Yongjie Hu

Date of Graduation


Document Type



Three futures contract pricing models have been commonly used for futures commodity market analysis. Empirical studies find no sufficient evidence of a systematic risk premium for holding futures contracts as suggested by Hedging Pressure Theory and the Capital Asset Pricing Model. Scholes (1981) creates a new way of applying the CAPM to the analysis of the futures market. This study develops Scholes' original work and generalizes the futures contract pricing model by deriving essentially the same model as the cost of carry relationship. The study supports the hypothesis that the present true futures price is the present true spot price with a present value of future total carrying costs. The general specification of the futures pricing model is further incorporated with measurement error in price variables. Cointegration between spot and futures prices should exist as the market equilibrium condition requires. Therefore, the derived model can be tested by checking if there is long-run equilibrium between the two variables. Heteroskedasticity robust Johansen cointegration test procedure (1988, 1991) is adopted to examine 10 pairs of spot and futures dollar prices of 5 currencies. The findings reveal that most pairs are cointegrated. The finding also shows evidence that it is rational to include measurement errors in the model. The cointegration analysis can be extended to examine the lead-lag relationship between spot and futures prices as Engle and Granger (1987) suggest. This study provides several causality tests. The results show that in general there is a futures to spot lead in price. The finding is basically consistent with previous studies. The leadership of futures price may result from the market imperfection. In order to evaluate the validity of the error correction model, out of sample forecasting is conducted to compare the error correction specification with other alternative models. The measurement criteria demonstrate that the ECM does not always exhibit the best forecasts which indicates that the magnitude of the error correction is rather weak. Spot and futures prices appear to be mostly simultaneously related on a daily basis, and the lagged interaction is not strong enough to present better predictive power.