What Lasts

Margins: seven laws on the thinnest votes

Exhibit 43 ranks 16 landmark enactments by the yes share, among members voting, in the narrower chamber on the last recorded vote before enactment, from the Social Security Act of 1935 to P.L. 119-21 of 2025. They are prominent laws chosen by judgment, not a random sample or the output of a rule, and each has a recorded vote or a documented voice vote in a publisher source. The findings apply to these 16.

Nine of the 16 cleared both chambers with at least 68.2% on the recorded votes. The lowest of the nine is the Tax Reform Act of 1986, which passed the House 292–136 on September 25, 1986 (Voteview, roll call RH0990818). The Civil Rights Act of 1964 passed the Senate 73–27 on June 19 and the House 289–126 on July 2, a 69.6% share in the narrower chamber (U.S. Senate Historical Office, Civil Rights Act of 1964; Voteview, roll call RH0880182). Medicare (1965), the Clean Air Amendments (1970) and the Americans with Disabilities Act (ADA, 1990) cleared the House 307–116, 375–1 and 377–28, and the ADA the Senate 91–6 (Voteview, roll calls RH0890097 and RH0910268; House Clerk, roll call 228, July 12, 1990; U.S. Senate, roll call vote 152, July 13, 1990).

Seven laws did not reach 57%. The Affordable Care Act (ACA) passed the House 219–212 on March 21, 2010 (50.8%; House Clerk, roll call 165). Dodd-Frank passed the House 237–192 on June 30, 2010 (55.2%; House Clerk, roll call 413). The Tax Cuts and Jobs Act (TCJA) cleared the Senate 51–48 on December 20, 2017 (51.5%; U.S. Senate, roll call vote 323). The infrastructure law passed the House 228–206 on November 5, 2021 (52.5%; House Clerk, roll call 369), and CHIPS and Science the House 243–187 on July 28, 2022 (56.5%; House Clerk, roll call 404). The Inflation Reduction Act and P.L. 119-21 each passed the Senate 50–50, on August 7, 2022 and July 1, 2025, with the Vice President breaking the tie; tie-breaks are not counted in the shares (U.S. Senate, roll call votes 325 and 372). P.L. 119-21 then cleared the House 218–214 on July 3, 2025 (House Clerk, roll call 190).

Four of the nine rest on partial records. The House approved the 1956 highway act and the CARES Act of 2020 by voice vote, so their Senate votes (89–1 and 96–0) stand in (Federal Highway Administration, This Day in Highway History: June 26; U.S. Senate, roll call vote 80, March 25, 2020). The 1935 and 1970 shares use original passage votes: both chambers adopted the 1935 conference report by voice vote (Social Security Administration, 1935 Congressional Debates on Social Security), and the 1970 conference-report tally was not found.

The narrowing runs along party lines. Exhibit 44 splits the final House vote by party for the 11 laws with a party table. Before 2010 the less supportive party gave at least 49.7% on each of four laws: the Social Security Act of 1935, Medicare, the ADA and the 1996 welfare law, on which Democrats gave 98 of 197 votes (Social Security Administration, 1935 tally and 1965 tally; House Clerk, roll calls 228 (1990) and 383 (1996)). Since 2010 it has given 11.4% or less (Republicans on CHIPS, 24 of 211; House Clerk, roll call 404). Four laws drew no votes from the other party in either chamber: the ACA, the TCJA, the Inflation Reduction Act and P.L. 119-21. The last three were budget reconciliation laws (Congressional Research Service, R48640, R48290 and R48611), a procedure that limits Senate debate to 20 hours and so removes the need for 60 votes to end debate (Congressional Research Service, RL33030). The ACA passed the Senate 60–39 on December 24, 2009 on the votes of the 60 members of the Democratic caucus alone (U.S. Senate, roll call vote 396).

Surface transportation bills show a similar decline since 2005 (Exhibit 45). The House passed the 2005 authorization 412–8 (98.1%) and the 2021 infrastructure act 228–206 (52.5%); the Senate’s shares went from 95.8% (91–4) to 69.7% (69–30) (Congressional Research Service, R48937, September 11, 2026).

These exhibits describe margins and do not explain them. A 2019 study of all laws enacted from 1973 to 2016 finds enacting coalitions nearly as bipartisan as in the 1970s and most laws, landmark laws included, still drawing substantial bipartisan support (Perspectives on Politics, 2019). Five of the seven narrow laws here were enacted after 2016, outside that sample.

Afterward: three markers of durability

Durability can mean three things: a law survives efforts to repeal it, survives in court, or keeps its funding. Exhibit 46 records one marker for each in the sources opened: a later statute, a Supreme Court ruling and a dated clock or cut.

Fifteen of the 16 have been amended, extended or partly repealed by a later public law. The Social Security Act gained disability insurance in 1956 and, in 1983, a higher retirement age and taxation of benefits (Congressional Research Service, R42035). Medicare was expanded in 1972 and gained Part D in 2003 (Centers for Medicare & Medicaid Services, History; Congressional Research Service, R40425). The Civil Rights Act of 1991 added compensatory and punitive damages to the 1964 act, and the Clean Air Act was amended in 1977 and 1990 (EEOC, Civil Rights Act of 1991; EPA, Evolution of the Clean Air Act). Five later acts changed the rate structure of the Tax Reform Act of 1986 (Congressional Research Service, RL30007). The sources opened document no full repeal of any of the 16. P.L. 119-21, enacted in 2025, is the one law with no later amending statute in them.

The ACA faced the most explicit repeal effort. A repeal bill, H.R. 3762, passed both chambers and was vetoed on January 8, 2016 (Congressional Research Service, R43289). The House passed the American Health Care Act 217–213 on May 4, 2017 (House Clerk, roll call 256). The Senate’s last recorded vote on it, 49–51 on July 28, 2017, was on an amendment (U.S. Senate, roll call vote 179; GPO, bill status of H.R. 1628). P.L. 115-97 then effectively eliminated the mandate penalty from tax year 2019 (Congressional Research Service, R45244).

P.L. 119-21 reopened four earlier laws, in both directions. It made the TCJA’s individual rates permanent, raised the CHIPS semiconductor investment credit from 25% to 35%, repealed three electric vehicle credits the Inflation Reduction Act had enacted or expanded and changed its clean electricity and manufacturing credits, and cut the funding cap of the Consumer Financial Protection Bureau (CFPB), created by Dodd-Frank, by 46% for fiscal 2025, from $823 million to $446 million (Congressional Research Service, R48611, R49031, IN12625, IN12624 and R48295). Some amending laws drew votes from both parties. The ADA Amendments Act of 2008, which restored the ADA’s scope, passed the House 402–17, with 174 Republicans and 228 Democrats in favor (EEOC, ADA Amendments Act of 2008; House Clerk, roll call 460). P.L. 115-174, which eased parts of Dodd-Frank, passed the Senate 67–31, with 50 Republicans, 16 Democrats and one independent in favor, and the House 258–159, with 225 Republicans and 33 Democrats (Congressional Research Service, R45073; U.S. Senate, roll call vote 54, March 14, 2018; House Clerk, roll call 216).

In the sources opened, five of the 16 have reached the Supreme Court. The Court upheld the Social Security Act’s old-age benefits on May 24, 1937 and Title II of the Civil Rights Act in Heart of Atlanta Motel v. United States on December 14, 1964 (Supreme Court, Helvering v. Davis, 301 U.S. 619, and 379 U.S. 241). It upheld the ACA’s individual mandate as a tax and limited the Medicaid expansion in the 2012 challenge brought by the National Federation of Independent Business (June 28, 2012), and held in California v. Texas (June 17, 2021) that challengers to the zeroed mandate lacked standing (Supreme Court, 567 U.S. 519, and Nos. 19-840 and 19-1019). In Seila Law v. CFPB (June 29, 2020) it ruled the director’s removal protection unconstitutional and severed it from Dodd-Frank (Supreme Court, No. 19-7). Two decisions, in 1999 and 2002, narrowed the ADA, according to the findings Congress wrote into the 2008 act (EEOC, ADA Amendments Act of 2008).

Ten of the 16 laws have a dated clock or a cut in the sources opened. A clock is a documented date, past or future, at which provisions of the law or a program it created expire, are depleted or need extension; a cut is a documented reduction in its funding. Each item is counted under one law, so the 1956 highway act and the ACA are not counted: the surface transportation extension and the premium credit expiry that touch them are recorded under the 2021 infrastructure law and the Inflation Reduction Act. On October 18, 2026, five items had dates ahead, four had dates that had passed or been resolved, and one was a cut.

In their 2026 annual reports (the Medicare report transmitted June 9, 2026), the Social Security and Medicare Boards of Trustees project the Social Security retirement fund depleted in the fourth quarter of 2032, when 78% of scheduled benefits would be payable, the combined retirement and disability funds in the third quarter of 2034 (83%), and Medicare Hospital Insurance in the second quarter of 2033 (89%) (Exhibit 47; Social Security Administration, A Summary of the 2026 Annual Reports; Centers for Medicare & Medicaid Services, 2026 Medicare Trustees Report). The TANF block grant created by the 1996 welfare law has run on short-term extensions since fiscal 2010; the latest, P.L. 119-75, funds it through December 31, 2026 (Congressional Research Service, R44668, February 10, 2026). The infrastructure law’s surface transportation authorizations were extended through December 11, 2026 (Congressional Research Service, R48937; GPO, bill status of H.R. 6500), and the CHIPS investment credit is limited to projects that begin construction before December 31, 2026 (Congressional Research Service, R49031).

The other five items are past, resolved or a cut. The TCJA’s individual rate cuts were written to expire after 2025, and P.L. 119-21 made them permanent (Congressional Research Service, R48640 and R48611). Many CARES Act provisions were set to expire in 2020 or with the national emergency; some expired, and P.L. 116-260 extended others (Congressional Research Service, R46634 and IN11568). The enhanced ACA premium tax credits, extended through 2025 by the Inflation Reduction Act, expired; KFF reports that marketplace enrollment fell 13% afterward, from 22.1 million in 2025 to 19.2 million in February 2026, and notes that the administration attributes the drop to anti-fraud efforts (Congressional Research Service, R48290; KFF, June 29, 2026). P.L. 119-21’s own one-year ban on Medicaid payments to certain reproductive health providers was set to expire on July 3, 2026 (KFF, June 25, 2026). The fifth is the CFPB cut above. Annual appropriations also lapsed: CRS counts 42 full days from October 1 to November 12, 2025, and 75 for the Department of Homeland Security from February 14 to April 30, 2026 (Congressional Research Service, RS20348, May 26, 2026).

The count of 15 of 16 shows that landmark laws are rarely left alone. It does not show which kind is reopened more. The nine broad-margin laws have had roughly 6 to 91 years to be reopened and the seven narrow ones roughly 1 to 16; the count does not separate an expansion from a repeal; and the set has no comparison group of laws that were not chosen. Six of the seven narrow laws have been changed, and all nine of the others, so the data do not support a claim that thin-margin laws are reopened more often. In the sources opened, none of the Supreme Court rulings on the five laws struck a law down whole.

Formal studies tie durability to conditions at enactment and to later turnover. The abstracts read for this study do not test a 50.8% law against a 69.6% one. A 2008 study in the American Journal of Political Science argues that divided government at enactment and House-Senate ideological disagreement influence the likelihood that a law is amended, and reports that laws crafted by diverse political coalitions are less durable than those crafted by strong, unified ones (American Journal of Political Science, 2008). A 2010 study of landmark laws enacted from 1951 to 2006 finds that divided government constrains repeal and yields more durable legislation over the long term (American Politics Research, 2010). A study of programs established from 1971 to 2003 finds them likelier to be restructured, ended or cut when a later Congress’s partisan composition differs substantially from the enacting Congress’s (American Journal of Political Science, 2009).

What was tried and dropped

The stalled list holds 13 efforts that reached recorded floor votes. They were selected by judgment from the recorded votes in the sources opened, so efforts that never reached a vote are absent, which partly explains the majorities. Of the 12 other than the Equal Rights Amendment (ERA), 11 won a majority of those voting in at least one chamber. The exception is the Border Act of 2024 (S. 4361), on which cloture on the motion to proceed failed 43–50 on May 23, 2024 (U.S. Senate, roll call vote 182).

Comprehensive immigration bills passed the Senate 62–36 (S. 2611, May 25, 2006) and 68–32 (S. 744, June 27, 2013); the bill-status record of each ends with Senate passage and lists no House floor action (U.S. Senate, roll call votes 157 of 2006 and 168 of 2013; GPO, bill status of S. 2611 and S. 744). Constitutional amendments need two-thirds of each chamber. The House passed a balanced budget amendment 300–132 on January 26, 1995 (House Clerk, roll call 51). The Senate then voted 65–35 on March 2, 1995, two votes short of the 67 needed, and 66–34 on March 4, 1997, one vote short (U.S. Senate, roll call votes 98 of 1995 and 24 of 1997). CRS reported in August 1998 that the House Judiciary Committee had begun a markup on February 5, 1997 and recessed without conclusion, with no further action (Congressional Research Service, 98-671). A term-limits amendment drew 227–204 in the House on March 29, 1995, 61 votes short of the 288 needed (House Clerk, roll call 277).

Congress proposed the ERA on March 22, 1972, and in 1978 extended its ratification deadline to June 30, 1982 (House 233–189, Senate 60–36; Congressional Research Service, R47619). Virginia became the 38th state to ratify on January 27, 2020, after five states had rescinded (Congressional Research Service, R47619). On December 17, 2024 the National Archives stated that the amendment cannot be certified (National Archives, December 17, 2024).

Several bills passed the House and stopped in the Senate (Exhibit 48). D.C. statehood passed 232–180 on June 26, 2020 and 216–208 on April 22, 2021; the Senate version received a committee hearing on June 22, 2021 and no floor vote (House Clerk, roll calls 122 of 2020 and 132 of 2021; GPO, bill status of S. 51). The For the People Act passed the House 220–210 on March 3, 2021, and the Senate vote to begin debate was 50–50 on June 22, 2021, against 60 needed (House Clerk, roll call 62; U.S. Senate, roll call vote 246). The House-passed Build Back Better Act (220–213, November 19, 2021) received no Senate vote on its text (House Clerk, roll call 385). The House passed S. 1383 with an amendment titled the SAVE America Act, 218–213, on February 11, 2026; the Clerk lists the vote under the bill’s own title, the Veterans Accessibility Advisory Committee Act (House Clerk, roll call 69; GPO, bill status of S. 1383). In the Senate, cloture on a photo-identification amendment to S. 1383 drew 53–47 on March 26, and cloture on the motion to proceed to S. 5271, a separate photo-identification bill, drew 52–46 on August 8, 2026; both fell short of 60 (U.S. Senate, roll call votes 73 and 231 of 2026). The Senate rejected the background-check amendment (S.Amdt. 715 to S. 649) 54–46 on April 17, 2013, also needing 60 (U.S. Senate, roll call vote 97 of 2013).

The seven Senate votes fell 1 to 17 votes short of their thresholds, and five of the seven had a majority of those voting in favor. The list covers efforts that reached a recorded vote in the sources opened and omits those that never did, so it cannot give a failure rate.

Two readings of a narrow margin

Supporters of lawmaking by narrow majorities start from the principle stated in Federalist No. 22, writing about the Articles of Confederation: the “sense of the majority should prevail,” and a rule requiring more than a majority gives “a minority a negative upon the majority” (The Avalon Project, The Federalist No. 22). On this reading the thin margins in Exhibit 43 show elected majorities doing what they were chosen to do, and the cost of the alternative is visible in Exhibit 48, where five of seven blocked Senate votes had a majority in favor. The academic record gives them partial support: the 2008 study reports that laws from strong, unified coalitions are better placed to entrench their policies than laws from diverse ones, though its abstract ties diversity to divided government and House-Senate disagreement, not to the minority party’s share of votes (American Journal of Political Science, 2008).

Critics start from the argument of Federalist No. 62 that “a continual change even of good measures is inconsistent with every rule of prudence and every prospect of success,” and that under “incessant changes” no one can say what the law will be tomorrow (The Avalon Project, The Federalist No. 62). On this reading a law passed without the other party invites reversal at the next change of control. The record in Exhibit 46 contains instances: the vetoed ACA repeal bill of 2016, and P.L. 119-21’s repeal of Inflation Reduction Act credits and its cut to the CFPB’s funding cap. The finding on partisan turnover points the same way.

The data settle the margins and the facts of each reopening. They do not settle whether narrowness caused any of it, since the nine broad-margin laws were reopened too, or whether durability is the right goal. Stability and responsiveness are values, and no table here ranks them. P.L. 119-21 shows the difficulty: it made one party-line law permanent, raised a credit in a second narrow-margin law, and cut or repealed parts of two others. A later majority sustained some of what earlier majorities built and removed the rest.

Options for making laws last

Four changes are commonly proposed to make major laws more stable or better understood. This study contains no direct test of any of them.

Supermajority requirements for major changes. The strongest case for is that a law changeable only with three-fifths support must carry part of the other party, which raises the cost of reversal after a change in control and serves the concern for stability in Federalist No. 62. The strongest case against is the one in Federalist No. 22: a supermajority rule gives a minority a negative, and existing thresholds already blocked five majority-backed Senate votes in Exhibit 48. Reconciliation shows how a threshold shapes drafting. CRS notes that Byrd-rule waivers need 60 votes and that the TCJA’s individual rate cuts were made temporary to avoid violating the rule, so a procedural limit produced a sunset (Congressional Research Service, R48640).

Sunset reviews. The strongest case for is that expiry forces a vote. A 2007 law-review study argues that timing rules reduce agency problems within the legislature and mitigate deliberative pathologies, and an expiry date is one kind of timing rule (Harvard Law Review, 2007). CBO counted 1,447 authorizations of appropriations that had expired before 2026 began. Of these, 512 were associated with FY2026 appropriations totaling $494 billion, and 308 more were set to expire during 2026 (Exhibit 49; Congressional Budget Office, Expired and Expiring Authorizations of Appropriations: 2026 Final Report, August 28, 2026). CBO adds that the chamber rules on appropriating for expired authorizations have often been waived. The strongest case against is the record of short extensions: TANF ran on 13 temporary extensions from 2002 to 2006, and the 2005 highway act on 10, so expiry has produced patches, and the TCJA’s individual provisions were set to expire for a procedural reason, not an evaluative one (Congressional Research Service, R44668 and R48937).

Post-enactment evaluation. The strongest case for is that the two readings above argue from outcomes neither can measure, and evaluation supplies them. GAO reported on August 20, 2024 that of 23 agencies it reviewed, 21 said the assessment process required by the Evidence Act helped them better understand what they were doing to collect and use evidence (GAO-24-106982). The strongest case against is capacity and comparability. Eighteen of the 23 had trouble interpreting guidance and doing their assessments, and GAO found variation in what agencies assessed and how they presented their findings (GAO-24-106982). The assessments also address agencies’ evidence capacity, not the effect of a given law, and evaluation informs votes without binding them.

Regular-order amendment. CRS describes regular order as a traditional, committee-centered process of hearings, markups and committee-reported bills, and notes that Congress has become largely party-centered (Congressional Research Service, R46597). The strongest case for is that wider participation is the route to wider coalitions. The strongest case against is that both chambers have adapted to polarization by setting regular order aside as circumstances warrant, so a rule would meet the same incentives. The 2008 study finds laws crafted by diverse coalitions less durable, but its abstract ties diversity to divided government and House-Senate disagreement, not to committee process, so it does not test whether wider participation makes laws last.

Dated clocks run in the coming months: the surface transportation extension ends December 11, 2026 (Congressional Research Service, R48937), TANF funding runs through December 31, 2026 (Congressional Research Service, R44668), and the 35% CHIPS credit is limited to projects that begin construction before December 31, 2026 (Congressional Research Service, R49031). Fiscal 2026, which ended September 30, 2026, was the final year for which the Commerce Department is to receive CHIPS incentive funding (Congressional Research Service, R49031).