Can environmental information infrastructure improve cleaner urban production? Evidence from China’s Smart Monitoring Pilot

Introduction Existing monitoring studies mainly examine compliance and pollutant outcomes, whereas green total factor productivity (GTFP) studies rarely isolate monitoring-specific digital infrastructure from broader smart-city policies. This study evaluates whether environmental information infrastructure improves cleaner urban production using China’s Smart Monitoring Pilot of the Ecological Environment as a quasi-natural experiment. Methods We use a city-level panel of 268 prefectures from 2014 to 2023 and estimate difference-in-differences models with city and year fixed effects. Identification is assessed through event-study estimates, alternative codings of the partial-exposure year 2022, winsorization, sample exclusions, an alternative GTFP measure, placebo tests, and spatial-exposure specifications. Results Under province-level clustered standard errors to reflect the level of treatment assignment, we find that pilot designation is associated with a modest short-run increase in GTFP. The estimate is stable across alternative codings of 2022 and sample exclusions; an alternative GTFP measure yields a directionally consistent but statistically insignificant result. Reduced-form channel regressions provide evidence consistent with potential enforcement-credibility and green-innovation pathways, including higher green invention patenting. Both pre-treatment digital infrastructure and pre-existing environmental regulation intensity positively moderate the treatment effect. Spatial exposure estimates are statistically insignificant, providing no detectable evidence of spillovers in the observed sample. These results are consistent with, but do not establish, the no-interference assumption. Discussion Smart environmental monitoring may work as productive information infrastructure, with gains that depend on data-processing capacity and the pre-existing regulatory environment. The modest magnitude and sensitivity to inference level caution against strong causal claims from this short post-treatment window.