We investigate the burstiness of star formation histories (SFHs) of galaxies at 0.4<z<1 by using the ratio of star formation rates (SFRs) measured from Hβ and FUV (1500 Å) (Hβ-to-FUV ratio). Our sample contains 164 galaxies down to stellar mass (M * ) of 10 8.5 M e in the CANDELS GOODS-N region, where Team Keck Redshift Survey Keck/DEIMOS spectroscopy and Hubble Space Telescope/WFC3 F275W images from CANDELS and Hubble Deep UV Legacy Survey are available. When the ratio of Hβ-and FUV-derived SFRs is measured, dust extinction correction is negligible (except for very dusty galaxies) with the Calzetti attenuation curve. The Hβ-to-FUV ratio of our sample increases with M * and SFR. The median ratio is ∼0.7 at M * ∼10 8.5 M e (or SFR∼0.5 M e yr −1 ) and increases to ∼1 at M * ∼10 10 M e (or SFR∼10 M e yr −1 ). At M * <10 9.5 M e , our median Hβ-to-FUV ratio is lower than that of local galaxies at the same M * , implying a redshift evolution. Bursty SFH on a timescale of a few tens of megayears on galactic scales provides a plausible explanation for our results, and the importance of the burstiness increases as M * decreases. Due to sample selection effects, our Hβ-to-FUV ratio may be an upper limit of the true value of a complete sample, which strengthens our conclusions. Other models, e.g., non-universal initial mass function or stochastic star formation on star cluster scales, are unable to plausibly explain our results.