Preliminary Regression

According to the quantity theory of credit, credit multiplied by the turnover rate of credit is equal to nominal gdp, growth in the total stock of credit should lead to a growth in nominal gdp (this is the relationship of interest in my regression). Regressing log(GDP)(nominal gdp growth) on log(TCMDO)(total credit market debt owed growth) gives the following results:

Dependent Variable: LOG(GDP)
Method: Least Squares
Date: 03/21/13   Time: 14:54
Sample (adjusted): 1968Q1 2012Q4
Included observations: 180 after adjustments
Variable	Coefficient	Std. Error	t-Statistic	Prob.
C	1.636129	0.038789	42.17971	0.0000
LOG(TCMDO)	0.738261	0.004144	178.1325	0.0000
R-squared	0.994422	Mean dependent var		8.491701
Adjusted R-squared	0.994390	S.D. dependent var		0.867738
S.E. of regression	0.064992	Akaike info criterion		-2.618068
Sum squared resid	0.751856	Schwarz criterion		-2.582591
Log likelihood	237.6261	Hannan-Quinn criter.		-2.603684
F-statistic	31731.18	Durbin-Watson stat		0.019569
Prob(F-statistic)	0.000000

According to this regression, a 1% increase in TCMDO is correlated to .74% growth in GDP. However there are problems with this regression. The Durbin-Watson statistic indicates that there is a high level of autocorrelation in this model. This could be solved for by adding a first and second order autoregressive scheme. If the t value for LOG(TCMDO) is still significant and a Breusch-Godfrey serial correlation test indicates that there is no autocorrelation, then we cannot reject the hypothesis that there is no autocorrelation in the model.

Though it must be noted that this model does not prove causality, only correlation. It could be that credit growth drives GDP growth or that GDP growth drives credit growth. Causality is beyond the scope of this model.