We find strong evidence of a funding risk premium in the cross-section of asset returns. Using funding shocks identified in our previous paper, we obtain estimates for the price of funding risk that are robust across Treasury bonds, corporate bonds, equities, and hedge funds. Funding shocks pose a risk to investors because they exacerbate the illiquidity and volatility of securities, increase the dispersions of asset illiquidity and volatility, and decrease contemporaneous returns. Our price-of-risk estimates are also robust to using mimicking portfolio returns, alternative portfolio sorts, traditional test assets, monthly returns or quarterly returns. Funding shocks are not subsumed by common proxies for market-wide illiquidity or dealers’ balance sheet risk.
Management Science, forthcoming.
You can find the paper here.
We extend results in Dai and Singleton (2002). Cochrane and Piazzesi (2005) show that (i) lagged forward rates improve the predictability of annual bond returns, adding to current forward rates, and that (ii) a Markovian model for monthly forward rates cannot generate the pattern of predictability in annual returns. These results stand as a challenge to modern Markovian dynamic term structure models (DTSMs). We develop the family of Conditional Mean DTSMs where the yield dynamics depend on current yields and their history. We match these bond risk premium stylized facts. A small Markovian factor “hidden” in measurement error (Duffee, 2011) helps but it is not sufficient to match the evidence.
Updated June 2013.
A recent paper with Hector Perez and Joshua Slive, also from the Bank of Canada. We study how CCP membership rules and risk controls affect the competitive structure of an OTC market. These constraints reduce systemic risk (i.e., CCP default risk) but they also increase the market power of its members. We find that risk controls can serve as an anticompetitive instrument, allowing members to coordinate around a more profitable outcome at the expense of non-members. You can find the paper here : Competition and Strategic Control of a Central Counterparty
Review of Finance, 2014.
With Bruno Feunou. Previous versions circulated as Forecasting Inflation and the Inflation Risk Premium.
Standard Gaussian macro-finance term structure models impose that the conditional mean is a function of the risk factors. We relax this assumption: yields are linear in the conditional mean (but not in the risk factors). To illustrate, if inflation is one of the factors, then yields should span expected inflation but not inflation. Second, expected and surprise yield changes can have opposite contemporaneous effects on expected inflation. Third, the inflation survey forecasts and the inflation rate can be used consistently within the state equation. These three features are inconsistent with the Markov assumption. Our results hold for the US and for Canada. The paper is here.
With Bruno Feunou. Central banks around the world restrict changes in their policy rate to a small set of possibilities. Typically, these are a few multiples of 0.25% away from the current policy rate. In economics, this is an example of the well-known class of ordered choice problems. This paper bridges the gap between discrete-choice problems and asset pricing. We generalize the standard probit-logit case and provide the implications from the absence of arbitrage to develop a wide class of term structure models. Empirically, we find significant non-linearitites in the Fed’s policy function and in US interest rates. The paper is here, Discrete Choice Term Structure Models, and the abstract follows.
The predictive power of Fed funds futures contracts for overnight federal funds rates has not been exploited in term structure models. The combination of data on Fed funds futures with a dynamic term structure model can identify the distribution of monetary policy shocks following future FOMC meetings. The distribution of policy shocks exhibits cyclical changes in the uncertainty surrounding future FOMC meetings. The uncertainty is mainly driven by a “path” factor signaling information about future policy actions. The uncertainty is lowest when the path factor
signals a tightening cycle and highest in a loosening cycle. The uncertainty raises the risk premium in loosening cycle and lowers the risk premium in a tightening cycle, reducing the transmission of target changes to longer maturities. The results trace the information content of futures to hedging demand.
Update (2016): Fed Funds Futures and the Federal Reserve. Some older results were published in Developments in Macro-Finance Yield Curve Modelling (2014).
Review of Finance 18 (1), 2014
With Bruno Feunou, Abderrahim Taamouti and Roméo Tédongap. The paper is here: The Equity Premium and the Maturity Structure of Uncertainty. The trade-offs between risk and returns vary with the investment horizons. The term structure of moments, measurable from option prices, can reveal the risk-returns trade-off associated with Long-Run Risk or Stochastic Volatility factors. Intuitively, we can study the dynamics of hard-to-measure risk factors via their impacts on the term structure of variance and other moments. As predicted by theory, we extract a small number of risk factors with substantial predictive power for the Equity Premium and the Variance Premium. These summarize the compensation for risks implicit in the Equity Premium and the Variance Premium at different horizons.
Review of Financial Studies 25(4), 2012.
Funding Risk Factor updated to December 2013
With René Garcia at the EDHEC Business School. Our main contribution is to show that the value of funding liquidity is an aggregate risk factor driving a substantial share of risk premia across fixed-income markets. The paper can be found here : Bond Liquidity Premia, here is the online appendix and the abstract follows.
Here is the factors used in the published version of the paper: FundingLiquidityFactor19862009. The updated funding liquidity and term structure factors can be found here: FundingLiquiditFactor_19862015Q4. These factors are obtained using the model in the paper above, but re-estimated with recent data. The funding liquidity factor is standardized to a mean of zero and standard deviation of one. Strikingly, using parameters estimated using recent data yield very similar results to that usin data until to the end of 2007 (as in the paper). This Figure –CompareFundingLiquidityFactors– compares the series obtained with data until 2007, 2009 and 2012. The model continue to perform exceptionally well out-of-sample.