data validation

It’s important to reconcile the various sources and claims we’ve gathered, as not all numbers align perfectly. Here we highlight a few key potential discrepancies and how we resolved them: Discrepancy in 2003 production figures: We have multiple sources citing ~340 tons for 2003, which we used as a cornerstone. However, one often-quoted activist figure said “Three U.S. duck farms...produce 400 tons each year”[26] (NYT 2003 via PETA). Why the ~60-ton difference? Likely because the 400 figure was a rounding up or inclusion of imports. The NYT article might have been simplifying by saying 400 tons consumed (which was true) and phrasing it as produced. Our decision: we trust the detailed industry-backed figure of 340 tons produced[43] and 420 consumed[19]. The activist number 400 conflates the two or uses a high estimate. So we explicitly cite 340 for production in 2003, and ~420 for consumption, rather than 400 for production (which no industry or official source supported directly). Current production uncertainty: We estimated ~200–250 tons now. The farms themselves don’t publicize tonnage, only duck counts. In 2019, HVFG was ~500k ducks/year[24] and La Belle ~182k[22]. If each duck yielded a 0.65 lb liver average, that’s (500k+182k)*0.65 = 443k lbs = 222 short tons. If average liver weight is higher (some claim ~0.8 lb for Moulard ducks at Hudson), the total could be ~273 tons. If lower (some ducks processed early or different grades), maybe ~180 tons. Given COVID disruptions and possible downsizing, the true 2023 figure could be anywhere from ~150 to ~250 tons. We gave a range and lean towards the lower end (~200) considering ongoing challenges. In cross-check, recall Wikipedia’s table had U.S. at “250 (tons, year?) = 0.9% of world in 2014”[87]. If world was ~26,396 tonnes in 2014[88], 0.9% is ~238 tonnes (which might have been rounded to “250” in the table). That lines up with our estimate for around that time. Extrapolating, a decline from ~238 tonnes (2014) to ~180–200 tonnes (2022) is plausible given the ban/pandemic impacts. Number of farms vs output share: Earlier sources said “four U.S. facilities as of 2012”[31] (HVFG, La Belle, Sonoma, + the small MN farm). Yet output was dominated by two (NY) by then. We double-checked that no other unknown farms existed. It appears not – often folks mis-cited “four” including a defunct farm or counting a second Hudson Valley plant. The Minnesota farm is the 4th. So we confidently state only 2 significant producers after 2012. Any claim of more farms is usually a confusion. Global share: Some sources in 2018 said “US and Canada use 500k birds/year”[32]. We know HVFG was 350k and La Belle ~150k then, and Canada maybe 100k, which sums ~600k. That’s in the ballpark. Another source said “US production ~2% of world”[13]. In 2015 world production ~27,000 tonnes[11], 2% would be 540 tonnes – far above actual. But likely they meant earlier or including Canada. More accurately, in 2003 US was 1.4% of world[89]; in 2014 maybe ~0.9%[87]; in 2020 possibly similar ~1%. We have reconciled by noting US <2% global in general, and not over-emphasizing that stat. Import/export figures reliability: Official USITC data (for example, via DataWeb) was used wherever possible. We cross-verified Shepstone’s import value graph[77] with OEC’s recent data[14] and narrative accounts (like French trade reports noting US volume). All signs point to consistency: imports were never huge (max few million $), and lately under $1m/year. The import breakdown (France vs Canada ~50/50 in value in 2003[49]) also fits: French product is higher unit value (goose, or canned luxury), while Canadian is more volume of duck at lower price. If anything, we erred on side of caution not to overestimate import tonnage. Conflicting economic impact statements: The farms have an interest in emphasizing their significance (jobs, local economy). For instance, Shepstone’s report claims foie gras was 45% of NY’s meat poultry production by value in 2002[51] – a striking number showing how niche the rest of poultry is in NY. We include those context bits as color but rely on the physical volumes for the quantitative core. When industry said “we supply 85–90% of the US market”[44], we checked that against import share to ensure it aligns (it did, imports ~10-15% share). Mortality and welfare stats: While not central to production quantity, some sources (HSUS, Farm Sanctuary) mention up to 20% pre-slaughter mortality on foie gras farms[90], etc. If true, that means for every duck that yields a liver, 0.2 ducks died prematurely. This could slightly affect the “ducks used” count vs liver output. We didn’t delve deep into that, but our duck counts likely refer to ducks processed, not including on-farm losses. The differences aren’t enough to change tonnages notably (maybe needing 10% more ducks input to get same output), but it’s a nuance to note in fine print. Methodological consistency: Whenever possible, we kept internally consistent assumptions. For example, using short tons consistently for U.S. measures. If a source gave metric, we converted. This avoids confusion like one report citing 300 metric tons vs another 300 short tons (which differ by ~10%). We’ve clarified in text where needed. Final estimates vs exact data: In presenting “best-estimate ranges,” we aimed to cover the plausible span. For instance, stating current production as “~200–300 short tons per year” acknowledges uncertainty. We lean on midpoints for narrative (e.g., “around 250 tons in mid-2010s”), but tables show ranges or approx. At each key juncture – 2003, 2012, 2019, present – we triangulated: - 2003: Shepstone/AVMA (340 prod, 420 cons)[43][19] vs activists (400 prod)[26] vs FAO (1.4% world = ~340)[87]. Conclusion: use 340/420. - 2012: HSUS (250k ducks at largest farm)[31] + 4 farms listed[31] vs industry (Sonoma closure, 85% NY share) – indicates ~240 tons production. - 2019: Duck counts (500k + 182k) vs reported % to NYC (25%) vs global context. We deduced ~300 tons max, and consumption ~320 with imports. - 2022: Reduction narrative + OEC trade confirmation (low trade) told us consumption and production both down somewhat. Conclusion on credibility: The sources used (USDA/AVMA, industry report, government letters, reputable journalism) generally align on orders of magnitude. There is no glaring conflict like one saying “1000 tons” vs another “100 tons.” The differences were at most 15–20%, which we accounted for in ranges. Therefore, the reconstructed time-series is robust in that even if any single source was off, the overall trends and approximate levels are supported by multiple data points. We thus have high confidence in stating, for example: - U.S. apparent consumption in the early 2000s was on the order of 400 short tons/year[19], - Current consumption is on the order of 250 short tons/year (by extrapolation and recent trade data), - Peak production ~340 tons[43], current production ~200–250 tons, - Imports peaked at tens of tons (never hundreds)[77], exports always just a trickle[15]. Where uncertainty remains (like current exact tonnage), we are transparent about it.