No outliers emerged that appear to interfere with the outcome of this study.

Regression Results: Dependent Variable = TSE-300


Variable	Estimated Coefficient	T-Ratio	P-Value
Monetary conditions index	-286.65	-3.2148	0.0033
Consumer sentiment index	47.932	2.8011	0.0091
Personal debt rate (change over last quarter)	-25767	-1.7112	0.0981
Canadian GDP growth (last quarter)	-154.69	-1.0008	0.3255
Constant	232.29	1341.4	0.17317
N=33, R2=0.7276, adj. R2=0.6887

Our estimating lines are then:
1. TSE-300 = -286.65(MCI) + 47.932(SENT) - 25767(DEBT)
2. TSE-300 = -286.65(MCI) + 47.932(SENT) - 25767(DEBT) - 154.69(CAGDP) + 232.29
   where MCI = monetary conditions index, SENT = consumer sentiment index, DEBT = change in personal debt rate since last quarter, CAGDP = annual growth rate in Canadian GDP.

Though line one is simpler, it is likely to be somewhat off the actual value. We retain the non-significant variables because their presence helps to improve the t-scores we see for the other variables. Another series of lines including government debt as an independent variable influencing the TSE-300 produced the program with a t-score greater than 8, which raises suspicions. The successful elimination of that factor involved the inclusion of its growth rate.

The negative correlation between the Monetary Conditions Index and the index is not surprising; from macroeconomics, it is known that a weaker currency invites greater investment from abroad. It is also not surprising to find a positive correlation between consumer sentiment and the stock index; in fact, it is almost surprising that the correlation isn't stronger.

Despite extensive trials, nothing appears to make the personal debt rate significant; though it comes close, it never breaks the important t-ratio of 2. This is somewhat surprising (nearly as much so as the result that the level of government debt appears to have no crowding-out effect on the TSE Composite.

Also surprising is the result that GDP rates appear to have little or no influence on the result. Given GDP's wide acceptance as a measure of inome, it seems unlikely that there would be no significant correlation between GDP and stock values.

Multicollinearity does not appear to be a problem for this model, though suspicions about autocorrelation have been raised. Our Durbin-Watson value is 1.0925, which falls between d_L=0.996 and d_U=1.510. The test is inconclusive on the subject of autocorrelation. A number of regressions involving other independent variables were tried in order to establish whether concerns about autocorrelation could be completely eliminated; however, none succeeded in raising p-values enough to justify a re-design of the model. There may simply be no choice but to accept a slight chance of autocorrelation. Heteroscedasticity does not appear to be a problem for this time-series study.

Summary and Conclusion >>