PE_L1_(without Appendix)

Question 1

What is Econometrics?

Answer 1

• Econometrics uses statistical tools to analyse economic data
• Helps answer causal and predictive questions
• Bridges economic theory and real-world data

Question 2

Steps in Econometric Analysis

Answer 2

Step-wise procedure:
- Step 1: Ask a causal question, e.g. “How does a job training programm affect the wage and employment
of an individual?“
- Step 2: Consult theory and translate question into hypothesis: E.g. human capital theory tells us that
training should improve employment and wage levels
- Step 3: Use data to answer the question scientifically: income and employment data (monthly) of
training participants and non-participants
- Data in different forms: cross-sectional, repeated cross-sections, panel data
- Step 4: Is the theory accurate or do we need to refine it?

Question 3

Types of Econometric Data

Answer 3

• Cross-sectional: one snapshot in time (Observations are drawn randomly from a population)
• Time series: data over multiple time periods (Observations are consecutive and hence not random)
• Combining cross-section and time series data
  -> Pool cross-sections: Two or more cross sections are combined in one data set (often policy changes)
  -> Panel (longitudinal)**: follows the same units over time

Question 4

Cross-sectional Data

Answer 4

• Observations on individuals/firms/etc. at a single point in time
• Order of observations often irrelevant
• Random sampling is typically assumed

Question 5

Time Series Data

Answer 5

• Observations over sequential points in time
• Order matters (trends, seasonality, autocorrelation)
• Common in macroeconomics and finance

Question 6

Panel Data

Answer 6

• Same cross-sectional units followed over time
• Combines both cross-sectional and time series aspects
• Can help control for unobserved, time-invariant factors

Question 7

Causality vs Correlation

Answer 7

• Correlation: variables move together but not necessarily cause each other
• Causality: change in one variable directly affects another
• Requires ceteris paribus comparisons (keeping other factors constant)

Question 8

Randomised Controlled Trials (RCTs)

Answer 8

• Randomly assign participants to treatment/control
• Ensures treatment and control groups are similar on average
• The gold standard for establishing causality

Random allocation into treatment and control group ensures that we compare, on average,
individuals who are similar in terms of observable (e.g. gender, age) AND unobservable
characteristics (e.g. motivation, ability) -> In other words: the ceteris paribus comparison is really valid!

Question 9

Natural Experiments

Answer 9

• Exploit external or policy changes that mimic random assignment
• Often used when true RCTs are unfeasible
• Example: lotteries or sudden policy reforms

Question 10

Key Goal of Econometrics

Answer 10

• Answer causal questions about human behaviour
• Make predictions or forecasts of economic outcomes
• Use statistical methods (e.g. regression) on real data

Question 11

Example of a Causal Question

Answer 11

• Does a job training programme increase wages?
• Compare participants vs. non-participants while controlling for motivation, ability, etc.
• Ideally, use RCT or robust observational techniques

Question 12

Notation for Variables

Answer 12

• Often we write outcomes as Y_i (e.g. wage)
• Regressors as X_i^k (e.g. education, experience)
• Aim: estimate effect of X on Y keeping all else constant

Question 13

In which situations in your life do you generate data?

Answer 13

(Register) data collected by the public sector

Data collected by private organizations

Question 14

Pool cross-sections

Answer 14

• Two or more cross sections are combined in one data set
• Cross sections are drawn independently of each other
• Can often be treated similar to a normal cross section
• Pooled cross sections are often used to evaluate policy changes
• Example:
• Evaluating effect of change in property taxes on house prices
• Random sample of house prices for the year 1993
• A new random sample of house prices for the year 1995
• Compare before/after (1993: before reform, 1995: after reform)

Question 15

Causality (Experiment)

Answer 15

• The only way to answer causal questions with data is to use data that were generated by an
  experiment
• “Test, learn, adapt” cycle of evidence-based policy making
• Step 1: Select individuals are eligible for the new program
• Step 2: Randomly allocate into two groups
• Treatment group: will participate in the new program
• Control group: will not participate in the new program (but can participate in status-quo/old program)
• Step 3: Carry out program with treatment group only
• Step 4: Evaluate difference in outcome variable (e.g. wage or employment rate) between treatment and
  control group
• Step 5: Improve new program with features that worked and drop features that did not work
• Iterate

Question 16

There are different types of experiments that allow answering causal questions

Answer 16

1. Randomized controlled trials (RCT)
2. Natural experiments (e.g. lotteries)