Patel, Mona Dinesh (1997). Testing for cointegration between the New York and London futures markets for coffee. Master's Thesis. Thesis uri icon


  • Arbitrage between the New York and London futures markets is an active event. The price spread between the New York and London futures markets, is affected by factors concerning the supply and demand of each commodity and the world coffee market. The New York futures price for coffee is the premium price and this reflects both short run differences in expected demands and supplies, as well as long run factors, such as tastes and preferences. If the supply of London coffee falls short, the premium between the two markets narrows and this allows for arbitrage possibilities. In the long run the premium widens, as exporters will purchase comparatively more of the New York coffee, driving its price up and therefore returning the premium to its usual level. To determine whether or not arbitrage is possible between the New York and London coffee futures markets, a test for cointegration was performed. Usually, if two markets have a cointegrating relationship, arbitrage opportunities exist. Cointegration signifies a long-term equilibrium relationship between two or more series. The process of testing for cointegration between the New York and London coffee futures prices involved two steps. The first is to determine stationarity and the second is to test for cointegration. Using the Dickey-Fuller and Augmented Dickey-Fuller tests, both of the data series were tested for stationarity. The data were transformed to induce stationarity. The second step is to test for cointegration, and this was performed using the Durbin-Watson test and the Dickey-Fuller test. These tests concluded that a cointegration relationship is present. To determine when the arbitrage opportunities occur, two forecasts models were developed. The error-correction model and the vector autoregression model were considered for the New York market, and the error correction model forecasts outperformed the vector autoregression model.

publication date

  • 1997