Assessing the effectiveness of targeted poverty alleviation policies in Xinjiang, China

The study addresses whether and how China’s targeted poverty alleviation policies implemented from 2016 have improved county-level economic development in Xinjiang, a region with a high concentration of ethnic minorities and historical underdevelopment. Motivated by global and national imperatives to eradicate poverty, and recognizing the shift in China from absolute to relative poverty reduction, the paper evaluates the policies’ effectiveness and mechanisms, using night-time lights as an objective proxy for economic activity. It further examines heterogeneity between southern and northern Xinjiang (and minority vs. nonminority areas) and explores the mediating roles of industrial and education-based poverty alleviation. The authors posit three hypotheses: H1—pilot poverty alleviation policies improve county economic outcomes; H2—policy effects are heterogeneous between minority/nonminority (and southern/northern) counties; H3—industrial and education measures act as mediators with differing economic benefits.

The review summarizes global evidence that targeted public policies can alleviate poverty, with examples from India and the Philippines. In China, policy focus evolved from eliminating absolute poverty to addressing relative poverty, with intensified efforts since 2011, including targeted measures (cadres stationed in villages, relocation, industrial support). Economic growth is highlighted as a key driver of poverty reduction, with debates on conditions for trickle-down effects and the role of policy design. Prior Chinese studies investigated poverty structures, industrial and education-driven alleviation, and regional disparities, particularly relevant to Xinjiang where a resource-dependent economy and educational deficits contribute to persistent poverty. Based on this literature, the study formulates H1–H3 regarding policy effectiveness, heterogeneity, and mediating mechanisms of industry and education.

Design: A quasi-experimental Difference-in-Differences (DID) approach, supplemented by Propensity Score Matching (PSM) to construct comparable control groups (PSM-DID), and an event-study specification to test the parallel trends assumption. A mediation (intermediary effect) model examines mechanisms through industrial and education channels. Setting and sample: 80 Xinjiang counties (32 national key poverty alleviation counties as treatment; 48 other counties as control). Policy start year set to 2016. Outcome measure: Economic development proxied by the natural logarithm of county-level night-time light intensity (VIIRS/NPP, NOAA), denoted Inlight. Core variables: DID = treatment × time, with treatment=1 for national key poverty alleviation counties; time=1 for post-2016. Heterogeneity assessed via a dummy Minc for southern Xinjiang (1=southern; 0=otherwise) interacted with DID. Controls: Fiscal expenditure level (FEL = fiscal expenditure/GDP), savings rate proxy (Sav = residents’ savings balance/GDP), social welfare capacity (Wel = social welfare adoption beds/population), population density (Pop). All nominal variables deflated to 2013 prices via CPI and log-transformed where appropriate to mitigate heteroscedasticity. Models:

• Baseline DID: Inlight_ct = β0 + β1 DID_ct + X′_ctβ + μ_c + γ_t + ε_ct.
• Heterogeneity: Inlight_ct = β0 + β1 DID_Minc_ct + β2 DID_ct + α Minc_c + X′_ctβ + μ_c + γ_t + ε_ct.
• Event study: Leads/lags of DID around 2016 to test pre-trends and dynamic effects.
• PSM: Control group matched on FEL, Sav, Wel (matching ratios tested 1:30–1:70; main 1:50). Post-match DID estimated with county and year fixed effects.
• Mediation: Three-equation framework where mediators M include industrial output (ln added value of primary, secondary, tertiary industries: lnins-1, lnins-2, lnins-3) and education (ln number of secondary school students: lnedu). Total, direct, and indirect effects identified via: (4) outcome on DID; (5) mediator on DID; (6) outcome on DID and mediator. Data: VIIRS night lights 2013–2019 (NOAA); socioeconomic data (FEL, GDP, savings, population, social beds, area, secondary school students, industry value added by sector) from Xinjiang Statistical Yearbooks 2013–2019. VIF diagnostics (<10) indicate no serious multicollinearity. Parallel trends validated via insignificant pre-policy interactions.

• Policy effectiveness (baseline DID): DID coefficient on ln(night lights) is positive and significant. Without year FE: 0.46 (t=10.75). With county and year FE: 0.19 (t=6.06), implying ≈19.13% higher regional economic development in treated counties post-policy, supporting H1.
• Heterogeneity: Interaction terms indicate stronger effects in southern Xinjiang relative to northern Xinjiang; minority/southern counties experienced greater gains, supporting H2.
• Parallel trends: Pre-policy interactions (two and one years before 2016) are not significant; post-policy years are significant: post_1=0.22***, post_2=0.24***, post_3=0.21***, confirming policy impact begins after implementation and persists up to three years.
• Robustness (alternative outcome): Using ln(GDP) as the dependent variable yields significant positive DID coefficients: 0.22***, 0.08***, 0.09*** across specifications, aligning with the main results.
• PSM-DID: After matching (1:50), covariate balance is achieved (post-match t-tests not significant). DID remains significant and positive: e.g., 0.51*** (t=8.26) without year FE; 0.20*** (t=4.15) with county and year FE, reaffirming H1.
• Mechanisms (mediation): Industrial channels show significant mediation—secondary (lnins-2) and tertiary (lnins-3) industries have larger positive effects than primary (lnins-1). Education (lnedu) exhibits a positive but smaller mediating effect (coefficient ≈0.08*, weaker than industry), supporting H3.
• Dynamics: Effects generally increase over the first three years post-implementation; discussion notes potential attenuation beyond three years.

The findings confirm that China’s targeted poverty alleviation policies significantly boosted county-level economic activity in Xinjiang, addressing the central research question on effectiveness. Stronger impacts in southern/minority counties suggest that tailoring and preferential measures for ethnic areas enhanced policy efficacy. Mediation analysis demonstrates that industrial development—particularly secondary and tertiary sectors—drives much of the economic gains, while education contributes positively but more modestly in the short run. Event-study evidence validates causal interpretation by meeting parallel trends, and robustness checks (alternative outcome and PSM-DID) strengthen credibility. The results imply that policy design emphasizing industrial capacity, infrastructure, and market linkages can rapidly elevate local economies, while sustained investments in human capital (education and vocational training) are necessary for longer-term poverty reduction and inclusive growth. The observed waning of effects after three years highlights the need for policy evolution from short-term capital projects toward durable institution- and capability-building.

The study provides causal evidence that targeted poverty alleviation policies implemented from 2016 significantly improved county-level economic development in Xinjiang, with larger gains in southern/minority counties. Industrial poverty alleviation is the primary driver (secondary and tertiary sectors outperforming primary), while education-focused measures also contribute but require longer horizons to translate into economic growth. Practical recommendations include: continuing and upgrading industrial development in former poverty counties; strengthening infrastructure for industrialization; expanding market access and sales channels for local specialties; fostering product innovation; and deepening human capital investments by enhancing teacher compensation, expanding higher and vocational education, and ensuring employment support for adult vocational trainees. Post-2020, policies should flexibly transition from short-term infrastructure-heavy interventions to mechanisms that sustain growth, inclusivity, and human capital formation. Future research should extend time horizons and integrate additional dimensions of poverty to assess long-term sustainability and the role of fund allocation quality.

• Time horizon: Data span 2013–2019, potentially insufficient to capture long-term and post-2020 effects of poverty alleviation and rural revitalization.
• Scope: Poverty alleviation is multidimensional (economic, political, cultural, social, ecological); not all aspects or determinants could be incorporated.
• Policy components: The analysis focuses on designation of key national poverty counties; due to data constraints, it does not quantify the specific effects of poverty alleviation fund amounts, allocation efficiency, or use quality.
• Short-run education effects: Education benefits may be underestimated in the short run; longer panels are needed to capture human capital’s lagged returns.