If for any reason you are unable to view the training or demonstration videos, please contact our support staff at support@insureware.com and we will arrange to send you a copy of the videos on CD-Rom. You will be able to run the videos from the CD.
The training videos should be used for hands on training. We suggest you run the videos on a separate computer using a data projector, and train as a group.
The only way you will learn all the new concepts and be able to exploit all the immense benefits is by using the system. Experiential learning is imperative.
It is important that you study the videos in sequential order as set out below.
Table of Contents
1. Introduction to ICRFS-Plus™ 10.3 and modelling modules
2. Modelling using the Link Ratio Techniques and Extended Link Ratio Family modules
3. Introduction to the Probabilistic Trend Family modelling framework
4. Modelling real data (CTP) in the PTF modelling framework
8. Further PTF Modelling Examples
10. Uneven Sampling periods, updating, and various other topics
13. Capital Management of long tail liabilities
- 14.1 Credibility Modelling Compa Maa951 and All 1M revisited
- 14.2 Alpha differences: Case Study Sad and Sam
- 14.3 MPTF Net versus Gross and Clusters Revisited
- 14.4 Credibility modelling small arrays
14. Advanced MPTF modelling
14.1 Credibility Modelling Compa Maa951 and All 1M revisited
Example of modelling a company with high process variability.
In this case the final development and calendar trends are both zero, after optimisation. The calendar trend is especially crucial in forecasting. We show how to find the underlying trends which were rejected as statistically insignificant in the last stages of optimisation. At times, as in this example, a more prudent forecast is produced by reintroducing the insignificant trend going forward.
A more systematic approach to this problem, where available, is to use credibility modeling.
In this case CompA represents data from a single company representing about 3% of the market. We also have a triangle Maa951 of the total industry in this line of business, and a model for this with much lower process variance than CompA.
The industry model has positive calendar year trends and a final negative development year trend.
When we run the two lines together as a composite in MPTF we find a correlation of around 0.25 between the datasets. The credibility model for CompA is then formed by the MPTF model which takes the correlations into account and is formed by freshly evaluating all the parameters on this basis.
After some exploration we find that the credibility model for CompA contains a positive calendar (inflation) trend, but does not support a negative final development trend. The forecast from this model is more conservative than the one based on following through on the insignificant trend.
A similar analysis is carried out with the composite TG 1M 1Mxs1M 2M. This CTG contains three triangles which represent three layers in a reinsurance contract, the retention, the ceded layer and the retrocession. The correlation between the layers in this case is quite high, being over 0.9 in all cases. Each of the individual models for the layers has a single calendar year trend. In the case of the two outer layers it is positive, but the middle layer optimses to zero after removal of a negative trend which proved insignificant.
We test whether this zero trend holds up under credibility modelling, when the information from all three triangles is pooled to make a more accurate determination.
After careful testing we find that not only does the zero trend for the ceded layer persist, but the positive trends for the other two layers become more pronounced.
The video Credibility Modelling is available here (20 minutes).
14.2 Alpha differences: Case Study Sad and Sam
In this video we discuss alpha parameters, alpha constraints, and alpha difference constraints in MTPF.
We begin with 1M 2M 1Mxs1M. The alpha difference constraint in the optimal model is explained.
We then continue with the Sad and Sam data. We first show that the accident year correlation is very high. We subsequently measure this correlation in MPTF to be extremely high.
Once the accident level changes are modelled, the process correlation drops significantly as you now have parameter correlation between the accident levels. Note that the red bars in the accident direction indicate that the change in alpha level (ie the trend) is the same for each dataset.
The reserve correlation is significantly lower than the process correlation.
It is shown how to calculate the combined table for the reserve and a future underwriting year. This is extremely important for calculating risk as often much of the future underwriting year is renewal business. If they are considered separately, the risk charge is typically far too high.
The relationship between the VaR and the T-VaR is explained.
The video illustrating case study Sad and Sam is viewable here (20 minutes).
14.3 MPTF Net versus Gross and Clusters Revisited
One application of ICRFS-Plus is the evaluation of Reinsurance programs. In this video we examine an example of Gross data versus Net of Reinsurance data. The correlation between the two sections are very high as expected.
There is an accident level parameter constraint for the first year.
SDFx40 is revisited. The optimisation routine is demonstrated along with a discussion of clusters.
This case study is accessible here (15 minutes).
14.4 Credibility modelling small arrays
In some circumstances we have a small array containing data with high volatility, but we also have a larger array of data for the same LOB which we believe represents a good approximation to the larger context of our original dataset. We want to use credibility modelling to extend our model and hence our forecasting ability beyond the limitations of the data.
We explain how to place the data in a larger array filled out with zeros so that MPTF modelling is possible. This is carried out for illustrative purposes with sample data called Company A and Company B which is not found in the workbook.
In this case Company B has very little data so we show how to use the model for Company A to extend the model based on Company B data so that we can produce a plausible forecast.
We end by discussing clusters in MPTF modelling. We show how clusters are derived from the matrix of correlations between the lines, using the example of LOB A B ... J. We show how the cluster in MPTF behaves exactly the same way as a standalone MPTF model consisting of only the lines in the cluster.
The video on credibility modelling small arrays is viewable here (20 minutes).
