A series of acreage surprises in Crop Production reports last year has put the integrity of USDA’s data in the spotlight, with the department moving to update its methodology in an effort to reassure producers and policy makers. Crop markets tend to get the most attention but livestock reports are market movers, too, which means it’s worthwhile to examine their accuracy.
Each year the USDA is required to publish the Livestock Historical Track Records, a report that compares the difference between their initial forecast that impacts markets the most to the final estimate after any revisions are made in the following years.
Head counts
Estimates for the January 1 cattle herd have been remarkably accurate in recent years. In absolute terms, from 2021-2025 USDA made an average revision of 148,800, or 0.16% of the total herd size. In real values, the agency has typically pegged their initial estimate too high, with three out of the five years seeing revisions to the downside. Since 1980, the average adjustment from initial to final estimate has been 603,000 head, or 0.61% percent of the total herd. The track record shows the accuracy in recent years has improved with smaller shifts, especially compared to the multi-million head adjustments seen in the 1990’s.
Estimates for the December 1 inventory in hogs have been slightly less accurate, with deviation between the initial and final figure being 768,000 head, or 1% of the herd in absolute terms. That’s right in line with the longer-term historical averages that see the initial estimate missing the true herd size by 1% since 1980.
Modernization focuses on improving crops more than livestock
USDA has announced plans to review the statistical process and modernize data collection. So far, action items spelled out in USDA releases have focused on the potential to increase the use of satellite imagery for crop acreage and yield estimates, and modernization of the reporting process to reduce the need for National Agricultural Statistics Service (NASS) surveys where Farm Service Agency and Risk Management Agency data are being collected already.
While those changes have the potential to improve accuracy on the crop side, they offer little change to livestock estimates. Identifying cattle via satellite imagery introduces new operational issues, such as animals hiding under shaded areas or blending into the surrounding environment, which make the method less reliable on livestock than it is for crops.
Similar issues arise with the use of administrative data. Crop farmers are heavily incentivized to report acreage and in some cases production data to agencies outside of NASS with some government insurance programs and other financial tools mandating reporting to be eligible. Programs for livestock farmers also require detailed inventory records to collect payments in most cases, but that data is rarely reported to the agencies at regular intervals. Often times data and documentation are only requested to confirm the validity of disaster or indemnity payments.
Given the accuracy of crop forecasts compared to those in livestock, the agency may be on the right track leaving cattle and hog estimation largely unchanged and choosing to focus resources on the crop side. In the Crop Historical Track Records released earlier this year, which provides similar data to the livestock release, data showed that the agency has missed corn production by an average of 2.36%, or 347 million bushels on average over the last five years from the August estimate to the January final. Soybeans have faced slightly bigger misses as a percentage of the crop, with the initial estimate missing by 3.1%, or 128 million bushels, over the last five years.
While the agency is still facing criticism, the difference between initial and final estimates has closed in recent years compared to the longer-term averages since 1980. The five-year average miss of 2.36% on corn production is down from 3.92% since 1980, with the 3.1% miss on soybeans also down from 4.68% seen over the longer-run. Given that commodity markets can be sensitive to even the smallest changes at times, and technology has improved over time, further gains to accuracy would still be welcomed and could be possible.