Unemployment and Enterprise Births in European Countries: A Sectoral Approach

Abstract

Different types of entrepreneurial activities are more or less conducive to socio-economic development. Among others, opportunity entrepreneurs are found to have a greater impact on economic growth, innovation, and employment compared to necessity entrepreneurs (pushed by the risk of being unemployed). In this context, the main goal of the study is to find the answer to the following research question: Which business sectors in European countries are sensitive to the unemployment push effect and thus more prone to necessity entrepreneurship? The paper provides further insight into the unemployment push hypothesis by investigating this issue in the group of 20 European countries across 11 business sectors (NACE Rev. 2 classification): manufacturing, construction, and services of business economies (nine sectors). The issue is examined by analyzing Eurostat and World Bank data for 2004–2020 using the panel vector autoregression (p-VAR) approach. The results confirm the unemployment push effect only in wholesaling, retailing, and repair services. The effect, however, seems to be triggered by entry behaviors in the Central Eastern European countries, which are more active in creating new businesses and more prone to necessity entrepreneurship. In addition, the interplay between unemployment and new enterprise dynamics in wholesaling, retailing, and repair services seems to be relatively more robust to existing heterogeneity within entrepreneurs, countries, and estimation techniques. The implications of the results could assist policymakers responsible for active labor market instruments promoting entrepreneurial attitudes.

1. Introduction

There are several obvious but widely recognized and empirically confirmed claims about entrepreneurship. First of all, entrepreneurial activity associated with creating economic and non-economic value through the leveraging of resources and capabilities [1,2] is an example of a very complex and multidimensional phenomenon [3,4,5,6,7]. It is reflected by the numerous categorizations of entrepreneurship types along such dimensions as: innovation spurs (innovative versus imitative/replicative entrepreneurs); dynamic of growth (a fast-growing business versus a slow-growing business); professional status (self-employed with employees, independent own-account workers, and dependent self-employed workers); the phase in the life cycle (start-ups vs. incumbent entrepreneurship); gender (female vs. male entrepreneurship); start-up motives (necessity versus opportunity entrepreneurs) [5]. Another source of entrepreneurship heterogeneity involves the different sectors in which entrepreneurial activity occurs (e.g., NACE classification).

Moreover, entrepreneurial activity manifests in many ways at various levels of observation (individual, enterprise, industry, local, regional, and national) [3,4,5,6,7,8]. Different types of entrepreneurial activities, however, might be more or less conducive to economic growth, social development, or technological change. Innovative entrepreneurs, advanced manufacturing and ICT services, self-employed with employees, and fast-growing businesses provide a greater economic contribution (in terms of job creation, innovation, earnings, and survival rate) [6,7,8,9,10].

Various motives for starting a business also lie behind the observed differences in socio-economic outcomes between entrepreneurs. Santarelli and Vivarelli [10] distinguish between progressive factors (e.g., profit expectations) and regressive factors (e.g., fear of being unemployed) for starting one’s own business. It correlates with the distinction between positive “pull” motivations to start a new business (e.g., need for achievement, seizing market opportunities) and negative “push” motivations to become an entrepreneur (risk of unemployment, family pressure) [11]. Pull (progressive) factors drive opportunity entrepreneurship, whereas push motivations (regressive factors) form the basis for necessity entrepreneurship. These “push” and “pull” factors are related not only to the individual characteristics of the potential founders but also to environmental (sectoral, macroeconomic) features [10]. Nevertheless, opportunity entrepreneurs are found to perform relatively well compared to other types of businesses, producing a greater impact on economic growth, innovation, and employment [12,13,14,15,16,17]. Therefore, the search for a deeper understanding of the factors that trigger and drive entrepreneurial activity constitutes an important field of research.

Among the various factors “pushing” individuals to start up a business, unemployment plays an important role [18,19,20,21,22]. It is captured by the unemployment push hypothesis, according to which increasing unemployment forces individuals to engage in entrepreneurial activity as a response to diminishing work opportunities [23]. Hence, at the macro and regional levels, the positive relationship between unemployment and setting up new enterprises is a proxy for necessity entrepreneurship [24,25]. This kind of relationship might affect the quality of entrepreneurship and, consequently, economic development. From a macro-level point of view, the entry of new firms facilitates a reallocation of production factors and structural adjustments in the economy. Economic development is stimulated when the creation of new companies triggers flows from the less productive to the more productive sectors [26]. Furthermore, a higher dynamic in new establishments may amplify boom periods or spur economic recovery during a recession [21]. Therefore, it is important to identify and explore the main forces (such as unemployment) behind firm dynamics and the quality of entrepreneurship in general and, in particular, in business sectors. Understanding how entrepreneurial activity in different business sectors responds to unemployment changes could help to design more effective entrepreneurship-supporting programs, in particular within active labor market policy [17].

The Purpose of the Study and the Research Hypothesis

The relationship between unemployment and new business dynamics is still ambiguous. Empirical investigations (e.g., [25,26,27,28,29,30,31]) so far have provided mixed results regarding the occurrence of the unemployment push effect. Further research is still underway, with different approaches and data from other countries to deepen the discussion.

The current study contributes to the nascent literature on the unemployment push hypothesis by investigating this issue in the group of 20 European countries (Austria, Bulgaria, the Czech Republic, Finland, Germany, Hungary, Estonia, France, Italy, Latvia, Luxemburg, the Netherlands, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden) (due to the availability of data for the entire period). Furthermore, the comparative analysis is enriched by considering the sectoral diversity of entrepreneurial activity, which provides further insight into the research subject. It is, to the author’s knowledge, the first empirical study verifying the unemployment push effect in such a group of European countries and such a wide range of business sectors: 11 business sectors (NACE Rev. 2 classification): manufacturing, construction, and services of business economies (nine sectors). Thus, the paper brings added value to the literature on the stochastic analysis of European companies in general.

The main goal of this study is to find the answer to the following research question: Which business sectors in European countries are sensitive to the unemployment push effect and thus more prone to necessity entrepreneurship? The research hypothesis assumes that the occurrence and intensity of the unemployment push effect vary by business sector. In particular, start-up dynamics in operational services are relatively more sensitive to the impact of unemployment. The issue is examined by analyzing Eurostat and World Bank data for 2004–2020 using the panel vector autoregression (p-VAR) approach.

The rest of the paper is organized as follows. The next section presents a brief literature review related to the unemployment push hypothesis. The following part informs about the research methods. Finally, the research results and their discussion, as well as the final conclusions, are presented.

2. Literature Review

Existing research suggests that the interplay between unemployment and entrepreneurial activity is manifold and spans a wide spectrum of explanations [10,18,26,32]. Not only does entrepreneurship impact unemployment, but also unemployment may have an impact on new business dynamics. This paper focuses on the latter, namely the impact of unemployment on enterprise births.

The relationship between unemployment and the creation of new enterprises is part of a broader debate on the factors that determine the establishment of new businesses. This issue can be studied from different theoretical viewpoints: the theory of Schumpeterian creative destruction; regional development theories; business cycle theories, including the dynamic stochastic general equilibrium framework; and the occupational choice theory [10,18,20,26,32]. This paper adopts the occupational view of entrepreneurship commonly used in studies on the relationship between unemployment and entrepreneurship [10,18,27,33].

According to the occupational choice model, people decide to become entrepreneurs if it appears to be more rewarding than being employed or unemployed [34]. The outcome of this cost-benefit analysis is influenced by the opportunity costs (e.g., wage level, unemployment benefits, entry costs) and the individual characteristics (abilities, skills). At the individual level, when the probability of being unemployed is relatively high (for example, during a recession, in lagged regions), starting your own business might be an interesting option [27]. This positive relationship between unemployment and new business dynamics is a manifestation of the unemployment push effect. The typically lower entrepreneurial skills and lower financial capital of the unemployed make them more able to enter business sectors with low entry barriers [18]. Given that the lower entrepreneurial skills of the unemployed may also manifest in their lower ability to identify market opportunities, a negative relationship between unemployment and the creation of new companies could also occur [35,36].

The negative links between unemployment and the creation of new enterprises are also grounded in macroeconomic theories, in particular in dynamic stochastic general equilibrium modeling. From this point of view, the main forces behind the formation of new firms are exogenous and take the form of positive exogenous productivity shocks [37,38]. During the market upturn (usually accompanied by lower unemployment rates), increasing market opportunities may encourage people to start a business [39]. In this theoretical framework, the formation of new firms is procyclical. This phenomenon is also known as the unemployment “pull” mechanism [26].

At the macro-level, the association between unemployment and entrepreneurship dynamics is also linked to the demand- and supply-side factors of economic growth. During a market downturn, the operating costs of businesses tend to decrease due to a greater supply of labor (rising unemployment) and lower labor costs. It may increase incentives to start a business, particularly in large-scale industries [21,40]. However, from the demand side, an increase in unemployment rates might decrease income and, consequently, consumer demand. It may undermine the willingness to start a new business and result in a negative correlation between unemployment and the formation of new firms [26].

The contradictory phenomena described by the unemployment “push” and “pull” motivations coexist in all industries. However, they do not always balance each other. The relative strength of the unemployment “push” and “pull” mechanisms is conditioned by:

• time, and more precisely, cyclical economic fluctuations: unemployment push mechanism plays a more significant role during an economic downturn: rising unemployment may cause more people to opt for having their own business for reasons of necessity [10,24];
• industry-specific characteristics, such as scale economies and the endowment of innovative capabilities: the unemployment push effect manifests primarily in small-scale businesses that have low capital and knowledge requirements (are easier to enter for the unemployed) [41,42,43];
• country-specific features: necessity-driven entrepreneurship is more frequent in countries/regions with a lower level of development, whereas opportunity entrepreneurship is recorded more often in countries/regions with higher levels of economic freedom and strong formal institutions [44,45,46].

There is substantial empirical literature on the relationship between unemployment and new firm formation. Two main approaches to investigating this issue may be distinguished:

• Exploring the microeconomic determinants of new enterprise formation: analyses focus on the microeconomic and individual characteristics of entrepreneurs that favor the occurrence of specific motives for starting a business. This type of analysis is mostly based on data provided by surveys among a sample of entrepreneurs (e.g., the Global Entrepreneurship Monitor). This research has so far been able to provide answers to questions on individual entrepreneurial features associated with opportunity (necessity) entrepreneurship and pull-push motivations. For example, van der Zwan et al. [11] and Källner and Nyström [47] indicate that opportunity entrepreneurs are more likely to be found among males, younger, wealthier, more aware of financial start-up support, and optimistic business owners;
• Investigating the macro-level, exogenous determinants of new enterprise formation: analyses are focused on the relationship between macroeconomic determinants and the process of new enterprise formation [25,26,27,28,29,43,44,47,48,49,50]. Unemployment is among the explanatory variables that are frequently identified as significantly correlated with entrepreneurship [23,48]. The empirical investigations of this relationship are based on econometric panel models, such as the fixed/random effect panel approach or the panel vector autoregression models (p-VAR).

The current study adopts the latter research approach and focuses on sectoral variation in new business formation caused by unemployment rates in European countries.

The multifaceted effect of unemployment on entrepreneurship, as well as the heterogeneous nature of entrepreneurship itself, may explain why empirical studies in this field have so far led to ambiguous conclusions. While some researchers confirm the unemployment push effect (e.g., Konon et al. [21] and Fritsch et al. [27] in Germany, Failrie [51] in the USA, Baptista and Preto [14] in Portugal, Wosiek et al. [24] in Poland, Dvoulety [28] in the Czech regions; Arauzo-Carod and Teruel-Carrizosa [52] in Spanish municipalities, Caree et al. in Belgian municipalities [18], and O’Leary in higher performing European regions [19]), other studies indicate that such an effect is not significant (e.g., Hajek et al. [29] in the Czech Republic, Santarelli and Tran [53] in Vietnam, Naudé et al. [54] in South Africa, Calá [30] in Argentina, and Carree [55] in the USA), or even works in reverse (e.g., Carree et al. [25] in Italian provinces or O’Leary in lower performing European regions [19]).

The ambiguous results on the links between unemployment and new enterprise formation highlight the need for further specific empirical investigation. As is clear from the above list of studies, empirical analyses are mostly based on a single country (at the regional or local level) and consider the overall entry rate. Some of them adjust for industry structure [18,21,24,25,55]; however, generally only a few types of business activity are taken into account. Carree and Dejardin [18] focus on retail and consumer services, Konon et al. [21] distinguish between small-scale and large-scale innovative and non-innovative industries, Wosiek et al. [24] consider different branches of the service sector, and Caree et al. [55] analyze retail and service industries with low entry barriers. This research provides further insights into the unemployment push hypothesis by investigating the effects of unemployment on new enterprise dynamics in European countries across various types of entrepreneurial activities (to control both for cross-country and entrepreneurship heterogeneity).

3. Methods

In relation to the main goal of the study, the research hypothesis assumes that the occurrence and intensity of the unemployment push effect vary by business sector. In particular, start-up dynamics in operational services are relatively more sensitive to the impact of unemployment. Operational services are represented by wholesale and retail trade, repair of motor vehicles and motorcycles, transportation and storage, and accommodation and food service activities. These service businesses are characterized by lower entry barriers and costs, which may result in their higher responsiveness to unemployment changes.

As the unemployment push effect refers to the formation of new enterprises, as a proxy for this process in European countries, the number of enterprise births per 1000 persons in the labor force (15–64 years) was adopted (Y_i) [19]. According to Eurostat: An enterprise birth occurs when an enterprise starts from scratch and begins operations, amounting to the creation of a combination of production factors with the restriction that no other enterprises are involved in the event. An enterprise birth occurs when new production factors, in particular new jobs, are created [https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:Enterprise_birth] (accessed on 21 October 2022). To capture the heterogeneity of business activity, the enterprise births were analyzed separately in the specific business sectors according to a statistical classification of the economic activities in the European Community (NACE Rev. 2). The following business sectors were considered: manufacturing (Y₁); construction (Y₂); wholesale and retail trade; repair of motor vehicles and motorcycles (Y₃); transportation and storage (Y₄); accommodation and food service activities (Y₅); information and communication (Y₆); financial and insurance activities (Y₇); real estate activities (Y₈); professional, scientific, and technical activities (Y₉); administrative and support service activities (Y₁₀); Information and Communication Technology (Y₁₁).

Unemployment changes (U) were measured through unemployment rates (LFS survey, in the population from 15 to 64 years). In the regression equation, this variable was included in the logarithmic form to improve its distribution in terms of normal distribution.

As is clear from the literature review, there is a two-way relationship between unemployment and entrepreneurship that might cause an endogeneity problem in the regression analysis. The results of the Durbin and Wu–Hausman tests for endogeneity (p < 0.05) confirm this problem for the majority of variables (except Y₆–Y₈). The endogeneity problem is not fully solved by typical regression models. Therefore, in this study, a panel vector autoregression model (p-VAR) was applied, which is designed for the presence of endogenous variables [56]. In the p-VAR approach, all variables are typically treated as endogenous and interdependent. The current level of each variable is explained by its own past and by the past observations of the other variables. In addition, the inclusion of the lagged dependent variable as a regressor minimizes the risk of an omitted variable problem.

The following p-VAR system of linear equations was specified:

$$Z_{j,t=}{\mu }_j+\left(L\right)Z_{j,t}+f_j+{\epsilon }_{j,t}$$

(1)

where:

-

${\mu }_j$—vector of constant terms for each variable,

-

(L)—lag operator,

-

Z_j,t—vector of endogenous variables (Y_i, U), where Y_i—number of enterprise births per 1000 persons (15–64 years) in the specific business sector (i); U—unemployment rate. To reflect that new enterprise formation is not only determined by the unemployment rate but also by the current economic situation, the GDP growth rate (GDP_gr) was also included in the regression equation as a robustness check;

-

f—fixed effect;

-

${\epsilon }_{j,t}$—vector of residuals;

-

i, j and t—type of business activity, country and time, respectively.

A relevant requirement for p-VAR estimations is for each of the endogenous variables to be stationary. As the Breitung and CADF Pesaran test results indicate that the Y_i variables have a unit root, they were included in the regression equation in the first difference [57].

The choice of an optimal lag order in p-VAR was based on the MMSC criteria for the generalized method of moments (GMM) models based on Hansen’s J statistic of over-identifying restrictions [58]. For most of the business sectors, the second-order p-VAR model was preferred (only for operational services, Y₃–Y₅, the first-order p-VAR model has a preference). For comparability reasons, the first- and second-order p-VAR equations were estimated.

The panel consists of heterogeneous European countries. The results of the Hausman test (p < 0.05) confirm that fixed effect modeling is appropriate. Introducing fixed effects allows for capturing the unobserved individual heterogeneity of countries, resulting in improved consistency of estimators [56]. The panel-specific fixed effects were removed using the forward orthogonal deviation.

The panel GMM technique was used as the estimation method, where lagged regressors were used as instruments in order to estimate the coefficients more consistently. Post-estimation tests were also performed: Hansen’s J test for over-identifying restrictions (to assess the consistency of the estimations); the Granger causality Wald test (whether unemployment Granger-causes enterprise births); and the stability condition of the estimated p-VAR (by calculating the modulus of each eigenvalue of the estimated model).

In addition, to deal with the problem of heteroscedasticity and other types of misspecification, robust standard errors were applied. Moreover, the Generalized Method of Moments (GMM) was used as it allows efficient estimation in the presence of heteroscedasticity of unknown form [59].

The adopted research procedure is an example of a macro-scale view of the process of establishing new enterprises. In this approach, the microeconomic determinants of starting a business, as well as other macroeconomic factors of entrepreneurship (apart from unemployment and economic growth), are neglected. Moreover, the present study focuses on the short-term, one-way effects of unemployment on new business formation, as a longer time frame is required to capture the feedback effect that newly established businesses have on employment [17,27]. The proposed analysis, however, is able to verify the links between unemployment and enterprise births.

4. Results

Table 1. Enterprise births per 1000 persons (15–64 years), unemployment rate, and GDP growth rate (average 2004–2020).

Country	Y1	Y2	Y3	Y4	Y5	Y6	Y7	Y8	Y9	Y10	Y11	U	GDPgr
Mean (average 2004–2020)
Bulgaria	0.62	0.65	3.70	0.55	0.79	0.32	0.11	0.56	0.99	0.24	0.26	8.7	2.7
Czech Republic	1.81	1.85	3.39	0.36	0.79	0.52	0.36	0.70	2.06	0.32	0.43	5.5	2.4
Germany	0.21	0.53	0.94	0.19	0.43	0.25	0.18	0.38	0.82	0.38	0.21	6.4	1.1
Estonia	0.80	1.41	2.66	0.74	0.37	0.82	0.23	0.58	1.66	1.09	0.72	8.4	2.8
Spain	0.39	1.62	2.35	0.47	1.12	0.24	0.27	0.47	1.17	0.74	0.23	16.9	0.7
France	0.46	1.45	2.00	0.47	0.58	0.51	0.16	0.41	1.62	0.64	0.43	9.0	0.8
Italy	0.53	1.25	1.96	0.20	0.60	0.25	0.19	0.28	1.58	0.45	0.24	9.6	−0.4
Latvia	0.85	1.21	2.27	0.69	0.36	0.64	0.23	1.04	1.66	0.77	0.59	11.0	2.6
Luxembourg	0.11	0.87	1.70	0.27	0.70	0.62	0.31	0.84	1.69	0.55	0.59	5.4	2.4
Hungary	0.53	1.23	1.89	0.40	0.51	0.60	0.77	0.55	1.51	0.90	0.50	7.4	1.8
Netherlands	0.42	1.38	2.16	0.38	0.46	0.84	0.22	0.22	2.91	0.73	0.70	5.2	1.2
Austria	0.25	0.51	1.21	0.27	0.64	0.36	0.12	0.14	1.12	0.43	0.30	5.3	1.1
Poland	0.92	1.85	2.81	0.70	0.41	0.52	0.36	0.18	1.20	0.46	0.49	9.1	3.7
Portugal	0.86	1.55	3.66	0.36	1.93	0.40	0.20	0.61	2.12	6.22	0.34	10.4	0.2
Romania	0.47	0.69	1.87	0.48	0.31	0.34	0.05	0.15	0.74	0.30	0.33	6.6	3.6
Slovenia	0.86	1.44	1.63	0.45	1.03	0.69	0.14	0.20	2.30	0.66	0.58	6.9	1.9
Slovakia	2.10	3.02	3.06	0.63	0.51	0.67	0.39	0.40	2.00	1.08	0.55	11.7	3.4
Finland	0.43	1.12	1.45	0.31	0.34	0.44	0.18	0.81	1.40	0.55	0.37	8.2	1.1
Sweden	0.44	1.13	1.60	0.28	0.43	0.79	0.11	0.51	2.04	0.57	0.63	7.5	1.9
Norway	0.33	1.51	1.18	0.47	0.25	0.60	0.08	0.78	1.48	0.70	0.46	3.7	1.4
Descriptive statistics (N = 340)
Mean	0.67	1.31	2.17	0.43	0.63	0.52	0.23	0.49	1.60	0.89	0.45	8.1	1.8
Median	0.49	1.22	1.94	0.36	0.51	0.46	0.18	0.44	1.51	0.58	0.39	7.4	2.2
Max	3.43	5.28	4.86	1.58	3.10	1.84	1.43	2.45	3.73	8.20	1.63	26.2	12.0
Min	0.06	0.28	0.64	0.13	0.09	0.11	0.00	0.07	0.31	0.09	0.10	2.1	−14.6
Std. Dev.	0.53	0.68	0.89	0.24	0.42	0.27	0.19	0.31	0.67	1.30	0.23	3.9	3.6
Vs	0.79	0.52	0.41	0.55	0.67	0.51	0.83	0.63	0.42	1.46	0.51	0.5	2.0
Skewness	2.13	1.66	0.71	1.28	2.67	1.15	2.41	1.43	0.75	3,95	1.28	1.61	−0.91
Kurtosis	8.15	8.28	2.79	4.92	12.5	5.28	10.6	7.85	3.16	18.3	5.87	6.49	5.75
Jarque-Bera	633.3	550.9	29.5	145.5	1693	149	1144	449	31,9	4215	210.4	319.5	153.7

Source: own calculations based on Eurostat data: Business demography by legal form (NACE Rev. 2) [BD_9AC_L_FORM]; Unemployment rates [lfsa_urgaed]; World Development Indicators [GDP growth annual; NY.GDP.MKTP.KD.ZG].

presents the descriptive statistics regarding enterprise births, unemployment rates, and GDP growth in the sample of 20 European countries.

In 2004–2020, start-up activity in Europe varied across countries, business sectors, and over time (Table 1). The new business dynamics were strongly related to the type of business activity (according to the mean values). Most of the new businesses were established in services of business economy (6.97), mainly in wholesale and retail trade (2.17), professional, scientific, and technical activities (1.60), and administrative and support service activities (0.89). Thus, in the European countries, entrepreneurial activity has been developing dynamically not only in operational services (wholesaling, retailing, and repair services), but also in those intensively using knowledge. In manufacturing, the average number of business births (per 1000 persons 15–64 years) was lower, at about 0.67, and in construction, it was 1.31. These numbers confirm that the service sector constitutes a fundamental segment of the European economy and nowadays is a crucial factor shaping the relationship between entrepreneurship and the labor market [60].

At the same time, the formation of new enterprises in the most popular business sectors (trade and repair services (Y₃), administrative and support service activities (Y₁₀)) had the highest volatility (based on the standard deviation values), while the lowest volatility was characteristic of the number of new companies in financial and insurance activities (Y₇). It should be noted that the observed fluctuations among the variables representing the economic situation (unemployment, GDP growth) were greater.

Regarding the skewness values, only the GDP growth rate was skewed to the left, whereas the rest of the variables were skewed to the right. According to the kurtosis values, all variables had leptokurtic distributions (above 3). In addition, the Jarque-Bera test indicates that all variables were not normally distributed.

Considering the spatial dispersion, the numbers of enterprise births were higher in the countries of Central and Eastern Europe, the Baltic States, and Southern Europe. The average annual number of business births in 2004–2020 (per 1000 working-age population) was about 1.054. The exceptions to this rule are Bulgaria (0.85) and Romania (0.54), both with lower levels of new establishments. Relatively fewer new businesses were also formed in Western Europe and in the Scandinavian countries. The lowest numbers were observed in Austria (0.51) and Germany (0.43). These observations are in line with previous studies [61,62] reporting that less developed economies display a greater tendency to create new businesses as a response to the dearth of alternative employment opportunities.

Considering the main business sectors (services, manufacturing, and construction), among the countries with relatively higher numbers of new enterprises in services, the following should be listed: Portugal, Slovakia, the Czech Republic, Estonia, and the Netherlands. This group of countries was distinguished by a higher number of business births, especially in operational services (Y₃–Y₅). Among the leading countries in setting up new enterprises in knowledge-intensive services (KIS; Y₆–Y₁₁) were Estonia, the Netherlands, Sweden, Slovenia, the Czech Republic, and Portugal. In the case of manufacturing activities, relatively higher numbers of new enterprises were observed in Slovakia, the Czech Republic, Poland, and Portugal, and in construction in Slovakia, the Czech Republic, Poland, and Spain (Table 1).

The number of enterprise births also fluctuated over time. These changes were to a different extent related to unemployment and the GDP growth rate, depending on the type of business activity (

Table 2. Correlation matrix between variables.

Variable	Y1	Y2	Y3	Y4	Y5	Y6	Y7	Y8	Y9	Y10	Y11
U	0.098	0.009	0.344	0.014	0.235	−0.301	0.078	−0.055	−0.149	0.128	−0.272
(p-value)	0.071	0.866	0.000	0.801	0.000	0.000	0.152	0.313	0.006	0.018	0.000
GDPgr	0.116	0.057	0.090	0.041	−0.099	−0.013	−0.024	−0.011	−0.105	−0.118	−0.002
(p-value)	0.033	0.292	0.099	0.455	0.068	0.810	0.663	0.838	0.053	0.030	0.974

The correlation between unemployment and the GDP growth rate: −0.109 (p = 0.045). * Correlation is significant at the 0.05 level (marked in bold). Source: own calculations based on Eurostat data.

).

In the analyzed European countries, the unemployment rate could shape the process of new enterprise formation in such business activities as wholesale and retail trade; repair of motor vehicles and motorcycles (Y₃); accommodation and food service activities (Y₅); information and communication (Y₆); professional, scientific, and technical activities (Y₉); administrative and support service activities (Y₁₀); and Information and Communication Technology (Y₁₁). However, the analysis only revealed positive correlations for trade and repair services (Y₃), accommodation and food services (Y₅), and administrative and support service activities (Y₁₀) that are consistent with the unemployment push effect. Thus, these service activities might be more exposed to necessity entrepreneurship. In addition, the GDP growth rate played an important role in establishing new manufacturing companies and administrative and support service businesses. Specifically, higher economic growth was conducive to expanding manufacturing businesses.

Moreover, the correlation analysis revealed that the correlation values were relatively low, indicating that multicollinearity would not affect the estimated results.

The p-VAR estimation results (est. 1a–11a) partially support the prior conclusions (

Table 3. First-order panel VAR estimates of Equation (1).

Estimation	est1a	est2a	est3a	est4a	est5a	est6a	est7a	est8a	est9a	est10a	est11a
Variable	dY1	dY2	dY3	dY4	dY5	dY6	dY7	dY8	dY9	dY10	dY11
L1.dYi	0.087	0.204	−0.135	0.27	0.194	0.325	0.048	0.294	0.383	−0.209	0.611 *
	(0.192)	(0.168)	(0.210)	(0.25)	(0.281)	(0.232)	(0.197)	(0.199)	(0.332)	(0.348)	(0.321)
L1.lnU	0.136	0.194	0.291 **	−0.009	0.203 ***	(0.027)	0.032	−0.002	0.262 **	0.219 **	0.03
	(0.086)	(0.151)	(0.146)	(0.57)	(0.058)	0.047	(0.034)	(0.071)	(0.131)	(0.092)	(0.043)
Variable	lnU
L1.dYi	−0.223 *	−0.32 ***	−0.29 ***	−0.713	2.071 **	1.8 **	−4.49 ***	−0.818	0.015	−0.277	3.38
	(0.115)	(0.081)	(0.10)	(0.433)	(0.841)	(0.754)	1.38	(0.39)	(0.103)	(0.174)	(1.2)
L1.lnU	0.902 ***	0.891 ***	0.93 ***	0.914 ***	1.02 ***	0.971	1.03 ***	0.952 ***	0.995 ***	0.922 ***	1.02 ***
	(0.061)	(0.6)	(0.08)	(0.109)	(0.109)	(0.101)	0.133	(0.078)	(0.061)	(0.067)	(0.13)
Descriptive statistics (N = 280)
J	22.26	23.62	17.6	7.05	11.85	17.16	13.07	15.82	22.22	19.69	13.57
J_pval	0.03	0.02	0.13	0.85	0.46	0.14	0.36	0.2	0.04	0.07	0.33

*, **, *** indicate statistical significance: * p < 0.1; ** p < 0.05; *** p < 0.01; robust standard errors in parentheses. All models meet stability conditions of panel VAR estimates (the modulus of each eigenvalue of the estimated model is less than one). Source: own calculations.

). A rising unemployment rate contributed, ceteris paribus, to a subsequent increase of enterprise births in the trade and repair services (est. 3a), accommodation and food services (est. 5a) and administrative and support service activities (est. 10a). The results are statistically significant, satisfy the stability condition, and the estimations seem to be consistent, as indicated by Hansen’s J test (p > 0.05).

A positive relationship was also noted between unemployment and the number of enterprise births in professional, scientific, and technical activities (est. 9a). In the case of the Y₉ variable, the low value of Hansen’s J test (p < 0.05) indicates a possible misspecification in the model. A similar problem appeared in estimations 1a and 2a. To solve this problem, two-year lags of the endogenous variables were included in the p-VAR estimations. This estimation technique did not result, however, in a satisfactory improvement of Hansen’s J statistic. Hence, in the next step, the third endogenous variable, the GDP growth rate, was included in the p-VAR model. This finally brought satisfactory results of Hansen’s J test for all estimations (est. 1b–11b) (

Table 4. Second-order panel VAR estimates of Equation (1).

Variable	est1b	est2b	est3b	est4b	est5b	est6b	est7b	est8b	est9b	est10b	est11b
	dY1	dY2	dY3	dY4	dY5	dY6	dY7	dY8	dY9	dY10	dY11
L1.dYi	−0.89 ***	−1.3 ***	−0.32	−0.43	0.26	−0.51 ***	−0.11	−0.39	−0.08	−0.95 **	−0.49 **
	(0.34)	(0.4)	(0.27)	(0.28)	(0.48)	(0.19)	(0.32)	(0.36)	(0.24)	(0.29)	(0.24)
L2.dYi	−0.54 ***	−0.47 ***	−0.22 *	−0.3 **	−0.08	−0.45 ***	−0.41 **	−0.03	−0.28 **	−0.27	−0.4 ***
	(0.19)	(0.24)	(0.11)	(0.14)	(0.24)	(0.09)	(0.18)	(0.25)	(0.12)	(0.2)	(0.11)
L1.lnU	−0.497	−0.569	1.368 ***	−0.132	0.183	−0.031	0.001	0.106	0.849	−0.087	0.04
	(0.36)	(0.7)	(0.59)	(0.19)	(0.2)	(0.16)	(0.13)	(0.18)	(0.55)	(0.59)	(0.16)
L2.lnU	−0.558	0.697	−1.163 *	−0.038	−0.027	0.056	−0.04	−0.098	−0.699	0.377	−0.017
	(0.38)	(0.7)	(0.62)	(0.17)	(0.2)	(0.17)	(0.15)	(0.19)	(0.57)	(0.63)	(0.17)
L1.GDPgr	−0.03 *	0.06	0.06	0.005	−0.01	0.001	0.003	0.01	0.02	0.001	0.003
	(0.02)	0.04	(0.03)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.03)	(0.04)	(0.01)
L2.GDPgr	−0.01	−0.01	−0.003	−0.004	0.01 *	−0.004	−0.001	−0.001	0.01	−0.002	−0.002
	(0.01)	0.01	(0.01)	(0.003)	(0.004)	(0.004)	(0.003)	(0.01)	(0.01)	(0.01)	(0.002)
lnU
L1.Yi	0.38	0.24	0.27	1.18 ***	0.95	1.33 ***	0.35	0.3	0.29 ***	0.86 ***	1.59 ***
	(0.27)	(0.15)	(0.18)	(0.45)	(0.79)	(0.37)	(0.9)	(0.27)	(0.11)	(0.25)	(0.47)
L2.Yi	0.18	0.03	0.16 **	0.55 ***	1.19 **	0.49 ***	1.31	0.01	0.13 ***	0.3 ***	0.66 ***
	(0.14)	(0.07)	(0.06)	(0.19)	(0.52)	(0.18)	(0.73)	(0.17)	(0.05)	(0.11)	(0.21)
L1.U	1.55	1.42 ***	1.12 ***	1.66 ***	1.79 ***	1.64 ***	1.08 ***	1.22 ***	1.39 ***	1.59 ***	1.78 ***
	(0.22)	(0.26)	(0.28)	(0.34)	(0.4)	(0.26)	(0.28)	(0.22)	(0.23)	(0.32)	(0.27)
L2.U	−0.72	−0.69 ***	−0.34	−0.57 *	−0.91 **	−0.8 ***	−0.25	−0.5 **	−0.57 **	−0.8 **	−0.97 ***
	(0.25)	(0.26)	(0.29)	(0.33)	(0.42)	(0.28)	(0.32)	(0.23)	(0.26)	(0.34)	(0.29)
L1.GDPgr	−0.001	−0.02	−0.02 *	−0.02 **	0.02	−0.02 *	−0.03	−0.02	−0.02 **	−0.02	−0.01
	(0.01)	(0.01)	(0.01)	(0.01)	(0.02)	(0.01)	(0.02)	(0.01)	(0.009)	(0.02)	(0.01)
L2.GDPgr	0.01	0.01	0.014 **	0.02 ***	0.013 **	0.02 ***	0.01	0.01	0.02 ***	0.02 **	0.02 ***
	(0.01)	(0.01)	(0.007)	(0.007)	(0.006)	(0.01)	(0.01)	(0.01)	(0.006)	(0.01)	(0.01)
GDPgr
L1.Yi	−6.49	−6.05	−8.32 *	−24.31 **	0.71	−21.59 **	13.11	−6.2	−4.03 *	−18.8 ***	−28.06 **
	(6.79)	(3.77)	(4.73)	(9.39)	(14.67)	(8.48)	(22.58)	(6.75)	(2.39)	(5.24)	(11.5)
L2.Yi	−4.36	−1.22	−4.88 ***	−13.5 ***	−20.77 *	−13.4 ***	−26.93	0.32	−3.6 ***	−6.29 **	−17.3 ***
	(3.65)	(2.004)	(1.78)	(4.59)	(9.76)	(4.34)	(17.09)	(4.5)3	(1.09)	(2.45)	(5.27)
L1.U	−3.2	1.54	11.88	−4.33	−2.48	−1.46	7.73	8.26	1.49	−3.26	−5.25
	(5.32)	(6.37)	(7.8)	(7.11)	(7.85)	(6.003)	(5.82)	(5.14)	(5.07)	(7.53)	(6.3)
L2.U	10.08 *	7.43	−3.06	6.2	6.74	8.83	0.27	0.36	6.01	11.32	12.84 *
	(5.88)	(6.03)	(8.01)	(6.61)	(8.15)	(6.37)	(6.77)	(5.25)	(5.42)	(7.97)	(6.64)
L1.GDPgr	0.01	0.6 *	0.82 *	0.58 **	−0.15	0.52 *	0.7 *	0.56 **	0.55 **	0.64	0.33
	(0.29)	(0.32)	(0.4)	(0.24)	(0.37)	(0.29)	(0.37)	(0.27)	(0.26)	(0.44)	(0.3)
L2.GDP gr	−0.07	−0.03	−0.15	−0.27 *	−0.11	−0.34 **	−0.11	0.02	−0.26 **	−0.3	−0.39 **
	(0.15)	(0.15)	(0.18)	(0.14)	(0.15)	(0.15)	(0.19)	(0.14	(0.13)	(0.19)	(0.15)
Descriptive statistics (N = 260)
J	18.84	20.79	16.94	20.13	14.08	19.71	21.86	14.75	20.17	22.79	17.4
J_pval	0.4	0.29	0.53	0.33	0.72	0.35	0.24	0.68	0.32	0.2	0.5
G_pval	0.348	0.537	0.03	0.052	0.075	0.841	0.527	0.843	0.111	0.097	0.791

*, **, *** indicate statistical significance: * p < 0.1; ** p < 0.05; *** p < 0.01; robust standard errors in parentheses. All models meet stability conditions of panel VAR estimates (the modulus of each eigenvalue of the estimated model is less than one). Source: own calculations.

).

The results of the analyses are sensitive to the inclusion of the additional control variable (GDP growth rate) and applying the second-order p-VAR model. The unemployment push effect was only confirmed for trade and repair activities (est. 3b). The push effect of unemployment on enterprise births occurs with a one-year lag. Considering a 2-year lag, unemployment hampers the establishment of new enterprises. However, this effect is weaker than the previous stimulus. Altogether, the results reveal that unemployment in Granger causes enterprise births in the trade and repair services (p = 0.03).

The previous estimation results for the accommodation and food services and administrative and support service activities were not robust. The analyses (est. 5b, 10b) do not confirm that unemployment affects the number of establishments in these business sectors.

Altogether, new firm births in manufacturing (est. 1a, 1b), construction (est. 2a, 2b), and most of the knowledge-intensive services (est. 6a−9a, 11a, 6b−9b, 11b) in European countries seem to be ambivalent towards unemployment changes.

5. Discussion

The results are consistent with the previous findings reporting a positive effect of the unemployment rate on the subsequent new enterprise formation in small-scale operational services (e.g., [18,27,42,43]). Specifically, the unemployment push hypothesis in wholesaling, retailing, and repair services was confirmed in Poland [24] and in Belgian municipalities [18]. Moreover, in Italy and the USA, the commerce sector was found to be more likely to be affected by unemployment, but the effect was not statistically significant for Italy [25] and very weak in the USA [55].

Other studies [25] also do not support the unemployment push hypothesis in the hotel and restaurant sector. They demonstrate, however, the negative effects of unemployment on entry into this sector in Italian provinces in 1997–2003 [25]. Nevertheless, entrepreneurship in the accommodation and food service industry is strongly related to location factors (tourism attractiveness) [63], which might be of greater importance than unemployment. Thus, sector-specific factors may mitigate the effect of the unemployment rate on new enterprise births in the accommodation and food service industry.

Another example of the impact of industry-specific factors could be construction (est. 2a, 2b) not influenced by unemployment changes. A specific attribute of entrepreneurial activity in the construction industry is its relatively greater exposure to the shadow economy [64,65]. Thus, institutional factors and legislation play an important role in business dynamics in this sector. For example, the introduction of subsidies for the repair and restoration of buildings in Italy forced several construction firms to switch from the shadow economy to the official economy [25]. A similar effect was observed in Poland as a consequence of VAT digitization and the tightening of the VAT tax system in 2017–2019 [65]. These observations, however, are not in line with the previous studies that have shown a positive (in the USA) [66] or negative (in Italy) [25] effect of unemployment on the formation of new firms in construction.

Furthermore, the results regarding the new firm births in manufacturing, and most of the knowledge intensive services, are also partially confirmed by the previous studies. According to Konon et al. [21] in German regions in 1995–2013, new firm formation in large-size manufacturing companies, including innovative manufacturing enterprises, was influenced by changes in the cyclical component of unemployment. Such an effect, however, was not observed in financial services. The opposite results were observed in the Italian provinces by Carree et al. [25]. They found no significant relationship between unemployment and manufacturing and transport, but a negative effect in financial services. In addition, Plehn and Dujowich [66] noticed the unemployment push effect in professional and business services and in financial services in the USA.

Horta et al. [67] and Audretsch et al. [20] shed some light on the ambiguous relationship between unemployment and the creation of new enterprises in knowledge-intensive and high-tech industries. They highlight the importance of unemployment structure in terms of the human capital of the unemployed (skilled unemployment versus unskilled unemployment). Although the unemployment rate negatively affected the creation of academic spin-offs in Italy [67] and knowledge-intensive firms in Germany [20], the higher share of “skilled unemployment” had a positive impact on the foundation of knowledge-intensive firms. Thus, the structural characteristics of unemployment also have an important impact on start-up activity and the unemployment push effect.

There are a few other explanations behind the mixed empirical findings for manufacturing, construction, and knowledge-intensive services. First, the effect of unemployment on business dynamics is not only sector- but also country-dependent [48]. Although the unemployment push effect was revealed for all the analyzed European countries, the positive Ganger casual effect of unemployment on entrepreneurship in trade and repair services was demonstrated mainly in the group of Central-East European and Baltic states (p = 0.039). In the subsample of Western Europe and Scandinavian countries, unemployment did not Granger-cause enterprise births (p = 0.963). This may be a consequence of the higher rate of necessity entrepreneurs among the self-employed in the Central-East European countries (ca. 26–27%) in comparison to the Western European countries (15.1%) or the Scandinavian states (6.4%) (Figure 1).

Second, the lack of clear-cut results concerning the relationship between unemployment and new enterprise dynamics might be linked to the existing heterogeneity within entrepreneurs. Van Steel et al. [5] and Caliendo and Kritikos [7] suggest that, apart from opportunity and necessity entrepreneurs, there is a need to distinguish hybrid opportunity-necessity entrepreneurs, who are simultaneously driven by both opportunity and necessity motives to start a business. The hybrid opportunity-necessity entrepreneurs constitute ca. 10–20% of the European self-employed (Figure 1).

Third, different studies apply different entrepreneurship measures (number of new business registrations, total early-stage entrepreneurial activity (TEA), gross entry rate, net entry rate, and enterprise births) and use different estimation techniques. Such empirical heterogeneity (model uncertainty) is a reason for the inconsistent empirical findings [48]. It could be stated that the interplay between unemployment and the new enterprise dynamics in trade and repair services (in comparison to other business sectors) is relatively more robust due to the existing heterogeneity within entrepreneurs, countries, and estimation techniques.

6. Conclusions

The establishment of new enterprises between 2004 and 2020 in Europe varied by country, business sector, and time. In addition, the current research confirms the existence of different responses of newly established enterprises to unemployment changes. Considering the heterogeneity of countries and business sectors, only the findings for wholesale and retail trade; the repair of motor vehicles and motorcycle businesses (Y₃) could be considered robust. The results reveal the unemployment push effect in this service sector, manifested by a positive relationship between rising unemployment and a subsequent increase in the number of enterprise births.

This conclusion, however, is not valid for other representatives of the operational services: transportation and storage, and accommodation and food service activities. While there was no statistically significant relationship between unemployment and enterprise births in transportation, the estimation results for hotel and restaurant businesses are mixed. Thus, the research hypothesis was only partially confirmed. In addition, the unemployment push effect was not observed in manufacturing, construction, and knowledge-intensive services. The numbers of newly formed businesses in these sectors were not affected by unemployment rate changes.

These findings contribute to the nascent literature on the relationship between unemployment and entrepreneurial activity but also add to it by being the first empirical attempt (to the author’s knowledge) to verify in comparative analysis the unemployment push effect in such a group of European countries and such a wide range of business sectors.

The results have some valuable implications for policy. First, the unemployment push effect might be viewed as a mechanism in the macroeconomic adjustment processes, particularly on the labor market. Among a variety of business sectors, retail, wholesale, and repair services are significantly related to the unemployment changes that affect the current state of the labor market. The results indicate the stabilizing effect of this service sector on the European economy as these service businesses react counter-cyclically. Although operational services have a smaller growth potential, they might play an important role in alleviating tensions in the labor market (in particular during a recession) compared to other industry businesses. On the other hand, however, the results suggest a large presence of necessity entrepreneurship within retailing, wholesaling, and repair services. As a result, the stabilizing effect of this service sector may be temporary.

Second, the research results support the importance of the sectoral approach in exploring the effect of unemployment on new firm establishments. From this point of view, policymakers should pay more attention to the type of business sector when designing programs to address unemployment through entrepreneurship (e.g., through self-employment grants). This might be important for active labor market policy instruments and their outcomes. From this perspective, self-employment grants for unemployed individuals are conducive to establishing businesses in traditional operational services. Even if self-employment grants could be an effective instrument in terms of labor market reintegration (they encourage unemployed individuals to start new enterprises), their impact on socio-economic development and innovation (due to the quality of entrepreneurship) could be limited. To increase the effectiveness of policy initiatives that promote self-employment among the unemployed, a more nuanced policy approach would be needed. In particular, the heterogeneity of the business sectors and the unemployed should be more fully taken into account (the entrepreneurship-supporting grants or trainings addressed to the highly qualified unemployed offer a greater chance for new businesses in knowledge-intensive industries).

Although a driving role of unemployment in new business formation in the trade and repair services was observed in the group of all analyzed European countries, the effect seems to be triggered by entry behaviors in the Central Eastern European countries, which were more active in creating new businesses and more prone to necessity entrepreneurship. This issue requires further investigation and constitutes a field for further research.

Although the paper confirms the validity of the sectoral approach in exploring the role of unemployment as a factor in new enterprise dynamics, it has some limitations. An important limitation is that the study focuses on the short-term, one-way effects of unemployment on new business formation and neglects other (micro-, macro-, industry-specific) factors of starting a business (apart from unemployment and economic growth). Furthermore, the results might be affected by the adequacy of the proxy used for entrepreneurial activity and data limitations. Therefore, more research is needed to determine whether alternative indicators of entrepreneurial activity and the inclusion of other European countries reinforce the robustness of the results. Moreover, it would be interesting to take into account not only the sectoral heterogeneity of the new businesses but also the heterogeneity of the unemployed (e.g., skill structure or share of long-term unemployment) and size classes of the new businesses (solo self-employed versus self-employed persons with employees).