Scribd
Upload
2 views

Lecture 5 - Panel data models

The document discusses various panel data methods including fixed effects estimation, random effects estimation, and difference-in-differences. It explains the fixed effects transformation and its implications for unbiased estimation, as well as the structure of random effects models. Additionally, it addresses further issues such as handling non-continuous dependent variables, endogeneity, and non-linear functions in panel data analysis.

Uploaded by

linhngoc9204
Copyright
© © All Rights Reserved
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
2 views

Lecture 5 - Panel data models

The document discusses various panel data methods including fixed effects estimation, random effects estimation, and difference-in-differences. It explains the fixed effects transformation and its implications for unbiased estimation, as well as the structure of random effects models. Additionally, it addresses further issues such as handling non-continuous dependent variables, endogeneity, and non-linear functions in panel data analysis.

Uploaded by

linhngoc9204
Copyright
© © All Rights Reserved
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 14

Panel Data Methods

 Fixed effects estimation


 Random effects estimation
 Difference-in-differences
 Further issues
Fixed effects estimation
 Consider a model with a single explanatory
variable: for each i:

 For each i, average this equation over time. We


get:

 Where and similar denotation for


other variables.
Fixed effects estimation
 Subtract two above equations, we get:

or:

where is the time-demeaned data on y,


and similarly for xit and uit.
 The fixed effects transformation is also called the
within transformation. The importance is that the
unobserved effect, ai has disappeared.
Fixed effects estimation
 Under a strict exogeneity assumption on the
explanatory variables, the fixed effects estimator is
unbiased: roughly, the idiosyncratic error uit should
be uncorrelated with each explanatory variable
across all time periods.
 The fixed effects estimator allows for arbitrary
correlation between ai and the explanatory
variables in any time period.
 Any explanatory variable that is constant over time
is removed by the fixed effects transformation.
Fixed effects estimation
 We can estimate the between-effects model:

 In Stata output, there are some R-squared:


 R-sq: within:

 R-sq: between

 R-sq: overall
Fixed effects estimation
 The composite error eit is decomposed into:

 Since ai is constant, eit is correlated with uit and


this correlation is computed by
 2
 a
 
2
a
2
u
Fixed effects estimation
 We should not perform data transformation and run
OLS on the transformed data. Although the point
estimates are correct, the standard error is invalid.
Since:
 OLS on transformed data by ourselves will use

(NT – k) degree of freedom.


 However, because of transformation of data, the

correct degree freedom is (NT – N – k). In Stata we


use xtreg command for fixed-effects model.
Fixed effects estimation
 We can perform fixed-effects model by running
OLS on dummy variables of observations.
 See computer example
Panel data format
id year age wage
1 2002 43 2.3
2 2002 45 4.4
3 2002 54 5.3
4 2002 32 2.3
1 2003 44 3.2
2 2003 46 4.4
3 2003 55 4.2
4 2003 33 5.1
1 2004 45 3.2
2 2004 47 4.8
3 2004 56 5.7
4 2004 34 4.1
Random effects model

• Instead of treating β1i as fixed, we assume that it is a


random variable with a mean value of β1
• The intercept value for an individual company can be
expressed as

where εi is a random error term


• Four companies have a common mean value for the
intercept ( =β1) and the individual differences in the
intercept values of each company are reflected in the
error term εi
Random effects model

• Substituting β1i into Yit, we obtain

where

• The composite error term wit consists of two


components, εi , which is the cross-section, or individual-
specific, error component, and uit, which is the combined
time series and cross-section error component.
Difference-in-differences
 Consider a model:
yit 1 xit   2 Dit  uit
where :
 t = 1, 2, i.e., two-periods panel data (or pooled

cross sectional data)


 D is a binary variables, which is equal 0 for all
observations at the time t1.
 D is correlated with D.

 Then we can estimate  by:


2

yit 1 xit   2 DitTit   3 Dit   4Tit  uit


Further issues
 Dependent variables are not continuous:
 Multinomial logit, ordered logit.

 Count model

 Fractional logit model

 Limited dependent variable:


 Tobit models

 Truncated/censored models

 Sample selection models


Further issues
 Endogeneity issues:
 Randomized design

 Simultaneous equation model

 Regression discontinuity

 Non-linear functions:
 Non-linear models

 Matching

 Panel and time-series models

You might also like