The principal-agent problem arises when a principal hires an agent to pursue the principal's interests but the agent develops priorities of his/her own.
The problem is found to different extents in all large organisations, including in the way that shareholders and senior executives find it very difficult to ensure that middle managers, foremen etc. work to the corporate agenda as seen from head office. There is a strong tendency for humans to align their goals and behaviour to that of the team or work group around them. As a result, it often seems that corporations are in practice run in the interests of the firm's managers who have grown to regard their firm's economic interests as entirely coincident with their own. (See Note 4 below for a comment on organisational design.)
It is for instance almost universally common for middle managers to focus on what they see as the long term good of their factory, office or other small part of the organisation. When asked to find efficiency savings they will fight hard to retain their budget and their staff numbers. They see no point in implementing all those 'stupid directives from head office', and find record keeping a bore. Boxes are meant to be ticked, whether or not the associated action has been carried out. They will often enter into an implicit bargain with their teams, allowing the use of material for personal ends, providing generous expense accounts, etc. so as to generate a better (non-confrontational) climate within the team - often characterised as 'high morale'. Such teams resist change - especially if it might lead to greater efficiency (working harder and/or job losses).
Large organisations often also suffer from the physical and geographical separation of head office from operations. Boeing was a particularly termer and tragic example of this, in the years before the 737-Max crashes. Michael Godfrey wrote this in the FT:
'The crisis at Boeing indeed had its beginnings in the merger with defence contractor McDonnell Douglas. Prior to the merger, management and engineering at Boeing were all located at the company’s headquarters in Seattle and McDonnell Douglas had its headquarters in St Louis. As a result of the merger a compromise was reached: the new management headquarters would be in Chicago. By this choice both parties lost.
By now it is clear that no one in Boeing Chicago has any idea what has been going on in Seattle. This was inevitable. Before the merger, management at Boeing Seattle worked well, including by means of management walking around the engineering areas. It is a long walk from Chicago to Seattle.'
The problem gets even worse when the organisation's Board of Directors (or equivalent) is formed of out-of-touch mates of the Chief Executive and/or chosen from the 'Great and Good'. It is no longer the case that investors own corporations and appoint a board (a) to set corporate strategy and (b) to hire a management team to execute that strategy. The current reality is that managements almost always choose who will be on their board, and managements choose the strategies which will most likely benefit themselves. It is not a huge over-simplification to say that the 2008 financial crisis was caused by risk-hungry managements and weak boards - as well as weak regulation. The problems at Lehmans are summarised here, and Theranos is another good (non-financial) example:-
The company was set up, with the best of intentions, by Elizabeth Holmes who was determined to find a way to make blood testing more convenient for both doctor and patient. John Carreyrou's book Bad Blood describes how it all went terribly wrong. He notes that Ms Holmes had recruited the following big names to her Board: George Shultz, James Mattis, Henry Kissinger, William Perry, Sam Nunn and Gary Roughead. "They were men with sterling, larger-than-life reputations who gave Theranos a stamp of legitimacy. ... Elizabeth had methodically cultivated each one of them and offered them board seats in exchange for grants of stock." Needless to say, none of them had either the knowledge or experience to understand what the company was doing, nor were they inclined to be critical of their apparently fabulous Chief Exec.
It is also the case, of course, that some middle managers are quite weak, and incapable of standing up to staff who are not doing a good or reliable job.
And senior managers often engage in a sort of magical thinking, believing that - by 'banning' something in their organisation - they actually get rid of it, as if they had pointed a magic wand and the 'banned' thing actually disappears in consequence. Spoiler - it doesn't!
There are two consequences for regulators.
- First, it can be very hard to obtain reliable information from the intermediate levels in any hierarchy.
- Second, regulators need to be aware that agents and their teams often resist the introduction of (what they see as) tedious protocols aimed at improving quality or safety. Rail and marine accident reports include many examples of such behaviour. It is quite common for teams to falsify data, including quality and safety data. New recruits are told by experienced colleagues that formal instructions - or things they learned in an induction program - can safely be ignored as out-of-date or impractical.
- One of the most far-reaching examples of the new-recruit problem was the 2020 Minneapolis killing of George Floyd by a police officer Derek Chauvin that triggered the Black Lives Matter protests. Mr Chauvin was a training officer for new recruits, and one of the three policeman that failed to intervene in the killing was on only his fourth day in the force. The latter's lawyer argued that "They're required to call him 'Sir'. He has 20 years' experience. What is my client supposed to do but to follow what the training officer said?"
- The New York Times noted subsequently that "the police department has been recruiting a new crop of trainees who will face the same challenge as every rookie: navigating the dramatic difference between what is preached in the academy and what is practised in the street.
- In the UK rail industry, there was concern in 2020 about dangerously ambiguous communication between signaler and train driver. "A signaler straight out of training school [using proper protocols} may have the professionalism knocked out of them by the prevailing culture among the old hands on the signalling force. ... [Proper protocols] are still seen as slightly embarrassing, slightly 'ooh, get him, who does he think he is?''
It can therefore be very hard for a regulator to find out what is really happening inside a regulated entity, not least because the entity's own senior executives probably do not themselves know what is happening and, if they do, will be very reluctant to admit their partial loss of control. Much the same applies to compliance officers, whose creation is too often regarded by their Board as a box which has been ticked, and so a good reason to ignore regulatory issues.
It is tempting to assume that the failure of front line staff to comply with safety, financial and other protocols is because they are responding to pressure from above, such as to meet financial or other targets, to complete work quickly, to maintain production etc. However ... leaders of rule-bending teams may themselves be keen to get through work quickly, and/or to impress seniors with their achievements, or maybe simply to get home on time, or to avoid a small amount of work running over into the next day. It is truly very difficult for regulators and company directors to know that this is happening unless there are robust and unpredictable inspection arrangements backed up by a strong compliance culture, and opportunities for whistle-blowing.
The principal-agent problem also means that regulators need to take into account how the middle and lower reaches of an organisation might react to regulatory pressure. The need for change and/or improvement may be accepted by senior executives, but this does not mean that the necessary changes will be accepted or implemented at the working level. The regulatory intervention needs to be designed with this problem in mind.
Here are some other examples of the principal-agent problem. (Others, which have caused major problems, may be found here.)
- Thames Water was fined £14m after a malfunction at a New Malden sewage works triggered nearly 50 remotely-monitored alarms, none of which was checked. 79 million litres of raw sewage flooded local land and a river.
- 'Rogue trader' Kweku Adoboli nearly destroyed UBS in 2011, at one point exposing the bank to losses of £7.4bn. A colleague had reported Mr Adoboli's breach of financial trading limits in 2010 but found that this soured their relationship and so did not report him again when (a) the same thing happened in 2011, and (b) he learned that Mr Adoboli was using 'an umbrella fund' to hide his losses. "I felt as though I had stitched him up ... I think that maybe he didn't trust me as much [after he was reported] ... "I went to a school where people didn't grass," said the colleague.
- Manhattan's prestigious Stuyvesant High School provided another good example in 2012 when caught up in what was described as a pervasive cheating scandal. The New York Times reported that "Although students enter the school knowing they are among the best in the city, they must compete with hundreds just like them. ... They described teachers as being relatively sympathetic, discouraging cheating, but not always punishing it as severely as school policy dictates." One consequence may have been that Stuyvesant's results became less trusted after this episode, thus damaging the academic and career chances of all its alumni.
- And investigations into the LIBOR scandal unearthed some juicy emails showing junior (but extremely well paid) staff's willingness to disregard ethical and other rules. 'Its just amazing how LIBOR fixing can make you that much money ... it's a cartel now in London' wrote one trader. The Financial Services Authority subsequently concluded that some bankers had decided that 'the rules did not apply to them' and noted that 'In March 2011, RBS attested to the FSA that its controls and systems were adequate. The attestation was inaccurate.' It took another year before RBS concluded that the bonuses of the back-room staff who submitted LIBOR figures should not have their bonuses linked to those of the traders who might make money out of manipulating those same LIBOR rates.
- There are innumerable examples in formal accident reports. Here is a nice one, following a collision in the English Channel in which one ship's collision avoidance technology had been partly disabled:
- "The settings in the 'guard zone' and 'target alarms' areas ... were locked. The adjustment of these settings was password protected and [the] deck officers ... were unaware of the password. The crew considered the resulting absence of alarms to be beneficial."
- An HSE Inspector reported that a contributory factor to a serious roller-coaster accident at Alton Towers had been that the frequency of false alarms meant that staff tended to assume that all alarms were false, and so restarted the ride without carrying our proper checks.
- An investigation into an accident on a heritage railway, in which a child was nearly killed, found that there had never been an audit of the Safety Management System.
- "Had SDR [(the train operator) carried out this audit], it would have identified that the engineering staff within the carriage and wagon workshop had not been signing the fitness to run forms for some years. Although SDR staff conducting the fitness to run examinations were confident in their own ability and competence to identify if a fault required a carriage to be removed from service, they did not feel comfortable in signing and dating the forms to show that they had examined the condition of the carriages and were authorising that they were fit to run. They stated that this was due to the number of faults being identified, the ‘fix and patch’ culture, and the perceived possible personal implications for them of signing the form and an incident or accident occurring subsequently."
- Railway accident inspectors reported in 2018 that
- "Members of the [track maintenance] team told us that [their supervisor] neither fully briefed them on the safety arrangements, nor checked their track safety qualifications. Nevertheless, they all signed ... to acknowledge that they had received a briefing."
- Ian Prosser and David Keay's History of Her Majesty's Railway inspectorate contains several examples of 'blind eyes' being turned to 'local arrangements' (i.e. breaches of regulations) the most damaging causing 224 deaths at Quintishill in 1915.
- The consequences were commercial rather than damaging to health, but I was struck by Nick Butler's comment, in the FT in January 2014, about a profits warning by Royal Dutch Shell:
- 'At the heart of the .. problem seems to be the gap between operational reality and top management, including the board. Big problems capable of triggering a profits warning in a company this size do not arise over 80 days. They grow more slowly and are often invisible to boards meeting every six weeks who have to trust whatever data they are given. .. The rest of the industry should avoid gloating. It would be much more useful for the boards of the other majors to ask themselves if they really know what is happening in the companies for which they are legally responsible'.
- Sivaraj Tharmalingam suffered from seizures and should have been checked every five minutes by SERCO custody staff whilst he was in a cell at a magistrates court. Instead, according to the Coroner inquiring into Mr Tharmalingam's death:
- 'There was a failure to carry out any checks by SERCO ... and the observations recorded on the computer system bore no relation at all ... to the observations actually made of the [other] detainees.'
- 16 year old mental health patient William Jordan was left unattended for several hours before he hanged himself in a Priory Clinic - despite clear instructions that he was to be checked every 15 minutes. Staff then falsified logs so it appeared as if they had in fact checked he was OK.
- White House security was compromised when the 'crash box' intruder alarm was muted at the request of White House ushers who complained that the boxes were too noisy.
- Buckingham Palace security was famously compromised when Michael Fagan broke in and spent some time talking to Queen Elizabeth in her bedroom. Police ignored several alarms believing them to be faulty.
And the principal-agent problem is hardly a new one. CE Montague tells the delicious story, in his book Disenchantment, of the old hand sergeant-major who, instead of taking his young recruits on a long training march, takes them to a pub where they are surprised to find that 'arrangements for serving a multitude are surprisingly complete', and that their sergeant-major is clearly well acquainted with the publican. I suspect that armies have suffered from, and Generals will have cursed, this sort of problem right back to the beginning of armed conflict.
Notes
See Jean Tirole's Hierarchies and Bureaucracies: On the Role of Collusion in Organizations for a lengthier discussion of the Principal-Agent problem.
It can be argued that the principal-agent problem explains the central issue of organisational design. Although delegating decision-making is an important part of good organisation design, it is of little use unless the decision makers share the organisation’s objectives. That is why incentives are so important. The individual with valuable local knowledge should be incentivised to act as if the objectives of the organisation are his or her own.