The Price Tag Is Not the Appropriation: How to Read a Congressional Cost Estimate

1. Three numbers, three different questions

A congressional proposal can have a large authorization, a smaller near-term spending estimate, and a deficit line that appears to point in the opposite direction. Those numbers need not contradict each other. They may be measuring different kinds of authority, different flows of money, and different periods. The difficulty begins when a headline presents one of them as the bill's single, complete price.

Consider an actual document: the Congressional Budget Office's September 29, 2026, estimate for H.R. 9338, the Pipeline Safety Authorization Act of 2026, as ordered reported by the House Committee on Energy and Commerce on July 21. Its table shows a net estimated authorization of $471 million. Estimated outlays subject to appropriation total $103 million over 2026–2031 and $471 million over 2026–2036. Separately, the summary shows an $11 million deficit reduction over the longer period from direct-spending and revenue effects. CBO assumes subsequent appropriations consistent with its estimate; it also explains how collecting user fees before spending them affects the timing of net outlays. These are findings for that particular bill version, not a statement that Congress enacted or appropriated those amounts. CBO's H.R. 9338 estimate.

Calling this simply a “$471 million spending bill” drops the condition attached to the appropriations. Calling it an “$11 million savings bill” drops the separate discretionary spending estimate. Neither description explains why money associated with the same proposed program can appear in different columns. A useful account must identify the question before choosing the number.

The purpose of this article is to make those questions concrete. Two real estimates provide reference points. A separate, explicitly fictional project follows money through authorization, appropriation, commitments, and payments. Other invented examples test what happens when a reporting window closes, a baseline changes, or an offset arrives before the spending it helps finance. None of the invented numbers is a finding about an actual federal program.

Reading a cost estimate is not an exercise in guessing which political side has discovered the “real” total. It is closer to reconciling several views of one transaction. An agency's permission to commit funds is not its invoice register. Its invoice register is not a ten-year change against a projected baseline. The distinctions are precisely what allow an estimate to answer more than one question without turning every answer into a different alleged scandal.

2. Follow one project through three stages

Budget authority, obligations, and outlays describe different stages. Budget authority permits financial commitments; an obligation is a binding commitment, such as a contract; an outlay generally occurs when the government makes the payment that settles an obligation. CBO's terminology guide separates those stages rather than treating each as another name for cash spending. CBO, Common Budgetary Terms Explained.

Imagine a fictional federal archive renovation. Congress has provided $100 million of budget authority for an account, and the project has the necessary authority to operate. The agency signs an $80 million construction contract. During the first year, the contractor completes milestones that produce $20 million in payments. For this simplified example, assume no other contracts, payments, adjustments, or expired balances.

At year-end the account has $100 million of budget authority, $80 million obligated, and $20 million paid. These are not three costs to add together. Adding them would produce a fictitious $200 million total by counting the authority, commitment, and partial settlement of the same project as separate expenditures. The difference between authority and commitments is $20 million. The difference between commitments and payments is $60 million. Those two residuals answer different questions.

The first residual tells us how much of the assumed authority has not been committed in this example. The second tells us how much of the contract remains unpaid. Neither residual establishes that the project is late, over budget, or wasteful. If the contract calls for work over several years, a large unpaid balance may be exactly what the schedule predicts. Conversely, rapid payment would not by itself demonstrate that the work is good. Cash timing and performance are separate measurements.

Now suppose an observer photographs an empty construction site after $20 million has been paid. That photograph might raise a useful question about what the payment purchased, but the aggregate numbers cannot answer it. Perhaps the contract covered design, off-site fabrication, or other agreed milestones; perhaps it did not. Investigating that issue requires contract terms and delivery evidence, not a substitution of budget authority for cash expenditure.

GAO's budget glossary makes another useful distinction: an appropriation is legal authority, not a pile of cash set aside in a Treasury drawer. That helps explain why the budget records stages of authority and execution rather than tracing a particular labeled banknote from a statute to a supplier. GAO, A Glossary of Terms Used in the Federal Budget Process.

3. “Authorized” does not always mean “funded”—but the exception matters

An authorization can establish a program and specify an amount that may be appropriated later. An appropriation typically provides budget authority. But it is misleading to turn that common sequence into a universal rule that authorization laws never provide funding themselves: CBO explains that authorization laws can directly provide funding, producing mandatory spending. The operative provisions, rather than the informal label attached to a bill, determine what has actually been supplied. CBO's explanation of authorization and appropriation acts.

Return to the fictional archive. Before the $100 million appropriation, imagine an earlier law authorized appropriations of $150 million for the renovation program. In this scenario, that provision did not itself make the money available. Congress later provided $100 million; the agency then committed $80 million and paid $20 million. The four figures now form a sequence:

The authorization-to-appropriation difference is $50 million. It is not automatically an unpaid government debt to the project. The appropriation-to-obligation difference is $20 million. It is not the same shortfall. The obligation-to-outlay difference is $60 million. That is not a third version of the first difference. Each subtraction compares a particular pair of stages.

Suppose a campaign advertisement says the archive “received $150 million but spent only $20 million.” In this invented case, “received” conceals the distinction between contemplated funding and supplied authority. A more exact statement would say that the program was authorized for $150 million, received a $100 million appropriation in the scenario, and made $20 million in first-year payments against an $80 million contract. The longer sentence is not an effort to excuse anything. It identifies the records a critic would need to examine.

Now alter the fictional legislation so that it directly supplies $150 million of budget authority instead of merely authorizing future appropriations. The original correction no longer works. A reader who memorized “authorization is never money” would be wrong. This is why reading the funding provision matters more than identifying the word “authorization” in a title. The question is whether further legislation is required to supply the authority at issue, not whether a preferred slogan can be attached to the proposal.

4. Build a spending waterfall, not a pile of annual totals

Payments from different funding years can overlap. CBO's discretionary spending tool, updated May 26, 2026, lets users vary budget authority and spendout rates to illustrate that relationship. Its rates are broad illustrative assumptions, not a promise about any particular account. It also distinguishes outlays associated with earlier authority from spending associated with new authority. CBO's interactive discretionary spending tool.

Here is an original miniature model, unrelated to the archive contract. Suppose a fictional program receives three appropriations: $100 million in Year One, $120 million in Year Two, and $80 million in Year Three. Assume each appropriation is eventually paid out in full. For every cohort, 20 percent is paid in the year authority is provided, 50 percent the following year, and 30 percent in the third year. Assume no earlier balances, cancellations, receipts, inflation adjustments, or changes in those rates.

The model's unit is millions of dollars. Each row follows one appropriation through time; each column adds payments arriving in the same year.

Year Three's $106 million in outlays is larger than that year's $80 million appropriation. That is not an arithmetic error: it combines $30 million from the first cohort, $60 million from the second, and $16 million from the third. An analysis that compares only this year's new authority with this year's payments misses the overlapping cohorts.

Year Four has $76 million in payments despite no new appropriation in the example. That fact is equally unsurprising. The payments settle commitments financed by earlier authority. A claim that the program was “secretly funded again” would not follow from the payment column. The relevant question would be where the payment's underlying authority came from, something the cohort rows already make visible here.

A simple cross-check catches several spreadsheet mistakes. Sum across each cohort: its payments should equal its assumed appropriation because this model assumes full eventual spendout. Sum down each annual column: that gives the year's payments. Sum the annual totals: $300 million. Both directions should reconcile. A model that shows $300 million in authority but $340 million in lifetime payments needs an additional assumption or a correction; the excess cannot be explained by merely calling it a timing issue.

The same model also distinguishes a funding change from a scheduling change. If all three appropriations are reduced by one-fifth and the rates stay fixed, lifetime outlays fall to $240 million. If the amounts stay unchanged but payments are delayed, lifetime outlays remain $300 million under the model's assumptions. A near-term table could show a reduction in either case. Only the cohort view reveals whether the reduction is smaller eventual spending or slower payment.

This toy model does not reproduce any agency's actual spending machinery. It deliberately strips that machinery down so the reader can inspect the relationship between annual authority and annual cash flows. Before applying it to a real account, one would need the account's applicable rules and evidence about how quickly its funds are used.

5. A window is not the lifetime of a policy

The fictional waterfall produces $200 million in payments during its first three years: $20 million plus $74 million plus $106 million. Another $100 million arrives in Years Four and Five. If a table stops after Year Three, its $200 million subtotal is correct. Calling that subtotal the program's entire eventual cost is not.

Now move the last cohort's start from Year Three to Year Four without changing its $80 million amount or its spendout rates. First-three-year outlays drop from $200 million to $184 million because the cohort's initial $16 million payment moves outside the window. The lifetime total stays $300 million, with the last payment now arriving in Year Six. The comparison shows a $16 million reduction within the selected window and no reduction in ultimate payments in this model. Both statements can be true.

This is not proof that real legislation uses timing changes deceptively. It is a diagnostic example. A lower subtotal may result from a smaller program, a later start, slower payments, additional collections, or several changes together. Those possibilities have different practical implications. The table needs enough years or explanatory text to distinguish them.

CBO's cost-estimate guide presents estimated authorizations, outlays, and other effects as separate information, and explains the structure of its estimates. The format is designed to expose those distinctions, not to make every displayed subtotal interchangeable. CBO's Cost Estimates Explained.

There is also a language problem. “Over ten years” does not identify which ten years. An estimate can include the current fiscal year alongside a following multiyear period, and different displayed totals can have different endpoints. A reader should reproduce the actual year labels rather than silently converting them into a generic decade. The model's Year One through Year Five labels are intentionally abstract; real analysis should use the document's fiscal years.

A window boundary becomes especially important when comparing two versions issued at different times. If one table ends a year earlier, subtracting its total from the other's can mix a policy change with an extra year of activity. To isolate the policy difference, put both versions on the same period and identify any changed assumptions. Otherwise the subtraction answers a hybrid question that neither estimate was designed to address.

6. The baseline supplies the missing “compared with what?”

A cost estimate usually describes a change from a benchmark, not the entire size of the affected program. CBO's baseline primer explains that its projections generally assume current laws remain in place and provide a benchmark for considering proposed changes. Its separate statutory primer explains that baseline construction also follows specified conventions; it is not simply an unrestricted forecast of what lawmakers are most likely to do next. CBO on developing the baseline; CBO on its statutory foundations.

Consider a fictional program with projected spending of $400 million under the stated benchmark. A proposed change produces projected spending of $430 million over the identical period. The estimated increase is $30 million. Reporting that the program “costs $30 million” would confuse the change with the level. Reporting that the proposal “adds $430 million” would count the benchmark spending as though none of it would occur otherwise.

Write the relationship explicitly: proposal level minus baseline level equals estimated change. In this example, $430 million minus $400 million equals $30 million. That identity is simple enough to seem unnecessary until a debate alternates between program totals and legislative changes without announcing the switch.

Suppose a second analyst uses a different benchmark of $450 million while retaining a $430 million proposal level. The calculation now shows a $20 million reduction. This does not establish that either analyst made a mathematical error. It establishes that they did not answer the same question. To decide which comparison is appropriate, one must identify why their benchmarks differ. Perhaps one uses different policy assumptions, a different period, or a different forecast. Our invented numbers do not specify which, so they cannot establish a winner.

A useful reconciliation would hold the period fixed and separate the changes. Starting from the first calculation, increasing the benchmark by $50 million moves the estimated effect from positive $30 million to negative $20 million, with no change in the proposal level. That is a complete explanation of this particular sign reversal. It does not require a theory about manipulated arithmetic.

A baseline also should not be mistaken for a moral judgment that existing spending is deserved or that changing it is undesirable. It is a comparison framework. A policy could be worth pursuing despite increasing spending relative to that framework, or unwise despite reducing it. Those are additional judgments, requiring evidence about effects beyond the subtraction itself.

The practical reading habit is to keep three quantities visible: the reference policy, the projected outcome under that policy, and the projected outcome under the proposal. A single delta is meaningful only after those pieces are identified. Without them, “saves” and “costs” are verbs missing half their sentence.

7. A real zero that does not mean a zero-cost program

A second actual document sharpens the point. CBO's April 24, 2026, estimate for H.R. 7567, the Farm, Food, and National Security Act of 2026, addresses the version reported by the House Committee on Agriculture on April 21. Its summary shows a $162 million increase in direct-spending outlays over 2026–2031, but a zero net change over 2026–2036. Separately, it estimates $15.785 billion in outlays subject to appropriation over the shorter period and $21.401 billion over the longer one. A footnote limits that latter estimate to specified authorizations. These figures describe that estimate's scope and bill version, not the bill's subsequent legislative status. CBO's H.R. 7567 estimate.

The important word beside the zero is “net.” A subtotal can combine increases and decreases. It is not a statement that every provision has no effect, that every year is unchanged, or that the underlying programs consume no resources. The separate discretionary line is another reason not to read the zero as the complete resource picture.

To see the distinction without reconstructing the farm bill, use an unrelated fictional reform. It increases direct spending by $200 million on one provision and reduces it by $200 million on another over the same period. Net change is zero. The reform can still change who receives payments, when payments occur, or what activities are financed. Its fiscal balance alone does not show that those changes cancel in people's lives.

Nor does net zero establish that there is no forecast uncertainty. If the two components depend on different assumptions, either might turn out differently. An exact arithmetic cancellation of central estimates is not a guarantee that realized outcomes will cancel. The uncertainty belongs to the estimated components even when the final displayed total is small.

The real document and the invented example serve different purposes. The former demonstrates that a zero net figure can coexist with other substantial, separately classified estimates. The latter explains the arithmetic possibility of offsetting components. Neither permits an unsupported claim about which real beneficiaries gain or lose; that requires examining the substantive provisions and their effects.

8. Reconcile a deficit line without adding the same money twice

For a deliberately simple hypothetical, assume a proposal increases direct outlays by $30 million and increases revenues by $12 million over the same period. Its direct-spending-and-revenue effect increases the deficit by $18 million. If revenues instead fall by $12 million, the deficit increase is $42 million. The subtraction changes sign because reduced receipts leave a larger financing gap, not because reduced revenues somehow count as cash spending by an agency.

Now add a provision authorizing future discretionary appropriations of $100 million. Assume an analyst estimates that supplying those appropriations would produce $70 million of payments within the displayed window and $30 million afterward. A responsible summary keeps the $18 million direct-and-revenue deficit effect separate from the $70 million conditional outlay estimate. If the reader wants a scenario in which all the assumed appropriations are subsequently supplied, the within-window effects combine to $88 million before any additional effects excluded by the setup. That is a scenario with an extra funding assumption, not an unconditional consequence of the authorization provision alone.

The incorrect total would add $100 million of authorized funding, $70 million of resulting payments, and $18 million of direct-and-revenue effects to claim $188 million. That calculation counts the prospective discretionary funding and part of its prospective spendout twice. The $100 million and $70 million are different views of the same modeled funding stream, not independent components of one cash-flow sum.

CBO's current FAQ explains why its estimates distinguish direct spending from spending subject to appropriation: the latter depends on subsequent funding legislation. It also notes a complication that defeats simple labels—authorizing legislation can affect appropriations already provided, and those effects are treated as direct spending. CBO's cost-estimate FAQ.

The fictional arithmetic assumes all components are measured in the same units and years. Before repeating it on an actual table, inspect those headings. Millions and billions are easy to mix; so are annual and cumulative figures. A discrepancy of a thousandfold may be a unit error rather than a policy disagreement. A smaller discrepancy may come from rounding. Neither possibility should be resolved by silently changing a table's units until a preferred total appears.

9. Negative net spending does not mean an agency did nothing

Some collections are recorded as offsets to spending rather than as revenue in the budget's revenue category. GAO's glossary distinguishes these treatments. The classification matters because a minus sign in a net outlay series can reflect incoming collections, not a negative quantity of work or an absence of payments. GAO's federal budget glossary.

Consider a fictional inspection service, not the pipeline program discussed earlier. Assume it collects $60 million in fees during Year One, makes $20 million in payments that year, and makes $60 million in payments in Year Two. Assume those fees are properly classified as offsets to outlays for purposes of this illustration, with no other flows. Net outlays are negative $40 million in Year One and positive $60 million in Year Two. Across both years, gross payments total $80 million, collections total $60 million, and net outlays total $20 million.

A chart showing only Year One can therefore look like a $40 million saving while $20 million is actually being paid to operate the service. A chart showing only Year Two can look like a $60 million cost without displaying the earlier collections. The combined view does not invalidate either annual number; it identifies what was left outside the frame.

There are three different quantities a reader might legitimately want here. Someone studying agency operations may need the $80 million in gross payments. Someone studying the burden on fee payers may need the $60 million in collections. Someone studying the net budget effect may need the $20 million difference. Choosing the third number does not make the first two irrelevant. Adding all three together does not make a more complete estimate.

This is also why “paid for” needs a definition. In the invented example, fees cover three-quarters of gross payments over the two years. They do not cover the remaining quarter. They also do not eliminate the costs experienced by the entities paying the fees. The phrase might refer to federal financing, distribution of costs, or a political judgment about whether the arrangement is acceptable. Arithmetic can answer the financing question under stated assumptions. It cannot turn that answer into every other meaning of the phrase.

10. A point estimate is a model output, not a promise

CBO's February 2025 uncertainty primer identifies several reasons estimates can differ from eventual outcomes, including limited evidence, broad legislative language, behavioral responses, the size of eligible populations, baseline uncertainty, and dependence on future events. It explains that a published point estimate is the agency's best assessment rather than a guaranteed realization. CBO on uncertainty in legislative cost estimates.

An original benefit example shows how uncertainty enters before any complicated economic model is needed. Suppose a fictional proposal offers an annual $1,000 payment to an eligible population of 100,000 people. Assume 60 percent participate. Ignore administration, taxes, interactions with other benefits, and payment timing. Estimated annual payments are 100,000 multiplied by 0.60 multiplied by $1,000, or $60 million.

Changing participation to 40 percent produces $40 million; changing it to 80 percent produces $80 million. These are sensitivity scenarios, not a statistical confidence interval. We have supplied no evidence that the alternatives have particular probabilities, that they bound all plausible outcomes, or that participation is independent of other inputs. Calling the interval “95 percent certain” would manufacture information the example does not contain.

A second change illustrates interaction. If the eligible population rises to 120,000 and participation rises to 80 percent, payments reach $96 million. Starting from $60 million, increasing eligibility alone adds $12 million, and increasing participation alone adds $20 million. Adding those two isolated changes gives $92 million, not $96 million. The missing $4 million is their interaction: the additional 20,000 eligible people also experience the additional 20 percentage points of participation.

That interaction is not an obscure accounting convention. It follows directly from multiplication. It explains why separate amendments or assumptions cannot always be added as though each acts on an unchanged world. A combined estimate can differ from the sum of two stand-alone estimates without anyone having altered the arithmetic unfairly.

The example also identifies what evidence would be useful. If eligibility is well measured but participation is poorly known, another count of the eligible population may add less information than reliable enrollment evidence. If the benefit changes incentives to participate, a rate from a materially different program may be a weak guide. Those are empirical questions, not answers supplied by the elegant appearance of a spreadsheet.

An honest account of uncertainty therefore names the sensitive input and shows its effect. It need not declare every estimate unknowable. The difference between a model's checked multiplication and its uncertain assumptions is itself valuable information. A calculation can be exactly correct given its inputs and still be an imperfect forecast of what will happen.

11. Interest depends on timing, and it needs its own label

A conventional legislative estimate should not automatically be described as including every financing consequence. CBO's debt-service report explains that interest effects depend partly on when changes in borrowing occur. Its current FAQ says cost estimates do not typically include debt-service effects, while also identifying separate analyses in which such effects have been provided. “Never includes interest” is therefore too absolute; the particular estimate and accompanying analysis must be checked. CBO on debt-service effects; CBO's current FAQ.

Use a small fictional financing model to see the timing issue. Assume a fixed annual interest rate of 4 percent, no principal repayment during the period, no compounding, and borrowing at the start of each year. Interest is paid from outside this model rather than borrowed. These simplifying assumptions are not a forecast of Treasury rates or CBO's methodology.

Under Scenario A, $100 million is borrowed at the start of Year One and remains outstanding for five full years. Annual interest is $4 million, for $20 million across the period. Under Scenario B, $20 million is borrowed at the start of each of five years. The outstanding principal during those years is $20 million, $40 million, $60 million, $80 million, and $100 million. Interest is $0.8 million, $1.6 million, $2.4 million, $3.2 million, and $4 million: $12 million in total.

Both scenarios end with $100 million of principal outstanding. They do not produce the same interest within the five-year window, because the money has been outstanding for different lengths of time. Looking only at the ending principal would conceal that distinction. Treating the $8 million difference as a permanent saving would also overreach: the model says nothing about what happens after Year Five.

This exercise is intentionally separate from the earlier appropriation waterfall. Not every dollar of budget authority maps mechanically onto immediate borrowing, and this simplified loan-style model is not a substitute for federal financing analysis. Its narrow purpose is to demonstrate why a financing estimate needs a timeline as well as a total. If interest is added to a published estimate, its assumptions, period, and status as an additional calculation should travel with it.

12. “Dynamic” is not a synonym for “people respond”

Debates sometimes contrast an allegedly frozen estimate with one that finally allows anyone to change behavior. The official distinction is more specific. In their April 2025 explanation, CBO and Joint Committee on Taxation staff say conventional estimates already incorporate behavioral responses. Dynamic analysis adds effects operating through changes in the broader economy. The document also discusses exceptions and details; these labels do not by themselves validate a particular growth assumption. CBO and JCT staff on dynamic analysis.

In the invented benefit example, changing participation from 60 percent to 80 percent is a behavioral response. Including that response does not, by itself, show that the model estimates changes in national output, wages, investment, or interest rates. It is perfectly possible to model enrollment decisions while leaving the broader economic forecast unchanged.

Conversely, someone could append a claimed growth effect without adequately modeling participation. The presence of a macroeconomic adjustment would not cure that omission. The analytical question is what responses are modeled, how they interact, and what evidence supports their size. A one-word label cannot answer those questions.

Consider another entirely fictional calculation. A proposal has a $100 million direct-and-revenue deficit effect before a separate macroeconomic adjustment. An analyst claims that broader economic changes would generate $15 million in additional receipts over the same period. If those effects are truly additional and consistently defined, the adjusted figure would be $85 million. But if the $15 million already includes receipts counted in the original $100 million estimate, subtracting it again double-counts the adjustment.

The necessary comparison is not “optimistic” versus “pessimistic” arithmetic. It is a mapping of components: which receipts were already counted, which new economic channel produces an additional change, and whether the estimates use compatible periods and assumptions. The example supplies no evidence that $15 million is realistic. Its purpose is to show what must be established before that subtraction is justified.

A claim that a policy pays for itself therefore needs more than a growth-related number somewhere in an appendix. It needs a coherent comparison in which all relevant components are defined and counted once. The same requirement applies to a claim that growth effects are necessarily zero. Neither conclusion follows merely from choosing a favored scoring label.

13. The bill version is part of the number

An estimate is an analysis of specified legislative text under stated assumptions, not an evergreen sticker attached to a bill number. CBO's process primer describes how analysts examine the actual provisions, develop estimates, and explain the basis of their results. It also distinguishes CBO's advisory role from the Budget Committees' role in enforcing budgetary rules. CBO Describes Its Cost-Estimating Process.

Imagine an original proposal that would provide a fictional benefit for three years, followed by an amendment extending it for two more. The earlier estimate cannot establish the amended proposal's cost merely because the bill number stayed the same. Nor can a reader assume the cost rises by exactly two-thirds. The added years could involve different enrollment, prices, implementation, or interactions. Our example has not specified those factors.

A more useful comparison would align two records: the earlier text and its estimate, then the amended text and its estimate or a clearly labeled independent calculation. The changed provision should be named. If no updated estimate exists in the material examined, that is a missing result, not permission to silently transfer the old one to the new version.

The same discipline applies to legislative status. A committee's approval, passage by one chamber, and enactment are different events. The two real examples in this article are identified by the versions stated on CBO's documents. This article does not claim to establish their later status. That boundary is important because even a perfect explanation of a committee-version estimate does not prove what a subsequently enacted law contains.

Finally, passage of funding legislation and execution of funding are different events too. A prospective estimate cannot establish how much an agency has actually paid after implementation. To compare a forecast with realized spending, the analyst must match the program scope, accounting category, and period. Comparing a projected change against actual total program spending would repeat the baseline error in a new form.

These are document-matching problems before they are ideological disputes. A reader can often identify the mismatch without taking a position on whether the policy is good. Fixing the reference does not settle the debate; it ensures that the participants are debating the same proposal.

14. Write the sentence that the numbers actually support

Bring the fictional examples together and consider this invented headline: “Congress spends $150 million on archive project; only $20 million used.” It combines an authorization amount with an early payment figure and uses a verb that does not identify either stage. A corrected account would distinguish the $150 million authorization, the later $100 million appropriation assumed in the scenario, the $80 million contract, and the $20 million in first-year payments. Whether the project performs well remains a separate question requiring evidence about its work.

A second invented headline says: “Program cut saves $16 million.” In the delayed-cohort example, the statement is incomplete. Payments fall by $16 million within the first three years because one cohort starts later, while lifetime payments remain $300 million under the assumptions. A precise sentence would report the within-window reduction and the unchanged eventual total. It would not call the delay either harmless or deceptive without additional evidence about its practical effects.

A third says: “Net-zero reform has no cost to anyone.” The hypothetical offsetting provisions support only a zero net change in the specified federal spending measure. They do not establish that recipients, taxpayers, providers, or other affected people experience no change. Likewise, a negative net outlay in the fee example does not mean that no payments occurred or no one bore a cost.

The objective is not to make every article unreadable with caveats. It is to keep the few qualifiers that determine what a number means: the measure, the comparison, the period, the bill version, and any subsequent funding assumption. Omitting one of those can change the proposition rather than merely simplify its wording.

A budget estimate is useful because it makes a structured prediction about a defined policy change. It does not, by itself, rank the policy's benefits, establish the quality of its implementation, or authorize an agency to spend. Those are different inquiries. Keeping them distinct makes criticism sharper: it becomes possible to challenge an assumption, expose a timing mismatch, or demand evidence of results without first misdescribing the estimate.

The price tag is therefore not the appropriation, and the appropriation is not the payment. A total is not a baseline change, a window is not a lifetime, and a balanced federal subtotal is not a complete verdict on public value. Once those distinctions are visible, the most useful question becomes much more specific than “What does the bill cost?” It becomes: “Which effect are we measuring, against what alternative, over which years, and what has to happen for that estimate to become actual spending?”