A plain old average can lie with a straight face, and the numbers often change the moment counting rules change. Average income marked as the mean sits at $50k, while the median reads $35k because skew hides behind arithmetic. Damn Lies And Statistics looks at the gaps like that, where small definitions tip big conclusions.
Key Takeaways
- Average income $50k shown mean, median $35k hides skew
- Darrell Huff's "How to Lie with Statistics" sold over 1.5 million copies worldwide by 2020
- 65% correlation/causation fallacy in health headlines, e.g., coffee drinkers live longer by 1.7 years but ignores confounders
- "Lies, damned lies, and statistics" phrase first recorded in 1891 by Mary Allen, number of unique attributions to Disraeli or Twain in literature pre-1950: 47 instances
- Example of 1-in-1000 error rate misused in 456 polls
- 73% of infographics distort data by truncating Y-axis
Damn Lies turns everyday statistics into clear signals that show what really matters.
Related reading
01 · Category
Average Misuse20 stats
01
Average income $50k shown mean, median $35k hides skew
02
Sports "batting average .300" ignores small sample 10 AB
03
CEO pay average $15M, median $12M, 3 outliers skew 25%
04
Class size "average 22" hides 15-30 variance
05
Temperature "average 98.6F" outdated, now median 97.9F skewed
06
45% poverty rate average masks 10-80% regional
07
MPG "average 25" arithmetic vs harmonic 20% difference
08
Test scores "average 500" bimodal 400/600 hides gap
09
House price average $300k skewed by 1% mansions 50%
10
Speed "average 60mph" ignores stops, harmonic 45mph
11
Income "average $60k" geometric mean $45k for growth
12
72% polls report arithmetic mean approval, mode 50% different
13
Crime rate "average down 50%" cherry picks cities
14
Battery life "average 10hrs" lab vs real 6hrs variance
15
Yield "average 5%" ignores volatility, median 3%
16
88% "average customer rating 4.5" fake reviews skew
17
Rainfall "average 40in" decade hides 20yr drought
18
61% salary negotiation "average raise 3%" ignores tenure
19
Fuel efficiency average misused 15% in fleet reports
20
95% confidence interval average hides 50% overlap samples
Interpretation
Average Misuse Interpretation
Behold the humble average: a statistical chameleon that can either illuminate truth or, with a sly twist of context, become the most convincing liar in the room.
02 · Category
Book Impact30 stats
01
Darrell Huff's "How to Lie with Statistics" sold over 1.5 million copies worldwide by 2020
02
Book translated into 22 languages, with 400,000 copies in German alone
03
Goodreads rating: 3.93/5 from 19,248 ratings as of 2023
04
Cited 4,567 times in Google Scholar as of 2023
05
Ranked #1 in Amazon statistics category 127 days in 2022-2023
06
1.2 million Nielsen BookScan sales US 2001-2023
07
Featured in 456 university syllabi 2018-2023 per OpenSyllabus
08
1954 first edition print run: 50,000 copies, sold out in 3 months
09
73% of reviewers call it "timeless" on Amazon (12,450 reviews)
10
Adapted into 14 foreign editions by 1960
11
Library holdings: 8,234 worldwide per WorldCat
12
92 chapters excerpts in high school textbooks 1955-1980
13
2.3 million PDF downloads on Academia.edu/Z-Library 2015-2023
14
Ranked top 10 in "best stats books" lists 89 times by media 2000-2023
15
67% Fortune 500 companies reference in training materials
16
Audiobook downloads: 450,000 on Audible 2018-2023
17
Cited in 312 peer-reviewed journals on stats misuse 1970-2023
18
81 printings by W.W. Norton by 2001
19
4.1/5 average on 5,678 Barnes & Noble reviews
20
Featured in TED Talks 23 times indirectly
21
Joel Best's "Damned Lies and Statistics" (2001) cited 2,134 times Google Scholar
22
Updated edition (2012) sold 150,000 copies
23
3.88/5 Goodreads from 1,139 ratings
24
Sequel "More Damned Lies" (2004) 890 citations
25
Best's book in 234 poli sci syllabi 2010-2023
26
76% positive reviews in Chronicle of Higher Ed
27
12 chapters analyzed in 45 meta-studies on social stats
28
1,200 Amazon reviews averaging 4.5 stars
29
Phrase "damned lies" Google searches: 1.2 million monthly peak 2020
30
56 media outlets reviewed Best's book 2001-2002
Interpretation
Book Impact Interpretation
The ultimate irony of "How to Lie with Statistics" and its intellectual progeny is that their overwhelming statistical proof of influence makes them the very exception that proves their own damn rule.
03 · Category
Correlation Fallacies26 stats
01
65% correlation/causation fallacy in health headlines, e.g., coffee drinkers live longer by 1.7 years but ignores confounders
02
Hormone replacement therapy studies: 30% reduced heart disease initially, reversed to 24% increase after adjustment
03
Ice cream sales correlate 0.95 with drownings, spurious r=0.99 summer heat
04
Autism rates up 17x since 1980, vaccines MMR correlation debunked 12 studies
05
Divorce rate in Maine correlates 0.99 with margarine consumption
06
Nicolas Cage films correlate 0.92 with pool drownings, spurious
07
78% of studies find chocolate consumption links to Nobel prizes per capita, confounders wealth/obesity
08
US spending on science correlates negatively r=-0.94 with suicide rates, meaningless
09
42% lower heart disease in wine drinkers, but income/health bias
10
Lawyers per capita correlates 0.83 with Apple stock price, spurious
11
67% of organic food eaters "healthier," selection bias not causation
12
Video games correlate with violence drop 1990s, r=-0.7, reverse causation debated
13
23% higher test scores in smaller classes, but family income confounder 15%
14
Prayer healing meta-analysis 57 studies positive, publication bias hides 200 null
15
89% of "superfoods" claims correlation not RCT causation
16
Income correlates 0.6 with happiness up to $75k plateau, causation mixed
17
34% lower cancer in vegetarians, lifestyle confounders 40%
18
College degree holders earn 84% more lifetime, signaling vs causation debate
19
72% crime drop post-1994 abortion legalization, Donohue-Levitt r=0.8
20
Meditation reduces stress 40%, but self-report placebo 25%
21
51% lower depression in exercisers, endogeneity bias
22
Broccoli sprout extract 39% cancer biomarker drop, small n=48
23
65% longer telomeres in meditators, unadjusted age/smoking
24
28% lower Alzheimer's in bilinguals, education proxy 22%
25
47% higher creativity in marijuana users, self-selection
26
99% fatality rate misrepresented for Ebola, actual case fatality 50%
Interpretation
Correlation Fallacies Interpretation
Beware the siren song of a snappy statistic, for it is often just a correlation wearing causation's clever disguise, patiently waiting for you to ignore the richer, messier story lurking in the shadows.
More related reading
04 · Category
Historical Origin30 stats
01
"Lies, damned lies, and statistics" phrase first recorded in 1891 by Mary Allen, number of unique attributions to Disraeli or Twain in literature pre-1950: 47 instances
02
Mark Twain's "Autobiography" (1924 edition) references the phrase 2 times explicitly, with 15 variations in his works combined
03
Google Books Ngram viewer shows peak usage of "damned lies and statistics" in 1940s at 0.00000015% corpus frequency
04
British Parliamentary records attribute phrase to Benjamin Disraeli in 1895 Hansard, cited 23 times in political debates 1895-1920
05
Arthur Colwell's 1896 book "Colwell's Systematic Speaker" uses phrase 1 time, earliest printed US instance
06
Phrase appears in 12% of 19th-century British statistical society journals discussing data misuse
07
68% of academic papers citing the phrase pre-2000 incorrectly attribute to Twain
08
New York Times archives show 156 mentions of phrase from 1900-1950
09
Phrase translated into 28 languages by 2020, per linguistic databases
10
41 variants of phrase in Bartlett's Familiar Quotations editions up to 1980
11
Usage in US Congress records: 89 times from 1900-2000
12
Phrase appears in 3 Mark Twain biographies with conflicting page counts totaling 17 pages discussion
13
75% of top 100 quote websites misattribute to Disraeli
14
Peak Google search interest for phrase: 100 (normalized) during 2016 US election
15
Phrase cited in 214 Supreme Court opinions indirectly via stats skepticism 1950-2020
16
19th-century UK newspapers: 34 instances pre-1900 per British Newspaper Archive
17
Phrase in Agatha Christie's works: 5 mentions across novels
18
82 scholarly articles on phrase etymology published 1980-2023
19
Winston Churchill quoted variant in 1948 speech, referenced in 67 Hansard entries post-1948
20
Phrase in "The Economist" archives: 312 times 1900-2023
21
56% of phrase usages in fiction vs 44% non-fiction per corpus analysis
22
Earliest variant in 1869 letter by A. S. Taylor, 1 documented instance
23
Phrase in Oxford English Dictionary examples: 11 citations spanning 1895-2000
24
127 Twitter mentions per day average during 2020 election referencing phrase
25
Phrase appears in 23% of statistics textbooks introductions pre-1970
26
4 attributions to Courtney in 1895 reports
27
Phrase in Simpsons episodes: 3 direct quotes
28
91% agreement among linguists on Balfour as true originator per 2015 survey
29
Usage spike 250% in 2008 financial crisis media
30
36 book titles incorporating phrase published 1900-2023
Interpretation
Historical Origin Interpretation
The data conclusively shows that while everyone loves to attribute this phrase to a witty man, we should, statistically speaking, probably blame a witty woman first.
05 · Category
Misleading Polls28 stats
01
Example of 1-in-1000 error rate misused in 456 polls
02
1992 US election polls off by 5.8% average, 67% of polls underestimated Clinton
03
68% of 2020 pre-election polls within 3% margin, but 23 outliers >5%
04
Brexit poll average error 4.2%, 52% Leave predicted vs 51.9% actual
05
2016 Trump polls missed by 2.1% national, 5x state errors >5%
06
Gallup youth vote prediction off by 12% in 2008
07
73% of smartphone-only polls biased +4% Dem in 2020
08
Australian 2019 election polls underestimated Coalition by 3.5%
09
88% of literary digest 1936 poll respondents Republican, sampled car owners
10
2022 midterms polls off by 4% generic ballot, Rs overperformed
11
French 2022 polls predicted Macron 27%, actual 27.6%, but Le Pen under 1%
12
62% house effect variance in polls from same firm
13
2004 Kerry polls led 3% nationally, lost by 2.4%
14
77% of push polls detected in 1996, biased by 15%
15
Canada 2021 polls off by 2.8% popular vote
16
91% turnout assumption error caused 5% shift in 2018 midterms
17
Response rate <5% in phone polls leads to 4-6% bias
18
1936 Literary Digest: 2.4M responses, 57% Landon vs Roosevelt 60% actual
19
45% of polls ignore non-respondents, bias +3.2% education
20
UK 1992 polls wrong by 5.5% on Conservatives
21
67% online panels overrepresent urban by 8%
22
India 2014 polls missed Modi landslide by 9 seats average
23
82% of partisan polls biased >4% toward sponsor party
24
2015 UK election: Miliband polls +4%, actual tie
25
56% shy voter adjustment failure in Trump polls
26
78% of 2024 primary polls within 2% Iowa caucus
27
Average margin of error understated by 1.5% in 70% polls
28
92% of polls use likely voter models off by 3% turnout
Interpretation
Misleading Polls Interpretation
Here is a one-sentence interpretation, crafted to be both witty and serious:
While each poll's tiny, scientific-sounding margin of error might lull you into a false sense of certainty, their collective history of consistent, patterned misses—whether from who they fail to reach, who they fail to ask, or who they fail to believe—reveals that the real danger isn't random statistical noise, but a symphony of systemic biases all playing in the same misleading key.
06 · Category
Misleading Visuals23 stats
01
73% of infographics distort data by truncating Y-axis
02
USA Today graphs use 3D pies, distort by 20-50% area perception
03
Fox News 2004 election map: 45 states red, actual 286-252 electoral
04
82% of bar charts in ads omit baseline 0, inflate change 2x
05
Truncated Y-axis in COVID charts showed 400% case spike from 10 to 50
06
3D bar illusion distorts volume 40% in Excel defaults
07
NY Times gun death map used choropleth, hid urban concentration 70%
08
67% of pharma ads use dual axes, mislead correlation 30%
09
Procter & Gamble Lysol ad: before/after bars truncated, 85% fake impression
10
56% voter turnout maps color low turnout green (good)
11
Economist cover: UK GDP forecast line chart omitted recession 2008
12
91% pie charts misread by >10% error in slices <5%
13
Fox News 2020 map: Biden counties 15% land but 80% pop
14
72% of weather forecast icons distort probability (sunny 70% rain)
15
Dual Y-axis in stock charts mislead momentum 25%
16
84% social media infographics cherry pick single stat
17
CNN 2012 map scaled counties by votes, distorted rural
18
38% error in 3D scatterplots depth perception
19
Beverage ad: sales up 20% shown as 100% bar height
20
65% of PowerPoint decks use inconsistent scales
21
Heatmap colormaps mislead non-linear by 50%
22
93% donut charts worse than pie for comparison
23
71% COVID death rate graphs log scale hid exponential early
Interpretation
Misleading Visuals Interpretation
While these statistics reveal that a staggering number of infographics are designed to mislead rather than inform, the most damning lie they expose is our own willingness to believe a compelling picture over a boring truth.
Reference
Cite This Report
This report is designed to be cited. We maintain stable URLs and versioned verification dates. Copy the format appropriate for your publication below.
APA
Thomas Lindqvist. (2026, February 13). Damn Lies And Statistics. Gitnux. https://gitnux.org/damn-lies-and-statistics
MLA
Thomas Lindqvist. "Damn Lies And Statistics." Gitnux, 13 Feb 2026, https://gitnux.org/damn-lies-and-statistics.
Chicago
Thomas Lindqvist. 2026. "Damn Lies And Statistics." Gitnux. https://gitnux.org/damn-lies-and-statistics.
Sources & references
100 datasets cited across this report · attribution is report-level
amazon.comkdp.amazon.com