How to Lie

with
Statistics
 1
J

,.
.,
,;
There are three kinds of lies: lies, damned Iies, and statirtk:s..
-Disraeli

Stati stical thinki ng will one day be as necessary for efficient

citizenship as the ability to read and write.
- H. G, Wells

It ain't so much the th ings we don't know that get us in trouble.

It's the things we know that ain't so.
-Artemus Ward

Round num bers are always false. - Samu el Johnson

] have a grea t subject (statistics ] 10 write upon, but fee l keenly

my literary incapacity to make it ea sily intelligible without
sacrificing accuracy and th oroughness.
- Sir Francis Galton
 !I
,i
i
~1
Also by Darrell H ufJ
,
:~

HOW TO TAKE A CHANCE

I
u-'ith illustraticns by Irving Gei.1
CYCLES IN' YOUR UFE
How to Lie with ·i
0;

with illustrations by Anatol Kovar sky

I By
DARRELL HUFF
Pictures by IRVING GElS

w· W· NORTON & C O M P A N Y · IN C · New York

 DEDALUS • Acervo • IME
QA276 How to he with statistics.
H889h

To my wlf.
Contents

Acknowledgments 6
Introduction 7
1. The Sample with the Built-in Bias II
(,
2. The Well-Chosen Average '7
3. The Little Figures That Are Not There 37
4. Much Ado about Practically Nothing 53
5. The Cee-Whiz Craph 60
6. The One-Dimensional Picture 66
7. The Semiattached Figure 74
C O P Y R1G HT J9S4 BY DARRELL RUFF AND IR VIN G GE lS
8. Post Hoc Rides Again Il7
ISBN 0 393 OS264 8 Clot h Edition
ISBN 0 393 09426 X Paper Edition
9. How to Statisticulate 100

PRI:ST£ D IN THE U NITlCD STATES OF AM ERICA 10. How to Talk Back to a Statistic 1>'
s.
 !

Acknowledgments
THE PRETIY little instances of bumbling and chicanery
with which this book is peppered have been gathered
widely and not without assistance. Following an appeal
of mine through the American Statistical Association, a
number of professional statisticians-cwho, believe me, de-
plore the misuse of statistics as heartily as anyone alive-
sent me items from their own collections. These people,
I guess, will be just as glad to remain nameless here, I
found valuable specimens in a number of books too, pri-
marily these: Business Statistics, by Martin A. Brumbaugh
and Lester S. Kellogg; Gauging Public Opinion, by Hadley
Cantril; Graphic Presentation, by Willard Cope Brinton;
Practical Business Statistics, by Frederick E. Croxton and
Dudley J. Cowden; Basic Statistics, by George Simpson
and Fritz Kafka; and Elementary Statistical Methods, by Introduction
Helen M. Walker.

"THERE's a mighty lot of crime around here," said my

father-in-law a little while after he moved from Iowa to
California. And so there was-in the newspaper he read.
It is one that overlooks no crime in its own area and has
been known to give more attention to an Iowa murder
than was given by the principal daily in the region in
which it took place.
My father-in-law's conclusion was statistical in an in-

6 7
 .' ,' ~ 't, "' .'> I v . _,_ . ""0-,

I BOW TO LIE WITH STATISTICS INTRODUCTION 9

lonna] way. It was b ased on a sam ple, a remarka bly biased hero laborin g overtime without tfm e-and-a-half in an Ill-
one. Like many a mor e sophisticated statistic it was guilty lit laboratory. Like the "little dash of powder, little pot
of semiattachment : It ass umed that newspaper space of pai nt, " statistics are making many an impo rta nt fact
given to crime re porting is a measure of crime rate . "look like wha t she ain't." A well - '!.~.Jl p~ sta tistic is
A few winters ago a dozen investigators independently better th an Hitler's D ig lie"; it misleads, ye t it cannot be
reported figures on antihistamine pills. Each showed that e.i~~ on you. ", ". . .,
a conside rab le perce nta ge of co lds cleared up afte r treat- This book Is a sort of primer in wap to use sta tistics to
ment. A great fuss ensued. at least in th e advertisements , decei ve. It may seem altogether too much like a manu al
and a medical-product boom was on . It was based on an for swindlers. Perha ps I ca n justify it in the manne r of the
eternally springing hope and alSo on a curious ref usal to retired burglar whose published remin iscences am ounted
look past the statistics to a fact that has been known for to a graduat e course in how to pick a lock and mufBe a
a long time. As Henry G. Felsen , a humorist an d no medi- footfall: The crooks already know these tricks; honest
cal autho rity , pointed out qu it e a while ago, proper treat. men must learn them in self-defense.
ment will cu re a cold in seven days, but left to itself a cold
will hang on for a week.
So i t is with much that you rea d and hear. Averages
and. relat ionships and tren ds and graphs are not al ways
.I
what they seem. There may be more in them th an meets
the e}'c. and there may be a good deal less.
T he secret lan guage of sta tistics, so appealing in a fact-
minded culture, is employed to sensa tionalize. Inflate,
confuse, and oversimplify. Sta tistica l methods and stat is-
tical terms are necessary in reporting the mass data of
social and economic trends, business conditions, "op inion"
polls, th e census. But without writers who use th e words
with honesty and understanding and readers who know
what they mean, th e result can only be semantic nonsense.
In popular writin g on scientific matters the abused statt s-
tic is almost crowding out the picture of the white-jacketed
to H OW TO LIE WTI'H STATISTICS

J/ I IJ
h 1
",.-
f\ I 1/ CHAPTER

/ ~ The Sample with

~/ .j the Built-in Bias
V}3
P

-\ h
J v IRE AVERAGE Yaleman, Class of '24," Time magazine
noted once, commenting on somethi ng in the New York
)
~
V I
J
Sun, "makes $25,111 a year."
Well, good fa, him!
But wait a minute. What does this impressive figure

VIf mean? Is it, as it appea rs to be , evidence that if you send

your boy to Yale you won't have to work in your old age

V;
and neither will he?
Two things about the figure stand out at first suspicious
glance. It is surprisingly precise. It is quite improbably

~ r-.....
salu brious.
V There is sma ll likelihood th at the average income of any
far-flung group is ever goin g to be known d own to the
/'
dollar. It is not particularly proba ble that you know your
n
I'

now T O LIE WITH STATI STICS THE SAMPLE wrm TIlE BtiILT~IN BIAS '3
own income (or last year so precisely as th at unless it was This is the sampling procedure. which is th e heart of the
aUderived. from salary. And $25,000 incomes are not often greater part of the statistics you meet on all sorts of sub-
all salary; people in that bracket are likely to have well- jects. Its basis is simple enough, although its refinements
scattered investments . in practice have led into all sorts of by-ways, some less
Furthermore, this lovely average is undoubt edly calcu- th an respectabl e. If you have a barrel of beans, some red
lated from the amounts the Yale men said they earned. and some white, th ere is ooly one way to God out exactly
Even if they had the hODOr system in New Haven in '24, how many of each color you have : Count 'em. However.
we cannot be sure th at it works so well after a quart er of you can find ou t approximately how many are red in m~ch
a ce nt ury that all these reports are honest o nes. Some easie r fashion by pulling out a handful of beans and coun t-
peo ple when asked their incomes exaggerate out of vanity ing just those, figuring that the proportion wi ll be th e same
all through th e barrel. If your sample is lar ge enough and
selected properly, it will represent th e who le well enough
for most purpo ses. If it is not, it may be far less accurate
than an intelligent gu ess and ha ve nothing to recommend
. .',.
......._ . ~ -~
it but a spurious air of scienti6c precision . It is sad truth
th at conclusions from such samples , biased Or too small or
or optimism. Others minimize, especially, it is to be fea red, both. Lie beh ind mu ch of wha t we read or think we know .
on income-tax returns; and having do ne this may hesita te The re port on the Yale men com es from a sam ple. We
to contra dict themselves on any other paper. Who knows ca n be pretty sure of that because reason tells us tha t no
what th e revenuers may see ? It is possible th at th ese two one ca n get hold of all the living members 01 th at class of
tendencies, to boast and to un derstate, ca ncel each othe r '24. There are bound to be many who se addresses are un -
out. but it is un likely. One tendency may be far stro nge r known tw enty-five years lat er .
than th e other. and we do not know whi ch one.
We have begun th en to account for a figure th at com-
mon sense tells us can hardly represent the truth . Now
let us put our Bnger on the likely source of the biggest
error, a source th at can produce $25,111 as the "average
income" of som e men whose actual average may well be
nearer half th at amount .
14 HOW TO l.JE WITH STATISTICS THE SAMPLE '\\'ITH THE BUILT-IN BlAS '5
And, of those whose addresses are 1.00","1) , many will not the men who , twenty-five yea rs or so after becoming Yale
reply to a que stionnaire, particularly a rather personal bachelors of arts, have not fulfilled any shinin g promise.
one. With some kinds of mail questionnaire. a five or ten They are clerks, mechanics, tramps, unemployed alco-
per cent response is qu ite high. This one should have holics, barely survi ving ....-rite rs and art ists .. . people of
done better than that, but noth ing like one hundred per whom it wo uld take half a dozen or more to add up to an
cent. inco me of $25,111. T hese men do not so often register at
So we find th at t he income figure is based on a sample class reunions, if only because they cannot afford the trip.
composed of all class members wh ose addresses are known
and who replied to th e qu estionnaire. Is th is a repr esenta-
tive sample? That is , ca n th is group be assumed to be ~~are -poor little Iambs
equal in income to th e unrepresen ted group. th ose who
cannot be reached or who do Dot reply? r~ho ~ave lost om- way

Who are the little lost sheep down in th e Yale rolls as Who are those who chucked the questionnaire into th e
"address unknownb? Are th ey the big-i ncome earn ers-. nearest waste basket? We cannot be so sure about these,
the Wall Street men, the corporation direct ors, the man u- bu t it is at least a fair guess th at many of th em are just
facturing and utility executives? No; the addresses of not making enough money to brag abo ut . They are a
the rich will not be hard to com e by. Many of th e most little like the fellow who found a note clipped to his first
prosperous mem bers of th e class can be found th rough pay check su ggesting that he consider th e am ount of his
Who's Who in Americn and oth er reference volumes even salary confldenttal and not mat erial for th e interchan ge of
i.E they have neglected to keep in touch with th e alumni office confidences. "Don 't worry," he told the boss. "I'm
office. It is a good guess th at the lost names are those of just as ashamed of it as you are:'
16 HOW TO LIE wrm STATISTICS THE SA MPLE wrru THE BUB.T-L'! BIAS r']

It becomes pretty clear th at th e sample has omitt ed two could learn a good deal more by going to their hous es and
groups most likely to depress the avera ge. The $25,111 saying you wanted to buy old magazines and wha t could
fi gure is be gin ning to explain itself. If it is a true fi gure be had? Then all you had to do was count the r ale Re-
for anyt hing it is one mer ely for that special group of the t.-iews an d th e Love Romance.s. Even that dubious device.
class of'24 whose addresses are known and who are willing of course, does not tell you what peo ple read. only what
to stand up and tell how mu ch th ey earn . Even that re- they h ave been exposed to.
quires an assumption that the gentl emen are telling the Similarly, the next time you learn from your reading
truth. th at the average American (you hear a good deal about
Such an assumption is not to be made lightl y. Experi- him th ese days, most of it faintly improbable ) brushes his
ence from one breed of sampling study , th at called market teeth 1.02 time s a day-a Bgure [ have just made up. but
research, suggests that it ca n hardly ever be made at all. it may be as good as anyone else's-cask yourself a qUe5-
A bouse-to-house survey pu rporting to study magazine tion. How can anyon e ha ve found out such a thing? Is a
readership was once made in which a key ques tion was: woman wh o has read in coun tless advertisements that DOn-
\Vhat magazines does )"our household read? \Vben the brushers are social offenders going to confess to a stranger
results were tabulated and ana lyzed it appeared th at a that she doe s not brush her teeth regularly? The statistic
great many people loved Harpers an d not very many read
True Story. Now there were publishers' figures around at
the time th at showed very clearly that True Story ha d
more millions of circulation th an Harper's had hundreds
of th ousands. Perhaps we asked the wron g kind of people,
the designers of th e survey said to them selves. But no,
the qu estions had bee n asked in all sorts of nei ghborhoods
all around th e country. The only re asonable conclusion
th en wa s that a good many of the respondents, as people
are called when th e)' an swer such qu estions, had not told
the truth. About all th e suntey had uncovered was snob-
o
I
bery.
In the end it was found th at if you wanted to know
what certain people read it was no use asking them. You
 B OW 1'0 LIE WITH !lTATIS11CS TIlE SAMPLE WITH TIlE Bun.T-L~ BIAS

may h ave meaning to one who wants to know only what zines reveal th eir inherent lack of meaning.
people say abou t tooth- brushing but it docs not tell a A psychiatrist reported once th at pr actically everybod y
great deal abou t the frequency with whic h bristle is ap- is neurotic. Aside from th e fact that such use des troys any
plied to incisor. meaning in the word "neurotic," take a look at the man's
A river cannot, we are told . rise above its source. Wen, samp le. That is, whom has the psychiatrist been observ-
it can seem to if there is a pumping sta tion concealed ing? It turns out th at he has reached this edifying con-
somewhere about. It is equa lly true that the result of a clusion from studying his patient s, who arc a long, long
sampling study is no better than the sample it is base d on. way from being a sample of the population. If a man were
By the time the da ta have been filtered th rough layers of normal, our psych iatrist would never meet him.
statistical manipulation and reduced to a decimal-pointed
average, the result begins to take on an aura of conviction
that a closer look at the sampling would deny.
Does early discovery of cancer save lives? Probably.
But of the figures conunonly used to p rove it the best that
can be said is that they don't. These, the records of the
Connecticut Tumor Registry. go back to 1935 and appear
to show a substantial increase in the five-year survival rate
from that yea r till 1941. Actually those records were be-
gun in 1941, and everything earlier was obtained by
tracing back. Many pat ients had left Connecticut, and
whether they had lived or died could not be learned . Give that kind of second look to the th ings yo u read,
According to the medical repo rter Leonard Engel, th e and you can avoid learning a whole lot of things that are
built-in bias th us created is "enough to accoun t for nearly not so.
the whole of the claimed improvement in survi val rate." It is worth keep ing in mind also that the dependability
To be worth much, a report based on sampling must of a sample can be destroyed just as easily by invisible
use a rep resentative sample. which is one from which sources of bias as by these visible ones. That is. even if
every source of bias has bee n removed. Tha t is where ow you can't find a source of d emonst ra ble bias. allow your.
Yale figure shows its worthlessness. It is also wher e a great sell some degree of skepticism abo ut the results as long as
many of the things you can read in newspapers and maga- there is a possibility of bias somewher e. There always is.
.. HOW T O LIE WlTII STATISTICS

The presidential elections in 1948 and 1952 were en ough to

THE SAMPLE WITII TRE BtJILT-IN BIAS

sample is a part. Every tenth name is pulled from a file

"
prove that, if there were any doubt. of index cards. Fifty slip s of pap er are tak en from a hat-
For further evidence go back to 1936 and the Literary fuI. Every twentieth person met o n Market Street is in-
Digest's famed fiasco. The ten million tel ephone and terviewed. ( But remember that this last is not a sample
Digest subscribers who assured the ed itors of the doomed of the population of the world. o r of the Un ited States, or
of San Francisco. but only of the people o n Market Street
at the time. One interviewer for a n o pinion poll said that
she got her people in a railroad station be ca use "all kinds
of peo ple can be found in a sta tion ." It had to be pointed
out to her that mothers of small children, for instance,
might be underrepresented there. ]
The test of the random sample is this: Does every name
or thing in the whole group have an equal cha nce to be in
the sample?
The purely random sample is the only kind that can be
examined with entire confidence by mea ns of statistical
magazine that it would be Landon 370, Roosevelt 161
theory. but there is one thing wrong with it . It is so diffi-
came nom the list that bad accurately pred icted the 1932
cult and expensive to obtain for many uses that sheer cost
election. How could th ere be bias in a list already so
eliminates i.t. A more economical substitute, whi ch is al-
tested? There was a bias, of course, as college theses and
most universally used in such Gelds as opinion polling and
other post mortems found: People woo could aHord tele-
market research, is called stratified ran dom sampling.
phones and magazine subscriptions in 1936 were not a
To get this stratified sample yo u di vid e yo ur universe
cross section of voters. Economically they were a special
into several groups in proportion to their known preva~
kind of people. a sample biased because it was loaded
Ience. And right there your trouble can begin : Your in-
with wh at turned ou t to be Republican voters. The sample
fonnation about their proportion m ay not be correct. You
elected Landon , but the voters thought otherwise.
instruct yow interviewers to see to it that they talk to so
The basic sample is the kind called "random," It is se-
~any Negroes and such-and-such a percentage of people
lected by pure chance from the "universe," a word by
m each of several income brackets, to a specified number
which the statistician means the whole of which the
of fanners, and so on. All the while the group must be

....
 now T O LIE WITH STATISTIC) THE SAMPLE WITH THE Bu:u..T~IN BlAS
'J
divided equally between persons over forty and under
forty years of age.
Th at sounds fine- but what happens? On the que stion
of Negro or white the interviewer will judge correctly
most of the time. On income he will make more mistak es.
As to farmers-how do you classify a man who farms part
lime and works in the city too? Even the question of age
can pose some problems which are most easily settled by
choosing only responde nts who obviously are wen under
or well over fort y. In that case th e sample will be biased
by the virtual absence of the late-thirt ies and early-forties
age groups. You can't win.
On top of all this, how do you get a random sample
within the stratification? Th e obvio us thing is to start
with a list of everybody and go after names chosen from
it at random: but tha t is too expensive. So you go into the
streets-and bias your sample against stay-at-homes, You problem, as with anything based on sampling, is how to
go tram door to door by da y-and miss most of the em- read it ( or a popular summary of it ) without learn ing too
ployed people. You switch to evening interviews-and much that is not necessarily so. Th ere are at least three
neglect the movie-goers and night -clubbers. levels of sampling involved. Dr. Kinsey's samples of the
The operation of a poll comes down in the end to a popul ation ( one level ) are far from random ones and may
running battle against sources of bias, and this battl e is not be particularly repr esent at ive, but th ey are enormous
cond ucted all the time by all the reput able polling organ i- sample s by comparison with anyth ing done in his fiel d be-
zations . What the reade r of the reports must remember is fore and his figures must be accepted as revealing and im-
th at the battl e is never won. No conclusion that "sixty- portant if not necessarily on the nose, It is possibly more
seven per cent of the American people are against" some- important to remember that any qu estionnaire is only a
thin g or other should be read without the lingering sample (another level ) of the possible questions and that
questi on, Sixty-seven per cent of which American people? the answer the lady gives is no more than a sample (third
So with Dr, Alfred C. Kinsey's "female volume." The level) of her attitudes and experiences on each question.

i ,.
 wrrn
r
HOW TO LIE STATlSTIal THE SAMPLE wrra THE BUILT-IN BIAS

The kind of people who make up the interviewing staH sounded good rather than what he actua lly believed? It is
can shade the result in an interestin g fashion. Some years also possible that the different group s of interviewers
ago, during the war, the National Opinion Research Center chose different kinds of people to talk to.
sent out two sta ffs of intervi ewers to ask three q uestions In any case the results are obviously so biased as to be
of five hundred Negroes in a Southern city. White inter- worthless. You can judge for yourself how many other
viewers made up one staff, Negro the other. poll-based conclusions are just as biased, just as worthless
One question was, "Would Negroes be tr eated better -but with no check available to show them up.
or worse here if the Japanese conquered th e U.S.A.?"
Negro interviewers report ed that nine per cent of those
they asked said "better." White intervie.....ers found only
two per cent of such responses. And while Negro inter-
viewers found only twenty-Eve per cent who thought
Negroes would be treated worse, white interviewers turned
up forty-five per cent.
When "Nazis" was substituted for "Japanese" in the
question, the results were similar.
The third qu estion probed att itudes that might be based
on feelings revealed by th e first two. "Do you think it is
more important to concent rate on beating the Axis, or to
make democracy work bett er here at home?" "Beat Axis"
was the reply of thirty-nine per cent, accord ing to the
Negro interviewers; of Sixty-two per cent, according to
th e white.
Here is bias introduced by unknown factors. It seems
likely that the most effective factor was a tendency that
must always be allowed for in readin g poll results, a desire
to give a pleasing answer. Would it be any wonder if ,
when answering a question with connotations of disloyalty
in wartime, a Southern Negro would tell a white man what

•
 H OW TO LIE WTnJ STATISTICS

You have pretty fair evidence to go on if you suspect

that polls in general are biased in one specific direction.
the direction of th e Literary Digest error. This bias is
toward the person with more money. more edu cation.
more illformation and alertn ess. better appearance. more
<!J----- CHAPTER 2
conventional behavlor, and more settled habits th an th e
average of the population be is chosen to rep resent.
You can easily see what prod uces this. Let us say that
. lThe Well - Chosen Average
you are an intervi ewer assigned to a sheet com er, with
One interview to get . You spot two men who see m to fit ~
the catego ry you must complete : over forty , Negro, urban.
One is in clean overalls, decently patched, neat. The other
is dirty and he looks surly. With a job to get done, you
approach the more likely-looking fellow, and your col-
leagues all over the country are making similar decisions.
Some of the strongest feeling against public-opinion You. I trust, are not a snob, and I certainly am not in the
polls is found in liberal or left-wing circles, where it is real-estate business. But let's sa)" that you are and I am
rat her commonly believed that polls are genera lly rigged . and that you are looking for property to buy along a road
Behind this view is th e fact that poll results SO often fail that is not far from the Calrfomla valley in which I live.
to square w ith the opinions and desires of those whose Having sized you up, I take pains to tell you th at the
thinking is not in the conservative direction. Polls, the y average income in this neighborhood is some S1 5,000 a
point out, seem to elect Republicans even when voters year. Maybe that clinches your interest in living here;
shortly thereafter do otherwise. anyway, you buy and tha t handsome figure sticks in your
Actually, as we have seen, it is not necessary that a poll mind. More than likely, since we have agreed that for tbe .
be rigged-that is. that the results be deliberately twisted purposes of the momen t you are a bit of a snob. vou toss / . t , '} <> (

in order to crea te a false impression. The tendency of the it in casually wh en tellmg yow friends about where you
sample to be biased in this consistent direction can rig live.
it automatically. A year or so later we meet again. As a member of SOme
taxpayers' committee I am circulating a petition to keep

I
1_
 HOW TO LIE wrru St"ATlSTlCS
the tax rat e down or asses sments down or bus fare down.
My plea is th at we cannot afford the increase: After a.n.
II THE WELL-CHOSES AVERAGE

H in this neighborhood there are more families with in.

comes of $5,000 a year than with any other amount,
29

the average income in th is neighborhood is only $3.500 a $5,000 a year is the modal income.
year. Perhaps you go along with me and .my .C?~~~tt~ In th is case. as usually is tru e with income ligures. an
in this- you're not only a snob, you're ~i ngr too- but you unqualified "average" is virtually meanin gless. One factor
can't help being surprised to hear about that measly that adds to the confusion is that with some kind s of in-
$3,500. Am 1 lying now, or was] lying last year? formation all the averages fall so close together that, for
You can't pin it on me eithe r time. That is the essential casual purposes, it may not be vital to distinguish among
beauty of doing your lying with statistics. Both those I· them.
figures are legitimate ave rages, legally arrived at. Both If you read that the ave rage height of the men of some
represent the same da ta. the same people. the same in - primitive tribe is only five feet, you get a fairly good idea
comes. All the same it is obvious that at least one of of the sta ture of these people . You don't have to ask
them must be so misleading 35 to rival an out-and-out lie. whether that average is a mean, median, or mode; it
My trick was to use a d ifferent kind of ave rage each would come out about the same. ( Of co urse. if you are in
time, the word "average" ha ving a very loose meaning. It
is a trick commonly used, somet imes in innocence but
I the business of man ufacturing overalls for Africans yOt)

often in guilt. b v fellows wishing to influence public opin-

ion or sell advertisin g space. When you are told that
I
something is an average you still don't know very much
about it unless )'ou can find out which of the common
kinds of average it is-mean, median. or mode.
Th e 815,000 figure I used when I wanted a big one is a I . -:_

mean , the arithmetic average of the incomes of all the

families in the neighborhood. You get it by adding up
all the incomes an d d ividi ng by the number there are .
The smaller figure is a med ian. and so it tells you that
half the families in question have more than $3,500 a
year and half have less. I might also have used the mode,
which is the most frequently met-with figure in a series.
 JO H OW TO LIE WITH STATISTICS THE WELL-CHOSEN AVERAC£
3'
would want more information than can be found in any a ne ar-by village or elderly retired people on pen sions. But
average. T his has to do with ranges and deviati ons, and th ree of the inha bit ant s are million aire week-enders and
well tackle that one in the next chapter. ) th ese three boost th e total income, and therefore the arith-
Th e different average s come out close togeth er when
you d eal with data, such as those having to do with many
human charac teristics. th at have the grace to faU close
to what is called. the normal distribution. If you d raw a
curve to represent it yo u get something shaped like a bell.
and mean, median , and mod e fall at th e same point.
Cons equently one kind of average is as good as another
for describing the heights of men, but for describ ing th eir
pocketbooks i t is no t . If you sho uld list the annual incomes meti c average, eno rm ously. They boost it to a figure that
of all th e families in a given city you might find th at th ey practical ly everybody in the neighborhood has a good deal
ranged from not much to perhap s $5O,(XX) or so, and you less th an . You have in rea lity the case that sounds like a
might fi nd a few very large ones. More th an ninety-five joke or a figure of speech : Nearly every body is below
per cent of th e incomes would be under $10,000, putting avera ge.
th em way over toward the left-hand side of the curve. T hat's why whe n you read an anno unceme nt by a cor-
Instead of be ing symme trical, like a bell, it would be poration executive or a bus iness prop rietor th at the aver-
skewed. Its sha pe would be a littl e like th at of a child's age pay of th e people who work in his establi shme nt is so
slide, th e ladder rising sharply to a pea k, th e worki ng part much , th e figure may mean someth ing a nd it may not.
slop ing gradually down. The mean would be qu ite a d is- H th e average is a med ian , you ca n learn some thing sig-
tance from the medi an . You can see what thi s would do nificant from it : Half the employees make more than that ;
to th e validity of any comparison mad e bet ween th e half make Jess. But if it is a mean ( and believe me it may
"average" ( mea n ) of one year and th e "a verage" ( media n) be th at if its nature is unspecified ) you may be getting
of anothe r. nothing more re vealing th an the average of one $45,000
In the neighborhood where I sold you some p roperty the income-the propnetor's-cand the salaries of a crew of un-
two averages are p articularly far apart be cause th e distri- derp aid workers. "Avera ge annual pay of $5,700" may
but ion is markedly skewed. It happens that most of yo ur conceal both the $2,000 salaries and the owner's profits
neighbors are small farmers or wage earners employed in taken in the form of a whopping salary.

I
I
 H OW TO LIE WITH STATISTICS 11
Let's take a longer look at that one. The fadng page
shows how many people get how much. Th e ross might
like to express: the situation as "average wag e $5,700"- Iitt
using that deceptive mean . The mode, however, is more
revealing : most common rate of pay in this business is
$2,000 a year. As usual, the median tells more about the
situa tion than any othe r single figure does ; half the people
get more than $3,000 and half get less.
,
$45,000

$IS,OOO

~f
How neatly this can be worked into a whipsaw device
in which the worse the story. the better it looks is Illus-
tra ted in some compa ny statements. Let's try our hand at $10.000
one in a small way.
You are one of the three partners who own a small
manufacturing business. It is now the end of a very good
4..-ARITHMETICAL AVERAGe
$5,700
year. You have paid out $198,000 to the ninety emp loyees

Jiti~
who do the work of makin g and shipping the chairs or
what ever it is that you manufacture. You and your part .
ners have paid yourselves $11.000 each in salaries. You $5,000
find there are profits for the year of 345,000 to be divided
equally among you. How are you going to describe th is'!
To make it easy to und erstand, you put it in the form of
averages. Since all the employees are doing about the
same kind of work for similar pay. it wo n't make much
difference whethe r you use a mean or a median. nlis is
what yo u come out with :
Average wage of employees $ 2,200
Average salary and profit of owne rs 26,000
Th at looks terrible. doesn't it? Let's try it another way.
 H OW TO LIE WITII STATIS·.n CS TIIE WELL-(:HOSEN AVERAGE 35
Take $30,000 of the pro fits and distribute it among the of employees ran ging all the way from beginn ing typist
three partners as bonuses. And th is time when yo u aver- to president with a seve ral-hu ndrcd-thousand-dollar bonus,
age up the wages, includ e yourself and your pa rtners. And all sorts of thin gs can he covered up in th is manner.
be sure to use a mean. So when you see an avcra ~e-pay figure. fi rst ask: Aver-
age of wha t? w ho's included? The United States Steel
Average wage or salary __ __ $ 2,806.45
Corporation once said that its employees average weekly
Average profit of owners _ __ 5,000.00
ea rnings went up 107 per cent between 1940 and 1948.
Ah. That looks be tte r. Not as good as you could make it So they did-but some of the punch goes out of th e magni-
look, but good enough. Less than six per cent of the ficent increase when you not e th at the 1940 fi gure includes
money available for wages and profits has gone into a mu ch larger number of partially employed peo ple. If
profi ts. and you can go further and show that too if you you work half-tim e one year and full -time the next. your
like. Anyway , you've got figures now that you can pub- earnin gs will double, hut tha t does n't ind icate anyth ing at
lish, post on a bulletin board . or use in bargaining. all about your wage rat e.
You may have read in the paper th at th e income of the
average American fam ily was $3,100 in 1949. You should
not try to make too much out of th at figure unless you also
know what "family" has bee n used to mean , as wen as
what kind of average th is is. ( And who says so and how
be kno~"S and how accu rate the figure is. )
This one happens to have come from the Bureau of the
Ce nsus. If you have the Bureau's repo rt you'll ha ve no
trouble finding the rest of the information you need right
there : This is a med ian; "family" signifies "two or more
persons related to each other and living together." ( If
persons living alone are included in the group the median
slips to $2,700, which is quite dlfferent. ) You will also
This is pretty crude beca use th e example is simplified, learn if you read back into the tables that th e fi gure is
but it is nothing to what has been done in the name of based on a sample of such size that there are nineteen
accounting. Given a complex corporation with hierarchies chances out of twenty th at the estimate- $3,l 07 before it
 H OW TO L IE WITH STATISTICS

wa5 round ed-is correct within a margin of $59 plus or

min us.
That probability and that margi n add up to a pretty
good estim ate. The Census people ha ve both skill enough
and mone y enough to bring their sampling studies do wn
to a fair degree of precision. Presumably they have no CHAPTER 3
particular axes to grind . Not all the figures you see are
born under such happy circumstances, nor are all of them The Little Figures
accompanied by any inform ation at all to show how pre-
cise or unprecise th ey may be. We'U work that one over That Are Not There
in the ne xt chapter.
Meanwhile you may want to try your skepticism on
some items from "A Letter from the Publisher" in Time
magazine . Of new subscribers it said, "Their median age
is 34 years and their aver age family income is $7,270 a
year ." An earlier survey of "old TIMErs" had found that USERS report 23% fewer cavities with Doakes' tooth paste,
their "median age was 41 years. • . • Average income was the big type says. You could do with twenty-three pe r
$9,535. . • ." The natural question is why, when median cen t fewer aches so you read on . These results, you find
is given for ages both times, the kind of average for In - come from a reassuringly "indepe ndent" laboratory, and
comes is carefully unspecified. Could it be that the mean the account is cert ified by a certified public accountant.
was used Instead because it is bigger, thus seeming to What more do you want?
dangle a richer read ership before advertisers? Yet if you are not outstandingly gullible or optim istic,
)"ou will recall from experience that one tooth paste is
seldom much better than any other. Then how can the
Doakes people report such results? Can they ge t away
with tellin g lies, and in such big type at that ? No, and
You might also try a game of what-kind-of-average-are- they don't ha ve to. There are easier ways and more effec-
you on the alleged prosperity of the 1924 Yales reported at tive ones.
the beginning of Chapter I. The principal joker in this one is the inadequate sample

37
 ROW T O LIE WUH STATISTICS 39
-cstatistically inadequate, that is; for Doakes' purpose it The importance of using a sma ll group is this: With a
is just right. That test group of users, you discover by large group any difference prod uced by chance is likely to
reading the small t)'Pe, consisted of just a dozen persons. be a small one and unworthy oCbig type. A two-per-cent-
(You have to hand it to Doakes, at that , for giving you a improvemen t claim is not going to sell much tooth paste.
sporting chance. Some ad vertisers would omit this infor- How results that are not indicative of anythin g can be
mation and leave even the statistically sophisticated only produced by pmc chan ce-given a small enough number
a guess as to what species of chicanery 'A'aS afoot. His of cases- is something you can test for yourself at small
sample of a dozen isn't so bad either, as these th ings go. cost. Just start tossing a penny, How often will it com.e
Something called Dr. Cornish's Tooth Powder came onto up head s? Half the time, of course, Everyone knows that.
th e market a few years ago with a claim to have shown Well, let's check that and see. . . , I have just tried ten
"considerable success in correction of . .. dental caries." tosses and got heads eight times, which proves that pennies
The idea was that the powder contained urea, which
laboratory work was supposed to have demonstrated to BY ACTUAL TEST (on, test]
be valuable for th e purpo se. The pointl essness of this was
th at the experimental work: had been purely preliminary
and had been done !'n precisely six cases.}
But let's get back to how easy it is for Doakes to get a
headline without a falsehood-in it and everythi ng certified
•••••
e•
(=> " .
• ~~ e
Sci.", I"'" that toss.d
at that. Let any small group of persons keep count of pe nnies come up bea ds
cavities for six months, th en switch to Doakes'. One of 80 I" C1 nt of the tim•.
three things is bound to happen: distinctly more cavities,
distinctly fewer, or about the same numbe r. If the first
or last of these possibilities occurs, Doakes & Company
files th e fi gures ( well out of sight somewhere ) and tries come up heads eighty per cent of th e time. Wen, by tooth
again, Sooner or later, by the operation of chance, a test paste statistics they do, Now try it yourself, You may get
group is going to show a big improvement worthy of a a fifty-fifty result, but probably you won't; your result,
headline and perhaps a whole advertising campaign. This like mine, stands a good chance of being quite a ways
will happen whether they adopt Doakes' or baking soda away from fifty-fifty. But if your patience holds out for
or just keep on using their same old dentifrice. a thousand tosses you arc almost ( though not quite) cer-
 wrrn STATISTICS
1I0W TO LIE
1 TIlE UTILE F1CURES TIL\T ARE NOT THI:Rli. i'
start to have no mean ing. Something like fifteen to twenty-
five times this many childr en would have been needed to
obtain an answer signifying anything.
Many a great, if fleeting, med ical discovery has been
launched similarly. "Make haste," as one physician put it,
"to use a new remedy before it is too late."

tain to corne out with a result very close to half heads-a

result, that is, which represents the real probability. Only
when there is a substantial number of trials involved is
the law of averages a useful description or prediction.
How many is enough? Th at's a tricky one too. It de-
pends among other things on how large and how varied
a population you are studyi ng by sampling. And some-
times the number in the sample is not what it appears
to be.
A remark able instance of thi s came out in connection
with a test of a polio vaccine a few years ago. It appeared
to be an impr essively large-scale experiment us medical
ones go: 450 children were vaccinated in a community
and 680 were left unvaccinat ed, as cont rols. Shortly
thereafter the community was visited by an epide mic.
Not one of th e vaccinated children cont racted a reco g-
nizable ew e of polio. Th e guilt does not always lie with the medical pro-
Neith er did any of the controls. \Vhat the experimenters fession alone. Publi c pressure and hasty journalism often
had overlooked or not understood in selling up th eir launch a tr eatm ent that is unpr oved, particularly when
project was the low incidence of paralytic polio. At the the demand is great and the statistica l background hazy.
usual rate, only two cases would have been expected in So it was with th e cold vaccines th at were popular some
a group this size, and so the test was doomed from the years back and the antihistamines more recently. A good
 HOW TO LIE W ITH STAT ISTICS THE LITrLE FIGURES rncr ARE NOT THERE 43
deal of the popula rity of these unsuccessful "cures" sprang lion th at it is worse than useless. Knowing nothing about
from the unreliable nature of the ailment and from a de- a subject is frequently healthier than knowing what is not
feet of logic. Given time, a cold will cure itself. so, and a little learn ing may be a d angerous thing.
How can you avoid being fooled by unconcluslve Altogether too much of rece nt American housing. for
results? Must every man be his own sta tistician and study instance, has been planned to fit the statistica lly average
the raw data for himself? It is not tha t bad; there is a test family of 3.6 persons. Translated into reality this means
of significance that is easy to understand . It is simply a

~~
way of reporting how likely it is that a test figure repre- TMf PROCRUSTEAn
STffflSTIC
sen ts a real result rather than someth ing produced by
chance. This is the littl e figure tha t is not there-on the
assumption that you, the lay reader, wouldn 't unde rstand
it . Or that, where there's an axe to grind, you would.
~_...JUl_ )LJUU
If th e source of your Information gives you also the three or four persons, which, in turn , means two bedrooms.
degree of significance, you'll have a better idea of where And thi s size family, "average" th ough it is, actua lly makes
you stand. This degree of significance is most simply up a minori ty of all families. "We buil d average houses
expressed as a probabil ity, as when the Bureau of the for average families," say th e builders- and neglect the
Census tells }'OU that there are nineteen chances out of majority that are larger or smaller. Some areas, in con-
twenty that their figures have a specified degree of preci- seq uence of this, have been overbuilt with two-bedroom
sion. For most purposes nothing poorer than this five per houses, unde rbuilt in respect to smaller and larger units.
cent level of significance is good enough. For some the So here is a sta tistic whose misleadin g incompleteness has
demanded level is one per cent , which means that there had expe nsive consequences. Of it the American Public
are ninety-nine chances out of a hundred that an apparent Health Association says: "When we look beyond the arith-
d ifference, or whatno t, is real. Anythi ng this likely is metica l average to the actua l range which it misrepresents,
sometimes described as "pract ically certain." we find that the three-person and four-person families
There's another kind of little figure tha t is not there, one make up only 45 per cent of the total. Thirty-five per cent
whose absence can be just as damaging. It is th e one that are one -person and two-person; 20 per cent have more
tells the rang e of things or their deviation from the aver- than four persons: '
age that is given. Often an average-whether mean or . Common sense has somehow failed in the face of the
med ian, speclfled or unspecified-cis such an oversimplifica, convincingly precise and authoritati ve 3.6. It has some-
 HOW TO LIE WITH STATISTICS TIlE LJTJLE FJCL"RES THAT ARE NOT TIIERE

how outweigbed what every body kno ws from observat ion: of th e other fifty per cent of parents in discovering that
that many families are small and quite a few are large. th eir ch ild ren are "advanced." But harm can corne of the
In somewha t th e same fashion th ose little figures that efforts of the unh appy parents to force their ch ildren to
are missing from what are called "Gesell 's norms" have conform to th e norms and thus be backward no longer.
produced pai n in papas and mamas. Let a parent read. All this does not reflect on Dr. Arnold Gesell or his
as many have done in such places as Sunday rotogravure methods. The fault is in the filtering-down process from
sections, that "a child" learns to sit erect at the age of so the researcher through the sensational or ill-informed
writer to the reader who fails to miss the figures th at have
disappeared in the process. A good deal of the misunder-
stand ing can be avo ided if to th e "norm" or average is
added an indicati on of th e range. Paren ts see ing th at their
youngs ters fall with in the normal range will qu it vv-o rrying
about small and meanin gless differences. Hardl y anybody
is exactly norm al in any way. just as one hundred. tossed
pe nnies will rar ely come up exactly fifty heads and fifty
tail s.
Confusing "norma l" with "desirable" makes it all the
worse, Dr. Gesell Simply stated some observed facts; it
was th e parent s who, in reading the books and arti cles,
concluded that a child who walks lat e by a day or a mouth
must be infer ior.
A good deal of th e stu pid cri ticism of Dr. Alfred Kinsey's
many months and be thinks at once of his own child . Let well-known ( if hardly well-read ) repo rt came from taki ng
his chil d fail to sit by th e specified age and th e parent must normal to be equivalent to good , right, desirable. Dr.
conclude that h is offspring is "retarded" or "subnormal" Kinsey was accused of co rru pting youth by giving them
or something eq ually invidious . Since half th e chil dren ideas and part icularly by calling al1 sorts of popular but
are bound to fail to sit by th e time mentioned, a good una pproved sexua l practices normal. But he simply said
many parents are made unhappy , Of course, speaking that h e ha d found these act iviti es to be usual , which is
mathematically, this unhappiness is balanced by the joy what normal me ans, and he did not stamp them with any
46 now TO LIE WITH STATISTICS THE LITn.E FIG~ES T HAT AR lo: NOT THERE 47
seal of approval. Wh ether they were na ughty or not did indus trial developments" in th e news magazin e Fortnight :
not come within what Dr . Kinsey considered to be his "a new cold temper bath wh ich trip les the hardness of
prov ince. So he ran up aga inst something th at has plagu ed steel, from Westinghouse:
ma ny anothe r ob ser ver: It is dangerous to men tio n any Now that sounds like quite a development . .. until you
subject having high emotional conten t withou t hastily try to p ut }'OUT finger on what it means. And then it be-
saying where yo u are for or agin it. comes as elusive as a ball of quicksilver . Does the ne w
The deceptive th ing abou t the little 6gure that is not ba th make just any kind of steel th ree times as hard as il
there is th at its a bsence so often goes unnoticed. T ha t, of was before treatment? Or does it produce a steel three
times as hard as any previous steel? Or what does it do?
It appea rs that the reporter has passed along some words
without inquiring what th ey mean, and you are expected
to read them just as uncritically for th e h app y illusion
th ey give you of having learned something. It is all too
,. reminiscent of an old definit ion of the lecture method of
classroom instru ctio n : a process by wh ich the contents of
the textbook of th e Instructor are transf erred to the note -
book of the student without passing thro ugh the heads of
either pa rty .
A few minu tes ago, while looking up some thing about
Dr. Kinsey in Time, I came upon another of th ose state-
men ts that collap se under a seco nd look. It appeared in
an advertisemen t by a group of electric companies in 1948.
"Today, electric power is available to more than th ree -
course , is the secret of its success. Critics of jou rnalism as
quart ers of U. S. farms ... :' T ha t sounds pretty good .
practiced toda y hav e deplored the paucity ot good old-
Those power companies are really on th e job. Of course,
fashioned leg work and spoken harshly of "Wash ington's
if you wanted to be ornery you could paraphrase it into
armchair correspo nde nts," who live by uncritically re -
"Almost one-quarter of U. S. farms do not have electric
writing government h and outs. For a sample of un enter-
power available today." Th e real gimmick , however, is in
prising journ alism take this item from a list of "new
that word "available," and by using it th e compa nies have
 now TO LIE WITH STATISTICS THE LIT'ILE FIGURES TIIAT ARE NOT THERE
~
been able to say just about anything they please. Obvi- is. Not all children grow in th e same way. Some start
ously th is does not mean that all those farmers: actually slowly and then speed up ; others shoot up quickly for a
have power, or the ad vertisement surely would have said while, then level off slowly; for still othe rs growth is a
so. They merely have it "available"-and that, for all I relatively steady process. The chart, as you might guess,
know, could mean that the power lines go past their farms is based on averages taken from a large number of meas-
or merely with in te n or a hundred. miles of them . urements. For the total, or average, heights of a hundred
youngste rs taken at random it is no doubt accurate enough,
WORLD WIDE AVAILABILITY o(How 10 Lie with Slotisli,," but a parent is interested in only one height at a time, a
purpose for which such a chart is virt ually worthless. If
you wish to know how tall your child is going to be, you
can probably make a better guess by taking a look at his

. -:
_ Areol within 25 m il., o f 0 ro ilroad, mot orobleo rood,
port o r nOW';!JlIlIbl. wo teorwo y (dog "ed routeos not ,hown}

Let me quote a title from an article published in Collier'&

tn 1952 , 'You Can Tell Now HOW TALL YO UR CHILD
WILL GROW." With the article is conspicuously dis- ., parents and grandparents. That method isn't scientific
played a pair of charts, one for boys and one for girls. and precise like th e chart, but it is at least as accura te.
showing what percen tage of his ultimate height a child I am amused to note that, ta king my height as recorded
reaches at each year of age. "To determine your child 's when I enrolled in high-school military training at four-
height at maturity," says a capti on, "check present meas- teen and ended up in the rear rank of the smallest squad.
urement against chart." I should eventually have grown to a bare five feet eight.
The funny thing about this is that the art icle itself- if I am five feet eleven. A three-inch error in human height
you read on-tells you what the fatal weakness in the chart come, down to a poor grade of guess.
 H OW T O LIE WITH STATISTICS THE LnTLE F IGTJ R ES T HA T ARE: NOT T I£.ERI.:
5" 5'
Before me ar e wrappers from t wo boxes of Grape-Nuts Such foolishn ess could be found only on material meant
Flakes . They are slightly d ifferent editions. as indicat ed for the eye of a juvenile or his morning-weary parent , of
by th eir testimon ials: one cites Two-Gun Pet e and th e course. No one would insult a big businessman's intel-
other say s. "If you want to be like Hoppy . . . you' ve got ligen ce with such sta tistical tripe , . . or would he? Let me
to ea t like Hoppyl" Both offer chart.s to show ("Scientists tell you about a graph used to advertise an ad vertising
proved it's true!") that th ese Oakes "start giving you • agency ( I hope this isn't getting con fusing ) in the rather
energy in 2 minutes!" In one case th e cha rt hidden in th ese special columns of Fortune magazine. The line on this
forests of exclama tion poi nts has numbers up th e side ; in graph showed th e impressive up wa rd trend of the agency's
th e other case th e numbers have been omitted. T his is business yea r by rear. Th ere were no numbers. With
just as well, since th ere is no hint of wha t the nu mbers eq ual honesty this chart cou ld have repr esented a tremen-
mean . Both show a steeply cllmbi ag red line ( r energy dou s growth, with business doubling or increasing by

I MJIlUn une 1 MINum nne 1923 1924 1925 1926 1921 1928 J1n9 1930 1931

release" ) , but one h as it starting one minute after eating millions of dollars a year, or th e sna il-like progress of a
Crape -Nuts Flakes, the other two minutes later. One line static conc ern adding only a dollar or two to its ann ual
climbs abou t twice as fast as the other, too, suggesting billings. It ma de a striking picture, th ough.
that even the draftsman didn't think these graphs meant Place little faith in an average or a gr aph or a trend
anythin g. when those important figures are missing. Otherwise you
 now TO LJE WITII STATISTICS

are as blind as a man choosing a camp site from a report

of mean temperature alone. You might take 61 degrees as
a comfortable annual mean. giving you a choice in Cali-
fornia betwe en such areas as the inland desert and San
Nicolas Island off th e south coast . But you can freeze or
roa st if you ignore the ra nge. For San Nicolas it is 47 to
C HA P T ER 4
S7 degrees but for th e desert it is 15 to 104.
Oklahoma City ca n claim a similar average temperature Much Ado about
for the last sixty years : 60.2 de grees. But as ),ou can see
from the chart below, that cool and comfortable figure Practically Nothing
conceals a range of 130 degrees.

Record Temperatures in Oklahoma City

1"'- 1952
~~ IF YOU don't mind, we will begin by endow'tng you with

~
two children. Peter a nd Lind a ( we might as well give
th em modish names while we're abo ut it ) have been given
( intelligence tests, as a grea t ma ny c hild ren a re in the

.
1lI1~ ,
course of their schooling . Now th e mental test of any
va riety is one of the prime voodoo fetish es of our time,
so you may ha ve to ar gue a litt le to find out th e results 01
411 H+-l-l-J,r+N-t-H th e tests; th is is inform ati on so eso teric th at it is often held
to be safe only in the hands of psycho log ists and edu cators,
201-\-H-l++t+l,\LH and they may be right at th at. Anyway , you team some-
HH,'t RecGrd tows
how that Pet er's IQ is 98 and Linda's is 101. You know,
o H.J+++++++t~
Low." of course, th at the IQ is based on 100 as average or
-,7' -20~;'-;\+!:-++H+~ "normal."
J fMAMJJASOII --'
Aha. Linda is your brighter child, She is, Iurthermore ,
53
54 HOW TO LIE wrrn ST~TISTICS MUCH ADO ABOUT P RACTICA LLY NOT IUNG 55
above average. Peter is below average, but let's not dwell defined capacity to handle canned abstra ctions. And Peter
on th at. and Linda have been given what is generally regarded as
Any such conclusions as these are sheer nonsense. the best of the tests, the Revised Stanford-Binet, which is
Just to clear the air, let's note 6rst of all that whatever administered individually and doesn't call for any par-
an intelligence lest measures it is not quite the same thing ticular reading ability.
as we usually mean hy intelligence. It neglects such im- Now what an IQ test purport s to be is a sampling of the
portant th ings as leadership and creati ve imagination. intellect. Like any other product of the sampling method,
the IQ is a figure with a statistical error, which expresses
the precision or reliability of that figure.
Asking these test questions is something like what you
might do in estimating the quality of the corn in a field
by going about and pulling off an ear here and an ear
there at random . By the time you had stripped down and
looked at a hund red ears, say, you would have gained a
pretty good idea of what the whole field was like. Your
Informat ion would he exact enough for use in comparing
this field with another field- provided the two fields were
not very similar. If they were, you might have to look
at many more ears, rating them aUthe while by some pre-
cise standard of quality.
It takes no account of social judgment or musical or artistic How accurately your sample can be taken to represent
or other aptitudes, to say nothing of such pe rsonality the whole field is a measure that can be represented in
matters as diligence and emotional balance. On top of figures: the probable error and the standard error.
that, the tests most often given in schools are the quick- Suppose that you had the task of measuring the size of a
and-cheap group kind that depend a good deal upon good many fields by pacing off the fence lines. The first
reading facility; bright or not, the poor reader hasn't a thing you might do is check the accuracy of your measur-
chance. ing system by pacing off what you took to be a hundred
Let's say that we have recognized all that and agree yards, doing this a number of times. You might find that
to regard the I Q simply as a measure of some vaguely on the average you were off by three yards. That is, you
 now TO LIE wrra STAT1$TlCi M UCH ADO ABOUT PRACTICALLY NOTIIINC

as likely that it is above or below that figure. Similarly

Lind a's has no be tter than a fifty-fifty probability of being I
within the range of 98 to 104. From this you can quickly
see that th ere is one chance in four that Peter's IQ is really
above 101 and a similar chance that Linda's is below 96.
Then Peter is not inferior hut supe rior, and by a margin of
( anywh ere from three points up ,
y
IOO Y ""' RO ~
What this comes do wn to is that the only way to think
came withi n three yards of hitting th e exact one hundred about 1Qs and many other sampling results is in ranges.
in half your trials, and in the other half of them you missed "Norm al" is not 100, but the range of 90 to no, say, and
by more tha n th ree yards. there would be some point in comparing a ch ild in this
Your probable error then wo uld be three yards in one ran ge with a child in a lower or higher ran ge. But COIl).-
hun dred , or three per cent . F rom then on, each fence line parisons between figures with sma ll differences are mean-
that measured one hundred yards by your pacing might ingless. You must always keep that plus-or-minus in mind,
be rE'COrded as 100 ± 3 yards. even (or especially ) when it is not stated.
(1-10st statisticians now prefer to use an other, but com- Ign oring these errors, which are implicit in all sampling
parable, measurement called the standard error. It takes studies, has led to some remarkably silly behavior. There
in about two-th irds of the cases instead of exactly half and are magazine edi tors to w hom readership surveys are
is conside rably hand ier in a mathemat ical way. For our gospel. mainly because th ey do oot understand them,
purpo ses we can stick to the prob able error, whi~b is the With forty pe r cent male readership reported for one
one still used in connection with the Stenford-B inet. } article and only thirty-five per cent for another, th ey
As with our hypothetica l pacing, the probable erro r of deman d more art icles like the first .
the Stanford-Binet IQ has been found to be three per Th e difference between thirty-five and fort y per cent
cent. This has nothing to do with how good the test is readership can he of impo rta nce to a magazine, but a
basically, only with how consistently it measures what- survey d ifference may not be a real one. Costs often hold
ever it measures. So Pet er's indicat ed IQ might be more readership samples down to a few hundred persons, par-
full y expressed as 9S = 3 and Lin da's as 101 :!: 3. ticularly after those who do not read the magazine at all
This says th at there is no more than an even chance that have bee n eliminated. For a mag azine that appeals
Peter's IQ falls anywhere be tween 95 and 101; it is just
\ primarily to women the number of men in the sample may
I
 now TO LIE WITH STATlS'I"ICS Mt:CH ADO ABOUT PRAcrICALLY NOTHING
~
be very small By the time th ese hav e been divided among But somebod y spotted someth ing. In the lists of almost
those who say they "read all," "read most," "read some," identical amount s of poisons, One cigarette had to be at
or "didn't read" the article in question, the t hirty-five per the bottom, and the one was Old Gold. Out went the
cent conclusion may be based on only a handful. The telegrams, and big advertisements appeared in news-
probable error hidd en beh ind the impressively presented papers at once in the biggest type at hand. The headlines
figure may be so lar ge that the editor who relies on it is an.d the copy .simply said that of all cigarette s tested by
grasping at a thin stra w. thls great national magazine Old Gold had the least of
Sometimes the big ado is made about a difference that these undesirab le thi ngs in its smoke. Excluded were all
is mathematically real and demonstrable but so tiny as to figures and any hint that the difference was negligible.
have no importance. This is in defiance of the fine old
saying that a difference is a difference only if it makes a
i
~-
In the end, the Old Gold people were ordered to "cease
and desist" from such misleading advertising . That didn't
di fference. A case in poin t is the hullabaloo over p rae· t make any difference; the good had been milked from the
tically nothing that was raised so effectively, and so profit.
ably. by the Old CoM cigarette people. I idea long before , As the New Yorker says. therell always
be an ad man.
It started innocently with the edit or of the Reader's
Digest, who smokes cigarettes but takes a dim view of
,!(
them all th e same. His magazine went to work and had
a battery of laboratory folk analyze the smoke from sev-
eral brands of cigarettes. Th e magazine published the
results, giving the nicotine and whatnot content of the
smoke by brands. The conclusion sta ted by the magazine
and borne out in its deta iled figures was th at all th e brands
were Virtually identical and that i t d idn't make any dlf-
ference which one you smoked. I
Now you migh t think this was a blow to cigarette I
manufacturers and to the fellows who think up the new I
copy angles in the advertising agencies. It would seem
to explode all advertising claims about soothing throats
I - -_ _!V

and kindness to T-zones.

 ~',, THE CEE-Wmz CRAPH 61

,
,.
something practically every body is interested in showing
or knowin g about or spotting or d eploring or forecasting .
,
Well let our graph show how nat ional income increased
ten pe r ce nt in a year.
Begin wltli pape r ruled into squ are s. Name th e months
CHAPTER 5 along the bottom. Indicate b illions of dollars up th e side .
Plot your points an d draw your line, an d your graph will
look like this :
tl±tlt:t:1The Gee - Whiz Graph
I ./,t l ~
..~ 1\~I.\-1lr ~~
.
~

.s
''''
,6
14-
~
,f
--< ~
THERE is terror in numbers. Humpty Dumpty's confidence ~ , ~. ,
12- '.
in telling Alice that he was master of the words he used C
0 10
'" ,

would not be extended by many people to numbers. Per-

- a
'\ j;;>,
haps we suHer from a trauma induced by grade-school
'" 6
I
arithmetic. ,,
Whatever the cause, it creates a real problem for the -+
writer who yeams to be re ad , the advertising man who ~
2.
expects his copy to sell goods, the publisher who wants
his books or magazines to be popular. Wh en num bers
o
J FM A IIl J
""
J A S O N D
in tabular Iorm are taboo and words will not do the work
Now th at' s clear enough. It shows what happened
well. as is often the case, there is one answer left: Draw
d uring th e year and it shows it month by month . H e who
a picture. runs may sec and und ersta nd . because th e whole graph
About th e simplest kind of sta tistical pictu re, or gra ph,
is in proportion and there is a zero line at the bottom for
is the line variety. It is very useful for showing trends,

60
 H OW TO LIE WIn I S'TATIS'I1CS THE GEE - WHIZ GRAPH

comparison . Your ten pe r cent looks like ten pe r cent-an look. Simply cha nge the proport ion be tween th e ordinate
upward trend th at is substantial but perhap s not ove r- and th e absc issa. T here's no rule against it, and it does
whelming. give your graph a prettier shape. All you ha ve to do is let
That is very well if all you want to do is convey infonna- each mark up the side stand for only one-tenth as many
tion. But sl:ppose you wtsh to win an argument, shock a J oUars as be fore.
reader, move him into action, sell him something, For
:2'% .0
that , this chart lacks schmaltz. Chop off the bott om.
2. ' .8

;g!2~
zz ~
o
c • 21 ·6

J ;2.0 --0
~
21.+
~ 16 ' ,
I JFMAM JJ "' SON D a 21·1
C
I Now that 's more like it . (You've saved pape r too, some-
-
0 2.1.0

I thing to point out if any carping fellow objects to your

2 0.8

I misleading graphics.) The figu res are the same and so is

the curve. It is the same graph , Nothing has been falsi-
'" 2 0.6

Ij fied- except the impression that it gives. But what the

2M
hasty reader sees now is a national-Income line that bas
climbed halfway up the paper in twelve months, all be- 2a t
"
"
"
i:
I cau se most of the cha rt isn't th ere any more. Like the miss-
i I I It
':/'
20. 0
I ing pa rts of sp eech in sente nces tha t you met in grammar J F M A M J J A 5 0 N 0

I classes, it is "understood ." Of course, the eye doesn't "un- That is imp ressive, isn't it? Anyone looking at it can just
derstand " what isn't there. and a small rise has become. feel prosperity th robbing in the arteries of the country.
Visually, a big one. It is a subtle r equivalent of ed iting "National income rose
Now that )'ou have practiced to deceive, why stop with ten per cent" into ... . . climbed a whopp ing ten per cent."
truncat ing? You have a further trick available t hat 's worth It is vastly more effective. however. be cause it contains
a dozen of th at . It will make your mod est rise of ten per no ad ject ives or ad ver bs to spoil the illusion of objectivity.
cen t look livelier th an one hundred per cent is entitled to There's nothi ng anyone can pin on you.
 HOW TO LIE wrru STA TISTICS TlU: CEE-WHr.l c n aen

And you're in good. or at least respecta ble, company. honesl red line th at rose just four pe r cent , und er this
Newsweek magazin e used thi s method to show that cap tion : GO VERNME NT PAY ROLLS STABLE .
"Stocks H it a 21-Year Hig h" in 1951, truncating the gra ph
at the eighty mark. A Columb ia Gas System ad vertise-
ment in Time in 1952 reproduced a chart "from our new
., .,
Annual Report." If you read th e lit tle numbers and an-
alyzed th em you found that d uring a ten-year period
I I I I I
'!G.!~~,!;.o_ l _ l _ l. I
,
I
I I
II
I
I
I
I
living costs wen t up abo ut sixty per cent and the cost of I I
+-_±_.J.,
I I I

gas dropped four per cent . Th is is a favor able picture, I iiiI :

• iF ...J- - -l-- -
iiI i
bu t it appar ently was not favora ble enough for Columbia II I, ' I I i!
I I : : I
Gas. They chopped off their cha rt at ninety per cent
( with no gap or other ind icatio n to warn you) so th at thi s
I, I, :I
I
,--+-+-+-+-+--
, , ,
,

I
I ,

.
•

I
I I I
$19.5:0.000
was what your eye told you : Living costs have more th an --~
:.I I,, II ,
tripled, and gas ha s gone down one-third! I , , •

Steel com panies have used Similarly misleading graphic

met hods in att empts to line up public opinion aga inst
wage increases. Yet the method is far from new, and its
Collier's has used this same treatmen t with a bar chart
impropriety was shown up long ago-nol [ust in techni cal
in nt"wspaper advertisements. Not e especially that th e
pu blications for stati sticians either. An edi torial write r
middle of the chart has bee n cut out:
in Dun's Review in 1938 reprodu ced a chart from an
ad vertisement advocating ad vertising in Wash ington . 3,205,000 .
D. C., th e arg ument being nicely expressed in the head- 3,200,000 ,.-- - - - - -
line over th e chart : GOVE RI\'ME NT PAY ROLLS UP l
T he line in the graph went along with th e exclamation 3.150,000 1-- - -
point even tho ugh th e figures behi nd It d id not. What they
showed was an increase from a bout $19,500,000 to $20.·
200,000. But tbe red line shot from near the bott om of th e
graph clear to th e lap. makin g an increase of under four
o fWol ___
~ ..... f ... _ _
,~,

per cent look like more than 400. The magazine gave its "" ""
From an April 24, 1953, new s-
own graphic version of the same figures alongside-an paper advertisement for C OLJ. IFJI·S
 THE ONE-DIMENSIONAL PIC'I'tlRE

ordinary bar chart, a Simple and popular method of repre-

senting quantities when two or more are to be com pa red.
A bar cha rt is capable of deceit too. Loo k w ith suspicion
on any versio n in which the ba rs chan ge th eir wid ths as
we ll as their lengths while repr esent ing a single factor

CHAPTER 6 o r in wh ich th ey picture three-dim ensional ob jects th e vel -

umes of which are not easy to co mpare. A trunca ted bar
chart has, and d eserves, e xactly th e same reputation as th e
The One ~ Dimensional Picture truncated line graph we have been talking about. The
habitat of th e bar chart is th e geography book, the cor-
poration statement, and th e news mag azine. This is true
also of its eye-appealing offspring.
Perh aps I wish to show a compariso n of two figures-the
average weekly wage of carpenters in the United Sta tes
and Hotundi a, let' s say. Th e sums might be $60 an d $30.
I wish to catch your eye with th is, so I am not satisfied
A DECADE or so ago you hea rd a good deal abou t the little me rely to print the numbers. I make a ba r chart . ( By
people, meaning pra ctica lly all of us. When th is began to the way, if that $60 figure doesn't sq ua re with the hu ge
sou nd too condescending, we became the common man. sum you la id out when your porch needed a new railing
Pretty soon that was forgotten too, which was probably last sunune r, rem ember th at your carpenter may not have
just as well. But the little man is still with us. He is the done as well every wee k as he did whil e wo rking for you.
character on the chart.
I And anyway I didn't 53Y wh at kind of ave rage I ha ve in
A chart on which a little man represents a million men, mind o r how I arri ved at it, so it isn't going to get you
a moneybag or stack of coins a thousand or a billion
\ anywhere to quibble. You see how eas y it is to hide behind
dollars. an outline of a stee r your be ef supply for next year, the most disreputable statistic if you do n't include any
is a pictorial graph. It is a useful device. It has wh at I am other Information with it? You probably guessed I just
afraid is known as eye-appeal. And it is capable of be- made this one up for purposes of illustration, but 111 bet
coming a fluent, devious, and success ful liar . you wo uldn't have if I'd used $59.83 inste ad. )
The daddy of the pictorial chart, or pictograph, is the
66
 uow 1'0 LIE

"'- - - - - - -
< "' - - - - - -- -
~
wrra STATISTICS

l THE ONE- DlMEN SIO S At. Plcrt1RE

dian , and th e more I can dram ati ze the d ifference between

thirty and sixty the better it will be for my argument . To
tell the truth ( which, of cour se. is wh at I am plann ing
~ "' - - - - - - - not to do ) , I wa nt you to infer some thing. to come away
••< with an exaggera ted impression , but I don 't want to be
« • caught at my tricks. Th ere is a way. and it is one that
•• is bein g used every d ay to fool you.
•
o I simply draw a moneyb ag to represent the Hotun dian's
o
thirt y dollars, and then I draw another on e twice as tall
to represent the American's sixty, T hat' s in proportion,
Th ere it is, with dollars-pe r-wee k indicated up the left isn't it?
side. It is a clear an d honest picture . T wice as much
mon ey is twice as big on th e chart and looks it .
Th e chart lacks th at eye-appeal though, doesn't it? I
can easily sup ply th at by using something that looks more
like money than a bar does: moneyba gs. One moneybag l

for the unfortunat e Hotundian's pittance, two for th e Now that gives the impression I'm after. The American's
American's wage. Or three for the Hot und lan, six for the wage now d warfs the foreign er's.
American . Eith er way. the chart remains honest and
\ The catch, of course. is this. Because th e second bag
clear, and it will not deceive your hasty glance. T hat is is twice as high as the first, it is also twice as wide. It
the wav an honest pictograph is made. occupies not twice but four tim es as much are a on th e
That would sati sfy me if all I wanted was to communi- page. The numbers still say two to One, but the visual
cate Informati on . But I want more. I want to say th at th e impression, which is the dominating one most of the time,
American workingman is vastly better off than the Hotun- says the ratio is four to one. Or worse. Since th ese are
 HO W TO I..Jl: w rm STATISIlCS
THE ONE-DIMENSIONAL PICIVRE
7'
pictures of objec ts havin g in reality three dimension s, the
second must also be tw ice as th ick as th e first. As YO UT STEEL CAPACITY ADDED
geometry book put it, the volumes of similar solids vary
as the cube of any like dimension. T\\'O limes two times
two is eight. Jf one moneybag holds $30. the other. having
eight tunes the volume, must hold not $60 but $240.
And that ind eed is the imp ression my ingenious little
chart gives. While sayin g "twice," I have left th e lasting
impression of an overwhelming eight-to-one rat io.
You'll have trouble pinning any criminal intent On me,
too. I am only doing what a grea t many other peop le do.
Newsw eek magazine ha s do ne it-With money bags at th at
The Amer ican Iron and Steel Institut e has done it , with Adapted by courtesy of S n:ELWA YS.

a pair of blast furnaces. T h e idea was to show how th e neigh bor, and a black bar. suggesting molten iron, had
tndustrv's stee lmaking ca pad ty had boomed betwee n the beco me two and one-half times as lon g as in th e earlier
19305 and th e 19405 and so indi cat e tha t the industry was decade. Here was a 50 per cent increase give n. th en
doing such a job on its own hook that any governmental dr awn as 150 per cent to give a visual impression of-
int erference was uncalled for . There is more merit in the un less my slide rule and I are getting out of th eir depth
principle than in t he way it was presented. The blast -over 1500 per cent. Arithmetic becomes fantasy.
furnace representing the ten-million-to n capacity ad ded in ( It is almost too unkind to mention th at j he same glossy
the '305 was drawn just over two-th irds as tall as the one four-eolor page offers a fatr-to-pnma speci men of the
(or th e fourteen and a qu arter million tons added In the truncat ed line graph. A curve exaggerates the per-cap ita
'40s. The eye saw two fur naces. one of them d ose to growth of steelma king capacity by get ting along with the
three times as big as th e ot her. To say "almost one and lower half of its graph missing. Th is saves paper and
one-half" and to be heard as "three't-cthat's wh at the one- doubl es the rat e of climb.)
d imensional picture can accomplish. Some of thrs may be no more than sloppy draftsmanship .
This piece of art work by the steel people had some But it is rather like be ing short-cha nged : When all the
other points of interest . Somehow the second furnace had mistak es are in the cas hier's favor, yo u can't help wonder-
fattened out horiz ontally beyond the proportion of its ing.
l' now TO LIE WITH STATISTICS
73
Newsweek once showed how "U. S. Old Folks Grow
Olde r" by mean s of a char t on which appeared two male THE DIMINISHING RHI NOCEROS
figures, one representin g th e 68.2-year life expectancy of
today, the other the Sd-year life expecta ncy of 1879·1889.
It was the same old story : On e figure was twice as tall as
the other and so wo uld have had eight times the bulk or
weight. This picture sensationalized facts in orde r to make
a bette r story. I would call it a Iorm of yellow journalism.
The same issue of th e magazin e contained a trun cated , or
gee-whiz, line graph.

THE CRESCIVE COW

W1860
There is still another kind of da nger in varying the size
of objects in a chart . It seems th at in 1860 ther e were
something over eight million milk eOW5 in the United
States and by 1936 there were more tha n twenty-five
m illion Showing this increase by draw ing two cows,
one th ree times th e hei ght of the oth er, will exaggerate
th e impr esstc n in the ma nne r we have been discussing. Apply the same decepti ve technique to what has hap-
But th e effect on th e hasty scanner of th e page may be even pened to the rhinoceros populati on and th is is what you
stranger : He may easily come away with th e idea that get. Ogden Nash once rhymed rhln osterous with prepos-
cows are bigger now th an th ey used to be. terous. That's the word for th e method too .
 TIlE SEM IATIACIfEI) n c UR£

It is not up to you-is it?- to po int ou t th at an antisepti c

that works well in a test tube may not perform in the
hu man throat, especially after it has been diluted ac-
cord ing to instructions to keep it from huming throat
tissue. Do n't confuse the issue by tellin g wha t kind of
CHAPTER 7 germ you killed . Who knows what germ causes colds.
part icularly since it proba bly isn't a germ at all?
In fac t, there is no known connection between assorted
The Semiattached germs in a te st tube and the whatever-it-is that prod uces
colds, but people aren't going to reason that sharply,
Figure especially while snilIling.
Maybe that one is too obvious, and people are be ginning
to catch on, although it would not appea r so from the
ad vertising pages. Anyway, here is a trickier version.
Let us say that during a period in which race pre judice
IF YOU can 't prove wha t you wa nt to prove, demonstrate is growing you are employed to "prove" otherw ise. It is
something else and pretend th at they are the same thing. Dot a difficult assignment. Set up a poll or, better yet , have
In the daz e that follows the collision of statistics with the th e polling done for you by an organization of good
human mind , hardl y anybody will notice the difference. rep utat ion. Ask th at usual cross section of the populat ion
The semiattach ed figure is a de vice guaran teed to stand if they thi nk Negroes have as good a cha nce as white
you in good stead. It always has. peop le to get jobs. Repe at your polling at intervals so that
You can't prove tha t your nostrum cures colds, but you you will have a trend to report,
can publish ( in large type ) a sworn laboratory report Princet on's Office of Public Opinion Resear ch tested
that h all an ounce of the stuff killed 31,108 genus in a this ques tion once. Wh at turn ed up is interesting evidence
test tube in eleven seconds. While you are abou t it. make that things, especially in opinion polls, are not always
sure th at the labora tory is reputable or has an impressive what they seem. Each person who was asked th e ques~
name. Reproduce the report in full. Photograph a doctor- lion about jobs was also asked some questions designed
typ e mod el in white clothes and put his picture alongside. to discover if he was strongly prejudiced against Negroes.
But don't mention the several gimmicks in your story. It tu rne d out that people most str ongly prejudiced were

74
 HOW TO UE WITH STATISTICS THE SEMIA'ITACl IED . · ICURE
n
most likely to ans wer Yes to the question about job op- 'brand." The figure itself may be ph ony, of course, in am
portunittes. ( It worked out that about two-thirds of those of several ways, but tha t really doesn't make any diffe:-
who were sympathetic toward Negroes did not th ink the ence. The only answer to a figure so irrelevant is "So
Negro had as good a chance at a job as a white person did, what?" With all prope r respect toward the medical
and abo ut two-thirds of those showing prejudice said tha t profession, do doctors know any more abo ut tobacco
Negroes were getting as good breaks as whit es. ] It was brands than you do? Do they have any inside information
pretty evident that from t his poll you would learn very tha t permits them to choose the least harmful among
litt le abo ut employment condit ions for Neg roes, although cigarettes? Of course they don't, and you r doctor would
you might learn some inte resting things about a man's be the first to say so. Yet that "27 per ce nt" somebow
racial attitudes . manages to sound as if it meant someth ing.
You can see, then, that if prejudice is mounting during Now slip back one per cent and co nsider the case of the
your polling period you will get an increasing number of juice extractor. It was widely adver tised as a device that
answers to the effect that Negroes have as good a chance "extracts 26 per cent more juice" as "proved by laborator,
at jobs as w lut es. So you announce your results: Your test- and "vouched for by Good Housekeeping Institute....
poll shows that Negroes are getting a fairer shake all the That sounds right good. If you can buy a juicer that is
time. twenty-six per cent more effective, why buy any other
You have achieved something remarkable by careful kind? \VeU now, without going into the fact that "labora-
use of a serniatta ched Sgure. The worse things get. the to') tests" (especially "independe nt laboratory tests" )
better your poll makes them look. have proved some of the da mdest things, [u...t what does
Or take th is ones "27 per cent of a large sample of that figure mean? Twenty-six per cent more tha n what?
eminent physicians smoke Th roaties-morc than any oth er When it was finally pinned down it was found to mean
only that this juicer got out that much more juice than
an old-fashioned hand reamer could. It had absolutely
nothing to do with th e d ata you would wan t before
purchasing; this juicer might be th e poorest on the market.
Besides being suspiciously p recise, th at twenty-six per
cent figure is tot ally irrelevant.
Advertisers aren't the only people who will too! vou with
uumbcrs if you let them. An articl e on drivin~ safety,
18 HOW TO LI E WITH STAT ISTI CS
r
THE SE MIArrACHED YICURE 79
published by Th is W eek magazine undoubtedly with your
foggy weather. More accidents occu r in clea r we ather,
best Interests at heart, told you what might happen to you
because there is more clear weather than foggy wea th er .
if you went "hur tling down thc highwa y at 70 miles all
All the same, fog may be much more dangerous to dr ive
hour, careening from side to side." You would have, t h e in.
art icle sa id, four tim es as good a chance of staying alive
You ca n use accident sta tistics to scare yourself to death
if th e time were se ven in th e morning th an if i t wen: seven
in connec t ion with any kmd of tran sportat ion . .. if you
at night. The evidence : " Four times more fata lities occur fail to note how poorly att ached the figures are.
on th e highways at 7 P. M . than at 7 A .M : ' Now that is
More peo ple were killed by airplanes last )'ear tha n in
a pproxim ately true, but the conclusion doesn't follow. 19 10. T herefore modem planes are more da ngerous?
More people are killed in the evening than in th e morning Nonsense. T her e are hundreds of tim es more people Bying
simply beca use more people are on the highways then to now. that's all.
be killed . You, a single driver, may be in greater da nge r It was repo rted that the number of deaths cha rgea ble
in the evening, but there is nothing in the figures to prove to steam railroads in one rec ent year was 4,712. That
it either way. soun ds like a good argument for sta ying off trains , perhaps
By th e same kind of nonsense th at th e article writer used for sticking to your automobile instead . But whe n you
you can shew that clear weath er is more dangerous than investigate to find what th e figure is all abo ut. you learn
it means something quit e d ifferent. Nearly half those
victims were people whose automobiles collided with
trains at crossings . The grea te r part of the rest were riding
the rods. Only 132 out of the 4,712 were passengers on
trains . And even th at figure is worth little for purposes of
comparison unless it is attached to informatio n on tot al
p assenger miles.
If you are worried about your ch ances of be ing killed.
on a coast-to-coast trip, you won 't get much releva nt
information bv asking wh eth er trains, plan es, or cars
s "-
-- ~ killed the greatest numb er of peop le last year . Get the
rate , by inquiring into the number of fata lities for each
million passenger miles, That will come closest to telling
80 HOW TO LIE WITH STATIS T ICS TIlE SE MIATT ACHED FI CURE s,
you where your great est risk lies. latrine. To replace it. the company has to spend 30 cents. Ju~t
There are many other fonns of coun ting up someth ing like that, there is th e profit on 20 sales dollars. Makes a man
want to go easy on the paper towels.
and then reporting it as something else. Th e general
But, of course, the truth is, what the company reports as
method is to ptck two thi ngs tha t sound the same but are pro6ts is only a haH or a third of the pro6ts. The part that isn't
not . As pe rsonnel manager for a company th at is scrapping reported is hidden in depreciation, and special depreciation.
wi th a un ion you "make a su rvey" of employees to find and in reserves for contingencies.
ou t how many have a complaint against the union . Unless
th e union is a ba nd of angels with an archangel at th eir Equally gay fun is to be ha d with percentages. Fo r a
head ),ou can ask and record wi th pe rfect honesty and recent nine-month period General Motors was ab le to
come out with proof that the greater part of the men do report a relatively mod est profit ( after taxes ) of 12.6 per
have some complaint Or other. You issue your informa tion cent on sales. But for that same period C M's p rofit on its
as a report that "a vast majority-78 per cent-are opposed investment came to 44.8 per cen t. which sounds a good
to the union ," What you have done is to add up a bu nch of deal worse-or bett er. depe ndin g on what kind of argu-
undifferentiated complaints an d tiny gripes and then call ment you are trying to win.
them someth ing else that sounds like th e same thing. You Similarly, a reader of Harper', magazine came to th e
ha ven't proved a thi ng, bu t it rather sounds as if you defense of the A &: P stores in that ma gazine's lett ers
have, doesn't it? column by pointin g to low net earn ings of on ly 1.1 pe r cent
It is fair enough, thou gh , in a way. The unio n ca n just on sales. He asked. "would any American citizen fear
as read ily "prove " th at practically all the wo rkers object pu blic condemnation as a profiteer ° ° for realizing II little
0

to the way th e plant is being run . over S10 for e\Oery $1,000 invested d uring a yea r?"
If you' d like to go on a hunt for semia ttached figures, Offhand th is 1.1 per cent sounds almost distressingJy
you might try running th rough corpora tion financial state. small Compare it with th e four to six per cent or more
merits. Watch for profits th at might look too big and so interest that most of us are familiar with from FHA
are concea led under another name. The Unite d Auto- mo rtgages an d bank loans and such . Would n't the A &: P
mobi le Workers' magazine Ammunition describes th e be better off if it went out of th e grocery business and
device this way: put its capital into th e bank and lived off inte rest?
The catch is that annual return on investment is not the
The statement says, last year the company made $35 million same kettle of fish as earnings on total sales. As another
in profits. Just one and a half cents out of every sales dollar.
You feel sony for the company. A bulb bums out in the reader replied in a later issue of llarper's, "If I purchase
 H O'W TO L IE WITH STATI STICS T HE SEMlATTACiJED n CURE

an article every mornin g for 99 cents and sell it each and apparently important change th at took place in just
aft ernoon for one dollar, I will make only 1 per cent on a few years. But all that has happe ned in actuality is tha t
total sales, but 365 per cent on invested money durin g th e cases are now recorded only when proved to be ma laria,
year." where formerly the word was used in much of th e South
There are often many ways of expressing any figure. as a colloquialism for a cold or chill.
You can, for instance, express exactly the same fact by The death ra te in th e Navy d uring the Spanish-Ameri-
calling it a one per cent return on sales, a fifteen per cent can War was nine per thousand. For civilians in New York
ret urn on investment , a ten-million-dollar profit, an in- City during the same period it was sixtee n per thousand.
crease in profits of forty per cent (compared with 1935- Navy recruiters later used these figures to show that it was
39 average ), or a decrease of sixty per cent from last year. safer to be in th e Navy tha n out of it. Assume these figures
The method is to choose the one tha t sounds best for the to be accu rate, as they probably are. Stop for a moment
purpose at hand and trust that few who read it will and see if you can spot what makes them, or at least the
recognize how imperfectly it reflects the situation . conclusion the recruiting people dr ew from them, virtuall y
Not all semiattached figures are products of intentional meanin gless.
deception. Many statistics, including medical ones that
are pretty importan t to everybody, are distorted by in-
consistent reporting at the source. There are startlingly
contradictory figures on such delicat e matters as abortions.
iUegitimate birth s, and syphilis. If you should look up the
latest available figures on influenza and pneumonia, you
might come to the strange conclusion that these ailments
are practically confi ned to three southern sta tes, which
account for about eighty per cent of the reported cases. The groups are not comparable. The Navy is made
What ac tually explains this percentage is the fact that up mainly of young men in known good hea lth. A civilian
th ese th ree states required reporting of the ailments after population includes infants, the old, and the ill, all of
other states had stopped doing so. whom have a higher death rat e wherever they are. These
Some malaria figures mean as little. Where before 1940 figures do not at all prove that men meeting Navy stand-
there were hundreds of th ousands of cases a year in the ards will live longer in the Navy than out. They do not
American South there are now only a handful, a salubrious prove the contrary either.
 THE SE MlATrACHED FIGURE
HOW TO LIE WITH STATISTICS

You may have heard the d iscoura ging news that 1952 appear to be attached to each other but are not r
was the worst polio year in medical history. This conclu- When Dewey was elected Governor in 1942, the minimum
sion was bas ed on what might seem all the evidence any- teacher's salary in some districts was as low as $900 a year.
on e could ask for : There were far more cases reported in Tod ay th e school teachers in New York Sta te enjoy th e high-
that year than ever before. est salaries in the world. Upon Governor Dewey's recommen-
da tion. based on the fi ndinF:s of a Committee he appointed,
But when expe rts went back of the se figures they found th e Legislature in 1947 appropriated .s,'32,OOO.OOO out of a sta te
a few th ings th at were more encouraging. One was tha t surplus to provide an immediate increase in the salaries of
there were so many ch ild ren at the most susceptible ages school tea chers . As a result the minimum salaries of teachers
in 1952 that cases were bound to be at a record numbel in New York City ran~e from $2.500 to $5,325.
if th e rat e remained level. Another was that a gen eral It is entirely possible that Mr. Dewey ha s proved him-
consciousness of polio was leading to more frequ ent diag self th e teacher's friend, but these figures don't show it.
nosis and recording of mild cases. Finally, there was at It is the old before-and-after trick, with a number of un-
increased financial incentive, there being more polio in- mentioned factors introdu ced and made to appear what
surance and more aid available from the National Founda- they are not. Here you have a "be fore" of $900 and ao
tion for Infantile Paralysis. All this threw considerable "after" of $2,500 to $5,325, which sounds like an improve-
doubt on the notion that polio had reached a new h igh , ment indeed . But the small figure is the lowest salary io
and the tot al num ber of de aths confirmed the doubt . any rural d istrict of the state, an d the b ig one is th e range
It is an interesting fact th at the dea th ra te or number in New ¥OTk City alone. There may have been an im-
of deaths ofte n is a better measur e of the incidence of an provement under Governor Dewey, and there may not.
ailment than direct incidence figures - simply because the This statement illustr at es a sta tistical form of th e be fore.
qu ality of reporting and record-keeping is so mu ch higher and-after photograph th at is a fam iliar stunt in magazines
on fatalities. In this insta nce, the obviously semiattached and advertising. A living room is photograp hed twice to
figure is better than the one that on the face of it seems show you what a vast impro vement a coat of paint can
fully attached. make. But between the two exposures new fu rnitur e has
In America the semiatta ched figure enjoys a big boom been added, and sometimes the "before" picture is a tiny
every fourth year Thi s ind icates not that th e figure is one in poorly lighted black-and-white and th e "after"
cyclical in nature, but only that cam paign tim e has ar- version is a big photograph in full color. Or a pair of pic~
rived. A campaign statement issued by the Rep ublican tures shows you what happened when a girl began to use
party in October of 1948 is built entirely on figures that

t
B6 now TO LIE WITH STATlSTICS

a hair rinse. By golly, she do es look better afterwards at

that. But most of th e change, you note on careful inspec -
tion, has been wrought by persuading he r t.o smile and
throwing a back light on her hair. More credit belongs
'.
to the photographer than to the rinse,
," 1:&
'-g'1'."'_. 8
.{i : a !\\ Post Hoc Rides Again
I. ~ ate two af.cd,.J,WhicM,"f'f+1 tim4.
GJJJkll a~ f.IJ f.h.e Iwv:t<"f.d1fkl,.
;D,d,i :'w&7"f.IJ~Wte?

S OMEBODY once went to a good deal of trouble to fin d out

if cigarette smokers make lower college grades than non-
smokers. It turned out that th ey did. This pleased a good
many people and they have been making much of it ever
since. The road to good grades, it ,'vould appear, lies in
givin g up smokin g ; and. to carry the conclusion one
reasonable step further, smoking makes dull minds.
This particular study was, I believe. prop erly done:
sample big enough and honestly and carefully chosen,
correlation having a high significance, and so on.
The fallacy is an ancient one which, however, has a
powerful tend ency to crop up in statistical material, where
it is disguised by a welter of impressive figu res. It is the
one that says that if B follows A, then A has caused B.
87

I
 now TO LIE W ITH STATISTI CS POST HOC R IDES AGAIN'

apparently than the one between grades and intelligence?

Maybe extroverts smoke more than intr overts. The point
is that when there are many reason able explanations you
are hardly entitled to pick one that suits your taste and
insist on it. But many people do.
To avoid falling for the post hoc fallacy and thus wind
,, up believing many things that are Dot so, you need to put
,,• any statement of relationship through a sha rp inspection.
,,
I
The correlation, that con vincingly precise figure that seems
to prove that something is because of something, can ac-
tually be any of several types.
One is the correlat ion prod uced by chance. You may
be able to get together a set of figu res to prove some un-
likely thing in this way. but if you try again, your nen
AD unwarranted assumption is be ing .made that since set may not prove it at all. As with the ma nufac tur er of
smoking and low grades go together. smoking causes low the tooth paste that appeared to red uce decay, you shnply
grades. Couldn't it just as we ll be the other way around? thr owaway the results ) 'OU don't want and p ublish widely
Perhaps low marks drive stude nts not to drink but to to- those you do. Civen a small sa mple, you are likely to find
bacco. When it comes right down to it, this conclusion is some substantial correlation between any pair of charac-
about as likely as the other and just as well supported by teristics or events that you can think of.
th e evide nce. But it is not nearly so satisfactory to propa- A common kind of co-variation is ODe in which the re-
gandists. lationship is real hut it is not possible to be sure which of
It seem s a good dea l more probable, however, that the varia bles is the cause and which th e effect. ] 0 some
neither of these things has produced the other, but both of these instances cau se and effect may change places
are a product of some thi rd factor. Can it be th at the from time to time or ind eed both may be cause and effect
sociable sort of fellow who takes his books less than seri- at the same tim e. A corre lation between income and
ously is also likely to smoke more? Or is there a clue in ownership of stocks migh t be of that kind. The more
the fact that somebody once established a corre lat ion be- money you make , the more stock you buy, and the more
tween ertrcversion and low grades-a closer relationship stock you buy, the more income you get; it is not accurate
 HOW TO LIE WITH STATISTICS POST HOC RlDF..$ AGAIN 9'
to say simply that one has produced the other.
Perh aps the trickiest of th em all is the very common
instance in which neither of the variables has any effect
at aU on the other, yet th ere is a rea] correlation. A good
deal of dir ty work has been done with this one. The poor
grades among cigarette smokers is in this category, as are
all too many medical statistics that are quoted without
the qualificatio n that although the relationsh ip has been
shown to be real, the ca use-and-effect nature of it is only rosier in the spring gives up when June comes and he still
a matt er of speculation. As an instance of the nonsense feels terrible.
or spurious correlation that is a real statistical fact, some- Another thi ng to wat ch out for is a conclusion in which
one has gleefully pointed to this : There is a close relation- a corre lation has been inferred to continue beyond the
ship between the salaries of Presbyterian ministers in dat a with which it has been demonstrated. It is easy to
Massachusetts and the price of rum in Havana. show th at the more it rains in an area, the taller the corn
Which is the cause and which the effect? In other grows or even the grea ter the crop. Rain, it seems, is a
words, are the ministers benefiting from the rum trade or blessing. But a season of very heavy rainfall may damage
supporting it? All right. That's so farfet ched tha t it is or even ruin the crop. The positive correlation holds up to
ridiculous at a glance. But watch out for other applica- a point and then quickly becomes a negative One. Above
tions of post hoc logic that di ffer from this one only in be- so-many inches. the more it rains the less com you get.
ing more sub tle. In the case of the ministers and the rum we're going to pay a little atten tion to the evidence O D
it is easy to see that both figures are growing because of the money value of education in a minute. But for now
the influence of a th ird factor: the histone and 'wo rld -wide let's assume it has been proved th at high-school graduates
rise in the price level of pr actically everyt hing . make more money than those who d rop out, that each
And take the figures that show the suicide rat e to be year of undergraduate work in college adds some more in-
at its maximum in Ju ne. Do suicides produce June brides come. Watch out for the general conclusion th at the more
-cor do June wedd ings precipitate suicides of the jilted? A you go to school the more money you'll make. Note that
somewhat more convincing (though equally unproved) this has not been shown to be tru e for th e years beyond
explanation is that the fellow who licks his depression all an undergraduate degree, and it may very well not apply
through th e winter with the thought that things will look to them either. People with Ph.D.s quite often become
I.

I
I
9' HOW TO LIE WITH STATISTICS POST HOC RIDES AGAIN

college teachers and so do not become members of the may easily tum out to be the financial ruination of Joe over
highest income groups. there. Keep in mind that a correlation may be real and
A correlation of course shows a tendency which is not based on real cause and effect-and still be almost worth-
often the ideal relationship described as one-to-one. Tall less in determining action in any single case.
boys weigh more than short boys on the average, so this Reams of pages of figures have been collected to show
is a positive correlation. But you can easily find a six- the value in dollars of a college education, and stacks of
~j
footer who weighs less than some five-footers, so the cor- pamphlets have been published to bring these figures-
relation is less than 1. A negative correlation is simply a !J and conclusions more or less based on them-to the atten-
statement that as one variable increases the other tends tion of potential students. I am not quarreling with the
to decrease. In physics this becomes an inverse ratio: intention. I am in favor of education myself, particularly
The further you get from a light bulb the less light there if it includes a course in elementary statistics. Now these
is on your book; as distance increases light intensity de- figures have pretty conclusively demonstrated that people
who have gone to college make more money than people
who have not. The exceptions are numerous, of course,
but the tendency is strong and clear.
The only thing wrong is that along with the figures and
facts goes a totally unwarranted conclusion. This is the
post hoc fallacy at its best. It says that these figures show
that if you (your son, your daughter) attend college you
will probably earn more money than if you decide to
spend the next four years in some other manner. This un-
warranted conclusion has for its basis the equally unwar-
ranted assumption that since college-trained folks make
more money, they make it because they went to college.
Actually we don't know but that these are the people who
creases. These physical relationships often have the kind- would have made more money even if they had not gone
ness to produce perfect correlations, but figures from to college. There are a couple of things that indicate
business or sociology or medicine seldom work out so rather strongly that this is so. Colleges get a dispropor-
neatly. Even if education generally increases incomes it , tionate number of two groups of kids: the bright and the

I
I

:l
 now TO LIE WITH STATISTICS

rich. The bright might show good earning power without

college knowled ge. And as for the rich ones . .. well,
I1 POST HOC RIDES AGAIN

not, although a hasty reader may come away with the idea
th at they are.
95

money breeds money in several obvious ways. Few sons Here again a real correlation has bee n used to bolste r up
of rich men are found in low-income brackets whether an unproved cause-and -effect relationship. Perhap s it all
the y go to college or not. I works the othe r way aro und and those women would have
The following passage is taken from an article in qu es-
tion-and-answer form that a ppeared in This W eek ma ga- ! rem ain ed unmarried even if th ey had not gone to college.
Possibly even more 'would have failed to marry . If th ese
zine, a Sunday supplement of enormo us circulation. possibilities are no better th an th e one th e writer insists
Maybe you will find it am using , as I do , that th e same upon, th ey are pe rhaps just as val id conclusions: th at is,
writer once prod uced a piece called "Popular Notions : guesses,
True or Falsej" Ind eed th ere is one piece of evide nce suggesti ng that
a p rope nsity for old-maidh ood may lead to going to col-
Q: What effect does going to college have on )'our chances lege, Dr. Kinsey seems to have foun d some correlation
of remaining unmarried?
A: If you're a woman, it sk-yrockets your chances of beco m- between sexuality and education , with traits perhaps being
ing an old maid. But if you're a man. it has the opposite effect fixed at pre-college age. That m akes it all th e more qu es-
- It mintmizes )'our chances of staying a bachelor. tionab le to say tha t going to college gels in the way of
Cornell University made a study of 1,500 typical middle- marryin g.
aged college graduates. Of the men. 93 pe r cent were mar-
Note to Susie Brown . It ain't necessarily so.
tied ( com pa red to 83 per cent for the general population ).
But of the middle-aged women graduates only 65 per cent A medi cal article once po inted with great alarm to an
were married . Spinsters were relatively three times as numer- increase in cancer among milk drinkers. Cance r, it seem s,
ous among college graduates as among women of the general was becoming increasingly frequen t in New E ngland,
population.
Mtrmesota, W isconsm, and Switzerland, wher e a lot of
Wh en Susie Brown, age seventeen , reads this she learns milk is produced and consumed, whil e remain ing rare in
Ceylon . where milk is scarce. For furt her evidence it was
that if she goes to col lege she will be less likely to get a
man th an if she doesn't, That is what the article says, and pointed out that can cer was less frequ ent in some Southern
there are sta tistics from a reputable source to go with it. states where less milk was consumed . Also, it was pointed
They go with it, but th ey don't bac k it up ; and note also out, milk-drinking English women get some kind s of can--
that while th e sta tistics are Cornell's the conclusions are cer eighteen tim es as frequently as Japanese women who
seldom drink milk.
 HOW TO LIE WITII STATISTICS POST H OC RIDES AGAIN

A little diggin g might uncover quite a number of ways gating the relationship between age and some physical
to account for these figures, but one factor is enough by characteristics of women, begin by measuring the angle of
itself to show the m up. Cancer is predominantly a d isease the feet in walking. You will find th at the angle tends to
th at strikes in middle life or after. Switzerland and th e be greater among older women. You might first consider
states mentioned first are alike in having populations with wheth er this indicates that women grow older because
relatively long spans of life. English women at the time the)' toe out, and ),ou can see immediately that this is
the study was made were living an average of twelve ridiculous. So it ap~ars that age increases the angle be-
years longer than Japanese women. tween th e feet, and most women must rom e to toe out
Professor Helen M. Walker has worked out an amusing more as they grow older.
illustration of th e folly in assuming the re must be cause Any such conclusion is probably false and certai nly un-
and effect whenever two thin gs vary together. In Invesu- warranted. You could only reach it legitimately by study-
ing the same women-cor possibly equivalent groups-over
a period of time. That would eliminate the factor re-
sponsible here . Which is that the older women grew up at
a time when a young lady was taught to toe out in walk.
'J
tng, while the members of the younger group were leam-
II
mg posture in a day when that was discouraged .
When you find somebody-usuaUy an interested party
-enaking a fuss abo ut a correlatio n, look first of all to see
if it is not one of this type, produced by the stream of
events, the tre nd of the times. In our time it is casy to
show a positive correlation bet ween any pair of things like
th ese: number of students in college, num ber of inmates
in ment al institutions, consumption of cigarettes, incidence
of heart disease, use of x -rey machines, production of
false teeth, salaries of California school teachers, profits
of Nevada gambling halls. To call some one of th ese the
cause of some other is manifestly silly. But it is done
every day.

I
] HOW TO LIE WlTII STATISTICS POST HOC RIDES AGAIN
)
Permitting statistical treatment and the hypnotic pres· As we have alread y noted, scantier evidence th an th is-
ence of numbers and decimal point s to befog causal rela- treated in the statistica l mill unt il common sense could no
tionships is little better than supersti tion. And it is often longer penetrate to it- has made ma ny a med ical fortune
more serious ly misleading. It is rath er like th e conviction and m2.JlY a medi cal art icle in rnagaztnes, includ ing Pr(}-
among th e people of the New Hebrides th at body lice pro- Fessional ones. Mere sophisticated observers fi nally got
d uce good health. Observation over the centuries had things straighte ned out in the New Hebrides. As it turned
taugh t them th at people in good health usually had lice out. almost everybody in those circles had lice most of the
and sick people very ofte n d id not. The observation itself time. It was, you might say. the normal cond ition of man.
was acc ura te and sound, as observations made Informally When, howe ver, anyone took a fever ( q uite possibly car.
over the years surprisingly often are . Not so much can be ried to him by those same lice ) and his body became too
said for the conclusion to which these primitive people hot for comfortable habitation, the lice left. There you
carne from th eir evidence: Lice make a man healthy. have cause and effect altogether confusin gly distorted,
Everybody should ha ve th em . reversed, and intermingled.

II

•
 HOw TO STATlSTlCULAn; ro r

am DOt cert ain th at on e assumption will be less offensive

to statisticians th an th e other. Possibly more important to
keep in m ind is that th e d istort ion of sta tistical d ata and
its man ipulation to an end ar e not always th e wo rk of pro-
fessionaI sta tisticians. What comes full of virtue from the
CHAPTER 9 statisticia n's desk m ay 6nd itself twis ted, exaggerated ,
./
/'\ ,,, .,. oversimp lified, and distorted-through -selection by sales-
How to man, public-relat ions expert, journalist , or adv ertising
,
; :, copywr iter.
Statisticulate
,i -- But wh oever the guilty party ma y be in aoy instance, it
is ha rd to grant him th e status of blunder ing innocent .
~ ~~ False charts in magazines and newspap ers freq uently
sensatio nalize by exaggera tion, rarely minimize anything.
Those who p resent statistica l arguments On be half of in-
dustry are seldo m foun d, in my experience, giving labo r
or th e customer a bett er break t11."10 the facts call for, and
MISINFORMIN G people by the use of statistical material ofte n th ey give him a worse one , When has a union em-
might be called statistical manipulation; in a word (tho ugh ployed a statistical work er so in competent th at he made
not a very good one), statisticulation . labor's case out weaker than it was?
The Litle of this book and some of the th ings in it might As lon g as th e errors remain on e-sided, it is not easy to
seem to imply that all such operat ions are the product of attribute th em to bungling or accident.
intent to decei ve. The president of a chapter of the Ameri- One of the trickiest ways to misrepresent statistical data
can Statistical Association once called me down for that . is by means of a map. A map introduces a fine ba g of vari-
Not chicanery much of the time, said he, but incompe- ables in which facts can be concealed and relationship s dis-
tence. Ther e may be some th ing In wba t he says, " but J
It reminds me of th e ntiDistl::f who achiev ed gr l'llt popularity amoog
• Author Louis Bromfield is said to have a stock f"Ply to C%1tIeal COI"- lDOUlen iIi his rongreg.io.tion by his B.attt2"ing rommt'llts on b.hies brought
responde ntS when h is mail kromes too heavy for individua l atteDtlOli In lOt ctu1 st~r';ng. But w h<:n til e rllot hf'l'"s compareo.l notes not on e rould
remember wh at the man had ~ id. only th at it ha d bee n "liOUlet hing nice."
With out conced ing anything and without eecc uragmg furth er correspond -
Turned out his invariable rem ark 'lWU, "Myr" ( beaming) "This if a baby,
ence it still satisfies almost everyone. 'fht, key sentence : "Thne may
Isn't it!"
be ~eth.ing ill. wha t you 53.y,"
. 00
 HOW TO LIE "WITH STATISTICS HOW TO STATISTICULATE

torted. My favorite trophy in this field is "The Darkening

Shadow." It was distributed not long ago by the First
THE DARKENING SHADOW
National Bank of Boston and reproduced very widely-by
so-called taxpayers groups, newspapers, and Newsweek
magazine.
The map shows what portion of our national income is
now being taken, and spent, by the federal government,
It does this by shading the areas of the states west of the
Mississippi (excepting only Louisiana, Arkansas, and part
of Missouri) to indicate that federal spending has become
equal to the total incomes of the people of those states,
The deception lies in choosing states having large areas
but, because of sparse population, relatively small in-
comes. With equal honesty (and equal dishonesty) the
map maker might have started shading in New York or
New England and come out with a vastly smaller and less
impressive shadow. Using the same data he would have
produced quite a different impression in the mind of any-
one who looked at his map. No one would have bothered
to distribute that one, though. At least, I do not know of
any powerful group that is interested in making public
spending appear to be smaller than it is.
H the objective of the map maker had been simply to
convey information he could have done so quite easily.
He could have chosen a group of in-between states whose
total area bears the same relation to the area of the coun-
try that their total income does to the national income.
The thing that makes this map a particularly flagrant
effort to misguide is that it is not a new trick of propa-
 HOW TO UE WITll STATISTICS HOW TO STATISTICULATE l OS

ganda . It is something of a classic, or chestnut. Th e same you were pleased to learn that folks wer e doing so well,
bank long ago published versions of th is map to show but you may also have been struck by how poorly tha t
federal expenditures in 1929 and 1937, and th ese shortly figure squared with your own observations. Possibly you
cropped up in a standard book, Graphic Presentation, by know th e wr ong kind of people.
Willard Cope Brinton, as horrible examples. Th is method Now h ow in th e world can Russell Sage and th e Bureau
"d istorts the facts," said Brinton plainly. But the First of the Census be so far apart? Th e Bureau is talking in
National goes right on drawing its maps, and News- med ians, as of course it should be, but even if the Sage
week and other people who should know better-and p0s- people are using a mean the diHerence should not be
sibly do-go right on reproducing them wi th neither quite this great. Th e Russell Sage Foundat ion, it t urns
warning nor apology. out, discovered this remarkable prosperity by producing
Wh at is the average income of American families? As what can only be described as a phony family. Their
we noted earlier, for 1949 the Bureau of the Census says method, they explained (when asked for an explan ation ),
that the "income of the average family was $3,100." But was to divide the total personal income of the American
if you read a newspap er story on "philanthropi c giving" people by 149,000,000 to get an average of $1,251 for
handed out by the Russell Sage Foundation you learned each pe rson. "Wh ich," they added, «becomes $5,004 in
that, for the same year, it was a notable $5,004. Possibly a family of four."

How to Make. .t-22,500ayear(.....,')

:L Ac9u;re
13 c h ild re n .
ot leost 1(one) wife onc1
Q .c olc u lo.Te thelA .~ . per ""Pit. income .
c....,,'!o _ .-. ~ I. ~ O O re r '1 t".4.. .. Q.rrn)'lC,.)
3 . M4lt'ply by 15. C,., " x $l. 5 co.t22.5<JO>
 H OW TO UE wrrn STATISTICS HO W TO STATJST ICULATE
"'I
Thi s odd piece of sta tistical manipulation exaggerates as to declare only that people sleep 7.8 (or "almost S")
in two ways. It uses the ki nd of average caned a mean hours ~ night, there would have bee n nothi ng striking
inst ead of the smaller and more Informative median . . . about It. It would have sound ed like wha t it was a
. . ' poor
something we worked ove r in an earlier chapter. And then approximatton and no more instructive than almost any.
it goes on to assume that the income of a family is in direct body's guess.
proportion to its size. Now I have four child ren, and I
lABOR AND REST OF A PEASANT WOMAN
wi sh things were disposed in that way. but they are DOt.
Families of four are by no means commonly twice as
wealthy as families of 1\\"0.
1923
In fairness to the Russell Sage sta tisticians, who may
be presumed innocent of desire to de ceive, it should be R",' Sleep
said that they were primarily interested in making a pic. 2h.26m. 5h.43m.
ture of giving rather than of getting . The funny figure for
family incomes was just a by-product. But it sp read its
deception no less effective ly for that, and it remains a
prime example of why littl e faith can be pla ced in an nn -
1936
qualified stat ement of average. 4h.43m.
For a spurious air of precision that will lend aUkinds of
weight to the most d isreputable statistic, consider the
decimal. Ask a hundred citize ns how many hours they slept
last night . Come out with a total of. say, 783.1. Any such
da ta are far from p recise to begin with. Most people will
miss their guess by fifteen minutes or more, and there is
DO assurance that the erro rs will balance out. We all know

someone who .....i ll reca'll five sleepless minutes as half a .- - -

night of tossing insomn ia. But go ahead , d o your arith-
met ic, and announce tha t people sleep an average of 7.831
hours a night. You will sound as if you knew pr ecisely
C~rt, (ldapte~ from U.S.S.R. (Scientific
what you were talking abo ut . If you had been SO foolish PubllShlFl g lneusute of Pictorial Statistics)
 HOW TO LI)!'; WITH STATISllCi HOW TO STATlSTICVLATE
"'9
Per centag es offer a fertile f ield for confusion. And like
the ever- impressive decimal they can lend an aura of
precision to the inexact. Th e United Sta tes Department
of Labor's ,Monthly Labor Review once stat ed that of the
offers of part- time household employm ent with provisions
for carfare, in Washington, 0 , C" during a specified
month, 4,9 per cent were at $18 a wee k. T his percen tage,
it turned out, was based on precisely two cases, there hav-
ing be en only fort y-one oflers altogethe r. Any pe rcen tage
figure based on 3 small n umbe r of cases is likely to be rnis-
leading. It is more informative to give the fi gure itself.
And when the percentage is carried ou t to decimal places
you begin to run the scale from the sill)' to th e frau du lent.
Karl Marx was not above achievi ng a spurious air of
"Buy your Christmas p resent s now and save 100 per
precision in the same fashion. In figuri ng th e "rate of
cent," advises an advertisement. This sounds like an offer
surplus-value" in a mill he began wi th a splendid collection
worthy of old Santa himself, but it turns out to be merely
of assumptions, guesses. and round numbers: "we assume
a confusion of base. The redu ct ion is only fifty per cent.'
the waste to be 6$ . . , th e row mat erial costs in round
The saving is one h undred per ce nt of the red uced or new
num bers £342. The 10,000 sp ind les cost, we will as-
price, it is tru e, but that isn't wha t the offer says.
sume, £1 per sp indle, . , . The wear and tear we put at 100;.
Likewise when the president of a Bower growers' asso-
. , , Th e rent of the buildi ng we !.u ppose to be £300.. , ,"
ciation said, in a ne\\ospape r interview , that "flowe rs are
He says, "The above data, which may be relied upon, were
100 per cen t chea per than four mon ths ago," he d idn't

.,
given me by a Man chester spinner."
mean that florists were now giving them awar. But that's
From the se approximat ions Marx calculates that : "The
what he said ,
rate of surp lus-value is therefore lO I n ~ 15311 / 13 %." For a
In her Hist ory of the Standard O il Com pany, Ida M.
ten-hour day this gives him "necessary labour - 3"/ as
Tarbell went even further. She said th at "price cutting in
hours and surplus-labour .,." 62/s~ ."
the southwest ... ranged from 14 to 220 pe r cent: ' Th at
There's a nice feeling of exactness to that two thirty-
thirds of an hour, but it's all bluff.
",1,• would call for seller paying buyer a considerable sum to
haul the oily stuff away.
 r
HOW TO STATISTI CULATE m
no HOW TO LIE WITH STATISTICS

The Columbus Dispatch declared th at a manufactured

product was selling at a profit of 3,BOO per cent, basing
50%
PAY CUT
this on a cost of $1.75 and a selling price of $40. In cal- RIlSTOREO
culat ing percenta ge of profits you have a choice of
methods (and you arc obligated to indi cate wh ich you
are using ). If figured on cost, this one comes to a profit
of 2,185 per cent ; on selling price, 95.6 per cent . T be
Dispatch apparently used a meth od of its own and , as so
often seems to happ en, got an exaggerated figure to report.

so z
PAYeur with the Indianapolis Building Trades Unions were given a
5 per cent increase in wages. That gave back to the men nne-
fourth of the 20 per cent cut they took last winter.
Soumis reasonable on the face of it- but the decrease has
bee n figured on one base- the pay the men were gett ing in
the first place-while the increase uses a smaller base the
pay level after the cut. '
You .c an che ck on this hit of statistical misltguring by
supposmg, for sim plicity, tha t the orlgtnal wage was $1 an
hour. Cut twenty per cent. it is down to 80 cents. A fi ve
per cent increase on that is 4 cents, which is not nne-fourth
Even The New York Times lost the Batt le of the Shift- but one-fifth of the cut. Like so many presumably honest
ing Base in publi shing an Associated Press story from mistakes, this one somehow managed to come out an ex-
aggeration which made a better story.
Indianapolis:
All this illustrat es why to offs et a pay cut of fifty per
The depression took a stiff wallop on the chin here today. cent you must get a raise of one hundred pe r cent.
Plumbers, plasterers, carpen ters, painters and others affiliated

l
 HOW TO LIE WITH STATISTICS HOW TO STAT1S'T1C'ULATE
" l
It was the Times also that once repo rted th at, for a fiscal
year, air mail "lost through fire was 4,863 pounds, or a
percentage of but O.()(XX)3." The story said that planes
had carri ed 7,715,741 pounds of mail during the year.
An insurance compa ny basing its rates in that way could
get into a pack of trouble. Figure the Joss aod you'll find
that it came 10 0.063 pe r cen t or one hundred times as The wide, blue yonde r lie also turns up during every strike.
great as th e newspape r had it. Every time there is a strike, the Chambe r of Commerce ad-
vertises that the strike is costing so many miWons of dollars
It is the illusion of the shift ing base th at accounts for a day.
the trickiness of add ing d iscou nts. 'Wh en a hardware job- They get the :Ggure by adding up all the cars tha t would
ber offers ".50"1 and 20$ off list," he doesn 't mean a seventy have been made if the strikers had worked full time. They add
per cent discount. The cut is sixty per cent since the in losses to suppliers in the same way . Everything possible is
added in, including street car fares and the loss to merchants
twenty per cent is fi gur ed on the smaller base left after in sales.
taking off fifty per cent .
The similar and equally odd notion that percentages
A good deal of bumbling and chicanery have come from
adding together things that don't add up but mer ely can be added together as freely as apples has been wed
against authors. See how convin cing this one, from T1u:
seem to. Children for generations have been using a form
New }'OTk Times Book Review, sounds.
of this device to prove that they don't go to school.
You probab ly recall it . Starting with 365 days to th e The gap between advancing book prices and authors' earn-
year you can subtract 122 for the one-third of the time ings, it appears, is due to substan tially bfgher production and
material costs. Hem: plant and maeofecturmg expeoses a1onl'
you spend in be d and another 45 for the three hours a day
have risen as much as 10 to 12 per cent over the last decade.
used in eating. From the remaining 198 take away 90 for materials are up 6 to 9 per cent, selling and advertising ex-
summer vacation an d 21 for Christmas and Easter vaca- penses have climbed upwards or 10 per cent Combined boosts
tions . The days that remain are not even enough to pro- add up to a minimum of 33 per cent (for one company) and
to nearly 40 per cent for some of the smaller houses.
vide for Saturdays and Sundays.
Too ancient and obvious a trick to use in serious bust- Actu ally, if each item making up the cost of pubJishing
ness, you might say. But the United Automobile Workers this book has risen around ten pe r cen t, the total cost must
insist in their monthly magazine, Ammunition, that it is have climbed by about that proportion also. The logic
still being used against th em. that permits adding those pe rcentage rises together could
"4 HOW TO LIE WITH STATISTICS HOW TO STATISTlCULATE lI5

lead to all sorts of Bights of fancy. Buy twenty things to- profits should climb from three per cent on investment
day and find that each has gone up five per cent over last one year to six per cent the next, you can make it sound
year. That "adds up" to one hundred per cent, and the quite modest by calling it a rise of three percentage points.
cost of living has doubled. Nonsense. With equal validity you can describe it as a one hundred
per cent increase. For loose handling of this confusing
pair watch particularly the public-opinion pollcrs.
Percentiles are deceptive too. When you are told how
Johnny stands compared to his classmates in algebra or
some aptitude, the figure may be a percentile. It means
his rank in each one hundred students. In a class of three
hundred, for instance, the top three will be in the 99 per-
centile, the next three in the 98, and so on. The odd thing
about percentiles is that a student with a 99-percentile
rating is probably quite a bit superior to one standing at
90, while those at the 40 and 60 percentiles may be of
almost equal achievement. This comes from the habit
that so many characteristics have of clustering about their
It's all a little like the tale of (,e roadside merchant who own average, forming the "normal" bell curve we men-
was asked to explain how he could sell rabbit sandwiches tioned in an early chapter.
so cheap. "Well," he said, "I have to put in some horse Occasionally a battle of the statisticians develops, and
meat too. But I mix 'em fifty-fifty: one horse, one rabbit," even the most unsophisticated observer cannot fail to smell
A union publication used a cartoon to object to another a rat. Honest men get a break when statisticulators fall
variety of unwarranted ad~ing.up. It showed the boss out. The Steel Industry Board has pointed out some of
adding one regular hour at $1.50 to one overtime hour at the monkey business in which both steel companies and
$2.25 to one double-time hour at $3 for an average hourly unions have indulged. To show how good business had
wage of $2.25. It would be hard to find an instance of an been in 1948 (as evidence that the companies could well
average with less meaning. aHord a raise), the union compared that year's productlv-
Another fertile field for being fooled lies in the COD- Ity with that of 1939-a year of especially low volume.
fusion between percentage and percentage points. H your The companies, not to be outdone in the deception derby,
,,6 HOW T O LlE WITH STATISTICS HO W TO ST ATISTICU LA TE
"7
insisted on making their comparisons on a basis of money
received by the employees rather than average hourl y
earnings. The point to this was tha t so many workers
had been on part time in the earlier year tha t their in-
comes were bound to have grown even if wage rates bad
not risen at all.
Time magazine, notable for the consistent excellence of
its graphics, published a chart th at is an amusing example
of how statistics can pull out of the bag almost anything
th at may be wanted. Faced with a choice of methods,
equally valid, one favoring the man agement viewpoint and
the other favoring labor, Time simply used both. The
chart was really two charts, one superimposed upon the
other. They used the same data .
One showed wages and profits in billions of dollars. It
was evident th at both were rising and by more or less the
same amount. And that wages involved perhaps six times
as many dollars as profits did. The great inHationary
pressure, it appeared, came from wages.
The other part of the dual chart expressed the changes
as percentages of increase. The wage line was relatively
Hat. Th e profit line shot sharply upward. Profits, it might
be inferred, were principally responsible £Or inDation.
You could take your choice of conclusions. Or, perhaps
better. you could easily see that neither element could
properly be singled out as the guilt y one. It is sometimes
a substantial service Simply to point out that II sub ject in
controversy is not as open-and-shut as it has been made to
Redrawn wilh the kind permission of Tpa.
seem. magazine as an example of a non-lying chart.

,,
J
 HOW TO LlE WITH STATISTICS

Index numbers are vital matters to millions of people

HOW TO STATlSTICULATE ..,
DOW that wage rates are often tied to them . It is per haps 200

.
'worth noting what <Yom be d one to make them dance to any
.
man s mUS IC'. 150
Milk
$ down - ,
100 BlO.d 1 I
50

To take the simplest possible example, let's say that

milk cost twe nty cents a qu art last year and bread was a last I'" Th is ltll
nickel a loaf. This year milk is down to a dime and bread Try it again, taking th is year as base period. Milk used
has gone up to a dime. Now wha t would you like to prove? to cost 200 per cent as much as it does now and bread was
Cost of living up ? Cost of living down? Or no change? selling for 50 per cent as much. Average : 125 per cent.
200 t-+--: Prices ,used to be 25 per cent higher than they are now.
To !nove that the cost level hasn' t changed at all we
%1U G I simply swi tch to the geometric average and use either
150
'=1. prices up period as the base. This is a little different from the arith-
metic avera ge, or mean, that we hav e been using but it is
100 Ifill a perfectly legitimat e kind of figure and in some cases th e
most use ful and rev ealing. To get the geometric average
50
of three numbers you multi ply them together and derive
the cube root. For four items. the fourth roo t; for tw o. the
Last year This year square roo t. Lik e that.
Consider last yea r as the base period , makin g the prices Take last year as the base and call its price level 100.
of that time 100 per cent. Since the price of milk has since Actu ally you mu ltiply the 100 per cent for each item to-
dropped to half (50 per cent ) and the price of bread has gether and take the root, which is 100. For th is year, milk
d oubled (200 per cent ) and the average of 50 and 200 is being at 50 per cent of last year and bread at 200 per cent ,
125, prices have gone up 25 per cent. multiply 50 by 200 to ge t 10,000. The square root, which is
 BOW TO LI!: WITH STATISTICS HOW TO STATISTICULATE

the geometric average, is 100. Prices have not gone up or Even the man in academic work may have a bias
down. ( possib ly unconscious ] to favo r, a point to prove. an axe
The fact is that, d espite its mathemati cal base , statistics to grind .
is as much an art as it is a science. A great man y manip- T his suggests giving statistical material, the facts and
ulations and even distortions are possible within the figures in ne\\--spape rs and books, magazines and ad vertis-
bounds of propriety. Often the statistician must choose ing. a very sharp second look be fore acce pting any of
amo ng methods, a subjective process. and find the one that them. Someti mes a careful squint will sharpen the focus.
he will we to represent th e facts. In commercial practice But ar b itrarily rejecting statistical methods makes no sense
b- is about as unlikely to select an unfavo rable method as either. That is like refusing to read becau se writers some-
a copywriter is to call his sponsor's produ ct flimsy and tim es use words to hid e facts and relationships rat her than
cheap when he might as well say light and economical. to reveal them. After all, a political candidat e in Florida
not long ago mad e considerable capital by accu sing his
opponent of "practicing celibacy." A New York exhibitor
of the motion picture QUO Vadis used hug e type to quote
I.lUST /.OYE MY NEW The Ney' York Tim es as calling it "historical pretentious-
M/KEDIIP
, £GGBEATJIR ness." And the makers of Crazy Water Crystals, a pro-
ITS ~ FI.IMSY prietary medicine. have been ad vertising their product
-AND O/fEAP as providing "quick. ephemeral relief."

How to LIE

How to llE l'lithout$MiJticS

 HOW TO TALK BACK TO A STATISTIC

About the first thing to look for is bias-the laboratory

10 with somethin g to prov e for th e sake of a th eor y, a repu ta-
tion. or a fee; th e newspaper whose aim is a good story;
labor or manageme nt with a wage level at stake.
Look for COllSC' lO US bias . The met hod may be di rect mis-
sta tement or it may be ambiguous statement that serves
to a Statistic as well and cannot be convicted . It may be selection of
favora ble dat a and suppression of unfavorab le. Units of
meas urement may be shifted, as w ith th e practi ce of using
one year for one comparison and sliding over to a more
fav orable year for another. An improper measure may be
used: s..•nean wh er e a median would be more informative
So F AR. [ have be en addressing you rather as if you were ( pe rhaps all too Informative }, with tb e tr ickery covered
a pirate with a yen for instruction in the finer points of by the unqualified word "aver age:
cu tlass work. In this concluding chapter 111 drop that Look sharply for unconscious bias. It is often more
literary device. 111 face up to the serious purpose that I dan gero us. In the charts and predictions of many stat is-
like to think lurks just beneath the surface of this book: ticians and economists in 1928 it ope rated to produce
explaining how to look a phony statistic in the eye and face remarkable things. The cracks in the economic structure
it down ; and DO less important, how to recognize sound were joyously overlooked, and all sorts of evidence was
and usable d ata in th at wilderness of fraud to which the adduced and sta tistically supported to show that we had
previous chapte rs have been lar gely devoted. no more than entered th e stream of prosperity.
Not all the statistica l informat ion th at you may come up- It may take at lea st a secon d look to find ou t who-says-
on can be tested with the sureness of chemical analysis so. The who m ay be hidden by wh at Stephen Potter, the
or of wh at goes on in an assayer's laboratory. But you can Lifemanship man, would probably call th e "O .K. name."
prod the stuff with five simple questions, and by finding Anything smacking of the medi cal profession is an O.K.
the answers avoid learning a remarkable lot that isn't so. name. Scientific laboratories h ave O, K. names. So do
 BOW TO LIE WITH STATISTICS HOW T O T ALK BACK T O A STATISTIC
'24
colleges, especia lly un iversities, more especially ones
em inent in techn ical work. The writer who proved a few
chapters back that higher educa tion jeopa rdizes a girl's
cha nce to marr}' made good use of th e O.K. nam e of Cor-
nell. Please note th at while the d ata came from Cornell,
th e con clusions were entirel y the writer's own . But the
O.K. n ame helps YOII carry away a misimpression of
"Corne ll University says . . ."
When an O .K. name is cited, make sure th at th e author-
ity sta nds behind th e information, not merely somewhere
alongside it.
You may have rea d a proud announcement by the
Chi cago Journal of Com merce. That publication had
mad e a survey. Of 169 corpora tions that repli ed to a poll
on pr ice gouging and hoarding, two-thirds declared that
th ey were absorbing price increases produ ced by the
Korean war. "The su rvey shows," said the l oum ol ( look
sha rp whenever rou meet those words!) , "that corpora-
tions have done exactly the opposite of what the enemies
o of the American business system have charged." This is
an obvio us place to ask, "Who says so?" since th e Journal
of Commerce might be regarded as an Int erest ed party.
It is also a splendid place to ask our seco nd test question:

BiG NO$ Il H 0
" T H Ro ....T ""' N 0

It turns out th at the Journal had begun by send ing its

questionnaires to 1,200 large comp anies. Onl y fourteen
 HOW TO Lill WITH STATISTICS HOW TO TALK BACK TO A STATISTIC

per cent had replied. Eighty-six per cent had not cared
to say anything in public on whether they were hoarding
or price gouging.
The Journal had put a remarkably good face on things,
but the fact remains that there was little to brag about.
It came down to this: Of 1,200 companies polled, nine
per cent said they had not raised prices, five per cent said
they had, and eighty-six per cent wouldn't say. Those
that had replied constituted a sample in which bias might
be suspected.
Watch out for evidence of ~ biased sample, ODe that
has been selected improperly or-as with this one-has
selected itself. Ask the question we dealt with in an early
chapter: Is the sample large enough to permit any reliable
conclusion?
Similarly with a reported correlation: Is it big enough
to mean anything? Are there enough cases to add up to
any significance? You cannot, as a casual reader, apply
tests of significance or come to exact conclusions as to the You won't always be told how many cases. The absence
adequacy of a sample. On a good many of the things you of such a figure, particularly when the source is an inter-
see reported, however, you will be able to tell at a glance ested one, is enough to throw suspicion on the whole thing.
-a good long glance, perhaps-that there just weren't Similarly a correlation given without a measure of
enough cases to convince any reasoning person of any- reliability (probable error, standard error) is not to be
thing. taken very seriously.
Watch out for an average, variety unspecified, in any
matter where mean and median might be expected to
diHer substantially.
Many figures lose meaning because a comparison is
missing. An article in Look magazine says, in connection

:1
i,
 HOW TO LIE WITH STATISTICS
HOW TO TALK BACK TO A STATISTIC
with Mongolism, that "one study shows that in 2,800 cases. time, and one of them had married a faculty man.
over half of the mothers were 35 or over." Getting any A couple of years ago the Boston Chamber of Commerce
meaning from this depends upon your knowing something ebose its American Women of Achievement. Of the six- f
about the ages at which women in general produce babies. teen among them who were also in Who's Who, it was I
Few of us know things like that. I,
announced that they had "sixty academic degrees and
Here is an extract from the New Yorker magazine's eighteen children." That sounds like an informative
"Letter from London" of January 31,1953. picture of the group until you discover that among the
The Ministry of Health's recently published figures showing women were Dean Virginia Gildersleeve and Mrs. Lillian
that in the week of the great fog the death rate for Greater M. Gilbreth. Those two had a full third of the degrees
London jumped by twenty-eight hundred were a shock to the
between them. And Mrs. Oilbreth, of course, supplied
public, which is used to regarding Britain's unpleasant climatic
effects as nuisances rather than as killers .... The extraordinary two-thirds of the children.
lethal properties of this winter's prize visitation ~ A corporation was able to announce that its stock was
held by 3,003 persons, who had an average of 660 shares
But how lethal was the visitation? Was it exceptional
each. This was true. It was also true that of the two mil-
for the death rate to be that much higher than usual in a
week? All such things do vary. And what about ensuing
lion shares of stock in the corporation three men held
weeks? Did the death rate drop below average, indicating
that if the fog killed people they were largely those who
would have died shortly anyway? The figure sounds
impressive, but the absence of other figures takes away
most of its meaning.
Sometimes it is percentages that are given and raw
figures that are missing, and this can be deceptive too.
Long ago, when Johns Hopkins University had just begun three-quarters and three thousand persons held the other
one-fourth among them.
to admit women students, someone not particularly en-
amored of coeducation reported a real shocker: Thirty- If you are handed an index, you may ask what's missing
three and one-third per cent of the women at Hopkins there. It may be the base, a base chosen to give a distorted
had married faculty members! The raw figures gave a picture. A national labor organization once showed that
clearer picture. There were three women enrolled at the indexes of profits and production had risen much more
rapidly after the depression than an index of wages had
 •
J30 H OW TO LIE WITH STATISTICS HOW TO T A.I..K BACK TO A STATISTIC 131

A5 an argum ent for wage increases this demonstration lost

its potency when someone du g out the missing figures.
It could be seen then th at profits had been almost bound to r"
rise more rapidly in percentage than wages simp ly because When assaying a statistic, watch out for a switch some- I ,
i,
profits had reached a lower point. giving a smaller base. where betw een the raw figure and th e conclusion. One
Sometimes what is missing is the factor tha t caused a thi ng is aU too often reported as another.
change to occur. This omission leaves the implication tha t / ' 0;: just indicated, more reported cases of a disease are
., i
J
some other. more desired, factor is responsible. Figures DOt always the same thing as more cases of the d isease.
publ ished one year attempted to show that business was A straw-vote victory for a candidate is not always ne gon-
on the upgrade by pointing out that April retail sales were abl e at th e polls. An expressed preference by a "cross
section" of a magazine's readers for articles on world
affairs is no final proof th at th ey ....-ould read the articles
if th ey were published.
Enceph alit is cases report ed in the central valley of Cali-
forni a in 1952 were triple the figure for the worst previo us
year. Many alarrned residents shipped th eir children
greater than in the year before. \ Vhat was missing was away. But when the reckoning was in, there had bee n no
th e fact that Easter had come in March in the earlier year great increase in deaths from sleeping slck.n ess. What
and in April in the later year. had happened was that state and federal health people
A report of a great increase in deaths from cancer in the had come in in great numbers to tackle a long-time prob-
last qu art er-century is misleading unless you know bow lem; as a result of their efforts a great many low-grade
much of it is a produ ct of such extraneous [actors as these: cases were recorded th at in other years would have been
Cancer is often listed now where "ca uses unknown" was overlooked , possibly not even recognized .
formerly used; autopsies are more frequent, giving surer It is all remin iscent of the way that Lincoln Steffens and
diagnoses; reporti ng and compiling of medical statistics Jacob A. Riis, as New York newspapermen, once created
are more complete; and people more frequently reach th e a crime wave. Crime cases in th e pa pers reached such
most susceptible ages now. And if you are looking at total proportio ns, both in numbers and in space and big type
deaths rather than the death rate, don't neglect the fact given to them, that the public dema nded action. Theodore
that th ere are more peop le now tha n there used to be. Roosevelt, as president of the reform Police Board, was
131 1I0W TO LIE WITt, STATISTICS
H OW TO T AU: BACh; TO A STATISTIC 133

seriously embarrassed. He pu t an end to the crime wave The figures would be more infonnative if the re were
simply by asking Steffens and Riis to lay off. It had all some indication of whethe r th ey are means or media ns.
come about simply because the report ers, led by those However. the major weakness is that the subject has been
two. had got into competition as to who could dig up the changed. What the Ministry really found out is how often
most burglaries and whatnot. Th e official police record these people said they bathed, not how often they did so.
showed DO increase at all. W .en a sub ject is as intimate as this one is, with the
"The British male over 5 years of age soaks himself in a British bath-taking tr adition involved, saying and doing
hot tub on an average of 1.7 times a week in th e winter may not be th e same thing at all. British he's may or may
and 2.1 times in the summ er," says a newspaper story. not bathe oftener than she's; all that can safely be con-
"British women average 1.5 baths a week in the winter and cluded is th at they say they do.
2.0 in the summer." The source is a Ministry of Works Here are some more varieties of change-of-subject to
hot-water survey of "6,(XX) representative British homes.... watch out for.
The sample was representative. it says, and seems quite A back-to-th e-farm movement was discerned when a
adequate in size to justify the conclusion in the San Fran- census showed half a million more Iartns in 1935 than five
cisco Chronicle's amusing headline : BRITISH HE'S years earlier. But the two counts were not talking about
BATH E MORE T HAN SHE'S. the same thing. The definition of farm used by the
Bureau of the Census had been changed; it took in at least
300,000 farms that would not hav e been so listed under the
1930 definition.
Strange things crop out when figures are based on what
people say-even about things that seem to be objective
facts. Census reports hav e shown more people at thirty-
Bve years of age, for instance, than at eith er thirty-four
(Ir' thirty-six. The false picture comes from one family

member's reporting th e ages of the others and, not being

o
1
I sure of the exact ages, tending to round them off to a
familiar multiple of five. One way to get around thi s:
ask birth dat es instead.
'The "population" of a large area in China was 28 million..

1
'34 HOW TO LIE WITH STATISTICS HOW TO TALK BACK TO A STATISTIC 135

Five years later it was 105 million. Very little of that in- cheaper...." The North Dakotan was referring to earlier
crease was real; the great difference could be explained statements that it cost eight dollars a day to maintain a
only by taking into account the purposes of the two enu- prisoner at Alcatraz, "thet cost of a room at a good San
merations and the way people would be inclined to feel Francisco hotel." The subject has been changed from
about being counted in each instance. The first census total maintenance cost (Alcatraz) to hotel-room rent
was for tax and military purposes, the second for famine alone.
relief. The post hoc variety of pretentious nonsense is another
Something of the same sort has happened in the United way of changing the subject without seeming to. The
States. The 1950 census found more people in the sixty- change of something with something else is presented as
flve-to-seventy age group than there were in the fifty-five- because of. The magazine Electrical Warld once offered
to-sixty group ten years before. The difference could not a composite chart in an editorial on "What Electricity
be accounted for by immigration. Most of it could be a Means to America," You could see from it that as "elec-
product of large-scale falsifying of ages by people eager trical horsepower in factories" climbed, so did "average
to collect social security. Also possible is that some of wages per hour," At the same time "average hours per
the earlier ages were understated out of vanity. week" dropped. All these things are long-time trends, of
Another kind of change-of-subject is represented by course. and there is no evidence at all that anyone of them
Senator William Langer's cry that "we could take a pris- has produced any other,
oner from Alcatraz and board him at the Waldorf-Astoria And then there are the firsters. Almost anybody can
claim to be first in something if he is not too particular
what it is, At the end of 1952 two New York newspapers
were each insisting on first rank in grocery advertising.
Both were right too, in a way, The World-Telegram went
on to explain that it was first in full-run advertising, the
kind that appears in all copies, which is the only kind it
runs. The Journal-American insisted that total linage was
what counted and that it was first in that. This is the kind
of reaching for a superlative that leads the weather
reporter on the radio to label a quite normal day "the
hottest June second since 1949."

i
1)6 HOW 1'0 LD: WITH STATISTICS BOW 1'0 TALE: BACK: TO A STATJSI1C IJJ

Change-of-subject makes it difficult to compare cost Accountants have decided that "surplus" is a nasty word.
They propose eliminating it from corporate balance sheets.
when you contemplate borrowing money either directly
The Committee on Accounting Procedure of the American In-
or in the form of installment buying. Six per cent sounds stitute of accountants s;.:.~'S: . . . Use such descriptive terms
like six per cent-but it may not be at all. as "retained earnings'" or "appreciation of fixed assets."
If you borrow $100 from a bank at six per cent interest This one is from a newspaper story reporting Standard
and pay it back in equal monthly installments for a year, Oil's record-breaking- revenue and net profit of a million
the price you pay for the use of the money is about $3. dollars a day.
But another six per cent loan, on the basis sometimes Possibly the directors may be thinking some time of splitting
called $6 on the $100, will cost you twice as much. That's the stock for there may be an advantage ... if the profits per
share do not look so large..•.
the way most automobile loans are figured. It is very
tricky.
The point is that you don't have the $100 for a year. By
the end of six months you have paid back half of it. If "Does it make sense?" will often cut a statistic down to
you are charged at $6 on the $100, or six per cent of the size when the whole rigmarole is based on an unproved
amount, you really pay interest at nearly twelve per cent. assumption. You may be familiar with the Rudolf Flesch
readability formula. It purports to measure how easy a
Even worse was what happened to some careless pur-
chasers of freezer-food plans in 1952 and 1953. They were piece of prose is to read, by such simple and objective
items as length of words and sentences. Like all devices
quoted a figure of anywhere from six to twelve per cent.
It sounded like interest, but it was not. It was an on-the- for reducing the imponderable to a number and substitut-
dollar figure and, worst of all, the time was often six ing arithmetic for judgment, it is an appealing idea. At
months rather than a year. Now $12 on the $100 for least it has appealed to people who employ writers, such
money to be paid back regularly over half a year works as newspaper publishers, even if not to many writers them-
out to something like forty-eight per cent real interest. selves. The assumption in the formula is that such things
It is no wonder that so many customers defaulted and so as word length determine readability. This, to be omery
many food plans blew up. about it, remains to be proved.
Sometimes the semantic approach will be used to A man named Robert A. Dufour put the Flesch formula
change the subject. Here is an item from Business Week to trial on some literature that he found handy. It showed
magazine. "The Legend of Sleepy Hollow" to be half again as hard
 HOW TO LIE wrru STATISTICS HOW TO TALK BACK TO A. STATISn c 139

to read as Plato's Republic. The Sinclair Lewis novel bom can expect to live longer th an that. The figure, in-
Cas! Tlm berlatle was ra ted more difficult th an an essay cidentally, is from the latest official complete life table
by Jacques Maritain , "The Spiritual Value of Art." A and is correct for the 19.'39-1941. period. An up -to-date
like ly story. estima te corrects it to sixty-live-plus. Maybe that will
Many a sta tistic is false on its face. It gets by only produce a new and equally silly a rgum ent to the eflect
because th e magic of numbers brings about a suspens ion th at practically everybody now lives to be Sixty -live.
of common sense. Leonard Engel, in a Harper's article, Postwar planning a t a big electrica l-appliance company
ha s listed a few of the medical variety. was going great guns a few years ago on the basis of a
declinin g bi rth rat e. some thing that had been tak en for
An example is the calculation of a well-known urologist that
granted for a long time. Plans caned for emphas is on
there are eight million cases of cancer of the prostate gland
in the United States- which would be enough to provide 1.1 small-cap acity appliances, apartm en t-size refrigerators.
carcinomatous prostate glands for every male in the susceptible Then one of the planners had an attack of common sense :
age group! Another is a prominent neurologist's estimate that H e came out of his gr aphs and charts long enough to
one American in twelve suffers from migraine; since migraine
notice th at he and his co-workers and his friend s and his
is responsible for a third of chronic headache cases, this would
mean that a quarter of us must suHer from disabling headaches. neighbors and his form er classma tes wi th few exceptions
Still another is the fi gure of 250.000 often given for the number either had three or four childr en or pla nned to. This led
of multiple sclerosis cases; death data indicate that there can to some open-minded investiga ting a nd cha rting-and the ..,
be, happily, no more than thirty to fort)" thousand cases of this
company shortly turned its emph asis most profitably to
paralytic disease in the country.
big-fam ily models.
Hear ings on ame ndments to the Social Security Act The impressively precise figure is some thing else that
have been haunted by va rious forms of a stateme nt th at contradicts common sense. A study reported in New York
makes sense only when not looked at closely. It is an City newsp apers announced th at a worki ng woman living
argu men t th at goes like th is: Since life expecta ncy is only with her family needed a weekly pay check of $40.13 for
about sixty-th ree year s. it is a sham and a fraud to set up adequ ate support. Anyone who ha s not suspended all
a social-security pl an with a retirement age of sixty-five, logical p rocesses whil e reading his p ap er will rea lize th at
because virtually every body dies befo re th at . th e cost of keeping body and soul togeth er ca nnot be
You ca n rebut that one by looking around at people calculated to the last cent. But the re is a dreadful te mpta-
you kn ow. The basic fallacy, however, is that the figure tion; "$40.13" sounds so much more knowing than "a bout
refers to expectancy at birth, and so about half the babies $40."

.'

;I
 HOW T O LIE WITH STATISTICS HOW TO TAl..X. BACK TO A STATISTIC
ti'
You are entitled to look with th e same suspicion on the
report. some years ago, by the American Petroleum Indus-
tri es Committee that the average yea rly tal: bill for auto-
mobiles is $51.13.
Extrapolations arc useful, parti cularly in that form of
sooths aying called forecasting trends. But in looking at
the figures or the cha rts mad e from them. it is necessary
to remember one th ing constantly : The trend-to-now may
he a fact, but the futu re tren d re presents no more th an an
ed ucated guess. Implicit in it is "everything else bei ng
equal" and "present trends con tinuin g." And some how
everything else refuses to remain equal, else life would
be dull indeed .
For a sample of the nonsen se inherent in uncontrolled
extrapolati on, consider th e trend of television. Th e num-
ber of sets in Ameri can homes increased around lO,()()()%
from 1947 to 1952. Project thi s for the next five years and
you find that ther el1 soon be a couple billion of the things.
Heaven forbid, or forty sets pe r family. If you want to be .,
even sillier, begin with a base year that is earlier in the -,
television scheme of things than 1947 and you can just as
well "prove" that each family will soon have not forty hut
fort)' thousand sets.
A Government research man . Morris Hansen , caned dential commission loaded with experts do ubted that the
Callup's 1{;48 elect ion forecasting "the most pu blicized u. S. population would ever reach 14!} million; it was
stat istical error in h uman history," It was a paragon of 12 million more than that just t welve yea rs later. Th ere
accuracy, however, compared with some of our most are textboo ks pu bli...hed so recent ly that they arc still in
widely used estimates of future population, which have college use that predict a peak population of not more
earned a nationwide horselau gh , As late as 1938 a prest- than 150 million and figure it will take until about 1980
14' HOW TO LIE WITH STATISTICS

to reach it. These fearful underestimates came from as-

suming that a trend would continue without change. A
similar assumption a century ago did as badly in the
opposite direction because it assumed continuation of the
population-increase rate of 1790 to 1860. In his second
message to Congress, Abraham Lincoln predicted the
u. S. population would reach 251,689,914 in 1930.
Not long after that, in 1874, Mark Twain summed up
the nonsense side of extrapolation in Life on the Missis-
sippi:
In the space of ODe hundred and seventy-six years the Lower
Mississippi has shortened itself two hundred and forty-two
miles. That is an average of a trifle over one mile and a third
per year. Therefore, any calm person., who is not blind or
idiotic, can see that in the Old Oolitic Silurian Period, just a
million years ago next November, the Lower Mississippi River
was upward of one million three hundred thousand miles long,
and stuck out over the Gulf of Mexico like a fishing-rod. And
by the same token any person can see that seven hundred and
forty-two years from now the Lower Mississippi will be only
a mile and three-quarters long, and Cairo and New Orleans
will have joined their streets together, and be plodding com-
fortably along under a single mayor and a mutual board of
aldermen. There is something fascinating about science. One
gets such wholesale returns of conjecture out of such a trilling
investment of feet,