Whicker (2008) - The Use of Deductive and Inductive Tools in The Study of Politics - A Comparison of Alternate Methodologies

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

THE USE OF DEDUCTIVE A N D INDUCTIVE

TOOLS IN THE STUDY OF POLITICS:


A COMPARISON OF
ALTERNATE METHODOLOGIES

Marcia Lynn Whicker


TWO TYPES OF REASONING

Deductive Reasoning

There are two types of scientific reasoning--a


fact so elementary it often escapes from our con-
sciousness Deductive reasoning requires a premise, a
conclusion, and a justification linking the two.
Sometimes called axiomatic reasoning, invention of
deduction has long been attributed to the ancient
Greeks Euclid's Elements, written around 300 B.C.,
displayed a highly developed system of axiomatic
deductive reasoning. Deductive reasoning is often
mathematical, but where sets of axioms come from in
the first place is not a mathematical question, since
mathematics as a tool of inquiry is concerned with
the consequences and not the origin of axioms Sets
of axioms may be derived from two sources:
idealizations of observed phenomena, such as Euclid's
abstractions of the characteristics of physical space,
and more recently and more commonly, modifications
of previous sets of axioms (Giere 1979; Meyer 1974).
Deductive axiomatic systems should exhibit two
characteristics: consistency, so that it is not possible
to prove both a theorem and its negation, and
independence so that any one of the basic axioms
cannot be removed from the set and proved by
knowing only the remaining axioms Axiomatic
systems may also but not always exhibit a third
characteristic of completeness If every statement
concerning the basic terms can either be proved or
disproved in the system, the axiomatic system is said

201
to be complete. While this characteristic is some-
times desirable, the best known systems from modern
algebra (groups, rings, fields, and vector spaces), are
not complete (Meyer 1974).
Inductive Reasoning

Deductive arguments are attractive, f o r t h e y


are perfect transmitters of not only justification (the
logic) but of truth (the conclusions) a s well. Yet
deductive reasoning alone is insufficient to develop
scientific theory. By contrast, an argument is a good
inductive argument when the truth of its premises
guarantees an appropriately high probability for the
truth of its conclusions One characteristic of
inductive arguments are that they do not always
convey or preserve "truth." A good inductive argu-
ment may have a false conclusion even though all its
premises are true. Otherwise, inductive arguments
would be deductive. A second characteristic is that
inductive arguments are knowledge expanding, so that
their conclusions contain more information than the
premises combined. With the spread of scientific
inductive reasoning, the view that probabilistic
knowledge was not really knowledge, but a makeshift
for the sake of action has been crumbling, partially
due to developments in physics incorporating stochas-
tic uncertainty into that field (Kaplan 1964; Giere
1979).
The Uses of Inductive and Deductive Reasoning
Science involves building theories-sets of con-
sistent coherent propositions or theorems about the
phenomenon of inquiry, and testing those theories
The greatest utility of deductive reasoning in the
scientific process is in developing through the ma-
nipulation of an axiomatic system, new propositions
The logic of the deductive axiomatic system produces
theorems which become propositions in the scientific
theory. Deduction as a reasoning process facilitates

202
theory-building. If theory testing has accompanied
the deductive process so that the axioms and justifi-
cations in the system bear a resemblance t o the real
world, then the derived theorems from deductive
reasoning will be more useful in building predictive
theory.
Ideally each proposition embodies a testable
hypothesis, stating a relationship between two or
more theoretical constructs Through measurement,
each theoretical construct is operationalized into a
variable. The hypotheses may be directional or non-
directional. Hypotheses may also be correlational,
stating merely that two concepts covary, or causal,
stating that covariation in one concept depends on
covariation in the other. While statistical analysis
facilitates rigorous testing of both types of hypoth-
eses, the internal validity necessary to prove causa-
tion must be built into the research design. Only
experimental designs--incorporating a control group,
random selection and assignment to groups, and
experimentor manipulation of the independent vari-
able-are sufficiently internally valid to test causal
hypotheses (Cook and Campbell 1979). Inductive
reasoning, including probability and statistical rea-
soning, is critical in theory testing. Inductive
reasoning involves repetitive studies and 'iests t o
establish the long run probabilities of the dependent
variable (the conclusion of the inductive aqpment),
given the truth of the premises (the presence of the
independent variable s).
The rapid growth of science (commonly known
statistics such a s the fact that of all the scientists
ever living on earth, over half are alive today come
to mind) has heighthened the importance of a previ-
ously less prestigious role: that of theory communi-
cator. Scientific information brokers who convey
theories, methods, and findings from one scientific
field to another are increasingly common and serve t o
faciiitate the dissemination of ideas across fields

203
In political science, such individuals may study
ndditional fields (eg. mathematics statistics, sociolo-
gy, psychology, economics, anthropology, history).
The cross fertilization between political science and
the other social sciences has been considerable, and
the cross fertilization between statistics and political
science has been improving. The interactions be-
tween political science and the life sciences have
been restricted mainly to theories in biopolitics,
attempting t o explore the biological foundations of
behavior, and the interactions with the physical
sciences has been minimal. The type of reasoning
required for this scientific activity is both inductive
and deductive. Expertise is required in the dominant
reasoning procedures and techniques of both or all the
fields being cross fertilized. Additionally, populizers
of scientific findings who communicate with inter-
ested lay audiences are also increasingly common (See
Table 1).

Type of Dominant Compat I b l 0


9cientl f i c Reasoning Methodologleal
ACtlVlty Procraa Tools

Theorv b u i l d i n q Deduct 1on Math


LSymbolltlnq Computer
tFroposi t 1 on st m u 1 a t I on
devel opment

Theorv t e s t l n q Induction Statlstlcr


ttleasnr ement Computer
*Test I n q sl nu1a t I on
r e 1 a t I onshi pa

Theory comnunl c a t 1 on M i xed Verbal


(10 o t h e r
s c i ent I f i c f 1 el d r
* T o the p u b l i c

Methodological Approaches And Tools


Verbal Analysis
The oldest employed methodological approach in
political science is verbal reasoning. During pre-

204
behavioralism, it was the only employed methodol-
ogical approach. Verbal analysis may be both induc-
tive and deductive. Case studies provide an example
of verbal analysis in a predominantly inductive
mode-words are used to capture the richness and
complexity of the case being described (see Table 2).
TABLE 2. THE RDVnNTAGES AND DISADVANTAGES OF CILTERNATE
METHODOLOGICAL TOOLS

Tvpe of Tool RClltivR Merits

Ve rb al Advantaqesr
11) Rcqulres no s p e c i a l i z e d q u a n t l t a t l v e t r a i n g i n g
I 2 1 I s more e a s l l v under s t ood a c r o s s a v a r i e t y of
audiences
13) Can be a p p l l e d b r o a d l y leg. I d l o s y n c r a t l c
cases, e l l t e s t u d i e s . n o n - r e p l i c a b l e e v e n t s )

Dlsadvantages:
( 1 ) Used p r i m a r i l y f o r I d l o s y n c r a t l c case s t u d i e s
12) Does n o t employ r e l i a b l e measurement, d e s p l t e
a t t e m p t s t o use words p r c c l s c l y
13) Research d e s i g n cannot be r i g o r o u s o r
Internally valld
( 4 ) Research d e s l g n cannot be e x t e r n a l l y v a l i d
( l a c k s p r o b a b i l i t y s am pllng pr oc edur es )

Statistical Advant ages:


W l d e l y used and u n d e r s t o o d among t h e
s c l e n t i f l c community
Employed f o r t h e o r y t e s t i n g phase of
s c l e n t i f l c process
C a l c u l a t l o n s a r e s t a n d a r d i z e d t h r o u g h canned
s t a t l s t l c a l programs? o n l y knowledge of u s e
and I n t e r p r e t a t i on r e q u l r e d
I n c l u d e s some enphas ls on measurement I s s u e s
leg. s c a l i n g and c l u s t e r l n g )
More r l g o r o u s t h a n v e r b a l a n a l y s l s
Can b e used t o t e s t b o t h c o r r e l a t i o n a l and
caus al hy pot hes es
Can I n c w p o r t a t e b o t h p h y s i c a l and
s t a t l s t l c a l controls

Dl sadvantagesr
( 1 ) Less u s d u l f o r t h e o r y development
I 2 1 Common a v a i l a b 1 1 I t y t hr ough canned
s t a t i s t l c a l packaqes can produce abuse
as w e l l a s use
( 5 ) C r e a t e s a b l a s t owar d r ? x i s t l n g d a t a bas es
( 4 ) Measurement pr oc edur es f o r some of t h e
most i n t e r e s t l n g c onc ept s r c m a l n e l u s l v e
(eg. u t i l l t y . power1
( 5 ) The more e l a b o r a t e t h e s t a t i s t i c a l t e e h n l q u e
t h e l c r s i n t e r p r e t a b l e t h e r e s u l t s (eg.
f a c t o r analysls. cannonlcal c o r r e l a t i o n
mu1 t l - v a r l a b l e IWIX”01
16) P r o b a b i l l s t l c r a t h e r than d o t c r m l n l s t l c
concluslcnm must be r o p l l c a t o d I n numerous
st udies
171 O f t e n t her e is l a c k of agrccment on mtandard
drslgns, tcchnlquec, and ncasurcmcnts f o r
r r p l l c a t l o n s t udles
.................................................................

205
Mathmaticbl Advantapes:
( 1 ) C b n bc uscd I n thmory b u l l d l n g
( 2 ) Oftmn produccs d c t c r m l n i s t i c coneluslonm
IS) I n c w p o r a t c s p r e c i s l o n bnd rlqor

Dlsadvantaqcsr
( 1 ) D i f f i c u l t f o r many p o l i t i c a l c c l m t l s t s
t o understbnd
(2) Mbv obsurm r l c h n e s s of rcbl w o r l d phcnomcnon
( 5 ) NO s u b c t l t u t c f o r t h m w y t c s t l n q bnd
r e p l l c b t i o n mtudlcs
.................................................................
Computcr Cldvantaqcs I
sIrmlation ( 1 ) M w m f l e x i b l c t h a n mbny mathcmbtficbl t o o l s
12) Can bm uscd f o r b o t h t h c o r y b u l l d l n g and
theory tcmtlnq
1 3 ) Does n o t r c c t u i r e foreknowledge o f t h e undec-
l y i n g s t a t l s t l t c a l d l s t r i b u t l o n of thr
m a j o r dcpcndcnt v r r l b b l c
14) Al1ouv f o r a b a t r a c t c o n t r o l o f f o r c r s b s i l d c
t h e indepcndcnt v a r l a b l c s a c t i n q on t h e
dcpmndmnt v a r i a b l cs
IS) Can I n c o r p o r a t c models o f p r l o r p r o b a b i l l t l c s
and m t o c h a s t i c d i s t u r b a n c e s
( 6 ) Can d y n a m l c b l l y modcl l o n q i t u d i n r l l t c r b t l v e
p r occsscs

D I sadvmtbges:
1 1 ) Clbstractncss makms communicatlon o f a modrl
Mro dlfficirlt
12) Each model m u s t be r i n l q u e l y c o n s t r u c t e d ,
c u r r e n t l y w i t h o u t b c n c f l t o f canncd computer
p r ogr a m s
1 3 ) Modml v c r l f l c a t l o n o f t e n provms C l U 6 i V C

Obviously verbal analysis requires no prior


quantitative training and can be applied inductively in
a variety of idiosyncratic non-replicable settings
Yet as a measuring device, verbal analysis
remains imprecise, often lacking the validity and
reliability achieved by more quantitative measure-
ment approaches As an inductive testing device,
verbal analysis also may be confounded more easily
by researcher subjectivity than a variety of quanti-
tative approaches. And as a deductive reasoning
device-a use of verbal analysis often employed in
judicial decisions and legal brief s--verbal analysis
lacks the simplistic precision and rigor of symbolic
logic and other mathematics. This is partially due to
a slippage in the meaning of verbal terms across
times, despite attempts in fields employing deductive
verbal reasoning to carefully and precisely define key
terms
Statistical Analvsis

206
While the development of new statistical tech-
niques is deductive, using mathematical axiomatic
systems, statistical analysis applied to data is an
inductive approach to testing scientific proposit ions
Statistics have several functions: measurement of
theoretical concepts, addressing questions of statisti-
cal significance in samples t o impute observed rela-
tionships are not manifestations of sampling error,
assessing the strength (association or correlation) and
direction of relationships between theoretical con-
cepts, and prediction. Statistics may be univariate
(one variable), bivariate (two variables), or multi-
variate (several variables); parametric (assume an
underlying probability distribution), or nonparametric
(makes no assumption about an underlying probability
distribution).
Statistical analysis has several advantages
Statistics are increasingly widely used and understood
among political scientists as well as the larger
scientific community. They are most useful in the
theory testing phase of the scientific process The
dissemination of canned statistical programs for use
on mainframes alleviated the burden of calculation,
making the use of statistical analysis dependent only
on familiarity w i t h a prepared statistical subroutine
and data interpretation. The proliferation of ever
more powerful microcomputers with prepared statis-
tical packages will no doubt further accelerate this
trend. Statistics are more rigorous than verbal
analysis and can be employed with probability sam-
pling to obtain external validity and generalizability.
Statistics used in conjunction with random sampling
are necessary and sufficient t o test correlational
hypotheses; statistics are necessary but not sufficient
to test causal hypotheses. While statistics may be
employed on data collected with physical controls on
additional variables impacting on the dependent vari-
able, they may be used t o impose statistical controls
for confounding factors when physical controls are
not possible.

207
Statistical analysis also has several disadvan-
tages: it is less useful for theory building than for
theory testing. Its use in theory building is indirect:
in showing the relationship of premises and derived
theorems in axiomatic systems to the real world, thus
legitimizing further derivations of more propositions
The wide spread availability of statistical packages
has sometimes encouraged the widely known GIGO
factor in analysis (garbage in, garbage out), as the
power of analysis has often exceeded the quality of
the data being analyzed. The spread of statistics
throughout political science has created a bias toward
existing data bases, while measurement procedures
for many political concepts remain unrefined. Inter-
pretation of statistical output is sometimes a creative
art, since often the more sophisticated the statistical
technique, the less interpretable the results, a fact
demonst rated by factor analysis, cannonical correla-
tion, and multivariable ANOVA. Statistical analyses
do not definitively answer questions, but rather
provide one more piece of evidence supporting or
refuting a proposition in a larger theory. Replication
studies are necessary, but often there is lack of
agreement on standard designs, techniques, and mea-
surements for replication studies. Replication is
often regarded as more mundane than innovative
pioneering work.
Mathematical Analysis

Mathematics is largely a deductive analytical


tool. Two fields of mathematics which have some-
what filtered into political sciences are set theory.
(often employed in game theory), and calculus (used
to describe rates of change). Symbolic logic has
rarely but occasionally been used. Mathematical
analysis has been used t o extend statistical tech-
niques. Advantages include the potentially useful role
of mathematics in theory building, the deterministic
conclusions produced from mathematical models, and
the greater precision and rigor embodied in axiomatic

208
mathematical systems Yet mathematics also re-
mains difficult for many political scientists t o under-
stand, requires substantial prior training, may obscure
the richness of the complexity of the real world, and
is not alone a substitute for theory testing and
replication s t u d i e s
Compute r Simulation
Compared t o vel-bal, mathematical, and statis-
tical analyses, computer simulation is a relatively
new technique. Predecessors of computer simulations
were "thought experimentqlVamong the more histori-
cally famous being Einstein's speculation about the
consequences of riding a beam of light through space,
and a classical thought experiment in physics pon-
dering the aging implications of one identical twin
going into space on a rocket while the other identical
twin remained on earth. A computer simulation is a
computer program which simulates concepts and
processes in the real world. Computer simulations
models have the potential for being dynamic-of
simulating multiple iterations of a process t o observe
changes across time. Computer simulations have five
parts inputs (independent variables), outputs (depen-
dent variables), parameters (control variables), as-
sumptions (restricting premises), and algorithms (step
by step decision rules for converting inputs into
outputs).
Simulations can be both inductive and deduc-
tive. The deductive use of computer simulation
involves sensitivity analysis and hypothesis genera-
tions Once a computer simulation model is devel-
oped, prespecified ranges of inputs can be introduced
during sensitivity testing. The range of output is t h e n
linked t o the range of input via testable propositions
When simulation models incorporate random sto-
chastic variation t o simulate real world uncertainty,
propositions are stated probabilistically, rather than
deterministically. The inductive use of simulation
involves comparing simulation output t o real world
209
data. The greater the similarity between the two,
the greater the confirmation of the process al-
gorithms incorporated in the simulation model.

Advantages of computer simulat ion modeling


include greater flexibility than is provided by many
mathematical tools (fewer restrictive unrealistic as-
sumptions), no prerequisite knowledge of the under-
lying statistical distribution of the major dependent
variable as is required by parametric statistical
analysis, abstract control of forces besides the inde-
pendent variable acting on dependent variables (ie.
many of the advantages of an experimental design),
the ability to incorporate models of prior probabilities
and uncertainties, and a dynamic capacity to model
events across time. Disadvantages include com-
munication problems which result from newness of
the technique, its abstractness, and the uniqueness of
each simulation model; the high degree of program-
ming skills required to build computer sirnulation
models; and the difficulty of verifying complex
models

Simulation has spread slowly throughout politi-


cal science. When simulation has been employed, it is
often as a sensitizing or pedagogical tool, rather than
as a research tool t o deductively derive theoretical
propositions or to test causal hypotheses. Exceptions
to this generalization include Cohen, March, and
Olsen's simulation of organizations decision-making
(1972) and simulations of legislative processes (Fran-
cis 1970; Shapiro 1968). Other social phenomena
stimulated include' human thought (Newell and Simon
1971); international relations (Guetzkow 1962); pro-
duction and pricing decisions (Cyert and March 1962);
and presidential behavior (Cole and Wayne 1980).
Massive policy simulations have also been developed
(eg. Forrester's DYNAMO and the Urban Institute's
TRIM models).
210
A SURVEY OF METHODOLOGIES I N POLITICAL
SCIENCE
Journals Examined and Categories of Methodologies

To assess current usage of alternate meth-


odologies, articles in leading political science journals
were categorized by their most sophisticated meth-
odology for the five year period 1978-1982. The six
journals evaluated were the American Political Sci-
ence Review, and the five regional political science

+
association journals: American Journal of Political
Science (Midwest), Journal of Politics (South), Polit
(Northeast), Social Science Quarterly (Southwest , and
Western Political Quarterly (West). These journals
are listed in alphabetical order.
Some judgment calls were made in the catego-
rization. Articles classified as verbal used no
quantitative analysis or data. The category of simple
statistics included measures of central tendency,
proportions and percents, coefficients of variation,
and even siriiple tests of significance on sample data.
The correlation category included nominal and ordinal
measures of association (lambdas, taus, gammas, etc.)
as well as simple and partial Pearson's R s The
multivariate predictive categoly included multiple
regression, probit, and multiple discriminant analyses
Since causal modeling and path analysis incorporate a
series of consistent hypotheses about the interactive
impact of several variables across time (Blalock
1971),- these techniques were placed in a separate
category. This technique tests already developed sets
of propositions and is inductive.

Articles which solely used factor analysis were


) placed in a separate category, but those which
incidentally used factor analysis in order to later use
regression or other multivarible techniques t o test
relationships between several variables were placed in
the multivariate prediction category. Articles which
211
employed already developed but not widely used
measurement or indexat ion techniques or compared
and contrasted several measurement techniques were
separated, as were articles which boldly attempted to
develop new measures of theoretical constructs
Classified as mathematical were articles which used
math to develop new propositions, tried to symbolize
phenomena and events previously communicated ver-
bally, and which laid down a mathematical justifi-
cation for subsequently employed statistical tech-
niques Simulation articles included those with real
data used for inductive purposes and those primarily
deductive which did not.
Findings

The results of this survey are presented in


Tables 3-5. Percents were rounded t o the neamst
whole number. Table 3 displays overall results
TABLE 3. PIETHOWLOSIES EPPLOYED IN LEADIM POLITICAL SCIENCE
JOURN(VS, NUREER OF ARTICLES AND PERCENTS. CUfRJLRTIVE
-I- -----___------------
F I V E YEAR T0TM.S

TYPW O+ # of PUCrntS
Rmthodol OOY art i cl ms

Vwbal 391 (31%)

Total 1282

Table 3 shows that 31% of all journal articles in


leading political science journals employed only ver-
bal analysis. Sixty-four percent employ statistical

212
analysis: 23% used simple statistics, 34% tested
relationships, and only 7% emphasized measurement
(including the factor analysis category). Five percent
were mathematicaL The number of computer sim-
ulation articles was so small that when rounded off,
it came out t o 0%. Despite the fact that simulation
may be used as both an inductive and deductive tool,
almost all of the simulations published in leading
political science journals were inductive. Of the six
simulation articles, five employed empirical data t o
test the fit of a process model, as opposed to using
simulation deductively to derive new propositions
Table 4 breaks out the data by category and by
year.

Slnol. .t.tl.ClC. 72
129%)

Corrolatza 27
110x1

41
(1%)

12
14%)

S
(2%)

F a c t o r .nalyri8 LO
14%)

4
(1%)

6
(1%)

Most percentage of articles in most categories


appears somewhat stable across the five year period.
Exceptions to this are articles which use correlation
alone which appear to be decreasing in number, and

213
articles which use time series, which although still
few in number, appear to be increasing.
Table 5 examines a breakout of the data by
specific journal. Four of the journals (APSR, JOP,
SSQ, and the Western) all approximate the overall
average of 31% of articles employing strictly verbal
analysis The two notable deviations from this field
wide average weE AJPS which was significantly
under the average (6961, and Polity, which was
significantly above the average (58%). The overall
percentage of 23% of articles using only simple
TABLE 5. MTHOWL061ES EPlfLOYED I N LEADING P O L I T I C W SCIENCE
JOURNIILS. NUMBER of ARTICLES AND PERCENTS. BY JOURNa

Sinplm s t a t i s t i c . 32 34 80 44 ss
(IS%) (16%) (32%) (30%) (19%)

Ccrrmlataan

Multivariatm
prmdiction

Causal modml inq 21 12 12 3 E


(10%) (6%) (SX) (2%) (3%)

T i w srri n 12 13 S I 1
(6x1 (6%) (2%) (1%) (0%)

Factor analvsas 9 a 7 1 8
(3%) (1%) (X)
-
(4%) (4%)

Othu n r a s u r a n t 7 7 6 0 3 i

t u h n i qums (3%) (3%) (2%) (0%) (1%) (1%)

D m v r l o p m d nru 3 9 1 2 6 3
maswmmnt (1%) (4%) (0%) (1%) (2%) (2%)
----_------I---

Mathmmatical 27 72 6 0 7 S
(13%) (10%) (2%) (0%) (2%) (3%)

statistics was exceeded b y J O P (32%), Polity (30%),


and the Western (33%); AJPS (15%), APSR (16%), and
SSQ (19%) published fewer articles using only simple
statistics than the overall average.

214
Articles using multivariate predictive tech-
niques were most likely t o be published in SSQ (31%)
and AJPS (22%). By contrast, APSR (12%), J O P (5%),
Polity (l%),and the Western (8%) published fewer
studies using multivariate prediction. Causal models
were published more frequently in AJPS ( l o % ) , while
AJPS (6%) and APSR (6%) exceeded the field in
publishing time series studies. AJPS and APSR were
also slightly more likely to publish articles using
factor analysis o r other measurement techniques
The APSR was somewhat more likely to publish
articles developing new measurements. AJPS (13%)
and APSR (10%) published 49 of the 66 mathematical
articles published in leading political science journals
during the five year period, or three-fourths of the
total. While three general Journals published no
simulation articles, AJPS, APSR, and S S Q published a
negligible number (1% each).
CONCLUSIONS

Only five percent of the articles in leading


political science journals during the five year period
examined published articles which employed deductive
theory building tools Over 64% of articles did
employ some type of theory testing statistical analy-
sis Only 7% of these articles emphasized measure-
ment issues. A substantial 31% employed non-
quantitatively rigorous verbal analysis. Despite the
fact that computer simulation has both inductive and
deductive potential, simulations in political science
articles were predominantly inductive tests of the fit
of process models to empirical data. A negligible
number of articles used simulation, leaving it an
mderutilized tool. The field of political science
remains heavily weighted toward theory testing at the
expense of deductive reasoning and tools and mea-
su rement concerns

215
REFERENCES

Blalock, H. M. Jr., ed. 1971. Causal Models in the


Social Sciences Chicago: Aldine-Atherton.
Burton, R. M.; and B. Opel. 1980. "A Computer
Simulation of the M-Form Hypothesis" Admin-
istrative Science Quarterly. 25: 457-466.
Cohen, Michael D.; James G. March; and Johan P.
Olsen. 1972. "A Garbage Can Model of
Organizational Choice.tr Administrative Science
Quarterly. 1 7 (March): 1-25.
Cole, Richard L., and Stephen J. Wayne. 1980.
"Predicting Presidential Decisions on Enrolled
Bills A Computer Sirnulation." Simulation and
Games 11 (September): 313-325.

Cook, Thomas D.; and Donald T. Campbell. 1979.


:- Design and Analysis
Issues for Field Settings Boston, Mass:
Houghton Mifflin Co.
Cyert, R ; and James March. 1962. A Behavioral
Theory of the Firm. Englewood Cliffs, N.J.:
P rent ice-Hall Inc.

Francis, Wayne L. 1970. Y3irnulation of Committee


Decision-Making in a State Legislative Body."
Simulation and Games 1 (September): 235-262.

Garson, G. David 1976. Political Science Methods


Boston: Holbrook Press
Giere, Ronald N. 1979. Understanding Scientific
Reasoning. New York: Holt, Rinehart, and
Winston.
Guetzkow, H. 1962. "The Use of Computer Sirnula-
tion in the Study of International Relations" in
216
Simulation in t h e Social Sciences: R e a d i n g s
ed. H. Guetzkow. Englewood Cliffs, N. J.:
P r e n t ice-Hall.
Johnson, Richard A.; and D e a n W. Wichern. 1982.
Applied Multivariate Statistical Analysis
Englewood Cliffs, N. J.: Prentice-Hall.

Kaplan, Abraham. 1964. T h e C o n d u c t of Inquiry:


Methodologv for Behavioral Science. S c r a n t o n .
"I

Pennsylvania: C h a n d l e r Publishing Co.


Kemeny, J o h n G.; and J. Laurie Snell. 1972. Math-
e m a t i c a l Models in t h e Social S c i e n c e s
Cambridge, Mass: M a s s a c h u s e t t s I n s t i t u t e of
Technology P re ss

Meyer, Burnett. 1974. An Introduction t o Axiom-


a t i c Svstems.
- ~ ~ Boston. M a s s : Prindle, Weber &
Schmidt, Inc.
Miller, Warren E. 1981. "The Role of R e s e a r c h in
t h e Unification of a Discipline." American
Political S c i e n c e Review. 75 (March): 9-16.

Newell, A ; and Herbert A. Simon. 1971. "Simula-


tion of Human Thought.ff in C o m p u t e r Simula-
tion of H u m a n Behavior. e d s M. D u t t o n a n d
W. H. Starbuck. N e w Yo&: John Wiley.
Orcutt, Guy; S t e v e n Caldwell; and R i c h a r d
W e r t h e i m e r 11. 1976. Policy Exploration
Through Microanalytic Simulation. Washington,
D.C.: T h e Urban Institute.
Paris, David C.; and James F. R e y n o l d s 1983.
T h e Logic of Policy Inquiry. New York:
Longman Inc.
R u m m e l , R. J. 1970. Applied F a c t o r A n a l y s i s
Evanston, Illinois: Northwestern University
P re ss

217
Shapiro, Michael J. 1968. "The House and the
Federal Role: A Computer Simulation of Roll-
Call Voting." American Political Science Re-
- 62 (June): 494-517.
view.
Spegele, Roger D. '1980. "Deconstruct ing Method-
01% ic al F a1sif ic at ion ism in In t e rnat ional Re la-
tions" American Political Science Review. 74
(March): 104-122.
Wahlke, John C. 1979 "Pre-Be haviorialism in Po-
litical Science." American Political Science
Review. 73 (March): 9-31.

Watson, Hugh J.; and David P. Christy. 1982. "The


Evolving Use of Simulation" Simulation and
Games 13 (September): 351-363.
Whicker, Marcia Lynn; and Alfred R. Mauet. 1983.
Tomputer Simulation Modeling: A Tool for
Studying Organizations" Administ rat ion and
Society. 14 (February): 481-505.

218

You might also like