What Federal Reserve Chairs, US Presidents, and S&P 500 CEOs emphasize most—and what they rarely say—measured by exact word counts and normalized per 1,000 words across MouthMath’s public-transcript corpus.
Powell’s “inflation” peaked in his big framework speeches, then faded as his term wound down
“inflation” frequency across landmark Powell speeches (260 total mentions in 20 docs)
Aug 23, 2024 JH
23.4/1k (54 / 2,312w)
Aug 22, 2025 JH
16.9/1k (54 / 3,199w)
Oct 14, 2025 Bal.
2.7/1k (7 / 2,554w)
Jan–Mar 2026 End
0 & 4 mentions
Across Jerome Powell’s 20 indexed documents (22,557 total words from August 2024 to March 2026), “inflation” is his #1 content word at 260 mentions (11.53 per 1,000 words). Nearly half of those uses fall in two annual Jackson Hole policy addresses: his August 23, 2024 “Review and Outlook” speech (54 uses in 2,312 words, or ~23 per 1,000) and his August 22, 2025 framework review speech (54 uses in 3,199 words, or ~17 per 1,000).
By his October 14, 2025 speech “Understanding the Fed’s Balance Sheet,”“inflation” drops to 7 mentions in 2,554 words (~2.7 per 1,000), and his final two indexed documents—a brief January 11, 2026 statement and March 21, 2026 award remarks—contain 0 and 4 mentions.
“Tariff” entered Powell’s vocabulary in spring 2025
All 2024 Speeches
0
“tariff” or “tariffs” mentions
Mar 7, 2025
1
First mention in sample
Apr – Jun 2025
5 · 4 · 5
Apr 4, Apr 16 & Jun 24 testimony
Aug – Sep 2025
6 · 4
Aug 22 Jackson Hole & Sep 23
In all of Jerome Powell’s 2024 speeches on MouthMath, “tariff” and “tariffs” appear 0 times. The word enters his remarks on March 7, 2025 (1 mention), then becomes a regular fixture across spring and summer 2025: 5 mentions on April 4, 2025, 4 on April 16, 2025, 5 in his June 24, 2025 semiannual testimony to Congress, 6 on August 22, 2025, and 4 on September 23, 2025.
From Volcker to Powell, each Fed Chair in our sample said “inflation” more often than the one before—then Warsh broke the streak
“inflation” mentions per 1,000 total words across six Fed Chairs
Warsh 2026
5.11/1k (119 uses)
Powell 2024–26
11.53/1k (260 uses)
Yellen 2016–17
5.84/1k (282 uses)
Bernanke 2013–14
2.71/1k (131 uses)
Greenspan 2005
1.37/1k (60 uses)
Volcker 1986–87
0.96/1k (70 uses)
Measured per 1,000 total words, “inflation” rises in every consecutive Chair’s corpus: Paul Volcker at 0.96 (70 uses in 1986–87), Alan Greenspan at 1.37 (60 uses in 2005), Ben Bernanke at 2.71 (131 uses in 2013–14), Janet Yellen at 5.84 (282 uses in 2016–17), and Jerome Powell at 11.53 (260 uses in 2024–26). Kevin Warsh breaks that run at 5.11 (119 uses in 23,297 words, June–October 2026)—see Finding #16.
While each Chair’s 20-speech sample captures the economic backdrop near the end of their tenure, the steady climb from Volcker through Powell also aligns with the institutional shift toward explicit inflation targeting and modern “forward guidance,” where public communication is itself part of monetary policy. These are sample-level word counts, not a full record of each Chair’s remarks, and Warsh’s sample so far is only six documents, mostly press conferences.
#04Federal Reserve Chairs · 1980s vs. 2020s Policy Vocabulary
Volcker talked about banks and deficits, not inflation
Volcker “banks”
3.00 / 1k
Over 3× his “inflation” rate
Volcker “deficit”
1.61 / 1k
1986–87 fiscal & trade focus
Volcker “inflation”
0.96 / 1k
Post-disinflation late tenure
Powell “deficit”
0.00 / 1k
Modern Fed avoids fiscal commentary
Paul Volcker is remembered for breaking the high inflation of the late 1970s and early 1980s, but by his 1986–87 speeches on MouthMath, inflation had been subdued and banking stability and fiscal deficits led his agenda. Volcker said “banks”3.00 times per 1,000 words and “deficit”1.61 times per 1,000 words, compared with 0.96 for “inflation.”
Jerome Powell’s 2024–26 corpus shows the opposite pattern: “deficit” appears at 0.00 per 1,000 words, reflecting how strictly modern Fed Chairs avoid commenting on fiscal budget policy.
AI has taken over Fed governors’ speeches—and Warsh is the first Chair in our sample to say it
“ai” mentions per 1,000 words across Federal Reserve officials
Michael Barr
6.84/1k (300 uses)
Lisa Cook
6.42/1k (280 uses)
Christopher Waller
3.92/1k (167 uses)
Philip Jefferson
2.95/1k (128 uses)
Michelle Bowman
2.54/1k (87 uses)
Kevin Warsh
1.03/1k (24 uses)
Jerome Powell
0.00/1k (0 uses)
Artificial intelligence has become a central research and policy theme for Federal Reserve governors: Michael Barr leads at 6.84 mentions of “ai” per 1,000 words (300 total uses, including single-speech peaks of 99 on February 17, 2026 and 85 on July 14, 2026), followed by Lisa Cook at 6.42 per 1,000 (280 uses, with peaks of 66 on July 17, 2025 and 71 on May 27, 2026), Christopher Waller (3.92, 167 uses), Philip Jefferson (2.95, 128 uses), and Michelle Bowman (2.54, 87 uses).
Across Jerome Powell’s 20 indexed documents, “ai” appears 0.00 times per 1,000 words, as it does for Yellen, Bernanke, Greenspan, and Volcker. Kevin Warsh changes that: 24 uses across his first six documents (1.03 per 1,000)—still well below the governors’ rates (see Finding #17).
Michelle Bowman uses “banks” at 8.02 per 1,000 words (275 uses across 34,306 words) and “risk” at 5.95 per 1,000 words—well ahead of Waller, Barr, and Jefferson on “banks” (1.5 to 3.1 per 1,000) and more than ten times Powell’s rate on “risk” (0.58 per 1,000). Meanwhile, her “inflation” rate (2.04 per 1,000) is less than a fifth of Powell’s (11.53), matching her bank-supervision role.
#07Federal Reserve Payments & Digital Assets · Barr & Waller
Stablecoins are almost entirely a Barr and Waller topic
Combined mentions of “stablecoin” and “stablecoins” reach 1.78 per 1,000 words for Michael Barr (0.64 singular + 1.14 plural) and 0.59 per 1,000 words for Christopher Waller (0.14 + 0.45). Across the rest of the Board, Jerome Powell, Lisa Cook, and Philip Jefferson are at 0.00, while Michelle Bowman is at 0.06 per 1,000 words.
Presidents of both parties each have a signature word
Highest-rate signature words among all 8 indexed US Presidents (per 1,000 words)
Clinton (D) “people”
5.78/1k (458 uses)
Bush 43 (R) “america”
5.19/1k (#1 of 8)
Bush 41 (R) “world”
4.66/1k (#1 of 8)
Carter (D) “energy”
3.82/1k (200 uses)
Reagan (R) “tax”
3.52/1k (255 uses)
Trump (R) “great”
3.08/1k (272 uses)
Obama (D) “jobs”
2.57/1k (232 uses)
Biden (D) “democracy”
1.35/1k (91 uses)
Comparing roughly 20 major speeches per president (~52,000 to 90,000 words each), every president from Jimmy Carter to Joe Biden leads all eight presidents in at least one defining word per 1,000 words:
Jimmy Carter (D):“energy” at 3.82 per 1,000 (200 uses, highest of all 8) and “soviet” at 2.67 (140 uses).
Ronald Reagan (R):“tax” at 3.52 per 1,000 (255 uses, highest of all 8) and “freedom” at 2.11 (153 uses).
George H. W. Bush (R):“world” at 4.66 per 1,000 (highest of all 8) and “war” at 1.77.
Bill Clinton (D):“people” at 5.78 per 1,000 (458 uses, highest of all 8).
George W. Bush (R):“america” at 5.19 per 1,000 (highest of all 8), “freedom” at 2.35 (147 uses), and “terrorists” at 1.93 (121 uses).
Barack Obama (D):“jobs” at 2.57 per 1,000 (232 uses, highest of all 8).
Donald Trump (R):“great” at 3.08 per 1,000 (272 uses) and “tremendous” at 0.37 (33 uses)—both highest of all 8.
Joe Biden (D):“democracy” at 1.35 per 1,000 (91 uses) and “folks” at 0.99 (67 uses)—both highest of all 8.
#09US Presidents · Pronoun Frequency (Stopwords Included)
Presidential “I” versus “we” cuts across party lines
Biden “I”
11.63 / 1k
Highest “I” rate in sample
Clinton “I” & “we”
11.26 · 20.74
#2 on “I”, #1 on “we” per 1k
Trump “I”
10.39 / 1k
Close behind Biden & Clinton
Bush 43 “I”
6.64 / 1k
Lowest “I” rate of all 8
When common stopwords are included, every president in the sample uses “we” more often than “I,” and the variation cuts across party lines. Per 1,000 words, Joe Biden (11.63) and Bill Clinton (11.26) use “I” most frequently, with Donald Trump close behind at 10.39, while George W. Bush is lowest at 6.64. First-person plural “we” ranges from 15.57 per 1,000 (George H. W. Bush) to 20.74 per 1,000 (Bill Clinton).
The Cold War disappears from presidential language as “China” rises
Carter “soviet”
2.67 / 1k
Bush 41: 1.18 · Reagan: 1.05
Post-1991 “soviet”
0.01 – 0.05
Clinton (0.05) through Biden
Reagan “china”
0.03 / 1k
Cold War-era baseline
Trump & Biden “china”
0.57 · 0.47
Nearly 20× Reagan’s rate
Mentions of “soviet” per 1,000 words track the end of the Cold War almost to the year: Jimmy Carter (2.67), George H. W. Bush (1.18), and Ronald Reagan (1.05) drop to 0.05 under Bill Clinton and 0.01–0.02 for every president thereafter. Meanwhile, “china” climbs from 0.03 per 1,000 under Reagan to 0.57 under Donald Trump and 0.47 under Joe Biden.
AI dominates tech CEOs’ remarks, with old-economy CEOs as the exceptions
“ai” mentions per 1,000 words in CEO prepared remarks & letters
Chuck Robbins Cisco
14.19/1k
Cristiano Amon QCOM
13.88/1k
Sundar Pichai GOOGL
13.40/1k
Lisa Su AMD
12.03/1k
Mark Zuckerberg META
7.85/1k
Satya Nadella MSFT
7.78/1k
Abel / Ortberg BRK / BA
0.00/1k
Among networking, chip, and cloud CEOs, more than one in every hundred words is “ai”: Chuck Robbins (Cisco) leads at 14.19 per 1,000 words, followed by Cristiano Amon (Qualcomm, 13.88), Sundar Pichai (Alphabet, 13.40), Lisa Su (AMD, 12.03), Mark Zuckerberg (Meta, 7.85), and Satya Nadella (Microsoft, 7.78).
At the other end of the S&P 500, Greg Abel (Berkshire Hathaway’s February 28, 2026 letter) and Kelly Ortberg (Boeing) use “ai”0 times, while Brian Niccol (Starbucks, 0.50) and Joaquin Duato (Johnson & Johnson, 0.74) mention it less than once per 1,000 words.
“Customers” per 1,000 words is led by John Furner (Walmart, 7.80), Marc Benioff (Salesforce, 7.23), Chuck Robbins (Cisco, 7.10), John Stankey (AT&T, 6.27), Brian Niccol (Starbucks, 6.01), and Andy Jassy (Amazon, 5.06). In Jassy’s shareholder letter, “customers” (13 mentions) and “customer” (8 mentions) still outrank “ai” (7 mentions).
#13Berkshire Hathaway · Greg Abel (Feb 28, 2026 Letter)
Abel’s first Berkshire letter centers on the company, not himself
“berkshire”
17 uses
#1 content word (1,730w letter)
“businesses” & “operating”
10 · 9
Core operating focus
“insurance” & “operations”
7 · 7
Followed by “capital” (6)
“warren” vs. “ai”
4 vs. 0
4 nods to Buffett; 0 mentions of AI
In Greg Abel’s February 28, 2026 shareholder letter (1,730 words), the highest-frequency content words are “berkshire” (17), “businesses” (10), “operating” (9), “insurance” (7), “operations” (7), “capital” (6), and “warren” (4). “AI” appears 0 times.
Across Jamie Dimon’s two indexed annual shareholder letters (April 7, 2025 and April 6, 2026, totaling 13,347 words), “should” appears 46 times and “need”25 times, right alongside “capital” (33), “global” (31), “bank” (30), “america” (27), and “world” (27)—reflecting how much of his annual letter is policy prescription alongside banking results.
#15SpaceX & Tesla · Elon Musk (2006 – 2016 Documents)
Musk’s early documents are about launch, not cars
“launch”
38 uses
#1 content word across 5 docs
“spacex” & “space”
28 · 26
Aug 2, 2006 – Jul 20, 2016
“nasa” & “commercial”
25 · 19
Congressional testimony focus
“falcon” & “competition”
15 · 15
“ai” appears 0 times
Elon Musk’s five indexed documents on MouthMath span August 2, 2006 to July 20, 2016 (Congressional testimony and master-plan essays). Because three of those five texts are SpaceX testimony before Congress, spaceflight vocabulary dominates his counts: “launch” (38), “spacex” (28), “space” (26), “nasa” (25), “commercial” (19), “falcon” (15), and “competition” (15). Across those 2006–2016 documents, “ai” appears 0 times.
#16Federal Reserve Chair · Kevin Warsh (Jun – Oct 2026)
Warsh breaks the Chair-by-Chair climb in “inflation” talk
Warsh “inflation”
5.11 / 1k
119 uses in 23,297 words
Powell “inflation”
11.53 / 1k
260 uses, 2024–26
Warsh prepared texts
6.98 / 1k
34 uses in 4,871 words
Warsh press conferences
4.61 / 1k
85 uses in 18,426 words
Across Kevin Warsh’s six indexed documents (June 17 to October 28, 2026: four FOMC press conferences, one Congressional testimony, and one speech), “inflation” appears 119 times in 23,297 words, or 5.11 per 1,000—tied with “policy” (also 119) as his most-used content word, and less than half of Powell’s 11.53. That ends the run in Finding #03, where each Chair from Volcker through Powell came in higher than the one before.
Caveat: Powell’s sample is speeches and testimony, while most of Warsh’s words come from press-conference answers. Counting only Warsh’s prepared testimony and speech, his rate is 6.98 per 1,000—still below Powell, but closer. This is a word-count pattern in a small, early sample, not evidence of a change in policy.
Warsh is the first Fed Chair in our sample to say “AI”
Warsh “ai”
24 uses
1.03 / 1k across 6 documents
Five earlier Chairs
0
Volcker through Powell samples
Warsh Aug 28 speech
8 uses
in 3,563 words
Michael Barr
6.84 / 1k
Governors still lead
“ai” appears 0 times in MouthMath’s samples for Powell, Yellen, Bernanke, Greenspan, and Volcker. Kevin Warsh says it 24 times (1.03 per 1,000 words), and it shows up in every one of his six documents: 12 across four FOMC press conferences, 4 in his July 14 testimony, and 8 in his August 28 speech.
The governors are still far ahead—Michael Barr at 6.84 per 1,000 and Lisa Cook at 6.42 (Finding #05). The earlier Chairs’ samples are from before AI was a common topic, so the zero partly reflects when they spoke.
Warsh’s signature word is “think”—mostly in live Q&A
Warsh “think”
3.86 / 1k
90 uses in 23,297 words
Five earlier Chairs
0.28 / 1k
65 uses in 235,199 words
Warsh press conferences
4.67 / 1k
86 uses in 18,426 words
Warsh prepared texts
0.82 / 1k
4 uses in 4,871 words
Kevin Warsh says “think”90 times (3.86 per 1,000 words), versus 0.28 per 1,000 for Volcker, Greenspan, Bernanke, Yellen, and Powell combined. Nearly all of it comes from his FOMC press conferences: 29 on June 17, 26 on July 29, 12 on September 16, and 19 on October 28, compared with just 2 each in his July 14 testimony and August 28 speech.
Caveat: this contrast likely reflects format as much as personal style. About 79% of Warsh’s counted words are press-conference answers, while the earlier Chairs’ samples are mainly prepared speeches and testimony, where “I think” is rare. On prepared texts alone, Warsh’s rate (0.82) is much closer to the others.
Tokenization: MouthMath lowercases all text, converts curly apostrophes to straight apostrophes, strips trailing possessive ’s (so Fed’s is counted with fed, while contractions like it’s stay intact), preserves u.s. as a single token, and excludes standalone numbers by default.
Normalization: Because speakers have different total word counts (from single 1,700-word shareholder letters to 90,000-word presidential speech collections), cross-speaker comparisons above are reported as mentions per 1,000 total words alongside raw occurrence counts. Singular and plural forms (such as tariff and tariffs, or customer and customers) are counted as distinct word tokens unless noted as combined.
Verify in context: Every finding above can be inspected directly in the MouthMath interactive explorer—select any speaker, choose All documents in the timeframe bar, and tap any word to read the sentences behind the count.