Ranking the Speakers: All 56, and the Method Behind It

The presidency has been ranked for eighty years. Schlesinger started it in 1948, Siena has run it seven times, C-SPAN four times, and there are two dozen surveys in between. You can find a president’s standing on ten named characteristics, with the per-category scores, on a live web page.

The Speakership has been ranked once. In 1998. By a master’s student.

That is the whole of it. Dennis Roy Still, at Iowa State, mailed a questionnaire to sixty congressional scholars and asked them to rate fifty Speakers on four questions. His abstract calls it, accurately, “the first attempt to rank the U.S. House Speakers” (Still, Ranking the American Speakers, Iowa State University Digital Repository). Nobody has repeated it in the twenty-eight years since. Still’s scholars could rate only thirty-eight of his fifty, and the six Speakers who have held the gavel since 1998 were not in his survey at all. Eighteen of the fifty-six people who have been Speaker have never been rated by anyone.

So this is the ranking that does not exist. I have graded all fifty-six individuals who have served as Speaker, from Frederick Muhlenberg in 1789 to Mike Johnson, using a rubric stated in full, and I want to show the work so that someone else can attack it.

One note before the method, because it is the most useful thing in this piece. When I tested the one dataset that does exist against the raw facts of tenure, a hard number emerged: the correlation between how long a Speaker served and how well the experts remembered him is 0.64, and the correlation between how well he was remembered and how many scholars were willing to rate him at all is 0.68. Obscurity is not neutral in this record. It reads as failure. Any ranking of Speakers has to say what it does about that, and most would rather not.

1. Provenance

Still modeled his instrument on the presidential surveys, and said so: Schlesinger, Maranell, Neal, and the rest. His four questions were the use of formal and informal powers, interest in domestic policy agendas, interest in foreign policy agendas, and overall performance while in office. He scaled each one to five.

This is not a primary-source assessment of the fifty-six men. It is three documented components, assembled by me, with the 1998 expert survey retained separately as a validation column. Every statute and rule underneath belongs to its cited source. The grades, the cut points, and the argument are mine.

2. The problem with the only dataset

Still’s thesis is a genuine contribution and it is also, in one place, broken. His Table 5 (foreign policy interest) lists the Speaker CARLISLE twice, once at 3.07 with fourteen responses and once at 2.14 with seven, and it omits Champ Clark entirely. His own text calls the table a survey of 37 Speakers. Counting the entries below 3.00 gives exactly the 28 his text claims, which is how I know the duplicate is a typographical defect in the printed table rather than an error of my reading.

A second, smaller discrepancy: his prose says “17 of the 38 Speakers received ratings greater than 3.00.” Strictly counted, sixteen are above 3.00 and Macon sits at exactly 3.00. The correct statement is 3.00 or above.

Neither defect touches Table 6, the overall-performance ranking, which is the column I use. I checked it three ways: it has thirty-eight rows, no duplicates, and its own internal consistency holds. But it is worth saying that the single scholarly ranking of the Speakership has a table with a Speaker listed twice, because it tells you how thin the attention has been.

3. The three components

I grade each Speaker on three things, each scored 0 to 4, for a maximum of 12. The profile string is those three digits in order.

The cut points are a choice, stated here so they can be argued with.

T (cumulative years)      4: >= 9     3: >= 5     2: >= 2.5   1: >= 0.5   0: < 0.5
L (landmark statutes)     4: >= 6     3: >= 4     2: >= 2     1: >= 1     0: 0
I (institutional change)  4: transformed the office   3: major   2: real   1: minor   0: none

Tenure is arithmetic and fully reproducible. Legislative output is a count of verified statutes, banded; the statutes are named in the sources and each was verified as passing the House under that Speaker and becoming law. Institutional change is the one component that is a judgment, and it is the component most of the argument rests on, so I state what earns a 4: Reed’s 1890 rules ending the disappearing quorum, Cannon’s consolidation and then the 1910 revolt that broke it, Clay’s conversion of the chair from referee to national policy command post. Those four points are the office being rebuilt, not the office being used.

Why not simply use Still’s four categories? Because they cover 38 of 56 men and stop in 1998, and because his own data shows that memory, not merit, is the strongest predictor of the score. A ranking that inherits his instrument inherits that. The three components are facts about what a Speaker did. Still’s survey is retained as the fifth column, and it is the check on whether the facts track what experts saw.

4. What the three components find on their own

The three components do not tell the same story, and the differences are the substance.

The three lists barely overlap in order. A Speaker who served nearly ten years (O’Neill) is not the same as a Speaker who passed seven landmark statutes (Foley), and neither is the same as a Speaker who rebuilt the office (Cannon). The composite is not an average that flattens these into mush; it is a sum of three separate claims about what the job is for.

Against the 38 Speakers the 1998 scholars rated, the components disagree with the experts in an informative pattern:

rho(tenure,       expert score) = 0.61
rho(legislation,  expert score) = 0.27
rho(institutions, expert score) = 0.62
rho(composite,    expert score) = 0.76

Legislation correlates worst with expert standing, at 0.27. That is the finding worth pausing on. A Speaker can run a floor that produces major law and still not be remembered as great, because the law is credited to the president who signed it. The Speakers who passed the most (Foley, Hastert, McCormack, Pelosi) are not the Speakers the scholars ranked highest (Clay, Rayburn, Reed, Cannon). The composite sits above every single component against the experts, which is the basis for treating it as more than any one of them.

5. The ranking

Fifty-six individuals, sorted by the sum of the three grades. The 1998 expert mean is shown where one exists.

Rayburn is first because he is the only Speaker who scores near the top of all three columns at once: seventeen years, six landmark statutes, and the 1961 enlargement of the Rules Committee that broke the conservative bottleneck. Clay and Cannon tie on the sum, and share the identical profile 334. The coincidence is exact: Clay served 6.6 years and passed five landmark statutes, Cannon served 5.8 and passed four, and each altered the office itself, Clay by making the chair a policy instrument and Cannon by consolidating powers the House then stripped from him in 1910. Reed is seventh, and almost entirely on the institutional column: his legislative count is four tariff and antitrust statutes across his two Speakerships, but his four on institutions is the disappearing quorum itself.

6. What the string encodes, and what it hides

The profile string says which component earned the grade. Two Speakers with the same sum have different records, and the string is the only thing that shows it.

340 and 223 both appear high in the table, and they are opposites. Tom Foley’s 340 is three points of tenure, four of legislation, none of institutional change; he ran a floor that produced the ADA, the Clean Air Act Amendments, NAFTA, the Family and Medical Leave Act, and the Brady Act, and he rebuilt nothing. Newt Gingrich’s 223 is two, two, three; fewer laws and a shorter tenure, but he is credited with the 1995 rules package that cut committee staff by a third, limited chair terms, and abolished three committees. Foley did more of the job the office exists to do. Gingrich changed the office itself.

The string also exposes the cases where the sum is misleading. Polk’s 202 puts him twenty-ninth, well below the experts’ 3.78. The reason is visible in the digits: a zero on legislation, because the central fact of his Speakership was the gag rule, a resolution that tabled antislavery petitions and never became law, which is a use of power rather than an output. Still’s scholars rated his overall performance highly. My rubric, which counts output, does not. That disagreement is not a bug; it is the rubric saying what it measures.

And the string exposes the Speakers whose rank is entirely tenure. Seven of the fifty-six have a legislative grade of 0 and an institutional grade of 0, and every one is a man whose Speakership lasted a single term or less: Barbour, Cheves, Kerr, Orr, Pennington, Taylor, and Pomeroy. Five of the seven served before 1861, Pomeroy served a single day in 1869, and Kerr died in office in 1876. Their placement on the list is a statement about how long they held the gavel, and about nothing else. The table should be read that way.

7. Sensitivity: what the weights decide

The sum weights the three components equally, each contributing a third. As with any equal weighting, that is a position nobody argued for, so I varied it. Four thousand random weightings drawn from a Dirichlet distribution, and the count of how often each Speaker lands in the top four:

Speaker      in top 4      in top 10
Rayburn       100.0%        100.0%
Cannon         76.4%        100.0%
Clay           61.4%        100.0%
Pelosi         53.7%        100.0%
O'Neill        49.7%        100.0%
McCormack      25.2%         65.9%
Reed           21.9%         71.7%
Foley          11.7%         45.4%

Two facts hold at once, and together they are the finding. The top ten is weight-invariant: Cannon, Clay, Rayburn, Pelosi, O’Neill, and Albert appear in the top ten under every one of the four thousand weightings. The top four is not. Seven distinct top-four sets appear across the weightings, and only Rayburn reaches the top four always; eight individuals do so at least once. Reed, who is seventh in the equal-weight table, is a top-four Speaker under any weighting that emphasizes institutions, because he is the man who made the House a majority-rule body; he drops out of the top four under any weighting that emphasizes legislation, because his laws are four tariff and antitrust statutes across his two Speakerships.

The weights decide the ordering inside the top tier. They cannot dislodge the tier itself. That is the honest claim: a top ten, fixed, and a top four that is a statement about what you think the Speakership is for.

8. The two cases the method cannot settle

Nancy Pelosi. She is fifth on the equal-weight sum, and she is fourth in the 2026 analytical ranking by another hand (US Political Rank, The 15 Most Consequential Speakers, July 2026). She has the highest legislative count in the table, eleven verified landmark statutes across her two Speakerships, including the Affordable Care Act, Dodd-Frank, the American Rescue Plan, and the Inflation Reduction Act. She has three points of institutional change, from the 2007 and 2019 rules packages. And she is the only Speaker in the top ten whom the 1998 scholars never rated, because they finished their survey nine years before she took the gavel. Whether she belongs above Reed and Cannon or below them is exactly the question the weighting argument in §7 leaves open; she is a top-four Speaker in 53.7% of weightings.

Mike Johnson. He is twenty-seventh, and the number should be read as provisional. His Speakership is unfinished, so his tenure component will rise, and his legislative output is the two statutes verified so far (the Laken Riley Act and the One Big Beautiful Bill Act). No ranking can grade a sitting Speaker against completed careers, and pretending otherwise would be the same error the 1998 survey made in reverse when it rated Gingrich mid-congress. I have included him because excluding him would make the table fifty-five, not fifty-six, and because his provisional position is itself informative: a Speaker with a two-digit legislative count would rank far higher within a year, and one with none would not.

9. The recognition problem, stated plainly

I opened with a correlation and I will close with it, because it is the thing this project will not let go.

r(Still score,   number of scholars who rated him) = 0.684
r(years served,  Still score)                      = 0.636
r(years served,  number of scholars who rated him) = 0.675

The Speakers the scholars declined to rate are almost all antebellum. Twelve of Still’s fifty have no score at all, and they are Dayton, Sedgwick, Varnum, Barbour, Stevenson, Bell, Jones, Davis, Orr, Grow, Pomeroy, and Kerr. A Speaker who served one Congress in 1843 is not a failure; he is a stranger, and the survey could not separate the two. The mechanism is documented in Still’s own numbers: Speakers who served before 1860 averaged 6.1 responses apiece, and Speakers who served after 1860 averaged 16.0. Ten of the rated Speakers sat at the bare minimum of four, and they are the earliest names in the table.

I cannot fix that with a survey that does not exist. What I can do is state which component does not depend on memory. Tenure is a fact from the House Historian’s own service dates, and it is the same fact whether or not anyone remembers the man. Legislation is a fact, from the statute. Institutional change is where memory enters, and it is the component I have argued for most openly, so that it can be attacked. The composite sits above every single component against the experts, at 0.76, which is the best evidence I have that the facts and the memory are measuring something real together rather than one of them alone.

10. Summary of the method

1. Take the roster of 56 individuals from the Office of the Historian, with official service dates.
2. Grade tenure on the official cumulative years at stated cut points.
3. Grade legislative output as a count of verified landmark statutes per Speakership, banded.
4. Grade institutional change from the sourced record of rules and precedents, and state what earns a 4.
5. Sum the three grades; emit the three-digit profile string.
6. Report the 1998 expert mean alongside, for the 38 Speakers who have one.
7. Vary the weights; treat the top ten as invariant and the top four as judgment-dependent.
8. Report tenure separately from any memory-dependent score.

11. Stated limitations

• One component is a judgment. Institutional change is mine, argued in the open at stated thresholds. A reader who weighs it differently gets a different top four, which §7 quantifies.
• The rubric rewards the measurable. Tenure and statute counts favor Speakers with long service and large majorities. The Speakers who exercised power by preventing action (Polk with the gag rule, Johnson with the recess) score low by construction, and the table should not be read as saying they were unimportant.
• The 1998 survey is the only external check, and it covers 38 of 56. It is a mail survey of 60 scholars with no political-bias question, a gap Still named himself.
• Memory is confounded with merit at r = 0.68 on the scholars’ willingness to rate, and no correction exists for it.
• A sitting Speaker is unresolved. Johnson’s position is provisional by definition.
• The one scholarly table of the Speakership has a printing defect, documented in §2, which is a measure of how little the field has been tended.

Sources

All figures in this paper are transcribed from the sources below and checked against each other. The 1998 survey data was read from the deposited full text of the thesis, including the printed tables (Table 3 use of powers, Table 4 domestic interest, Table 5 foreign interest, Table 6 overall performance) and the questionnaire reproduced in Appendix B.

• Still, D. R. (1998). Ranking the American Speakers: A survey of congressional scholars’ perceptions (Master’s thesis, Iowa State University). Iowa State University Digital Repository. https://dr.lib.iastate.edu/handle/20.500.12876/KrZJ855r The only scholar survey of the Speakership. Source of the 1998 expert column, the four questions, the N>4 threshold, and the Table 5 printing defect noted in §2.
• Office of the Historian, U.S. House of Representatives. Speakers of the House by Congress. https://history.house.gov/People/Office/Speakers-List/ The roster of 56 individuals and their service dates, with footnotes on each resignation, death, and intra-term election.
• Office of the Historian, U.S. House of Representatives. Speaker of the House Fast Facts. https://history.house.gov/Institution/Firsts-Milestones/Speaker-Fast-Facts/ The count of 56, the seven nonconsecutive Speakers, Rayburn’s 17 years 2 months 2 days, and the five Speakers who died in office.
• Office of the Historian, U.S. House of Representatives. “A Chair Made Illustrious”: A Concise History of the U.S. House Speakership (2024). https://history.house.gov/Education/Publications/concise-history-speakership-2024/ The institutional-change record: Clay’s conversion of the chair, Reed’s 1890 rules, the 1910 revolt against Cannon, Rayburn’s 1961 Rules enlargement, Gingrich’s 1995 package, and the Pelosi-era rules changes.
• Congressional Research Service. The Speaker of the House: House Officer, Party Leader, and Representative (Report 97-780). https://www.everycrsreport.com/reports/97-780.html The structural history of the office, including the 1910 prohibition on the Speaker sitting on the Rules Committee.
• Congressional Research Service. Speakers of the House: Elections, 1913-2025 (Report RL30857). https://www.everycrsreport.com/reports/RL30857.html Ballot counts and vote tallies for the modern Speaker elections, including the 1923 Gillett election and the 2023 McCarthy and Johnson elections.
• Office of the Historian, U.S. House of Representatives. Speaker Elections Decided by Multiple Ballots. https://history.house.gov/People/Office/Speakers-Multiple-Ballots/ The exhaustive list of the sixteen multi-ballot Speaker elections, from the 3rd Congress through the 118th.
• US Political Rank. The 15 Most Consequential Speakers of the House, Ranked by the Record (July 2026). https://uspoliticalrank.com/rankings/speakers-of-the-house-ranked The only other systematic ranking of Speakers found, cited in §8 as a comparison point. Its own method note states that “No official score for Speakers exists.”

PRH | huffmanwrites.org | © Philip Huffman