Siyu Bian

with Teresa Serra, Philip Garcia, and Scott H. Irwin

Market responses to public news announcements are commonly measured by their impact on price returns variance, which allows inference on the value of information and the length of the price discovery process. Recently published articles based on high-frequency data fail to disentangle efficient market price variance from microstructure noise, which produces biased estimates of announcements’ market impacts. By using a Markov Chain framework, we address the shortcomings of previous research and assess the market response to key public information releases affecting agricultural markets. We compare two mechanisms to release public information that have been used in these markets: the trading halt and the real-time. We show how the value of microstructure noise can be used to improve public policy decisions. We find that the real-time release of information brings faster efficient price discovery at the cost of large microstructure frictions. Increases in the cost of noise are not compensated by the improvements in the speed of efficient price discovery. Overall, our findings are highly relevant to public policy and have implications for market design.

with Teresa Serra, and Philip Garcia

Our findings show that the linear models outperform nonlinear machine learning algorithms in out-of-sample volatility forecasting. We then design a hybrid neural network that chooses between linear and nonlinear structures when forecasting volatility. The hybrid neural network decomposes its forecasts into linear and nonlinear components. The decomposition results suggest that nonlinear components amount to between 0% to 25% of total volatility.

In 2014 hog and corn market prices reflected the consequences of the Porcine Epidemic Diarrhea Virus (PEDV) outbreak. This research studies, for the first time, the impact of the PEDV outbreak on the daily covariance of lean hog and corn intraday futures price returns. My results suggest that the PEDV crisis caused relevant increases in the lean hog (increase by 250%) and corn (increase by 150%) price return variances, and much stronger changes in the spillovers between the two markets (increase between 400% to 500%). Results also suggest that, due to substantial increases (about 600%) in market microstructure noise after the PEDV outbreak, observed return covariances constitute a relevant biased estimation of efficient return covariances.