Kinetic Component Analysis

Abstract

We introduce Kinetic Component Analysis (KCA), a state-space application that extracts the signal from a series of noisy measurements by applying a Kalman Filter on a Taylor expansion of a stochastic process. We show that KCA presents several advantages compared to other popular noise-reduction methods such as Fast Fourier Transform (FFT) or Locally Weighted Scatterplot Smoothing (LOWESS): First, KCA provides band estimates in addition to point estimates. Second, KCA further decomposes the signal in terms of three hidden components, which can be intuitively associated with position, velocity and acceleration. Third, KCA is more robust in forecasting applications. Fourth, KCA is a forward-looking state-space approach, resilient to structural changes. We believe that this type of decomposition is particularly useful in the analysis of trend-following, momentum and mean-reversion of financial prices.

An instrument exhibits financial inertia when its price acceleration is not significantly greater than zero for long periods of time. Our empirical analysis of 19 of the most liquid futures worldwide confirms the presence of strong inertia across all asset classes. We also argue that KCA can be useful to market makers, liquidity providers and faders for the calculation of their trading ranges.

Keywords: Kinetic Component Analysis, Time Series, Principal Component Analysis, LOWESS, Fourier Analysis, Kalman Filter

JEL Classification: G0, G1, G2, G15, G24, E44