Promises and Passage

What platforms say and what laws do

A party platform is adopted at the presidential nominating convention every four years. The Comparative Agendas Project splits each Democratic and Republican platform since 1948 into quasi-sentences, a new unit after each period or semicolon, and gives each one a policy topic from the same scheme it uses for public laws and roll-call votes (Comparative Agendas Project, Democratic and Republican platform files; U.S. Party Platforms Data Codebook, 2016). That common scheme allows a direct comparison of what a platform talks about with what the next Congress votes on and enacts.

Platforms have grown and shrunk (Exhibit 1). The Democratic platform ran 193 quasi-sentences in 1948 and 2,028 in 2024, its longest (Comparative Agendas Project, Democratic platform file). The Republican platform peaked at 1,910 in 1988; the 2024 text ran 320, the party’s shortest since 1952 (Comparative Agendas Project, Republican platform file). In 2020 the Republican convention adjourned without adopting a new platform and kept the 2016 text, which ran 1,645 (American Presidency Project, Resolution Regarding the Republican Party Platform, August 24, 2020). Length is not commitment, and a short platform can carry more specific pledges than a long one.

The comparison of topics uses an overlap index: 100 minus half the sum of the absolute differences between a platform’s topic shares and the topic shares of the non-commemorative public laws of the next Congress, across 20 major topics. Two identical agendas score 100 and two with no topic in common score 0. Winning platforms scored between 37.3 (1948) and 71.0 (2016) (Comparative Agendas Project, Democratic and Republican platform and public law files). The winner’s platform scored higher than the loser’s after 10 of the 19 elections, lower after 8, and the two were equal to one decimal place after 1964 (Exhibit 2; Comparative Agendas Project, Democratic and Republican platform and public law files). On a 0 to 100 overlap scale built from Comparative Agendas Project topic codes, the averages were 55.0 for winners and 55.2 for losers (the same files). Republican platforms scored higher than Democratic ones in 14 of the 19 elections, whichever party won, averaging 57.1 against 53.2, so part of what the index measures is how each party writes its platform (the same files).

Control of Congress does not sort the results either. In the 10 elections after which the winner’s party held the House and the Senate, its platform was closer to the next Congress’s laws than the loser’s in 4, farther in 5 and level in 1, by an average of 1.8 points less (Comparative Agendas Project, public law file; Office of the Historian, Party Divisions of the House of Representatives; U.S. Senate, Party Division). In the 8 elections that produced divided government, and in 2000, when the Senate opened at 50–50, the winner was closer in 6, by an average of 1.5 points more (the same sources). On this measure, the topics of a Congress’s laws look about as much like the losing platform as the winning one.

Floor votes are closer to platforms than laws are. For all 19 elections, the winning platform’s overlap with the next Congress’s House roll calls averaged 66.9, against 55.0 with its laws, and the roll-call overlap was the higher of the two in every case (Exhibit 3; Comparative Agendas Project, public law and roll-call vote files). The House votes on the topics the parties campaign about more often than it enacts laws on them.

The gap has a structure (Exhibit 4). Averaged over every platform and every Congress from the 81st to the 117th, platforms gave 16.3% of their policy text to international affairs, 8.7% to macroeconomics and 7.6% to civil rights; laws gave those topics 4.4%, 2.3% and 1.1% (Comparative Agendas Project, Democratic and Republican platform and public law files). Laws went the other way on public lands (20.9% of laws, 4.0% of platforms) and government operations (18.0% and 5.7%), the topics that hold most land conveyances, park boundaries, agency housekeeping and similar measures (the same files). Defense took about a tenth of both (the same files). Part of the gap reflects what laws are: many laws make small administrative changes that no platform would mention, and much foreign policy is made by the executive and by treaty rather than by statute.

The topic test is a test of attention, not of promises kept. A platform paragraph opposing a policy and a law enacting it can fall under the same topic. The pledge test in the next section asks the narrower question.

Pledges, votes and laws

The standard measure of promise-keeping counts pledges: statements committing a party to one specific action or outcome that can be clearly determined to have occurred or not. A 2017 comparative study in the American Journal of Political Science applied that definition, or a broader one in the US and UK studies, to 18,743 pledges from 54 election campaigns in 12 countries, and scored fulfillment from legislation, decrees, budget data and other sources (American Journal of Political Science, 2017, accepted manuscript). Governing parties fulfilled 59% of their pledges at least partially (5,044 of 8,547) (the same study). Single-party majority governments in the United Kingdom fulfilled 86% (494 of 575) and single-party minority governments in Sweden 87% (112 of 129) (the same study). For US governments from 1977 to 2001, the authors’ summary reports that “on average over 60 percent of the pledges made by the party of the president were fulfilled at least partially,” comparable to some coalition governments (Exhibit 5; authors’ summary, AJPS, June 7, 2017). They treat US presidents as single-party executives who “must negotiate with the other main party in the legislature on each piece of legislation,” and they note that parties “do not necessarily tailor their pledges to foreseeable constraints” (accepted manuscript).

This study applies the narrow definition to the three most recent platforms whose parties then held the presidency and both chambers: the 2016 Republican platform (115th Congress, 2017–18), the 2020 Democratic platform (117th, 2021–22) and the 2024 Republican platform (119th, through October 26, 2026) (American Presidency Project, 2016 Republican, 2020 Democratic and 2024 Republican platforms). It scores only enacted law, so its rates are not comparable with that study’s, which also count executive action. To avoid choosing pledges by hand, the universe is every quasi-sentence the Comparative Agendas Project codes to taxation (subtopic 107), voting rights (206) or campaigns, elections and ethics (2012): 125 quasi-sentences in all (Comparative Agendas Project, Democratic and Republican platform files). The two areas were chosen because they usually differ in route. A change in tax rates can travel in a reconciliation bill, which needs a simple majority in the Senate. A change in election law generally cannot: the Senate’s Byrd rule treats a provision as extraneous when it produces no change in outlays or revenues or when its budget effect is merely incidental, and waiving the rule takes 60 votes (Congressional Research Service, The Budget Reconciliation Process: The Senate’s “Byrd Rule,” RL30862, September 28, 2022).

Two independent coding passes, each made without sight of the other under a written protocol, marked whether every quasi-sentence was a pledge. They agreed on 120 of 125 (96%; Cohen’s kappa 0.91) (the two passes over the same platform files). The five disagreements were resolved by rule: three sentences urging states or election officials to act were not counted as party pledges, a heading that repeats a later sentence was not counted, and a sentence promising “oversight” of election vendors was judged too vague to test. Area follows the project’s topic code; the two passes placed two pledges differently (repeal of the Affordable Care Act and repeal of the Johnson Amendment). Route agreed for 37 of the 38 pledges both passes found (the two passes). That left 38 sentences coded as pledges. Four restate an earlier pledge, one is an executive-branch pledge and one is a sentence on vehicle mandates that the project coded to elections; setting those six aside leaves 32 distinct pledges, 11 on taxes and 21 on elections and ethics (the three platforms and the project’s platform files, as above; this part’s pledge calls). Two coders then placed each pledge, independently and from the record, in one of four categories for the Congress that followed: enacted as pledged, enacted in part, voted and failed, or never brought to a vote. They placed 27 in the same category; the other five, where they differed or found that the record does not settle the pledge, are not scored and are listed under Limitations (this part’s pledge coding rubric).

Of the 11 tax pledges, two were enacted as pledged and three in part; two were never brought to a vote, and four are not scored (Exhibit 6; this part’s pledge calls). The 2017 tax law cut the corporate rate to 21% and adopted a deduction for foreign dividends (Public Law 115-97, GovInfo). The same law set the individual mandate penalty to zero (Public Law 115-97). The 2021 rescue plan enlarged refundable credits for 2021 only (Public Law 117-2), and the 2022 law imposed a 15% minimum tax on the book income of large corporations while leaving the 21% rate unchanged (Public Law 117-169). The 2025 reconciliation law made the 2017 individual provisions permanent and created a deduction for tips of up to $25,000 that phases down above $150,000 of income and ends after 2028 (Public Law 119-21, GovInfo).

Of the 21 election and ethics pledges, none was enacted as pledged and two were enacted in part; 14 were voted and failed, four were never brought to a vote, and one is not scored (Exhibit 7; this part’s pledge calls). Congress appropriated $75 million in each of fiscal 2022 and 2023 for election security grants to states (Public Law 117-103, Consolidated Appropriations Act, 2022; Public Law 117-328, Consolidated Appropriations Act, 2023). The Courthouse Ethics and Transparency Act of May 13, 2022 required federal judges’ financial disclosure reports to be published in a searchable online database (Public Law 117-125, sec. 2(b), GovInfo). The House passed the For the People Act in March 2021, 220–210, and the 2022 combined voting bill; the Senate voted 49–51 on cloture for the latter (Clerk of the House, Roll Call 62, March 3, 2021; Senate.gov, record vote 9, 117th Congress, January 19, 2022). The House passed the SAVE America Act, requiring documentary proof of citizenship to register and photo identification to vote, 218–213 in February 2026; a Senate cloture vote on a photo-identification bill failed 52–46 in August 2026 (Clerk of the House, Roll Call 69, February 11, 2026; Senate.gov, record vote 231, 119th Congress, August 8, 2026).

The sample is small and narrow. It covers three platforms and two topics, and a single law can carry several pledges, so the outcomes are not independent. Two of the 11 tax pledges and one of the 21 election pledges come from the 2024 platform, which made few pledges in these areas, so its percentages rest on one or two pledges (American Presidency Project, 2024 Republican Party Platform, July 8, 2024; this part’s list of the in-scope pledges).

Why the route matters

The Senate’s cloture rule requires three-fifths of senators duly chosen and sworn, normally 60, to end debate on most measures (Congressional Research Service, Filibusters and Cloture in the Senate, RL30360, April 7, 2017). Reconciliation bills are an exception: the Congressional Budget Act limits debate on them, so they pass with a simple majority, and the Byrd rule limits their contents to provisions with budget effects that are not merely incidental (Congressional Research Service, RL30862). Exhibit 8 shows the final Senate votes on four reconciliation laws, the 2017 tax law, the 2021 rescue plan, the 2022 Inflation Reduction Act and the 2025 reconciliation law, and the cloture votes on six election bills. The four reconciliation laws passed the Senate with 51, 50, 50 and 50 senators voting yea, the last two with the Vice President’s vote (Senate.gov, record vote 323, 115th Congress, December 20, 2017; 110, 117th Congress, March 6, 2021; 325, August 7, 2022; 372, 119th Congress, July 1, 2025). The six election bills drew 49 to 52 yeas on cloture and failed (Senate.gov, record votes 246, 420 and 459, 117th Congress, 2021; 9 and 346, 2022; 231, 119th Congress, August 8, 2026). On that count the election bills had as much support as the tax laws; they lacked the 60 votes that their route required.

The 51-vote route is not automatic. The final 2017 Senate attempt to repeal parts of the Affordable Care Act through reconciliation, an amendment to the House-passed reconciliation bill, failed 49–51 (Senate.gov, record vote 179, 115th Congress, July 28, 2017). In 2026 Republican senators tried three times to move the voter-identification pledge through reconciliation, each on a motion to waive budget points of order: an amendment adding reconciliation instructions and reserve funds to the budget resolution failed 48–50, and two amendments to the reconciliation bill S. 2 failed 48–50 and 50–49; only the last had a majority, and it lacked the 60 votes the waiver needed (Senate.gov, record votes 96, April 23, 2026; 144, June 4, 2026; 151, June 5, 2026). S. 2 passed 52–47 without them and became Public Law 119-98 on June 10, 2026 (Senate.gov, record vote 163, June 5, 2026; Public Law 119-98, Secure America Act, GovInfo).

The procedural reading is one account of Exhibit 6: the five tax pledges enacted as pledged or in part all came through reconciliation laws (Public Laws 115-97, 117-2 and 119-21), while of the 14 election and ethics pledges voted and failed, 13 fell on Senate cloture votes short of 60 and one passed the House in a measure the Senate never agreed to (this part’s pledge calls). There are three others. Election rules divide the parties sharply: on the six Senate cloture votes in Exhibit 8, the bills drew one yea in all from senators of the other party, and all three 2026 reconciliation attempts drew nays from the majority party (four, four and three), so the same pledges might fail under any procedure (Senate.gov, the record votes above). Tax cuts and credits are easier to pass than tax increases whatever the route, and the two Democratic tax pledges never brought to a vote, on carried interest and the estate tax, both sought increases, as did the two not scored, on the 2017 cuts and the corporate rate (this part’s pledge calls). And parties may write pledges they expect to pass into the parts of the platform that reconciliation can carry. In this sample route and topic coincide almost completely, so the data cannot separate these readings. The election security grants came through an omnibus appropriations law that passed the Senate 68–31 with votes from both parties (Senate.gov, record vote 78, 117th Congress, March 10, 2022).

Whether a low enactment rate is a failure depends on what a platform is for. One view holds that platforms are signals: they are written by convention delegates rather than by the legislators who must pass bills, nominees often distance themselves from them, and a separated system makes many pledges unreachable even for a party holding both branches. On that view, scoring a platform like a contract measures the Senate’s rules more than the party’s good faith. The other view holds that the platform is the only party-wide statement voters can consult, that parties fulfill most of their testable pledges in comparative studies, and that a party choosing to move some priorities by reconciliation and not others is making a choice voters can judge (American Journal of Political Science, 2017). Both views accept that a fair score counts only testable pledges, reports partial fulfillment and shows the route each pledge needs.

Options

Score platforms on a common basis. A published, repeatable count of testable pledges by route, using the definition in the previous section, would let voters compare a party’s record with its platform. The case for it is transparency and comparability across parties and years; the case against is that any coding involves judgment about what counts as a pledge, and two independent passes still disagreed on 5 of 125 sentences here (the two passes over the Comparative Agendas Project platform files). Publishing both passes’ calls and the sensitivity of each count addresses part of that.

Write fewer, testable pledges. The 2024 Republican platform opened with a list of twenty promises and ran 320 quasi-sentences; the 2024 Democratic platform ran 2,028 (American Presidency Project, 2024 Republican Party Platform; Comparative Agendas Project, Democratic and Republican platform files). A short list is easier to score and to hold a party to. The case against is that a short list narrows what a coalition can promise its members, and that the length of a platform says little about how many of its pledges are testable.

Widen or narrow the 51-vote route. Proposals to let reconciliation carry more kinds of policy, or to change the cloture threshold for legislation, would change which pledges a governing party can enact alone. CRS describes the Byrd rule’s six tests and the 60-vote waiver, and the Senate’s cloture rule and its exceptions (RL30862; RL30360). Those who favor wider majority routes argue that a party that wins the presidency and both chambers should be able to enact its program and answer for it at the next election. Those who oppose argue that the 60-vote threshold forces measures to win support across parties, which makes them more durable, and that rules a majority writes for itself today will serve the other party tomorrow. In January 2022 the Senate sustained the chair’s ruling 52–48 against a change in the cloture rule for one voting bill, keeping the 60-vote threshold for it (Senate.gov, record vote 10, 117th Congress, January 19, 2022).

What is coming. The House adopted an FY2027 budget resolution on July 22, 2026, 216–214, with reconciliation instructions to four House committees: Armed Services ($60 billion), Intelligence ($13 billion), Agriculture ($12 billion) and House Administration ($10 billion), each the most the committee’s changes may add to deficits over 2027 to 2036; the Senate’s 2026 roll-call list, whose latest vote is dated September 30, shows no vote on it (Clerk of the House, Roll Call 281, July 22, 2026; H.Con.Res. 113 as engrossed in the House, GovInfo; Senate.gov, Roll Call Votes, 119th Congress, 2nd Session). If adopted by both chambers, it would open the way to a third reconciliation bill in the 119th Congress, and any election provision in it would face the Byrd rule tests described above. The next platforms will be written in 2028.