I study organizational design under informational interdependence—the fact that a single piece of information affects many decisions. In my model, a principal must decide on the allocation of authority over two decisions in order to aggregate information dispersed among biased agents. The allocation of decision rights alters the interaction between interdependence and biases, shaping individual incentives for communication. I find the principal prefers to delegate a high-conflict decision if that improves information transmission on the other dimension. When agents have extreme preferences, centralization can discipline decision-specific biases and improve communication. I also analyze how interdependence affects information acquisition. Agents can decide to specialize, which signals commitment not to manipulate information and, hence, enhances credibility. Finally, I show that delegation reduces the initiative to acquire information: it will be lower overall and concentrated on states that are more important for the decision at hand.

I examine a cheap talk game with two decisions, two payoff-relevant states, and two senders. The model features informational interdependence because information about each state affects both decisions. Senders are imperfectly informed, and communication depends on the nature of their information. I first analyse the case in which each sender observes a signal that fully reveals a state. I show communication depends on how the interdependence aggregates decision-specific biases. This aggregation can lead to positive or negative informational spillovers, which affect communication. Secondly, I analyse the case in which each sender observes noisy signals about both states. Because some realizations influence decisions in different directions, a sender is tempted to follow the most favourable of them. In equilibrium, this leads to a loss of credibility that harms communication. Finally, I show how this credibility loss leads to beneficial congestion effects.

"Effort-inducing Promotions" (with Z. Aboutalebi)

Promotions are signals of worker ability. In our framework, a firm designs the amount of information a promotion conveys to make employees of different abilities work harder. The model features different types of workers and firms. Workers differ in their cost of effort, while firms differ in their marginal productivity of labour. High productivity firms can afford to pay higher wages but only make offers to worker with high expected ability. We show that low-productivity firms want to distort the promotion signal away from full revelation of a worker's type. Doing so increases the effort from low ability workers and decreases that of high ability ones (relative to the fully separating contract). Conditions on the convexity of effort costs guarantee such signalling contracts are optimal. Our result qualifies the traditional notion of "no distortions at the top" in Contract Theory—the firm substitutes informational rents from the high-type for similar rents from the low-type worker. We also show that a worker's career prospect is path-dependent.