VoxEU Column Frontiers of economic research

The science of apologies with experimental evidence

Apologies are often hard – that’s the point. An apology is due when trust is broken, and to restore trust the apology must be hard. This column discusses a model of apologies as costly signals with some recent experimental evidence.

Following the recent wave of apologies by politicians,1 celebrities,2 and in particular by firms,3 there have been numerous commentaries about the nature of apology – in particular how it is pointless and overused. Recent research in the social science of apologies can help us understand their logic, and shed light on the purpose of the rituals of repairing social transgression.

A market economy depends on the strength of weak ties, and only rational fools (Sen 1977) ignore the value of relationships in support of economic transactions (Granovetter 1973). But from time to time, mistakes are made that temporarily break these ties. Apologies work to restore such frayed relationships.

As a behavioural economist, my own research has focused on three goals, namely showing that:

  • apologies can have real economic consequences;
  • models based on economic incentives can explain how and when people give and receive apologies; and
  • economic theory can help us understand what it means when we say the words “I am sorry.”

Corporations frequently become the target of public outrage, like General Motors in its handling of recent safety issues and automobile recalls. The public apologies by GM CEO Mary Barra have come under attack and scrutiny. Do such apologetic words have any meaning? Lee et al. (2004) analysed annual reports from publicly traded companies and found that companies that admitted responsibility for bad earnings had higher stock prices one year later than those who did not.

Although compelling, Lee et al. can only establish a correlation between apologies and economic outcomes. In recent work, Elaine Liu and I find a causal link between doctors’ apologies and their patients’ inclinations to litigate (Ho and Liu 2011). At the time of writing, 36 US states had passed laws that encourage doctors to apologise. These laws are premised on two ideas:

  • Doctors are often hesitant to apologise because they are scared of lawsuits.
  • Patients most often sue out of anger, sometimes because their doctor never apologised.

By making a doctor’s apology to the patient inadmissible in court, these states hope to break this vicious cycle and improve the doctor-patient relationship.

By looking at malpractice claims in each state before and after the laws were passed, we find that after a state passes an apology law, malpractice cases settle 19-20% faster (especially in the most severe cases) and we see a 16-18% reduction in the number of claims filed.

This research was inspired by a model of apologies I developed in Ho (2012). The model studies a class of ongoing principal agent relationships where the principal and agent engage in repeat interactions, but the agent’s suitability is unknown to the principal, and contracts are incomplete. Relationships between a doctor and a patient, a home owner and a contractor, a politician and the electorate, and an investor and an entrepreneur all share these feature, and therefore they all rely on a degree of trust. From time to time, mistakes are made and that trust is broken.

The fundamental insight of the model is that for an apology to restore the broken trust, the apology must be hard. The idea is based on a key insight of game theory that finds that our ability to signal a desirable quality depends on the cost of performing that signal. Signaling theory has traditionally been applied in contexts such as explaining a peacock’s extravagant plumage or understanding the value of an expensive college degree. The work here applies the idea of signaling to apologies and human relationships.

When we are wronged, we all want the transgressor to apologise. However, often when they apologise, we punish them for it. We make them feel bad. The reason? If an apology were easy, it would no longer have any meaning.

This simple insight that apologies work to restore relationships but are costly for the apologiser has powerful implications borne out in experiments I conducted with students in a lab. Subjects played a simple investment game that depended on trust. If the investment failed, they were allowed to apologise for that failure. The theory and experiment both show that apologies should be more common in long relationships, more common early in a relationship, and more common when there is a better match between the two parties.

Some say these recent examples of public apologies followed by public jeering and humiliation are just a waste of everybody’s time. But it is precisely the public jeering and humiliation that make the apologies effective.

What does this research say about what makes for a good apology? Essentially, anything that makes the apology costly or difficult. Here are some types of apologies to consider:

  • “I’m sorry about your grandmother’s illness”. Recognition of the pain is a start. Demonstrating that you at least have the empathy to recognise the damage caused and an acknowledgment that the rules that were violated.
  • “I’m sorry – I will never do it again”. Often people will offer forgiveness for the first transgression, if the transgressor accedes to being held to a higher standard in the future. As the saying goes: “Fool me once, shame on you; fool me twice, shame on me.”
  • “I’m sorry – I am an idiot”. Admitting your own incompetence means you give up some of your reputation in exchange for forgiveness. Tiedens (2001) find that voters liked Bill Clinton more after seeing a video of him apologise about the Lewinsky scandal, but then they became less likely to want to vote for him because they think he is less competent.
  • “I’m sorry – here are some flowers”. The more expensive the better. Offering reparation for the harm done is a way to a pay a tangible cost to make up for the mistake.
  • “I’m sorry – it wasn’t my fault”. This is perhaps the least effective as it is the least costly to say. But it could work if you can prove it wasn’t your fault in a way that is costly to fake.

Of course, much more remains to be done to understand this complex social institution, but this analysis offers a start. As for any shortcomings in what I write here, my apologies.


Granovetter, M (1973), “The Strength of Weak Ties”, American Journal of Sociology 78(6): 1380-1380.

Ho, B (2012), “Apologies as Signals: With Evidence from a Trust Game”, Management Science 58(1): 141-158.

Ho, B and E Liu (2011), “Does sorry work? The impact of apology laws on medical malpractice”, Journal of Risk and Uncertainty 43: 141-167.

Lee, F, C Peterson and L Tiedens (2004), “Mea Culpa: Predicting Stock Prices From Organizational Attributions”, Pers Soc Psychol Bull 30(12): 1636-1649.

Sen, A (1977), “Rational Fools: A Critique of the Behavioral Foundations of Economic Theory”, Philosophy & Public Affairs 6.

Tiedens, L (2001), “Anger and advancement versus sadness and subjugation: The effect of negative emotion expressions on social status conferral”, Journal of Personality and Social Psychology 80(1): 86-94.

1 President Obama for healthcare.gov, Governor Christie for the George Washington Bridge scandal.
2 Shia LaBoeuf for plagiarism, Ted Nugent for racism.
3 General Motors for its safety issues, Netflix for its pricing, Apple for its Maps, JP Morgan for its regulatory mishaps.

8,504 Reads