Abstract
This study aims to estimate the net effect of China’s entrepreneurship policy on back-home migrant workers since 2015. Using survey data from the provinces of; Zhejiang, Henan, and Guizhou with propensity score matching to control selection bias, the study indicates that the overall entrepreneurship policy has a significant positive impact on the probability of entrepreneurial entry. The poverty alleviation effect of the overall policy is greater than its industrial development effect, while the employment effect is not significant. The infrastructure policy’s effect on employment, industrial development, and poverty alleviation is greater than the other policies, and the financial policy’s effect on these three aspects is not significant. This study contributes to the existing literature by addressing a well-defined gap regarding the net effect of entrepreneurship policy on entrepreneurship entry and provides well-supported and plausible explanations for the inconsistencies among previous studies on policy effect indicators. As a practical contribution, this study provides China and other developing economies with compelling empirical evidence to support the design and amendment of entrepreneurship policy at the national level.
Introduction
Evaluating the effectiveness of public policy can be accomplished by an assessment of how well the policy objectives have been met. Known as the ex-post evaluation, this process is typically done after the policy implementation has been completed and the results are apparent. Effectiveness evaluation is the most meaningful among the three kinds of evaluation, which include prior, in-progress, and ex-post, and is the best way to test a policy’s effectiveness, efficiency, and benefit as well as the basis for policy adjustment or suspension. Importantly, ex-post evaluation is frequently used as the basis for future resource allocation. However, public policy evaluation is often difficult to complete due to political conflicts of interest, financial budget constraints, methodological difficulty in identifying the policy causation, difficulty in obtaining policy data, and the diversity of policy impact.
The same difficulties exist in assessing entrepreneurial policies. Due to the lack of large samples and high-quality data, most of the existing effectiveness assessments of public policies on entrepreneurship are descriptive (Mirzanti, et al., 2018; Goetz, et al., 2010), and some studies attempt to make up for the data deficiencies by constructing new evaluation models (Dewi and Hanifah, 2022; Smallbone, 2020). The literature on evaluating the effect of entrepreneurship policies with methodological rigor is sparse (Rigby and Ramlogan, 2016). There appear to be no relevant studies in the recent literature on the net effect of entrepreneurship policies on “new venture creation,” the most important objective of entrepreneurship policies. Moreover, as presented in the literature review, the very limited existing research on the impact of entrepreneurial policies on the performance of new ventures has not reached a consensus, which means that there is a need for further exploration in this area.
We address these gaps by evaluating the effect of China’s policies for back-home migrant workers’ entrepreneurship (BHMWE).Footnote1 There are several methodological advantages in evaluating China’s policies for BHMWE. Firstly, there is a large policy portfolio covering the entire process of preparation, implementation, and development of entrepreneurship which makes it possible to evaluate the net effect of the policies on entrepreneurship entry, and of the main policy instruments on nascent firms’ economic performance as well. Secondly, most of the major policies in this policy portfolio were promulgated in 2015–2016. By the time the data was gathered for this study in early 2021, sufficient time had elapsed for the policy effects to become apparent, and also recent enough to enable entrepreneurs to provide accurate information on corporate performance and policy participation. Thirdly, the entrepreneurship policies cover a large population group, which enables us to obtain samples large enough to compare the treatment group and the control group, thus, technically eliminating the self-selection bias. The differences in economic development in eastern, central, and western China also make it possible for us to test the heterogeneity of policy effects.Footnote2
Given that we need to assess the effectiveness of China’s policies for BHMWE, our research will be evidence-producing, rather than hypothesis-testing. That is, the research question in this study is to what extent the implementation of the entrepreneurship policy has achieved the intended policy objectives, or more precisely, what is the net increment in the behavioral performance of entrepreneurship policy participants minus the behavioral performance had he or she not participated in the entrepreneurship policy, rather than why and how the policies worked. Our evidence indicates that the awareness of entrepreneurship policies has a significant positive impact on the probability of BHMWE, the net effects between different matching methods and in different regions are different though, there is a robust consistency in the estimated values. The impact of entrepreneurship policies, including the overall and specific policy instruments on the performance of new ventures is quite different. Most policy instruments have significant positive effects on employment, sales, and pretax profit of new firms, with the exception of financial policy, which we find had no significant effects on either of the three economic indicators.
The major contribution of this study to the existing literature is the estimation of the net effect of entrepreneurship policy on the “new venture creation”. This is a key indicator in measuring the effect of entrepreneurship policy, bridging the gap on this key indicator in the existing literature, and providing clear evidence for the effectiveness of entrepreneurship policy. Secondly, by taking advantage of a large sample size and multi-covariate analysis, the study conducts strict matching, and robust estimates of the impacts of policy on other indicators, such as the number of employees. This enables the development of plausible explanations for the inconsistencies among previous related research. Thirdly, this study includes an in-depth and thorough test of the effectiveness of a policy portfolio. Policy variables include not only policies such as fiscal subsidies and financial support, but also non-financial policies such as entrepreneurship training and policy consultation, and comprehensive policies with entrepreneurship incubation as the main content. This provides systemic implications for the design and adjustment of entrepreneurship policies in China, and in other developing countries as well.
Literature review
Given that the purpose of this study is to identify the “additionality” or “net” values of entrepreneurship policy effect, the literature was screened in two steps: First, the keywords “entrepreneurship policy”, “policy effect”, “business performance” and“policy impact” were used to conduct a preliminary screening of the existing published research both in WOB and CNKI over a 20 year period, from 2002 to 2022. Second, after reviewing the flagged publications, those articles that were not methodologically related to the“net effect” were eliminated from further inclusion. Therefore, papers included in the literature analysis are those methodologically employing either randomized experiments, pre-test and post-test comparison of an experimental group, or the data of the experimental group and the matching comparison group, or any econometric approach by which self-selection bias could be eliminated. The papers included in the literature analysis must be methodologically rigorous, or at least those that could be classified as a causal inference analysis in a relatively rigorous sense.
Assessments of Financial-related policies
Caliendo and Kunn (2011) examined the employment outcomes for participants in the Programme of startup subsidies, and bridging allowances in Germany using propensity score matching and difference-in-differences methods. Their study found that the probability of being not-unemployed was 15.6% higher among recipients of the startup subsidy and 10.6% higher among those receiving the bridging allowance, relative to the control group. In terms of labour market integration, the probability of employment was 22.1% higher among participants in the startup subsidy Programme and 14.5% higher among participants in the bridging allowance Programme than the non-participant group. In terms of income, the monthly wage income of startup subsidy and bridging allowance participants was €435 and €618 higher, respectively, than that of non-participants, and the total income (income from self or paid employment plus transfer payments) was €270 and €485 higher, respectively. Figueroa-Armijos and Johnson (2016) applied a spatial difference-in-differences statistical technique to calculate the effects of the Entrepreneurship Community Partnership Tax Credit Programme on participating counties in Kansas, USA. The study’s dependent variables included: proprietors’ income growth per capita, personal income growth per capita, employment growth, growth in the number of proprietors per capita, and growth in average earnings per job. Their findings showed that only one indicator, personal income growth per capita, was significant (p < .01), and even that had a diminishing effect over time. However, given the overall result of the regression, the researchers considered this solitary effect of per capita income to be inconclusive. All results of the spatial lag model, spatial error model, and non-spatial model were consistent. Wu and Huang (2018) analysed the entrepreneurship policies included in the “Mass Entrepreneurship and Innovation” platform. Upon completion of a Ridit analysis, they reported that “optimizing fiscal and tax policies” was the focus of the current entrepreneurship policies at all administrative levels. When the same study subsequently analysed the average treatment effect on the treated (ATT) of the fiscal and tax policy with propensity score matching, the results indicated that the ATT of the fiscal and tax policy was 147.63, meaning that the average profit of enterprises benefiting from the fiscal and tax policy was RMB 1.4764 million higher than that of the matched enterprises that were not. The average effect of fiscal and tax policy awareness was −97.58, meaning that the average profits of enterprises that “did not know” about the fiscal and tax policy were RMB 975,800 lower than that of the matched enterprises that did.
Assessments of non-financial policies
Wren and Storey (2002) examined the impact of the UK Enterprise Initiative, a publicly supported advisory Programme, on the performance of SMEs and found that the scheme had positive impacts on the performance of the experimental group of firms. The authors observed that the initiative had no effect on the survival rates for smaller firms, but medium-sized firms were 4% more likely to survive over the longer term than the control group. Their corrected regression analyses indicated that the policy support impacted sales and employment but varied by firm size. Roper and Hart (2005) evaluated the effects of the Business Link Programme on small firms in England during the 1996 to 1998 period. They found that the Programme had no significant effect on small firms’ sales, employment, or productivity growth. By contrast, excluding the control for selection bias resulted in positive employment growth from the Business Link assistance during that period. Mole et al. (2008) conducted a follow-up study on the Business Link Programme by investigating the types of firms using the advisory services Programme for SMEs, the types of firms benefiting most from such support, and the impact of Programme participation on sales and employment growth. They employed a non-experimental approach with a regression model to control for group differences and Difference-In-Differences methods to eliminate potential bias from the unobserved variables. Ordinary Least Squares (OLS) results indicated that more intensive application of assistance from the Business Link Programme had a positive and significant impact on employment growth but not on the growth of sales. A Difference-in-Differences analysis showed that the average number of employees in the firms receiving assistance from the Business Link Programme increased by 4.4% relative to the control group. However, the employment effect varied significantly across firms with differing types of corporate strategies. Denmark’s North Jutland Entrepreneurial Network, an advisory and mentoring Programme for entrepreneurs, offers three different levels of counselling products: L1-Basic counselling; L2-Follow-up counselling and L3-Extended counselling for entrepreneurs during the startup process. Rotger et al. (2012) analysed the marginal effects of different levels by using a matching method and found that the two-year survival rate of L2 participants increased by about 8%, and the four-year survival rate increased by 5.2%, while for the L3 participants the two-year survival rate increased by 7.6% and the four-year survival rate by 6.4%. For the 2002–2003 cohort, the average effect of L2 participants on employment was 0.5, but this increase was short-term, and the coefficient was not significant in the longer term. The employment effect of L3 participants was not significant in the early stages but became significant after the second and third years. The output effect of L2 and L3 was positive and significant, but the output effect of L2 started to decline after one year, at which time that of L3 started to rise. Fairlie et al. (2015) used experimental data from the GATE (Growing America Through Entrepreneurship) project to analyse the effect of entrepreneurship training. GATE is a pilot Programme Organised by the Department of labour and the Small Business Administration. In the Fairlie et al. research, 4198 applicants were randomly assigned to either a treatment or a control group. Their findings suggest that the GATE project’s impact on final output would be limited. Initially, they found a positive effect of business training which dissipated over the longer term. Fang (2021) empirically tested the income effect of entrepreneurship training based on the survey data of BHMWE in 2019 and found that entrepreneurship training could significantly improve the performance of BHMWE. Additionally, participation in entrepreneurship training improved the ability of back-home migrant workers to gain access to government policies regarding support Programmes and regulatory matters related to new venture development, which improved the performance of BHMWE.
Assessments of multiple instrument policies
Amezcua (2010) examined 950 incubators and 19,000 incubated firms in the United States as the treatment group and a group of matched non-incubated firms as a control group to examine the survival, employment growth, and sales growth of new firms. The findings indicate that incubation reduced the firms’ life expectancy but increased the firms’ employment and sales growth rate. The same study also found that certain types of incubators produced better-performing new businesses and that women-owned businesses benefited more from incubators than male-owned ones. A similar study by Schwartz (2013) using a sample of 371 firms in each of the experimental and control groups, incubated and non-incubated, respectively, in Germany found no significant difference in survival rates between startups that were incubated and those that were not. Their analysis also showed that incubated firms had a statistically significant lower survival rate at three different incubator locations. Autio and Rannikko (2016) analysed the impact of a six-year high-growth entrepreneurship policy in Finland on the firms’ sales growth. Using an eight-year panel that started two years before the initiative began, and propensity score matching to control selection bias, they found that the sales growth of the treated group was 120 percentage points higher than that of the control group over a two-year time span, and 130 percentage points over the control group over three years.
Comments
A comprehensive review of related literature reveals that the existing research using causal inference analysis in the strict sense is limited. Within a 20-year period, only 12 such papers were found, among which there are three papers on the effects of financial-related policies, six on the effects of non-financial policies, and three papers on the effects of multiple instruments policies. Among the nine micro-level studies, indicators to measure the effects of entrepreneurship policies include the number of employees (five papers), sales revenue (six papers), firm’s survival rate (five papers), profit (one paper), and productivity (one paper). The three papers on regional-level Programme effect used other indicators. No research was found in the literature on the impact of entrepreneurship policies on entrepreneurial entry. There is no clear consensus among the current studies on the effects of entrepreneurship policies. From the perspective of firm survival, one study (Rotger, Gørtz, and Storey 2012) found that entrepreneurship policies could effectively improve a firm’s survival. Two studies (Fairlie, Karlan, and Zinman 2015; Wren and Storey 2002) found that entrepreneurship policies had no significant impact on a firm’s survival. From the perspective of sales revenue, one study (Fang 2021) indicated that entrepreneurship policies could significantly improve a firm’s sales revenue. Two studies (Mole et al. 2008; Roper and Hart 2005) found that entrepreneurship policies had no significant impact on sales revenue. Wren and Story (2002) found that the impact of entrepreneurship policies on sales revenue depended on firm size; policy participation could increase the sales revenue of small enterprises but had no significant impact on the sales revenue of large enterprises. In terms of the number of employees, all studies examined found that entrepreneurship policies could significantly expand the size of enterprises.
The relative under-exploration of the net effects of entrepreneurship policies is likely related to the difficulty in accessing large samples and high-quality data. In terms of the indicators measuring the net effect of entrepreneurship policy, regrettably, it appears there are no studies on entrepreneurial entry because, in essence, the most critical indicator to measure the effect of entrepreneurship policy should be the extent to which it increases the number of entrepreneurial ventures.
The lack of consensus among the existing research can likely be explained by three factors. Firstly, the heterogeneity of the research subjects. For different entrepreneurship policies, different economic and social environments in different countries and regions, and different entrepreneurial subjects, the size and direction of the effect of policy are bound to be different. For these reasons, there exists a need to explore the effects of different entrepreneurship policies in different countries. Secondly, the rigour of methodologies. For example, Fang (2021) used discrete values of 1–9 to represent the level of enterprise income, which could not clearly explain the economic meanings of the average treatment effect on the treated (ATT) of training participation. PSM-DID was used to estimate the ATT of training participation, but there were only four covariates in the study, which was barely sufficient to meet the CIA condition of PSM estimation and thus inevitably impaired the accuracy of statistical inference. Thirdly, in some studies, the subjects in question were beyond the category of entrepreneurial enterprises. For example, Autio and Rannikko (2016), where despite overcoming the selection bias problem and having good internal validity, the average maturity of the sample companies was between 4.2 years and 6.7 years. If measured by the Global Entrepreneurship Monitor’s criteria, those companies would not be considered startups but would be classified as small and medium-sized enterprises (SMEs). Making causal inferences based on SME performance indicators will likely distort and amplify the effect of entrepreneurship policy.
This study will examine the net effect of entrepreneurship policies on new venture creation, a key indicator of entrepreneurship policy effect, to address the gap in the existing research on entrepreneurship entry. This study also tests the impact of entrepreneurship policies on other policy effect indicators such as the number of employees, providing compelling evidence through the use of large samples, multiple covariates, and rigorous matching procedures. By doing so, this study provides plausible explanations for the inconsistencies among prior studies on the topic.
Methodology
The area of exploration in this study is essentially the estimation of the average treatment effect of entrepreneurship policy. Because the sample data cannot meet the requirement of randomness, the average treatment effect of entrepreneurship policy cannot be conclusively ascertained by subtracting the outcome of the non-participants from the outcome of the participants. If Y1 represents the average outcome of the entrepreneurship policy participants, Y0 represents the average outcome of the non-participants, D is the participation variable of entrepreneurship policy, when participating in policy, D = 1, not participating, D = 0. Theoretically, the average treatment effect on the treated (ATT) of the entrepreneurship policy should be equal to the average outcome of the policy participants minus the average outcome of the same cohort of participants if they had not participated: