ABSTRACT A series of public administration reforms were implemented in Japan to cope with the secular stagnation since the 1990s, some of which took the form of the incorporation of public organisations. Drawing on the incorporation of Kohsetsushi, public innovation intermediaries for SMEs, this study evaluates its average treatment effect on the treated (ATT) by applying the difference-in-differences (DID) model to panel data (2000–2021). Local governments decided whether and when to incorporate Kohsetsushi, which suggests staggered treatments and heterogeneous treatment effects. Applying the conventional two-way fixed-effects DID model to panel data with staggered treatments may yield biased ATTs due to contaminated comparisons where early-treated units are used as a control group. This study adopted the DID model to correct the bias, which clarified the correlation between the treatment timings and treatment effects on technology diffusion. Kohsetsushi incorporated late enhanced research and inventive activities with geographically broad spillover while reducing technical consultation with localised spillover. Sources of heterogeneity are discussed from the perspectives of agglomeration externalities, industrial knowledge bases, and learning capacity.