Do pod counts predict soybean yield?
Nobody quotes a pod count when they talk soybean yield. The reason usually given is that pod counts don’t convert, and that is half true. They do not convert through any published formula. But they carry real signal, and it can be measured. We could not find anyone who had published that measurement, so here it is.
Want a yield for your own field? The soybean yield calculator works it from plants, pods per plant, seeds per pod and seed size.
What the tour actually does with soybeans
Pro Farmer’s scouts measure corn as a yield in the field: ear counts, grain length, kernel rows. They never measure a soybean yield at all. The published method is to count pods: “Count all pods on the three plants, any pod ¼″ or larger, and average.” That count is normalised to a three-foot square and published per state. No state soybean yield is published, ever.
Their stated reason is exact, and it is the reason this page has to be careful: “The number of pods it takes to make a bushel is different in each state and varies from year to year.”
The measurement
Twenty-eight state-years: seven tour states, 2022 through 2025. Pod counts as Pro Farmer published them. Soybean yields as USDA NASS finally published them, in the Crop Production Annual Summary each January, cross-checked against the state field-office releases.
Each dot is one state in one year, shown as a difference from that state’s own four-year average. Pooling every state together gives r = +0.85, but most of that is just South Dakota having both low pod counts and low yields. It measures the gap between states, not the thing anyone wants to know. Subtracting each state’s own average isolates the real question: when a state’s pod count moves, does its yield move with it?
The corn side of the same tour, run through exactly the same arithmetic, is the control. Shaded band is 95%.
| Fit | n | r | R² | Typical miss | of mean yield |
|---|---|---|---|---|---|
| Pods → soybean yield, within state | 28 | +0.66 | 0.43 | 2.89 bu | 5.3% |
| Tour corn → corn yield, within state | 28 | +0.85 | 0.73 | 5.55 bu | 3.0% |
| Pods → soybean yield, pooled | 28 | +0.85 | 0.72 | 3.95 bu | 7.2% |
| Year-on-year % change in pods | 21 | +0.70 | 0.49 | 5.6 pts | n/a |
The misses above are in-sample. Each state’s mean, which the within-state fit subtracts, includes the year being predicted, so a call made ahead of the count would miss by more than 2.89 bu. Both within-state misses are divided by 20 degrees of freedom, not 26: the seven state means take the place of the one intercept, so they spend six more.
The answer. Pod counts are about 1.8× looser than the tour’s corn reading, a typical miss of 5.3% of mean yield against 3.0%, in-sample. That is worse, and it is nothing like useless. The tour measures corn directly and never measures a soybean yield at all, so a factor of 1.8 is a small price for a number that is only ever counted, never weighed. Pod counts deserve to be quoted. They deserve to be quoted with a band around them.
The 2026 tour, read this way
Added 16 October 2026. AGSIST calibration, not a Pro Farmer number. To be checked against NASS’s January 2027 Annual Summary.
Each state’s August 2026 pod count, set against its own 2022–2025 average and turned into bushels with the within-state fit above (0.0227 bu per pod), with the ±6 bu band. The answer is a difference from that state’s own 2022–2025 average soybean yield, not a yield.
| State | 2026 pods | 2022-25 average | Pods vs average | Bu vs state average | 95% band |
|---|---|---|---|---|---|
| Ohio | 1,197.25 | 1,225.45 | -28.2 | -0.6 | -6.6 to +5.4 |
| South Dakota | 945.98 | 1,024.68 | -78.7 | -1.8 | -7.8 to +4.2 |
| Indiana | 1,318.64 | 1,315.38 | +3.3 | +0.1 | -5.9 to +6.1 |
| Nebraska | 1,219.62 | 1,186.13 | +33.5 | +0.8 | -5.2 to +6.8 |
| Illinois | 1,430.37 | 1,354.66 | +75.7 | +1.7 | -4.3 to +7.7 |
| Iowa | 1,362.93 | 1,265.38 | +97.5 | +2.2 | -3.8 to +8.2 |
| Minnesota | 1,257.80 | 1,092.40 | +165.4 | +3.8 | -2.2 to +9.8 |
What it says. Every state’s band crosses zero. On pod counts alone, 2026 cannot be told apart from an average year in any of the seven states. Minnesota leans highest and South Dakota lowest. The band is if anything too narrow: 2026 is not in the fit, and the fit’s own misses are in-sample.
Pod counts: Pro Farmer’s 2026 state results, Aug 17–20 (the state tables, checked against the percent changes in the same articles). 2022–2025 averages: the counts in the measurement above.
The divisor everyone repeats is wrong
The rule of thumb you will hear is “divide the pod count by 27.” It traces to a single 2013 DTN blog post quoting an unnamed source, and the journalist who printed it attached her own warning: “I just wouldn’t put too much stock in that yield estimate’s accuracy.” It is not a Pro Farmer number and never was.
Our twenty-eight state-years put it at 22.0 as a simple ratio of pods to bushels, and 23.5 as the slope of the pooled fit, which mixes differences between states with changes within a state. To turn one state’s change in pods into bushels, use the within-state fit from the table above instead: about 44 pods to the bushel (0.0227 bu per pod). The 22.0 and 23.5 sit inside the 21.8–25.8 range implied by Pro Farmer’s own published national pairs. Three separate routes land in the same place, and it is not 27.
Worth knowing if you are tempted by a textbook: the extension formulas (Purdue, Ohio State, Iowa State, and USDA’s own objective-yield conversion) give 15 to 18 on this basis at 3,000–3,600 seeds a pound; with larger seed (2,500 a pound) the same arithmetic gives about 12. Divide a Pro Farmer count by 15 and you will overstate yield by 47 to 57% (22.0 and 23.5 against 15); at 18 the overstatement is 22 to 31%. The reason is that Pro Farmer counts any pod a quarter-inch or larger off three plants on fields chosen by the route. Different scale, on purpose. Never import those constants.
What would break this
Four years is four years. Leaving any one state out moves the correlation between +0.60 and +0.74, which is stable. Leaving out a year is the worry: drop 2025 and it falls to +0.47. One season is carrying a lot of the signal, which is exactly what four points per state should be expected to do. More years would help, and the data exists: every Pro Farmer nightly report carries a year-ago comparison, so each report read back adds two years.
Pod weight compensates, and the tour cannot see it. USDA’s own eleven-state pod counts fell 12.6% from 2018 to 2019 while yield fell only 6.3%, because the average pod got heavier. Seed size alone swings a yield estimate by about 17% with no change in pod count whatsoever. The tour counts pods and never weighs them, so that whole term is invisible to it. This is the structural ceiling, and no amount of extra data removes it.
One state, one year, is a coin flip and a half. On year-over-year change, pod count and yield move the same direction 14 times in 21. The corn control manages 16 of 21. Read a single state’s pod count as a yield signal and you will be right about two times in three.
What we will not do
Print a single number and call it a soybean yield. On this fit the 95% band is roughly ±6 bushels an acre, and pooled it is a little over ±8. A point estimate inside a band that wide tells you very little. If we ever publish an estimate it will be a range, it will carry the band, and it will be labelled as ours: a calibration, not a translation of anything Pro Farmer said.
Where the numbers came from
Pod counts: Pro Farmer Crop Tour state results, 2022–2025, as published each night of the tour. Soybean yields: USDA NASS Crop Production Annual Summary, the January-following edition each year, cross-checked against NASS state field-office releases and Executive Briefings. Every cell sourced; none interpolated or filled in. Two cautions: Ohio and South Dakota 2025 come from the NASS State Agriculture Overview rather than a January field-office release, and NASS revised several 2021 state figures upward between the January 2022 and January 2023 prints, so 2021 is not used in any fit on this page. Arithmetic is ours; the underlying figures are not.