Table of Contents

Table of Contents

Introduction 1

Install the Software 1

Associate a Worksheet with MetaRisk Reserve 2

Choose a Model 3

Link Data to the Model 3

Adjust the Settings 7

Refresh the Data 9

Fit the Data to the Model 10

Residual Graphs 10

Paid Incremental Actual Triangle 12

Age Factors 12

Scaled Deviance graphs 13

Paid Scaled Deviance Residual Triangle 14

Paid Incremental Fitted Triangle 14

Matrices 15

Improve the Fit 16

Evaluate the Fit with Interactive Charts 17

Set and Evaluate Parameter Breaks 17

Declare Specific Values as Outliers 19

Declare Negative Values or Zeros as Outliers 19

Track Outliers and Parameter Breaks in Excel 19

Reports 20

Run a Simulation 22

Run an Aggregation 23

Update to the Latest Version 25

Utilities 25

Global Preferences 26

View the MetaRisk Reserves Log 26

Help 26

MetaRiskReserve Start-Up Guide / iii

Introduction

MetaRisk® Reserve™ is a software application intended to help insurers measure their reserve risk and parameterize it for their capital models. This solution streamlines the process of identifying the drivers and implications of claims inflation in the portfolio, which allows the optimization of reserving practices. Designed to be easily used with the base MetaRisk application, MetaRisk Reserve can also be used on a standalone basis and with other economic capital models.

MetaRisk Reserve works as a Ribbon in Microsoft Excel, as shown below. You start with a cumulative loss triangle in an Excel spreadsheet and then use stochastic models to fit the data and run simulations.

Install the Software

Prerequisites:

§  Administrative rights on the local Windows operating system

§  Microsoft Excel 2007 or 2010

Install the MATLAB runtime:

  1. Download MCRInstaller.exe to your local machine. We recommend that you save this file to your C:\Temp folder.
  2. Double-click MCRInstaller.exe and follow the installation wizard. We recommend that you use all of the default settings.
    After you complete the wizard, you can delete MCRInstaller.exe from your C:\Temp folder.

The newly installed files reside in:
C:\Program Files\MATLAB\MATLAB Compiler Runtime\v717

Install the MetaRisk Add-In:

  1. Make sure Microsoft Excel is not running.
  2. Download setup.exe to your local machine. We recommend that you save this file to your C:\Temp folder.
  3. Double-click setup.exe and follow the installation wizard. We recommend that you use all of the default settings.
    After you complete the wizard, you can delete setup.exe from your C:\Temp folder.

The newly installed files reside in:
C:\Program Files\MetaRisk Reserve

Add the MetaRisk Add-In in Excel:

  1. Start Microsoft Excel.
  2. Open the dialog box for setting Excel options:
    From Excel 2007: Click the Microsoft Office button, and then click Excel Options.
    From Excel 2010: Click the File tab, and then click Options.
  3. In the left pane, click Add-Ins.
  4. Select Go > Browse.
  5. Go to C:\Program Files\MetaRisk Reserve. Click MetaRiskReserve.xll and then click OK.
    MetaRisk Reserve now appears selected in the Add-In list.
  6. Click OK to close the Add-Ins dialog box.
    A MetaRisk Reserve tab will appear on the right side of the Excel Ribbon.

Associate a Worksheet with MetaRisk Reserve

The first step in using MetaRisk Reserve is to associate your Excel file with MetaRisk Reserve:

  1. In Excel, open the file that contains your loss triangle.
  2. On the Excel Ribbon, click the MetaRisk Reserve tab.
    The MetaRisk Reserve ribbon appears.
  3. Click the MetaRisk Reserve button to associate the worksheet with MetaRisk Reserve and enable MetaRisk Reserve functionality.
    The MetaRisk Reserve button turns from blue to amber, as shown above.

Make sure that you keep the MetaRisk Reserve button on when you save and close your file. If you disassociate your loss triangle from MetaRisk Reserve, you lose all data links and must start from scratch. However, associating and disassociating data does not affect the files in the export directory or the worksheets that contain fit and simulation data.


Choose a Model

MetaRisk reserve offers several models for estimating reserves.

On the MetaRisk Reserve Ribbon, click the Model Choice button. From the drop-down list, select one of the following:

Paid Chain Ladder / Industry-standard model for paid losses, which uses bootstrapping to create variability. This model fits by accident year and development year.
Paid GLM / Generalized Linear Model (GLM) for paid losses. This model fits by accident year, development year, and calendar year, and is more flexible than the Paid Chain Ladder model. With this model, you can improve the fit by using interactive charts, as described on page 16.
Incurred Chain Ladder / Chain Ladder model for incurred losses. Because incurred losses are the sum of paid losses plus loss reserves, you must first run a simulation on the data using the Paid Chain Ladder model before you can use the Incurred Chain Ladder model.
Incurred GLM / Generalized Linear Model for incurred losses. Because incurred losses are the sum of paid losses plus loss reserves, you must first run a simulation on the data using the Paid GLM model before you can use the Incurred GLM model.
Aggregation / A combination of the distributions created using the Paid Chain Ladder model or the Paid GLM model on different books of business. To use this option, you must have already run simulations on two or more sets of data. For information on how to run an aggregation, see page Error! Bookmark not defined..

Link Data to the Model

In this step, you specify the range of data in your worksheet that you want to link to the model. For each model, you must link the cells that compose the cumulative paid loss triangle. You can also link in additional cells for other parameters in your model.

Instructions for each model are given below.

Chain Ladder Models

  1. On the MetaRisk Reserve ribbon, click the Link Data button.
    The Link Data pane appears.
  2. In your Excel worksheet, select the area that contains the cumulative paid loss triangle. Do not include any row or column labels in the selection.

  1. In the Link Data pane, double-click Paid Triangle.
    Optionally, you can click Paid Triangle once and then click the Add Range button.
    A green checkmark appears next to
    Paid Triangle when the link process was successful.
    If a flag appears next to
    Paid Triangle, see the warnings that appear in red at the bottom of the pane. Fix any problems, and then repeat steps 2 and 3.

4.  (Optional) If your worksheet contains data on amounts of earned premium dollars in the exposure periods, you can add it to your dataset. Use the following procedure to link this data to MetaRisk Reserve. (Linking premiums does not affect the fit to the model; it only allows loss ratios to be shown in the simulation output):

ú  In the Link Data pane, look at the formula next to Premiums. This tells you the size of the range (number of rows × number of columns) that you must select for premium values in your worksheet. The example below shows that you must select a range of cells in your worksheet that is 7 rows wide × 1 column high if you want to add premiums to your data.

ú  In your worksheet, select the correctly sized range of cells.

ú  In the Link Data pane, double-click Premiums (or select Premiums and then click the plus sign).
A green checkmark appears when the link process is successful. If a flag appears instead, see the warnings that appear in red at the bottom of the pane. Fix any problems, and then repeat.
If you have already linked a range of cells in your spreadsheet and want to confirm that you linked the correct values, click Premiums and then click the arrow key.
The range of cells in your spreadsheet that are linked to premiums appear highlighted.


If you want to remove the premium values from your model, or if you want to remove the current values and select new ones, click Premiums in the Link Data pane, and then click the minus sign.
The checkmark next to Premiums disappears, and the premium values are no longer linked (although they are not deleted from the Excel worksheet).

  1. (Optional) You can also link values in your worksheet to the following parameters. Use the procedure described in Steps 3 and 4 to link these values:

Paid Tail Mean / The mean tail factor. You can use multiple tail factors to apportion the tail loss into several future development periods. The tail factor is lognormally distributed.
For example, if Paid Tail Mean is a 1 ´ 3 entry of 1.10, 1.05, 1.01, and Paid Tail St Dev is a 1 ´ 1 entry of 0.10, then this will be simulated from a lognormal distribution with a mean of 1.17 (=1.10 ´ 1.05 ´ 1.01) and a standard deviation of 0.10, and each simulated value will be re-allocated over the 3 years proportional to the original 1 ´ 3 Paid Tail Mean entry.
Paid Tail St Dev / The standard deviation of the cumulated nperiod tail factor. MetaRisk Reserve assumes a lognormal distribution for the cumulated tail factor.
The standard deviation of the tail factor. If there are multiple Paid Tail Mean tail factors, then this is the standard deviation of the product of those factors. The tail factor is lognormally distributed.
Systemic Risk CV / The coefficient of variation of a Gamma distribution with a mean of 1.0. to provide systemic risk to the results of the bootstrap model. Each simulation of unpaids from the model will be multiplied by a simulation from the gamma distribution. It will result in a wider distribution with the same mean as without systemic risk. For benchmarks, click the Help button, click the Help option from the drop-down menu that appears, and then open this file:
2011GC Analytics Benchmarks for Evaluation.xlsx

GLM Models

To link data in your worksheet to the Paid GLM model, follow steps 1–4 starting on page 3 for the Chain Ladder models.

You can also link these optional parameters:

Prosp Dev Trends
(Optional) / Mean and standard deviation of prospective development period decay parameters (similar to the tail factor in the Chain Ladder models).
Type and link these in, or, once the triangle has been linked in, click the Utilities button, select the desired number of periods, and then select Prosp Dev Trends.
1.  Select an empty cell in your spreadsheet.
2.  On the MetaRisk Reserve Ribbon, click Utilities.
3.  If you want to add columns to show additional periods, enter the number you want next to Number of Periods.

4.  Click Prosp Dev Trends.
Labeled rows for data are automatically created along with the proper links. The top row is the expected prospective development trend (such as 0.80), matched with the bottom row, which is the standard deviation of that parameter, assuming a lognormal distribution. Each column represents a future development period. If a column is blank, then it is assumed that the development trend simulated for the prior development period is used.
The example below shows the Prosp Dev Trends rows that are added if three periods are specified.
Prosp Cal Trends
(Optional) / Mean and standard deviation for calendar year trend for the unpaid portion of the triangle. Type and link these in, or, once the triangle has been linked in, click the Utilities button, and then select
Prosp Cal Trends.
To create rows for these parameters, see the steps in Prosp Dev Trends, above.
The top row is the expected prospective calendar trend, matched with the bottom row which is the standard deviation of that parameter, assuming a lognormal distribution. For example, if a 5% inflationary trend is assumed, then the calendar trend is 1.05. Each column in this input represents the next calendar period. If a column is blank, then it is assumed that the calendar trend simulated for the prior calendar period is used. The range must include as many columns as there are future calendar periods.
The example below shows the rows created for a 4 × 4 triangle.

If there is no prospective calendar year trend entered, then MetaRisk Reserve assumes that the future trend is the same as the calendar period trend for the most recent diagonal. MetaRisk Reserve also assumes that the variability is the same as the variability of the most recent calendar period parameter, with no variability correlation to any other parameters.
Prosp Cal Trends External
(Optional) / This field appears if the Prosp Cal Trend Method field on the Settings pane is set to External.
Enter a range of prospective calendar trend scenarios [number of scenarios ´ number of prosp cal periods]. The model will select randomly from this range of scenarios when it simulates.

Exposures
(Optional) / A column of exposure measures to account for changes in experience by exposure period. Examples are on-level premium, policy count, and payroll. If exposure data is linked in, the model will fit to the exposure-adjusted loss, weighing credibility by relative exposure.
Although this field is optional, we recommend that you use it with GLM models.

Adjust the Settings

Before MetaRisk Reserve fits the data, you can make adjustments as follows:

  1. On the MetaRisk Reserve Ribbon, click the Settings button.
    The Settings pane appears.
  2. For all models, you can make the following adjustments if you want to change the default values (some fields are read-only):

Simulation
No. of Iterations /
Enter the number of iterations for the simulation. The value can range from 1 to 50,000.
Seed / Enter the seed value. Use the same number each time you run a simulation to get the same results. Keeping the default value of zero will give you a random seed each time the simulation is run. The value in this field can range from 1 to 1,000,000.
Triangle Characteristics
As of Date /
The "as of" date of the paid triangle. If you want to run an aggregation, make sure the value in this field is the same across all modeled books of business.
Dev Length / (Read-only) Length of the development period. This value updates automatically when Expo Length is updated.
Expo Length / Select the length of the exposure periods: Monthly, Quarterly, Semi-Annually, or Annually. If you want to run an aggregation, make sure the value in this field is the same across all modeled books of business. Changes to this field also change the Dev Length and Initial Dev Length values.
Initial Dev Length / (Read-only) Length of the development period. This value updates automatically when Expo Length is updated.
Start Date of Last Expo Period / (Read-only) Start date of most recent exposure period.