chau	0.97353	59
Ha Noi	0.97618	28	Hai Phong	0.86141	60
Dong Thap	0.97615	29	Vinh Long	0.84432	61
Ninh Thuan	0.97612	30	Can Tho	0.81404	62
Quang Binh	0.97611	31	Quang Nam	0.76145	63
Quang Ninh	0.97610	32

Variable		Mean	SD	Min	Max	Observations
Year	Overall	2,016	2	2,013	2,019	N = 441
	Between		0	2,016	2,016	n = 63
	Within		2	2,013	2,019	T = 7
Idprovince	Overall	32	18.20	1	63	N = 441
	Between		18.33	1	63	n = 63
	Within		0.00	32	32	T = 7
Pop	Overall	1,473.29	1,310.05	305.20	9,038.60	N = 441
	Between		1,316.73	315.73	8,332.51	n = 63
	Within		78.16	960.77	2,188.33	T = 7
Labor	Overall	865.57	715.89	208.70	4,826.00	N = 441
	Between		719.47	213.66	4,473.54	n = 63
	Within		43.70	514.33	1,218.03	T = 7
Business	Overall	8,731.46	26,913.22	399.00	239,623.00	N = 441
	Between		26,282.79	521.14	178,562.10	n = 63
	Within		6,554.08	−49,106.68	69,792.32	T = 7
Fdinum	Overall	40.85	132.23	0.00	1,365.00	N = 441
	Between		124.02	0.00	827.86	n = 63
	Within		48.10	−330.01	577.99	T = 7
Fdicap	Overall	470.61	1,069.49	−7.40	8,669.70	N = 441
	Between		886.16	0.00	4,938.64	n = 63
	Within		607.64	−2,484.93	5,387.33	T = 7
Buscap	Overall	349,608.60	1,154,116.00	7,330.00	9,614,692.00	N = 441
	Between		1,117,492.00	9,054.00	6,550,992.00	n = 63
	Within		316,584.80	−1,746,360.00	3,716,170.00	T = 7
Buslabor	Overall	209,075.70	455,691.90	7,156.00	2,958,127.00	N = 441
	Between		456,661.50	7,606.00	2,781,411.00	n = 63
	Within		44,261.89	−171,825.00	416,418.10	T = 7
Asset	Overall	151,286.70	458,784.00	2,352.00	3,820,547.00	N = 441
	Between		444,861.10	2,981.29	2,787,345.00	n = 63
	Within		123,612.20	−885,958.30	1,184,489.00	T = 7
Revenue	Overall	276,015.90	696,490.80	4,047.00	5,993,159.00	N = 441
	Between		675,960.90	4,630.14	4,228,986.00	n = 63
	Within		185,491.60	−1,047,053.00	2,040,189.00	T = 7
Profit	Overall	9,743.66	26,789.07	−23,498.00	190,636.00	N = 441
	Between		25,829.74	−4,108.29	149564.30	n = 63
	Within		7,718.58	−45,915.63	61,602.09	T = 7
Profitrate	Overall	2.14	2.69	−8.40	16.06	N = 441
	Between		2.16	−2.10	9.91	n = 63
	Within		1.63	−4.85	11.60	T = 7
pci	Overall	60.63	4.14	48.96	73.40	N = 441
	Between		2.68	54.99	68.51	n = 63
	Within		3.17	51.32	68.45	T = 7
pci1	Overall	7.89	0.69	5.86	9.47	N = 441
	Between		0.27	7.40	8.60	n = 63
	Within		0.63	6.21	9.24	T = 7
pci2	Overall	6.29	0.75	4.12	8.68	N = 441
	Between		0.49	4.95	7.33	n = 63
	Within		0.58	4.85	8.41	T = 7
pci3	Overall	6.19	0.52	4.45	7.63	N = 441
	Between		0.31	5.43	7.01	n = 63
	Within		0.41	4.85	7.22	T = 7
pci4	Overall	6.63	0.80	4.83	8.90	N = 441
	Between		0.66	5.49	8.35	n = 63
	Within		0.46	5.04	7.77	T = 7
pci5	Overall	5.66	0.98	2.81	8.94	N = 441
	Between		0.69	4.32	7.08	n = 63
	Within		0.71	4.02	8.53	T = 7
pci6	Overall	5.43	0.99	3.12	8.19	N = 441
	Between		0.61	3.90	6.85	n = 63
	Within		0.79	3.41	8.06	T = 7
pci7	Overall	5.33	0.96	3.08	8.37	N = 441
	Between		0.58	4.11	7.06	n = 63
	Within		0.77	3.58	7.31	T = 7
pci8	Overall	5.87	0.75	3.68	7.82	N = 441
	Between		0.46	4.73	7.28	n = 63
	Within		0.59	3.92	7.46	T = 7
pci9	Overall	6.06	0.84	4.10	8.24	N = 441
	Between		0.67	4.79	7.66	n = 63
	Within		0.50	4.68	7.41	T = 7
pci10	Overall	5.91	0.81	3.68	7.99	N = 441
	Between		0.52	4.83	7.17	n = 63
	Within		0.63	3.77	8.13	T = 7