In the basement of the Kellogg School, there is a cafe. It’s a busy cafe, which says more about the available alternatives than about its absolute quality. Because it gets busy and because a good number of its customers are polite enough to walk out of class five minutes early to beat the crowd, I and my colleagues have learned that it is a much better to plan to go down for a sandwich a little before noon than a little after noon. According to CNBC, Goldman Sachs faces similar issues with queuing in its cafeteria and it actively tries to manage the system (The creepy capital efficiency of Goldman’s cafeteria, Oct 17).
The most crowded time of the day to eat lunch is, naturally, during lunch time. For most people, this falls around noon. This creates the phenomenon of the lunchtime rush hour. You know this all too well if you’ve ever tried to stop in your local chopped salad place at, say, 12:30 in the afternoon.
Goldman didn’t like the idea of its people waiting on long lines to get their lunch. People are capital to Goldman. It wants to use its capital efficiently. Standing on line waiting for dumplings or salad or a burger is not an efficient use of Goldman’s capital. …
The cafeteria has a set of timed discounts. If you show up in the cafeteria before 11:30 or after 1:30, you get a 25 percent discount on your food. Goldman incentivizes employees to avoid the rush hour.
So there are a couple of intriguing things about this. First, if Goldman employs really valued their time, a discount shouldn’t really be necessary. Kellogg faculty don’t a discount for getting lunch before the students get out of class. We just get a shorter wait. I can go beyond an anecdotal example on this. Jan and I have a paper that shows that if all customers agree on what are preferred and less-preferred times for service they will spread themselves out. In equilibrium, the preferred times are more crowded while the shorter wait in the less-preferred time just compensates for giving up on going at the preferred time.
If Goldman finds it necessary to offer a financial carrot to level load the cafeteria, then there must be a little more going on. Arguably, this implies something about how professionals go about their day. While many have days in which every minute is scheduled with meetings, most still have days in which they have some discretion over their time. When a task is finished, the question is whether to immediately move on to the next task or to return a phone call or go get lunch. The discount plan then serves to nudge people to level load the cafeteria.
This could help explain why saving a few bucks at lunch is enough to sway the behavior of workers who are generally well-paid. A priori, I would not have guessed that a small discount would be enough to get Masters of the Universe to vary their lunch time routines, but by making early and late focal, I can see how this will work. Thinking about these incentives in this way also highlights that the workers you would have guessed would be most responsive to these incentives may have the hardest time taking advantage of it. Upfront, it would be natural to expect that support staff would be most likely to adjust their schedule to save a few bucks but they are also the most likely to be on a schedule that tells them when they can go to lunch.
A final point. Goldman’s scheme introduces its own inefficiencies.
If you find yourself in the cafeteria sometime around 1:20 pm, you’ll notice that the lines at the pay registers are empty. So are many of the tables.
But the cafeteria area between where the food is collected and where you pay is quite crowded. The Goldman lunchers are chatting with each other, waiting for the final minutes to tick down until they can save a dollar or two.
I am not sure that there is a good fix to this. On any given day, if there are half a dozen people milling around at 1:20, it would be efficient to cut the price immediately and get them on their way. But if that happens every day, there is an incentive to come at the tail end of the peak time but still in the peak and the whole thing will unravel.