Policy and Choice
Public Finance through the Lens of Behavioral Economics
By William J. Congdon Jeffrey R. Kling Sendhil Mullainathan
BROOKINGS INSTITUTION PRESS
Copyright © 2011 THE BROOKINGS INSTITUTION
All right reserved.
Chapter One Introduction
When should the government intervene in the economy? When do markets fail? How do we craft policies that maximize social welfare? How do we design policies to minimize unintended consequences? Traditional public finance provides a powerful framework to tackle those questions. This framework, however, relies on an overly simple model of human behavior. This book revisits the core questions of public finance but with a psychologically richer perspective on human behavior. We do not merely apply psychology to economic problems; instead, we explore how psychological factors reshape core public finance concepts such as moral hazard, deadweight loss, and incidence.
To build our case, we construct a single analytical framework that encompasses both traditional policy levers—taxes and subsidies—and psychologically informed ones—such as defaults and framing. Three examples—health insurance, taxes, and externalities—illustrate how this approach alters our understanding of basic policy problems.
Models of health insurance emphasize moral hazard. Individuals choose care by comparing the price of care with its benefits. Since under insurance the price of care is often below its actual cost, people may overuse it. For example, because a consumer pays only a fraction of the full cost of an MRI, he or she may decide to get one even if it provides minor benefits. Insurance design seeks to balance the benefits of insurance against inefficient overuse, such as through copayments, health savings accounts, or consumer-directed health plans. For our purposes, notice how the logic of overuse relies crucially on individuals making choices in a narrow, calculating fashion: it occurs because consumers make a trade-off between the price of care and its true benefit.
Medical studies, however, suggest that health care choices are significantly more complex. Take the case of a diabetic who is prescribed medication. The cost-benefit calculus for taking the medication is clear cut. Diabetes is a serious disease, and insulin provides an important tool to manage it: the long-term health benefits drastically outweigh the monetary and "hassle" costs of buying and taking the medication. Human psychology can short-circuit that calculus. A patient focused on day-to-day concerns may simply forget to take his medication; another patient may simply "feel good" and decide that taking the medication is not worth it; and still another may decide to skip a dose simply because the benefits are in the future and not salient right now. Missing a single dose may not feel especially costly relative to the salient hassle costs ("I really don't feel like experiencing the pain of an injection right now"). Medication use by diabetics is not a unique example. Psychology affects decisions about nearly all types of medical care. In other words, the "psychic" cost-benefit calculus may be very different from the economic calculus.
For our purposes, we are particularly interested in how such deviations interact with traditional economic concepts, in this case moral hazard. We must now look beyond overuse of care. We must also consider the possibility of underuse: care that patients fail to use even when their benefits exceed the cost. That has important implications for policy design. Take the case of copayments—the payments made by an insured person each time he or she uses a medical service. The usual policy logic dictates that we can use elasticity of demand for a category of care to set copayments. A high demand elasticity means that the care is of low value. If small changes in price (which bring it closer to true cost) dissuade many people, the value of that care must not have been very high: a high demand elasticity signals overuse. As a result, copayments should increase with the elasticity of demand.
That logic fails in a behaviorally augmented model. A high elasticity of demand no longer indicates overuse. When a copay increase reduces demand, we can no longer infer that the care is actually of low social value; perhaps people were underusing it and we are worsening the problem. When individuals do not choose optimally, a change in demand tells us only that people choose as if they do not value the care. Return to the case of insulin treatment for diabetes. A patient who was non-adhering on some days because he feels that medication is optional on days when he "feels good" will show price sensitivity: he will skip more doses on those days if prices are high. In effect, he feels that the care is optional. Increasing copays for such a patient on the basis of that elasticity would, however, be worsening a behavioral bias. In effect, psychology forces us to reinterpret empirical data on demand. Empirical studies can no longer simply use demand elasticities to measure moral hazard. We must understand more about the category of care where the elasticity appears. The demand elasticity is no longer sufficient for setting policy. The optimal amount of a copayment must be based on both knowledge of elasticity and an external assessment of the value of the treatment. In some cases, optimal copayments may even be negative—for example, in cases in which it is worthwhile to pay people to take their medication because of the positive spillover effects of doing so.
We can also examine nonprice levers. Consider the provision of "nudges," the label given by Richard Thaler and Cass Sunstein to psychologically astute interventions that influence behavior. In the case of drug adherence, an example would be simple reminders to take medications. Once we recognize that nudges can affect use of care, we must examine insurance design more broadly. When will insurers nudge patients to use care? When will they nudge patients to reduce use of care? The answer depends on how profits align with health outcomes. Take again the case of drug adherence. Patients' failure to adhere to treatment regimes has long-term costs: hospital admissions, for example, will be higher. A long-term insurer will bear those costs; as a result, a profit-maximizing long-term private insurer will have incentives to devise and implement nudges to increase adherence. Investments in disease management—which many companies increasingly make—can be understood from that perspective. In contrast, a short-term insurer bears none of the costs of patients' non-adherence. They not only have zero incentives to provide nudges to improve adherence, they also have perverse incentives to find nudges that discourage use, even when use has high long-term benefits for the patient. For example, the short-term insurer can create costs by making it a hassle to schedule a doctor's appointment or to refill a prescription. The psychological perspective therefore can add to our understanding of why health insurance is structured in certain ways when provided in a private market.
A fuller integration of behavioral economics and public finance allows us to go beyond just suggesting specific psychologically astute policies to experiment with. It provides a different framework for understanding such traditional public finance levers as copayments and market structure.
Governments must raise revenues to provide services. Traditional public finance has a well-developed framework for determining how to set taxes optimally. Models of incidence help us understand who bears the burden of taxes; models of efficiency help us understand how taxation can hinder economic activity. Together they offer practical insights for designing policy for taxes of all stripes: income, sales, and so on. For example, one broad insight is that efficient taxes are those that minimally distort consumer choices. Since individuals were choosing optimally in the absence of taxes, a change in the choices that they make represents a welfare cost. Concretely, one should raise revenues by, for example, taxing low-elasticity goods—taxes on, for example, cigarettes are often justified in part for this reason.
Behavioral economics complicates that logic. One recent study finds, for example, that individuals may fail to perceive sales taxes that are not included in the prices posted on store shelves but are computed at the register. People may simply fail to attend to them—they are not salient at the time of choice. Applying traditional logic, tax non-salience represents an opportunity for governments: they can raise revenues without distorting behavior. That logic, however, is incomplete. Lack of response to a non-salient tax is not the same as lack of response to a salient tax. When people fail to respond to a non-salient tax, there is an error: they make consumption choices as if an item costs $X, but in purchasing the item they actually spend $X + $Y. As a result, they have $Y less to spend in the future than they had planned.
How that affects all other consumption must now enter the welfare calculation. Consider two polar cases. The lost money could be treated as a pure income effect: individuals see that they have $Y less to spend on all other goods and adjust accordingly. That would, in effect, turn the non-salient tax into a lump-sum tax, and governments therefore should use non-salient taxes heavily. Alternatively, suppose that the $Y is taken out of a narrower mental account. For example, rather than thinking of their overall budget as depleted by $Y, individuals think of $Y as depleting their grocery budget specifically, and they may spend $Y less on their next trip to the grocery store. Or they may never change consumption and instead simply end up saving less. In such cases, the low demand response to non-salient taxes is misleading: though it does not generate distortions in the demand for the good being taxed, it is creating possibly higher distortions elsewhere. As a result, governments would need to take into account other potential distortions before using non-salient taxes.
In this case we also see that it is impossible to think about the implications of a nudge on tax salience—for example, excluding taxes from posted prices—in isolation from the public finance framework. The simple application of traditional logic suggests that one should always use nudges to reduce tax salience. In an integrated framework, that is no longer the case. The effects of reduced salience must include all the demand responses that it elicits.
One of the triumphs of public finance is to provide a clear understanding of how to deal with external costs (externalities). Take the case of carbon emissions, which contribute to global warming, a typical negative externality. Individuals and firms do many things that affect carbon emissions, from driving automobiles to retrofitting factories; in making their choices, they impose costs on society. Traditional public finance provides an elegant solution to ensuring that those externalities are internalized in choice: individuals and firms must face the full costs of their carbon-emitting activities. The prices that they pay must include not only the marginal cost of the goods that they consume but also the cost of the carbon emissions that those goods produce. Put simply, we can achieve economic efficiency by placing a carbon tax on goods that is equal to the social cost of the carbon emissions produced by those goods. There are technical and political challenges in implementing such a tax, but the conceptual solution is clear.
As with the other examples, decisions involving carbon emissions may not be made in accordance with standard assumptions. For example, psychological studies suggest that social comparisons can drive behavior. Being told, for example, that "you used x kilowatt hours last month, but your neighbors used y kilowatt hours" can reduce a person's consumption of electricity. Based on that insight, a company called OPOWER has implemented a large-scale program that charges utilities to send social comparison reports to consumers. In randomized, controlled trials with hundreds of thousands of utility customers across the United States, the reports have been shown to reduce electricity consumption in the average household by about 2 percent. Notice several interesting aspects of this example. First, even with a traditionally efficient carbon tax, there may be inefficiency if consumers do not choose their energy consumption levels optimally. Second, in addition to the role the prices play in affecting behavior, nudges or other interventions can play a powerful role. Third, and most important, in this case the private sector has generated a nudge—social comparison reporting—in order to affect energy consumption.
The last point is especially interesting because it suggests that policy levers besides carbon taxes and government-imposed nudges can be devised. Can the government somehow induce firms to nudge effectively? They have levers that can be used for that purpose. Consider decoupling for utilities, under which the profits of electricity retailers are no longer directly related only to the volume of electricity sold; they also receive revenues for reducing consumption. That type of lever, if it encourages utilities to nudge consumers toward reducing energy use (as it has in the case of OPOWER), is a powerful tool.
The logic here involves both economics and psychology. Psychology recognizes the power of nudges; economics recognizes the power (and peril) of markets. Firms may have nudges available to them that the government does not. So the government can do better than just implement its own nudges; it can look for policy levers such as decoupling that encourage firms to create and use nudges to improve consumer well-being.
These examples suggest that, first, psychological insights must be applied more deeply and broadly in public policy. They can be more than an added-on tweak at the end of a predetermined economic policy—they can alter the basic policy framework, from deadweight loss to moral hazard. Second, many of the policy suggestions based on behavioral insights are not especially behavioral. Decoupling is a traditional economic policy lever, but the behavioral approach enriches our understanding of its impacts. Finally, these examples illustrate that the law of unintended consequences continues to be important: policy changes (nudges or otherwise) must continue to be analyzed within the broader system in which they operate.
While integrating the psychological insights of behavioral economics into public finance policy holds great promise, as illustrated above, doing so also introduces a set of potential stumbling blocks for analysis and policy design. The approach described in this book overcomes or at least alleviates two of the major challenges.
Can the number of potential psychological factors be made manageable?
Psychology is, naturally, a very rich discipline, full of insights. That richness creates an overload of information. For any policy problem, it seems that an endless array of psychological phenomena could be relevant. The length of unemployment spells could be influenced by cognitive dissonance, hyperbolic discounting, anchoring, overconfidence, and loss aversion, to cite just a few examples. How do we handle such a vast array of possibilities?
We believe that the answer lies in abstraction. Knowing the specific psychological factor that drives a behavior is important in designing nudges. For example, job seekers may procrastinate in searching for jobs for a variety of reasons. Some activities (going out with friends, watching TV) may be enjoyable and therefore hard to resist. On the other hand, unemployment can sap a person's motivation, making it hard to exercise the self-control needed to engage in day-to-day activities such as sending out resumes. Those factors suggest different interventions: should we reduce procrastination by offering people a chance to commit themselves to searching for work in the future, or by finding a way to remotivate the unemployed?
Our insight is that despite the differences in those two examples, they have much in common: in both, the unemployed individuals recognize the future benefits of searching for a job; in both, they would like to and plan to search for a job; in both, they are unable to implement their desires because, at the moment, something (a tempting activity, lack of motivation) intervenes. We can lump those factors and other phenomena into a particular category labeled bounded self-control—the category of psychological factors that reflect a general tendency whereby people would like to take an action with future benefits but fail to do so. Categorizing helps us to craft policy principles. For example, when bounded self-control is a problem, we would argue that one must be very careful about the structure of incentives. Giving a person a bonus to leave unemployment will have weak effects if the benefits are realized far in the future. Those with bounded self-control already recognize and would like to capitalize on the future benefits of searching for a job. Their problem is implementing their desires; adding a modest bonus to those future benefits will not help much.
Excerpted from Policy and Choice by William J. Congdon Jeffrey R. Kling Sendhil Mullainathan Copyright © 2011 by THE BROOKINGS INSTITUTION. Excerpted by permission of BROOKINGS INSTITUTION PRESS. All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.