A Global Entrepreneurship Efficiency Benchmarking and Comparison Study based on National Systems of Entrepreneurship and Early-Stage Business: A Data Envelopment Analysis Approach

National Systems of Innovation (NSI) literature comprised the holistic relationships toward interactive, iterative, and cumulative innovation processes at a country level (e.g., Freeman, 1987; Lundvall et al., 2002). NSI’s institutional and structural focus inspired policymakers with a framework-based performance evaluation (Nelson, 1993) and made stakeholders consider the influence of regulatory intervention in national entrepreneurship (Inacio Junior et al., 2021). As a result, entrepreneurship literature has strived to study systemic context and system-level conditions of entrepreneurial action (Ács et al., 2014).

National Systems of Entrepreneurship (NSE) concept is about institutional resource allocation closely related to the entrepreneurial attitudes, abilities, and aspirations of individuals leading to the creation and operation of new businesses (Ács et al., 2014; Tasnim & Afzal, 2018). In summary, the NSE depicts national capacity leading to entrepreneurial activities and presents the dynamic, inter-correlated relationships among the multi-dimensional economic, social, and institutional frameworks (Ács et al., 2014), and also helps explain the national economic capability for potential entrepreneurial activities (Ács, Szerb, et al. 2014; Inacio Junior et al., 2021; Tasnim & Afzal, 2018). Further recently, the NSE concept evolved into the “entrepreneurial ecosystem” concept (i.e., Szerb et al., 2019), which views the regional ecosystem may affect different types of entrepreneurial activities (Kirznerian or Schumpeterian). A healthier ecosystem grants greater capacity for higher business formation, while a weak ecosystem leads to innovation of entrepreneurs due to a lack of support policies or economic support (Szerb et al., 2019). NSE also represents the dynamism and interaction among individual attitudes, aspirations, and abilities that stimulate resource allocation via new business creation and operation (Ács et al., 2014). In a country, individuals pursue entrepreneurial opportunities and evaluate the desirability and feasibility of the pursuit, which is also determined by a particular country’s institutional (contextual) factors, such as distribution of available resources, social norms, and attitudes (Ács et al., 2014). Some NSE output indicators evaluate the degree of emergence for the newly born entrepreneurs within a given population, and the most widely used example is GEM’s total early-stage entrepreneurial activity rate, TEA (Autio, 2007). Global entrepreneurship indicators such as GEDI (Global Entrepreneurship and Development Index) and GEI (Global Entrepreneurship Index) deliver multi-faceted information on the NSE performance at a system level by profiling each NSE. However, these indicator-based approaches might be less valuable in contemplating the relationships between input factors and the productivity or efficiency of analyzed units (Inacio Junior et al., 2021). Also, considering the countries with the highest values in NSE indicators as benchmarks could lead to severe bias or failure in decision-making processes (Bogetoft, 2012).

Understanding the system performance requires a holistic approach (Inzelt, 2004; Kuhlmann, 2003), and quantifying the national systems may not be sufficient if not followed by the productivity analysis considering inputs and outputs (Edquist & Zabala-Iturriagagoitia, 2015; Inacio Junior et al., 2021). It is reasonable to identify the national system’s performance based on the efficiency concept, where a unit’s productivity is based on the rate of a total of outputs by a total of inputs (Ramanathan, 2003). Also, it contributes to the existing literature to evaluate and review if a particular country’s institutional condition leads to the high efficiency of entrepreneurial activities. This paper seeks to provide an empirical analysis of entrepreneurial efficiency based on the relational assumptions among the national framework conditions and individual entrepreneurial outcomes as hypothesized in the NSE concept.

Different studies suggested the methodological approach for efficiency comparison and analysis for global entrepreneurship. Bygrave et al. (2003) examined the correlation of entrepreneurship from a push-pull perspective in 29 countries participating in the GEM program, finding the relationship between opportunity-pull entrepreneurship and variables such as informal investment, entrepreneurial capacity, and perception of start-up opportunities. Sternberg and Wennekers (2005) provided empirical evidence that the role of entrepreneurial activity varies according to economic development stages. Also, they suggested that regional framework conditions, such as infrastructure and policies, need to be considered to investigate the regional entrepreneurial context. Lafuente et al. (2016), based on the data of 63 countries from 2012, tested the knowledge spillover theory of entrepreneurship using DEA, highlighting the relationship between NSE and knowledge spillovers leading to better efficiency. Tasnim and Afzal (2018) compared country-level efficiency based on the NSE perspective, hypothesizing that individual-level actions and country-level framework conditions interact. Das and Kundu (2019) compared countrywide efficiency based on GEDI and GEM datasets from 2012 to 2016 to understand small-sized enterprise performance. They utilized attitudinal parameters as independent variables and adopted dependent variables, such as the nascent entrepreneurship rate. Also, Lafuente et al. (2021) conducted an efficiency analysis using the non-parametric technique to figure out how country-level entrepreneurship relates to total factor productivity using global entrepreneurship data from 2002 to 2013. Faghih et al. (2021) investigated entrepreneurship “overall efficiency,” assessing national entrepreneurship attitude toward the entrepreneurship system, based on GEM 2018 data. Inacio Junior et al. (2021) conducted an efficiency analysis of the GEI data to demonstrate that the global entrepreneurial rankings may misinterpret NSE dynamics with the impression that countries showing higher national framework indicators are more productive in generating actual entrepreneurial activity. The authors claimed that there might be another perspective that prioritizes the efficiency of actual entrepreneurial activities. They pointed out that lower-positioned countries in framework conditions often present higher efficiency, producing more entrepreneurial business opportunities at the individual level.

The previous studies can be divided into three streams; (a) Research that dealt with the level of NSEs without efficiency analysis among countries (i.e., Ács et al., 2014), (b) Research that compared efficiency among countries without longitudinal or time-series consideration (e.g., Inacio Junior et al., 2021; Tasnim & Afzal, 2018), and (c) Research that needs further probe in setting input-output relationship (i.e., Faghih et al., 2021). In Faghih et al.’s efficiency study (2021), GDP is defined as the output of the overall entrepreneurial activity. Under the research setting, the author consequently prioritizes a country with less entrepreneurial activity (less input) and higher GDP (more output) as an efficient benchmark. If the researchers adopt this theoretical assumption, entrepreneurship-nurturing implications for developing economies with less-supportive institutional conditions might be limited. If a country’s NSE efficiency is defined by the total input-output rate (Ramanathan, 2003), the variable selection might be reconsidered (Inacio Junior et al., 2021). Consequently, providing objective logic for national entrepreneurial efficiency and criteria for global comparison benchmarks is expected to contribute valuable insights into fostering early-stage businesses.

Data envelopment analysis (DEA) was proposed by Charnes et al. (1978). DEA is a part of the linear planning methodology and can evaluate the relative efficiency of each DMU (decision-making unit) by applying multiple input and output factors (Casu & Molyneux, 2003). DEA is a non-parametric statistical technique used in various fields because it can be applied even when the measurement scale of each variable is different (Barros & Alves, 2004; Nguyen et al., 2016; Wu et al., 2009). DEA evaluates units that produce maximum (weighted sum) outputs at minimum (weighted sum) input levels as efficient DMUs (Smith & Street, 2005; Stone, 2002). It informs the level of improvement and decrease of the input variable, has the advantage of expressing the growth potential of the output variable numerically, and is used in various research fields (e.g., Nguyen et al., 2016; Wu et al., 2009).

DEA directly compares efficient and inefficient DMUs and provides information on which DMUs can be benchmarked to improve efficiency. It has two models, input-oriented and output-oriented. After fixing the output value, the former calculates the efficiency index based on the input factor values, focusing on the input reduction. The latter fixes the input factors and pays attention to how much the output can be improved. In addition, DEA has CCR (Charnes-Cooper-Rhodes) model (Charnes et al., 1978) that assumes the Constant Return to Scale (CRS) and BCC (Banker-Charnes-Cooper) model (Banker et al., 1984) based on Variable Return to Scale (VRS) assumption.

This study evaluated global entrepreneurship’s static and dynamic efficiency and explored the benchmarking countries for entrepreneurship efficiency. The analytical procedure is as follows. First, the data collection process chose the primary data for the research. Second, the pre-processing stage identified the input and output variables. Third, static efficiency analysis calculated the technical efficiency (TE), pure technical efficiency (PTE), and scale efficiency (SE) of DMUs. Fourth, dynamic efficiency analysis covered the longitudinal examination using the DEA/Window method. Fifth, the efficiency-stability matrix based on quadrant analysis was drawn, grouping the similarly-characterized countries. The overall research framework is presented below (see Figure 1).

This study utilized the Global Entrepreneurship Monitor (GEM) dataset from 2015 to 2020. GEM is a comparative study tracing entrepreneurial thoughts, behaviors, and activities having over 150,000 participants from over 50 countries (GEM, 2020). The research consists of Adult Population Survey (APS) and National Expert Survey (NES) and provides primary data on the competitiveness of each country’s NSE by surveying the adult population and experts (Das & Kundu, 2019). The countrywide dataset from 2015 to 2020 consisted of 335 NES and 320 APS instances. The 24 countries consecutively participated throughout the whole period (6 years, the time frame of this research) were chosen as DMUs, and a total of 144 data instances were finally collected for analysis. As for the DMUs, four countries were from Latin America and Caribbean (17%), fourteen were from Europe and North America (58%), two were from Africa (8%), and four were from Asia and Oceania (17%) (see Table 3).

This study adopted a computational analytical tool to eliminate human errors. Efficiency Measurement System (Scheel, 2000) was utilized as a reliable DEA statistical tool (Afzal & Lawrey, 2012). This package provides the CCR, BCC, DEA/Windows, and Malmquist analysis.

A country’s institutional factors (framework conditions) are known to be fundamentally correlated to formal structure/the norms of regulatory bodies and cultural/social practices while playing a decisive role in promoting individual entrepreneurial behaviors (Bianchi et al., 2015; Boudreaux et al., 2019; Bylund & McCaffrey, 2017; Sendra-Pons et al., 2022). Therefore, it is of theoretical value and interest to view the entrepreneurship phenomenon from an institutional theory standpoint regarding the remarkable contextual influence on promoting entrepreneurial activities stemming from the institutional environment (e.g., Bruton et al., 2010; Sendra-Pons et al., 2022; Sinha et al., 2019). The institutional theory assumes that the regulatory, social, and cultural elements affect organizations to promote legitimacy and be poised for survival (e.g., Ahlstrom & Bruton, 2003; Fang, 2010). The theory has been broadly accepted as a theoretical foundation in various research topics, recently widening its presence in studying determinant factors leading to the success of new entrepreneurial activities (e.g., Bruton et al., 2010; Peng, 2001). The institutional environment, such as accountability, effectiveness, and quality of related authorities, policies, and support organizations, may enhance or hinder the development of businesses (Sendra-Pons et al., 2022). In this regard, the paper hypothesizes and selects variables based on the assumptions from the institutional theory point of view to build an efficiency model. Determining valid inputs and outputs is substantial in analyzing efficiency (Chan & Karim, 2012). The following variables were selected to create the efficiency model. The variables were determined by the consensus of three researchers and 2 Ph.D. holders after a peer review for validity. Finally, three input variables and one output variable were selected (see Table 4).

(a) Total Early-stage Entrepreneurial Activity (TEA): TEA, a part of the GEM dataset, was selected as the output variable. GEM describes entrepreneurship activity in a country through a variety of indicators. TEA is defined as the proportion of those who participate in early-stage business activities among the adult population aged 18 to 64. GEM identifies entrepreneurs into four levels; (a) potential entrepreneur, (b) nascent entrepreneur (less than 3 months), (c) new business (less than 42 months), and (d) established business (more than 42 months). TEA represents the proportion of nascent entrepreneurs and new business owners among all entrepreneurs. TEA is a valuable benchmark for entrepreneurial activity and an indicator adopted by many economies to broadly gauge the degree of newly born entrepreneurs across the country (Wright, 2019). The emergence of new businesses could be viewed as an output measure under NSE (Ács et al., 2014), and there have been academic approaches that adopted TEA as a dependent variable to investigate its affecting factors or to compare entrepreneurial performance among countries (e.g., Micozzi, 2013; Rusu & Roman, 2017) based on the institutional theory. Thus, comparing the TEA level with national framework conditions might be a legitimate way to compare national entrepreneurship efficiency. Consequently, this study set TEA as output and analyzed the efficiency based on the weighted averaged inputs.

(a) GDP per capita: A country’s population and GDP are absolute indicators that could help understand its socioeconomic resources. Each country has a different institutional framework resulting from various aspects, and one of the significant elements is the economic development stage of the country (Eijdenberg et al., 2019). Some researchers viewed that entrepreneurship affected economic growth (e.g., Doran et al., 2018; Stoica et al., 2020), while other researchers presented a different way of understanding the causal direction between the two variables, hypothesizing that GDP is among the other factors that may significantly affect the emergence of the nascent entrepreneurs (e.g., Micozzi, 2013; Rusu & Roman, 2017). Being a country’s economic development stage, GDP could be considered a legitimate input factor in an efficiency analysis model. Additional empirical research shows that GDP per capita significantly impacts entrepreneurial activity (Shane, 2010). Since the research model adopts TEA, a ratio index, as an output variable, GDP per capita, which enables countrywide comparison, was chosen as an input variable. Except for Taiwan, the 2015-2020 GDPPC data were secured from the Worldbank (2021). Separate GDPPC data for Taiwan was obtained from IMF (2021).

(b) Established Business Ownership (EBO): EBO refers to the proportion of businesses that have been in operation for over 42 months and is calculated by the proportion of established business owners among the adult population aged 18 to 64. GEM recognizes the EBO rate as a part of the supportive environment leading to new businesses and a health indicator for a country’s entrepreneurial ecosystem (GEM, 2021). After establishing a new business, it is essential to continue and scale it. Comparing the EBO (input, the existing and established business fundamental) and TEA (output, new businesses) can signal how the outcome of the institutional environment is resilient or productive regardless of the current entrepreneurial context within the country (Wright, 2020), highlighting the country’s capacity to create new business opportunities. It is also considered one of the factors affecting new entrepreneurial activities (Almodóvar-González et al., 2020). By comparing EBO and TEA rates, the viability and vitality of a new business could be measured and translated into efficiency.

(c) National Expert Composite Index (NECI): GEM’s sub-item, National Expert Survey (NES), supports the NSE concept by evaluating expert opinions on national entrepreneurial conditions through a scale measure. Over 36 experts in 9 fields from each country participate in evaluating 12 items such as the economic environment, market access, government policies, and entrepreneurship education. This study calculated NECI by averaging the results of 11 items, excluding the recently included item, entrepreneurial post-school education. GEM considers NECI as a measure to gauge the easiness of beginning and developing a new business (GEM, 2020). Overall, NECI summarizes the evaluation results of a country’s entrepreneurial framework conditions into a composite score (GEM, 2020) and may help efficiently compare the processes or the mechanisms leading to entrepreneurial behaviors, such as TEA. As a result, NECI was selected as a legitimate input variable for the efficiency analysis model.

The total number of inputs and outputs in DEA should be limited to a minimum to avoid unrealistic evaluation results. The number of input and output factors is closely related to the number of DMUs, and for validity, the number of DMUs should be over twice the product of the number of input and output factors (Banker et al., 1984). This study chose three inputs, one output factor, and 24 DMUs, satisfying the condition.

The descriptive statistics of the variables are as follows (see Table 5). From the data for 24 countries from 2015 to 2020, the mean of GDPPC was 30,489 USD (min: 1,606, max: 116,654). For NECI, the mean value was 2.8 (min: 2.1, max: 3.6), and the EBO rate was 7.9% (min: 1.5, max: 20.3). TEA, the output variable, was 11.3% on average (min: 1.9, max: 36.7).

A strategic quadrant matrix analysis presented efficiency-stability axes based on dynamic efficiency results (see Figure 3). A model with four quadrants consisting of the X-axis (Mean, average = 0.471) and Y-axis (LDP, average = 0.266) was created. DMUs belonging to Quadrant 2 (High efficiency–Low stability) and 3 (Low efficiency–High stability) need to balance their framework conditions to pursue efficiency and stability. Seven countries are in Quadrant 2, and twelve are in Quadrant 3. Three countries (Guatemala, Brazil, and Croatia) located in Quadrant 1 with higher average entrepreneurial efficiency and stability are considered benchmarks for DMUs in other quadrants. DMUs in Quadrant 1 are considered to have shown higher-than-average entrepreneurial efficiency continuously regarding their framework conditions. The two countries included in Quadrant 4 display low efficiency and low stability. These countries may need attention and efforts at the national level to promote and foster entrepreneurship activities.