Imagine walking down a street looking for dinner. Two restaurants sit next to each other. Their menus are similar, their prices almost identical, and you have never eaten at either. One is half empty. Outside the other, twelve people are waiting for a table. Without tasting a single dish, you have already learnt something. Or at least, you think you have. Twelve strangers have apparently decided that one restaurant is worth waiting for. Suddenly, joining the queue does not feel irrational. It feels like using information.

Economists have a name for part of what is happening here: herd behaviour. In a classic 1992 model, economist Abhijit Banerjee showed that perfectly rational individuals can end up following the decisions of people before them because those decisions may contain information they themselves do not have. If I know very little about two restaurants but see ten people choosing one of them, their behaviour becomes a signal about its quality. I do not need to know what they know. I only need to believe that they might know something.

That creates a strange possibility. The restaurant may have a queue because it is good. But it may also appear good because it has a queue. Research using restaurant settings has found that longer waiting lines can increase consumers' perceptions of product quality, particularly when consumers have relatively little prior knowledge. More recent research similarly finds that simply seeing a queue can signal brand popularity and raise expectations of quality. The queue is therefore doing two jobs at once. It is the consequence of demand, but it can also create more demand.

This matters because restaurant choice contains an information problem. Before eating somewhere new, the restaurant knows considerably more about the quality of its food and service than I do. I cannot taste dinner before deciding where to have dinner. Searching for more information takes time, so observing what other consumers have chosen provides a shortcut. Herding, in this context, does not have to mean consumers mindlessly copying one another. It can be a rational response to uncertainty. The problem begins when the decisions I am observing were themselves based on somebody else's decision rather than independent information.

This is where herding can turn into an information cascade. Suppose the first diner happens to choose Restaurant A because a friend recommended it. The second person sees someone inside and also chooses A. A third arrives, sees two people choosing A and assumes they probably know something. By the time I arrive, I see a busy restaurant and interpret the crowd as ten separate votes for quality. But perhaps only the first customer actually knew anything. Classic models of information cascades show how individuals observing previous choices can rationally begin ignoring their own private information, allowing surprisingly little original information to generate widespread behaviour. Ten decisions can look like ten pieces of evidence when they are partly repetitions of the same one.

The internet was supposed to make this problem easier. Instead of staring through a restaurant window, I can now see hundreds or thousands of previous customers' experiences. Economically, online reviews reduce information asymmetry by giving consumers information about quality before they purchase. And they have measurable effects on demand. Harvard Business School economist Michael Luca found that a one-star increase in a restaurant's Yelp rating was associated with a 5 to 9 percent increase in revenue, with the effect concentrated among independent restaurants. The result makes intuitive sense. A familiar restaurant chain already carries a reputation. An unknown independent restaurant needs consumers to learn whether it is worth trying.

Reviews therefore perform much the same economic function as the queue, but at enormous scale. A line outside a restaurant tells me that twenty people chose it tonight. A 4.8-star rating tells me that hundreds of people supposedly approved of it before I arrived. Research published in Management Science found that greater review activity helps consumers learn restaurant quality faster, raising revenue for higher-quality independent restaurants and lowering it for poorer-quality ones. In that sense, social information can make markets more efficient. Good restaurants that consumers might otherwise never risk trying can build reputations without belonging to a famous chain or spending heavily on advertising.

But making everyone else's choices visible creates another problem. Reputation can become self-reinforcing. A restaurant that receives strong early reviews appears higher quality, attracts more customers, generates more reviews and acquires an even stronger informational advantage. A restaurant with little initial attention faces the opposite problem. Even if it is equally good, consumers have less evidence on which to trust it. This produces something resembling increasing returns to reputation: being popular today makes it easier to become popular tomorrow. Research on online restaurant platforms finds that consumers respond not only to ratings but also to how much information sits behind them, including the number of reviews. Five stars from 2,000 people simply feels like a safer bet than five stars from two.

This means ratings do more than describe demand. They can redirect it. A small difference in reputation can affect where the next consumer eats, whose choice then becomes another data point influencing the consumer after them. The process is not necessarily inefficient. If the better restaurant consistently receives better information signals, social learning helps customers find it. But the original insight from information-cascade theory is more uncomfortable: herds can form even when the crowd is wrong. Once enough people appear to favour one option, following them can become individually rational even when the popularity itself began with surprisingly weak evidence.

There is also a limit to how valuable popularity can become. Waiting has an opportunity cost. Every additional minute spent outside a restaurant is time that could have been spent somewhere else. Research using data from more than 94,000 customers at an Indian restaurant found that longer waits were associated with customers abandoning the queue, returning less quickly and spending less time dining once seated. So the queue produces two competing economic signals. Its existence can say, "This restaurant must be good." Its length can eventually say, "It isn't worth it."

The consumer is therefore making a more complicated calculation than it first appears. Waiting carries a cost, while popularity supplies information. When uncertainty is high, the informational value of the crowd may outweigh the cost of joining it. When I already know which restaurant I prefer, the queue tells me much less. Research on restaurant waiting supports exactly this distinction: consumers with less prior knowledge appear more influenced by queues when judging product quality. The value of the herd is greatest when my own information is weakest.

And restaurants are only the easiest place to see this mechanism. The same logic can influence which app we download, which book we buy, which stock suddenly attracts attention or which product becomes impossible to find after going viral. In each case, other people's choices contain potentially useful information. But once enough people begin reacting to those choices, popularity can start producing more popularity. The economic question is no longer simply whether consumers have information. It is whether the information they are receiving is genuinely independent.

Perhaps this is the real paradox of the queue. Markets work better when consumers can learn from one another. Reviews, ratings and visible demand can reduce uncertainty and allow unfamiliar businesses to build reputations. Yet the same mechanisms can make markets path dependent, allowing early popularity to shape later demand. The crowd can reveal quality, but it can also manufacture confidence in it.

So when twelve people are waiting outside one restaurant and the place next door is empty, joining the queue may be perfectly rational. The mistake is assuming that twelve people necessarily represent twelve independent reasons. Sometimes a queue is evidence that everyone knows something you do not. Sometimes it is evidence that everyone thinks everyone else does.

Sources

  • Banerjee, "A Simple Model of Herd Behavior" (1992), Quarterly Journal of Economics — academic.oup.com
  • Bikhchandani, Hirshleifer and Welch, survey on information cascades — tamuz.caltech.edu
  • Research on queue length and perceived restaurant quality, Asia Marketing Journal — amj.kma.re.kr
  • Penn State research on the positive effects of waiting and perceived popularity — pure.psu.edu
  • Michael Luca, "Reviews, Reputation, and Revenue: The Case of Yelp.com," Harvard Business School — hbs.edu
  • Research on review activity and restaurant demand, Management Science — pubsonline.informs.org
  • Research on wait times and customer behavior using data from an Indian restaurant, Journal of Operations Management — doi.org