CAT Modeling 

Using GIS Tools to Improve the Catastrophe Modeling Process at USAA

Abstract: USAA has developed a web based, graphical model that provides users flexibility in estimating the number of potential, active, and remaining claims for any type of catastrophe. This presentation will present and demonstrate the tool, needs and benefits of developing this functionality for USAA. The model uses Esri software applications with some custom development on our base GIS infrastructure.

CAT Model Solution Overview

Solution Concept Definition

Improve the Cat Modeling process by enhancing the current process with additional, more timely data, in order to develop more accurate catastrophe reserves to help ensure the continued financial stability of the Association; and to make more accurate Cat Operation decisions to better serve our members.

What is the CAT Model application?

The CAT Model is an ArcIMS (Internet Map Server) application designed to assist Loss Reserving and Cat Operations personnel in the viewing and analysis of Homeowners, Fire, and Renters policy in force data, as well as other catastrophe-specific data.

What does the CAT Model application provide?

There are four options with which to view the policy in force information:

• Quick View:  An interactive tool which provides tools for simple viewing  and querying of the U.S. policy in force data.

• Java View:  An interactive tool similar to the Quick View, but with additional analysis capabilities. 

• Pre-Processed Policy In Force Maps:  This option shows policy in force maps at the U.S., state, county and ZIP code levels.

• Policy In Force (PIF) Reports & Geocoding Hit Rates:  This option shows policy in force counts (both total PIF and geocoded PIF) at the state, county and ZIP code levels.  The geocoding hit rate is the percentage of PIF that were successfully geocoded to the address level, however other levels of geocoding are available in the application.

What data sources are used in the CAT Model application?

• Homeowners, Fire and Renters policy in force:  This data layer originates from the Property & Casualty system and the files are updated once a month.  Each blue dot depicts a single risk address. 

• Nightly Claims Data Layers: Nightly claims layers are created by CAT-ID to provide up to date information about a particular event.

• State, County, ZIP Code, City, Military boundaries, and Highways/Streets:  These layers come from GDT Dynamap/2000 data and are updated on a quarterly basis.

• Other Data Layers:  Other relevant data layers, such as hurricane wind bands or earthquake shake intensity zones, will appear on an as-needed basis. 
  

Project Definition

Project Scope

This project will implement GIS functionality to support:

1. All catastrophic situations including hurricanes, earthquakes, snowstorms, tornadoes, fires, floods and hailstorms.

2. Various scenarios for a particular catastrophe, including data at the individual policy location level.

3. The incorporation of actual claim information, as it becomes available.

4. A single daily cat data source that will serve both claims management and loss reserving.

5. A single graphical representation of the model through CAD based tools.

Business Requirements

1. Provide necessary specifications to allow the Cat Model access to Policy in Force (PIF) geographic layer stored in SDE/Oracle.
   
2. Provide an automated ability to store Cat claims file in Oracle on GISPROD. The loss file must be able to relate to the geocoded PIF file by Member #,     Line of business, and policy # (MLP).
   
3. Incorporate a plotting tool so that the users can define and manipulate damage polygon(s). Polygons should be able to be created from HURRTRAK, user creation (ArcView), FEMA polygons, Hail Swath or an Arc/Info Generate file. This would be accomplished by overlaying damage polygons (i.e., wind bands) on top of the exposures and then producing estimates of the expected losses, LAE and claim counts within these data layers.
   
4. Define a damage polygon overlay process. This process should be able to overlay damage polygons with Cat claims/PIF geocoded layers. A new Oracle table will be created which contains the results from the overlay process.
   
5. MAPPING Requirements:
   
   1. Research Mapping/Spatial Analysis tool to provide polygon creation capabilities including the exception polygons. These polygons should be able to select PIF/Cat claims file information based on their extent.
      
   2. Policies In Force (PIF) and Loss data – need to be able to display:
      
      1. PIF dots on the map
      2. PIF counts by zip code or county
         
   3. Layers:
      
      1. States
      2. Counties
      3. Zip Codes
      4. City Boundaries
      5. Landmarks – parks, monuments
      6. Residential streets, major roads and major highways
      7. Hydrology
      8. Military bases
         
   4. Annotation:
      
      1. County names
      2. ZIP Codes
      3. City names
      4. Street and Highway names
      5. Ability to create text labels and text boxes
         
   5. Field of view manipulation (zoom in, zoom out, pan, etc.)
      
   6. Select of PIF/Loss information – need dynamic link to the attributes that includes the ability to open up the select table and export this table to a spreadsheet. There needs to be the ability to select by:
      
      1. individual selection
      2. damage polygon
      3. user defined polygon
      4. county
      5. city (Metropolitan Statistical Area)
      6. ZIP Code
         
   7. Scale (miles)
      
   8. Printing – Need ability to print, through the network, to:
      
      1. 8˝ x 11 color printer
      2. E-size plotter (HP Design Jet 750C)
      3. Ideally, this capability would also include printing (and plotting) at Cat Sites locations or through Remote LAN Access
         
   9. Overlay damage polygons with:
      
      1. Policies In Force (PIF)
      2. Loss data (This will include both the daily cat claims data and the ‘on demand’ data that can be run every few minutes at the beginning of a large scale CAT)
         

Figure 1: Overview of a hurricane track on southern Florida

               
                Figure 2: Hurricane track showing PIF and claim data.

               
                Figure 3: Hurricane wind bands with PIF (blue) and claims (red)

               
                Figure 4: Close-up of claims analysis within a cat zone.

Benefits

Overview

When a major catastrophe occurs, senior management will rightfully expect accurate, timely (daily) information that they can use to make critical decisions.Sending the wrong number of adjusters with the wrong skill sets to the wrong places can result in poor service to the membership when they need us most, and can result in unnecessarily high expenses. Inaccurate estimates of loss and LAE costs can lead to improper decision-making by senior management.

Also, the push towards shorter financial closing schedules (eventually, to a “virtual close”) makes it imperative that robust catastrophe data be accessible on a daily basis to facilitate the accurate estimation of catastrophe liabilities. Waiting until delivery of correct month-end data to analyze catastrophes is not an option.

Early Benefits – San Diego County (Alpine) Fire

Around 5:00 AM on Wednesday, January 3, 2001 a fire started about 30 miles east of San Diego. With the winds blowing at 65 mph toward the west the fire quickly got out control. It started threatening the moderately populated city of Alpine. By mid-morning, evacuations began to take place. Looking at a 3-hour visible satellite loop the smoke from the fire could be seen blowing across San Diego and then out over the Pacific.

By late morning the WESTRO (USAA Sacramento office) requested that Cat Operations plot a map showing the southwest portion of San Diego county. They asked that the map include the Homeowner and Fire Policies In Force, by risk location, so that they could assess the exposure. This was a standard request, but given that the fire was out of control and moving quickly WESTRO had need for immediate information. They expressed concern about how quickly that need could be met.  Plotting a hard copy map in the home office generally takes about an hour, if the map is fairly simple. Then the map must be shipped over night to the region which means they wouldn’t receive it until the next morning.

With the new GIS web application in place, WESTRO did not have to wait. It only took about 5 minutes for them to pull up a web browser and “plot” their own map on the screen. This new application provides the same information (plus more) that is available in a hard copy plotted map and it is available immediately.

Fortunately, the fire did not continue to rage. By the next day, Thursday, the winds were down to a mere 30 mph and then by Friday, the fire fighting teams were able to start controlling the fire. USAA only received 6-7 claims and only one of those actually had any fire damage. The rest were for smoke damage or Additional Living Expenses. But it could have easily been much worse.

Acknowledgements

My sincere appreciation to the CAT Model project team for your dedication, hard work and commitment to GIS tools, technologies and the industry.

Randy Murray, P&C Business Project Manager 
Reginald Warren, IT Senior Design Analyst
Connie Cheng, IT Senior Systems Analyst
Joe Sterling, IT Systems Analyst
John Clark, IT Associate Systems Analyst
Julie Brown, IT Associate Systems Analyst

Author Information

Craig Hopkins
Information Technology Director, USAA Information Technology
Enterprise Solutions, Corporate Real Estate & Services
9800 Fredericksburg Road
San Antonio, TX 78288
(210) 913-8168
Email: craig.hopkins@usaa.com