ch10 doubts

keshav1996keshav1996
in W-10-Multivariate
  1. Fall 2019 (Q11) part c.--> Isn't the first sample answer given wrong? It is as as follows- "One way results can lose profit for factors below the base. Industry 1-4 are given lower factors than the GLM. They can also lose profit from charging too high a factor above the base and losing renewals and new customers from this. " I have concluded the following from other answers: "GLM output (by considering other variables simultaneously and correcting for exposure correlations b/w them) differentiates between levels of the risk better (as shown by a steeper upward GLM output line compared to relatively flatter one way output line) and thus by using one way results insurer is over-charging low risks and undercharging high risks, leading to adverse selection and ultimately reduced profitability." Is that correct?
  2. Spring 2017 (Q8) part c.--> "An insurer may operate in a territory or LOB where regulators deem the multivariate ratemaking is inequitable, and may require the insurer to use community rating." How can multivariate ratemaking be inequitable? It may not have large public acceptance which makes sense, but doesn't it lead to more equitable rates in general ?
  3. Fall 2015 (Q10) part a. Sample Answer for Graph 2-->"The LR relativity is considerably above the indicated GLM relativity for the “>10,000” level, implying that “mileage” may be correlated with other exposure variables (the candidate may then argue the merits of this observation)". How is correlation implied by this observation? Also, what merits of this observation can one argue?

Thanks.

Comments

  • GrahamGraham

    Question 1: Fall 2019 (Q11) part c.--> Isn't the first sample answer given wrong? It is as as follows-  "One way results can lose profit for factors below the base. Industry 1-4 are given lower factors than the GLM. They can also lose profit from charging too high a factor above the base and losing renewals and new customers from this. "  I have concluded the following from other answers: "GLM output (by considering other variables simultaneously and correcting for exposure correlations b/w them) differentiates between levels of the risk better (as shown by a steeper upward GLM output line compared to relatively flatter one way output line) and thus by using one way results insurer is over-charging low risks and undercharging high risks, leading to adverse selection and ultimately reduced profitability." Is that correct?

    • It looks like sample answer #1 stated the reverse of what's shown in the graph, although the general reasoning is valid.
    • The relevant observation is that the GLM estimate of relativities is different from the one-way result. It doesn't matter here whether the one-way analysis produces lower or higher factors than the GLM analysis because undercharging and overcharging both cause adverse selection which impacts profitability.
    • The steepness of the GLM line is not really relevant, only that it's different from the one-way result.
  • GrahamGraham

    Question 2: Spring 2017 (Q8) part c.--> "An insurer may operate in a territory or LOB where regulators deem the multivariate ratemaking is inequitable, and may require the insurer to use community rating." How can multivariate ratemaking be inequitable? It may not have large public acceptance which makes sense, but doesn't it lead to more equitable rates in general?

    • I think what they mean here is that it results in rates that are too accurate. It may be that a certain low-income or racialized territory receives high rates based on loss experience. This could be considered inequitable however, because residents of this territory may have other disadvantages that make it difficult to directly improve their loss experience. It's as if they are being penalized for being low income.
    • Side note: In your question you are referring to sample answer 4. If you understand the first 2 sample answers, I would skip sample answers 3 and 4. If you try to understand the details of every sample answer in the examiner's reports, your progress will be slowed down. In general, it's better to try to keep moving ahead.


  • GrahamGraham

    Question 3: Fall 2015 (Q10) part a. Sample Answer for Graph 2-->"The LR relativity is considerably above the indicated GLM relativity for the “>10,000” level, implying that “mileage” may be correlated with other exposure variables (the candidate may then argue the merits of this observation)". How is correlation implied by this observation? Also, what merits of this observation can one argue?

    • The loss ratio method is a univariate method and univariate methods don't account for correlation between variables so the indicated relativities would generally be less accurate than for multivariate methods such as GLMs.
    • The fact that the loss ratio method is so much higher than the more accurate GLM method for high mileage is indirect evidence that there could be correlation between mileage and other another variable. That's just a general fact about univariate versus multivariate methods.
    • Arguing the merits means providing an explanation as I did above.
Sign In or Register to comment.