Regression and data

[krose]

Hi

I have two questions that I hope you can answer.

First Q
I've been doing a few regressions and have been wondering if I can use a specific type of data. I would like to analyse if a change in the unemployment rate has an effect on consumption. However, I'm not sure if I can treat the unemployment rate like I normally do and calculate the percentage change from the last period? The thing that makes me confused is that the unemployment rate is a percentage of a total population that probably also changes each period (mayby not alot, but still).

Second Q
When you use the build in Analysis Toolpak in Excel to do the calculations you get a lot of estimates and I am unsure of one of these. Just beneath "adjusted r-squared" there is a standard error. Which of the estimates above does it refer to?

Thanks,

Kenneth

 

[David Harper CFA FRM]

Hi Kenneth,

On 1, that exceeds my expertise, I fear a real econometrician would know better. (it seems to raise interesting technical issues; e.g., why not parse population and number of unemployed as two independent vars?). I did notice that Gujarati gladly uses unemployment rate (percentage) as independent (not change in unemployment rate). See his example 9.9 page 266: he uses the functional form 1/X as the explanatory variable where X = unemployment rate(%). This is very interesting, it appears on the assumption that uemployment is asymptotic to the intercept (the natural rate of unemployment)

On 2, I *think* it is the standard error of regression (SER; aka, SEE) and you can confirm by checking to see it equals SQRT[SSR/d.f.]. Actually, the ANOVA table should show MSE of the residual. If SQRT[MSE of residual] = your number, you are looking at SER.

Thanks, David