H.F. 844 (Smith); S.F. 667 (Pogemiller): Various Plans; Making Current Full Actuarial Value Method and Various Service Credit Purchase Provisions Permanent Rather than Temporary, does the following:

1. The following provisions in Teacher Retirement Association (TRA) and first class city teacher fund association law will be permanent provisions of law rather than expiring on May 16, 2003. These are the full actuarial value payment provisions governing:

• military service;

• prior-out-of-state teaching service;

• maternity leaves and maternity-related breaks-in-service;

• private, parochial, and quasi-public teaching periods;

• Peace Corps and VISTA service;

• charter school teaching periods;

• uncredited part-time teaching section of law (found only in first class city teacher plan law); and

• a provision of law which permits use of college supplemental fund to purchase service credit under all these provisions
  

2. The parental and family leave full actuarial value service credit purchase provision is made permanent rather than expiring on May 16, 2003. The pension plans covered by this provision are:

• General State Employees Retirement Plan of the Minnesota State Retirement System (MSRS-General);
• State Correctional Employees Retirement Plan (MSRS-Correctional);
• General Employee Retirement Plan of the Public Employees Retirement Association (PERA-General);
• Public Employees Police and Fire Retirement Plan (PERA-P&F);
• Teachers Retirement Association (TRA);
• the first class city teacher funds (DTRFA, MTRFA, SPTRFA);
• Minneapolis Employees Retirement Fund (MERF);
• Minneapolis Police Relief Association (MPRA); and
• Minneapolis Fire Relief Association (MFRA)

3. The methodology currently used (found in Minnesota Statutes, Section 356.55) to compute the full actuarial value service credit purchase prices for all the above provisions of law is made permanent rather than expiring on July 1, 2003.

The parental and family leave provision, which applies to members of MSRS-General, MSRS-Correctional, PERA-General, PERA-P&F, TRA, the first class city teacher funds, MERF, MPRA, and MFRA, is coded as Section 356.555. This provision permits active members of the applicable plans to purchase up to five years of service credit by making the required payment computed under section 356.55. Purchases are permitted for family leaves or family-related breaks-in-service, after which the individual returns to the same employer or to an employer covered by the same pension plan.

The TRA and first class city teacher fund association full actuarial value payment provisions governing military service, prior-out-of-state teaching service; maternity leaves and maternity-related breaks-in-service; private, parochial, and quasi-public teaching periods; Peace Corps and VISTA service; charter school teaching periods; uncredited part-time teaching, and a provision of law which permits college supplemental funds to be used to purchase service under these provisions, are coded in TRA law as Sections 354.533 through 354.539. The comparable provisions in first class city teacher plan law are coded as Sections 354A.097 to 354A.104, and Section 354A.106.

To purchase service credit under the above provisions, the individual must be an active member of the applicable Minnesota public pension fund association, must be vested (have at least three years of service credit), and have one or more periods of the type of service or breaks-in-service (in the case of periods to provide military service, or breaks-in-service due to maternity) covered under the applicable provision. To receive service credit under any of these provisions, the eligible individual must pay to the applicable pension fund the purchase price that is computed by the pension fund administration using the method in Section 356.55. The individual must provide any documentation necessary to determine eligibility to purchase service credit for the applicable period or periods. Some of these provisions further stipulate that the service credit purchase cannot be made if the period of time is already covered by another pension plan.

The provisions permit individuals to purchase considerable amounts of service credit, with purchases of up to ten years of service credit typically being permitted. Exceptions are the maternity leave/break-in-service provision, the military service credit provision, and the previously uncredited part-time teaching service provision. The maternity leave/break-in-service provision permits a maximum of a five-year service credit purchase, and the previously uncredited part-time teaching service provision (which is found only in first class city teacher law) puts no limit on the amount of service that can be purchased. The military service credit provision limits the purchase to the initial period of induction or call to active duty without any voluntary extension. Also, the individual cannot purchase service credit under the military provision if the individual is entitled to receive a military pension, or if service credit for the period was already purchased in another plan.

In 1998, the Legislature passed a revised full actuarial value methodology, which is coded as Minnesota Statutes, Section 356.55, although the provision had an expiration date. This is the method to compute full actuarial value service credit purchase prices that would now be made a permanent provision rather than a temporary provision. The methodology is used for all purchases of service credit in Minnesota public pension plans unless another approach is specified in the applicable special or general law. The provision temporarily replaced another method for computing full actuarial value service credit purchases, which is coded in statutes as Section 356.551. The newer provision was requested by various pension funds, particularly TRA and various other teacher funds, which contended that the prior approach produced price estimates that were too high, unjustly discouraging service credit purchases. Milliman USA, the actuary retained by the Legislative Commission on Pensions and Retirement (Commission), played a key role in developing the revised methodology, but on behalf of the pension fund administrations rather than at the direction of the Commission.

The revised full actuarial value methodology in Section 356.55 includes a requirement (Subdivision 7) that the Commission-retained actuary must include in a report to the Commission information on service credit purchases. The information must include the purchase price paid and the resulting increase in the pension plan’s liability due to the purchase as reflected in the next actuarial valuation. The actuary provides this information in a table in the Summary of Valuations, provided annually to the Commission. The applicable tables appeared in the 2000, 2001, and 2002 actuarial valuation summaries.

The information in those tables was intended to permit the Commission to determine whether the revised full actuarial method is producing reasonably accurate results. If the Commission was satisfied with the results, the Commission could recommend a repeal of the expiration date, making the provision permanent. When Section 356.55 passed in 1998 it included a July 1, 2001, expiration date. The expiration date was later extended, first to July 1, 2002, and then to July 1, 2003. However, the Commission has not conducted a thorough review of the procedure’s results. The current bill H.F. 844 (Smith); S.F. 667 (Pogemiller): Various Plans; Making Current Full Actuarial Value Method and Various Service Credit Purchase Provisions Permanent Rather than Temporary, removes the expiration date from Section 356.55 and from numerous full actuarial value service credit purchase provisions, making them all permanent provisions. It is reasonable to review the results of the revised full actuarial value methodology before considering any action.

This memo in part summarizes the results of a Commission staff study of the revised full actuarial value methodology. Before beginning that data review, let us first consider the concept of a full actuarial value service credit purchase. This provides some standards by which to evaluate the results of the purchase of service credit prices produced using Section 356.55.

Every legislative session, the Commission reviews requests for service credit purchases in Minnesota public pension plans, either as special or general legislation. In considering these requests, the LCPR had been guided by the following statement in the Commission’s Principles of Pension Policy. The applicable principle, II.C.10, reads as follows:

If the situation involves Minnesota public or quasi-public employment, if the pension fund is not harmed by the purchase, if the purchase does not raise broader policy concerns, and if the purchase does not violate equity considerations, the Commission generally recommended that the bill be enacted. The comments in the statement about the appropriate purchase price describe the full actuarial value concept.

Following the passage of the revised full actuarial value methodology (Section 356.55) in 1998, the 1999 Legislature passed several general law service credit purchase provisions for teachers using the method in Section 356.55 to compute the price. Like the revised full actuarial value method, these service credit purchase authorization provisions had an expiration date. The original expiration date was in May 2002, but this was later extended to May 2003. These service credit purchase provisions have been discussed above.

All these general law service credit purchase provisions are inconsistent with the Commission’s policy statement and with typical requirements for service credit purchases. The conflicts with the pension policy statement are as follows:

1. The provisions apply to whole classes of individuals rather than requiring individual legislative review;
2. Some of the service permitted to be purchased has no connection to Minnesota (such as the out-of-state teaching provisions); and
3. Purchases are permitted for service that is not public service or quasi-public service (such as the purchases of service for parochial or private school teaching service).

Some reasons why the Legislature may have been willing to pass legislation that was inconsistent with prior policy are:

1. The provisions may have been viewed as temporary measures to deal with short-term employment situations; and
2. The Legislature was given repeated assurances that the provisions would cause no financial harm or windfall gains to the individuals or to the pension funds, because the purchases were to be at full actuarial value as computed under the revised methodology.

The Commission’s pension policy statement states that a service credit purchase should not harm the pension fund. If the purchase is to not harm the pension fund, and also not harm the individual, the price must be accurately computed. The price must be set to match the pension plan’s true additional liability created by the additional service credit the individual will receive. If the purchase price accurately reflects the true full actuarial value of the service credit purchase, no liability in excess of the purchase price will be shifted to the plan (which would occur if the purchase price is too low), and no windfall should be created for the plan (which would occur if the purchase price is too high, exceeding the additional liability that is created). Any windfall to the individual would harm the plan. Any harm to the individual would create a windfall to the plan.

The general notion of full actuarial value pricing is straightforward. The problem is that things are bound to go wrong in practice. Any flaw or bias in the procedure used to compute the full actuarial prices will create error. While it may be possible to further refine a flawed procedure, another source of error will never be overcome. To produce an estimate that proves to be correct over time, it is necessary to know the future with certainty. Since no one has a crystal ball, the calculations must be based on assumptions about the future which inevitably will prove to be incorrect. These will create unintended gains for some individuals and losses for others.

Computing a full actuarial service credit purchase price requires estimating what the full required reserves will be for an individual’s pension at the time of retirement with and without the additional service credit. The present value of the difference is the price of the full actuarial value of the service credit purchase. To make these estimates one must make guesses about future events. One must assume when the individual will retire, and for some individuals who purchase service credit that date may be many years into the future. The expected pension will depend upon individual’s high-five average salary at the expected time of retirement. Therefore, it is necessary to build into the full actuarial value calculation a salary increase assumption so that the estimated high-five average salary at the predicted time of retirement can be computed. It is also necessary to estimate future investment returns. This annual rate of return assumption is used in various liability estimates within the actuarial valuations and is used as a discount factor in a full actuarial value service credit purchase to obtain the price today of the additional annuity amount estimated to be received in the future. The full actuarial value method also assumes that there will be no changes in the pension plan benefit package prior to the individual’s retirement.

It follows that any estimate made today will prove to be incorrect when viewed in the future because reality will depart from the assumptions used in the calculations. Actual investment returns will differ from predicted returns. It will be rare to have individuals with salary increases over time that exactly match the salary increase estimates. The individual may retire before or after the expected date. Benefit improvements may occur prior to retirement. Benefit increases lead to an underestimate of the full actuarial purchase price. Even if all of the other assumptions used in the calculations were exactly met, the estimated liabilities will be based on the pension plan as it existed when the price estimate was provided. Any improvements in the pension plan made after the individual makes the purchase payment and before that individual’s retirement will create a purchase price that is too low. Due to the benefit improvements, the liabilities will be higher than previously estimated, creating a windfall for the individual.

A. Covered Funds, Nature of Data

In the sections that follow are conclusions about the accuracy of the full actuarial value method found in Section 356.55 given the data on service credit purchases found in the Commission-retained actuary’s Summary of Actuarial Valuations for fiscal years 2000, 2001, and 2002. This is a somewhat shorter version of the Commission staff’s study of this methodology, as found in a Commission staff memo dated February 26, 2003, titled "Review of Full Actuarial Value Service Credit Purchase Results, Fiscal Years 2000-2002."

The data cover service credit purchases made during fiscal years 2000-2002 in the following plans:

• TRA Teachers Retirement Association

• DTRFA Duluth Teachers Retirement Fund Association

• MTRFA Minneapolis Teachers Retirement Fund Association

• SPTRFA St. Paul Teachers Retirement Fund Association

• MSRS Minnesota State Retirement System

• MSRS-Correctional MSRS Correctional State Employees Retirement Plan

• PERA Public Employees Retirement Association

• PERA-P&F Public Employees Police and Fire Retirement Plan

• MERF Minneapolis Employees Retirement Fund

The information is grouped under two categories, "active member" and "retired member," although all the purchases were made by active members of the applicable funds. These terms refer to the status of the individual on the actuarial valuation date. The "active member" category includes all purchases during the year made by active members where the member remained active on the actuarial valuation date. The "retired member" category includes all purchases made during the year by an active member who had retired by the actuarial valuation date. The actuary’s tables also include a "deferred member" group, applicable in cases were the individual was active when the service credit was purchased but was in deferred status by the actuarial valuation date. There were few of these cases, and some plans had no individuals in that category during the three years under study. The "deferred member" category is not included in this analysis.

An example of results for a single service credit purchase follow. The example is a PERA active member service credit purchase made during the period ending June 30, 2000. This information is the first entry found in the Commission-retained actuary’s Summary of Actuarial Valuations-2002, in Table 1-E:

In the above example, the individual purchased 3.417 years of service. Like most of the service credit purchases, the individual did not purchase service credit in even, full-year increments. The purchase payment amount determined under Section 356.55 was $19,582, which is the amount the individual paid for the service credit as shown in the "employee payment" column (there was no employer payment). The next column indicates the change in accrued liability recognized in the pension fund due to that purchase. To compute that figure, the actuary determined the liability recognized in the plan on the actuarial valuation date for that individual including the service credit just purchased. The actuary also computed the liability that would have been recognized in the actuarial valuation related to that individual if the purchase of service credit had not occurred. The difference between these two amounts is the change in plan liability due to the service credit purchase as recognized in the subsequent actuarial valuation. This amount appears in the "change in liability" column. It would be reassuring if the change in accrued liability equaled the service credit purchase amount determined under Section 356.55. In this case, the change in liability was $26,747, but the purchase of service price was $19,582, resulting in a loss to the plan (indicated in the gain/loss column) of $7,165.

Because of the length of those tables (over 40 pages in total), all the service credit purchase data found in the actuary’s summary documents are not included with this memo. However, a few tables summarize results, beginning with Table 1 below. Results derived through the actuarial report approach do not well track the service credit purchase prices computed under Section 356.55. The indicated gain/loss is never zero, and the deviation, whether positive or negative, is often large.

Table 1 summarizes the information on all the service credit purchases that occurred in all the applicable plans during fiscal years 2000-2002. The information includes the total number of purchases that occurred, the number (and percent) that had either a gain or a loss when the purchase price is compared to the increase in liability in the next actuarial valuation due to the purchase. The information also includes the average payment, average liability change, and average gain or loss. All this information is provided for active and retired groups, as previously defined.

Table 1

Summary of Service Credit Purchase Data

The information in Table 1 indicates that TRA has far more service credit purchases than any other plan. In fact, TRA members made more than three times more service credit purchases than all other plans combined. For the three-year period, there were 612 TRA service credit purchases by individuals who remained active at the next actuarial valuation date, and 103 purchases by individuals who retired by the next actuarial valuation date, for a total of 725 purchases. The total number of purchases in all other plans combined was 230.

Table 1 also indicates that in no case did the service credit purchase amount exactly match the additional liability due to the purchase noted in the next actuarial report, and the amounts are rarely even close. The pattern observed is that most active purchases (purchases by individuals who remained active members at the next actuarial valuation) resulted in a gain to the plan (the added liability due to the purchase as recognized in the next actuarial valuation was less than the purchase price). The opposite result is seen for those individuals who purchase service credit and then retired by the next actuarial valuation (the "retired purchases"). When the purchase price is compared to the change in liability reflected in the next actuarial valuation due to the purchase, most retired purchases resulted in losses to the fund.

For the three-year period as a whole, this pattern of active purchase gains and retired purchase losses held for all funds.

• For TRA, for the period as a whole, 82 percent of the active purchases resulted in a gain to the fund compared to the additional liability due to the purchase recognized in the next actuarial valuation, while for the retired purchases most cases (62 percent) resulted in a loss to the fund. The typical active purchase caused a gain of $4,277 for the fund, while the typical retired purchase resulted in a $7,800 loss.
• For DTRFA, 58 percent of the active purchase were gains, while 67 percent of the retired purchases recorded losses.
• In the MTRFA, 74 percent of active purchases were gains for the fund, while 67 percent (two out of three) retired purchases were losses, creating an average loss for the retired group in that fund of $20,189 per case.
• For the SPTRFA, 60 percent of the active purchases were gains, while 80 percent of the retired purchases were losses, creating an average loss per retired purchase of $20,012.
• For MSRS, 62 percent of the active purchases were gains, and all retired purchases (there were only three) were losses. In that plan, the average loss per retiree purchase was $43,949.
• In MSRS-Correctional, where very few purchases occurred, both active purchases were gains and the one retired purchase was a loss.
• In PERA-General, 57 percent of active purchases were gains and 71 percent of retire purchases were losses, with an average loss per retiree purchase of $5,452.
• In PERA-P&F, 67 percent of the active purchases were gains and the one retired purchase was a $31,254 loss.
• In MERF, one active purchase occurred, which was recorded as a sizable gain to the fund ($39,857) and no retired purchases occurred.

Viewing each year separately rather than as a three-year total, there were a few cases where the "active gain/retired loss" pattern did not hold, but that is, in part, a consequence of a small sample size. A few purchases could significantly shift the averages. In PERA, there were more active losses than gains in 2000 and 2001, but most purchases occurred in 2002 and a strong percentage of the 2002 active purchases were gains (70 percent). In DTRFA, two-thirds of the active purchases during 2000 were losses, but there were only three active purchases in total. In SPTRFA, most active purchases in 2001 and 2002 were recorded as fund losses, but in 2000, 95 percent (19 out of 20) of purchases were gains, resulting in more active gains than losses for the three-year period as a whole.

B. Size of Reported Gains

Table 2 focuses on TRA service credit purchases where gains were reported. While this information covers only TRA, there is no reason to believe that the TRA results are not indicative of the results for other funds. And in examining TRA, the vast majority of all the service credit purchases that have occurred are examined. In some of the other funds the sample size is too small to support any general conclusions.

Table 2

Examination of TRA Gains

The first four columns of data in Table 2 provide information on all TRA gains by year and for the three years combined. There were 612 TRA active purchases in total, and 503 of those (82 percent) were noted as creating gains to the fund. If most of those gains were small, suggesting a close tracking between the purchase price and the liability that results from the purchase as indicated in the next actuarial valuation, it would seem to provide some assurance that the members purchasing the service credit are not being harmed.

However, most gains were large, not small. For active purchases:

• The table indicates that only 14 percent that were gains had purchase prices within $1,000 of the change in liability indicated in the next actuarial valuation due to that purchase.
• In 46 percent of the cases, the difference between the purchase price and the added liability recognized due to that purchase was between $1,000 and $5,000.
• In 24 percent of the cases, the difference was between $5,000 and $10,000.
• Eleven percent of the cases had differences between $10,000 and $20,000.
• In three percent of the cases, the difference was between $20,000 and $40,000.
• In two percent of the cases, the difference was more than $40,000.

For the retired group, although losses occurred far more frequently than gains, there were 39 cases where gains occurred over the three-year period. For some unknown reason, a smaller percentage of these retiree gains were comparatively small, under $5,000. For the active group for the three years as a whole, 60 percent of the gains were $5,000 or less (this results from combining the first two active groups noted above, the 14 percent of the cases with gain under $1,000 and the 46 percent of cases where the gain was $1,000 to $5,000). For the retired group, only 38 percent of the gains were less than $5,000. When recorded gains occurred they tended to be large. Twenty-three percent were between $10,000 and $20,000, 13 percent were between $20,000 and $40,000, and 18 percent were $40,000 or more.

The system Milliman USA uses to compute the actuarial valuations deals in full-year increments only. The actuary has stated that estimates of gains or losses may be distorted (overstated) whenever individuals purchase service credit that is not in full-year increments. Therefore, the last four columns of Table 2 examine TRA cases representing only full-year purchases.

When the gain situations are examined, using the full-year-only purchase data tends to increase variability rather than reducing it. Using full-year-only data, the sample size decreases considerably, from 612 active purchases to 176 and from 103 retired purchases to 15. For the active group, 14 percent of full-year-only purchases resulted in gains less than $1,000, the same percentage as occurred when all records were used, and 30 percent created recorded gains of $1,000 to $5,000, compared to 46 percent of total purchases. The percent of the gains in the $5,000 to $10,000 range actually increases to 34 percent for full-year-only purchases from 24 percent previously, and there continues to be a large percentage of even higher value gains, including six percent (six cases) in excess of $20,000.

For the retired group the full-year-only sample size is very small; only 15 purchases occurred and only seven of these were gains. There is a high percentage increase in high value gains in the $10,000 to $20,000 range, the $20,000 to $40,000 range, and above, although those percentages are based on very few cases.

C. Size of Reported Losses

Table 3 provides the same type of comparisons as Table 2, except that it deals with cases where losses to the fund occurred. For the active group using all data, most losses were clustered in the lower value ranges. Most of the losses were less than $10,000, and 61 percent of the losses were in the $1,000 to $5,000 range. For the full-year-only data, there again is a very high percentage of losses (69 percent) clustered at the $1,000 to $5,000 range.

In contrast, retired group losses are clustered in the highest value ranges. When all data are used, only 13 percent of the losses were less than $10,000, while 87 percent were losses of $10,000 or greater. For the full-year purchase data, all losses are at least $5,000 or more. The highest percentage of losses (38 percent) occurs in the $20,000 to $40,000 range, but the sample size is very small.

Table 3

Examination of TRA Losses

D. Interpretation of Results

Little consistency is observed between the full actuarial value payments computed under Section 356.55 and the results reflected in the actuarial valuation—the difference between the liability in the valuation for the individual with and without the service credit purchase. This may be due to flaws in the full actuarial value methodology or due to inconsistencies between the two approaches. It is difficult to determine the relative impact of these two factors.

Part of the observed differences is likely to stem from differences in the nature of the calculations. The full actuarial value methodology recognizes that this is a self-selected group. While for the plan’s covered membership as a whole, turnover assumptions (probabilities of terminating employment at each given age) are relevant, individuals who purchase service credit are expecting to remain with the fund for the long term and to draw an annuity upon retirement. Therefore, the full actuarial value calculation does not discount computed amounts by probabilities of turnover, and the procedure further specifies that no mortality decrements will be used (Section 356.55, Subdivision 2, paragraph (b), clause (6)). Computed amounts are discounted only by the assumed long-term investment rate of return, 8.5 percent annually, to obtain a present value. The actuarial valuation, in contrast, is based on the hypothetical average employee, with various probabilities of withdrawing from service at each age, and probabilities of death. The actuarial valuation liabilities with and without the service credit purchase presumably would be discounted by expected turnover, by mortality, and by the expected rate of return. The difference between the two calculations derived from the actuarial valuation (the computed present value of the liability with the service credit purchase, and the expected present value of the liability without that purchase, both heavily discounted) might tend to be less than the amount computed under the full actuarial value methodology.

If the results seen in the tables were largely due to the turnover and mortality decrement assumptions reflected in the actuarial value methodology but not in the full actuarial value methodology, then one would expect a greater difference the longer the period of years in which those decrement assumptions were used in a calculation. In other words, the younger the age of the purchasing employee (the longer the period prior to retirement), the more likely it is that the full actuarial purchase price would exceed the change in liability reflected in the actuarial valuation approach, resulting in a computed gain. Unfortunately, information in the actuary’s tables does not include information on age or length of service at the purchase date. What we do have is information based on the status of the individual on the actuarial valuation date. The active group members remained as active employees on the actuarial valuation date. The retired group members were retired on that date, indicating that they retired within a year of purchasing the service credit.

If this decrement issue is impacting results, it would be expected to see gains for the active group, and in earlier tables there was a tendency for active purchases to produce reported gains. However, if the decrement argument is valid, it is far from a complete explanation of the results. Flaws in the full actuarial value methodology and various other factors, whatever they may be, are also important. While gains were more likely on active member purchases, there are many individual cases where that did not occur, and for some funds in some years the active member purchases as a group produced a net loss rather than a net gain. Years where this occurred in DTRFA, PERA, and SPTRFA are noted above.

The retired member results are troubling. Differences between the actuarial valuation derived results and the full actuarial value purchase method, assuming decrements are an issue, should be disappearing near retirement, as probabilities of turnover approach zero and mortality, even near retirement, remains a negligible factor. One would expect similar results without any bias (neither a tendency to produce gains nor a tendency to produce losses). One would also expect less dispersion (smaller gains or losses when they do occur) than occurred in the active member results. However, that is not what is observed. When purchases occur within a year of retiring, the full actuarial value method tends to produces a loss compared to the results derived from the actuarial valuation. This raises a concern that the full actuarial value method in use is producing a purchase price that is too low. While that is the tendency noted in prior tables, there are also numerous cases of gains. The dispersion is unexpected and troubling. Gains and losses when they occur tend to be far larger than the typical active member gain or loss. For TRA retired cases, which were examined in detail in Table 3, 62 percent of retired group purchases for the three-year period as a whole resulted in losses. The average loss per case where losses occurred was $30,786. Of those losses, only 13 percent were less than $10,000. Forty-eight percent of those losses were in the $20,000 to $40,000 range, and 27 percent of the losses were greater than $40,000. Examining full-year purchases only did not significantly change any general tendencies, although sample size is quite small. For the three-year period, 53 percent of those purchases (eight cases) were losses, and none of those losses were less than $5,000. Three losses were in the range of $20,000 to $40,000, and one was in excess of $40,000.

E. Observations and Conclusions

Review of the full actuarial value purchases suggests a number of observations:

1. Lack of Consistency Between Actuarial Valuation and Full Actuarial Value Method. The most obvious conclusion is that results derived from the actuarial valuation and the full actuarial purchase price method do not agree. The approaches are not consistent. No cases were found where there was exact agreement. Differences tended to be large rather than small, whether or not the purchases are occurring just before retirement. The actuarial model used by the actuary, Milliman USA, has a shortcoming in that it cannot handle fractional years of service credit, only full years. Examination of only full-year service credit purchases is hampered by sample size because most purchases are for fractional years, but the results do not significantly change except that the dispersion appears to worsen.

2. Implications about the Fairness of the Method and Fairness of the Service Credit Purchase Provisions. The review of service credit purchases raises serious concerns that the method used to compute service credit purchase prices is not sufficiently accurate and is not providing fair outcomes, either to the individuals or the pension fund. The Legislature intended that individuals should pay for the service credit purchase an amount equal to the true additional liability created in the pension fund due to the purchase. There is no reason to believe that the current method is doing that.
   
   Given results presented here, the Commission may wish to consider whether it has sufficient confidence in the approach to continue its use. The Commission may also wish to consider whether to continue the numerous general law service credit purchase provisions in TRA, first class city teacher plan law, and laws involving other plans, which require an accurate determination of full actuarial value in order to produce a fair outcome. Fund members considering a purchase of service credit may also wish to make a similar assessment. Perhaps in some cases this system is treating them with reasonable fairness, but the data suggest that they might be undercharged or overcharged by large amounts, and the outcome may not be predictable in all cases by the member purchasing the service.

3. Other Related Public Pension Policy Issues Raised by the Review. In deciding whether to recommend that the full actuarial value methodology currently used (Section 356.55) and the various service credit purchase provisions in existing law should be made permanent, the Commission may wish to consider the very small percentage of plan members who purchase service credit. In TRA, which had by far the most service credit purchases, an average of 200 purchases occurred per year from 2000 through 2002. TRA has 71,700 active members. Thus, in any given year, considerably less than one percent of the active membership (0.27 percent of active members) purchased service credit. Those who did purchase service credit had the financial resources, often many tens of thousands of dollars, to make those purchases. The Commission may wish to consider whether it is good public policy to continue these provisions which are used by only a fraction of one percent of fund members. These purchases harm some purchasers while creating a windfall for others. The net financial impact of all these purchases, which could create a net loss to the fund, must be financed by the employers and the other employees covered by the plan, employees who do not have the resources to consider a purchase of service credit.
   
   The Commission may wish to consider that when individuals purchase service credit just before retiring, the data reviewed in this memo suggests that the purchase is likely to be subsidized (a gain to the employee, a loss to the fund).
   
   The Commission may also be concerned about purchases by younger members, individuals who are several years or many years from retirement. Despite some data in this report that suggests that the fund is likely to gain on those purchases, that may be an illusion. The longer before retirement that a purchase of service credit occurs, the more error is involved in estimating the purchase price. The reason is that any estimate must be based on assumptions about salary increases and investment returns which will prove incorrect over time. The longer those must be projected, the more likely an error—possibly of substantial proportions—will occur.

Finally, the Commission should be aware that the service credit purchase price is estimated based on the plan at the time of the purchase. By purchasing service, the individual gets free access to all future benefit improvements that will apply to that purchased service. If benefit improvements occur after the purchase (and that is increasingly likely the longer before retirement that a purchase occurs), the purchase price is too low and the individual receives a windfall, although it is impossible to detect that viewing data soon after the purchase occurs.

These bills raise several pension policy issues:

1. Implications of the Proposed Change. The Commission may wish to consider whether the method used to compute purchase prices is too faulty and the service credit purchase provisions themselves are too much in conflict with previous service credit purchase policy to warrant continuing these provisions.

2. Problems with the Computation Procedure. The bills would make permanent the method currently used (Section 356.55) to compute full actuarial value service credit purchase prices. The data suggest that method is not sufficiently accurate. The Commission’s pension policy statement requires that service credit purchases should not harm the pension fund or the individual. The review of the results from three years of service credit purchases suggests that these purchases often harm the individual or the fund. Review of the data suggests that the method often produces estimates that may be thousands of dollars too high or too low, and in many cases these differences can be tens of thousands of dollars.

On average, the method produces estimates of purchase prices by individuals about to retire which are too low, in many cases much too low. This suggests that a substantial subsidy is being provided by other fund members and employers.

On average, the method produces estimates of purchase prices by individuals further from retirement which appear to be too high, based on reported results. However, that result may not be credible due to the inability of the actuarial valuation procedures to properly account for liabilities likely to be imposed on the fund by the subgroup who purchase service credit. A related problem for this group is that the individuals receive free access on the purchased credit to future benefit improvements that occur prior to retirement. What looks like an overcharge based on review of the current data may prove to be a subsidy overtime.

3. Conflict of the Service Credit Purchase Provisions with Commission Principles Statement. The issue is the conflict between these service credit purchase provisions and the Commission’s Statement of Pension Policy Principles. Many of the purchase of service credit provisions enacted in 1999 or later are not consistent with those policy principles. It is possible that the Commission and the Legislature initially authorized these provisions, despite their conflict with longstanding policy, because they were temporary measures intended meet a short-term employment situation. As permanent provisions, the Commission may wish to consider whether the provisions are acceptable. The provisions are in conflict with established policy because they permit purchases by broad groups or categories of people without requiring review of the individual case, they permit service credit for periods which are clearly not public employment in nature (such as the provisions authorizing service credit purchase for private school or parochial school teaching), or have no connection to Minnesota (as is the case with purchases for out-of-state teaching service, or out-of-country service).

4. Impact on Commission’s Policy Statement. If these service credit purchase provisions, which are in conflict with the Commission’s policy statement, are made permanent, the Commission may need to revise its policy document. The Commission may find it difficult to deny any request from plan members in Minnesota public pension plans for service credit purchase authority, regardless of the nature of the service, if any, provided during the service credit purchase period.

5. Small Group Utilizing the Provision. The issue is whether it is good policy to have the employers and employees bear the risk for provisions utilized by a minute fraction of members. The data suggest that in any given year, considerably less than one percent of active pension plan members utilize these service credit purchase provisions. If these provisions are continued, that fraction is likely to decrease.

6. Conversion of Assets from Savings and Defined Contribution Plans to Defined Benefit Plans. The various service credit purchase provisions amount to an opportunity for members of those defined benefit plans to convert personal savings and defined contribution plan assets, where there are no guarantees, into a defined benefit plan asset that is guaranteed. The Commission may wish to consider whether that is appropriate. The Commission may also wish to be concerned that continuing this policy will encourage members of defined contribution plans, like the higher education Individual Retirement Account Plan (IRAP), or the PERA Defined Contribution plan, to seek legislation allowing them to transfer into defined benefit plans. Individuals are in effect purchasing an annuity from the state through the pension plan, rather than from a private insurance company. All investment risk is being shifted from the individual to the state taxpayer.

7. Various Risks Shifted to the Pension Plan. The provisions shift mortality risk and investment risks to the pension plans. Regarding mortality risk, the accuracy of the full actuarial valuation methodology depends upon an accurate prediction of life spans. If individuals live longer than expected, the service credit purchase price computed under the method will prove in the long term to be insufficient. The mortality assumptions may be reasonable estimates of current mortality conditions, but future life expectancy increases will cause pension funds to suffer mortality losses because benefit payments must continue until death. This risk that more assets may be needed to cover the benefit payouts through these extended lifetimes is understood and accepted. It is part of the risk borne by having a defined benefit pension plan to reward public employees for their service. What the Legislature may wish to ask is whether that risk should be accepted when it stems from service credit being purchased for periods having nothing to do with Minnesota public employment, as is the case with many of these service credit purchase provisions.

Also, net liabilities will be created in the pension funds from this service credit purchase authority due to the action of investment markets, and the way teachers and other public employees will react to the incentives that are created by those markets. The full actuarial value methodology assumes that the pension fund will earn an 8.5 percent annual return on the money after the purchase price is received. If the pension fund return is less than 8.5 percent annually from the date the purchase price is received through the date of the individual’s retirement, the purchase price will not be adequate to cover the fund’s added liabilities. If that occurs, the fund is subsidizing the purchase and the individual receives a windfall.

The responsibility for investing the assets, and all investment and mortality risk, is shifted from the member to the pension fund. Given that all individuals can be expected to act in their best financial interest, it is reasonable to expect many to pursue service credit purchases in the defined benefit plan when investment markets are weak. By making the purchase, the individuals will receive an additional annuity amount consistent with earning an 8.5 percent annual return on that money until retirement, and they have no worry of outliving that income stream—it will be paid until death. In a period of weak investment markets, this may leave them far better off financially than they can expect by investing on their own. The pension fund is likely to be harmed by these purchases because the fund will not be able to achieve an 8.5 percent return. In recent years, pension fund returns have been negative, not positive.

There will be less interest in making service credit purchases in periods of strong investment markets. If individuals expect to earn a return well above 8.5 percent by personal investing or other tax deferred investments, they will generally not be interested in purchasing an additional defined benefit plan incremental annuity that can be supported with only an 8.5 percent return. Therefore, the most reasonable expectation, if these service credit purchase provisions become permanent, is that there will be considerable interest in these purchases when markets are weak (when the fund is most likely to be hurt by the purchases) and less interest in these purchases when markets are strong (when the fund may be able to realize a net gain on the purchases). There is no reason to expect these periods to offset each other. The most likely outcome is that over time, these service credit purchase provisions will not be cost neutral—they will add to plan contribution requirements.

Another risk to the fund is the risk of benefit improvements. The purchase price cost estimate is based on the cost of the benefits offered by the plan at the time of the purchase. If the plan is enhanced after that date but before the individual retires, the purchase price will not cover the full liabilities that ultimately are created.

8. Encouraging Purchases by Permitting Tax-Free Transfers. The current bills would make permanent rather than temporary provisions of law which allow MnSCU members to use assets from their supplemental fund accounts to purchase service credit in defined benefit teacher plans. These tax-free transfers encourage service credit purchases. The issue is whether the Legislature should encourage these purchases. If the Commission feels that the full actuarial value service credit purchase should not be encouraged, then the transfer provisions should be permitted to expire.

9. Permitting Purchases of Service Given Current Pension Fund Actuarial Condition. For several reasons outlined in this memo, the service credit purchase provisions are likely to harm the pension funds. The issue is whether the service credit purchase provisions should be permitted to expire due to the financial condition of the defined benefit pension plans. Some of the plans are in a weak condition to absorb any harm. PERA-General is considerably less than fully funded. TRA appears to have a funding ratio greater than 100 percent in current actuarial work, but that is an illusion. The result stems from an overoptimistic estimate of assets caused by the actuarial value of asset methodology in current use. All defined benefit pension plans have been seriously hurt in recent years by the financial markets, which have been providing negative returns. When market value rather than actuarial value is used, TRA is not fully funded. It actually has an unfunded liability. The MTRFA, in particular, is in weak financial condition and is less able than TRA to bear the net liabilities that can be created by these provisions and the full actuarial value service credit purchase price estimation method. The MTRFA is 65 percent funded based on actuarial asset value. It is only 46 percent funded based on market value.

10. Interaction with Benefit Increase Proposals. Last year TRA and first class city teacher plan administrators and various teacher union officials worked on a proposal to merge TRA and the three first class city teacher fund associations. If a merger occurs in the future, it is likely that teachers and possibly other public employees will seek a substantial benefit improvement as part of that package. If employees anticipate that a benefit improvement will occur, members will seek to purchase service credit prior to the effective date of the improvement in order to receive the windfall from the service credit purchase.

This final section outlines alternate actions that the Commission may wish to consider. Several amendments are provided. Commission staff can prepare others at the Commission’s direction.

In considering the current bills, the Commission may wish to also address an apparent oversight that occurred during the 2001 First Special Session. During that session, a provision was added to TRA law (Section 354.541) and a comparable provision to first class city teacher plan law (Section 354A.109). These two provisions authorize members of the applicable teacher plan, if the individuals previously taught at the University of Minnesota without retirement coverage, to purchase up to ten years of service credit in TRA or the applicable first class city teacher plan by paying the full actuarial value price estimated using Section 356.55. These two provisions do not have a sunset date. These provisions are comparable to all other full actuarial value service credit purchase provisions which are set to expire under current law on May 16, 2003. The parental or family leave provision passed in the same bill as the University of Minnesota service provisions (Laws 2001, First Special Session Chapter 10), and the parental/family leave provision does have a May 16, 2003, expiration date. Since all similar provisions passed as temporary provisions with an expiration date, it appears that the lack of an expiration date on the University of Minnesota service provisions is an error.

Action 1: Because the Commission may not have time to fully consider the full actuarial value study, the Commission may wish to consider an amendment which would revise the bills to extend all the provisions covered by the bills for an additional one-year period. The Commission could then study full actuarial value method issues over the interim, and consider any further necessary action next year. LCPR03-103 extends all provisions covered by the bills for an additional year, rather than expiring during 2003, and it adds a May 16, 2004, repeal date to the two University of Minnesota teaching service purchase of service credit provisions (Sections 354.541 and 354A.109).

Action 2: Recommend the bills to pass without amendment. If the Commission chooses to recommend that H.F. 844 (Smith); S.F. 667 (Pogemiller) should pass without amendment and the provisions are enacted into law, the revised full actuarial value estimation method (Section 356.55), the various general law service credit purchase provisions discussed in this memo, and the provisions which permit use by MnSCU faculty of their supplemental fund account to purchase service credit under the general law service credit purchase provisions all become permanent rather than expiring during 2003.

Action 3: Recommend that the bills should not pass. The effect will be that the revised full actuarial value method, the various general law service credit purchase provisions included in the bills, and the provisions authorizing use of supplemental fund assets to purchase service credit will expire during 2003. After July 1, 2003, any full actuarial value service credit purchase under any general law provision that remains in law, or any service credit purchase authorized by the Legislature through special legislation, will be computed using the prior full actuarial value computation method, coded as Section 356.551. It generally produces a purchase price that is as high or higher than the current method, lessening the extent of subsidies.

Action 4: Using amendment LCPR03-060, revise the bills to extend all the provisions covered by the bills for a period to be determined by the Commission.

Action 5: Using amendment LCPR03-061, revise the bills to extend all the provisions covered by the bills for a period to be determined by the Commission, except for the revised full actuarial value estimation method, which is permitted to expire. (This causes any full actuarial value service credit purchase payment to be computed under the methodology in Section 356.551.)

Action 6: Using amendment LCPR03-104, the revised full actuarial value estimation method and the military break-in-service provisions are extended for one year, while the remainder of the provisions are permitted to expire. A repeal date of May 16, 2004, is added for the two University of Minnesota teaching service purchase of service credit provisions (Sections 354.541 and 354A.109).