Data on the incidence and prevalence of hate crimes in the United States come primarily from the Uniform Crime Reports (UCR). Although UCR data undercount most crimes, hate crimes are particularly underreported, especially for some groups. We compare 2000-2011 UCR data in Pennsylvania to data from a state agency that came from police, media, and citizen reports. First, we find that the state-agency database is generally consistent with the UCR data, in terms of absolute counts, correlations, and predictors. Second, we find that UCR data underestimate hate crime rates by a factor of about 1.6 overall and by a factor of 2.5 for rural areas. Moreover, although UCR data on hate crimes show a decrease in the most recent 5-year period, the state agency data show that hate crime incident counts have not dropped. We suggest that using a broader index that includes both the UCR and a database like that in Pennsylvania will give a more complete picture of hate crime.