Comparing Political Systems

1
Comparing Political Systems
Session 2 Key Concepts and Major Lines of
Thinking in Comparative Research
Why Compare?
2
What is science and what makes the sciences
scientific?
Science is about knowledge (latin scientia).
But what makes scientific knowledge different
from the knowledge that people use in their
everyday lives?
Scientific knowledge is generated by standardized
methods of data compilation and data analyses.
Scientific knowledge requires data (i.e.,
information about characteristics of phenomena)
and analytical methods to interpret these
data. Data must be documented and data analyses
must be reproducible.
Scientific knowledge occurs in terms of facts
(descriptions of phenomena), definitions
(classifications of phenomena) and theories
(explanations of facts). Most profoundly,
theories aim at answering why questions in that
they try to explain the occurrence of one class
of phenomena (effects) by the occurrence of
another class of phenomena.
In principle the social sciences (and political
science as one of their areas) do not deviate
from these general characteristics of science.
They only differ in the subject matter They
apply scientific methods to social and political
phenomena.
Political scientists, for example, ask for the
causes of revolutions, the causes of war and
peace, or the causes of regime change and regime
stability.
3
Understanding Causality as Conditionality
Causality X causes Y means that if there is X,
there will also be Y. X ? Y Alternatively Y is
present on the condition that X is present.
Example If the temperature falls below zero
degrees Celsius, water will be freezing to ice.
Other conditions being equal!
Control Boundary Conditions (Experiment)
4
Example of an Experiment
A possible example from experimental psychology
and research on media effects Does exposure to
violence in media make people more supportive of
the death penalty?
Explanatory variable (EV) Exposure and
non-exposure to violence in media.
Dependent variable (DV) Support and non-support
of the death penalty.
Experiment Expose a test-group to the stimulus
(violent movies) compose a control-group that
is not exposed to the stimulus. Analyze both
groups support for the death penalty.
If members of the test-group show stronger
support for the death penalty than members of the
control-group, one may conclude that exposure to
violence increases support of hard punishment.
However, this conclusion can only be drawn, if
members of the two groups differ solely in the
characteristics one wants to analyze. Thus, in an
experiment groups have to be designed such that
their composition is equal in all other relevant
characteristics. Holding these characteristics
constant (i.e., other Conditions being equal)
means to isolate the effect of an EV on a DV.
5
The Comparative Method as a Quasi-Experiment
Experiments give the researcher control over the
stimuli she/he wants to analyze and allow
her/him to hold boundary conditions constant,
such that causes and effects can be isolated from
boundary conditions.
(Un)fortunately, in the social sciences--and in
political science in particular-- experiments are
not possible to the extent that would allow
researchers to answer their most important
research questions. Power holders do not allow
political scientists to instigate revolutions or
economic shocks in order to examine what happens
if
Thus, as a surrogate for natural experiments,
political scientists make systematic
comparisons.
What does systematic comparison mean? Isnt it
rue, as a German proverb says, that one cannot
compare apples with pears in a meaningful way
Of course, one can compare two objects, as
different as apples and pears, if one only has a
common reference (tertium comparationis) under
which such comparison makes sense. It makes
sense, for instance, to compare apples and pears
for their nutrition values.
6
The Comparative Method as a Quasi-Experiment
Comparison resembles an experiment in that
researchers look for cases that have been under a
certain stimulus (test-group) and compare them
with other cases that have not been under this
stimulus (control-group) in order to see, if the
stimulus brings a certain effect.
For example you may want to test the hypothesis
that economic shocks lead to political
revolutions. Accordingly, you look for cases in
which there has been an economic shock
(test-group) and compare them with cases in which
there has not been an economic shock, in order to
see whether the incidence of political
revolutions is systematically linked with
economic shocks.
In addition, one must select ones test-group and
control-group such that the cases of these two
groups differ solely in the stimulus, implying
that they are rather similar in other basic
characteristicssuch as level of economic
development and type of political system.
This is most difficult to achieve because nations
are never so similar that they only differ in a
certain stimulus. Thus, the requirement of
holding boundary conditions constant cannot be an
absolute one in CP.
7
Difficulties in Research Practice
Comparative research is often confronted with
design problems, based on case selections not
sufficiently taking care about the control of
boundary conditions (the too many variables-too
few cases problem).
Apart from these design problems, there are a
number of practical difficulties that have to be
handled in a competent manner. For example, one
needs theoretically meaningful concepts of what
economic shocks and political revolutions are
about. You also have to have straightforward
criteria that enable you to decide on the basis
of empirical data whether or not an instance is
counted as an economic shock or not and whether
it is counted as a political revolution or not.
And, of course, you need data to make such
decisions.
Requirements for data quality are high throughout
all the sciences. Data have to be reliable in
the sense that they must not be faked and that
any researcher applying the same data generating
procedure produces the same data. Data have to be
valid in the sense that they indicate what they
pretend to indicate. It must be clear how
representative data are for the class of
phenomena under study. Considering data on
political revolutions, it must be clarified
which portion in time and space of the totality
of political revolutions these data represent.
For general conclusions can only be drawn on the
basis of data that represent a class of
phenomena. Finally, data themselves as well as
their sources and sampling procedures must be
properly documented in order to open ones
research for scientific debate.
8
The Great Divide in the Comparative Politics
Phenomenological or reconstructive approach (Area
Studies)
Human communities are unique in their totality.
The only way to understand them is to reconstruct
their whole complexity. To understand a community
one must know as much about it as possible.
Comparing communities makes little sense
because they are unique. There are no general
relationships between variables that could be
dissolved from the complexities of single
communities. The process of understanding
(hermeneutics) cannot be organized by methods.
The social sciences cannot have the same
methodological canonization as the natural
sciences.
Comparative or analytical approach (Comparative
Politics)
Human communities are complex but there are
general relationships and tendencies that cut
through the complexities of single societies. The
only way to make these general relationships
visible is to dissolve case-bound complexities
into variables whose linkages can be analyzed by
standardized techniques. Scientific analysis
is about reduction not reconstruction of
complexity, since generalization is at the
heart of scientific explanation. In principle,
social science methods do not differ from
methods in the natural sciences, since social
phenomena are susceptible to measurement and as
are natural phenomena.
9
Some Basics of the Comparative Method
If one wants to compare one has to have
Cases (units of observation)
Characteristics or parameters on the basis of
which one compares the cases (cases being the
carriers of these characteristics).
If a characteristic occurs in only one
specification it is called a constant.
If a characteristic has various specifications it
is called a variable.
General patterns structuring social and political
reality can be conceptualized as associations
between variables, such that a particular
specification in one variable occurs in
association with a particular specification in
another variable.
It has been argued, for example, that a
particular specification in the variable type of
political system, namely democracy, occurs in
association with a particular specification of
the variable level of economic development,
namely industrial and postindustrial level of
economic development.
10
Qualitative and Quantitative Variables
All variables not reflecting natural quantities
are qualitative variables by their very nature.
Indeed, most of the variables in political
science, such as party preferences, types of
political systems, governmental coalitions, are
qualitative variables.
Considering their level of information
(statisticians talk about the scaling level),
qualitative variables can be nominal or ordinal.
They are nominal, if their specifications simply
indicate mutual exclusiveness. An example is the
variable type of political system, if we
simply dichotomize democracies and dictatorships.
Qualitative variables are ordinal, if their
specifications can be considered as rank ordered.
An example is the variable national power, if
we operationalize the rank-ordered specifications
minor power, medium power, major power, super
power.
Quantitative variables, by contrast, represent
natural quantities. Examples are per capita GDP,
measured in units of a currency, or voter
turnout, measured in percentages of the total
electorate.
But even though qualitative variables do not
represent natural quantities, for the purpose of
statistical analyses their specifications can be
coded numerically. Actually, this is what is
always done in representative surveys. Measuring
simply means to translate information into a
scale of numerical codes, such that each code
represents a distinctive specification of the
measured variable.
11
Deterministic and Probabilistic Variable Relations
Comparative Politics compares cases (usually
national societies) in order to identify the
patterns that structure political reality. To the
extent to which such patterns exist, they are
reflected in systematic associations between the
variables that characterize the units of
observation (cases).
The associations between variables can be
conceived in terms of deterministic relationships
or probabilistic relationships.
Deterministic relationships allow for no
exceptions. Thus, deterministic relationships
are falsified by only one observation in
contradiction with them. The conjecture, for
example, that all swans are white is definitely
falsified by only a single black swan, the
millions of observations supporting this
conjecture notwithstanding.
12
Sufficient and Necessary Conditions
Deterministic relationships can be expressed in
terms of sufficient or necessary conditions. A
condition is sufficient to produce a certain
outcome, if the presence of this condition
suffices to produce this outcome. A condition
is necessary, if its presence is needed to
produce the outcome.
Lets consider one of the most debated hypothesis
in political science the democracies dont
fight each other thesis. This thesis involves a
statement on the association between two
variables type of political system of two
interacting states (both democratic vs. not
both democratic) and kind of interaction
(peace vs. war), such that the specification
both democratic in the first variable is
associated with the specification peace in the
second variable.
13
Necessary Condition
Formulating this association as a necessary
condition points at the following statement
Only if two interacting states are democratic,
they do not fight wars against each other. This
implies that all pairs of states that dont fight
wars against each other must be composed of a
pair of democracies. Hence, if we find a pair of
states that doesnt fight wars against each
other, of which one is a non-democracy, the
condition will be falsified. In order for the
condition to hold, the lower left cell in the
four-fold matrix must remain empty.
14
Sufficient Condition
Formulating the association as a sufficient
condition points at the following statement
Always if two interacting states are
democratic, they do not fight wars against each
other. This implies that all pairs of states
of which both are democratic do not fight wars
against each other. Hence, if we find a pair of
democracies states that did fight wars against
each other, the condition will be falsified. In
order for the condition to hold, the upper
right cell in the four-fold matrix must remain
empty.
15
Both Necessary and Sufficient Condition
If we face a total determination (only if and
always if) of a phenomenon, the distribution
would be as follows
16
Qualitative Comparative Research (QCA)
Some researchers commit themselves to
deterministic explanations of social phenomena.
They try, for example, to explain the occurrence
of revolutions, of wars and of democracies
exhaustively, leaving no room for
exceptions. Since one condition can almost never
explain a class of phenomena completely, these
researchers look for additional conditions and
various combinations of conditions, until no more
exception is left. Unfortunately, however, this
endeavour often ends up in as many
combinations as countries (cases), implying that
complexity is hardly reduced and little
is explainedgiven that explaining country
variation requires to operate with fewer
arguments than countries.
17
Probabilistic Relationships
The social world is composed of individuals and
their interactions. As individual behavior is
never totally predictable so are their
interactions. Hence, the best we can expect is
an incomplete explanation of social phenomena
that captures major tendencies. This is the
realm of quantitative comparative analyses. It
does not look whether specifications in one
variable determine specifications in another one.
Instead, it examines the degree to which
variation in one variable is associated with
variation in another variable, whereby the
degree of uncertainty in such associations is
precisely quantified.
18
(No Transcript)