Chapter Four: Data Collection Methods

Introduction
 Research in Political Science and International Relations begins with
identifying a clear problem.
 Defining this problem helps determine the broader context in which it
exists and guides the researcher toward the relevant population of
interest—such as states, institutions, communities, or individuals.
 Since it is usually impossible to study an entire population, researchers
select a smaller group, or sample, that represents the whole.
 This process is known as sampling, and different sampling techniques
can be used depending on the nature of the research.
 This chapter explains the idea of a sample and outlines the main
sampling methods used in research.

4.1 Sample and Sampling Techniques

4.0 Population and Sample

 In research methods, the population refers to the entire set of units

(people, countries, organisations, etc.) to which we hope to generalise
our findings. A sample is a subset of that population which we
actually observe and collect data from.
 In political science and IR, for example, if you are studying the
attitudes of all elected parliamentarians in Ethiopia, your
population is that group; if you survey 200 of them, that is your
sample.

Why this matters in Political Science/IR

 Suppose you are studying “democratic transitions in Sub-Saharan

Africa since 1990”. Your population could be all Sub-Saharan African
states that have held multiparty elections since 1990. If you then select,

1
 say, 25 of those states and survey elite policymakers, that’s your
sample.
 If you wish to study “citizens’ attitudes toward foreign military bases
in their countries”, your population might be all adult citizens in
countries hosting foreign bases. If you survey 1,200 respondents
spread across three countries, that’s your sample.

Key considerations:

 You must clearly define your population: what are the temporal,
spatial, inclusion/exclusion criteria? E.g., “all UN member states
between 2000-2020” or “all registered voters in Ethiopia in 2025”.
 Ensure you have a sampling frame, or at least a way to approximate it:
e.g., a list of all eligible units (voters list, list of organisations, list of
countries).
 Distinguish between census and sample: If you manage to include the
entire population (e.g., all 193 UN member states), that’s a census (or
near-census), and sampling error from selection is essentially zero—
but you still may face non-response or missing data issues.

 Example (IR): Suppose you are studying how small-island

states perceive the influence of major powers in the
Pacific.
 Your population might be “all small-island states in the
Pacific recognised by the UN as of 2025” (say N = 15).
 If you interview 10 of these states’ foreign ministry
officials, that is your sample. You would need to discuss
how representative those 10 are of all 15.

4.1.1 Characteristics of a Good Sample

What makes a sample “good” in political science/IR research?

A “good” sample in political science/IR research will have several

characteristics, including:

2
 Representativeness: The sample should mirror key characteristics of
the population. For example, if the population of states includes low-
income, middle-income, and high-income countries, then your state
sample should include each kind in roughly proportional numbers (or
else you must justify oversampling/weighting).
 Adequate sample size: Enough units to allow meaningful inference
(statistical or qualitative). For instance, if analyzing parliamentary
voting behaviour across 54 African states, interviewing only 3 states
would limit generalisation.
 Low selection bias: The method by which units are selected should
minimise systematic exclusion of certain relevant units (e.g., excluding
closed/hard-to-reach states or elites).
 Clear sampling frame: The list from which you select your units must
match the defined population as closely as possible. E.g., if your
population is “all registered voters in Addis Ababa”, you should have
an up-to-date electoral register or list.
 Feasibility and cost-effectiveness: While theoretical ideals matter, in
practice you must balance quality with resources, time, access, and
ethics.
 Transparency and ethical clarity: You must report how you selected
the sample, the response rate, any exclusions, and limitations.

 Practical Example: Suppose a researcher aims to survey

diplomatic elites in the Horn of Africa on climate change
cooperation.

 The population is “all foreign ministry directors (or equivalent)

in Djibouti, Eritrea, Ethiopia, Somalia, and Sudan”.
 A good sample might include one or two directors per country
plus a backup list, selected via official directories, ensuring
language-access, and adequate representation of each country.
 A poor sample would include only Ethiopian officials (because
they were easier to contact) and ignore the other four, thereby
introducing bias.

3
  Example (Political Science): Imagine you are studying voter
attitudes toward foreign aid in Ethiopia.
o If you only sample university students in Addis Ababa, you
would lack representativeness of rural voters, older voters, or
non-students.
o Thus your sample would be biased.

4.1.2 Sample Size

 Sample size matters because of its effect on precision (smaller

sampling error) and statistical power (for hypothesis testing).
 However, in political science/IR automatic “bigger is better” is not
always true: you must balance cost, logistics, and theoretical
relevance.
 In cross-national studies (e.g., N countries = 195) your sample might
be constrained by the number of countries.
 In survey research you might calculate minimum sample size based
on expected effect size, desired confidence level, and variance.

In political science/IR contexts:

 If you survey citizens (large-N survey) you typically aim for large
samples (hundreds to thousands) to allow subgroup analysis (e.g.,
gender, region, political party).
 If you survey elites (e.g., ministers, ambassadors) the sample size will
be small by necessity (maybe dozens). In such cases you emphasise
depth and qualitative validity over large-N statistical power.
 Cross-national studies: The “N” itself may be limited (e.g., number of
countries). For example, if you study “all Arab League states”, you
have at most ~22 units. You need to discuss how this limits inference.
 You may use power calculation (in quantitative work) or saturation
logic (in qualitative research) to justify size.

 Example: You want to test whether regime type affects support

for multilateral climate treaties among 130 countries.
4
  Suppose you expect a medium effect size and you wish 95%
confidence, you might calculate that you need ~80 countries
sampled.
 If you manage only 50 due to data constraints, you must
acknowledge increased margin of error and reduced power.

 Example: If studying elite perceptions (say foreign ministers)

across 54 African states, you might aim to interview 54 or as many
as possible.
o If resource-limited you may sample 20 states, but you must
discuss how you selected those states and the limitations in
inference.
 Political/IR Example: A survey of 10,000 voters in Ethiopia may
allow you to estimate regional and demographic differences in
foreign policy attitudes.
o On the other hand, a study of 15 foreign ministers is small
but still valuable for elite insights.

Key Factors Affecting Sample Size

1. Population size (N) – the total number of individuals in the

population.
2. Margin of error (E) – the amount of error you are willing to tolerate
(e.g., ±5%).
3. Confidence level (Z) – the degree of certainty that the population
parameter lies within the margin of error (e.g., 95% confidence = Z =
1.96).
4. Population proportion (p) – the estimated proportion of the
population that has the characteristic of interest (often assumed to be
0.5 when unknown).

5
Formula for Sample Size (for Large Populations)

When the population is large (infinite or > 20,000), use:

⋅ ⋅ (1 − )
=

Where:

 = required sample size

 = Z-score corresponding to your confidence level
 = estimated proportion of the population
 = margin of error (in decimal form)

Example:
If confidence level = 95% (Z = 1.96), margin of error = 5% (E = 0.05), and
= 0.5:

1.96 × 0.5 × (1 − 0.5)

= = 384.16
0.05

So, you’d need about 385 respondents.

Adjusted Sample Size (for Finite Populations)

If your population is small (finite), adjust the sample size as follows:

=
−1
1+

Where:

 = adjusted sample size

6
  = total population size
 = sample size from the large-population formula

Example:
If = 385and = 2,000:

385
= = 322
384
1+
2000

So you’d need about 322 samples.

Summary Table

Parameter Symbol Typical Value

Confidence level Z 1.96 (95%), 2.58 (99%)
Margin of error E 0.05 (5%)
Population proportion p 0.5 (if unknown)
Population size N Known total

Recommended
Research Type Example
Sample
Nationwide political Afrobarometer Kenya poll =
384–1200 respondents
survey ~1600
Elite interviews 10–30 interviews Interviews with AU diplomats
Rwanda genocide tribunal;
Case study 1–5 cases
ICC Kenya cases
Conflict event Armed Conflict Location &
All available events
analysis Event Data (ACLED)

7
  Qualitative studies prioritize depth over breadth (King, Keohane &
Verba, 1994), while quantitative studies prioritize generalization.

Example: let’s calculate your sample size step-by-step for a population of

N = 800, using the standard assumptions (unless you specify otherwise):

 Confidence level = 95% (Z = 1.96)

 Margin of error = 5% (E = 0.05)
 Estimated population proportion = 0.5 (p = 0.5 — most conservative,
gives the largest sample size)

Step 1: Compute initial sample size for large population

⋅ ⋅ (1 − )
=

Substitute values:

1.96 ⋅ 0.5 ⋅ (1 − 0.5)

=
0.05
3.8416 ⋅ 0.25 0.9604
= = = 384.16
0.0025 0.0025

So for a large (infinite) population, n = 384.16.

Step 2: Apply finite population correction

=
( −1)
1+
8
Substitute values:

384.16 384.16 384.16

= = = = 259.7
(384.16−1) 1 + 0.47895 1.47895
1+
800

Final adjusted sample size = 260 respondents (rounded up).

Summary

Parameter Symbol Value

Population size N 800
Confidence level 95%
Margin of error 5%
Proportion 0.5
Large-population n 384
Adjusted for finite population 260

4.2 Sampling

4.2.1 Reasons for Sampling

 Why do researchers sample rather than study the whole population?

In political science/IR research the reasons include:

 Practicality/Feasibility: It is often impossible or prohibitively

expensive to study the entire population.
o E.g., surveying all citizens in all UN member states would be
impossible.
 Time and cost constraints: Fieldwork, interviews, translation and
travel all cost time and money — sampling reduces these burdens.

9
  Manageability: Working with a manageable sample allows for better
control over data quality, interview depth, and logistical coordination
(especially for elites or international fieldwork).
 Access/ethics/logistics: Some units are hard to reach (closed
regimes, secretive organisations) or politically sensitive — sampling
helps focus on accessible units.
 Generalisation: When done properly, sampling enables you to make
inferences about the larger population (with caveats).
 Resource optimisation: Focusing your resources on fewer units may
allow deeper data collection (richer interviews, higher quality survey)
rather than superficial enumeration of all units.

In IR research for instance, you cannot interview every diplomat in every

state on every issue, so you sample a subset of states or officials.

4.2.2 Sampling Error

 Sampling error refers to the difference between the parameter (true

value in the population) and the statistic (value from the sample)
caused by the sampling process.
 Sampling error is the difference between a statistic computed from
your sample and the true parameter of the population, solely because
you sampled rather than measured the whole population. It’s a
function of sample size, variation, and sampling design.
 Sampling error arises even when sampling is unbiased; the mere
randomness introduces variation.
o It is distinct from systematic bias (which arises from poor
sampling method or frame).
 In political polling (e.g., election polls), sampling error is a key metric
in determining how confident we are in poll results.

10
Important in political/IR context:

 Even with perfect random sampling, you cannot escape sampling

error — only reduce it (by increasing size).
 Non-response, missing data, or biased frames contribute to bias
(systematic error) rather than sampling error per se, but both reduce
validity.
 Many political polls emphasise margin of error; however, real-world
political science research often faces additional non-sampling-error
complexities (coverage issues, elite non-response, hidden
populations).
 In some IR studies where the population is small (e.g., 20 states),
traditional notions of “margin of error” become less applicable — you
may treat it more like a case-study set than a large-N inference.

 Example: Suppose you sample 500 respondents in Kenya about

foreign aid. You find 40 % support for increasing aid.
o Your margin of error may be around ±4.4 % (if design is
simple random).
o This difference arises because you did not survey all citizens.
If the true support is 43 %, your sample statistic is 40 % —
the 3 % difference is sampling error (plus any bias).
o In a state-level sample: If you study 30 out of 54 African
states, and find a certain average diplomacy index of 7.2,
there’s sampling error (but harder to quantify) and you must
discuss how missing states might differ.
 Example: Suppose you sample 1000 registered voters in Ethiopia
and find 60% support for a given policy.
 Your margin of error (say ±3%) reflects sampling error; the true
population proportion could be between ~57% and 63%.
 If your sample was poor (non-representative, high non-
response) then systematic bias may dominate.

11
4.3 Sampling Methods

 Sampling methods fall broadly into two categories: Probability

Sampling and Non-Probability Sampling.
 Additionally, there are “other” or hybrid methods.

4.3.1 Probability Sampling

 Probability sampling means every unit in the population has a

known, positive probability of being selected.
o This enables strong inference and estimation of sampling error.
 Probability sampling means that every unit in the population has a
known, non-zero probability of selection.
o This allows for calculation of sampling error, valid statistical
inference, and generalisation to the population (if other
assumptions hold).

Key types of probability sampling with IR/Political Science:

 Simple Random Sampling (SRS): Every unit has equal chance.

o For example, if you have a list of 1,000 registered voters in
Addis Ababa, you randomly select 200 via random number
generator.
o This method is conceptually straightforward.
o Example: A researcher has a list of all 450 parliamentarians in
the EU and randomly selects 100 to survey about EU
enlargement.
o Example in IR: Suppose you have a list of all 193 UN member
states, you randomly select 40 states to survey their foreign
policy officials on a climate change question.

 Stratified Random Sampling: You divide the population into sub-

groups (strata) based on key characteristics (e.g., region, income level,

12
 regime type) and then sample from each stratum (often
proportionally or equally) to ensure representation.
o Moreover, the population is divided into homogeneous sub-
groups (strata, e.g., regime type, region, income group), and
you sample within each stratum (often proportionally) to ensure
representation.

 Example: If studying democratic transitions in Africa, you

may stratify by regime type (democracy, hybrid,
autocracy) and sample 10 states from each stratum.
 Example: Studying attitudes toward multilateralism
among diplomats: You might stratify by region (Africa,
Asia, Americas, Europe), and sample 5 diplomats from
each region to ensure regional representation.

 Cluster Sampling / Multi-stage Sampling: Useful when individual

units are numerous or scattered and a complete frame is lacking.
o You sample clusters first (e.g., regions, countries) then sample
units within clusters.
o This sampling is useful when a sampling frame of individual
units is not readily available but clusters (e.g., states, regions)
are. You randomly sample clusters, then sample units within
clusters.
 IR Example: Surveying NGOs across South Asia: select 4
countries (clusters) randomly, then within each country
randomly select 15 NGOs.
 IR Example: You want to survey civil society organisations
across countries. You randomly select 8 countries
(clusters), then within each country randomly select 10
organisations.

13
  Systematic Sampling: You list units in some order then select every
th unit after a random start. E.g., every 10th person on a voter list.
o Moreover, it is selecting every th unit after a random start
from an ordered list.
o Some treat as probability sampling when the frame is
appropriate.
 Example: From a list of 500 embassy staff in a region, you
select every 10th person after random start to construct your
survey sample.
 Example: A researcher uses a list of 10,000 diplomats and
picks every 20th name after a random start to form a sample
of 500.

Practical Observations in Political Science/IR

 Probability sampling offers the strongest basis for inference, but often
faces constraints: incomplete frames (e.g., list of all rebels in a civil
war), access problems (closed regimes), cost/time limits. Researchers
must consider trade-offs.
 Although probability sampling is ideal for generalisation, in many IR
settings complete frames are unavailable, response rates are low, and
access is constrained.
o Researchers must note these limitations.
 Recent work has explored online probability sampling among activists
or elites (e.g., via social-media ads) and discussed how coverage bias
and sampling error interact in those contexts.
 Researchers sometimes oversample smaller strata (e.g., small states)
and then apply statistical weighting in analysis to restore
representativeness.

14
4.3.2 Non-Probability Sampling

 Non-probability sampling means the probability of selection of each

unit is not known.
 This reduces ability to generalise and to compute standard errors.
 Non-probability sampling means the probability of selection of each
unit is unknown and/or zero for many units.
 This reduces generalisability and often does not allow calculation of
sampling error in the usual way.

Common types:

 Convenience Sampling: selecting units simply because they are

easily accessed. E.g., surveying students in a classroom.
 Selecting units that are easiest to reach.
o For example, interviewing participants at a political science
conference.
o Example in IR: Interviewing whichever embassy officials
happen to be in town during your visit — easy, but not
representative.
o Example in IR: Researcher surveys embassy staff in Addis
Ababa because they are accessible; but staff in remote postings
are omitted.

 Purposive (Judgement) Sampling: researcher selects units based on

specific purposes or criteria (experts, case studies).
 Selecting units based on the researcher’s judgement about which are
most informative.
o Example: Selecting 5 key opposition parties in a set of regimes
to compare their strategies.
o Example: Selecting 8 foreign ministers from emerging powers to
interview about global order change because they are
influential and accessible.

15
  Snowball Sampling: Initial participants recruit others respondents.
o Particularly useful for hidden or hard-to-reach populations.
 Example: Studying clandestine transnational activist
networks: you begin with one known activist who then
refers other network members.
 Example: Studying rebel groups: you interview one
fighter, who then identifies others.

 Quota Sampling: you choose units to fill predefined quotas or you

set quotas for specific sub-groups (e.g., 30% women, 20% minorities)
and then fill them by non-random selection within each quota.
o Example: Surveying party activists: you decide to interview 50%
male/50% female activists, and you recruit until quotas are
filled, though within each quota you may choose conveniently.
o Example: In surveying parliamentarians, you ensure you
interview “at least 30% women” but choose them non-
randomly.
 Practical note: Non-probability methods are common in qualitative
and exploratory research in political science/IR (e.g., elite interviews,
case studies), or when probability sampling is infeasible.
o But researchers must be very transparent about limitations and
avoid over-generalisations.

When non-probability is appropriate:

 When you are doing exploratory, qualitative, or hypothesis-

generating research rather than formal statistical inference in political
science/IR (e.g., elite interviews, case studies), or when probability
sampling is infeasible.
 When the population is hidden, dangerous, or hard to define (e.g.,
insurgent groups, illicit networks).

16
  But researchers must be very transparent about limitations and avoid
over-generalisations.
 When logistical or ethical constraints prevent probability sampling.

Limitations:

 You cannot compute a meaningful margin of error or make strong

population generalisations.
 The sample may suffer strong bias (self-selection, referral bias, limited
coverage).
 Researchers must clearly report limitations and avoid overstating
generalisability.

4.3.3 Other (Hybrid / Advanced) Sampling Methods

Beyond the basic probability vs. non-probability dichotomy, political

science/IR research often uses hybrid or sophisticated designs:

 Mixed-mode or mixed-method sampling designs: Combining

probability for one stage and purposive/non-probability for another.
Example: You randomly sample 30 countries (probability) and within
each country you purposively select the foreign ministry official best
positioned to answer questions (non-probability).
 Oversampling and weighting: You deliberately sample more units
from a subgroup (e.g., small states) and then apply statistical weights
so that the overall sample reflects the population distribution.
Example: In a survey of 154 countries, you sample 20 small island
states specifically (oversample) because they are under-represented,
then weight them so the analysis reflects global distribution.
 Multi-phase or multi-stage sampling: For instance, you first sample
clusters (countries), then sample units within them (organisations),
then sub-units (individuals).
Example: A study of civil society organisations: Stage 1 select

17
 countries; Stage 2 select NGOs; Stage 3 select staff within NGOs for
interviews.
 Balancing and probability-balanced designs: Some recent research
outlines how to select samples that balance across multiple variables
(regime type, GDP, region) and satisfy “model-assisted” criteria.
 Online sampling via social media/cohort panels: As one study
shows, researchers used Facebook ads to sample political activists
across countries and time, acknowledging the sampling error and bias
issues of online platforms.

Practical IR Example: Let’s say you are researching “how domestic public
opinion in emerging powers influences foreign policy decisions”. You
might:

1. Use stratified sampling: select emerging powers by GDP bracket

(middle-income, lower-middle) across regions.
2. Within each selected country, use quota sampling to survey 1,000
citizens ensuring gender, urban/rural quotas.
3. For elite interviews in each country, use purposive sampling to select
one policymaker, one opposition member, one civil society leader.
4. In analysis you weight the citizen survey responses to match
population demographics and clearly state the limitations of the elite
sample.
5. Suppose you are investigating elite perceptions of foreign influence
in African states. You use stratified random sampling for states
(probability), then within each state you use purposive sampling of
the foreign ministry experts (non-probability). You clearly label the
methods and interpret findings cautiously.

Conclusion

 Sampling is a foundational element of empirical research in political

science and international relations.

18
  A well-designed sampling strategy enhances the validity and
credibility of your findings; a poor sampling strategy undermines
them.
 Sampling is at the heart of empirical research in political science and
international relations.
 Whether you are conducting a large-scale survey of citizen attitudes
or in-depth interviews with foreign policy elites, your sampling
decisions significantly influence your validity, reliability, and
generalisability.
 In political science/IR contexts, practical constraints (access, cost, data
availability, sensitive topics) often force compromises: the key is to be
transparent about your sampling design, to justify your choices, and
to interpret findings in light of limitations.

Key take-aways:

 Clearly define your population and sample.

 Strive for a good sample: representative, appropriate size, minimal
bias, transparent frame.
 Understand the trade-offs: probability sampling is ideal for
generalisation; non-probability may be necessary for elites or hidden
populations.
 Always discuss sampling error, bias, and limitations in your work.
 Choose the sampling method matching your design, context,
resources, and research goals.
 Probability sampling remains the gold standard for generalisation,
but non-probability methods have their place—especially in
qualitative, exploratory, or elite research settings.
 Use hybrid or advanced designs when needed, but be transparent.
 In political science/IR contexts, you will frequently face constraints
(access, cost, frames) so your discussion of limitations is just as
important as your methodological choices.

19
By carefully documenting your sampling process, justifying your choices,
and situating them within your theoretical and practical context, you
strengthen the credibility of your research.

20
4.2: Research Instruments

4.0 Measuring Instruments

 In empirical political science and IR research, a measuring instrument

is any tool (questionnaire, interview guide, observation checklist,
experiment protocol, etc.) used to collect data on concepts of interest.
 The link between abstract theoretical concepts (e.g., “political
tolerance”, “foreign policy activism”, “regime legitimacy”) and
observable data is established through instruments.
 These tools bridge the gap between theory and observation (Babbie,
2020).
 For example, when studying “political trust,” we cannot directly
observe trust but can measure it through carefully designed
questions or behavioural indicators.

Key points:

 Concept → Operationalisation → Measure (instrument)

 The instrument must be designed to ensure the data collected
correspond meaningfully to the theoretical concept.
 Poor instruments weaken the validity and reliability of your research.

Why Instruments Matter

Every research question requires an appropriate instrument. A mismatch

between concept and instrument leads to invalid results. For instance:

 Studying foreign policy decision-making through public opinion polls

is a mismatch, since policy decisions are made by elites, not the
public.
 Conversely, investigating citizens’ support for international alliances is
best approached through a survey instrument targeting the general
population.

21
Instruments Commonly Used in Political Science & IR

1. Questionnaires and surveys (quantitative)

2. Interviews (qualitative or mixed)
3. Observations (behavioural)
4. Experiments (controlled interventions)
5. Task analysis (procedural)
6. Delphi technique (expert forecasting)

Each tool has distinct advantages and challenges, often shaped by the
political and cultural context of the study area.

4.2.1 Questionnaire

 A questionnaire is one of the most widely used tools in Political

Science and IR.
o It allows researchers to systematically gather standardized
information from a large number of respondents, enabling
statistical analysis and cross-national comparisons (de Vaus,
2014).
 A questionnaire is a structured set of questions presented to
respondents, typically used in surveys.
o In political science/IR it might survey voters, elites, diplomats,
NGOs, or citizens across states.

Design Considerations

1. Question wording: Clear, unambiguous wording, avoiding political-

loaded or confusing phrasing, avoid leading, double-barrelled, or
ambiguous questions.
o Poor: “Do you agree that corruption and inefficiency in the
government must end?”
o Better: “Do you agree that government corruption must be
reduced?”

22
 2. Response format:
o Closed-ended (Likert scales, yes/no, multiple-choice) for easy
quantification.
o Open-ended questions for richer qualitative insights.
3. Translation & Pretesting: In multilingual contexts like Nigeria or
India, instruments should be translated and back-translated to
preserve meaning (Chhibber & Verma, 2018).
4. Pre-test for comprehension, translation (if cross-national), cultural
appropriateness.
5. Order of questions: avoid priming effects, social desirability biases.
6. Mode of administration: face-to-face, telephone, online.

Practical Examples

 Afrobarometer Survey (Africa) uses standardized questionnaires to

assess citizens’ attitudes toward democracy, governance, and foreign
influence (Bratton et al., 2020).
 Asian Barometer gathers data on democratic values, trust, and
governance in Asia. Questions such as “How satisfied are you with
your government’s foreign policy?” allow comparisons across regimes.
 A survey of citizens in Ethiopia on attitudes toward foreign military
bases: questionnaire includes items such as “On a scale from 1-5, how
favourable are you toward foreign troops stationed in your country?”
 A questionnaire for diplomats in the Pacific on great power
competition: “How often in the last year has your ministry engaged
with China’s Belt and Road Initiative? (0 = never, 1 = once, 2 = 2-3
times, 3 = 4+ times)”.

Strengths

 Facilitates cross-country comparisons.

 Standardised, allows large-N analysis, comparisons across
groups/countries.
 Economical for large samples.
 Enables statistical modelling (e.g., regression, correlation).

23
Limitations

 Social desirability bias—respondents may give politically correct

answers, especially in authoritarian regimes (Maoz, 2012).
 Limited depth; complex motives behind responses may remain
hidden.
 May miss depth of responses; responses may be influenced by social
desirability or non-response bias.

4.2.2 Mail Questionnaire

A mail (postal) questionnaire is a subtype of questionnaire distributed via

postal mail (or email in modern versions) and returned by the respondent.

A mail questionnaire is distributed and returned by mail (or increasingly

by email or online platforms like Qualtrics). It’s ideal for literate, dispersed
populations such as diplomats, legislators, or NGO leaders.

IR and Political Science Examples

 A researcher studying ASEAN diplomacy might mail questionnaires

to foreign ministry officials in Indonesia, Singapore, and Malaysia
asking about their perceptions of regional integration.
 An African scholar might mail survey instruments to MPs across
ECOWAS states to explore inter-parliamentary cooperation.
 A researcher studying national parliamentarians in Asia sends a
mailed questionnaire to MPs in 10 countries asking about their views
on inter-parliamentary cooperation.
 The questionnaire includes a cover letter explaining purpose,
confidentiality assurances, and a pre-paid return envelope.
 In a comparative study of lobbying laws, researchers might mail
questionnaires to regulatory bodies across EU, Africa and Asia to
capture their transparency practices.

24
Considerations:

 Response rates tend to be lower than face-to-face; follow-up

reminders may be needed.
 In some Global South contexts, postal systems may be unreliable;
email may substitute if addresses are available.
 For cross-national IR work, translation and local adaptation are vital.

Advantages

 Reduces interviewer bias.

 Allows respondents to answer at their convenience.
 Reaches geographically dispersed elites.

Limitations

 Low response rates (typically below 30%).

 Limited control over who completes it.
 Requires strong follow-up procedures.

Solution: Follow-up emails or institutional endorsements (e.g., letters from

universities or ministries) can increase response rates.

4.2.2 Interview

 Interviews provide deeper data by engaging respondents in verbal (or

sometimes video/phone) discussion guided by an interview protocol.
 Interviews generate in-depth qualitative data, often used when
studying elite behaviour, foreign policy processes, or sensitive
political topics (Mosley, 2013).

Types of Interviews

1. Structured Interviews: Same questions asked identically across

respondents.
2. Semi-Structured Interviews: Core questions with flexibility for
follow-up (common in IR).
25
 3. Unstructured Interviews: Open conversation guided by general
themes.

Examples in Political Science/IR

 Elite Interviews:
o Interviews with Kenyan diplomats about AU peacekeeping
missions.
o Interviews with Chinese foreign policy experts about Belt and
Road objectives.
 Citizen Interviews:
o In Myanmar, interviews with civil society activists on their
engagement with international NGOs during political reforms
(Schroeder & Sjoberg, 2022).

Strengths

 Rich, nuanced data; ability to probe, clarify; useful for elite or sensitive
topics.
 Provides depth and context.
 Allows exploration of complex motives and causal mechanisms.

Limitations

 Interviewer bias and social desirability effects.

 Limited generalisability.
 Time-consuming transcription and coding.
 Not easily standardized for large-N comparisons; more resource-
intensive; interviewer bias risk.

Tip: Always pilot the interview guide, record (with consent), transcribe, code
for themes. Use multiple coders for reliability in qualitative coding.

Ethical Note: In repressive environments (e.g., authoritarian states),

anonymity and data security must be prioritised (Abrahamsen, 2019).

26
4.2.3 Observation

 Observation involves systematically watching behaviour, events or

conditions and recording them.
o In political science/IR this might involve observing legislative
debates, protests, diplomatic negotiations, media coverage, or
international conferences.
 Observation involves systematically recording behaviours or events in
natural settings, helping researchers understand political actions as
they occur (Johnson & Reynolds, 2020).

Types of Observation:

 Participant observation: Researcher engages directly (e.g., attending

rallies) or researcher is present and may participate in the context.
 Non-participant observation: researcher watches without
involvement. The observer remains external.
 Structured observation: pre-defined categories/checklists.
 Unstructured observation: open-ended field notes.

Practical Examples:

 Protest Observation: Documenting youth mobilisation during the

“End SARS” protests in Nigeria. Researchers might record crowd
dynamics, leadership patterns, and protest symbols.
 Diplomatic Observation: Observing UN Security Council sessions to
analyse negotiation tactics or speech patterns.
 Election Observation: International observer missions (e.g., EU or
ECOWAS observers) systematically record procedural fairness during
elections.
 A researcher sits in on sessions of the African Union Peace and
Security Council in Addis Ababa to observe negotiation behaviour
among member-states.
o They record frequency of interruptions, alliance signals, tone of
speech.
27
  A researcher observes street protests in Manila to code protester
demands, police responses, crowd size, and uses a checklist to ensure
systematic data.

Design issues:

 The presence of an observer may change behaviour (observer effect).

 Coding must be consistent across observers (inter-rater reliability).
 Ethical clearance if observing sensitive contexts (e.g., armed conflict).

Advantages

 Provides behavioural data rather than self-reports.

 Context-rich and flexible.

Limitations

 Observer presence can alter behaviour (Hawthorne effect).

 Subjective interpretation may reduce reliability.

4.2.4 Experimentation

 Experiments involve the manipulation of one or more independent

variables to observe effect on a dependent variable, controlling for
other factors.
 In political science/IR, field experiments, survey experiments, or lab
experiments are increasingly used.
 Experiments test cause–effect relationships by manipulating
independent variables and observing outcomes.

Types of Experiments

 Lab Experiments: Controlled environment (common in political

psychology).
 Field Experiments: Real-world setting (used in development or
electoral studies).
 Survey Experiments: Randomly assign vignettes in a survey.
28
Practical examples:

 Field Experiment (Kenya): Voters randomly receive information

about anti-corruption campaigns. Researchers then measure shifts in
voting preferences (Kramon & Posner, 2013).
 Survey Experiment (India): Respondents exposed to differing
narratives about China’s trade policy—measure effects on attitudes
toward foreign policy cooperation.
 Lab Experiment (US): Participants read diplomatic crisis scenarios to
test theories of deterrence or alliance formation.
 A survey experiment in Nigeria where respondents are randomly
assigned to receive different vignettes about foreign aid (positive vs
negative framing) and then asked for policy support.
 A field experiment in India where a treatment group receives
information about EU trade negotiations, and then researcher
measures change in support for free trade.

Design considerations:

 Random assignment ensures comparability of treatment and control

groups (internal validity).
 Must consider ethical issues (consent, deception).
 External validity: Can results generalize beyond the experimental
sample? ([turn0search7])
 In IR/Global South contexts, field experiments often require
partnership with local organisations, language translation, logistical
coordination.

Strengths

 Strong internal validity (causal inference).

 Enables hypothesis testing.

Limitations

 Artificiality in lab settings.

29
  Ethical issues—deception or unequal treatment.
 Difficult logistics in unstable political contexts.

4.2.5 Task Analysis

 Task analysis examines the tasks or activities of participants,

organisations or systems to understand how they perform, what
decisions they make, and what steps they follow. In political
science/IR it might be used to analyse decision-making processes,
diplomatic protocols, or legislative workflows.
 Task analysis involves decomposing complex political or bureaucratic
activities into sequential steps to understand process flow and
decision-making.

Practical examples:

 In a study of peace negotiations, task analysis might map each step of

the negotiation process: agenda-setting, mediation, drafting,
ratification. Researchers may interview diplomats and refer to archival
negotiation minutes to construct the task flow.
 In analysing parliamentary committees, task analysis may code tasks
performed by committees (hearings, drafting legislation, oversight
visits) and time taken for each.
 Foreign Policy Analysis: Mapping the step-by-step decision process
during Ghana’s mediation role in the ECOWAS crisis (identifying
actors, timing, and communication).
 Legislative Processes: Analysing parliamentary budget approval
procedures—who initiates debate, what committees review, how
amendments are negotiated.
 Conflict Mediation: Studying peace negotiation tasks in Sudan
(agenda-setting, mediation, ceasefire drafting).

Why useful:

 Clarifies procedural bottlenecks.

30
  Helps to operationalize complex organizational or institutional
processes.
 Useful for comparative institutional research (e.g., comparing
domestic policy-making processes across states).
 Encourages systematic breakdown of process into measurable pieces.
 Aids in policy design and process optimization.

4.2.6 Delphi Technique

 The Delphi technique is a structured communication method originally

developed as a forecasting tool.
o It involves iterative rounds of expert surveys, feedback, and
revision until consensus emerges.
o In political science/IR it can be used to forecast conflict,
international cooperation, or policy trends.
 The Delphi method is a structured, iterative process of gathering expert
opinions until consensus is reached (Linstone & Turoff, 2002).

Steps

1. Identify expert panel.

2. Round 1: Collect responses anonymously.
3. Summarize and share results.
4. Round 2: Experts revise views.
5. Continue until convergence.

Practical example:

 A Delphi study of Asia-Pacific security experts predicts likely

outcomes of China-US competition in Southeast Asia over the next 10
years.
o Experts complete a first-round questionnaire; their aggregated
responses are summarised and circulated; a second round
allows experts to revise their views in light of group feedback;
this continues until consensus or stability is reached.

31
  Another example: A Delphi panel of African foreign policy analysts
forecasting the evolution of the African Continental Free Trade Area
(AfCFTA) implementation.
 Forecasting African Union’s peace and security trajectory through
rounds of consultation with policy analysts.
 Predicting ASEAN responses to U.S.–China rivalry by surveying
regional security experts over multiple rounds.

Advantages

 Reduces groupthink via anonymity.

 Incorporates expert judgment on emerging issues.
 Ideal for policy forecasting and strategic planning.
 Access to expert insight, structured iterative refinement, can generate
forecasts or consensus.

Limitations

 Requires careful expert selection.

 May take long to reach consensus.
 Not suitable for large-N studies.
 Depends on expert selection (sampling bias), may not guarantee
accuracy, time-consuming.

32
4.2.7 Validity and Reliability of Research Instruments

 Instruments must be evaluated for validity (do they measure what

they aim to?) and reliability (do they measure consistently?).
o Without both, your findings will be questionable.
 Measurement quality depends on two essential properties: validity
(accuracy) and reliability (consistency) (Creswell & Creswell, 2018).
 Validity concerns the accuracy of inference from the data collected to
the concept intended. There are multiple types of validity:

Characteristics of Sound Measurement

 In Political Science and International Relations (PSIR), sound

measurement is crucial for producing credible and replicable
findings—whether we are measuring democracy, political attitudes,
institutional performance, or levels of interstate conflict.
 Two key characteristics of sound measurement are validity and
reliability (Babbie, 2013).

1. Validity

 Validity refers to the degree to which a measurement tool actually

measures what it intends to measure (Adcock & Collier, 2001).
 In PSIR, this means ensuring that indicators of concepts such as
power, legitimacy, or sovereignty truly reflect the underlying
theoretical ideas.
 Validity determines whether an instrument truly measures the
concept it intends to measure.
o For example, when constructing indices such as the Polity IV
Democracy Score or the Corruption Perceptions Index,
researchers must ensure that indicators correspond accurately
to the theoretical concept (Adcock & Collier, 2001).

33
1. Internal Validity

 Internal validity refers to whether the causal relationship observed

between variables is trustworthy and not influenced by confounders
(Shadish, Cook & Campbell, 2002).
 Internal validity concerns whether observed effects in a study result
from the variables being investigated rather than external or
confounding factors (Shadish, Cook & Campbell, 2002).

 Example: When studying the impact of military alliances on state

security, high internal validity means we can confidently attribute
changes in security levels to alliance membership—not to
economic conditions or regime type.
 Example: A researcher examines whether UN sanctions reduce
authoritarian resilience.

 High internal validity requires ruling out alternative explanations such

as economic decline, domestic protests, or external military pressure.

2. External Validity

 External validity refers to the extent to which research findings can be

generalized beyond the specific case, sample, or setting studied (King,
Keohane & Verba, 1994).
 External validity concerns the extent to which findings can be
generalized to other cases, populations, or time periods.

 Example: A case study of democratization in Tunisia may have

limited external validity if its conclusions cannot be generalized to
other Middle Eastern states.
 Example: A case study on South Sudan’s peace process may have
limited external validity if its findings cannot be applied to
peacebuilding efforts in Colombia or Sri Lanka (George & Bennett,
2005).

34
Types of Validity of Measurement

 Beyond internal and external validity, the accuracy of measurement

instruments is assessed through several specific types of validity.

i. Content Validity

o Content validity evaluates whether a measurement

covers the full range of the concept being studied
(Babbie, 2013).
o A measure has content validity when it captures the
full domain of a concept (Babbie, 2013).

o Example: A State Fragility Index should

include multiple dimensions:
 Political (e.g., regime legitimacy)
 Social (e.g., human development)
 Economic (e.g., inequality)
 Security indicators (e.g., internal conflict)
o Omitting any major dimension weakens
content validity.
o Example: A democracy index has strong
content validity if it includes political rights,
civil liberties, electoral competitiveness, and
rule of law.

ii. Construct Validity

o Construct validity assesses whether a measurement

aligns with theoretical expectations and correlates
appropriately with related variables (Adcock &
Collier, 2001).
o Construct validity evaluates whether a measure
behaves as expected in relation to other theoretical
variables (Adcock & Collier, 2001).

35
  Example: If a measure of nationalism
correlates with anti-immigration attitudes,
this supports its construct validity.
 Example:
 A new measure of soft power should:
o Positively correlate with cultural
exports
o Negatively correlate with global
distrust ratings
o Align with theoretical expectations in
Nye’s soft-power framework
 If these patterns appear, construct validity
is supported.

iii. Face Validity

o Face validity refers to whether a measurement

appears, on the surface, to represent what it is
supposed to measure.
o Although subjective, it is often a first step in
instrument development.
o Face validity asks whether the measure appears
logical "on its face."

 Example: A survey question like “How

often do you participate in political
discussions?” seems to have face validity as
an indicator of political engagement.
 Example: The question “How often do you
vote in national elections?” has face validity
as an indicator of political participation.

iv. Criterion-Related Validity

 Criterion-related validity measures how well an indicator correlates

with an established external criterion (Punch, 2014).
36
  Criterion validity tests whether a new measure correlates with an
established “gold standard.”

 Example: If a corruption perception index correlates strongly with

Transparency International’s rankings, it exhibits strong criterion
validity.
 Example: If a new Authoritarianism Scale correlates strongly with
Freedom House’s “Not Free” category, it demonstrates criterion
validity (Punch, 2014).

A simple formula for criterion validity is the Pearson correlation

coefficient (r):

∑( − ¯ )( − ¯ )
=
∑( − ¯ ) ⋅ ∑( − ¯ )
Where:

 X = your measure
 Y = benchmark measure

 The Pearson correlation coefficient (\(r\)) measures the strength and

direction of the linear relationship between two continuous variables,
with a value ranging from -1 to 1.
 A value of 1 or -1 indicates a perfect positive or negative linear
correlation, respectively, while 0 indicates no linear relationship.
 The sign of \(r\) shows the direction (positive or negative), and the
absolute value shows the strength (from weak to strong).

Understanding the value of \(r\)

 Sign:
 Positive (\(r>0\)): As one variable increases, the other variable
tends to increase.
 Negative (\(r<0\)): As one variable increases, the other variable
tends to decrease.

37
  Zero (\(r=0\)): There is no linear relationship between the two
variables.

Magnitude (strength):
 Strong (| |≥0.5): A strong correlation indicates that the data
points are close to a straight line.
 Moderate (| |≈0.3−0.5): A moderate correlation shows a less clear
but still present relationship.
 Weak (| |<0.3): A weak correlation means the data points are
more scattered and the linear relationship is weak.

Example:

a. Positive Correlation (as one increases, the other increases)

 Education level and voter turnout

In many democracies, individuals with higher educational attainment
are more likely to vote.
→ As education increases, voter turnout also tends to increase.

 Campaign spending and vote share

Candidates who spend more on campaigns often receive a larger
share of the vote.
→ More campaign spending is associated with more votes.

b. Negative Correlation (as one increases, the other decreases)

 Government approval ratings and economic inflation

As inflation rises and prices increase, approval of the incumbent
government tends to drop.
→ Higher inflation relates to lower government approval.

 Strength of democratic institutions and corruption levels

Countries with stronger, more transparent democratic institutions
generally report lower corruption.
→ Stronger institutions correlate with lower corruption.
38
c. No Correlation (no consistent relationship

 Voter age and support for specific foreign policy issues (in some
contexts)
In some countries, age does not reliably predict whether someone
supports a specific foreign policy stance (e.g., a trade agreement).
→ Age does not predict position on the issue.

 Party identification and views on local zoning laws

Partisanship may not have a meaningful relationship with opinions on
very localized, technical zoning decisions in some communities.

3. Reliability of Measuring Instruments

 Reliability refers to the consistency or stability of a measurement

across time, forms, or observers (Bryman, 2016).
 Reliability indicates whether a measurement instrument yields
consistent results over time or across items (Bryman, 2016). A
measure can be reliable without being valid, but validity requires
reliability.
 In PSIR, researchers require reliable measures to ensure that shifts in
variables such as conflict intensity or public opinion reflect real
change—not flaws in measurement.
 Example: The Global Peace Index must consistently measure
conflict intensity yearly; otherwise, changes may reflect
measurement error rather than real-world changes.

Tests of Reliability

i. Test–retest Reliability

 This method involves administering the same instrument to the same

respondents at two points in time.
 The same test is administered to the same subjects at two points in
time.

39
  Example: A political trust scale given to citizens twice within two
weeks should yield similar results if trust is stable (Babbie, 2013).
 Example: A survey measuring citizen trust in parliament should
produce similar results if administered again within a short
period—assuming no major political crisis occurred.

Formula: Pearson correlation between Time 1 and Time 2 scores

A common way to assess test–retest reliability is the Pearson

correlation coefficient between Time 1 and Time 2 scores:

∑( − ¯ )( − ¯ )
=
∑( − ¯ ) ⋅ ∑( − ¯ )

Where:

 = scores on the measure at Time 1

 = scores on the measure at Time 2
 ¯ and ¯ = mean scores at Time 1 and Time 2

This coefficient (rₜₜ) shows how consistent a political measure is over time.

Examples:

a. High Test–Retest Reliability (strong positive correlation)

 Political ideology scales (e.g., liberal–conservative self-placement)

People usually provide similar ideology scores when surveyed twice
within a short period.
→ If someone rates themselves a “3” (liberal) at Time 1, they tend to
give a similar rating at Time 2.
This would produce a high rₜₜ, indicating stable ideology.

40
b. Moderate Test–Retest Reliability

 Trust in government
Public trust fluctuates with events (political scandals, economic news),
but is not completely random.
→ Scores may be somewhat consistent over weeks but still shift.
This produces a moderate rₜₜ.

c. Low Test–Retest Reliability (weak or near-zero correlation

 Short-term policy opinions influenced by recent events

Opinions on rapidly changing issues (e.g., immediate reactions to a
foreign policy crisis or scandal) may vary widely between Time 1 and
Time 2.
→ People might answer very differently after new developments.
This results in a low rₜₜ, indicating instability.

c. Negative Test–Retest Correlation (rare, but possible)

 Survey items interpreted differently over time

Suppose a survey question about “support for political reform” is
interpreted positively at Time 1 but, after a news event, respondents
interpret “reform” negatively at Time 2.
→ Those who rated high at Time 1 may now rate low at Time 2.
This could produce a negative rₜₜ, indicating inconsistent
measurement.

ii. Alternate (Multiple) Forms

 Researchers create two equivalent versions of an instrument and

administer them to the same respondents (Neuman, 2014).
 Two equivalent versions of the same instrument are administered.

41
 o Example: Two parallel tests measuring knowledge of
international treaties (e.g., the Geneva Conventions, UN
Charter) should yield similar scores.
o Example: Two parallel forms of an international relations
knowledge test should correlate highly.

iii. Split-Half Technique

 A single instrument is divided into two halves (e.g., odd vs. even
items), and the results are correlated. The instrument is split into two
halves, and results are compared.
o Example: A political efficacy questionnaire’s two halves should
produce similar scores.
o Example: A political ideology scale divided into odd and even
items should produce highly correlated scores.

The Spearman–Brown prophecy formula is used to estimate the

reliability of a full test based on the correlation between two halves of the
test:

2
SB =
1+

Where:

 = the correlation between the two halves of the measurement (e.g.,

two halves of a survey scale).

The formula predicts how reliable the full-length measure is when

combining both halves.

Examples:

a. Measuring Political Ideology with a Multi-Item Scale

A researcher splits a 10-item ideology scale into two halves (items 1–5 and
6–10).

42
  Correlation between halves: = 0.70

Using the Spearman–Brown formula:

2(0.70)
SB = ≈ 0.82
1 + 0.70

Interpretation: The full ideology scale is highly reliable, meaning

respondents answer consistently across items.

b. Measuring Trust in Government

A survey includes 8 items measuring trust in government. The two halves

correlate only moderately.

 Correlation between halves: = 0.45

2(0.45)
SB = ≈ 0.62
1 + 0.45

Interpretation: The full scale has moderate reliability, suggesting trust in

government is somewhat consistent but also influenced by current events.

c. Measuring Support for Environmental Policy

A political scientist uses a short 6-item scale about support for

environmental regulation.

 Correlation between halves: = 0.20

2(0.20)
SB = ≈ 0.33
1 + 0.20

43
Interpretation:
The scale shows low reliability, possibly because the items are too broad
(e.g., mixing climate policy, pollution controls, and energy policy).

d. Very High Reliability Example (Elite Survey)

A survey of political experts includes two sets of items measuring attitudes

toward democratic norms.

 Correlation between halves: = 0.90

2(0.90)
SB = ≈ 0.95
1 + 0.90

Interpretation: Experts respond very consistently, indicating a highly

reliable measure of democratic-norm support.

iv. Cronbach’s Alpha Test

 Cronbach’s Alpha assesses internal consistency for multi-item scales.

 Cronbach’s Alpha assesses the internal consistency of multi-item
scales.
 Values above 0.70 are generally considered acceptable (Tavakol &
Dennick, 2011).
o Example: A scale measuring attitudes toward global
governance may include several items; Alpha tests whether
these items measure the same underlying concept.
o Example: A 10-item scale measuring attitudes toward regional
integration in ECOWAS should show high alpha if all items
measure the same underlying orientation.

Cronbach’s Alpha Test Formula:

44
  Cronbach’s alpha estimates the internal consistency reliability of a
multi-item scale:

∑
= 1−
−1

Where:

 = number of items in the scale

 = variance of each individual item
 = variance of the total scale score

A higher α means the items measure the same underlying construct more
consistently.

Interpretation:

 Above 0.70 = acceptable

 Above 0.80 = good
 Above 0.90 = excellent (Tavakol & Dennick, 2011)

Examples from Political Science:

a. Measuring Political Efficacy

A researcher uses a 5-item scale to measure internal political efficacy (e.g.,

“I feel confident in understanding political issues”).

 Items vary moderately but are strongly related.

 Suppose:
o =5
o The sum of item variances is relatively low compared with the
total score variance.

This produces, for example:

= 0.85
45
Interpretation: The items work well together and consistently measure
political efficacy.

b. Measuring Support for Democracy

A 7-item scale asks respondents about democratic principles (free

elections, civil liberties, accountability, etc.).

 Some items tap different sub-aspects (e.g., elections vs. minority

rights), so item variances may be somewhat higher relative to total
variance.

This yields something like:

= 0.70

Interpretation: The scale has acceptable reliability, but items may reflect
slightly different components of democratic attitudes.

c. Measuring Populist Attitudes

A 10-item populism scale includes questions about “the people,” “the

elite,” and anti-establishment ideas.

 Because populism is multidimensional, item variances may not align

well.

Example:

= 0.55

46
Interpretation: Low internal consistency: the items may not measure a
single construct well. The scale might need revision.

d. Measuring Trust in Government

A short 3-item scale asks about trust in Congress, the executive, and
government overall.

 Short scales often yield lower α because fewer items contribute to

total variance.

Example:

= 0.60

Interpretation: Moderately reliable, but reliability might improve by

adding more items or refining wording.

Practical Application in PSIR Research

Example 1: Measuring Democracy

 Validity issue: Does the indicator capture both political rights and
civil liberties?
 Reliability issue: Does the democracy index produce similar scores
across consistent time periods?

Example 2: Measuring Conflict Intensity

 Validity: Including fatalities, duration, displacement, and weapons

used improves content validity.
 Reliability: Uppsala Conflict Data Program (UCDP) ensures reliability
through standardized coding procedures.

Example 3: Measuring Public Support for Foreign Policy

47
 Survey items measuring trust in international institutions should
correlate (construct validity).
 Re-administering the survey two weeks later checks test–retest
reliability.

48
4.3: Literature Review and Theoretical Framework

4.3.1 Meaning and Purpose of Literature Review

Meaning

 A literature review is the systematic examination, interpretation, and

synthesis of existing scholarly knowledge related to a specific
research problem.
 It involves summarizing what scholars, policymakers, and practitioners
have already written or discovered about a topic, identifying key
debates, and highlighting areas requiring further exploration (Booth,
Sutton & Papaioannou, 2016).

In Political Science and International Relations, the literature review

helps you anchor your study in established scholarship. It is not just a
summary but a critical conversation with previous researchers (Hart, 2018).

It asks:

 What do we already know about this phenomenon?

 What don’t we yet understand?
 How can my research add new insights or challenge existing
paradigms?

Purpose

The literature review serves several purposes:

1. Contextualization:
It positions your research within existing scholarly debates.
o Example: When examining China’s Belt and Road Initiative (BRI)
in Africa, you locate your study within literature on South-South
cooperation, dependency theory, and global power shifts
(Alden & Large, 2019).
2. Identifying Gaps:
Helps to highlight what’s missing in the literature.
49
 o Example: Many studies explore China’s influence in Africa, but
few examine how African states exercise agency and negotiate
conditions — your study might fill that gap (Tull, 2006).
3. Avoiding Duplication:
Prevents repeating existing research unnecessarily.
4. Refining Research Questions:
Reading existing works helps clarify what specific angle you should
take.
o Example: You might narrow a broad topic like “African foreign
policy” to “Nigeria’s regional security diplomacy in ECOWAS.”
5. Theoretical and Conceptual Framing:
Guides the selection of appropriate theories or conceptual
frameworks (e.g., realism, constructivism, dependency theory).
6. Demonstrating Scholarly Competence:
A strong review shows that you understand the key debates and
schools of thought in your field.

Practical Illustration

Suppose you are researching “The Impact of Chinese BRI Loans on

African States’ Foreign Policy Autonomy.”

 You review prior works on Chinese economic diplomacy (Brautigam,

2011), dependency and development theory (Frank, 1967), and
African agency in IR (Brown & Harman, 2013).
 You find that while scholars discuss “debt diplomacy,” few provide
empirical evidence on how debt shapes African voting behavior in the
UN General Assembly.
 Your study then aims to close that gap.

50
4.3.2 Scope of the Literature Review

Meaning

 The scope defines the boundaries of your review — what to include

and what to exclude. It ensures focus and coherence.
 Without scope, your review risks becoming too broad and unfocused
(Creswell & Creswell, 2018).

Components of Scope

1. Thematic Scope:
Defines the main themes — e.g., foreign aid, diplomatic behaviour,
regional integration, etc.
o Example: For a study on African peacekeeping, you might
include literature on AU operations, UN mandates, and regional
security, but exclude unrelated works on domestic politics.
2. Temporal Scope:
Specifies the period under review.
o Example: You might focus on 2013–2025 for studies on the BRI
or 1990–2024 for post-Cold War African diplomacy.
3. Geographical Scope:
Defines your area of interest.
o Example: Focusing only on East Africa or ASEAN member states.
4. Conceptual Scope:
Clarifies key concepts and variables.
o Example: In studying “foreign policy autonomy,” decide how it
will be measured (e.g., UN voting alignment, aid dependence).
5. Methodological Scope:
States what kind of studies are reviewed — qualitative, quantitative,
or mixed methods.

Practical Example

If your study is “African Agency in the Belt and Road Initiative”:

51
  Temporal scope: 2013–2024 (BRI period)
 Geographic scope: East and Southern Africa
 Conceptual scope: Agency, economic diplomacy, sovereignty
 Methodological scope: Peer-reviewed journals, policy briefs, and
think-tank reports

This clarity ensures that your review remains focused and relevant.

4.3.3 Systemizing the Search

A literature review must be systematic, not random. This means using a

clear, replicable process to find and document relevant studies.

Steps to Systemize the Search

1. Define Keywords and Phrases

o Combine core concepts and synonyms:
Example:
“Foreign policy autonomy” AND “Africa” AND “China”;
“BRI” AND “small states” AND “diplomacy”
o Use Boolean operators (AND, OR, NOT) for precision.
2. Select Databases and Sources
o Academic: JSTOR, Scopus, Web of Science
o Policy/Think Tanks: Brookings, SAIIA (South African Institute of
International Affairs), CSIS, ISEAS–Yusof Ishak Institute
o Regional: African Journals Online (AJOL), Asian Political Science
Review
o Institutional repositories (e.g., AU, ASEAN, UN archives)
3. Apply Inclusion and Exclusion Criteria
o Include: Peer-reviewed, English-language studies, 2010–2025.
o Exclude: Duplicates, non-scholarly blogs, outdated Cold War
material (unless historically relevant).
4. Record Search Results
o Create a log: database, keywords, number of hits, selected
works, reasons for exclusion.
5. Use Reference Management Tools
52
 o Zotero, Mendeley, or EndNote to manage and tag articles.

Practical Example

Researching “African Foreign Policy toward China”:

 Search “China Africa diplomacy,” “BRI Africa,” “African foreign policy,”

“agency in asymmetrical relations.”
 Use JSTOR and Scopus; filter for 2010–2025; review top-cited works.
 Record in Excel or Zotero — note methodology, key findings, and
gaps.

Outcome: A systematic, transparent, and reproducible search strategy that

enhances credibility.

4.3.4 Reading the Literature and Making Notes

Once collected, reading and note-taking must be strategic — focused on

extracting what matters.

Stages of Reading

1. Scanning: Quickly review abstracts, keywords, and conclusions to

assess relevance.
2. Deep Reading: Read selected works carefully, noting theoretical
approach, methodology, findings, and limitations.
3. Critical Analysis: Evaluate strengths, weaknesses, and biases.
4. Synthesis: Integrate insights thematically.

Note-Taking Techniques

 Summary Tables:

53
 Author Year Theory Method Region Findings Gaps
BRI Lacks
Alden Case
2019 Dependency Africa enhances African
& Large Study
influence perspective

 Concept Mapping: Visually link authors and themes to see

convergences and divergences.
 Annotation: Use digital tools (Zotero highlights, sticky notes) to
record thoughts while reading.
 Analytical Memos: Write reflective notes on how each piece
connects to your research.

Synthesizing vs Summarizing

Avoid writing a “book report.”

Instead of saying:

“Author A says this, Author B says that,”

do this:
“Scholars diverge on whether small states exhibit agency in asymmetric
relations. Dependency theorists (Frank, 1967; Amin, 1976) emphasize
structural limits, while constructivists (Wendt, 1999) highlight normative
autonomy. This study integrates both perspectives.”

Practical Example

In your study on ASEAN and China’s BRI:

 Group literature under themes like Economic Dependence, Political

Agency, Regional Cooperation, and Security Autonomy.
 Compare findings:
o Malaysian scholars emphasize economic pragmatism (Liow,
2019).
o Western scholars stress vulnerability and dependency
(Ferdinand, 2016).

54
 o Your synthesis might propose a nuanced model: “Negotiated
Autonomy in Asymmetric Partnerships.”

4.3.5 Theoretical Framework

Meaning

A theoretical framework is the foundation of analysis — the lens through

which you interpret data. It outlines the relationships among key concepts
and variables and explains why phenomena occur (Punch, 2014).

While the literature review summarizes what is known, the theoretical

framework organizes your argument.

Functions

1. Guides data collection and interpretation.

2. Connects your study to established theories.
3. Provides explanatory mechanisms for observed phenomena.
4. Helps formulate hypotheses.

Developing the Framework

1. Identify Core Concepts

Example: Dependency, Agency, Power Asymmetry, Foreign Policy
Autonomy
2. Select Theories Relevant to Your Study
o Realism: Power and national interest dominate state behaviour.
o Liberalism: Cooperation and institutions mitigate conflict.
o Constructivism: Identities and norms shape interactions.
o Dependency Theory: Unequal economic relations constrain
autonomy.
o Postcolonial Theory: Power and knowledge reproduce
hierarchies (Acharya, 2014).

55
 3. Define Variables and Relationships
Example:
o Independent variable: External debt to China
o Dependent variable: Policy alignment with Chinese positions
o Moderating variable: Institutional strength or domestic
opposition
4. Develop Hypotheses or Propositions
o “Higher economic dependence leads to reduced foreign policy
autonomy.”
o “Domestic political contestation mediates external influence.”
5. Illustrate the Model

External Dependence ───► Policy Autonomy

▲
Domestic Institutions

Practical Example

A researcher studying Kenya’s foreign policy behavior under the BRI may
build a framework combining Dependency Theory (Frank, 1967) and Small
State Agency Theory (Hey, 2003):

 Argues that Kenya’s high economic dependence constrains

autonomy.
 However, domestic political coalitions sometimes allow strategic
negotiation — showing partial agency.
This dual framework better captures Africa’s nuanced engagement
with great powers.

Conclusion

A strong literature review and theoretical framework are the intellectual

backbone of any research project in Political Science and IR.

They help:

56
  Situate your study within existing debates.
 Clarify research gaps and questions.
 Justify theoretical and methodological choices.
 Ensure your work contributes meaningfully to scholarly discourse.

By systematically searching, critically synthesizing, and grounding your work

in theory, your research achieves both rigor and relevance — especially in
the complex, multipolar world of the Global South.

57
4.4 Ethics in Political Science Research

4.4.1 Ethics in Political Science Research

 Meaning and importance

Ethics in research refers to the set of moral principles guiding how
researchers design, conduct, interpret and report their studies.
 In political science and IR, ethical considerations are crucial because
research often involves human subjects (citizens, elites, refugees, etc.),
sensitive political topics (conflict, authoritarianism, human rights) and
data that can affect reputations, policy, or people’s lives.

Key ethical principles include:

 Informed consent: Participants must understand what the research

is, what their involvement entails, and voluntarily agree to participate.
 Anonymity and confidentiality: When dealing with vulnerable
groups (e.g., dissidents in repressive states, refugees), the researcher
must protect identities and personal data.
 Minimising harm: The research should not expose participants to
physical, psychological, social or political risks.
 Transparency and honesty: Data must be reported honestly; no
fabrication, falsification or misrepresentation.
 Avoiding conflicts of interest: Researchers must disclose funding
sources, affiliations, and potential biases.
 Respect for political and cultural contexts: Political science
research often crosses national, cultural or ideological boundaries;
respect for local norms, power dynamics and histories is necessary.

Political Science / IR practical examples

1. Research with elites in authoritarian states

Suppose a researcher interviews opposition activists in a one-party
state about foreign policy dissent. Ethical issues include: Are
participants safe? Does being interviewed expose them to surveillance

58
 or repression? Ensuring anonymity and safe data storage becomes
critical.
2. Survey research in conflict zones
When surveying citizens in a war-torn country (e.g., South Sudan,
Myanmar), ethical issues include: Are respondents a vulnerable
population? Could their answers expose them to reprisals?
Researchers may have to use remote/anonymous methods or avoid
certain questions.
3. Using archival / big data sources
For example, a researcher analysing diplomatic cables (e.g., from
Wikileaks) or social media posts of political actors must consider
privacy, consent, and whether public posts still require ethical
treatment. While the data may be public, the political consequences
might not be.
4. Plagiarism and research integrity
In political science, ethics go beyond human subjects to include
integrity in writing, citing and data use. For instance, if a scholar uses
large portions of another’s article without proper attribution, that is
unethical. A recent case shows elected officials’ doctoral theses were
found to contain major plagiarism, raising issues about political
accountability.

Why ethics matter in IR

 Political science research often influences policy, public debates,

human rights, and governance. Unethical research can mislead policy,
harm populations or damage public trust.
 Ethical lapses (fabricated data, plagiarised texts, mis-represented
results) weaken the discipline’s credibility.
 IR involves power inequalities (North/South, developed/developing,
major/minor states). Ethical reflection helps avoid perpetuating
colonial or exploitative research practices.

59
Checklist for Ethical Research in Political Science / IR

 Did you obtain and document informed consent (written or verbal)

for each participant?
 Is your data collection sensitive to the participants’ context (political,
cultural, linguistic, security)?
 Are anonymity and confidentiality assured, especially for vulnerable
groups?
 Have you assessed potential harms to participants or communities
and planned mitigation?
 Are your data storage, access, sharing protocols secure?
 Are you transparent about your funding, affiliations, assumptions, and
potential bias?
 Do you ensure credit and attribution for others’ ideas/data, avoid
plagiarism and properly cite sources?

4.4.2 Practical Tips for Avoiding Plagiarism

What is plagiarism?
Plagiarism is the use of another person’s ideas, words or data without
proper attribution — presenting them as your own. It includes direct
copying, paraphrasing too closely, and failing to cite.

Why it matters in Political Science / IR

 Scholars may reuse others’ theoretical frameworks, case descriptions,

data sets or legal texts. Failure to attribute undermines scholarly
ethics and can lead to reputational damage (for individuals and
institutions).
 In IR, plagiarised political analyses or policy briefs can mislead
governments, donors or publics.
 Plagiarism scandals have affected politicians, researchers and
universities — e.g., major dissertations found to contain large
portions of un-credited text.
60
Common types of plagiarism and examples

 Verbatim copying: Copying text word-for-word without quotation

marks or credit. Example: A student writes: “NATO’s founding
Washington Treaty is permanent …” without quotation marks or
proper citation.
 Paraphrasing too closely: Changing only a few words but keeping
original structure, without citation. Example: Taking an article’s
sentence, changing a few words, and presenting it as your own.
 Idea plagiarism: Using someone’s unique idea or framework without
citation.
 Self-plagiarism: Re-using your own previous work without
disclosure/attribution.
 Data plagiarism or fabrication: Presenting someone else’s data as
your own, or altering data. Ethical issues extend beyond wording.

Practical tips to avoid plagiarism

1. Keep track of sources from the start

o Maintain detailed notes of all articles, books, chapters,
webpages you consult.
o Use reference-manager software (e.g., Zotero, Mendeley) to tag
and organise.
2. Use quotation marks and citations for direct quotes
o Any time you use text verbatim, use quotation marks and
include author, year, page number.
o Example: “The modernization school … argued that developing
countries would have to modernize their traditional social forms
…” (Rostow, 1960: 23).
3. Paraphrase properly and cite
o Re-write the idea in your own words and still cite the source.
Don’t just change a word here and there.
o Example: Original: “One would expect that the afflictions of the
Anglo-American model would boost confidence in … social

61
 market capitalisms of Western Europe.” (Öniş & Güven,
2011:475) — Plagiarised by slight alteration and no citation.
4. Cite ideas, not just direct quotes
o If a concept, framework or argument is someone else’s, cite it
even if you did not quote directly.
o Example: If you adopt dependency theory (Frank, 1967) to
analyse small-state foreign policy, cite Frank’s work.
5. Use plagiarism-detection tools but don’t rely solely on them
o Tools like Turnitin or iThenticate help detect overlaps, but they
don’t replace sound citation practices.
6. Be transparent about data, methodology, collaboration
o If you worked with others, acknowledge their contributions. If
you reused datasets or translated interviews, state so.
7. Understand institutional and cultural norms
o Plagiarism perceptions vary globally, but core principles remain.
Even if one country’s citation culture differs, international
standards expect proper attribution.
8. Avoid last-minute “patchwork” writing
o Rushed writing often leads to poor citation, improper
paraphrasing and unacknowledged borrowing. Allow time for
revision and properly integrate sources.

Example of best practice in Political Science / IR

A PhD student researching “Small State Diplomacy in the Indian Ocean”
writes a chapter on theory. They include the following:

According to dependency theory, peripheral states often face structural

constraints that limit their diplomatic autonomy (Frank, 1967). In the
context of small island states, Smith (2015: 88) argues that external
economic leverage diminishes policy choice.
Here the student cites both the foundational theory and a more recent
application. They then proceed with their own analysis, referencing these
ideas correctly.

62
Consequences of plagiarism in IR/Political Science

 Academic sanctions (failed thesis, revoked degree)

 Loss of credibility in scholarly and policy circles
 For politicians and public figures, such scandals can become matters
of public trust and governance accountability.
 Undermines research reliability: if sources are mis-represented, entire
arguments may collapse.

Conclusion

 Ethical conduct is foundational in Political Science and International

Relations research. It extends from how we treat people (participants,
communities) to how we treat ideas (attribution, honesty).
 By adhering to ethical principles—obtaining informed consent,
minimizing harm, ensuring confidentiality—and by vigilantly avoiding
plagiarism, researchers maintain credibility, integrity and contribute
responsibly to scholarship and policy.
 In your projects: build ethics into your design from day one, and
cultivate good citation habits throughout your writing.

63