SOCOL CRISTIAN| MARINAS MARIUS MINIMUM WAGE AS A PUBLIC POLICY INSTRUMENT – PROS AND CONS Figure 22. VAR model stability 1.5 I n v e r s e R o o t s o f A R C h a r a c t e r i s t i c P o l y n o m i n a l 1.0 0.5 0.0 -0.5 -1.0 -1.5 -1.5-1.0-0.5 0.0 0.5 1.0 1.5 According to the Granger stability, if the prior values of variable X provide statistically significant information about future values of Y, then X is said to Granger-cause variable Y. The null hypothesis is that X does not Granger-cause Y, so that an associated probability below a threshold of 1% or 5% will result in its rejection. By applying the Granger causality as simultaneous influence of three of the lagged variables on the fourth, we obtained the following results: Inflation rate, minimum wage and employment rate with lags 1, 2, 3 and 4 do not influence together, in terms of Granger causality, the average wage; the acceptance probability is 57%; Average wage, minimum wage and employment rate with lags 1, 2, 3 and 4 influence together, in terms of Granger causality, the inflation rate; the probability of rejection of this hypothesis is 3.80%, i.e. below the 5% threshold; Average wage, employment rate and inflation rate with lags 1, 2, 3 and 4 influence together, in terms of Granger causality, the minimum wage; the probability of rejection of this hypothesis is 0%, i.e. below the 1% threshold; Average wage, minimum wage and inflation rate with lags 1, 2, 3 and 4 influence together, in terms of Granger causality, the employment rate; the probability of rejection of this hypothesis is 0%, i.e. below the 1% threshold. The last set of conditions that a VAR model needs to satisfy refer to the econometric validity of the residue as a result of normal distribution, presence of homoscedasticity and absence of autocorrelation in errors. Table 9 shows the probabilities of validity of the residue of the VAR model with four variables and four lags. As the probabilities are above the significance threshold of 5%, the H0 hypotheses associated to the 3 tests are accepted, supporting the accuracy of the VAR model. Probabilities associated to the VAR residue specific tests Table 9. Tests applicable to the VAR residue LM autocorrelation test H0 – no autocorrelation in errors for the selected lag Normalization test H0 – the VAR residue has a normal distribution 0. 6714 0. 1793 Heteroscedasticity test H0 – no heteroscedasticity 0. 1067 To test the intensity of reaction of a variable to a minimum wage shock impulse of one standard unit of deviation, as well as the relative importance of that shock to the other three variables, we used the impulse response function and forecast error variance decomposition. An impulse of one standard unit of deviation of minimum wage causes a positive cumulative response of the average wage in the next three years, although low in intensity, particularly from the seventh quarter 3 8