Working papers:


Heterogeneous Firms, Wages, and the Effects of Financial Crises
[working paper here]

The Great Recession manifested itself mostly as a decline in hours and employment in the US, but a decline in Total Factor Productivity in the UK. Why did these two economies experience the crisis differently? I argue that a third fact, that real wages fell further during the crisis in the UK, may provide an explanation. I show that how a financial crisis manifests in the real economy in standard heterogeneous firm models depends crucially on assumptions on wage adjustment, and use this result to explain the divergent experiences of the two countries during the recent financial crisis. Theoretically, while a decline in wages protects the labour market, I show that, in the presence of financial frictions, it also causes a fall in TFP by misallocating resources across firms. Financially unconstrained firms are able to take greater advantage of a decline in wages than constrained firms, leading to a relative reallocation of resources towards the unconstrained, which I show must always reduce TFP in my model. On the other hand, if wages are fully rigid, I show that a financial crisis will have no effect on TFP. Quantitatively, the model is able to explain the greater fall in hours during the crisis in the US, the lack of a significant fall in TFP in the US, and 1/3 of the UK’s TFP decline, or “productivity puzzle”.


Fiscal Policy with Limited-Time Commitment (with Andrea Lanteri)
[Revise and resubmit, Economic Journal. Working paper here]

We consider models where the Ramsey-optimal fiscal policy under Full Commitment (FC) is time-inconsistent and define a new notion of optimal policy, Limited-Time Commitment (LTC). Successive one-period lived governments can commit to future plans over a finite horizon. We provide a sufficient condition on the mapping from finite policy sequences to allocations, such that LTC and FC lead to the same outcomes. We then show that this condition is verified in several existing models, allowing FC Ramsey plans to be supported with a finite commitment horizon (often a single period). We relate the required degree of commitment to the economic environment: in economies without capital, the minimum degree of commitment required is given by the government debt maturity; in economies with capital and government balanced-budget constraints, the required commitment is given by the horizon over which the budget has to be balanced. Finally, we solve numerically for the LTC equilibrium of an economy where the equivalence result fails and show that a single year of commitment to capital taxes provides substantial welfare gains relative to the No-Commitment time-consistent policy.


Growth and Business Cycle Effects of Future Financial Crises
[working paper here]

I study the ex-ante effects of the fear of future financial crises. I show theoretically that this “crisis fear” has both negative growth and business cycle effects, and can overturn the conventional view of the trade-offs of prudential policy. In a continuous-time framework, I model crises as multiple-equilibria events where the net worth of the financial sector is wiped out by a self-fulfilling fall in asset prices. I study the effects of allowing agents to understand the probability distribution over future crises, by solving the model with a sunspot (modelled as an endogenous jump process) determining equilibrium selection. The fear of crises lowers investment and growth today, even if experts are currently well enough capitalised to survive a crisis. The possibility of future crises also creates a state-dependent financial accelerator. Prudential policy can simultaneously increase growth and stabilise the economy, in contrast with common arguments that prudential policy should decrease growth.



Work in progress:


Sticky Wages and Financial Frictions
[slides available here]

Recent empirical work has highlighted that the wages of new hires are more flexible than those of existing matches, which casts into doubt the ability of the standard search and matching model to generate reasonable fluctuations in unemployment. Schoefer (2015) proposed a “financial channel of wage rigidity” by which the stickiness of existing wages affects firm cashflow, makes hiring (and hence unemployment) more volatile in the face of financial frictions. I extend this idea in two directions, allowing both existing and new wages to simultaneously affect hiring decisions. Firstly, I show that this financial channel may be amplified or dampened when one recognises that firms borrow against future profits, which depend on the wages of new hires. Secondly, I introduce capital into the model and show that it is the wages of new hires which determine how firms allocate scarce cash between hiring and investment.


Sudden Stops and Financial Fragility (with Sweder van Wijnbergen)
[preliminary draft available on request]

We study the interaction of sudden stops and domestic financial fragility in a small open economy macro model with a domestic banking sector. We present two main results. Firstly, sudden stops are more painful when the banking sector is undercapitalised. When the banking sector is well capitalised, sudden stops lead to depreciation, but when it is undercapitalised they also lead to the disruption of domestic production. Secondly, twin price spirals amplify the effects of shocks: depreciation reduces the real value of bank net worth, leading to fire sales of assets and declines in domestic asset prices. Declines of domestic asset prices also reduce bank net worth, reducing the ability to borrow from abroad, leading to further depreciation.


(Exact) Non-Stochastic Simulation of Heterogeneous Agents using Endogenous Grids
[notes and example code available on request]

I develop an algorithm which extends Young’s (2009) Non-Stochastic Simulation algorithm to overcome the curse of dimensionality. The algorithm simulates the population distribution on a set of endogenously chosen nodes. If the population distribution has finitely many nodes, the algorithm constructs and only simulates on the true (potentially time-varying) nodes. If the distribution has uncountably or countably infinite (or simply too many) nodes, the algorithm discards low-density nodes and projects the distribution onto the remaining nodes.