The impact and effectiveness of China's entrepreneurship policy for back-home migrant workers

The paper evaluates the effectiveness of China’s entrepreneurship policies targeted at back-home migrant workers (BHMWE) using an ex-post approach. Public policy effectiveness is best assessed after implementation, but evaluations are challenging due to political, budgetary, methodological, and data constraints and the diversity of policy impacts. Similar challenges exist for entrepreneurship policies, where rigorous causal evaluations are sparse and evidence on net effects—especially on new venture creation—remains limited and inconclusive. The study focuses on China’s BHMWE policies because: (1) a comprehensive policy portfolio covers preparation, implementation, and development stages, enabling assessment of the net effect on entrepreneurship entry and on nascent firms’ performance; (2) most major policies were promulgated in 2015–2016, allowing sufficient time for outcomes to materialize by the 2021 survey; and (3) policies target a large population, enabling large samples and regional heterogeneity analysis. The research is evidence-producing rather than hypothesis-testing, asking: to what extent have entrepreneurship policies achieved intended objectives, specifically what is the net increment in entrepreneurial entry and firm outcomes attributable to policy participation? The authors find policy awareness significantly increases the probability of entrepreneurial entry, while policy instruments differ in their impacts on employment, sales, and pretax profit. Financial policy shows no significant effects on these indicators, whereas infrastructure-related policy shows the strongest positive effects. The study contributes by estimating net effects on new venture creation, reconciling inconsistencies in prior research through large-sample, multi-covariate matching, and by assessing a complete policy portfolio including financial and non-financial instruments.

The authors systematically reviewed 2002–2022 literature in WOB and CNKI using keywords such as entrepreneurship policy, policy effect, business performance, and policy impact, retaining only methodologically rigorous causal-inference studies (randomized experiments, pre/post with controls, matched comparisons, or econometric designs addressing self-selection). They summarize three groups: (1) Financial-related policies: e.g., Germany’s startup subsidies and bridging allowances increased employment probabilities and incomes (Caliendo & Kunn, 2011); a Kansas tax credit program had limited, inconclusive effects (Figueroa-Armijos & Johnson, 2016); in China, fiscal/tax policy awareness and participation correlated with higher profits via PSM (Wu & Huang, 2018). (2) Non-financial policies: UK advisory programs showed mixed impacts on survival, sales, and employment with effects varying by firm size and intensity of assistance (Wren & Storey, 2002; Roper & Hart, 2005; Mole et al., 2008). Denmark’s staged counselling improved survival and had short- to medium-term employment and output gains (Rotger et al., 2012). The GATE RCT found short-lived training effects overall (Fairlie et al., 2015). A study on BHMWE training in China found significant income/performance improvements and better access to government support (Fang, 2021). (3) Multiple-instrument policies: US incubators increased employment and sales growth but sometimes reduced survival (Amezcua, 2010); German evidence found no survival advantage and even lower survival for some incubators (Schwartz, 2013); Finland’s high-growth policy boosted sales growth markedly (Autio & Rannikko, 2016). Across 12 rigorous studies in 20 years, outcome measures most often included employees, sales, survival; very few examined profits or productivity; none studied entrepreneurial entry as a direct outcome. Findings are heterogeneous and sometimes conflicting, attributable to differences in policy types, contexts, methodological rigor (e.g., limited covariates potentially violating CIA in PSM), and samples that sometimes included mature SMEs rather than startups, potentially biasing estimated policy effects. The authors aim to fill the gap by estimating net effects on entrepreneurial entry and by providing robust evidence on employment, sales, and profits using large samples and extensive covariates.

Design: Estimation of average treatment effects using Propensity Score Matching (PSM) to address selection bias. The treatment variables include overall policy awareness/participation and specific policy instrument usage. The primary outcome for entry is the probability of starting a new firm; for firm performance, outcomes are ln(number of employees), ln(sales revenue in 2020), and ln(profit before tax in 2020). Causal framework: ATT = E(Y1|D=1) − E(Y0|D=1). Because E(Y0|D=1) is unobserved, PSM uses the propensity score P(D=1|X) to match treated and control units under two conditions: (1) Conditional Independence Assumption (CIA) and (2) Common Support. Covariates: To support CIA, the authors include up to 18 covariates correlated with participation and/or outcomes: gender, age, marital status, ethnicity, CPC membership, cadre status, dependents, education, self-assessed health, years of working, work location (municipality/provincial vs lower-level), monthly income, employment contract nature, social security coverage, job position (technical/managerial vs frontline/clerical), pilot county, and regional dummies (east, central). For entrepreneur subsamples in performance analyses, founding year (ln), own funds (ln), and financing amount are also included. Common support is enforced; post-matching differences in means are kept within 10%. Estimation: Propensity scores are estimated via logit models separately for national, eastern, central, and western samples. Two matching algorithms are used for robustness: nearest neighbor (K=5) and kernel (bandwidth=0.06). Standard errors and significance are derived via bootstrap. Balance tests (Ps R2, LR chi2, MeanBias) confirm improved covariate balance post-matching. Sample and data: Primary survey of migrant workers and back-home entrepreneurs in Zhejiang (east), Henan (central), and Guizhou (west). Data collection methods: face-to-face, telephone, and guided online surveys conducted mid-November 2020 to mid-March 2021 by trained student enumerators native to rural areas. Total questionnaires: 4176; valid responses: 3693 (1188 east, 1194 central, 1311 west); 868 respondents identified as entrepreneurs for firm-performance analyses. Questionnaire contained 38 items across demographics, work history, entrepreneurship, and policy participation; internal reliability was high (e.g., Cronbach’s alpha for entrepreneurial awareness 0.984; financial policy scale 0.906; others >0.8). Policy instruments studied: overall policy and six specific instruments—fiscal, financial, land, infrastructure (incubators/parks access), talent (training and related support), and green channel (business environment facilitation).

• Entrepreneurial entry: Policy awareness/participation significantly increases the probability of starting a business. ATT (percentage point increases) by kernel/nearest-neighbor indicate robust effects. Preferred estimates based on matching quality are: National 16.77%; East 15.82%; Central 16.72%; West 16.33%.
• Employment (ln employees): Overall policy not significant. Significant positive effects for: fiscal policy ATT 22.56%; land policy 37.81%; infrastructure policy 40.47%; talent policy 14.83%. Financial policy and green channel policy not significant.
• Industrial development (ln sales revenue): Significant positive effects for overall policy ATT 25.08%; fiscal 17.86% (marginal at ~5–10%); infrastructure 39.55%; talent 26.73%; green channel 22.47%. Financial and land policies not significant.
• Poverty alleviation (ln profit before tax): Significant positive effects for overall policy ATT 36.19%; fiscal 27.26%; land 28.05%; infrastructure 38.95%; talent 21.67% (marginal); green channel 25.79%. Financial policy not significant.
• Comparative magnitudes: The overall policy’s poverty alleviation effect exceeds its industrial development effect by 11.11 percentage points (36.19% vs 25.08%), while its employment effect is not significant. Across instruments, infrastructure policy delivers the largest and most consistent gains across employment, sales, and profits; financial policy shows no significant impact on any outcome.
• Regional heterogeneity: Marginal effects on entrepreneurial entry are somewhat higher in central and western regions than in the east, suggesting larger marginal returns to policy in less-developed regions, consistent with diminishing marginal effects where baseline entrepreneurship is high (e.g., Zhejiang).

The findings directly address the central question of whether entrepreneurship policies for back-home migrant workers increase new venture creation and improve early firm outcomes. Evidence shows a robust, sizable increase in the probability of entrepreneurial entry attributable to policy awareness/participation, filling a major gap in prior literature. For firm performance, policy effects are heterogeneous: infrastructure, land, fiscal, talent, and green channel policies generally improve sales and/or profits, with infrastructure policy exhibiting the strongest, broadest effects. In contrast, financial policy shows no significant effects on employment, sales, or profits, suggesting misalignment between financial instruments and the needs or realities of rural and nascent entrepreneurs (e.g., lack of tailored products and limited access to venture capital). The overall policy’s non-significant employment effect likely reflects the nascent age and small scale of firms (mean age ~3.7 years; mean employees ~9.8) and the offsetting lack of employment impact from financial and green channel policies in the short term. Distinguishing entrepreneurship policy from SME policy is crucial: entrepreneurship policies target nascent ventures (≤42 months per GEM), prioritizing entry and early survival over immediate large-scale job creation. Policy implications include reinforcing SME-oriented supports over time to enhance employment impacts, increasing policy inputs in central and western regions where marginal effects are larger, and redesigning financial policy instruments to better serve back-home entrepreneurs. The strong effects of infrastructure policies validate investments in incubator parks and business incubation services.

This study provides rigorous causal evidence that China’s post-2015 entrepreneurship policies for back-home migrant workers significantly increase the probability of new venture creation. Using large-scale survey data from Zhejiang, Henan, and Guizhou and PSM with extensive covariates and robustness checks, the authors show that: (1) policy awareness/participation raises entrepreneurial entry by about 16–17 percentage points nationally, with slightly larger effects in central and western regions; (2) infrastructure, land, fiscal, and talent policies enhance employment (with infrastructure strongest), while overall employment effects are not yet significant; (3) overall, fiscal, infrastructure, talent, and green channel policies increase sales; (4) all but financial policy increase pretax profits, with infrastructure and land showing particularly strong effects. The poverty alleviation impact surpasses the industrial development impact, underscoring the social benefits of these policies. Contributions include the first rigorous estimate of net policy effects on new venture creation, reconciliation of prior mixed findings via robust matching with many covariates, and a comprehensive evaluation of a full policy portfolio. Future research should extend geographic coverage, explicitly examine heterogeneity across provinces and pilot vs non-pilot counties, capture the organizing/coordination role of local governments quantitatively, and assess longer-term employment outcomes as firms mature.

• Geographic scope limited to three provinces (Zhejiang, Henan, Guizhou) due to funding and pandemic constraints; broader coverage would strengthen generalizability.
• Policy effects likely heterogeneous across provinces and between pilot and non-pilot counties; these differences warrant deeper, explicit heterogeneity analysis.
• The organizing and coordinating role of local governments appears important but was not quantitatively measured in the survey; future instruments should capture this policy dimension.
• Short firm age and small scale in the sample may understate longer-term employment effects; follow-up studies are needed to assess dynamics over time.