The objective of this contribution is to discuss some aspects of system identification of cascade systems. Models of such systems are important in for example cascade control applications. We will restrict our attention to systems with one input signal and two output signals. First, we will analyze some fundamental limitations regarding the statistical properties of such estimates, and in particular why it can be difficult to identify cascade systems where the sub-transfer functions are close to each other. We will then show how an unstructured SIMO estimate can be used to find a cascade system model using an indirect prediction error method or balanced model reduction.