<p>This paper analyzes the role of network structure in shaping employment dynamics within rural development policies, focusing on the LEADER program in Asturias (Spain). Using municipal-level employment data and a Poisson Network Autoregressive (PNAR) framework, we examine how alternative network definitions condition the detection and interpretation of policy effects in a context characterized by strong spatial and temporal interdependence. We construct and compare three network configurations: an institutional network based on Local Action Groups, an investment-based similarity network, and an economically grounded network capturing similarity in primary sector characteristics. While direct network spillover effects are generally not statistically significant across specifications, the results show that network structure plays a relevant indirect role by altering the magnitude and significance of key policy-related covariates, particularly those associated with LEADER funding intensity and beneficiary participation. Our findings indicate that economically grounded networks provide a more nuanced representation of rural interactions and improve the empirical characterization of employment dynamics, even in the absence of strong direct spillover parameters. This highlights the importance of network design in evaluating place-based policies and cautions against relying solely on institutional boundaries when assessing territorial policy impacts. Overall, the paper contributes to the literature by demonstrating how network econometric models can enhance policy evaluation in rural settings, offering methodological insights with direct implications for the design and territorial targeting of development interventions. </p>