I'm graduating on economics this year and, unfortunatelly, I took econometrics for granted three years ago, and now I have sort of a pickle on my hands, and I would like to check if you guys could help me close this paper.
So my goal is to understand the effect of antycyclical policies on the sales of vehicles in my state* during the 2009 economical crisis. The government took some measures aiming to hold the sales number during that time: on the fiscal side they lowered IPI (tribute over industrialized products) to zero in some cases, forcing the prices of new vehicles down. On the monetary side they lowered the interest rate on the central bank, forcing the comercial banks to offer lower rates on loans, and also expanding the offert of credit overall.
Following some existent metodologies and economic models, I've come up with the following model:
s = f(p, r, c), where: s: sales; p: a price index; r: interest rate; c: credit released to purshase new autos
- all series are deseasonalized using a GDP index
- I used the X12-Census method with trend filter set to 12 terms and ARIMA specification 2 1 0 to adjust sales and credit.
- I deleted 4 negative entries on the price index in order to use log (not sure if this is appropriate)
- I attached the workfile I'm using, the variables i used to estimate are s, cr, r and p.
- The other series in there, credit, price and sales are the raw untreated data.
- To identify the crisis period, I used a dummy variable on my model, and this is where things are getting hard to interpret for me. The econometrical model I'm running on Eviews is as it follows:
ls log(s) c log(cr) log(r) log(p) dummy dummy*log(s) dummy*log(cr) dummy*log(r) dummy*log(p)
The output on eviews:
So, can you guys assist me interpreting the results? Is this output worth anything?
As far as I can tell:
The t statistics are good at 95% for the constant, credit, and interest, and good at 90% for the price index;
the f statistic is good at 99%
Durbin-Watson is inconclusive
The white test indicates the model is homocedastic
the LM series indicates ther eis no serial correlation (I'm not sure of this one)
*I live in Paraná - Brazil.