On 10th October 2016 Bengt Holmstrom and Oliver Hart, two academics who developed modern contract theory, were awarded the 2016 Nobel memorial prize for economics. Contracts are everywhere in modern economies and the investment industry is no exception. They determine how investment professionals are remunerated for their work and how investment institutions are paid to provide services to other institutions, eg managing a portfolio. How does current practice in the industry compare with best practice as defined by the modern contract theory framework? Does it provide appropriate incentives and alignment? I address these questions in this blog post by reference to a small selection of Holmstrom’s work which focuses on applied mechanism design (ie how to design a contract). Hart’s work is more concerned about contracts relating to the ownership of firms.
In one of his earliest publications “Moral hazard and observability” (1979), Holmstrom introduced the so-called “informativeness principle” which is now widely recognised as one of the key principles in addressing principal-agent problems. In essence it means that if there exists any information that can reduce the uncertainty with regards to what an agent actually does, then this information should feature in the contract. The example he uses concerns CEO compensation. Holmstrom suggested that because share prices reflect factors in the economy outside the CEO’s control, simply linking compensation to the firm’s share price will reward the CEO for good luck (or punish him/her for bad luck). It is therefore better to link his/her pay to the firm’s share price relative to those in the same industry. I would argue that the practice of linking performance fees to absolute returns, while having significant exposure to the market, is in violation of this principle. In many real-world situations it is simply very difficult to separate good (or bad) luck from the effect of actions, even with the assistance of all available information. Arguably, assessing an asset manager’s skill in beating the market is one of these situations. What would modern contract theory say about it? According to Holmstrom’s findings (cited by this FT.com article) the more difficult it is to observe the effect of one’s actions (whether individual or institutional), the less remuneration should be performance-based. Where there is significant uncertainty, it is simply better to make fixed payments.
Holmstrom’s 1991 paper with Paul Milgrom, “Multitask Principal Agent Analyses – Incentive Contracts, Asset Ownership and Job Design”, considers situations in which the agents’ tasks are multi-dimensional. I believe that it is a fair and accurate description of most, if not all, real-world situations given that I have not yet seen a job description document with only one line in it. In this situation, Holmstrom argued, performance-based incentives have an important role in directing the allocation of the agents’ attention among their various duties. This is why a school teacher’s compensation should never include a variable component linked to the results of his/her students’ standardised exams – in which situation the teacher would be incentivised to only teach the narrowly defined skills that are tested, at the expense of activities such as promoting curiosity and creative thinking, which are harder to measure. What are the implications for the investment world? There is the issue of how sales people are, generally, compensated in our industry. I think it is reasonable to say that an asset manager’s sales team’s task is at least two-dimensional: 1) to meet clients’ needs by recommending suitable products and 2) to increase the firm’s revenue by selling more products. Arguably the current practice seems to only link incentive pay to factor #2. Holmstrom’s work provides us a strong theoretical underpin to understand where this design would lead our industry to. Given the complexity of the issues, any contract design will need to account for the unique circumstances, and the incentive problems must be analysed in totality. As a result I am not suggesting there is a panacea here. However Holmstrom suggested that, in certain situations, it would be better to fix the compensation (no incentive component) than to base one’s compensation only on a subset of the dimensions (the ones that can be effectively measured).
Our congratulations to both Bengt Holmstrom and Oliver Hart for their richly-deserved prize. And well done to The Simpsons (read the story here) for correctly predicting the winner six years prior!
