Coffee Disappearance Recovered After 1995 — Soft Drinks Peaked, Then Prices Outran Both

Did Americans keep drinking more coffee and soda while shelf prices jumped? The short answer from USDA disappearance and BLS price indexes is no — not as a single rising volume story. Coffee green-bean pounds per capita recovered after a mid-1990s trough but never reclaimed the early-1970s levels. Carbonated soft drinks rose hard through the 1980s–90s, peaked near 53 gallons per person in 1997 on the USDA Statistical Abstract series, and then drifted lower on the sparse industry extension that fills the gap after ERS removed Beverage Marketing Corporation soft-drink files from the Food Availability system. Meanwhile the BLS nonalcoholic beverages CPI (CUUR0000SAF116), rebased to 1990 = 100, sits near 229 in 2025 — more than double the base year — and coffee’s own item index has accelerated again in 2025–26.

The interactive dashboard above races volume indexes against beverage CPI, then flips through decade bars, a price–volume scatter, era snapshots, volume−price gaps, and a selected-year table. The rest of this piece walks the sources, the split between coffee and soda, and the caveats that keep the story honest.

What “disappearance” actually measures

USDA Economic Research Service food availability (often called disappearance) is a balance-sheet residual: production plus imports and stock changes, minus measurable nonfood uses and exports, divided by population. It is a national supply available for consumption proxy, not a household diary or scanner panel. For coffee, ERS converts roasted and instant forms onto a green-bean equivalent and also publishes retail-weight pounds. Fluid gallons in this post follow ERS’s documented conversion of 60 six-ounce cups per retail pound of regular roasted coffee (and the parallel instant factor in the source table), which works out to 2.8125 gallons per retail pound.

That framing matters when prices enter the frame. CPI tracks what shoppers pay for a fixed market basket. Disappearance tracks how much coffee (or soda) the marketing system absorbs. The two can diverge for years: prices can spike while volume stagnates, or volume can boom while real prices ease. The question here is specifically whether per-capita volumes kept rising through the same decades when beverage CPI climbed — and especially through the post-2019 inflation spike.

Coffee: a trough, then a partial climb

On the archived ERS coffee table (series through 2015), green-bean per capita starts the modern window high: about 13.6 lb in 1970. It falls through the frost-shock 1970s, drifts near 10 lb in the 1980s, then slides to a trough near 7.9 lb in 1995. From that low to the last ERS print in 2015 (10.2 lb), coffee disappearance rises roughly 30%. Indexed to 1990 = 100, 2015 coffee volume sits near 99 — essentially flat versus 1990 — while nonalcoholic beverage CPI on the same base is already near 185.

So coffee’s post-trough recovery is real, and it is not a continuous boom from 1970. Specialty café culture, single-serve pods, and foodservice roast demand can coexist with a per-capita green-bean series that never recovers the Vietnam-era peak. The dashboard’s dual-axis race makes that visible: the amber volume area rebounds after 1995; the rose CPI line keeps walking higher.

ERS has not refreshed the coffee availability workbook past 2015 in the archived FADS download. That is a hard stop for official disappearance, not a claim that Americans stopped drinking coffee. Any 2016–2026 coffee-volume story would need import, roast, or industry panel data with different definitions — outside this post’s USDA spine.

Soft drinks: the boom that stopped

Carbonated soft drinks tell the opposite mid-period story. The USDA / Statistical Abstract gallons series (sourced to ERS Food Consumption, Prices, and Expenditures and related unpublished worksheets) rises from 24.3 gallons per capita in 1970 to 46.3 in 1990 and 53.0 in 1997 — the peak year in that published government table. Diet and regular both participated; the Abstract’s diet share grew from a couple of gallons in 1970 to more than 11 by the mid-1990s while regular kept climbing.

Then the official series ends. At Beverage Marketing Corporation’s request, ERS removed BMC soft-drink (and related) series from the Food Availability website, and the Census of Manufactures path that briefly replaced BMC also ages out of the Abstract’s annual refresh. For the post-1998 window this post therefore uses sparse industry-extension markers (openly cited Beverage Digest / IBISWorld-style levels): roughly 53 around 2000, drifting toward the low-40s by the early 2020s (42.9 in a 2021 print; 41.9 in a 2025 estimate). Those points are labeled as extensions in the dashboard table — useful for direction, not for pretending ERS still publishes annual CSD disappearance.

The directional punchline does not depend on the tenth of a gallon: soda volumes peaked before the big CPI acceleration of the 2020s, and the extension sits below the 1990 USDA level even as beverage CPI more than doubles.

Prices: the line that did keep rising

BLS CPI-U nonalcoholic beverages (CUUR0000SAF116) annual averages move from 129.3 in 1990 to 252.5 in 2019, 295.6 in 2025 (eleven published months; October missing during the appropriations lapse), and about 300.1 for the 2026 year-to-date average through July. On a 1990 = 100 scale that is roughly 195 by 2019 and 229 by 2025 — a steady march, then a sharper post-2021 step.

Item indexes sharpen the caffeine lens. Coffee CPI (CUUR0000SEFP01) is famously jagged: the mid-1990s commodity spike, the 2011 run-up, a softer late-2010s stretch near the 2019 average of 193, then a jump to 266 in 2025 and a 2026 YTD average near 298. Carbonated drinks CPI (CUUR0000SEFW01) is smoother: 123.6 in 1990, 230.1 in 2019, 278.5 in 2025. From 2019 to 2025, coffee’s item index rises about 38%, soda’s about 21%, and the broader nonalcoholic beverage index about 17%.

Volumes did not mirror that price path. Coffee disappearance was already flat-to-recovering on a 1990 base when CPI was compounding. Soda disappearance was rolling over while soda CPI kept grinding higher.

Decade math: when volume and price disagreed

The dashboard’s decade bars summarize the mismatch:

• 1990s: Coffee volume ends the decade still below the 1990 base after the mid-decade trough; CSD gallons (through the 1997 USDA peak) are up mid-teens percent versus 1990; beverage CPI is up about 31% by 1999.
• 2000s: Coffee green-bean per capita edges lower into 2009; the CSD extension falls from the ~53 peak region toward the mid-40s; beverage CPI rises another ~26%.
• 2010s: Coffee recovers into the 2015 ERS close; CSD extension keeps easing; beverage CPI adds another ~13% by 2019.
• 2020s (partial): No ERS coffee print; CSD extension is roughly flat-to-down from 2021 to 2025; beverage CPI rises about 15% from 2020 to 2025.

That is the opposite of “volumes kept rising while prices jumped.” Prices jumped on a volume path that was recovering selectively (coffee) or already in long decline (soda).

Price–volume scatter and the widening gap

On the scatter panel, each coffee year is a point with beverage CPI index on the X axis and coffee volume index on the Y axis (both 1990 = 100). Early-1990s points cluster near the diagonal. As X marches past 150 and 180, Y stays near 90–100. The 45-degree reference line is the “volume kept up with prices” fiction; coffee points fall below it for most of the post-2000 window.

The gap panel subtracts beverage CPI index from volume index. Coffee’s gap deepens into large negative territory by 2015. CSD’s gap turns negative once the boom ends and CPI keeps compounding — and the sparse extension years sit deeper still.

Selected years at a glance

Coffee gallons (ERS retail conversion) move with retail pounds: about 22 gal per capita in 1990 and 22 gal again in 2015 — another reminder that the green-bean recovery from 1995 did not recreate a 1970s fluid boom.

Caveats and confidence

Disappearance ≠ intake. Plate waste, preparation waste, and foodservice vs at-home splits are not fully observed. Coffee’s green-bean series is sensitive to roast and instant conversion factors.

Soft-drink continuity is broken. USDA’s published gallons through 1997 are the high-confidence spine. Post-1998 markers are industry composites used only to show direction after ERS removed BMC files. Do not treat 2010 or 2025 CSD cells as ERS official disappearance.

CPI basket vs volume mix. Nonalcoholic beverages include juices, bottled water, and other non-caffeine items. Coffee and carbonated item indexes are the tighter caffeine price lenses; the headline beverage CPI is the broad shelf-price context the brief requested.

Calendar quirks. 2025 CPI annuals omit October during the appropriations lapse. 2026 figures in the data module are January–July averages, not full-year prints.

No causal claim. This post does not estimate demand elasticities or attribute volume decline to sugar taxes, bottled-water substitution, or energy-drink cannibalization. It answers a narrower question: whether the official volume series kept rising through the price climb. They did not — coffee recovered from a trough without reclaiming old highs; soda peaked in the late 1990s; beverage CPI kept rising into 2025–26.

Bottom line for the aisle

If the mental model is “Americans drank ever more caffeine drinks while prices exploded,” the USDA–BLS tape rejects it. Coffee disappearance partially healed after 1995 and then froze in the official record after 2015. Soft-drink gallons topped out near the turn of the millennium on the government series and look lower on the industry extension. Prices, especially the broad nonalcoholic beverage CPI and the coffee item index after 2021, are the series that kept climbing. Gallons and pounds did not keep pace with that CPI path.