Prisoners of Their Own Device

This article appears in the April 2023 issue of The American Prospect magazine. Subscribe here.

Phil Swagel had good news and bad news for Bernie Sanders.

It was the spring of 2022, and the director of the Congressional Budget Office (CBO) was testifying virtually while recovering from a case of COVID-19. His audience was the Senate Budget Committee, where Sanders served at the time as chair. Yet, as Swagel was the last of the committee’s witnesses that day, that audience had been reduced to Sanders alone. The topic of the testimony was the Vermont democratic socialist’s signature policy proposal: Medicare for All.

The good news, according to Swagel, was that his office had determined that a single-payer health care system would vastly expand the availability and quality of coverage in the United States, while potentially reducing national health expenditures (the total amount that the country and its citizens pay for health care).

The bad news: Enacting single payer, Swagel said, would reduce gross domestic product, the key measure of economic growth, between 1 and 10 percent over the next ten years.

Sanders could have taken the opportunity to grill Swagel on these projections. Instead, he accentuated the positives. “Are we in agreement,” the chairman asked, “that if everybody in this country were guaranteed comprehensive health care, people would be freer to leave their jobs and do what they really wanted to do?” Swagel agreed. “That is right,” he said, “and that would boost productivity and entrepreneurship.”

Once computed, the “hard numbers” found in CBO’s baseline tables conceal all the assumptions and uncertainties involved in producing them.

Sanders’s Republican colleagues, by contrast, pounced. “Democrats’ math does not add up,” read the press release published by the office of Budget Committee ranking member Lindsey Graham. “When you examine CBO’s analyses on both the creation of a single-payer system and how it would be financed … the economic outlook is bleak.” And so, the endless “war of the estimates” rolled ever onward.

MAKING POLICY INVARIABLY MEANS trying to predict the future. Every time an idea is proposed, its champions engage in a kind of forecasting, highlighting how their legislation will usher in a glorious new reality. This bill will protect seniors from the ravages of poverty. That one will reduce supply chain bottlenecks, lowering costs for consumers.

Some of these forecasts are relatively easy to make. Tax cuts will reduce government revenues (though don’t tell that to conservatives). Lowering the Medicare eligibility age will increase program enrollment.

Others require answering far harder questions about the relationship between government and the economy. How will federal laws affect the number of hours people work, or the amount of money they spend and save? What about firms’ investments in new technology? Or carbon emissions? Or how many children will be born in a given year, as CBO was recently tasked with estimating?

To answer these questions, policymakers employ experts to construct models of the world, each of which relies on a long chain of assumptions about how governments, people, and firms will interact with one another. Practically speaking, as the British statistician George Box once noted, “all models are wrong, but some are useful.” Even an inaccurate employment forecast might nevertheless provide policymakers some perspective or trend line on how the future labor market might look under different conditions.

A model might cease to be a benign source of information, however, if it systematically ignores important dimensions of public problems. For example, economic models produced by the CBO assume, in the face of significant empirical evidence, that federal investments deliver half the rate of return of private investments. Equally important, the budget office admits that it has no basis for estimating the effects of climate change mitigation efforts on future economic realities. Taken together, progressive economists argue, these assumptions stack the deck against public solutions to major problems.

Importantly, CBO’s economic forecasts are not merely one set of information among many that members of Congress consider. Rather, they shape policy in profound ways, regardless of their accuracy or utility. They become a taken-for-granted component of the budget baseline, a benchmark projection of federal spending and revenues that sets the terms for every congressional debate over fiscal policy. Once computed, the “hard numbers” found in CBO’s baseline tables conceal all the assumptions and uncertainties involved in producing them.

The root of the problem here is not necessarily CBO, however, or even the models its economists build. After all, the CBO is a legislative support agency, created, funded, and directed by Congress. Congress also determines the way that CBO’s numbers are used. Yet while Republicans in Congress have heaped scorn on the CBO when its projections threaten their policy ambitions, and attempted overtly to change the way those projections are computed, the same cannot be said for Democrats. Even progressive congressional leaders—who know well how CBO scores can affect the prospects of legislative success—have largely refrained from calling out questionable assumptions found in its models. Why?

HOWEVER ESSENTIAL TO THE CRAFT OF LEGISLATION, projecting policy outcomes is a notoriously fraught human enterprise. As the authors of a leading graduate-school textbook on the subject bluntly assert, “even veteran policy analysts do not do it very well” and many “often duck it entirely, disguising their omission by a variety of subterfuges.”

To see why, let’s imagine a hypothetical legislative proposal to invest $50 billion in federal grants-in-aid per year over the next ten years—all of which would be financed through federal borrowing. Roughly half of this investment would go to fixing the country’s crumbling transportation infrastructure. The other half would be divided almost evenly between investments in education and R&D.

Given the poor—often perilous—conditions of American highways and railroads, one might assume that this kind of investment would be a boon to the economy. Yet, according to the CBO, this is not the case. At best, the policy would have a negligible positive impact on gross domestic product. To see why, we can consult CBO’s 2016 report on “The Macroeconomic and Budgetary Effects of Federal Investment,” which argues that the average short-run rate of return for federal investment is roughly 5 percent; this is one-half as large as the return on investment in the private sector.

CBO makes this claim for two reasons. First, the average productivity of investments by federal, state, and local government, the report asserts, is 8 percent—three-fourths as high as the average productivity of private investments. Second, each dollar of investment by the federal government increases total public investment by only two-thirds of a dollar. That is because federal spending generally leads to a decrease in state and local spending. The rest, we are told, is simple multiplication: “Two-thirds times three-fourths equals one-half.”

It all sounds simple enough. Yet scratch the surface even slightly, says Rutgers University economist Mark Paul, and CBO’s assumptions turn out to be “very wrong.”

One fundamental flaw, Paul suggests, is CBO’s claim that a dollar of public-sector investment is far less productive than an equivalent level of investment in the private sector. “In fact,” he points out, “the evidence shows that public investment often has a higher economic rate of return than private investment.” The evidence in question is not a cherry-picked set of outliers. It is a widely cited survey of 68 studies published between 1983 and 2008. That review, published by economists Pedro Bom and Jenny Ligthart in the Journal of Economic Surveys, finds that the average rate of return for public investment is 10 percent in the short run and 16 percent in the long run. That is double the rate the CBO model employs. When Bom and Ligthart narrow the inquiry to examine only “core” public investments—in roads, highways, airports, and utilities—they find that the rate of return is higher still.

To its credit, CBO does mention Bom and Ligthart’s findings in its 2016 report. It dismisses their overall point, however, because “studies examining relatively recent periods” found that the output effects of government spending had “declined over time” and might be far lower than their average estimate. Yet even if returns from government investment have decreased somewhat over time, recent studies—including studies restricted to the U.S. context—still suggest they are far higher than CBO’s baseline. Still, the report concludes that “in CBO’s judgment” its assumption is consistent with the “range of estimates found in the research literature.”

This is not the only questionable assumption found in CBO’s models. As the Roosevelt Institute’s Emily DiVito and Mike Konczal point out, the models routinely assume that deficit-financed federal investment invariably raises interest rates and results in negative economic growth, “independent of the economic trends we’ve seen in the past 20 years.”

We have an actual real-world study to consult here: the Biden administration’s investments in advanced manufacturing. Since the White House declared its intent to onshore critical components like semiconductors, and incentives were approved by Congress, electronics manufacturing has soared. It’s an example of public investment not crowding out private spending, as CBO often assumes, but crowding it in.

SUCH ARGUMENTS MIGHT NOT PASS MUSTER in an academic journal, but then again, CBO is not really an academic institution. While the office’s estimates may stimulate intellectual debate, that is not their central purpose. They exist as a benchmark Congress can use to compare competing proposals to one another, using the same models and rules. Accuracy is explicitly sacrificed for consistency.

It is often impossible, however, to fashion those rules in a neutral manner. Despite the powerful effect of CBO cost estimates on the chances of passing legislation, the agency’s work remains focused primarily on analyzing the costs of major policy reforms, not their benefits. Yet even where costs are concerned, the office’s lack of subject-matter expertise imperils its usefulness for members of Congress.

The budget office admits, for example, that it “has no basis for estimating” the future savings of climate change mitigation policies “because many of the linkages between climate change and the federal budget require further assessment.” Lacking such information yields conclusions that border on the cruelly absurd. A 2021 CBO report suggests that climate change might amount to a savings for federal programs like Social Security, Medicare, Medicaid, and Supplemental Security Income. The costs of those programs will decline, the report concludes, “to the extent that participants die at younger ages.”

Another target for criticism has been the guidelines CBO relies on to score legislation. For example, CBO does not treat the effects of changes in IRS tax enforcement as a source of “revenue” when analyzing the costs of legislation, even though common sense (and real-world observation) would dictate that investing more in enforcement would uncover more tax cheats.

The Biden administration’s proposed Build Back Better Act increased funding for tax enforcement, ramping up audits of corporations and wealthy individuals who benefited from the gradual defunding of the IRS in the previous decades. CBO estimated that this would bring in more than $100 billion over the next decade. Yet in the midst of the debate over the legislation, CBO announced that it would not count the change as a cost savings, even though it would be reflected in the office’s baseline budget projections once the legislation was enacted.

As Sen. Chuck Grassley put it in 2006, “CBO is God around here.”

Yet another limiting principle is the CBO’s adherence to a strict ten-year window for evaluating the benefits of federal investments. To be sure, longer-range forecasts come with far greater levels of uncertainty, greater at any rate than CBO economists are willing to tolerate. Nevertheless, when policy proposals make significant changes to society—such as climate change mitigation measures—their full impact may not be realized for several decades, rendering any benefits invisible in the ten-year window of analysis.

While CBO’s assumptions are often “wildly out of line with empirical studies,” Paul says, they are still very much in line with what gets taught in graduate macroeconomics courses, which have not yet caught up to the literature. In the midst of this disciplinary chaos, conventional wisdom dies hard—especially inside CBO.

Economists, Paul suggests, need to admit that “we’re not great at forecasting the economic effects of policies.” While he acknowledges that CBO has improved its efforts at transparency in recent years, the budget office “has to do a better job at educating people on the assumptions its models use” and how the models contain a “lot more uncertainty” than CBO reports portray.

TO MODEL THE EFFECTS OF SINGLE-PAYER REFORM, CBO economists first had to craft multiple stylized versions of the policy, mixing and matching different payment rates, levels of cost sharing, and benefit packages. Using the budget office’s “life cycle” growth model, they then simulated how these policy choices would affect a dizzying array of interactions between households, firms, and the government. Adding to the complexity, the models attempted to consider how policy would affect population health and, in turn, life expectancy. With each added layer of complexity comes more choices about assumptions and hence greater potential for error. The more groundbreaking the legislation, the harder it will be to model with any level of certainty.

You can see the inherent difficulty associated with forecasting the long-term daily economic decisions of hundreds of millions of not-entirely-rational humans in CBO’s baseline budget projections. For the entirety of the 2010s, CBO projected long-term interest rates significantly higher than reality, assuming a return to an equilibrium that never happened. This put many policies out of reach, because of the expectation that deficit spending would cost more than it would have. The same problems pop up elsewhere; CBO has an entire section of analysis on its website called “Accuracy of Projections” assessing its failings. This doesn’t stop news outlets from promoting the latest CBO budget analysis with rock-solid sureness.

Therein lies the problem: The assumptions behind CBO models are rarely contested before the public eye. One reason for this is the process by which the budget office’s research is transformed from working papers—replete with caveats, limitations, and acknowledgments of uncertainty—into the individual point estimates that become the official “scores” for major bills and the headlines that typically advertise the legislative “price tag.” The lawmaking process has little tolerance for footnotes or ranges of possible estimates.

Moreover, members of Congress who consume CBO reports, and under current rules must live under their dictates of how much a policy costs and how much revenue they would have to find to cover it, have few incentives to look “beneath the hood” of the estimates to evaluate their assumptions. A colloquy about output elasticities hardly makes for good political theater. When committee chairmen are given a “mixed bag” report from CBO, they are naturally inclined to salvage the good, polish up the bad, and leave the ugly as a consolation prize for the ranking member.

If what the CBO does is a blend of art and science, members tend to describe the budget office in terms that frequently bend toward the religious. “CBO is God around here,” as Sen. Chuck Grassley (R-IA) put it in 2006, “because policy lives and dies by CBO’s word.” The history of health reform, Sen. Ron Wyden (D-OR) agreed, “is congressmen sending health legislation off to the Congressional Budget Office to die.”

Few institutions in American political life are described with such alpha-and-omega portent. Little wonder that the event that seems to have prompted then-congressional candidate Greg Gianforte’s 2017 body-slamming of a reporter was that reporter asking him about the latest CBO report.

But the CBO is just an organization of 275 staff members, many of whom have Ph.D.s, and most of whom spend their time projecting how much federal programs cost. It took members of Congress—together with the Beltway press—to make it into a god.

Following impoundment controversies in the Nixon administration in 1974, congressional policy entrepreneurs attempted to design the CBO as an agency that would counterbalance the power of the president in the budgetary process. It would be an alternative source of information that would be loyal to Congress, checking the proclivities of the Office of Management and Budget as well as other executive branch organizations. In that role, the CBO frequently came in for criticism, specifically when it called out the Reagan administration’s argument that a massive tax cut would lead to a budget surplus.

During that first decade, few members of Congress would have called the CBO a “god” on Capitol Hill. Instead, they frequently slashed the office’s budget, creating half a dozen “near-death experiences,” in the words of former director Robert Reischauer. Former Dixiecrat Sen. Russell Long (D-LA) was so incensed about CBO’s projections on tax cuts that he vowed to “find somebody who knows more how to put the answer in the computer so that it comes out the right way.”

But the CBO would soon win an important set of champions. First, the House and Senate Budget Committees—which relied on the office for leverage over congressional appropriators—defended it from its attackers in the Reagan administration and, later, protected it from Gingrich-led efforts to bleed Congress of any independent legislative support. During the 1980s, increasingly conservative Democrats came to value the CBO as a tool for critiquing the Reagan administration’s budget-busting policies. The 1984 Democratic platform mentions the deficit over 40 times.

The strongest defense of the budget office came from an increasing elite preoccupation with the federal budget deficit. While the CBO continued to face cuts throughout the 1980s, mounting deficit pressures gave the office a window to influence the policy process in a new way. As former director Rudy Penner recalled, while the CBO could not take policy positions, he felt it was “safe for me to be against deficits.”­ In interviews with journalists and public speeches, Penner frequently warned about fiscal irresponsibility, which netted the office “a lot of credibility.”

By the early 1990s, congressional efforts to set automatic deficit reduction targets made CBO scores an “obligatory passage point” for legislation, the central standard by which new policy ideas are adjudicated. Following the enactment of “pay-as-you-go” (PAYGO) requirements in the early 1990s, members of Congress began to seek out “under the table guesstimates” of program costs from CBO before legislation was completed. As Robert Reischauer recalls, PAYGO had an almost “psychological” effect on members of Congress. Republicans instrumentally used the CBO’s scores to attack Bill Clinton’s national health reform proposal. Democrats attacked Republican tax cuts by citing projections by the CBO of their deficit effects. Members would soon begin shelving or heavily revising costly proposals so as not to “screw up the PAYGO scorecard and piss off their colleagues.”

As the deficit obsession increased, the CBO also became a go-to source of policy information for journalists. My research suggests that citations to the budget office in The New York Times are highly correlated with increases in the federal deficit, and the largest share of op-eds mentioning the CBO in The Washington Post between 1975 and 2018 are focused on deficit effects.

THE CBO’S TRANSFORMATION INTO A DEFICIT ORACLE did not mean congressional leaders always passively accepted scores or reports that they disagreed with. As former CBO director Douglas Holtz-Eakin—who later founded the conservative American Action Forum—recalls, “everyone makes mistakes.” His was rushing out an “undisguised free trade manifesto” on CBO letterhead before the office’s editorial staff could ensure that it was “balanced in the usual CBO fashion.” Fifteen minutes after the release of the letter, he found himself in the office of Sen. Robert Byrd (D-WV), who asked him if he’d like to “run CBO on a dollar a year.”

Holtz-Eakin’s predecessor, Dan Crippen, found out the hard way what happens to a CBO director who stands up to concerted congressional opposition. A former domestic-policy adviser in the Reagan administration, Crippen was appointed by Republicans to direct the CBO in early 1999. Yet within several years, he became embroiled in a long-simmering controversy over so-called “dynamic scoring”—a method of analyzing the macroeconomic feedback effects of legislation.

Since the early 1990s, congressional Republicans had insisted that economists at the CBO should apply the technique to analyses of major tax-cut legislation, which they believed would give them proof for their long-sought belief that tax cuts pay for themselves. After all, budget analyses produced by the President’s Council of Economic Advisers—Democratic and Republican alike—employed this technique. Historically, however, the House of Representatives had forbidden the CBO from using dynamic scoring. CBO leaders were also averse to using a technique that was likely to stir up partisan controversies and which they personally believed was—in the words of Robert Reischauer— “ideology, not economics.”

With the appointment of Crippen, Republican leaders thought that they might finally get movement on dynamic scoring. Soon after his appointment, rumors began to surface that Crippen had abandoned the apolitical pretenses of prior directors and was prepared to take a more explicit partisan stand on proposals to reform Medicare and Social Security. Yet when it came to dynamic scoring, Crippen did not budge, much to the chagrin of House Budget Committee chair Jim Nussle. During a particularly tense 2002 closed-door meeting on the Republican budget proposal, Nussle bluntly announced: “CBO sucks, and you can quote me on that.”

Holtz-Eakin, appointed by Republicans at the end of Crippen’s term in 2003, saw things differently from his predecessor. Dynamic scoring was hardly voodoo, as Democrats seemed to believe. Rather, it was just another set of assumptions about how government policy affected the economy. To the former Syracuse University professor, it made a great deal of sense for the CBO to check its own suppositions and admit uncertainty. The CBO should, he thought, help Congress “to look at the world without the policy, look at the world with the policy, and compare all the differences.”

The bigger an effect a lawmaker wants to have on society, the harder CBO makes it to accomplish.

Soon after his arrival, Holtz-Eakin convened a meeting of his staff to inform them of his decision. “We’re going to do dynamic scoring,” he told them. When this invited a litany of objections, he brought the hammer down. “Let me explain this to you,” he said, “if I have to write in every number personally, I will. Or you can do the work. Which way is this going to happen?”

Still, Holtz-Eakin did not believe that dynamic scoring was the “magic elixir” Republicans thought it would be. “They’re not numerate in the Congress, uniformly,” he told an interviewer in 2011, “so my solution to this was they want the forbidden fruit, have an apple and beware.” In any case, his experience told him that it wouldn’t make much of a difference. Yet when CBO’s dynamic scores produced what Holtz-Eakin calls a “big nothing,” Republicans hardly snapped out of it. “What they concluded, instead, was I did it wrong.”

When the House of Representatives is under Republican control, its rules now routinely require the CBO, as well as the Joint Committee on Taxation (JCT), to provide a dynamic score of major tax changes “to the extent practicable.” In theory, this creates a double standard for how major legislation is evaluated. Dynamic-scoring requirements, after all, do not extend to spending bills. In practice, given the extensive costs and time required to produce valid dynamic scores, the CBO has sometimes found it difficult to provide them, especially when the legislative process moves quickly.

Yet the idea of dynamic scoring is arguably of greater utility to Republicans than the scores themselves. In 2017, for example, the CBO provided only a static score of Republicans’ signature Tax Cuts and Jobs Act, because it was “not practicable for a macroeconomic analysis to incorporate the full effects of all of the provisions in the bill, including interactions between these provisions, within the very short time available between the completion of the bill and the filing of the committee report.”

A dynamic analysis released by the JCT two weeks later also revealed that the tax bill would result in significant deficit increases. In response, congressional Republicans circulated a set of talking points attacking the “substance, timing, and growth assumptions of JCT’s ‘dynamic’ score,” and highlighting prediction errors in the CBO’s prior analyses. According to Republicans, the JCT’s dynamic score was insufficiently dynamic, because of its assumptions about how consumers and workers would respond to lower levels of taxation and its assumptions about the pace at which the Federal Reserve would raise interest rates. Doubts about these scores, if anything, enabled swift legislative action. In the final hours before the bill’s passage, Senate Majority Leader Mitch McConnell announced that he was “totally confident that this is a revenue-neutral bill.”

Incidentally, CBO’s eventual score for the Tax Cuts and Jobs Act, with a ten-year cost of $1.5 trillion, ended up being too conservative. Four months later, the budget office revised its update; it would now cost $1.9 trillion.

FOR ALL THE GODLIKE PROPERTIES ascribed to the legislative scorekeepers, Republicans have realized that gods that fail can be ignored. Holtz-Eakin seems to agree: “The CBO can’t stop anything that Congress really wants to do. You can tell them anything and they’ll just go ahead.”

But if scorekeeping is not a binding constraint on policy, Democrats seem not to have received the memo. It is virtually impossible to imagine Chuck Schumer dismissing a CBO score out of hand—even on “priority” legislation. Nor are Democrats as aggressive as Republicans when CBO scores spell trouble for their preferred bills. When the CBO’s analysis revealed that 2017 legislation to “repeal and replace” the Affordable Care Act would cause millions to lose health insurance coverage, Republicans hauled the CBO director before the agency for a series of grueling oversight hearings. Two members of the House Freedom Caucus introduced amendments—ultimately unsuccessful—to gut the office’s budget.

One reason for this is that, during the Obama years, Democratic leaders were arguably more enamored with deficit reduction—or at least the appearance of deficit reduction—than their Republican peers. As Robert Saldin highlights in his book When Bad Policy Makes Good Politics, lawmakers obsessed with generating projected “savings” for the Affordable Care Act — making good on President Obama’s commitment to not pass a plan that “adds a dime to the federal deficit” — pursued an ill-fated proposal for a voluntary long-term services and supports program called the CLASS Act.

To anyone who knows anything about how risk pools work, the phrase “voluntary long-term care insurance” will sound titanically stupid. And it is. But in the land of pay-fors, it was a golden opportunity. As designed, the program’s long vesting period meant that CLASS would generate nothing but increased revenue in the first half of CBO’s ten-year legislative scoring window—helping Democrats to make good on their misguided promise of deficit-neutral health reform. But because too few healthy people would sign up for the plans, CLASS would ultimately end up generating an insurance “death spiral” of increased costs and declining participation. Congress repealed the program within a few years of its creation. (And paying for long-term care is still a disaster for most families.)

This is not an isolated case. In the first few months of the COVID-19 pandemic, House Speaker Nancy Pelosi explicitly rejected including automatic stabilizer provisions in relief legislation—which would have allowed the federal government to tie benefit levels to the duration of the crisis rather than arbitrary cutoff dates. Her reasoning? The CBO’s scorekeeping rules for automatic stabilizers would have inflated the total price tag for the legislation beyond Pelosi’s proposed $3 trillion ceiling. “If we put every good idea people wanted in the bill, it would be an $8 trillion bill,” as one staffer put it.

And Democratic leaders’ obsession with a “good score” is reportedly one reason why their 2021 plan to expand Medicare dental, vision, and hearing benefits delayed the phase-in of benefits for eight years. Consider that Medicare Parts A and B were initially implemented in 11 months. In 1972, Congress required the new end-stage renal disease benefits to take hold in one year. Even Medicare Part D and the Children’s Health Insurance Program took less than three years to implement. The Affordable Care Act took four.

Progressives are hardly unaware of how CBO scores—which emphasize the costs, though not the benefits, of major legislation—imperil their chances at reform. The last decade has seen the introduction of a handful of legislative proposals that require the CBO to perform new sorts of policy analysis. The Poverty Impact Trigger Act, introduced by Rep. Barbara Lee (D-CA), requires the CBO to forecast the effects of major legislation on poverty and establishes a Poverty Impact Division of the budget office. Prior to chairing the Senate Budget Committee, Bernie Sanders introduced legislation requiring the CBO to estimate the effects of legislation on carbon emissions.

Still, progressive efforts to reform the CBO have hardly made it past the drafting stage.

Nor have progressives made it a priority to publicly investigate the assumptions underlying CBO’s models. As ranking member on the Senate Budget Committee, Sanders focused his attention on attacking the Trump administration’s budget. When he became the chair of that committee in 2021, Sanders held hearings on the need to expand Social Security and to enact Medicare for All, as well as on the costs of inaction on the climate crisis.

There is a clear political logic here. Oversight hearings rarely persuade the CBO to alter its methods or assumptions. So why waste valuable public attention on the arcane mechanics of macroeconomic models that might be better spent at growing support for a popular, progressive agenda? On the other hand, this means that dubious assumptions about the value of public investment remain submerged in the back of obscure government reports, or in closed-door meetings without public pressure. This only perpetuates CBO’s oracular mystique.

That mystique comes at a cost. Consider the CBO’s February 2021 finding that increasing the minimum wage to $15 an hour would cause 1.4 million Americans to lose their jobs. While major news sources treated the report as the official word on the matter, they ignored that the CBO decided to weight studies with larger job losses more heavily in its analysis, and chose not to examine a variety of high-quality studies on policy trade-offs included in a systematic review commissioned by the British Treasury. Among other things, CBO included an initial study of the Seattle minimum-wage experiment that found evidence of major job losses, but excluded the authors’ follow-up study, which produced more positive results.

FORECASTING ERRORS AND DEBATABLE ASSUMPTIONS would not be a problem if CBO’s scores were treated as one potentially useful source of information among others. Yet by transforming the office into an oracle, Congress has become a kind of supplicant. CBO scores are now all too easily used as an excuse for legislative inaction rather than a tool for decision-making. The limitations of that tool become more apparent when applied to proposals for more sweeping changes to the economy or society. The bigger an effect a lawmaker wants to have on society, the harder CBO makes it to accomplish.

However hard-won, the CBO’s reputation as an honest broker of hard truths in a polarized age—or, as one Washington Post editorial put it, a “skunk at the congressional picnic”—has had costs of its own. It is far easier to report the office’s point estimates as stylized facts rather than what they are: conditional, assumption-laden projections of the future. “The fact that CBO is just so solidly in Wonkville,” as economist Mark Paul puts it, “makes it harder to go after and also it looks more partisan going after it,” even when its assumptions are off base. “In reality,” Paul says, “how you crunch the numbers is a deeply political question.”

Still, as Republicans demonstrated during the passage of the Tax Cuts and Jobs Act, scorekeeping is only as powerful an obstacle to legislation as its audience allows it to be.

Members of Congress aren’t handcuffed to the current scorekeeping regime, or even to the results of a single commissioned study. These institutions are, at best, trick handcuffs: Congressional coalitions create them and can alter them when they want to. Or, more appropriately, when they receive intense, cross-cutting pressure to do so.

In other words, to build better models, we might need to start with building a better politics.