TEXAS MUNICIPAL RETIREMENT SYSTEM ACTUARIAL EXPERIENCE INVESTIGATION STUDY AS OF DECEMBER 31, 20 14 November 18, 2015 Board of Trustees Texas Municipal Retirement System Austin, Texas Dear Members of the Board: Subject: Results of the 2015 Experience Study We are pleased to present our report of the 2015 Actuarial Experience Investigation Study for the Texas Municipal Retirement System (TMRS). Our report includes a discussion of the recent experience of the System, it presents our recommendations for new actuarial assumptions and methods, and it provides information about the actuarial impact of these recommendations on the liabilities and other key actuarial measures of TMRS. With the Board's approval of the recommendations in this report, we believe the actuarial condition of the System will be more accurately portrayed. The Board’s decisions should be based on the appropriateness of each recommendation, not on their collective effect on funding periods or unfunded liabilities. This study was conducted in accordance with generally accepted actuarial principles and practices, and with the Actuarial Standards of Practice issued by the Actuarial Standards Board. The undersigned meet all of the Qualification Standards of the American Academy of Actuaries. In addition, all of the undersigned have extensive experience as retained public sector actuaries for several large, statewide public retirement systems. We wish to thank Ms. Leslee Hardy, ASA, EA, FCA, MAAA, Director of Actuarial Services, and the entire TMRS staff for their assistance in this project. Sincerely, Joseph P. Newton, FSA, EA, MAAA Mark R. Randall, FCA, EA, MAAA Brad Stewart, ASA, EA, MAAA Yi Chen, ASA, MAAA J:\3052\2015\Exp\Report\ExperienceStudy Linked.docx Texas Municipal Retirement System Table of Contents Table of Contents PAGE COVER LETTER SECTION I 2 INTRODUCTION SECTION II 7 SUMMARY OF RECOMMENDATIONS SECTION III 10 ANALYSIS OF EXPERIENCE AND RECOMMENDATIONS SECTION IV 35 ACTUARIAL IMPACT OF RECOMMENDATIONS SECTION V 38 SUMMARY OF NEW ASSUMPTIONS SECTION VI 50 SUMMARY OF DATA AND EXPERIENCE APPENDIX I TERMINATION AND POPULATION EXPERIENCE BY CITY APPENDIX II ACTUARIAL IMPACT BY CITY SECTION I INTRODUCTION Section I Texas Municipal Retirement System Introduction Introduction A periodic review and selection of the actuarial assumptions is one of many important components of understanding and managing the financial aspects of the Texas Municipal Retirement System (TMRS). Use of outdated or inappropriate assumptions can result in understated costs which will lead to higher future contribution requirements or perhaps an inability to pay benefits when due; or, on the other hand, produce overstated costs which place an unnecessarily large burden on the current generation of members, employers, and taxpayers. A single set of assumptions is typically not expected to be suitable forever. As the actual experience unfolds or the future expectations change, the assumptions should be reviewed and adjusted accordingly. It is important to recognize that the impact from various outcomes and the ability to adjust from experience deviating from the assumption are not symmetric. Due to compounding economic forces, legal limitations, and even moral obligations, outcomes from underestimating future liabilities are much more difficult to manage than outcomes of overestimates, and that asymmetric risk should be considered when the assumptions are set and the investment policy and funding policy are created. As such, the assumption set used in the valuation process needs to represent the “best estimate” of the future experience of the System and be at least as likely, if not more than likely, to overestimate the future liabilities than underestimate them. Using this strategic mindset, each assumption was analyzed compared to the actual experience of TMRS and general experience of other large public employee retirement systems. Changes in certain assumptions and methods are suggested upon this comparison to remove any bias that may exist and to perhaps add a slight margin for future adverse experience, where appropriate. Next, the assumption set as a whole was analyzed for consistency and to ensure that the projection of liabilities was reasonable and consistent with historical trends. The following report provides our recommended changes to the current actuarial assumptions. Summary of Process In determining liabilities and contribution rates for retirement plans, actuaries must make assumptions about the future. Among the assumptions that must be made are: • Retirement rates • Mortality rates • Turnover rates • Disability rates • Investment return rate • Salary increase rates • Inflation rate 2 Section I Texas Municipal Retirement System Introduction For some of these assumptions, such as the mortality rates, past experience provides important evidence about the future. For others, such as the investment return assumption, the link between past and future results is much weaker. In either case, actuaries should review the plan’s assumptions periodically and determine whether these assumptions are consistent with actual past experience and with anticipated future experience. The last such actuarial experience investigation was performed in conjunction with the December 31, 2010 actuarial valuation. For this experience study, we have added TMRS’ experience for the four-year period from December 31, 2010 through December 31, 2014 (FY 2011 – FY 2014). In conducting experience studies, actuaries generally use data over a period of several years. This is necessary in order to gather enough data so that the results are statistically significant. In addition, if the study period is too short, the impact of recent economic conditions may lead to misleading results. It is known, for example, that the health of the general economy can impact salary increase rates and withdrawal rates. Using results gathered during a short-term “boom or bust” will not be representative of the long-term trends in these assumptions. Also, the adoption of legislation, such as plan improvements or changes in salary schedules, will sometimes cause a short-term distortion in the experience. For example, if an early retirement window was opened during the study period, we would usually see a short-term spike in the number of retirements followed by a dearth of retirements for the following two-to-four years. Using a longer study period mitigates giving too much weight to such short-term effects. On the other hand, using a much longer period could “water down” real changes that may be occurring, such as mortality improvement or a change in the ages at which members retire. For this analysis, we used between five and twenty years of data, depending on the assumption being studied as follows: Assumption Data Used Comment Payroll/Population Growth 10-20 Years Long term trends are needed, but more importantly, prospective changes must be considered Individual Salary Increases 9 Years 9 years is the combination of the two most recent experience studies. Longer period will capture a longer economic cycle Termination 9 Years Longer period will capture a longer economic cycle Post-Retirement Mortality 4 Years This assumption reacts the quickest to changing trends and is best studied using an even number of years for comparing to generational mortality tables. More years were used to analyze the rate of improvement over time All other 5 Years The assumptions react quicker to changing trends and are less correlated with the economic cycle. Five years provides more credibility to some of the assumptions that have smaller incidence, such as active mortality and disability 3 Section I Texas Municipal Retirement System Introduction In an experience study, we first determine the number of deaths, retirements, etc. that occurred during the study period. Then we determine the number expected to occur, based on the current actuarial assumptions. The number of “expected” decrements is determined by multiplying the probability of the occurrence at the given age, by the “exposures” at that same age. For example, let’s look at a rate of retirement of 15% at age 55. The number of exposures can only be those members who are age 55 and eligible for retirement at that time. Thus, they are considered “exposed” to that assumption. Finally, we calculate the A/E ratio, where "A" is the actual number (of retirements, for example) and "E" is the expected number. If the current assumptions were “perfect”, the A/E ratio would be 100%. When it varies much from this figure, it is a sign that a new assumption may be needed. (However, in some cases we prefer to set our assumptions to produce an A/E ratio a little above or below 100%, in order to introduce some conservatism.) Of course we not only look at the assumptions as a whole, but we also review how well they fit the actual results by gender, by age, and by service. If the data leads the actuary to conclude that new tables are needed, the actuary may "graduate" or smooth the results, since the raw results can be quite uneven from age to age or from service to service. Please bear in mind that, while the recommended assumption set represents our best estimate, there are other reasonable actuarial assumption sets that could be supported. Some reasonable assumption sets would show higher or lower liabilities or costs. For example, while our analysis may conclude that an overall 6.75% investment return assumption is appropriate, others might argue that a different rate is also appropriate. Organization of Report Section II summarizes our recommendations. Section III contains our findings and recommendations for each actuarial assumption. The impact of adopting our recommendations on liabilities and contribution rates is shown in Section IV. Section V discloses all of the actuarial assumptions and methods. Finally, tables summarizing the analysis of the assumptions are in Section VI. Appendix I provides the detail for the termination load and population decline assumption by City and Appendix II provides the impact on individual employer’s contribution rates. Section VI Exhibits The exhibits in Section VI should generally be self-explanatory. For example, on page VI-10, we show an exhibit analyzing the termination rates for members with 10 or more years of service. The second column shows the total number of members with at least 10 years of service who terminated during the study period. This excludes members who died, became disabled or retired. Column (3), labeled “Exposure Weighted by Salary” shows the total exposures of this group. This is the number of members who meet the criteria who could have terminated during any of the years. On this exhibit, the exposures exclude anyone eligible for unreduced retirement. A member is counted in each year they could have terminated, so the total shown is the total exposures for the ten-year period. Column (4) shows the probability of termination based on the raw data. That is, it is the result of dividing the actual number of terminations (col. 2) by the number exposed (col. 3). 4 Section I Texas Municipal Retirement System Introduction Column (5) shows the current termination rate and column (6) shows the new recommended termination rate. Columns (7) and (8) show the expected numbers of terminations based on the current and proposed termination assumptions. Columns (9) and (10) show the Actual-to-Expected ratios under the current and proposed termination assumptions. 5 SECTION II SUMMARY OF RECOMMENDATIONS Section II Texas Municipal Retirement System Summary of Recommendations Summary of Recommendations Our recommended changes to the current actuarial assumptions, and supporting rationale, may be summarized as follows: Economic Assumptions 1. Lower the inflation assumption from 3.00% to 2.50%. 2. Decrease nominal investment return assumption from 7.00% to 6.75%. Based on our analysis, the expected arithmetic return based on the target asset allocation is 6.80% while the median expected geometric return over the next 20 years is 6.66%. (The Board has already voted to reduce the investment return assumption to 6.75% based on preliminary information provided before publication of this report). 3. Set the general wage inflation assumption at 3.00% (0.50% above inflation). 4. Modify the existing salary scale assumption based on recent trends and experience with an ultimate increase for long service members equal to 3.50%. 5. Make no changes to the current payroll growth rate assumption of 3.00%. The payroll growth assumption does not impact the liabilities, only the development of the amortization of the unfunded actuarial accrued liability. However, include a reduction for some cities based on patterns of population decline. 6. Reduce the assumed future cost-of-living increases in conjunction with the lower inflation assumption. Mortality Assumptions (Valuation Purposes Only – No Impact on Annuity Purchase Rates) 7. Make no change to the current post-retirement mortality tables for non-disabled retirees. 8. Change the disabled post-retirement mortality assumption to be the same as the healthy tables, except with a 3 year set-forward. In addition, add a 3% minimum mortality probability to reflect impaired mortality for this group. 9. Assume members who become disabled will choose a 50% Joint and Survivor payment option. 10. Update the pre-retirement mortality tables to be 50% of the post-retirement assumptions. 7