In evaluating the effectiveness of R&D subsidies, the literature has focused on potential crowding out effects, while the possibility of misappropriation of public funds that results from moral hazard behavior has been completely neglected. This study develops a theoretical framework with which to identify misappropriation. Using Chinese firm-level data for the period 2001-2011, we show that misappropriation is a major threat. 42% of grantees misused R&D subsidies for non-research purposes, accounting for 53% of the total amount of R&D subsidies. In a second step, we study the loss of effectiveness of R&D subsidies in stimulating R&D expenditures that is due to misappropriation. We measure the loss in effectiveness by estimating the causal effect of R&D subsidies in the presence of misappropriation using an intention-to-treat (ITT) estimator and comparing it to the ideal situation (without misappropriation) using the complier average causal effect (CACE). We find that China's R&D policy could have been more than twice as effective in boosting R&D without misappropriation. R&D expenditures could have been stimulated beyond the subsidy amount (additionality), but noncompliant behavior has resulted in a moderately strong partial crowding out effect. We find significant treatment heterogeneity by period, subsidy size, industry, and ownership. Notably, the loss in effectiveness has diminished following a policy reform in 2006. Nevertheless, the misappropriation of public funds considerably undermines the impact of R&D policies in China.