The accurate determination of electrocaloric coefficients in nanometer-thin, polycrystalline doped HfO2 is challenging and has led to very different values reported in the literature. Here, we apply two different methods in order to compare and analyze reversible and irreversible or metastable contributions to the electrocaloric effect. The indirect method is based on temperature-dependent ferroelectric hysteresis characteristics. Furthermore, we apply a direct method, where electrocaloric temperature variations are observed using a specialized test structure. A comparison of both methods reveals that the indirect method dramatically overestimates the response due to thermal fatigue effects, which are caused by the migration of charged defects to the electrode interfaces. The partial transition to the antiferroelectric-like tetragonal phase is not immediately reversed to the polar Pca 2 1 phase upon cooling. An electrocaloric coefficient of - 107 μ C m - 2 K - 1 is determined for a 20 nm thick Si-doped HfO2 film with the direct method, which corresponds to a Δ T of 4.4 K.