Genuine and Spurious Serial Correlations in Asset Returns — An Illustration: Real Estate versus Equity

Abstract

Asset returns in efficient markets should not display serial correlations. Otherwise, asset prices would be predictable to a certain extent and arbitrage opportunities would appear, contradicting the assumption of efficiency.

Lack of serial correlation is considered to be true for most sufficiently traded financial assets. For other assets, such as real assets, and in particular for real estate, a substantial level of serial correlation is suspected, and, as will be seen, verified in practice. There are many factors that may contribute to a high level of serial correlation: high transaction costs, information imperfections, scarcity of transactions, and role of surveys instead of transactions in the process of price formation...

This short paper focuses on one hidden factor: averaging. As transactions are scarce, and are related to heterogeneous assets, price indices must be built so as to encompass a minimal period of time: one month at least, generally a quarter, and the resulting index is computed as an average over that period.

We show that averaging introduces a .25 serial correlation for basically non-serially correlated asset price moves. For genuinely serially correlated series, averaging increases the measured serial correlation. We check those findings on two large US asset classes: residential real estate and equity.

Keywords: Serial Correlation, Asset Prices, Asset Returns, Efficient Markets, Predictability

JEL Classification: C10, G11, G12, G14, G17