Early harvest low?
A telephone call from a reader prompted me to do some research I have been considering for a long time. He questioned whether the harvest low in soybeans had ever been in September. My immediate reaction was that it has a very strong tendency to be in October, but that the information to give a precise answer was easily available.
To understand how I come about this and other seasonal information, you need to know how my data is acquired and organized. Soybean trends are based on the closing price of May futures at the CBOT. I keep these in a big Excel spreadsheet with the years in columns and the days in rows, beginning with May 1, 1979 in the upper left corner. As of the end of April this year, there are 27 columns. Since there are 20 trading days in most months, there are 240 lines for each year. This makes the spreadsheet 27 columns wide by 240 lines long. That is a lot of prices to sort through.
To get the information I wanted to answer this persons question, I went to the chartroom link on my website and determined the approximate date for the harvest low and top of the "dead cat bounce" in each year. That made it much easier to pinpoint the seasonal turn dates and prices for each year from the massive number of prices available.
Marketing research is not an exact science. Some months have 19 trading days. Some have 21 or 22. That means I have to shift prices up or down in the column occasionally to fit the calendar. Going back to 1986 in my observations, there are at least two years when prices did not fit the typical pattern. The marketing year 2003-2004 saw May futures rally from the summer through the following March to over $10 per bushel. The year 1991-1992 showed a bounce that lasted only five days. In some of the other years I had to make a judgment call as to when the exact low or high was.
For most recent years, I also have a chart of the dead cat bounce to see how cash prices stacked up compared to futures. Individuals who follow my work know that I track cash prices beginning around October 1 because in most years, a big part of the bounce is basis-driven. Futures prices do not always give a true picture of what happens in the country. Unfortunately, I could not find the chart for 1999-2000 or 1990-1991. Back a few years ago, I had a disk drive crash. I suspect that information might have been a victim of a computer breakdown. It is probably stored on a backup disk somewhere which I could not find.
Research shows that 16 times in the last 21 years the harvest low has been in October. It was in November four times. In 1999-2000, the harvest low was the last week of September. Even that year, there was a low in October that was very close to the one in September. This does not predict when it will be this year. It only indicates what happened historically. Still, to be this consistent indicates there are fundamental factors that pressure the market in October.