AN ENVIRONMENTAL MODEL OF RISK IN CONSUMER CREDIT Bernie
Grablowsky'
THEORY DEVELOPMENT As behavioralists
in many fields
knowledge
of man and his
researchers
in several
or synthesized (14:10-14) of datd
on behavioral
and the
problem
behavior that the
adaptation
ledge
of consumer
been devoid
of
place.
research
various
behavior;
of any integrated
yet,
the
(3:29-33)
gap between
narrowly
defined models
is now some evidence in the
received the
integrated
to more general
and consumer have
field
theorization
our
environment,
theories.
specific,
There
advanced
his
have successfully
to bridge
findings
of marketing
have
with
application
operations
of
Many areas
into
research
of their
business
study
of business
have begun
in the market
both
areas
new knowledge Theorists
of
interdependerce
theory
cases and agreement from
sciences.
benefits
(27:27-32)
of our
of consumer of its
plethora
of consumer
can benefit
behavioral the
(12:22-27) the
credit
subject
knowhas
matter.
Although the marriage of research and practice, taken place, the hoped for offspring of theory
in some cases, has has not previously
However,
an original
been forthcoming.
this
paper
ceptualization of a viable theoretical credit. The proposed model is based 'Bernie Old Doninion
Grablowsky University,
now presents model of risk on a concept of
is an Assistant Professor Norfolk, Virginia.
con-
in consumer credit character
of Business
Management
at
108
and is developed from
finance,
combined
from
a synthesis
marketing,
with
of knowledge
psychology,
industry
practice
on consumer
and sociology.
and knowledge
to
credit
This
is then
produce
the
suggested
model. The theory the concept,
and its
which
credit
scoring
Risk
Measurement
model,
The problem
transaction.
based
and profitably
evaluate
is an application awarded credit
Usually, through
attention
quantitatively
research
of personal
(18:799-806) numerical
to which relative by an analysis
This
can serve
based
on a sound
variables
that
produce paper
as the
basic
the
credit based
businessman
a
effectively used today
numerical
information
scores
provided
models
are
by the
concept more
of the consistent for
and
is replete with data type of model. A
a person's
development
underpinnings
use financial
values are assigned in a experience in granting credit
literature of this
influence the
the
he could The method
point of past
a better,
presents
credit
since
decisions.
where
scoring
system possibly
a
(25:327-340)
scoring
risk.
of
in extending to extend
credit
by which
good and bad risks. Credit the inconsistent performance
could
Then
form
own judgement,
to providing
system
to both showing
psychological
first.
in the
businessman
ability
previous
applicants.
these
demographic data manner determined
the
credit
items
involved the
creditor's
has turned rating
(9)
Generally,
risk
of operations
to certain applicant.
the
upon the
gained
Recently,
be advanced tested
has been with
has depended
upon experience
will
be presented.
borrower
credit
profitably
will
of determining
to a particular first
development
has been successfully
basic
sociological
credit prediction
of a general such a credit
and
performance of theory scoring
credit which model.
109 The Need for
Theory
From the three
C's
early
writings
of credit
--
important
concepts.
textbooks
on consumer
in consumer
character,
capacity,
They have credit
credit
today,
and capital
persisted
from
until
in claiming
the
early
1900's
the
--
have
space to the
remained
in most present
time. Typically, applicant's the
capacity personal
lender
however,
and capital
balance
an idea
of the
relatively
easy
to make him appear
a better
obtaining
bureau
a credit
of most of the However, nearly credit
sheet
for
half of checks. To determine
banks
applicant risk
with
the
offering
estimate
The term
"character"
presents
meanings
to different
people.
to supply than
lender
character,
from
very
Because
of the
of the
three
the
any,
validity
has had to information.
because it has different deal depends on the evaluator's and his life of character
experiences. is a very
nature of present evaluative and because of the general
character, C's character
that
additional
is
the most
research
important
in this
area
processes agreement
(4:59), is
will
study reveal
of psychological
it
necessary.
several
accepted
and sociological theories
on character
is,
information By
Theory
A thorough sources
intangible
to credit
seems logical Behavior
is.
concept.
as applied that
It
make these
lender if
gives
Association,
do not
the
limited,
problems A great
This debts.
of accuracy.
Bankers
cards
sociological and psychological background (7:196) Thus, a minimum acceptable level nebulous
degree
credit
erroneous
can check
American
credit
the
he actually
a reasonable by the
an applicant's
make a subjective
to pay his
the
to a study
the (5)
statement.
ability
credit report,
from
and income
borrower's
information
according
are determined
literature formation.
110
As might
be expected,
these
theories
are
closely
related
to one
another and may be viewed with equal amounts of credulity. the development of our credit literature seems to be very aligned
and consistent
of character Relevant are
with
formation. portions
presented
here
what
of the
in order
is called
the
environmental to lay
the
environmental
theory
realistic
sense
Environmental ---
Theory
One of the influences
of the
earliest
on man's
on behavior
foundation
writers
behavior
and molded consisted
groupings
for
its
of his
Man's
groups.
group
vations
was not
importance
profound
(26)
were
He theorized
viewed
to the
social
and his Thus,
as being
face-to-face
behavior
standards
was
of an individual's
membership. first
for
partially of the
Veblen's of sociology, attitudes environment, reference
environmental
investigator
of
social
influences
and, in fact, many of his conclusions could hardly to much of our society today. However, Veblen's were
studies
the
Both so that
an individual's
in general,
of conformity
Veblen
wherein
his
seen to be a result
of
Veblen.
actions
culture
present
and aspired
subject
was Thorstein
peer
extension
be evident.
on the
by society,
or specifically
behavior applicable
will
in the marketplace.
influenced contacts
theory
theory
of character
to a more specific environmental theory of credit risk. theoretical references and empirical tests are correlated the
However, closely
his
because
time
and can be recognized
of the
stimulus
he gave
on be obser-
for
their
to further
subject. model
of man drew
cultural and behavior consisting groups,
on what
anthropology, are viewed of his
and face-to-face
are
now called
and social as being
culture, groups.
psychology.
influenced
subculture, (13:40)
the by his social
fields Man's social classes,
111
A more modern ment can be seen
treatise in the
researchers
indicate
character:
a persisting
produce
a rather
that
ment,
Peck and Havighurst are set into school,
and peer
Family Influences - -~-__ Character appears emotionally outside guide
powerful the
behavior
child's
character
to Peck,
"to
and act,
and mother
learned
Although there or inherited,
passed which
from
mirror
by the
child
but
each as just
tend
and
interaction.
A
character. child
does
According
learns
the
relationship
character
Forces
to shape
to feel
kind
of
with
him."
has been discussion as to whether it does not seem to be a trait but
intimate,
and parents.
they
a parent's degree,
and morally,
to child;
may be
segments:
person
his
(19:177)
character genetically
is
seem to be a trait
may be learned. Evidence
conclusion. though
it
by Rettelheim (21)(22)
is possible
improved
personalities,
treatment.
(19:179)
School
influences
by parent-child
have been in their
parent
and organization
shaped
between
startling
psychologically
father
content
develop-
influences.
formed
may indeed
an almost
although
of character
identifiable
negligible,
previously
as
which
behavior,
Cultural
to be predominately
are not
of study
readily
relationships
family
such a thing
and motives
the
(19)
but
group
is
attitudes
that
culture.
broad,
These
a lengthy
conclude
by the
develop-
and quality
After
three
of character
by Peck and Havighurst.
show there of
kind
inconsistencies.
classified family,
studies
predictable
are
influences
reported
pattern
there
of values
on social study
and by Red1 and Wineman
Many therapists to
teach it
children
requires
support
have demonstrated to have different, exhaustively
long,
this
that
even dramatically
intensive
Influence? An additional
environmental
influence
is felt
in the
classroom
112 as an individual evidence
has not
to whether induced Yet, it that
it
from
been able
merely
schools
to reinforce
discriminate
character
of doubt
character
as
structure
already
and on-going family experiences. empirical evidence and observation
between
maturity
The available
curtain
the
children
who outwardly
who show good character.
show poor
Possibly
seems to be highly
as a result
correlated
with
grades. Peck and Havighurst
possess social
precisely which of their
are
likely
likely
of outside
influence.
Peer ---
Influences
Group
The evidence
upper
of peer
end of
structure,
even after
Usually, behavioral
achievers
character
to the knowledge
in school
scale
also
tend
to
able to use it effectively. in anything they undertake.
goals
character
informal
is, than
influence then, than
group
and do it
independently
the
peer
period
partly,
what
peer their
group
already
a supporting
relationship
peer
group
of
that
is
interaction
j%fundamentalchanges
a lengthy
conclusion
tendencies force
the
their
to produce
and more effective can
antagonism or to assimilate
and they are to do well
whether
strong
parents
character
(19:151)
sufficiently this
poor
instability,
to make them poor
to achieve
is questionable
behind
of
with
intelligence.
have superior intelligence; They have an active incentive They are more
people
to concentrate
native
at the
that
qualities
and inability
regardless Subjects
report
those
system,
rationally,
causal
to adolescent.
to sweep away the
tends
and those
of this,
It
child
by the child's previous is fairly obvious from
character school
matures
group
influence
and,
child
gets
the
forces present. force
are partly,
peer
can be seen acting The peer in the
a force
The reasoning
influences from
clear.
is
in character
influence.
family
less
group
development
earlier because
group. to reinforce
is of
less
a
character.
113
This discussion emphasizing the primacy influences in character development is not to as indicating that the peer group is never a in character development. On the contrary . probable that the peer group might have been skillful guidance of interested adults, as a to change their character . . . (19:141) This
kind
of human influence
character,
although
Asch notes a person's "outward fact,
opinions force"
Riesman
increasingly the
Less
is
developed,
which
or reshaping
by their
peers,
values. the
that
do tend
in effect
character.
make up the
(1:151-162)
(6:430)
individuals rather
to influence In
are becoming than
by parents,
in
(23:4) influence
research
does
has a fairly
of peer indicate
permanent
groups
that
on an adult's
character,
structure
once
throughout
a
life. Consistency
Inter-temporal have tended subjects
new social to maintain toward
memberships
to signs
known about However,
-Character
the
group
of a person's
of their
way of shaping
an uncommon method.
and attitudes,
influenced
character. person's
is
informal
has pointed
definition
firmly
it
that
is a possible
of the family be interpreted formative force . . it seems used, under the treatment agency
life,
studies
on character
to show a stable, reached
adolescence.
and intellectual very
persistently
development
and maintenance
predictable pattern of character once Although many subjects learned
skills
as they
their
and in the mode of
deeply
reacting;
grew older, held that
feelings is,
their
they
appeared
and attitudes character
structure. In short, the ratings and the actual case histories both suggest that whatever pattern of moral behavior and character structure a child shows at ten years of age, he is far more likely than not to display into late adolescence; Both the case and, our belief is, for the rest of his life. records and the ratings which were based on them show that
114
there is room for change in later life; but, . . . they suggest that prolonged deep-going influences would be necessary to effect such a change, and that such influences are not likely to occur in the average person's life. (19:157) The evidence early his
in life
suggests
and once
direction
that
of character
simply
makes him more of
is the
possibility
of
or relationship
on the
is
kind
from
relatively
of
person.
of the
very
childhood
fixed.
Growth
emotional
emotional
changes
begins
early
An exception
due to a deeply
order
Basic
development
passes
growth that
change
child relationship. circumstances.
character
an individual
experience
intensity
in character
to this
of the
may occur
parent-
under
these
CHARACTER DEFINED
less
Character, permanent
which
in general, may therefore structure of an individual's
are reflected
is based
in his
on a cause-effect
dividual's
character.
of social
forms;
In this
concept credit
must
Cole
for
be oriented
toward
provides
is apparently "Character
the
between
context itself,
specific
this
definition
and an in-
character
purposes
credit
a starting speaking
social
This society
man is made by his
of character
However, textbooks
sense
as the more or attributes
and satisfactions.
relationship
in that
For a discussion definition.
drives
be defined personal
is
society.
the product (23:4)
is a satisfactory of this
character.
study
the
An examination
of
point. generally
is an intangible
when he says:
sum of
personal
attributes
. . .
(7:194) Beckman and Bartels The character and moral qualities
state
that:
of an individual which identify
is the aggregate him . . .
Character thus becomes credit character qualities combine to make one conscientious debts. (4:54)
of mental
when these concerning his
115 More specifically, Character comprises those makes him want or intend
which (4:54) With the
these
specific
credit
definitions
model
of
qualities of a credit risk to pay when a debt is due . . .
in mind,
credit
character
we will as it
now proceed
to develop
may be used to evaluate
risk. PERTINENT CREDIT CLASSIFICATION In a description
character,
Bartels
of those provides
qualities
a concise
that statement
VARIABLES make better valuable
credit for
future
reference. Positions of responsibility, trust, professional certification, and mental and physical skill generally engage people with qualities This is due largely to the personal which make for better credit risk. integration, and group development . . . as well as to the continuity, references which they involve. (2:312) debtors through
Willingness to pay is related to the social class of . . . Relations with groups affect credit character the attitudes toward debt which they engender . . .
What occupation may not reveal about credit character other activities of an individual sometimes do . . . Activity with welfare, cultural, educational, political, religious, and recreational groups, while it may not produce good credit character, evidences conspicuous involvement, which is generally compatible with debt responsibility. (2:314) In addition,
Bartels
gives
some evidence
of credit
capacity.
As ability to buy and to pay in a continuing existence is dependent primarily upon continually incoming purchasing power, earnings and the ability to sustain and increase them are of major importance in credit capacity . . . Information concerning earnings alone, however, is insufficient basis for determining personal credit capacity. Factors underlying earnings must be considered. Earnings are a reflection of one's ability and capacity to earn . . .; the type of employ-
116
ment in which one is engaged, occupational position for advancement, . . . industriousness, continuity and attitudes toward work. (2:316) In addition evidence
to these
of the
marketing
researchers
a new vantage
statements
relevancy
of
and opportunity of employment,
of a theoretical
behavioral
data
and psychologists
is
nature,
empirical
now available
explore
consumer
as
credit
from
point.
Plumner
has investigated
life
styles
and credit
card
usage
through Activity, Interest, and Opinion (AIO) research. (20:35) Life style research is designed to indicate the difference between heavy
users
and light
purposes.
The heavy
profile. light
In this users A wide
in life
membership,
and light
study
and community.
Opinions
were
the
study
of who uses
credit
questions
rather
indicates
than
a movement
others
and opinions
travel, areas
is basically cards
work,
style
heavy
and
are covered as club
and entertainment.
as interest
in home,
as economics,
politics,
the
indication
of a definite
chological
variables
the
family, and
factor
profiles
significant
variables,
relationship card
credit
was used for
between usage.
card
income
users.
to variables groups
by Plumner
and since analysis,
and and
similar there
sociological Further
AI0
variables
are similar
of low
analysis
segmentation
style
as independent
of classifying
up as significant and credit
marketing
use of life
used by Plumner
to build
Since
keep turning
now.
a descriptive
and why,
demographics
variables
researchers
to determine
is presented
into
such activities
in the method
bankrupts.
variables
market
interests,
such
segmentation
by some life
measured
on such topics
pure
The independent used by other
the
market
investigated.
Although
personal
for
are identified
organization, include
usually
segregated
Plummer
represented
business
users
Plummer
research. community
Interests
of a product,
of credit cards. range of activities,
style
study
users
evidence
is an and psyof this
117 According man is simply
toMartineau, economic models of man assume a rich a poor man with more money. Economics overlooks psydifferences between individuals which may result from
chological different
social
class
Martineau differences
memberships.
suggests between
that
past
different
the use of money. (15:122-123) richer dimension than income As a result of past work Martineau has identified Dolphin,
held by the bankrupt's linked to the choice difficulty. tudes the
project.
and little
study
desire
found
and relatives as an escape of the
touched
of
personal
the
troubles,
of desire
study
evidence
Goble
lack
to take
the
attitudes
debtor's
not
for
low
the
atti-
focus
of
a credit He built
behavior
on a combination
of biographical,
aptitude,
tests.
factor
Goble
number
low income
user reported
and Matthews
of factors of
credit. by Plumner
in their
major
bankruptcy
that bankruptcy screened.
a large
based
analysis,
money
or budgeting,
Matthews concluded could be effectively
groups.
bank-
and creditors'
ledge
By using
Other
toward
debt,
found
in handling
rating. planning
toward
in 1967 that income
of thrift action
of financial
to pay or attitude
hypothesized
be obtained
to those
work, classes.
that
Matthews
a good credit
debtor
collection actions. (16:283-284) cases due to attitudinal factors
a smaller
view
were possibly from financial
on and were
brankrupts,
little
to maintain
influencing
were marital lack
only
to have established
factors
bankrupts,
limitations
were
way they
(8:107)
In another rupts
of consumer
of
credit
are psychological
and the
area and his own original traits of various social
peer groups of bankruptcy
Because toward
show there
classes,
Social class, therefore, becomes a class in which to view a person's actions
in this several
in a study
studies
social
profile
and credit was able
could
set of questions know-
to separate
which effectively identified the typical Many of his factors were very similar in his
bankruptcy
life
studies.
style
study,
and by Dolphin
118 The applicability through method type
answers,
the
applicant
would
be unable
as he can now do with
of credit
mation
for
application.
reliability,
A correctly the
few attempts
at credit
classification
attitude measurement seems to hold promise as a possible of classifying potential credit users. The usefulness of this of measure would be even more important if a test could be set
up so that type
of these
worded
necessity
the
presently
Although it
costs
attitude
used
a creditor
bureau
could
bias
his
financial-demographic can check
money to obtain
measure
of a credit
to consciously
credit
possibly
this
infor-
bureau
reduce
reports.
or eliminate
report.
AN ENVIRONMENTAL MODEL OF CREDIT CHARACTER This model
paper
based
earlier, described.
the
A brief several
applications
paper
structure
will
This
somewhat
theory study,
consisting
of this forces.
as they concept paper.
the
to the
an attitude
patterns. flow
present
As noted
diagram
detailed
credit
on character
divergent
concepts,
However, as the
will
be
empirical granting
for
this
formation
each with study
one most applicable
its
the
process
attributes of parents,
which school,
is
apparent
as a foundation
formed
and peer
personal and actions
theories
environmental
of credit character may be built. credit character will be defined
which are reflected in the individual's responsibility as shown by his intent obligations
behavior
interaction
literature
of an individual's forces
for
structure.
of the
has been selected
mental
induced
and devotees.
on which a relevant Thus, for this
foundation
contributions
market
review
basic
model's
beneficial
in a dichotomous
theory
general
A subsequent
and its
reveals
the
on environmentally
only
model
now presents
as that
by environgroups,
concept toward
and
of debt payment
of
come due. of credit
Character
The environmental
character is assumed forces
are
broadly to
describes
be a function
listed
as parents,
the
nature
of environmental school
and
119
peer
groups,
process. his
in that
order
concept
is assumed debts
evidence
in the
character
is assumed
toward
payment
of an individual's
by that
as they
importance
character
of responsibility that
may be represented his
of
The individual's
person's
of debts. concept
intentions
formation
to be mirrored
by
Finally,
of credit
and actions
it character
in meeting
come due.
Mathematically,
the
definition
individual's concept of debt responsibility
may be seen as shown below: individual's character
=f(
credit
L-11
1
where individual's If
we combine
proposals
submitted
for
credit
equation
risk Czl,
credit
character
existing
theory
here
on character,
may be completed.
= f (environmental
on credit the Thus,
capacity
with
functional equation
forces) the
relationship Cl1 now becomes
such that concept = f ( individual's of debt responsibility
;r"e;;l";y;;s
)
individual's ability to pay
+f(
c21
1
where individual's in equation individual's ability to Figure character
1 combines to give
the
concept of debt Cll, and
pay
=f(
responsibility
non-behavioral mental
the
effects
of credit
total
model
relationship
environforces
is given
1
capacity to credit
with
credit risk,
as
-------
peer
groups
Forces
- 4-I
1
1
matters (a) income (b) budgeting (c) responsibility Education (a) mental ability (b) self (c) children Self Image (a) moral values
.-_ from existing theory on credit capacity
f
BEHAVIORAL MODEL OF CREDIT RISK
Figure
character
Credit Character and Capacity
L------------___
r ---
and other members
parents family
,+@
school
r
Environmental
Variables Representing Individual's Concept of Debt Responsibility and Capacity to Pay
121 expressed in equation C21. An examination of the not
now used
a graphic
in credit credit
by credit
risk;
analysts
reveals
evaluation
representation
determining
model
presence
of variables
decisions.
of
influential
the
theory
in their
the
quest
The model provides variables involved in
encompasses for
a reliable
many areas risk
neglected
model.
HYPOTHESES The previous of this
few pages
have
thus
suggested
the
first
hypothesis
paper. Hypothesis I: Sociological be used as discriminating segregating active credit "bad" credit risks.
Hypothesis
I will
test
the
model.
In order
proposed results
model, obtained
its discriminating by using financial
provides Hypothesis variables applicants variables.
the
to test
second
effectiveness
&SOZU~~
scoring
comparison
and psychological data can variables for effectively card users into "good" and
the
relative
of the
effectiveness
proposed
of the
ability will be compared to the and demographic variables. Th
hypothesis.
II: A credit scoring model using behavioral will provide better discrimination of credit than a model based on financial and demographil
TESTING THE MODEL Data
Sources In order
were
collected
response of
to test
from
people
associations credit
from
the
two groups
questionnaire
who had recently for
theoretical
model
of people
with
a Likert
borrowed
home mortgage Group investigations.
of credit
via
risk,
data
a self-administered Group I consisted
scale.
money from
loans and were I received their
savings
and loan
subjected to thorough questionnaires via
122
U.S.
mail.
In addition,
collected
on each
behavioral credit
financial
respondent
questionnaire. card
ultimately
returned,
The 9 unusable
Each member of Group mailed
misunderstanding
Group
data
II
are considered
criteria.
Group
cases
also
to the
risks
II
based
Group
who were
being
counseled
poor
on their for
out
their
of
being
purchase.
or defaulted
questionnaires
and demographic
rating
exclusively
position
one credit
of
these
by most credit
financial
the
Because
credit,
to be composed
least
at the
Columbus.
risks
present at
Financial
sessions.
credit
considered
have many delinquent
on their recent records. Group II members filled counseling
of Greater
both
selected
to pre-select
people
thus
II members
to return were bank.
of over-extending
study.
answers,
or failure respondents
was no attempt
Service
112 were
the
the data
very
or in default
for
incomplete
into
of
is
usable
from
questions,
There
I members,
inclusion
experiences
people
delinquent
study.
Counseling
past
103 were
100 complete
before
consisted
Credit
one or more
of poor
of the
in the the
Consumer
were
I was an active
to Group
resulted
The first
inclusion
or screen
out
and of those
questionnaires
questionnaires. for
data attached
user.
Of 130 questionnaires
obvious
and demographic
by a questionnaire
In most purchases
at one of information
their
of the
types usually requested for credit applications was also available for each subject. Each member of Group II was also an active credit card
user
up to the
soon as the
results
was halted. included
There
time
his
from
100 respondents
was no attempt
in the Group
The Oiscriminant The collected
counseling
II
sessions
were
were available,
to pre-select
the
initiated.
As
data
collection
respondents
sample.
Function data
were
randomly
separated
into
an analysis
123
sample
of 75 observations
from
each of Groups
I and II;
a validation
sample of 25 observations from each group was also formed. in accordance with the split-sample approach for validating equations as reconended The analysis sample multiple
discriminant
analysis
tically significant step of the program adding the
most
second;
explanatory
contributing
power next
ability,
two were
examination
of
their
selection.
Eight
Variable
best
the
model
entered
first;
power
independent further
should
of the
were study. the
entered the equations,
others.
selected,
based
Of the
equations
"best"
clarify
statis-
At each variable
had been entered into produced 36 behavioral
considered
selected
classification
is presented
models.
the
on
An
reasoning
efficiency
with
as equation a minimum
Cal was able validation
discriminant
model
and several
two equations
several
procedure. with the
most explanatory
for
eventually
(11:250-258) via a
behind
Equation
The first Equation
produced
of the
necessarily
was examined
classificatory
examined,
in the
the
each one was not
Each equation
the
which
and so on until all variables This procedure, in effect,
equation. although their
program
and Morrison. was analyzed
equations using a stepwise a new variable was entered,
to the
variable
by Frank, Massey, of 150 observations
This is predictive
power
to correctly classify Only one other sample. as equation
Cal.
zh = eAl - eA2 eA1 + A2
number
Csl.
It
had
of variables.
48 of 50 observations equation had as good a
124
where:
A, = -38.12
+ 4.71X1
+ 4.07X,4
+ 0.07X,,
+ 2.61X2o
+ 2.08X12
+ 3.05X28
+ l.83X34
- O.38X35 A2 = -19.41
+ 3.80X1
t 1.40X14
+ 0.9SXll
+ l.92x2o
+ 0.60X12
+ 2.36X28
+ 0.41X34
+ l.lox35 e = natural The value the
of Zrritical
subject
than
each
the
the
observation,
had been correctly is
be classified
0.00
By examining for
= 0.00.
should
to or less
presented
log
subject
confusion it
of 2.72.
Thus,
if
Zh is greater
as a good risk. is considered matrix
If
to determine
The confusion
than
0.00
Zh is equal
a bad risk.
and posterior
was possible
classified.
as Table
value
probabilities which
matrix
for
observations equation
C31
1. Table
1
CONFUSION MATRIX OF VALIDATION SAMPLE-EIGHT VARIABLE EQUATION
True Classification Group Group
I (Good Risk) II (Bad Risk)
Number Classified As Group I Group II (Good Risk) (Bad Risk) 24 1
2:
Total
I:
125
By normalizing total)
it
will
is
Table
possible
be classified
1 (dividing
to
show the
under
each table probability
any of the
listed
Table
that
Group Group
I (Good Risk) II (Bad Risk)
the
probability
are due to real be tested
of correct classification by using since both samples are of equal size,
classification
is only
50 percent.
results
of using
differences data,
in the the
F-test
two groups
(17:156-163) equation
C31
and not due to chance
is performed.
The hypothesis
to
is: There is no difference and Group II. [31 with
52.09.
is significant
This
a 1 percent have occurred
between
8 and 141 degrees
probability by chance.
at
1 percent.
that the results The hypothesis
that the equation represents To test the significance 3.
1.00 1.00
classification
the
For equation
Table
0.04 0.96
that
of the Ho:
is
Total
of correct
To make sure occurrences
SAMPLE--
of Classification Group II (Bad Risk)
0.96 0.04
There is a 96 percent probability equation C33. By chance alone,
row
2
Probability Group I (Good Risk)
Classification
by its
any observation
groups.
NORMALIZED CONFUSION MATRIX OF VALIDATION EIGHT VARIABLE EQUATION
True
value
the means of Group
of freedom, Thus,
the there
F value is less
I is
than
indicated by Table 1 could is rejected and the conclusion
a true difference in the two groups. of each variable in the equation see
126
Table
3
F VALUES FOR EQUATION VARIABLES-EIGHT VARIABLE EQUATION
Degrees Variable Number
Variable
1
:;: 32: 35
values
that C33
= 1141
Name
the
have a high variables'
are
because
level
of
to the
variables
0.03
4.92 8.11 10.98 43.76 3.67 3.56 6.56 9.68
significance
contributions the
Significance Level
F Value
education budgeting expense planning savings and checking accounts living beyond means increasing income achievements determining course of life
;:
All
of Freedom
0.01 0.01 0.01
0.06 0.06 0.01 0.01
and the
conclusion
explanatory
represent
true
power
differences
is
of equation in the
two
groups. All equation rather
the
preceding
C33
represents
than
We should equation
differences be able
C33
with
Seventeen --._ _-.. Variable -.- -...In addition efficiently variables
true
were
made to statistically
differences
resulting
to get
equally
similar
data
from
in Group chance
prove I and Group
occurrences
good classification collected
II
of the
of data
in a similar
that data.
by using
manner.
Equation -___ to equation
discriminate were
tests
required
[al,
between by the
equation the
C41
two groups,
function
zh = eA1 - %2 eA1 + A2
to attain
also
was able
although
to
additional
such a good dichotomy.
127
where:
A, = -69.41
+ 3.03X,
- 0.02X3
+ 2.46X7
- 3.16X,o
+ 1.13X,,
+ 2.07X12
+ 4.92X14
t o.39X,5
+ 3.O6X,g
- 3.02Xzo
+ 3.48X2,
t 5.55Xz2
t 0.96Xz6
+ 1.87Xz8
- O.52X34
t 1.O8X35
+ 3.42X36
A2 = -50.65
- 1.76X,
- 0.69X3
+ 3.51X7
- 2.61X,o
+ 1.84X,,
t O.5OX,2
+ 2.56X,4
+ 0.10X,5
+ 4.35X,g
+ 2.44Xzo
+ 3.17X2,
+ 4.92Xz2
+ 1.40Xz6
+ 0.76Xz8
- 1.65X34
+ 2.49X35
+ 2.72X36 Z
The confusion validation
= 0.00.
critical
sample
matrix
showed
that
two observations
from
the
were misclassified. Table
4
CONFUSION MATRIX OF VALIDATION SAMPLE-'SEVENTEEN VARIABLE EQUATION
True Group Group
Classification I (Good Risk) II (Bad Risk)
Number Classified Group I (Good Risk) 24 1
As Total 1 24
25 25
128 By normalizing
Table
4 the
classificatory Table
efficiency
can easily
5
NORMALIZED CONFUSION MATRIX OF VALIDATION SEVENTEEN VARIABLE EQUATION
True Group Group
I (Good Risk) II (Bad Risk)
Equation
['r]
50 percent
also
Classification
As Group II (Bad Risk)
by chance
Total
0.04 0.96
has a 96 percent
the
classification
1.00 1.00
efficiency
versus
alone.
significance
of the
equation,
the
F-test
procedure
used. There is no difference and Group II.
Ho: For equation 26.51.
with
C41
This Ho is
rejected
Group
I and Group To test
Table
6.
the
These
exceptions.
Equations variables
level
represents
the of
true
F value
I is
significance. differences
in
of each variable have a high
even with
equation
the
itself
four is
in equation
significance variables
still
C43,
level,
highly
see
with
four
of questionable significant
and has
efficiency. C31
these from
means of Group
freedom,
1 percent
equation
significance
the
of
the
II.
variables
discriminating
indicating
at the
and the
However,
significance,
between
17 and 132 degrees
is significant
Thus,
high
of
SAMPLE--
0.96 0.04
expected
To test is
Probability Group I (Good Risk)
Classification
be seen.
and [Q] two subjects
the
other
both
misclassified were
48 subjects.
remarkably
the
same two observations,
different
in the
relevant
129 Table
6
F VALUES FOR EQUATION VARIABLES-SEVENTEEN VARIABLE EQUATION
Degrees Variable Number
Variable
:
of
Freedom
= 1132
Name
Significance Level
F Value
student education ability
2.05 6.73
0.15 0.01
1;
ability spending
4.09 1.01
0.32 0.05
;: 14 :9"
budgeting expense planning savings and checking accounts record of expenses of importance budgeting
9.85 2.89 30.76 4.36 0.50
0.09 0.01 0.01 0.04 0.48
2': 22 26 28
importance living beyondof means credit rating acceptability of bankruptcy secure job increasing income
0.53 1.92 1.87 0.91 6.54
0.17 0.47 0.17 0.34 0.01
3": 36
determining achievements course success in carrying
3.45 6.52 1.79
0.07 0.01 0.18
Of the the
better
variables
Risk
two equations model
and is,
from
on the
average,
Class-Variable Because
of the of risk
equation
relationship class
of life out plans
reviewed
above,
an operational
therefore,
The variables
c41. level,
toplanlearn
than
point
simpler
in equation those
equation
[33
of view.
and easier also
C31
It
has less
to use than
have a higher
in equation
seems to be equation
significance
CLII.
Relationship [33
between
is presented.
was the
best
behavioral
its significant This will help
equation,
a discussion
variables and the prediction to show the importance of the
130 individual's the
environments
difference
sample
the
direction
By comparing variables,
as factors
between the direction
of answer
of each
equation
coefficients,
it
variables
and their
influence
instrumental
1
group's
each net
the
:;: 28 zz
A person's credit
risk;
converse higher
education the
result
credit
is his
presented
credit
in Part
The amount expenses I spent
risk.
Net Result Group I Group II (Good Risk) (Bad Risk)
surface,
this
be raised
risk.
of time
time
11)
t
budgeting
is contrary is the
a person inversely their to the
effectiveness
is
related
better
educated
the the
in agreement influences
spends
expenses proposed of Group
to his
risk.
The
subject
with
the
is,
the
relationship
on character.
on budgeting
related
t
t t t t t t
positively
the
The less This
is
1) is
education,
I on environmental
(variable less
the
the
FOR EACH RISK GROUP
Name
be stated.
of
7
(variable
better
may also
of the
equation
education budgeting expense planning savings and checking accounts living beyond means increasing income achievements determining course of life
::
to
and the
RELATIONSHIP
Variable
for
to analyze
on each Table
Variable Number
scores
question,
is possible
DIRECTION OF VARIABLE
in explaining
distributions.
to his than risk II’s
or planning risk
Group
class. II.
model. budgeting
Group
On the
The question methods.
131
Group
II may have expended
but with
less
competence.
contradictory
results.
additional
more
time
This
is
planning a reasonable
Although it presented below
evidence
their
expenditures,
explanation
for
the
may not be the correct explanation, does tend to confirm this con-
clusion. The relationship to pay all
his
one.
This
indicates
they
cannot
spend
so that
they
expenses
as they especially Group
the
of
The latter
money is
a direct
make so much money that themselves
income range
sufficiently
to cover
explanation
wide
enough class
their
seems more rea-
of incomes
represented
II. control
income
a checking risk.
budgeting
of a person's
risk
budget
from
of the
to credit to live to risk
either
they
money left
in both
related
The ability directly related his
or else
in view
a balance
importance
measure
good risks
come due.
having
12) and his
that
to successfully
is directly
respondent
all,
I and Group
Ability maintain
it
the
(variable
have enough
sonable, in both
between
expenses
and a savings This
ability
one's This
financial
income
(variable of
health. 20) is also
be considered
and willingness
therefore,
evidence
(variable
could
and,
account
is additional
to a person's within class.
and expenses
another
to successfully
budget
expenses. Change
related, a better end result has,
in income
with risk of
income
change
his education, ambition, The ability of a person
directly
related
to education, factors determine related
(variable
to risk. mental
not measured the to risk
28) and credit
risk
are directly
a person whose income has been steadily rising than a person whose income has not been rising.
course class.
ability,
could
be due to the
type
of job
being The a person
and other factors. to achieve his goals (variable 34) is This, like variable 28, could be related
here.
ambition, The ability
of his
life
(variable
perseverance, of the subject 35) was also
and other to independently directly
14)
132 Except [a]
were
for
variable
positively
congruent
with
11,
related
the risk
all
the
significant
to risk
model
class
proposed
variables
and the
in Part
in equation
relationships
I.
The discrepancy
with variable 11 may be more illusory than real; expecially, effective results of the subjects' time spent in budgeting that
is,
his
real
In order
to
time,
rather
than
his
were
expended
time,
if the expenses,
were
measured
directly. probe
the
very
basic
causes
here, a researcher would have to interview more extensive psychological testing than The variables used use. for more basic underlying tested. to probe both
For a credit these
federal
and state
on variable-risk
are
forces.
rarely Present
sponsored,
tends
relationships
are
contained
only
surrogates
of the
individuals
are the
resources
truth-in-lending to legally
is detrimental
to better protect Within the scope
results
possibly
and motives model,
The legislation
but it does serve users of credit.
available legislation,
prohibit to credit
such research
the rights of active and potential of this paper, the above observations sufficient.
I
The most accurate equations possible
study
drives
scoring
underlying
investigations.
Hypothesis
in this
of the
subjects in depth using this author was able to
and most
have been discussed to draw a conclusion
statistically
significant
behavioral
and tested for validity. for the first hypothesis
It is now of this paper.
Hypothesis I: Sociological and psychological data can be used as discriminating variables for effectively segregating active credit card users into "good" and "bad" credit risks. Based on the (equations their
high
hypothesis
results
of the
C31 and C41) with level should
of significance, be accepted.
their
first
two behavioral
accuracy the
equations
of classification
conclusion
is that
and the
first
133
THE FINANCIAL The first order
hypothesis
to test
the
and financial provided
of
second
this
paper
hypothesis,
discriminant
and tested
DISCRIMINANT
for
has now been proven.
the
models,
FUNCTION
comparison
a financial
statistical
In
of behavioral
model
must first
be
significance.
The scoring model used as a base for the proposed financial discriminant equation was obtained from a local company which has had many years has revised
experience
with
and updated
The company's classifying
its
approximately
credit
their
models.
times rode1
applicants.
does
a model
applicable
equations
model
are the
same, not
as a starting excess
financial
used by the
data
the
point
population
on the
Fifteen
subjects
analysis containing
procedure. 13 variables,
behavioral
equations.
functions
which
were
is not
mentioned
coefficient
equations
as
basic model is fairly However, in order to
to the
equation
company
in order
number of questions that possibly also contributed
of has
specific
sample
in
as required produced
and C41.
C31
The resulting exact
decade.
company
this paper, the variable coefficients have been modified by using the same discriminant analysis technique which behavioral
last
a good job
the
Columbus'
proper to say their used in this study.
The company
in the
Since
of metropolitan
it is sample
produce
several
feel card
50 percent
active card users, applicable to the
own scoring
models
representatives their
more clearly
its
values. to eliminate
used
were produced After examination produced Equation statistically
study.
the
the
variables
model
necessity
was required questionnaire
This
was used
of collecting reduced
the
to answer response
and rate.
using the stepwise discriminant of the data, only one equation,
results [s]
Only
An existing the
in this
each subject to the high
now, therefore,
above.
close
was also significant.
to those
one of the
of the
two
few financial
134
z
= eA1 - eA2 h
where:
A, = -79.67
+ 3.8OY, + l.43Y6
t 0.50Yg
t l.07Y,o
A2 = -53.72
+ 1.78Y4
+ 1.98Y7
+ 0.19Y8
+ 0.46Y,,
+ 0.73Y,2
+ 2.22Y,5
+ 2.78Y1
+ 0.29Y2
+ 0.53Y5
+ 1.33Y6
+ 0.31Yg
+ 0.42Y,o
+ 0.06Y14
+ 2.05Y4
•t 1.79Y7
- 0.14Y8
+ 0.32Yll
+ 0.29Y,2
+ 2.47Y15
= 0.00.
Z critical The confusion
+ 0.69Y2
+ 0.67Y5
+ 0.18Y,4
dichotomous
c51
eA1 + A2
matrix
relationship
in Table of
the
8 shows the
financial
Table
fairly
well
defined
data.
8
CONFUSION MATRIX OF VALIDATION SAMPLE-THIRTEEN VARIABLE EQUATION
Number Classified True Group Group
Classification I (Good Risk) II (Bad Risk)
(G%?~i~k) 24 2
As Group II (Bad Risk)
2:
Total
E
135
By normalizing ificatory
Table
efficiency
8 it
is easier
of equation
to view
the
Table
Group Group
I (Good Risk) II (Bad Risk)
It both This
is
0.96 0.08
possible
to see the
from a pure classification equation will classify of the
good risks
of 12 percent
versus
8 percent
is compounded
when the
costs
The null
hypothesis
is
because
of true
discriminating of
chance Ho:
For equation
C51 with
35.81. and the
This
because
of true
is
that
Total 1.00 1.00
of this
as bad risks, the
or a total
behavioral to insure
differences
in the
results
in Group
of each variable
equation
[51
two groups
is
and not
data. between
at 1 percent. the
The error
are considered.
tested
of the
misclassification
equations.
of misclassification again
function,
and from cost viewpoint. of the bad risks as good risks
13 and 136 degrees
differences
The significance
efficiency
is no difference I and Group II
is significant
conclusion
lower
for
occurrences
There Group
As
0.04 0.92
ability 8 percent
and 4 percent
because
SAMPLE--
Probability of Classification Group I (Good Risk)
Classification
class-
9
NORMALIZED CONFUSION MATRIX OF VALIDATION THIRTEEN VARIABLE EQUATION
True
relative
C51.
the means of
of freedom,
the
The hypothesis indicated
I and Group is tested
in Table
F value
8 occurred
II. in Table
is
is rejected
10.
136
Table
10
F VALUES FOR EQUATION VARIABLES-THIRTEEN VARIABLE EQUATION
Degrees Variable Number
of
Variable
Name
0.01 0.01
dependent ofstatus
5.08 2.54
0.03 0.11
1.31 1.61 1.73
0.25 0.21 0.19
loan
4.68 8.42
0.03 0.01
ofwithsecond non-local loan stores with credit jewelers
27.13 1.91 1.90
0.01 0.17 0.30
accounts purpose accounts
two variables is highly
have
children
loan
source ofof second purpose first
but
a high
(equation accurate
hypothesis
of this
a financial
and statistically
equation
CSI) which
and highly than
significance
level.
are both
in dichotomizing paper, model,
that the
(equation highly the
There
are
as
of the
and a financial
statistically
data.
a behavioral
results
[sl)
significant
To prove
model
the
second
is more accurate
two models
must be compared
tested.
COMPARISON OF BEHAVIORAL AND FINANCIAL
and financial
The equation
significant.
We now have a behavioral equation
Significance Level
F Value
14.83 10.72 62.37
weekly telephone earnings status source of first
All
= 1136
type bills ofpaidbank accounts length of time at residence marital number
a whole
Freedom
several
criteria
on which
functions
could
be examined.
the
ANALYSES
efficacy The first,
of the
behavioral
and of the
137 greatest
importance,
classifies in the
subjects sense
contained Relative
that
would from the
in the
be the
accuracy
an independent
functions
validation
with
which
validation
themselves
were
not
their
from
data
sample.
Classificator_y_Efficiency --I----._
into
correctly
independent
built
_
From Table 2 and Table g it is obvious that the eight-variable equation has a slightly better ability subjects
the model
sample;
correct
groups.
The degree
is 8 percent versus 12 percent for On the surface this difference model.
of
behavioral to classify error
equation [57, is interesting,
C31
the
for
equation
financial but not con-
clusive evidence in itself that there is any real difference in the discriminating ability of the two functions. However, by using the Kilmogorov-Smirnov to determine Using the
null
if the
calculated
one function
"D"
by using behavioral
is
is that the
better
there
behavioral function is,
than
the
it
is possible
other. by Siegel
(24:130),
"D"
is
less
proportion of errors C31 and equation Csl. than
1.0.
With
"D"
of significance.
=
less
than
as presented
in the equation
if
Jp
samples,
is:
level
1.36
large
test
is rejected
=
Since
is truly
to be tested
a 1 percent
D = 1.36
conclusion
for
There is no difference resulting from using
hypothesis for
test
Kolmogorov-Smirnov
hypothesis Ho:
The null
two-tailed
J
1.0,
is a real
1.36
5o 625
the
da1
=
c61
0.575
hypothesis
difference
is rejected in the
model rather than the financial therefore, more accurate than
results
and the obtained
model. The the financial
138
function
and the difference
pelative
Cost
investigation
However,
it
is
is also
cost
consequences
of using
data
available
are not
generally good loans
(27:g-13) that
to cover
loans,
for
nearer
of
is
of while
would
if
credit
and he is allowed
only
be the
sold,
C53
ratio
for
in-
not
unlike
equally
misclassifies
misclassified
a greater equation
C31 is
which
misclassifies
basis.
we could
profit
from
the
purchaser
if
research,
the
in addition
The exact
maximization
to this
to
one bad loan.
probably
two functions, have a model
to make credit
At any rate,
expenses
action. for
possible
goods
payments. other
It
four
loans. C31
cost
some infor-
a higher proportion of as a bad risk, he is
and the profit which would have been made on credit However, if a true bad risk is classified as a good
risk of the
Even though is from
cost is
possible
costs.
from
a higher proportion of good risks rather than bad risks. If a true good risk is classified refused credit sales is lost.
other.
profit
on the
academic
the
there
the
to loss
equation
on a relative
the
resulting
equation
Of the
be better
over
reporting
consumer
most purely
to examine
takes
of profit
bad risks.
to optimal It
level.
misclassification
it
losses
2 and g show that
good and bad risks; proportion
the
ratio
types
for
researcher,
relative that
Zaegel the
other
Tables
to this
by Zaegel
Although
stallment
sufficient interesting
one equation
known about
has been reported five
0.01
I
cost mation
at the
Efficiency
The above purposes.
is significant
the will lost
ratio
of subjects depends
a part
of
sale
but
should
seller to the
of profit not
purchases,
this
the also
default
loss the
incur
collection on the
to reject study.
(10:406)
value
credit costs
credit
in each factors
not
entire
on his
profits
on other
will
not
risk
and trans-
class
available
139
Hvoothesis
II
A decision
is now possible
paper.
II: A credit scoring model using behavioral will provide better discrimination of credit than a model based on financial and demographic
Hypothesis variables applicants variables. The evidence
II of this
on Hypothesis
indicates
that
this
hypothesis
should
be accepted
as
true. CONCLUSION The first hypothesis be used as classificatory
tested
using
Several
equations
equations
Two equations results. were
tested
for
significance levels
further
promise
of offering as well
variable
acceptable
levels.
as the
base
general the
equations groups
two best
which question
the
shows they
of questions. behavioral
for
items groups
Table
are contained
0.06
the
results
in the
two
each of the
variables question
questionnaire.
to each equation
best from
was chosen
against
from
11 shows the final
the
levels
included
and the major
on the
have contributed
items
of
had significance
equation
are
equations,
had a level
variables
comparison which
significant
in the
equation.had
contain
equations
and highly
themselves
This
program.
and testing.
significance
obtained from financial models. An examination of the variables behavioral
positive
The individual
with
testing
analysis
An eight-variable
areas
analysis
as each variable
The equations 0.01.
to 0.01.
within
additional
for
discriminant
two functions,
in all
well
paper--that behavioral data can in a credit risk model--was
a stepwise
examined
validity.
0.49
results
to 0.01, for
were
of at least
from
test
from
showed
These
of this variables
and the
used group
three in in
All significant
variables are fairly well distributed The financial and demographic
throughout each group. variables produced only one equation
which
and had high
was both
statistically
valid
classification
power.
140
Table BEHAVIORAL VARIABLES
Equation
Variable Number
EightVariable
1 :: :i 5: 35
SeventeenVariable
1 : 10 :: ii :i I: fi z9 36
Variable
11
USED IN SIGNIFICANT
Name
education budgeting expense planning savings and checking accounts living beyond means increasing income achievements determining course of life education student ability ability to learn spending plan budgeting expense planning savings and checking accounts record of expense importance of budgeting living beyond means importance of credit rating acceptability of bankruptcy secure job increasing income achievements determining course of life success in carrying out plans
EQUATIONS
General Classification Formal Schooling Philosophy II of Life II II Achievements II I, Formal
in Lil
Schooling II II Philosophy of Life ,I II II 0, ,I II II 10 Achievements in Lil 0 II I, lb
141
Since
this
is not the
function
surprising
was based to find
15 variables--l3
variables
this
variable significant risks
behavioral as the
than
in misclassifying had equal is not
the
not
Solely
result
because
bad risks
both
but
a cost
it
inclusion
is
better
The conclusion the behavioral
and a classification
eight as
is contrary usually
involved equations
good and bad risks. than
13
as the
more bad
The behavioral
both
of
variables
this
losses
it
all of
as well
financial
basis,
higher
for
model, nearly
to misclassify
on a cost
rates
equation. test, that from
the
tended
of the
the
subjects
as good risks.
methodology,
15 variable contains
Even with classify
also
misclassification best
equation
nor, were variables.
equation
from the financial Kolmogorov-Smirnov equations
did
equation; behavioral
good risks.
desired
best
to be exact. equation
The financial to the
on an existing
the
that
This
resulting
was, as confirmed by the equations were better basis.
142
REFERENCES 1.
Asch,
S.E. "Effects of Group Pressure Upon the Modification and Oistrotion of Judgements," Group Dynamics. Evanston, Illinois: Row, Peterson and Company, 1953, pp. 151-162.
2.
Bartels, R. Credit Company, 1967,
3.
Management.
"The General Theory Vol. 32, January
Marketing, 4.
Beckman,
T.N.,
New York:
The Ronald
Press
p. 312. of Marketing," Journal 1968, pp. 29-33.
of
and Bartels, R. Credit and Collections in Theory New York: McGraw-Hill Book Company, p. 59.
and Practice. 5.
Brimmer, A.F. "Bank Credit Cards: The Record of Innovation and Growth." Presented at the Annual Seminar of Puerto Rican Bankers Association, San Juan, Puerto Rico, March 29, 1971.
6.
Charters, W.W., and Newcomb, T.M. "Some Attitudinal Effects of Experimentally Increases Salience of a Membership Group," Readings in Social PsychoZogy. New York: Henry Holt and Company, 1952, p. 430.
7.
Cole,
R.H. Illinois:
Consumer and Corrnnercial Credit Management. Richard
0.
Irwin,
Inc.,
1968,
Homewood,
p. 196.
8.
Dolphin, R.O., Jr. "An Analysis of Economic and Personal Factors Leading to Consumer Bankruptcy." Occasional Paper No. 15. Michigan State University: Bureau of Business and Economic Research, Graduate School of Business Administration, 1965, p. 107.
9.
Ourand, D. York:
10.
Frank,
Risk Elements in Consumer Installment
Financing.
National
Inc.,
Bureau
of Economic
Research,
New
1941.
R.E., Kuehn, A.A., and Massy, W.F. Quantitative Techniques for Marketing Decisions. Homewood, Illinois: Richard D. Irwin, Inc., 1962, p. 406.
143 11.
Frank,
R.E., Massey, W.R., and Morrison, D.G. "Bias in Multiple Discriminant Analysis," Journal of .Marketing Research, Vol. 2, August 1965, pp. 250-258.
12.
Grubb,
E.L., andGrathworth, H.L. "Consumer Self-Concept, Symbolism and Market Behavior: A Theoretical Approach," Jowvlal of Marketing, Vol. 31, October 1961, pp. 22-27.
13.
Kotler,
P.
"Behavior Vol.
Marketing, 14.
Lewis,
Models for 29, October
R.J., and Erickson, Systems: A Symthesis," 1969, pp. 10-14.
L.G.
Journal
Analyzing Buyers," 1965, p. 40. "Market Functions of Marketing,
Journal
of
and Marketing Vol. 33, July
15.
Martineay, P. "Social Classes and Spending Behavior," of Marketing, Vol. 13, October 1958, pp. 122-123.
16.
Matthews, H.L. "Socio-Economic Indicators and Attitudinal Determinants of Personal Bankruptcy," AM4 1967 Winter Conference Proceedings, 1967, pp. 283-284.
17.
Morrison,
D.G.
Jownal 18.
Myers,
19.
Peck,
of
"On the
Interpretation
Marketing
Research, Vol.
of Discriminant 6, May 1969,
Journat
Analysis," pp. 156-163.
J-H., and Forgy, E.W. "The Development of Numerical Credit Evaluation Systems," American Statistical Association Journal, Vol. 58, September 1963, pp. 799-806. R.F.,
The Psychology of Character and Havighurst, R.J. New York: John Wiley and Sons, 1960.
Development . 20.
Plumner, J.T. "Life Style Patterns and Comnerical Bank Credit Card Usage," Journal of Marketing, Vol. 35, April 1971, p. F., and Wineman, D. The Free Press, 1951.
Children
Who Hate.
21.
Redl,
22.
Rettelheim, Free
23.
New Haven Conneticut: Riesman, D. The Lonely Crard. University Press, 1950, p. 4.
24.
Siegel,
B. Love is Not Enough. Press, 1950.
Glencoe,
Glencoe, Illinois:
for the Behavioral S. Nonparametric Statistics New York: McGraw-Hill Book Company, 1956, p. 130.
Illinois: The Yale
Sciences.
35.
144 25.
Smith,
P.F. "Measuring Management Science,
Risk on Consumer Installment Vol. 11, November 1964, pp.
26.
Veblen,
R. The Theory of the Leisure Macmillan Company, 1899.
27.
Zaegel,
R.J. "A Point The Credit WorZd,
28.
Zaltman, G. Journal
Class.
New York:
Rating System for Evaluating Vol. 52, October 1963, pp.
"Marketing Inference in the of Marketing, Vol. 34, July
Credit," 327-340.
Behavioral 1970, pp.
The
Customers,' 9-13. Sciences," 27-32.