Lecture 9: measurement
how to answer/do applied research?
Examples:
Who do people pay not to go to the gym?
Does present bias effect credit card borrowing?
Why do people buy insurance?
Would you accept a 50-50 gamble to win 200, lose 100?
Steps in Empirical Economic Analysis:
1.Careful formulation of the question of interest has to be a question that hasn’t been answered
yet
2. Formulation of an econometric model (sometimes, but not always, based on a formal economic
model)
3. Formulation of hypotheses in terms of unknown model parameters
4.Datacollection and use of econometric methods to estimate the parameters in the econometric
model and formally test hypotheses of interest
5. Carefully interpret the results. Are they economically meaningful?
Step 1: Careful formulation of the question of interest
Empirical analysis is integrated with economic theory
Economic theory provides us with explicit models for how agents make decisions
- Helps us asking precise questions
- Helps us interpreting what we see in the data
Empirical evidence in turn informs theory
- Helps us rejecting theories with weak support in the data
- Tells us how strong these relationships are
- Raises new puzzles to be explained
Why do we need economic theory:
Suppose we observe an increase in prices in a market with no change in quantities sold. This is
consistent with:
A shift in the demand curve in a market where supply is perfectly inelastic
A shift in the supply curve in a market where demand is perfectly inelastic
A combination of shifts in the supply and demand curves in markets where neither supply nor
demand are perfectly inelastic
We need to let theory guide us and impose assumptions about the structure of supply and demand
in order to interpret the data. Not necessarily an economic model: “theory" can refer to any kind of a
priori reasoning/intuition.
,Step 2: Formulation of an econometric model (sometimes, but not always, based on a formal
economic model)
Example of an econometric model:
What are the α? (you have to make it by yourself, what you think is suitable)
What is ε? (the error term, you have to define it by yourself)
Present bias and future bias are the economic aspects
Before formulating an econometric model, you need to describe the data.
- Summary statistics
- Use of graphs
Summary statistics:
Descriptive graphs:
Trend in general: the more years
of schooling, the more salary you
get
Step 3: Formulation of hypotheses in terms of unknown model parameters
Hypotheses testing:
, 1.We want to reject the Null Hypothesis (H0) and provide evidence in favor of the Alternative
Hypothesis (H1).
2. Statistical test: depends on the type of data (nominal, ordinal, Interval) and hypothesis (difference
or association).
→in empirical economics we often focus on interval data (difference between two elements has a
meaning) and hypothesis of association(one element is related to another).
Correlation VS causality:
Correlation: X and Y move together
Causality: X causes Y to move
Example: The presence of more police officers in high crime areas (correlation) does not imply
that the police cause crimes (causality)
Difficult to establish causation: which one is cause and which one is effect? More in a few slides..
Significance level and power:
Significance level(α): probability
of rejecting H0 when true (Type I error).
- E.g. “Significant at a 10% level" means that there’s 10% probability that you find evidence
against the null hypothesis when it is actually true.
P-value: probability thatH0 is true. When a p-value is less than or equal to the significance level,
you reject the null hypothesis.
Power(1-β): probability of rejecting H0when false.
Step 4: Data collection and use of econometric methods to estimate the parameters in the
econometric model and formally test hypotheses of interest
Data collection:
1. Experimental data
2. Observational (non-experimental) data
Randomized Experiments:
Experiments are used to estimate the effect of X on Y (causation).
They are based on randomization:
- Only some individuals are treated.
- The choice of the treated people is random.
- Randomization is key to identifying the causal effect. A simple difference between two groups is
only correlation.
Very common in medicine and psychology. See next example.
Randomized Experiments example
how to answer/do applied research?
Examples:
Who do people pay not to go to the gym?
Does present bias effect credit card borrowing?
Why do people buy insurance?
Would you accept a 50-50 gamble to win 200, lose 100?
Steps in Empirical Economic Analysis:
1.Careful formulation of the question of interest has to be a question that hasn’t been answered
yet
2. Formulation of an econometric model (sometimes, but not always, based on a formal economic
model)
3. Formulation of hypotheses in terms of unknown model parameters
4.Datacollection and use of econometric methods to estimate the parameters in the econometric
model and formally test hypotheses of interest
5. Carefully interpret the results. Are they economically meaningful?
Step 1: Careful formulation of the question of interest
Empirical analysis is integrated with economic theory
Economic theory provides us with explicit models for how agents make decisions
- Helps us asking precise questions
- Helps us interpreting what we see in the data
Empirical evidence in turn informs theory
- Helps us rejecting theories with weak support in the data
- Tells us how strong these relationships are
- Raises new puzzles to be explained
Why do we need economic theory:
Suppose we observe an increase in prices in a market with no change in quantities sold. This is
consistent with:
A shift in the demand curve in a market where supply is perfectly inelastic
A shift in the supply curve in a market where demand is perfectly inelastic
A combination of shifts in the supply and demand curves in markets where neither supply nor
demand are perfectly inelastic
We need to let theory guide us and impose assumptions about the structure of supply and demand
in order to interpret the data. Not necessarily an economic model: “theory" can refer to any kind of a
priori reasoning/intuition.
,Step 2: Formulation of an econometric model (sometimes, but not always, based on a formal
economic model)
Example of an econometric model:
What are the α? (you have to make it by yourself, what you think is suitable)
What is ε? (the error term, you have to define it by yourself)
Present bias and future bias are the economic aspects
Before formulating an econometric model, you need to describe the data.
- Summary statistics
- Use of graphs
Summary statistics:
Descriptive graphs:
Trend in general: the more years
of schooling, the more salary you
get
Step 3: Formulation of hypotheses in terms of unknown model parameters
Hypotheses testing:
, 1.We want to reject the Null Hypothesis (H0) and provide evidence in favor of the Alternative
Hypothesis (H1).
2. Statistical test: depends on the type of data (nominal, ordinal, Interval) and hypothesis (difference
or association).
→in empirical economics we often focus on interval data (difference between two elements has a
meaning) and hypothesis of association(one element is related to another).
Correlation VS causality:
Correlation: X and Y move together
Causality: X causes Y to move
Example: The presence of more police officers in high crime areas (correlation) does not imply
that the police cause crimes (causality)
Difficult to establish causation: which one is cause and which one is effect? More in a few slides..
Significance level and power:
Significance level(α): probability
of rejecting H0 when true (Type I error).
- E.g. “Significant at a 10% level" means that there’s 10% probability that you find evidence
against the null hypothesis when it is actually true.
P-value: probability thatH0 is true. When a p-value is less than or equal to the significance level,
you reject the null hypothesis.
Power(1-β): probability of rejecting H0when false.
Step 4: Data collection and use of econometric methods to estimate the parameters in the
econometric model and formally test hypotheses of interest
Data collection:
1. Experimental data
2. Observational (non-experimental) data
Randomized Experiments:
Experiments are used to estimate the effect of X on Y (causation).
They are based on randomization:
- Only some individuals are treated.
- The choice of the treated people is random.
- Randomization is key to identifying the causal effect. A simple difference between two groups is
only correlation.
Very common in medicine and psychology. See next example.
Randomized Experiments example
