Herding in Queues: Why we Can’t resist Long Lines for Bagels

This Week’s Focus: The Psychology and Economics Behind Long Lines

This week, we explore the curious case of Apollo Bagels, a West Village favorite whose immense popularity has put it at risk of eviction. Despite efforts to manage its long lines, the crowds remain, driven by more than just a love of bagels.

Today’s article delves into the psychology behind such phenomena, examining how herding behavior, social proof, and the fear of missing out (FOMO) influence consumer decisions. We discuss how individuals use public signals, like line length, to infer private information about quality—often leading to information cascades where early public decisions outweigh private judgment. Using concepts like representativeness heuristics, bounded rationality, and mimetic theory, we unpack the deeper dynamics of collective behavior and the role of social relationships in shaping desire.

Apollo Bagels, a beloved West Village institution, is under threat of eviction—not for failing to pay rent, but because its long lines have allegedly become a nuisance.

The landlord has cited the congestion caused by the bagel shop’s popularity as the primary reason. Despite efforts by Apollo Bagels to manage these lines by increasing staff and introducing online ordering options, the crowds persist.

This situation begs the question: Why are people willing to endure such long waits for bagels, especially in New York City, a place already overflowing with high-quality bagel shops?

Apollo Bagels is no ordinary bagel shop. Ranked among the top 16 in New York—the birthplace of the modern bagel—it has received rave reviews for its perfectly chewy texture and innovative toppings. Customers frequently share their enthusiasm on social media, with posts often featuring lines stretching around the block. It’s very clear that for many, the wait is part of the experience.

Of course, this phenomenon is not unique to bagels. Think about the last time you joined a line at a popular food joint or chose a course at Wharton (for example) simply because it was popular— assuming it must be amazing (only to realize how wrong you were).

Interestingly, Dan Frommer’s consumer trend review underscores this point: 74% of customers are willing to return after a long wait, which shows that for many, the experience feels worth it. The allure of exclusivity and the social proof of others waiting in line seem to outweigh the inconvenience of standing idle.

This behavior speaks to a broader human tendency: equating scarcity or popularity with quality. But doesn’t this defy rationality? Shouldn’t we just recognize this as a clever social trick and move on to the next best bagel shop, returning later when the line is shorter?

When is the “wisdom of the crowd” just individual stupidity multiplied?

The persistent popularity of such lines suggests that rationality often takes a back seat to the powerful influence of social proof and the fear of missing out. This interplay of rational and emotional decision-making is at the heart of why we see long lines—and why we often join them.

Let's go deeper.

Herding Behavior: A Rational Perspective

The first explanation for this behavior lies in the concept of information cascades, first formalized by Bikhchandani, Hirshleifer, and Welch (1992).

Many people use the term herding, as if it’s an irrational behavior, but the research shows that herding is actually fully rational. When faced with uncertainty, individuals often use public signals (the line’s length), to infer private information (whether they will like the bagel or not).

The model has several important factors:

Private Signals: Each customer has a private belief about the bagels’ quality, represented by a signal (e.g., good or bad).

Public Signal: The length of the queue provides observable information. A longer queue suggests that others’ private signals are positive.

Decision Rule: A customer chooses to join the queue if the expected utility of waiting (accounting for both the perceived quality and the waiting cost) exceeds the utility of not waiting.

Let’s walk through an example:

Consider a new bagel shop where customers must decide whether to join the line or not. Each customer receives a private signal (either from their own intuition or from others) regarding the bagel’s quality (“good” or “bad”), and observes the decisions of previous customers (public signal). Let’s look at how herding develops.

Initial Conditions

The true quality of the bagels is either good or bad.

Prior probability: 50% chance of good quality, 50% chance of bad.

Each customer receives an independent private signal that is 70% accurate:

If the bagels are good: 70% chance of “good” signal, 30% chance of “bad.” If the bagels are bad: 70% chance of “bad” signal, 30% chance of “good.”

Customers know their private “signal” is not perfect, so they are willing to observe (“listen to”) the public signal.

Customer 1: Private Signal Only

The first customer receives a “good” signal. We’ll be using Bayes’ Rule (which I’m sure you ALL remember from second grade…right?).

I considered skipping this part and speaking “from authority” to simply say, “Trust me, the math works.” But I know that nothing, and I mean nothing, will complete your holiday cheer more than some casual statistics and Bayes’ rule, especially in a year where the first night of Hanukkah coincides with Xmas —a once in 15 years occurrence.

Back to the math…

P(Bagel is Good|Signal = “good”) = [P(Signal = “Good”|Bagel is Good) × P(Bagel is Good)] / P(Signal= “good”) = (0.7 × 0.5) / [(0.7 × 0.5) + (0.3 × 0.5)] = 0.7

Customer 1 joins the line having received a “good” signal with 70% probability.

Customer 2: Combining Private and Public Information

Let’s say Customer 2 receives a “bad” signal but observes Customer 1’s decision to join. Customer 2 must combine the following:

Private signal (“bad”): suggests 70% probability of bad quality.

Public information (Customer 1 joined): suggests 70% probability of good quality.

Using Bayes’ Rule with both pieces of information:

P(Bagel is Good|Join1, Bad_Signal) = [P(Join1|Bagel is Good) × P(Bad_Signal|Bagel is Good) × P(Good)] / P(Join1, Bad_Signal) = (0.7 × 0.3 × 0.5) / [(0.7 × 0.3 × 0.5) + (0.3 × 0.7 × 0.5)] ≈ 0.5

Customer 2 is now indifferent (50-50) and, following convention, joins the line.

Customer 3: The Beginning of a Cascade

Let’s assume Customer 3 also receives a “bad” signal but observes the two previous joins. The public information (two joins) now strongly outweighs their private signal:

P(Good|Join1, Join2, Bad_Signal) > 0.7

Customer 3 rationally ignores their private signal and joins.

At this point, an information cascade begins.

All subsequent customers will join regardless of their private signals, and the public information (three joins) dominates any private signal. In fact, private information is no longer incorporated into the public pool of knowledge and while each individual acts rationally, the collective outcome may be inefficient.

This example demonstrates how herding can emerge from rational decision-making when individuals have access to both private and public information. It illustrates the power of public decisions in shaping behavior. Even if individual signals are equally distributed between good and bad, early public decisions can lead to an information cascade where subsequent customers disregard their private signals.

Scary huh?

Accounting for Waiting Costs

The model above, however, overlooks two important aspects of queues:

(i) people may dislike waiting (and most do),

(ii) in a queueing setting people don’t really observe all the previous decisions (customers leave after being served, and occasionally, the queue clears almost completely).

Debo and Veeraraghavan (2011) expand our understanding of queue behavior by introducing the balancing act between positive externalities (the informational value of queue length) and negative externalities (the time cost of waiting). Their analysis sheds light on how waiting costs influence herding behavior and reshape the dynamics of queue formation.

Their main findings are that herding becomes pronounced when the difference in queue lengths between competing options crosses a certain threshold. In these cases, consumers interpret longer queues as signals of higher quality.

However, as waiting costs increase, consumers’ willingness to join longer queues diminishes. This creates a natural limit to herding, as individuals weigh the incremental perceived quality against the mounting cost of waiting time.

Positive externalities amplify herding when the queue’s informational value outweighs the inconvenience of waiting. For example, observing a long queue at Apollo Bagels reinforces the belief that the bagels are exceptional, tempting more customers to join.

Negative externalities grow in importance as queues become excessively long, deterring marginal participants and restoring balance.

The interplay between waiting costs and quality perception explains why herding is more likely in environments with manageable congestion. For example, Apollo Bagels may continue to attract long lines because the perceived quality boost from the crowd outweighs the inconvenience of waiting—up to a point. However, as congestion becomes excessive, the deterrent effect of waiting costs curtails further growth in the line.

Behavioral Biases: Experimental Evidence

But until now, research has been primarily theoretical, trying to explain anecdotal evidence of herding.

Debo and Kremer (2011) conducted laboratory experiments to rigorously examine deviations from rational behavior in queue-joining decisions. Their experimental setup provided a controlled environment to isolate and analyze human biases when facing uncertain service quality.

But why did the researchers choose lab experiments?

Real-world queue-joining behavior is influenced by a multitude of factors, such as advertising, pricing, and location. Controlled experiments allow researchers to strip away these confounding elements and focus on how individuals perceive and respond to queue lengths as public and private signals regarding quality.

The experiment consisted of multiple rounds where participants made decisions about whether to join one of two queues for an uncertain service quality:

Setup: Participants were presented with two service options, each with an observable queue length. The true quality of the services was unknown but drawn from a pre-specified distribution.

Private Signals: Each participant received an independent private signal about the quality of the services, with a known accuracy (e.g., 70%).

Public Information: Participants could observe the queue length of both services before making a decision.

Costs: Waiting in a queue incurred a cost proportional to the queue length. Participants aimed to maximize their payoff, which depended on receiving the higher-quality service while minimizing waiting costs.

The main results are quite interesting:

Bias Toward Long Queues: Participants exhibited a strong preference for joining longer queues, even when their private signals suggested otherwise. This behavior highlights an overly strong mental mapping between queue length and perceived quality.

Judgmental Errors: Short queues were overly associated with low-quality services. Participants often neglected the fact that high-quality services could occasionally have short queues due to stochastic variation.

Systematic Deviations: While random errors occurred, the primary driver of deviations from rational behavior was judgmental bias rather than pure randomness. Participants failed to account for the probabilistic nature of queue formation and relied excessively on case-specific evidence.

Negative Impact on Welfare: The experimental results showed that herding based on biased perceptions of queue length often led to suboptimal outcomes. Participants wasted time in longer queues and incurred higher waiting costs without a corresponding increase in service quality.

What drove these results?

Mental Shortcuts: Participants relied on heuristics like “longer queues mean better quality,” a simplification that ignored the variability inherent in queue formation.

Neglect of Base Rates: The base rate of queue lengths for high- and low-quality services was underutilized in decision-making. Instead, participants disproportionately focused on the immediate public signal.

Behavioral Biases: A combination of representativeness heuristics and bounded rationality explained the majority of observed behavior. Participants showed an excessive reliance on the observable queue length, reinforcing herding tendencies.

This experiment highlights the inefficiencies introduced by behavioral biases in decision-making. While long queues might signal quality in some cases, their overemphasis leads to wasted time, lower overall welfare, and misallocation of resources. This finding highlights the importance of understanding and mitigating biases when designing systems that rely on user behavior.

The implications of these findings are quite relevant for stores like Apollo Bagels and the consumer trends highlighted by Dan Frommer. The Frommer survey reveals that 74% of customers are willing to return after waiting in a long line, suggesting that the experience itself—not just the product—plays a significant role in customer satisfaction. This reinforces the notion that long lines act as both a social proof mechanism and a marketing tool.

For Apollo Bagels, the presence of long lines can enhance its reputation, turning it into a destination for those seeking “the best bagels in New York.” However, this benefit comes with trade-offs. Excessive waiting times can lead to customer frustration and diminish the overall experience, especially if the perceived quality of the bagels does not meet the heightened expectations created by the line.

The willingness of customers to endure long waits reflects the power of perceived value shaped by social proof and exclusivity.

In other words:

A rational customer herds.

A status-seeking, social-proof hungry, mental-shortcut customer herds even more, wasting not only their own time, but also everyone else’s.

Mimetic Theory and the Role of Preferences

But since we’re already on the subject, it’s worth asking: Do you really desire bagels? Why do you desire bagels? Do you have any agency in desiring bagels?

René Girard’s (whom Peter Thiel considers to be his main intellectual mentor) mimetic theory offers a framework for understanding the underlying mechanisms of social desire and collective behavior. At its core, the theory posits that human desires are fundamentally mediated through social relationships rather than arising autonomously.

This mediation occurs through what Girard terms “mimetic desire” —a complex process whereby individuals unconsciously adopt and internalize the desires of others whom they observe or consider as models.

René Girard’s mimetic theory builds on the idea that humans learn through imitation—a process we all recognize in actions, similar to how babies learn to walk or speak. Girard’s key insight is that this imitation goes deeper: we don’t just mimic actions; we mimic desires. We want things because others want them, creating a social web of imitation that shapes our values and conflicts.

Girard uses religious rituals and taboos to illustrate his theory, explaining that in early societies, shared desires often led to rivalry and violence—what he calls a “mimetic crisis.” To manage this, communities developed rituals, myths, and taboos to channel or suppress dangerous rivalries. For example, the scapegoat mechanism—a cornerstone of his theory—explains how communities defuse tensions by uniting against a single individual or group, often through sacrifice.

In short, Girard argues that human culture and conflict stem from the deep, often unconscious imitation of others’ desires, with religion and taboo arising as mechanisms to manage the resulting tensions.

In queueing contexts, mimetic theory provides several testable predictions about behavior. The visible presence of others waiting doesn’t merely signal product quality or scarcity - it actively shapes the underlying desire for the product itself. This helps explain several empirical observations.

The mimetic framework has significant implications for understanding preference formation in complex social environments. In educational and professional contexts, for instance, the theory suggests that perceived prestige operates not merely as a signal of quality but as a generative mechanism for desire itself. This helps explain the persistence of certain career paths’ popularity even when market conditions no longer support their traditional appeal, the clustering of educational choices around specific courses ... or specializations and the difficulty individuals face in distinguishing between authentic personal preferences and socially mediated desires.

One must admit that while mimetic theory offers powerful explanatory mechanisms, several important caveats warrant consideration: Some aspects of the theory prove challenging to test empirically, particularly the distinction between “authentic” and mimetic desires, and as I have shown above, conventional economic and psychological theories of information cascades and social proof may explain some observed phenomena without requiring mimetic desire as a mechanism. But overall, it’s an interesting theory worth exploring.

Conclusion: The Cost of Long Lines

The allure of long queues, whether for bagels or broader life choices, illustrates the deep influence of social proof, behavioral biases, and mimetic desire on human decision-making. From Apollo Bagels to classroom enrollments or career paths, people often follow the crowd, assuming the collective knows best. Yet, as the aforementioned research and theories reveal, herd behavior can result in inefficiencies and suboptimal outcomes.

For businesses like Apollo Bagels, leveraging these dynamics strategically—by managing perceptions of scarcity and enhancing the waiting experience—can maintain desirability while avoiding customer dissatisfaction.

For individuals, the lesson is clear: step back, question the source of your preferences, and decide whether the queue you’re standing in (literally or figuratively) truly aligns with your values and goals.

In the end, the spectacle of a long line might be as much about performance as it is about product. Understanding this can help businesses optimize their strategies and individuals make more authentic, deliberate choices—because sometimes, the best bagel is the one that doesn’t require the wait. And Wharton’s best course is …. Let’s just enjoy the holidays, and get back to this after classes resume…