back


Explanation of the Tests


When comparing macroeconomic parameters over time, clear correlations can be observed. These correlations are highly interesting for economic policy. Some empirical observations contradict prevailing theoretical expectations.


We test these correlations primarily with regression tests (SPSS Statistical software). Whether the correlations are random or systematic, and whether they are strong or weak, we judge on the basis of statistical parameters. The significance, also called the P-value, tells whether a correlation is random or not. This is also called the error probability. The randomness is between 1.000 = random and 0.000 = free of randomness. And the beta value measures the strength of the influence of one variable on another dependent variable, between 0.0 = no influence and 1.0 = completely determining.


Finally, in the case of relevant results, it must be clarified whether a cause and effect relationship exists. If significant time lags are observed, then the trailing variables can be considered as effects. Otherwise, additional facts must be used to establish cause or effect. Otherwise, it remains with the mere determination of correlations.


We define:

Significance (sig.):    

    0.000                                                    Highly significant

    Up to 0.050                                         Significant    

    0.051 - 0.099                                        Weakly significant

    0.100 and greater                               Not significant


Beta:             

    0.8 and greater                                   Highly dominant correlation

    From 0.6 up to but excluding 0.8    Dominant correlation

    From 0.4 up to but excluding 0.6    Medium correlation

    From 0.2 up to but excluding 0.4    Low correlation

    Under 0.2                                            Weak/no correlation


Durbin-Watson (DW):

Normally, the estimation errors of regression tests, i.e., the deviation of the estimated values from the values to be tested, are random. However, if there are systematic correlations between them, this is called autocorrelation.


The Durbin-Watson test determines whether autocorrelation exists for the correlation tested. If yes, then the calculated quality values are too high. To avoid such estimation errors of the regression, the Cochrane-Orcutt corrected estimation procedure is applied.


Abbreviations list:

    SPSS AREG     Cochrane-Orcutt autoregression test                    

    N                       Series length                        

    R2K                  Corrected R-squared                        

    DW                   Durbin-Watson measure of autocorrelation                

    Regressor         Quantitative relationship between the

                             independent variables and the dependent variable                        

    Beta                   Strength of the influence of the independent 

                             variables on the dependent variable        

    sig.                    Significance, size of the probability of error of the 

                            test result.    


In the text, we summarize the findings of regression tests: 

     Significance    The quality of the observed correlation, the 

                              smaller the significance value, the lower the 

                              probability of error                

    Beta                   Highly dominant means that only minor other 

                              influences can exist in addition to the identified 

                              influence.                    

                              Dominant means that weaker other influences can

                              exist alongside it.

                              Medium means that other, similarly strong 

                              correlations may exist alongside it.                    

                              Minor means that other stronger influences may 

                              exist alongside it.     

                                   

For economic policy, only clear and weighty correlations are of interest, the influence of which have noticeable effects. Therefore, we limit ourselves to the examination of such weighty economic correlations. In addition to the correlations identified here, there are certainly multiple weaker correlations that we do not deal with.


back