A cost estimate is the number the budget process runs on. A bill can command speeches, hearings, and a majority in search of a path to passage, and a single figure attached by an estimating office can decide whether that path exists. The figure tests a bill against committee allocations and against the totals in a budget resolution. The figure decides whether a point of order lies against consideration on the floor and whether a waiver must be assembled under pressure. The figure certifies whether a committee has satisfied a reconciliation instruction or has fallen short of a directed change in revenues or spending. In the Senate, the figure also maps a bill against the Byrd rule, which reaches past the budget window and penalizes a measure that increases deficits in the years beyond that window. Passage, procedure, and price therefore arrive together, bound in one document that opponents and supporters must both use.

The measuring rod behind the figure carries the central claim of the account that follows. The baseline is written in statute, not chosen. Congress enacted the conventions that govern the projection against which every bill is measured, specifying how expiring programs, expiring tax provisions, and annual appropriations are treated before any estimator examines any particular proposal. A familiar complaint in budget debate holds that the assumptions behind an estimate are unrealistic, that a tax provision certain of extension ought to be treated as permanent, or that a large program facing a scheduled end ought to be treated as finished. Each complaint sounds like an objection to an analyst and lands, on inspection, on a convention Congress enacted and Congress alone can amend. An estimator who abandoned the statutory background in favor of a more plausible background would replace a public rule with a private judgment, and every score built on that substitution would lose the comparability that a fixed rod supplies.

That distinction frames the questions the article answers. Who builds the measuring rod, and under what appointment, duty, and order of service does the builder operate. Who scores which title of a bill, and why a tax title and a spending title carry different authorship inside a single total. Why the count stops at ten years, and what an estimator says, in weaker and more qualified language, about the decades the count excludes. How an estimate is constructed from statutory text, agency information, comparable programs, and behavioral responses, and how drafters answer the construction through dates, phase ins, expirations, and shifts in payment timing. What dynamic analysis adds to a conventional estimate, under what House rule that addition entered official figures between 2015 and 2018, and where informed argument divided over completeness, governance, and design. How accurate the record proves when the estimating office grades earlier projections against outcomes, separating forecast error from legislative change and from behavioral surprise.

Read as a sequence, those questions move from institution to arithmetic to judgment. The office comes first, because a score deserves trust only when the charter, the appointment standard, and the professional staff behind the score are understood. The baseline and the division of labour come next, because attribution and background decide what a total means before any dispute over magnitude begins. The window, the construction, and the dynamic contest follow, because each shapes the total in a direction a headline conceals. The accuracy record closes the sequence, because confidence in a number ought to rest on evidence about earlier numbers rather than on agreement with a preferred result. A reader who holds that sequence in mind reads an estimate as an auditor reads a ledger, tracing each figure to a source, a convention, a frame, and a stated limit.

The Office Congress Built to Price Its Own Bills

A major bill in Congress carries a price, and the price travels with the bill through every stage that follows committee action. Before a committee reports a measure that changes taxes, entitlement spending, or other federal outlays, an estimate must attach a figure to the proposal, expressed as a change in revenues, spending, and the deficit against a common measuring rod. The figure does far more than inform debate in a general sense. It determines whether the bill fits inside the totals and allocations that the budget resolution has set, whether a point of order lies against consideration on the floor, and whether supporters must find offsets elsewhere or accept a recorded increase in the deficit. A proposal described by its sponsors as modest can become procedurally impossible once the estimate records a large cost over the budget window, and a proposal attacked as extravagant can gain room to move once the estimate records savings or revenues that opponents had discounted. The number, in that sense, governs what the rules allow, and control over the number amounts to control over a gate in the legislative process itself.

The estimating function also disciplines the language of legislation. Drafting choices that read as minor in statutory text, a start date moved by a year, a benefit formula adjusted by a fraction, an eligibility rule widened to a new class of recipients, acquire a concrete fiscal meaning once an estimator translates text into projected cash flows. Committees therefore draft with the estimate in prospect, testing language against the likely score and revising language when the projected cost exceeds the allocation available. A debate conducted without an agreed figure leaves each side free to assert a different cost for the same text, and enforcement collapses into competing assertions. A debate conducted with a single figure supplied by a single office forces disagreement onto the assumptions and the policy, where a presiding officer, a budget committee, and a recorded vote can act on it.

Why did Congress create its own budget office?

Congress created the office in 1974 because budget decisions required an estimator that Congress controlled and trusted. Reliance on executive branch figures placed legislative debate on presidential arithmetic. A nonpartisan office housed in the legislative branch supplied independent cost estimates and baseline projections under rules Congress itself had enacted by statute.

Before 1974, Congress possessed deep expertise in pieces of the budget and no institution charged with the budget as a whole. The appropriations committees controlled annual spending bills, the tax writing committees controlled revenues, and authorizing committees created programs whose costs extended far beyond a single fiscal year, while no committee owned the combined effect of those separate decisions on totals for revenue, spending, and borrowing. The executive branch, by contrast, planned as a whole. The President submitted an annual budget prepared through the Office of Management and Budget, and executive estimates framed the fiscal argument that Congress then contested line by line. When a congressional committee disputed an administration figure, the committee relied on agency testimony, outside analysis, or staff work assembled for that contest alone. The imbalance was structural rather than personal. The presidency argued from a unified set of books kept by its own budget office, and Congress answered from fragments.

The conflict over impoundment during the administration of President Richard Nixon exposed the cost of that imbalance in constitutional terms. Nixon asserted a power to refuse to spend funds that Congress had appropriated, and the dispute joined a broader argument about which branch set fiscal policy when the two branches disagreed. Congress responded with a statute that addressed procedure and information together. The Congressional Budget and Impoundment Control Act of 1974, enacted as Public Law 93-344, created a congressional budget process with its own timetable, its own committees, and its own source of numbers, while also restricting presidential impoundment. The statute treated arithmetic as a question of institutional power. A legislature that borrowed the executive branch estimate for every major decision argued on ground prepared by its rival, and a legislature that generated an independent estimate could test presidential claims, score legislative alternatives on an equal footing, and enforce limits of its own choosing.

The settlement of 1974 therefore placed three linked creations inside a single law, the framework described in the Congressional Budget Act of 1974. House and Senate Budget Committees received responsibility for a concurrent resolution on the budget that set totals and divided spending authority among the other committees. A timetable tied the resolution, the appropriations bills, and reconciliation to dates in the legislative year. The Congressional Budget Office received responsibility for the estimates and projections on which the resolution and its enforcement depended. A resolution without independent estimates rested on figures supplied by the administration whose budget Congress sought to amend or replace, and an estimating office without a budget process produced reports that no rule required anyone to use. Joined together, the committees set the limits, the office measured compliance with the limits, and the floor enforced the measurements through points of order sustained or waived by vote.

The new office began operating in 1975, and its first director was Alice Rivlin, an economist whose appointment announced the professional character that the statute demanded. The office opened inside the legislative branch, accountable to Congress and separate from the executive departments whose programs the office priced. Members of Congress who received an unwelcome estimate received the estimate from an institution that Congress had created for exactly that unwelcome task. The charter that Congress wrote in sections 201 and 402 of the 1974 act then specified, in unusual detail for an analytical agency, who would lead the office, how that leader would be chosen and removed, whom the office would serve first, and what the office owed the institution as a matter of duty rather than discretion.

Section 201(a) of the 1974 act, codified at 2 U.S.C. 601(a), vested appointment of the director jointly in two officers drawn from the two chambers, the Speaker of the House and the President pro tempore of the Senate. Joint appointment prevented either chamber from claiming the office as a House institution or a Senate institution, and bicameral ownership matched the bicameral character of the budget resolution itself, which required agreement between the chambers before the totals bound anyone. The appointment did not proceed from the two presiding officers alone. The statute directed the appointing officers to act after considering recommendations from the House Budget Committee and the Senate Budget Committee, the two committees whose work depended most directly on the estimates. Advice from the budget committees tied the choice to the committees with the largest stake in the credibility of the product, and the formal power remained with officers answerable to the full membership of each chamber.

The same paragraph imposed the condition that defined the office from its opening day. Selection was to be made without regard to political affiliation and solely on the basis of fitness to perform the duties of the office. The language excluded the ordinary currency of Washington appointments, in which party service, campaign loyalty, and ideological agreement supplied the usual tests of suitability. Fitness, as the statute framed the term, pointed toward professional competence in economics, budgeting, and administration, toward a record that members of either party could inspect and accept. The command operated as a promise made by Congress to a future Congress as well. A majority tempted to install a loyalist whose estimates favored the majority program faced a statutory standard that opponents could quote and that the director could invoke as protection after appointment. A standard written into law does not enforce itself, but a standard written into law gives every later defender of the office a text to stand on.

The term and removal provisions of section 201 balanced independence against accountability in a deliberate pairing. The director served a term of four years, a span longer than the two year life of a single House of Representatives, so a director appointed at the start of a Congress remained in place after an election had changed the membership and possibly the majority of one or both chambers. Continuity across elections protected institutional memory, preserved estimating methods from wholesale replacement after each change in party control, and signaled to the staff that professional standards outlasted the political cycle. At the same time, the statute preserved a congressional remedy. Either house could remove the director by resolution, a power exercisable by a single chamber acting alone, without presidential involvement and without the concurrence of the other chamber. Reappointment carried no statutory limit, so a director who retained the confidence of successive appointing officers could serve through several terms. The combination gave the director security against casual dismissal and gave Congress an unmistakable, if drastic, instrument against a director who lost the confidence of the institution.

Practice added an informal convention to the statutory mechanics. The two budget committees alternated in taking the lead in recommending directors, so the initiative in selecting the estimator passed back and forth between the House committee and the Senate committee across successive appointments. Alternation spread ownership of each appointment, reduced the risk that a single committee would treat the directorship as a preserve, and gave each new director a constituency in both chambers from the moment of selection. Formal law supplied the joint appointment, and repeated practice supplied a norm of shared custody that the text had implied rather than commanded.

The professional rule extended below the director through the staff of the office. Analysts were chosen on professional grounds, for training and competence in economics, statistics, health policy, tax analysis, and the other disciplines that the estimating workload required, rather than for party service or political connection. A career analyst corps gave the charter a daily reality that appointment language alone could not supply. Directors arrived and departed on four year terms, while the analysts who built the models, maintained the data series, and remembered the treatment adopted for a similar provision in an earlier bill remained in place across directorships. Consistency of method, a quality on which the credibility of the office depended, lived in that continuity.

Section 402 of the 1974 act imposed the operational duty at the center of the charter, a duty framed as an obligation owed to committees rather than as a service offered at the discretion of the director. The office was to provide cost estimates for bills reported by congressional committees, so the estimating function attached to the formal act of reporting, the point at which a committee presented a finished text to its chamber and asked the chamber to act. Estimates prepared under section 402 measured the change that the reported text would produce in federal spending and revenues, a formulation that made the baseline, rather than the total size of a program, the starting point of the inquiry. A bill that reauthorized an existing program at an existing level could carry a small estimate even though the program itself spent a large sum each year, and a bill that created a modest new benefit could carry a large estimate if the benefit grew quickly across the window. In the original design of the 1974 act, those estimates were framed over five years, a horizon long enough to capture phase in periods and early growth in a new program and short enough to rest on economic and demographic assumptions that analysts could defend with some precision. The frame lengthened in later decades as budget resolutions and enforcement rules lengthened the horizons over which Congress chose to bind itself, and the principle survived the change in length intact, because a window too short rewarded drafting that pushed costs just beyond the first fiscal year.

Section 202 of the 1974 act assigned the second great duty, the production of the baseline against which every individual estimate was measured. The office was required to publish annual baseline projections of federal spending, revenues, deficits, and the debt, together with the economic assumptions on which the projections rested, in the volume issued as The Budget and Economic Outlook. The baseline described the path of the budget under laws already enacted, projected forward on stated assumptions. Budget committees used the baseline to construct the annual resolution, since a resolution total acquired meaning only in comparison with the spending and revenue path that existing law already implied. Committees drafting new legislation used the baseline as the denominator of every estimate, since section 402 measured change and change requires a starting path. A baseline altered to flatter a pending bill would appear in a published volume read far beyond the committee rooms, which multiplied the cost of any manipulation.

An order of service governed the allocation of analytical labor inside the office, and the order followed the logic of the budget process itself. The budget committees stood first in line, because the resolution came first in the legislative year and because every other committee allocation descended from the resolution totals. The money committees stood next, the appropriations committees with their annual spending bills and the committees with jurisdiction over revenues and major entitlement programs, because estimates for those committees carried direct enforcement consequences under the resolution. Other committees and individual members received estimates, analyses, and technical assistance as the capacity of the office allowed after the claims of the budget and money committees had been met. Congress directed the scarce resource of trusted analysis toward the points in the process where enforcement attached, and Congress accepted delay elsewhere as the price of reliability at those points.

The design that the charter embodied can be stated as a pair of constraints working in opposite directions. From one direction, statute constrained the estimator. The duties were mandatory, the frame for estimates was set by law and by the enforcement rules built on the law, the baseline rested on a statutory duty, the director selection standard was written into section 201, and the order of service subordinated the preferences of any single member to the needs of the budget process. An estimator constrained in that fashion could not freely select a method because the method flattered a sponsor, could not withhold an estimate because the estimate embarrassed a committee chair, and could not redefine the baseline in the middle of a debate because a new definition improved the apparent position of a pending bill. From the other direction, structure insulated the estimator. Joint appointment by officers of two chambers, a selection standard that excluded political affiliation, a four year term crossing an election, removal only by formal resolution of a house, a professional staff, and a norm of alternating recommendation between the two budget committees combined to raise the cost of retaliation and to lower the benefit of attempting influence. Pressure remained possible, as the removal power demonstrated on the face of the statute, and pressure carried a public price that the charter deliberately made high.

Congress built the office in that fashion because the number had to be believed by both sides before the number could govern anything. Enforcement by point of order depends on losers accepting the measurement that defeated a proposal, and a minority accepts an adverse measurement from an estimator that the minority helped to select, under a statute that the minority can quote, far more readily than an adverse measurement supplied by the majority staff or by the administration in power. The majority gains from the same arrangement, since a favorable estimate from an independent office persuades undecided members, outside audiences, and the other chamber in a way that a favorable estimate from a partisan source cannot match. Credibility, in the economy of the charter, was not a decoration added to competence after the fact. Credibility was the asset that converted an analytical product into a governing instrument, and every element of sections 201, 202, and 402, from the fitness standard to the service queue, protected that asset as the condition of the whole enterprise. The office would price bills whose sponsors included the very committee leaders who recommended its director, would publish baselines that embarrassed administrations and congressional majorities in alternating seasons, and would attach unwelcome figures to measures carrying the hopes of presidents and the reputations of chairs. A weaker design, an estimator serving at the pleasure of a single officer or housed inside the executive branch whose budget Congress existed to revise, would have bent under that load or would have been dismissed as bent, which for enforcement purposes amounts to the same failure. The 1974 settlement answered the problem by giving the estimator a statutory home, a professional staff, a mandatory docket, and a leader chosen under a rule that denied party as a qualification.

The Baseline and the Conventions Written Into Statute

A baseline is a projection, built under rules fixed in law, of what the federal budget would look like if the laws already enacted were left to operate as enacted. It projects budget authority, the legal authority to enter into obligations that will result in payments from the Treasury. It projects outlays, the disbursements that flow from that authority when obligations are liquidated. It projects revenues, the receipts collected under the tax laws and other revenue laws in force. It projects the deficit or surplus that results from the difference between those outlays and those revenues. A cost estimate therefore does not report how much a program would spend in absolute terms after enactment. It reports how much more, or how much less, the government would spend, and how much more, or how much less, the government would collect, if the bill became law and everything else in the projection remained as the baseline assumed. The subtraction is the whole of the exercise. Spending that appears only because the bill creates it, expands it, accelerates it, or prolongs it remains in the result, and that remainder is the score that enters committee reports, floor debate, and the enforcement procedures of the budget process. A change in the baseline can enlarge a score, shrink a score, or erase a score altogether without a single word of the bill having changed, and for that reason the rules for building the baseline could not be left to habit or preference.

The Congressional Budget Office was created by the Congressional Budget and Impoundment Control Act of 1974, Public Law 93-344, and the baseline is produced under section 202 of that act. The office publishes the baseline in The Budget and Economic Outlook, the periodic volume in which its economic and budget projections are presented together. Cost estimates prepared for individual bills measure the effects of those bills against that baseline. The sequence deserves emphasis, because it reverses the order that public argument often suggests. An estimator confronting a health provision, a tax provision, or an authorization does not begin by deciding what assumptions seem plausible for the budget as a whole and then apply those assumptions to the provision. The general assumptions have already been settled, in a document prepared under statutory direction before the particular bill arrived, and settled in a form that applies in the same way to bills of every kind. The estimator’s judgment operates inside a frame that was built earlier and built elsewhere. The frame within which the work proceeds, however, is not the estimator’s creation. The conventions that define the frame shape every score before any estimator exercises judgment on the particular language of a bill, and a dispute that presents itself as a dispute about a number is very often, on inspection, a dispute about the frame.

Who chooses the assumptions behind a cost estimate?

Congress chooses the governing assumptions through statute. Section 257 of the Balanced Budget and Emergency Deficit Control Act of 1985 prescribes how the baseline treats direct spending, receipts, expiring programs, and discretionary appropriations, and the Congressional Budget Office applies those rules rather than selecting a preferred forecast for each estimate.

Section 257 bears the title “The Baseline,” and its citation, 2 U.S.C. 907, places it among the provisions through which Congress instructed its own budget office how the measuring rod was to be constructed. Congress did not merely direct that a baseline be prepared, or that it be prepared on reasonable assumptions, or that it reflect the best judgment of the director. Congress specified, category by category, what the projection was to assume about the operation of spending law, about the operation of revenue law, about programs whose authorizations were scheduled to end, and about the annual appropriations that Congress itself had made for a single year and might enlarge, reduce, or abandon in the next. Each specification removed a question from the realm of estimation and placed it in the realm of instruction. A convention, in this setting, is not a custom that estimators happened to share, and it is not a modeling choice defended in a methodological paper. It is a direction from the institution that created the office, addressed to the office, about how the institution wanted its proposals measured. If the baseline could be rebuilt estimate by estimate, then two identical provisions, presented in different months or by different committees, could receive different scores because the background projection had shifted for reasons unrelated to either provision. Comparability across bills, across committees, and across sessions depends on a background that holds steady for reasons other than an estimator’s view of the moment, and a statutory baseline supplies that steadiness.

The first convention in section 257 concerns the great body of law that operates without annual appropriations. Laws creating direct spending and receipts are assumed to operate in the manner those laws specify. A statute that defines eligibility for a benefit, sets a payment formula, and directs that payments be made to persons who satisfy the definition is projected as though the definition, the formula, and the direction mean what they say. Funding for entitlement authority is assumed adequate to make all payments the laws require. That adequacy assumption closes off a line of speculation that would otherwise corrupt every projection built on mandatory law. Without it, an estimator would have to guess whether some future shortfall or administrative failure might leave eligible persons unpaid, and the baseline would begin to record guesses about breakdowns in the machinery of government rather than the operation of the law as written. With the assumption in place, the projection follows the entitlement. If eligibility expands under a formula already in law because wages, prices, caseloads, or demographic facts move in a projected direction, outlays in the baseline move with the expansion, because the law requires the payments and the convention requires the estimator to assume that the payments can be made. The convention treats enacted mandatory law as operative law, fully funded for the obligations it creates, and it leaves to a proposed bill the burden of accounting for any departure from that operation.

The treatment of programs scheduled to expire extends the same logic, through a rule whose date and threshold reveal how concrete the statutory drafting was. No program established by a law enacted on or before the Balanced Budget Act of 1997, with estimated current year outlays greater than $50,000,000, is assumed to expire in the budget year or the outyears. The date separates an older stratum of programs, created by laws enacted on or before the 1997 act, from programs created afterward. For a program on the older side of the date and above the threshold, the baseline refuses the scheduled ending. Authorization language may state that the program terminates on a specified date, and the baseline, following section 257, projects the program past that date as though the operation of the program continued. For newer programs, those established by laws enacted after the 1997 act, section 257 does not impose the same automatic answer. The scoring of newer programs is settled in consultation with the Budget Committees and the Office of Management and Budget. That consultation is a revealing exception, because it shows what happens on the margin where the statute declines to dictate a single treatment. The question returns to institutional negotiation among the office that prepares the estimate, the committees charged with the budget process, and the executive branch office charged with the President’s budget, precisely because no enacted convention has already answered it. Where Congress legislated a convention, consultation is unnecessary. Where Congress legislated none, consultation is the mechanism that fills the gap.

The largest case under the expiration rule is the Supplemental Nutrition Assistance Program, the largest expiring program that the baseline assumes to continue. The program operates under authorizations that Congress enacts for a period of years and then must enact again, a structure familiar from farm legislation, in which nutrition assistance and agricultural programs move through Congress on a recurring cycle. The baseline follows the statute instead. Because the program satisfies the conditions of section 257, its outlays continue in the projection across the scheduled expiration, at levels driven by the eligibility and benefit rules of the governing law rather than by the date on the authorization. Congress had watched programs of this scale move through expiration dates and reauthorizations often enough to know that a baseline built on literal termination dates for major programs would lurch downward at dates that had no corresponding expectation of actual termination, and would then lurch upward again when reauthorization occurred. A measuring rod that lurched for calendar reasons would attribute to reauthorizing bills enormous costs that represented no change in policy at all, since the policy, in substance, had never been expected to lapse. Section 257 chose stability of the rod over literalism of the calendar for programs large enough to disturb the totals.

Discretionary spending presents the opposite structural problem, and section 257 solves it with a different convention. Discretionary programs receive their funding through annual appropriations acts, so no enacted law supplies a payment formula or an entitlement that a projection can simply follow into future years. The appropriation for a given year is exactly that, an amount for that year, voted by a Congress that cannot bind its successors to repeat it. The convention adopted assumes that discretionary appropriations continue at current levels adjusted for inflation. The projection takes the enacted level and moves it forward by a price measure, neither rewarding a program with assumed growth beyond inflation nor punishing it with an assumed real decline. Two price measures do this work, each assigned to the kind of cost it fits. The GDP price index, a broad measure of prices across the economy, is used generally. The employment cost index, a measure of the cost of labor, is used for personnel costs, where the expense being projected is dominated by compensation rather than by the general run of goods and services an agency purchases. Even with that refinement, the convention remains deliberately unambitious as prophecy. It does not anticipate that Congress would expand a favored activity, terminate a disfavored one, respond to a war, a disaster, or a recession with supplemental appropriations. It projects the government that the most recent appropriations created, carried forward at prices that keep its real scale roughly constant, and it offers that projection as the background against which change is to be measured.

Receipts receive a treatment that stands in sharp contrast to the treatment of expiring spending programs, and the contrast is written into the same section. Receipts follow the law as written. An expiring tax provision expires on its scheduled date in the baseline, even when every participant in the budget debate expects an extension; the section carries a single exception, for excise taxes dedicated to a trust fund, which are assumed to continue at their current rates. A rate reduction enacted with a sunset loses its effect on projected revenues at the sunset. The estimator is not invited to weigh the political strength of the provision’s sponsors, the history of earlier extensions, or the statements of leaders who promise that the provision will never be allowed to lapse. All of that evidence might support an excellent forecast, and none of it is admissible under the convention, because the convention for receipts is fidelity to the enacted text. The asymmetry that results is therefore not an oversight, a drafting accident, or a bias discovered after the fact in the office’s practice. Expiring provisions on the spending side and expiring provisions on the tax side are treated asymmetrically by design. A large spending program rides through its expiration date inside the baseline, while a tax provision of comparable political durability falls out of the baseline on schedule.

The asymmetry shapes the politics of extension in opposite directions on the two sides of the budget, and understanding that shaping is essential to reading any score that involves a sunset. Consider first the path of a tax provision approaching its scheduled end. Inside the baseline, revenues after the sunset are projected at the higher level that the expired provision no longer reduces. A bill that extends the provision for additional years is then measured against that higher revenue path, and the measurement registers a revenue loss roughly equal to the full value of the provision across the extension period. It follows arithmetically from a baseline that obeyed the statute and removed the provision on schedule. The score for the extension announces, in effect, that keeping a provision that taxpayers had been using all along counts as giving up revenue, because the measuring rod had already banked the revenue from its disappearance. Consider next the path of a large spending program approaching an authorization expiration that section 257 instructs the baseline to ignore. Inside the baseline, outlays for the program continue across the date as though reauthorization had already occurred. A bill that reauthorizes the program in substantially its existing form is measured against that continuing path, and the measurement registers little or no additional cost, because nearly everything the bill provides was already present in the background projection. The extension appears nearly free, not because the program costs nothing, and not because an estimator decided to excuse its sponsors, but because the baseline had assumed the very continuation that the bill enacts. The program’s real cost sits inside the baseline itself, in the totals that precede any bill, rather than in the score of the reauthorizing measure. A provision’s score can therefore be manufactured or destroyed by its interaction with a convention. Those interactions determine, years in advance and before any estimator studies the extension bill, whether an extension will arrive on the floor carrying a heavy measured cost or almost none at all.

These consequences explain why the baseline is frequently unrealistic by design, and why that unrealism, properly understood, is a property of a measuring instrument rather than a defect in a forecast. The baseline assumes that tax cuts expire while large expiring spending programs continue. It assumes that appropriations grow only with inflation regardless of stated plans, announced priorities, or pressures that any observer could identify in advance. It takes no position on what Congress will do next, even in cases where the next action is broadly anticipated by the members who will take it. Assembled into a single projection, those assumptions describe a future that informed participants would seldom describe as the likeliest future. Revenues in the projection jump at sunset dates that Congress may never allow to take effect. If the baseline were offered as a prediction, those features would be grounds for discarding it. The baseline is not offered as a prediction. It is the baseline, the fixed background published under section 202 of the 1974 act in The Budget and Economic Outlook, against which the cost of changing the law is struck. A measuring rod does not fail because it declines to bend in anticipation of the object being measured. Its value lies in holding a known shape, so that differences in measurements can be attributed to differences in the proposals rather than to movements in the instrument. The statutory conventions give the rod its known shape. Transparency of construction substitutes for realism of prediction, and the substitution is deliberate.

That substitution supports the central claim that a reader of cost estimates should carry into every dispute about a score. The baseline is written in statute, not chosen. The estimators of the Congressional Budget Office did not select the assumption that a large expiring program continues, and they did not select the assumption that an expiring tax provision vanishes, and they did not select the assumption that appropriations advance with inflation while plans for expansion or retrenchment are ignored. Congress selected those assumptions, enacted them in section 257 of the Balanced Budget and Emergency Deficit Control Act of 1985, and directed its budget office to build the baseline accordingly. The most common criticism of cost estimates, that the assumptions are unrealistic, is usually a criticism of conventions Congress itself enacted. That criticism may have force as an argument for amending the conventions, since Congress retained the authority to rewrite section 257 whenever a majority preferred a different measuring rod. As an argument against a particular estimate, however, the criticism misidentifies its target. An estimator who departed from section 257 in order to produce a more plausible background would not be improving the estimate. The estimator would be substituting a private judgment about future legislation for the background that the law prescribed, and every subsequent score would inherit that substitution without the visibility that a statute provides.

One qualification completes the account, and it concerns the difference between the publications in which the office explores the budget outlook and the estimates in which the office scores legislation. The estimating office has at times published alternative presentations alongside the baseline to show how the path changes under different policy assumptions. In those presentations, a reader can see the budget trajectory that results if expiring tax provisions are extended, if discretionary spending follows a path other than inflation adjustment, or if other conventions are relaxed in favor of assumptions closer to expected policy. Such presentations serve a genuine analytical purpose. Their existence does not soften the rule that governs estimates. Scores are struck against the statutory baseline alone. Every bill receives the background that section 257 constructed, published in the outlook, applied uniformly, and defended on a single ground that recurs throughout the budget process. The ground is that Congress wrote the conventions, Congress can rewrite them, and until Congress does so, the measure of a bill is the distance between the bill and the law that already exists, projected exactly as Congress directed that the law be projected.

Two Scorekeepers and the Division of Labour

A major bill often reaches public discussion carrying a single headline number, and that number often arrives carrying a single attribution. News reports, floor speeches, and committee summaries compressed the figure into a convenient formula in which the Congressional Budget Office had scored the bill at a given cost or a given saving over a given period. The formula sounded complete because the budget office had issued the cost estimate, had signed the letter that transmitted it, and had combined the components into the total that debate quoted. Compression, however, concealed authorship. Inside the single total, two distinct bodies of professional work had been joined, and for a bill with a substantial tax title the larger or more contested share of the arithmetic frequently belonged not to the office that issued the estimate but to the staff of the Joint Committee on Taxation. Congress had written the division into law, had assigned different parts of the federal ledger to different estimators, and had directed that the parts be combined before enforcement rules consumed the result. A reader who treated every dollar in the headline figure as a budget office product therefore misunderstood both the document and the institution that had produced the disputed part of it.

The conflation persisted because the finished product looked unitary. A cost estimate letter presented one bill, one set of budgetary effects, and one bottom figure for the deficit, organized by year and by category so that the Budget Committees, the leadership, and the Congressional Record could use the same numbers for the same procedural purposes. Enforcement depended on that unity. Scorekeeping, budget resolution allocations, and points of order could not operate on two rival totals issued in parallel by two offices answering to different committees. The system therefore required a single official estimate for a bill even when the expertise behind the estimate was divided. That shorthand was harmless for a bill that changed only spending programs, because the budget office had estimated the outlays, the administrative costs, and the offsetting receipts within its assigned sphere, and attribution to the budget office described the work with reasonable accuracy. The shorthand became misleading when a bill rewrote the Internal Revenue Code. In that setting, the budget office letter functioned as a container for a revenue estimate that another staff had built from tax data, tax models, and assumptions about taxpayer behavior, and the container received credit for contents it had deliberately not duplicated.

Which body scores the tax title of a bill?

The staff of the Joint Committee on Taxation scores the tax title. Under section 201(f) of the Congressional Budget Act of 1974, its estimates are the official estimates for changes to income, estate and gift, excise, and payroll taxes, and the Congressional Budget Office incorporates them into its cost estimates.

The statutory settlement began with section 201(f) of the Congressional Budget and Impoundment Control Act of 1974. That provision stipulated that the revenue estimates of the staff of the Joint Committee on Taxation were the official estimates for legislative changes to income taxes, estate and gift taxes, excise taxes, and payroll taxes considered by Congress. The language settled more than the source of a number. It settled finality. For the covered taxes, an estimate prepared elsewhere, however sophisticated the model or distinguished the economist, lacked official standing for the legislative process. The figure that entered the cost estimate and then entered enforcement was the figure supplied by the Joint Committee staff. The budget office incorporated that figure rather than producing a rival revenue number of its own. The statute thereby prevented a form of institutional competition that would have damaged the budget process at its foundation, namely two official scorekeepers publishing two official revenue totals for the same provision and leaving committees to choose the total that made a bill easier to pass.

The choice of the Joint Committee staff as the revenue estimator rested on a concentration of expertise that had accumulated around the tax writing committees long before the budget office existed. Estimating a change in the income tax required more than projecting a rate against a static base. The estimator had to understand how the provision interacted with deductions, credits, exclusions, filing behavior, business organization, timing of income recognition, and the administrative capacity of the Internal Revenue Service. The Joint Committee staff, composed of economists, lawyers, and accountants who worked continuously on the structure of the tax code, held that knowledge as a daily professional discipline rather than as one assignment among many. Its lawyers read the statutory language as drafters of tax legislation read it, its accountants traced the language into returns and reporting, and its economists translated the combined judgment into revenue effects. The budget office, created in 1974 to give Congress an independent source of budgetary and economic analysis, had no reason to reproduce that specialized apparatus inside a second building and every reason to rely on it under a statutory rule that made reliance mandatory.

History reinforced the logic of the statute because the division predated the budget office itself. The Joint Committee estimating role had grown with the income tax across the twentieth century, as federal revenue came to depend on a tax whose yield turned on intricate statutory definitions and on the responses of millions of filers to changes in those definitions. By the time Congress enacted a comprehensive budget process in 1974, the tax committees already possessed a professional staff with an established estimating function, an established relationship to the legislative drafting process, and an established body of models and data. The 1974 act confirmed that arrangement in statute rather than creating it. Confirmation mattered because the figure became an input into allocations, enforcement, and deficit totals for the government as a whole. Section 201(f) connected the older tax institution to the newer budget institution without displacing either, assigning the revenue judgment to the staff that had long made it and assigning the integration of that judgment into a complete cost estimate to the office Congress had created for budget-wide work.

The budget office nevertheless produced revenue estimates of its own, and the residue assigned to it followed a discernible logic. The office estimated proposals affecting unemployment insurance taxes, customs duties, miscellaneous fees, and the earnings of the Federal Reserve System. Each item in that residue sat closer to the spending and economic analysis that formed the ordinary work of the budget office than to the core Internal Revenue Code provisions that the Joint Committee staff scored. Unemployment insurance taxes operated inside a system of state programs, federal accounts, and employment cycles that the budget office modeled as part of its labor and income security work. Customs duties depended on trade volumes, import prices, and tariff schedules that intersected with the office’s economic forecast. Miscellaneous fees arose from programs the budget office costed on the spending side, where a charge for a service, a license, or a regulatory activity could not sensibly be severed from the cost of the activity itself. Federal Reserve earnings turned on interest rates, the central bank portfolio, and monetary conditions that belonged to the macroeconomic framework the budget office maintained for its baseline. It marked the boundary at which a receipt ceased to be principally a product of tax law administered through returns and became principally a product of a program, a market, or an economic condition that the budget office already modeled for other parts of the estimate.

Assembly turned the two products into one enforceable document. In a typical cost estimate letter, the budget office prepared the estimate, sourced the revenue figures to the Joint Committee staff, and combined the revenue total with the outlay total and any other budgetary effects into the single deficit figure that enforcement rules consumed. The credit appeared on the face of the document rather than in an obscure acknowledgment, so a careful reader could see which institution had supplied which component before accepting the combined result. That transparency served a practical purpose in committee work. A member questioning a spending assumption addressed the question to the budget office analysts who had modeled enrollment in a program, the cost of a benefit, or the pace at which an agency would obligate funds. A member questioning a revenue assumption addressed the question to the Joint Committee staff, where the dispute might concern elasticity of reported income, the timing of realizations, or the expected shift between taxable and tax preferred forms of compensation. The letter joined the answers for procedural use without erasing the separate responsibility for them. The internal attribution preserved accountability inside the unity that enforcement demanded.

Collaboration complicated the neatness of the boundary because many policies affected taxes and spending at the same time, and the two staffs worked closely on such policies rather than exchanging finished numbers across a wall. A provision could reduce revenue in one account, increase outlays in another, and alter the behavior that determined both effects. The budget office relied on the tax committee staff’s expertise in developing estimates for those provisions, and the tax staff in turn depended on budget office work on the populations, markets, and programs that a tax provision touched. Consistency was the central requirement. If one staff assumed a given level of participation, a given price, or a given response by employers and households, the other staff had to build its side of the estimate on the same assumption, or the combined total would count a world that could not exist. The single figure at the end of that process represented not only addition but agreement, an agreement that the revenue effect and the spending effect in the estimate described the same bill operating on the same economy during the same years.

Health legislation exposed the interlock more completely than any other field because premiums, subsidies, and tax credits sat on both sides of the ledger at once. A premium tax credit illustrated the problem in its purest form. The credit reduced revenue when a taxpayer claimed it against tax liability, and the same credit generated an outlay when the amount exceeded liability or when the Treasury advanced payment to an insurer on the taxpayer’s behalf. Enrollment in health insurance, the choice of a plan, the premium charged for that plan, the income reported by the household, and the decision by an employer to offer coverage all influenced how much of the credit appeared as forgone revenue and how much appeared as direct spending. The Joint Committee staff brought the tax side of that calculation, including the structure of the credit, the interaction with other provisions of the code, and the response of filers. The budget office brought the health side, including insurance markets, Medicaid interaction, employer coverage, and the uninsured population. Neither side could be estimated credibly in isolation, because an assumption about enrollment adopted for the outlay estimate changed the population claiming the credit in the revenue estimate, and an assumption about how employers responded to a tax provision changed the number of people who needed subsidized coverage in the first place. Scoring a health bill therefore required one consistent set of assumptions built jointly, tested against both ledgers, and revised until the revenue figure and the outlay figure told a single coherent story about the same households.

Institutional character explained part of the authority that the arrangement carried inside Congress. The Joint Committee on Taxation was a joint committee of Congress, drawn from both chambers and supported by its own professional staff rather than by a department of the executive branch. Its members came from the House Committee on Ways and Means and the Senate Committee on Finance, the two committees with jurisdiction over tax legislation, so the estimating staff served the legislators who wrote the tax laws and who bore responsibility for defending the revenue consequences of those laws in committee and on the floor. The chief of staff of the Joint Committee led the estimating work and functioned, in practical terms, as the revenue side counterpart to the director of the Congressional Budget Office. Each led a professional staff, each answered to Congress rather than to the President, each guarded a reputation for technical competence that depended on producing estimates that members of both parties could use, and each presided over numbers that could determine whether a bill survived a procedural challenge. The functional resemblance clarified the standing of the tax estimate. The figure did not arrive as a subordinate contribution to another office report. The figure arrived as the product of the senior revenue estimating institution of Congress, incorporated by law into the budget office document.

That institutional standing shaped how disputes unfolded, and the consequences reached both legislators and readers of the published estimate. A revenue dispute was litigated before a different referee than a spending dispute. A senator who believed that a tax cut would induce more economic activity, more reported income, or less avoidance than the official estimate allowed had to persuade the Joint Committee staff, work through the Finance or Ways and Means Committee, or challenge an estimating convention that the tax committees treated as their own. A senator who believed that a new benefit would enroll fewer people or cost less per participant had to make that case to the budget office. The tax committees guarded the Joint Committee estimate as an instrument of their jurisdiction as well as a technical product. Control over the official revenue number meant control over the starting point from which amendments were judged, compromises were measured, and the cost of a chairman mark was announced. Surrendering that number to the budget office would have weakened the tax committees’ position inside the legislative process, quite apart from any question about which staff modeled a provision better. The statutory division therefore endured because technical competence and committee power pointed in the same direction, and a reader tracing a controversy over a tax estimate without grasping that alignment mistook a jurisdictional contest for a purely methodological one. The place of the tax title in the wider legislative journey, from the originating committee through floor consideration in both chambers, receives fuller treatment in how tax bills move through Congress, and that journey helps explain why the tax committees defended the estimating arrangement that section 201(f) had confirmed.

For readers outside the committee rooms, the diagnostic that followed from the division was simple to apply and difficult to overstate in value. The first question to ask of any headline number was which body had produced which part of it. A headline that reported a ten-year cost invited a division of that cost into revenue losses, outlay increases, and interactions between the two, followed by an attribution of each part to the staff that had estimated it. The attribution changed the questions that a skeptical reader pressed. Revenue skepticism led toward tax bases, behavioral responses, timing shifts, and the treatment of income that could move between years or between forms. Spending skepticism led toward participation, benefit levels, provider prices, state responses, and administrative capacity. Confusing the two led debate toward the wrong office, the wrong assumptions, and often the wrong remedy, because a member could denounce the budget office for a revenue convention the office had been required by statute to accept, or praise the Joint Committee staff for an outlay projection its analysts had never made. The cost estimate letter itself supplied the correction to that confusion on its face, in the credit that named the source of the revenue figures inside the budget office document. Reading that credit was the smallest act of literacy the budget process demanded, and it transformed a single attributed number into what the number had been from the start, a joint product of two scorekeepers whose separate judgments Congress had bound together into one enforceable total.

The Ten Year Window and What Lies Beyond It

Every figure in a cost estimate carries a date, and the collection of dates carries a boundary. A table that reports a change in outlays or revenues for a fiscal year assigns that change to a particular year, sums a defined run of years into a total, and leaves every year after the run outside the total. The boundary is not discovered in the policy being measured. A tax credit does not cease to reduce revenues on a date chosen by a budget rule, a health program does not cease to pay benefits because a reporting frame ends, and a loan program does not cease to carry credit risk because the last column of an estimate has been filled. The dating is a rule of the process rather than a property of the policy, and the distinction controls everything that follows in a reading of any score. The Congressional Budget Office measured the change that a bill would produce against a baseline, under the assumption that the bill was enacted as written and that the economy followed the forecast the office had published, and it reported that change inside a frame Congress had selected for its own decisions. The frame also gave every reader a partial number that looked complete, because a total printed without its boundary invites treatment as a total effect when it is in fact a total within a period.

Why do cost estimates stop at ten years?

Cost estimates stop at ten years because budget resolutions and enforcement rules adopted a ten year horizon, and the Congressional Budget Office reports the change from its baseline within the frame those rules consume. The limit is a process rule, not a finding that a policy ends.

The ten year horizon acquired its authority by accumulation. Section 402 of the Congressional Budget Act of 1974 framed the original estimating duty over five years, requiring the office to provide the five year cost of legislation reported by committees. That assignment reflected the scale of the process created in 1974. The budget resolution, the allocation of spending authority among committees, and the cost estimate attached to a reported bill all required a period long enough to show a program reaching operation and short enough to keep the figures tied to the forecast that supported them. Five years served that purpose in the statute as enacted, and early estimates presented their results inside a frame half the length of the frame that later became standard. The five year frame also expressed a caution that survived its replacement. An estimate grows less certain as it extends, because enrollment, prices, wages, interest rates, caseloads, and behavior all move away from the forecast conditions under which the first years were projected.

The lengthening occurred in steps, and no single amending statute performed it. During the 1990s and 2000s, budget resolutions came to be written against ten year projections, so the resolution that set aggregate levels of revenues, spending, and deficits for a decade required estimates capable of speaking to the same decade. Enforcement followed the resolution outward. The pay as you go scorecard measured the deficit effect of direct spending and revenue legislation over periods that included a ten year horizon, reconciliation instructions directed committees to produce changes within a ten year frame, and points of order tested measures against budgetary limits that extended across ten years. Each element of the enforcement architecture consumed figures in the form the architecture prescribed. An estimate confined to five years could not answer a point of order written against ten, could not fill a scorecard kept on a ten year basis, and could not certify compliance with a reconciliation instruction drafted for a decade. The estimating practice therefore moved with the rules that used the estimates. By the 2010s, a cost estimate presented year by year figures and a ten year total as a matter of course, and the budget outlook publications of the Congressional Budget Office projected ten years ahead, so that the baseline against which bills were measured and the measurements themselves occupied the same span.

The ten year total became, in consequence, the number that enforcement consumed and that headlines repeated. A bill was described as costing or saving a stated sum over ten years, and the qualifier that located the sum inside a window receded behind the figure itself. That compression had an institutional logic. Enforcement requires a determinate test, and a determinate test requires a closed period with a beginning and an end. A point of order either lies or does not lie against a measure under the applicable rule, and the determination depends on figures for the period the rule names. The ten year total supplied the quantity for those judgments, neatly, repeatedly, and identically across proposals that differed in every other respect. The same compression distorted public understanding in a predictable direction. Two bills with identical ten year totals could carry radically different obligations in the eleventh year and afterward, one expiring cleanly and the other entering its most expensive phase, and the headline figure registered no difference between them. Two bills with different ten year totals could converge to the same long run cost once a temporary provision in one of them lapsed. The frame treated those distinctions as information outside the score, even when the distinctions formed the central fiscal fact about the legislation.

The window also acted directly on legislative design, and its pressure operated with mechanical regularity. A provision whose costs began in year eleven was free inside the frame, because the total recorded no outlay for a benefit, a subsidy, or a tax reduction scheduled to commence after the final counted year. A provision whose savings arrived in year eleven bought nothing inside the frame, because the total recorded no reduction for a restraint, a fee, or an efficiency scheduled to take effect after the same boundary. Between those two poles, the placement of a start date, the slope of a phase in, and the insertion of a sunset acquired fiscal meaning that had nothing to do with program logic. A delayed start lowered the measured cost without lowering the annual cost once the program reached full operation. A phase in spread expenditures so that only the smaller early amounts fell inside the count. A sunset cut off the counted years at the point where the provision became most expensive, while leaving open the possibility of a later extension whose cost would be scored in a new window from a new baseline. The gravitational pull fell on every date in a bill. Drafters who ignored the edge paid for permanence inside the frame, and drafters who attended to the edge could present a design whose measured cost understated its steady state cost by construction. The drafting response to that pressure, the catalog of delayed effective dates, expiring provisions, and back loaded savings that the window encouraged, receives its own treatment later in the article.

Effects beyond the window did not vanish from the record, because Congress and the estimating office built machinery to report them without enlarging the score itself. For major legislation, the Congressional Budget Office reported whether costs would grow or shrink in the second decade, the ten years that followed the formal window. Such a statement placed on the record an informed characterization of direction and approximate magnitude, so that a member voting on the basis of the ten year total also had before the chamber a statement about the period the total excluded. The practice acknowledged a plain fact about the composition of major bills. Entitlements for health care, retirement, and income support, along with permanent changes in the revenue code, rarely reached a steady state within ten years of enactment. Caseloads built, eligibility phased in, provider payment formulas compounded, and demographic change pressed harder in the second decade than in the first. An estimate that stopped without comment at year ten therefore omitted the period in which the largest structural forces operated, and the second decade statement supplied a disciplined comment without pretending to the precision of the year by year table.

The estimate for the health care legislation enacted in March 2010 illustrated the form. The Congressional Budget Office, in its March 2010 letter on the legislation, stated that the bill would reduce deficits in the second decade by a small share of gross domestic product, in a range around one half of one percent. The formulation deserves attention for what it refused to claim. The office gave a direction, a rough magnitude expressed as a share of the economy rather than as a dollar total, and a range rather than a point. Each choice reflected the distance of the projection. A share of gross domestic product scaled the claim to the size of the economy that would bear it, and a range around one half of one percent communicated that the reduction was material without being large relative to the fiscal pressures of the period. The letter thus extended the analysis past the window while marking, in its own language, the lower resolution of everything past the edge.

A separate test gave the years beyond the window a harder edge in Senate enforcement. Under that test, a measure failed if it would increase deficits by more than $5 billion in any of the four consecutive ten year periods following the window. The four periods together covered forty years beyond the estimate frame, divided into blocks so that a large increase in any single block triggered failure rather than being averaged away against smaller increases or decreases in neighboring blocks. The test operated as a Senate point of order written into budget resolutions, which a senator could raise against a bill, an amendment, or a conference report that breached the stated beyond window limit, subject to waiver by the supermajority the applicable rule required. Reconciliation bills faced a distinct and stricter discipline in the same territory. The Byrd rule, the provision of Senate procedure named for Senator Robert C. Byrd that also excluded material extraneous to budgetary change from the expedited reconciliation process, treated a provision that would increase the deficit in any year beyond the window covered by the reconciliation measure as extraneous, with no tolerance of the $5 billion kind at all. Formal estimates for major bills stated the result of the test on their face, so the conclusion appeared in the estimate itself rather than in a separate parliamentary memorandum. The design of the test revealed its purpose. The $5 billion threshold stood far below the scale of a major entitlement or a major tax change, so the test operated less as a tolerance for minor spillover and more as a near prohibition on deficit increase in the out years for measures passed under the protections the test guarded. A bill could comply inside the window by pairing permanent costs with temporary savings and then fail beyond the window when the savings expired and the costs continued, which was precisely the configuration the test existed to catch.

The long term outlook supplied the institution with its own answer to the blindness of the window. In The Long-Term Budget Outlook, the Congressional Budget Office projected revenues, spending, deficits, and debt over thirty years, a horizon three times the length of the cost estimate frame and long enough to contain the retirement of the baby boom generation, the compounding of interest on accumulated debt, and the full effect of health care cost growth on the federal budget. The outlook did not score individual bills. It described the path of the government as a whole under specified assumptions about law and the economy, and it separated that path into the part driven by demography and health costs and the part driven by policy choices. The separation mattered because it assigned responsibility with care. Population aging raised spending on Social Security and Medicare under any plausible policy, as the ratio of beneficiaries to workers rose on a schedule fixed largely by births that had already occurred by the time of any given outlook. Health care costs per beneficiary grew under pressures from technology, prices, and utilization that legislation influenced without fully controlling. Against those forces stood choices that Congress retained in full: the level of revenues, the design of benefit formulas, the age of eligibility, the structure of subsidies, and the decision to borrow the difference between spending and receipts or to close it. An outlook that blended the two categories into one rising line of debt would have taught a false lesson of inevitability. An outlook that separated them showed where pressure originated, where choice remained, and how much of a projected deterioration a stated policy change could or could not reverse.

Read together, the window and its supplements described a single measurement system with a hard center and a softer periphery. The hard center, the ten year estimate, carried legal consequences through scorecards, instructions, allocations, and points of order, and its figures were stated year by year to the nearest stated unit. The periphery, the second decade characterization, the test across four consecutive ten year periods, and the thirty year outlook, carried information without the same legal consequence, and its figures were stated as ranges, thresholds, and paths under assumptions. Confusion between the two zones produced the most common errors in the public use of a score. The estimate itself warned against each error in its structure. Its baseline was a construct, its economic assumptions were dated, its treatment of behavior followed the conventions the office disclosed, and its window was printed on the face of every table for any reader willing to attend to it.

The diagnostic that survives the detail is therefore a habit of questioning rather than a rival number. A ten year figure is a measurement inside a frame Congress chose, assembled from a baseline Congress instructed the estimator to use, under rules Congress wrote and could rewrite. The figure answers the question the process asked, which concerned the change in deficits during a specified decade under specified assumptions, and it answers no other question without assistance. The honest questions about any bill include the questions the frame excludes by construction. What did the estimator say about the second decade, and in what terms of direction, range, and share of the economy was that statement cast. Did the formal estimate state a result under the test for the four consecutive ten year periods beyond the window, and did the measure pass that test or require a waiver of the point of order it triggered. Does the provision that drives the ten year result expire, phase in, or reach a steady state inside the frame, or does its principal effect begin after the last counted year. Does the long term outlook locate the pressure the bill addresses in demography, in health costs, in revenues, or in the accumulation of interest, and does the bill act on the source the outlook identifies. A score read with those questions in hand regains its proper standing. It remains the indispensable common measure that allows a legislature to compare proposals, enforce its own budget, and record the fiscal consequence of a vote. It ceases to masquerade as a prophecy about a policy whose life, by design and by demographic fact, extends far past the edge of the page on which its first ten years were counted.

How an Estimate Is Built and How Drafters Answer It

The estimating office scores the legislative text as written. The words that define who qualifies, what benefit or tax change applies, when a provision takes effect, and when a provision ends are the raw material of the calculation, and the speeches, summaries, and stated intentions that surround a bill enter the calculation only insofar as the enacted language carries them. A floor statement describing a program as narrow does not narrow a score if the statutory definition is broad, and a sponsor describing a provision as temporary does not limit its cost if the text contains no ending date. The office therefore reads a bill the way an administering agency and a court would be obliged to read the resulting law, attending to definitions, cross references, effective dates, authorization levels, and the interaction of one section with another. Ambiguity in the text becomes a question about the most plausible administration of the language as drafted, and the estimate records the cost of that plausible administration rather than the cost of the policy as advertised.

That reading of text is the foundation, and the construction proceeds outward from the foundation to the institutions that would have to operate the law. The office obtains information from the agencies that would administer the program, because those agencies hold the operational knowledge that a statutory sentence compresses. Administrators know the size and composition of an existing caseload, the unit cost of a benefit or a service, the capacity of a payment system, the time required to issue regulations and modify computer systems, and the points at which a new eligibility rule would collide with an existing one. Agency information does not dictate the result. The estimating office weighs the information against other evidence, tests the assumptions embedded in an agency projection, and retains independent judgment over the final figure, a separation that protects the score from becoming a restatement of the administering department’s preference.

Experience with comparable programs supplies the next layer of evidence, and in many estimates that layer carries much of the weight. Few provisions are entirely without precedent. A new benefit resembles an older benefit in its eligibility structure, its payment method, or the population it reaches, and the recorded behavior of the older program offers a guide to the behavior that the new provision would produce. Enrollment in a comparable program shows how quickly eligible people learned of a benefit and completed an application in the past, payment records show how costs per participant moved across the early years of operation, and administrative records show how error, delay, and attrition affected the difference between the population eligible on paper and the population served in practice. The estimating office adjusts that experience for differences in generosity, outreach, eligibility thresholds, and economic conditions, rather than transplanting an old result unchanged.

Upon those foundations the office builds the behavioral layer, the part of a conventional estimate that separates a modeled response from a mechanical multiplication. A mechanical calculation would apply a new tax rate to an unchanged base, or multiply a new benefit level by an unchanged caseload, and stop there. A conventional estimate does not stop there. The people and firms a provision touches respond to the provision, and the estimate incorporates the responses that evidence and economic reasoning support. Taxpayers alter the timing of income, accelerating a transaction into one year or deferring a transaction into another when a rate change makes the timing advantageous. Income and activity shift between categories when the law treats categories differently, moving between forms of compensation, between business organizations taxed under different rules, and between taxable and favored activity. Participation changes as well, since a more generous benefit draws in eligible people who had not enrolled under an earlier design, and a new condition or a reduced benefit discourages enrollment that would otherwise have occurred. Each response alters the federal receipt or outlay associated with a given statutory rate or benefit level, so the score reflects the law operating on a population that reacts to the law.

The largest boundary on that behavioral layer is the treatment of the economy as a whole, and precision on this point matters because the convention is often misstated. A conventional estimate holds the overall size of the economy at the level of the office’s economic forecast. Gross domestic product, aggregate income, and the broad path of the economy in the forecast do not change in the estimate because a bill has been introduced, so feedback from a provision to the total size of the economy, and from a changed total back to federal receipts and outlays, does not enter the conventional figure. A conventional estimate is therefore not a zero response estimate, a phrase that would describe the mechanical calculation the behavioral layer rejects. A conventional estimate is a fixed economy estimate, responsive at the level of the taxpayer, the beneficiary, and the firm, and deliberately unresponsive at the level of aggregate output.

The construction then separates effects that operate under different budgetary controls, because a single total that merged every category would misdescribe the decisions Congress would face after enactment. Direct spending effects flow from the law itself, through entitlement authority, mandatory payment formulas, and provisions that obligate the government without a further appropriations decision. Spending subject to appropriation follows a different path. An authorization may permit or direct an agency to carry out an activity, but the money reaches the agency only through a later appropriations act, and the timing and level of that later decision remain in congressional hands. For provisions of that second kind, the estimating office estimates the cost of implementing the authorization on the assumption that the money is provided, since an assumption of no funding would convert every authorization estimate into a figure of zero and would hide the cost of the commitment Congress was considering. The estimate reports that figure separately from direct spending effects, preserving the distinction between a cost that a bill imposes by its own terms and a cost that a bill contemplates subject to a future vote.

Federal credit programs require a distinct accounting treatment, because the cash pattern of a loan misstates the cost of the loan. Under a cash presentation, a direct loan would appear as a large outlay in the year of disbursement and as receipts in later years as the borrower repaid, while a loan guarantee would appear nearly costless at issuance and expensive in the year a default produced a federal payment. The Federal Credit Reform Act of 1990 replaced that treatment with an accrual basis for federal credit. Under the accrual approach, the estimating office records the subsidy cost of a direct loan or a loan guarantee up front, in the year of the commitment, as an estimate of the long term cost to the government after expected repayments, defaults, fees, and recoveries have been taken into account. The reform made the budgetary treatment of credit comparable to the treatment of grants, so a decision to deliver assistance through a loan and a decision to deliver assistance through a payment could be weighed on a common basis at the time of decision.

A further construction obligation attaches to mandates, under a statute that made the effect of reported legislation on other governments and on private parties a part of the formal record. The Unfunded Mandates Reform Act of 1995 requires the estimating office to identify intergovernmental mandates and private sector mandates in reported bills and to estimate the costs those mandates would impose on state, local, and tribal governments in one case and on private entities in the other. The act set thresholds in 1995 at $50 million for intergovernmental mandates and $100 million for private sector mandates, with both figures adjusted for inflation after that year, and a mandate whose estimated cost exceeded the applicable threshold triggered the act’s disclosure and procedural consequences. The mandates component differs from the federal cost estimate in its object. The federal score measures receipts and outlays of the United States government, while the mandates statement measures costs borne outside the federal budget by governments and firms that the legislation would oblige to act. A provision could impose a substantial burden of that external kind while recording a modest federal cost, and the 1995 act ensured that the difference appeared in the estimate rather than outside the estimate.

An estimate assembled through those layers remains a conditional calculation rather than a prediction, and the conditions are part of the result. The figure assumes enactment as written, on the schedule the text specifies, administered in accordance with the language, in the economy described by the forecast that framed the estimate. The figure does not price later amendments, because an amendment changes the text and a changed text requires a new estimate. The figure does not price the actions of later Congresses, which could extend, repeal, fund, or refuse to fund elements of the law in ways the original text could not anticipate. The figure does not price administrative choices that the text leaves open, beyond the plausible administration the office adopted in building the estimate, and a different regulatory choice within the statute’s discretion could produce a different cost from the same enacted words.

Can a bill be drafted to change its own score?

Yes. Because an estimate follows the language enacted, the placement of dates, expirations, and payment timing inside that language changes the total recorded inside the budget window. Delaying a start, ending a provision early, or moving a payment across a fiscal year line alters the score without necessarily altering the underlying policy.

The answer follows directly from the method of construction. If the office scores the words enacted, then the choice of words is a choice about the score, and drafters who understand the construction can arrange dates and durations to place costs where the budget window registers them lightly or registers them not at all. The arrangement does not require concealment, since the devices operate through the visible terms of the bill itself, in the effective date on the page, in the expiration clause, and in the schedule of payments that any reader of the text could inspect. The central reader skill is recognition, an ability to look past a headline total for a ten year period and identify the drafting features that produced the total, distinguishing a provision that is inexpensive because its design restrains cost from a provision that appears inexpensive because its design postpones, truncates, or relocates cost in relation to the window.

Delayed effective dates offer the clearest illustration, because the mechanism requires nothing beyond a calendar. A provision that takes effect several years after enactment consumes few of the years in the estimate window, while a provision effective on enactment consumes all of them, and the ten year totals diverge even when the annual cost of the fully operating provision is identical. The health care legislation enacted in March 2010 used that structure at a large scale. The legislation was enacted in 2010, and its main coverage provisions began in January 2014, so four of the ten years in its estimate carried little coverage cost while the later years carried the cost of a program in operation. Delay also serves purposes apart from scoring, since a new program requires regulations and enrollment systems that a distant start date accommodates, and both explanations can be true of the same provision at once. The scoring effect exists regardless of motive, because the window is a fixed span of fiscal years, and every year of delay removes a year of operating cost from that span.

Expirations work in the opposite direction on the calendar and to the same effect on the total, by cutting a provision off before the window closes rather than by starting the provision after the window opens. A provision scheduled to expire records costs only for the years of operation inside the window, and the recorded total understates the cost of the same provision maintained permanently at the same annual level. Phase ins complement expirations, spreading the arrival of a provision across the early years so that the first years record partial costs while later years record rising costs, and the expiration then ends the series before a full decade of the mature provision has been counted. The tax act of 2001 combined those elements, phasing in several reductions over a period of years and expiring at the end of 2010, so the estimate reflected a schedule of partial and complete reductions inside the window and no cost for the policy beyond the scheduled ending.

The expiration in the tax act of 2017 shows the device operating under a direct procedural compulsion rather than under a preference for a lower headline figure alone. The individual provisions of the 2017 act expire after 2025, and that sunset was imposed by the pressure of the Byrd rule, whose beyond the window test for a reconciliation bill penalizes a measure that would increase the deficit in years beyond the budget window. Drafters who wished to move the 2017 reductions through reconciliation therefore had to design the reductions so that the measure did not fail that test, and an expiration of the individual provisions after 2025 supplied the design. The sunset resolved the procedural problem and lowered the recorded cost inside the window at the same time, at the price of placing the fiscal effect of any later extension outside the original estimate entirely.

Timing shifts form a quieter class of device, smaller in scale than a delayed start or an expiration and dependent on the structure of the fiscal year rather than on the life of a program. Federal enforcement watches costs year by year, and a receipt or a payment moved across a fiscal year boundary changes the score in the years on either side of the boundary without changing the substance of the underlying policy. A payment scheduled near the end of a fiscal year and moved into the opening days of the next fiscal year reduces recorded outlays in the first year and raises recorded outlays in the second by the same amount, while the beneficiary receives the same payment separated by an interval of days. Devices of this kind redistribute recorded costs among years in ways that matter when a particular year carries a binding target or limit, and the estimating office records the shift in the year by year figures, so the movement is visible to a reader of the full estimate.

The Medicare physician payment history supplies the purest case in the record of a score governing the shape of legislation across a long period, because Congress returned to the same problem on a nearly annual schedule and each return was dictated by the difference between the cost of a patch and the cost of a repair. A payment formula for physician services under Medicare produced scheduled reductions in payment rates, reductions that Congress was unwilling to allow to take effect and equally unwilling to repeal at the price a permanent repeal commanded in an estimate. The result was the practice known as the doc fix, in which Congress repeatedly postponed the scheduled reduction for one year at a time. Each postponement scored at the cost of one year of higher payments measured against the formula, a figure far lower than the cost of a permanent repair, which would have recorded higher payments in every year of the window. Legislation took the form it took, a succession of temporary overrides, not because a temporary override was a considered health policy, but because the estimate attached to the durable alternative made the durable alternative the hardest option to enact.

Appropriations practice produced a related device inside the annual spending process, where the boundary between mandatory and discretionary spending created an opportunity that drafters learned to use. Appropriations bills set discretionary spending for the year, subject to limits, while mandatory programs operate under separate authorizing law, and a change in a mandatory program enacted inside an appropriations bill came to be known by the acronym CHIMP, for a change in a mandatory program. Where a CHIMP produced scored savings, by delaying a mandatory payment, capping a mandatory account for a year, or rescinding mandatory budget authority, those savings could offset discretionary spending inside the appropriations process and allow a bill to carry a higher level of discretionary appropriations while remaining within its allocation on paper. The device varied in substance from case to case. Some changes altered real program operations and saved money across the full window, while other changes chiefly moved spending beyond the year of the appropriations bill, producing savings in the scored year that later years repaid.

The CLASS Act stands as the limiting case among the devices, the instance in which a favorable fit between a program and the budget window coincided with a program that could not operate as designed. The Community Living Assistance Services and Supports program, known as the CLASS Act, was a long term care insurance program included in the health legislation of 2010, structured to collect premiums from enrolled participants during the early years of operation while the benefits owed to participants lay mostly beyond the estimate window. Within the window, the program therefore recorded substantial premium receipts against limited benefit payments, and that pattern improved the ten year figure for the legislation as a whole. The difficulty lay beyond the window, and the administration suspended the program in 2011 after concluding that the program was unworkable on the terms the statute had set. Congress repealed the CLASS Act in January 2013. A score can certify the effect of a text on the federal budget across a specified decade without certifying that the institution the text creates can survive contact with administration and with obligations maturing after the window has closed.

The judgment the record supports about these practices is an even handed one, and the judgment begins with the observation that drafting to the score is not deception. The estimating office publishes the year by year figures behind every headline total, so a delayed start appears as years of low cost followed by years of full cost, an expiration appears as a series that ends inside the window, and a timing shift appears as a movement between adjacent years. Estimate letters flag the devices as well, drawing attention to the features of a design that drive the figure, and a reader of the complete estimate encounters the same calendar facts that the drafters used in constructing the bill. The appropriate criticism of a particular device concerns its fiscal substance, whether an expiration disguises a permanent intention, whether a patch accumulates a cost larger than the repair it avoids, or whether a program’s finances cohere beyond the window, and that criticism proceeds best from the disclosures the estimate itself supplies. The skill, for every reader of a score, lies in reading the estimate’s own disclosures rather than its headline, tracing the year by year pattern to the drafting features that produced the pattern, and asking of any favorable total how much of the favor belongs to the policy and how much belongs to the calendar.

The estimate anatomy gives a reader a single reference for the moving parts that the preceding sections developed in sequence. Each row isolates one component of a typical major bill, names the body that scores that component, states the assumption the baseline applies to that component, records what the ten year window counts, and identifies the drafting feature that moves the recorded number most directly. Placed after the account of construction and drafting, the table serves as a findable artifact for use during an actual reading, a compact map from a line in an estimate letter back to the institution, the convention, and the calendar choice that produced that line. For a provision involving credit, mandates, or macroeconomic feedback, the row also separates the federal budget effect from costs and effects recorded elsewhere in the estimate letter or in a separate analytical statement altogether for that bill.

The Estimate Anatomy

Component	Scoring body	What the baseline assumes	Inside the window	Where drafting moves the number
Revenue title	Joint Committee on Taxation staff	Expiring tax provisions expire on schedule	The full ten year revenue effect is counted	A sunset or a phase-in reshapes the counted loss
Mandatory spending title	Congressional Budget Office	Large expiring programs are assumed to continue	The ten year outlay path is counted	A delayed start pushes cost toward the edge of the window
Discretionary authorization	Congressional Budget Office	Appropriations are assumed to continue with inflation	Spending subject to appropriation is reported apart from direct spending	The authorized level sets the scored requirement
Federal credit program	Congressional Budget Office	The subsidy cost is recorded up front under credit reform	The lifetime subsidy is counted at commitment	Fees and default terms move the subsidy rate
Mandates on states and private parties	Congressional Budget Office	No entry exists until a mandate is enacted	Costs are tested against the mandate thresholds	Threshold design decides how hard the statement binds
Macroeconomic feedback	Both estimating bodies for major legislation	The economy is held at the forecast path in a conventional estimate	Feedback is counted only in a dynamic estimate	Deficit financing in the design feeds back through interest costs

Dynamic Analysis and the Contest Over Feedback

A conventional estimate and a dynamic estimate begin from the same bill, the same budget window, and much of the same evidence about how people and firms respond to a change in law. The difference lies in the boundary drawn around those responses. A conventional estimate incorporates a behavioral layer. Taxpayers may change the timing of income or deductions, shift activity from one taxed form to another, or alter participation in a program whose rules have changed. Those responses enter the estimate because they affect revenues and spending directly, even though the economy as a whole is held at the level projected in the economic forecast used by the estimating office. Gross domestic product, aggregate employment, the capital stock, and other macroeconomic variables remain on the forecast path. The bill changes behavior inside that economy, but the estimate does not allow the bill to change the size of the economy itself.

A dynamic estimate relaxes that convention. The behavioral layer remains, and a second layer is added on top of it. If a bill is projected to change output, employment, the capital stock, or interest rates, the estimate traces those macroeconomic changes back into the budget. Higher or lower output changes the tax base, and a change in employment changes payroll and income tax receipts. A change in the capital stock changes investment and the income generated by capital. A change in interest rates changes the cost of federal borrowing and therefore changes debt service on the existing stock of debt and on any new borrowing that the bill produces. The feedback loop runs from law to economy and from economy back to the budget, and the dynamic score is the conventional result after that loop has been run through the estimating model.

The contrast, stated precisely, is not between an estimate with responses and an estimate without responses. Both conventions assume that people respond. The contrast is between two scopes of response. The conventional scope stops at responses that can be measured against a fixed macroeconomic background. The dynamic scope extends to responses that alter the background itself. That extension grows large as a matter of practice, because the macroeconomic background is not observed in the same way as a statutory rate or an eligibility rule. Output after a major tax or spending change depends on judgments about labor supply, saving, investment, and the way deficits affect interest rates and private capital formation. Different defensible models produce different paths for those variables, and the budgetary feedback inherits that spread.

What does a dynamic estimate add to a conventional one?

A dynamic estimate adds the budgetary consequences of a bill’s effects on output, employment, the capital stock, and interest rates to the conventional score, tracing higher or lower economic activity back into revenues and spending rather than holding macroeconomic aggregates fixed at forecast levels.

The institutional history of that broader question in the tax field runs through the staff of the Joint Committee on Taxation. The staff has supplied macroeconomic analyses of major tax proposals since 2003, after a House rule change in that year required the analysis. Those analyses examined how a proposal might affect the wider economy. Their status mattered as much as their content. A macroeconomic analysis from the Joint Committee staff was presented as discussion, accompanied by ranges of model results, and it stood outside the official estimate that governed the legislative process. The official figure remained the conventional estimate, and the macroeconomic material did not replace the number used for budget enforcement.

That separation reflected a practical distinction between exploration and scoring. A range can show that modeled feedback varies with assumptions about how strongly labor and capital respond, how quickly investment adjusts, and how deficits interact with interest rates. A range can also preserve disagreement among models instead of compressing that disagreement into a single figure.

The arrangement changed in the House in January 2015. The House adopted a rules package that replaced its former provision with Rule XIII, clause 8. The new rule directed the Congressional Budget Office and the staff of the Joint Committee on Taxation, to the extent practicable, to incorporate the budgetary effects of changes in economic output, employment, capital stock, and other macroeconomic variables into estimates for major legislation. The qualifier mattered. The obligation applied to the extent practicable, a phrase that left room for the estimating offices to judge whether the data, the models, and the schedule for a particular bill allowed a dynamic component to be incorporated responsibly.

The same rule imposed a disclosure discipline alongside the estimating duty. The offices were to identify the critical assumptions behind an estimate and the sources of data used in it. Where practicable, an estimate was also to include a qualitative assessment of long term effects. Disclosure performed a different function from incorporation. Incorporation changed the number. Disclosure changed the ability of legislators and outside readers to inspect the number, to see which assumptions carried the result, and to compare the treatment of one bill with the treatment of another. A qualitative assessment of long term effects reached beyond the budget window in words, without converting those distant effects into a precise figure inside the window.

The rule defined major legislation through three doors, and the definition decided which bills received the new treatment. The first door was size. A bill qualified if its conventional estimate showed a gross budgetary effect of at least 0.25 percent of projected gross domestic product in any year, a threshold that corresponded to about $43 billion at 2014 output levels. The test used the conventional estimate as the trigger, so the older method determined whether the newer method applied. The second door was designation for mandatory spending legislation by the chair of the House Budget Committee. The third door was designation for revenue legislation by the chair or vice chair of the Joint Committee on Taxation. A bill could therefore enter the dynamic process because it was large, or because a designated official placed specified categories of legislation into the process. Discretionary spending legislation was excluded, which kept bills funded through the appropriations process outside the rule even though appropriations bills can be large in absolute terms.

Directives in budget resolutions reinforced the House rule during the same period. The budget resolutions agreed to for fiscal years 2016 and 2018 carried provisions requiring dynamic estimates in both chambers for the same two Congresses covered by the House rule. The directives widened the reach of the 2015 approach beyond the House rule alone. The House rule and the budget resolution provisions together formed the framework under which dynamic scoring operated during the 114th Congress and the 115th Congress.

Under that framework, dynamic estimates for the first time carried a point estimate. Earlier macroeconomic analyses from the Joint Committee staff had presented discussion and ranges of model results outside the official estimate. The 2015 framework moved feedback inside the estimate and expressed the result as a single figure incorporating that feedback. A point estimate has administrative advantages that a range lacks. Budget enforcement operates on numbers, and a single figure can be entered into scorekeeping, compared with a budget target, and tested against a rule. A single figure can also look more certain than the modeling behind it, because the figure does not display the alternative outcomes that other assumptions or other models would have produced.

The rule was in effect through 2018. Its operation covered the 114th Congress, which sat in 2015 and 2016, and the 115th Congress, which sat in 2017 and 2018. The rule lapsed with the 115th Congress at the close of 2018. Stated as a dated fact, the House dynamic scoring rule of January 2015 governed estimates during those two Congresses and did not govern estimates in 2019 or 2020.

The tax legislation enacted in 2017 supplied the most closely watched worked example of the framework. The example is useful because the conventional and dynamic figures were both published for the same act, allowing a direct comparison of what feedback changed and what feedback left intact. The staff of the Joint Committee on Taxation prepared the conventional estimate for the act’s revenue provisions. That estimate put the revenue loss at about $1.46 trillion over the budget window, before added interest costs. The figure represented the conventional scope. Behavioral responses of the familiar kind were incorporated, and macroeconomic aggregates remained at forecast levels, so the $1.46 trillion loss was the measured cost of the act’s rate reductions, deductions, credits, and other revenue provisions against a fixed economic background.

The macroeconomic analysis of the same act estimated that feedback from higher output would raise revenues by about $451 billion over the window. Gross domestic product was estimated to be about 0.8 percent higher on average as a result of the act. The $451 billion figure was not a separate benefit recorded outside the budget. It was revenue feedback, the portion of the conventional loss that higher output was projected to offset through larger tax bases. The analysis also accounted for added interest costs. A larger deficit increases federal borrowing, and the interest on that borrowing is itself federal spending. Those added interest costs clawed back part of the feedback. After the feedback was counted and the added interest was charged against it, the macroeconomic analysis left a net deficit increase on the order of $1 trillion on the dynamic basis.

The comparison disciplines several loose ways of describing dynamic scoring. Dynamic analysis did not make the 2017 act free in the published estimate. A revenue loss of about $1.46 trillion on the conventional basis became a net deficit increase on the order of $1 trillion on the dynamic basis, a substantial reduction in measured cost and a substantial remaining cost at the same time. Feedback moved the figure by hundreds of billions of dollars, large enough to matter for budget enforcement and for political argument, and feedback fell well short of offsetting the full conventional loss in that analysis. The two numbers were not two measurements of exactly the same object, because each number embodied its convention.

The 2017 act also showed how the estimating system and the enforcement system interlock. The act was moved through the reconciliation process, and its individual provisions were set to expire after 2025. The reason lies in the Byrd rule, which bars a reconciliation bill from increasing deficits beyond the budget window. A permanent set of individual provisions with the act’s revenue profile would have produced deficit increases outside the window and collided with that prohibition, so the provisions were given an expiration date inside the window, as described in the Byrd rule article. Scoring and enforcement met at exactly that point. The estimate measured the deficit path, and the Byrd rule converted the shape of that path into a drafting decision, trading permanence for compliance with the reconciliation rules. The sunset was the mechanical consequence of fitting a large revenue bill into a procedure that forbids deficit increases beyond the window.

Against that record, the contest over dynamic estimating concerned three separate questions that public argument often compressed into one. The first question asked whether a complete estimate must include macroeconomic feedback at all. The second question asked who decides which bills receive dynamic treatment and which models produce the feedback. The third question asked whether growth effects belong mainly to the estimating convention or to the design of the bill being estimated.

Douglas Holtz-Eakin, a former director of the Congressional Budget Office, argued the completeness position. In that argument, the macroeconomic consequences of a sufficiently large bill are part of the economics of the bill, and an estimate that ignores those consequences is incomplete rather than neutral. The force of the position rests on scale. A bill large enough to change incentives across the labor market, the capital market, and the federal borrowing path presents a different case in Holtz-Eakin’s argument. Holding the economy fixed for such a bill does not avoid an assumption. The fixed economy is itself an assumption, one that sets the macroeconomic effect at zero for scoring purposes. From that premise, the completeness argument treats dynamic estimating not as an embellishment added to a finished score but as the removal of an artificial zero from the score.

Robert Greenstein of the Center on Budget and Policy Priorities warned about a different feature of the 2015 framework. Greenstein warned that the rule’s designation mechanism and the choice among models invited pressure on the estimators and could tilt results toward the preferences of the majority that wrote the rule. The designation concern followed from the structure of the major legislation definition. Two of its three doors depended on designation by specified chairs, so the decision to apply dynamic treatment to mandatory spending or revenue legislation below the size threshold rested with officials of the majority. The model concern followed from the spread of outcomes that macroeconomic models produce. If several models or assumption sets are defensible, the selection among them can move the feedback figure by large amounts, and a process that leaves that selection exposed to political pressure places estimators in a difficult institutional position. In Greenstein’s warning, the architecture of the rule, designation plus model choice, created points at which the preferences of the rule’s authors could enter the production of the number itself.

William Gale of the Brookings Institution and Andrew Samwick located the growth question in a different place. Writing in 2014, Gale and Samwick argued that reforms improving incentives, reducing existing subsidies, avoiding windfall gains, and avoiding deficit financing carry the most auspicious growth effects. That formulation directs attention to the design of a bill rather than to the label of the estimating convention applied to it. A reform that improves incentives changes the reward to work, saving, or investment at the margin. A reform that avoids windfall gains denies a benefit for activity that would have occurred anyway, so public money is not spent purchasing no change in behavior. In the Gale and Samwick account, a bill with those design features has a stronger claim to growth effects than a bill without them, whatever estimating convention is used, and a bill lacking those features does not acquire growth effects because an estimator applies a dynamic label to it.

The three positions do not answer one another directly, because each addresses a different stage of the process. Holtz-Eakin addressed the scope of the estimate and argued that scale makes macroeconomic effects part of the object being measured. Greenstein addressed the governance of the estimate and warned that designation and model choice expose the estimator to pressure from the majority that designed the rule. Gale and Samwick addressed the substance being estimated and argued that growth follows design features such as improved incentives, reduced subsidies, no windfalls, and no deficit financing. The 2015 rule joined all three stages in a single mechanism, so disagreement at any one stage became disagreement about dynamic scoring as a whole.

The estimating bodies answered the methodological problem through practice rather than through polemic. The Congressional Budget Office and the staff of the Joint Committee on Taxation presented dynamic results as model dependent. Their analyses published the critical assumptions behind an estimate and, in their analyses, presented a range of outcomes across models alongside any point figure. Publication of assumptions allowed a reader to identify the channels through which feedback was supposed to run, whether through labor supply, capital formation, or the interest cost of added debt. A range across models allowed a reader to see how much of the feedback survived a change in modeling approach and how much belonged to a particular model.

By the close of the period covered by the rule, dynamic analysis had acquired a settled place in congressional estimating practice and an unsettled place in congressional argument. The settled place was procedural. Macroeconomic analysis had been required for major tax proposals since 2003, dynamic incorporation had been required for major legislation under the House rule of 2015 and under the budget resolution directives for fiscal years 2016 and 2018, and the 2017 tax act had shown the arithmetic of feedback at full scale, with about $451 billion of projected revenue feedback set against a conventional revenue loss of about $1.46 trillion and added interest costs reducing the gain. The unsettled place concerned interpretation, and the three attributed positions each survived the lapse of the rule at the end of 2018. A convention that holds the economy fixed answers a narrow question consistently. A convention that lets the economy move answers a broader question dependently on models, assumptions, and institutional safeguards. The choice between those conventions, as the record from 2003 to 2018 presents it, is a choice about what a score is for, and the figures produced under each convention carry that choice inside them.

How Accurate the Estimates Prove and Why They Are Attacked

Every cost estimate issued by the Congressional Budget Office enters a political environment that rewards certainty and punishes qualification. A sponsor wants a number that fits inside an announced deficit target, a committee allocation, or a promise made during a campaign. An opponent wants a number that exposes a hidden cost or a coverage loss the sponsor preferred to leave vague. The estimate itself is built for a narrower purpose. It measures the change in federal revenues, outlays, and deficits that would follow if a particular text were enacted and if the economy performed as the office had forecast. That construction gives the number its usefulness and explains much of the disappointment that later attaches to it. Judging accuracy therefore requires separating the parts of a later difference that belong to the estimate itself from the parts that belong to economic movement, legislative change, and administrative choice.

That separation is not an excuse invented by an office protecting its reputation. It is the method the office itself applied in public, repeatedly, across decades of evaluating its earlier projections. The practice deserves attention as an institutional fact before any individual figure is examined, because political actors rarely grade themselves in the same way. The budget office, by contrast, turned its own record into a published subject, with named reports, dated comparisons, and explanations that identified where its expectations had missed. They decomposed the differences, source by source, and provided the material by which readers could judge the office on evidence rather than on the convenience of any single estimate.

Does the budget office grade its own work?

Yes. The Congressional Budget Office published formal evaluations of its earlier baselines, economic forecasts, and revenue projections, comparing what it had projected with what occurred and explaining the sources of the differences. Those reports appeared over decades and placed the office’s record beside the records of other forecasters.

The fiscal year 2018 evaluation offers a compact example of the method and of the scale of error involved in a year without a major economic turning point. In its December 2018 report, The Accuracy of CBO’s Baseline Estimates for Fiscal Year 2018, the office compared the projections it had published in June 2017, adjusted for legislation enacted after those projections were prepared, with the amounts the fiscal year actually produced. The estimate of discretionary outlays came in 0.8 percent above the amount actually spent, a small proportional difference applied to a very large category of spending, measured against projections prepared before the fiscal year began. The estimate of net interest came in 2.8 percent below the outcome, a larger proportional miss in the opposite direction, and the office assigned most of that difference to a specific mechanism rather than to a general failure to add the accounts correctly. Inflation had run higher than expected, and higher inflation raised payments on inflation protected securities, the federal obligations whose principal adjusts with price changes. A forecast input, the path of prices, moved; a category of outlays tied mechanically to that input moved with it; and the interest estimate, built on the earlier price path, fell short of the payments the Treasury made.

The same report recorded a Medicaid underestimate with a different kind of cause, one that sits between forecasting and behavioral judgment. The office had observed a slowdown in the program and had treated that slowdown as lasting. The slowdown proved temporary. Spending therefore ran above the path implied by an assumption that an observed pattern would continue. The choice between treating such a change as a new pattern or as a brief disturbance cannot be escaped by better arithmetic, since the arithmetic begins only after the judgment about persistence has been made. The significance of the report lies partly in its willingness to name that judgment in print, attached to a program as large and politically sensitive as Medicaid, rather than burying the difference inside an unexplained residual.

The 2018 figures, taken alone, do not settle the larger question of accuracy, because a single calm year is the easiest test a forecaster can face. The office’s broader evaluations addressed harder conditions, and their central finding was consistent. The largest errors clustered at particular moments in the economic cycle, and that clustering carries the main lesson of the accuracy record.

In The Economic Forecasting Record, 2019 Update, published in October 2019, the office assessed its economic forecasts, the projections of output, income, prices, and related variables on which every budget estimate rests. The report found the errors largest around business cycle peaks, just before recessions begin. Near a peak, the observable data describe an expanding economy, and models estimated from recent experience project much of that performance forward. The turning point is the moment at which the recent past becomes a poor guide and the moment least visible in data published with lags. A central path prepared just before a recession will generally embody more of the expansion than the economy delivers. The office placed its forecasting record beside those of the Administration and of private sector forecasters in the same report, presenting the comparison as part of the evaluation rather than confining the examination to its own numbers in isolation. That frame converts the question into whether its errors were distinctive in size or direction.

The revenue evaluations reached a matching conclusion from the tax side of the budget, with the sharpest figures in the office’s retrospective literature. The largest errors appeared in the projections released in early 1982, in early 1990, in early 2001, and in early 2008, each made near the start of a recession. In each case the office overestimated budget year revenues by between about 9 percent and about 25 percent. Revenues respond with particular force to declines in capital gains, corporate profits, and high income earnings, all of which can contract faster than the broader economy. The evaluations also identified a timing mechanism that pushed the largest discrepancies into the year after each recession began. Tax payments lag economic activity. Estimated payments, final settlements, and refunds reflect income earned in an earlier period, so the full revenue consequence of a downturn registers in collections after the downturn itself has started. A projection prepared at the onset therefore misses twice, first because the recession was not in the forecast and then because the payment lag delivers the recession’s revenue effects into the following fiscal year.

Against that record of large cyclical misses, the same evaluations found the overall accuracy of the office’s revenue projections similar to that of the Administration’s projections. The finding cuts in two directions. It denies the office any claim to a method so superior that its revenue numbers escaped the errors common to official Washington. It also denies the charge that the office’s numbers were uniquely unreliable, since the executive branch forecaster produced projections of comparable overall accuracy across the same difficult history. Turning points defeated nearly every central projection, and the office’s evaluations said so with its own figures included in the indictment.

A careful reader of these reports keeps three categories distinct, because each points to a different remedy and a different judgment about the office. The first category is forecast error, in which the economy moved away from the path the estimate assumed. The inflation surprise behind the fiscal 2018 interest underestimate belongs there, as do the recession driven revenue overestimates of 1982, 1990, 2001, and 2008. The second category is legislative drift, in which Congress changed the policy after the estimate was prepared. A committee that raises a benefit, narrows an eligibility rule, delays an effective date, or substitutes a different tax provision after receiving a score has altered the object being measured, and the later fiscal outcome measures the enacted law, not the earlier draft. Treating the difference as an error by the estimator confuses authorship. The third category covers classification and behavioral surprise, in which the text left a question open that implementation or human response answered in an unexpected way. Whether a stream of payments counts as federal receipts and outlays is a classification question of the kind that decided the presentation of the Clinton health plan in 1994, and how employers, insurers, patients, or taxpayers adjust to a new rule is a behavioral question embedded in nearly every major estimate.

The conclusion the office’s own reports support is that the largest errors are economic rather than arithmetical. Arithmetical error would imply that the machinery connecting assumptions to totals was unreliable, and it would justify discarding scores even when the underlying economic outlook was sound. Economic error implies something different. It implies that the score faithfully translates a forecast and a legislative text into budget totals, while sharing the forecast’s vulnerability to the events that defeat all forecasters. Congress can respond to that vulnerability by demanding sensitivity analysis, by phasing provisions, or by revisiting legislation after outcomes arrive. What Congress cannot coherently do is demand that an estimate eliminate the uncertainty of the economy itself and then denounce the estimate as defective when a recession arrives on a schedule no published forecast had contained.

The distinction between a score and a prediction gives that conclusion its precise form. A score is not a prediction of what will happen. An estimate measures the change from the baseline if the bill is enacted as written and the economy follows the forecast. Both conditions sit inside the definition of the number. Much of what later gets called estimation error consists of legislation changed after the estimate, economic movement no forecaster caught, or administrative practice the text did not determine. A comparison of a ten year estimate with a ten year outcome that ignores those causes misreads what the number was at the moment it was issued. The estimate was a conditional measurement, offered so that legislators could compare the fiscal consequences of alternative texts on common assumptions before choosing among them.

None of this places any individual estimate beyond criticism. Methodology can be challenged on its treatment of labor supply, of insurance take up, or of the response of state governments to a federal matching formula. Baseline conventions can be challenged as well, particularly where statute directs the office to apply assumptions about provisions expiring or continuing on terms Congress has repeatedly altered in the past. Those criticisms name an assumption, propose an alternative, and can be answered and tested against evidence. An institution that publishes the decomposition of its own misses has given its critics the vocabulary for serious criticism, and the serious criticism in the record has overwhelmingly concerned method and convention rather than the loyalty of the people producing the numbers.

The enforcement setting in which scores operate raises the stakes, because an estimate does not merely inform debate; in defined circumstances it determines whether a procedural barrier stands or falls. When a measure moves under budget enforcement rules, the estimate is measured against committee allocations, against aggregate spending and revenue levels, and, in the case described by the reconciliation process explained, against the instructions that directed committees to produce specified changes in spending and revenues. A score on one side of an instruction permits the measure to proceed under the protection of the reconciliation procedures; a score on the other side exposes it to points of order. Participants therefore experience the estimate as a gate, and resentment at a gatekeeper whose gate swings against a favored bill is a predictable feature of the system Congress designed. That resentment is evidence about the importance of the gate rather than evidence about the politics of the person tending it.

The claim that the estimates are political deserves to be stated fairly before it is tested, because a weak version of the claim is obviously true and a strong version requires evidence. In the weak version, an estimate is political because it has political consequences, because it can strengthen or destroy a coalition, and because every participant understands those effects before the number appears. In the strong version, the office shapes its numbers to help one party or one president, shading assumptions toward the outcome its leadership prefers, and the pattern of its estimates should therefore favor a consistent side. That version makes a prediction of its own, and the historical record supplies the test. An office captured by one party should spare that party’s prized bills and burden the other party’s. The record from 1994 through 2017, read in chronological order, fails that test in turn.

In February 1994, under Director Robert Reischauer, the office determined that the premiums flowing through the health alliances in President Clinton’s health plan would be recorded as federal revenues and outlays. The determination was a classification decision, not a forecast of medical costs, and its consequences were severe for the plan’s financing presentation. Premiums that the administration had depicted as private transactions outside the federal budget would instead swell both sides of the federal ledger, enlarging the measured size of government. The ruling damaged the plan in the year the plan failed. The office applied a classification judgment about the degree of federal control exercised through the alliances, reached a conclusion the White House did not want, and published it at the moment of maximum political cost to the administration.

In July 2009, Director Douglas Elmendorf testified that the health bills then in Congress would not reduce the growth of federal health care spending. The testimony complicated the Obama administration’s central claim for its legislation, which held that expanding coverage and restraining long term cost growth belonged to the same reform. The testimony stated the office’s judgment in terms that opponents of the bills quoted immediately and that the administration had to answer. The institutional posture was the same as in 1994, under a different Democratic president, separated by fifteen years and by a change in directors.

The sequence then turns, and the turn gives the record its evidentiary force. In March 2010, the office estimated that the enacted health legislation would reduce deficits by 143 billion dollars over the period from 2010 to 2019. That estimate aided the bill’s passage, supplying the administration and the congressional majority with an authoritative figure that answered the fiscal objection at the center of the opposition’s case. An explanation built on partisan intent cannot comfortably hold both the 2009 testimony and the 2010 estimate. An explanation built on method holds both without strain. The bills changed between the testimony and the final estimate; the enacted text contained specific revenue provisions, Medicare payment changes, and an implementation schedule that the office scored as written. The variable that changed was the legislation, not the allegiance of the estimator.

In July 2012, the office estimated that repealing the health legislation would increase deficits by 109 billion dollars over the period from 2013 to 2022. The estimate ran against the position of the House majority that had voted for repeal. Repeal, in the office’s accounting, would restore spending and surrender revenues the enacted law had set in place, and the net effect was a larger deficit even though repeal eliminated programs the majority opposed. A Republican led House confronted an estimate that raised the fiscal cost of its own priority, two years after a Democratic led Congress had received an estimate that lowered the measured cost of its priority, in each case on the same statute scored in opposite directions. Symmetry of that kind is the natural signature of a baseline method applied consistently.

The 2017 estimates completed the pattern by turning the office’s authority against a Republican bill in the full view of a Republican Congress. In March 2017, the office estimated that the American Health Care Act would leave 24 million more people uninsured by 2026 and would reduce deficits by 337 billion dollars. The coverage figure dominated the debate that followed, while the deficit figure simultaneously recorded that the bill achieved substantial budgetary savings. In May 2017, the office’s estimate of the amended bill put the uninsured figure at 23 million and the deficit reduction at 119 billion dollars. The amendments had moved both numbers, narrowing the coverage loss slightly and shrinking the savings considerably, and the estimate reported both movements without selecting the one convenient to either side. The bill passed the House and failed in the Senate. Across twenty three years, the office had constrained a Democratic president’s health plan through classification, contradicted a Democratic administration’s cost claim in testimony, certified a Democratic law’s deficit reduction, priced a Republican repeal as a deficit increase, and attached a politically lethal coverage number to a Republican replacement bill. Each party’s prized health legislation met the office as an obstacle at the moment that party held power and needed the number most.

The appointment record reinforces the same conclusion from the institutional side. The first director, Alice Rivlin, served from 1975, and the directorship thereafter passed through appointments made under Congresses controlled by both parties. Douglas Elmendorf took office in 2009 under a Democratic Congress, the same director whose testimony that year complicated the Obama administration’s health care argument. Keith Hall took office in 2015 under a Republican Congress. Phillip Swagel took office in June 2019 with the Senate under Republican control and the House under Democratic control, an appointment completed under divided government. A directorship that rotates through selections made by different congressional majorities, and whose occupants proceed to issue estimates damaging to the appointing side’s successors and allies, fits poorly with a theory of partisan capture. Those constraints do not make error impossible. They make sustained favoritism difficult to hide, because every estimate is documented and every inconsistency between the treatment of one party’s bill and the other’s would appear in print for the disadvantaged side to cite.

The attacks themselves follow a rhythm that the partisan theory cannot explain and the inconvenience theory explains completely. Criticism of the office intensifies when an estimate blocks a bill, recedes when the same office certifies a favored measure, and returns with the next inconvenient number, with the identity of the critics rotating as control of Congress and the White House rotates. The office’s estimates have constrained both parties’ prized bills, its directors have been appointed under both parties’ Congresses, and the substantive criticisms that survive scrutiny concern methodology and the statutory baseline conventions the office is required to apply. A challenge to the assumed response of employers to a coverage mandate is a methodological criticism. A challenge to a baseline convention that Congress wrote into statute is a criticism Congress could answer by amending the statute. Both criticisms identify something real, and both locate the remedy in a place more specific than the motives of the estimator, in a better model, a better body of evidence, or a better written budget law.

That closing judgment is the more precise and the more interesting critique the record supports. Partisanship, as an explanation, predicts a pattern the estimates do not show and excuses legislators from examining the rules they themselves enacted. Methodology and baseline conventions, by contrast, are written down, debatable in detail, and connected directly to the numbers that decide whether legislation proceeds, which is why the debate surveyed in budget process reform and its impact returns repeatedly to the rules the estimates enforce rather than to the personnel who compute them. Congress chose the baseline, chose the conventions that define it, and chose an office to apply them without regard to which bill benefited in a given month. The estimates that resulted were sometimes wrong in the ways the office’s own evaluations documented, and sometimes decisive in the ways the health care record from 1994 to 2017 displayed, against whichever party held the pen. Accuracy, judged by the separated causes the evaluations teach readers to separate, is a record of ordinary professional fallibility under extraordinary economic uncertainty. The attacks, judged by their timing, are a record of something simpler. They arrive when the number bites.

Reading an Estimate Like an Auditor

A disciplined reading of a cost estimate begins with authorship. The first diagnostic question asks which body produced each part of the number. Revenue effects on income, estate and gift, excise, and payroll taxes belong to the staff of the Joint Committee on Taxation under section 201(f), incorporated into the estimate that the Congressional Budget Office issues. Outlay effects, mandate costs, credit subsidy costs, and the residue of receipts involving unemployment insurance taxes, customs duties, miscellaneous fees, and Federal Reserve earnings belong to the budget office. A dispute addressed to the wrong body wastes its force. Attribution therefore precedes argument, and the credit line inside an estimate letter supplies attribution on the face of the document.

The second question asks which baseline conventions did the quiet work. A total that appears generous toward a reauthorized program may reflect the section 257 instruction that a large expiring program continues inside the baseline, so reauthorization adds little measured cost. A total that appears severe toward an extended tax provision may reflect the opposite instruction, that receipts follow enacted law and an expiring provision expires on schedule, so extension registers a full revenue loss. None of those results requires a hidden preference by an estimator. Each follows a convention Congress enacted in the Balanced Budget and Emergency Deficit Control Act of 1985, and a serious objection to any result names the convention and addresses the statute that contains the convention.

The third question asks what the window includes and what the estimator said about the years beyond the window. A ten year total records a decade and excludes the eleventh year onward by construction, so a delayed start, a phase in, and an expiration can lower a total without lowering the annual cost of a fully operating provision. A careful reading checks the year by year pattern behind a total, looks for a second decade characterization expressed as direction and rough magnitude, and checks the result under the test that fails a measure increasing deficits by more than $5 billion in any of the four consecutive ten year periods beyond the window. The fourth question asks whether the figure is conventional or dynamic, and, for a dynamic figure prepared under the House rule in effect from 2015 through 2018, which assumptions about labor supply, capital formation, and interest costs the feedback rests on, and whether a range across models accompanies the point figure.

The fifth question asks what the accuracy record implies about the confidence a number deserves. The December 2018 evaluation of fiscal year 2018 found discretionary outlays estimated 0.8 percent above the outcome and net interest estimated 2.8 percent below the outcome, with higher inflation raising payments on inflation protected securities. Revenue evaluations found the largest errors near recession starts, with budget year revenues overestimated by between about 9 percent and about 25 percent in the projections released in early 1982, in early 1990, in early 2001, and in early 2008. That record supports confidence in method and humility about magnitude.

Working through those five questions on a real estimate produces a file of material rather than a single impression: a statute text marked by provision, a baseline convention noted beside each major figure, a window pattern traced year by year, and a chronology of amendments that changed a score between committee action and enactment. A reader building that file can keep your statute notes, citations, and case chronologies together free on VaultBook, holding the section 257 conventions, the section 201(f) attribution, and the dated accuracy findings in one organized place beside the estimate that depends on those authorities. A reader consolidating the wider framework, from the budget resolution and reconciliation to the civics of separated powers that placed an estimating office inside the legislative branch, can practice and revise US government and civics material on ReportMedic, testing recall of the institutions and procedures that give a score procedural force. The figure governs passage because the process assigns the figure that power, and the figure earns examination because a statute, a convention, a window, and a documented record stand behind every digit.

Frequently Asked Questions

Q: How does CBO score a bill?

The Congressional Budget Office measures the change a bill would produce against the baseline, reporting effects on revenues, outlays, and the deficit year by year across the budget window. The office scores the legislative text as written, builds evidence from administering agencies and comparable programs, and incorporates behavioral responses by taxpayers, beneficiaries, and firms while holding the overall economy at the level of the published forecast. Revenue effects from the covered taxes enter from the staff of the Joint Committee on Taxation. Direct spending, spending subject to appropriation, credit subsidy costs, and mandate costs appear as distinct components, so the combined deficit figure preserves the separate character of each part behind the total.

Q: What is the CBO baseline built on?

The baseline is a projection of budget authority, outlays, revenues, and the deficit or surplus under laws already enacted, published in The Budget and Economic Outlook under section 202 of the Congressional Budget Act of 1974. Conventions in section 257 of the Balanced Budget and Emergency Deficit Control Act of 1985 govern construction. Direct spending and receipts operate as enacted law specifies, funding for entitlement authority is assumed adequate for required payments, discretionary appropriations continue at enacted levels adjusted for inflation, and specified treatments govern expiring programs and expiring tax provisions. A cost estimate measures distance from that published path rather than the absolute size of a program.

Q: Are CBO assumptions set by statute?

The governing baseline assumptions are set by statute. Section 257 of the Balanced Budget and Emergency Deficit Control Act of 1985 prescribes how the baseline treats direct spending, receipts, expiring programs, and discretionary appropriations, and the Congressional Budget Office applies those prescriptions rather than selecting a preferred forecast for each estimate. Estimator judgment operates inside that frame, on questions involving enrollment, unit costs, behavioral responses, and plausible administration of a particular text. A dispute about a statutory convention therefore addresses Congress, which enacted section 257 and retained authority to amend section 257. A dispute about judgment inside the frame addresses the estimating office and the evidence behind a modeling choice.

Q: Who appoints the CBO director?

Section 201(a) of the Congressional Budget Act of 1974, codified at 2 U.S.C. 601(a), vests appointment jointly in the Speaker of the House and the President pro tempore of the Senate. The appointing officers act after considering recommendations from the House Budget Committee and the Senate Budget Committee. Selection is made without regard to political affiliation and solely on the basis of fitness to perform the duties of the office. The director serves a term of four years, either house may remove the director by resolution, and reappointment carries no statutory limit. Practice added alternation between the two budget committees in taking the lead on recommendations across successive appointments.

Q: Does CBO or JCT score revenue provisions?

The staff of the Joint Committee on Taxation scores the principal revenue provisions. Under section 201(f) of the Congressional Budget Act of 1974, estimates from that staff are the official estimates for legislative changes to income taxes, estate and gift taxes, excise taxes, and payroll taxes, and the Congressional Budget Office incorporates those estimates into a cost estimate rather than producing a rival revenue figure. The budget office estimates a residue of receipts connected to programs and economic conditions, including unemployment insurance taxes, customs duties, miscellaneous fees, and the earnings of the Federal Reserve System. Attribution inside a combined total follows that statutory division of labour.

Q: What is dynamic scoring at CBO?

Dynamic scoring adds the budgetary effects of macroeconomic feedback to a conventional estimate. A conventional estimate incorporates behavioral responses while holding gross domestic product and other macroeconomic aggregates at forecast levels. A dynamic estimate traces a bill through output, employment, the capital stock, and interest rates, and then traces changed activity back into revenues, spending, and debt service. House Rule XIII, clause 8, adopted in January 2015 and in effect through 2018, directed the estimating bodies, to the extent practicable, to incorporate such feedback into estimates for major legislation, defined in part by a gross budgetary effect of at least 0.25 percent of projected gross domestic product in any year.

Q: How long is the CBO scoring window?

The standard scoring window measures ten years. Section 402 of the Congressional Budget Act of 1974 originally framed estimates over five years, a horizon suited to the budget process created in that statute. During the 1990s and 2000s, budget resolutions, budget scorekeeping, reconciliation instructions, and points of order moved to a ten year horizon, and estimating practice moved with the rules that consumed estimates. An estimate confined to five years could not answer an enforcement test written against ten. By the 2010s, a cost estimate presented year by year figures and a ten year total as a matter of course, matching the span of the published baseline.

Q: How accurate are CBO projections?

The Congressional Budget Office grades earlier projections in published evaluations. The December 2018 report on fiscal year 2018 found discretionary outlays estimated 0.8 percent above the amount spent and net interest estimated 2.8 percent below the outcome, with higher inflation raising payments on inflation protected securities. Revenue evaluations found the largest errors in projections released in early 1982, in early 1990, in early 2001, and in early 2008, each near the start of a recession, with budget year revenues overestimated by between about 9 percent and about 25 percent. Overall revenue accuracy resembled the accuracy of Administration projections across the same history.

Q: What is the difference between a cost estimate and a prediction?

A cost estimate is a conditional measurement rather than a prediction of events. The measurement states the change from the baseline on two stated conditions, enactment of the text as written and performance of the economy along the published forecast path. A prediction, by contrast, would weigh whether Congress enacts a text, whether later Congresses amend a law, whether administrators choose one permissible course over another, and whether a recession interrupts the forecast. Much of the difference between a ten year estimate and a ten year outcome belongs to those excluded conditions. Legislative change after an estimate alters the object measured, and economic movement alters the background assumed, without either change describing a failure of arithmetic.

Q: Why do tax bills carry expiration dates?

An expiration date confines a tax provision to part of the budget window, so an estimate records revenue effects only for the years of operation inside the window and records no cost for the provision beyond the scheduled ending. A sunset can also carry a procedural function. Under the Byrd rule, a reconciliation bill faces a prohibition on increasing deficits beyond the window, and an expiration inside the window answers that prohibition. The individual provisions of the tax act of 2017 expire after 2025 for that reason. The baseline adds a further consequence, because receipts follow enacted law and an expiring provision expires on schedule, making a later extension score as a fresh revenue loss.

Q: What does subject to appropriation mean in a cost estimate?

Subject to appropriation describes spending that reaches an agency only through a later appropriations act. An authorization may permit or direct an activity, and the money follows only when Congress provides the money in a separate decision. For such provisions, the Congressional Budget Office estimates the cost of implementing the authorization on the assumption that funding is provided, because an assumption of no funding would reduce every authorization estimate to zero and conceal the commitment under consideration. The estimate reports that figure separately from direct spending, preserving the distinction between a cost a bill imposes by the terms of the bill and a cost a bill contemplates subject to a future vote.

Q: How are federal loan programs scored?

Federal credit programs are scored on an accrual basis under the Federal Credit Reform Act of 1990. The Congressional Budget Office records the subsidy cost of a direct loan or a loan guarantee up front, in the year of commitment, as an estimate of long term cost after expected repayments, defaults, fees, and recoveries enter the calculation. A cash presentation would misstate the decision, showing a direct loan as a large outlay followed by receipts and a guarantee as nearly costless at issuance and expensive upon a default years later. Accrual treatment makes credit comparable with grants, so assistance delivered through a loan and assistance delivered through a payment can be weighed on a common basis.

Q: What is a mandates statement in a cost estimate?

A mandates statement identifies mandates a reported bill imposes on governments and private parties outside the federal budget and estimates the cost of compliance. The Unfunded Mandates Reform Act of 1995 requires the Congressional Budget Office to identify intergovernmental mandates and private sector mandates in reported bills. The act set thresholds in 1995 at $50 million for intergovernmental mandates and $100 million for private sector mandates, with both figures adjusted for inflation after 1995, and a cost above the applicable threshold triggers disclosure and procedural consequences under the act. A provision can record a modest federal cost while imposing a substantial external burden, and the statement places that burden inside the formal estimate.

Q: Why does the baseline assume some expiring programs continue?

Section 257 directs continuation for a defined class of programs. No program established by a law enacted on or before the Balanced Budget Act of 1997, with estimated current year outlays greater than $50,000,000, is assumed to expire in the budget year or the outyears. The Supplemental Nutrition Assistance Program is the largest case. The convention gives the measuring rod stability, because a baseline that terminated a major program on a scheduled date and restored the program upon reauthorization would attribute an enormous cost to a reauthorizing bill that changed no policy in substance. For programs established by laws enacted after the 1997 act, treatment is settled in consultation with the Budget Committees and the Office of Management and Budget.

Q: What happens to a bill’s effects after the ten year window closes?

Effects after the window continue in operation and fall outside the ten year total by construction. A benefit formula, a tax provision without an expiration, and a credit commitment each carry consequences past the final counted year, and the headline figure records none of those consequences. The estimating system supplies softer instruments for that territory. For major legislation, the Congressional Budget Office reported whether costs would grow or shrink in the second decade, expressed as direction and rough magnitude. A separate test fails a measure that increases deficits by more than $5 billion in any of the four consecutive ten year periods following the window, and The Long-Term Budget Outlook projects the government as a whole over thirty years.

Q: Who asks CBO to prepare a cost estimate?

The duty attaches by statute to committee action. Section 402 of the Congressional Budget Act of 1974 obliges the Congressional Budget Office to provide cost estimates for bills reported by congressional committees, so a committee presenting a finished text to a chamber triggers the formal estimate. An order of service governs allocation of analytical capacity. The House and Senate Budget Committees stand first, because the budget resolution comes first in the legislative year. The appropriations committees and the committees with jurisdiction over revenues and major entitlement programs stand next, because estimates for those committees carry direct enforcement consequences. Other committees and individual members receive estimates and technical assistance as capacity allows.

Q: Does CBO score a bill as written or as intended?

The Congressional Budget Office scores a bill as written. Definitions, eligibility rules, benefit formulas, cross references, effective dates, authorization levels, and expiration dates in the legislative text form the raw material of the calculation. Speeches, summaries, and stated intentions enter only insofar as enacted language carries those intentions. A floor statement describing a program as narrow does not narrow a score when the statutory definition is broad, and a sponsor describing a provision as temporary does not limit cost when the text contains no ending date. Ambiguity becomes a question about the most plausible administration of the language as drafted, and the estimate records the cost of that plausible administration.

Q: What is a timing shift in budget scoring?

A timing shift moves a receipt or a payment across a fiscal year boundary, changing the recorded score in the years on either side of the boundary without changing the substance of the underlying policy. A payment scheduled near the end of a fiscal year and moved into the opening days of the next fiscal year reduces recorded outlays in the first year and raises recorded outlays in the second year by the same amount, while the beneficiary receives the same payment separated by an interval of days. Such shifts matter when a particular year carries a binding target or limit, and the year by year figures in a complete estimate display the movement for a reader examining the pattern behind a total.

Q: Why do cost estimates change as a bill moves through Congress?

An estimate follows a specific text, and movement through Congress changes texts. Committee amendment, floor amendment, action by the second chamber, and conference agreement each can alter a definition, an eligibility rule, a benefit formula, an effective date, or an expiration date, and each alteration changes projected revenues or outlays under the baseline conventions. A changed text therefore requires a new estimate, and differences between an earlier figure and a later figure often record drafting decisions rather than a change of mind by an estimator. The health legislation record illustrates the sequence, as the bills examined in testimony during 2009 differed from the enacted text scored in March 2010, and the fiscal results differed with the language.

Q: How does CBO use information from federal agencies?

The Congressional Budget Office obtains information from the agencies that would administer a program, because those agencies hold operational knowledge compressed by a statutory sentence. Administrators know the size and composition of an existing caseload, the unit cost of a benefit or a service, the capacity of a payment system, and the time required to issue regulations and modify computer systems. Agency information does not dictate a result. The estimating office weighs agency information against other evidence, tests assumptions embedded in an agency projection, and retains independent judgment over the final figure. That separation protects a score from becoming a restatement of a preference held by the department charged with administering a law.