empirical. lecture 1. archival research

Vocabulary practice flashcards covering research validity, experimental and archival research designs, panel data concepts, and difference-in-differences analysis.

Archival research

Involves analysing existing records or documents for sources of pre-existing data.

Internal validity

How confident we can be that $X$ causes $Y$ (and not something else).

Lab experiments (validity)

The type of research that usually has high internal validity.

External validity

How well the results apply to the real world beyond the study setting.

Archival research (validity)

The type of research that usually has high external validity.

Ordered from least externally valid (most internally valid) to most externally valid (least internally valid)

1. Lab experiment

2. Field experiment

3. Survey

4. Archival research.

Experimental research (research design)

Manipulate variables and control the environment to observe the effect.

Archival research (research design)

Analyze pre-existing data or records.

Experimental research (data collection)

Data is collected in real-time, often through surveys, tests, or observations within a controlled setting.

Archival research (data collection)

Data is gathered from existing records.

Experimental research (control over variables)

Researchers have a high level of control over independent variables and can design the study to isolate specific effects.

Archival research (control over variables)

Researchers have little or no control over variables because they are working with pre-existing data collected for other purposes.

Spurious correlation

A misleading relationship between two variables that appears to be causal (false correlation).

Endogeneity problems

Problems consisting of omitted variables and reverse causality.

Reverse causality

Occurs when two variables influence each other, making it difficult to determine which one is the cause, and which is the effect.

Panel data / longitudinal data

Contains observations on multiple entities, where each entity is observed at two or more points in time.

Balanced panel data

Panel data where units are followed for an equal number of periods.

Unbalanced panel data

Panel data where units are followed for different number of periods.

Solution for panel data with observable variables

Control variables.

Solution for panel data with unobservable variables

Fixed effects.

Time fixed effects

Effects that vary across unit, but are constant over time.

Unit fixed effects

Effects that vary across time, but are constant across units.

Main advantage of panel data

Very flexible and can help us control for many types of unobserved variables (fixed effects).

Main disadvantage of panel data

Reverse causality.

Diff-in-diff

A method that uses a natural experiment that caused a random assignment, compares with a similar group as control group, and compares difference within group with difference between group.

Diff-in-diff key assumption

Both treated and control groups' trend were the same in the "pre-period"; otherwise, the two groups are not comparable (pre-existing differences that could bias the estimate).