Orange juice volatility, not what you think

Trading Places (Image via RottenTomatoes.com)

by Arthur Field

It is New Year’s Eve. The Harrisburg Express speeds from Philadelphia to New York City. It thunders down the rails carrying passengers Inga from Sweden, Naga Eboko from Cameroon, Lionel Joseph, an Irish priest, a gorilla and Beeks, with his briefcase. The case contains the crop report for orange juice.

Possibly every single reader of Futures can remember every major scene from the iconic movie "Trading Places." Indeed, it made many of us take up trading futures and commodities. We’ve all likely said some variation of the great line: "Pressure? Here it’s kill or be killed. Make no friends and take no prisoners. One minute you’re up half a million, the next — boom! — your kids don’t go to college and you’ve lost the Bentley!"

"Trading Places" culminates in a wonderful scene in the frozen concentrated orange juice futures pit in New York. Back in the old days, all trading was pit trading. It was hectic and exciting. How exciting? There were times in the 1980s when, in an effort to get an edge, the runners were on roller skates.

The upshot in the movie was the main characters, Winthorpe and Valentine, waiting until Wilson drove the price of orange juice way, way up. They then sold into the market, the Secretary of Agriculture announced the real crop report and the price plummeted. Our heroes cover their shorts and make millions, putting the Dukes into the poorhouse at the same time. It’s a fun story that defined trading for a generation. The movie never gets old.

It also has some good lessons, particularly for the orange juice trader. One topic is the issue of volatility and some widely held beliefs. Does orange juice really swing wildly in January because of the crop report? More important, if such volatility does exist, can we use it to our benefit?

To answer these questions, the monthly range of orange juice prices was recorded starting with January 1996. The monthly range is defined as the high price of the month minus the low price. The closing prices were disregarded for the purposes of this test. Through Nov. 30, 2011, there were 191 pieces of data. The highest trade during the sample period was $2.0940, observed during the month of March 2007. The lowest price paid was $0.5420 in May 2004. The range of the orange juice contract, therefore, is $2.095 – $0.542, or $1.553. In percentage terms, the high is an increase of 286.7% over the low price.

In performing this test, it is important to make uniform comparisons. A 10¢ range with a low of 70¢ is not the same as a 10¢ range when the market is trading at $1.65. The former is a 14.28% range, which is far more volatile than the 6.06% range of the latter. Consequently, all of the range values were converted into percentages by dividing the range by the low price to normalize them across the sample universe. With these data, it now is possible to construct a meaningful statistical test.

If the calendar plays a significant role in orange juice price volatility, we would expect certain months to have much more volatility than others. To test this, we use an Analysis of Variance with an F-test statistic. The null hypothesis is, there is no effect.

For those whose statistics knowledge is a bit rusty, the F test examines different dependent values and assumes there is no significant difference between them based upon the observation of independent values. In other words, all of the fixed variables (that is, the different calendar months) will produce essentially the same dependent observations (that is, the average volatility during the month). The null hypothesis of the test is there will be no statistically significant difference between the mean volatilities of any of the months. The alternate hypothesis is that, yes, there is. Although the F statistic measures the variance, if the variance is low, it follows the means are essentially similar. If the variance is high, the mean volatility must be significantly different among the 12 months.

The Analysis of Variance test has 12 independent variables and 16 observations in each month. (Although December is missing for 2011, it is easy to perform this slightly unbalanced test using any statistical software package.) The F statistic will have 11 and roughly 180 degrees of freedom. This is a powerful test for discovering patterns, and we can have much confidence in its results.

The average volatility for each month is shown in "O.J. monthly" (below). The results of our Analysis of Variance test was p<0.00001. Essentially, there is virtually no significant chance that the means of the months are equal. We can conclude with great confidence that the calendar does indeed play a major role in the average volatility experienced in the orange juice market over time.

Click to enlarge

It is essential to realize that this does not prevent high volatility from taking place in any month. The test simply compares the averages and finds there is more variation between the months than within the given months.

In a blow to Hollywood, and perhaps running against the assumptions of most orange juice traders, the "Trading Places" months of December and January were not the top of the volatility list. Examining the table, it is easy to discern that October, August, January and December experience greater than average volatility, but October is much higher than December. Those four months account for 17.17% of the volatility of the orange juice market.

March, April and November are relatively quiet months by comparison, experiencing only 12.3% of the volatility. It may be concluded that the best trading opportunities are likely to be presented when volatility is greatest. It might also be concluded that the greatest risk occurs at those times as well. Short-term traders might find lots of trading opportunity in orange juice during October or August. March and April may be too quiet, but also present less risk.

On average, volatility throughout the year is 14.64%. As a side note, we can examine if extremely low volatility has been a harbinger of any movement. Indeed, it turns out it is, statistically speaking. When volatility in any month dropped below 7.25%, the market often changed direction within the three months thereafter. Often, the low volatility immediately preceded a major market shift.

For example, in December 2006, the monthly low was $1.96 and the high was $2.0940 (see "Calm before the storm," below). This range of 13.4¢ was only a 6.8% swing for the month. Given December is expected to be a pretty wild month, the volatility was less than 45% of what we might normally see. January 2007 saw a major drop in the market, with volatility falling to 13.35%. The all-time high was hit only two months later in March 2007, and then the orange juice market dropped precipitously, falling to $1.1060 in the next six months.

Similarly, in April 2004, the volatility dropped to 4.42% and the market low for the month was $0.5580. The following month, orange juice bottomed at $0.5420 before beginning its wild climb into the 2007 high. Several other examples of such predictive low-volatility activity may be found.

The lessons of the orange juice market may be extrapolated to other markets and shorter time frames. The trader easily can test different hours of the trading day to see if certain periods present higher volatility than others. Similarly, it may be tested to see if sudden contraction in the range of any time bar vs. its historical norm predicts a sudden change in market direction. After all, the goal, as Valentine and Winthorpe say at the end of the movie is: "Looking good; feeling good!"

See the original article >>