Task 4

[HD]

Discussions for task 4 of the CDEF scenario.

[Tai duong dang]

Hi,
I am doing my task 4 and stucking at the interdependency of assumptions, anyone have ideas ? My point is that the economics assumptions are the most important and have greatest interdependency, it can affect inflation student population, ect, but since the model is stochastics int rate, what should we comment ?

Thanks and Best Regards

[Katherine Garner]

Does anyone know if we use the cost per employed person we recommended in task 2 for our sensitivity tests or do we keep the "GoalSeek" amount already in there (it's like 824)?

[HD]

The GoalSeek amount is only for the most recent scenario run through the model (ie. scenario 999), so that value is fairly meaningless for most of our analysis. I think sticking with our recommended CPEP is more appropriate.

[said.t...@gmail.com]

We should do sensitivity test using mean , standard deviation, CTE,... ?

[Katherine Garner]

I was wondering the same thing...I tested the assumptions in blue but I don't know what they mean by saying "use your recommended metrics" to perform the sensitivity tests?? Did that sentence mean use the cost you calculated from your recommended metric and then test based off of that test amount?

[Jason]

I could be wrong but my understanding of "using the same metrics recommended in Task 2" was that we should compare our recommended risk/return metric prior to changing any assumptions and after we test an assumption. 

For example, if my recommended risk/return metric for task 2 was the Sharpe Ratio, I would take the value of the maximum sharpe ratio as my baseline for all sensitivity tests. Then after I change an assumption like the student growth rate, I would compare how the new Sharpe ratio value (with the updated student growth rate) fairs against the baseline Sharpe Ratio.

[Katherine Garner]

Hmm, I see. I just compared the ending balance. Did you just ignore that value completely then?

[HD]

I think the issue with comparing the ending balance is that it is only considered for the final stochastic scenario run through the Macro (ie. Scenario ID 999)? You won't know what the ending balance of the other scenarios (0-998) were.

[Melissa]

Unless the spreadsheet changed a lot - the cost per employed person and ending balance is in a section called "Single Scenario View". HD is correct - your outputs of mean, CTE, VAR etc. are calculated based on a set of ~1000 scenarios. Scenario 999 is just result of the last scenario that was run through the macro to produce the risk/return metrics for a specific asset mix you picked.

I think sensitivity testing using either your risk/return metric individually, some combination thereof (everyone seems to be mentioning sharpe ratio for some reason), or even your recommended contribution amount are reasonable approaches. Pick one and run with it. 

[Katherine Garner]

Oh wow, that actually makes complete sense. I'm going to compare my metric rather than the ending balance!

[Jason]

HD, does that mean the YoY returns in the <Model> tab also values generated from the 999th run?

[Jason]

Katherine, yes I did ignore the ending balance since I totally forgot that it existed. 

[HD]

Yes, I think the YoY return is most likely calculated from the stochastic returns assumptions defined for each asset class (all hidden) applied to your asset mix for each scenario.

[Katherine Garner]

Is it okay for me to only run the tests with the asset mix I chose? Running for 10+ portfolios seems excessive

[Jason]

I plan on only running the asset mix I chose from Task 2 since that is my optimal portfolio

[Tim Hall]

How are you guys sensitivity testing attrition rate? The difference from 1% to 10% and 30% is throwing me off.

[HD]

I figured attrition rate is partially related to the 65% grade average to receive tuition for the next year, so the percentage of people trying harder to reach that goal would result in a general % decrease in attrition rate. Basically I shocked everything by the same amount across all years.

[AL]

Hello all.  Quick question. Are you all thinking of including starting cost of college tuition and starting fund balance as part of your assumptions?  Seems to me these are more of an input item since they will be known at inception.  

[Tim Hall]

Hi AL,

I did, just on the basis that they are specifically listed under the assumptions section.

[AL]

Ok, thanks for the info, Tim!

[AL]

Sorry one more question.  So did you also test the number of years funding will be made as this is also under the assumption section?

[Jared]

So... I'm having trouble with arguably the simplest thing. I moved on to task 5, and I've written up everything except for this one little bit missing from task 4:

For the life of me, I am having difficulty coming up with factors excluded from the model that may have a material effect on results presented. For right now, I've included wage growth. I justified this because although things are measured as cost per person, this figure (cost per person) will be used to create a payroll tax which will depend on a person's income. Therefore, fluctuating incomes will affect funds received for CDEF.

First, do you guys think that's acceptable, given my explanation?

Second, do you have any other baits/hints to throw at me for another thing missing from the model?

[Jared]

I've considered fertility rates, but that would be reflected in student growth.

I've considered grades earned by students, but that would be reflected in student attrition

" " unemployment, but that would be reflected in "new unemployment claimants"

What other factors could affect solvency of the fund?

[Eduardo Sebayan]

Does it make sense to test new hires and new unemployment separately? Seems like testing one is the same as testing the other? Or would it make sense to treat employment as one single assumption? 

[Katherine Garner]

Hey Jared, i'm having the same issue. I thought about immigration rate but that would also be in the student growth...Also, in regards to your wage assumption, the assignment says "the amount of payroll tax will be fixed in advance." I was considering adding income as well but wasn't sure if this statement assumed a flat tax amount? Not sure if fixed=flat. 

[Jared]

I'm fairly certain that payroll taxes by definition are a percent of income. At least, according to wikipedia, payroll taxes "are usually calculated as a percentage of the salaries that employers pay their staff."

[Jason]

I also considered wage increases but I think the annual tax contribution is a fixed dollar amount and not a percentage of income so an increase in the population's earning doesn't necessarily increase the tax contribution amount. Perhaps you can somehow argue that some citizens might have more disposable income and wiling to pay more taxes...? 

I assumed that the fund management fee was the expense only related to managing the fund and discussed how administrative costs are not accounted for. Do you guys think that management fee already incorporates that or no?

[Jason]

I have a silly question that I would like your guys' input. Did you shock all the assumptions as a percentage of the base value (ie. 110% of growth rate) or did you add/subtract a certain amount of basis points (ie. growth rate increases by 3%) 

[Jared]

Looking again at the wording on page 1 of the FA overview, it does seem to suggest a flat tax. So I'm back to square one then. Oh well, I had only written up <100 words on that specific factor to begin with.

[Jared]

I shocked all assumptions as a percentage of base value to keep things "fair". I also mentioned this in the word doc.

[Jared]

And by word doc, I mean my submission. Weird wording there, sorry.

[AL]

what about GPA criteria?  I haven't gotten to that question yet, just brainstorming.

[Jared]

I think GPA criteria is supposedly accounted for with attrition rates...

[Katherine Garner]

Do you think it's a bad idea to suggest assumptions that could be interrelated? Technically the model already has anyway. 

[Jared]

We're supposed to identify factors excluded that may affect the results.

If your assumptions are moderately correlated, then fire away. If they move almost perfectly in tandem, you wouldn't be adding much value to the model, would you?

[Jared]

The model does include interrelated assumptions, but they're somewhat correlated rather than perfectly correlated. I would stay away from factors that correlate too closely to assumptions already included.

[Melissa]

I'll confirm that through this exercise, the intention is that the payroll tax will be a flat dollar. Not the smartest idea but that's how the task was structured.

Jared - here are some hints based on submissions I've seen. The current model assumes that a student can only enter the school system at grade 1 and will remain in the school system until they withdraw. Another hint is that most assumptions in the model are static over time, including attrition, tuition, etc. Also, only expenses related to investment management are considered currently within the model.

Of course, there are other ones you can suggest. Hope that helps. 

[Zoo Keeper]

"Does anyone know if we use the cost per employed person we recommended in task 2 for our sensitivity tests ..." 

I use my recommendation of asset mix in task 2 to do my sensitivity test. 

[Zoo Keeper]

"We should do sensitivity test using mean , standard deviation, CTE,... ?"  I think it's up to you...depends on how you justify it.

[Zoo Keeper]

"First, do you guys think that's acceptable, given my explanation?

Second, do you have any other baits/hints to throw at me for another thing missing from the model?" 

How about the immigrants or population of foreign students in Cascadia? Foreign students are always paying much more for college tuition, right? 

How about some cultural/social external factors? Factors under these categories won't overlap with the demographic/economic factors which are already in the model. 

[said.t...@gmail.com]

Demographic factor in not in the model because the assumption "the Opening employed persons" is constant over the 30 years.
I think this assumption should be adjusted with an the mortality table of Cascadia.

[Eduardo Sebayan]

Do we have any idea about what interest rate assumption they are using to discount cash flows?  I am thinking that is something that should be more explicit in the model. 

I am still struggling with the 2 material factors that are missing in the model. I thought of mortality, entry of foreign students, etc, but I am thinking that would all be captured by attrition/student growth etc.