5B Flashcards
Why is correlation not the same as causation?
Correlation does not establish causality because of possible omitted variables, spurious relationships, or reverse causation.
What is a panel study?
A longitudinal study where multiple cases (e.g., people, countries) are followed over time, collecting repeated observations.
What is an example of a panel study?
The LISS-panel in the Netherlands surveys the same 15,000 respondents every year to track changes in political attitudes.
What are within-case effects in panel studies?
Examining changes within the same individuals over time, rather than comparing different individuals at a single time point.
Why are within-case effects useful?
- They bring us closer to causality by reducing omitted variable bias.
- They provide more data points, especially for cross-country research where the number of countries is limited.
What is omitted variable bias?
When an unmeasured factor influences both X and Y, creating a spurious relationship.
Why do within-person effects reduce omitted variable bias?
Since individuals serve as their own control, time-invariant factors (e.g., genetics, upbringing) cannot cause bias in within-person comparisons.
What is an example of a within-person effect in political science?
A person’s trust in politics decreases in the years when they vote for a populist party, suggesting a relationship between political trust and populist voting.
What is the problem of reversed causality?
In cross-sectional studies, we cannot tell if X causes Y or if Y causes X.
How do panel studies with lagged effects address reversed causality?
They examine whether X at an earlier time predicts Y at a later time, reducing the possibility that Y influences X instead.
What is an example of a lagged effect?
If political trust in 2008 predicts populist voting in 2009, it suggests that trust influences voting rather than the other way around.
What is an experiment?
A study where researchers manipulate an independent variable and randomly assign participants to treatment and control groups.
What are the two defining features of experiments?
- Manipulation – The researcher actively changes the independent variable.
- Randomization – Participants are assigned to groups by chance.
Why is randomization important in experiments?
It ensures that treatment and control groups are similar on all characteristics except the treatment, eliminating omitted variable bias and reverse causation.
What is an example of an experiment in political science?
An online survey experiment where respondents are randomly assigned to read different news articles before answering political attitude questions.
What is a limitation of experiments?
Some variables cannot be manipulated due to ethical or practical reasons (e.g., we cannot randomly introduce authoritarianism to study its effects).
What is external validity, and why is it a problem in experiments?
External validity refers to how well the results apply to the real world. Many experiments occur in artificial settings, making them less generalizable.
What is a natural experiment?
A situation where randomization occurs naturally, allowing researchers to study causal effects without direct manipulation.
How can natural experiments help with policy evaluation?
They provide real-world causal evidence where full randomization is not possible, such as in legal or economic policy changes.
What is an instrumental variable in natural experiments?
A variable (e.g., assigned judge) that indirectly influences the outcome (e.g., recidivism) only through its effect on the independent variable (e.g., type of sentence).
What is a quasi-experiment?
A study that lacks full randomization but still resembles an experiment, often using pre-test and post-test comparisons.
What are common features of quasi-experiments?
- A treatment and control group.
- A pre-test and post-test design.
- A manipulation, though not fully randomized.
How are quasi-experiments different from full experiments?
They do not have completely random assignment, making them inferior to full experiments but stronger than observational studies for causal inference.
Why are quasi-experiments important for public policy research?
Policy changes cannot be fully randomized, so quasi-experiments help evaluate their effectiveness in real-world settings.
What is an example of a quasi-experiment?
A universal basic income program in an American Indigenous community allowed researchers to compare pre- and post-policy effects on child mental health and crime.
What is an interrupted time series design?
A quasi-experimental method that examines changes before and after a policy intervention, comparing trends over time.
How do the different research designs rank in terms of causal inference?
- Experiments (highest causal certainty).
- Natural experiments (strong, but dependent on real-world conditions).
- Quasi-experiments (good, but lack full randomization).
- Panel studies with lagged effects (better than cross-sectional but still limited).
- Cross-sectional studies (weakest in establishing causality).
