Benefits, Compensation and HR Consulting

Transcription

1 Benefits, Compensation and HR Consulting University of California Retirement Plan ACTUARIAL EXPERIENCE STUDY Analysis of Actuarial Experience During the Period July 1, 2006 through June 30, 2010 Copyright June 2011 THE SEGAL GROUP, INC., THE PARENT OF THE SEGAL COMPANY ALL RIGHTS RESERVED

2 THE SEGAL COMPANY 100 Montgomery Street, Suite 500 San Francisco, CA T F June 13, 2011 Mr. Dwaine B. Duckett University of California Vice President, Human Resources 1111 Franklin Street, 5 th Floor Oakland, California Re: Review of Actuarial Assumptions for the July 1, 2011 Actuarial Valuation Dear Vice President Duckett: We are pleased to submit this report of our review of the actuarial experience of the University of California Retirement Plan (UCRP or Plan). This study utilizes the census data of the last five actuarial valuations to analyze the demographic experience for the four-year period ending on June 30, It includes the proposed actuarial assumptions, both demographic and economic, to be used in future actuarial valuations starting with the July 1, 2011 actuarial valuation. We are members of the American Academy of Actuaries and we meet the Qualification Standards of the American Academy of Actuaries to render the actuarial opinion herein. We look forward to reviewing this report with you and answering any questions you may have. Sincerely, Paul Angelo, FSA, MAAA, FCA Senior Vice President and Actuary John Monroe, ASA, EA, MAAA Vice President and Associate Actuary ST/hy v4/ Benefits, Compensation and HR Consulting ATLANTA BOSTON CALGARY CHICAGO CLEVELAND DENVER HARTFORD HOUSTON LOS ANGELES MINNEAPOLIS NEW ORLEANS NEW YORK PHILADELPHIA PHOENIX PRINCETON RALEIGH SAN FRANCISCO TORONTO WASHINGTON, D.C. Multinational Group of Actuaries and Consultants MEXICO CITY OSLO PARIS BARCELONA BRUSSELS DUBLIN GENEVA HAMBURG JOHANNESBURG LONDON MELBOURNE

3 TABLE OF CONTENTS Page I. INTRODUCTION, SUMMARY, AND RECOMMENDATIONS...1 II. BACKGROUND AND METHODOLOGY...6 III. ECONOMIC ASSUMPTIONS...8 A. INFLATION...8 B. INVESTMENT RETURN...10 C. SALARY INCREASE...22 IV. DEMOGRAPHIC ASSUMPTIONS...29 A. RETIREMENT RATES...29 B. MORTALITY RATES NON-DISABLED (HEALTHY)...41 C. MORTALITY RATES - DISABLED...45 D. TERMINATION RATES...49 E. DISABILITY INCIDENCE RATES...58 F. ELIGIBLE SURVIVOR ASSUMPTIONS...64 G. SERVICE FROM UNUSED SICK LEAVE CONVERSION...68 H. LUMP SUM CASHOUT TAKE-RATE...71 I. FUTURE BENEFIT ACCRUALS...79 J. ADMINISTRATIVE EXPENSES...80 V. COST IMPACT OF ASSUMPTION CHANGES...81 APPENDIX A CURRENT ACTUARIAL ASSUMPTIONS...83 APPENDIX B PROPOSED ACTUARIAL ASSUMPTIONS...89

4 I. INTRODUCTION, SUMMARY, AND RECOMMENDATIONS To project the cost and liabilities of the Plan, assumptions are made about all future events that could affect the amount and timing of the benefits to be paid and the assets to be accumulated. Each year actual experience is compared against the projected experience, and to the extent there are differences, the total funding policy contribution rate is adjusted. If assumptions are modified, the total funding policy contribution rate is adjusted to take into account a change in the projected experience in all future years. There is a great difference in both philosophy and cost impact between recognizing the actuarial deviations as they occur annually and changing the actuarial assumptions. Taking into account one year s gains or losses without making a change in the assumptions means that that year s experience was temporary and that, over the long run, experience will return to what was originally assumed. Changing assumptions reflects a basic change in thinking about the future, and it has a much greater effect on funded status than recognizing gains or losses as they occur. The use of realistic actuarial assumptions is important in maintaining adequate funding, while paying the promised benefit amounts to members already retired and to those near retirement. These assumptions will also be utilized in estimating future costs and projecting the funded status of the Plan. Therefore, matching the assumptions as closely as possible to expected Plan experience will best inform planning for setting contribution amounts to the Plan. The actuarial assumptions used do not determine the actual cost of the Plan. The actual cost is determined solely by the benefits and administrative expenses paid out, offset by investment income received. However, it is desirable to estimate as closely as possible what the actual cost will be so as to permit an orderly method for setting aside contributions today to provide benefits in the future, and to maintain equity among generations of members and funding sources. This study was undertaken in order to review the economic and demographic actuarial assumptions and to compare the actual experience with that expected under the current assumptions during the four-year experience period from July 1, 2006 through June 30, The study was performed in accordance with Actuarial Standard of Practice (ASOP) No. 27, Selection of Economic Assumptions for Measuring Pension Obligations and ASOP No. 35, Selection of Demographic and Other Noneconomic Assumptions for Measuring Pension Obligations. These Standards of Practice put forth guidelines for the selection of the various actuarial assumptions utilized in a pension plan actuarial valuation. Based on the study s results and expected near-term experience, we are recommending various changes in the current actuarial assumptions. -1-

5 Many of the proposed demographic assumptions were a joint effort between Segal and Deloitte Consulting (consulting actuary for the retiree health plan). To the extent possible, we incorporated both retiree health and pension considerations into the proposed assumptions. We are recommending changes in the assumptions for real ( across the board ) salary increases, promotional and merit salary increases, retirement from active membership, non-disabled (healthy) life mortality, disabled life mortality, termination, disability incidence, service from unused sick leave conversion and Lump Sum Cashout take-rate. In some cases we have changed the structure of the assumption. For example, we observe that termination rates correlate better with years of service than with age. Therefore, the proposed assumptions for termination rates are by years of service instead of by age, which was the current practice. We also found that retirement rates for Staff members also correlate well with years of service in addition to age and so our proposed assumption is based on both variables. Our recommendations for the actuarial assumption categories for the University of California Retirement Plan (UCRP or Plan) are as follows: Inflation Future increases in the Consumer Price Index (CPI) which drive investment returns and active member salary increases, as well as cost-of-living adjustments (COLAs) for retired members. Recommendation: Maintain the rate at 3.50% per annum as discussed in Section III(A). Investment Return The estimated average future net rate of return on assets over the projected lifetime of the Plan as of the valuation date. This rate is used to discount liabilities. Recommendation: Maintain the rate at 7.50% per annum as discussed in Section III(B). Individual Salary Increases Increases in the salary of a member between the date of the valuation and the date of separation from active service. This assumption has three components: Inflationary salary increases, Real across the board salary increases, and Promotional and merit increases. Recommendation: Maintain the current inflationary salary increase at 3.50% and increase the real across the board salary increase from 0.25% to 0.50%, as discussed in Section III(C). This means that the combined inflationary and real across the board salary increases will increase from 3.75% to 4.00%.. In addition, reduce the promotional and merit increases to those developed in Section III(C). -2-

6 Retirement Rates The probability of retirement at each age at which members are eligible to retire. Recommendation: For active members, adjust the current retirement rates to those developed in Section IV(A). For active Staff members, the rates will also depend on years of service in addition to age. For deferred vested members, maintain the assumed retirement age of 59. Mortality Rates The probability of dying at each age. Mortality rates are used to project life expectancies. Recommendation: For non-disabled (healthy) pensioners, decrease the mortality rates by using the RP-2000 Combined Healthy Mortality Tables projected to 2025 with a two-year set back as developed in Section IV(B). For disabled pensioners, decrease the mortality rates by using the RP Disabled Retiree Mortality Tables projected to 2025 with a two-year set back for males and no set back for females as developed in Section IV(C). For pre-retirement mortality, use the same mortality tables that are used for healthy pensioners. Termination Rates The probability of leaving active membership at each age and receiving either a refund of member contributions ( accumulations ) or a deferred vested retirement benefit. Recommendation: Implement service-based termination rates and increase the current termination rates overall to those developed in Section IV(D). In addition, implement a new assumption that a member will choose between a refund of contributions and a deferred vested benefit based on which option has the greater present value at termination. Disability Incidence Rates The probability of becoming disabled at each age. Recommendation: Decrease the current disability rates overall to those developed in Section IV(E). Eligible Survivor Assumptions The probability of having a survivor at death. Recommendation: Maintain the current percentages for Eligible Survivors, as described in Section IV(F). Also, maintain the current assumption for number of Eligible Survivors per Active Member with Eligible Survivors. Service from Unused Sick Leave Conversion Increases in Service Credit due to conversion of unused sick leave. Recommendation: Slightly decrease the current assumption for Faculty and Safety members retiring from active membership and slightly increase the current assumption for Staff members retiring from active membership, as developed in Section IV(G). -3-

7 Lump Sum Cashout Take-Rate The probability of electing a Lump Sum Cashout at retirement. Recommendation: For active members, implement service-based take-rates and adjust the assumption as described in Section IV(H). For deferred vested members, maintain the assumption of 45% of members electing a Lump Sum Cashout. For retiree crossovers from disability status, increase the assumption from 12% to 13% of members electing a Lump Sum Cashout as described in Section IV(H). Future Benefit Accruals Amount of Service Credit projected to be earned by active members in years after the valuation date. Recommendation: No change to assuming that all active members earn one year of Service Credit each year in the future, as discussed in Section IV(I). Administrative Expenses Fees for administrative, legal, accounting, and actuarial services, and other functions carried out by the Plan. Recommendation: No change to the percentage loading to the normal cost of 0.50% of payroll, as developed in Section IV(J). We also suggest that for any assumption changes being recommended for UCRP, the assumption will also be changed for the University of California 415(m) Restoration Plan and the PERS Plus 5 Plan actuarial valuations, as applicable. The main exception to this is for the administrative expense assumptions where these two Plans will continue to have their own distinct assumptions. Section II provides some background on basic principles and the methodology used for the experience study and for the review of economic and demographic actuarial assumptions. A detailed discussion of each assumption and reasons for the proposed changes is found in Section III for the economic assumptions and Section IV for the demographic assumptions. Section V shows the cost impact of the proposed changes. Note that if these assumptions are adopted by the Board of Regents, then the annuity option factors and Lump Sum Cashout factors should be reviewed for consistency with the assumptions proposed in this report. Based on our proposed change to the mortality assumptions, we have also included a proposed mortality table to use in the determination of the annuity option factors and Lump Sum Cashout factors. We would propose that a July 1, 2012 effective date be used for the changes to the annuity option factors and Lump Sum Cashout factors. -4-

8 We have estimated the impact of the proposed assumption changes (including the effect of the proposed changes to the assumptions used for annuity option factors and Lump Sum Cashout factors) as if they were applied to the July 1, 2010 actuarial valuation. If all of the proposed assumption changes were implemented, the Plan s Normal Cost as a dollar amount would have increased by $14 million (1.0%) and the Actuarial Accrued Liability (AAL) would have increased by $1.77 billion (3.7%). The total funding policy contribution would have increased from 23.25% to 25.50% of payroll. The change to the mortality table was significant as it increased the Normal Cost by 0.5% of payroll and the AAL by $1.6 billion. Changes to the assumptions for termination rates and disability incidence offset most of this increase in Normal Cost, but only slightly offset this increase in AAL. We stress that this is an illustration based on applying the proposed assumptions to the previous valuation (2010). -5-

9 II. BACKGROUND AND METHODOLOGY In this report, we analyzed both economic and demographic ( non-economic ) assumptions. The primary economic assumptions reviewed are inflation, investment return, and salary increases. Demographic assumptions include the probabilities of certain events occurring in the population of members, referred to as decrements, e.g., termination from service, incidence of disability, service retirement, and death after retirement. Economic Assumptions Economic assumptions consist of: Inflation Increases in the price of goods and services. The inflation assumption reflects the basic return that investors expect from securities markets. It also reflects the expected basic salary increase for active members and drives increases in the allowances of retired members. Investment Return Expected return on the Plan s investments after expenses. This assumption has a significant impact on the total funding policy contribution rate. Salary Increases In addition to inflationary increases, it is assumed that salaries will also grow by real across the board pay increases in excess of price inflation. It is also assumed that active members will receive raises from promotions and step increases. These are sometimes referred to as promotional and merit increases. The setting of these economic assumptions is described in Section III. Demographic Assumptions In order to determine the probability of an event occurring, we examine the decrements and exposures of that event. For example, taking termination from service, we compare the number of active members who actually terminate in a certain age and/or service category (i.e., the number of decrements ) with those who could have terminated (i.e., the number of exposures ). For example, if there were 500 active members in the age group at the beginning of the year and 50 of them terminate during the year, we would say the probability of termination in that age group was or 10%. -6-

10 The reliability of the resulting probability is highly dependent on both the number of decrements and the number of exposures. For example, if there are only a few people in a high age category at the beginning of the year (number of exposures), we would not lend as much credence to the probability of termination developed for that age category, especially if it is out of line with the pattern shown for the other age groups. Similarly, if we are considering the death decrement, there may be a large number of exposures in, say, the age category, but very few decrements (actual deaths); therefore, we would not be able to rely heavily on the probability developed for that category. One reason we use several years of experience for such a study is to have more exposures and decrements, and therefore more statistical reliability. Another reason for using several years of data is to smooth out fluctuations that may occur from one year to the next. However, we also calculate the rates on a year-to-year basis to check for any trend that may be developing in the later years. Please note that any active member experience during the four-year period for members who worked at the Los Alamos National Laboratory (LANL) or the Lawrence Livermore National Laboratory (LLNL) was excluded from the determination of any prospective assumptions that affect active members. This is because there are currently no active members in UCRP working at LANL or LLNL. We believe that in general it would not be appropriate to include the experience for a subset of the Plan s active members in developing assumptions for future events when that subset of members is no longer active in the Plan. We also believe we should not let these extraordinary one-time events impact the proposed assumptions. Similarly, the temporary Furlough/Salary Reduction Plan approved by The Regents in July 2009 also occurred during the four-year experience study period. The salary and service credit data upon which this experience study is based excludes the temporary Furlough/Salary Reduction Plan. This is because of the temporary nature of the Furlough/Salary Reduction Plan and also due to the amendment to UCRP adopted by The Regents to ensure that the Furlough/Salary Reduction Plan has no impact on the calculation of member benefits. -7-

11 III. ECONOMIC ASSUMPTIONS A. INFLATION Unless an investment grows at least as fast as prices increase, investors will experience a reduction in the inflation-adjusted value of their investment. There may be times when riskless investments return more or less than inflation, but over the long term, investment market forces will require an issuer of fixed income securities to maintain a minimum return which protects investors from inflation. The inflation assumption is long term in nature, so it is set using primarily historical information. Following is an analysis of 15-year and 30-year moving averages of historical inflation rates: Historical Consumer Price Index 1930 to 2010 (U.S. City Average - All Urban Consumers) 25 th Percentile Median 75 th Percentile 15-year moving averages 2.7% 3.5% 4.8% 30-year moving averages 3.3% 4.2% 5.0% The average inflation rates have continued to decline gradually over the last several years due to the relatively low inflationary period in the 1990s and early 2000s. Also, the later of the 15-year averages during the period are lower as they do not include the high inflation years of the mid-1970s and early 1980s. The current inflation assumption of 3.50% is comparable to most retirement systems, not only in California, but nationally. Here are some relevant comparisons: System Inflation Assumption Los Angeles City Employees Retirement System 3.75% Los Angeles City Fire & Police Pensions 3.50% Los Angeles County Employees Retirement Association 3.50% CalPERS 3.00% Median from NASRA 2010 Public Fund Survey 3.50% -8-

12 Regarding the last entry, in a 2010 public fund survey published by the National Association of State Retirement Administrators (NASRA), the median inflation assumption used by 125 large public retirement funds in their 2009 actuarial valuations was 3.50%. UCRP s investment consultant, Mercer Investment Consulting, anticipates an annual inflation rate of 2.8%. Note that in general, the investment consultants time horizon for this assumption is shorter than the time horizon used in the actuarial valuation. Based on all of the above information, we recommend that the current 3.50% annual inflation assumption be maintained at 3.50% for the July 1, 2011 actuarial valuation. Note that the UCRP Cost-of-Living Adjustment (COLA) provision for annuitants generally provides for 100% of Consumer Price Index (CPI) increases up to 2% per year plus 75% of CPI increases above 4% per year. This means that based on the 3.50% inflation assumption being recommended we will continue to value this COLA provision as a flat 2% per year. -9-

13 B. INVESTMENT RETURN Investment return is an important component in the pension funding equation: contributions plus investment return equals benefits plus expenses. The investment return assumption is intended to reflect the long-term return that will be achieved on the Plan s assets in future years. The use of a higher investment return assumption increases the risk that the Plan will not achieve its assumed return over the long run, causing a future shortfall of Plan assets and an increase in total funding policy contributions. Conversely, a lower investment return assumption increases the chance that the Plan will exceed its assumed return over the long run, leading to more than expected Plan assets and a decrease in total funding policy contributions. Since no amount of analysis can predict future returns with certainty, setting the investment return assumption generally involves considering an acceptable range of expected returns and then selecting a specific point within that range consistent with the Plan s tolerance of the risks described just above. Historical Returns For reference, UCRP actual rates of return from July 1, 1990 through June 30, 2010, on a market and actuarial basis, are shown in Chart 1, along with the assumed earnings rates during that same period. Please note that while historical Plan performance is one data element that may be reviewed, caution should be exercised to avoid relying on that data too heavily. The relevant Actuarial Standard of Practice (ASOP No. 27, Section 3.3) states with regard to selecting any of the economic assumptions: [T]he actuary should consider recent economic data. However, the actuary should not give undue weight to recent experience. For example, if the recent history was largely attributable to a significant change in bond yields or inflation, it may be unreasonable to assume that such investment returns will continue over the measurement period. This Standard is particularly relevant when setting the investment return assumption because UCRP s investment experience over the past twenty years has been heavily influenced by extraordinary investment market events, including periods of unprecedented market gains and losses. For that reason, our investment return assumption is not explicitly based on the actual return history of UCRP. -10-

14 30% 25% 20% 15% 10% 5% 0% -5% -10% -15% -20% 9.8% '90- '91 Chart 1 Historical UCRP Investment Rates of Return 26.1% 21.3% 16.8% 15.5% 25.8% 21.6% 12.3% 12.7% 14.5% 10.3% 5.6% -2.6% -5.5% Market Value of Assets (MVA) Assumption (7.5% starting 1994) -9.0% Actuarial Value of Assets (AVA) '91- '92 '92- '93 '93- '94 '94- '95 '95- '96 '96- '97 '97- '98 '98- '99 '99- '00 '00- '01 '01- '02 '02- '03 '03- '04 '04- '05 '05- ' % '06- ' % 13.3% -5.6% -19.2% '07- '08 '08- '09 '09- '10

15 However, these returns do provide information about the historical practice of The Regents regarding its selection of a specific investment return assumption. The Regents adopted the 7.50% earnings assumption in 1994 and maintained that assumption during the 1990s when market returns were very high and many funds were increasing their earnings assumptions. It has also been maintained throughout the 2000s during which several years of market losses were experienced. Comparison with Other Public Retirement Systems One general test of the current investment return assumption is to compare it against those used by other public retirement systems, both in California and nationwide. The following table compares the current UCRP net investment return assumption against a sample of large California state, county and city public retirement systems: Assumption Current UCRP Sample California Public Retirement Systems* Low Median High Net Investment Return 7.50% 7.50% 7.75% 8.00% * Includes CalPERS, UCRP, 13 county systems and 7 major city/municipal systems The next table compares the current UCRP net investment return assumption against those of 126 nationwide public retirement systems that participated in the National Association of State Retirement Administrators (NASRA) 2010 Public Fund Survey: Assumption Current UCRP NASRA 2010 Public Fund Survey Low* Median High Net Investment Return 7.50% 7.25% 8.00% 8.50% * After eliminating the very lowest as an outlier Based on the above, the current investment return assumption of 7.50% falls toward the lower end of a reasonable range when compared to other retirement systems. This provides relatively greater protection against the risk of future asset shortfalls and future increases in the total funding policy contributions. This is consistent with The Regents historical practice, noted above. -12-

16 A Risk Adjustment Methodology In addition to the historical perspective and comparisons against other public retirement systems, Segal s standard approach for its California public sector plans is a more quantitative analysis that starts with the very common building block method (as described in ASOP No. 27, Section a) which looks at the components of the investment return assumption: inflation, real rate of return and expenses. It then includes a Risk Adjustment, which is an empirically based method of measuring and comparing risk tolerances among different plans. We will develop each of these building block components in turn. Inflation As previously discussed in Section III(A), we are recommending a 3.50% inflation assumption. Real Rate of Investment Return This component represents the portfolio s incremental investment market returns over inflation. Theory has it that as an investor takes a greater investment risk, the return on the investment is also expected to be greater, at least in the long run. This additional return is expected to vary by asset class and empirical data supports that expectation. For that reason, the real rate of return assumptions are developed by asset class. Therefore, the real rate of return assumption for the UCRP portfolio will generally vary with The Regents asset allocation among asset classes. The following table shows the UCRP target asset allocation and a set of expected real rates of return assumptions by asset class. The column of returns represents the average of a sample of real rate of return expectations. The sample includes the expected annual real rates of return provided to us by ten investment advisory firms retained by California public retirement systems. We believe these assumptions reasonably reflect a consensus forecast of long-term market returns. -13-

17 Asset Class UCRP Target Asset Allocation and Assumed Arithmetic Real Rate of Return Assumptions by Asset Class and for the Portfolio Percentage of Portfolio Average Real Rate of Return from a Sample of Consultants to Segal s California Public Clients* US Equity 23.0% 6.75% Developed International Equity ** 24.0% 6.89% Emerging Market Equity 5.0% 9.29% Core Fixed Income 12.0% 1.45% High Yield Bonds 2.5% 3.74% Emerging Market Debt 2.5% 4.02% TIPS 8.0% 1.26% Real Estate 7.0% 5.41% Private Equity 6.0% 10.42% Absolute Return/Hedge Funds/Real Assets 10.0% 4.14% Total 100.0% 5.56% * These are based on the projected arithmetic returns provided by the investment advisory firms serving the county retirement systems of Alameda, Contra Costa, Fresno, Orange, Sacramento, San Bernardino, San Diego and the LA City Employees Retirement System, City of Fresno Retirement Systems and the LA Fire & Police Pensions. These return assumptions are gross of any applicable investment expenses. ** Includes a 2% allocation to Global Equity Please note that the above are representative of indexed returns and do not include any additional returns ( alpha ) from active management. This is consistent with the Actuarial Standard of Practice No. 27, Section e, which states: Investment Manager Performance - Anticipating superior (or inferior) investment manager performance may be unduly optimistic (or pessimistic). Few investment managers consistently achieve significant above-market returns net of expenses over long periods. The following are some observations about the returns provided above: -14-

18 1. The investment consultants to our California public sector clients have each provided us with their expected real rates of return for each asset class, over various future periods of time. However, in general, the returns available from investment consultants are projected over time periods shorter than the duration of a retirement plan s liabilities. 2. Using an average of expected real rates of return allows the Plan s investment return assumption to include a broader range of capital market information and it should help reduce year to year volatility in the Plan s investment return assumption. 3. Therefore, we recommend that the 5.56% portfolio real rate of return be used in this analysis of the UCRP investment return assumption. This is 0.34% lower than the corresponding real rate of return that was calculated four years ago. This is due to lower expected real returns by asset classes provided to us by the investment advisory firms. Plan Expenses The real rate of return assumption for the portfolio needs to be adjusted for investment expenses to be paid from investment income. Note that the valuation assumptions include a separate loading for administrative expenses, as discussed in Section IV(J) of this report. We obtained information on investment expenses from the Treasurer s Office to be used in the setting of this assumption. Based on the average of the investment expenses for the years ending June 30, 2009 and June 30, 2010 it appears that a future investment expense assumption of 0.65% is reasonable. Risk Adjustment As noted above, this model adjusts the real rate of return assumption for the portfolio to reflect the potential risk of shortfalls in the return assumptions. The UCRP asset allocation also determines this portfolio risk, since risk levels are driven by the variability of returns for the various asset classes and the correlation of returns among those asset classes. This portfolio risk is incorporated into the real rate of return assumption through a risk adjustment, which in turn corresponds to a statistical confidence level of meeting or exceeding the assumed return. The purpose of the risk adjustment (as measured by the corresponding confidence level) is to increase the likelihood of achieving the assumed investment return in the long term by factoring market volatility into the assumption. The 5.56% expected real rate of return developed earlier in this report was based on mean or average returns. This means there is a 50% chance of the actual return being at least as great as the average (assuming a symmetrical distribution of future returns). The risk adjustment increases that probability. This is consistent with our experience that retirement plan fiduciaries would generally prefer that returns exceed the assumed rate more often than not. -15-

19 Four years ago, The Regents adopted an investment return assumption of 7.50%. In combination with the inflation, real return and expense components from four years ago, that return implied a risk adjustment of 1.75%, reflecting a confidence level of 74% that the actual average return over 15 years would not fall below the assumed return, assuming that the distribution of returns over that period follows the normal statistical distribution. 1 In our model, the confidence level associated with a particular risk adjustment represents the likelihood that the Plan s actual average return would equal or exceed the assumed value over a 15-year period. For example, if we set our real rate of return assumption using a risk adjustment that produces a confidence level of 60%, then there would be a 60% chance (6 out of 10) that the average return over 15 years will be equal to or greater than the assumed value. The 15-year time horizon represents an approximation of the duration of the Plan s liabilities, where the duration of a liability represents the sensitivity of that liability to interest rate variations. If we use the same 74% confidence level to set this year s risk adjustment based on the current longterm portfolio standard deviation of 13.97% provided by the Treasurer s Office (as determined by Mercer Investment Consulting), the result is a large risk adjustment of 2.40%. Together with the other investment return components, this produces a net investment return assumption of 6.01%, which is substantially lower than the current assumption of 7.50%. Alternatively, and for comparison, if the return assumption is left at 7.50% the corresponding risk adjustment of 0.91% represents a lower confidence level of 60%. We note that the risk adjustment model and associated confidence level is most useful as a means for comparing how The Regents have positioned themselves over periods of time. 2 This range of values for the confidence level should be considered in context with other factors, including: 1. As noted above, the confidence level is more of a relative measure than an absolute measure, and so can be reevaluated and reset for future comparison. The prior level of 74% was substantially higher than the corresponding values for Segal s other California public sector retirement system clients, which range from 55% to 62%. 1 Based on an annual portfolio return standard deviation of 10.13% provided in Strictly speaking, future compounded long-term investment returns will tend to follow a log-normal distribution. However, we believe the Normal distribution assumption is reasonable for purposes of setting this type of risk adjustment. 2 In particular, it would not be appropriate to use this type of risk adjustment as a measure of determining an investment return rate that is risk-free. -16-

20 2. The confidence level is based on the standard deviation of the portfolio that is determined by UCRP s investment consultant and provided to us by the Treasurer s Office. The standard deviation is a statistical measure of the future volatility of the portfolio and so is itself based on assumptions about future portfolio volatility and can be considered somewhat of a soft number. 3. We also note that the reported standard deviation increased from around 10% to 14% between 2007 and This is a very large increase, and by itself accounts for about a 0.85% (85 basis points) reduction in the expected return or 6% (6 percentage points) of the decrease in the confidence level. Based on discussions with the Treasurer s Office, we understand that this change reflects a reassessment of the amount of risk taken on by the UCRP portfolio, where that reassessment is primarily a result of the significant market downturn during As with any model, the results of the risk adjustment model should be evaluated for reasonableness and consistency. This is discussed in the following Test of Risk Adjustment section that includes a discussion of the relationship between the inflation assumption and the risk adjustment. Taking into account all of the above, our recommendation is to maintain the net investment return assumption at 7.50%. In terms of our risk adjustment methodology, this return implies a risk adjustment of 0.91% reflecting a confidence level of 60% that the actual average return over 15 years would not fall below the assumed return. -17-

21 Recommended Investment Return Assumption The following table provides the calculated net investment return assumption that results from the previous discussion. Components of the Net Investment Return Assumption Assumption Component Recommended Value Inflation 3.50% Plus Portfolio Real Rate of Return 5.56% Minus Expense Adjustment (0.65)% Minus Risk Adjustment (0.91)% Total 7.50% As noted above, Segal s other California public retirement system clients have risk adjustments corresponding to confidence levels in the range of 55% to 62%. A 60% confidence level is a substantial reduction compared to the confidence level implicit in this assumption four years ago. However, it does continue to provide some protection against the risk of future returns falling short of the assumed return and future increases in the total funding policy contribution. Test of the Risk Adjustment The original development of the risk adjustment component of our investment earnings model arose from our experience with many retirement boards over many years. We consistently observed that combining the board s inflation assumption with the real return and expense components (i.e., using no risk adjustment) produced and produces a substantially higher assumed return than what the boards actually adopt, regardless of the consulting actuary or the methods involved in the process. This led to the development of a risk adjustment component for our model. There is a range of risk adjustment methodologies that may be incorporated in the development of an earnings assumption. Ideally, the particular risk adjustment selected should reflect the downside risk tolerance of the boards making the decision. This is similar to the volatility risk that boards consider when selecting an appropriate asset allocation. In addition to the generally risk adverse attitude of retirement plan boards as noted above, we believe another reason for the use of a risk adjustment is to control the risk of overstating the effect of the inflation assumption on the assumed investment return. As noted earlier, the inflation assumption for -18-

22 actuarial valuations is generally longer term than that used by investment consultants. For many years, that has led to higher actuarial valuation inflation assumptions. A higher inflation assumption has a conservative effect higher current cost on the wage increase and COLA assumption, but is less conservative as part of the investment earnings assumption. In effect, the risk adjustment compensates for this by offsetting the effect of the higher inflation assumption on assumed investment earnings. One way to test the reasonableness of the risk adjustment incorporated in our recommendation is to compare our risk adjusted investment return (i.e., 7.50%) against the expected net investment return that would result from using the average of all the capital market assumptions including the lower inflation assumptions of the investment consultants in our sample. The following table shows that comparison. It shows that the difference between our recommended return and that derived using the average of all the capital market assumptions of the investment consultants in our sample is relatively small, and can be attributed to the relationship between the two different inflation assumptions and the risk adjustment. Assumption Element Risk Adjusted Investment Return Average of Investment Consultant Sample Difference Inflation 3.50% 2.66% 0.84% Risk Adjustment (0.91)% 0.00% (0.91)% Real Rate of Return 5.56% 5.56% 0.00% Expenses (0.65)% (0.65%) 0.00% Total 7.50% 7.57% (0.07)% The 0.07% (7 basis points) difference between the two calculations represents about a 1% greater confidence level under the risk adjusted method. This indicates that the risk adjustment more than offsets the effect of using an inflation assumption higher than that used in the capital market assumptions, and so produces a somewhat greater confidence level. Finally, as one additional point of comparison, we did receive information from the Treasurer s Office that was prepared by Mercer Investment Consulting and showed that based on their long-term capital market assumptions, the expected geometric mean return of the long-term UCRP asset allocation is 7.96%. We stress that this information was prepared not for purposes of setting the investment return -19-

23 assumption, but rather for use in their role as investment consultant. Nonetheless, we observe that when adjusted for expenses this result is comparable to our recommendation. Stochastic Approach One common methodology for setting the actuarial investment rate of return assumption involves stochastic modeling. While this type of analysis is routinely used for asset allocation studies, we believe some caution should be exercised for using it to determine the actuarial investment rate of return assumption. This is because most often stochastic modeling is based entirely on a single set of capital market assumptions about the level and volatility of future returns. In other words, this approach involves setting one assumption based on another set of assumptions. Aside from the apparent circularity in this approach, this can lead to the undesired result of expected investment returns that vary significantly depending on which investment consultant develops the capital market assumptions used by a retirement plan. As discussed earlier in this report, this is why we use a broader average of capital market returns when using our risk adjustment model. With that reservation noted, we did receive information from the Treasurer s office on stochastic returns that were used solely for purposes of setting UCRP s asset allocation. This stochastic modeling is scenario based and has a very short-term time horizon of three years. Four most likely economic scenarios were developed and asset returns were estimated for each scenario. A key component of that modeling was to examine the downside risk under each scenario. The four scenarios studied were described as Most Likely, Good, Bad and Disaster. The graph below shows the probability that a given level of compound annual geometric returns is exceeded over the three-year period. 1 Probability Return Exceeds x% Frequency % 20% 0% 20% 40% 60% Compound Annual 3 Year Return Most Likely Good Bad Disaster -20-

24 The median compound annual geometric returns over the three-year period for each of these scenarios are respectively, 7.7%, 8.7%, 5.4% and 2.7%. This is indicative of the significant variation in expected returns among the four scenarios. Besides the wide range of results, the other drawback of relying too heavily on this information for setting the actuarial investment rate of return assumption is that it is based on a very short time horizon (three years), which is much shorter than the duration over which the long-term actuarial investment rate of return assumption is used to project returns for a retirement plan. Nonetheless, these results are at least generally consistent with the 7.50% investment earnings assumption recommended above. Conclusions and Recommendation In summary: When compared to other systems, the 7.50% assumption is conservative and consistent with The Regents historical practice in setting the investment earnings assumptions. The risk adjustment model indicates a 60% confidence level that the average future return of the UCRP portfolio will be no less than 7.50%. While this is a substantial reduction in confidence level from four years ago, it is consistent with corresponding results from other systems. The 7.50% risk adjusted investment return assumption is neither overly conservative nor overly aggressive when measured against the market and inflationary expectations of a sample of investment consultants. Based on these results, we recommend that the investment return assumption remain at 7.50%. -21-

25 C. SALARY INCREASE Salary increases impact Plan costs in two ways: (i) by increasing members benefits (since benefits are a function of the members Highest Average Plan Compensation (HAPC)) and future normal cost collections; and (ii) by increasing total active member payroll over which UAAL payments (or credits if the UAAL is negative) can be amortized. These two impacts are discussed separately below. As an active member progresses through his or her career, increases in pay are expected to come from three sources: 1. Inflation Unless pay grows at least as fast as consumer prices grow, active members will experience a reduction in their standard of living. There may be times when pay increases lag or exceed inflation, but over the long term, labor market forces will require employers to maintain the members standard of living. As discussed earlier in this report, we are recommending no change to the current 3.50% inflation rate assumption. This inflation component will be used as part of the salary increase assumption. 2. Real Across the Board Pay Increases These increases are typically termed productivity increases since they are considered to be derived from the ability of an organization or economy to produce goods and services in a more efficient manner. As that occurs, some portion of the value of these improvements can provide a source for pay increases. These increases are typically assumed to extend to all active members across the board. The State and Local Government Workers Employment Cost Index produced by the Department of Labor provides evidence that real across the board pay increases have averaged about 0.7% to 1.0% annually during the last ten to twenty years. We also referred to the annual report on the financial status of the Social Security program published in August In that report, real across the board pay increases are forecast to be 1.2% per year under the intermediate assumptions. The real pay increase assumption is generally considered a more macroeconomic assumption. However, we note that the actual average inflation plus across the board increase (i.e., wage inflation) over the four-year experience period was 4.7%. This is nearly 1.5% greater than our proposed price inflation assumption. -22-

26 Considering these factors, we recommend increasing the real across the board salary increase assumption from 0.25% to 0.50%. This means that the combined inflation and across the board salary increase assumption increases from 3.75% to 4.00%. 3. Promotional and Merit Increases As the name implies, these increases come from an active member s career advances. This form of pay increase differs from the previous two, since it is specific to the individual. For UCRP, this assumption is structured as a function of a member s service, and it is derived from employer- and member-specific information as part of the experience study. The annual promotional and merit increases are determined by measuring the actual salary increases by active members, net of the inflationary and real across the board components. Increases are measured separately for Faculty and Staff/Safety members. This is accomplished by: Measuring each continuing member s actual salary increase over each year of the experience period; Excluding any members with increases of more than 50% or decreases of more than 10% during any particular year; Categorizing these increases according to member demographics; Removing the wage inflation component from these increases (estimated as the increase in the members average salary during the year for all members); Averaging these annual increases over the four-year experience period; and Modifying current assumptions to reflect some portion of these measured increases reflective of their credibility. -23-

27 The following table compares Faculty members actual average promotional and merit increases by years of service over the four-year experience period from July 1, 2006 through June 30, The actual increases were reduced by the actual average inflation plus real across the board increase (i.e., wage inflation) for each year over the four-year experience period (4.4% on average). FACULTY Years of Service Current Assumptions July 1, 2006 Through June 30, 2010 Average Faculty Promotional and Merit Increases Proposed Assumptions Less than % 0.25% 2.75% & over Average 2.56% 1.36% 2.07% -24-

28 The following table provides similar information for Staff and Safety members. The actual average promotional and merit increases were determined by reducing the actual average total salary increases by the actual average inflation plus real across the board increase (i.e., wage inflation) for each year over the four-year period (4.7% on average). STAFF AND SAFETY Years of Service Current Assumptions July 1, 2006 Through June 30, 2010 Average Staff and Safety Promotional and Merit Increases Proposed Assumptions Less than % 0.65% 2.75% & over Average 1.86% 0.82% 1.43% -25-

29 Charts 2 and 3 provide a graphical comparison of the actual promotional and merit increases, compared to the current and proposed assumptions. Chart 2 shows this information for Faculty members and Chart 3 is for Staff and Safety members. The charts also show the actual promotional and merit increases based on an average of both the current and previous four-year experience period. This is discussed below. We realize that the four-year experience period from July 1, 2006 through June 30, 2010 consisted of some years where salary increases were suppressed and other years where catchup pay increases occurred for many members. Therefore, we also examined the prior four-year experience period. We believe that when the two four-year experience periods are combined into an average result it provides a reasonable representation of potential future increases over the long-term. Based on this experience, we are proposing reductions in the promotional and merit salary increases for all members. As mentioned earlier in this report, the temporary Furlough/Salary Reduction Plan approved by The Regents in July 2009 occurred during the four-year experience study period. The salary data upon which this experience study is based excludes the temporary Furlough/Salary Reduction Plan. This is because of the temporary nature of the Furlough/Salary Reduction Plan and also due to the amendment to UCRP adopted by The Regents to ensure that the Furlough/Salary Reduction Plan has no impact on the calculation of member benefits. -26-