Rhode Island SIC August 1, - PowerPoint Presentation

1. Rhode Island SIC August 1, 2016Allan Emkin | John Burns, CFA PCA 2016 Capital Market Assumptions

2. Today’s discussionCapital market assumptions BackgroundDeveloping Capital market returns Summary of proposed 2016 Capital Market Assumptions Modeling ConstraintsModel Return Resampling ProcedureAppendixSetting capital market risk premia – detailSurvey of 2016 Capital Market return assumptions Agenda

3. Review PCA / ERSRI capital market assumptions How were return and risk assumptions derived? SIC approve model input dataReview model asset class constraints (mix / max allocations)SIC approve model asset class constraintsReview model Return Resampling process Today’s Discussion

4. 4Capital Market Assumptions Capital Market return expectations have declined in the past decade Primarily due to the decline in interest rates Riskier assets’ expected returns are also lower Their expected returns are based off a risk premium over interest ratesLower inflation expectations also dampen expected total returns for most classesLong-term investors face difficult choices as they invest in a diversified portfolio to meet their long-term return expectations

5. Objective: Establish consensus expectations for how ERSRI’s investment classes and their markets are expected to perform in the future2016 PCA Capital Markets Assumptions (CMAs) are the basis for developing ERSRI model inputsTo the extent an asset class is structured in a traditional fashion PCA CMAs will be used To the extent the portfolio utilizes risk or functional classes – the CMAs will be customizedThe capital markets expectation-setting exercise is subjective/qualitative and may contain significant forecast errorIn forming expectations, PCA reviews a broad range of economic, fundamental, and investment industry dataPCA Capital Market Assumptions

6. Assumes 10 year investment horizon No premium assumed for active managementExpected returns are net of feesStandard Classes: no customization required U.S. Equity Portfolio Non-U.S. Portfolio Core Fixed Income Portfolio Private Equity Portfolio Inputs for functional classes are customized – starting point is PCA standard assumptionsOther: New Concept Classes Yield-Driven Portfolio – mix of assetsRisk mitigation Class – customized role Other ?PCA Capital Market Assumptions:

7. Begin with the role of the functional classIdentify assets or strategies that fill the role and their respective weighting in the functional class Create capital market assumptions for each assets or strategy, including correlations Combine the individual classes to create one capital market assumption for the Class Understanding the role of the asset is key Fewer model inputs (asset classes or functional classes) produces clearer output – focused on what’s importantWhich risks are assumed, and Impact on plan financial health PCA Capital Market Assumptions: Modeling Functional Classes

8. Capital market assumptions are based on capital market theory and investment experienceGreater risk, greater expected return – consistent with historical long-term capital market returns Illiquid and infrequently priced assets have risk even if it’s not perceived in monthly or quarterly return volatility We assign increased incremental risk commensurate with the underlying risk of the investmentAdjustments are made for leverage employed Real Estate portfolio PCA capital market assumptions are similar to those of other investment practitionersLong investment horizon smooths out differences among market practitionersPeer comparisons examined to identify potential outliersBackground on PCA Capital Market Assumptions

9. Summary of 2016 capital market expected returns Expect interest rates to rise from historically low levelsExpected returns are relatively low due to starting point Fixed income: interest rates are very low U.S. Equity securities: are trading at high Price / Earnings ratios Non-U.S. Equity: are trading at more modest valuations Low income generated by most capital market assets PCA and GRS have different inflation assumptions, driven mostly by differences in time horizonPCA 10-year assumption = 2.25%GRS 30-year assumption = 2.75% Two options for the inflation assumption during the modeling process (CMA’s will be adjusted to reflect the option chosen)Use either the PCA or GRS assumption, ORUse PCA’s assumption for the next 10-years and move to GRS’s assumption for the remainder of the time horizonBackground on PCA Capital Market Assumptions

10. Basic Building Block framework for estimating arithmetic returns(Inflation) + (Real Risk Free Rate of Cash) + (Premium over Real Risk Free Rate)All risky-asset class return expectations are built as risk premium over cash Expectation for cash returns are 2.0%, 75 basis points below inflation assumption Expected compound return estimates are the result of first estimating arithmetic average asset class returns and volatilities, which are then converted to geometric return estimates.Expected Arithmetic return: the return expected in any one yearExpected Geometric return: the return expected over a multi-year holding periodExpected Geometric return is also less than the expected Arithmetic return due to incurring investment risk (return volatility penalty) PCA Asset Class Return Assumptions

11. Inflation RatePCA uses both market-based fundamental measures and other sources of inflation expectations to determine an expected long-term rate of inflationFor the purposes of the asset liability study PCA uses the system’s actuary’s expected inflation assumption, 2.75%  Real Risk Free Return The real risk-free rate can take two forms: (i) a short-term rate of return based on default-free government debt and (ii) a real rate of return or real yield on a default-free, zero-coupon bond whose duration closely matches the horizon of an investor’s cash flow requirements. In addition we have examined the trends of their annual time series utilizing exponential smoothing techniques. Premium Over Risk Free RateInvestment class’s return associated with various risks above and beyond the risk-free return Often the largest and most volatile component of expected return, hence the most difficult to forecast.Building Block Components

12. Asset classes with long histories: U.S. Equity and Fixed IncomeBegin with asset classes’ historical volatilities. Adjust for trend in rolling 5 year standard deviations ending with 2011 - 2015. Asset classes with shorter histories Fewer rolling 5 year periods available to discern trend.Compute historical standard deviations, weighting the most recent periods heavier than prior decades, combine these estimates with shorter trendsReturn Volatility

13. PCA’s standard Capital Market Assumptions include correlation data for all pairs of major investment classes. This correlation data would be used in a standard Mean Variance Optimization Model.However, since the asset liability model is using a return sampling technique, correlations are an output of the process and not a model input After simulations are run (>1,000 scenarios), return correlations (output) are reviewed for robustness and reasonableness  The Model accounts for variations in risks and correlations within an investment horizon Historically, asset class correlation relationships have not been constant.Incorporating correlation variation in the modeling process is a key advantage of the resampling return technique vs. a Mean Variance Optimization model Correlations

14. PCA Capital Market Assumptions: Arithmetic Returns  Arithmetic Returns %2004200520062007200820092010201120122013201420152016Change since 2008Inflation2.502.502.502.503.003.003.002.753.003.002.502.002.25-0.75Cash3.753.754.004.504.004.003.503.003.002.252.003.152.00-2.00Core Bonds4.504.654.755.255.505.254.003.302.903.103.102.753.00-2.50Real Estate*7.507.006.757.007.006.757.257.007.006.406.256.005.50-1.50U.S. Stocks8.759.009.009.009.2510.009.258.759.358.758.508.508.50-0.75Non-U.S. Stocks8.759.009.009.009.2510.009.759.009.859.259.009.259.500.25Private Equity12.7513.0013.0012.5011.7512.5012.5012.0012.6012.0011.7511.8512.100.35PCA capital market assumptions have declined since 2008 primarily due to the decline in inflation expectations and interest rates Real Estate assumes no leverageBased on PCA inflation assumptions

15. Preliminary Capital Market Assumptions and Constraints Preliminary modeling constraints and capital market assumptionsCore Real Estate assumes leverageAssumptions use GRS Inflation of 2.75%

16. Constraints Influenced by:Client’s statutory investment requirementsBoard preferences (e.g., home bias, risk allocation intentions, fiduciary comfort with an investment, etc.) Liquidity and incomeMinimum required total fund rate of returnMaximum risk tolerance Modeling Constraints – Primary Factors Influencing Constraints

17. Incorporate any statutory restrictions governing the portfolio management into the modelFocus analysis on investable / realistic outcomes Extreme allocations are not going to be adopted However, constraints can’t be too restrictive Leads to predetermined outcome thus reducing the value of the analysis Set investment constraints to be consistent with prudent investment pacing for illiquid assets: Private Equity or Real EstatePrudent investment in Private Equity requires that allocations be diversified over time (vintage year)Analysis will also be run unconstrained using Mean Variance Optimization Highlights the capital market assumptions return and risk profile Modeling Constraints

18. Return sample data starts 1970 (46 years of annual data)Returns are modeled market returnsReturn history is modified to reflect custom ERSRI Capital Market AssumptionsMore realistic distribution of returns than mean variance optimization – which assumes normal distribution Return data preserves historical behavioral patterns of each strategic classCorrelation patterns vary, particular during crisis periods Non-normal distributions, better reflecting reality, particularly over shorter investment periodsEach simulation randomly selects a series of annual return data (returns for all asset classes for that one year)Each portfolio weighing is evaluated and ranked across each 30 year return simulation Model Return Resampling Procedure

19. Model Return Series Sample

20. Appendix

21. Like many market participants, PCA utilizes a “Building Block” approach to capital market assumptions Combines both fundamental and historical information and dataDeveloping expectations for risks relies more heavily on analysis of historical dataExamine the trends of historical data across investment classesIn addition, PCA may use statistical procedures to adjust data and/or emphasize more recent data rather than utilizing simple computational techniques that treat all asset class history as equivalent.Where does the assumption construction process start?

22. The equity risk premium is an important input variable Typically the largest portfolio allocation – will influence the total fund expected returnBuilding block for other asset class returns We compare estimates derived from analysis of historical risk premiums and risk premium trend extrapolations to risk premium estimates derived via fundamental models. Our first step is to compute estimates of long-term equity risk premium utilizing a basic dividend discount model:RPe = D/P + g - Rf ± [impact due to valuation changes]where:RPe is the estimated equity risk premiumD/P is the current dividend yieldg is the long-term real dividend growth rate, andRf is the real risk-free rate.Setting Risk Premium: U.S. Equity

23. For strategic asset allocation purposes, PCA believes that it is difficult to predict whether one large developed public equity capital market (multi-trillion dollar market with thousands of publicly-held companies) will outperform another over an extended investment horizon. We assume that the rate of economic growth in emerging markets will continue to outpace that of developed markets for the foreseeable future. Since Emerging Markets are included with non-U.S. equity they collectively have a higher return than U.S. equity.The valuation level of developed markets outside of the U.S. is currently considerably lower than that of the U.S. markets. Therefore, the equity risk premium for non-U.S. equities is set to be 1.0% higher than the U.S. equity premium, and the global equity risk premium is set to be 0.60% higher.Setting Risk Premium: Non-U.S. Equity

24. PCA applies the same general approach for estimating the expected fixed income risk premium return as that applied in establishing equity risk premium returns: Examine trends of the historical fixed income risk premium and Assess market-based fundamentals. Within fixed income, cash flows and cash flow growth are less uncertain than in the equity markets and long-term appreciation of underlying principal does not occur under equilibrium conditions.As a result, current yields-to-maturity across the fixed income spectrum provide key baselines from which to begin projecting long-term return expectations. From this point, analyses of risk premium trends and the current interest rate environment are then used to adjust the yield-to-maturity to arrive at a final estimate for the Fixed Income risk premium return.In this current investment cycle we believe a major factor influencing the Core Fixed Income return will be rising interest rates in futureThe recent past saw big return contribution from price appreciation relative to income for Fixed Income assets We assume that over the next 10 years that income will be a bigger contribution to return as appreciation turns marginally negative In addition we assume credit will account for the majority of Fixed Income real returns Setting Risk Premium: Fixed Income

25. Neither of these asset classes lend themselves well to statistical procedures utilized by the Capital Asset Pricing Model. A key reason for this problem is that these asset classes are not marked-to-market on a near-continuous basis as is the case with the other publicly-traded asset classes. As a result, more reliance on qualitative and fundamentals-based procedures is necessary for developing return and risk expectations for these classes. As with the other asset classes, PCA examines the trends in each of these asset classes’ risk premium returns. Setting Risk Premium: Real Estate and Private Equity

26. The trend of the real estate risk premium return has been to exhibit highly cyclical characteristics, largely attributable to the trending behavior associated with real estate appraisals and capital discount rates that fluctuate only modestly over time compared to other market-based rates.For the core real estate asset class PCA models its risk premium return as falling between the risk premium returns of stocks and bonds. This approach reflects the common acceptance that real estate is a hybrid asset class offering both potentially high levels of current income (greater than fixed income), while also providing for potential long-term capital appreciation via income growth. One other attractive aspect of real estate is that since leases on commercial real estate are typically re-negotiated over time, lease cash flows should grow along with inflation.Given the recent strong Real Estate market returns – the Real Estate risk premium has been reduced 50 bp from 2015 assumptions Setting Risk Premium: Real Estate

27. The excess returns expected from private equity typically range from 3.0% to 5.0% annually over public equity counterparts. This premium is often associated with an “illiquidity premium” required by investors. As with the real estate asset class, PCA begins by assigning a “default position” for the private equity illiquidity premium. PCA then adjusts this illiquidity premium based on its current trend and any key fundamental factors impacting the asset class. The private equity illiquidity premium has varied cyclically over the last 40+ yearsSetting Risk Premium: Private Equity

28. To begin our risk projection process, we first review asset classes’ historical volatilities. For the 89 years ending 2014 (beginning with 1926), the standard deviation of annual returns for U.S. Equity and U.S. Core Fixed Income asset classes were 20.3% and 5.7%, respectively. We then compute standard deviations for each discrete 5-year period ending with 2010-2014. Using statistical procedures, we then map out the trend of those discrete observations. Examples:While the secular trend for U.S. Equity risk appears to be downward, U.S. Core Fixed Income risk appears to be cyclical.We expect the long-term secular trend of lower equity risk to continue as the volatility of global output continues to moderate.We believe the volatility of U.S. core fixed income has likely bottomed and could rise in response to a more volatile (less predictable) level of inflation, should that occur. Return Volatility: U.S. Equity and Fixed Income

29. Several asset class benchmarks have 40 years or less of history (e.g., international equities, non-U.S. and global bonds, real return, private real estate, and private equity). As a result, the number of 5-year risk data points is too few to perform any meaningful statistical analysis.In these cases, PCA computes historical standard deviations, weighting the most recent periods heavier than prior decades, and combines these estimates with visual inspection of shorter trends to develop future expectations for risk of the strategic class. These procedures are applied to all other asset classes lacking ample history for further statistical trend analysis.Return Volatility: Other Asset Classes

30. 302016 10-Year Capital Market Assumptions: OverviewPCA’s capital market assumptions are in-line with peers and generally near the median US EquityFixed IncomeReal EstatePrivate EquityFirmCompound ReturnArith. ReturnStDevFirmCompound ReturnArith. ReturnStDevFirmCompound ReturnArith. ReturnStDevFirmCompound ReturnArith. ReturnStDevBNY7.20%8.40%16.10%JPM3.72%3.80%4.00%PNC7.14%8.65%18.05%PNC9.63%13.00%27.40%Callan7.18%8.80%18.70%Wilshire3.48%3.60%5.00%NEPC6.50%7.55%15.00%BNY9.14%10.80%19.10%JPM6.98%8.10%15.50%Callan3.03%3.10%3.80%Callan6.02%7.30%16.50%PCA9.04%12.10%26.00%PCA6.91%8.50%18.50%PCA2.92%3.00%4.00%BlackRock5.61%7.00%17.20%Wilshire8.98%12.40%27.50%Wilshire6.50%7.85%17.00%BlackRock2.91%3.00%4.30%JPM5.47%6.10%11.50%JPM8.53%10.70%21.80%PNC6.40%7.50%15.35%VOYA 2.60%2.85%7.10%PCA5.12%5.50%9.00%NEPC8.49%10.90%23.00%BlackRock6.33%7.50%15.80%BNY2.54%2.60%3.50%Wilshire5.03%6.40%17.00%Callan8.24%13.10%32.80%NEPC5.96%7.40%17.50%Sellwood2.48%2.60%5.00%Sellwood4.73%6.40%18.75%Sellwood8.10%12.60%31.50%VOYA 5.32%6.70%17.10%NEPC2.42%2.60%6.03%BNY4.07%4.20%5.30%BlackRock8.01%11.00%25.60%Sellwood5.20%6.90%19.00%PNC2.41%2.50%4.35%VOYA 3.96%6.10%21.20%Data is sorted by expected compound return from highest to lowest