Using bond futures data, we test whether high-frequency trading (HFT) is engaging in back running, a trading strategy that can create costs for financial institutions. We reject the hypothesis of back running and find instead that HFT mildly improves trading costs for institutions. After a rapid increase in the number of HFTs, trading costs as measured by implementation shortfall decrease by 27 basis points for smaller-sized positions ($2–$10 million notional). For larger-sized positions there is no significant effect. We explain the improvement as being the consequence of HFT reducing effective spreads and per-trade price impacts.