In the solution, point 2 mentioned running the model with past data to ensure objectives are met; I thought they have not built a model yet?
Hi Jensen The data exist though. I think it is suggesting you use the year-before's actual in-force data to project the copmany's assets and liabilities a year later, and see how that compares with what the actual assets and liabilities turned out to be at the end of that year. This will test how "good" the model is at predicting actual events (ie it tests the model's validity). Robert
Hi Robert So, the "model" here refers to the newly completed one they are planning to build in the beginning? Cheers.
New model, old data. I think the idea is that you try to make a model, then check it to see if it works ok. Whilst you would have built the model to be consistent with theory, it's a good idea to see the working in practice. It's not much help to set it up now and wait till next year to see how well it did (since it defeats the purpose of needing the model now) so one way around this is to use last year's data, model one year and see how well it compares to what actually happened in practice. Of course the past isn't always representative of the future, but it's the best we have (short of waiting or time travel)