However, JCDEA was also one of the variables which Benston excluded from the final regression because of the possibility of collinearity with LLEA (17).(2)
However, due to potential multi-collinearity problems analyses were presented with omission of ownership structure (SSTOCK, FEDERAL), large liabilities to earning assets (LLEA and LLEAC), Net Worth (NWTA, NWTAC) and combinations of these variables.
In some instances Benston (1985) did not simultaneously test for combined effect on nominal levels and changes in variables such as NWTA, LLEA, BRKDEA, however, we choose to retain both the nominal and percentage changes in the regression to capture any interactions.
In regressions [TABULAR DATA FOR TABLE 5 OMITTED] (4) and (5), when NWTA, NWTAC, LLEA, and LLEAC are excluded from the regressions, change in return on total assets (RETTAC) becomes significantly and positively related to failure.
Large liabilities to earning assets (LLEA) was found to be significantly and positively related to failure in all the regressions in which it was included.