Research

Research Papers:

Can the value of a cryptocurrency be uniquely determined by the fundamentals, such as the rule for money growth implicit in the design of the protocol? To answer this question, we construct a recursive asset-pricing model for a single fiat cryptocurrency, similar to actual Bitcoin. We think of our model as an ideal laboratory, in which equilibria correspond to model solutions that can generate actual data. Our approach stresses the role of the value function as an object of rational choice and hence rests on solid micro-foundations. By imposing enough economically motivated restrictions on that choice, we are able to pin down a unique equilibrium and hence demonstrate that the value of our cryptocurrency is immune to self-fulfilling expectations. This result depends only on the design of the cryptocurrency protocol.

Cryptocurrency Research eJournal (SSRN), Vol. 1, No. 31 (2021)

Presented at:

European Financial Management Association 2022 Annual Meeting, Rome, 29 Jun 2022.
38th International Conference of the French Finance Association (AFFI), St Malo, 23 May 2022.
Brown Bag Seminar, Humboldt University of Berlin, 25 Apr 2022.
12th Financial Markets and Corporate Governance Conference, Monash, 20 Apr 2022.
Mini-Conference on cryptocurrencies at Weizenbaum Institut, Berlin, 4 Nov 2021.

The Purchasing Power of Money in an Exchange Economy (under revision)

Fiat money requires no backing to be accepted at a uniquely determined positive value. I show this using an equilibrium model with realistic frictions and rational households allowed to freely interact in a competitive environment. The model portrays a modern 'cashless' economy relying on electronic payments. Monetary policy is run by a generic authority, which most closely corresponds to a consolidated banking system or monopoly central bank issuing digital currency. The policy defines the accumulation laws for nominal net worth and features two exogenous dimensions corresponding to the nominal interest rate and the rate of new net worth creation via a continuous helicopter drop. The equilibrium concept of (Lucas1978) is generalized by adding a no-arbitrage condition by which the value of money cannot fall to zero if agents know that it can be positive. The model disproves the claims that the price level is generally subject to self-fulfilling expectations or that monetary policy can be reduced to controlling the interest rate via a feedback rule. There is no role for a fiscal authority, either running a non-Ricardian policy or providing implicit backing. Models that remain silent on the nature of money can be misused to draw dramatically incorrect conclusions.

Presented at:

53rd Annual Conference of the Money, Macro and Finance Society, Canterbury, 5 Sept 2022.
World Finance and Banking Symposium, Budapest, 17 Dec 2021.

Babylonian Risk Aversion (with Alex Stomper)

We use crop prices recorded over 2000 years ago in ancient Babylon (Seleucid period) to estimate the coefficient of (relative) risk aversion of ancient crop traders. In order to do this, we assume a simple consumption-based model, according to which crops could either be consumed or stored. We focus on barley and dates, by far the most important sources of consumption in Babylon (at that time), and derive a consumption-based moment condition according to which ex-ante (and hence average ex-post) returns on storing dates relative to barley should be related to the difference in storage risk. Identification is allowed by the cyclical nature of agricultural risk, and the fact that the calendar of harvests is known to us from archeological sources. Indeed, we document that the risk of storing dates relative to barley, measured by the volatility of price changes, was larger at the time of the year corresponding to the harvest of dates. Although the estimated coefficient of risk aversion is of course subject to error (for numerous reasons), the point estimates across specifications are surprisingly similar to what others typically obtain in modern datasets. (Online Appendix)

Estimating Discrete-Time Gaussian Term Structure Models in Canonical Companion Form

Term structure models are generally difficult to estimate. In this paper, I focus on the very popular class of essentially-affine models with flexibly specified prices of risk (Duffee, 2002), formulated in discrete time (Ang and Piazzesi, 2003). The main problem with these models is that there are too many parameters in the most general parametrization, and hence not all of them can be uniquely identified. This problem can be solved by applying the main result of the paper - that all factor dimensions must be spanned by forward rates of n shortest maturities (for an n-factor model). This leads to a convenient 'companion parametrization', which allows for straightforward estimation (for example, by Kalman filter). In contrast to popular 'eigenvalue' parametrization, one needs no prior knowledge of the configuration of eigenvalues, since repeated, or complex eigenvalues, are automatically taken care of (coefficients of the minimal polynomial associated with a companion matrix are real). Empirical application of the estimation procedure reveals apparent violations of no-arbitrage in popular empirical datasets, since long-maturity forward rates appear to move independently of short-maturity rates. This provides an explanation for the predictability evidence of Cochrane and Piazzesi (2005), and suggests that some investors might have enjoyed extremely high realized Sharpe ratios.

Understanding Bond Risk Premia Uncovered by the Term Structure

This paper documents a puzzling conditional correlation between ex-ante excess bond returns, measured by the Cochrane-Piazzesi (2005) factor (CP factor), and regression-based forecasts of inflation. More concretely, the level factor, and the inflation-forecasting factor constructed as linear combination of forward rates, almost exactly span the CP factor. Excess bond returns are high precisely when these two factors move away from each other in the data. Such deviations are present mainly during the inflationary period of 1970's and 1980's. Before and after, expected inflation closely tracks the level factor, and there is almost no excess return predictability.