FROM QUESTIONS TO ANSWERS:

GETTING IN TOUCH WITH DATA

SUI 251 – Week 6
TOBB ETU Summer 2025
 Exercise & discussion on Long and Leeds (2006)
In our last lecture before the vacation I told you that we would have an in-class
discussion on this article...
Q1 What is the research question? Under which subdiscipline would you
categorize it?
Q2 What is the dependent variable (DV)? How is it defined?
Q3 Do the authors adopt a deductive or inductive approach? Why?
More advanced questions:
Q4 What is the key independent variable (IV) which they focus on? Do the authors
expect/demonstate a positive or negative relationship between the key IV and
DV?
Q5 Do the authors present a causal mechanism (an explanation into how the IV
affects the DV)? If there is, what might be the assumptions and analytical
priorities behind it? Can you link them to any of the theoretical perspectives
which you are familiar with?
 Last week, we also learned how we cite
academic works in the APA style. (Please
take the APA Style Handout seriously,
prepare and check your in-text citations and
references according to the instructions
given there.)
Proper referencing is an indispensable
component of academic writing and it is the
basic measure against ‘unintended
plagiarism’.
 Suppose, you are going to share some arguments from
Long and Leeds (2006) in your literature review.
• Make sure that you prepared the proper full reference and put it in the
‘References’:
Long, A. G., & Leeds, B. A. (2006). Trading for Security: Military alliances and economic
agreements. Journal of Peace Research, 43(4), 433–451.
When it comes to in-text citations:
• Things are easy when you are interested in conveying the broad message
Long and Leeds (2006) find that not all military alliances but only those which contain trade-
specific clauses improve the bilateral trade between allies.
OR
The content of the treaty, i.e., presence of specific clauses and terms for economic cooperation,
is argued to affect the market behavior (Long & Leeds, 2006).
 Quote or Paraphrase?
• If you are interested in using a specific section of the work, you have two
options.
• You may quote it
Leaders have diverse motivations for expanding the alliance by making commerce an additional
area of partnership. For instance, Long and Leeds (2006) highlight that if “the desire for the alliance is not
equally shared, it is possible that sweetening the pot, for instance, by offering to reduce trade barriers,
would make the deal more attractive” (p. 436).

We do not prefer this, unless the quoted expression has a worth-keeping formulation. If
you quote repeatedly, what you write will not be your work!
• Therefore, paraphrasing is a better solution
A leader may see a military alliance with another state as beneficial, but if the other state is not equally
interested, the leader may sweeten the deal by offering other incentives, such as lower trade barriers (Long
& Leeds, 2006, p. 436) .
 What if I summarize a common argument, idea or
finding that I came across in multiple works?
• An example:

A number of studies have found that military alliances affect trade (e.g. Gowa &
Mansfield, 1993, 2004; Gowa, 1994; Mansfield & Bronson, 1997).
You order them (1) alphabetically and (2) chronologically.
G comes before M
If the name is the same, a work published in 1993 comes before the other
one that was published a year later.
 Identify what type of document the following
references are for, and explain what is wrong with
them.
• Foa, R. S., & Mounk, Y. (2016). The Danger of Deconsolidation: The democratic
disconnect. Journal of Democracy, 5-17.
• King, G., Keohane, R. O., & Verba, S. (1994). Designing Social Inquiry: Scientific
Inference in Qualitative Research.
• M. V. Kauppı & P. R. Viotti (2023). International Relations Theory. Rowman &
Littlefield.
• Bennett, A. (2010). Process Tracing and Causal Inference. Rethinking Social
Inquiry: Diverse Tools, Shared Standards, 207-221. Rowman & Littlefield.
No matter what our inferential intent or methods
are…
WE HAVE TO THINK ABOUT HOW WE ARE GOING TO OBSERVE THE
CONCEPTS WHICH WE TALK ABOUT WHILE EXPLORING THE REAL WORLD…
How are we going to OBSERVE AND MEASURE some CONCEPTS- which are
abstract constructs- in empirical research?
Some concepts which we might be interested in as political scientists are
Economic development,
Support for a political party or a candidate,
Regime type or quality of democracy, etc.
Using concepts in empirical research
• To describe, we have to get in touch with facts and we need some appropriate
and reliable data.
• To analyze the variation in the electoral support that Erdogan garnered across
different parts of the country in the previous presidential election, I may collect
some election results data, e.g., vote percentages in the 2023 Presidential
Elections across provinces.
• To determine how the cost of living has changed in ten years, we can collect
annual inflation rates in the past decade.
 Three inevitable steps in inference-making
• In question-driven research, one should go through three steps regardless
of the type of inference s/he will make. These are: (1) conceptualization, (2)
measurement, or operationalization, and (3) case selection, or sampling.
• The first step for drawing a valid inference is to be clear about just what it is
we are making inferences about.
Defining the variables of interest is known as conceptualization.
• If we are to study corruption –it doesn’t matter whether we are after a
descriptive or a causal inference!!- we must come up with a clear and
precise definition. REMEMBER MY RECOMMENDATION! Check out the
literature!!! Is there a widely used - authoritative definition out there?
 How to conceptualize “corruption”
• The non-governmental organization Transparency International, one of the
leading institutions in corruption studies, offers a simple and clear definition:
“the abuse of entrusted authority for private gain.”
• However---
• Whatis a private gain? Does private gain require a monetary exchange? Do
non-material gains like “influence” or “sexual assault” count?
• Who defines abuse? Is it a legal or ethical problem? Reference: laws or values?
Are they universal?
• The latter question poses a crucial challenge - especially when we are interested
in studying global trends (in almost all countries). We should find a
generalizable set of criteria in order to operationalize and measure our
concepts! We are after operational definitions, just like Singer and Small’s
definition for war.
 Concepts and sub-types:
Ladder of generality/abstraction
• When we determine the
conceptual boundaries of
our research, our mind
should be very clear on the
analytical scope of our
prospective explanation.
What are we going to
study? We should set the
focus of our conceptual
preferences according to
the puzzle we are
interested in.
Measurement (i.e., operationalization of the
concepts)
• After we define our concepts – we have to discuss and find out how we are
going to observe and measure the concepts we refer to in our research
questions and hypotheses.
• To this end, we MUST answer the following questions:

• What is the unit of analysis?

• What is the level of analysis?
• Do the data cover the entire population or are they based on a sample drawn
from a larger population?
• Are the data cross-sectional or longitudinal?

• Are the data qualitative or quantitative?

 The unit of analysis is what is being studied or
compared; in political science research, the units of
analysis are typically political actors, individuals, political
entities, or political incidents. (To put it simply, it refers
to the "identity" or "essence" of the observations in the
data.)
Unit of Analysis
• What percent of the vote did Erdogan receive in the
2023 Presidential Elections across provinces? What is
the unit of analysis?
Provinces of Turkey
• Party preferences of the respondents in a survey.
What is the unit of analysis?
Individuals
 Level of Analysis
The level of analysis refers to the scale of the data, whether or not they have been
aggregated. In other words, does the data measure particular entities like individuals
directly; or does it represent a broader phenomenon about a group or society.
Depending on the research question and/or theoretical lenses we rely on, we may
have micro-level analysis or macro-level analysis.
• For example, a country is made up of millions of households and households are
made up of several individuals. At the micro-level, an individual has an income
(some wages, salaries, pensions, etc.) . At a higher level of analysis, all the
individual incomes in a household can be added to yield the household income.
Scaling up yet another level: Gross Domestic Income (GDI) = all income earned in a
country (which is some macro-level data).
• Another example: number of years that a person has attended school(=Level of
education at individual level) --> mean years of schooling (Average education level in
a country, indicating the quality of the human capital as macro-level property)
 Compare what? Temporal or cross-sectional variation?

• A cross-sectional study can measure the support for Erdogan in the 2018 Presidential
Elections across provinces .
• A time-series data, longitudinal data, can illustrate the inflation rates across years in
Turkey.
Time-series
cross
sectional
data
• It merges
longitudinal and
cross-sectional
aspects and
compares same
observations
(cross-sectional
data) across
years (as in
time-series
data)
.
 Quantitative and qualitative data...
Quantitative data: A data that attributes some numerical value to the observations
or cases (GDP growth rate, GDP per capita, vote percentage, democracy level of a
state in an index, number of battle-related casualties ..etc.)
• HOWEVER, all our observations cannot be attributed with some numerical values for
our variables of interest. Do you remember Doganay (2007) from last week?
• She studies the MPs’ discourses on democracy and democratization. To this end, she
analyzes the speeches delivered by the parties’ spokespersons during the plenary
debates on 1995 Constitutional Amendments… Is it possible to “quantify” their
rheoteric? Sometimes it is possible but for some, it is not preferable albeit doable. As
social scientists we often deal with verbal content – which remain unquantified.
Therefore, qualitative data is always an option –that we will discuss during
comparative designs and case studies later in this class.

.
 Unit of analysis: Years
Longitudinal data

Unit of analysis: US States

Cross-sectional data
Unit of analysis: Countries Unit of analysis: French departments

Time-series cross-sectional data Cross-sectional data

 Operationalization…
Operationalization is the process of making a concept observable and/or measurable.
It is a crucial step for applying the concepts to the related facts in an empirical
research. Simply put, operationalization is a process of defining the measurement of a
phenomenon.
How do we measure ______?
• Economic growth: Positive or Negative / Annual GDP growth (3.2%, -6.7%..etc.)
• Electoral performance of a party:Vote percentage in the elections (17%, 28%...etc.)
• Trade openness: Imports+Exports / GDP (in billion $)
• War: War / Not war (1 / 0)
• Democracy: Democracy/ not democracy ( 1 / 0)
OR level of democracy (5, 1, 8, 10…etc.)

.
 Indices (Index in plural): Quantifying the “non-
numerical”
In social science, some phenomena are not measured easily…
Complex concepts, like democracy…
How are we to “measure” democracy?
We consider each and every aspect or dimension of a working democracy and
assess them to achieve a “democracy” score. An index (indices in plural) is a
solution for concepts which we cannot measure directly.
Example: V-Dem Index, Polity IV, Freedom House Index, etc.
https://www.v-dem.net/
Types of data…
• To make inferences from data and to present them we should know some
major characteristics of the data.
Nominal data
Nominal data presents items which are distinguished by a simple naming
system. They are data that imply no numeric value, such as profession,
nationality, religion, ethnicity or regime type. They just name a thing –
categorize it, or they simply describe some quality of things.
• B. Geddes' typology is amenable to collect some nominal data in case all
autocracies are categorized as military regime, Single-party system,
Personalist Dictatorships, and Monarchy.
 Ordinal data
Ordinal data refers a set of categories that can be placed into
some kind of order by their position on a scale. For example,
they may indicate superiority.
• The Freedom House’s measurements on civil liberties and
political freedoms, for example, are often divided into three
categories: free, partly free, and not free.
• Economic development – when measured categorically: High
income countries, middle income countries, low income
countries
 Interval level data
• The third level of measurement is interval level data, where values can be
both ordered (from the highest to lowest) and communicate precise
differences. They are mostly inherently numerical variables.
• Financial and economic indicators are usually interval level data. According
to the World Bank, Mexico’s GDP per capita in 2011 was estimated at
$10,047 and Argentina’s was $10,942: a precise difference of $895 per
capita.
• Other examples: war casualties (number of dead and wounded people),
vote share of political parties in the elections (in percentages).
 Some more examples
• Annual coal production (in million short tonnes) INTERVAL

• Election
results (name of the winning candidate in every NOMINAL
department)
• Level of democracy (v-dem’s Liberal Dem. Index that varies from 0 INTERVAL
to 1)
• State capacity (High, intermediate, low) ORDINAL

• Satisfaction with the gov’t performance (very satisfied, somewhat

ORDINAL
satisfied, neither satisfied nor dissatisfied, somewhat dissatisfied,
very dissatisfied)
• Geographic location of the country (Africa, America, Asia… etc.) NOMINAL
REMEMBER! In any research, we have to go through following stages…
1. Conceptualization – defining our concepts. Prefer an authoritative definition. If
there is none, use a tentative one. Something is better than nothing – as long as it is
"OK".
2. Measurement or operationalization: “How am I to observe/measure my concepts
as variables, so that I can use them in an empirical research?”
Quantitatively or qualitatively? Even some very challenging (i.e., hard to measure)
concepts like corruption and level/quality of democracy can be quantified. But when
dealing with verbal content as data sources (especially speeches, statements,
“rhetoric” and “discourses”) we may prefer not to.
3. Case/sample selection for empirical research: What am I going to analyze? After
having determined our unit of analysis, we have to devise and justify a strategy
while selecting the cases and/or observations, about which we are going to collect
data. (In small-N designs, we select cases; whereas in large-N designs we define
and draw samples. I will explain what these "N"s mean, after the midterm. DO NOT
WORRY ABOUT THEM YET.)
 What is "sample" and why do we need
“samples”?
• In most of the micro-level studies, we –as political scientists and analysts- are unable
to collect data about each and every citizen, houshold or individual.
• States are exceptional bodies of data collection. They have the means and capability
to conduct censuses (or censi in plural)
A systematic method that collects and records data about the members of the
population is called census (for each and every unit of the population).
• Think about the electoral behavior and political culture literatures? How do we
collect data about the Turkish electorate or the Spanish society?
We conduct surveys on samples! (Sampling means taking a portion –i.e., a subset- of
the population selected to represent the entire group, in all its characteristics)
POPULATION WHICH WE WANT TO
UNDERSTAND (AND MAKE
INFERENCES ABOUT)

SUBSET THAT WE COLLECT

DATA ABOUT… WHAT WE
OBSERVE…
Sample & population
• In many cases, particularly when we are dealing with large populations (like all
Turkish voters, for example), it is impossible to collect data about every single
member of the population which we are interested in. Under these circumstances,
we take a subset of the population, i.e., a sample. Any research that relies on a
sample data faces an additional challenge to measurement error. That’s sampling
error.
• Remember, although we collect data about the sample, we make inferences about
the population. What social scientists really care about is not the percentage of
the sample that approves of the government’s foreign policy (This is the sample
statistic).
• What we want to know is population parameters - i.e., qualities and
characteristics of the POPULATION- such as the percentage of AKP voters in the
population.
Making inferences based on survey data…
• Remember! In surveys, we infer about the population (what we do not know) by
using the data we collect about a sample (we know). In an election survey, the
population we want to infer about is huge! 55.319.222 people for the 2018 General
Elections in Turkey.
• In this case, if a survey examines a small sample of 1000 (a relatively limited
number of observations), it seems likely that his sample would fail to capture all
the variation in the population as it is. Therefore, we need to acknowledge/admit
that what we observe in the sample does not perfectly reflect the distribution in
the population, although we do our best to ensure that it is somewhat
representative of the population. In other words, observations in the data
mirror the population – albeit with a slightly blurry image of it (We will
elaborate on “representativeness” in SUI 252)
• To make some inference about the population based on the sample statistic,
we take the margin of error into account within some confidence interval…
 • This range of the sample statistic
Confidence interval… plus and minus the margin of error
is known as a confidence interval
because pollsters are confident that
the true population parameter,
the actual percentage of likely U.S.
voters intending to vote for Obama,
lies within the range of 45.5 to 52.5
percent.
• Rule of thumb: We assume the
confidence interval as 95% in such
graphs or data outputs, if not
indicated by the author.
• That means we are 95% confident
that, based on the data the we
collect, percentage of Obama
voters in the real population is
between 45.5 and 52.5.
 Confidence interval…
• The larger the size of your survey
sample, the greater the likelihood it will
match the target population. However,
once the sample size reaches 2000, the
increases in precision is marginal. Unless
statistically significant results from a
specific subgroup of the population are
needed or we are not working on a tight
election, surveying 1000-1500 randomly
selected individuals will suffice for most
purposes because the margin of error (i.e.,
confidence interval) remains at about ± 3
percentage points around any percent.

How to interpret the findings of a poll conducted to observe the vote preference of the
Brits on Brexit…