The Determinants of SME Success in the Long Run: An Ecosystem Perspective

ABSTRACT

Small and medium-sized enterprises (SMEs) drive economic growth especially in developing countries. Nonetheless, it is empirically challenging to identify the key attributes that predict the long-run success of SMEs. We analyse a detailed nationwide innovation survey of 4,075 companies in Korea, as well as their performance metrics, and demystify the key features that predict outperformance in the long horizon. We draw conditional causal inferences for SMEs by utilizing coarsened exact matching. Both the least absolute shrinkage and selection operator regularization technique and the elastic net method unveil that, among a variety of factors that include innovative activities, government support, and other external factors in the SME ecosystem, in-house research and development (R&D) are the most important factors for the success of SMEs. Government support in the form of product purchase plans or government-wide acquisition contracts also matters for SME success, but the association is transient and short-lived. Hence, accumulating intangible capital by internalizing R&D activities is of the utmost importance for SME success in the long run.

KEYWORDS:

• Small and medium-sized enterprises
• innovation
• government support
• feature engineering
• machine learning

JEL CLASSIFICATION:

• L25
• L26
• O30

Acknowledgement

This study was supported by the Research Program funded by Seoul National University of Science and Technology (SeoulTech).

Disclosure statement

No potential competing interest was reported by the author(s).

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Notes

¹ https://www.nationalgeographic.org/encyclopedia/ecosystem.

² Refer to Sexton and Bowman (1985), Chandler and Jansen (1992), Robinson and Sexton (1994), and Watson et al. (1998) for further details.

³ See Malecki (2018) for further details.

⁴ For example, Van Gelderen, Thurik, and Bosma (2006) analyses 517 companies for three years after the establishment. Using logistic regression, they document that that risk perception of the market plays a role to explain SME/venture success.

⁵ The raw data set is available at the following link: www.stepi.re.kr/kis/service/sub02_data_application.do.

⁶ The survey questions are identical to the community innovation survey (CIS) carried out by the Statistical Office of the European Union (Eurostat). The survey questions are in the Appendix.

⁷ The CIS 2014 can be found out at the following link: https://ec.europa.eu/eurostat/cache/metadata/en/inn_cis9_esms.htm. The link to the STEPI survey in Korean is as follows: https://www.stepi.re.kr/kis/index.do.

⁸ The Bank of Korea reports that the total number of companies in Korea in 2014 is 530,641 and that the number of manufacturing companies is 122,097. The STEPI innovation survey choose1s 4,075 firms from 122,097 companies in a random manner.

⁹ In Korea, SMEs are those firms that satisfy the following two criteria: 1) the size of assets in the balance sheet is less than KRW 150 billion (approximately equal to USD 134 million) and 2) the sales amount in the income statement is less than the average sales in the industry that the company belongs to.

¹⁰ The STEPI surveys conducted before 2014 are not consistent with Eurostat’s community innovation surveys.

¹¹ We utilize the KIS Value database in Korea.

¹² IBK is a leading bank in the SME sector in Korea.

¹³ Propensity score analysis is another widely used approach when econometricians need to assess treatment effects from non-experimental data. However, there has been a sharp debate on the caveats of propensity score matching, e.g. Guo, Fraser, and Chen (2020) in that ‘propensity score matching may accomplish the opposite of its intended goal-increasing imbalance, inefficiency, model dependence, and bias’ (Guo, Fraser, and Chen 2020, 463).

¹⁴ The $L_1$ norm ${\left|\mathrm{X}\right|}_1$ of a vector $\mathrm{X}$ is ${\left|\mathrm{X}\right|}_1=\sum _{i=1}^n\left|{\mathrm{X}}_{\mathrm{i}}\right|$.

¹⁵ Empirical studies (e.Ahn 2022; Ahn and Ahn 2023; James et al. 2013) have shown that the elastic net method can outweigh the LASSO method on data with correlated explanatory variables.

¹⁶ As a robustness check, we draw a new set of results using sales growth, asset growth, and return on assets. We confirm that the key results remain unchanged.

¹⁷ We use 3-digit industry classification codes to control for the industry fixed effect.

¹⁸ A kitchen sink regression is a typical regression with a long list of explanatory variables.

¹⁹ We define in-house R&D as the number of in-house R&D staffs over the total number of employees. We do not use R&D expenditure from income statement.

²⁰ The complementary evolution dynamics of intangible capital through R&D has been discussed in Hall and Hayashi (1989), Klette (1996), Hall (2002), and Ahn (2019), among others.

²¹ In line with the findings, it should be noted that two institutions play crucial roles in helping SMEs further develop innovative pathways in the U.S.: The U.S. small business innovation research (SBIR) program and the small business technology transfer (STTR) program. SBIR aims to ‘provide funding for some of the best early-stage innovation ideas – ideas that, however promising, are still too high risk for private investors, including venture capital firms’ (Roland Tibbetts, the founder of SBIR). STTR is an initiative to help SMEs have greater access to funding in the federal R&D arena. This type of government supports is indeed prevalent across countries. The U.K. government also provides SMEs with public procurement through ‘GovTech Catalyst’, which utilizes innovative digital technology to solve social issues. A similar line of government support policies has been implemented in Israel, China, France, and Germany, among others.

²² https://www.worldbank.org/en/topic/smefinance.