Recent breach in foreign report due to low-level mistake USDA crop forecasters weren't always so careful with their numbers. Early last century, a conniving USDA employee schemed with a market trader to make some ill-gotten money.
"As I understand it, this guy was pulling a shutter trick. He'd raise or lower a window shutter and that was a signal to his partner outside the building about what the forecasts were. His partner was taking that information to the market - he was doing insider trading," says Tommy Gregory, the head of Mississippi's Agricultural Statistic Service office in Jackson.
When the scheme was discovered, USDA decided security precautions were a must. Those stringent precautions - improved upon over the decades - are still in place today, making a recent lapse by NASS' sister agency, the Foreign Agricultural Service, newsworthy.
On Oct. 11, a crop forecast for five countries was released a day early. What did the forecast contain? Among other things:
- The 2000-01 Argentine soybean crop was pegged at a record 22.6 million tons. That number is up from the September estimate of 21.5 million.
- The 2000-01 Brazilian soybean crop was estimated at 33.5 million bushels, up from the September forecast of 32.8 million.
- The Australian wheat crop is expected to be 21 million tons, which is 2 million less than September estimates.
- The Pakistani cotton crop is estimated to be 8.3 million bales - up 600,000 from September.
The breach According to USDA sources, the breach didn't occur for nefarious reasons, but due to a simple mistake. An FAS analyst wrote a "highlights-type" report based on the early estimates coming from a committee meeting. That report was placed on the desk of an information specialist who, failing to realize it was meant to be sent out the following day, issued the report immediately.
"FAS prepared some of their numbers before lock-up. That's how it got the chance to be released early. If they'd waited until lock-up, it would have been impossible for this to have occurred," says Gregory.
What happened was against regulations.
"FAS had an inexperienced employee - a computer technician essentially - whose job was to disseminate information. He messed up and the person who trained him apparently didn't do too great a job either."
Although the National Agricultural Statistics Service issues reports only on U.S. estimates and had nothing to do with the recent lapse, Gregory says his agency and FAS compile reports in a similar manner.
"NASS doesn't even begin preparing survey pages in D.C. (at USDA headquarters) until we lock up. We lock all doors, flip a switch that shuts off all communication lines, shut down elevators, use special wire shutters on the windows, and post armed guards. No one gets in without proper passes - including the secretary of agriculture," says Gregory.
Tight security There's a national design to what NASS does. Take cotton, for example. From Gregory's office in Jackson, data is sent to USDA headquarters containing the numbers of what Gregory and colleagues believe the Mississippi cotton yield to be.
Those recommendations are submitted electronically and encrypted. The files are kept in an encrypted form until the office locks up late at night the evening before a report is released.
"In the lock-up area there's an armed guard outside the door. All phone lines are cut off. It's impossible to get in or out. The elevators even have large doors on hinges that swing shut and are locked. If an elevator comes up to the floor and the door somehow opens, the person inside the elevator is then confronted with a locked gate."
Security is so tight, says Gregory, that "it's kind of spooky" under lock-up.
Once the statisticians are locked inside, the files are unencrypted. Until that time, the statisticians have seen no data.
"At that point, they get all state recommendations and add them up to see what the U.S. levels are. The surveys we do at state level are pooled there also."
What about security in state offices?
"We don't go to the D.C. extent, obviously. However, when we make recommendations, only a few staff members are allowed to work with the numbers. The data we work with is locked up. We don't let people wander our offices. Doors are kept locked. We take quite a few precautions."
A typical month In Mississippi, cotton is the only crop for which NASS sends people out into fields to make counts. "We first go out in June for a base survey. The entire state is divided into one-square-mile segments. There are about 12,000 of those in the state. My office building here in Jackson is in one of those segments," says Gregory.
NASS draws a sample of those segments and weighs it towards cropland. So it's much more likely for a farmer in the Delta to have his land selected for a survey than for Gregory's office building to be picked.
"But I have a chance of being selected and if I'm growing cotton in the yard here, there's a known probability of being selected."
NASS employees go to about 420 segments out of the 12,000. People are sent to interview farmers in person. After speaking with farmers, they identify every field, whether cotton, soybeans or corn.
"That gives us a probability survey. Then, inside the computer, we put all the 420 segments and randomly select cotton fields to watch. We have sample units falling across all the segments. Since we weighted those segments to cropland, more of the cotton samples are going to fall in the Delta than will fall in the hills," says Gregory.
Setting up "In the first month of doing this, we tell employees to go to the field's most accessible corner (the `starting corner'). We're setting up a small section to watch counts on. They then walk a random number of rows across the field and a random number of steps into the field. At their last step, they stop and put a yardstick down across the row. They then take a tape measure and pull out 5 feet beyond that and begin their count on a small marked-off area."
On about the 25th of the month, the plants within each sample unit are studied. "They count plants, small bolls, squares, large bolls. They use a 1-inch boll gauge. If a boll won't go through the 1-inch diameter, it's labeled a large boll. If it passes through the ring, it's called a small boll."
When the cotton opens, it's picked and mailed to Jackson where NASS dries and weighs it. This process is called the "objective yield survey." It starts on the 25th and the process is completed on the 5th of the next month.
"In addition to that, on the first of the month we call around 1,000 farmers, asking what they expect their yields on cotton, soybeans and corn will be. We weight this survey to large growers."
What's a large grower? "The higher your acreage, the more likely you'll get a call from us. I don't know what the strata boundaries are exactly. But farmers with around 10,000 acres will all be called. Fifty percent of the next group down will be called. It stair-steps down from there. People with 100 acres or less will have a call ratio of one out of 50. It's heavily weighted toward larger producers."
The first calls for this survey are on Aug. 1. If NASS can't get a farmer on the phone, someone is sent out to knock on his door. Gregory says such efforts lead to a high response.
Gins are also contacted - usually starting in September. "We ask how many bales they've ginned so far and how many they expect during the season."
Shipping data After gathering the data and punching it into a computer, Gregory and colleagues have to submit their crop estimates to USDA on the 7th or 8th day of the month. All except cotton. "Numbers for cotton - since we're a leading cotton state - aren't submitted until just before the report is released. The reports are published on the 12th of each month. My recommendations are due at noon on the 11th."
Those at the national headquarters spend the rest of the day making sure all states' encryption files are there. Later that night, they lock up and open the encrypted files.
"They'll read my numbers, my comments and review our office's work. They then release a report at 7:30 a.m. our time the following day."
Mississippi's NASS office provides estimates on corn, soybeans, cotton, hay, sweet potatoes and wheat.
"Cotton is the only one that we're important enough to send numbers right before the report comes out. With the other crops, we're able to send recommendations in two or three days prior to the report coming out."