Start with the number that just happened, because it's the clearest possible proof the mechanism is live, not historical. On August 10, 2026, Lake Mead -- the largest reservoir in the United States, formed by the Hoover Dam -- hit a new all-time-low elevation of 1,040.09 feet, 26.7% of capacity, breaking the previous record set in July 2022.^[1] Hoover Dam's hydropower output is already running about 40% below its maximum. Below elevation 1,035 feet -- a threshold the Bureau of Reclamation's own July 2026 study says could be crossed within weeks -- the dam's generating capacity drops by 70%.^[2] The seven Colorado River basin states couldn't agree on a joint plan for what happens after their current operating guidelines expire October 1, 2026; the Bureau of Reclamation finalized its own federal framework instead, on a deadline, because the states themselves ran out of time to agree.^[3]

None of that is bad luck. It's the direct, traceable consequence of a decision made in 1922, before any of the people managing the crisis today were born. The Colorado River Compact allocated a fixed volume of water -- 7.5 million acre-feet a year to each of the Upper and Lower Basins, 15 million acre-feet total, plus another 1.5 million promised to Mexico by treaty in 1944 -- based on flow measurements taken during an unusually wet stretch of years. The river's actual long-term average is closer to 14.6 million acre-feet; since the current megadrought began around 2000, it's run closer to 12.5 million.^[4] The compact promised more water than the river reliably carries. That gap has a name in the field -- the "structural deficit" -- and it's been baked into the compact's own founding math for over a century, not something drought created on its own. Drought just made the gap impossible to paper over any longer.

Who actually absorbs that gap is its own real, ongoing fight, and it runs through California's own farmers specifically. Within California, a 1931 agreement among the state's major water users set an internal priority order -- and the Imperial Irrigation District, which irrigates the Imperial Valley's farmland, holds 3.1 million acre-feet a year of that allocation, 2.6 million of it with a water right dated to 1901, among the single most senior rights anywhere on the entire Colorado River system.^[5] That's a bigger, more senior claim than most of Arizona's entire state allocation. It's the concrete reason Arizona's farms and cities have absorbed earlier and deeper mandatory cuts than California's have, and it's why California's Imperial Valley agriculture keeps showing up in the same fight, compact after compact, negotiation after negotiation -- not a new dispute, the same one 1931 already decided the outcome of.

Now put a different compact next to it, governing a different river, written by a different formula. The Republican River Compact -- negotiated in 1943 among Colorado, Kansas, and Nebraska after a real, lived 1930s drought and a 1935 flood, with its own administering body (the RRCA) created in 1959 -- didn't promise a fixed volume at all. It allocates a percentage of whatever the basin's actual average annual water supply turns out to be: roughly 49% to Nebraska, 40% to Kansas, 11% to Colorado.^[6] When the river runs low, every state's share runs low with it, automatically, by the same formula. There's no structural deficit built into the math, because the math was never a fixed promise in the first place.

That different founding choice produces a genuinely different kind of dispute, not a smaller version of the same one. Kansas v. Nebraska reached the Supreme Court in 2014-2015 -- Nebraska had knowingly overused its share by roughly 70,869 acre-feet in 2005-2006, and the Court awarded Kansas $3.7 million in damages plus $1.8 million in disgorgement.^[7] That was the second time in a decade the same two states had gone to the Supreme Court over this compact. But both fights were technical accounting disputes over a formula everyone still agrees is fair -- not an existential fight over whether the formula itself ever made sense. Nobody is measuring Republican River reservoirs against an all-time-low headline the way Lake Mead gets measured every summer now.

Two more compacts show the same tool can be built with entirely different enforcement machinery, independent of the allocation formula itself. The Delaware River Basin Compact (1961, New York/New Jersey/Pennsylvania/Delaware) made the federal government itself a full, equal voting member alongside the four governors -- one of only two interstate water compacts built that way.^[8] The Pecos River Compact (Texas/New Mexico) went further still: after its own repeated Supreme Court fights, including a 7-1 ruling as recently as 2020, it now runs with a permanent, court-appointed "River Master" who recalculates every state's compliance every single year -- a standing enforcement mechanism neither the Colorado nor the Republican River compact has ever needed to build.^[9]

Same legal tool, four rivers, four different results -- and the difference isn't luck or regional character. It's two design choices, made at each compact's founding and never revisited: whether the allocation formula promised a fixed amount or a share of whatever actually showed up, and what happens when someone takes more than their share. The Colorado River Compact chose a fixed number, in a wet decade, with no binding cross-basin enforcement body -- and a hundred years later, that choice is what's actually drying up Lake Mead, not the drought alone. The Republican River Compact chose a percentage and a real administering body from the start, and its worst outcome, a century in, is still just a bill for $5.5 million and an argument about how to count imported water. Whoever wrote the math in the founding document is still winning or losing the argument today, whether or not anyone negotiating it in 2026 remembers that the math was ever a choice at all.