psid3.do

1
psid3.do

• Panel study of income dynamics
• Longitudinal data of 5000 households started in
  1969
• HH members followed annually since
• This example has 5-years worth of data for 789
  continuously employed full time male workers
• 5789 3945 obs

2

• 1st exercise, ignore panel nature of data and
  estimate human capital earnings function
• Regress log hourly wage on
• Education
• Tenure and tenure2
• Experiene and experience 2
• Race
• Union status

3

• generate new variables
• gen expage-educ-5
• gen exp2expexp
• gen wagelaborinc/hours
• gen wagellog(wage)
• gen tenure2tenuretenure

4
(No Transcript)
5

• get OLS estimates
• reg wagel exp exp2 tenure tenure2 union educ
  black

6
Notice R2 and compare to FE in next regression
7
Areg constructs within Panel means rather
than Estimate LSDV model
Data must Be sorted by ID
Big jump in R2
ID is the variable That identifies Groups for
Fixed-effects
Number of groups (N in class notation)
8

• . now estimate a model with fixed effects
• . the xi command is used to construct a series
  of
• . dummy variables. suppose you have a variable
  
• . x2 that has four values, 1,2,3,4. i.x2 will
• . construct a set of 3 dummy variables, one
  for
• . x21, x23 and x24.
• gt
• gt notice that in the within group model, you
• . must delete variables without any
  within-panel
• . variation. in this case, educ and black
• . xi reg wagel exp exp2 tenure tenure2 union
  i.id

9
(No Transcript)
10
Summary of Results (Xs)
Variable OLS Fixed
Exp 0.0336 (0.0040) 0.055 (0.006)
Exp2 -4.2E-4 (1.0E-4) -6.5E-4 (1.6E-4)
Tenure 0.0307 (0.0029) 0.0178 (0.0032)
Tenure2 -7.5E-4 (1.1E-4) -5.0E-4 (1.4E-4)
Union 0.129 (0.017) 0.052 (0.024)
11
Rate of return to tenure

• Yit ß0 Eit ß1 Eit2 ß2 Tit ß3 Tit2 ß4
• Unionit ß5 EDUCi ?1 Blacki ?2 eit
• dY/dT ß3 2 Tit ß4
• Return to tenure OLS FE
• 5 years 0.0232 0.0128
• 10 years 0.0157 0.0078
• 20 years 0.0007 -0.0022