Distributed Lag Fixed Effects and Non-Additive Response Surface Inference for Intergovernmental Fiscal Transfers: Evidence on Food Security

Abstract

This study examines how intergovernmental fiscal transfers relate to food security when effects may be delayed and instruments may operate non-additively. Using panel data on 435 Indonesian districts/cities in 2018–2023 (N = 2,610), we estimate fixed-effects distributed-lag models for DAK-Irrigation, DAK-Agriculture, and the Village Fund, then fit post-2020 interaction models. To keep interactions interpretable, we evaluate marginal effects over observed transfer combinations and complement the baseline with checks for common shocks, extreme values, and placebo threats. Three findings emerge. First, irrigation associations are timing-sensitive: near-term coefficients are negative, whereas the two-year lag is more favorable. Second, Village Fund associations are most consistent with a one-cycle (t-1) pattern, although they remain sensitive to aggregate time variation. Third, agriculture terms are weak. The post-2020 response surface suggests flattening at high joint transfer intensity, but implied FSI shifts are small, marginal effects remain indistinguishable from zero, and placebo diagnostics do not support a causal interaction block. Because 2020–2023 coincides with the COVID-19 crisis and recovery, any non-additivity may reflect crisis-era coordination stress rather than a normal-state relationship. Overall, the paper offers a framework for evaluating lagged fiscal programs in short panels and interprets estimates as within-district associations rather than causal effects.