Transaction Costs In Dealer Markets PDF Download

Utilizing the large subset of trades in which dealers act purely as agents, we decompose dealer spreads in U.S. corporate bond OTC markets into components arising from: 1) dealers' market-making role, and 2) their role as agents for their non-dealer customers. We investigate the determinants of both components. We find that agent-related spreads are large and comparable in magnitude to market-making spreads. In their role as agents, dealers face liquidity-search and customer interface costs, while in their role as market makers they face inventory and asymmetric information costs. Consistent with this, we find that while market-making spreads are strongly correlated with market risk variables, agent-related spreads are not, depending instead on liquidity driven variables. While market-making spreads are inversely related to trade size, agent-related spreads increase with trade size before leveling off and then declining -- possibly indicating that agent-dealers devote lesser search time to relatively smaller trades. While market-making dealer spreads are positively correlated with risk variables, whether trading directly with non-dealer customers or with dealers acting purely as agents, the difference between the former and the latter is negatively correlated with risk variables implying that market-making dealers benefit more from direct interaction with traders when risk and information asymmetry is higher, consistent with dealers deriving information-related benefits from their customer interface. Except for very small trades, explicit transaction costs of non-dealer customers are lower when they trade directly with market-making dealers than when they route trades through a dealer acting purely as an agent. Finally, we show that existing studies have underestimated average overall trading costs in the corporate bond market by conflating the spreads of dealers acting purely as agents with full dealer spreads that include both agent and market making costs. Given our findings on the size and economic determinants of agent-related dealer costs, our results have significant implications for the extensive empirical literature on dealer spreads in other OTC markets.