An Integrated Approach for Evaluating the Efficiency of FDI Attractiveness: Evidence from Vietnamese Provincial Data from 2012 to 2022

Abstract

In Vietnam, foreign direct investment (FDI) is an important capital flow for sustainable socio-economic growth and international economic integration, contributing to the supplementation of capital, modern technology, management and business capacity, and the ability to organize and participate in the global supply chain. In this paper, a novel hybrid model combining simple average combination of SMA (Simple Moving Average), IFTS (Improved Fuzzy Time Series), and DEA window analysis is proposed to evaluate FDI attractiveness in Vietnam. Five crucial indicators, including labor force (LF), gross regional domestic product (GDP), the Provincial Competitiveness Index (PCI), FDI by capital, and FDI by cases, were employed to explore their impacts on the efficiency of attracting FDI into provinces, for sustainable economic growth. First, the future values of indicators for 2021–2022 were predicted based on collected historical data from 2012 to 2020. Then, the DEA window was employed to evaluate the efficiency of the provinces in terms of their FDI attractiveness during the period 2012–2022. From the results, Binh Duong, Ho Chi Minh, and Vung Tau were found to be the most efficient localities at attracting FDI, while An Giang, Tuyen Quang, and Can Tho had the lowest FDI attractiveness indexes. The proposed model was proven to be effective in identifying both the provinces which should be targeted for further improvement, and the provinces which should serve as role models for other provinces. In this direction, this paper can serve as a significant guideline for decision-makers and the Government to improve FDI attractiveness, with solutions to attract FDI in a sustainable way.

1. Introduction

Foreign direct investment (FDI) can significantly contribute to the sustainable development of both home and host countries in numerous critical ways, including by enhancing market access, bringing in foreign currency, fostering the growth of skills and human capital, transferring technology, and boosting competition in local markets [1]. Additionally, it can aid in the modernization of industry and more easily enable small and medium-sized businesses to join regional and international value chains. In Vietnam, FDI has been a key driver of sustainable economic development and international economic integration. After the introduction of reform policy, called Doi Moi, in 1986, Vietnam emerged as one of the most attractive locations for FDI in Southeast Asia, bringing various benefits to the host nation such as capital, finished products, components, new technology, organizational and managerial skills, distribution channels and markets. The flow of FDI by multinational corporations into Vietnam has continuously increased, making it one of the essential sources of domestic economic growth for the last 30 years. In ASEAN, since 2015, Vietnam surpassed Malaysia to become the third largest recipient of FDI in the region, after Singapore and Indonesia. Amid the recent impact of the COVID-19 pandemic in 2020 on the world’s economic growth, while FDI to Singapore, Thailand, Indonesia, and Malaysia decreased sharply, FDI into Vietnam was not significantly affected [2]. In 2010, annual investment capital was USD 11 billion; by 2015, it was USD 14.5 billion; and by 2016, it was USD 15.8 billion [2]. By the end of 2017, Vietnam had attracted over 25,000 foreign direct investment projects with a total registered investment of more than USD 333 billion. To date, 129 countries/territories have invested in Vietnam. FDI projects have been present in 63/63 provinces of Vietnam, and FDI capital has also been invested in 19/21 production and business lines [3]. According to the Foreign Investment Department—Ministry of Planning and Investment [3], in 2019, FDI capital in Vietnam was USD 38.95 billion, up 7.2% compared with 2018. There have been 3883 projects registered to contribute newly granted investment registration certificates, with a value of USD 16.75 billion, making Vietnam one of the most attractive countries for foreign investors. In particular, in 2020, Vietnam was in the top 20 countries attracting the most FDI globally, ranking 19th (https://unctad.org/, 2021).

According to Hanh et al. [4], the majority of the current legal FDI projects in Vietnam include traditional types of investment. The investments could be conducted through joint ventures, build–operate–transfer (BOT), build–transfer (BT), build–transfer–operate (BTO), and business cooperation agreements, which are all possible forms of foreign investment. However, most FDI projects in Vietnam focus on areas which are low value-added, less environmentally friendly, and have high emissions. Green FDI projects have received little attention. Thus, attracting green FDI has been a keystone solution to help Vietnam maintain competitiveness and sustainable development [5]. Recent challenges—from the COVID-19 pandemic to natural disasters to climate change—have forced businesses to flexibly transform their business models and embrace global trends to overcome difficulties, which are critical factors in their socio-economic and environmental success. In recent years, Vietnam has advocated for selectively attracting FDI towards quality projects for sustainable industrial development. FDI inflows are shifting to high value-added industries and occupations, prioritizing areas with high technology, projects with advanced, modern, and green technologies, and a gradual decrease in labor-intensive ones. Projects with extensive capital but a higher risk of causing environmental pollution, involving outdated technology, and using a large amount of labor, are being rejected. As of 31 December 2020, Vietnam had attracted 33,062 FDI projects with a total investment capital of over USD 386,233 billion [6], in which FDI projects with modern and environmentally friendly technologies gradually increased.

Although Vietnam’s efforts to attract FDI have been successful, FDI inflows into Vietnam are unevenly distributed among provinces (Institute of Development Strategy, Ministry of Planning and Investment) [7]. FDI concentration is near major economic centers such as Hanoi, Ho Chi Minh, and Da Nang. Meanwhile, some northern mountainous provinces and the Central Highlands have low FDI. This leads to very wide development gaps between regions in Vietnam. Provinces and cities that attract a large amount of FDI grow at a high rate and modernize as associated with urbanization. Meanwhile, other localities remain underdeveloped, relying mainly on agriculture and resource exploitation [8]. Vietnam, similar to other countries, has differences in natural and social conditions between regions. Hanoi is the capital, and Ho Chi Minh City is the economic center of the south, so investment capital has been concentrated with special policies and mechanisms in these two regions. Hai Phong, Da Nang, and Can Tho are three cities directly under the Central Government; their technical and social infrastructure has developed quite synchronously. The provinces adjacent to major economic centers have favorable conditions, attracting more FDI and achieving a higher level of development. Uneven FDI distribution among industries, fields, and territories has negative impacts on the sustainable development of the whole national economy, such as creating dependence on technology and markets, and putting pressure on domestic enterprises [9]. Thus, the performance evaluation of an individual region in attracting FDI is crucial to understand the causes of variation among them, and can provide policymakers with a foundation for further actions to enhance FDI attractiveness uniformly and sustainably.

Labor has become one of the important factors for FDI-attractive location selection, and countries or regions with low wages are more likely to attract FDI. Notably, many developing countries have no trade unions, and workers lack the ability to negotiate wages [10]. In countries where there are appropriate policies for training human resources and maintaining the health of human resources, such countries have a competitive advantage and can attract FDI inflows, in addition to having a workforce to meet the needs of foreign investors. In a recent study by Nguyen (2021) [11], investigations in Vietnam showed that there was a positive correlation between FDI and labor force, GDP, inflation, index of business freedom, and index of investment freedom. Vietnam has an abundant and young human capital, and in recent years has quickly shifted from unskilled labor to skilled labor by changing the growth model and improving the quality of human resource training, which has helped to attract FDI into Vietnam. Thus, labor resources affect the country’s ability to attract FDI, as also indicated in numerous empirical studies focusing on FDI attractiveness in Vietnam [11,12,13]. The Provincial Competitiveness Index (PCI) has proven to be an index that many domestic and foreign investors refer to when deciding to invest in Vietnam, as well as an essential source of information for policymakers to improve the investment environment. Nguyen and Ho [14] indicated a number of PCI components to FDI attraction such as labor training, business support, transparency, legal institutions, administrative documents, and land access, to name a few, which are factors that increase the profitability of investors, thereby ensuring the attraction of FDI.

In this paper, a two-stage methodology is proposed to evaluate FDI attractiveness in Vietnam’s provinces, to understand their past, current, and future performance, and benchmark them against the efficiency frontier. Initially, a set of indicators was selected through a literature review and experts’ opinions, which were considered to be key performance factors contributing to measurable values that demonstrate how effectively a host economy is achieving their efficiency of FDI attractiveness; these indicators are necessary data for policymakers to develop strategic plans for attracting and developing the FDI sector in association with Vietnam’s sustainable development strategy. The future patterns of FDI, as well as other indexes such as GDP, are important inputs for policymakers, even more so during severe economic downturns, such as the one caused by the COVID-19 pandemic. Thus, in the first stage of the suggested model, two forecasting methods, simple average combination of SMA (Simple Moving Average in State Space form) and IFTS (Improved Fuzzy Time Series), are applied to predict future values (2021–2022) based on collected historical data (2012–2020). In the second stage, data envelopment analysis (DEA) is employed to evaluate the efficiency of the provinces in terms of their FDI attractiveness during the period 2012–2022. In this direction, to deal with problems of unfair measurement during the 11 year analysis (2012–2022), we use an extended DEA approach, specifically, DEA window analysis is used to detect performance trends over years. While traditional DEA has been used in previous empirical studies which have mainly focused on static efficiency evaluation, which lacks evaluation and analysis in different time periods, DEA window analysis was followed in the development of dynamic DEA by Malmquist (1953) [15] in estimating catch-up and innovation effect. Thus, DEA window analysis is utilized to assess and quantify efficiency change on a yearly basis. To our best knowledge, the integrated approach proposed in this research has never been reported in the existing literature, especially for evaluating FDI attractiveness in the Vietnam context.

The suggested model was applied to data from 42 of Vietnam’s 63 provinces. The provinces were categorized into the country’s seven key economic regions. The aggregate data were chosen based on the consistent availability of yearly reports, covering five dimensions of the provinces: labor force (LF), gross regional domestic product (GDP), provincial competitiveness index (PCI), FDI by capital (cumulative FDI), and FDI by cases (cumulative cases). The 21 missing provinces are mostly located in the highland areas, where the economic indicators are not primarily focused on foreign investors. Considering the huge differences in FDI attractiveness among destinations that Vietnam faces, this paper is the first attempt to analyze the provinces’ FDI attractiveness in Vietnam using the novel integrated approach. Findings from this paper can serve as an important guideline for local policymakers and the Government to improve FDI attractiveness with solutions to attract FDI in a sustainable way.

The rest of the paper is organized as follows: Section 2 presents the literature review. In Section 3, methodology concepts are provided. Section 4 presents the case study. Section 5 presents the results analysis, and Section 6 gives concluding remarks.

2. Literature Review

In this section, we aim to build the literature regarding two aspects: (1) FDI attractiveness evaluation around the world and province-level investigations in Vietnam, and (2) applications of DEA measurement. Therefore, by providing relevant studies, we point out the gap in the literature.

2.1. Determinants of FDI, FDI Impact and Attractiveness and Investigations in Vietnam

There is a wide range of studies on FDI impact and attractiveness, in which FDI attractiveness is defined as the performance/efficiency of the host economy to attract FDI. FDI can boost macroeconomic management mechanisms such as growth prospects, skilled labor, natural resources, fundamental and advanced facilities, and export channels [16]. Traditionally, researchers have examined various economic factors of FDI attraction, including GDP growth, human capital, infrastructures, exchange rate and trade openness [17,18,19,20,21]. Deng et al. [22] stated that the balance of international payment was one of the essential factors, besides the economic development of a country and human capital, to attract FDI. Some studies have focused on technology and equipment transfer, job opportunities, increased export, and international management expertise [23,24,25,26,27]. Basile et al. [28] considered training and R&D as the determinants of attracting foreign subsidiaries in 50 European regions. The institutional impact of the host economy in attracting FDI has been approved by a vast literature [29,30,31]. Mihaylova [32] considered the most important policy-related dimensions, such as institutional framework, labor force quality, cost competitiveness, and infrastructure endowment, to rank central, eastern, and southeastern European countries in their FDI attractiveness. Cui et al. [33] studied the regional advantages as province-level competitiveness indexes for attracting and retaining foreign direct investment in China. According to Yao et al. [34], the improvement of local financial intermediation was preferable to concessional arrangements for attracting FDI. In a study examining the determinants of FDI attractiveness with evidence from ASEAN-7 countries [35], Dang and Nguyen indicated that economic growth, economic institution quality, tax burden, and inflation were major determinants that significantly attracted FDI, while population growth and political institution quality were inversely associated with FDI intake.

In the Vietnam context, vast research provides comprehensive analysis of numerous determinants of FDI attraction at the sub-national level. Meyer and Nguyen (2005) [36] studied the geographical characteristics of the regions attracting FDI; the authors estimated a binomial regression model to show the important role of population, infrastructure, industrial zones, education, FDI stock and economic growth in attracting FDI to provinces in Vietnam. Using a simultaneous equations model, Anwar and Nguyen (2010) [37] investigated the relationship between FDI inflows and economic growth in 61 Vietnamese provinces from 1996 to 2005. Hoang and Goujon (2014) [38] used spatial econometric models to determine the drivers of FDI distribution in Vietnamese provinces after the Asian crisis in 1997; cross-sectional data from 2001 to 2010 were used in their empirical investigation. Esiyok and Ugur (2015) [39] studied the FDI attraction in 62 Vietnamese provinces using a spatial regression approach. Nguyen (2016) [40] studied the FDI determinants in all provinces and cities in Vietnam from 2008 to 2012, using the fixed effect and random effect methods. According to empirical findings, market potential, labor cost, labor quality, infrastructure, provincial policy efficacy, and the previous year’s FDI concentration all have a major impact on FDI allocation across provinces and cities in Vietnam. Furthermore, market potential and wage rates have been statistically proven to influence the size of FDI projects. Ngo et al. (2018) [41] identified the FDI attractiveness of Vietnam at the subnational level with both traditional and emerging factors suggested by economics and international business theories using a longitudinal dataset of 63 provinces/cities from 2008 to 2013. Minh (2019) [42] investigated the effect of institutional quality on FDI inflows in Vietnam, using a set of panel data from a Provincial Competitiveness Index (PCI) survey, and inward FDI to 59 provinces and cities of Vietnam in 2010–2017. The authors applied the GMM estimation with period-specific predetermined instruments to investigate the relation between the quality of provincial governance and FDI inflows. Using a spatial econometric method, Hoang et al. (2021) [43] investigated the factors that influenced foreign direct investment in Vietnam’s Southern Central Coast (SCC) area from 2007 to 2016. The estimation findings demonstrated that FDI inflows into the SCC area were drawn by the host province’s low labor costs, and that provinces with national seaports had a significant advantage in attracting FDI, validating the SCC region’s vertical FDI structure. Surprisingly, skilled labor from neighboring provinces had a beneficial effect on FDI inflows to the host province. Regarding the importance of local institutional quality in attracting FDI, the estimation suggested that the host province’s legal institutions and degree of social security had a significant effect.

2.2. Applications of SMA, IFTS, and DEA Methods

Simple Moving Average (SMA) is a well-known forecasting method. It is easy to understand and interpret, easy to use and has been applied in many studies. Lauren and Harlili [44] used the SMA technique for stock trend prediction. Svetunkov and Petropoulos [45] examined the performance of statistical and judgmental forecasts for monthly sales data from a multinational pharmaceutical company. Ali et al. [46] applied SMA for a problem in the supply chain where information was not shared. Muangprathub et al. [47] employed SMA for observing price trend and potential changes, from data on approximately three hundred stocks retrieved from the Stock Exchange of Thailand. Nayak et al. [48] used SMA for a case study of solar power battery use in India, to smooth the fluctuating power output of solar panels.

The Improved Fuzzy Time Series (IFTS) forecasting method has been widely used to solve forecasting problems in which the historical data were vague, imprecise, or were in linguistic terms. Wang and Hsu [49] developed and applied an Improved Fuzzy Time Series model for forecasting for the tourism industry, from Taiwan to the United States, using annual data for the period of 1991–2001. Duru [50] proposed an Improved Fuzzy Time Series approach for dry bulk shipping index forecasting. Chou et al. [51] implemented tourism demand forecasting based on an Improved Fuzzy Time Series model. Ghosh et al. [52] presented an improved fuzzy time-series method of forecasting based on L–R fuzzy sets for foodgrain production forecasting. Vovan [53] employed the Improved Fuzzy Time Series forecasting model using variations of data in the forecasting of two real problems in Vietnam: the penetration of salt, and the total population.

DEA is a nonparametric assessment and production theory-based mathematical method for benchmarking similar element types based on predetermined inputs and outputs [54]. A decision-making unit (DMU) to be compared can be peers such as schools, manufacturing companies, hospitals, countries, states, and cities, to name a few. The motivation behind DEA is to use linear programming-based optimization to measure efficiency from either an output maximization, or input minimization, perspective [55]. DEA has been widely used to evaluate the performance of diverse entities in various areas: energy efficiency [56], city performance [57,58], system efficiency [59,60], and environmental performance [61]. Regarding FDI efficiency assessment, DEA has been effective in defining the efficiency score of a host economy in attracting FDI, which has successfully been utilized in the various following studies. Lei et al. (2013) [16] used DEA to analyze the FDI attractiveness for the sustainable development of 30 provinces in China. Mastromarco and Simar (2015) [62] introduced the time-dependent conditional DEA estimators to study the effect of FDI and time on catching up in 44 countries during 1970–2007. Dai (2016) [63] conducted a study on FDI efficiency between Korea’s and China’s major cities using the group method of data handling (GMDH) and DEA methods. Teplova and Sokolova (2019) [64] employed the DEA method and DEA–Malmquist to measure the efficiency of FDI transformation on a sample of 31 developed and developing countries. Zhang et al. (2019) [65] used the DEA scale return model to calculate the efficiency of environmental protection spending in China and test the linkages among FDI, fiscal decentralization, and government expenditure efficiency. Pan et al. (2020) [66] studied the impact of FDI quality on energy efficiency using the slacks-based measure data envelopment analysis (SBM-DEA) method. Wang et al. (2022) [67] presented a dynamic direction distance function–data envelopment analysis (DDF-DEA) model to investigate the efficiency of attracting FDI with the inclusion of environmental regulation in 31 Chinese provinces over the period 2015–2017.

2.3. Research Gaps

In view of the above discussion which exhaustively reviewed relevant studies, a hybrid approach that combines the forecasting model of simple average combination of SMA, IFTS, and DEA methodologies to predict and evaluate the past, current, and future FDI attractiveness, has never been reported. DEA is widely regarded as a powerful and practical tool when considering various approaches for assessment and prediction. More specifically, DEA window analysis is suitable for the present research because policymakers must identify efficiency trends over time. It has been used in a variety of economic endeavors, demonstrating the applicability of these techniques and their acceptance as research tools [68,69]. By observing efficiency scores, DEA identifies provinces which should be targeted for further improvement, and provinces which should serve as role models for other provinces. Additionally, implementation of simple average combination of SMA and IFTS for the determined indicators has proved to be very efficient and accurate.

Attempts to apply these three methods to FDI attractiveness assessment in Vietnam were completely missing in the ample relevant academic literature, and this lack attracted our attention. Therefore, this paper aims to provide an evaluation method for policymakers to determine key factors, drawbacks, and opportunities that affect their locality’s attractiveness for FDI, with new orientations and a system of synchronous solutions. It is expected that the study’s results will help policymakers attract and develop the FDI sector in association with the sustainable development strategies of Vietnam.

While the application of two forecasting models is first proposed, in such problem, by using DEA to process the performances of Vietnamese provinces over economic and provincial competitiveness index criteria, this paper aims to gauge the levels of relative efficiency of FDI attractiveness, as well as appraise the potential for provinces to obtain and maintain FDI inflow. By benchmarking provinces that are ‘efficient’, according to DEA criteria, this work also seeks to shed light on other regions that warrant the greatest attention for increasing efficiency, improving sustainability, and regional and national economic security.

3. Materials and Methods

3.1. Forecasting Techniques

In this paper, the simple average combination of SMA (Simple Moving Average in State Space form) and IFTS (Improved Fuzzy Time Series), is proposed. The statistical software “R”, together with the packages “smooth” and “AnalyzeTS” were used to simplify this batch-forecasting process by implementing different customized forecasting scripts. Figure 1 shows the forecasting flowchart.

3.1.1. IFTS Fuzzy Time Series

We propose a new FTS model called the Improved Fuzzy Time Series (IFTS) model [70]. This model is suitable for non-seasonal time series that can interpolate historical data and forecast the future. In the proposed model, all of the parameters are surveyed and determined by suitable methods and algorithms. Performing for many data sets with different characteristics, the proposed model shows its effectiveness in comparison with existing models.

The IFTS model has advantages in being able to process time series with abnormal changes and short history [53]. Another difference to conventional methods, such as ARIMA, is the value used in forecasting, which is the fuzzy set of real numbers for a given set of universes. The fuzzy sets can be interpreted as a number class with the same boundaries.

The definition of fuzzy time series is presented as follows.

Definition 1.

If the set $U$ is the set of universes $U=\left\{u_1,u_2,\dots ,u_n\right\}$, then a fuzzy set $A_i$ of the set $U$ is defined as:

$$A_i=\frac{{\mu }_{A_i}\left(u_1\right)}{u_1}+\frac{{\mu }_{A_i}\left(u_2\right)}{u_2}+\dots +\frac{{\mu }_{A_i}\left(u_n\right)}{u_n},k$$

(1)

where${\mu }_{A_i}\left(u_j\right)$is the degree of membership$u_j$of the fuzzy set$A_i$, with$i,j$=$1,2,\dots ,n$are linguistic values.

Definition 2.

Let $X\left(t\right), \left(t=1,2,\dots \right)$ , a subset of real numbers be the universe of discourse by which the fuzzy sets $f_i\left(t\right)$ are defined. If $F\left(t\right)$ is a collection of $f_1\left(t\right), f_2\left(t\right),\dots ,$ then, $F\left(t\right)$ is called an FTS defined on $X\left(t\right)$ .

Definition 3.

Given a chain of historical data $\left\{X_i\right\}$ and predictive values $\left\{X_i\right\},i=1,2,\dots ,n,$ we have the popular parameters to evaluate the built models as follows:

Mean squared error:

$$MSE=\frac{1}{n}{\displaystyle \sum }_{i=1}^n{({\widehat{X}}_i-x)}^2$$

(2)

Mean absolute error:

$$MAE=\frac{1}{n}{{\displaystyle \sum }}_{i=1}^n\left|{\widehat{X}}_i-X_i\right|$$

(3)

Mean absolute percentage error:

$$MAPE=\frac{1}{n}{{\displaystyle \sum }}_{i=1}^n\left(\frac{{\widehat{X}}_i-X_i}{X_i}100\right)$$

(4)

Symmetric mean absolute error:

$$SMAPE={\displaystyle \sum }_{i=1}^n\left(\frac{\left|{\widehat{X}}_i-X_i\right|}{\left(X_i+{\widehat{X}}_i\right)/2}100\right)$$

(5)

Mean absolute scaled error:

$$MASE=\frac{{\sum }_{i=1}^n\left|{\widehat{X}}_i-X_i\right|}{\frac{n}{n-1}{\sum }_{i=2}^n\left|X_i-X_{i-1}\right|}$$

(6)

Steps of the IFTS fuzzy model:

Step 1: Calculation of variations of data between two consecutive periods of time:

$$U_i=X_{i+1}-X_i,i=1,2,\dots ,n-1$$

(7)

Find the universal set $U$ containing the interval between Min $\left\{U_i\right\}$ and Max $\left\{U_i\right\}$.

Step 2: Divide the universal set $U$ into $m$ equal-length intervals $u_i, i=1,2,\dots ,m.$

Each interval $u_i$ contains the variation values corresponding to different data. The midpoints of each interval will be determined $\left(u_i^0, i=1,\dots ,m\right).$

Step 3: Fuzzification of time series data.

The fuzzy set $A_1,A_2,\dots ,A_n$ is defined on the universal set $U$ by the formula:

$$A_i=\left\{\frac{{\mu }_{A_i}\left(u_i\right)}{u_{i,}},u_i\in U,{\mu }_{A_i}\in \left[0,1\right]\right\}$$

(8)

with:

$${\mu }_{A_i}\left(u_i\right)=\frac{1}{1+{\left[C\times \left(U_i-u_i^0\right)\right]}^2},i=1,2,\dots ,m$$

(9)

where,

• $C$ is a constant and $C\in \left(0,1\right),$
• $U_i$ is the variation of data between two periods of time determined in Step 1, and,
• $u_i^0$ is the midpoint of each interval identified in Step 2.

Step 4: Selection of basis $w (1<w<n)$ corresponding to the intervals of prior time.

Calculation of fuzzy relationship based on the value $w$.

Establishment of operation matrix.

Step 5: Forecast for time $t$ using the formula:

$$\widehat{X}\left(t\right)=X\left(t-1\right)+V\left(t\right)$$

(10)

with:

$$V\left(t\right)=\frac{{\sum }_{i=1}^w{\mu }_t\left(u_i\right)\times u_m^i}{{\sum }_{i=1}^w{\mu }_t\left(u_i\right)}$$

(11)

where ${\mu }_t\left(u_i\right)$ is an element of $F_t$, $X\left(t-1\right)$ is the value at time $t-1,$ and $\widehat{X}\left(t\right)$ is the forecasting value at time t.

Calculation with statistical Software “R”:

All calculations are performed with the statistical Software R and the additional package “AnalyzeTS”.

The following steps need to be executed within the “AnalyzeTS” package of the statistical Software “R”:

• Determination of the value $w$, the number of elements used in the data set as prior information to forecast the data at $t=w+1$;
• Determination of the value $n$, which represents the number of fuzzy levels. The Sturges formula is used to calculate the required number of fuzzy levels $n=1+3.322 log$(n);
• Running the implemented DOC algorithm to obtain the C value;
• Computation of forecast.

3.1.2. Simple Moving Average (SMA)

Simple Moving Average (SMA) is a simple and reliable forecasting method from the R “smooth” package [45]. Moving averages are able to smooth fluctuations and identify trends. SMA can be used for very short data sets with only 1 parameter, the order of the moving average, to be estimated [71].

The Simple Moving Average from the “smooth” package for the statistical Software “R” constructs the AR model in the single source of error state space form as:

$$y_t=\frac{1}{n}{{\displaystyle \sum }}_{j=1}^n{y}_t-j$$

(12)

where the AR(n) process is modelled as:

$$y_t=w^{\prime }{v}_{t-1}+{\in }_t$$

(13)

$$v_t=F{v}_{t-1}+g{\in }_t$$

(14)

$v_t$ is the state vector.

The selection of the order of the moving average is based on the lowest AICc value.

3.1.3. Forecast Combination

Forecast combinations have been found to produce better forecasting results than individual forecasting methods [72,73,74]. Simple forecast combination methods tend to be more accurate than complex forecast combination methods; as these methods use equal weights, no weightings are learned.

In this paper, we use the simple average combination method (SA), with the equation written as follows.

Suppose $y_t$ is the variable of interest, there are N not perfectly collinear predictors:

$$f_t={\left(f_{1t},\dots ,f_{Nt}\right)}^{\prime }$$

(15)

The simple average gives equal weights to all predictors:

$$w=1/N$$

(16)

The combined forecast is then obtained by:

$${\widehat{y}}_t={\left(f_t\right)}^{\prime }w$$

(17)

3.1.4. Forecast Evaluation

Forecast errors are the difference between the actual value and predicted (forecasted) value. We calculated the difference between scale-dependent and scale-independent forecast errors. Scale-dependent errors are on the same scale as the data and cannot be compared with other data sets. As opposed, scale-independent errors can be compared across different data sets and are easy to understand and interpret [75].

In this paper we utilize MAPE (mean absolute percentage error) as scale-independent error:

$$\mathrm{MAPE}=\frac{100}{n}{\sum }_{i=1}^n\left|\frac{x_i-y_i}{x_i}\right|$$

(18)

where $y_i$ is the prediction and $x_i$ the true value.

Returned forecast errors values can be interpreted as:

• <10% is a highly accurate forecast;
• 10%~20% is a good forecast;
• 20%~50% is a reasonable forecast;
• 50% is an inaccurate forecast.

3.2. DEA Window

To measure the effectiveness of decision-making units (DMUs) across time while considering carry-over activities between different windows, a window DEA was created. Window analysis improves the chances of understanding how the efficiency level changes as a result of overlapping window sequences. This model uses a moving average analogue, which includes observations from the whole research period, to estimate the efficiency change over time. As a consequence, the efficiency ratings are achieved with findings that are more trustworthy, and accurate to a greater extent. The following is an explanation of the model [54,76].

It assumes that there are N decision-making units (n = 1, …, N), which are observed in T periods (t = 1, …, T) with r inputs used to generate s outputs. Therefore, the sample has N × T observations, and an observation n in period t, $DM{U}_t^n$ has an r-dimensional input vector $x_t^n={\left(x_{1t}^n,x_{2t}^n,\dots ,x_{rt}^n\right)}^{\prime }$ and an s-dimensional output vector.

$$y_t^n={\left(y_{1t}^n,y_{2t}^n,\dots ,y_{st}^n\right)}^{\prime }$$

(19)

The window starting at time k, $1\le k\le T$ with the width w, $1\le w\le T-k,$ is denoted by $k_w$ and has N × w observations. The matrix of inputs for this window analysis is written as:

$$X_{k_w}=\left(x_k^1,x_k^2,\dots ,x_k^N,x_{k+1}^1,x_{k+1}^2,\dots ,x_{k+1}^N,\dots ,x_{k+w}^1,x_{k+w}^2,\dots ,x_{k+w}^N\right)$$

(20)

The matrix of outputs is as follows:

$$Y_{k_w}=\left(y_k^1,y_k^2,\dots ,y_k^N,y_{k+1}^1,y_{k+1}^2,\dots ,y_{k+1}^N,\dots ,y_{k+w}^1,y_{k+w}^2,\dots ,y_{k+w}^N\right)$$

(21)

The input-oriented DEA window problem for $DM{U}_t^{\prime }$ under a constant return to scale (CRS) assumption, is subject to:

$$\begin{gathered} {\mathsf{\theta }}_{k_{w}t}^{\prime }=\underset{\mathsf{\theta },\lambda }{\min }\mathsf{\theta } \\ \mathrm{s}.\mathrm{t}. \\ -X_{k_w}\lambda +\mathsf{\theta }{x}_t^{\prime }\ge 0 \\ {Y}_{k_w}\lambda -y_t^{\prime }\ge 0 \\ {\lambda }_n\ge 0 (n=1,\dots ,N\times w \end{gathered}$$

(22)

4. A Case Study

4.1. Selection of DMUs

We examined the FDI attractiveness of 42 out of 63 provinces in Vietnam using data from 2012 to 2020 and projected values for 2021–2022. The twenty-one missing provinces were largely in the highland regions, where foreign investors are not the main focus of the economy. According to Figure 2 and

Table 1. The list of DMUs.

Province	DMU	Key Economic Region	Province	DMU	Key Economic Region
Bac Giang	FDI-1	Northern Midlands and Mountains	Binh Dinh	FDI-22	South Central Coast
Phu Tho	FDI-2	Northern Midlands and Mountains	Binh Thuan	FDI-23	South Central Coast
Quang Ninh	FDI-3	Northern Midlands and Mountains	Da Nang	FDI-24	South Central Coast
Thai Nguyen	FDI-4	Northern Midlands and Mountains	Khanh Hoa	FDI-25	South Central Coast
Tuyen Quang	FDI-5	Northern Midlands and Mountains	Ninh Thuan	FDI-26	South Central Coast
Yen Bai	FDI-6	Northern Midlands and Mountains	Phu Yen	FDI-27	South Central Coast
Bac Ninh	FDI-7	Red River Delta	Quang Nam	FDI-28	South Central Coast
Ha Nam	FDI-8	Red River Delta	Quang Ngai	FDI-29	South Central Coast
Ha Noi	FDI-9	Red River Delta	Lam Dong	FDI-30	Central Highlands
Hai Duong	FDI-10	Red River Delta	Binh Duong	FDI-31	South East
Hai Phong	FDI-11	Red River Delta	Binh Phuoc	FDI-32	South East
Hung Yen	FDI-12	Red River Delta	Dong Nai	FDI-33	South East
Nam Dinh	FDI-13	Red River Delta	Ho Chi Minh	FDI-34	South East
Ninh Binh	FDI-14	Red River Delta	Vung Tau	FDI-35	South East
Thai Binh	FDI-15	Red River Delta	An Giang	FDI-36	Mekong River Delta
Vinh Phuc	FDI-16	Red River Delta	Ben Tre	FDI-37	Mekong River Delta
Ha Tinh	FDI-17	North Central Coast	Can Tho	FDI-38	Mekong River Delta
Hue	FDI-18	North Central Coast	Hau Giang	FDI-39	Mekong River Delta
Nghe An	FDI-19	North Central Coast	Kien Giang	FDI-40	Mekong River Delta
Quang Binh	FDI-20	North Central Coast	Long An	FDI-41	Mekong River Delta
Thanh Hoa	FDI-21	North Central Coast	Tien Giang	FDI-42	Mekong River Delta

, the researched provinces were divided into the seven major economic areas of the nation.

4.2. Selection of Inputs and Outputs

The annual reports encompassing five provincial indicators—labor force (LF), gross regional domestic product (GDP), the Provincial Competitiveness Index (PCI), FDI by capital (cumulative FDI), and FDI by cases—were used to determine the aggregate data of inputs and outputs (cumulative cases). The definitions and information sources for the gathered indicators are displayed in Figure 3: Three inputs were used: labor force (LF), gross regional product (GDP), and the Provincial Competitiveness Index (PCI). Two outputs were obtained: FDI by capital (cumulative FDI), and FDI by cases (cumulative cases).

As in Figure 3, the definition of input and output factors are described as follows.

Input factors:

• Labor force (1000 people): the annual number of employees of a province;
• Gross regional domestic product (VND billion): the annual GDP of a province;
• Provincial Competitiveness Index (PCI): This indicator provides information on where a province excels in economic governance, as well as areas where improvement is needed to become more competitive and achieve greater socio-economic growth, including legal institutions, land rental prices, priority conditions, administrative documents, loans, and settlement of legal procedures, etc. The index also examines local efforts toward national green growth strategies in response to climate change.

Output factors:

• FDI by capital (USD million): the annual amount of FDI of a province;
• FDI by cases (USD million): the annual number of FDI projects of a province.

5. Results Analysis

5.1. Results of Forecasting

Forecast accuracy of input and output variables are presented in

Table 2. Forecast accuracy of input variables (unit: percentage).

Province	Forecast Errors—Labor Force			Forecast Errors—GDP			Forecast Errors—PCI
	MAPE Proposed Model	MAPE SMA	MAPE IFTS	MAPE Proposed Model	MAPE SMA	MAPE IFTS	MAPE Proposed Model	MAPE SMA	MAPE IFTS
Bac Giang	3.50	4.12	2.88	15.29	23.63	6.94	0.91	2.36	2.50
Phu Tho	2.63	1.11	4.16	0.68	10.58	9.22	1.79	4.31	4.46
Quang Ninh	0.41	1.89	2.70	17.45	25.12	9.79	3.37	5.05	1.69
Thai Nguyen	0.43	1.06	0.51	1.54	10.61	13.68	2.44	4.98	1.94
Tuyen Quang	1.11	1.93	0.88	0.38	10.65	11.42	1.63	1.98	1.42
Yen Bai	0.76	1.83	0.43	7.48	14.62	0.35	3.50	3.02	3.98
Bac Ninh	1.57	4.48	1.34	15.34	18.82	11.86	5.08	6.12	4.03
Ha Nam	0.83	2.19	3.85	10.30	18.74	1.87	1.46	2.32	1.87
Ha Noi	4.48	1.53	10.50	3.10	7.11	13.30	2.18	3.99	2.03
Hai Duong	2.95	0.97	4.94	6.29	14.92	2.34	1.70	3.48	1.82
Hai Phong	1.83	0.87	2.80	12.44	19.62	5.26	5.44	4.55	6.32
Hung Yen	0.41	0.71	0.37	24.37	32.15	16.59	6.89	9.45	4.33
Nam Dinh	1.00	0.92	1.09	12.40	20.00	4.81	1.70	1.67	1.88
Ninh Binh	0.95	1.31	0.59	16.93	24.55	9.32	3.09	2.06	5.71
Thai Binh	2.02	1.11	2.92	27.67	19.94	35.39	4.03	4.91	3.16
Vinh Phuc	1.67	1.27	2.60	16.81	21.05	12.57	2.14	2.20	3.20
Ha Tinh	1.06	1.18	0.94	16.08	19.97	12.20	2.28	1.56	3.00
Hue	0.45	0.36	1.04	10.10	10.91	9.29	2.18	7.51	3.15
Nghe An	0.91	0.38	1.77	14.18	21.73	6.63	2.60	2.42	2.78
Quang Binh	3.50	4.12	2.88	15.29	23.63	6.94	0.91	2.36	2.50
Thanh Hoa	0.59	0.96	0.50	19.64	27.27	12.00	2.21	2.12	5.15
Binh Dinh	0.81	0.72	0.91	2.22	8.89	13.33	2.96	3.20	3.44
Binh Thuan	0.46	1.17	2.09	17.60	25.01	10.18	4.37	7.86	1.48
Da Nang	0.81	3.03	4.66	6.78	15.27	9.20	4.69	2.48	5.85
Khanh Hoa	1.25	2.40	1.09	7.80	5.93	14.69	2.46	1.07	5.31
Ninh Thuan	1.83	2.66	0.99	8.74	19.91	2.42	1.49	3.03	1.74
Phu Yen	2.66	1.12	4.20	8.32	16.19	1.35	5.09	6.78	3.39
Quang Nam	0.92	1.08	2.93	6.76	4.24	17.76	3.30	2.75	3.85
Quang Ngai	0.52	0.53	0.52	2.43	11.12	6.26	1.32	1.69	1.92
Lam Dong	4.53	4.10	4.96	0.95	10.52	8.62	3.03	2.34	3.71
Binh Duong	3.05	8.58	2.47	13.69	23.01	4.38	4.32	3.86	4.77
Binh Phuoc	0.71	2.53	2.69	1.47	11.34	9.77	5.81	5.68	5.94
Dong Nai	0.56	2.42	3.54	10.10	17.60	2.60	4.34	6.53	2.15
Ho Chi Minh	2.14	4.27	0.92	12.19	1.59	25.98	2.06	1.71	2.41
Vung Tau	3.21	1.59	5.10	54.90	51.35	58.45	2.51	6.95	2.28
An Giang	3.38	6.65	2.09	2.95	9.03	3.12	2.95	9.03	3.12
Ben Tre	1.80	2.33	1.26	10.65	16.87	4.42	3.26	10.07	3.56
Can Tho	0.61	0.64	0.59	6.61	7.31	5.91	5.43	8.11	2.74
Hau Giang	3.00	2.82	3.25	2.33	11.76	7.11	1.91	5.44	1.75
Kien Giang	0.39	0.38	0.40	9.04	8.56	9.52	4.50	4.01	5.00
Long An	3.02	3.32	2.72	6.37	18.03	5.29	0.67	2.15	2.20
Tien Giang	0.63	2.31	3.20	7.32	16.25	3.24	1.94	1.70	5.58
Average	1.65	2.12	2.39	11.02	16.80	10.37	3.00	4.16	3.31

and

Table 3. Forecast accuracy of output variables (unit: percentage).

Province	Forecast Errors—FDI by Capital			Forecast Errors—FDI by Cases
	MAPE Proposed Model	MAPE SMA	MAPE IFTS	MAPE Proposed Model	MAPE SMA	MAPE IFTS
Bac Giang	13.68	25.14	4.11	5.28	18.58	29.14
Phu Tho	18.43	27.25	9.61	3.91	19.00	26.83
Quang Ninh	2.49	4.84	7.43	5.15	11.09	21.38
Thai Nguyen	2.54	8.92	3.84	9.49	15.55	34.53
Tuyen Quang	2.93	10.24	4.39	10.64	51.18	29.90
Yen Bai	1.31	2.02	4.64	2.97	12.87	11.47
Bac Ninh	2.36	10.86	15.59	6.45	17.24	30.13
Ha Nam	13.86	27.82	3.85	3.37	22.43	29.18
Ha Noi	5.91	1.53	17.18	4.21	17.21	25.64
Hai Duong	4.27	8.46	1.35	1.53	12.40	15.13
Hai Phong	5.44	9.29	20.17	6.66	12.61	25.93
Hung Yen	2.82	12.35	4.47	1.09	11.47	13.64
Nam Dinh	3.31	4.27	10.89	10.12	10.63	30.87
Ninh Binh	12.46	13.00	11.91	6.26	12.85	25.37
Thai Binh	2.43	11.99	7.43	9.13	11.70	29.95
Vinh Phuc	5.24	13.79	3.30	2.07	16.42	19.23
Ha Tinh	2.95	0.50	6.40	0.39	10.24	11.03
Hue	3.61	27.00	22.34	1.78	12.76	16.31
Nghe An	15.39	17.52	13.26	7.12	15.79	30.03
Quang Binh	13.68	25.14	4.11	5.28	18.58	29.14
Thanh Hoa	4.78	3.51	13.08	3.50	21.78	28.78
Binh Dinh	15.51	25.40	29.70	2.93	8.89	7.68
Binh Thuan	3.25	5.31	11.81	2.91	8.31	14.13
Da Nang	2.07	7.56	3.42	4.25	21.63	30.13
Khanh Hoa	5.33	19.47	8.81	1.21	6.93	9.35
Ninh Thuan	21.31	4.95	38.69	4.10	6.79	15.00
Phu Yen	3.47	10.87	8.66	1.41	1.04	1.78
Quang Nam	3.16	1.10	7.41	2.56	31.50	19.39
Quang Ngai	15.06	8.10	22.02	4.63	1.64	10.91
Lam Dong	2.57	5.04	1.14	1.27	0.49	2.86
Binh Duong	1.11	9.20	6.99	3.44	8.89	15.76
Binh Phuoc	8.46	19.88	2.96	6.98	22.14	36.11
Dong Nai	6.11	9.34	2.87	1.51	8.42	11.13
Ho Chi Minh	1.50	5.64	2.63	4.00	15.24	23.25
Vung Tau	4.67	12.27	2.93	2.12	14.47	18.70
An Giang	19.61	26.30	12.92	2.44	5.63	10.51
Ben Tre	16.11	18.52	16.08	2.70	4.69	10.09
Can Tho	2.57	9.95	11.68	1.62	4.14	0.89
Hau Giang	22.22	51.03	6.59	1.37	8.70	11.43
Kien Giang	5.77	1.67	13.21	2.62	14.23	19.47
Long An	14.43	14.39	13.70	4.01	13.66	21.67
Tien Giang	12.09	17.61	6.57	8.93	6.04	23.90
Average	7.77	13.07	10.00	4.13	13.47	19.71

, respectively.

For the labor force input variable, the difference of the error metrics between our forecasting methods was very small. On average, the MAPE for the forecast combination still performed better than the single methods of SMA and IFTS; namely, 1.65% for the forecast combination, 2.12% for SMA, and 2.39% for IFTS. For GDP input variable, the average MAPE for all 42 provinces was 11.02% with the SMA–IFTS forecast combination, 16.08% with SMA, and 10.37% with IFTS. On average, the IFTS method was almost on par with the forecast combination. This was attributed to the well-behaved nature of the GDP data, with IFTS showing a better performance than SMA. For the PCI input variable, 42 forecasts were more accurate than the single forecasting methods, with an average MAPE of 3.00% for the forecast combination, 4.16% for the SMA, and 3.31% for IFTS.

With regard to the output variables, for FDI by capital, the difference of the forecasting errors for each method was marginal as reflected by the average MAPE. The average MAPE for all 42 provinces was 7.77% with the SMA–IFTS forecast combination, 13.07% for the SMA method, and 10.00% for the IFTS method. For FDI by cases, the average MAPE for all 42 provinces was 4.13% with the SMA–IFTS forecast combination, 13.47% with the SMA method, and 19.71% with the IFTS method.

The average MAPE of 42 provinces is visualized in Figure 4. From the result, we can conclude that for all five data categories, the average MAPE for the forecast combination was lower compared with Simple Moving Average and IFTS as a single forecasting method. This forecasting method shows huge potential for batch forecasting and forecasting of short time series with fluctuating values over time.

An example of the curves of the proposed forecast model (the combined SMA and IFTS forecast model) for output variables including FDI by capital (total FDI) and FDI by cases (example of three biggest provinces: Ho Chi Minh, Da Nang, and Ha Noi) is depicted in Figure 5, below.

5.2. Results of DEA Window

Prior to the DEA window analysis, the Pearson correlation test was implemented to confirm inputs and outputs to obtain “Isotonicity”, meaning that, if the input quantity increased, the output quantity could not decrease under the same condition. As shown in

Table 4. Pearson matrix of input and output variables (2012–2022).

Variables	Labor Force	GDP	PCI	FDI by Capital	FDI by Cases
Labor Force	1
GDP	0.876 **	1
PCI	0.108 *	0.254 **	1
FDI by Capital	0.717 **	0.858 **	0.254 **	1
FDI by Cases	0.844 **	0.951 **	0.209 **	0.827 **	1

Note: ** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed).

on the Pearson correlation indexes for the whole period, including the historical period from 2012 to 2020 and the future period from 2021 to 2022, the inputs and outputs had a significantly positive relationship.

Table 5. FDI attractiveness efficiency of provinces (2012–2022).

Province	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022	2012–2022	Rank
Bac Giang	0.26	0.27	0.30	0.32	0.35	0.36	0.36	0.35	0.37	0.38	0.38	0.34	18
Phu Tho	0.16	0.16	0.16	0.15	0.17	0.17	0.17	0.19	0.21	0.21	0.22	0.18	28
Quang Ninh	0.27	0.27	0.28	0.25	0.25	0.23	0.22	0.19	0.20	0.21	0.21	0.23	25
Thai Nguyen	0.06	0.35	0.54	0.45	0.41	0.38	0.37	0.37	0.38	0.38	0.38	0.37	15
Tuyen Quang	0.05	0.04	0.04	0.04	0.03	0.03	0.03	0.04	0.04	0.04	0.04	0.04	41
Yen Bai	0.08	0.08	0.08	0.07	0.06	0.09	0.07	0.07	0.07	0.06	0.06	0.07	39
Bac Ninh	0.35	0.35	0.46	0.59	0.66	0.77	0.85	0.87	0.92	0.97	1.00	0.71	7
Ha Nam	0.18	0.21	0.25	0.33	0.39	0.38	0.41	0.45	0.48	0.51	0.52	0.37	14
Ha Noi	0.68	0.67	0.67	0.72	0.74	0.65	0.75	0.75	0.79	0.81	0.81	0.73	6
Hai Duong	0.48	0.48	0.47	0.51	0.49	0.49	0.43	0.40	0.42	0.42	0.42	0.46	12
Hai Phong	0.44	0.56	0.53	0.53	0.57	0.53	0.57	0.56	0.64	0.66	0.68	0.57	9
Hung Yen	0.44	0.45	0.49	0.53	0.50	0.47	0.46	0.37	0.38	0.38	0.38	0.44	13
Nam Dinh	0.07	0.07	0.08	0.12	0.12	0.25	0.24	0.21	0.23	0.23	0.24	0.17	30
Ninh Binh	0.14	0.15	0.14	0.16	0.16	0.15	0.14	0.13	0.13	0.13	0.13	0.14	31
Thai Binh	0.05	0.06	0.06	0.08	0.08	0.08	0.08	0.10	0.11	0.12	0.12	0.08	38
Vinh Phuc	0.26	0.26	0.29	0.32	0.30	0.31	0.32	0.32	0.34	0.35	0.36	0.31	20
Ha Tinh	1.00	1.00	0.86	0.82	0.92	0.86	0.75	0.64	0.64	0.63	0.62	0.80	5
Hue	0.28	0.28	0.27	0.34	0.28	0.26	0.33	0.34	0.33	0.34	0.34	0.31	21
Nghe An	0.11	0.11	0.10	0.10	0.10	0.10	0.09	0.09	0.09	0.09	0.09	0.10	37
Quang Binh	0.26	0.27	0.30	0.32	0.35	0.36	0.36	0.35	0.37	0.38	0.38	0.34	18
Thanh Hoa	0.42	0.53	0.49	0.46	0.44	0.51	0.48	0.45	0.47	0.46	0.46	0.47	11
Binh Dinh	0.10	0.16	0.15	0.14	0.07	0.08	0.08	0.07	0.08	0.08	0.08	0.10	36
Binh Thuan	0.26	0.41	0.38	0.34	0.32	0.30	0.23	0.24	0.24	0.25	0.25	0.29	22
Da Nang	0.45	0.45	0.43	0.43	0.46	0.45	0.50	0.57	0.62	0.67	0.69	0.52	10
Khanh Hoa	0.17	0.15	0.21	0.20	0.11	0.28	0.26	0.25	0.27	0.28	0.29	0.22	26
Ninh Thuan	0.28	0.28	0.25	0.26	0.26	0.27	0.30	0.27	0.24	0.23	0.22	0.26	24
Phu Yen	0.22	0.22	0.71	0.61	0.58	0.56	0.21	0.20	0.20	0.21	0.22	0.36	17
Quang Nam	0.48	0.43	0.40	0.39	0.35	0.33	0.32	0.31	0.33	0.33	0.33	0.36	16
Quang Ngai	0.31	0.28	0.27	0.29	0.09	0.11	0.12	0.11	0.12	0.12	0.12	0.17	29
Lam Dong	0.16	0.15	0.15	0.11	0.10	0.10	0.08	0.07	0.07	0.07	0.07	0.10	35
Binh Duong	1.00	0.98	0.97	1.00	1.00	1.00	1.00	0.99	0.98	1.00	1.00	0.99	1
Binh Phuoc	0.21	0.21	0.23	0.27	0.25	0.27	0.29	0.31	0.34	0.35	0.35	0.28	23
Dong Nai	0.80	0.73	0.77	0.83	0.86	0.84	0.86	0.88	0.90	0.91	0.92	0.84	4
Ho Chi Minh	0.84	0.88	0.91	1.00	1.00	0.94	0.93	0.95	0.97	0.99	1.00	0.95	2
Vung Tau	0.86	0.90	0.91	0.90	0.89	0.82	1.00	0.91	0.97	0.99	1.00	0.92	3
An Giang	0.02	0.02	0.03	0.03	0.02	0.02	0.02	0.02	0.02	0.02	0.02	0.02	42
Ben Tre	0.08	0.09	0.09	0.12	0.11	0.12	0.13	0.12	0.15	0.16	0.17	0.12	33
Can Tho	0.07	0.07	0.06	0.09	0.10	0.06	0.05	0.06	0.05	0.05	0.05	0.07	40
Hau Giang	0.15	0.15	0.14	0.22	0.13	0.12	0.07	0.08	0.08	0.09	0.09	0.12	34
Kien Giang	0.18	0.16	0.15	0.19	0.19	0.24	0.24	0.23	0.23	0.24	0.24	0.21	27
Long An	0.59	0.59	0.57	0.72	0.67	0.67	0.63	0.59	0.61	0.61	0.62	0.63	8
Tien Giang	0.11	0.12	0.12	0.12	0.15	0.15	0.14	0.14	0.15	0.15	0.16	0.14	32
Average	0.32	0.33	0.35	0.37	0.36	0.36	0.36	0.35	0.36	0.37	0.37	0.36

shows the FDI attractiveness efficiency scores of 42 provinces obtained from the DEA window for the whole time period (2012–2022). On the one hand, it was found that, on average, Binh Duong (0.99), Ho Chi Minh (0.95), and Vung Tau (0.92) ranked first, second, and third, respectively. These three provinces all belong to the South East region. According to statistics from the Foreign Investment Agency (Ministry of Planning and Investment), in the first quarter of 2022, Binh Duong led the country with a total investment capital of USD 2.32 billion, mainly in the fields of industrial and commercial products and services [77]. Due to a focus on implementing solutions to improve the investment environment and constantly improving competitiveness, Binh Duong has significantly improved in attracting FDI. Ho Chi Minh, which has a population of nine million, is the biggest city and largest consumer market in the nation and offers a skilled and dynamic workforce and a well-developed commodity distribution system. The city’s per capita income and purchasing power quadruple the national average [78]. Ho Chi Minh is always on the list of localities attracting the most FDI in the country, due to the attractive investment environment as well as the attention and effective support of the city government for foreign investors. Ho Chi Minh has also increased the application of information technology and digital technology to standardize and simplify business processes and correct the limitations and shortcomings of civil servants and public employees to improve quality, serve, and bring satisfaction to the FDI business community. Vung Tau is a “destination of FDI” because the province has a convenient location, has the largest deep water seaport in Vietnam, and has synchronous and smooth transportation, electricity, and water infrastructure systems. Importantly, Vung Tau has always aimed for a sustainable investment environment and open procedures [79].

On the other hand, Can Tho (0.07), Tuyen Quang (0.04), and An Giang (0.02) were found to be the least efficient localities to attract FDI during the research period, mainly because of their traffic problems: bottleneck traffic hampered the ability of industrial parks to attract investors, and the cities did not have a focus on developing a synchronous transport infrastructure development plan. Being the least efficient, means that these provinces have not used their resources to attract FDI, which is implied by having the poorest efficiency. There is, therefore, a large potential for these provinces to increase their performance in terms of efficiency. Instead of adding additional resources, the improvement might be achieved by using available resources more effectively.

Figure 6 is a diagrammatic view of the efficiencies’ evolution for each province. As can be seen, most of the provinces exhibited relatively stable performance over 2012–2022. This was expected since none of the five indicators could change dramatically in a short period of time. There were a few prominent instances that showed noticeable differences, namely, Bac Ninh, Thai Nguyen, Ha Tinh, and Phu Yen. Bac Ninh’s efficiency in attracting FDI has accelerated ever since 2012, when the first Samsung Electronics Vietnam (SEV) factory in Vietnam was established in Bac Ninh [80]. This was a historic decision that laid the foundation for the significant investment process of Samsung Vietnam. After more than a decade, of a total of nearly USD 19 billion of investment capital by Samsung in Vietnam, nearly half was invested in Bac Ninh, reaching more than USD 9.3 billion. Samsung chose Bac Ninh to build its first factory due to favorable factors of politics, economy, people, and geographical location; particularly, the province’s policies and support mechanisms removed difficulties in the process of investments, production, and businesses. In 2013, after focusing on investing in Bac Ninh and having achieved substantial success, Samsung officially decided to continue investing in Thai Nguyen province, which is why, with reference to Figure 6, the FDI attractiveness in Thai Nguyen rocketed from 2012 to 2013 then remained at good efficiency ever since. Total FDI in Thai Nguyen increased 70 times between 1993 and 2011, and 2012 to 2020, due to the spillover impact of the Samsung project [81]. In an opposite scenario, FDI attractiveness in Ha Tinh dropped gradually since 2012. This performance was very likely to have been hampered by the 2016 Vietnam marine life disaster (Formosa Steel, Ha Tinh) [82] and the delay in the Vung Ang II Thermal Power Plant project during 2017–2018 [83]. Figure 6 also depicts Phu Yen’s performance during the research period, which was the most unstable. The Vung Ro Refinery and Petrochemical Complex project may have been responsible for the province’s performance changing significantly between 2013–2014 and 2017–2018. The project undertaking began in 2014, when the factory’s construction was completed, however, the site development made little headway and was stuck for a while, until the project was finally abandoned in 2018 [84].

Figure 7 plots the distribution of the provinces’ efficiency scores. Based on Figure 6, over the period 2012–2022, most provinces were able to maintain their efficiency level. However, on average, three-quarters of provinces achieved low efficiency (less than 0.5), a quarter of provinces had middle efficiency (range from 0.5 to 1), and none obtained high efficiency (greater than 1). This pattern also appeared mostly in annual distributions.

To observe the FDI attractiveness efficiency regarding each key economic region,

Table 6. FDI attractiveness efficiency trend (2012–2022).

Key Economic Region	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022
Northern Midlands and Mountains	0.15	0.20	0.23	0.21	0.21	0.21	0.20	0.20	0.21	0.21	0.21
Red River Delta	0.31	0.32	0.34	0.39	0.40	0.41	0.43	0.42	0.44	0.46	0.47
North Central Coast	0.42	0.44	0.40	0.41	0.42	0.42	0.40	0.37	0.38	0.38	0.38
South Central Coast	0.28	0.30	0.35	0.33	0.28	0.30	0.25	0.25	0.26	0.27	0.27
Central Highlands	0.16	0.15	0.15	0.11	0.10	0.10	0.08	0.07	0.07	0.07	0.07
South East	0.74	0.74	0.76	0.80	0.80	0.77	0.81	0.81	0.83	0.85	0.85
Mekong River Delta	0.17	0.17	0.17	0.21	0.20	0.20	0.19	0.18	0.18	0.19	0.19

shows the efficiency indexes of the regions, while Figure 8 plots their FDI attractiveness’ evolution paths. As shown in Figure 8, all regions had stable trends of efficiency in attracting FDI over 2012–2022. The South East region achieved the highest pattern, followed by the Red River Delta and the North Central Coast. The Red River Delta not only ranked second in attracting FDI in the country after the South East region, but it is also considered as the place of convergence of many global brands from major world corporations such as Honda, Toyota, LG, Samsung, and Canon, to name a few [85]. Among them are projects with registered capital of up to several billion U.S. dollars, including: Samsung Display Vietnam, a Korean investment project in Bac Ninh province with a total registered capital of USD 6.5 billion, with the goal of manufacturing, assembling, processing, marketing, and selling electronic equipment screens; LG Display Hai Phong, a Korean investment project with a total registered capital of USD 4.65 billion; Smart City, a Japanese investment project in Hanoi, with a total registered capital of USD 4.138 billion; Samsung Technology Complex a Singaporean investment project in Bac Ninh, with a total registered capital of USD 2.5 billion; and Jaks Hai Duong Power Company Limited (Hai Duong BOT thermal power plant), a Hong Kong investment project with a total registered capital of USD 2258 billion, to name a few. In recent years, the North Central region is also gradually becoming a positive highlight in attracting FDI due to the continuous efforts of the provinces to complete infrastructure projects in a synchronous, modern, and cohesive manner to link with many regions and the world, taking the initiative in investment promotion, and issuing many attractive investment incentive policies, etc. In contrast, the Central Highlands region is the least efficient locality in attracting FDI because the investment environments of those provinces still have many shortcomings and difficulties, such as poorly developed infrastructure, lack of human resources, transparency, and administrative procedures. Therefore, provinces in this region must focus on building and implementing orientations and breakthrough solutions to improve the investment environment, to successfully and sustainably attract FDI in the future.

6. Discussions and Conclusions

The attraction of FDI to Vietnam has contributed to the achievement of multiple critical socio-economic development goals of the country. However, compared with preferential policies and advantageous resources, the attraction of FDI to the economy is still not commensurate, and remains uneven across regions. Particularly, in the context of the current COVID-19 pandemic, the assessment of FDI attraction to Vietnam is an important task and poses many problems that need to be resolved. In this paper, a novel hybrid model combining simple average combination of SMA (Simple Moving Average) IFTS (Improved Fuzzy Time Series), and DEA window analysis was proposed as a means to evaluate the current performance (2012–2020) and future performance (2021–2022) of the FDI attractiveness of Vietnamese provinces. Their efficiency in attracting FDI in different time periods was assessed by measurements covering five provincial dimensions: labor force (LF), gross regional domestic product (GDP), provincial competitiveness index (PCI), FDI by capital (cumulative FDI), and FDI by cases (cumulative cases).

To our best knowledge, the approach proposed in this research has never been reported in the existing literature, especially for evaluating FDI attractiveness in the Vietnam context. Accordingly, the main results are summarized as follows. As there is variation in FDI attractiveness among the provinces, there is huge potential to improve their performance by learning from successful models. Throughout the research periods, Binh Duong, Ho Chi Minh, and Vung Tau were found to be the localities most efficient at attracting FDI, while An Giang, Tuyen Quang, and Can Tho had the lowest FDI attractiveness indexes. Due to its efforts to improve its investment environment, vigorously implementing the goals of improving the quality of FDI inflows, along with many open mechanisms and policies and a series of practical solutions, the South East Region achieved the best FDI attractiveness. Within economic circles, policymakers in other provinces should learn from successful provinces with high FDI attractiveness. They can revise policies or exert their competitiveness to increase FDI attractiveness, new development patterns should be considered and governments should emphasize the development of knowledge-intensive industries, high-technology, advanced manufacturing and information technology, energy-saving, new energy, and modern service industries, to obtain investors’ attention and continue developing towards the goal of sustainability. Government should encourage neighboring provinces and regions to maximize their input factors across geographical boundaries and shift their attention to other areas, such as the Central Highlands. To ensure that factor inputs freely circulate among investors and provinces, local government can establish economic circles and policies. Additionally, investors should seize the opportunity to take advantage of provinces with potential markets. All in all, this paper can serve as an important guideline to provide insight for policymakers in the context of a developing country by investigating its FDI attractiveness.

However, this research has some limitations. First, some localities were missing in the investigation due to lack of data, so in the future, a complete research could be implemented to obtain a more detailed overall view. Second, more input and output variables can signify avenues for future studies, especially, novel factors related to the COVID-19 pandemic and environmental aspects. The use of other input and output variables, as well as employing other methods, may lead to different results and important findings. In terms of methods, for forecasting, Grey systems methods can be utilized for insufficient data, while multi-criteria decision-making techniques can be useful to evaluate efficiency performance.