The most recent Cattle on Feed report was released Friday, March 23. Though this was after press time, predictions abounded last week on what the report would hold.
Average predictions from the Urner Barry survey reported in the CME Daily Livestock Report placed on-feed populations in feedlots with a capacity of 1,000 head or greater on March 1 at 108.1 percent of March 1, 2017; placements of cattle into feedlots during February were estimated at 104.5 percent of last February; and marketings of cattle out of feedlots during February were estimated at 101.4 percent of February 2017.
Pre-report predictions, particularly of placements, have been points of concern lately. In the last 10 months, average placement predictions have been 2.5-12.6 percentage points off actual placement numbers. Since large differences between pre-report expectations and reality can have jarring effects on the feeder cattle futures, this seeming trend has drawn increasing attention.
In all but one case, the average predictions underestimated placements. In some cases, such as May, July, and December of 2017, these underestimations were quite severe at 9 or more percentage points off reality. In four cases—the three previously mentioned and February 2018—the actual placement number was higher than the highest end of the estimate range.
But this seeming trend of underestimating placements might not be what it seems.
“It is correct that we, as analysts, have underestimated, but that’s kind of why USDA does the survey,” Jim Robb, Director and Senior Agricultural Economist at the Livestock Marketing Information Center (LMIC), told WLJ speaking of the Cattle on Feed report.
“Placements are very difficult to estimate, especially when you’re having a drought year in a significant part of the country. Last year, we had drought in the northern Plains and this year we’re having it in the southern Plains.”
The pre-report estimation process is generally a survey of a variety of analysts coupled with historical data and recent events. The Urner Barry survey, for instance, reflects 11 different analysts’ expectations. Bloomberg and Reuters also put out pre-report estimates with slightly more and different analysts.
“Largely, nobody’s got proprietary data doing these pre-report estimates,” Robb explained. “Groups are doing this largely off of secondary data sources and past USDA reports with those seasonal patterns.”
He also explained that pre-report estimates reflect the quality of the data used to make the estimates, as well as many outside factors.
“We look at the feeder cattle auction movement, we look at the data that comes out from CattleFax that they publish weekly, we look at historical relationships, we look at year-to-year patterns, we look at drought—there’s about eight factors that we tend to look at.”
Heifer placement, availability of feed or pasture, and the point in the cattle cycle can also affect pre-report estimates.
Robb agreed that the past 10 months of pre-report placement predictions so consistently underestimating reality was “getting to the point of a little bit of interest” and gave the impression of a pattern on the short term.
“But let me play statistician with you: How many times can you flip a coin in 10 minutes where you’re going to come up with five tails in a row? It will actually be fairly likely if you flip it long enough. And you don’t have to flip it very long.”
In short, the appearance of a pattern could be due to a relatively short-term focus. A longer-term perspective would reduce the appearance of a pattern. When it comes to market trends, longer-term perspectives are best.
“If you look long term, [the accuracy of predictions] tends to be bouncing around,” Robb noted.
He also pointed out that, in the case of placement predictions, how wide the prediction range is can be very telling. In general, the wider the range, the less certain the surveyed analysts are about the prediction.
Based on records compiled and analyzed by WLJ, the average range of predictions for placements in pre-report estimates since February 2017 has been 10.5 percentage points. By comparison, the average ranges for on-feed populations and marketings estimates during that same time were 2 and 3.5 percentage points respectively. This shows much greater uncertainty about predicting placements than the other two categories.
Robb called the pre-report estimates a moving target, but ultimately not where industry focus should lie. Instead, he stressed value of the actual USDA surveys like Cattle on Feed the estimates try to predict.
“If the industry went on pre-report estimates only, they’d have been wrong for the last year,” he said, again acknowledging the recent predictions that didn’t closely reflect placement reality.
“This is why you shouldn’t rely on pre-report estimates; you should rely on what USDA says.” — Kerry Halladay, WLJ editor
