Pulse Market Insight #260

Crop Estimates and Track Records

It’s common practice to debate and disagree with StatsCan’s production estimates. One issue some people seem to have is the idea that StatsCan should be able to nail down exactly how much crop was produced. And if it can’t do that, the numbers are worthless. Keep in mind, StatsCan’s numbers are estimates, by definition “a rough or approximate calculation”. Unless StatsCan would come out and measure every grain bin and bag (and who wants that), these will always be approximations.

Once we’ve accepted that the StatsCan numbers won’t be exact, we can step back and look at them objectively to see if there are signals about the actual crop size, a bit like reading tea leaves. Just as a quick refresher, StatsCan issues three sets of production estimates. The first one in late August uses a computer model that evaluates satellite images of crops from late July to come up with yield and production numbers. The second estimate released in mid-September is similar, except that it uses crop conditions from late August. StatsCan releases a third estimate in early December, but this one is based on farmer surveys taken in November after the harvest is (hopefully) finished.

Each of these estimates has limitations. As a result, StatsCan will sometimes revise production numbers after the fact, sometimes as much as two years later. And contrary to the common perception that StatsCan crop estimates are too large (and depresses the market), these later adjustments are almost always higher.

When we look at StatsCan’s track record for pea production estimates, we see there hasn’t been a consistent pattern over the past few years. Since StatsCan started using its model-based approach in 2017 for the September estimates, the December number came in higher three times and was lower four times. But what is unique for peas compared to almost all other crops is that StatsCan made no or only minor adjustments after the fact. This suggests that StatsCan’s December estimates have been close to reality.

StatsCan’s September pea yield estimate came in at 37.0 bu/acre, well above last year at 32.3 and higher than the 5-year average of 33.9 bu/acre. Keep in mind that two of those five years used in the average were 2021 and 2023, with very disappointing yields. This latest StatsCan yield isn’t huge and based on other reports, could be close to the actual outcome.

Since StatsCan switched to a model-based approach in 2017 for its September estimate, the December survey-based number for lentils has tended to be lower more often than not. In some cases, by a lot. That said, there have been cases where the December estimate wasn’t the final word and the crop size needed to be bumped back up after the fact.

The latest StatsCan lentil yield estimate for 2024 is 1,379 pounds (23.0 bushels) per acre, a solid increase from the 1,100 pounds (18.3 bushels) per acre last year and better than the 5-year average of 1,218 pounds (20.3 bushels) per acre. This year, we’ve heard lentil yield estimates all over the map, making it hard to sort out the prairie-wide yield, but we wouldn’t be surprised to see a slightly smaller estimate in December.