Posted by Kathryn Schwartz on May 16, 2014

The time series properties of the consumption growth of a group of individuals classified as shareowners at a single point in time might not be indicative of the properties of the IMRS relevant for past asset prices. This is important in the US, and particularly in the UK, where levels of share ownership and the composition of the group of shareholders have changed dramatically in recent years. this

This paper studies the dme series properties of shareholders’ consumption and introduces a new way of controlling for the effects of compositional change. Panel data with a sufficiently long time-series dimension that allow us to identify groups of shareholders and non-shareholders over time, which also contain information on total consumption, do not exist. Instead we use a grouping estimator to repeated cross-section data and condition on past information to hold the composition of the group constant in looking at changes over time. This is an application of synthetic panel estimation which, to our knowledge, has not been used before. We define groups of shareholders and nonshareholders in each time period on the basis of predicted probabilities of share ownership. Furthermore, we define these probabilities on the basis of variables that are perfecdy predictable from one period to the next. In computing consumption growth we compare the same group of households in adjacent periods, i.e. we compute the IMRS between time t and /+1 using the consumption of households predicted to be shareholders at time t. This technique controls for changing composition of the group of shareholders between periods. It also solves the problem that the decision to own shares in each period is likely to be endogenous with respect to consumption.

Tags: , ,