Efficient Frontier Analysis:

1
KALSON ASSOCIATES
Efficient Frontier Analysis Focus on Municipal
Operating Portfolios FGFOA Annual
Conference Orlando, Florida May 22, 2007


2
Agenda

• Background Why should a finance officer
  grapple with
• asset allocation? Isnt cash the only game in
  town?
• Asset Allocation Model Inputs
• Running the Model
• Output of the Model
• Key Findings, Recommendations, Limitations
• Questions

Kalson Associates
3
Why Bother to Conduct an Asset Allocation Study?
4
Background for a Typical Municipal Operating
Portfolio

• Many years since the last allocation study was
  performed
• if at all and newer bond sectors and
  asset classes have
• become mainstream
• Current money market rates are acceptable, but
  these
• relatively high rates wont last
• Very few knowledgeable forecasters expect gt
  5/year
• returns from typical bond sectors
• Municipal budgets are tight and every penny
  counts
• City Council has an understandably conservative
  risk
• tolerance

Kalson Associates
5
Our expectations going in

• It will be difficult to generate 5 return per
  year during
• the next 5 years
• Therefore, prudent diversification is
  appropriate to
• achieve a potentially higher rate
• Extending duration will be beneficial, assuming
  the
• return to an upwardly sloping yield curve
  environment

Kalson Associates
6
Asset Allocation Model Inputs
7
The Asset Allocation Model

• Model output is only as good as the input!
• The three key inputs
• Five-year expected returns for appropriate bond
  asset classes we believe the use of historical
  returns is NOT the way to go
• Expected volatility (standard deviation) around
  the average expected returns
• Expected relationships (correlations) between
  pairs of asset classes as they fluctuate in the
  marketplace over time

Kalson Associates
8
Inputs 1 2 Expected Returns and Volatility

• Step 1 We collected forecasts of asset class
  returns and risk from respected bond management
  organizations, including
• WAMCO
• BlackRock

Kalson Associates
9
Inputs 1 2 Expected Returns and Volatility

• Step 2 Aggregate the forecasts we obtained
• Step 3 Make minor adjustments to the
  forecasts to achieve a more rational
  set of outcomes

Kalson Associates
10
Input 3 Correlations

• Cross-correlations, or relationships between
• asset classes, change slowly
• Therefore, we merged historical and forecasted
• correlations to arrive at our assumptions.
  

Kalson Associates
11
What are Correlations?

• Cross correlation statistics range between -1.0
  and 1.0.
• A correlation of 1.0 suggests the highest degree
  of co-movement between two asset classes.
• A correlation of -1.0 indicates two asset classes
  moving in opposite return directions over time.
• A correlation of 0.0 indicates no co-movement
  between asset classes.

Kalson Associates
12
Sample Correlations for the latest 10-years
Kalson Associates
13
Running the Model
14
Asset Classes Considered
Low or Zero Current Exposures In the Portfolio
Significant Current Exposures In the Portfolio

• T-bills
• 1-3 year Gov/Credit
• Intermediate Gov/Credit
• Mortgage-backed (MBS)

• CMBS
• Broad market Gov/Credit
• Emerging market debt (EMD)
• High Yield
• Sovereign Risk

Kalson Associates
15
How the Allocation Model Works

• The model blends the three forecasted inputs and
  develops an efficient set of outputs.
• For instance, given a certain risk level, what is
  the highest possible combined return that could
  be obtained?
• or
• Given a specific return, what is the lowest risk
  outcome?
• Each run tries approximately 1,000 combinations
  of asset classes
• Review 11 efficient portfolios with lower to
  higher risk levels

Kalson Associates
16
OPERATING PORTFOLIO ALLOWABLE ASSET CLASSES
Lehman 1-3 yr Govt/Credit
Lehman Sovereign
Lehman CMBS
Lehman Intermediate Govt/Credit
Lehman Mortgage-Backed
Lehman Govt/Credit
Lehman High Yield
Lehman Emerging Markets
Lehman 3-Month T-Bill
Citi 1 Year Treasury
17
Running the Allocation Model

• We started with the operating portfolios actual
  allocations as our baseline.
• We ran the model unconstrained key outputs
• 7.1 expected return versus the actual
  portfolios expected
• 5.4. The offset is a 5.8 standard
  deviation vs. 2.2.
• 50 allocation to high yield bonds is far too
  concentrated
• 21 allocation to EMD also is too concentrated

Kalson Associates
18
Asset Class Constraints

• We placed mild and then strong constraints on
  several asset classes
• Then, we determined if the returns, risk and
  asset class allocations made sense
• Overall, our modeling tested 3 separate scenarios
• We believe Run 2 and Run 3 provides the best
  return/risk tradeoff, with allocations that make
  common sense.

Kalson Associates
19
Run 2 Constraints

• We reduced our maximum allowable high yield
  exposure to 15
• We reduced our maximum allowable EMD exposure to
  5

Kalson Associates
20
Run 2 Results

• Of the 11 portfolios generated by the allocation
  software, we chose to focus on Portfolio 3
• Portfolio 3 provided a 6.3 expected return,
  and a 3.9 risk (standard deviation)
• For reference, the plans current allocation
  projects 5.4 expected return with a 2.2 risk
• This portfolio suggests meaningfully adding three
  asset classes high yield, broad market
  Gov/Credit and EMD

Kalson Associates
21
Run 3 Constraints

• All of the same constraints as Run 2 except
• we reduced our maximum allowable high yield
• exposure to 10 (down from 15 in Run 2).
  

Kalson Associates
22
Run 3 Results

• Of the 11 portfolios generated by the allocation
  software, we chose to focus on Portfolio 3
• Portfolio 3 provided a 6.2 expected return,
  and a 3.9
• risk
• Similar to Run 2, this portfolio adds high
  yield, broad market G/C and EMD

Kalson Associates
23
Our Focus

• The best way to choose one portfolio over another
  is
• to look for a combination of the following
• higher expected returns
• lower expected risk
• 3) Finance Committee/City Council comfort with
  scenario constraints
• 4) common sense allocations
• 5) Avoid an abrupt overhaul of current asset
  classes

Kalson Associates
24
Actual Output of the Model (Scenario 3)
25
Lehman Intermediate Govt/Credit
Lehman Mortgage-Backed
Lehman Govt/Credit
Lehman High Yield
Lehman Emerging Markets
Lehman CMBS
Lehman 1-3 yr Govt/Credit
Lehman Sovereign
Lehman 3-Month T-Bill
Citi 1 Year Treasury
Forecasted Return
Forecasted Risk

• Scenario 3 Constraints
• 0 - 10 in Lehman High Yield
• 0 - 5 in Lehman Sovereign
• 3. 0 - 5 in Lehman Emerging Markets
• 4. 0 - 50 in Lehman Mortgage-Backed
• 5. 0 - 30 in Lehman CMBS
• 0 - 50 in Lehman Inter. Govt/Credit
• 0 - 50 in Lehman Govt/Credit

26
(No Transcript)
27
Portfolio 3
Lehman Aggregate
Initial Portfolio
28
Key Findings Recommendations Limitations
29
Key Findings

• Increasing targeted duration from 1.7-years 1-3
  year index to 4/5-years L Agg index increases
  expected return by 1 at an acceptable 1.7
  percentage point higher risk
• Since lengthening duration doesnt dramatically
  increase forecasted risk , the model is clearly
  partial to broad market exposure
• Due to relatively high expected returns and low
  correlations, the model also likes high yield
  and EMD
• The model introduces a higher level of
  Intermediate G/C to reduce overall portfolio risk
  

Kalson Associates
30
Recommendations

• Shift from a lower to a higher duration stance
• Introduce a significant allocation to high yield
  and a
• small direct exposure to EMD
• Maintain a separate portfolio for immediate cash
  needs
• Hire managers that you expect to outperform the
  
• generic forecasts without higher risk the
  impossible
• dream?
• Provide duration latitude to the active managers
  when
• they choose to be defensive

Kalson Associates
31
Inherent Limitations

• The five-year estimates of return and risk are
  just that
• estimates
• KA adjusted some of the forecasted data for
  better fit
• An unanticipated need for the bulk of the
  portfolio due to
• an infrastructure or weather-related crisis
  could result in
• market losses

Kalson Associates
32
Questions