Concepts for a New Economics: Complexity

Messing with the link between actions and consequences

Paul Samuelson’s Foundations of Economic Analysis cast a long shadow over economics. It centered economic reasoning around optimization and equilibrium. And economists treated economic policy as an exogenous variable, meaning that government behaves as a totally independent actor, capable of wise and benevolent intervention.

But the Samuelson paradigm is inadequate. Both before and after Samuelson, economists working outside of that paradigm have provided a better description of economic behavior.

Individual behavior consists of habits and experimental changes to those habits. The individual or firm owner responds to three levels of influence: its own preferences; the norms and expectations of the small groups to which the individual belongs; and the customs and formal rules given by the larger society. As the group norms or social rules change, individuals must react.

The overall behavioral system is complex, in the technical sense of that term. It is impossible to know how a change in behavior will affect the overall system. Every change of individual habits, group norms, or formal rules is an experiment, and people learn by trial and error. Learning is a challenge, because many factors affect outcomes, and no one can be certain how the causal chains interact.

Mortgage Credit Scoring

To illustrate complexity, let me give an example in which I was involved in the early 1990s. This was the introduction of credit scoring as a decision tool in mortgage lending. At Freddie Mac, the immediate purpose of this experiment was to make better decisions about which loans to make and which loans to turn down. But by the time the process change had reverberated throughout the system, it had contributed to a worsening of credit quality at Freddie Mac, and during the financial crisis in 2008 the company had to be taken over completely by the government.¹

Before Freddie Mac used credit scoring, the process of deciding whether or not to approve a mortgage loan included having an individual, known as an underwriter, examine the borrower’s credit history. The underwriter would approve or deny the loan, based on rules supplied by the underwriter’s organization and the underwriter’s own experience and judgment.

Credit scoring took the loan decision away from the individual human underwriter. Instead, every potential borrower was assigned a numerical score, based on an algorithm developed by a specialist firm based on what we now would call Big Data.

Using scores rather than human judgment improved mortgage lending decisions. The scores allowed lenders to make low-risk loans that a human would have rejected and identify high-risk loans to avoid that a human would have accepted.

As an immediate, local change, credit scoring improved the performance of loans backed by Freddie Mac. But the long-term systemic reaction differed.

Credit scoring made borrower’s credit quality legible to anyone interested in the loan. If you wanted to measure the risk of a mortgage default based on human underwriting, all you had to go by was the human’s yes-or-no decision to approve the loan. This was insufficient for potential investors to know whether loans they might purchase were unusually risky or not.

With credit scoring, a numerical risk score was available to any potential purchaser of the loan. Using credit scores, Wall Street was able to enter the business of packaging mortgage loans into securities while providing institutional investors in those securities with usable measures of the underlying risk. This eventually led to the creation of the exotic mortgage security instruments that brought down many financial institutions during the crisis of 2008.²

With credit scoring, Wall Street was able to securitize loans from low-income borrowers. This caused political embarrassment for Freddie Mac, which enjoyed government backing in part because politicians wanted to expand access to mortgage lending. Congress and the President set goals for Freddie Mac to demonstrate that it was providing mortgages to low-income and minority borrowers. In its effort to meet those goals, Freddie Mac allowed the credit quality of its loan portfolio to erode.³ The credit scoring innovation, as it spread through the industry and affected its political position, thus rebounded on Freddie Mac to the detriment of its risk management.

Financial Deregulation

In a complex economy, regulation is not a simple, straightforward process. For example, the development of Large Language Models such as ChatGPT has re-opened questions about intellectual property rights that have no easy answers. Is training the models on copyrighted books and proprietary web sites “fair use” of those materials or not?

Many pundits say that there was too much deregulation of the financial industry in the 1980s, as if regulation were simply a dial that you turn in the direction of “more” or “less.” In fact, the regulations related to bank safety and soundness during that period were made stronger, including risk-based capital requirements and more rigorous accounting and reporting standards.

The deregulation that took place in the 1980s was intended to make financial markets more competitive, by allowing different banks to move into each other’s territory. Banks were allowed to cross state lines. The attempt to separate commercial banking from investment banking was abandoned.

The goal of this deregulation was to reduce market power, thereby providing consumers with better services at lower costs. This in fact happened, as can be seen in the ability of consumers to earn higher rates on savings, have lower interest rates when they borrow, and buy stocks and other securities at little or no brokerage cost.

But opening up competition in banking also lowered the “franchise value” of having an exclusive market in a particular service or location. This changed the behavior of banks, making them more willing to risk, because they had less to lose if they failed. In a complex system, these consequences were not well anticipated by regulators.⁴

Complexity and Feedback

The moral of these stories is that in a complex economy, businesses and regulators who try new experiments will not receive immediate, clear feedback. Instead, the consequences of experiments will play out over time, and they will interact with other changes taking place in ways that are difficult to disentangle. Complexity in this sense is one of the most important concepts for economists to keep in mind.

1. I was long gone from Freddie Mac when the process played out. As an individual, one might have predicted that by pushing a useful innovation, I would gain personally. Instead, when the project to implement credit scoring was finally begun in 1994, I was rudely demoted by the individual who had maneuvered himself into position in charge of the project. At the time, I felt considerable resentment. But by motivating me to quit Freddie Mac, this actually worked out for the best for me.

2. The full story of the financial crisis is too long to be retold here. One little-known causal factor was the Recourse Rule adopted by regulators in 2004, which was intended to make financial institutions safer, but ultimately had the opposite effect. For more on the crisis in general see my Not what they had in mind. For more on the role of the Recourse Rule, see Engineering the Financial Crisis.

3. See Hidden in Plain Sight, Peter Wallison’s analysis, which made him quite unpopular and has been challenged in some quarters.

4. For more on the problem of imperfect information in regulation, see The regulator's Calculation Problem