Creating a Company-Wide Picture - MPTF |
|
Multiple triangle models allow you to detect relationships (or the lack of them) at many different levels between lines of business, segments, layers and different types of triangles (eg case reserve estimates, number of claims reported, paid losses). |
Multiple Lines of Business |
|
ICRFS-Plus™ allows you to:
|
Dynamic Financial AnalysisIf you want to do Dynamic Financial Analysis, you need ICRFS-Plus™ to tell you the actual variability and correlations between lines of business in your company's experience. An example of this calculation is given in Case Study 1, where we compare the inflationary trends and residuals in two lines of business to see whether they are independent or correlated. |
|
To assess the liabilities of the company as a whole, it is necessary to evaluate the diversification benefit, which depends on
the amount of correlation between the different lines of business. See Case Study 1 for an example of how to calculate the liabilities
for the aggregate of two related lines of business.
The MPTF module of ICRFS-PLUS™ allows you to quantify two types of correlation by estimating them from your historical data. Correlations in the forecast distributions come from two sources -correlations between residuals (process correlation) and correlations between parameter estimates (parameter correlation). When you create PTF models for your different lines of business, you may notice that there are similarities in the patterns of residuals. This is an indication of process correlation - the “random” part of the losses, due to process variability, in one line is related to the “random” part of the losses in another line. In addition, some lines may have calendar trend changes in the same years. Process correlation also induces correlation in these parameter estimates. If the trends are not significantly different, you may set them equal, and then the correlation between the parameter estimates becomes 1. ICRFS-PLUS™ allows you to design models that represent what is going on in the data. The model is then used to forecast a probability distribution for each future accident and development period combination (a cell), together with correlations between every pair of cells. In MPTF, that means there are also correlations between the cells in different triangles. From these correlations between cells we can calculate correlations between any pair of accident or calendar years, and between any pair of triangles. These correlations are driven by what is in your data, and do not rely on other companies’ or industry data, which may behave very differently. |
Understanding the Relationship Between Segments |
It is important to understand the amount of diversification (or lack of it!) in different segments, for example:
|