Excerpt from Side Effects: The Economic Consequences of the Health Reform
by Casey B. Mulligan
(c) copyright 2014 by JMJ Economics

Chapter 1

The Paradox of Affordability

. . . in our unbelievably rich land, the quality of health care available to many of our people is unbelievably poor, and the cost is unbelievably high.

Senator Edward M. Kennedy (December 9, 1978, speech at the Democratic National Committee midterm convention)

As the nation has become richer and health technologies have advanced, health care has become capable of delivering results ranging from new knees and hips for the elderly to a good chance at a long life for premature babies, to peace of mind for patients and their families. At the same time, spending on health care has grown faster than the economy itself. Millions of people pay a significant portion of their income, often in the form of voluntary deductions from their paychecks that go toward health insurance premiums, so that they and their families can access good doctors, hospitals, pharmaceuticals, and medical devices when needed. Many people take jobs solely for the purpose of paying for their health insurance. The magnitude of their sacrifices demonstrates the importance that people ascribe to health care.

In order for patients to receive the healthcare results that they value, society has to dedicate workers to diagnose illnesses, administer treatments, think and experiment with ways to improve health care, manufacture medical devices, adjudicate disputes, administer payment systems, and produce structures, equipment, and software to assist with these tasks. Those working in health care and supporting industries are workers who cannot instead be producing goods and services such as food, transportation, and entertainment.

Because people differ in terms of their income, health status, family situation, and priorities, invariably a segment of the population is unable or at least unwilling to pay for their own health care. After Senator Kennedy’s speech, the share of the population without health insurance continued to grow, at least in part because health care continued to get more expensive (Cohen et al. 2009).

In many other markets, it is tolerated and maybe even welcomed when a customer segment stops buying in response to high costs. Many households, for example, have stopped subscribing to cable television as the monthly cable bill grows more expensive and they are no longer able or willing to dedicate the funds to such an expense. The trend toward dropping a cable subscription is not universally viewed as a serious problem. But health care is said to be different: people are supposed to get quality health care even if they have not taken steps to purchase health care themselves.

In the past, the federal government stepped in to deliver health insurance to populations that were least likely to get it on their own. In the 1960s it created the Medicare insurance program, which heavily subsidizes payments to health providers on behalf of elderly patients. The scope of the program has expanded over time, most recently with the 2003 addition of prescription drug benefits. Medicaid is another public health insurance program created in the 1960s and expanded in recent decades, in this case by federal and state governments on behalf of poor families, especially those with children. The Affordable Care Act of 2010 (hereafter, ACA) is the most recent and perhaps most significant federal law intended to reduce the fraction of the population without health insurance.

Medicaid and Medicare help poor and elderly individuals avoid some of the tough sacrifices that would be necessary for them to purchase health insurance without assistance. On average, Medicaid and Medicare permit beneficiaries to work less, spend more on nonhealth-related goods and services, or both. But no program can change the fundamental reality that society has to pay for health care with more people in the workforce, fewer nonhealth goods and services, more productivity, or all of the above. Thus, while Medicare and Medicaid help their target populations and give them a bigger slice of the economic pie, the programs also diminish the pie itself. The programs reduce, among other things, how much program participants work on average, exacerbating the societal problem that the economy as a whole cannot expand its health sector without giving up something else of value. In effect, people who are not receiving assistance from Medicare or Medicaid are paying twice for the programs: once as the total economic pie gets smaller and a second time as they receive a smaller piece.

A reasonable person might conclude that Medicare and Medicaid do relatively little to shrink the economic pie, because even without the programs the vast majority of work would be done by people who are neither poor nor elderly. But the economic effects of the ACA are a different matter since the program involves the bulk of the U.S. population that has been excluded from Medicare and Medicaid, and thereby the people who have been doing most of the work in the economy. The purpose of this book is to use economic reasoning to measure the taxes in the ACA and to offer quantitative predictions about its effects on the labor market, capital accumulation, and total production in the U.S. economy.

A. Hidden Taxes

At first glance it might appear that the ACA helps people get access to health care and disproportionately benefits low-income households without many new taxes. By one estimate, the ACA’s tax increases are less than 0.5 percent of gross domestic product, and less than several other hardly memorable tax increases of the postwar period (Klein 2012). The White House suggested that health reform would largely pay for itself, without mentioning taxes that, individually or in combination, would have more than a “little effect” on the labor market (Council of Economic Advisers June 2009).

Politicians and journalists use the term tax more narrowly than economists do, but the economic definition is needed to understand the effects of the ACA. Suppose, hypothetically, that the government provided a “universal” $2,000 health benefit to every person and paid for it with a tax, in the narrow sense of the word, of $4,000 per employee. Employees are half the population, so the employee taxes average $2,000 per person and are enough to pay for the universal benefit.

Now consider an alternative “targeted” approach that pays the $2,000 health benefit only to people who do not work and gets the revenue from a $2,000 tax per employee. By excluding workers from the benefit, the targeted approach appears to spend and tax less: only $1,000 per person. But the economic result is the same because, in both systems, employees pay $2,000 more than they receive. In both systems, people who are not employed receive more than employed people do; in the universal system their lack of employment exempts them from a large tax whereas in the targeted system it exempts them from a smaller tax plus it gives them access to a benefit that is withheld from workers.

Withholding benefits from people who work or earn is hardly different than telling them to pay a tax. For this reason, the field of economics refers to benefits withheld as “implicit taxes.” What really matters for labor market performance is the reward to working inclusive of implicit taxes, and not the amount of revenue delivered to the government treasury according to economically arbitrary distinctions between implicit taxes and other taxes. The targeted system gives the same economic results, including the economic harms from taxes, as the universal benefit system does but without the (politically ugly) appearance of bringing significant revenues to the government treasury.

The ACA resembles the targeted approach because it is full of implicit taxes. Many of them have remained hidden “in the fog of controversy” surrounding the law and their effects excluded from economic analyses of it. As of 2014, essentially the only place to find an economic analysis of the ACA’s large and hidden employment taxes is this book (or drafts of its chapters). No investigation of the economic effects of the ACA should be considered accurate unless it accounts for the ACA’s implicit taxes.

Figure 1.1 puts the ACA’s new taxes in perspective of federal tax increases over the past seventy years. The taxes include federal personal income taxes (Form 1040, shown in pink), social insurance payroll taxes (gray), and various employment and implicit taxes (red). The figure does not show revenue for the U.S. Treasury—that statistic is vulnerable to some of the arbitrary distinctions noted above—but instead shows the effect of various tax laws on the incentives for workers to earn more labor income rather than less as measured by a marginal labor income tax rate (by marginal labor income tax I mean the extra taxes paid, and subsidies forgone, as the result of working). During a period that included more than a dozen tax increases, the ACA is arguably the largest as a single piece of legislation, adding about six percentage points to the marginal tax rate faced, on average, by workers in the economy. The only way to cite larger marginal tax increases would be to combine multiple coincident laws, such as the Revenue Acts of 1950 and 1951 and the new payroll tax rate that went into effect in 1950. The four payroll tax rate increases between 1970 and 1980 are another example of a large rate increase if we also include the personal income tax rate changes that occurred during the decade owing to inflation causing taxpayer incomes to creep into higher tax brackets without any new legislation. Even with these adjustments, the ACA is still the third largest marginal tax rate hike during the seventy years. Another feature of the ACA that distinguishes it from other large marginal tax rate rises is that the former is, by law, entirely permanent whereas essentially the only other permanent ones shown in Figure 1.1 are the payroll tax rate changes.


Figure 1.1 represents the ACA as a single number, but underlying that number are multiple economically distinct taxes. Chapter 2 of this book is an introduction to a dozen or so provisions in the law and offers some indicators of their relative importance. The employer penalty is explained and measured in Chapter 3. Chapter 4 features the employee penalties that are hidden in the ACA’s arrangements for subsidizing health insurance assistance. The subsidy rules also include a couple of new implicit income taxes, which are discussed and measured in Chapter 5.

Economists generally acknowledge that taxes have side effects, and this book is not unusual in terms of its representation of the amount of unintended consequences per dollar of taxation. The real surprises are the wide range and astonishing size of the taxes measured in Chapters 3, 4, and 5. Each of the chapters shows how the ACA’s new taxes can put millions of workers in a “100 percent tax” situation in which full-time work pays less than part-time work, or less even than not working at all. This is without counting the workers who can raise their disposable income by working less and thereby climbing one of the “cliffs” that are part of the ACA’s rules for determining federal assistance on the basis of household income. Anyone interested in the evolution of the labor market in the United States has to understand the ACA’s new taxes—they’re vastly more important than, say, the interest rate on fed funds—and the first half of this book is so far the only comprehensive and user-friendly introduction to the topic.

B. Using Economics to Forecast Policy Consequences

A complicated law like the ACA has forces pushing in multiple directions. For example, the ACA contains an implicit tax on unemployment benefits. This by itself tends to reduce unemployment and increase employment, which is the opposite result of some of the other taxes in the law. But it does not mean that anything is possible, or that we must wait for more data before having any idea as to which forces dominate. The ACA’s various taxes can be quantified individually and collectively. Chapters 2 through 5 show that the law’s employment taxes far outweigh its employment subsidies, and its income-earning disincentives far outweigh its income-earning incentives. When the government taxes something, the usual result is to get less of it, which is why I expect national income per person and hours worked per person to be less than they would be if the ACA had not been passed.

In writing this book, my exemplar was The Economic Consequences of the Peace, in which John Maynard Keynes (1919) offered his predictions for the effects of the 1919 Treaty of Versailles (between World War I Allied Powers and Germany) on the German economy. Keynes believed that several economic consequences of the Treaty were knowable ahead of time, and he wrote his book after the hotly debated document was written but before it was fully executed. He carefully quantified the economic provisions of the Treaty, and the economy that would be affected by them. Keynes wrote little about “the ideal question”—what should have been done—and instead focused on the consequences of what was actually written. I tried to include the same basic ingredients in Side Effects, and treated the implementation of the ACA as an opportunity to learn about the economy and applied economic theory, operating under the assumption that the ACA would inevitably go into effect without significant modification of the law (as interpreted by the Obama administration as of early 2014, with exceptions noted in the book). As it turns out, I reached conclusions analogous to Keynes’: that full execution of the document (the Treaty in his case, the ACA in mine) would create significant economic side effects, and that advocates of the document were not fully aware of, or forthright about, the costs created.1

After Keynes, many detailed economic analyses of public policies and the national economy—Friedman and Schwartz (1963), Costa (1998), Cole and Ohanian (2004), and Goldin and Katz (2008) are good examples—have been executed after the policy of interest is in place for a number of years and thereby with the benefits of hindsight and ample time for reflection, debate, and synthesis. The historical approach is valuable and should be continued, but it sometimes gives the impression that policy consequences are inherently unknowable until after the fact, if ever. With its reliance on hindsight, the historical approach arguably fails to profit from one of the major strengths of economic theory, which is to identify common elements of economic events on the basis of incentives instead of outcomes. I deliberately completed this book before mid-2014, and therefore before the release of any of the data needed to execute a historical evaluation of the ACA, in order that we may have a detailed and comprehensive analysis of the law’s effects on the economy that cannot rely on hindsight.

To the economic theorist, few events are fundamentally new; rather, they tend to be new collections of familiar economic forces. Thus plenty of data were available to assess likely consequences of the ACA even before the law went into effect. The law itself is data on the types and magnitudes of incentives that would be created when (and if) it is enforced. A wealth of historical episodes provides quantitative information about how employers, employees, and consumers react to employment taxes, insurance costs, and other economic forces unleashed by the new law. This book presents and refers to much of that data, and it organizes them with some of the basic models in economics such as the theory of substitution effects, Adam Smith’s theory of equalizing differences, and the neoclassical growth model.

Quantitative forecasting with economic theory also offers a kind of discipline against excessive simplification or exaggeration of the main effects because they will likely be punished with obvious contradictions by future events. Forecasting also has a way of honing one’s attention on what really is unknown—for example, the degree to which people and businesses will comply with the new law—and the possible effects of the unknowns on the outcomes of interest (Tetlock 2006, p. 217). In the years ahead, studies will be released that have more data than I have, more complex models built on those data, and new ideas for sifting through the old data. New studies will undoubtedly teach us about the consequences of the ACA and about economic theory itself, but we may never know how many of their successes rely on hindsight and the freedom from future events that might contradict their conclusions.

Forecasting is frequently attacked as imprudent or immodest. It is well known that price changes on the stock market are difficult, if not impossible, to predict (Malkiel 1973). This result is sometimes taken a step further to suggest that changes in gross domestic product and other aggregate indicators of economic performance cannot be predicted (Campbell and Mankiw’s 1987 results might be interpreted this way) and that it is foolish to try. Yet even the harshest critics of expert predictions of policy consequences admit some economic policy hypotheses are reliable: for example, that communist planning delivers terribly low living standards compared to market alternatives (Huemer 2012, p. 16). What, then, should be said about a law that takes the economy 80 percent of the way to communism? or 20 percent? or 2 percent? The real issue, I think, is the size of a policy impulse relative to the many other changes that would be occurring in the economy even without the policy—the smaller and more uncertain is the policy impulse relative to the other changes, the more difficult it is to predict how the economy will be different after the law is in effect than it was before—which is why the first half of this book deals exclusively with measuring the size of the new incentives contained in the ACA. The size is great enough that I would be shirking my responsibility if I concluded that anything is possible.

Short-term forecasts can readily consider some of the forces that are unrelated to the ACA, and thereby cut down on forecast errors due to the omission of non-ACA factors. Population growth and inflation are two of them, which I handle by offering predictions for hours worked per person, employees per person, and inflation-adjusted national income per person. I consider the aging of the population by making age adjustments to the hours and employment data. I also assume that technological progress or something like it tends to put inflation-adjusted national income per person on an upward trend, and that the ACA and other factors cause deviations around the trend. Also important are non-ACA tax changes that are coincident with the ACA, especially the December 2013 expiration of extended unemployment benefits, which are considered in the same way the ACA tax changes are. These strategies are needed for predictions relating to a single economic variable, such as national income, but other predictions just refer to relationships among variables: the ACA’s impact on weekly employment rates is greater than its impact on the average weekly hours worked by those employed, the ACA reduces some measures of productivity more than others, etc. For the “relationships” predictions, inflation, population growth, technological advancements, and other factors may hardly matter, in which case I do not need estimates of their magnitude.

The purpose of the predictions in the second half of this book is not only to learn about the ACA, but also to learn about economic theory and policy analysis themselves, so that analysis of future policies can be improved. In this regard, the ultimately false predictions are just as important as the “correct” ones. This book lays out clear criteria for readers in the future, who will have the advantage of additional data, to use to determine which of the predictions turned out to be incorrect in the sense that reliable measures of the relevant economic outcomes will fall outside of the predicted range. In some scenarios, predictions fail because of unanticipated events unrelated to the health reform, such as a new and costly war. Or the ACA may be significantly amended without ever being executed in its current form, and without being executed in an alternative form that is examined in this book. But the more interesting failures will be economic effects of the law that are not adequately treated here, or improper quantification of the economic effects that are treated. For the reasons cited above, these lessons for economic theory and policy analysis would be harder to obtain without being able to look back at a detailed and comprehensive policy analysis that did not rely on hindsight.

Chapter 6 begins to formulate the predictions by adapting a model from the labor economics literature to offer predictions for the magnitude, direction, and incidence of the long-run impacts on weekly work schedules of the ACA’s various taxes. Here the focus is on the impact of the ACA, which is a first step deliberately ignoring population aging, technological progress, and other non-ACA factors that would affect the economy regardless of whether the law had passed (those are added in later chapters). The model features, among other things, a couple of sharp incentives for employers and employees to limit weekly work hours to twenty-nine or fewer. Nevertheless, it is possible that the ACA may increase, or hardly change, the average number of hours worked per employee, although the law also tempts employers and employees to change their methods for reporting hours and incomes even if they do not change their actual work habits.

A more visible impact of the law will be to reduce the fraction of the population that is employed during a given week. Although the model in Chapter 6 accounts for many occupations and types of taxes, its quantitative conclusion for aggregate work hours can be closely approximated by mere multiplication of the total amount of taxation (documented in Chapters 3 through 5) and a “reward coefficient” that summarizes the sensitivity of labor market behavior to tax rates. This part of Chapter 6 contains the book’s clearest demonstration that, without my findings as to the size and scope of the new taxes, my conclusions about the ACA’s aggregate employment and hours effects would hardly be different from others in the literature. The literature uniformly fails to mention or quantify the large new implicit employee taxes (my Chapter 4), fails to adequately quantify the new implicit income taxes (my Chapter 5), and usually fails to consider the special tax treatment and anticompetitive effects of the employer penalty (my Chapter 3).

C. Adam Smith’s Theory of Equalizing Differences

Because the ACA does not tax all workers and sectors uniformly, its effects extend well beyond the employment rate and the average length of the workweek. Chapters 7 and 8 quantify and interpret the ACA’s deviations from uniform taxation from the perspective of the theory of equalizing differences. The theory is a workhorse in labor economics today and dates back to the beginning of economics, when Adam Smith (1776/1904, chapter I.10.1) explained how the free flow of workers among occupations and sectors can spread “disadvantages” located in one part of the economy—the ACA’s employer penalty is an obvious disadvantage of that sort—across the entire economy. Indeed, Smith’s theory is the foundation of the health-economics concept of a “value of a statistical life,” which is a critical part of cost-benefit analysis in the healthcare field (Viscusi and Aldy 2003). The two chapters begin with an illustration of the profound effects of equalizing differences on everyday life, and then they obtain quantitative results with basic arithmetic (adding and multiplying pairs of numbers) that represents the theory as applied to the ACA. Chapter 7 focuses on insurance coverage and relative wages, whereas Chapter 8 examines productivity (that is, the aggregate value of production per hour worked and related measures) building on recent research on “misallocations” of labor across sectors, regions, and firms.

Chapter 7 concludes that the ACA is likely to achieve the goal of significantly expanding the fraction of the population with health insurance, and it may do so to a surprisingly large extent, because it puts so many people in a position of having to pay more to be uninsured than insured. Workers having no direct contact with the ACA’s penalties and subsidies will nonetheless experience the law’s provisions, often in the form of lower wages, because they compete and produce with workers who do. Chapter 8 shows that the ACA will reduce the demand for both low- and high-skill labor because it reduces productivity and imposes penalties. The results in Chapters 7 and 8 are sharply different from those obtained by other authors on the basis of “health insurance simulation models” because the simulation models fail to consider the theory of equalizing differences or otherwise put the employer-employee relationship in a market context.

Chapters 6 through 8 show why there is a paradox of affordability. The ACA makes health care more affordable for segments of the population, but in doing so it makes health care less affordable for the nation as a whole. The ACA will have the nation working fewer hours, and working those hours less productively, so that its nonhealth spending will be twice diminished: once to pay for more health care and a second time because the economy is smaller and less productive.

D. Economic Dynamics and Conclusions

It’s one thing to estimate how much the ACA will eventually (say, within three or four years) depress employment rates, but it’s another thing to say how those effects will unfold from month to month. The dynamics depend on several additional and unknown (to me) factors, such as the rate at which employers and employees will learn about the new incentives they face and the rate at which they process updates to the rules made by regulators who are themselves updating their understanding of the law’s effects. Employees will form opinions as to the desirability of the new individual insurance plans, whose characteristics will change during the early years of the ACA. Chapter 9 therefore forms predictions for the labor market and the entire economy for a three- or four-year horizon following the beginning of the ACA’s exchange subsidies by adding my estimates of the impact of the ACA to estimates of other factors creating economic change. The chapter also discusses possible scenarios for marginal tax rates, especially changes to the ACA itself through legislation or possible announcements by the federal government as to its enforcement of ACA provisions.

The federal government and other advocates of the Affordable Care Act have dismissed concerns that the coming labor market contraction would be significant, or even noticeable, by pointing to Massachusetts’ recent experience with a reform also designed to expand insurance coverage. Chapter 10 looks at the Massachusetts experience and comes to some surprising conclusions.

Chapter 11 summarizes the main tax findings, impact estimates, and economic projections. It also suggests areas for future research, perhaps the foremost of which would be to quantify the ACA’s incentives for human capital accumulation, including schooling, job training, health capital, and occupational choice. The end of the book also presents a list of the acronyms.

  1. Keynes (1919, p. 23) writes that “although the school of thought from which [the Treaty] springs is aware of the economic factor, it overlooks, nevertheless, the deeper economic tendencies which are to govern the future.” Keynes was also expecting larger economic effects from French Prime Minister Clemenceau’s Treaty than I am expecting from the ACA, championed by U.S. Senator Kennedy, President Obama, and their followers.

(c) copyright 2014 by JMJ Economics