01:220:322 Chap.1 Econometric Questions, Data and Causality
Econometric Questions, Data and Causality
What econometrics adds to statistics
Econometrics applies statistical methods to economic data so that models suggested by theory can be estimated and tested. It is treated as its own discipline because economic data are produced by people and firms who make choices and respond to one another.
Whoever chooses more schooling, a larger police force or a new product line usually differs in ways the analyst cannot see, and that is the root of most problems the course solves.
The ceteris paribus question
The target is almost always causal: what happens to an outcome when one factor changes and everything else is held fixed. A correlation is weaker.
Two variables can move together because a third factor drives both, as when richer districts both buy more technology and have students with more support at home. Learning to name that third factor is the first skill the course rewards.
Potential outcomes
The lectures define causality with potential outcomes.
For each person there is an outcome with the change and an outcome without it, and the individual effect is their difference. Only one of the two is ever observed, so every method in the course is a way to fill in the missing one for a group.
Randomised experiments do it by design; natural experiments and regression with good controls try to approximate it with observational data.
Why experiments are the exception
Economists rarely run controlled trials: they are costly, whole economies cannot be manipulated, and some treatments would be unethical to impose.
Most data are therefore nonexperimental, collected after people have already sorted themselves.
Four data structures
Cross-sectional data observe many units once, and their ordering does not matter. Time series follow one unit over many periods, and order matters. Pooled cross sections combine fresh samples from different periods.
Panel data follow the same units over time, which helps with causal questions because each unit can be compared with itself. Identifying the structure is the first step before choosing a method.
Reading a data set before modelling it
Before running anything, list what one row represents, which variables are outcomes and which are candidate causes, and whether the same unit appears more than once.
A listing of the first rows in Stata shows this immediately. The ordering of a cross section carries no information, while in a time series the order is the information.
What this chapter covers
- 01
Econometrics as the bridge from theory to data
- 02
Causality versus correlation
- 03
Potential outcomes and the missing counterfactual
- 04
Experiments and observational data
- 05
Cross sections, time series, pooled cross sections and panels
Worked example · free
Spot the confounder
- 2Name the reverse and common causes: cities with more crime hire more police, and city size or density raises both counts.
- 2State the ceteris paribus question: the change in crime if police rose while everything else about the city stayed the same.
- 2Propose a design: compare changes in crime in cities that received a hiring grant for reasons unrelated to crime with changes in similar cities that did not.
Key terms
- Econometrics
- The use of statistical methods on economic data to estimate relationships and test theories.
- Ceteris Paribus
- Holding all other relevant factors fixed while one factor changes; the condition behind a causal effect.
- Potential Outcome
- The outcome a unit would have with, or without, a given change; only one is observed for each unit.
- Observational Data
- Data collected passively after people or firms have made their choices, also called nonexperimental data.
- Panel Data
- Observations on the same units, such as people or schools, repeated across several time periods.
Econometric Questions, Data and Causality FAQ
Why is correlation not enough in economics?
People and firms choose their treatments, so the groups being compared usually differ in other ways. Those hidden differences move outcomes too, and the raw association cannot separate them from the effect.
What is the difference between pooled cross sections and panel data?
A pooled cross section draws new random samples in each period, while a panel follows the very same units over time. Only a panel lets a unit be compared with its own earlier values.
How do potential outcomes define a causal effect?
Each unit has one outcome with the change and one without; the effect is their difference. Because only one is observed, methods estimate average effects by building a credible comparison group.
What makes a natural experiment useful?
A rule, lottery or accident assigns the treatment for reasons unrelated to the outcome, so the untreated units show roughly what the treated units would have experienced. Later chapters use this logic for instruments and difference-in-differences.
Exam move
For every example in the course, write the causal question in ceteris paribus form, name the data structure, and list one factor that could create a spurious correlation.
Working through Econometric Questions, Data and Causality in 01:220:322? Sia is AskSia’s AI Economics tutor — ask any 01:220:322 Econometric Questions, Data and Causality question and get a clear, step-by-step explanation grounded in how 01:220:322 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
