Huge performance and scalability improvements are achieved for VARMAX and VARMAX-GARCH models that are estimated through the method of maximum likelihood. In some test cases, the estimation speed is thousands of times faster than in previous versions.

The BOUND statement is supported. That is, you can specify inequality constraints on parameters to be estimated.

The INITIAL statement is supported. Hence, you can set up the initial values for parameters to be estimated.

The matrix expressions, operators, and functions are supported in the BOUND, INITIAL, RESTRICT, and TEST statements.

The multivariate DCC GARCH model is supported.

Four new forms of univariate GARCH models, namely the exponential GARCH (or EGARCH) model, the power GARCH (or PGARCH) model, the quadratic GARCH (or QGARCH) model, and the threshold GARCH (or TGARCH) model, are supported through the new SUBFORM= option in the GARCH statement for CCC and DCC GARCH models.

The concentrated likelihood estimation (that is, estimating the constant correlation matrix through the standardized residuals by specifying CORRCONSTANT=EXPECT) is supported for CCC and DCC GARCH models.