COVID 19 and regional entrepreneurship: which regional features can mitigate the devastating effects of COVID-19?

ABSTRACT

Given the assumption that entrepreneurship is context-dependent and embedded in spatial features, the severity of the impact of the COVID-19 crisis on entrepreneurship may vary significantly across regions. This study, which is the first in the literature, categorizes 81 NUTS III level regions in Türkiye according to their entrepreneurship levels and examines how and to what extent regional characteristics determine entrepreneurship patterns, using MANOVA-ANOVA and Discriminant Function Analysis (DFA). Inspired by the resilience literature, it tries to reveal which regional attributes are more effective in keeping entrepreneurship resilient against shocks by comparing the pre-crisis (2012) and crisis periods (2020). Revealing critical differences between the pre-crisis and crisis periods, the article shows that population density, positive migration, high education levels, sectoral diversity, government incentives and natural geographical features, such as altitude and precipitation, are vital for regional entrepreneurship to remain resilient against COVID-19. Additionally, high population growth, young population rate, financial capital and income level, developed production, innovation, transportation and communication infrastructure combined with strong international networks and suitable climatic conditions are critical determinants that encourage regional entrepreneurship. On the contrary, specialization in certain sectors appears to be a regional weakness.

KEYWORDS:

• Pre-Crisis (2012) and Crisis Period (COVID-19)
• regional entrepreneurship
• demand-side and supply-side determinants
• political and natural environmental factors

SUBJECT CLASSIFICATION CODES:

• I18
• L26
• R11

Disclosure statement

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

Notes

1 First, the problem of multicollinearity was controlled, and no correlation greater than 0.8 was detected between variables (see Appendix Table 2). Second, the normality test revealed that all data were between −2 and +2 Kurtisos and Skewness values, thus having a normal distribution (Tabachnick and Fidell 2013). Third, homogeneity of variance was tested. Since the Levene’s test value was below the 0.05 significance level, the assumption of homogeneity of variances was violated. Therefore, the Games-Howell procedure was used as the Post Hoc method, as suggested by Andy Field (2013). Finally, Box's M test was used to ensure the homogeneity of the variance/covariance assumption. Homogeneous matrices hypothesis was rejected because p values were less than 5% significance level. Violation of this assumption is not a major problem for DFA and MANOVA, as there is sufficient sample size, as Field (2013) suggests. Therefore, Pillai's Trace values are used instead of Wilks' Lambda values when reporting MANOVA.