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.

8. Further PTF modelling examples

8.1 CompA, SDF, WCom

The first triangle group we examine is CompA. This data is quite volatile on the percentage scale. A good model will measure the volatility in the data and will project the volatility into the future.

Topics covered:

Diagnostics
Autovalidation
High process variability
Forecasts
Pricing

The ultimates for the forecasts are conditional on the previously observed data. There is no need to smooth the ultimates or use Bornhuetter-Ferguson. The ultimates are the true ultimates conditional on the data.

The method of clearing process variance changes is also shown. This is useful when you have 'patterns' found in the hetero which you do not want to project into the future.

There is no need to adjust for inflation with inflation vectors unless you need to know the inflation after economic inflation. That is, you want to measure social or superimposed inflation.

The video on CompA, SDF, and WCom is viewable here (35 minutes).

8.2 SDF and WCom revisited

We examine SDF in ELRF. The cumulative values are not indicative of the incremental value to be paid in the next development period. Ratios do not work for this data.

We revisit WCom and demonstrate how to set constraints manually. The wizard is also run and the wizard models compared.

The video on SDF and WCom is viewable here (10 minutes).

8.3 Case Study LR High

This video is on assessing the validity of an ELRF model using the residual plot and how to modify the ELRF model to improve the diagnostics.

We begin by analysing PL(C) in the triangle group LR High using ELRF. Examination of the calendar year residuals reveals a significant trend. The trend we see in the residuals is the trend in the data minus the trend in the method. If, as in this case, it goes downwards the method is overfitting and will give forecasts that are too high.

By using different model templates and by giving zero weight to certain early years we can customise the ELRF model so that the trend is no longer visible in the residuals. We illustrate this and show how the forecast is directly dependent on the degree to which we can capture the calendar year trend.

The same situation is illustrated for PTF models using triangles simulated from an SGI model in the triangle group SDF.

The relationship between the trend in the data and the trend in the method has a major impact on forecasts. The difference between PTF and ELRF in this regard is that in the PTF the individual trends are actually quantified, while in ELRF we can only make determinations about the relative size of the trends.

Returning to LR High we look at PTF models for Paid Losses, Case Reserve Estimates and Number of Claims Closed.

This leads to a discussion of the issues involved in forming a plausible forecasting scenario. The interpolation of trends from previous calendar years to future years is illustrated. We create a scenario in this way which matches the ELRF default forecast. This enables us to assess the implicit assumptions involved in that method.

This video on LR High is viewable here (35 minutes).

8.4 Accident year hetero

This video begins with a discussion of accident year hetero and moves on to issues of symmetry in modelling.

First a caveat. If there are significant differences in variance across two or more blocks of accident years it is likely that other trends also differ in these blocks and so the best way to model them is to use subdivided triangles in MPTF.

ICRFS-Plus 10.2 has the capacity to model accident year hetero in just the same way as development year hetero, except that there is no automatic hetero option.

We illustrate the consistency of accident year hetero with development year hetero by modelling a triangle and its transpose.

In triangle group ABC we show how to transpose a triangle so that the development and accident year dimensions are interchanged.

We show that the transposition is preserved if the models are transposed, or indeed if a model such as Statistical Chain Ladder is applied which is symmetric in development and accident directions.

In this case accident year hetero applied to the original triangle corresponds to development year hetero applied to the transposed triangle. We show how to do this and also which tables and graphs enable you to see exactly what hetero has been applied. The forecasts by calendar year are identical for these two transposed cases.

We consider the same situation in the case of link ratio based methods and demonstrate by a simple argument that the forecast values from a volume-weighted average ratio method are identical regardless of whether the ratios are calculated column-wise or row-wise. Hence Mack models predict exactly the same means for the original and the transposed triangle.

The Mack standard deviations however do not share in this symmetry. This fact is explained by the implicit conditioning on the values in the first column in the standard column-wise method. This break in symmetry indicates a flaw in the Mack methodology for computing standard deviations.

This video on Accident Year Hetero is viewable here (35 minutes).

8.5 Future accident year alphas and premiums and link ratios to ultimate

A new feature in ICRFS-Plus 10.3 is control (selection) of future alphas when forecasting future accident years in the context of pricing or assessing underwriting risks. Another new feature is link ratio to ultimate both in the LRT and ELRF modelling frameworks.

This video on Future accident year alphas and premiums is viewable here (14 minutes).

8.6 How can we tell that data modeled by Murphy et al is not real?

In the paper Manually Adjustable Link Ratio Model for Reserving by Murphy et al on the CAS web site, a dataset is modeled using the Mack method. Notwithstanding that the Mack method for these data is completely unsound we demonstrate effortlessly that the data are not real. Our submission on “Meaningful Intervals” explains this. The fact that the data are not real was subsequently corroborated by Daniel Murphy at the CLRS held in Washington 2008.

This video on the data modeled by Murphy is viewable here (7 minutes).

Top

Table of Contents