Cointegration and Market Efficiency

Abstract

Asset markets in emerging-market economies are generally thought to be less efficient than the larger, more liquid and thick markets in some rich economies. This suggests the possibility of profiting by detecting inefficiencies in emerging asset markets. A number of authors suggest that one way of looking for opportunities is to test asset prices for cointegration. This article points out that cointegration among asset prices, or its lack, has no impli cations by itself for efficiency. If risk premia are time-varying, then it is possible that cointegration and the efficient markets hypothesis (EMH) jointly hold. There are strong restrictions, however, on how risk premia must vary to ensure that cointegration and the EMH actually hold. These restrictions mean that the finding of cointegration allows sharper, more structured tests of the EMH. First, the parts of assets' rates of return that are predictable from cointegration must equal risk premia. Since the predictable parts of cointegrated assets' rates of return vary over time, risk premia must vary pari passu. Priced risk factors beyond the market are generally required; thus, the Capital Asset-Pricing Model (CAPM) is highly unlikely to be consistent with cointegration and efficiency. The betas, not the risk premia, of at least some priced non-market risk factors must depend on the cointegrating error. Second, the parts of assets' rates of return that are not predictable from cointegration must be explicable cross-sectionally in terms of risk-factor realisations and idiosyncratic error terms. Further, the forecast errors' dependence on realisations of non-market risk factors must vary over time with the cointegrating error.