Morningstar Direct Asset Allocation Certification ®

Name:

SM

___________________________________________________

Evaluators: ___________________________________________________ Date:

___________________________________________________

Asset Allocation Basics 1. Why should we care about coming up with an Asset Allocation and then implementing it with investable securities rather than just choosing investable securities regardless of Asset Allocation? It is fairly widely agreed that asset allocation accounts for most of the variability of return in an investor’s portfolio. Some research suggests that 91.5% of the variation between the returns of different portfolios is accounted for by the asset allocation decision rather than individual stock or fund picking. Whenever you invest you are inherently making an asset allocation decision either implicitly or explicitly. It is much easier to reliably model, forecast and compare different asset class behaviors than it is individual stocks or funds. Asset Allocation allows us to decompose the investment decision into two parts and isolate the asset class level decision from the individual security level decision. http://corporate.morningstar.com/ib/documents/MethodologyDocuments/IBBAssociates/RoleAssetAllocation.pdf INPUTS 2. What is an asset class? Name some common asset classes. An asset class is a group of investments with some common property or behavior. Equity, Fixed Income, Bonds, Government Bonds, Corporate Bonds, Large Cap Equity, Small Cap Equity, Growth Equity, Value Equity, US and International versions of the previous…. 3. What are inputs or capital market assumptions (CMA)? Input or Capital Market Assumptions (CMAs) are a set of information that can be used to describe the likely behavior of asset classes. This includes the choice of model and the information needed for that model.

1

Morningstar Direct Asset Allocation Certification ®

SM

4. What is an asset class model? What choices does AA offer for Asset Class models? An asset class model is a framework for predicting asset class behavior. In general it is a simplification where some real but hopefully inconsequential parts of real historical behavior are ignored and the user only needs to specify the things that really matter. AA offers the Johnson, Log-Normal and Bootstrapping Models. 5. What is parametric model? Which of our asset class models are parametric and which are not? A parametric model or distribution function is a mathematical formula that describes the likelihood of any return for an asset class. This is commonly communicated with a probability density function graph. This is the bell shaped curve that most people are familiar with. The shape of this curve is determined by parameters like arithmetic mean and standard deviation. The Log-Normal and Johnson models are parametric and the Bootstrapping model is nonparametric. 6. What are the advantages and disadvantages of Parametric models over Non-Parametric models? Parametric models allow us to isolate different decisions about behavior down to individual parameters so that we can make active decisions about behavior for one parameter and leave the other parameters alone, i.e. letting them be determined by historical values. For instance many investors have active feeling about how expected return may be different in the future. For instance they may think we are entering into a Bull market and will increase expected return from the historical values. However they probably don’t have a feeling for how the dispersion of returns might change. So by increasing expected return and leaving standard deviation alone they can get a model that meets their expectations. Parametric models also allow us to simplify behavior down to only the factors that matter and smooth out noise in historical data. For instance in the following picture the histogram bars represent the number of historical returns in the range that the bar covers. The curve represents what our distribution model based on this same historical data predicts. The generally-held thought is that if we had an infinite number of historical returns, the histogram would follow a bell shaped curve. In this case, the highlighted bar is short, probably because we just don’t have enough historical data to get a smooth curve. The parametric models smooth out this inconsistency in the data and predict a larger number of outcomes in this area. If you believe there is a real reason for this asset class not to have many outcomes in this area, then a parametric model is not appropriate for this asset class.

2

Morningstar Direct Asset Allocation Certification ®

SM

Parametric models are always limited in their flexibility, since they try to fit messy historical data into a clean mathematical function. Both of the parametric models in AA are continuous bell-shaped curves (single peaks in the distribution, constantly increasing likelihood of returns to a peak followed by constantly decreasing likelihoods). So if you want to model an asset class that for some reason has multiple peaks or gaps where no returns can occur, surrounded by areas where returns do occur, then AA’s parametric distribution will not model this behavior well. Bootstrapping can model behavior like this. 7. What are the parameters needed to complete a Log-Normal model? Arithmetic mean (a form of Expected Return), standard deviation and correlations 8. What is the difference between a Log-Normal and Normal model? For all practical purposes they can be used interchangeably. When you select Log-Normal in AA you are actually getting a little bit of both. Traditional MVO assumes normally distributed returns. For multiple period models like Monte Carlo forecasting, a log-normal distribution makes more sense. In the log-normal distribution the Natural Log of (1 + the decimal form of the returns) are assumed to be Normally distributed. The distributions are very similar. The main difference is that the log-normal distribution can’t result in you ever losing more than all of your money where theoretically a negative infinity return is possible in a Normal model. The log-normal model results in a slight positive skewness and kurtosis over the Normal model.

3

Morningstar Direct Asset Allocation Certification ®

SM

9. Why do we use the terms expected return and arithmetic mean interchangeably? In the standard asset allocation optimization model (MVO) the optimization is a one period model or the optimal result over the next one period. If you feel that the historical data is all relevant then the best way to set a one period expected return is to take the average or arithmetic mean of the historical data. 10. Why can’t I enter geometric mean rather than arithmetic mean? Geometric means are more familiar to people when reporting historical data. The arithmetic mean is the commonly expected parameter for asset allocation because of the wide adoption of normal distribution models and Mean Variance Optimization, both of which rely on a one-period expected return or arithmetic mean. The geometric mean and arithmetic mean are equal when standard deviation is 0. As standard deviation increases, the two diverge. The geometric mean is a result of the arithmetic mean and the standard deviation. We could use this relationship to give the option to enter geometric mean and then calculate arithmetic mean based on the standard deviation. But we have not done so. Entering a geometric mean in the arithmetic mean field will result in a double counting of the standard deviation later on, and will yield overly pessimistic results later on in optimization and forecasting. 11. What are the parameters needed to complete a Johnson model? Arithmetic mean, standard deviation, skewness, kurtosis and correlations 10. What is skewness? In a normal distribution the distribution is symmetric around the mean. If a model supports a controllable skewness, then the distribution can be made to lean positive or negative. Another way to say this is that the distribution will have a longer tail on one side then the other. A model that supports a skewness parameter is more flexible and can model accurately more situations then a normal distribution.

4

Morningstar Direct Asset Allocation Certification ®

SM

11. Is a positive or negative skewness a more attractive property for an asset class and why? A positive skewness is more attractive as it means the extreme returns or those that are far away from the mean are positive or above the mean. Another way to say this is a positive skewness means that most of the risk is on the upside and a negative skewness means that most of the risk is on the downside. 12. What is kurtosis? Kurtosis is a property of a model that controls the ratio of occurrences in the model that occur around the peak of the distribution and in the tails versus those that are just below and above the peak of the distribution. A positive kurtosis means that the distribution has fatter tails and a taller peak than a normal distribution. A negative kurtosis means that the distribution has thinner tails and a lower peak than a normal distribution, and is sometimes described as having shoulders, which makes sense if you visualize such a distribution curve. A negative kurtosis is rare. A model that supports a kurtosis parameter is more flexible and can model situations more accurately than a normal distribution.

13. Given no skewness, is a larger or smaller kurtosis a more attractive property for an asset class and why? Generally a smaller kurtosis value is more attractive as it means there is less chance of extreme values that are far away from the mean. If there is a strong positive skewness, then it could be argued that a large kurtosis is a good thing as there is large upside potential without the corresponding extreme negative values. 14. When I work with kurtosis in AA, do I enter in kurtosis or excess kurtosis? It is common practice to call what is technically excess kurtosis as just kurtosis. Direct follow this convention and actually expects the values entered to be excess kurtosis. Also, for background, a normally distributed set of returns naturally has a kurtosis of 3. Hence an excess kurtosis is what we usually talk about; if an asset class has a kurtosis of 4, we talk about an excess kurtosis of 1. If an asset class has a kurtosis of 2.5, it has an excess kurtosis of -0.5.

5

Morningstar Direct Asset Allocation Certification ®

•

SM

Large kurtosis (or excess kurtosis, for that matter…), means a lot of peakedness; relative to a normal distribution, a highly kurtotic distribution has a taller head, fatter tails, and smaller shoulders, so more small moves, more extreme events, and fewer moderate moves.

•

The permissible range of excess kurtosis is [-2, infinity]. Mathematically, it cannot be less than -2.

15. What information is needed to complete a Bootstrapping model? A set of date ranges. When you set the parameters for parametric models from historical data the date ranges chosen are used to calculate a summary of the behavior and once this is calculated the underlying data from that date range is no longer known to the model. The date ranges for bootstrapping identify the outcomes that will be possible in Bootstrapping when optimizing and forecasting. 16. What is the most common way to come up with the values needed like expected return and standard deviation for a set of capital market assumptions? The most common way is to associate each asset class with a set of historical returns (or a proxy series) and calculate the historical value of each parameter needed. 17. Can I use daily returns data for a proxy series to an asset class? If not why? No, we currently do not allow for daily returns. Asset Allocation is about separating out the investment decision into asset allocation and manager selection. The advantage of this is the stability of asset class behavior. This behavior is only likely to be predictable over a time horizon measured in years. There are challenges and additional tools needed to realistically work with daily data and there is very little advantage in terms of more accurately modeling asset class behavior in doing so. 18. When would you want to use something other than historical values to create a set of capital market assumptions? What assumption do you make when you use historical values? When you use historical values to create a set of capital market assumption you make the assumption that the historical characteristics of these asset classes will repeat in the future. This assumption is only valid if the underlying economic cycles that created those returns are still possible (no large structural or regulatory changes in how these asset classes behave) and you are investing for long enough to experience a variety of economic cycles and smooth out short term behavior. If you are investing for a shorter time horizon you may want to use specific knowledge about the current environment over pure historical data. Direct AA also provides several methodologies for determining expected return other than pure historical. 6

Morningstar Direct Asset Allocation Certification ®

SM

19. Why do we provide several methods for determining expected return? Using historical returns to base prediction of future behavior may not always be appropriate. There may be reasons to believe historical performance will not repeat. You may also not have enough historical data for an asset class to represent all market cycles. Building Blocks, for instance, breaks down an asset class return into the likely places that return is generated from. It may be that the prediction of these parts is easier to do than the asset class as a whole. 20. Explain the disadvantage of Log-Normal Distribution Log Normal distributions are only described by two parameters (Arithmetic Mean and standard deviation). This makes them limited in their flexibility. This limitation most commonly shows up in an under prediction of extreme events. Most of the time when a Log-Normal distribution is fitted to historical data there are returns in the historical data that occur outside of where the Log-Normal would predict any returns based on the number of historical observations we have. The Log-Normal distribution is also almost symmetrical. There is no way to control the skewness or asymmetry of the distribution. For some asset classes this is an important property that can make them much more attractive or less attractive than other asset classes that have approximately the same Log-Normal distribution (same standard deviation and arithmetic mean). 21. Are there any advantages of using Log Normal distribution? The Log-Normal distribution is intuitive, well known, and is easy to calculate and so results in less waiting in the software. Most people are familiar with the rule that 2/3 of the time, we get returns that are within one standard deviation of the mean; over 95% of the time, we are within 2 standard deviations off the mean and 99% of the time, we are 3 standard deviations off the mean.

22. When should I use the Johnson Distribution Model? The Johnson Model is a super set of the Log Normal distribution. The Johnson Model can be set to match and distribution that the Log Normal can create. Through the addition of two extra parameters (skewness and kurtosis) the likelihood of extreme events and the ratio of the number of distributions above and below the mean can be controlled. These are the two areas where the Log-Normal distribution most commonly doesn’t fit historical data well. So the Johnson Distribution Model can model a wider range of asset class behavior and thus produced more realistic results. 7

Morningstar Direct Asset Allocation Certification ®

SM

23. Why wouldn’t I use the Johnson Model? The Johnson Model is not as well knows as the Log-Normal. This means that you may have to explain to the end client why you are using it, how it works, and the additional assumptions that you are making (what to set skewness and kurtosis to). In many cases for traditional asset classes there are very few differences in terms of the efficient asset mixes that the Log-Normal and Johnson models predict. It may not be worth the education effort to use the Johnson model when the Log-Normal model is an adequate approximation for your asset class set. The Johnson model also takes additional calculation time. 24. When should I use the Bootstrapping model? The Bootstrapping model is the only model that can represent non-linear correlations. It is also the only model that can represent certain unusual patterns of returns. For most asset classes there is a return range that is very likely and the likelihood of returns falls off from that range in both directions. If an asset class has some gaps where returns just can’t occur between two other areas where returns are likely then the Bootstrapping model is the only model that could account for this. The Bootstrapping model also allow you to isolate individual economic scenarios or time periods and make your results based on those scenarios occurring again in proportions that you set. For example if you assume that your asset classes will behave like they have historical but that high inflationary times are going to be more prevalent in the future then they were in the past you could isolate the high inflationary times in the historical data and give them a larger weight then they have in the historical data.

25. Do these new models mean the old Log-normal model is wrong? No, the log-Normal model can model the vast majority of traditional asset class behavior. The log-normal model has been successfully used for 60+ years and found such wide adoption because it is a good estimation. The Johnson model is more flexible and thus can model more behavior but for most traditional asset classes the differences are a refinement and not large changes. When you have asset classes that are particularly risky, highly skewed, or have high kurtosis then the results of the Johnson model can be very different then the log-normal.

8

Morningstar Direct Asset Allocation Certification ®

SM

26. What is Non-Linear correlation and can AA model this? Non-linear correlation is when the correlation between two asset classes depends upon the return range in which the asset classes are in. One thing that is frequently talked about lately is the fact that in market downturns or extreme conditions all correlations go to 1. This is an example of non-linear behavior. In normal times correlations have one value but they go up in extreme times. Only the Bootstrapping asset class model can model non-linear correlations. Both the Log-Normal and Johnson models rely on a single correlation value to describe the behavior between asset classes. You can determine visually if historical data has a linear relationship or not in the correlation window. In this window if you turn on the graph version and the returns between two asset classes fall generally along a line then a traditional correlation value is a good measure of the behavior between asset classes. If the values fall upon a line that changes slope or not along a line at all then the asset classes exhibit non-linear correlations. 27. What is serial correlation and can AA model this? Serial correlation is how predictive is the last return value of the next return value for a series. None of the asset class models in AA can model serial correlation. Modeling serial correlation would not generally affect the types of output displayed in AA and so we don’t really worry about this but it is something that is sometimes asked by clients. 28. How can I compare historical data and the inputs/Asset Class model I’ve created? In the asset class or asset mix distribution window you can go to the edit menu and turn on the histogram option. The histogram shows the historical data and the curve shows the asset class model. How well the curve follows the histogram bar heights is a measure of how well he model fits the historical data. Question 6 discusses why a tight fit may or may not be a good thing. 29. What does a non-positive semi definite correlation matrix mean? Non-positive semi definite is a mathematical term that means there are overlapping asset classes or contradictory information in the correlation matrix. This makes it impossible to do many of the calculations in AA. The difficult thing about determining overlapping asset classes is that they may be combinations of asset classes. For instance if you use the following three indexes as proxies to asset classes you will have overlapping asset classes as one’s returns are a weighted average of the other twos returns: Russell 3000, Russell 1000, Russell 2000. The problem with contradictory information generally occurs because a non-common time period is used to set the correlations. This can result in situations like the following: Asset Class A equals Asset Class B. Asset Class B equals Asset Class C but Asset Class A does not equal Asset Class C. This is mathematically impossible. For instance the following 3 periods of data demonstrates non common time period data that would produce a paradoxical relationship like this. 9

Morningstar Direct Asset Allocation Certification ®

Asset Class A

SM

Asset Class B

Asset Class C

1

N/A

3%

5%

2

3%

3%

3%

3

5%

5%

5%

30. Does Morningstar provide Inputs or CMAs? AA doesn’t contain any input assumptions that can’t be created in the tool using historical data. The OneYear. FiveYear, and TwentyYear input files are provided as three input files with different investment horizons. These are created using Building Blocks methodology for arithmetic mean and historical data for standard deviation and correlations. 31. Does Morningstar provide Ibbotson Associates Inputs or CMAs? AA doesn’t contain any capital market assumptions from what was Ibbotson Associates. EnCorr contained an input file base on Ibbotson Consulting’s work but the methodology used to create it was not disclosed to EnCorr users. This input file is not available to AA users.

32. What is the difference between the Asset Class Historical Statistics and Asset Class Simulated Statistics? The asset class historical statistics table shows the results of calculating the statistics from the historical data of the proxy series to the asset classes over the time range that can be set in the edit\settings menu. Historical statistics are not based on the asset class model but are just calculated as they would be in custom calculations and other areas of Direct. The asset class simulated statistics window shows the statistics calculated according to the forward looking asset class model that has been set up in Estimates. The simulated statistics are based on calculating the statistics on a set of returns from a Monte Carlo simulation run from the asset class model.

10

Morningstar Direct Asset Allocation Certification ®

SM

33. What is the difference between arithmetic mean (refined) and arithmetic mean? What is the difference between standard deviation (refined) and standard deviation? When you create your arithmetic mean and standard deviation in Estimates they are calculated at the lowest common frequency of your proxy series (usually monthly) and then converted to the return display frequency chosen. This frequency conversion is done by the industry standard frequency conversion methods that assume returns are additive not compounding. This is thus an estimate. When you choose a return display frequency other than the data frequency to complete the rest of the calculations for AA that require simulation we simulate return at the data frequency and then compound together groups of these return to create returns at the display frequency. Since this process uses compounding and the annualization method in input creation uses the estimated method that assumes returns are additive the number don’t match. The refined are the additive method and the plain values are the compounding method. We are looking into providing options for the user to control which method is being used throughout the software. So… Refined = additive, not compounded (referred to as estimated in EnCorr) Regular = compounded (referred to as precise in EnCorr)_ 34. What is Black-Litterman? What are the two main steps in Black-Litterman? Why would I use this? Black-Litterman is a method for determining the expected returns of Asset Classes. The first step is to compute the returns (called implied returns) that would be necessary for the portfolio of the Asset Classes according to the market capitalization weights to be efficient. This puts every asset class on in a common risk/reward tradeoff which promotes diversification since no asset class has a level of return that is not commensurate with its risk. The second step of Black-Litterman is the ability to specify views on how the returns may differ from the implied returns either in absolute terms or in relative terms to another asset class. The big advantage here is that the views take into account the relationships between the asset classes and preserve the common risk/reward tradeoff. For instance specifying that Large Cap Equity will out preform the implied returns will probably also result in an increase in the return of Mid Cap Equity as the two are positively correlated to some degree. http://datalab.morningstar.com/knowledgebase/aspx/files/Step_by_Step_Guide_to_the_Black_Litterman_Model.pdf

11

Morningstar Direct Asset Allocation Certification ®

SM

35. What is Building Blocks? Why would I use this? Building Blocks is a method for determining the expected returns of Asset Classes. The premise of Building Blocks is that the return of an asset class can be separated out into several factors and that these factors are more predictable then the asset classes themselves. For Equity asset classes the factors are the risk free rate, the return you should expect for investing in equity over bonds, and the return you should expect for investing in large cap over small cap. For fixed income the factors are the return you should expect for a risk less asset with any maturity component removed, the return you should expect for investing for a given horizon and the return you should expect for investing in Corporate over Government bonds. 36. What is CAPM? Why would I use this? CAPM is a method for determining the expected returns of Asset Classes based on a regression of a broad market index versus the series in question. CAPM provides a consistent risk/reward framework that promotes diversification in the optimization results. 37. What does the return display frequency setting do? The return display frequency setting controls what time scale (Monthly Returns, Annual returns…) will be used to display the CMA parameters and the calculations throughout inputs and optimization windows in AA. 38. What does changing the currency of an asset class do? Changing the currency of an asset class takes the return stream of the proxy and adds the currency risk of funding an investment in that proxy from another currency. For example imagine as a US investor you want to invest in a Eurodenominated series. You would have to first sell dollars to get Euros to invest. Then you would at some point sell out of the investment and trade the Euros received for dollars. Any change in the spot rate between when you bought the investment and sold it would change your return in dollar terms. By converting a Euro denominated proxy series to dollars you would be getting a more accurate reflection of the behavior of that asset class for someone investing in it in dollars. Another way to say this is that when you convert, you are converting the return stream into a different currency. This converted return stream will reflect the impact of the exchange rate fluctuation. For instance, as a US-based investor, you would assume that the returns you achieve in a Euro-denominated asset will be adjusted for changes in the $/Euro exchange rate. This is a simple, realistic scenario. This means you are not hedging; you are experiencing all the currency fluctuations. 12

Morningstar Direct Asset Allocation Certification ®

SM

39. Can I enter my own values for AM, SD, skewness, kurtosis, and correlations? Yes, you can enter your own values in the Input Summary table in the Estimates window. In addition to typing in the values you can copy the values from excel and then click on a cell in the Inputs Summary table where you want the values to start being pasted in from and press Ctrl + V to paste them in. 40. What is the simulation setting for in the input settings? Many of the calculations in AA require running a Monte Carlo simulation to produces returns that conform to the asset class model you create and then calculating the values from that simulated data. For instance when calculating the arithmetic mean of an Asset Mix you can just take the weighted average of the components arithmetic means to get the Asset Mixes Arithmetic mean. For statistics like CVaR there is no such formula where we can take the CVaR’s of the asset classes and calculate the CVaR of an Asset Mix. So instead we have to simulate a series of returns for the Asset mix and then calculate the CVaR of that series of returns. The simulation setting determines how many values are in that simulated data. The fewer the simulations, the faster the program runs but some of the results may be just based on the random numbers used in simulation. Generally this value shouldn’t be changed. If you are using asset classes with very, very high standard deviations then this may need to be increased to get reliable results. To think about this in historical data terms think of calculating historical statistics. If we use a low number of periods we may not get a fair representation of the stream of data. We might happen to pick periods that are all from Bear market time. By using more periods we guarantee we have the full picture of what the series does. The number of simulations can be thought of in the same way. Also see question on random seed. 41. What is the random seed setting for in the input settings? Many of the calculations in AA require running a Monte Carlo simulation to produce a series of simulated returns. In a Monte Carlo simulation you join together a set of random numbers generated by the computer and your asset class model to produce the simulations. If random seed is on, then the random numbers used change each time you change any of the inputs in AA. This can cause changes in the output especially if your number of simulations is low. Generally we want the same exact output values every time we use the same input parameters. Turning random seed on and running a calculation multiple times can show us how much the results change just based on the random numbers chosen by the computer. If the results change in meaningful ways, then the number of simulations should be increased.

13

Morningstar Direct Asset Allocation Certification ®

SM

42. What is the inflation series used for in the input settings? Can I create capital market assumptions that take inflation into account? You can mark one of your asset classes as representing inflation. This series will then be used to adjust forecasting results for inflation if you turn on inflation adjustment in forecasting. This doesn’t affect optimization. If you want optimization to consider inflation, the best option is to create custom benchmarks for each asset class proxy that are a combined series of an index with an inflation index geometrically subtracted. If you do this then you would not also want to select an inflation asset class, as doing so would mean double counting inflation in forecasting. 43. What is the data frequency shown in the input settings? The data frequency shown in input settings is just for informational purposes. It is the lowest frequency of any of the proxy series you have chosen. Historical input parameter calculation is done in this frequency and then holding period converted to the display frequency. 44. Name the four methodologies that can be employed to develop expected returns? Historical, Building Blocks, CAPM, and Black Litterman 45. Can I create capital market assumptions that will take into account fees or taxes? There is no built-in functionality for this. You could for example just take a set percentage off of the Arithmetic mean for each asset class. What you should do would depend on the structure of the fees or taxes and would probably be just a simple approximation as no real functionality for this exists. 46. When you’re running a model, explain the correlation matrix condition number. At what breakpoint should you be concerned with the correlation matrix condition number? The condition number of a correlation matrix gives a measure of how close the matrix is to containing mathematically impossible relationships or asset classes that are identical to another asset class or combination of asset classes (known as collinearity). When either a mathematical impossibility exists or a collinearity then optimization and forecasting are not possible using this correlation matrix. Imagine if you ask the optimizer to make a decision between the Russell 1000 and Russell 2000 or the Russell 3000. There is no correct answer because the 3000 is a mix of the 1000 and 2000.

14

Morningstar Direct Asset Allocation Certification ®

SM

As a correlation matrix is still feasible but is approaching the point where it is not, the results of Optimization become very sensitive to small changes in the capital market assumptions. The point at which we warn about the correlation matrix becoming ill-conditioned is when it reaches a value of 20. The ideal range is from 1-20. When this number gets above 20 the user should consider if they have included overlapping asset classes and if so remove them. They should also try to use a common time period that all asset classes have data for if they are calculating the correlations from historical data. The point at which the correlation matrix is unusable for everything except MVO Optimization is when the condition number is undefinable or NA. This is also known as a matrix that is non-positive semi-definite. Also see Question 29 about the term positive semi-definite. 47. Cite examples of an overlap when dealing with correlations. Equity Indexes in same region that have similar type of market caps (Large Caps: SP500, R1000, etc..) Bond Indexes in same sector that have similar maturities Equity Indexes in same region that have the overall index, growth index, and value index (R1000, R1000G, R1000V) Equity Index that are Global but you have the US and Intl component as well.

48. Why should you not put mutual funds or other investable securities into the optimizer as asset classes? The results of the optimization are only as good as your capital market assumptions. The behavior of true asset classes is more likely to repeat historical behavior or be predictable based on current economic trends. Mutual fund or other securities performance can be wildly affected by anything from a change in managers, to a different strategy taken on by a manager or just bad luck in stock picking inside the same strategy.

15

Morningstar Direct Asset Allocation Certification ®

SM

49. What are the baseline settings used for? The baseline settings affect the various arithmetic mean methods. Refer to the screenshot below and the description of the arithmetic mean methods.

16

Morningstar Direct Asset Allocation Certification ®

SM

Optimizer 1. What is an asset mix? An asset mix is a portfolio of asset classes. The weights of the asset classes must add up to 100% but may contain negative numbers (short positions). 2. What does Optimization do? Optimization in AA is the process of identifying the asset mixes that according to the asset class model (CMAs) and the constraints have the highest level of return for every possible level of risk. The Optimization results in graphical form are known as the efficient frontier. 3. What do the points on the efficient frontier represent? They represent an efficient or optimal mix of asset classes in the opportunity set. Therefore, each point on the efficient frontier maximizes the expected return per unit of risk or, vice versa, minimizes risk per unit of return. 4. What does MVO stand for? What is MVO? Mean variance optimization. MVO is optimizing with arithmetic mean as your measure of return and standard deviation as your measure of risk. It makes the assumption that asset class returns are normally distributed. 5. How should you pick which point on an efficient frontier is best for a given client? All the points on the efficient frontier are optimal. They are just different choices. The question of matching a client’s utility function or comfort level with the spot of the efficient frontier that is appropriate for them has never been an exact science. Some common ways of doing this are through a risk tolerance questionnaire, presenting the range of outcomes for a few select optimal asset mixes, or by having a risk budget or maximum level of risk specifically given by the client. AA doesn’t contain a risk tolerance questionnaire as this is more used in an individual investor situation, which is not the primary focus of Direct.

17

Morningstar Direct Asset Allocation Certification ®

SM

6. Why might you pick something other than a point on the efficient frontier? Sometimes there are real world constraints or behaviors that can’t be modeled in AA. This may require tweaking an optimal asset mix to produce something near the frontier that is not technically optimal. For instance a frequent request is to not have weights in increments of less than 1% or 5%. 7. Can I optimize with short positions or leverage? Yes you can set your minimum holding for each asset class to a negative number and the maximum holding to 0.This will allow you to take a short position in that asset class. You may also want to allow your long asset classes to go higher than 100% in the situation where you allow short positions. Due to a quirk in MVO you can’t have an asset class constrained to be either long or short. You have to pick ahead of time. So if you want this to be possible you must create two versions of each asset class. One that is constrained to be short and one long and then in your optimization results you would net the two positions to get your weight for that asset class. 8. Can I do surplus optimization or liability-driven investing (LDI)? This has not been implemented in AA yet. 9. What Types of Return can you optimize on? You can optimize on arithmetic mean or geometric mean. 10. What are the advantages and disadvantages of Optimizing on arithmetic mean versus geometric mean? Arithmetic mean is the average return over the next period but gives no indication about the long term growth potential or the return a user could expect to compound year after year. It’s a one period average result of all past data. In the investment community we are used to seeing performance numbers, especially historical performance as annualized returns we can expect to compound period over period, also known as Geometric returns. An arithmetic average can be misleading and overly optimistic because two series whose returns exhibit the same arithmetic mean but different standard deviations can have very different cumulative returns over time. Therefore, arithmetic mean is not good for forecasting wealth accumulation whereas Geometric is a multi-period compounded expected return, which is better for forecasting wealth accumulation. 18

Morningstar Direct Asset Allocation Certification ®

SM

The advantage of arithmetic mean is that it has been used in traditional asset allocation tools for 60+ years, so it is the industry standard for optimization. Asset Allocation practitioners expect to see arithmetic mean. It is also often faster to calculate optimizations results using arithmetic mean than geometric mean. 11. What Types of Risk can you optimize on? Which of these are downside risk measures? You can optimize on standard deviation, CVaR (conditional value at risk), First Lower Partial Moment (FLPM, or LPM1), Downside Deviation, and SMDD Standard deviation. CVaR, FLPM, and Downside Deviation are downside risk measure. 12. Why would I use a downside risk measure versus standard deviation? Standard deviation penalizes for upside deviation or extreme positive returns as much as it does negative returns. By using a downside risk measure, you find asset mixes that minimize losses and are indifferent to upside behavior. Most investors are risk averse and feel worse about returns below the mean than they feel good about returns above the mean by the same amount. Down side risk measures thus find a set of portfolios that are more optimal in terms of an investor’s true preferences, rather than just being optimal based on standard deviation. 13. Why have standard deviation then? Despite its shortcomings, it is still the most commonly used measure of risk Arithmetic mean and standard deviation were chosen by Markowitz as the measures of risk and return for MVO because their mathematical properties made computation of the efficient frontier easy. Because of their use for 60+ years they have become ingrained in people’s expectations of what an asset allocation optimization does. The other nice thing about standard deviation in the normal distribution model is that most people are familiar with the 68-95-99 rule that provides a framework for people to understand what a standard deviation of any value means in practical terms. 14. What is CVaR? CVaR is the average loss in adverse times. You set what you consider to be adverse times by specifying what percentage of the outcomes to look at. This is the CVAR cutoff %. For example, if you set a cutoff of 5%, then you are getting the average loss in the worst 5% of outcomes. This means that 1 out of 20 times you are having a bad outcome. So to be comfortable with an annual CVaR of 5% over a 20 year time horizon you should be prepared for your worst one year on average to have a loss of 10%. CVaR is very intuitive because it is specified in loss terms unlike standard deviation which is the square root of the sums of the squared differences from the mean. Standard deviation is not very straightforward but only feels like it is because of it widespread adoption. 19

Morningstar Direct Asset Allocation Certification ®

SM

7A.Explain Target Allocation A, B, and C, provided your CVAR cutoff is 5%

th --If I select Target Allocation A, I'm going to lose 4% on average if I’m in the bottom 5 percentile of this allocation’s

outcomes.

th --If I select Target Allocation B, I'm going to lose 8% on average if I’m in the bottom 5 percentile of this allocation’s

outcomes.

th --If I select Target Allocation C, I'm going to lose 16% on average if I’m in the bottom 5 percentile of this allocation’s

outcomes. 7B. Finish this sentence: As I pick a more aggressive target allocation along the CVaR/geometric mean frontier, I will have ___________________ A higher chance of incurring larger losses short term, but long term, I will have more gains potentially.

15. Why can’t I optimize on VaR. It is available in the statistics table. VaR has mathematical properties that make it difficult to optimize on. It is also only a measure of behavior at one particular point in the distribution. This can cause the optimal weights to be very unstable and non-intuitive even when you can calculate the frontier. VaR really comes from a regulatory background and we feel CVaR is both mathematically more convenient and also matches investors’ real preferences better.

20

Morningstar Direct Asset Allocation Certification ®

SM

16. Why are my historical statistics table values so different from my simulation statistics table values? There are many reasons for this. The most straightforward is that the parameters for the asset class model may not have been calculated from historical data. The asset class model may be forward looking or based on active decisions by the creator about how future behavior may not match the past. The simulation statistics are based on the asset class model. Even when the inputs are based on historical data the model is a smoothed fitted representation of the historical data (for more info read question about parametric versus non-parametric models).

17. Explain what Downside Deviation means. Downside deviation is also a measure of downside risk but rather than focusing on the average loss of bad outcomes (CVAR), you're determining how far you fall below the target and how often. It's the standard deviation of the shortfalls to the target. Therefore, if you input a 2% Target, then that is your threshold (see calc below) This differs from the Sortino Ratio in that the Sortino Ratio completely ignores events that fall above the return target, whereas downside deviation counts them as events with zero returns. Hence, downside deviation implies lower volatility than the Sortino Ratio.

21

Morningstar Direct Asset Allocation Certification ®

SM

17A. Explain Target Allocation A, B, and C, provided your target is 2% for the downside deviation below a given target.

-- If I select Target Allocation A, The standard deviation of the shortfalls is 2% -- If I select Target Allocation B, The standard deviation of the shortfalls is 3% -- If I select Target Allocation C, The standard deviation of the shortfalls is 5% 17B. Finish this sentence: As I pick a more aggressive target allocation along the geometric mean/downside deviation frontier, ___________________ the times I miss meeting my target in the short term are more often and/or miss by more but I have a higher potential long term wealth growth.

18. Explain what First lower partial moment means? First lower partial moment is also a measure of downside risk but rather than the standard deviation of the shortfalls to the target (downside risk), it's the average of the shortfalls to the target. So if you input a 2% target, that is your threshold. It follows the same exact steps as downside deviation, but this time, it's taking the average and not the standard deviation of shortfalls.

22

Morningstar Direct Asset Allocation Certification ®

SM

18A. Explain Target Allocation A, B, and C, provided your target is 2% for the 1st lower partial moment below a given target.

--If I select Target Allocation A, my average shortfall to the target is 0.5% --If I select Target Allocation B, my average shortfall to the target is 1.0% --If I select Target Allocation C, my average shortfall to the target is 2.0% 18B. Finish this sentence: As I pick a more aggressive target allocation along the geometric mean/1st Lower Partial Moment Frontier, ___________________ the times I miss meeting my target in the short term are more often and/or miss by more but I have a higher potential long term wealth growth.

23

Morningstar Direct Asset Allocation Certification ®

SM

19. How does 1st lower partial moment differ from downside deviation? They both start with the shortfalls to target. 1st lower partial moment is just the average or arithmetic mean of these values. Downside Deviation is the standard deviation of these values. Downside deviation matches the preferences of

st an investor who is more risk adverse than 1 Lower partial moment since the penalty for a shortfall increases with the

square of the difference from the target, rather than linearly. For example for one shortfall of 4% has more of a penalty than two shortfalls of 2% under Downside deviation. 20. How is downside deviation different from the Sortino Ratio? Sortino ratio is like a Sharpe Ratio, (so excess return / standard deviation), where the standard deviation is replaced with downside deviation. Furthermore, that downside deviation in the Sortino ratio only utilizes the outcomes where you miss the target return, whereas the downside deviation we are using here in Direct counts the outcomes where you do not miss your target, assigning them zero. 21. How does optimizing on arithmetic mean versus standard deviation (MVO) differ from all other risk/return metric optimizations? Optimizing on arithmetic mean versus standard deviation uses MVO optimization whereas all other options are based on a Monte Carlo simulation. MVO is quicker to calculate than Monte Carlo-based optimization and doesn’t rely on random number generation, thereby avoiding the need to specify how many simulations to base the results on. 22. What is a Monte Carlo Simulation? Why are they used? The Monte Carlo simulations done in AA produce a series of hypothetical returns that conform to the asset class model. The returns conform both in the single asset class parameters (arithmetic mean, standard deviation, skewness and kurtosis) and the joint behavior (correlations) and so we can calculate statistics from the simulation data and assume those to be characteristic of the asset class model that produced them. Sometimes statistics can’t be calculated from the formulas that describe the asset class model directly, but by doing Simulation these calculations can be accomplished. Monte Carlo simulations are done throughout AA, from the Asset Class statistics (simulated) to Optimization and forecasting. The only time simulation is not involved is when Optimizing using MVO.

24

Morningstar Direct Asset Allocation Certification ®

SM

23. Do you really optimize on these other measures or just do MVO optimization and label the results with these other measures? Many clients have asserted that we must not really be optimizing on all these risk and return measures but just labeling results in these different scales. This is not true. Optimization actually searches to find the asset mixes that are optimal for the risk and return measures selected. 24. What problems does resampling address? Resampling is a combination of the traditional Mean Variance Optimization and Monte Carlo Simulations. It recognizes that Capital Market Assumptions are forecasts and not a “sure thing”. Traditional Mean Variance Optimization produces results that can change wildly with small changes in the Capital Market Assumptions. It also tends to produce results that are not well diversified. The resampled efficient frontier averages the optimal results under many different sets of Capital Market Assumptions that are similar to the original and thus produces more diversified and robust portfolios -- hence, asset allocations that remain more stable over time. 25. How does resampling work? Resampling runs a Monte Carlo simulation to produce a series of returns from the original input assumptions. Because this series of returns is fairly short when the arithmetic mean, standard deviation and correlations are calculated from these returns, they don’t exactly match the original assumption. This set of similar asset class assumptions are then optimized with MVO. We do this a few hundred times and then take all the optimal asset mixes from these similar frontiers and plot them back on the original efficient frontier graph according to the original input assumption. We then average all the asset mix weights over a small range of standard deviation to produce a resampled optimal asset mix. These asset mixes are then joined through interpolation to create a frontier. This averaging of asset mixes that come from a range of plausible future realized capital market assumptions produces asset mixes that do well on average across all the plausible outcomes.

26. How does AA’s resampling differ from Michaud’s/New Frontier Analytics? The only difference is in the step of identifying asset mixes to average together. There is no conclusive evidence of which method is better. Resampling is inexact by design. NFA has a more technical sounding way of identifying asset mixes to group together but there is no mathematical proof of its superiority. Our own testing produced very inconclusive results.

25

Morningstar Direct Asset Allocation Certification ®

SM

27. What is SMDD Standard Deviation? SMDD Standard Deviation is optimizing on standard deviation through Monte Carlo simulation rather than MVO. MVO assumes normally distributed asset classes, so if you optimize on plain standard deviation, then you will not take into account skewness or kurtosis even if you use the Johnson distribution model. 28. MVO is a single period optimization. Do we have multiple period optimization? Multi-period optimization is a term that is used to refer to a number of different things. Sometimes it refers to optimizing over several sets of capital market assumptions to come up with either one asset mix that performs the best across all, or a set of asset mixes where each is optimal for a set of CMAs (just multiple optimizations done at once, but with no linkage between them). Traditional MVO uses the next one period expected return as the measure of return. So this is clearly a single period model. Some people also refer to optimizing on geometric mean as a multi period optimization as the asset mixes’ return is reported in geometric terms which, unlike arithmetic mean, don’t change, regardless of time horizon. AA has optimization on geometric mean, but does not have the other multi period processes. 29. Why are Johnson model frontiers generally more pessimistic than Log-Normal? The normal/log-normal models fitted to historical data generally under-represent the likelihood of extreme events. The Johnson model fits historical data better and so is generally more pessimistic about the return that can be achieved at any risk level. 30. Why are my optimal asset mixes so concentrated in a few asset classes? Isn’t asset allocation about diversification? What can be done about this? This is a classic problem of asset allocation. The main cause is input assumptions. Input assumptions, especially those created from historical datamake a few asset classes much more attractive than other asset classes in a return per unit of risk sense. Using models for coming up with arithmetic mean like Black-Litterman or CAPM produce asset classes that pay approximately equally in risk for each unit of return and thus more asset classes are likely to be included in the optimal portfolios. Resampling also produces more diversity by considering all input assumption similar to the original as equally likely and producing “optimal” portfolios that do well under all these different input assumptions. Constraints are another common way of addressing this problem.

26

Morningstar Direct Asset Allocation Certification ®

SM

31. Why don’t the efficient asset mixes change much when I switch to geometric mean and one of the downside risk measures? If you are using the Log-Normal model, the downside risk is approximately the same as upside risk, so standard deviation and the downside risk measures produce basically the same optimal allocation. 32. Why don’t the efficient asset mixes change much when I switch to the Johnson model and geometric mean and one of the downside risk measures? MVO and the Normal/Log-Normal distribution model have been used for 60+ years as they are a very good approximation of asset class behavior for traditional asset classes. The biggest limitation of Normal distributions is that they underrepresent extreme event. Since they do this pretty much equally for all asset classes, the addition of this doesn’t change the optimal portfolios much. It changes their risk level but not their weights. It is only when you have particularly skewed or fat-tailed (kurtosis) asset classes that the use of a downside risk measure and the Johnson model will make a big difference. For instance the use of hedge fund indexes, commodities or other untraditional asset classes. For most users, the Johnson model is more useful for the difference in forecasting results because of the more realistic modeling of extreme events than it is for the changes it makes to the efficient asset mixes. 33. Why can the standard deviation of an asset mix be lower than the standard deviation of all the asset classes that make it up? When asset classes have inverse or negative correlations to each other, their risks can actually cancel out and lower the standard deviation of a combination of asset classes is below the weighted average standard deviation the individual asset classes. This is the key benefit of diversification. 34. Continuing from the previous question, when you switch the x-axis on the efficient frontier graph to another risk measure, what happens? The system re-optimizes using the selected risk measure. 35. What three types of optimization constraints does Direct support? Single, Group, and Relative constraints

27

Morningstar Direct Asset Allocation Certification ®

SM

36. What is the difference between individual, group, and relative constraints? Give examples. Individual: This is the ability to specify a minimum and maximum percentage that an asset class can have in any of the asset mixes on the efficient frontier. Max Holding of Real Estate is 20% or Min Holding of US Equity is 30%. Group: This is the ability to specify that the sum of the allocations to a group of asset classes fall in a range (min holding to max holding). This might be used for instance to guarantee that the allocations to all equity asset classes is never more than 50% of the entire portfolio.

Relative: This allows you to specify that one group of asset classes can have more or less allocation than another group of asset classes in all of asset mixes on the efficient frontier. This is most commonly used in circumstances like when you have three foreign equity asset classes and five domestic and you want the sum of the foreign equity allocations to be less than the domestic. Relative constraints also allow for coefficients. This allows, for example, for a situation where you want foreign equity to be less than 20% of the allocation of domestic equity.

28

Morningstar Direct Asset Allocation Certification ®

SM

Forecasting 1. What is forecasting? Forecasting takes your Capital Market Assumptions and predicts the wealth and returns of your portfolio over various time periods, taking into account real-world concerns like inflation and cash flows. 2. Why are the results all in percentile terms? Our input assumptions contain information about the potential range of returns (standard deviation). Therefore, the results are probabilistic, i.e. they show the range of outcomes. They show, for instance, the minimum return you could

th expect in a particularly good outcome (a top 10% outcome should return at least the 90 percentile value), or a th

particularly bad outcome (a bottom 10% outcome should return the 10 percentile value or worse). 3. Can I view forecasting results in real or inflation adjusted terms? Yes. This can be done in two ways. The most straightforward way is to create an inflation asset class. If you don’t want it included in optimization results it can be constrained to a 0% maximum holding. In other words, if you do not want to allocate funds in your strategy to this asset class, set the constraint to zero; typically, setting this constraint to zero is what people do. You then select the inflation asset class in the input\options dialog. When you forecast, in the forecast settings, there is then a checkbox for inflation adjust. If this is checked, the return and wealth percentile values will be in real terms. Also, you can then inflation adjust cash flows in the cash flow setup. If you inflation adjust a cash flow. the dollar amount entered should be in start of simulation dollars. If you don’t inflation adjust it, it should be in time of cash flow application dollars. 4. Why is inflation treated as an asset class? What is the advantage of this? By treating inflation as an asset class we model not only its own behavior, but also its correlation to other asset classes. This produces more realistic simulations than a constant inflation assumption. 5. What if I just want a constant inflation rate? Set the standard deviation for the asset class to a very small number (.00001, 0 is not accepted). Set the arithmetic mean to the constant inflation value that you want.

29

Morningstar Direct Asset Allocation Certification ®

SM

6. Can I model taxes or fees in forecasting? This can only be done through cash flows. There is no functionality designed specifically for this. 7. What is the difference between forecasting frequency and return display frequency? Forecasting frequency controls how often things happen in the simulation. For example in a monthly forecasting frequency you can apply cash flows and see wealth values monthly, in an annual you can only do this annual. Return display frequency controls the scale the returns are shown in. For instance, do you want to see an annualized return or the monthly return you could expect to receive? 8. What is use random seed? See random see question under inputs 9. Why would I change the number of simulations? See # simulation question under inputs

10. What order do things happen in during the simulation? How does this affect the result? AA takes the most conservative approach to simulation timing. At the beginning of each period the rebalancing rule is checked and rebalancing is done if necessary. Then any cash outflows are applied. Next the return is applied and finally any cash inflows are applied. So long as returns are positive, of course, this is the most conservative approach. 11. Can I change my asset mix over the course of the forecast? For instance if I’m looking to get more conservative as the forecast progresses can I do this? The only way an asset mix changes currently is because of not rebalancing. You can’t explicitly change the asset mix for instance to model a life cycle fund.

30

Morningstar Direct Asset Allocation Certification ®

SM

12. How does periodic rebalancing work? Assuming a quarterly forecast what would rebalance every 2 periods if 5% change mean? Periodic rebalancing checks ever x periods if any of the asset class weights is + or – y% from what it started from. In the example given every 6 months we would check if any of the weights differ from the original by 5% or more. If it does we rebalance and check again in 6 months. E.g. if the target allocation to US Large Cap was 24%, and the actual allocation was between 20 and 28%, we would not rebalance, but we would rebalance if the actual allocation had risen to 29% or dropper to 19%. If it doesn’t we wait 6 months and then check again. th

13. Why doesn’t the 50 percentile match the expected return of my asset mix? The only time they match is when your distribution is symmetric. Even the log-normal distribution is slightly positively skewed. Note: You can’t control the skewness (see normal/log-normal difference question). 14. What are the three types of cash flows in AA? Monetary Amount % of Initial Wealth % of Most Recent Value 15. What does it mean to inflation adjust a cash flow? When you inflation adjust a cash flow you can enter the cash flow values in start of simulation dollars, and have them automatically grow with inflation. 16. Can I inflation adjust the cash flows but not the return values? No, currently you can only adjust the cash flows if the whole simulation is adjusted. 17. Why don’t my return values reflect my cash flows? Return is not a function of cash flows. For instance if you had $1 and made a cash inflow of $1 and had no return should this be considered a 0% return or a 100% return. We believe it should be considered a 0% return; return is a measure of you asset class properties and the mix you have chosen, not how you contribute to the portfolio.

31

Morningstar Direct Asset Allocation Certification ®

SM

18. What does the cumulative probability of loss table show? When would you use this? For instance what does a 95% cumulative probability of a loss of 5% after 10 years mean? Cumulative probability answers a very risk adverse question. It is designed for someone who says I can’t lose more than x% at any time from my starting wealth. It shows you the probability that a loss of the specified magnitude will have occurred any time up to the current time period. So, in the example provided, 95% of the simulations were down 5% from their initial value sometime over the course of 10 years. Now this doesn’t mean that some or in fact most of these have probably rebounded by the end of 10 years. Most of the losses will occur at the beginning of a simulation as they are reported in annualized terms and it gets more and more difficult to achieve large loses in annualized terms due to convergence over time. 19. When are cash flows triggered? Are they triggered monthly between the start and end dates, annually or something else? A simulation period occurs every “forecasting frequency” after the initial date. So assume an initial date of 2012-05 and a forecasting frequency of quarterly. The first simulation period will be 2012-06 through 2012-08. To determine if any cash flows should be applied to this simulation period, each of the cash flows begin and end dates will be checked against this date range. If they intersect at all then the cash flow will trigger once for this simulation period, with outflows being applied to 2012-06 and inflows to 2012-08. Next we will move on to the 2012-09 through 2012-11 simulation period, and we will check for the intersection again to see if cash flows should be triggered. So they are triggered in the “forecasting frequency”. 20. What is the data frequency? This just tells you want the common frequency of our proxy series are. 21. What does show Back History do? It shows you how your asset mix performed historically up until the start of the forecast. 22. As you’re interpreting the forecasted results, how do you explain a 10% return for 5 years at the 95

th

percentile in the return percentiles table, if you’re explaining this in Dec 2012? This is saying that at the 95th percentile you would have an annualized return of 10% per year. A 5% chance of achieving a 10% annualized return or more over 5 years. A 95% chance of achieving an annualized return of less than 10% over 5 years. 32

Morningstar Direct Asset Allocation Certification ®

SM

23. Does forecasting reflect the choice of distribution model and the parameters set when creating inputs or does it assume a normal distribution? Forecasting is just like optimization. We use the distribution model chosen during input creation. We run the Monte Carlo simulation to produce individual period returns from your distribution model and then group them together into simulations through compounding to create cumulative return over time. We then calculate all of the percentile values off of this return stream. The asset class model you chose during input setup is the driver of the results. Limits 1. How many input files or efficient frontiers can be compared in one case file? 5. Since there is one efficient frontier per input file, this also means you can open up to five input files in a case file. 2. How many asset mixes can be added to each frontier or input file in a case file? 10 3. How many asset classes can I have in one input file? Up to 100 asset classes can be added to one input file. As the asset classes increase the number of simulations you can do is decreased automatically to prevent over-use of the servers. General Questions 1. What is the difference between an input file and a case file? An input file consists of your selected asset classes, your select asset class model, and the parameters that go into that model. A case file consists of your selected input files, which windows you are viewing in your workspaces, and all the display settings that go into showing those. In addition, all your forecasting parameters are saved in the case file. 2. What type of organization/client would use Asset Allocation Modeling? Wealth Managers, Tactical Asset Allocation Committees, Investment Policy Committees, Plan Sponsors, Advisors

33

Morningstar Direct Asset Allocation Certification ®

SM

RESEARCH 1. You’re a wealth manager and you have to identify new target allocations for an existing portfolio. Based on your client’s choice of funds, how do you begin researching the proxies to represent your asset class assumptions in Direct? You can create an investment list of the funds, run a Holdings Based or RSBA chart either in Workspace or in Presentation Studio BUT given we don’t have a global style index (like MPI) and the fact that we can’t bring in our own indexes and save them (like MPI/EnCorr), you have to piece meal this and separate out the asset classes (bonds vs. fixed income, domestic equities vs. intl equities) when running the regressions. Therefore, a more efficient way, is to run a correlation matrix in Excel API of the funds against their assigned or prospectus benchmarks for 3 Years. The result will show how well these funds are correlated to their benchmarks. Presuming they have high correlations, you can then use these indexes as the benchmarks. If their correlations are not high, then go back to workspace, grab their best fit indexes, and apply the best fit indexes Excel API to rerun the correlations. If Morningstar Index is the Best Fit, replace this with another comparable index. GENERAL QUESTIONS 1. Terms synonymous with Asset Allocation Modeling Strategic Asset Allocation Policy Portfolio Strategic Policy Benchmark 2. Once you’ve built your target allocations, what module can you use to apply your new target allocation weights to your investment policy? Explain why. Total Portfolio Attribution: Once you’ve created your asset allocation modeling in Direct’s Asset Allocation Tool, you can implement the weights by creating your model portfolio, bringing it into Total Portfolio Attribution to apply your new target weights, and ultimately to determine if you added value at the Tactical Asset Allocation decision level, at the Active Management decision level, or both. More to come in the future here, as we are working on interaction between AA and PMS, which will open the door for PR and PA, too.

34