• Selected
• Entire List

Abstract:
Forecasts of the federal funds rate target, the most anticipated indicator of the Fed's monetary policy stance, can be considerably improved when its evolution is modeled as a marked point process in discrete time. This paper motivates an alternative approach that combines Hamilton and Jorda's (2002) autoregressive conditional hazard (ACH) and Russell and Engle's (2005) autoregressive conditional multinomial (ACM) model. The paper also proposes a methodology to evaluate probability function forecasts of this model class. The ACH-ACM qualifies as a useful modeling framework. Goodness of fit and point forecasts are improved. Parsimonious specifications deliver useful probability function forecasts for the target.

Marked Point Process, Autoregressive Conditional Hazard Model, Autoregressive Conditional Multinomial Model, Probability Forecast, Forecast Evaluation

Abstract:
This paper proposes a new information share for price discovery based on Russell's (1999) autoregressive conditional intensity model. While previous studies rely on equally spaced high frequency data, we use the information conveyed by trade intensities to determine a market's contribution to price discovery. Thereby, we account for the irregular nature of transaction data. Moreover, in contrast to the commonly applied Hasbrouck (1995) approach, which yields lower and upper bounds for information shares, our model delivers a unique measure. Our empirical application to US-listed Canadian stocks supports previous evidence for the home market leadership in price discovery.

Price Discovery, Multivariate Autoregressive Conditional Intensity, Cross-Listed Stocks