The Long Lunch — Letters from the Floor

The standard story about the business lunch is a story of decline, and the standard mood of the story is regret. There was once, the story goes, a golden age of the long midday meal, when executives left their offices at noon and did not return until three, when deals were struck over the second course and sealed over the third, when the expense account was bottomless and the martinis were dry and the whole apparatus was, in some hard-to-specify way, more civilized than what we have now. The decline of this institution is mourned in roughly the same register as the decline of the handwritten letter: as evidence of a coarsening, a loss of grace, a surrender of the human to the merely efficient.

I want to resist this story, or at least complicate it, because I think it gets two things wrong. It is too sentimental about what the long lunch was, and it misunderstands what the long lunch actually did. The institution was never primarily about grace, and its decline is not primarily about coarsening. The long lunch was a piece of relationship infrastructure, an unusually effective one, and understanding why it worked tells you more about the present than any amount of nostalgia.

Begin with the tax code, because the long lunch was, to a degree that is now slightly embarrassing, a creature of tax policy. Business meals have been deductible, in one form or another, since the earliest days of the American income tax, originally not by any specific act of legislative design but simply as an ordinary and necessary cost of doing business. For most of the twentieth century the deduction was generous and loosely policed. The phrase “three-martini lunch” entered wide circulation in the postwar decades to describe exactly the kind of leisurely, alcohol-soaked midday meal that the deduction made functionally free. President Kennedy called for a crackdown on the practice as early as 1961, and got nowhere. The deduction survived, at one hundred percent, into the 1980s.

Then it began to erode. The Tax Reform Act of 1986 cut the deductible share to eighty percent and tightened how the deduction was policed. (The requirement that real business be discussed at the meal was older still, dating to the Revenue Act of 1962, a Kennedy-era reform that hedged the deduction with limits and paperwork without managing to shrink it.) The Revenue Reconciliation Act of 1993 cut it again, to fifty percent, where it has largely remained. The institution did not survive these cuts intact. It is difficult to prove a direct causal line from a tax rate to a cultural practice, but the timing is suggestive: as the subsidy shrank, so did the lunch. The thing that had been functionally free became merely half-priced, and then the calculus that had always quietly underwritten the long midday meal stopped working.

The tax history matters because it punctures the sentimental version of the story. The golden-age lunch was not sustained by a more relational business culture. It was sustained, in significant part, by a tax incentive that made it cheaper to talk to a client over a long meal than almost any other way. When the incentive went, the practice went with it, which suggests the practice was never quite as freely chosen, or as purely civilized, as the nostalgic account implies.

But here is where the nostalgic account, for all its sentimentality, is onto something real. The long lunch, whatever its grubby fiscal origins, did something that has turned out to be genuinely hard to replace. It built trust, and it built it through a mechanism that the research literature has since worked out in some detail.

The mechanism is shared food. In a series of studies at the University of Chicago, the behavioral scientists Kaitlin Woolley and Ayelet Fishbach found that people who eat the same food as their counterparts become measurably more trusting of them. In one experiment, participants in a simulated investment game who had eaten the same snack as their partner invested more money with them, and reached agreements faster, than those who had eaten different foods. The effect did not require conversation about the food, or even acknowledgment of it. The mere fact of having eaten the same thing produced a small, reliable increase in trust.

A related body of work points the same direction. Researchers at Stanford’s Graduate School of Business, the negotiation scholar Margaret Neale and her colleague Peter Belmi, found that sharing food from a common source, rather than from separate plates, changed the outcomes of negotiations. Pairs eating from a communal bowl in an adversarial negotiation created more value than those eating separately. A 2020 study by Jiyin Cao, Dejun Tony Kong, and Adam Galinsky found that shared eating produced higher levels of cooperation and lower levels of competition than eating apart, an effect that held up even when participants were allowed to negotiate in the natural back-and-forth way that real deals require. Across these studies the finding is consistent and slightly uncanny: the act of eating together, especially of eating the same thing, does measurable work on the relationship between the eaters that conversation alone does not do.

This is the thing the long lunch was for, whether or not anyone sitting at the table could have articulated it. The deal could have been done by memo. The terms could have been exchanged by phone. What the phone and the memo could not do was the slow, low-bandwidth work of two people sharing a table for two hours, ordering from the same menu, watching each other eat, talking about things that had nothing to do with the deal. The lunch was not the inefficient packaging around the real business. The lunch was the real business. The deal was the part that could have been handled any number of ways. The trust was the part that could only be built like this.

Seen this way, the long lunch was not an indulgence that happened to involve some relationship-building. It was a relationship-building technology that happened to involve some indulgence. The two-hour duration, which looks from the spreadsheet like waste, was the active ingredient. Trust does not form on a deadline. It forms in the unhurried stretches, the second cup of coffee, the digression about a daughter’s college search, the moment when one party reveals something slightly more personal than the transaction required and the other party receives it well. None of this can be rushed, and all of it was what the two hours were for.

I said at the outset that I would not write the nostalgic version of this piece, and I want to keep that promise, because the nostalgic version leaves out something important. The long lunch was also exclusionary in ways that should not be missed. It was, for most of its history, an institution of men, conducted in clubs and restaurants where women were unwelcome or barred outright, and where the real business of a company was transacted in a room that half the company could not enter. It ran on alcohol in a way that was bad for the people who drank too much of it and worse for the people who depended on them. It rewarded a particular kind of social performance, an ease with the menu and the wine list and the unhurried male sociability of the midcentury dining room, that functioned as a quiet gatekeeper. The skills it rewarded were not evenly distributed, and they were not distributed by merit. To mourn the long lunch without remembering this is to mourn an institution that worked beautifully for the people it was built for and not at all for everyone else.

So its decline is not a simple loss. Some of what we lost was worth losing. The fitness-for-duty programs, the diminished tolerance for midday drinking, the cultural shift that made the three-martini lunch look less like sophistication and more like a problem, these were not failures of civilization. They were, in part, corrections.

But the correction overshot, and what it overshot is the part worth attending to. We did not replace the long lunch with a better trust-building technology. We replaced it with no trust-building technology at all. We replaced it with the desk.

The numbers here are stark. In 2017, the Wall Street Journal reported that Americans had made some 433 million fewer lunchtime restaurant visits than the year before, the lowest level of midday restaurant traffic in at least four decades. The pandemic accelerated what was already well underway. By 2025, a Yahoo/YouGov poll found that the most common way for American workers to spend their lunch break was eating at their desk, reported by half of respondents, while only twelve percent said they went out to lunch with others during the week. The midday meal, for most workers, has become a private refueling stop conducted in front of a screen, eaten alone, often while continuing to work.

It is worth being precise about what this represents. Megan Elias, a food historian at Boston University who has written a history of lunch, points out that the social business lunch was never the norm for most workers in the first place. It was always a relatively narrow practice, concentrated among executives and the people who courted them. The mass experience of lunch was always more constrained, more rushed, more tied to the factory whistle and the timed break. So the desk lunch is not, for most people, a fall from some universal golden age. It is the latest form of a midday meal that was, for most workers, never leisurely to begin with.

But for the slice of business that ran on the long lunch, the people whose actual job was to build and maintain commercial relationships, something real has gone missing, and nothing has stepped into its place. The coffee meeting is shorter and does not involve shared food. The Zoom call removes the table entirely; you cannot break bread through a screen, and the research suggests this is not a trivial loss but a structural one. The conference, which I will write about elsewhere, has absorbed some of the function, compressing a year’s worth of relationship-building into three exhausting days on a trade show floor. None of these is a real substitute. They are all faster than the long lunch, and all of them skip the part that made the long lunch work.

This is the pattern worth noticing, because it recurs across the whole history of business relationships. A practice that looks inefficient turns out to be doing essential work, precisely through its inefficiency. The slack, the duration, the apparent waste: these are not bugs in the system that a more rational age can engineer away. They are the system. The long lunch was a machine for manufacturing trust, and it manufactured trust by being slow. When we optimized away the slowness, we did not get a faster trust machine. We got no trust machine, and a generation of business relationships that are thinner than they used to be, conducted between people who have never shared a meal and may never share anything but a calendar invitation and a screen.

What replaced the long lunch, in the end, was efficiency, and efficiency turned out to be the one thing the long lunch was never supposed to provide. The meal was where you went to be inefficient together, on purpose, because the inefficiency was the point. We have gotten very good at eliminating that kind of time from working life. We have not yet figured out what to do about everything it was quietly accomplishing while we thought it was just lunch.

Further Reading