NAPERVILLE, Illinois, Feb 8 (Reuters) - Top exporter Brazil’s soybean harvest will be smaller than originally thought after excessive heat and dryness clipped yields in key growing areas, but much of the recent crop cuts has been offset by stealthy expansions in last year’s output.

This trend with previous harvests has been ongoing for seven years, a span that contains both great and poor crops, and it likely reflects that demand, particularly exports, has been pushing maximum limits imposed by prior supply assumptions.

Additionally, benchmark estimates from the U.S. Department of Agriculture and its Brazilian counterpart Conab vary drastically when it comes to Brazil’s 2023-24 soybean potential, though the difference is not unprecedented.

USDA on Thursday reduced Brazil’s 2023-24 soybean harvest to 156 million metric tons from 157 million last month, disappointing analysts looking for something closer to 153 million. Earlier in the day, Conab slashed the crop to 149.4 million tons from 155.3 million in January.

In October, USDA had Brazil’s 2023-24 soy output at 163 million tons and the 2022-23 crop at 156 million, but the agency on Thursday bumped last year’s harvest to 162 million tons from 160 million estimated in January, basically erasing the effects of this year’s crop losses on the balance sheet.

That will mark seven consecutive harvests where USDA’s final Brazilian soybean figure is larger than what was reported in the October after harvest, the first month of the new global soybean marketing year.

USDA’s 2022-23 Brazil soy harvest peg is up nearly 4% in the last four months, more than in the previous six years, where final production was as much as 3% higher than in October. In its report on Thursday, USDA cited nearly finalized crush and export data causing the lift in last year’s output.

It is possible USDA’s reluctance to make the expected 2023-24 crop cuts is tied to this tendency for production to drift higher after the fact due to bigger demand numbers, and this phenomenon has triggered a full data overhaul in the past.

In February 2019, USDA revised nearly 20 years of data on Brazil's soybean stocks after the prior year’s exports far outpaced what was possible under previous supply ideas.

Conab did something similar a while later, withholding its soybean supply and demand table from its monthly reports from early 2020 to mid-2021 while data reevaluation was ongoing.

AGENCY DEVIATIONS

USDA’s 2023-24 Brazilian soy crop estimate sits 6.6 million tons (4.4%) above Conab’s, a difference equivalent to about 243 million bushels - slightly smaller than U.S. soybean ending stock levels in the past three seasons.

This is the second-largest February deviation between the agencies since 2022, when USDA’s figure sat 6.8% above Conab’s. Differences did not exceed 1.5% in other recent Februarys. It is not clear which agency was closer in the end because they currently hold a 5 million-ton difference in 2021-22 production, USDA being higher.

But Conab has a slight skill advantage over USDA when comparing each agency’s February number against the final number reported by that agency. Between 2016 and 2023, Conab’s February projection deviated by an absolute average of 4.4% from its reported final, and USDA’s February estimate held a 5.5% difference with its own final number.

For context, a 5.5% difference on a 155 million-ton crop amounts to 8.5 million tons (313 million bushels).

Conab has not changed its 2022-23 soy crop peg from 154.6 million tons since at least October, when it started issuing forecasts for 2023-24. Conab and USDA have begun maintaining increasingly different final Brazil soy production numbers in recent years, and it is not clear why. Karen Braun is a market analyst for Reuters. Views expressed above are her own.

(Writing by Karen Braun Editing by Matthew Lewis)