In order to test the model’s ability to reproduce the actual behavior of the trust investor — i.e., to simulate the trust investment process — the model was required to select portfolios for a specific set of actual trust accounts. In particular, the model was required to generate portfolios for four accounts with which the trust officer had dealt but which were not used in developing the model. These portfolios are presented below along with the selections made by the trust officer for the same accounts. The generated portfolios were then compared with portfolios generated by various random and naive models. The results of these tests indicate that the trust investment program selected a greater proportion of correct securities than did any one of the alternative models.
To obtain additional confirmation, the testing process was carried one step further — that is, the processes by which the portfolios are generated were submitted to empirical test. The test consisted of comparing the stream of output of the trust investment model to the recorded decision behavior of the trust investor. This test was applied to several of the mechanisms incorporated in the model. Although it is not possible to state that all the decision processes were unequivocally confirmed, the evidence strongly supports the hypothesis that the model’s mechanisms have captured a considerable portion of the trust investment process.
Source: Skyttner Lars (2006), General Systems Theory: Problems, Perspectives, Practice, Wspc, 2nd Edition.