It is generally believed that the existence of Granger causality between stock index futures prices and spot prices is inconsistent with the cost of carry theory. In this paper, we argue that the logical links between Granger causality and cost of carry have not previously been correctly set out. We show that causality tests must be set in the framework of an error correction model (ECM) and that this is also the appropriate framework within which to test cost of carry theory. We, therefore, use a single unified framework to test both causality and cost of carry. Within this framework, we show that, under reasonable assumptions, cost of carry requires that there must be some causal relationships between futures and spot prices, and that cost of carry may be consistent with either bi‐ or unidirectional causality. We use our specification to study the pricing of the French CAC‐40 (CAC: Compagnie des Agents de Change) stock index futures contract. We find that the data are consistent with cost of carry only in the first period of the data, when causality only runs from spot to futures but not later, when causality is either bidirectional or runs from futures to spot. These results provide little support for the view that stock index futures necessarily lead spot and suggest the need for a reconsideration of the theory and evidence on stock index futures pricing. Copyright © 2000 John Wiley & Sons, Ltd.

It is generally believed that the existence of Granger causality between stock index futures prices and spot prices is inconsistent with the cost of carry theory. In this paper, we argue that the logical links between Granger causality and cost of carry have not previously been correctly set out. We show that causality tests must be set in the framework of an error correction model (ECM) and that this is also the appropriate framework within which to test cost of carry theory. We, therefore, use a single unified framework to test both causality and cost of carry. Within this framework, we show that, under reasonable assumptions, cost of carry requires that there must be some causal relationships between futures and spot prices, and that cost of carry may be consistent with either bi‐ or unidirectional causality. We use our specification to study the pricing of the French CAC‐40 (CAC: Compagnie des Agents de Change) stock index futures contract. We find that the data are consistent with cost of carry only in the first period of the data, when causality only runs from spot to futures but not later, when causality is either bidirectional or runs from futures to spot. These results provide little support for the view that stock index futures necessarily lead spot and suggest the need for a reconsideration of the theory and evidence on stock index futures pricing. Copyright © 2000 John Wiley & Sons, Ltd.