Working Papers

Centralized vs Decentralized Markets: The Role of Connectivity (with Simone Alfarano, Albert Banal-Estanol, Eva Camacho, Giulia Iori and Burcu Kapar)

We consider a setting in which privately informed agents are located in a network and trade a risky asset with other agents with whom they are directly connected. We compare the performance, both theoretically and experimentally, of a complete network (centralized market) to incomplete networks with differing levels of connectivity (decentralized markets). We show that decentralized markets can deliver higher informational efficiency, with prices closer to fundamentals, as well as higher welfare for mean-variance investors.

Market Fragmentation and Contagion (with Jean-Pierre Zigrand)

We study the transmission of liquidity shocks from one sector of the economy to other sectors in a general equilibrium model with multiple trading venues connected by profit-seeking arbitrageurs. Arbitrageurs effectively provide liquidity to investors by intermediating trades between venues. The welfare impact on venue k of a liquidity shock on venue l can go in either direction, depending on whether intermediated trades on k behave as complements or substitutes for such trades on l. In addition to this direct effect through the arbitrage network, there is a feedback effect of an adverse shock reducing liquidity and arbitrageur profits, which leads to a lower level of intermediation, further reducing liquidity. We illustrate this contagion with examples of high-frequency trading in equity markets, shocks to one tranche of a collateralized debt obligation impacting investors in the other tranches, carry trade crashes, shocks to cross-country bank lending following the global financial crisis, and the bursting of the Japanese bubble in the early 1990s.

Arbitrage Networks (with Jean-Pierre Zigrand)

This paper studies the general equilibrium implications of arbitrage trades in segmented financial markets. Arbitrageurs choose a category of trades to specialize in. This results in an equilibrium network in which the various market segments are linked by arbitrageurs. Arbitrageurs exert externalities on each other depending on their position in the network. Due to these externalities, the complete network architecture, in which all links are feasible, is in general suboptimal for arbitrageurs; it is dominated by a hub-spoke architecture. The hub acts as a repository of liquidity, facilitating trades with minimal price impact. For an arbitrary architecture, as the mass of arbitrageurs grows, equilibrium prices converge to those of the frictionless economy with no segmentation. On the other hand, even if the architecture is complete, equilibrium networks may not be complete or even connected, regardless of the mass of arbitrageurs.

Efficiency Properties of Rational Expectations Equilibria with Asymmetric Information (with Piero Gottardi)

In this paper we provide a characterization of the welfare properties of rational expectations equilibria of economies in which, prior to trading, agents have some information over the realization of uncertainty. We study a model with asymmetrically informed agents, treating symmetric information as a limiting case. Trade takes place in asset markets that may or may not be complete. We show that equilibria are characterized by two forms of inefficiency, price inefficiency and spanning inefficiency, and that generically both of them are present. Price inefficiency arises whenever equilibrium prices reveal some information. It formalizes and generalizes the so-called Hirshleifer effect, by showing that generically an interim Pareto improvement is possible even conditional on the information that is available to agents in equilibrium; the primary source of the inefficiency is a pecuniary externality. Spanning inefficiency, on the other hand, arises if prices are not fully revealing and markets are incomplete relative to the uncertainty faced by agents in equilibrium. In this case, an ex-post improvement can generically be implemented by providing agents with more information, thus expanding their risk-sharing opportunities and reducing informational asymmetries, even though this additional information restricts the set of allocations that are incentive compatible and individually rational.