All entries for November 2007
November 24, 2007
November 16, 2007
November 14, 2007
In this file ex4-bis.doc(you need to log in to see the file) you can find most of the eviews output for the questions in exercise sheet 4 that we already covered in the classes.
The file also includes some comments that should help you go through the remaining questions in the same exercise sheet. We will definitely cover the remaining questions in the next class, but because that will only happen the day before your test you might want to make sure you work on these questions beforeyou come to classes.
As usual, feel free to contact me (especially if you spot mistakes) either by leaving a comment here or by emailing me.
November 07, 2007
(If you are an MSc student, you might be interested in how to use robust standard errors in eviews).
As mentioned in today's classes, we do not generally assume normality and homoskedasticity, so, provided we have a large sample, we employ a robust estimator for the variance of the OLS estimator which allows for heteroskedasticity.
The video below shows:
- how to tell eviews that you want the robust standard errors.
- how to obtain the correct test statistics for the hypothesis that all the slope coefficients are equal to zero (using the Wald option after estimation)
You should use the same command to carry out tests on a subset of coefficients as you are asked to do in excercise sheet 4.
Keep in mind that the equation being estimated is:
If the video below does not appear (you need a Flash reader installed, click on PLAY, you should also be able to zoom-in by right-clicking on it) or it takes ages to upload, this pdf briefly guides you through the same things. robuststuff.pdf.
Feel free to ask if anything is not clear. I have now enabled anonymous comments.
Please, do try to solve the rest of excercise sheet 4 before the next class.
November 04, 2007
Writing about web page http://www.voxeu.com/
I just wanted to point out this very interesting website which provides "research-based policy analysis and commentary from Europe's leading economists".
The site hosts articles by many prominent economists on a number of different topics. Although (or should I say luckily) the articles are short and non-technical, they contain references to research papers for those interested in finding out more about a specific topic.
Definitely interesting for a research student or researcher, but probably equally stimulating for undergraduates and postgraduates who might have this strange interest in the real world and who, at some point, we'll have to come out with some sort of research project.
I think it is worth keeping track of their RSS feeds so they will be on my side bar.
November 02, 2007
The last bit of excercise sheet 2, which we could not cover in some classes, asked you to estimate this model.
Note that if and only if the individual is BOTH female AND married.
How do we interpret ?
Well, we know that, ignoring whether a person is married or not, the effect of being female is the coefficient on the female binary variable, that is . But if that person is also married then overall effect of being female becomes .
Therefore, the ln(w) of a female differs by from that of a man and the difference becomes if the female is also married.
Therefore measures the additional effect of being married for a female. In other words, it measures whether the gender wage differential changes depending on whether the female is married or not.
Clearly, you can also interpret this focusing on the variable. In that case, the interpretation becomes: measures the additional effect of being female for a married person. In other words, it measures whether the "marriage wage differential" changes depending on whether the married person is female or not.
Because our dependent variable is ln(w), the coefficient on these dummies can all be interpreted as approximate percentage changes. So, for example, suppose that we find that , the coefficient on Female, is . That means that females on average earn approximately 7% less than males.
Note: because BOTH the experience variable and the dependent variable are in logs, the coefficient on ln(exp) is an elasticity as we showed in the class. Therefore the interpretation is that when experience changes by 1%, the wage (our dependent variable) changes by . No need to transform the coefficient in this case.
November 01, 2007
In 2 of the three classes yesterday I did not have time to go over how to use microsoft equation editor to type equations in Word.
As you know, you will have to submit your assignments electronically, so you need to familiarise yourself with this.
It is not particularly complicated. If you manage to reproduce the equation in this file equationeditor.doc, you should be ok.
Also, this webpage provides a simple guide (but that is probably more than you strictly need). Obviously, if you are having problems you can come and see me during my office hour or use this blog to interact.