during die growing period. Once planted, you would have a very good idea of the expected supply; onehatf of the supply-demand equation would be constant eadi year.

Each agricultural product has its own particular sensitivity to weather. Grain planting can be delayed due to rain, causing some farmers to switch from corn to soybeans; hot weather during pollination will significantly reduce yields; a freeze in September can stop the ripening

FIGURE 7-5 Soybean monthly price variation.

»pi Hay Jun Jul 1 Stt! Oct Nm Dec

process and damage production. Freezes are of greater concem than droughts and affect more products. In general, pattems for crop sensitivity to weather depend upon their location in the Northem or Southern Hemisphere. Active frade between world niarkets shows that crops grown primarily in the Northem Hemisphere are allected by weather developments in the Southem Hemisphere. In Table 7-8 major weather events are shown separated into Southem and Northem Hemisphere. Figure 7-6 combines the soybean and corn weather sensitivity charts to show the effects of both Southem and Northern Hemisphere weather.

Measuring Weather Sensitivity

While a weather sensitivity chart, such as shown in Figure 7-6, may appear to have a strong similarity to a standard seasonal chart, they are actually very different Weather sensitivity is a measurement of potential price volatility It could simply be the highest price that the market reached during a period in which the weather event occurred. Those more adept at statistics would want to record the increase in price from the average during those months in which weather was a factor, then show the price representing the 95" confidence level.

A thorough approadi to measuring weather sensitivity is to record the temperature eadi day and find out whether it is unusually far above or below the average, infonnation available from the U.S. weather or meteorological service will give you both regional weather and measurements of effective heat. It is the cumulative effect of heat that is damaging to a crop, rather than a high terrqaerature for one day. In addition, the amount of crop damage is not simplj a linear relationship with rising temperatures: rather, it is more likely to start slowly and increase quickly once critical levels of time and heat are readied.

6 jon Davis, Weather Sensitivity in the Markets (Smith Bamey, October I994j. TABLE 7-8 Weather-Related Events m the Southern and Northem Hemispheres

Januarj Februarj

March April

May June

Southem Hemisphere

Corn pollination in S. Africa

and Argentina

Soybean pod development

Northem Hemisphere

J peak for Florida freeze Heating oil cold or hot Heating oil cold or hot

Cotton planting Soybean planting

Cotton boil development

July

August Septeiriser

October November

Coffee freeze Brazil OJ freeze Brazil

Cocoa pod development in W Africa and pod rot in Brazil

Sugar: heat in Russia Corn pollination

Soybean pod development Atlantic hurricanes affect sugar, orange juice, heating oil Com freeze Soybean freeze Cotton harvest Coffee in Brazil rainy season Soybean harvest Com harvest Heating oil cold or hot

FIGURE 7-6 Weather sensitivity for soybeans and com.

USpod

SEASONAL FILTERS

Seasonal filters identifj periods during the year in which you would favor either a long or short position in a market based upon a clear seasonal component. These periods are normally chosen from the results of a calculation that uses detrended data; however, where there is a large difference between the median value and the average value, we should expect the median value to be more representative of expected price movement. The large difference between the median and average also tells us to expect extreme prices and high volatility during those months. This can be seen in Table 7-9, showing monthly soybean prices as a percentage of the annual average price. The double-digit values in 1973, 1977, and 1988 are enough to cause the averages at the beginning of the year to be lower than the median, and June, July, and August to be higher than the median. This bias is mainly the result of only 3 of 24 years.

Using the examples of heating oil and soybeans, which have been discussed in the previous section, we can select the months with the laiest and most reliable seasonal price moves as the best choice for seasonal filters. We can also separate the year into upward and downward cycles to allow for a wider range of trades. The following periods could be considered for heating oil and soybeans:

Choice ( )

Long from August 1 to November I ort from November I to March I LongfromMarchltoJulyl Short from July I to November I

Futf-rearSeosorxif Patterns

Long from April I to October I aiort from November I to March I Long from March I to July I Short from July I to March I

one praaical note needed here. Although ttiere may appear to be a statiatical benefit to selling healing oil in ttie later part of winter, ttiene is also extreme risk. Prices jumped 40>o from March to April 1982 and 20«o during ttie same period in 1983. We must also consider ttiat ttie Gulf War. which pushed prices higher from August 1990. conformed by chance to ttie seasonal pattern; in reality, it could hate happened any time during ttie year. These special situations do rot alwajs cause prices to rise. In 1986. Saudi Arabia decided lo dump oil in a polilically motivated elfort to drive prices down, successfijlly pushing cmde oil under $ 10 per barrel.

Years wiUi Similar Characteriatics

Seasonal studies often yield results ttiat are not as clear as desirable, and ttiese results may be rejected because of ttie obvious lack of consistency Often, ttiis is caused by a few years ttiat conflict witti ttie normal seasonal patterns °5 4»lling Oie data into analogous years can give strikingly bettei

7 David Handmaker. "Low-Frequency Filters in Seasonal Analjsis,"Journal of Futures MaAets, Vol 1 No 3 (John Wiley Sons, New York. 198 .

For example, crop production is primarily determined by weattier. Poor weattierwill cause sharp rallies during ttie growing season, whereas good weattier results in a dull, sidewajs maiket. Bad weattier develops slowly A drought IS not caused by ttie first hot dsy but prolonged dsjs of sunshine and no rain. Similarly, delayed planting due to wel fields or a late wirter will set ttie stage for an underdeveloped crop. A trader can see ttie characteriatics of a weattier market in time to act on it.

In Figure 7-7a. ttie seasonal com pattern has been separated into ttiose years witti good weattier and ttiose witti bad weather. In a later sludy by ContiCommodily, soybean seasonality was separated into bull and bear yeare (Figure 7-7b). Bull years include all bad weattier years but also years witti such incidents as ttie 1973 Soviet grain sale. Because ottier events are not confined to ttie growing monttis, ttiey may distort rattier ttian clariij ttie patterns. Bear would