Loss Forecasting: Part 1 Preparing your Data

1
Loss Forecasting Part 1Preparing your Data

• We are interested in forecasting both the
  frequency and severity of losses (separately).
• Hopefully, we have past data that will provide
  information to us for predictive purposes.
• We must prepare this data for use in the
  forecasting process.

2
Common Data Preparations/Adjustments

• Loss frequency development
• Conversion to frequency rates
• Loss severity development
• Inflation adjustment
• Conversion to severity per loss

3
Loss Development

• Usually, an organization does not have full
  information at the end of the financial reporting
  (or policy) period regarding what the ultimate
  loss frequency and severity will be stemming from
  that period.
• In other words, an organization often finds out
  about loss occurrences (or the loss amount
  changes) over the course of several reporting
  periods.

4
Loss Development (Contd)

• An organization may not yet know the ultimate
  frequency and severity figures from, say, last
  period, as the risk management team plans to
  forecast losses for the upcoming period.
• Past losses are developed so that the forecasters
  can see the likely ultimate outcome of prior loss
  periods. Loss development is, itself, a forecast.

5
Conversion to Frequency Rates

• Frequency of loss occurrence means little until
  we know how the frequency relates to how much was
  at risk
• 25 workers injured (poor information) versus 1.37
  workers injured per 100 full-time workers (better
  information
• Total frequency is divided by units exposed
  (e.g., total number of injuries divided by 100 FT
  workers)

6
Inflation Adjustment

• Severity of loss (loss amount) must be adjusted
  to control for the impact of inflation on loss
  dollars.
• Usually, severity is expressed in the current
  years dollars (brought up to date) for
  forecasting purposes, using
• inflation rates
• inflation indices

7
Conversion to Severity Per Loss

• Total severity of losses in a particular
  reporting period includes information about loss
  frequency. It is the typical severity per loss
  times the number of loss occurrences that renders
  this total severity number.
• If data are provided in the aggregate (by
  reporting period rather than by occurrence), the
  data must be converted into an average, or other
  measure of central tendency) for forecasting
  purposes.

8
The Data are ready

• Once you have prepared the data for forecasting,
  you need to begin eyeballing it for patterns,
  stability, outliers, etc. to decide which
  forecasting method might best fit your data.
• Now onto Part II
• The Actual Forecast!!!