Chapter 3 Content Flashcards
What is the difference between correlation and causation?
Correlation: When two variables move together but do not necessarily cause each other.
Causation: When a change in one variable directly results in a change in another.
The Identification Problem questions the direction of influence between two variables.
What is the Russian Peasant Fallacy?
A mistake where peasants killed doctors during a cholera epidemic, believing they caused the disease.
This illustrates misinterpreting causation.
What is the SAT Training Fallacy?
Students who took SAT prep courses performed worse on average, possibly due to weaker students self-selecting into prep courses.
This highlights issues in interpreting data related to causation.
What defines a randomized trial?
An experiment where individuals are randomly divided into treatment and control groups.
This is considered the gold standard for establishing causality.
Why do randomized trials work?
Random assignment removes bias and confounding variables, ensuring differences between groups are due to the intervention.
This allows for clearer causal inference.
What was a significant finding from Estrogen Replacement Therapy (ERT) trials?
ERT trials found that hormone therapy increased heart disease risk, leading to reduced usage.
This is an example of how randomized trials can reveal important causal relationships.
What is a limitation of randomized trials?
Results may not generalize beyond the study sample.
Attrition bias can also skew results if participants drop out.
What is observational data?
Data collected from real-world behavior rather than controlled experiments.
This type of data can be biased, complicating causal inference.
What is the challenge with observational data?
Observational data can contain bias, making it harder to determine causation.
This is due to uncontrolled external factors affecting outcomes.
What methods can address bias in observational data?
- Time Series Analysis
- Cross-Sectional Regression
- Quasi-Experiments
These methods help clarify relationships and potential causal links.
What does time series analysis track?
Changes in two variables over time.
It can help identify trends and correlations.
What example illustrates time series analysis?
Cigarette prices and youth smoking rates showed a negative correlation (higher prices, lower smoking rates).
This example demonstrates how correlation can be misleading.
What is cross-sectional regression analysis?
Examines relationships between variables at a single point in time.
It helps analyze data without considering changes over time.
What finding was observed with TANF benefits and labor supply?
A $1 increase in TANF benefits reduced annual work hours by 0.2 hours.
This suggests a correlation that may not indicate causation.
What is a quasi-experiment?
Natural changes in policies create treatment and control groups.
This method helps in understanding causal effects without randomization.
What does the difference-in-difference estimator do?
Compares changes in two groups before and after a policy change.
This helps isolate the effect of the policy change.
What was the net impact of benefit reduction on work hours in Arkansas compared to Louisiana?
Net impact of benefit reduction on work hours = 150-hour reduction.
Arkansas reduced TANF benefits while Louisiana did not, illustrating the policy impact.
What is the biggest challenge in policy analysis?
Establishing causality.
This is crucial for effective policy-making.
What alternative methods can be used when randomized trials are not possible?
- Time Series Analysis
- Cross-Sectional Regression
- Quasi-Experiments
These methods help analyze relationships and infer causality in different contexts.
Fill in the blank: Randomized Trials are the ______ standard for causality.
[gold]
This highlights the importance of randomized trials in research.
True or False: Correlation implies causation.
False
Correlation does not necessarily indicate that one variable causes the other.
empirical public finance
The use of data and statistics methods to measure the impact of government policy on individuals and market
correlation
to economic variables are correlated if they move together
causal
two economic variables are causally related if the movement of one causes movement of the other
empirical economics must distinguish between three explanations between A and B, what are the three explanations?
A causes B, B causes A, some third factor causes both, empirical economist must distinguish among three explanations, correlation alone does not imply causation
first solution to solve the identification problem
Random trials solve the problem of identification
why do randomized trials solve the problem to identification
randomized trials solve the problem to identification because they rule out reverse causation
randomized assignment means the treatment and control group differ only by treatment, this rules out any third factors causing both treatment and effects
gold standard for determining cauality
randomized trials because any different between treatment and control group must be due to treatment
another problem with randomized trials is that the results might be only valid for the sample of individuals participating in the trial not for the whole population
attrition
reduction in the size of a sample over time, which if not random, can lead to biased estimates
time series analysis
co-movement of two series overtime, can be used to identify and measure correlation, potentially providing support for existing theories regarding causation
time series analysis often produces striking patterns
what is the problem with time series analysis
does not distinguish causation from correlation
different sub periods can give different impressions, excluded variables may drive the results especially those like economic growth and wage subsidy programs
cross sectional regression analysis
statistical analysis of the relationship between two or more variables exhibited by many individuals at one point in time with other factors held constant
regression line
the line that shows the best linear approximation to the relationship between any two variables,
regression analysis find the best fitting linear relationship between two variables
control variables
variables included in the cross sectional regression models to account for differences between treatment and control groups that can lead to bias
Quasi experiments
a Quasi experiments is a research design that resembles a true experiment but lacks random assignment of participants to treatment and control groups
key features: no randomization, treatment and control groups but the assignment to these groups is not random, can help establish causal relationships
