An Interview with George B.

Dantzig: The Father of Linear Programming

Author(s): Donald J. Albers, Constance Reid, George B. Dantzig
Source: The College Mathematics Journal, Vol. 17, No. 4 (Sep., 1986), pp. 293-314
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2686279
Accessed: 04/11/2009 14:55
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=maa.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to
The College Mathematics Journal.

http://www.jstor.org

George Bernard DaatzJg.

292

An

Interview

The
by Donald

with
of

Father
J. Albers

B.

George

Linear

Dantzig:

Programming

with Constance

Reid

George Bernard Dantzig is rich in many ways, but as a boy growing up in the
shoes. Today he and his wife Anne live in a lovely
twenties he wore secondhand
home on the edge of the Stanford University
campus. His study walls are covered
several honorary doctoral degrees, the John von Neumann
with awards, including
Theory Prize, and the National Medal of Science.
are especially noteworthy in view of the trouble that he had
His accomplishments
in junior high school. "I was doing very poorly in my first course
with mathematics
in algebra. To be precise, I was flunking."
In 1947 Dantzig invented linear programming
and the simplex method. Dantzig's
environmental
led to an explosion of economic,
and statistical appli?
programming
The iron and steel industry has used his method to evaluate iron ores,
cations.
of coke ovens, and select products.
The Federal
the addition
Energy
explore
Linear
has used his method to explore energy policy alternatives.
Administration
for use to control water and air
has also been used or suggested
programming
to jobs, and racially balancing
refinery scheduling,
assigning personnel
pollution,
schools.
He responds
work is often thought of as applied mathematics.
by
Dantzig's
saying, "I have never been able to tell the difference between the so-called pure and
the nonpure and don't believe that there is any."
in his Stanford office in November
of 1984.
Dantzig was interviewed
What's

this

MP:

Professor

Dantzig:

What's

MP:

Okay,

what should

Dantzig:

Your

name is Don,

MP:

It's George.

Dantzig:

Do you remember

MP:

It's Bernard.

Dantzig:

What

MP:

The well-known

Dantzig,

"Professor

for taking

thanks

this "Professor

Dantzig"

time to talk about

yourself

today.

stuff?

I call you?
right

Do you remember

my middle

does George

Stuff?

Dantzig"

Bernard

my first name?

name?

suggest?

writer named

George

Bernard

Shaw?

That's right. My father, Tobias

was both a writer and a
Dantzig:
Dantzig,
He hoped that I would be a writer, and so he named me after
mathematician.
George Bernard Shaw. He named my younger brother Henri Poincare Dantzig after
Henri Poincare.
a
the great mathematician
My brother did, in fact, become
He worked as an applied mathematician
for the Bendix Corporation
mathematician.
until he died in 1972 at age 54.
293

MP:
I know your father's books, especially
his wonderful
Language of Science, but nothing much about him personally?except
born in Russia and studied in Paris with Poincare.

The
Number,
that he was

Paris is a place where professors don't pay much attention to students so

Dantzig:
I can't say precisely what my father's relationship
to Poincare was. I do know that
he attended
his lectures, studied his works, and admired him greatly. One of his
books is entitled Henri Poincare.

MP:

How

did your father get to the United

States?

He came to this country twice, once before he married my mother. On

Dantzig:
that first trip he visited his aunt in South Carolina whose family owned a general
He worked for them for a while as a peddler. He must have
apparel business.
decided it had no future, for he returned to Paris. My mother, Anja Ourisson, had
at the Sorbonne at the time she
grown up in Poland and was studying mathematics
met my father. After their marriage they emigrated
to Oregon, where my father
In those days, gangs of young men were hired to fell trees
worked as a lumberjack.
and nearly worked to death. Later he had a job as a road worker. On my birth
certificate he is listed as a painter, probably a house painter. As a result of all the
hard manual work, he developed huge arm muscles.
MP:

How

did he manage

to get back into mathematics?

Dantzig:
My father believed that he could never get a job in a university because
of his heavy Russian accent; but one day at the public library he ran into Frank
at Reed College, who told him he
Griffin, the head of the mathematics
department
was crazy to be working as a lumberjack and road builder. Griffin assured him that
with his academic
credentials
he could get a job in any university.
That was a
turning point in my father's career. He applied to Indiana University and was hired.
He didn't have a formal Ph.D. at the time, but he soon acquired one while a
there.
professor
MP:

Did he quit worrying

about his accent

after that?

I don't think he ever worried much about his accent. His spoken English
Dantzig:
was otherwise fluent, and he was known for his marvelous English writing style.
MP:
more

about

He obviously
him?

was very important

in your life. Can you tell me a little bit

He was a dynamic person with a very strong personality.

He knew the
Dantzig:
classics and could quote them in Greek, Latin, and a dozen other languages. He was
an excellent
teacher and quite a raconteur. In the 1920's many of his friends were
in the philosophy
of science. He used to hold salons in our home.
very interested
From age eight until I was sixteen, I used to sit in the corner and listen to the
smart-alec intellectuals
of the 1920's expounding.
I learned a lot but said little. After
a while I began to suspect that they really didn't know what they were talking
about. Perhaps this explains why to this day I can never get very excited about
ideas.
philosophical

294

Tobias Dantzig, father of George, as a young

man. Born in Russia, he studied at the Sorbonne
and later became a mathematics professor in the
U.S.

Anja Ourisson, mother of George Dantzig. She

studied mathematics at the Sorbonne and lan?
guages at Johns Hopkins University.

Baby George?age, one year.

The Dantzig brothers, Henry and George, in 1920.

295

Flunking
MP:

Can you remember

Junior

High

School

when your own interest

Algebra
in mathematics

was aroused?

Yes. I was in the ninth grade of Powell Junior High School in Washing?
Dantzig:
ton, D.C. My father was teaching nearby at the University of Maryland. I was doing
very poorly in my first course in algebra. To be precise, I was flunking. I remember
walking home one day, furious with myself. How is it, I asked myself, that I, a son
of a mathematician,
do poorly while all the other kids in the class do so much
better? I was very angry with myself. After that I sailed through algebra.
MP:

It sounds

as if you had a lot of self-confidence

Dantzig:
grade had
recovering
was good
remember

Confidence
came slowly. My interest
been zero. I then began to blossom
from my poor start in algebra, I began
in math and science in high school.
being interested in any other subjects.

MP:

Do you remember

any influential

in school
in science
to get top
I was on

by grade nine.
work up to the seventh
courses. Later on, after
I
marks in mathematics.
the chess team. I can't

teachers?

Mr. Gilbert, my mathematics

teacher at Central High
Yes, especially
I took geometry
from him. Geometry
really turned me on. Another
influence was Abe Seidenberg, who entered high school one year after I
important
did. We did our math together and became good friends. He is now a professor at
Dantzig:
School.

Berkeley.
MP:

I studied

from his book,

Lectures

on Projective

Geometry.

The Big Prize?a photo of his classmates in the second grade. Dantzig (center of rear row)
was awarded the photo for winning the long division contest. It was offered for sale but his
family was too poor to buy it. The only way he could get it was to win the contest.

296

Ten

Thousand

Problems

Geometry

was like mother's milk to me, to quote Eliza in

Dantzig:
Projective
geometry
I was brought up on it. My father taught me by giving me
Shaw's Pygmalion.
problems to solve. He gave me thousands of geometry problems while I was still in
high school.
MP:

Thousands?

I would say over ten thousand. After he gave me one and I came back
Dantzig:
with a solution, he would say, "Well, I'll give you another one." It seemed as if he
had an infinite storehouse of them. At first he would check my solutions, but after a
while he would accept them as correct and just give me another, and another, and
another problem. The mental exercise required to solve them was the great gift from
the time
my father. Solving thousands of problems during my high school days?at
when my brain was growing?did
more than anything else to develop my analytical
power.
MP:

Just working

those problems?

Dantzig:
have done

Yes, it was brain exercise.

as well.

MP:

Did you ever feel that your father was pushing

Problems

on any other subject probably

would

you into mathematics?

Never! It was I who asked for the problems. I believe he gave them to me
Dantzig:
just to get rid of me. It was almost as if he were saying, "Here's another problem.
Now go away and don't bother me." He was always busy with whatever he was busy
of
care of his students, writing, doing research, and so on. Eventually,
with?taking
course, he did run out of problems and had to go to the Library of Congress to dig
ones.
up additional
MP:

So you literally

Dantzig:

Eventually,

exhausted

yes. It seems

his supply

of problems?

that he didn't

have an infinite

supply

after all.

Tobias Dantzig gave his son George a

"great gift"?thousands of geometry
problems to solve.

297

Secondhand
MP:
You mentioned
were growing up.

to me earlier

Shoes
that there wasn't

much

money

when

you

We were always very poor. When my father taught at Johns Hopkins in

Dantzig:
we wore secondhand
shoes. In the 1920's my mother obtained a master's
1919-20,
degree in French from Johns Hopkins in order to qualify for a job at the Library of
She was a linguist and a specialist in Slavic languages.
Even with the
Congress.
income from two salaries, we were still poor. I don't remember ever having pocket
money. There were hardly any jobs for kids. I did have a paper route once. My
at
father never earned very much. When he retired as head of the math department
Maryland shortly after World War II, his pension was only $2,250 a year, which was
half his yearly salary. This happened just before the postwar inflation. He moved to
and that
Los Angeles where he tried to pick up extra money teaching, consulting,
sort of thing. Although his health was failing, he managed somehow. He never asked
us, his children, for money. I don't know how he ever managed. Although he had
been in Los Angeles only a short time when he died in 1957, hundreds of people
came to his funeral. Nobody who ever met him forgot him. There was something
magical about the way people were attracted to him.
"Applied

Mathematics?I

By the time you finished

in college?
in
mathematics
major
MP:

Didn't

high school

Seek

it Out"

was it clear to you that you would

Dantzig:

Yes.

MP:

You said your father was teaching at the University

was a convenient
place to go. Was that it?

Maryland

of Maryland.

So

I certainly had no
Yes, of course. It wasn't a time for high aspirations.
Dantzig:
dreams of going off to a fancy school which would require my family to support me
away from home.
MP:

Did your mathematics

study have a strongly

applied

flavor at that time?

in any of the
not. I don't recall a single application
No, absolutely
Dantzig:
What math there was in physics and
courses I took at Maryland.
mathematics
chemistry was pretty primitive. I did, however, encounter an interesting application
in a freshman chemistry course given by a Professor White. I wrote
of mathematics
a little applied mathematics
paper on how to efficiently extract iodine from a water
as an extractor. He looked at it and said that it
solution using carbon tetrachloride
the carbon-tet,
but that he was sure
idea to subdivide
was a very interesting
someone must have already published the idea. Two years later, when I was a junior,
on the
and showed me a paper just published
he came around, very shamefaced,
in the way of an
same idea. That was the only thing I ever did at Maryland
I wasn't opposed to doing applied mathematics?it
just never sought
application.
me out, and I didn't seek it out.
MP:
After getting your bachelor's at Maryland in 1936, you went to Michigan
for graduate study. Did you take any statistics at Michigan?
and R. L.
Dantzig:
Mainly I studied under G. Y. Rainich, T. H. Hildebrandt,
Wilder. I did take a statistics course with H. C. Carver. In the summer of 1936 I
married Anne Shimmer, and she came to Ann Arbor with me. We earned money
298

Dantzig as a graduate student in Michigan.

Newlyweds Anne and George Dantzig enjoying a picnic in Michigan in 1937.

on the side for a flour company. The
working for Carver. He did some consulting
wanted to buy large quantities of wheat in the commodity
market during
company
the year at prices that would average out to the average annual price. Carver worked
out a system of hedging that was supposed
to be sensational.
He swore me to
the
was
careful
to
the
work
sheets
so that I would
he
same,
secrecy; just
camouflage
not discover his secret regression formula.
Statistics as taught by Carver seemed to me just a bag of tricks?it
didn't have
I
I
rationale
that
could
discern.
else
took
was
at
any
Everything
Michigan
terribly
abstract?so
abstract that I had but one desire: to quit my graduate studies and get
a job, which I did.
299

MP:

So that's when you went to Washington,

D.C?

Yes. By luck?that
was 1937, still the Depression?I
Dantzig:
got a job as a
clerk with the Bureau of Labor Statistics. The job I took, although at a
statistical
lower civil service grade, was the one that had recently been vacated by Milton
Friedman. I was assigned to a project called "Urban Study of Consumer Purchases"
and asked to review a paper on double sampling by the famous statistician
Jerzy
who was then at University
Neyman,
College in London. It was my first encounter
with statistical theory based on a logical rationale. I was very excited by the paper.
it was the least representative
of Neyman's
contributions.
I am
Later I discovered
I would have become even
sure that if I had seen anything more representative,
more excited.
MP:

Did you like your work at the Bureau

of Labor Statistics?

Our group was very

Yes, I learned a lot about practical applications.
Dantzig:
good. My co-worker was Duane Evans. He and I became good friends. Later Evans'
model of the U.S.
Leontief's
work in World War II on Wassily
input-output
economy
changed the course of my career.
MP:
So then how did you get from the Bureau
you really moved around!

to Berkeley?

Like your father,

In retrospect
I don't think I moved around too much. While at the
Dantzig:
Bureau, I wrote to Neyman, who was by then at Berkeley, and told him that I would
like to finish my Ph.D. under him. At Berkeley in 1939, statistics was still part of the
so the focus was on pure mathematics
and not on
mathematics
department,
that I was ever
statistics.
The total number of courses in theoretical
statistics
exposed to was two given by Neyman.
MP:

What was Neyman

like as a person?

had a dominating
that he was able to assert long
Dantzig:
Neyman
personality
after he had been officially retired. In his seventies and eighties, he continued to run
the statistical laboratory at Berkeley. No one dared to contradict him. He was top
that he was a tyrant. He
dog in every sense. I don't want to give the impression
wasn't. He was very likeable?everyone
respected him as the leading mathematical
in the world, quite correctly, I think.
statistician

Jerzy Neyman was "top dog in every

sense" according to Dantzig.

300

How

to Get

a Ph.D.?Do

Your

Homework!

MP:
How did it happen that you did your Ph.D.
you took so few courses in statistics?

on a statistical

topic

when

It happened because during my first year at Berkeley I arrived late one

Dantzig:
classes. On the blackboard
there were two problems that I
day at one of Neyman's
assumed had been assigned for homework. I copied them down. A few days later I
for taking so long to do the homework?the
to Neyman
problems
apologized
seemed to be a little harder to do than usual. I asked him if he still wanted it. He
told me to throw it on his desk. I did so reluctantly because his desk was covered
with such a heap of papers that I feared my homework would be lost there forever.
About six weeks later, one Sunday morning about eight o'clock, Anne and I were
He rushed in
awakened
by someone banging on our front door. It was Neyman.
to one of your
with papers in hand, all excited: "I've just written an introduction
For a minute I had
papers. Read it so I can send it out right away for publication."
no idea what he was talking about. To make a long story short, the problems on the
blackboard
that I had solved thinking they were homework were in fact two famous
in statistics.
That was the first inkling I had that there was
unsolved
problems
anything special about them.
MP:

But you had apologized

to Neyman

for taking

so long to do them.

and you know how graduate

Well, there was no particular deadline,
take their time. A year later, when I began to worry about a thesis topic,
Neyman just shrugged and told me to wrap the two problems in a binder and he
would accept them as my thesis.
until after World
The second of the two problems, however, was not published
War II. It happened
this way. Around 1950 I received a letter from Abraham Wald
the final galley proofs of a paper of his about to go to press in the Annals
enclosing
Mathematical
Statistics. Someone had just pointed out to him that the main result
of
in his paper was the same as the second "homework"
problem solved in my thesis. I
He simply inserted my name as
that we publish jointly.
wrote back suggesting
coauthor into the galley proof.
Dantzig:
students

Homework

and

Religion

MP:
Is it true, as I have heard, that the story of your "homework
has been used by ministers in sermons?

problems"

so. The other day, as I was taking an early morning walk, I

Dantzig:
Apparently
was hailed by Don Knuth as he rode by on his bicycle. He is a colleague at Stanford.
was visiting in Indiana recently and heard a
He stopped and said, "Hey, George?I
sermon about you in church. Do you know that you are an influence on Christians
"
the
of middle America?" I looked at him, amazed. "After the sermon," he went on,
minister came over and asked me if I knew a George Dantzig at Stanford, because
that was the name of the person his sermon was about."
MP:

How

did that happen?

The origin of that minister's sermon can be traced to another Lutheran

Dantzig:
Schuler of the Crystal Cathedral in Los Angeles.
Several
the Reverend
minister,
he
I
and
years ago
happened to have adjacent seats on an airplane. He told me his
and I told him my story about the homework
ideas about thinking positively,
and my thesis. A few months later I received a letter from him asking
problems
301

to include my story in a book he was writing on the power of positive

permission
Schuler's
but essen?
thinking.
published version was a bit garbled and exaggerated
tially correct. The moral of his sermon was this: If I had known that the problems
were not homework
but were in fact two famous unsolved problems in statistics, I
would have become discouraged,
and
probably would not have thought positively,
would never have solved them.

World
MP:

War

II and

Did you finish your Ph.D.

Air

Force

Planning

at Berkeley?

Not quite. I had completed my course work, and my thesis was settled in
Dantzig:
June 1941. But I had not defended
my thesis or my minor thesis on dimension
theory. This was six months before Pearl Harbor. Many of us wanted to contribute
to World War II, which we believed the U.S. was about to enter. I went back to
and had an interview
with Charles Bates
during summer vacation
Washington
who had been selected by Secretary Lovett to set up Air Force
"Tex" Thornton,
The Air Force at that time did not have a good system for
Statistical
Control.
reporting the status of their aircraft. They didn't even know their total number of
planes, which at the time was less than 100. The interview took place on the corner
of 20th Street and Constitution
Avenue. He wanted me to join him right away. My
wife, Anne, who was with me at the interview, is very good at spotting talent. She
told me Thornton was a man who was going places and I should take the job. She
was right. As you know, after the war Tex founded Litton Industries.
MP:
subject,

So it was your work with the Air Force that got you into this now famous
linear programming?

Not exactly right away. I stopped my graduate studies and joined the Air
Dantzig:
Force as a civilian. I was put in charge of the Combat Analysis Branch of Statistical
Control. I set up a reporting system for combat units on the number of sorties flown,
aircraft lost and damaged, bombs dropped, and targets attacked. I became quite
methods
at programming
the only
ma?
expert
planning
using
"computing
chines" we had then?people
desk calculators.
using hand-operated
in the Pentagon
included
Brandon
Barringer
(a well-known
My colleagues
banker), Robert McNamara
(of World Bank fame), Edward Learned
Philadelphia
at New York
(of the Harvard Business School), and Warren Hirsch (the probabilist
who
was
my deputy.
University),
The

Challenge

That

Led

to

Linear

Programming

In spring 1946 I returned to Berkeley and finished my Ph.D. I turned down an

offer from Berkeley because it paid too little and returned to the Pentagon, where I
adviser to the U.S. Air Force Comptroller.
But I was really
became the mathematical
looking for an academic position. To entice me into remaining with the Air Force,
two of my colleagues, Dal Hitchcock and Marshall Wood, challenged me to see what
could be done to mechanize the planning process; that is, to find a more rapid way
to compute
a time-staged
training, and logistical
supply program.
deployment,
in those days meant using analog devices or punch-card
Mechanization
equipment.
I set out to formulate a model. I
with my training as a mathematician,
Consistent
was fascinated
by the work being done at the Bureau of Labor Statistics by Duane
and Marvin Hoffenberg
on the input-output
model of
Evans, Jerome Cornfield,
302

at
Wassily Leontief. I had learned about it during the war in telephone conversations
were much too busy during the day to talk.
night with Duane Evans?we
In my Linear Programming and Extensions you will notice that I pay great tribute
the Inter-industry
It was Leontief who around 1932 first formulated
to Leontief.
Model of the American Economy, organized the collection of data during the Great
and finally tried to convince policy makers to use the output from the
Depression,
and
All of these things are necessary
steps for successful
applications,
analysis.
Leontief took them all. That is why in my book he is a hero.
Leontief s model had a matrix structure which was simple enough in concept with
sufficient detail that it could be useful for practical planning. I soon saw that it had
model and what was needed was a
Leontiefs
was a steady-state
to be generalized.
time. In his model there was a
over
one
could
that
change
highly dynamic model,
and the items pro?
the
between
one-to-one
production
processes
correspondence
What was needed was a model with many alternative
duced by the processes.
the application
had to be large scale?with
activities.
hundreds, perhaps
Moreover,
In other words,
of activities and items. Finally, it had to be computable.
thousands
there had to be a practical way to compute what
once the model was formulated,
with their
to engage in so as to be compatible
of these activities
quantities
would be
characteristics
and given resources. The model I formulated
input-output
described
dynamic linear program with a staircase matrix
today as a time-staged
structure. Initially there was no objective function, in other words no explicit goal.
Such goals did not exist in any practical sense because planners simply had no way
them.
to implement

An

Earth

Filled

with

Computers

a planning
A simple example illustrates the fundamental
difficulty of formulating
of assign?
Consider
the
an
such
problem
approach.
activity-analysis
program using
ing 70 men to 70 jobs. An "activity" consists of assigning the /th man to the y'th
are (a) that there are 70 men, each of whom must be assigned,
job. The restrictions
and (b) that all of the jobs, also 70, must be filled. The level of an activity is either 1,
meaning it will be used, or 0, meaning it will not. Thus there are 2 X 70, or 140,
zero-one
with 4900 corresponding
and 70 X 70, or 4900, activities
restrictions
or ways
there are also 70 factorial permutations,
decision variables.
Unfortunately
The problem is to compare these 70 factorial ways and to
to make the assignments.
select the one which is optimal, or "best" by some criterion.
Now in this example 70 factorial is a very big number. To get some idea of how
computer available at the time of the
big, suppose we had had an IBM main-frame
been
then and now?have
Would
it?between
million
fifteen
years ago.
Big Bang
No! But suppose that an even more
all the possible solutions?
able to examine
one billion
had been available, one that could have examined
computer
powerful
assignments
per second. The answer would still be no. Even if the Earth were filled
with nanosecond-speed
computers, all working in parallel, the answer would still be
no. If, however, there were ten Earths, all filled with nanosecond-speed
computers,
in parallel from the time of the Big Bang until the sun grows cold,
all programmed
then perhaps the answer would be yes. The remarkable
thing is that the simplex
method with the aid of a modern computer can solve this problem in a split second.
This example illustrates why, up to 1947 and for the most part up to this day, a
great gulf exists between man's aspirations and his actions. Man may wish to state
but there are so many ways to
his wants in terms of an objective to be extremized;
303

that it has been

go about doing the job, each with its advantages and disadvantages,
to
them
and
to
them
that
one
which
choose
is best. So,
impossible
compare
among
man has always had to turn to a leader whose "experience"
and "mature
invariably,
would guide the way. The leader's guidance
in the
judgment"
usually consisted
of a series of edicts or ground rules to those developing
issuance
the programs.
such methods are still widely used, the world today is far too complex for
Although
such simplistic methods to work, and they don't.
In late 1946, before we knew that high-speed
electronic
were soon
computers
a mathematical
model that satisfactorily
going to exist, I had formulated
repre?
in practice. However, in place
sented the technological
relations usually encountered
of any explicitly stated goal, or function to be extremized, there were a large number
of ad hoc ground rules issued by those in authority to aid in the selection of the
to choose from the astro?
solution.
Without
these it would have been impossible
nomical number of feasible solutions.
MP:
Most

That certainly has to be classed

mathematicians
prefer problems which

as a very messy real-world

are cleanly formulated.

problem.

for someone coming from a purely mathematical

It is almost impossible
Dantzig:
with
to
how to go about
little
to understand
exposure
background
applications
a real-world problem in mathematical
terms. There is a certain softness
formulating
in the definition of many "dirty" real-world problems, which
?a lack of precision?
When I say
mathematical
formulations.
them to have many equivalent
permits
I
in
mean
the
mathematical
sense
of
don't
one-to-one
equivalent
"equivalent,"
but equivalent for the purpose of the application.
From the point
correspondence,
of view of the person looking for an answer, one definition of the problem may be
as another. But one definition may turn out to be completely
just as satisfactory
to mathematical
amenable
analysis and solution while another may be mathemati?
of the problem can one decide
Only through detailed knowledge
cally hopeless.
is just as acceptable.
Linear programming
whether the more tractable formulation
with them many large problems
models
have been successful
because
can be
formulated
so that they are acceptable to planners and solvable on a computer.

The
MP:
you're

Young

You are often

comfortable
with?

Father
called

of Linear

the Father

Programming

of Linear

Programming.

Is that a title

I have to tell you a story. Twenty-five

Dantzig:
years ago I visited Japan for the
first time. When I got off the plane, the Japanese who met me were very surprised at
how young I was. Since I had been billed as the Father of Linear Programming,
they
to see an old man with white hair and a cane being helped
expected
apparently
down the ramp. I was 45 at the time.
MP:
The preface to your book Linear Programming
and Extensions
opens
with a provocative
statement: "The final test of a theory is its capacity to solve the
problems which originated it."
Dantzig:

Did I say that? It's a great quote.

MP:
Your second
the theory and solution
be just as interesting
to
Curiously
enough, until
304

Show me where.

with
paragraph isn't bad either: "This book is concerned
of linear inequality systems. On the surface, this field should
as its special case, linear equation systems.
mathematicians
1947 linear inequality
theory generated only a handful of

The Father of Linear Programminglecturing in Japan.

papers, while linear equations and the related subjects of linear algebra and
theory had developed a vast literature. Perhaps this disproportionate
approximation
interest in linear equation theory was motivated more than mathematicians
care to
admit by its use as an important tool in theories concerned with the understanding
of the physical universe."
isolated

I think of modern mathematicians

as a distinct race characterized
Dantzig:
by
in applications.
From a historical point of view mathematicians
their non-interest
let us say, 1820 were very closely tied to physics?or,
in the case of
before,
theory, to gambling. For the past 150 years, however, mathematicians
probability
have created their own abstractions and followed the mathematical
fads that happen
to be in fashion. The fact that there is a whole world of exciting new mathematics
out there in such fields as Operations
Research, Computer
Science, Optimization
I for one have no interest in trying to
has not excited their interest.
Theory
would be a hopeless task. The most we can hope for is that they
re-educate
them?it
to the point that they don't prejudice gifted students too much
can be educated
that goes by different names.
against that wonderful world of mathematics
MP:
Dantzig:
as they
matics
able to
believe
problem

How

do they bias students

against

applications?

By showing their contempt for anything which is not pure mathematics

into thinking that pure mathe?
define it. Students are being brainwashed
I have never been
is in some way purer than other forms of mathematics.
tell the difference between the so-called pure and the nonpure and don't
has its origin in a real
that there is any. Just because my mathematics
doesn't make it less interesting to me?just
the other way around, I find it
305

the puzzle I am working on all the more exciting. I get satisfaction

out of
I find that just as much
that I'm working
on a relevant
knowing
problem.
mathematical
ingenuity has to go into solving problems from a newly developing
area as from some old so-called pure math area.
are now being applied to almost every aspect of human activity.
Computers
name it?is
Every field of science, of medicine, engineering,
business?you
being
in some way. However, before you can put a problem into a computer
computerized
and efficiently find a solution, you must abstract it. To abstract it, you have to build
a model. Before you start to do anything with a model, you have to mathematize
it.
It is this process of abstracting applications
from every aspect of life that has given
rise to a vast new world of mathematics
that is being developed outside mathematics
This mathematics
is just as interesting and just as exciting and just as
departments.
as any mathematics
that is taught in the regular courses. Most mathe?
challenging
maticians
remain completely
unaware of it.
makes

MP:
One scenario I've heard a few times of late goes like this. Given the
innate nature of mathematicians
and given that there are all kinds of mathematics
outside mathematics
value is
stuff, where the applicational
departments?important
often very apparent to students?we
departments
may see mathematics
eventually
take on a role similar to that of philosophy
departments.
Dantzig:
Possibly. However there will always be a few core courses that need to be
taught: algebra, matrix theory, calculus, analysis, and some topology, so there will
for mathematics
to do even if they continue to be
always be something
departments
indifferent
to the developing
new areas of mathematics.
MP:
operations

Do you feel that mathematicians

research?

are biased

against

the people

who do

I myself have never experienced

Dantzig:
any bias. I believe I am well accepted as
In addition, I've got the right "union card," a Ph.D. in mathe?
a mathematician.
matics from Berkeley.

Discovering
MP:
programming

Can you remember

model?

the

Linear
your

Programming

Model

when

first proposed

feelings

you

the linear

To tell the truth, it was a very gradual awakening.

When the planning
Dantzig:
for the Air Force, the very notion of an objective
was first formulated
problem
the idea of a sharply defined goal, was nonexistent.
Of course we paid lip
function,
service to the concept of a goal. In the military setting I often heard it said, "Our
goal is to win the war." In a business setting one would hear, "Our goal is to make a
between the stated goal and
profit." But you could never find any direct relationship
the actions to achieve the goal. If you looked closely at the next step, you would find
that some leader in his conceit had promulgated
a bunch of ground rules to guide
the way to the goal. This is a far cry from honestly
looking at all alternative
of actions across the board and picking the best combination.
combinations
Those
in charge often do a hand-wave and say, "I've considered
all the alternatives,"
but
this is so much garbage. They couldn't possibly look at all possible combinations.
Before 1947 the possibility
that there could be a tool like linear programming
that
would enable one to examine millions of combinations
There
was inconceivable.
was no algorithm or computational
tool for doing so.
I didn't discover the linear programming
model all in a flash. It evolved. About a
whole year was spent deciding
whether my model could be used to formulate
306

as you know, were carried

practical scheduling
problems. Planning and scheduling,
out on a vast scale during the war. Running the Air Force was the equivalent
of
of a whole nation. Hundreds
of thousands
of people were
running the economy
involved in the process. The logistics were on a scale that is impossible
to convey to
an outsider. My colleague Marshall Wood and I reviewed thousands
of situations
drawn from our wartime experience.
The ground rules used in planning
were expressed
in a completely
different
format from the way we now formulate a linear program. What we did was review
these rules one by one and demonstrate
that almost all could be reformulated
in linear programming
format. Not all. In some cases discreteness
and
acceptably
also had to be taken into account.
nonconvexity
When I first formulated
model, I did so without an
my linear programming
I
function
or
for
a
while
with
objective
goal.
struggled
adding ground rules for
from the feasible solutions one that was in some sense "optimal."
But I
selecting
soon abandoned
this approach and replaced it with an objective function
to be
The model I formulated was not specialized
maximized.
to the military. It could be
all
to
kinds
of
one
had
to
do was change the names
applied
planning problems?all
of the columns
and the rows, and it was applicable
to an economic
planning
problem or to an industrial planning problem.
The general model was one which I assumed economists
had looked at and for
which they had developed solution techniques. Albert Kahn of the National Bureau
I visit T. J. Koopmans
of Standards
in
at the Cowles Foundation
suggested
I did so in June 1947. Koopmans
to my
at first seemed indifferent
Chicago.
but then he became very excited?it
was as if, in a lightning flash, he
presentation,
its
to
economic
saw
One
reason why linear program?
suddenly
significance
theory.
ming caught on so quickly outside the military can be traced back to the realization
in 1947, that a good part of economics
could also be translated into
by Koopmans,
the linear programming
format. Incidentally,
became the leader of a
Koopmans
brilliant group of economists
who developed
the theory of allocation
of resources
and its relation to linear programming.
in 1975 when he received
This culminated
the Nobel Prize.

Koopmans, Dantzig, and Kantorovich. Koopmans and

Kantorovich shared the Nobel Prize for economics in
1975. Dantzig's development of linear programming
was fundamental to their work. Koopmans expressed
regret that Dantzig was not named to share the honor.

307

Dantzig receiving the National Medal of Science from President Ford in 1975.

MP:
Do you think that economists
the significance
of linear programing?

other than Koopmans

would

have spotted

Hard to say. Economists

had been developing
economic models for over
Dantzig:
in spite of the fact that their
two hundred years without realizing its importance
field began with a linear model proposed by the Technocrats
back at the time of the
model of the
French Revolution.
It was a rather poorly formulated
input-output
of the peasant, the artisan,
Leontief type with various economic sectors consisting
the approach, economists
over the next
and the nobility. But instead of developing
created
more
more
nonlinear
models.
Walras's
hundred
and
years
sophisticated
model. From the
model, for example, was a very general nonlinear programming
is an anachronism.
It should have been
historical point of view, linear programming
the model that played a central role in economic thought from the beginning rather
came about because
than emerging at a late date as a throwback. The anachronism
to
until very recently mathematical
models were not being used by economists
answers. They were used instead as a convenient
substitute for
obtain quantitative
logical verbal argument. Leontief was the first economist to break away
long-winded
model
from this classical use by constructing
and solving a large scale quantitative
based on real data.
MP:

How

did the discovery

of the simplex

algorithm

come about?

in early 1947 that the

I am coming to that. I learned from Koopmans
Dantzig:
didn't have an algorithm, and that was bad news. The generals in the Air
economists
Force were paying us to solve real planning problems. By hook or crook, we were
expected to find a practical way to solve them.
that the
I set out in the summer of 1947 to invent one. I began by observing
we
could
set.
feasible region is a convex body?a
Therefore,
polyhedral
improve by
moving along edges from one extreme point to the next. But this procedure seemed
as a
In three dimensions,
the region could be visualized
inefficient.
hopelessly
diamond
with faces, edges, and corner points. In the cases of many edges, the
procedure might wander along improving edges for a long time before reaching the
optimal corner point of the diamond.
308

MP:

So that was your geometrical

representation?

Yes. There was nothing novel in the procedure.

Dantzig:
Any mathematician
I
but would immediately
would consider it as a possibility
discard it. So obviously
initially rejected the idea. I next looked at the problem using the geometry of the
columns instead of the rows. Curiously, in the column geometry the algorithm I just
described
looked efficient. I found it extremely difficult to create a problem in m
and
variables which I couldn't solve in m pivot steps; that
n non-negative
equations
is, in m moves along edges.
At first I thought that the method might be efficient but not necessarily practical.
as
For a big problem there could be many combinations
(corner points)?perhaps
many as the stars in the heavens. It might require a million steps to solve it. That
efficient, since this number is small relative to the number of
might be considered
but hardly practical. So I continued
to look for a better
combinations
involved,
alternative
algorithm.
sent one of his students, Leonid Hurwicz, to see me.
That summer Koopmans
Leo and I kicked around an idea we called "climbing up the beanpole," which was a
precursor of the simplex method. It assumed the variables summed to unity. Later I
the procedure by getting rid of the convexity constraint.
My branch at
generalized
with
it.
the Pentagon
We
looked
around
for
some
small
examples to
experimented
solve. One of them was a nutrition problem of George Stigler's. This problem
to be solved by the
became
famous
because it was the first practical problem
method.
simplex
Von

Neumann

to Dantzig:

"Get

to the

Point."

with the
That fall, while my group at the Bureau of Standards was experimenting
von
I
with
and
to
consult
the
Neumann
decided
"great" Johnny
simplex algorithm,
He was considered
see what he could suggest in the way of solution techniques.
by
in the world. On October 3, 1947, I visited
many to be the leading mathematician
him for the first time at the Institute for Advanced Study. I began by explaining the
model in terms of activities and items and so
formulation
of the linear programming
it to him as I would describe it to an ordinary mortal. He
forth. I described
in a way which I believe was uncharacteristic
of him. "Get to the point,"
responded
I said to myself, "Okay, if this man wants a quickie, then that's what
he snapped.
he'll get." In less than a minute I slapped the geometric and the algebraic versions of
He stood up and said, "Oh, that."
my problem on the blackboard.
For the next hour and a half he proceeded
to give me a lecture on the
mathematical
theory of linear programs. At one point, seeing me sitting there with
and my mouth open (after all, I had searched the literature and
my eyes popping
found absolutely
nothing), he said, "I don't want you to think that I am pulling all
a magician.
I have just
this out of my sleeve on the spur of the moment?like
a book with Oscar Morgenstern
on the theory of games. What I
recently completed
that the two problems are equivalent. The theory that I am
am doing is conjecturing
outlining for your problem is an analogue to the one we have developed for games."
That was the way I learned for the first time about Farkas's lemma and the
duality theorem.
On another visit to Princeton in June 1948 I met Albert Tucker. Soon Tucker and
his students,
Harold Kuhn and David Gale, began their historic work on game
and duality theory. Twelve years later AI Tucker,
theory, nonlinear programming,
of my book
and
who had been reading the manuscript
Linear Programming
asked me, "Why do you ascribe duality to von Neumann and not to my
Extensions,
group?" I replied, "Because he was the first to show it to me." "That is strange," he
309

has done."
said, "for we have found nothing in writing about what von Neumann
"True," I said, "but let me send you the paper I wrote as a result of my first
I sent him the report I wrote for my Air Force
with von Neumann."
meeting
dated 5 January 1948, which contains
branch, "A Theorem on Linear Inequalities,"
far
I
formal
as
the
first
of
(as
know)
duality. Later Tucker asked me, "Why
proof
it was not my result?it
didn't you publish it?" I replied, "Because
was von
All I did was to write up, for internal circulation,
Neumann's.
my own proof of
had outlined to me. It was my way of educating the people in
what von Neumann
in
office
the
as the originator of
Pentagon." Today everyone cites von Neumann
my
of the
the duality theorem and credits Tucker, Kuhn, and Gale as the publishers
first rigorous proof.
MP:
in contact

Von Neumann
with.

apparently

made a strong impression

on anyone

he came

Yes, people would come to him because of his great insight. In the initial
Dantzig:
atomic physics,
of a new field like linear programming,
stages of the development
his
advice
invaluable.
After
these
or
fields were
whatever,
computers,
proved
in greater depth, however, it became increasingly more difficult for him to
developed
make the same spectacular contributions.
I guess everyone has a finite capacity, and
Johnny was no exception.
"The

Simplex

Algorithm

Worked?I

Could

Stop

Looking"

MP:
better

So while you were off seeing von Neumann and trying to come up with a
with the simplex
algorithm,
your group at the Air Force was experimenting
had
weren't
about its useful?
that
them?
You
very
optimistic
algorithm
you
given
ness?

That's right. As I said, I thought the method might be efficient but not
Dantzig:
to look for a better algorithm. About a year later, in June
practical so I continued
me
to look elsewhere when the simplex
asked
1948, my group
why I continued
was
out
so
well
on
the
test
problems.
working
algorithm
MP:

So it was completely

unexpected?

Yes. Most of the time it solved problems with m equations in 2m or 3m

Dantzig:
was truly amazing. I certainly did not anticipate that it would turn out
that
steps?
at the time with problems
in higher
to be so terrific. I had had no experience
and I didn't trust my geometrical
intuition. For example, my intuition
dimensions,
from one
told me that the procedure
would require too many steps wandering
In
In
it
few
intuition
in
vertex
to
the
next.
takes
one's
brief,
steps.
adjacent
practice
higher dimensional
space is not worth a damn! Only now, almost forty years from
the time when the simplex method was first proposed, are people beginning
to get
some insight into why it works as well as it does.
MP:
done

A fellow at Bell Labs, Narendar Karmakar,

new with linear programming.
something

has been

reported

to have

on the theoretical result of Kachian that

It is an important improvement
Dantzig:
time. Kachian's
a linear program can be solved in polynomial
theorem states that
the computational
time is guaranteed to be less than a polynomial
expression in the
dimensions
of the problem and the number of digits of input data. The bound is

310

extremely
high, hence not a practical result. We will just have to wait and see if
in practice to the
interior algorithms,
such as Karmakar's,
will prove competitive
simplex method for general linear programs. I would not be surprised if it turns out
to be an efficient way to solve problems with special structure such as multistage
problems.

Any

"Apparently

Form

Can

of Government

be

Made

to Work..."

I notice that John D. Williams is one of the people whom you include in
MP:
How did he figure in your
of your book on linear programming.
the dedication
career?
In 1952 I left the Air Force to work for RAND. John was my boss. After
Dantzig:
I had worked for several months without receiving any direction, I went to see him.
to do?" He didn't say a word, not
I said, "John, what is it that I'm supposed
desk. Five minutes passed, and I
across
his
me
from
at
sat
one?he
looking
just
began to get uneasy. Still not a word. Ten minutes passed. Finally he said, "George,
I understood
what he meant and got
you know better than to ask that question."
out of his office fast. John's policy was to let his researchers do their thing. For
example, he tolerated me for nine years while I wrote my book. Of course, I also
wrote a lot of papers during the same period.
was
at RAND
Division
chart for the Mathematics
Williams's
organization
and
was
remarkable
The
research
to
horizontal.
output
anyone.
Nobody
reported
is
a
worldwide
RAND
the
however,
My
impression,
reputation.
Corporation
gave
came. Ray
never knew from whence the reputation
that its top administration
Olaf
Fulkerson,
Lloyd Shapley, Richard Bellman, Ted Harris, Selmer Johnson,
mad
like
all
were
name
but
a
to
few,
papers
producing
Helmer, George Dantzig,
and doing so without any direction whatsoever.
During that time, network flow
theory was developed by Fulkerson, game theory by Shapley, dynamic programming
by Dantzig. It was a complete contrast to the
by Bellman, and linear programming
was
Air
in
Force. In the Pentagon
the
with
worked
had
I
everything
group
Even
fashion.
in
the
from
came
down
Orders
so,
military
top
vertically.
organized
effective. I find it amusing that there
and remarkably
we were highly motivated
the
could be these two very different ways to organize research?one
anarchistic,
form
of
and yet both highly efficient. Apparently
other dictatorial,
government
any
can be made to work if the people are motivated enough.
MP:

What

caused

you to leave RAND

and return to the academic

world?

Dantzig:
My leaving had to do with the way we teamed up to do our research. In
I was part of a team with Ray Fulkerson and Selmer Johnson. For a
the beginning
time we did great things together. Then after a while, although we remained good
at that time had about
friends, each of us got busy doing his own thing. RAND
for
which
is
a
Mathematics
in
the
Division,
doing just mathematical
big group
thirty
research. There were no new people being hired to work with us as disciples.
up against

the experience

Yes, there must be change, dynamics.

Dantzig:
and working closely with researchers elsewhere.

My stimulus

MP:
senior

So you need young

people?

minds

and expertise
comes

of the

from students

311

The Simplex Algorithm

The Simplex Algorithm is the standard method of solving linear programs. A linear program seeks
to maximize (or minimize) some linear functions, subject to a system of linear constraints.
Example of a Linear Program. A factory can manufacture chairs and tables. Let jq be the
number of chairs produced and x2 the number of tables. Chairs sell for $40 apiece and tables for
$200 apiece We seek to maximize our sales income, subject to limited supplies of raw materials.
The inequalities below describe the production requirements and constraints on these materials:
Maximize z = 40jcx+ 200x2 subject to xx > 0, x2 > 0, and
Wood:
Labor:
Braces:
Upholstery:

xx + 4jc2<1400
2xx + 3jc2<2000
jc1+12jc2<3600
<1800
2Xl

(1)

The model (1) is called a linear program. The Figure below shows the feasible region (shaded area)
of points (x,x2) that satisfy the inequalities in (1). It is straightforward to show that a linear
objective function assumes its maximum value at a corner of the feasible region (assuming the
feasible region is bounded).

2xx= 1800

>-.vi

300
500
400
600
Feasible
regionof linearprogram

Labor
Upholstery

The Simplex Algorithm developed by George Dantzig starts at the origin (0,0), which is a
corner of the feasible region, and moves along boundary edges from corner to corner increasing
the objective function until a maximizing corner is reached.
to convert tne
The algorithm first introduces nonnegative slack variables x3,x4,x5,x6
constraint inequalities to equations:
Maximize z = 40^ +200x2 subject to xx > 0, x2 > 0, x3 > 0, x4 > 0, x5 > 0, x6 > 0, and
Wood:
Labor:
Braces:
Upholstery:

312

jci+ 4x2+x3
+x4
2*!+ 3x2
Xi+12jc2 +
+*5
2xx
+x6

=1400
=2000
=360?
= 1800

(2)

A vector x = (xl,x2,x3,x4,x5,x6)
satisfying (2) is called feasible. It can be shown that a
corner of the feasible region of (1) corresponds to a feasible vector x in which two of the six
variables are zero. For example, the corner where the wood and braces constraints intersect
corresponds to x3 = x5 = 0. We can recast (2) into matrix or tabular form:
Maximize z = 40;^ + 200x2 subject to x,-> 0 (1 < /' < 6) and
x2
4
..3
?12}
b
-200

-40

x3
1
0
0
0

x4
0
1
0
0

*5
0
0
1
0

X6
0
0
0
1

1400
2000
3600
1800

(20

In linear algebra terminology, the columns for the slack variables are a basis for the column
space; variables associated with the basis columns are called basic. Setting non-basic variables xx
and x2 equal to 0 in (2) gives us a feasible corner (the origin in (1)) because setting xl = x2 = 0 in
(2) makes the basic variables equal to the right-side values which are all positive.
The Simplex Algorithm repeatedly performs a change of basis to move to new a feasible corner
(at which the non-basic variables are set equal to zero). The variable that enters the basis (and
becomes positive) is the variable with the largest positive coefficient in the objective function. In
(2), x2 would enter the basis. Keeping xx = 0, we increase x2 (which increases the objective
function) until a basic variable is forced to zero. Since x5 = 0 when x2 = 300 in the braces
equation, the variable x5 leaves the basis. We can rewrite (2) in terms of the new basis by
performing a Gauss-Jordan pivot on the coefficient 12 of x2 in (2'). The new system is:
Maximize z = (y)xx - (y) jc5+ 60000 subject to xt > 0 (1 < i < 6) and

12
2

*2
0
0
1
0

*3
1
0
0
0

x4
0
1
0
0

0
0
0
1

200
1100
300
1800
1

(3)

60,000

Setting xx = x5 = 0 yields a new feasible corner of (1), the corner where the braces constraint and
the line xY= 0 intersect. From (3), we see that the objective function z = (y)xx -(y)*5 +60,000
is 60,000 at this corner.
The Simplex Algorithm continues with xx entering the basis (the coefficient of xx in the
objective function is positive). Since the smallest ratio of rightmost column entry divided by xx
entry occurs for row 1, we perform a Gauss-Jordan pivot on |. Hence, jc3 leaves the basis. This
yields:
Maximize z = - 35x3 - 5jc5+ 67,000 subject to jc,-> 0 (1 < i < 6) and
*2
0
0
1
0

*3

1
0
0
0
0

35

x4
0
1
0
0
0

*5
_ 1
2
5

300
575
275
1200

0
0
0
1
0

(4)

67,000

Since all coefficients in the current objective function z = ? 35jc3? 5xs +67,000 are negative,
the corner point where x3 = x5 = 0 (the intersection of the wood and braces constraints in (1)) is
the optimal corner. Therefore, the maximal value of the objective function is 67,000.
Alan Tucker, SUNY at Stony Brook

313

and

Politicians

Linear

Programming

to another topic. In an article about you in the Stanford

Campus Report, you say that policy makers often ignore powerful analytical tools.
Are you optimistic
about modeling playing a larger role in the future?
MP:

Let's

turn

I keep hoping it will. I have been involved since 1975 in the development
Dantzig:
which stands for Planning
of a macroeconomic
energy model called PILOT,
Level Over Time. It contains a lot of detail bearing on energy, such as
Investment
devices in house?
energy conservation,
energy supply, industrial use, energy-saving
holds, and so on. Recently we expanded the detail in the economy part of the model
in order to estimate the impact of innovation,
and foreign competi?
modernization,
tion. It is a long-term model useful for analyzing trends forty years into the future, a
tool that provides a tremendous insight into complex dynamic issues which face the
nation.
In spite of the fact that the PILOT model is the real McCoy?a
powerful tool for
makers do not line up to use PILOT or, for that
making policy decisions?decision
matter, any other model. Decision
society is a haphazard,
making in a complex
and undisciplined
unstructured,
process that doesn't lend itself to effective use of
models. Policy makers, instead, look for quick answers to very complex questions?as
maker is in a
a result the decisions
they make are bad. Even when a decision
to
use
the
numbers
models
as
for action,
produced
by
planning
guidelines
position
he is reluctant to do so because the policies produced by models are never the whole
answer. A model may help one to decide the best place to put a new airport, but
farmers or some other group, who
then something
unexpected
always happens?like
have not been considered
or even thought of when the model was formulated,
coming forth with objections.
Politicians
know the unexpected
so they tend to ignore the
always happens
models and engage in an ad hoc, haphazard decision process instead.
MP:
refinery

Why then
scheduling?

have

linear

programming

models

been

so

successful

Dantzig:
determine

A refinery is typically headed up by one person so he can use a model

what crude oil to buy and what to produce with it.

MP:

But you are not optimistic

about more complex

for

to

applications?

No. Not at all. Many enterprises such as a nation are so complex that no
Dantzig:
one is really in charge. It is here that our models have the greatest potential
for
and making a real contribution
to the national well
coming to grips with complexity
being. But this potential is frustrated by the lack of structure and discipline in the
decision-making
process itself.
bottleneck
to the effective
The approach I favor for addressing this fundamental
structure for
of models for national planning is to develop a disciplined
application
that will make use of
dialogue between all the special interest groups involved?one
a coordinating
group whose job is to facilitate the bargaining process by supplying
of proposed compromises
and
data from models on the feasibility and optimality
trade-offs.
MP:

314

Thank

you very much,

George.

That's

a good note to end on!