Uncertainty pervades financial markets. How financial market participants form expectations when faced with uncertainty is therefore central to the study of financial markets. This thesis contains three chapters each highlighting the role of expectations formation in determining outcomes in financial markets.

The first chapter, coauthored with Nicholas Vause and Nora Wegner, builds a theoretical model of self-fulfilling fire sales motivated by the dash for cash of March 2020. Investment funds fear being hit by a future liquidity shock and can choose to preemptively liquidate their bond holdings. However, funds face uncertainty about how many other funds will choose to preemptively liquidate. If funds wait to see if the liquidity shock crystalises, they risk selling their bonds into a depressed market if other funds have already chosen to liquidate. This creates the risk of a self-fulfilling fire sale where funds choose to preemptively liquidate because they expect that other funds will liquidate. Following the global games literature (Carlsson & van Damme 1993, Morris & Shin 1998), we derive the probability of a self-fulfilling fire sale and extend the model to include a central bank providing a market backstop. The central bank can choose the quantity of assets it is willing to purchase and the discount (relative to the bond return) that it charges to purchase the bond. When we introduce the central bank, we show that if the central bank can credibly commit to (i) set its discount low enough and (ii) the quantity of asset purchases high enough, then it can eliminate self-fulfilling fire sales. Moreover, it can achieve this without having to purchase any bonds. The aggressive policy works via expectations. In particular, it makes the pessimistic beliefs that drive the fire sale impossible for funds to rationalise because the funds know that the central bank stands ready to provide liquidity via asset purchases if needed.

Whereas in the first chapter I assume agents can costlessly absorb all available information, in the second chapter (solo-authored), I follow the rational inattention literature (Sims 2003) and relax the assumption that information is costlessly obtained. This enables me to examine whether fragilities can build up in the financial system simply because agents pay insufficient attention to each other. I build a model where bank values are interdependent within a finan- cial network. However, banks cannot costlessly observe other banks’ values. Instead, banks must choose to pay attention to developments in the value of other banks. Because paying attention incurs a cost, banks may optimally choose not to allocate significant attention to certain banks. In the model, banks that believe they have a higher value choose to supply more credit. Therefore, if inattentiveness causes banks to incorrectly infer their own value, their credit supply decisions will be distorted relative to the optimal. I show that banks that are moderately important for determining values in the network cause the greatest distortion in credit supply because banks do not deem them important enough to pay high levels of attention to, yet they are still important for determining bank values. This suggests that we should be more cautious about dismissing all but the most interconnected banks as being important for ensuring financial stability.

Thus far, agents have known the objective probability distributions relevant for decision making. In the third chapter, coauthored with Harjoat Bhamra and Raman Uppal, we follow Knight (1921) and distinguish between risk (known probabilities) and Knightian uncertainty (unknown probabilities). We argue that geopolitical uncertainty can often be viewed as Knightian uncertainty rather than risk and our objective is to examine the effects of this uncertainty. We do this by constructing a dynamic stochastic equilibrium production model of a world economy with two countries. Each country is characterised by a traded and a non-traded goods sector and a representative investor with Stochastic Differential Utility who is averse to Knightian uncertainty. We model geopolitical uncertainty as a loss of confidence in the correct model for the shocks to the efficiency units of capital where the investors cannot assign probabilities to the alternative models for the shocks. We solve this model in closed form and show how uncertainty operates by reducing households’ perceived expected returns on capital which, in turn, distorts portfolio and consumption choice decisions. We then examine the implications of these distortions for trade flows, exchange rates, growth, and the level of social welfare. We show that our model can match stylised facts of the UK economy following the Brexit referendum.