POSC 300 Scope/Methods of Poli Sci

Stages of the Scientific Research Method

Research question —> hypothesis —> operationalization —> collect/analyze data —> evaluate findings —> conclusion

Issues with the Scientific Research Method

Not always applicable to complex questions, change in logic to fix complexities

Operationalization

A working definition of a usually complex subject

Monroe’s 5 Criteria of Appropriate Political Science Research Questions

1. Clarity

2. Testability

3. Politically Relevant

4. Originality

5. Theoretical Significance

Clarity

Can you come up with and tell people in a plausible way this is because of this? Clear understanding of question will tell me about that data.

Testability

Empirial Not Normative. Must be falsifiable —> can be proven wrong. Must be feasible —> possible to happen & test

Politically Relevant

Speaks to research already done. Tends to value specifics.

Originality

Must build on existing research to some extent without stealing. Replication is key to theory development.

Theoretical Significance

Matters to Poli Scientists. Tends to value generalization/broadness.

Theoretical Significance & Political Relevancy trade-off

Difficult to balance both generalized laws = science, understanding cultural significance = theory. Contradict each other.

Null Hypothesis:

Independent & Dependent variable do not have relation to each other —> there is no relationship between these

Scientific Method of Research - Donovan & Hoover

Research Question —> Hypothesis —> Reality Test —> Generalizations about hypothesis —> Back to beginning

Reliability

Is this information reliable?

Validity

Reliability hits the core of this concept, can’t have reliable valid measure.

Correlational Design

Looks at whether two or more variables are related to one another. Observe & measure variables as they naturally occur. Don’t deliberately manipulate them.

Correlation DOES NOT EQUAL causation

Reverse causation, a third variable, etc

When to use correlational design?

When want to know whether variables are associated, especially when can’t or shouldn’t manipulate them

True Experimental Design

Manipulating the independent variable, randomly assigning participants to groups, measuring the dependent variable, holding other things as constant as possible

Why random assignment for True Experiment?

Help distribute characteristics between groups so that they are comparable enough to manipulate the IV and measure the DV

Problems with True Experiment

May not reflect real-world scenarios well, participants may behave differently because of being aware to a study, sample may not represent well all American voters, the way the test is presented could affect results.

Quasi Experimental Design

Designed to investigate potential casual relationships, but don’t have random assignment (don’t have same level of control as a true experiment). Don’t have random assignment to the relevant groups.

Why quasi-experimental?

Comparing groups based on something that already exists rather than randomly assigning subjects to conditions. Have less control over the groups.

Quasi-experimental weakness?

Can make a stronger casual argument than with a simple correlation in some quasi-experimental designs, but can still have confounding variable that are difficult to eliminate.

Case Study

Involves taking one particular case or event and studying it in depth. Instead of collecting data or randomly assigning participants, could focus heavily on different aspects as to what happened and why of an event/situation.

Strength of case study

Can investigate a case in much greater detail than you could if trying to study hundreds of cases.

Weakness of case study

Generalizability: less ability to generalize, the case could be unusual.

Research Question

What am I trying to find?

Hypothesis

What do I predict?

Independent Variable

What is the presumed cause

Dependent Variable

What is the outcome

Operationalization

How will I actually measure the IV and DV

Research design

Why is this particular design appropriate?

Control variables

What other factors could affect my DV that I should account for?

Causation problems

What prevents me from confidently saying X caused Y?

Select Sample —> Measure IV —> Measure DV —> Compare/Analyze Relationship —> Determine Correlation

Correlational Design

Random Assignment —> Experiment & Control (or other Experiment) —> Pre-test —> Treatment —> Post Test

True Experiment

Select Existing Groups —> EX: Pr Group & SMDP Group —> Measure DV —> Control or Other Differences —> Compare Groups —> Evaluate Causation

Quasi-Experimental Design

Select Case —> Identify Possible Explanations —> Collect Multiple Sources of Evidence —> Analyze Evidence —> Evaluate Explanations —> Draw Conclusion

Case Study Design

Independent Variable (IV)

The possible CAUSE, does (X) affect y? A specific value to measure

Dependent Variable (DV)

The possible EFFECT, does x affect (Y)?

Natural Experiment

The real world creates the experiment. Researchers don’t decide who gets the treatment —> researchers look at something that naturally happened & compare the results.

Before-and-after studies

Measure the DV before a policy change, a law/policy changes, Measure the DV after the change. Compare the two.

Problems with before-and-after studies

Other factors could have caused the change. A policy might not have an immediate effect. Can strengthen the study by comparing with other states that also OR DIDN’T change laws but can get complicated due to changes at different times.

Biggest weakness of ALL quasi-experiments

Something besides the IV might explain the change in the DV.

What makes a research question appropriate?

Clarity, Testability, Politically Relevant, Originality, Theoretical Significance