GOVT 301

WM, Dan Doherty Spring 2025

4 dichotomies- Different ways to categorize research

Descriptive/Casual, Idiographic/Nomothetic, Qualitative/Quantitative, Inductive/Deductive

Idiographic

Thorough explanation of a particular event incorporating “all” or as many as possible factors.

Nomothetic

Partial explanations of a general phenomenon. One or few factors to explain a general class of events/outcomes/conditions.

Intervening variables

Variable the impacts DepV and is itself affected by an IndV. Helps explain how ur X influences ur Y

Antecedent variables

An Ind V that precedes another Ind V in time and affects it.

Dichotomous Variables

Impacts of the level of a variable or the probability of a variable (Voting vs not voting)

Political Science

Identifying patterns in human political behavior, (voting, lobbying, marching, revolutionizing, etc) in political institutions (Congress, nation states, EPA), or in the social structures that relate to politics (families, internet, schools/colleges, etc.)

Theories

Zero in on one or some factors of a phenomenon and provide a casual story. 

Hypothesis

A statement proposing a relationship between two or more variable OR a testable expectation that follows from a theory OR a statement that ought to be observed in the real world if theory is correct.

Conceptualization

A precise definition of a concept for the purpose of a specific research project.

Conceptualization of a variable

A precise definition of a characteristic for the purpose of a specific research project.

Operationalization

The process of developing specific research procedures that result in actual data- in empirical observations of the concept

Indicator

A sign of the presence or absence, level or amount of the concept being studied. (Are there protests? How many? How often? How large?)

Freedom

The opportunity to act spontaneously in a variety of fields outside the control of the government and other centers of potential denomination according to broad categories- political rights and civil liberties.

Political Rights

Electoral Process, Political Pluralism and Participation, Functioning of Government, Additional Discretionary Political Rights

Civil Liberties

Freedom of Expression and Belief, Associational and Organizational Rights, Rule of Law, Personal Autonomy and Individual Rights

4 Levels of Measurement

Nominal, Ordinal, Interval, Ratio

Nominal (Separate)

Different values/numbers only indicate different categories. Also called categorical

Ordinal (and order)

Different values allow us to rank or order the cases in meaningful way

Interval (and space)

Different values represent consist degree of variation

Ratio

Same as interval, with zero indicating absence of the concept

Measurement errors

Difference between the real value of an attribute and what the researcher comes up with in the process of measuring

Random vs Systematic Measurement Errors

• Almost all measurement errors are systematic. Not knowing where the error is doesn’t make it random

• Bad: random measurement error in Indp Variable, systematic error in DV or IV (all create bias)

• Not so bad: error in Dep. Variable (no bias)

Accuracy in Measurement: Reliability

the extent to which an instrument yields the same results on repeated trails

Accuracy in Measurement: Validity

The correspondence between measurement and instrument yields and the concept it is supposed to measure

Face, Content, and Criterion/Predictive Validity

• Does it make sense on its face?

• Does a measure cover all of the dimensions we are interested in? Does it cover the full range of meanings included within the concept?

• Does the measure help us predict things we would expect it to help predict?

Literature Review

A somewhat brief written review of the existing research and theories relevant to one topic

Purpose of a Lit Review

to discuss existing bodies of knowledge to which your research would add

Sometimes Lit Reviews… (2)

presents contrasting perspectives

describes strengths and weaknesses of that existing research

Lit Reviews Often… (1)

presents the relevant body of knowledge without contrasting or pointing out strengths and weaknesses

Population

a collection of units about which you are interested in making claims

Census

a gathering of data about all cases in a population (not often used in poli sci)

Sample (3)

a subset of the pop. of interest.

Should be representative of the pop. in its entirety.

Generalizable, externally valid.

Selection bias

occurs when certain types in a pop are more/less likely to be chosen for a sample

Probability samples

 a sample selected from a population using some chance process

Samples…

allow us to make an estimate about the population

Probability samples (3)

• Simple random sample

• Stratified sample: choosing a demographic and dividing your pop. by it, then taking a proportional amount from each demographic

• Cluster sampling: divides a population into groups (clusters) and randomly selects entire clusters for the sample

What’s important (in sampling)…

is that each unit in your pop has the same probability of being chosen for your sample

Non-probability samples (3)

• Haphazard sample: go down to confusion corner and ask ppl

• Purposive sample: Actively choosing an unrepresentative sample to learn more from it

• Snowball/chain referral sample: find one person part of pop, and that person connects you to another person in the pop.

• Non-probability samples

a sample selected from a pop not using a method based on a chance process.

Selection bias is…

almost guaranteed

Content Analysis

a procedure by which communications (words or images) are transformed into quantitive data

Ex: speeches, debates, movies, political ads, social media

2 steps of content analysis

Creating a sample

coding

sampling plan (3)

Look at every case in your universe OR

sample randomly from your universe OR

restrict universe then sample randomly.

sampling plan has a…

goal of a representative sample.

Coding

Develop a coding scheme,

be clear what your variables are going to measure,

read/watch a sample of your sample with the goal of reliability and validity.

Coding: Reliability and Validity

Intercoder reliability as a reliability test.

Learn from others to test validity

Push Poll

A poll where the intention is to transmit (typically false) information under the guise of a question.

Ways to administer and construct

• Admin: self-administered, face to face, over the phone

• Construct: Open ended and closed ended

Open ended questions will…

tell you a lot more info

and allows people to be fit into more accurate groups

Problems to avoid when constructing a survey (4)

choice order effect

question order effect

question wording effect

social desirability bias

Choice order effect

if you really don't care you'll probably go with the last or first one

Question order effect

previous questions may affect the cognitive response process and respondents' answers

• Question wording effect

“are we spending too much/too little on welfare” vs “are we spending too much/too little on assistance the poor”

• Social desirability bias

Response based on what choice is more desirable. :”did you do your civic duty and vote” vs “did you vote”

Other probs with surveys (5)

• Interviewer effect

• Honesty

• Doorstep opinions:

• Guidance

• Probing

• Interviewer effect

something about the person interviewing you changes your response 

Honesty

Changes with the way the survey is administered

Doorstep opinions

• depth of opinions

guidance

• being able to ask clarifying questions

Probing

• Next question is determined by previous question

Open-ended interviews

a survey that opens up

The Problem of Causality (3)

• challenging to estimate a causal relationship w/o bias 

• A linear regression can easily show a causal relationship between two variables

• A Y can have multiples x’s, but if the main x and additional variable correlate it is a problem

Lurking Variable

When a Y has multiples x’s, but the main x and an additional variable correlate AND

a variable that is mathematically correlated to the IndV of interest and has a causal influence on the DepV of interest.

Confounding Factor

another name for lurking variable

All LV’s… (3)

• have to be properly controlled to estimate w/o bias.

• It is difficult to control them all

• Difficult to know all of them.  

The main problem with causality

estimating a casual relationship without bias

How to estimate a casual relationship without bias (4)

• X and y must be mathematically related (scatterplot)

• Relationship must be plausible

• Y (depV) cannot influence X (indV)

• Control of lurking variables

When the X (IndV) manipulates the Y (DepV), and vice versa

reverse causation

 Islam and Authoritarianism thesis:

• Is Islam the predominant religion in a country, and if so, how does that affect the country’s degree of authoritarianism or democracy?

 Islam and Authoritarianism: The main obstacle to the development of democracy

the subordination of women

 Islam and Authoritarianism: The 6 controlled lurking variables

• Economic development,

• economic performance,

• communist heritage,

• British colonial heritage,

• Natural resources/oil/OPEC,

• ethnic diversity

 Islam and Authoritarianism: Bias?

• He controlled for many lurking variables, but there’s no way to say whether or not he got all of them or got the ones he got accurately. Bias has been accounted for, but his work isn’t unbiased.

Unique Essential features of experiments (2):

• subjects randomly assigned to different groups, 

• researcher manipulates IndV of interest in the groups (doesn’t require complete control)

Bad Luck

Sometimes experiments show correlations that are just random and unassociated

Computers and Bad Luck

A computer can calculate the probability that an experiments is experiencing bad luck

Anything thats random…

is not correlated with anything else. Bad luck shows a correlation

Internal Validity

• The ability of research design to estimate w/o bias the impact of an IndV on a DepV

External Validity

• Ability to take research findings from one research project and validly apply those findings to a whole pop, other pops, other settings, other situations, etc.

Threats to External Validity (2)

• Unrepresentative sample (a pop validity concern) [college students are popularly studied bcs we often do studies]

• Unrepresentative situation or setting (ecological concern)

Primary shortcoming of Poli. Sci: 

• There are so many things you just can't study using an experiment

different method of creating treatment groups (2)

• Simple random assignment (as the precursor)

  • Creates groups that are very similar

• Blocking/Stratifying then randomly assigning

• threats to internal validity (3)

• LV’s, 

• History

• demand characteristics (ppl find out or gain a hunch of what the researcher is studying)

X, O, and R

Treatment, Measurement of DepV, Random Assignment into groups

Classic experiment:

•  R O X O

Pre/Post-test:

• R O   O

PostTest Only:

• R X O and R   O

  • Cheaper. Participants less likely to find out

Solomon four groups design

• ROXO, RO O, R XO, R  O

  • commonly used in psychology. expensive.

Lab v Field Experiments: Lab (6)

• Treatments can be carefully calibrated and dispersed better

• Can isolate groups, no cross contamination

• Exceptional high internal validity

• Many IndV’s that can be manipulated

• Questionable external validity

• Demand characteristics more likely

Lab v Field Experiments: Field (6)

• Maintain two necessary experimental components

• Takes place in a natural setting

• High external validity

• Sometimes cannot carefully calibrate treatments or isolate groups

• Internal validity is usually very high

• Fewer possibilities 

Natural Experiments (6)

• Random or atleast arbitrary assignment are handled by ‘nature’ (usually a government agency)

  • Government sets subsidized housing price, people apply for it because they feel they need

• Technically not experiments because research did not assign and manipulate

• Challenging to find good ones

• Challenging to track down all the subjects and to measure the depV

• take s place in a natural setting

• High internal and external validity