logwageit = β1femi+β2blki+β3edi+β4 exp+β5 exp2 +β6occit+β7indit+β8southit+ +β9smsai
Why is it important to include aggregate time effects in the wage
equation?
If i estimate equation by fixed-effects (FE) Why am i unable to identify the effects of black, fem, educ, and exper on wages?
Thank you
Here are two questions. Answering the main one, you should include time effects if you hypothesise that there might be overall changes of wages over time, e.g., due to changing economic situation in the country, and you'd like to account for them.
Regarding the FE model, the problem is that these variables don't vary across the time period when the individuals in your panel dataset are observed. Think about -black- and -fem- variables, for instance: they won't change over the individual's life, so you can't include them in the FE model.
For the first question, i can say that if i include agreegate time effects i am able to control de Xit? (Please see bolt definitions)
The definition of the variables are:
lwage=log(wage);
exp: years of full-time work experience;
exp2: experience squared;
occ: occ=1 if in a blue-collar occupation;
ind: ind=1 if working in a manufacturing industry;
south (residence): south==1 if in the South area;
smsa: smsa=1 if in the Standard metropolitan statistical area;
fem: if female;
ed: years of education;
blk: if black;
d77-d82: year dummies
Thank you
You're, most likely, using the Baltagi and Khanti-Atom dataset. There, the effects of the variables you put in bold on wages can be estimated with fixed effects regression regardless of whether you include aggregate time effects or not, because these variables vary over time. Adding aggregate time effects, however, is not a substitute for these bold variables (otherwise, Stata would have dropped some of them due to collinearity). So you can add both the four bold variables and aggregate time effects to your model.
If you want your model to show the effects of the time-invariant (within the panel id variable) variables, consider using random effects models. You may want to use Hausman test to see which of the two models (fixed effects vs. random effects) is more appropriate.
How can i estimate the model by pooled OLS ? In stata data file i do not have the β10d77t + . . . + β15d82t
lwage it = β1femi+β2blki+β3edi+β4 exp+β5 exp2 +β6occit+β7indit+β8southit+ β9smsait + β10d77t + . . . + β15d82t + ci + uit
Thank you
In order to take into account common macroeconomic shocks and cyclical effects affecting the labour market
Pooled OLS is a simple OLS not taking into account panel specifics (i.e., no fixed or random effects). So you run it by the standard -regress- command.
When using FE and time-invariant dummies, you run into singularity of X'X, which is why STATA drops these variables (similar to the dummy trap) , as you told it to use FE.
You should probably consider random effects.
Change over time
http://books.google.com/books?id=64vt5TDBNLwC&pg=PA485&lpg=PA485&dq=important+to+include+aggregate+time+effects+in+the+wage+equation&source=bl&ots=JzW_qIctrd&sig=dLAGgROqvev6Vn8r4JwNNcB_sJA&hl=en&sa=X&ei=EABWVJmuEYXiaoK9gtAP&ved=0CDYQ6AEwBA#v=onepage&q=important%20to%20include%20aggregate%20time%20effects%20in%20the%20wage%20equation&f=false
You should use fixed effect when you suspect one of your explanatory variables to be endogenous. Is it the case in your data? If not, feel free to use a random effects model (re option instead of fe option). If at least one of your explanatory variablees is endogenous, random effects will give biased coefficients for this variable but also for the exogenous variables correlated to the endogenous ones.
I do not understand why are you saying "aggregate" time series. please clarify this. Answering the question, I'd say that you might require to check the term "omitted variables"
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Analysis
- Data Analysis
- Dataset
- Big Data