• back
  • 1.3.10-THE WORLD COTTON MARKET-FUTURES MARKETS

  •  
  • Futures markets

    Chapter 1 - The world cotton market - Cotton prices 

     
     

    The Number 2 cotton contract traded in New York (ICE Futures U.S., formerly the New York Board of Trade and before that the New York Cotton Exchange) is the most widely used cotton futures contract in the world, but contracts are also traded in China (Zhengzhou Commodity Exchange – ZCE), India (National Commodity & Derivatives Exchange – NCDEX – in Mumbai) and Brazil (Bolsa de Mercadorias e Futuros – BM&F – in São Paulo). Cotton market participants operating in Brazil, as well as speculators, have access to both the New York and São Paulo markets, which is not the case in China or India. There are forward cash markets in several other countries.

     

    The primary economic purpose of the ICE cotton futures market is to provide a forum for price discovery and a tool for price risk management. Cotton futures prices are established throughout the trading day by open outcry through the actions of many diverse market participants with a large number of competing buyers and sellers. Price quotes are transmitted worldwide. These prices reflect the latest information about supply and demand and are determined in a trading pit with the narrowest possible spread between bids and offers. The current standard cotton futures contract is for 50,000 pounds (100 bales or 22.68 tons) of strict low middling grade with 1–1/16" staple length. The contract is traded for five delivery months: March, May, July, October and December.

    Detailed information on the ICE Futures U.S. and the other futures markets is presented in chapter 4.

    Relationship between New York futures and the Cotlook A Index

    There is no formal relationship between New York futures prices and the Cotlook A Index because there are times when United States cotton is not among the five cheapest growths quoted in order to calculate the Index. However, over time, the two price series have been highly correlated. The correlation between daily quotes for the A Index and daily closing values for the nearby New York futures contract over the last six years has been 94%. Nevertheless, significant divergences can occur between the two price series. Since 2000, the basis between the A Index and New York Futures (the difference between the two prices) has ranged from New York being 9 cents per pound higher than the A Index to New York being 9 cents below, a relative shift of 18 cents per pound. Over the same period, the A Index itself has ranged between 29 cents per pound and 83 cents per pound, a spread of 54 cents. Consequently, the change in the basis between New York and the A Index represents one-third of the variation in futures prices, thus limiting the usefulness of the New York contract for hedging cotton from outside the United States (see figure 1.30).


    1.3.10-en1 

    Source: ICAC

    Modeling cotton prices


    ICAC uses a statistical model to relate season averages of the Cotlook A Index to a ratio of stocks-to-consumption outside China and a ratio of stocks-to-consumption in China. Prices can be predicted with about 80% accuracy, provided that cotton supply and use in China and in the rest of the world can be predicted. Experience indicates that forecasts for two years in the future are not accurate, but that price forecasts made in March and April of each year are reasonably useful for the coming season.

    This statistical model is based on concepts that have been known for approximately a century. Cotton industry newsletters from before the Second World War talk about stocks and the availability of supply, and the modern econometric techniques used to quantify the relationships between supply, demand and prices were developed in the 1940s. Computers make these calculations easily today, but the basic theoretical concepts are the same as those understood a century ago.

    For most commodities, prices are related to a single variable, the ratio of world ending stocks to world use. As the stocks-to-use ratio rises, prices tend to fall, and vice versa. However, in recent years, not only the aggregate stocks-to-use ratio has been relevant for cotton but also its geographical distribution. In particular, because of the structural changes undergone by the world cotton sector in the late 1990s, the stocks-to-use ratio in China has had a significant effect on world cotton prices. Therefore, an innovation used in the ICAC price model is that the ratio of world stocks-to-use is disaggregated into two variables: the ratio of non-China stocks-to-use and the ratio of China stocks-to-use. Arithmetically, when weighted by their corresponding regional shares of world consumption, the two ICAC variables are equivalent to the world ending stocks-to-use ratio (see figure 1.31).

     

    1.3.10-en2 

    Source: ICAC


    In general, or as a simple rule of thumb, other things being equal, a 1% increase in the stocks-to-use ratio in China results in a decline of about one-third of a percentage point in the season average Cotlook A Index. Similarly, a 1% increase in the non-China stocks-to-use ratio results roughly in a 1.4% decline in the season average Cotlook A Index. This model explains about 80% of the year-to-year variation in average cotton prices, meaning that even if supply and use statistics were known perfectly, there could still be errors between forecasts and actual average prices of about 20%.

    Sources of price forecast error

    There are two sources of error in any statistical modelling exercise: the model itself and the variables used in the model. The ICAC model is statistically unbiased, meaning that the model itself does not tend to over-predict or under-predict. At the end of each season, when supply, use and trade statistics are known, the mean absolute difference between the predicted and the observed season average Cotlook A Index is about 4 cents, and the residuals of the model are random (see figure 1.32).

     

    1.3.10-en3 

    Source: ICAC

    The second source of error is the statistics that are used in the model, and this is the greater source of forecast error for cotton and for most commodities. The biggest problem for ICAC has been forecasting the ratio of China stocks-to-use. However, forecasts are highly linearly correlated to the observed stocks-to-use ratio for China and the rest of the world.

    In evaluating price-forecasting techniques for cotton, it is important to realize what is not included and what is not possible.

    • The ICAC price model does not explicitly include non-cotton market variables such as macroeconomic indicators and competing crop prices. Interest rates, inflation, prices of energy, GDP growth, the prices of competing crops such as wheat, soybeans, sugar and rice, and other variables affect the cotton market. However, those macroeconomic and cross-commodity impacts are linked to changes in prices of cotton through their impacts on cotton production and consumption. Therefore, ICAC considers GDP growth when estimating consumption, and soybean prices are considered when estimating production in Brazil for instance. To the extent that cotton supply and use are estimated correctly, the likely impacts on cotton prices of macroeconomic indicators and competing crop prices can be anticipated, but they are not explicitly included in the price model.
    • The ICAC price model does not acknowledge technical chart patterns, price cycles, random walk variables or lagged dependent variables as valid predictors of future cotton prices. Such models are often developed by mathematicians to predict future commodity prices based on patterns in past prices. Innumerable examples of correlations and patterns in prices can be proven after the fact. Since such models have limited foundation in theory and are of no use in explaining fundamental changes in cotton supply and use and their impacts on prices, they are not relevant to ICAC’s objective of providing greater transparency to the world cotton market. ICAC is not aware of any mathematical price model that can correctly forecast price changes any better than models based on market fundamentals.
    • The ICAC price model cannot be adapted to predict monthly or quarterly prices. The model is estimated based on annual data, and efforts to develop explanatory variables for a quarterly or monthly model have not yielded useful results. When the annual model indicates a season average price above the current price in any season, it is valid to infer that market forces will tend to cause prices to rise over the coming months, but the pattern of monthly price movements cannot be predicted solely with the annual model.

    Expectations of accuracy

    Accuracy in cotton price forecasting depends crucially on accuracy in forecasts of supply, use and trade. Therefore, improvements in forecasts of annual average prices will depend on improvements in forecasts of supply, use and trade, particularly for China.

    Structural changes can occur in the cotton market that require modifications to the price model itself. For instance, in the early 1990s, the breakup of the Soviet Union led to surges in exports from Central Asian countries of cotton previously held in a State reserve, with many of the exports moving under barter arrangements. The ICAC price model was modified to include a variable for barter sales for several seasons until barter sales were essentially discontinued. Price modelling is not a one-time exercise, and ICAC routinely re-estimates the model to update coefficients and test potential variables.

    Given that price forecasts tend to be wrong, it is reasonable to wonder what the value in making forecasts is. Price forecasts are accurate reflections of fundamental market conditions at the time they are made. By providing an explicit price forecast based on current best information available about likely supply and use, each price forecast serves as a valid indicator of where prices would tend if current information were correct.