Bank herding behavior is often hypothesized to increase systemic risk, but the actual effect is unclear ex-ante from the theory and unknown ex-post from the data. We expand the literature on this topic in several dimensions – posing alternative hypotheses regarding the effects of herding in asset, liability, and off-balance sheet portfolios; developing a novel set of bank-specific, time-varying measures of herding in these portfolios; and empirically testing the relations between bank herding for all three portfolios and bank systemic risk contributions. We find nuanced empirical results that differ by portfolio, bank size class, and periods before versus after TARP.