Moral Hazard and Measurement Hazard

One of the most fundamental features of insurance markets is the possibility that providing insurance against a specific hazard will increase the incidence of the hazard being insured. Someone who is at least partially protected from a specific loss will generally have a reduced incentive to take costly actions to avoid the loss — the consequence being a higher probability of a loss than if there were no insurance. This phenomenon has long been recognized by practitioners of insurance and academics who study insurance markets as the "moral hazard problem."

If the term sounds a bit, well, moralistic, that's because it's an old term and may have originally been used to describe things that we might be more likely to see in moral terms — intentionally setting a fire to make a fraudulent insurance claim, for instance. But in its modern usage, it applies more broadly to the incentive effects of insurance, including cases in which moral judgment might not be so obviously called for. Risk avoidance is costly, and neither maximum avoidance (which you would tend to get with no insurance) nor minimal avoidance (resulting from full insurance) is likely to represent the most efficient insurance contract.

This basic trade-off between incentives for risk avoidance and financial protection against risk shows up in any insurance setting, including those in which insurance is provided by the public sector. Deposit insurance for banks generally makes banks and their creditors less likely to pay attention to risks that could lead to balance sheet losses — pushing the banking industry at least somewhat in the direction of a greater probability of suffering losses. And unemployment insurance (UI) affects the job-seeking incentives of the unemployed, pushing the labor market at least somewhat in the direction of more unemployment.

In the unemployment case, as in any other, reasoning about the direction of the effect on incentives is one thing. Discerning the magnitude of the effect is more difficult. The question in the case of UI, especially since the Great Recession, is whether and to what extent insurance benefits — and in particular, extended benefits — have affected employment. During the Great Recession, the federal government extended the maximum duration for unemployment benefits to 99 weeks in most states. Thus far, studies have found fairly modest effects from this change on unemployment: It appears to have contributed between one-tenth and one-half of a percentage point to the overall unemployment rate. (See "Expanding Unemployment Insurance" in this issue.)

Here in the Fifth Federal Reserve District, effective July 1, 2013, North Carolina's legislature dramatically reduced its UI benefit payout and duration. Consequently, North Carolina became ineligible for federal extended UI benefits six months before they expired for the rest of the country. This has invited comparisons between North Carolina's labor market performance and the performance of other states with access to extended UI.

Supporters of North Carolina's decision argue that the swift decline in its unemployment rate since July 2013 is evidence that cutting UI benefits reduced moral hazard and prompted unemployed individuals to more actively seek work. But critics of the cuts note that labor force participation in North Carolina also fell during the same period, suggesting that some job seekers who lost UI benefits gave up looking rather than found work. They also note that North Carolina's performance was similar to neighboring states that did not cut benefits early.

In this debate, a few words of caution are in order. First, and most broadly, it's always tricky to draw conclusions from a single example. North Carolina's labor market is a small sample within the whole United States, and attempting to apply its experience to the other 49 states is unlikely to provide enough evidence to conclusively determine the effects of extended UI. Second, problems analyzing data tend to grow as geographic coverage shrinks, especially in the case of unemployment numbers. The Current Population Survey used by the Bureau of Labor Statistics to track unemployment relies on a sample of households designed to be representative of the entire country. Disaggregating these data to estimate state-level statistics introduces some imprecision. Furthermore, labor market data at the state level are often more "noisy" than countrywide data. For example, decisions by a single major employer can have a large impact on state employment and mask the effects of policy changes like adjustments to UI.

Finally, it's important to bear in mind that assessing the effects of extended UI benefits on overall employment is one input to, but not the same thing as, assessing the desirability of that policy. If society places greater value on UI as a means to improve the welfare of the involuntarily unemployed, it may be more willing to tolerate some broader negative effects like increased unemployment duration. As with all insurance problems where there is an element of moral hazard, an optimal insurance scheme is one that weighs the benefits of cushioning the insured from some losses against the costs of altered incentives.