Archives

# Tests Arellano Bond AR and AR have low and high

Tests Arellano–Bond AR (1) and AR (2) have low (0.011) and high (0.36) p-values, respectively. This, as mentioned in the previous model, is the expected result because it shows that there was a high correlation of the residue in AR (1), but no longer in AR (2). The Hansen Test was 15.16 with a p-value of 0.126, a high value and that leads to accept the null ldk378 that the instruments are valid and uncorrelated with the error term of the difference equation.
Take notice of the Difference in Hansen Test result presented at the end of Table 3, which shows that the p-value for this test is high, which means that the instruments in level are also valid, and therefore the GMM System model adds to the GMM Difference model valid information and should be chosen as the most suitable model between the dynamic models here displayed. Thus, the best choice seems to be the GMM System model, since this best represents the dynamic nature of economic relations.
To evaluate whether the coefficients found by dynamic models are acceptable, it is possible to compare them to those found in the results of a OLS-Pooled model that considers the lagged variable – which would determine the maximum value that this parameter could achieve – and a model of Fixed Effects that considers the lagged variable – which would determine the minimum value for the parameter. Table 4 shows the results found for these models.
Considering that the best model to be used is the GMM System, some considerations can be made. The variable lagged Spread showed great significance, demonstrating the high stiffness to lower banking spreads, which depends significantly on the passed value of this variable. The coefficient found for Rzpib stands for the λ parameter of the model of Nakane (2001), which attempts to capture the average degree of market power in the banking industry. Its value is negative, in the order of 0.04, and is consistent with the results found by Nakane (2001), which means that Brazilian banks do not behave competitively, but do not represent a cartel structure. However, the increase in the value of λ indicates that there was an important change in the practices of banks in the country, as they are shunning from the more competitive market structure.
The sign of the estimated coefficient to Rzpib is in accordance with that shown in the model of Nakane (2001). The banking spread will be reduced when there is an increase in defaults, for the reasons stated above. Despite the recent instability in international financial markets following the subprime crisis that has increased the liquidity preference of banks, the sign of the estimated coefficient indicates that the increases in profits of major commercial banks in the Brazilian economy could to allow them to reduce their interest rates on bank loan operations, but this variable has not presents statistically significant. Moreover, competition stimulated by public banks contributed to these recent reductions in banking spread in the country.
Although the signal of the variable Lnopcred is in accordance to what was found by Nakane (2001), it showed no statistical significance. The basic interest rate presented a positive and significant coefficient, as expected. This last statement reinforces the argument that the interest rate acts as a floor to the level of banking spread, forcing this to rise when the interest rate increases. Summarizing, it is observed that both the macroeconomic and microeconomic factors are able to influence the level of banking spread, and more, if the macroeconomic measures to reduce spread have reached their limit, as supposed by Afanasieff et al. (2002) it is necessary to intensify the use of microeconomic measures, as recently implemented by the Brazilian government, through the encouragement of competition in the banking sector promoted by public banks, among other measures.

Concluding remarks