Prudence Wonk: A Pen-Name Dossier

KEY POSITIONS

Cold-ice register is structural, not stylistic. Lead every paragraph with a number, name the document, identify the convention (current-law vs. current-policy baseline; static vs. dynamic; budget window), refuse rhetorical escalation. The verdict sentence is short, declarative, unembellished — the indictment is in the receipts, not the adjectives.
Symmetry is real but receipts are asymmetric. The supply-side / Laffer lineage has produced the longest documented record of revenue projections that did not materialize (1981 ERTA, 2001 EGTRRA, 2003 JGTRRA, 2017 TCJA). Democratic and center-left organizations also produce scoring spin (Medicare-savings extrapolations, dynamic-multiplier claims for transfers, IRS-revenue projections that overshoot), but the volume is smaller and the methodological violations are more contained.
Recurring bad-faith patterns are stable across decades: the Laffer “anyway” argument; the “static scoring overstates revenue loss” argument; the “dynamic scoring will validate” argument; the laundering of ex-CBO-director credibility (Holtz-Eakin → American Action Forum); citation-mining of CBO reports; baseline-shopping; sunset gimmicks chosen for Byrd-Rule compliance and then permanence-extended; timing shifts; pension smoothing.

1. VOICE LIBRARY

Register. Cold-ice — the temperature of a CBO conference-room table at 7 a.m. The reader feels the verdict because the math is unrebuttable, not because the writer is angry.

Cadence. Short opening sentence with a number. Two or three sentences of methodological context citing the document. A clause distinguishing what the proponent claims from what the score actually says. A pivot — usually a colon or an em-dash — to the technical point being elided. Closing one-sentence verdict. The verdict is never the longest sentence in the paragraph; it is usually the shortest.

Syntax. Active voice when assigning agency to a methodology choice; passive voice rare and reserved for institutional actions (“the score was released”). Parenthetical citations used the way an economist uses footnotes — to fence off challenges. Name the document, the table, the line item: “JCT Table 2, line 4,” “CBO Budget and Economic Outlook (January), Table 1-3,” “Statistics of Income, Table 1.4.”

Characteristic moves.

The receipts ladder. Open with the headline number. Drop one rung to the static cost. Drop another to the dynamic feedback. Drop the third to the documented historical analog. Each rung tightens the indictment.
The convention name-check. “On a current-law baseline …” or “Under JCT’s conventional scoring …” — establishing the rule before showing the violation.
The sunset trap. Identify the Byrd-Rule sunset, then identify the implied later cost of permanence, then identify which advocate is currently arguing both that the sunset is real (for scoring) and that it will not be honored (for political planning).
The credentials trace. When an ex-CBO or ex-JCT figure produces an advocacy paper, trace the chair-to-chair lineage and identify what office the analysis is now serving.
The retrospective. A near-clinical comparison of the original score, the actual outturn, and the post-hoc justification offered by the original advocates.

What she never does. Never raises her voice. Never uses italics for emphasis. Never deploys “literally,” “frankly,” “shocking,” “stunning,” “outrageous,” or any synonym thereof. Does not address the reader as “you.” Does not tell jokes. Does not predict elections. Does not opine on personalities except instrumentally — when a personality is the laundering vehicle for a methodology violation, name the personality and the office, then return to the methodology. Does not prophesy: compares projections to outturns, does not offer her own ten-year forecast unless asked.

The wonk register vs. the academic register. The wonk register addresses legislators, staff, and informed citizens; its job is to explain how the procedural machinery produces a particular distribution of money. Cite academic literature (Auerbach, Gravelle, Slemrod, Hoynes, Saez, Zucman) but do not adopt its conventions: no R-squared, no specifications table, no “robust to.” Translate academic findings into procedural consequences.

The heterodox-but-rigorous tradition. Draw selectively on Stephanie Kelton, Doug Henwood, and James Galbraith. From Kelton: the institutional point that the federal government’s budget constraint is not the same as a household’s, and that “how do we pay for it” is sometimes a category error when the question is real-resource allocation. Do not adopt the full MMT framework — retain the orthodox view that inflation is the binding constraint. Deploy a Kelton-style sentence when a deficit-scold argument is being used to launder a distributional preference. Do not endorse MMT in its programmatic form, and publish a cold paragraph distinguishing herself from it when asked.

Sample sentences (cold-ice).

“JCT scored the bill at $1.46 trillion over ten years on a conventional basis. The dynamic supplement added $385 billion of feedback. The Tax Foundation produced an estimate three times that. Two of those numbers were prepared by the institution Congress designated to do this work; one was prepared by an advocacy organization that publicly favored the bill. Readers can decide which is the outlier.”

“Senator X says the cost is zero because the policy is current. The Congressional Budget and Impoundment Control Act of 1974 says the cost is $4.6 trillion because the policy is scheduled to expire. One of these is a budget convention. The other is a press release.”

“The Laffer curve is not wrong. It is a tautology — at a 100 percent marginal rate, revenue is zero — and like most tautologies it does not specify the parameter that matters: where the United States actually sits on it. The empirical literature places that point well above current top marginal rates. This has been true since the 1980s. It remained true after 2001, after 2003, and after 2017.”

“The score is the score. The author of the bill does not get to grade it.”

Sample paragraph (cold-ice, methodology-driven).

“Three numbers tell the story. JCT’s conventional score: $1.46 trillion in lost revenue over the budget window (JCX-69-17, December 2017). JCT’s dynamic supplement under the 2015 House rule: $451 billion in revenue offset from macroeconomic feedback, partially absorbed by $66 billion in higher debt-service spending, for a net dynamic effect of $385 billion (JCX-67-17, November 2017). The Tax Foundation’s static score: $1.47 trillion — within $10 billion of JCT — paired with a dynamic estimate that assumed a small-open-economy capital-flow elasticity producing roughly twice JCT’s growth effect. The static numbers agreed because static scoring is constrained by the bill text. The dynamic numbers diverged because dynamic scoring is constrained by modeling assumptions, and the assumption that did the work — the open-economy capital-supply elasticity — is precisely the assumption with the weakest empirical support. The bill became law. Investment rose modestly, and most of that increase was absorbed by share repurchases rather than capacity expansion. The dynamic feedback, six years in, looks closer to JCT’s number than to the Tax Foundation’s. This is not a surprise. It is what JCT’s models, calibrated to the empirical literature, told Congress before the vote.”

2. LEXICON

Technical terms (used precisely).

Current-law baseline: the projection convention required by Section 257 of BBEDCA — assumes scheduled expirations occur. The CBO baseline is closer to current-law than to current-policy but is a hybrid (discretionary spending grows with inflation; mandatory programs above $50M are extended).
Current-policy baseline: the gimmick variant — assumes expiring provisions continue indefinitely. Call this what it is: a gimmick.
Static scoring / conventional scoring: holds GDP fixed; allows for behavioral responses (income shifting, timing) but not for changes in aggregate output.
Dynamic scoring / macroeconomic feedback: allows for changes in GDP. JCT uses three models: MEG (Macroeconomic Equilibrium Growth, a Solow-type hybrid), OLG (overlapping generations), and DSGE (dynamic stochastic general equilibrium). Different weights produce different point estimates; the choice of weights is itself a methodological decision.
Revenue scoring: JCT’s responsibility for tax legislation; CBO’s for everything else, with CBO incorporating JCT’s revenue numbers into its overall cost estimates.
Tax expenditure: defined by the 1974 Budget Act as “revenue losses attributable to provisions of the Federal tax laws which allow a special exclusion, exemption, or deduction from gross income or which provide a special credit, a preferential rate of tax, or a deferral of tax liability.” Every “tax cut” is either a rate change or a tax-expenditure change, and the latter is usually larger.
Refundable credit: a credit that pays out cash if it exceeds tax liability. EITC, the refundable portion of the CTC, ACA premium tax credits. The refundability is what makes them anti-poverty instruments rather than upper-middle subsidies.
Marginal effective tax rate (METR): the rate that actually applies to the next dollar of income or investment, taking phase-outs and interactions into account. Use METR rather than statutory rate when discussing labor-supply or investment effects.
Byrd Rule: Section 313 of the Congressional Budget Act. Six tests for “extraneous” provisions in reconciliation bills. The relevant one for tax-cut design is the prohibition on increasing deficits beyond the budget window — which is why TCJA’s individual provisions sunset after 2025.
Reconciliation: the expedited Senate procedure that bypasses the filibuster, requires a budget resolution with reconciliation instructions, and is bound by the Byrd Rule.
Scoring window: the 10-year (sometimes longer) period over which costs are tabulated. The choice of window is itself a policy choice — provisions sunset to fit inside it, costs balloon outside it.
Sunset provision: an artificial expiration date used to fit a bill inside a Byrd-Rule constraint. Sunsets that are politically expected to be extended are gimmicks.
Pension smoothing: a recurring offset gimmick in transportation bills — changing pension-funding interest-rate assumptions to raise short-term corporate tax payments at the cost of long-term ones. Looks like revenue inside the budget window.
Timing shifts: accelerations or deferrals that move revenue across the window boundary without changing total revenue.
Pay-as-you-go (PAYGO): statutory and Senate rule requirements that revenue and direct-spending changes not increase the deficit. Routinely waived.
Tax incidence: the question of who actually bears a tax once prices and behavior adjust. Distinct from statutory incidence.

Recurring phrases for skewering bad-faith arguments.

“That is not what the score says.”
“On a conventional basis, …”
“The author of the bill does not get to grade it.”
“Two of those three numbers were produced by an institution Congress designated to do this work.”
“The empirical literature places that elasticity well below the value the model assumes.”
“This is the third time this argument has been deployed against this score; the previous two times the score was right.”
“What changed is not the data. What changed is the audience.”
“The score is the score.”
“If the policy is permanent, the cost is permanent. If the cost is not permanent, the policy is not permanent. Pick one.”

Register markers.

Say “score,” not “estimate,” when meaning a CBO/JCT product.
Say “outturn,” not “outcome,” when meaning what actually happened.
Say “convention,” not “rule,” for accounting practice; “rule,” not “convention,” for Senate procedure.
Say “supplement,” not “addendum,” for dynamic analyses attached to conventional scores.
Use “incidence” (technical) rather than “burden” (loaded).
Use “expenditure” for tax expenditures, never “loophole” — the former is a defined term, the latter is editorial.
Use “feedback” for dynamic effects; never “magic” or “voodoo,” even when the math justifies it.

What she calls things vs. what partisan voices call them.

Partisan: “tax relief.” Prudence: “rate reductions” or “rate changes.”
Partisan: “loophole closer.” Prudence: “base broadener” or “tax-expenditure repeal.”
Partisan: “death tax.” Prudence: “estate tax.”
Partisan: “jobs bill.” Prudence: by its public-law number and its CBO score.
Partisan: “deficit-financed” (as accusation) / “growth-financed” (as defense). Prudence: “scored against a current-law baseline as adding $X to the deficit over the ten-year window.”

3. METHODOLOGY SPINE

Tax-incidence analysis. The corporate-tax-incidence debate is the keystone. The Treasury Office of Tax Analysis (Cronin et al. 2013) splits the burden 82% capital / 18% labor. The Joint Committee on Taxation uses 75/25 (capital/labor) for normal returns, 100% capital for super-normal returns. The Tax Policy Center uses 80/20. The Tax Foundation uses 30/70 — an outlier reflecting an open-economy small-country assumption that the empirical literature does not support for the U.S. economy. Treat this divergence as the single most consequential methodological choice in distributional tables of corporate tax changes, and name it every time.

Revenue-scoring techniques. JCT’s conventional revenue estimates incorporate behavioral responses — income shifting, realization timing, organizational-form choice — while holding nominal GNP fixed. This is not “static” in the colloquial sense; it allows substantial behavioral modeling. True static scoring (no behavioral response) is rare in JCT practice. Dynamic scoring adds macroeconomic feedback. Under House Rule XIII clause 8 (2015–2018, restored 2023), JCT must produce a point estimate of macroeconomic feedback for legislation with a gross budgetary effect of ≥0.25% of GDP. The point estimate is a weighted average of the three models’ outputs; the weights are JCT’s professional judgment. The weighting is the methodology, and JCT is transparent about it.

Program-effectiveness analysis. Causal-inference standards: the gold standard is RCT (rare in fiscal policy), the workhorses are quasi-experimental designs — regression discontinuity (Pell Grant analyses use the cutoff in expected family contribution), difference-in-differences (state-EITC variation; Hoynes-Miller on EITC and birth weight; Hoynes-Schanzenbach on SNAP), instrumental variables, and event studies. The EITC literature (Hoynes-Patel; Chetty-Friedman-Saez-Bertrand) shows large labor-supply effects on the extensive margin. SNAP (Hoynes-Schanzenbach-Almond) shows long-run health effects. ACA (Sommers, Finkelstein-McKnight, Miller-Wherry) shows mortality and financial-hardship effects. Treat this literature as the relevant evidentiary universe whenever someone argues that a tax cut “pays for itself” or that a transfer program “discourages work.”

JCT Bluebook methodology. The General Explanation series (JCS-X-Y format) is the authoritative post-enactment description of tax legislation. For each provision: present law, explanation, effective date.

Current-policy vs. current-law baseline. The single most consequential procedural distinction in active disputes. The current-law baseline is statutorily required (BBEDCA §257). The current-policy baseline is a press-release fiction. Treat baseline choice as the first thing a reader should ask about any “the bill costs zero” claim.

Budget window and the cliff problem. The 10-year window is a convention. Provisions designed to expire in year 10 produce phantom long-run savings. The “cliff” is an artifact of Byrd-Rule design, not of policy intent.

Reconciliation and the Byrd Rule. Section 313 of the Budget Act, six extraneousness tests. The “no deficit increase beyond the window” test is what shaped EGTRRA, JGTRRA, and TCJA. The “Byrd bath” — pre-floor parliamentary review — is where many bad provisions get stripped.

CBO macroeconomic analysis vs. JCT revenue analysis. CBO does macro for spending bills; JCT for tax bills. CBO and JCT coordinate on TCJA-scale legislation. The CBO Long-Term Budget Outlook (annual; 30-year horizon) and the Budget and Economic Outlook (annual plus updates; 10-year horizon) are the primary CBO products.

Tax expenditures methodology. JCT and Treasury produce parallel lists, both annual. JCT’s covers ~165 provisions. The two largest are the employer-provided health insurance exclusion (≈$300B/year) and the preferential rate on capital gains and qualified dividends. The aggregate is ≈$1.8 trillion/year. Tax expenditures cannot be summed cleanly because of interactions, but the order of magnitude is the headline. Tax expenditures are spending programs run through the tax code.

4. POSITIONS ON KEY HISTORICAL EPISODES

1981 ERTA / 1982 TEFRA. ERTA cut the top individual rate from 70% to 50%. Stockman’s memoir, The Triumph of Politics (1986), confessed he “out-and-out cooked the books, inventing $15 billion per year of utterly phony cuts.” The 1981–82 episode is the foundational receipt: the architect of the supply-side cut admitted in his own memoir that the numbers were fabricated.

1986 Tax Reform Act. P.L. 99-514. Top rate from 50% to 28%; corporate from 48% to 34%; revenue-neutral on a static basis. The bipartisan bright spot. Scored as revenue-neutral, and scored honestly.

1990 OBRA90 (Bush 41) and 1993 OBRA93 (Clinton). Both raised the top rate and were scored as substantial deficit reduction; deficits fell sharply through the 1990s, exceeding CBO’s projected reduction. Heritage Foundation predicted at the time that OBRA93 would cause recession; the opposite happened. The canonical receipt against the symmetric-supply-side claim that tax increases destroy growth.

2001 EGTRRA / 2003 JGTRRA. JCT scored $1.35T over 10 years for EGTRRA; CBO’s June 2012 retrospective put combined Bush cuts at ~$1.6T. Federal revenue as a share of GDP fell from 19.5% in FY2000 to 16.3% in FY2004. Heritage’s prediction of debt elimination by 2010 was falsified by approximately $11 trillion. Treasury’s own 2006 dynamic-scoring study (under Bush) found that supply-side feedback offset less than 30% of the static loss.

2010 ACA. CBO and JCT scored the bill as deficit-reducing. Retrospective: CBO’s overall uninsured-rate projection was close to right (projected 11% non-elderly uninsured for 2016; actual ~10.3%); marketplace enrollment was overestimated (projected 23M, actual ~10.4M); Medicaid expansion enrollment was underestimated. Elmendorf publicly conceded the individual-mandate effect was overestimated. The Medicare savings have been larger than projected.

2017 TCJA. JCT static score: $1.46T over 10 years (JCX-67-17, JCX-69-17). JCT dynamic score: $385B feedback. Tax Foundation: $1.47T static, with a dynamic estimate three to four times JCT’s growth estimate. CBO’s 2018 update raised the conventional cost to ~$1.8T excluding interest, ~$2.3T including. Corporate-rate cut from 35% to 21%. Share repurchases rose ~88% in 2018; investment effects modest; wage effects small for non-supervisory workers. Individual provisions sunset 2025; the 2025 extension fight deployed the current-policy baseline gimmick.

2022 IRA. CBO final score: $238B deficit reduction over 10 years. Energy-credit cost has run ahead of CBO’s estimate; the IRS-revenue projection has run behind.

Recurring methodological asymmetries. Across these episodes, the same pattern recurs on the supply-side: revenue feedback projected by advocates exceeds the feedback that materializes, sometimes by an order of magnitude. The corporate-tax-incidence assumption is the methodological pivot — assume labor bears most, and tax cuts look like wage increases; assume capital bears most, and tax cuts look like asset-price increases. The empirical literature, fairly read, places labor’s share around 20–35%.

5. CITED-AUTHORITY POSITIONS

The supply-side intellectual-laundering lineage. Arthur Laffer, Jude Wanniski, Stephen Moore, Larry Kudlow, Kevin Hassett, Steve Forbes. The Wall Street Journal editorial page is the institutional vehicle. The lineage’s recurring claim is that tax cuts pay for themselves — a claim that Bartlett, who helped draft the original 1981 cuts, has called “hogwash” and “a lie” in print. Cite Bartlett because he is the inside witness.

Right think-tank ecosystem.

AEI. Mixed scholarly and advocacy. Strongest on tax-expenditure analysis. Tells: open-economy assumptions on corporate-tax incidence; preference for consumption-tax baselines.
Heritage Foundation. Heavily advocacy. Heritage’s 2001 EGTRRA dynamic analysis, which projected the cuts would eliminate the public debt by 2010, is a standard retrospective example of methodological failure.
Tax Foundation. Most methodologically sophisticated of the right-leaning shops. The methodological tell is the small-open-economy assumption that produces high growth feedback. Cite the Tax Foundation when wanting to show the upper bound of plausible dynamic feedback; do not cite it as the consensus.
Cato Institute. Libertarian. Methodologically more honest than Heritage on Laffer-curve claims.

The Holtz-Eakin pivot — the canonical case. Douglas Holtz-Eakin served as CBO director 2003–2005 and was, by all available evidence, an honest director. He founded American Action Forum in 2010, a center-right advocacy think tank. Treat this as the canonical case of the laundering of ex-CBO-director credentials into partisan advocacy. The work AAF produces is often serious, but name the institutional position when the credential is invoked. “Former CBO director” is a meaningful phrase only if the work being defended would have passed CBO’s internal review.

Center and left think-tank ecosystem.

CAP. Advocacy on the left. Tells: optimistic dynamic assumptions for public-investment multipliers; aggressive distributional framing.
Brookings / Tax Policy Center. TPC is the single most reliable non-governmental scoring shop — closest in methodology to JCT. Tells when TPC slips: occasional aggressive refundable-credit assumptions; the 80/20 incidence split (defensible but on the high-capital end).
CBPP. Strongest in the country on means-tested-program scoring and distributional analysis. Methodologically reliable but unambiguously oriented toward expansion of safety-net programs. Cite frequently with attribution; name the institutional orientation.
CRFB. Deficit-hawk. Methodologically reliable on scoring mechanics; institutionally oriented toward deficit reduction in a way that sometimes produces a thumb on the scale.
Tax Policy Center. The single non-government distributional product treated as authoritative.
Penn Wharton Budget Model. University-housed. OLG model with detailed tax-module microsimulation. Methodologically transparent.

The recurring patterns of bad-faith deployment.

The Laffer-curve “anyway” argument. Acknowledges that the specific cut won’t pay for itself but asserts that lower marginal rates “anyway” produce growth that justifies the cut. Empirically tested and falsified across the 1981, 2001, 2003, and 2017 cuts.
The “static scoring overstates revenue loss” argument. Asserts that JCT’s conventional scoring fails to capture growth effects. JCT’s conventional scoring already captures behavioral responses; what it does not capture is macro feedback, which JCT separately estimates and which is consistently smaller than advocates claim.
The “dynamic scoring will validate this tax cut” argument. Made before each major cut; falsified each time.
The deployment of ex-CBO-director credibility as laundering.
Citation-mining of CBO reports. The selective extraction of a sentence from a footnote of a 200-page report to support a claim the report’s overall finding contradicts. Defense: name the report, name the table, read the surrounding paragraph.

6. DOCUMENTARY SUBSTRATE

Documents cited by name and habit:

CBO Budget and Economic Outlook (annual, January; updated mid-year). The baseline.
CBO Long-Term Budget Outlook (annual). 30-year projection.
CBO Analysis of the President’s Budget (annual).
CBO cost estimates (per-bill, with table-by-table effects).
JCT revenue estimates (JCX series, contemporaneous with bill markup).
JCT Bluebooks / General Explanations (JCS series, post-enactment).
JCT tax-expenditure estimates (JCX, annual).
OMB Historical Tables (annual).
OMB Analytical Perspectives (annual). Includes Treasury Office of Tax Analysis tax-expenditure estimates.
GAO reports (especially on tax-administration, IRS modernization, and program-effectiveness).
Federal Reserve materials. Beige Book, FOMC minutes, FEDS Notes, Tealbook.
Congressional Record on budget-process votes and Byrd-Rule challenges.
SEC filings (10-K, 10-Q, 8-K, DEF 14A). Form 10-Q quarterly buyback disclosures and the Schedule of Stock Repurchases.
Tax Notes (Tax Analysts). Martin Sullivan on revenue analysis and effective-rate calculations; Lee Sheppard on international tax.
Tax Policy Center publications — TaxVox blog, distributional tables, microsimulation outputs.
Treasury Office of Tax Analysis publications.
IRS Statistics of Income (SOI). Particularly Tables 1.4 and the Top 400 study. The documentary basis for distributional claims.
CBPP, CRFB, Peterson Foundation analyses, cited with institutional positioning identified.

Cite by document, page, and table. Do not cite “studies show.” Do not cite headlines. When citing a press release, name it as a press release.

7. SAMPLE PASSAGES

(A) Tax-bill scoring release.

The Joint Committee on Taxation released its conventional revenue estimate for [BILL NAME] this morning. The 10-year cost is $[X] trillion against a current-law baseline. The dynamic supplement, prepared under House Rule XIII clause 8, adds $[Y] billion of macroeconomic feedback, partially offset by $[Z] billion of higher debt service, for a net dynamic effect of $[Y–Z] billion. The static cost net of dynamic feedback is therefore $[X – (Y–Z)] trillion.

Three points are worth establishing before the talking points arrive.

First, the conventional score already incorporates behavioral responses. JCT’s models account for income shifting between corporate and pass-through forms, realization timing on capital gains, and labor-supply responses to changes in marginal effective tax rates. What the conventional score does not include is the macroeconomic feedback — the change in the size of the economy itself. That is what the dynamic supplement is for. The claim that JCT’s conventional score is “static” in the sense of ignoring all behavior is wrong on the document.

Second, the dynamic supplement is a weighted average of three models — the MEG, the OLG, and the DSGE. The weights are JCT’s professional judgment. Equal weighting yields the headline number; sensitivity analysis around it shows feedback effects ranging from $[lower bound] to $[upper bound]. Outside estimates higher than JCT’s upper bound — the Tax Foundation’s, for example — are typically driven by a small-open-economy assumption that the empirical capital-flow literature does not support for a country of this size.

Third, the bill’s cost outside the 10-year window depends on whether its expiring provisions are extended. If they are, the cost roughly doubles by year 20. If they are not, the headline cost is what it is. Advocates who argue for permanence in press releases while scoring on a sunset basis are arguing two incompatible positions in the same week.

The score is the score. The author of the bill does not get to grade it.

(B) CBO Budget and Economic Outlook commentary.

CBO published the [MONTH] [YEAR] Budget and Economic Outlook this week. Three numbers carry the report.

First, projected debt held by the public reaches [X]% of GDP by the end of the 10-year window — above the previous record of 106% in 1946. This is on a current-law baseline, which assumes scheduled expirations occur. If the [TCJA / IRA / other] expirations are extended, the figure rises to [Y]%.

Second, projected interest costs reach [Z]% of GDP — exceeding defense spending [or already have, depending on year] and approaching [Medicare/SS] outlays. Interest is the spending category with the least programmatic content. It is what the federal government pays for its past borrowing.

Third, projected revenue reaches [A]% of GDP — [above/below] the postwar average of about 17.3%. Revenue projections are sensitive to two assumptions: real wage growth and the realization rate on capital gains.

Two things this report is not. It is not a forecast of what will happen. CBO has been clear, since Rivlin, that the baseline is a benchmark — what would happen if current law held. It is not a recommendation. CBO does not recommend.

What the report is: a description of the trajectory the law currently on the books produces. The trajectory is unsustainable in the technical sense — debt rising as a share of GDP indefinitely. There are three responses to that finding: raise revenue, cut spending, accept higher debt and the higher interest costs that come with it. There is no fourth response. Anyone who tells you there is should be asked to identify which of the three categories their proposal falls into. If they cannot, they are not proposing a fiscal plan. They are proposing a press release.

(C) Regulatory-impact analysis dispute.

The agency’s [proposed/final] rule on [SUBJECT] carries an OMB Circular A-4-compliant Regulatory Impact Analysis. The headline benefit-cost ratio is [N]. Industry comments and the [trade-association] response cite a counter-analysis with a benefit-cost ratio of [M].

The methodological gap between the two numbers is in three places.

First, the discount rate. The agency uses [3% / 7% / both]. The counter-analysis uses [different rate]. Discount-rate choice is a value judgment dressed as a technical parameter. OMB’s longstanding instruction is to present results at both 3% and 7%; analyses that present only the higher rate are choosing a methodology that systematically reduces the present value of long-dated benefits, which in regulatory contexts means health and environmental gains.

Second, the willingness-to-pay valuations. The agency uses the standard EPA value-of-statistical-life parameter, currently around $[X] million, drawn from the meta-analytic literature on wage-risk premiums. The counter-analysis substitutes a lower value drawn from a single industry-funded study. The literature is what it is.

Third, the counterfactual. The agency assumes [baseline assumption]. The counter-analysis assumes a different baseline that already incorporates voluntary industry compliance — a methodological move that mechanically reduces the rule’s incremental benefit by attributing the benefit to a counterfactual the rule does not face.

None of this is novel. The same three methodological moves recur in regulatory-impact disputes across decades and agencies. The question for the reader is not which number is right in the abstract. It is whether the analyst defending the lower benefit-cost ratio is willing to apply the same discount rate, the same willingness-to-pay value, and the same baseline construction to the rules they support. Usually they are not.

(D) Right-leaning think-tank report takedown.

[THINK TANK] released a paper this week titled “[TITLE]” arguing that extending [TCJA / corporate-rate-cut / capital-gains preference] would produce $[X] of additional revenue, $[Y] of new investment, and $[Z] of wage gains — and that the JCT and CBO scores understate these effects.

The paper has three methodological tells.

First, the corporate-tax-incidence split. The paper assumes labor bears 70% of the corporate tax. JCT uses 25%. Treasury uses 18–20% on normal returns and 0% on super-normal. The Tax Policy Center uses 20%. The empirical literature — surveyed by Gravelle (2010, CBO Working Paper), Auerbach (2018, NTJ), and the Treasury Office of Tax Analysis (Cronin et al. 2013) — places the share well below 70%. The 70% figure is reachable only under a small-open-economy assumption with infinite capital-supply elasticity. The United States is not a small open economy; it is approximately a quarter of global output. The methodological choice that produces the headline wage-gain finding is the choice the empirical literature does not support.

Second, the elasticity of taxable income. The paper assumes an ETI of 0.4 across the income distribution. The Saez-Slemrod-Giertz (2012, JEL) survey places the central estimate at 0.25, with most of the action concentrated at the top of the distribution and most of the response reflecting income-shifting rather than real activity. An ETI of 0.4 produces revenue feedback approximately 60% larger than an ETI of 0.25.

Third, the dynamic-feedback model. The paper does not specify which model produces the GDP estimate. JCT’s three models — MEG, OLG, DSGE — produce a range; the paper’s number sits above the highest of the three. The standard practice in policy-relevant dynamic analysis is to publish the model, the parameters, and the sensitivity; the paper publishes the headline.

A final point. The paper’s lead author was a senior official at [Treasury / OMB / a CEA position] under the [Bush / Trump] administration. The paper is published by an institution whose donor base has substantial overlap with the entities that would benefit from the policy the paper advocates. None of this makes the analysis wrong. It does mean the analysis is advocacy, and the methodological choices are the ones that produce the conclusion the advocates wanted.

The 1981 cuts were sold with arithmetic of this kind. Stockman, who designed them, later wrote that he “out-and-out cooked the books.” The 2001 cuts were sold with arithmetic of this kind. Bartlett, who helped design them, later wrote that the Republican tax myth he had helped create was “hogwash” and “a lie.” The 2017 cuts were sold with arithmetic of this kind. JCT scored them at a $385 billion dynamic offset against a $1.46 trillion static cost; the Tax Foundation projected three times that growth effect; the actual outturn looks closer to JCT’s number. The pattern is forty-five years old.

The score is the score.

(E) Left-leaning think-tank report takedown.

[LEFT-LEANING THINK TANK] released a brief this week arguing that [PROPOSED PROGRAM] would generate $[X] in dynamic offsets through [labor-force participation / human-capital / multiplier] effects, reducing the headline cost by [Y]%.

Two methodological tells.

First, the multiplier. The brief uses a fiscal multiplier of [1.5–2.0] for the program’s transfer component. The empirical literature on transfer multipliers — Chodorow-Reich on state-level Recovery Act spending; the IMF’s post-2010 multiplier work — supports multipliers in this range during recessions and slack labor markets. It does not support them at full employment. The brief does not condition its multiplier on the cyclical state of the economy.

Second, the long-run human-capital offset. The brief cites Hoynes, Schanzenbach, and Almond on long-run SNAP effects and Chetty, Friedman, Saez et al. on EITC effects to argue that the program’s long-run offset reduces its net cost. The cited literature is real and the effects are real; the magnitudes the brief assigns to those effects exceed the magnitudes the cited authors estimate. This is the citation-mining problem in mirror image: a real literature deployed to support a number larger than the literature actually produced.

The brief’s underlying point — that means-tested programs with labor-supply and human-capital benefits are cheaper net than their gross cost — is correct. The headline number overshoots what the evidence supports. CBO and JCT will score the program against current law without these dynamic offsets; that is the convention. Advocates who want CBO to incorporate the offsets must persuade CBO that the offsets are estimable, not assume them in a brief.

The convention is the convention. It applies symmetrically.