Carter signed the Airline Deregulation Act on October 24, 1978, and American carriers got to set their own prices for the first time in forty years. Before that, fares came down from the Civil Aeronautics Board. After, an airline could charge what it wanted, when it wanted, on whatever conditions it could invent.
They invented a lot. By April 1992, American Airlines was carrying more than 500,000 fares in its reservation system. Robert Crandall, American's CEO, announced a simplification called Value Pricing that would cut the number to roughly 70,000, projecting at least $25 million a year in savings from reduced ticketing complexity alone. The existing structure, he said, had become too complex for customers to understand or to regard as fair.
The simplification lasted months. Competitors matched or undercut the new fares, a broad fare war erupted, and American retreated. The 500,000 came back. The complexity was durable because it wasn't incidental. It was the accumulated residue of competitive strategy, and no airline could unilaterally disarm.
The harder problem had already been named inside American's own operations research group. In a 1992 paper in Interfaces, Barry Smith, John Leimkuhler and Ross Darrow described the challenge that had driven the airline's revenue management work since the early 1960s and that deregulation made urgent: on every flight, decide how many seats to sell at each price level, across thousands of combinations of market and fare class. A fare class is the letter code attached to a given price and its conditions, and there were a great many of them. Controlling those combinations independently was, in the authors' word, infeasible.
Infeasible, not wrong. Each rule could be perfectly defensible on its own terms. The advance-purchase requirement made sense. So did the Saturday-night-stay restriction, which existed to separate business travelers from leisure travelers. So did the capacity controls. What nobody could do was hold the interactions: how many discount seats to protect on Tuesday's Chicago-to-Dallas given how demand was moving on Thursday's Denver-to-Miami, and on a few hundred other flights sharing connecting passengers.
I have inherited configuration systems that behave this way. Every value defensible in isolation, documented, owned by someone who can explain it, and nobody in the building able to tell you what happens when you change one.
American's answer, beginning in 1983, was virtual nesting. Rather than manage thousands of individual market and fare-class controls, they sorted the combinations by expected value and clustered them into eight inventory buckets. Statistical forecasting models then determined how many seats each bucket got. The system that came out of this work, DINAMO, fully implemented in 1988, didn't delete a single fare rule. It sat on top of them and decided which booking classes to make available on each flight based on forecast demand instead of rule-by-rule human adjudication. American put the quantifiable benefit at $1.4 billion over three years.
IATA's own technical literature still describes the resulting split: revenue management controls availability, while fare filing establishes prices and conditions. The rules stayed, the way the signature lists stayed. What moved was the decision, out of the hands of people reading rules and into a forecasting layer that could take the combinations as input rather than as something a person had to understand.

