Don't follow the herd

How can herd behaviour influence our choices?

In the wildlife animals move together as a group and herding behaviour is not uncommon. Humans are not that different and can build up group momentum following just one or two leaders. As it turns out this herd mentality and behaviour of humans can be observed in all walks of life. Every hour of the day we are bombarded by constant need to make decisions and most of the time these decisions are affected by what those around us are doing. The decisions that we take range from simple ones- “which restaurant should I go to?”, to the more complex of “which assets do I invest my savings in?”. When making those decisions people are often prone to irrationality with herd behaviour being a prime example of this.


Researchers have shown that we are heavily impacted by the information available in the immediate surroundings. Herding behaviour is so common to humans that it affects all aspects of life.

  • Fertility choices (e.g. how many children does one have) suggests that we are heavily influenced by what others in the same area are doing
  • Voters are impacted by the opinion polls released before elections, resulting in a vote that mirrors the direction of the polls.
  • Rush for cryptocurrencies fuelled by crowd’s investments without understanding the underlying value of the asset or not using available private information.
  • The rise of artificial intelligence, big data and machine learning indicates that a disproportionate proportion of the population is engaged in working on topics that are currently “hot”

As social animals, humans are known to go with the flow.


In an attempt to understand the rationale behind the decision making influenced by signalling from others, multiple economic models were suggested. This year’s economics noble prize laureate - Abhijit Banerjee proposed a simple model assuming that the decisions taken by others are based on rationality (i.e. decisions reflect information that isn’t available to us). The outcome of everyone going with the flow rather than using their own private information would indicate herd behaviour. Herd behaviour is exacerbated by a continuous loop where each person’s information is less and less informative to others as individuals become less responsive to their own private information. Banerjee finds that such reduction of informativeness is so severe that the society would be better off if it constrained individuals to only use their own private information.


Imagine a situation where you find yourself in front of two restaurants A and B that seem equally good. That night there are three other individuals that each have a private information for which restaurant is better. In a situation where three of these individuals prefer restaurant A and only 1 prefers restaurant B, it should be straightforward that restaurant A will end up with most customers. Now imagine a situation where that one person with a private signal for restaurant A (customer 1 in the illustration) arrives first and takes a sit in restaurant A motivated by own private information. The second person to arrive will then have two available signals, his own private information that B is better and the signal from the first person that A is better. Banerjee also assumed that restaurant A is marginally favoured a priori, leading to the second person choosing A over B despite own information signal. This then means that any subsequent customer that arrives will choose restaurant A, ignoring own signal, resulting in a herd mentality, as illustrated in the picture below. Everyone then ends up at restaurant A despite B being certainly better.

If in the presented situation the second person however decided to follow her own signal, the third would have known that signal of the second is B and would certainly choose B along with all the subsequent individuals. In such a situation, the second person’s choice to ignore her own information and join the herd has imposed a negative externality* on the rest of the individuals.

The model thus implies that it is not only the information signals and their validity but also the order of those signals that matters. The signals of the first few decision makers will have a direct impact on where the initial crowd will form, thereafter joined by everybody else. Herd mentality in part reflects the human laziness to assess the available information and ignore educated private signals in favour of what the majority is doing. Such laziness reflects the need for a heuristic - a decision-making shortcut that takes away the mental strain/toll of exerting costly effort.


Heuristics are mechanisms or approaches to problem solving that may not be rational or optimal but instead sufficient to fulfil an immediate goal. While simplifying the decision making process, heuristics have also been shown to be prone to cognitive biases. One of such heuristics would be to look around and to understand what others around us are doing and apply the same reasoning to solve immediate problems at hand.

Humans do not always act rationally and in many cases people act fairly irrationally. Many heuristics that are used daily to assist in decision-making cause cognitive biases resulting in systematic errors. Daniel Kahneman, a Nobel prize-winning economist, has proposed two mechanisms that drive the way humans think: System 1 and System 2. System 1 is fast and reactive, while System 2 is slow and deliberate where reason dominates. Because people are inheritably lazy, most of the times it is System 1 that dominates.

In case of herd behaviour it is also System 1 that would come into play. “If everyone is doing it then it must be a good idea” logic might work for non-important tasks but when more is at stake System 2 with slow deliberate reasoning needs to take over.


The abundance of easily available online reviews at present has a large impact on how we make decisions. Online information on the one hand allows to simplify the decision process by making the choices and reviews of others widely available. On the other hand, the herding behaviour trap needs to be taken into account with System 1 on the look-out for shortcuts.

The information signals of other previous customers are now largely available online and can allow individuals to form their private information signals based on such reviews. The online reviews can also be treated as information signals from other consumers. In such a case individual signals can be perfectly observed online.

To make an extension to Banerjeet’s model and incorporate it into the realities of the online world, we can assume that private signals are now more informed and individuals will have better private signals than before. This may mean that private signals may be given a greater value than the signals received from other customers in the restaurant example. If you are now the second customer in the example introduced above with a clear private signal for restaurant B, it is likely that you would ignore the signal from the first customer for restaurant A and go for restaurant B instead.


Online reviews have high transaction costs as exerting effort without monetary benefit is required. This means that a lot of the reviews written may either be a) fake or b) stimulated by extremely bad or good quality/service (outliers). Having polar opposite reviews then means that it may be difficult for a user to estimate the true quality of the product. In general people are also more affected by the negative reviews as humans dislike losses more than they like gains. Consumers thus tend to prioritise negative reviews in order to avoid a risky choice.

There may also be hidden costs to having others choose what you are choosing. In such cases, some will thus have an incentive to hide their true signals to avoid the crowding out effect that may arise as a result of increased demand. Take restaurants as an example, if you like the walk-in service you’re getting from a restaurant nearby your home, you may not be motivated to leave a good review online to attract more customers. This means that in some cases benefits accrued to customers might vary based on the order of choice. The opposite effect of this can be observed in the investment industry where higher the herding the greater the benefit to the early investors.

Uncertainty of the quality of the reviews would then result in a break-down of the extension to the Banerjeet’s model. Current abundance of fake online reviews would imply that general population’s private signals are not necessarily informative and signals from other consumers should not be given high weighting. Herding behaviour can be avoided if there is a wider awareness of the quality of reviews in the community. The way to avoid herding behaviour is thus to prioritise the logic of System 2 reasoning and ignore the signals of other individuals.

Ayrat M
Strategy Consultant

Expert for commercial applications of game theory, industrial economics, mechanism & market design.