Search Results
Working Paper
Mining for Oil Forecasts
In this paper, we study the usefulness of a large number of traditional determinants and novel text-based variables for in-sample and out-of-sample forecasting of oil spot and futures returns, energy company stock returns, oil price volatility, oil production, and oil inventories. After carefully controlling for small-sample biases, we find compelling evidence of in-sample predictability. Our text measures hold their own against traditional variables for oil forecasting. However, none of this translates to out-of-sample predictability until we data mine our set of predictive variables. Our ...
Working Paper
Can Spanned Term Structure Factors Drive Stochastic Yield Volatility?
The ability of the usual factors from empirical arbitrage-free representations of the term structure?that is, spanned factors?to account for interest rate volatility dynamics has been much debated. We examine this issue with a comprehensive set of new arbitrage-free term structure specifications that allow for spanned stochastic volatility to be linked to one or more of the yield curve factors. Using U.S. Treasury yields, we find that much realized stochastic volatility cannot be associated with spanned term structure factors. However, a simulation study reveals that the usual realized ...
Working Paper
Big Data Meets the Turbulent Oil Market
This paper introduces novel news-based measures for tracking global energy markets. These measures compress thousands of news articles into a parsimonious set of real-time indicators and are successful in-sample forecasters of oil spot, futures, and energy company stock returns, and of changes in oil volatility, production, and inventories, complementing and extending traditional (non-text) predictors. In out-of-sample tests, text-based measures predict oil futures returns and changes in oil spot prices better than traditional predictors, although the latter are more useful for forecasting ...
Working Paper
A Note on the Finite Sample Bias in Time Series Cross-Validation
It is well known that model selection via cross validation can be biased for time series models. However, many researchers have argued that this bias does not apply when using cross-validation with vector autoregressions (VAR) or with time series models whose errors follow a martingale-like structure. I show that even under these circumstances, performing cross-validation on time series data will still generate bias in general.
Working Paper
A Note on the Finite Sample Bias in Time Series Cross-Validation
It is well known that model selection via cross validation can be biased for time series models. However, many researchers have argued that this bias does not apply when using cross-validation with vector autoregressions (VAR) or with time series models whose errors follow a martingale-like structure. I show that even under these circumstances, performing cross-validation on time series data will still generate bias in general.