General CP2 queries (new student)

[Trevor]

Hi,

I just started trying out CP3 by jumping into past exam questions straightaway without studying any materials, as this was the advice from work colleagues given I have Excel modelling work experience.

After doing a few early years past exam questions, I think it is do-able, however I have a few comments:

1. Excel modelling can be subjective, there may be many ways to build a model / formula mechanics to make something work. For example, dynamic ranges or advanced functions using INDIRECT() or DATEDIF() functions. Are these acceptable?
   
   
2. Different treatments of calculation, for example converting annual interest rates to monthly rates
   A theoretical way is to take the 12th-root of it, such that taking the 12th power will give back the interest rates,
   however to tailor advice to non-technical audiences, we may simplify the calculations by simply dividing the rates by 12 (eg: 3% pa approximated to 0.25% instead of 0.2466%)
   This may yield different result after many years of projection. Will marks be penalised for getting a "wrong answer"?
   
   
3. Subjective judgements
   There are situations where we have to make a judgement whether to accept any unusual data or trends. For example if we were required to plot a graph to project future results, I might have decided the past data exhibits a trend, therefore use it as projection. However the sample solution does not consider that a trend.
   This again, leads to different final result. Will the "wrong answer" be penalised?
   
   
4. Different interpretation of question wordings
   At times I get a very different answer from the sample solution because I understood the question differently and had different calculation treatments.
   
   
5. Complex scenarios
   I remember coming across a modelling question which there are way too many possible combinations of events. As my first CP2 past paper practice, I realized it is very difficult to cover all scenarios (thousands of billions possibilities for that particular example).
   Not knowing I can simplify the scenario by setting assumptions, I was tempted to consider all the scenarios using Excel VBA macro (For loops).
   I had a feeling that VBA macro will not be acceptable (as it gives unfair advantage for candidates with work experience).
   However, in an exam setting I will not be sure if making such assumption is valid as it wouldn't have met the "client's" requirement fully.
   Is there a way to tell if any simplification is allowed for exam purposes?
   

I understand that the CP2 exam is not fully Excel modelling and getting the "correct answer" as there are audit trails and summaries too.
Can I know how much weight does a "correct" modelling answer carry?

I apologise in advance if any of the question seemed obvious, I had no briefing or background understanding of the exam but attempted past paper questions straightaway.

Thanks,
Trevor

 

[ntickner]

Lot of questions there, most of which don't come up here in the forum much. Nice to get something new to talk about

1,5. You're welcome to use more complicated ways of doing things. There are 2 marks in the Paper 1 schedule for 'simple techniques', which you'd lose if you use a technique that's unnecessarily complex. Particularly if there's an alternative simple way to do it (which there pretty much always will be). The exam questions are supposed to be doable by someone with a passing knowledge of Excel (and yes, Macros are definitely not required and will be penalised). There are certainly times where students have used more technical approaches (like data tables, or array formulas for example) and have made things simpler as a result, and those cases get kudos. But then, using array formulas when they're not necessary, and they just complicate the process is frowned upon. The point is: don't try and show off your knowledge of Excel. Try show off your communication skills in explaining what you've done. The Audit Trail is the key, which you can see from the mark schedule (I'd strongly recommend looking through this in the Examiner's report - it gives a good idea of the sort of things they're looking for). The model is really just there to a) show that you can read, understand, and follow instructions, and b) give you something to communicate (particularly in Paper 2)

2, 3. There's almost always more than one way to do things, and generally, all are acceptable unless the paper specifically prohibits it. It's a good opportunity to put in an assumption to explain why you're doing it that way. Even if it is the wrong way of doing it, if your audit trail explains what you're doing, you'll only lose a minimal amount of marks on the technique side. It won't be the difference between a pass & a fail unless you're doing stuff wrong elsewhere. Data changes are similar - the variety of ways to fix a data error is wide, and generally all are acceptable provided you can explain why you're doing it that way.

4. If you understood the question differently but still within the bounds of acceptable answers, then this is fine. If you understood the question incorrectly, then you'll lose marks for the bit you did wrong, but won't be penalised any further (so if your wrong answer leads to everything beyond that being wrong too, you'll get marks for the later stuff provided you didn't do anything else wrong).

On the 'thousands of billions' example - which paper was this? I suspect you misunderstood this part, as the examiners would never expect you to do anything that would require you to use macros.

 

[Trevor]

Hi ntickner,

Thanks for you advice, that greatly reduced my worry about getting a different numerical answer or "wrong answer".
Can I know though what are the weightage for each sections? How is 100% spread between the Excel model, Audit trail, and summary?

Also what are the recent passing marks and passing rates for this?
I have done the SPs and SA papers whose passing rates are around 40% each sitting. Although being told CP2 (and CP3) are simpler papers, I don't want to let my guards down as I have heard people failing them too.



Regarding the question with thousands and billion possibilities, I was referring to a rather early exam paper:
In the IFoA past paper website, CA2 2005-2018.
(Alternatively: https://www.actuaries.org.uk/docume...xam-papers-and-examiners-reports-2005-2018?aa)
From the downloaded folder, the exam question is:
Pre-2015 Exams > A Cat's Lives > 01 CA2_CatsLives Instructions for candidates.pdf

In this question, we were told a cat lives up to 15-20 years, and a cat have 9 lives.
Given the scenario, a suitable projection period is 0.5 years up to year 20, which gives 40 projection periods
We need to consider the expected cost until complete death of the cat, possible scenarios are:

• all 9 deaths within 0.5 years
• 8 deaths in 0.5 years, 1 death between year 0.5 -1
• 8 deaths in 0.5 years, no death between year 0.5-1, 1 death between year 1-1.5
• ...

Theoretically speaking, this gives (20*2*9)C9 = 253,037,064,783,855,000 possible selections of when the cat dies.
(Derivation: there are 20 years, each year with 2 periods. Within each period, there are 9 "death windows". Therefore we choose 9 out of 360 possible windows)
The later part of the question makes the modelling even more difficult, assuming different mortality rates across different lives.
However again, as my first attempt without any studying, I didn't know I was allowed to simplify the calculation by assuming only one death at most in each period. From a consultancy point of view though, this is a major assumption which will invalidate the calculations and therefore not answering Mrs. Macavity's concern (although cats having 9 lives is not logical to begin with).

In general, I've come across very straightforward modelling questions (Assurance Premiums) as it directly relates to life insurance calculations, also tough models (Auctioneers) which has no relation with actuarial calculations at all.

 

[ntickner]

Have a look at past examiner's report for the split between the various sections - it has the full marking schedule breakdown.

Pass rates are around 60-65%, I think. At least, they were before the IFoA stopped publishing pass lists.

I'd strongly recommend focusing on the more recent papers - the ones from after the change to have two separate papers have a very different structure and tend to be more real-life situations (rather than cats, tooth fairies, etc).