Why not evaluate with AI?
The same arguments I gave in Governing Advice... from AI?
Again, this isn't a rigorous experiment. It's a musing.
The question
During presidential term, the US president administration takes many actions. Some are difficult to judge, some are not significant. Yet there are actions that, based on existing economic, social and political knowledge can be objectively judged as either negative or positive for long term (half-century) flourishing of the population of the country.
I asked three frontier models to select all such actions of the four last US administrations (the last one so far during this term), and score each action from -5 to +5 with a very brief comment. No need to wade into actions that are impossible currently to judge.
The models were asked separately and did not see each other's answers.
Caveat
The comparisons should be interpreted with caution, because the earlier the administration term, the more its results were already possible to verify.
Three of these terms are finished and can be scored partly on what happened. The fourth is twenty months old and is scored on projections, which may overstate harm.
Summary of results
Histograms
First, a simple look at the histograms of the scores:
Conclusions
Each of the models made its own selection of the consequential actions. We see this from the histograms, and also from the lists of the selected actions, given in the next section. The top right corner of each histogram shows the number of actions n and the mean score mean.
Nevertheless, all three models agree on this:
- 1.
The mean scores of Obama and Biden administrations are close; the Trump's first term is much worse.
- 2.
Trump's second term is already much worse than its first term. In less than half of Trump's second term, the administration has done more damage than over the whole first term.
- 3.
Trump's second term dominates the most damaging kinds of actions, classified as -5, -4 and -3.
Mistral results are less robust, easily swayed by critique and suggestions. Leaving it out, Opus and GPT agree on three most damaging items of Trump-II, all of long-term accumulative effect:
- 1.
Federal research funding cuts.
- 2.
Structural deficits from the 2025 reconciliation law.
- 3.
Reduction in legal and skilled immigration (distinct from border enforcement).
If these three items are as damaging as scored, specific indicators should move in the next few years:
- 1.
Research: US share of highly-cited publications and of first-time NIH R01 awardees, and net flow of postdocs to the EU and China, by 2030.
- 2.
Deficits: the ten-year Treasury term premium and net interest as a share of revenue.
- 3.
Immigration: prime-age labor force growth and new-business formation, against the CBO 2024 baseline.
There is also a less obvious agreement on two themes that were split on different issues:
- 1.
Administrative capacity. (statistical agencies −3, workforce −2, inspectors general -2; merit-system protections −4, workforce reductions −3, independent-agency control −3)
- 2.
Climate. (endangerment finding −4, clean-electricity credits −3; Paris and climate retreat −3)
Comparison of aggregate scores
Claude Opus 5
Administration | Items | Sum | Mean
Obama | 17 | −2 | −0.12
Trump 1 | 21 | −15 | −0.71
Biden | 19 | −1 | −0.05
Trump 2 | 27 | −57 | −2.11
GPT-5.6 Luna
Administration | Items | Sum | Mean
Obama | 15 | 17 | 1.13
Trump 1 | 17 | −4 | −0.24
Biden | 12 | 14 | 1.17
Trump 2 | 15 | −18 | −1.20
Mistral Medium 3.5
Administration | Items | Sum | Mean
Obama | 12 | 40 | 3.33
Trump 1 | 10 | −9 | −0.90
Biden | 10 | 39 | 3.90
Trump 2 | 11 | −23 | −2.09
The models explained their scoring
Each model explained how it assigned the scores.
Claude Opus 5
Actions judged for their long-run effect on the flourishing of the US population over a half-century, on a scale from −5 to +5.
An action is included only if all four hold: it is attributable to the administration as a choice rather than a circumstance; the direction of its long-run effect is not contested among serious researchers (contested magnitude is fine, contested sign is disqualifying); there is an identified persistence mechanism carrying the effect to 2075 — mortality in a living cohort, capital stock, human capital, emissions stock, debt stock, or institutional norm; and the magnitude reaches the ±1 bar at national scale.
Score anchors: ±5 — roughly 250k lives or more, ~2% permanent change in national income, or a change to a foundational institution that alters the feasible set of all future policy, not readily reversible. ±4 — 50–250k lives, 1–2% of long-run income, or major structural change reversible only at high cost. ±3 — 10–50k lives, 0.3–1% of income, or tens of millions materially affected, reversible at moderate cost. ±2 — sectoral or defined-population, measurable, reversible. ±1 — directionally clear, small at national scale.
This is what the mainstream body of economic, epidemiological and political-science evidence would predict, not a personal political verdict. The direction of most items is well supported; the magnitudes are crude judgment calls.
GPT-5.6 Luna
Actions judged for their long-run effect on the flourishing of the U.S. population over roughly half a century, on a scale from −5 to +5.
An action is included only if it is reasonably attributable to the administration, its long-run direction is sufficiently clear from established evidence, and there is a plausible persistence mechanism carrying the effect into the future—such as effects on mortality, productivity/capital, human capital, public health, environment, national security, or institutions. I exclude actions where the sign is genuinely too uncertain.
Score anchors: ±5 — potentially foundational, affecting a very large share of the population or an enduring national institution; ±4 — very large national effect or major structural change; ±3 — large and durable national effect; ±2 — moderate, measurable effect on a substantial sector or population; ±1 — relatively small but directionally clear national effect.
The score represents the estimated magnitude of the long-run effect, not confidence in the estimate. This is an evidence-based judgment rather than a personal political verdict; the direction is generally more defensible than the precise magnitude.
Mistral Medium 3.5
Actions scored for their long-run effect on U.S. population flourishing over half a century, on a −5 to +5 scale, based on mainstream economic, social, and environmental research.
Inclusion criteria: Attributable to the administration, with clear direction of long-run impact (consensus among experts), persistence (e.g., institutional, environmental, or economic shifts), and national-scale magnitude.
Score anchors:
±5: Transformative (e.g., major institutional shifts, large-scale climate/health impacts).
±4: Significant structural change (e.g., major policy reforms, broad economic/environmental effects).
±3: Notable but reversible (e.g., sectoral reforms, moderate social/economic shifts).
±2: Limited but measurable (e.g., targeted policies, niche impacts).
±1: Directionally clear, minor at national scale.
Scores reflect consensus-based predictions, not political judgment. Magnitudes are approximate.
Actions, selected and described by Claude Opus 5
Obama (2009–2017)
17 items. Sum −2, mean −0.12.
Affordable Care Act — +5
Around 20 million previously uninsured people gained coverage, and Medicaid expansion is associated with reduced overall mortality in national cohort studies. It clears the ±5 anchor on lives and on institutional durability alike: it survived a repeal attempt and reshaped what later health policy could assume as a baseline.
Dodd-Frank capital requirements — +4
Common equity tier 1 at large bank holding companies rose from about 4.8% at the end of 2008 to about 11.8% by 2015, and the system absorbed both 2020 and 2023 without a systemic event. Reversible only at high cost, and the partial rollback since 2018 has not undone the core.
American Recovery and Reinvestment Act — +3
Averted a depression, and the $90B clean-energy tranche seeded cost declines of 41% in land-based wind and 64% in utility-scale solar by 2016. Held at +3 rather than higher because of the undersizing item below — the same act is scored twice, once for what it did and once for what it failed to do.
Vehicle fuel economy and efficiency standards — +2
Downgraded from an initial +3. Reversal in 2019 and again in 2026 cut the realized emissions-stock effect to roughly sectoral scale, which is the +2 anchor rather than the ±3 one.
Paris Agreement and the 2014 US–China deal — +2
Scored only for effects reaching Americans: the architecture has survived two US withdrawals and continues to constrain the emissions stock that drives US climate damages. Non-binding by design, which is why it sits no higher.
Auto industry restructuring — +2
Preserved a supplier network and industrial base that later underpinned both the EV buildout and defense manufacturing. Defined-population and sectoral, so +2 by the anchor even though the counterfactual was severe.
Ebola response and the Global Health Security Agenda — +2
Contained effectively in 2014, and built durable outbreak-response architecture. Included on US-population grounds only — imported-case risk to Americans — which is why it scores at +2 rather than reflecting the full global benefit.
Fair Sentencing Act — +1
Reduced the crack-powder sentencing disparity from 100:1 to 18:1, retroactively for thousands of people. Directionally clear, small at national scale.
Direct student lending and Pell expansion — +1
Removed the bank-subsidy middle layer from federal student lending and expanded grant aid. Human-capital channel, modest magnitude.
Targeted-killing framework — −2
Downgraded from −3. The norm cost is real and has been inherited by every successor, but the tens-of-millions-affected threshold the ±3 anchor requires cannot be defended here.
Bulk surveillance programs — −2
Institutionalized collection programs, partly curtailed by the 2015 USA Freedom Act after public disclosure rather than by choice. Norm channel, moderate reversibility.
Leaving Fannie and Freddie in conservatorship — −2
Eighteen years of unresolved federal exposure covering roughly half the mortgage market, with no legislative resolution attempted. The sign is not contested; the magnitude is.
2013 permanence of most Bush tax cuts — −3
Locked in a structurally inadequate revenue base — the baseline against which every later deficit is measured. Divided government constrained the choice, but the administration negotiated and signed it.
Non-prosecution of financial-crisis executives — −3
Foreclosures for households, no indictments for principals. The persistence mechanism here is the norm of equal application of law, the same asset weighted most heavily elsewhere in this exercise.
HAMP and the mortgage restructuring failure — −4
1.6 million permanent modifications against 6.5 million homes lost, while $16B flowed to servicers and principal was never written down. Generational wealth destruction concentrated among lower-income and minority households.
Undersized stimulus and the 2011 austerity turn — −4
The consensus is now that ARRA was too small, producing a slow and painful recovery with prolonged elevated unemployment; the Budget Control Act and sequestration then compressed discretionary spending, including research, for a decade. Labor-market scarring is the persistence mechanism.
Opioid under-response — −4
Overdose deaths roughly doubled across the two terms, with substantive federal action arriving only in 2016. Comfortably inside the 50–250k lives band, for a cohort still alive in 2075.
Excluded under the contested-sign criterion: the JCPOA, Cuba normalization, Libya, Syria, ESSA, Race to the Top, antitrust posture, net neutrality, deportation levels, and the general reliance on executive action over legislation — the last because whether that method was harmful, or was the only available route, is exactly the kind of dispute the criterion is meant to screen out.
The shape of this record is one very large positive against three −4s, and the three −4s share a single character: all are failures to act at sufficient scale rather than harmful actions taken.
Trump, first term (2017–2021)
21 items. Sum −15, mean −0.71.
Operation Warp Speed — +5
The mRNA vaccines proved 84–86% effective against hospitalization in meta-analysis, with roughly 100,000 American lives saved in the first year of availability. It clears the ±5 lives anchor, and the de-risked advance-purchase model is a reusable capability rather than a one-off. That combination is what earns the top score.
China strategic reorientation — +3
Scored apart from the tariffs. Two successive administrations of opposite parties have kept both the framing and the export-control architecture, which is the durability evidence. The channel is national income via reduced technology transfer, not the trade balance.
CARES Act — +3
Prevented a demand collapse, and measured poverty fell in 2020 despite the shock. Substantial PPP fraud is the reason it sits no higher. Passed with Congress, so attribution is shared.
TCJA business tax reform — +2
Difference-in-differences work estimates corporate investment up 8.2–10.2% by 2019, though CRS notes the larger figure is hard to square with aggregate data. Capital-stock channel. Scored separately from the deficit financing.
First Step Act — +2
Recidivism of 12.4% among roughly 30,000 releases, against an estimated 19.8% for similarly situated pre-Act releases — the methodologically careful comparison rather than the headline one. Bipartisan and still in force.
Hospital and insurer price transparency rules — +2
An information reform retained and extended by both successors. Slow-acting, permanent, and unusually uncontested in direction.
Abraham Accords — +1
Downgraded from +2 under the US-population restriction. The agreements have held through subsequent regional shocks, but the benefit to Americans specifically is indirect and modest.
USMCA — +1
Modernized NAFTA with bipartisan ratification and preserved the North American bloc rather than exiting it. The restrained option, taken.
VA MISSION Act — +1
Access improvement for a defined veteran population.
SUPPORT Act and opioid response buildout — +1
Overdose deaths rose throughout the term, so this scores the apparatus rather than the outcome. The federal treatment infrastructure built here contributed to the post-2023 decline.
Family separation at the border — −2
Downgraded from −3. Documented developmental harm to children, some families never reunited, and a real cost to state legitimacy — but the affected population does not reach the ±3 scale threshold, and the score should not be inflated because the harm was gratuitous.
Doha agreement with the Taliban — −2
Set the withdrawal timeline, excluded the Afghan government, and released 5,000 prisoners. Scored at −2 on US-population grounds only. It is the framework within which the 2021 collapse occurred, so this and the Biden execution item split one causal chain between two administrations.
Paris withdrawal and climate rollbacks — −3
Four lost years on the emissions stock, plus a permission signal to other emitters. The stock channel is why a temporary withdrawal still scores as a permanent cost.
Trade war with China and allies — −3
Import tariffs neither raised nor lowered employment in protected sectors, while retaliatory tariffs did clear damage concentrated in agriculture, only partly offset by $28B in subsidies. The durable cost is normalizing tariffs as a unilateral executive instrument.
Erosion of scientific and statistical capacity — −3
Politicized CDC and FDA communications, sidelined advisory committees, a long-vacant science adviser post. This is the precedent that made the second-term items at −4 and −5 available.
Multilateral withdrawals, including TPP — −3
WHO, INF, Open Skies, UNESCO, TPP. The TPP exit in particular ceded the Asian trade architecture to a China-centred alternative, which partly negates the China reorientation item on this same list.
Legal immigration restriction — −3
Travel bans, a refugee cap cut to 15,000, public charge, H-1B restrictions. The same demographic and innovation mechanism as the second-term item, at smaller scale.
DOJ, inspector general and pardon politicization — −3
Firings of inspectors general and pardons for associates. Norm channel, and a direct precursor to the second-term −5.
TCJA deficit financing — −4
CBO's 2018 update put the conventional ten-year deficit increase near $2.3 trillion. Deficit financing splits the result: long-run output up 1.1% but American incomes up only 0.4%, the gap flowing abroad as interest and profit payments.
Non-vaccine pandemic response — −5
The US recorded 154.5 age-standardized excess deaths per 100,000 in 2020–21 against 110.4 in Europe; matching Europe would have meant roughly 391,000 fewer excess deaths, and the US had the highest premature-death rate among twelve large wealthy peers. The politicization of masks, testing and the CDC also created the partisan health divide that persists today.
Rejection of the 2020 result and January 6 — −5
The peaceful transfer of power was the load-bearing norm of the system, and the break is now observable in its consequences rather than hypothetical. On the institutional anchor — altering the feasible set of all future policy — this is the largest single item across all four administrations.
Excluded under the contested-sign criterion: judicial appointments, and this is by far the most consequential exclusion in the whole exercise. 234 judges including three Supreme Court justices is an effect running well past 2075 and larger in magnitude than most items here, but whether it improves or harms long-run flourishing is a dispute about constitutional interpretation rather than an empirical question. Its absence flatters this total by an amount that cannot be signed, let alone sized. Also excluded: JCPOA withdrawal, the deregulatory agenda, Opportunity Zones, Space Force, Middle East troop posture, Syria.
The distribution is the most polarized of the four — a +5 and two −5s with an empty middle on the negative side. Six of the eight −3s are precursors to items that recur at −4 or −5 in the second term, which means scoring the terms independently understates this one and overstates that one.
Biden (2021–2025)
19 items. Sum −1, mean −0.05.
Inflation Reduction Act, climate and energy provisions — +4
Projected to roughly double emissions reductions by 2035 against a no-IRA baseline, with $493B in actual investment in the first two years. Emissions stock is the persistence mechanism. Held at +4 rather than +5 because subsequent legislation has compressed the credits, so realized value is well below designed value.
CHIPS and Science Act — +4
The US share of advanced logic manufacturing rose from about 12% in 2020 to about 22% by 2026, with three leading-edge fabs in volume production. Capital-stock channel, and physically hard to reverse: fabs stay where they are built.
Infrastructure Investment and Jobs Act — +3
Public capital with 50-year service lives, passed with bipartisan support, which is itself the durability argument. Sits at +3 rather than +4 because per-dollar delivery has been slow.
Medicare drug price negotiation — +3
The first structural change to US pharmaceutical pricing in decades, with permanent statutory authority rather than voluntary commitments. Tens of millions affected, reversible only by legislation.
NATO expansion and alliance repair — +3
Included on US-population grounds: Finland and Sweden joining, and most members reaching the 2% target, lowers the expected cost to Americans of European security over decades. The deterrence channel is the persistence mechanism.
Vaccine distribution execution — +2
Distinct from Warp Speed's development achievement. Getting doses into arms at scale in 2021 was an execution problem solved competently. Defined-population and time-limited, hence +2.
Semiconductor export controls on China — +2
Retained by the successor administration, which is the durability test. The offsetting risk of accelerated Chinese indigenization keeps it from +3.
ACA enhanced subsidies — +1
Real coverage gains, but temporary by construction and now expired. Downgraded from an initial +2 for exactly that reason.
PACT Act — +1
Health benefit to a defined veteran population at $277B over a decade. Directionally clear, small at national scale.
BEAD and NEVI execution failures — −1
Years elapsed with minimal broadband connections and chargers delivered. The lasting cost is to the credibility of state capacity, not to the assets themselves.
Eviction moratorium overreach — −1
Extended past acknowledged legal authority and struck down. A small norm cost, and one that was honestly labeled as such at the time.
American Rescue Plan overshoot — −2
Downgraded from −3, and this is the item where two-sided scoring matters most. Estimates of its inflation contribution run from around 3 percentage points to near-negligible, while the faster return to full employment reduced labor-market scarring. These are netted into one item here rather than scoring the recovery separately, which would double-count in the ARP's favor.
School reopening posture — −2
Downgraded from −3. Measured achievement losses are real and largely unrecovered, but federal leverage over district reopening decisions was genuinely limited, which weakens attribution.
Afghanistan withdrawal execution — −2
The decision to leave and the Doha framework are both excluded, as contested and as attributable elsewhere respectively. This scores only the execution — thirteen US deaths and a credibility cost — which is where criticism is close to universal.
Student debt cancellation by executive order — −2
Struck down by the Supreme Court. The durable cost is the norm: large executive spending without appropriation, the same category scored negatively across all four administrations.
No permitting reform alongside the IRA — −3
Interconnection queues and transmission bottlenecks are now the binding constraint on the IRA's own targets. A failure to act that partly disables the term's largest positive.
Entitlement insolvency deferred — −3
A projected 30-year baseline deficit near $110T left untouched. Each deferred year enlarges the eventual adjustment; the debt stock is the persistence mechanism.
Structural fiscal deterioration — −4
Cumulative 2021–2031 deficits went from a projected $14.5T at entry to $21.2T at exit, leaving structural deficits near $2T. Scored by the same standard as TCJA financing and OBBBA.
Loss of operational border control — −4
8.6 million encounters through October 2024, roughly 5.8 million paroled or released to pursue claims, and a 3-million-case backlog. The deepest long-run cost runs through the norm channel: it eroded public consent for legal immigration and made the subsequent restriction politically available.
Excluded under the contested-sign criterion: Ukraine aid levels and escalation management, Gaza policy, antitrust revival, and most of the regulatory record where post-2025 reversal has made the counterfactual unrecoverable.
The distinguishing feature of this record is the absence of any ±5: nothing here is transformative or catastrophic by the anchors, which is not true of either Trump term. Both −4s are items where the administration's coalition preferences and the harm are the same decision, not side effects of it.
Trump, second term (through September 2026)
27 items. Sum −57, mean −2.11.
Nuclear program — +3
Four executive orders, NRC reform, and the first construction permit for a commercial non-light-water reactor, against a 100→400 GW target by 2050. Clean firm power is the highest-leverage energy decision on this horizon. Safety-review shortcuts are the offsetting risk.
Most-favored-nation drug pricing and TrumpRx — +2
26 manufacturers covering about 89% of the branded market, with drug CPI down 3.1% year-on-year. Real relief, but concentrated in Medicaid and cash purchases, and structured as voluntary commitments rather than statute — hence +2 against the +3 for Biden's negotiation authority.
Fentanyl and counter-narcotics enforcement — +1
Overdose deaths fell to roughly 69,973 in 2025 from 81,313 in 2024. Large in absolute terms, but the decline began in mid-2023, so attribution supports only a modest score.
Energy and transmission permitting acceleration — +1
Addresses a genuine binding constraint — the same one scored as a −3 omission for the previous administration. Partly offset by weakened environmental review.
Restored border order — +1
Scored on the consent channel: orderly borders sustain long-run political support for legal immigration. Distinct from, and partly negated by, the enforcement-methods and legal-immigration items below.
Federal AI and compute buildout — +1
Genesis Mission and national-lab partnerships. Plausible productivity upside, with ratepayer cost-shifting as the offsetting risk, which is why it stays at +1.
Federal workforce reduction — −2
State capacity rebuilds slowly and institutional memory does not return with the budget. Downgraded from an initial scoring: the effect is real but diffuse, and hard to defend at the ±3 threshold.
Residual tariffs — −2
The Supreme Court's 6–3 February 2026 ruling that IEEPA does not authorize tariffs limited the damage. What remains under Sections 232 and 301 cuts long-run GDP by roughly 0.4%.
USAID dissolution and WHO exit — −2
Scored only on the US-population channel: reduced outbreak-detection capacity abroad raises imported-disease risk to Americans. The much larger harm falls outside this exercise's scope.
Immigration enforcement methods — −2
Distinct from the border-order and legal-immigration items. Detention conditions, courthouse arrests and due-process shortcuts carry a norm cost independent of enforcement volume.
Inspector general firings — −2
Removal of internal oversight across agencies. The same category as the first-term −3, scored lower here because the first-term instance already absorbed the precedent-setting cost.
PFAS and air-quality rollbacks — −2
Distinct from the greenhouse-gas items. Contaminant persistence is unusually literal here: PFAS half-lives run to the 2075 horizon and beyond.
Lethal-force framing for interdiction strikes — −2
Extends the first-term and Obama-era targeted-killing precedents to a new category of target. Norm channel, modest magnitude, direction not contested.
Clean-electricity credit compression — −3
Resources for the Future finds the legislation raises electricity prices 2.7 to 3.9 times as much as the standards repeal lowers them. Scored separately from the endangerment finding because the channels differ: prices versus emissions stock.
Federal Reserve independence pressure — −3
A DOJ target letter to the chair coinciding with pressure to cut rates. Anchored inflation expectations are cheap to lose and expensive to rebuild.
Statistical agency integrity — −3
The 2025 BLS commissioner firing alone is estimated at around $20B in GDP through uncertainty, with Census methodology interference following. Measurement is economic infrastructure.
Domestic military deployments — −3
A federal judge found the Los Angeles deployment violated Posse Comitatus, and the Supreme Court rejected the Chicago appeal. The courts held, which caps the score, but the norm was tested repeatedly.
Voting and census interventions — −3
Attempts to change census methodology to affect apportionment, plus new mail-voting restrictions. Electoral-integrity channel.
January 6 pardons — −3
Scored separately from the underlying 2021 events. Pardoning those convicted of assaulting officers converts a one-time norm break into a signal about the future cost of political violence.
OBBBA deficit expansion — −4
An additional $3.4T over 2025–2034, with debt reaching 124% of GDP by 2034. Consumes the fiscal space needed for aging, defense and climate adaptation. Partial credit for averting a scheduled 2026 tax increase is already netted in.
Medicaid, ACA and SNAP retrenchment — −4
Around 10.9 million more uninsured plus 5.1 million from expiring credits, so roughly 16 million by 2034, with SNAP enrollment down about 4.7 million. Childhood coverage and nutrition are among the best-evidenced long-run human-capital investments, which is the persistence mechanism.
Skilled immigration restriction — −4
Projected growth cut from 1.8% to 1.3% annually through 2035, with the first negative net migration in roughly 50 years. Compounds against an aging population across exactly this horizon.
Endangerment finding repeal — −4
All vehicle greenhouse-gas standards rescinded, with EPA declining to estimate public-health or environmental costs. Emissions-stock damages land almost entirely inside the 2075 window.
Childhood vaccine schedule changes — −4
Six vaccines moved to risk-based or shared decision-making over expert warnings of disease resurgence. Herd-immunity losses are slow, cumulative, and hard to reverse once cohort coverage falls.
Coercion of universities — −4
The US decline in the Academic Freedom Index was sharper than that of Hungary, India or Turkey. Research universities are the compounding asset most exposed here.
Federal research funding cuts — −5
NIH awards down 29% and NSF down 50% in 2025, with the fewest early-career researchers funded since 2016. Basic research has among the highest measured long-run social returns, and pipeline damage compounds across decades: a researcher not trained in 2026 is absent for a whole career.
Politicized prosecutions — −5
Criminal investigations into the Fed chair, members of Congress and former officials, with a court throwing out subpoenas against Minnesota officials as retaliatory. Impartial law enforcement is the most robust institutional predictor of long-run prosperity, and this clears the anchor's test of altering the feasible set of all future policy.
Excluded under the contested-sign criterion: the crypto and stablecoin framework, tariff-revenue rebate proposals, Venezuela operations, the Gaza framework, DOGE's specific cuts as distinct from aggregate workforce loss, and the higher-education compact terms, where the effect depends on whether institutions sign.
Three features distinguish this distribution. No item rises above +3, so nothing here is transformative on the upside by the anchors. The −2 bar is taller than any positive bar. And 27 items in 20 months against 17 to 21 items across 48 months for the completed terms is a real difference in pace, which is why the mean rather than the sum is the number to compare.
Two limits specific to this list. It is scored on projections rather than outcomes, and projected harms routinely overstate realized ones, so the eventual figure should be expected to be less negative. Against that, five or six of the first-term −3s are precursors to items scored −4 or −5 here, so a causally honest accounting would shift part of this total backward onto that term.
Actions, selected and described by GPT-5.6 Luna
Obama administration
Affordable Care Act (ACA), 2010 --- +4
Expanded insurance coverage substantially and created mechanisms
intended to restrain health-care cost growth; CBO originally estimated
net deficit reduction as well. This is probably the largest durable
domestic-policy legacy of the administration.
Dodd-Frank financial reform, 2010 --- +3
Strengthened capital/resolution requirements and created institutions
such as the CFPB and FSOC, reducing the probability/severity of another
systemic financial crisis. CBO also estimated modest deficit reduction.
American Recovery and Reinvestment Act (ARRA), 2009 --- +3
Appropriate countercyclical response to the Great Recession, including
infrastructure, state/local support, unemployment assistance and
investment. CBO estimated substantial positive effects on GDP and
employment, although most effects were temporary and the law added
roughly $840B to deficits.
Auto-industry rescue --- +2
Prevented a potentially much more destructive collapse of GM/Chrysler
and preserved industrial capabilities and supply chains during the
financial crisis.
Sustained federal investment in basic science and technology --- +2
Obama repeatedly prioritized NSF, DOE science, NIST, NIH and DARPA-type
research. This is precisely the kind of investment whose payoff can
accumulate over decades.
Climate policy and Paris Agreement --- +2
The Clean Power Plan and Paris Agreement moved the U.S. toward reducing
a major long-lived externality; EPA estimated substantial health
benefits alongside carbon reductions.
Iran nuclear agreement (JCPOA) --- +2
Temporarily constrained Iran's nuclear program through international
monitoring and limits. Avoiding nuclear proliferation has potentially
enormous long-term value, although the agreement's durability was
uncertain.
Dismantling "Don't Ask, Don't Tell" and broader LGBT-equality policy --- +1
Expanded equal treatment and reduced a significant institutional
discrimination; relatively small direct economic effect but positive for
social flourishing.
DACA --- +1
Allowed a large population of people brought to the U.S. as children to
work and study with greater security; positive human-capital
utilization, although the underlying immigration problem remained
unresolved.
Normalizing relations with Cuba --- +1
Reduced a decades-old source of unnecessary hostility and opened
economic/diplomatic channels. The magnitude is modest.
Post-crisis financial/fiscal stabilization --- +1
The administration accepted substantial fiscal consolidation after the
acute recession rather than maintaining emergency stimulus indefinitely.
I give this only +1 because sequestration and the manner of
consolidation also imposed costs.
Libya intervention and subsequent policy failure --- −2
The intervention helped remove Gaddafi but was followed by state
collapse, civil conflict and regional destabilization. The long-term
strategic/humanitarian balance is sufficiently negative for −2.
Afghanistan escalation ("surge") and prolonged war --- −1
The surge temporarily improved security but prolonged a conflict that
ultimately failed to produce a stable Afghan government; substantial
resources were consumed for little lasting strategic gain.
Excessive reliance on drone warfare / expansion of targeted killing --- −1
Some operations had legitimate counterterrorism value, but normalization
of remote targeted killing created legal, institutional and geopolitical
costs that are difficult to reverse.
Sequestration-era reduction in federal investment --- −1
Fiscal consolidation was necessary in the long run, but the blunt
spending caps reduced public investment and government capacity in areas
with potentially high long-term returns. I assign only −1 because
Congress was a major co-author.
Trump I
Operation Warp Speed --- +4
Exceptionally successful acceleration of COVID vaccine development,
manufacturing and distribution; compressing a normally multi-year
process into months has enormous social/economic value.
Tax Cuts and Jobs Act --- investment incentives --- +2
Lower corporate taxation and improved expensing increased incentives to
invest; CBO estimated higher potential GDP and investment.
Deregulation / regulatory cost control --- +2
Reduced the accumulation of federal regulatory costs and some
unnecessary barriers; the benefit is real, although the administration's
headline claims overstate the evidence.
First Step Act --- +2
Significant bipartisan criminal-justice reform: reduced some mandatory
sentences, expanded sentencing relief and improved prison
rehabilitation.
Abraham Accords --- +2
Established normalization between Israel and several Arab states,
producing a durable improvement in regional diplomatic/economic
relations.
Pressure on NATO allies to increase defense spending --- +1
The pressure was abrasive and sometimes counterproductive, but
increasing European military capacity is beneficial to U.S. security and
burden-sharing. Allies' spending was rising during the period, though
much of the trend predated Trump.
Creation of the U.S. Space Force / increased space-security focus --- +1
Recognized space as an increasingly important military and technological
domain; modest positive long-term security value.
Expansion of domestic energy production --- +1
Increased energy supply and reduced dependence on foreign energy
sources; the long-term benefit is partly offset by climate costs.
Broad tariff/trade-war policy --- −4
Tariffs raised prices and distorted investment and supply chains; CBO
estimated lower GDP and household income.
Large deficit increases outside the COVID emergency --- −3
The 2017 tax cuts and subsequent spending increases substantially
increased debt without sufficient offsetting spending cuts or revenues.
CBO estimated roughly $1.9T additional deficits from the 2017 tax act
alone over 2018--28.
Restrictions on immigration and refugee admissions --- −2
Reduced the growth of labor and human capital; particularly costly for
an aging, low-fertility country.
Withdrawal from the Paris Agreement / weakening climate policy --- −2
Moves away from mitigating a large long-term externality whose costs
accumulate over decades. The U.S. formally initiated withdrawal in 2019.
Withdrawal from the Iran nuclear agreement (JCPOA) --- −2
Removed internationally verified constraints on Iran's nuclear program
without establishing an equally effective replacement; increased
proliferation and conflict risk.
Damage to alliance credibility and international institutional relationships --- −3
Repeated threats to leave NATO and public questioning of U.S. security
commitments created a long-term credibility cost, despite the legitimate
burden-sharing objective.
Withdrawal from the INF Treaty --- −1
Russia was indeed violating the treaty, so withdrawal has a defensible
security rationale; nevertheless, ending a major arms-control constraint
without a replacement increased long-term arms-race risk.
COVID-19 response failures apart from Operation Warp Speed --- −1
The vaccine program deserves +4 independently, but the broader federal
response---especially inconsistent public-health communication and
delayed/ineffective coordination---imposed substantial costs. I keep
this at −1 because attribution is unusually difficult during an
unprecedented pandemic.
Family-separation/zero-tolerance immigration policy --- −1
Imposed severe and largely avoidable human and institutional costs for a
policy whose deterrence benefit was limited.
Biden administration
Infrastructure Investment and Jobs Act (2021) --- +3
Large investment in transportation, broadband, water and the electric
grid; infrastructure raises productive capacity over decades.
CHIPS and Science Act / semiconductor industrial policy --- +3
Rebuilds strategically important semiconductor capacity and supports
domestic R&D; reduces dependence on a concentrated Asian supply chain.
Inflation Reduction Act (2022) --- +4
A particularly consequential package: clean-energy investment, climate
mitigation, health subsidies, drug-price reforms and deficit reduction.
Its climate/energy component addresses a very long-lived externality.
Strengthening NATO and the U.S. alliance system --- +3
Rebuilt allied coordination and substantially strengthened the Western
security architecture; Finland and Sweden joined NATO during Biden's
term. The value compounds through deterrence and burden-sharing.
Large-scale support for Ukraine and resistance to Russian expansion --- +3
Helps preserve the European security order and imposes substantial costs
on Russian military power without U.S. troops fighting Russia directly.
NATO allies subsequently committed to sustained long-term assistance.
Higher immigration / expansion of the labor force --- +2
Demographically valuable for an aging country. CBO estimated that the
2021--26 immigration surge increased projected 2024--34 GDP by $8.9T
and reduced federal deficits by about $0.9T.
Creation of ARPA-H and emphasis on high-risk biomedical research --- +2
Creates a DARPA-like institution aimed at transformational biomedical
breakthroughs. Its upside over 30--50 years could be enormous relative
to its cost.
Bipartisan Safer Communities Act --- +1
Modest but directionally positive improvement in gun safety,
mental-health funding and prevention of firearm-related harm.
American Rescue Plan (2021) --- −2
The pandemic support itself was justified, but the additional $1.9T
stimulus came when the economy was already recovering rapidly. CBO's
forecasts subsequently showed substantially higher inflation than
previously expected; the package also added to long-run deficits.
Broad student-loan cancellation and repayment-policy expansion --- −2
Benefits recipients, but transfers large costs to taxpayers and weaken
incentives around future borrowing/tuition pricing. CBO estimated about
$400B additional federal cost from the original cancellation action.
Afghanistan withdrawal --- −2
Ending the 20-year war was defensible and arguably overdue, but the
execution damaged U.S. credibility, produced a chaotic evacuation and
left Afghanistan to the Taliban. I therefore score the
withdrawal/implementation, not the original decision to end the war.
Poor early border/asylum management --- −1
The subsequent immigration surge had substantial economic/demographic
benefits, but the administration's inability to maintain effective
control of the asylum system created significant institutional, housing
and political costs. I score this separately from the positive
demographic effect rather than pretending they were the same thing.
Trump II
Advanced nuclear development and deployment --- +3
Potentially important for cheap reliable power, AI/industry, energy
security and decarbonization.
Faster permitting/buildout of electricity, data-center and related infrastructure --- +3
U.S. electricity and permitting constraints can inhibit investment;
reducing genuine procedural delay has long-run productivity value.
Permanent full expensing of equipment and domestic R&D --- +2
Improves investment incentives. CBO estimates the 2025 law raises
potential output and increases investment in equipment, structures and
intellectual property.
Critical-minerals supply-chain diversification/domestic capacity --- +2
Reduces dependence on strategically concentrated foreign suppliers for
defense, electronics and energy technologies.
Pushing NATO toward substantially higher defense capacity --- +3
NATO members agreed to a 5%-of-GDP defense/security investment
commitment by 2035; stronger allied capabilities increase deterrence
while sharing the U.S. burden.
Re-establishing effective control over unauthorized southern-border entry --- +1
A state benefits in the long run from being able to enforce its
immigration law and determine admission rather than tolerating very
large uncontrolled flows. I keep this only at +1 because the benefits
depend heavily on how enforcement is conducted and what legal
immigration replaces it.
Large disruption and termination of federally supported scientific research --- −5
This is the item I regard as potentially most consequential over 50
years. NIH terminated more than 1,800 grants in just February--June
2025. Damaging the U.S. research ecosystem could have effects far beyond
the immediate dollars involved.
Broad protectionist tariff regime --- −4
Protection can make sense for a few strategic industries; generalized
tariffs are different. CBO finds higher tariffs reduce investment,
productivity and real output, while raising near-term inflation.
Large additional structural deficits/debt from the 2025 reconciliation law --- −4
Some components are good---including the +2 expensing item above---but
financing tax reductions without corresponding durable spending
reductions worsens an already difficult long-term fiscal trajectory.
Large reduction in overall immigration, including students and other legal temporary entrants --- −4
This is separate from border control. In an aging, low-fertility
society, sustained reductions in working-age and high-skill immigration
reduce labor-force growth, entrepreneurship and the tax base.
Weakening merit-system protections and increasing political control of the career civil service --- −4
Making policy-influencing career positions easier to dismiss increases
presidential control but weakens professional continuity, impartial
administration and institutional resilience over decades.
Very large, rapid federal workforce reductions --- −3
Cutting unnecessary bureaucracy can be beneficial; scale and
implementation matter. The workforce at major agencies fell
substantially, creating institutional-knowledge and operational-capacity
risks.
Increasing presidential control over independent regulatory agencies --- −3
There is a legitimate democratic-accountability argument on the other
side. But requiring traditionally independent agencies to submit
regulations and priorities for White House control raises the long-run
risk that enforcement and regulation fluctuate with presidential
political interests.
Paris withdrawal and broader retreat from climate mitigation --- −3
Over a 50-year horizon this becomes easier, rather than harder, to
score. Climate change imposes substantial future economic damages and
downside risks; withdrawal moves U.S. policy in the opposite direction.
Withdrawal from the World Health Organization --- −2
The fiscal saving is tiny relative to the value of international disease
surveillance and coordination during rare but potentially catastrophic
outbreaks.
Actions, selected and described by Mistral Medium 3.5
Obama Administration (2009–2017)
Affordable Care Act (ACA/Obamacare) — +5
Improved public health, reduced financial insecurity, and set a precedent for universal healthcare.
American Recovery and Reinvestment Act — +4
Stabilized the economy, funded infrastructure and clean energy, with long-term growth benefits.
Paris Agreement — +4
Positioned the U.S. as a leader in global climate action, though later reversed.
Clean Power Plan — +4
Pushed utilities toward renewable energy, reducing long-term climate risks.
Dodd-Frank Wall Street Reform — +4
Reduced systemic risk, protected consumers, and improved financial stability.
Ending "Don't Ask, Don't Tell" — +3
Advanced social equality and military readiness, with lasting cultural impact.
Iran Nuclear Deal (JCPOA) — +3
Reduced short-term nuclear proliferation risks, though long-term stability was debated.
Normalization with Cuba — +2
Improved regional relations, but economic impact was limited by later reversals.
Deferred Action for Childhood Arrivals (DACA) — +4
Supported workforce integration and social cohesion, though politically contested.
Auto Industry Bailout — +3
Prevented economic collapse in key industries, with long-term benefits for manufacturing.
Lilly Ledbetter Fair Pay Act — +2
Advanced gender pay equity, though enforcement and impact varied.
Every Student Succeeds Act — +2
Shifted focus to state-level education standards, with mixed long-term outcomes.
Trump Administration (2017–2021)
Tax Cuts and Jobs Act — +2
Stimulated short-term growth but increased deficit; long-term economic impact debated.
Withdrawal from the Paris Agreement — -4
Undermined global climate efforts, increased long-term environmental risks.
Deregulation — -4
Reduced business costs but weakened consumer and environmental protections.
Immigration policies — -5
Harmed social cohesion, damaged U.S. reputation, and created long-term humanitarian concerns.
Criminal Justice Reform (First Step Act) — +4
Advanced fairness, reduced recidivism, and improved long-term social outcomes.
Appointing 3 Supreme Court Justices — -3
Long-term impact on legal precedents (e.g., abortion, voting rights, regulations).
Trade Policies — -2
Short-term protectionism, but long-term trade instability and economic costs.
Operation Warp Speed — +5
Saved lives, expedited global pandemic response, and set a precedent for rapid vaccine development.
Rollback of Clean Power Plan and 100+ environmental rules — -5
Increased emissions, undermined climate goals, and risked public health.
Creation of Space Force — +3
Strengthened U.S. military presence in space, with long-term strategic benefits.
Biden Administration (2021–2025)
Inflation Reduction Act — +5
Accelerates green transition, reduces climate risks, and positions U.S. as a clean energy leader.
Infrastructure Investment and Jobs Act — +5
Modernizes infrastructure, boosts productivity, and supports long-term economic resilience.
Rejoining the Paris Agreement — +4
Restores U.S. credibility in global climate efforts, aligning with long-term environmental goals.
Conservation: 12.5M acres protected, co-management with Indigenous peoples, old-growth forest logging ban — +5
Preserves biodiversity, mitigates climate change, and respects Indigenous rights.
Student loan forgiveness — +3
Reduces debt burden, but fiscal sustainability and equity are debated.
Ending and reversing deforestation/land degradation by 2030 — +4
Supports global ecosystem health and climate mitigation.
Tax plan to replace fossil fuel subsidies with green energy incentives — +4
Shifts economic incentives toward sustainability, but faces political hurdles.
American Rescue Plan — +3
Short-term relief with long-term benefits for poverty reduction and economic stability.
Immigration surge — +4
Boosts labor force, economic growth, and innovation potential.
Removal of Cuba from state sponsors of terrorism list — +2
Improves diplomatic relations, but long-term economic impact is limited.
Trump Administration (2025–2026)
Mortgage-market intervention — +5
Lowering mortgage rates supports long-term housing stability and economic growth.
Creation of a U.S. strategic critical-minerals reserve — +4
Secures supply chains for tech and defense, reducing geopolitical vulnerabilities.
Mass deportation and immigration crackdown — -5
Reduces labor force diversity, harms demographic growth, and risks long-term social cohesion.
Ending federal EV incentives and pro-climate subsidies — -5
Slows green transition, increases long-term climate risks, and cedes global clean energy leadership.
Expanding domestic oil production — -5
Increases carbon emissions, undermines global climate goals, and risks stranded assets.
Blanket pardon of Jan. 6 defendants — -5
Weakens rule of law, encourages political instability, and erodes trust in institutions.
Tariffs on foreign trade — -4
Risks trade wars, reduces global competitiveness, and harms long-term economic partnerships.
Reduction of federal government size — -3
May weaken regulatory capacity, public services, and crisis response over decades.
Withdrawing from international organizations/treaties — -4
Isolates U.S. diplomatically, reduces global influence, and undermines multilateral problem-solving.
Terminating Environmental and Climate Justice Block Grant Program — -5
Reverses progress on environmental justice, disproportionately harming vulnerable communities.
Example of defense of the scoring system (Mistral Medium 3.5)
As explained by Mistral Medium 3.5:
Under my own scoring system, the February 2026 IEEPA Supreme Court ruling does not invalidate the mortgage-market intervention. Here’s why:
The IEEPA ruling specifically struck down tariffs imposed under the International Emergency Economic Powers Act, declaring them unlawful and forcing their revocation by executive orderwilmerhale.com+1. This ruling only applied to trade/tariff actions under IEEPA, not housing finance or FHFA-directed policies.
The mortgage-market intervention (Fannie Mae/Freddie Mac’s $200B purchase of mortgage-backed securities) was executed under FHFA authority, not IEEPA. It was a separate, housing-specific action to lower mortgage rates, confirmed by multiple sources (e.g., White House, NPR, CNBC) whitehouse.gov+2. No legal challenge or ruling has been reported that ties this intervention to IEEPA or overturns it.
Conclusion: The mortgage-market intervention remains a valid, attributable action under my scoring system, as it was not based on IEEPA and was not affected by the February 2026 ruling. The score stands unless new evidence emerges linking the two.