Pennsylvania Municipal Retirement System

Transcription

1 Pennsylvania Municipal Retirement System Experience Study Results and Recommendations For the period covering January 1, 2009 December 31, 2013 Produced by Cheiron July 2015

2 Table of Contents Section Page Letter of Transmittal... i Section I Board Summary...1 Section II Analysis of Economic Assumptions...17 Section III Analysis of Demographic Assumptions...36 Appendices Appendix A Salary Increase Analysis...46 Appendix B Current Actuarial Assumptions...48 Appendix C Proposed Actuarial Assumptions...53

3 LETTER OF TRANSMITTAL July 14, 2015 Mr. Stephen W. Vaughn Pennsylvania Municipal Retirement System Commonwealth of Pennsylvania 1010 North 7 th Street, Suite 301 Eastgate Center Harrisburg, PA Dear Board Members: At your request, we have completed an experience study of the Pennsylvania Municipal Retirement System (PMRS). Our study compares assumed versus actual experience with respect to all demographic and economic assumptions used in the preparation of the Actuarial Valuations for the five-year period from January 1, 2009 through December 31, The purpose of this report is to present the results of our study as well as alternative assumptions that will be employed for the January 1, 2016 Actuarial Valuation, as determined by the board. This report is for the use of PMRS and its auditors in preparing financial reports in accordance with applicable law and accounting requirements. We hereby certify that, to the best of our knowledge, this report is complete and accurate and has been prepared in accordance with generally recognized and accepted actuarial principles and practices which are consistent with the Code of Professional Conduct and applicable Actuarial Standards of Practice set out by the Actuarial Standards Board, and that we meet the Qualification Standards, as defined by the American Academy of Actuaries, to render the opinion contained in this report. Finally, in preparation of this report, we relied, without audit, on data provided to us by PMRS. We performed an informal examination of the obvious characteristics of the data for reasonableness and consistency in accordance with Actuarial Standard of Practice No. 23. To the extent there are material omissions or misrepresentations of any of the data elements we relied upon in formulating opinions, we could arrive at different conclusions. This report was prepared exclusively for PMRS for the purpose described herein. Other users of this report are not intended users as defined in the Actuarial Standards of Practice, and Cheiron assumes no duty or liability to any other user. Sincerely, Cheiron Kenneth A. Kent, FSA, FCA, MAAA Principal Consulting Actuary Karen Zangara, FSA, MAAA Principal Consulting Actuary

4 SECTION I BOARD SUMMARY At the request of the Retirement Board, Cheiron has performed a study of the experience of the Pennsylvania Municipal Retirement System (PMRS). This experience study examines PMRS s experience during the five-year period from January 1, 2009 through December 31, 2013, The Study Period. This report presents the results of our study and examines alternatives to several of the current actuarial assumptions for use in future valuations of PMRS based on these results as well as past trends and future expectations. We studied PMRS s experience with respect to both demographic and economic assumptions. Demographic assumptions deal with expected membership behavior or experience. These include the retirement rates, termination rates, disability rates, and mortality rates. Economic assumptions deal with System wide elements such as investment returns, inflation, salary increase rate (salary scale) and administrative expenses. Salary increases can be considered either demographic (membership oriented) or economic (given the inflation component). For this study, we included salary experience under the economic portion of the study. The overall experience during this period has been dominated by the recovery of the recession of 2008 and Typical of economic down-turns, public employers are frequently slow to react to reducing wage increases because of timing delays in communication and prior agreements. Similarly, participants are quick to react to the economic down-turn by reducing the number of terminations during the period of uncertainty following a significant recession. Therefore, experience study results which span a recession need to be carefully considered by the Board to consider which results are driven by short term reaction to the recession and which results reflect long-term trend changes and changes to the future plan experience. 1

5 SECTION I BOARD SUMMARY Table I-1 summarizes all key findings and recommendations arising from this study. Table I 1 Recommended Changes to Economic and Demographic Assumptions (All Municipal and Uniform Plans) Demographic Current Assumption Recommended Change Retirement Termination Municipal rates by age Uniformed rates by age by service and size of plan Municipal increase rates for ages and 67+ Uniformed reduce rates for ages 57+ Municipal (<25 actives) reduce rates for service < 5 and increase for service > 5 Uniform (<25 actives) reduce rates for service < 4 and increase for service > 3 Municipal (25+ actives) reduce rates for service < 3 and increase for service > Uniform (25+ actives) no change Disability by age Reduce rates for ages 25+ Mortality - Healthy Actives Mortality - Healthy Retirees Mortality - Disabled Males: RP-2000 Combined Healthy table setback one year Females: RP-2000 Combined Healthy table setback five years RP-2000 Combined Healthy table (sexdistinct) RP-2000 Combined Healthy table setforward ten years (sex-distinct) RP-2000 Employee Healthy table (sex distinct) projected to 2015 based on Scale AA mortality improvements, no setback for males, five year setback for females RP-2000 Retiree Healthy table (sex distinct) projected to 2005 / 2010 for males/females based on Scale AA mortality improvements No change Miscellaneous Demographic Marital Status 85 percent married with females 3 years younger than their spouses Change percent married to 80 percent Economic Inflation/Cost of Living Rate 3.00% Reduce the rate to 2.8% Investment Return/Discount Rate 5.50% No Change Salary Increase Rate Salary scale by age Decrease salary scale Expenses Based upon expected expenses for the current year Use prior year increased at 5% Methods for Experience Study Analysis We present the key findings of our experience review of the demographic and economic assumptions and provide alternative assumptions for consideration. When we look at demographic assumptions we are looking at the actual occurrences of typical events during a participants life. When a participant leaves a certain status like an active to retired event we refer to that as a decrement. The assumptions we test in performing this study are the probability of these events occurring or the rate of decrement from active to retired status once an active participant reaches eligibility to retire, for example. The actual experience refers to the number of people who decrement for a specific event. The expected 2

6 SECTION I BOARD SUMMARY number represents the number of participants we expected would decrement according to the assumptions in place during the study period. For each of the first four basic sets of demographic assumptions noted above in Table I-1, we provide graphs outlining the results of the experience study analysis and charts outlining the actual vs. expected number of participants for the occurrence. Graph Analysis The graphs within this report provide the results of the experience study for each demographic assumption under review. Along the left axis is the rate of the decrement occurring while the participant s age (or possibly service, depending upon the assumption) is provided along the horizontal axis. For all graphs presented, the data details are provided in the Section II of the report. There are three items provided in each graph for the demographic assumptions: 1. the actual rate of the occurrence (such as retirement) (illustrated by the black points), 2. the expected rate of the occurrence (such as retire) based on the current assumptions (illustrated in royal blue), and, 3. the expected rate of the occurrence based on the alternative assumptions (illustrated in teal) In addition to reviewing the ratios of actual versus expected, we measure the credibility of the data at each study period is also reviewed. Credibility refers to the question: Do we have enough data to make a judgement that the experience supports a change in trend or is sufficient to an assumption change? In any trend statistical analysis, one must consider whether the experience is sufficient to believe a true change is occurring over what was expected in the past. For example, if a coin is flipped twice and both times comes up heads, one will not necessarily conclude that both sides are heads. However, if the coin is flipped 100 times and all flips result in heads, one would have more confidence in believing that both sides of the coin show heads. The more incidences that occur at any date point, such as an age, the greater the confidence that the experience is real and will continue to occur at the observed rate. Thus, we put more credence on high confidence intervals. The credibility of the data is illustrated by the confidence interval at each study period. This represents the area of the box or bar around each data point. The wider the confidence interval (illustrated by gold bars on the graphs), then the analysis is less credible due to fewer participants exposed to the event being analyzed. The narrower the confidence interval, then the analysis is more credible due to more participants exposed to the event being analyzed. Typically we want to consider assumption changes around the data points with the narrowest bars because the data is sufficient to support the expectation that these results represent reasonable expectations of future participant behavior. The graphs within this study all show 90% confidence intervals. This implies a range where 90% of the time the actual experience is expected to fall within the range. Ratio Analysis - Actual vs. Expected Data 3

7 SECTION I BOARD SUMMARY In addition, for each of the first four basic sets of demographic assumptions, we determined an actual to expected occurrence ratio at each age (sometimes further segregated by gender or by plan type). For example, for Uniformed plans there are 104 participants who actually retired during the study period while there were participants expected to retire. Therefore the ratio of actual to expected retirees is 69% (104 divided by 151.5). Another way to say this is, 31% fewer members retired than expected during the study period. If the actual to expected ratio is greater than one, then the assumption may be too low; if it is less than one, then the assumption may be too high. When there is a trend of rates that are materially too high or low and the data is credible we provide an alternative assumption to bring the ratios closer to one, which means the number of people we expect for an occurrence under the alternative assumptions is closer to the actual number of people who had the occurrence. Changes (or no changes) in assumptions are supported when analyzing the total actual decrements versus expected number of decrements based upon the current assumptions over the five-year testing period. The ratio of actual divided by expected number of participants impacted by the decrement (in this case, retirements) during this period demonstrates how well the current assumptions meet the actual experience of the plan. Ideally, this ratio should be about 100% to show that the expected retirements approximately match the actual retirements. However, this ratio analysis must be balanced with the experience graphs and the data used to determine this ratio, because outlier age groups may skew this ratio accuracy. Economic Assumption Review The review of the economic assumptions is based upon the following, which is reviewed in more details in Section III of the report: o historical data in general (i.e. the markets), o historical experience of the plan, o outlook for the future, and o assumptions used by other public sector plans. The following section in the Board Summary provides an explanation of the principal experience study results. 4

8 SECTION I BOARD SUMMARY Summary of principal experience study results and recommendations: 1. Retirement. Participants meeting the eligibility requirements for normal retirement benefits are exposed to the likelihood of when they will choose to retire. Although some municipalities offer early retirement benefits, for PMRS plans all early retirement benefits are actuarially equivalent to the normal retirement benefits. This means there is no gain or loss associated with the value of the benefit paid to a participant electing early retirement. Therefore, the explicit valuation of early retirement benefits is not valued in the PMRS valuation because an early retirement benefit has the same value as a normal retirement benefit payable at an earlier date because the benefit is reduced to reflect the longer period of payments expected and the fact payments are starting sooner. of retirement were significantly lower than expected for Police and Fire Divisions (i.e. Uniform Plans) compared to the current assumption for ages greater than 55. This is a continuation of the trend first observed in the prior study. Whether this is due to economic realities or the ability of uniform service employees to be effective longer, it represents a trend we are seeing in other funds as well. Actual retirement rates were analyzed at each age compared to the expected retirement rates at each age. Based on this analysis, we have developed lower retirement rates for consideration for the Uniform Plans for ages greater than 55, which are more in-line with the retirement experience for this group. 100% 80% Uniformed Division Active Members: Retirement for Males and Females 90% Confidence Interval Actual Expected 60% 40% 20% 0% < Age For Municipal Plans, actual retirements are slightly higher than expected at ages 55 through about 62 and ages greater than 66. Similar to the prior study, there is volatility in the retirement rates for different ages. We suggest an increase to the retirement assumptions for ages and 67+ for the Municipal Plans. As illustrated by the graph, the confidence intervals for ages prior to 57 are very wide because the number of participants eligible for normal retirement at these ages is relatively low as most plans require attained age 55 as a minimum retirement age. 5

9 SECTION I BOARD SUMMARY 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% Municipal Division Active Members: Retirement for Males and Females 90% Confidence Interval Actual Expected < Age As evidenced by Table I-2 below, the ratio of actual to expected retirements for the Uniform Plans is significantly different from the target ratio of 100%, which supports, along with the graph above, changing this retirement assumption. For the Municipal Plans, the ratio is fairly close to 100%, but the data supports changes to this retirement assumption at certain ages. Table I 2 Plan Actual Retirements Expected Retirements Ratio: Actual/ Expected Uniform Plans % Municipal Plans 1,020 1, % 2. Termination The current termination assumption for the Municipal and Uniform Plans is based upon the number of active participants in the Plan. There are different termination assumptions for plans with less than 25 active participants and plans with 25+ active participants. These termination assumptions differ for the Municipal and Uniform Plans. For the Municipal Plans with less than 25 active participants, actual terminations are somewhat lower than expected for short service participants and somewhat higher for longer service participants. Due to the relatively high confidence intervals associated with these results, we recommend leveling the termination rates for this group by decreasing the rates for those with years of service less than five and increasing the rates for those with more than five years of service. 6

10 SECTION I BOARD SUMMARY 30% 25% 20% Municipal Division Active Groups with less than 25 members: Turnover for Males and Females 90% Confidence Interval Actual Expected 15% 10% 5% 0% Service Likewise, for the Municipal Plans with 25 or more active participants, actual termination rates were lower than expected for shorter service members and higher for longer service members. We recommend lowering the termination rates for participants with less than three years of service and increasing the rates for those with more than three years of service in Municipal plans with 25+ members. 30% 25% 20% Municipal Division Active Groups with 25+ members: Turnover for Males and Females 90% Confidence Interval Actual Expected 15% 10% 5% 0% Service For the Uniform Plans with less than 25 active participants, actual terminations were less than expected for those with very short service and larger than expected for those with longer service. We recommend lowering the termination rates for participants with less than four years of 7

11 SECTION I BOARD SUMMARY service and increasing the rates for participants with more than three years of service in Uniform Plans with less than 25 members. 30% 25% 20% Uniformed Division Active Groups with less than 25 members: Turnover for Males and Females 90% Confidence Interval Actual Expected 15% 10% 5% 0% Service For the Uniform Plans with 25 or more active participants, there were only 14 terminations over the study period. This is consistent with the results from the prior study which had only 18 terminations. Due to a lack of credible data, we suggest no changes to these termination rates. 30% 25% 20% Uniformed Division Active Groups with 25+ members: Turnover for Males and Females 90% Confidence Interval Actual Expected 15% 10% 5% 0% Service 8

12 SECTION I BOARD SUMMARY The following table shows the actual terminations based upon the experience and the expected terminations using the current assumptions, broken down by the number of active participants in the plans. Ideally, the ratio should be close to 100%. These ratios, along with the graphs provided above, support the recommended changes to the termination rates. Table I 3 Plan Plan Size Actual Expected Ratio: Terminations Terminations Actual /Expected Municipal Plans < % >=25 1,161 1, % Uniform Plans < % >= % 3. Disability While there were relatively few disabled participants, the results from this study are consistent with the prior study, which illustrated decreasing the disability rates to some extent. While the data is not credible at lower ages, the data for ages 50+ is credible. Assuming a smooth increase in the disability rates from ages 20 to 50, we suggest decreasing the disability rates for ages for participants from the Municipal and Uniform Plans, as illustrated in the graph below: Uniformed and Municipal Division Active Members: Disability for Males and Females 0.6% 0.5% 90% Confidence Interval Actual Expected 0.4% 0.3% 0.2% 0.1% 0.0% Age While the table below shows the relative few incidences of disabilities, the result was consistent with the prior year s study where there were 23 actual disabled participants and 76 expected disabled participants, producing a ratio of actual to expected of 30%. Therefore, we recommend lowering the current disability assumption to some extent for the Municipal and Uniformed plans. Currently, the disability rates for participants in Municipal and Uniform plans are 40% and 60% (respectively) of the 1964 OASDI (Social Security) Experience for Males. 9

13 SECTION I BOARD SUMMARY Table I 4 Actual Disabled Expected Disabled Ratio: Plan Participants Participants Actual /Expected All Plan % Mortality The RP-2000 Mortality table setback one year for males and five years for females is currently being used for active members. This means that the ages of males and females are decreased by one year and five years, respectively, when determining the probabilities of death. The RP-2000 Mortality table without adjustment is currently being used for all healthy inactive members. The RP-2000 Mortality table set forward ten years is the current assumption for disabled participants mortality. This means that the ages of disabled retirees are increased by ten years when determining the probabilities of death. For the mortality assumption, if the data is credible it is preferable to build into this assumption future mortality improvements, to some extent. People are living longer and not anticipating this trend for the retired population could create an underfunding of the plan. To account for future mortality improvements, the ratio of Actual to Expected deaths should exceed 100% for the retirees, where the data is credible. For the current actively employed participants where the data is not as credible, probably a 100% ratio would suffice with the understanding that as prescribed by law, this assumption will be reviewed again in four years. In 2014, the Retired Pensioners 2014 (RP-2014) mortality table was published. This study reviewed recent mortality experience of uninsured private retirement plans in the United States. While this is the most recent analysis of mortality rates, it does not include data and results from a significant portion of the public sector. Furthermore, the RP-2014 mortality table for active males actually has increased mortality rates for ages 50 and up when compared to the RP-2000 mortality table for actively employed males. Although these are the most recent mortality tables published by the SOA, we suggest the RP-2000 active male table with 15 years of projected mortality improvements for actively employed males because this provides a better fit for the PMRS experience. The data for the active female mortality rates is not credible, so we suggest applying 15 years of projected mortality improvements to the RP-2000 mortality table for actively employed females set back by 5 years. This provides consistency with the prior assumption and the actual to expected ratio attains 100% under these assumptions. Based upon the retiree mortality rates, the current assumptions for the male retirees appear to provide a mortality improvement assumption already because the actual number of deaths exceeded the expected number of deaths. For the female retirees, additional mortality improvement could be applied. Therefore, we recommend a slight update to the retiree mortality table by applying the RP sex distinct mortality table for retirees with five years of projections for males and 10 years for females. 10

14 SECTION I BOARD SUMMARY For the inactive participants, we excluded the vested terminated data from this study because it appears that the vested terminated employees are not reported as deceased until they are retirement eligible. This reporting delay in the deaths for this group would skew the results of the mortality analysis. Therefore, this data was excluded from the study. The following graphs provide the mortality rates for the following: (1) Healthy active participants (2) Healthy retirees Due to limited credible data for disabled deaths, we provide a table outlining the total number of deaths, actual versus expected. Healthy Active Participants: 1.5% Uniformed and Municipal Divisions Active Members: Mortality for Male Members 90% Confidence Interval Actual Expected 1.0% 0.5% 0.0% < Age 11

15 SECTION I BOARD SUMMARY 1.5% Uniformed and Municipal Divisions Active Members: Mortality for Female Members 90% Confidence Interval Actual Expected 1.0% 0.5% 0.0% < Age Healthy Retirees: 18% 15% 12% Uniformed and Municipal Divisions Retired Members: Mortality for Male Members 90% Confidence Interval Actual Expected 9% 6% 3% 0% Service 12

16 SECTION I BOARD SUMMARY 18% 15% 12% Uniformed and Municipal Divisions Retired Members: Mortality for Female Members 90% Confidence Interval Actual Expected 9% 6% 3% 0% Service Disabled Participants: Table I-5 Uniformed and Municipal Divisions - Disabled Mortality Males and Females Combined Expected Deaths Based Actual Disabled Exposures upon Assumptions: Deaths Current In addition to the graphs, the table below provides the actual versus expected deaths for healthy participants (i.e. excludes disabled participants) and the ratio of these values. For the active incidents of mortality, the ratio is significantly less than 100% implying that the current mortality tables are not reflecting improvements in life expectancy. For the inactive incidents of mortality, the ratios are above 100%, which implies that future mortality improvements are already provided with these tables and a slight change in the assumptions may be justified. Table I 6 Sex Status Actual Deaths Expected Deaths Ratio: Actual/Expected Male All Actives % Female All Actives % Male All Retirees % Female All Retirees % 13

17 SECTION I BOARD SUMMARY 4. Marriage Assumption Based on the study data, only 67% of the participants are married at retirement. The inclusion of children as beneficiaries brings the percentage of participants with spouses or children as beneficiaries are 73%. The current assumption is that 85% of the participants are married at retirement. Given the statistics from the current data, this suggests lowering the assumption part way to 80% and to track this more carefully between experience studies to see if the assumption holds up among new retirees in the next four years. The assumption also assumes a female spouse is, on average, 3 years younger than the male. The data continues to support this assumption. This affects the assumption of the projected form of benefit at retirement and can materially impact the liabilities as the average difference between a life annuity benefit and Joint and 100% Survivor annuity can be an increase of 15%. 5. Salary Increase - The salary increase rate represents the year over year increase in pay of continuing actives. The current assumption is based on a decreasing rate table as participants get older. Upon reviewing the salary increase by age groups, the overall rate of salary increase is lower than the current table especially for participants greater than 25. This was the opposite result of the prior study, where the salary increase rates were higher than expected for ages 20 to 45. This change could be explained in part due to the recession in Frequently during a recession, there is a timing delay for salary increase impact, often due to previously approved pay increases. Therefore, what is being observed could be the delayed impact on salary increases for years However because salary increases are also based on the underlying inflation rate which has also been historically low there is the potential that the lower growth in salaries may continue for some time. The alternative salary increase assumption continues to be an age-based table with higher salary rates of increase at younger ages to reflect the experience. The graph on the next page illustrates the five-year average rate of salary increases year over year by five year age groups and the alternative salary increase assumption. In the following section of the report, additional analysis is provided reviewing the pay increases before and after the recession impact. Under the current assumptions for final average pay plans, a 6% increase in pay is assumed in the final year of employment to reflect the experience of spiking of pay in the final year of retirement that has been experienced. This assumption is separate and apart from the balance of the salary scale assumption and therefore not reflected in the graph below. 14

18 SECTION I BOARD SUMMARY 12% 10% 8% Uniformed and Municipal Divisions Active Members: Salary Increase for Males and Females 90% Confidence Interval Actual Expected 6% 4% 2% 0% Age The alternative assumption does not completely bridge the gap between actual and expected increase rates, which may be appropriate due to the impact of the recession of 2009 on wage increases. The table below shows aggregate actual salary increase rates by year. Table I 7 Actual Salary Year Increase Rate % % % % % Total 3.80% Based on this data, employers may have been slow to reflect the recessionary pressures, possibly due to previously negotiated wage increases. The wage increase rates began to rebound to prior levels during As such, we have not fully reflected the recent experience in the alternative assumption. 6. Investment Return Assumption/Discount Rate - The current discount rate assumption is 5.50%. This rate is on the low end of assumed rates when compared to different public pension funds. On the other hand, it is comparable to the required rates that must be applied in the private sector. This rate was recently updated by the board for the 2013 actuarial valuation. Previously, this rate was 6.00%. By lowering the investment return/discount rate assumption, the board recognized the extraordinarily low market conditions for fixed income and the challenge to meet the 6.00% investment return provided to the plans that participate in PMRS. Furthermore, due to the 2008 significant market 15

19 SECTION I BOARD SUMMARY down-turn, the gap between the system s reserves and the market value of assets needed to be resolved. By lowering the discount rate assumption from 6.00% to 5.50%, an increased probability that the system can credit the investment return for the participating plans and close the gap between the reserves and market value of assets was created. We recommend no change at this time. 7. Inflation Assumption There continues to be a fair amount of speculation about inflation and when it will increase each year. At the same time the rate continues to be at relatively historic low levels and may be influencing our long term expectations. This assumption is a building block for long term return rates and salary growth rates. So part of the explanation for the reduction in these other assumptions is supported by the expectation that inflation rates may be projected to remain lower than the 30 or 40 year historic rates. It is also the assumption applied in measuring the liability for the cost of living provisions adopted by a number of the municipalities. The current assumption is 3.00% while currently the 20 to 30 year average inflation assumption is between 2.4% and 2.8%. We suggest consideration to reduce the current assumption to 2.8% and provide statistics, based upon CPI and the National Conference on Public Employers Retirement Systems (NCPERS) November 2014 study, regarding the current assumption used by the major public sector funds across the country as well as other geometric averages. 8. Administrative Expenses The source for covering administrative expenses continue to be from two sources, the statutory rate of $20/participant charged to the municipalities and the additional expense to cover the actual requirements of the System. This additional expense is covered each year by excess returns over the 5.5% credited return rate. To the extent the rate is not realized the administrative expenses need to be made up in future years. To determine the actuarial value of assets for the system, an assumed expense is added to the reserve accounts. The current expense assumption is to assume the actual expense during the year. However this only works because the valuation is performed as of the beginning of the year and delivered after the year is over. We suggest as an alternative that the administrative expense assumption be changed to the prior year s expenses increased by 5% as an expected expense as of the beginning of the year. While it is this change in assumptions will probably not have any impact on the individual municipalities, for projection purposes for the system, this assumption may be more realistic and allow for a determination of the return rate that must be achieved once the System if fully funded, to remain so after accounting for the 5.5% crediting rate. All of the alternative assumptions if adopted by the Retirement Board would be implemented effective January 1, 2016, based upon the recommendation from the Board. The balance of this report supports the rationale for the alternatives presented above. In Section II, we present detailed analysis and exhibits supporting the various demographic assumption changes. In Section III we present similar information with respect to the economic assumptions. 16

20 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS In this section, we present the key findings of our experience review of the demographic assumptions used by PMRS, including alternative assumptions for consideration. The demographic assumptions included in this review are: 1. Retirement 2. Termination from Active Employment (Other than Death, Disability, or Retirement) 3. Disability 4. Mortality (Active, Inactive Healthy, and Inactive Disabled) 5. Marriage Percentage and Spouse s Age For each of the first four sets of assumptions noted above, we determined an actual to expected occurrence ratio at each age (sometimes further segregated by gender). For example, for Municipal Plans there are 348 participants who were age 65 during the study period and eligible to retire, of which 85 retired. Based on the assumption in place during the study, about 101 of the 348 participants were expected to retire. Therefore the ratio of actual to expect retirees is 84%. Another way to say this is 84% of the number expected to retire during the study period actually did retire. It is worth restating that if the actual to expected ratio is greater than one, the assumption may be too low; if it is less than one, the assumption may be too high. The graphs presented in this section first appeared in Section I; however, we are repeating the graphs in this section along with the corresponding tables with the graph data for the reader s benefit. 1. Retirement A. Current Assumptions All Municipal Employees Assumptions are based on ages once normal retirement eligibility requirements are met. All Police and Fire Employees Assumptions are based on ages once normal retirement eligibility requirements are met. There are no early retirement assumptions for the valuation because none of the plans that participate in PMRS offer subsidized early retirement benefits. All benefits paid earlier than the normal retirement age are actuarially equivalent to the benefits paid at the normal retirement date. In the prior study, for the retirement experience review, early retirement eligible participants were exposed to the retirement decrement to determine whether or not there was a need for an early retirement assumption. Based upon the results of the prior study, it was 17

21 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS determined that participants do not retire early in sufficient numbers to retain an assumption for this decrement. For this study, due to the actuarial equivalence of the retirement benefits and the results of the prior study, only participants eligible for normal retirement benefits were exposed to the retirement decrement. Therefore, the number of participants exposed to retirement in this study at earlier ages has dramatically decreased when compared to the last study. B. Experience All Municipal and Police and Fire Employees The current assumptions for both employee groups are based on age. For the Municipal Plans, the rates were slightly higher than expected at the younger retirement ages and vary more at the higher ages than expected. Overall, they track very closely with the current rates, although some adjustments can be made to better align the assumptions with the experience. Overall, for the Uniformed employees, the actual retirements during the study period were much lower than expected (see the Results section outlined in item D below). C. All Municipal Employees Although overall expected retirements are close to actual retirements, we propose increasing the rates at selected ages to better match the pattern of actual retirements. All Uniformed Employees The current assumptions are significantly higher than current rates at older ages but somewhat lower at younger ages. We propose lowering the rates at older ages while increasing the rates at select younger rates to better reflect the actual experience. D. Results The tables and graphs on the following pages compare three items: the number of people eligible to retire, the number of people expected to retire based on the current assumptions, and the number of people expected to retire based on the alternative assumptions. The alternative assumptions bring the ratios closer to one at each age studied, which means the number of people we expect to retire under the alternative assumptions is closer to the actual number of people who retired. The exception to this goal of having the ratios close to one is at the final retirement age. For retirement decrements, a final retirement age is selected to assume 100% retirements will occur. For PMRS this age is 75. The actual experience does not reflect 100% retirement at this age, because typically a relatively small number of people choose to continue to work. This 18

22 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS difference in assumptions vs the plan experience for the system as a whole impacts such a few number of people that the impact on the liabilities is immaterial. One aspect of these results that will be prevalent throughout the study and exhibited by the graphs is that at later ages, the number of exposures to retirement can be relatively small. For example, at age 60 the Uniform Plans only had 30 participants that reached that age, there are 11 participants expected to retire from this group but only 2 participants retired. As a result, we cannot always have great confidence that the experience reflects a new retirement pattern. This uncertainty is reflected in the confidence intervals shown on the graphs. The larger bars indicate lower confidence in the actual experience. Table II-1 Municipal Division Active Members Retirement for Males and Females Age Exposed Actual Expected Actual Expected Ratio: Actual over Expected Ratio: Actual over < % 2.22% 500% % 500% % 8.00% 250% % 250% % 10.00% 0% % 0% % 15.00% 0% % 0% % 17.50% 0% % 0% % 15.00% 0% % 0% % 15.00% 167% % 167% % 17.22% 129% % 131% % 16.00% 125% % 118% % 17.50% 0% % 0% % 17.14% 83% % 84% % 14.82% 165% % 111% % 12.62% 116% % 105% % 10.18% 123% % 90% % 10.27% 127% % 93% % 10.27% 132% % 97% % 15.44% 103% % 89% % 13.85% 129% % 100% % 23.10% 95% % 122% % 25.83% 72% % 103% % 19.96% 109% % 121% % 29.00% 84% % 98% % 25.57% 132% % 169% % 15.06% 134% % 101% % 15.13% 151% % 115% % 15.16% 106% % 81% % 15.19% 156% % 119% % 14.94% 105% % 78% % 15.11% 193% % 145% % 14.93% 175% % 131% % 15.08% 305% % 230% % 94.89% 28% % 27% Total 5,116 1,020 1, % 19.5% 102% 1, % 101% 19

23 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% Municipal Division Active Members: Retirement for Males and Females 90% Confidence Interval Actual Expected < Age Table II-2 Uniformed Division Active Members Retirement for Males and Females Age Exposed Actual Expected Actual Expected Ratio: Actual over Expected Ratio: Actual over < % 0.0% 0% % 0% % 19.9% 158% % 105% % 14.8% 135% % 200% % 10.0% 163% % 162% % 10.0% 97% % 97% % 9.9% 245% % 243% % 16.4% 113% % 74% % 22.3% 91% % 102% % 24.9% 75% % 93% % 27.6% 60% % 83% % 29.8% 20% % 40% % 35.1% 19% % 44% % 45.2% 46% % 103% % 49.6% 67% % 111% % 54.1% 25% % 67% % 65.5% 24% % 77% % 75.6% 40% % 100% % 94.3% 47% % 44% Total % 28.4% 69% % 93% 20

24 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS 100% 80% Uniformed Division Active Members: Retirement for Males and Females 90% Confidence Interval Actual Expected 60% 40% 20% 0% < Age 2. Termination from Active Employment A. Current Assumptions All Employees Current termination assumptions for all employees are service and count based. Different assumptions apply for plans with less than 25 active participants and for plans with 25 or more active participants. In all cases, termination rates decrease as service increases. See the tables on the following pages for more details. B. Experience All Employees In general, actual terminations were lower than expected for shorter service participants and higher than expected for longer service members. The Municipal Plans with less than 25 active participants experienced lower actual terminations for those with shorter service and slightly higher rates for those with service between five and more years. The Municipal Plans with 25 or more active participants had lower termination rates for those with less than three years of service but generally higher termination rates for participants with more than three years of service. The Uniform Plans with less than 25 active participants experienced lower terminations at one year of service and higher terminations after three years of service. The Uniform Plans with 25 or more active participants had limited credible data upon which to assess that termination trends for these plans. 21

25 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS The recession of 2008 and 2009 also appears to have impacted termination rates for years Unlike the delayed impact of the recession on the wage increases, the recession had a more immediate impact upon lowering termination rates. During this period employees tended to have lower termination rates for plans with 25+ participants. There appears to be no impact on plans with less than 25 participants, maybe because there are fewer employment alternatives in these municipalities, regardless of a recession. C. Municipal Employees For plans with less than 25 active participants, we propose to lower rates for participants with less than five years of service and increase rates for participants with six or more years of service. For plans with 25 or more active participants, the proposed termination rates are lower for participants with less than three years of service and higher for participants with more than three years of service. The next section shows the proposed assumptions for municipal employees over the study period. Uniformed Employees The proposed Uniformed Plans termination rates for plans with less than 25 participants have been lowered for participants with less than two years of service and increased for those with more than three years of service. The current termination rates for the Uniform Plans with 25 or more active participants appear to be suitable based upon the experience overall (total ratio equals 59%) and the limited data. The next section shows the proposed assumptions for uniformed employees over the study period. D. Results The following tables and graphs compare three items; the number of people eligible for the termination decrement, the number of people expected to terminate based on the current assumptions, and the number of people expected to terminate based on the alternative assumptions. For proposed changes, the alternative assumptions bring the ratios closer to one, which implies the number of people we expect to terminate under the alternative assumptions is closer to the actual number of people who terminated. 22

26 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS Table II-3 Municipal Division Active Groups with less than 25 members Turnover for Males and Females Service Exposed Actual Expected Actual Expected Ratio: Actual over Expected Ratio: Actual over < % 20.0% 49% % 66% % 20.0% 57% % 76% % 12.0% 87% % 95% % 10.0% 70% % 88% % 8.0% 81% % 92% % 6.0% 69% % 69% % 4.0% 146% % 106% % 3.0% 170% % 92% % 3.0% 182% % 99% % 2.5% 124% % 77% 10+ 6, % 2.5% 121% % 121% Total 13, % 5.4% 90% % 94% 30% 25% 20% Municipal Division Active Groups with less than 25 members: Turnover for Males and Females 90% Confidence Interval Actual Expected 15% 10% 5% 0% Service 23

27 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS Service Exposed Actual Expected Actual Table II-4 Municipal Division Active Groups with 25+ members Turnover for Males and Females Expected Ratio: Actual over Expected Ratio: Actual over < % 20.0% 80% % 89% % 20.0% 72% % 80% 2 1, % 15.0% 80% % 86% 3 1, % 12.0% 90% % 90% % 7.0% 125% % 97% % 7.0% 129% % 100% % 6.0% 111% % 83% % 5.0% 158% % 105% % 5.0% 120% % 92% % 5.0% 87% % 87% 10+ 8, % 3.0% 143% % 107% Total 16,923 1,161 1, % 6.5% 106% 1, % 94% 30% 25% 20% Municipal Division Active Groups with 25+ members: Turnover for Males and Females 90% Confidence Interval Actual Expected 15% 10% 5% 0% Service 24

28 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS Service Exposed Actual Expected Actual Table II-5 Uniformed Division Active Groups with less than 25 members Turnover for Males and Females Expected Ratio: Actual over Expected Ratio: Actual over < % 14.0% 75% % 87% % 14.0% 54% % 63% % 12.0% 84% % 84% % 10.0% 92% % 102% % 6.0% 100% % 85% % 4.0% 88% % 71% % 3.0% 84% % 50% % 2.0% 272% % 109% % 2.0% 278% % 123% % 1.0% 296% % 74% 10+ 1, % 1.0% 359% % 119% Total 3, % 4.2% 123% % 92% 30% 25% 20% Uniformed Division Active Groups with less than 25 members: Turnover for Males and Females 90% Confidence Interval Actual Expected 15% 10% 5% 0% Service 25

29 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS Service Exposed Actual Expected Actual Table II-6 Uniformed Division Active Groups with 25+ members Turnover for Males and Females Expected Ratio: Actual over Expected Ratio: Actual over < % 13.0% 64% % 64% % 10.0% 0% % 0% % 7.0% 0% % 0% % 7.0% 119% % 119% % 6.0% 52% % 52% % 5.0% 0% % 0% % 4.0% 0% % 0% % 3.0% 83% % 83% % 3.0% 270% % 270% % 3.0% 0% % 0% % 3.0% 66% % 66% Total % 4.0% 59% % 59% 30% 25% 20% Uniformed Division Active Groups with 25+ members: Turnover for Males and Females 90% Confidence Interval Actual Expected 15% 10% 5% 0% Service 26

30 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS 3. Disability A. Current Assumptions Municipal Employees Current assumptions for municipal employees are based on age. The rates are 40% of the 1964 OASDI (Social Security) Experience for males. 15% of disablements are assumed to be service related. Uniformed Employees Current assumptions for uniformed employees are based on age. The rates are 60% of the 1964 OASDI (Social Security) Experience for males. 50% of disablements are assumed to be service related. B. Experience All Employees The study shows that the expected number of participants becoming disabled was lower than the actual. This is consistent with the results from the prior study. Due to the small number of disability incidences, disablements for all employees were reviewed in aggregate for municipal and uniform plans. A number of plans under PMRS provide separate benefit structures for service related disabilities. This is an even smaller sample size in terms of disability incident rates. It is assumed 15% of all municipal plan incidents and 50% of all uniform plan incidents of disability are service related, and we make no recommendation of changing these assumptions at this time. C. All Employees Although there are limited exposures, the rates of disablements at the later ages appear to provide credible data to warrant a change. We recommend changing the assumptions by lowering the disability incidence rates at a graded decrease from 0.02% reduction of the rates up to 0.27% reduction of the rates. D. Results The following tables and graphs compare three things; the number of people eligible to become disabled, the number of people expected to become disabled based on the current assumptions, and the number of people expected to become disabled based on the alternative assumptions. 27

31 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS Table II-7 Uniformed and Municipal Division Active Members Disability for Males and Females Ratio: Actual over Expected Ratio: Actual over Age Exposed Actual Expected Actual Expected < % 0.000% 0% % 0% % 0.026% 0% % 0% , % 0.034% 0% % 0% , % 0.048% 72% % 125% , % 0.069% 0% % 0% , % 0.104% 76% % 146% , % 0.171% 47% % 88% , % 0.283% 29% % 62% , % 0.453% 38% % 95% Total 39, % 0.241% 39% % 86% Uniformed and Municipal Division Active Members: Disability for Males and Females 0.6% 0.5% 90% Confidence Interval Actual Expected 0.4% 0.3% 0.2% 0.1% 0.0% Age 28

32 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS 4. Mortality A. Current Assumptions All Healthy Participants Males: RP-2000 Combined Healthy Mortality table setback one year Females: RP-2000 Combined Healthy Mortality table setback five years All Current or Future Healthy Inactive Participants RP-2000 Combined Healthy Mortality table for males and females Disabled Mortality: RP-2000 Combined Healthy Mortality table setforward 10 years for males and females B. Experience All Healthy Participants Deaths among active participants are typically small and may not provide meaningful statistics on pre-retirement mortality in a five-year period broken out between males and females. However, there were about 40,000 exposures for males (about 28,000 for actives; 12,000 for retirees) and 18,000 exposures for females (about 12,000 for actives; 6,000 for retirees) which provides a large enough sampling to analyze each group separately. The actual mortality rates were less than the expected for active males and females. For retirees, the actual mortality rates were higher than the expected rates for males but generally in line with expectations for females. For the active females, the data for the mortality experience is not credible due to the sample size of actual deaths that occurred. All Disabled Participants Mortality for disabled participants gives us an even smaller group to analyze actual versus expected experience. However, based upon the data, the actual mortality among disabled participants was lower than expected. Over the five year period of this study, the number of disabled male and female deaths was 19 and 6, respectively, illustrating the lack of credible data for this analysis. C. s All Healthy Participants Similar to the prior study, there is not enough data to statistically create an entirely new mortality table. However, the data is large enough to use a current mortality table and adjust accordingly to 29

33 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS the current mortality experience. We recommend updating the current tables with projected mortality improvements. Furthermore, the RP-2000 mortality table for current active employees can be applied to the active population and the retiree table can be applied to the inactive population. All active males and females would use the RP 2000 active employee mortality table with 15 year AA mortality improvement projections using the sex-distinct factors. For the active females, the five year set-back would continue. For the inactive participants, the males and females would use the RP 2000 annuitant morality table with five year mortality improvement projections for males and 10 year mortality improvement projections for females. The proposed experience for all participants is shown in the section D below. All Disabled Participants For the disabled participants, due to the limited credible data we recommend making no changes to the current mortality table. D. Results The following tables and graphs compare three items; the number of people exposed to the mortality assumption, the number of people expected to die based on the current assumptions, and the number of people expected to die based on the alternative assumptions. As you can see, the alternative assumptions bring the ratios to at least 100% of the actual experience for the active males and females, which means the number of people we expect to die is not less than the actual number of participants who died during the period studied. For the retirees, the ratios of actual over alternative assumptions are greater than 100%, showing that to some extent future mortality improvements are anticipated. Due to the lack of credible data for the disabled mortality deaths, we are providing a table on the number of actual versus expected disabled deaths. 30

34 Age Exposed Actual Expected PENNSYLVANIA MUNICIPAL RETIREMENT SYSTEM SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS Table II-8 Uniformed and Municipal Divisions Active Members Mortality for Male Members Actual Expected Ratio: Actual over Expected Ratio: Actual over < % 0.0% 0% % 0% % 0.0% 0% % 0% , % 0.0% 189% % 193% , % 0.0% 280% % 267% , % 0.1% 45% % 44% , % 0.1% 0% % 0% , % 0.2% 115% % 132% , % 0.2% 82% % 108% , % 0.4% 101% % 143% , % 0.7% 42% % 60% % 1.4% 48% % 93% Total 27, % 0.3% 73% % 100% 1.5% Uniformed and Municipal Divisions Active Members: Mortality for Male Members 90% Confidence Interval Actual Expected 1.0% 0.5% 0.0% < Age 31

35 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS Table II-9 Uniformed and Municipal Divisions Active Members Mortality for Female Members Age Exposed Actual Expected Actual Expected Ratio: Actual over Expected Ratio: Actual over < % 0.0% 0% % 0% % 0.0% 0% % 0% % 0.0% 0% % 0% % 0.0% 0% % 0% % 0.0% 0% % 0% , % 0.1% 130% % 156% , % 0.1% 128% % 158% , % 0.1% 36% % 47% , % 0.2% 95% % 117% , % 0.3% 66% % 74% % 0.6% 106% % 147% Total 11, % 0.2% 82% % 101% 1.5% Uniformed and Municipal Divisions Active Members: Mortality for Female Members 90% Confidence Interval Actual Expected 1.0% 0.5% 0.0% < Age 32

36 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS Age Exposed Total Actual Deaths Table II-10 Uniformed and Municipal Divisions Retired Members Mortality for Male Members Expected Deaths Actual Expected Ratio: Actual over Expected Deaths Ratio: Actual over % 0.2% 0% 0 0.2% 0% % 0.3% 606% 2 0.5% 327% , % 0.5% 191% 6 0.6% 158% , % 0.9% 186% % 187% , % 1.6% 141% % 147% , % 2.7% 135% % 147% , % 4.7% 119% % 127% , % 7.9% 124% % 127% % 15.6% 108% % 115% Total 12, % 3.4% 127% % 133% 18% 15% 12% Uniformed and Municipal Divisions Retired Members: Mortality for Male Members 90% Confidence Interval Actual Expected 9% 6% 3% 0% Service 33

37 SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS Total Actual Deaths Table II-11 Uniformed and Municipal Divisions Retired Members Mortality for Female Members Expected Deaths Expected Ratio: Actual over Expected Deaths Ratio: Actual over Age Exposed Actual % 0.1% 2000% 0 0.1% 2379% % 0.2% 526% 0 0.2% 494% % 0.4% 0% 1 0.4% 0% % 0.7% 142% 7 0.7% 138% , % 1.2% 153% % 157% , % 2.1% 74% % 78% % 3.4% 110% % 118% % 5.6% 96% % 103% % 11.5% 104% % 116% Total 5, % 3.0% 106% % 114% 18% 15% 12% Uniformed and Municipal Divisions Retired Members: Mortality for Female Members 90% Confidence Interval Actual Expected 9% 6% 3% 0% Service Table II-12 Uniformed and Municipal Divisions - Disabled Mortality Males and Females Combined Exposed Actual Expected

38 5. Marriage Percentage and Spouse s Age A. Current Assumptions SECTION II ANALYSIS OF DEMOGRAPHIC ASSUMPTIONS All Municipal and Uniformed Employees Currently, it is assumed that 85% of active employees are married at the time they retire and husbands are three years older than their wives. B. Experience All Municipal and Uniformed Employees The experience shows that 67% of current retirees are currently married. The inclusion of children as beneficiaries brings the percentage of participants with spouses or children as beneficiaries are 73%. Husbands are three years older than their wives. C. All Municipal and Uniformed Employees We suggest changing the percentage married to 80% and maintain the three year age difference. 35

39 SECTION III ANALYSIS OF ECONOMIC ASSUMPTIONS We considered the following to be economic assumptions in our analysis: 1. Inflation/Cost of Living Rate (if applicable) 2. Investment Return/Discount Rate 3. Salary Increase 4. Expenses Both the investment and salary increase assumptions are interrelated with the inflation rate, which can also be considered as the cost of living increase rate. The rate of investment return consists of two components: the "real rate" of return and the inflation component. Similarly, the rate of salary increase is separated into different components: the inflation rate, a merit increase (seniority) and, sometimes, component set aside for "productivity" gains. In developing recommendations for these assumptions, several factors are considered: 1. Inflation o historical data in general (i.e. the markets), o historical experience of the plan, o outlook for the future, and o assumptions used by other public sector plans. A. Current Assumptions The inflation rate is an underlying aspect of all economic assumptions. It is also the assumption used to project cost-of-living benefits for those municipalities that provide this benefit. The difference between other economic assumptions relative to the long-term underlying rate of inflation is an important measure. The current rate of inflation is 3.0%. The regional rate of inflation over the last five and ten year periods has been between 2.0% to 2.3% based upon the Consumer Price Index for All Urban Consumers (provided below). However for actuarial valuations we project the expected inflation out to longer periods including 20 to 30 years. The range of inflation in those terms is between 2.4% to 2.8%. Consistent with other changes being made to the salary scale assumption and the sensitivity of this assumption to the cost of living adjustment valuation, we would suggest considering moving this assumption to 2.8%. This would be consistent in part with the change in the discount rate assumptions from 6.00% to 5.50% in 2013 although the discount rate assumption was changed for reasons beyond the recent inflation experience. 36

40 SECTION III ANALYSIS OF ECONOMIC ASSUMPTIONS B. Experience 1. Historical Experience in General Based on the Consumer Price Index for All Urban Consumers U.S. City Average (all items), Table III-1 on the next page shows the inflation rates for the past 30 years: Table III 1 All Urban Consumers U.S. City Average (CPI-U ) Year Ending December 31 Increase in CPI-U % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % Compound Averages % % % % % % 37

41 2. Other Public Sector Plans SECTION III ANALYSIS OF ECONOMIC ASSUMPTIONS, The National Conference on Public Employers Retirement Systems (NCPERS) November 2014 study based on the 2014 survey of large public sector retirement systems shows an average inflation assumption of 3.2%, down from 2013 when the average was 3.5%. This can be seen by the graphic taken from this report below. This shows that the 3.0% assumption is still the prevalent rate, but the average is continuing to trend lower. Given the declining investment assumptions and the timing of this data, long term inflation rates below 3.0% can be anticipated given the continued low inflation environment. 38

42 SECTION III ANALYSIS OF ECONOMIC ASSUMPTIONS C. Based on the above factors, lowering the inflation rate to 2.80% should be considered in support of the other economic changes discussed in this report. 2. Investment Return/Discount Rate A. Current Assumptions PMRS assets are assumed to earn 5.50% net of expenses rate for the measurement of liabilities. The asset allocation has historically been developed to provide for achievement of at least this rate 90% to 95% of the time which is an important consideration given this rate is credited to each of the participating municipalities and counties asset account. Given the recent change from 6.0% to 5.5%, we do not recommend any additional changes at this time. However, the board may want to undertake a more comprehensive review of this assumption and the analysis and parameters required to be reviewed in setting this assumption in the future, especially in light of the unique structure of the PMRS system. We provide below additional information in support of the current trends where we are seeing regular reductions of the interest/discount rate. B. Experience 1. Historical Experience in General Table III-2 provides the rates of investment returns experienced by PMRS during the last 10 fiscal years. of return were computed as the ratio of the net investment earnings to market value of asset. Current Assumption: 5.50% per annum 39

43 SECTION III ANALYSIS OF ECONOMIC ASSUMPTIONS Table III 2 Net Investment Returns on Market Value of Assets Year Ending December 31 Return % % % % % % % % % % % Compound Averages up to December 31, 2014 Last 5 Years ( ) 9.79% Last 10 Years ( ) 6.55% Last 15 Years ( ) 5.84% The investment returns on a five and 10 year basis are significantly higher than the current assumption. However, historical returns should not provide the sole basis for setting the return assumptions. Historical returns are highly dependent on the period examined. Different periods can yield significantly different historical returns. Long-term investment return expectations on assets should not be the sole measure used in determination of the value of liabilities under PMRS. The higher this assumption, the greater the risk of understated liabilities and costs to PMRS will increase in the future. 2. Other Public Sector Plans The findings from the Public Fund Survey as of October 2014 performed by the National Association of State Retirement Administrators (NASRA) show that the median investment return used by public sector plans is 7.75%, a decrease from 8.00% as of the prior study. 40

44 SECTION III ANALYSIS OF ECONOMIC ASSUMPTIONS PMRS 0 Distribution of Investment Return Assumptions % < 7.0% 7.00% 7.0% % % % % 8.00% % % While PMRS rate is below all of the other systems in this study, PMRS is unique in that it credits this rate of return for participating municipalities. Because PMRS is in affect insuring an investment return it is essential this assumption be considered on a regular basis in terms of the long term risk of the assumption and the capacity of PMRS to continue to provide this level of return to its members. According to the investment manager, there is a 90% probability that the assets will provide a 5.5% rate of return during the year. Finally, since the NASRA study was completed, there continues to be a downward trend in the investment assumption used by many large public sector plans. 41