Invisible pension investments.
Low-asset DB plans make some significant use of insurance companies as financial intermediaries--either through a separate account investment (PSA) or a guaranteed interest in an insurer's general account--but reliance on insurance steadily declines as the level of plan assets increases. Insurance, and in particular PSA investment, is markedly more important to DC plans than to DB plans. Also unlike DB plans, DC plan utilization of insurers' general accounts remains fairly steady rather than markedly declining with increases in total plan assets. The allocation of a constant proportion of the portfolio to general account investments, regardless of DC plan asset level, might be explained by the popularity of guaranteed investment contracts (GICs), which are commonly offered as a stable value fund investment option under now-ubiquitous participant-directed 401(k) plan designs. (92)
The utilization of 103-12 IEs by either DB or DC plans is very limited. That observation might indicate that pension plans make realty and other investments through partnerships in which plan holdings are kept below twenty-five percent, thereby avoiding characterization of the partnership's property as plan assets. (93)
The broad categories of defined benefit and defined contribution plans mask large variations in plan characteristics. Defined contribution plan, for example, is defined by reference to whether benefits are based solely on the amount of contributions credited to an individual account maintained on behalf of the employee, plus any income, expenses, gains, losses, and forfeitures (of the accounts of other participants, if applicable) which may be allocated to the account. (94) DC plans are classified into a number of subsidiary types, according to the method of determining contributions and (sometimes) the nature of plan investments. The most common variety of DC plan today is the 401(k) plan, under which employee-participants may elect to contribute a portion of their wages or salary to the company's retirement savings program on a pre-tax basis, rather than receiving it as current (taxable) cash compensation; the employer typically contributes as well, either by matching all or part of the employee's elective deferral or by making non-elective contributions (i.e., contributions to all workers eligible to participate, regardless of whether or how much each chooses to defer). In contrast, under a money purchase pension plan (MPPP) the employer promises to make specified annual contributions to each participant's account in an amount that is commonly set as a percentage of the employee's current compensation (and which is independent of firm profits). Are these different DC plan types associated with significant differences in a plan's utilization of indirect investment vehicles?
Figure 5 presents the results for large 401(k) plans in 2008. The pattern is extremely similar to Figure 4, the composite picture for all DC plans. That result is not surprising, because in recent years the DC plan universe has come to be dominated by 401(k) plans. By 2008 about seventy-five percent of all DC plans were of the 401(k) type, and the proportion of active participants was about the same (see Figures 8 and 9). The authors also computed utilization rates of the same six types of independent investment vehicles by large profit-sharing plans and by large DC plans which allow participants to direct the investment of their accounts (not shown). In each case the results are essentially identical to the 401(k) plan results. The 401(k) elective contribution feature is usually a component of a profit-sharing plan, and over the last fifteen years most 401(k) plans sponsored by large companies have been amended to give participants control over the investment of their accounts, typically by allowing them to select among a menu of mutual fund investment options that offer a broad range of risk and return characteristics. Accordingly, the connection between 401(k) elective contribution features, profit-sharing plans, and participant-directed investments is so tight that one would expect the common pattern of utilization of indirect investment vehicles that is actually observed.
The money purchase plan is a different creature entirely. Figure 6 shows the extent of reliance on indirect investment vehicles by large MPPPs in 2008. Here the pattern is subtly but noticeably different. As with DC plans overall (compare Figure 4), hefty reliance on mutual funds is the norm, and with a dramatic drop in the top decile. Mutual fund usage by MPPPs--unlike DC plans in general or 401(k) plans--actually falls below master trust investments for plans in the largest-asset
category. Most significant is the surprisingly large reliance of MPPPs on insurance company general account investments (in all asset ranges except the top decile), giving insurer-intermediaries an importance in this sector of the pension plan universe that they do not have elsewhere. Money purchase plans are a dying breed, apparently as a result of changes in the limits on deductibility of contributions enacted in 2001, which eliminated employers' incentive to offer a MPPP in addition to a 401(k) or other profit-sharing plan. (95) (See Figure 8.) As the rapid demise of the MPPP overlaps the period of the EFAST filing system, the authors also looked for changes in the utilization of indirect investment vehicles by large MPPPs between 2000 and 2008. Although 4135 large MPPPs filed returns reporting non-zero assets in 2000, compared to only 2289 in 2008, the utilization pattern of indirect investment vehicles in 2000 (not shown) is very similar to Figure 6.
The cash balance plan is a DB subtype that is often referred to as a hybrid plan because the promised benefit is a lump sum payment based on a specified percentage of the participant's annual compensation (pay credits) augmented by an assumed rate of return (interest credits), and so mimics the yield of a DC plan (specifically, a MPPP). In contrast, most DB plans--often referred to as "traditional" pension plans--promise benefits in the form of a life annuity, the amount of which is specified by a formula that typically takes into account some measure of the worker's average compensation and length of service. The resemblance between cash balance and MPPPs raises the question whether cash balance plan utilization of indirect investment vehicles is more like the pattern for DB plans or MPPPs. Figure 7 displays the answer. Perhaps surprisingly, in the aggregate cash balance plan indirect investments are quite similar to DB plans generally (compare Figure 3), and notably dissimilar to the MPPP pattern (Figure 6). (96)
E. DFE Asset Holdings
Having investigated the general patterns of pension plan utilization of indirect investment vehicles, we now briefly review the general patterns of investment allocation by DFEs themselves. Aggregate descriptive statistics concerning asset holdings of the four types of DFEs were compiled by extracting from the Department of Labor's raw data files all annual reports by DFEs. (97) DFE-type codes were used to categorize the filings, (98) and Schedule H balance sheet data for each type of DFE were combined.
A simple comparison of the total net assets held in each type of DFE (i.e., MTIA, CCT, PSA, or 103-12 1E) with the aggregate amounts reported by all large pension plans as the year-end value of interests held in DFEs of the same type reveals some fundamental facts. Virtually all MTIA assets arc attributable to large pension plans; small plans (meaning plans with less than 100 participants) do not utilize MTIAs to any significant extent. (99) This fact is consistent with the finding discussed earlier that the utilization of MTIAs by "large" plans increases dramatically with the size of total plan assets (see Figures 3 and 4). The net assets of the other three varieties of DFEs substantially exceed the reported value of large pension plan interests. Large pension plans account for sixty-five percent of PSA net assets in 2008; the remaining thirty-five percent is apparently attributable to PSA interests held by small plans or governmental plans. (100) Unlike MTIAs and PSAs, ownership of interests in CCTs and 103-12 IEs is not limited to employee benefit plans. (101) Not surprisingly, therefore, interests held by large pension plans account for only a minority of the net assets of these two types of DFEs: specifically, large plans hold in the aggregate about a forty percent stake in all CCTs, and only a twelve percent interest in 103-12 IEs. (102) Small plans might own some of the remaining interests, but probably very little, because CCT assets not attributable to large plans greatly exceed the total net assets of all small plans in 2008, (103) while as seen earlier, interests in 103-12 IEs reported by large plans are concentrated in the very highest asset plans (Figures 3 and 4).
Somewhat surprisingly, given the dominance of MTIA investments by large plans (see Figures 1-2), in 2008 the largest number of DFE filings came from CCTs, (104) and CCTs also reported the greatest total assets ($2.13 trillion versus $1.28 trillion held in MTIAs). The higher CCT asset total is apparently attributable to two factors. (105) First, recall that a CCT that is a common trust fund is not restricted to holding assets from qualified plans, but may also contain funds attributable to private trusts, estates, and UGMA custodianships. (106) Second, CCT asset totals seem to be afflicted by substantial double counting, due to the fact (discussed below) that CCTs invest heavily in other CCTs. If CCT investments in other CCTs are subtracted, CCT total assets in 2008 are reduced to $1.47 trillion, while eliminating MTIA investments in other MTIAs only reduces MTIA total assets to 1.26 trillion.
Figure 10 presents a side-by-side comparison of the proportionate asset allocation of the four types of DFEs among the investment categories reported on Schedule H. (107) (Those categories of investments in which no type of DFE invested at least one percent of total assets are not displayed.) Consider first the investment by DFEs in other DFEs. As would be expected, some MTIA funds (only two percent) are invested in another MTIA, but because a master trust may pool assets only from plans sponsored by either a single employer or a group of commonly-controlled employers, other types of DFEs, which pool assets from unrelated employers, do not put funds in an MTIA. (108) While cross investment in other DFE varieties is not restricted, it is noteworthy that the only substantial DFE investment in PSAs is by other PSAs, and likewise the only substantial DFE investment in 103-12 lEs is by other such investment entities. (109) Because banks and insurance companies compete as financial intermediaries, the absence of cross investment between CCTs and PSAs may come as no surprise.
CCTs, however, are a special case--they place a large part of their funds in other CCTs (about thirty-one percent in 2008), and MTIAs also entrust a big chunk of their total assets to CCTs (about eighteen percent). (110) This link might be attributable to the fact that a bank must serve as trustee or custodian of a master trust, while a CCT is also a bank-maintained investment fund: perhaps bank-managed master trusts tend to invest in CCTs sponsored by the same bank. The heavy reliance of CCTs on investments in other CCTs would be consistent with banks creating a set of core funds with differing risk and return characteristics (e.g., corporate debt instruments having various ratings and maturities, domestic or foreign equities having varying levels of capitalization, dividend policies, industry concentration), and combining those core funds in various ways to build a broad range of feeder funds, each with a distinct investment policy or temporal horizon (such as target date funds). That CCTs invest an unusually large share of their total assets in common stock (nearly thirty percent), but put an unusually small amount (less than one percent) in registered investment companies (mutual funds) might suggest that other CCTs fill the role that mutual funds play in the investment portfolio of other DFEs.
Putting tiered or nested DFE investments to one side, perusal of Figure 10 reveals several additional differences in investment concentration between DFE types. Only MTIAs hold any employer securities (4.6%); (111) because other DFEs pool assets from plans of unrelated employers, they cannot hold "employer" securities per se. As noted previously, CCTs exhibit a distinctively high level of investment in common stock and low utilization of mutual funds. CCTs hold a lot more of their money in interest-bearing accounts (presumably in the same bank that is trustee (112)) than other DFEs (11% compared to 3-5%). Only insurance company PSAs take substantial stakes in real property (nearly 14%); (113) they also make more direct loans than other DFEs, and one might suspect that this is traceable to commercial real estate lending (construction and permanent financing).
The 103-12 IE is an outlier in its high levels of investment in U.S. government securities and "other investments." (114) The "other investment" category includes state and municipal securities, as well as options, index futures, repurchase agreements, collectibles, and other personal property. (115) Hedge funds that are owned twenty-five percent or more by benefit plan investors (so that the plan asset look-though rule applies) may qualify as 103-12 IEs, (116) and large holdings in debt securities, options, and derivatives would be consistent with the arbitrage-based short-term trading (based on proprietary models) that characterizes the business strategy of many hedge funds. (117) The asset category designated "joint ventures" should be mentioned in this connection, because most hedge funds and private equity funds are organized as partnerships or joint ventures, and if benefit plan participation in such a fund is under twenty-five percent (so that the assets of the fund are not deemed plan assets under the look-through rule), then an ownership interest in the fund would be reported as a "joint venture" interest. (118) While Figure 10 displays asset allocations, it should be noted that the 103-12 IE is also an outlier on the liability side. Total liabilities, as a share of total (gross) assets fall between 7% and 9% for MTiAs, CCTs, and PSAs, but the liability proportion for 103-12 IEs is more than 21%. Qualified plans rarely buy securities on margin or otherwise borrow to acquire or improve non-realty investment properties, because doing so would trigger the tax on unrelated debt-financed income. (119) Therefore one would expect that plan liabilities would typically constitute only a small share of gross assets. Even though leverage is a major component of most hedge fund investment strategies, (120) the larger liability proportion for 10312 IEs would be consistent with many of these investment vehicles being hedge funds, because hedge funds commonly take steps to insulate tax-exempt investors from the tax on unrelated-debt-financed income by interposing a corporation (a so-called "blocker") between the main fund and its tax-exempt investors. (121)
The DFE asset allocations presented in Figure 10 are dominated by a small number of high-asset DFEs. In the case of MTIAs, in 2008 the top 10% of MTIAs, ranked by total assets, accounted for 70% of all MTIA assets. Similarly, the top decile of CCTs reported 75% of all CCT assets, while the top deciles of PSAs and 103-12 IEs contributed 83% and 66% of all PSA and 103-12 IE holdings, respectively. Due to the dominance of a small number of high-asset DFEs, the average asset allocations presented here mask some variations in asset allocations by the large majority of DFEs of the same type. To take one striking example, virtually all PSA investments in real estate and "other loans" (likely real estate financing) is traceable to the 10% of PSAs with the largest total assets. In contrast, PSA investments in registered investment companies (mutual funds), which account for more than 90% of the total assets of half of all PSAs (i.e., 910 PSAs with non-zero but below-median total assets), falls to under 20% of the total assets of the 182 PSAs in the top asset decile.
Before linking DFE asset holdings to their investor-plans, it may be worthwhile to consider how much pooling DFEs accomplish. From the data in Table 1 we can determine minimum values for the average numbers of pension plans (large and small) that invest in each type of DFE: 3.1 for MTIAs, 19.4 for CCTs, 450 for PSAs, and 9.0 for 103-12 IEs. The corresponding minima for the average number of large pension plans that invest in each type of DFE are: 2.4 for MTIAs, 9.8 for CCTs, 96.2 for PSAs, and 5.2 for 103-12 IEs. (122) Interestingly, the number of pension plan investors does not greatly increase with increases in the size of MTIA total assets--on average, high-asset MTIAs have only a few more participating pension plans than low-asset MTIAs, but they pool much larger investments from each participating plan. The number of pension plans per DFE does increase substantially with the amount of DFE total assets for other types of DFEs, and dramatically so in the case of PSAs, as illustrated by the solid lines in Figure 11. The average number of participating plans grows from less than five in the lowest PSA asset deciles to 2407 in the highest decile! Although high-asset PSAs are pooling investments from many more plans, the size of the average plan investment also increases with the size of PSA total assets. As shown by the dotted lines in Figure 11, the average pension plan investment in MTIAs and PSAs rises with the size of the DFE's total assets. The graph also reveals that PSAs pool much smaller investments than MTIAs (note the logarithmic scale). Due to the prevalence of nonpension-plan investors in CCTs and 103-12 IEs, the average plan investment in these indirect investment vehicles cannot be determined simply by dividing total reported CCT or 103-12 IE assets by the number of participating pension plans. Therefore, DFE assets per plan is not a meaningful measure of the average pension plan investment and is not shown for CCTs or 103-12 IEs.
The complete protocol developed to accomplish first-order matching of DFE assets and liabilities to the pension plans that hold interests in DFEs is set forth in a web appendix to this article. (123) Here we merely highlight the major challenges encountered in the effort to link the balance sheet data and describe how they have been addressed.
The source data for large pension plans, including identifying information, plan characteristics, and balance sheet numbers, are taken from EBSA's Pension Research Files. The Pension Research Files have been subjected to a correction process involving both automated global edits (checking filings for internal consistency) and manual plan-specific edits to enhance data quality. (124) DFE filings are not included in the Pension Research Files (125) and therefore are more apt to be infected with errors, including missing information and reported asset/liability positions that do not match reported totals. In addition, pension plan Schedule D filings, which are necessary to link a plan to the DFEs in which it invests, are not part of the Pension Research Files and must also be taken from the unedited raw data sets. (126)
Perhaps due to errors in the raw data, a significant fraction of reported plan interests in DFEs could not be successfully matched to an identifiable DFE. For 2008, large pension plans reported 411,971 non-zero year-end interests in MTIAs, CCTs, PSAs, or 103-12 IEs. Of these 411,971 reported distinct investments, 184,089 actually represent interests in CCTs or PSAs that did not file Form 5500 and so were not DFEs. Excluding those interests in non-DFE CCTs and PSAs, 227,882 reported large plan interests in DFEs remain. Ultimately, 41,173 of those interests (18.1%) could not be matched.
We attempted reverse matching--that is, going from DFEs to the investor-plan--in an effort to raise the matching success rate. P7 Each DFE must file Schedule D. Part II, to identify all plans that hold an interest in the DFE. Unfortunately, however, Part II does not call for reporting current values of those interests. Consequently, a plan's failure to intelligibly identify a DFE in the plan's Schedule D, Part I, can be fixed where the DFE identifies the plan in the DFE's Schedule D, Part II, but only if the plan's filing contains only one garbled link. If there are multiple bad links, then even if each unidentified DFE adequately identifies the investor-plan, there is no way to know how much of each DFE's assets and liabilities to attribute to the investor-plan. For 2008 only 6190 bad links are attributable to plans that have exactly one bad link, and so reverse matching made only a modest improvement in our attribution success rate. In the end, this additional step salvaged only an additional 1320 links.
Even if a pension plan reports that it holds an interest in a specifically-identified DFE, often the data proved unusable due to inconsistencies between the amounts reported on the plan's Schedule D and Schedule H. For example, it is surprisingly common for an investor-plan to report on Schedule D a dollar value for its end-of-year interest in one particular CCT that is substantially greater than the amount the plan reports on its Schedule H balance sheet as the end-of-year dollar value of its interest in all CCTs! In such circumstances it is impossible to know how much of the DFE's assets and liabilities should be attributed to the investor-plan. Once this problem came to light the protocol was revised to filter out irretrievably defective filings. All large pension plans were sorted into three groups. Group I plans did not invest in any DFE at any time during the year (and so linking of indirect investments is not at issue). (128) Group 2 plans reported a DFE investment, but the amounts reported as the value of the plan's investment in each separate DFE of a given type (MTIA, CCT, PSA, or 133-12 IE) on Schedule D, when summed, did not match the amount reported on Schedule H as the plan's interest in DFEs of that type. Group 3 consists of those plans that utilized one or more DFEs and filed substantially consistent data on Schedules D and H. (129) The Group 2 plans, containing irretrievably defective financial data, were not run through the linking protocol, but a regression analysis was conducted to identify factors (plan, return, or investment characteristics) associated with bad filings. (130) Group 3 plans were linked to their DFEs and plan balance sheets were reconstructed to properly categorize assets and liabilities attributed from the DFEs in which they invested. This sorting of large pension plans reporting non-zero assets in 2008 produced the following breakdown: Group 1 (no DFE), 38,063 plans; Group 2 (inconsistent information on Schedules D and H), 15,198 plans; Group 3 (consistent DFE information), 27,247 plans. Accordingly, the large pension plan asset allocations (portfolio compositions) reported below exclude nineteen percent of all plans, or some thirty-six percent of plans that invested in one or more DFEs.
The accuracy of our matched results is also impaired somewhat by disparities in the reporting periods used by DFEs and their investors. Schedule D must be filed by a plan or DFE that invests in an MTIA, CCT, PSA, or 103-12 IE at any time during the plan year. (131) The beginning and end-of-year (BOY and EOY) values of a plan's interest in each category of DFE are reported on the plan's Schedule H, but on Part I of Schedule D, the plan reports only the EOY value of its interest in each separate DFE (as well as each CCT or PSA that does not file as a DFE). DFE investments attributed to a plan reflect the DFE's EOY Schedule H asset holdings and liability positions. (132) Consequently, if a plan uses an annual reporting period that differs from the reporting period used by a DFE in which the plan invests, then the information imputed from the DFE's balance sheet will correspond to a different date (the end of the DFE's reporting year) than the date that controls the plan's balance sheet (end of the plan year). Where the snapshot of a plan's indirect holdings is taken at a different time than the snapshot of the plan's direct holdings, pasting them together creates a composite picture with some distortion. Due to portfolio changes intervening between the end of the DFE's year and the end of the plan year, the linked results for a particular plan may generate categorical asset and liability allocations that never actually occurred. These inaccuracies for a particular plan are presumably mitigated in the summary data because other plans will be imputed indirect holdings that err in the opposite direction, leaving average overall asset allocation data largely unaffected. (In the interval between the end of the DFE year and the plan year, for example, one DFE may increase its holdings of common stock while another reduces its position.) More important, most large pension plans and DFEs use the calendar year as their reporting period, so the linked balance sheet results are dominated by apples-to-apples comparisons. (133)
Another limitation of our results concerns the matter of tiered DFEs. As explained previously, both MTIAs and CCTs report that a substantial share of their holdings consists of interests in other CCTs (in 2008, 18% and 31%, respectively). (134) To obtain a complete picture of the portfolio composition of a plan that invests in an MTIA that owns an interest in a CCT (for example), a share of the CCT's assets and liabilities (determined by the MTIA's proportionate ownership of the CCT) should first be attributed to the MTIA, then reattributed to the investor-plan (along with a portion of all the direct holdings of the MTIA) according to the plan's stake in the MTIA. Such multilayered indirect investments are not restricted to two tiers of DFEs, as illustrated in Figure 12. Multilevel asset and liability imputation requires identifying lowest-tier DFEs (those that hold no interest in another DFE) and applying an iterative process for tracing a share of their assets and liabilities up through chains of indirect ownership. DFE Schedule H filings are unedited (unlike the Pension Research Files, only the uncorrected raw data are available), as are all Schedule D filings, which makes multilevel attribution (tracing assets and liabilities through tiered DFEs) particularly susceptible to bad links and cascading reporting errors. To avoid those complications, our current protocol only links a share of DFE assets and liabilities to a large pension plan that directly owns an interest in the DFE. If the DFE in question (DFE1) holds an interest in another DFE (DFE2), that interest is attributed to the investor-plan but reported as an interest in either an MTIA, CCT, PSA, or 103-12 IE, according to DFE2's type: the DFE2 asset and liability composition will not be reflected in the composite balance sheet constructed for the plan that owns a share of DFE 1. Hence our current matching protocol preserves only second-level asset and liability categories. If multilevel asset and liability imputation could be comprehensively implemented (with all links properly identified), then the resulting pension plan balance sheets would have no entries in the asset categories reflecting interests in MTIAs, CCTs, PSAs, and 103-12 IEs. (135)
A. DB and DC Plan Allocations
1. Findings of the Current Study
Composite summary results of the linking of large single-employer defined benefit plans with their first-tier DFE investments based on 2008 returns are presented in Figure 13A. The graph shows the proportion of gross assets invested in the Schedule H asset categories by all 3620 single-employer DB plans that did not utilize a DFE in 2008 (Group 1 plans, shown as blue columns) alongside the corresponding average aggregate asset allocations (direct and indirect) of the 3348 linkable single-employer plans that invested some portion of their assets in one or more DFEs (Group 3 plans, shown as red and pink stacked columns). Of the twenty non-DFE Schedule H asset categories, the twelve that comprise the largest shares of gross assets for plans that do not use DFEs are displayed on the left side of the graph, with the four DFE types shown to the right. For plans using DFEs, the stacked column format displays the relative contribution of direct and indirect investments. The red column height represents plan assets owned directly and reported on the plan's Schedule H balance sheet, including (on the right side of the graph) reported interests in the four types of DFE. The pink portions of the columns show the incremental increase in asset shares resulting from attributing DFE assets to the proper balance sheet categories. By focusing on the pink increments it can be seen that tracing MTIA assets back to plans holding interests in master trusts translates into substantial increases in the shares of plan investments allocated to common stock and both corporate and government debt (recall that state and municipal bonds are included in "Other Investments"). Consistent with Figure 1, observe the dominance of MTIA interests prior to linking. Proper attribution reduces undifferentiated DFE interests, so there is no pink surplus atop the red DFE columns. Instead, the adjacent green and yellow stacked columns display post-linking DFE interests. The green component represents second-tier DFE interests, reflecting the fact that the assets of successfully linked first-tier DFEs include some interests in other DFEs, and the yellow part of the column shows the extent to which reported first-tier DFE interests of each type could not be successfully matched to the assets of a particular DFE (unmatched residue). Observe that linking actually increases proportionate CCT holdings because first-tier MTIAs and CCTs invest a significant portion of their assets in lower-tier CCTs. The side-by-side comparison of plans that do not and do use DFEs (blue and red/pink columns, respectively) indicates that for many asset categories proper identification of indirect investments substantially reduced the apparent disparity in portfolio composition between Group 1 and Group 3 plans, but there are a few striking exceptions. Indirect interests create or increase a disparity in the shares of plan portfolios allocated to two asset categories, preferred corporate debt and receivables. Also observe that linking barely narrows the wide disparity in DB plan utilization of RICs (registered investment companies, or mutual funds).
On average, single-employer defined benefit plans that use DFEs hold far more assets than plans that do not (by a factor greater than six in 2008). Consequently, the post-linking portfolio variation seen in Figure 13A might be attributable to differences in asset size rather than DFE utilization per se. To isolate those factors, Figure 13B shows the same comparison for a subset of single-employer defined benefit plans reporting total gross assets of at least $166.14 million but less than $383.66 million. Different numbers of Group 1 and Group 3 plans fall within this range, but the distribution throughout the range and the average assets of the groups are very similar. (Combining all Group 1 and Group 3 single-employer DB plans and ranking them by asset size, this range corresponds to the 90th to 95th percentile.) The figure reveals that the stark difference in mutual fund (RIC) utilization is not due to differences in plan asset levels.
Corresponding graphs showing the results of linking large single-employer defined contribution plans with their first-tier DFE investments based on 2008 returns are presented in Figures 14A and 14B. (The "imputed assets" category shown in Figure 14A reflects a common reporting error among DC plans, as explained in the margin. (136)) DC plans make less use of DFEs than DB plans do, and the linking failure rate for DC plans is somewhat lower. (Compare the yellow segments representing unlinked first-tier DFE interests with the red columns reporting direct DFE holdings in Figures 14A and 13A.) Just as for DB plans, single-employer defmed contribution plans that use DFEs tend to hold more assets than plans that do not, so a comparison of the portfolio compositions of a subset of DC plans of comparable size that do and do not employ DFEs is shown in Figure 14B. (137) Observe that the apparent difference in mutual fund (RIC) utilization between DC plans that do and do not invest through DFEs (seen in Figure MA) disappears when the comparison is limited to plans of comparable asset size (Figure 14B). This is contrary to the finding for DB plans. The overall average disparity in common stock investments also vanishes when plans with similar asset levels are compared, but only if common stock held in DFEs is taken into account. On the other hand, disparities in DC plan holdings of employer securities, interest bearing cash, and "other investments" (including state and local bonds, options, and derivatives) are seen in both Figures 14A and 14B.
2. Comparison with EBSA Summary Statistics
In April 2012, the Employee Benefits Security Administration (EBSA) of the Department of Labor released its first-ever DFE statistical report, which, like the current study, focuses on returns filed in 2008. The report contains counts of DFEs, counts of private pension plans invested in DFEs, and asset counts. (138) Most relevant to this study, the EBSA report also includes a table that takes assets reported by pension plans as invested in DFEs and distributes them into the other financial asset categories according to the composition of the DFE portfolios in which the pension plans invest. (139) These composite balance sheets, reporting the aggregate financial position of all large private DB plans and all large private DC plans in 2008, can be readily converted into proportionate asset allocations and compared to the DB and DC plan results obtained by the current study.
EBSA's results are compared with the current study in Figures 13C (DB plans) and 14C (DC plans). Each graph uses a four-column presentation, where the first (blue) column in each cluster displays the EBSA linked results. The following three columns show the results of the current study, as follows: the second (red) column shows the composite average asset allocation of all plans included in this study (after linking those that use DFEs); the third (green) column separately displays the asset allocation of those plans that do not use DFEs (Group 1); and the fourth (purple) column separately displays the asset allocation of those plans that use DFEs and report consistent information on Schedules D and H (Group 3). Thus, the last two columns in each cluster display the same information presented in Figure 13A in the case of DB plans, or 14A for DC plans, but without the internal breakdown (stacked columns) between direct and indirect investments for plans that use DFEs.
Concentrating on the DB comparison (Figure 13C), the first impression may be that the EBSA numbers do not square well with the current study. Some important differences in approach may explain the disparities, however. EBSA's numbers ostensibly reflect complete linking of DFE assets and liabilities to investor pension plans. Apparently, this includes tracing the assets of second (and lower) tier DFEs back to plans holding an interest in a first-tier DFE, such as when an MTIA or CCT has some of its assets invested in another CCT. Moreover, EBSA reports no unmatched residue resulting from unidentified DFEs (bad links). In sharp contrast to the current study, EBSA reports all post-linking DFE interests as zero. Observe the absence of blue columns in the four DFE categories on the right side of the graph. If EBSA's methodology is reliable, then the DFE interests shown in red (which represent stakes in lower-tier DFEs plus reported first-tier DFE interests that were not successfully associated with a uniquely-identified DFE) should account for all differences in the height of the blue and red columns reporting non-DFE investments. Presumably, EBSA addressed the pervasive problem of poor quality returns (inconsistent reporting of DFE interests on Schedules D and H, as well as inadequately identified DFEs) either by contacting filers and calling on them to correct the data, or by application of some sort of sampling technique. Unfortunately, EBSA has not yet released an explanation of its methodology. Besides EBSA's announcement of comprehensive successful linking, some of the disparity in Figure 13C may be attributable to EBSA's inclusion of all large DB plans, in contrast to the current study which focuses on large single-employer plans.
The DC plan comparison in Figure 14C shows much closer correspondence between this study and EBSA's report of portfolio composition. That correspondence presumably results from the lower overall utilization of DFEs by DC plans and our higher MTIA linking success rate relative to DB plans. Observe that the greatest disparity lies in the share of the portfolio devoted to common stock, while CCTs--which we have seen invest most heavily in common stock (see Figure 10)--represent the largest unlinked DFE interest.
The question that motivated this investigation is whether material differences in portfolio composition are associated with a pension plan's use of indirect investment vehicles. To a first approximation, Figure 13B presents an initial tentative answer for single-employer DB plans in 2008. This side-by-side comparison of plans holding similar amounts of assets which either do or do not use DFEs (shown by the red/pink and blue columns, respectively) reveals remarkably consistent average portfolio compositions once indirect investments are properly identified. Indeed, these two groups of plans allocate substantially equivalent average portfolio shares to most asset categories; dramatic disparities appear only for holdings of registered investment company shares and interests in insurance company general accounts. The data suggest that CCT interests might substitute for RIC holdings among plans using DFEs. The picture for single-employer DC plans in 2008 (Figure 148) also displays wide-ranging consistency. Here. RIC holdings constitute almost sixty percent of pension portfolios whether or not a plan uses DFEs; significant differences are found only in the share of assets allocated to employer securities (ESOPs, apparently, do not employ DFEs), interest-bearing cash, and other investments. Generally speaking, these data are consistent with DFE utilization for the purpose of obtaining increased diversification and economies of scale, not for a nefarious objective of obscuring pension plans' overall portfolio composition.
At this stage of the research, however, several big caveats must bracket such an optimistic conclusion. First, these results depict average aggregate asset allocations. At the individual plan level it is entirely possible that disclosed direct holdings of conservative investments (say, U.S. government bonds and blue-chip stocks) are combined with practically undisclosed indirect investments through DFEs in positions that have very different risk, return and liquidity characteristics (such as hedge funds, for example). As a further step, the authors investigated the degree of correlation between direct and indirect investments at the individual plan level, as reported below in Part IV.C.
Second, due to the extreme generality of' the asset categories reported to the Department of Labor by large pension plans and DFEs on their Schedule H balance sheets, even an apparently perfect correspondence between a particular pension plan's direct and indirect investments--equal proportions invested in "common stocks" for example--would not rule out wildly divergent investment characteristics. The common stocks held by the plan and a CCT in which it holds an interest might represent stakes in different industries, publicly-traded enterprises or privately-held operating companies, companies incorporated domestically or abroad, etc. This granularity of the digitized financial data forestalls attempts to drill down further into the truly consequential characteristics of pension plan portfolio composition. This limitation is explored further in Part V. (140)
Third, the results displayed here exclude all plans that invested in DFEs but reported inconsistent amounts as the current value of their year-end interests in DFEs on Schedule D and Schedule H (Group 2 plans, as defined earlier). If there were a deliberate effort to hide the nature of plan investments, such inconsistent reporting would surely facilitate it. The characteristics of unlinkable Group 2 plans are investigated below in Part IV.D.
Finally, finding corresponding overall asset allocations by plans that do and do not utilize DFEs does not rule out major disparities on the liability side of the balance sheet, and there are some. Among large single-employer DB plans using DFEs (Group 3 plans), indirect liabilities (that is, DFE liabilities linked to plans) are many times larger than the total direct liabilities reported on Schedule H by the investor-plans. Direct liabilities account for only 31% of the total liabilities (direct and linked from DFEs) of all Group 3 DB plans, but for plans in the bottom half of the DB plan asset distribution direct liabilities amount to only 1% of the total liabilities! Looking at liabilities as a share of assets, direct liabilities of all Group 3 DB plans are approximately 2.8% of reported Schedule H gross assets, while total liabilities (direct and linked from DFEs) come to 8.4% of total gross assets (direct and linked from DFEs). The ratio of direct liabilities to gross assets reported on Schedule H, it should be noted, is approximately the same for DB plans that do not use DFEs as for those that do (Groups 1 and 3, respectively). Therefore, the high levels of indirect liabilities are not simply substituting for lower direct liabilities in Group 3 plans. DB plan DFE investments are associated with much larger liabilities than direct investments, and that DFE debt is currently hidden from plan participants. (141)
The data displayed in Figures 13B and 14B represent weighted average portfolio allocations for a subset of high-asset (90th to 95th percentile) single-employer DB and DC plans in 2008. The investment behavior of plans of this size does not necessarily reflect the asset mix of a "typical" plan. As demonstrated previously, there are systematic differences in DFE utilization as the size of the investor-plan's total assets increase (Figures 37). Does the categorical composition of pension plan portfolios, taking into account linked first-tier DFE investments, also vary with asset size? To check for variations in portfolio composition by plan asset size, single-employer DB and DC plans were separately grouped into ten asset size ranges, each containing equal numbers of plans, and the average portfolio composition before and after linking with DFEs was computed for each such decile. The authors then compared the results for mid-range plans (fifth and sixth deciles) with the numbers for high asset plans and the overall weighted averages. (142) The unmatched results discussed earlier show extensive utilization of registered investment companies (RICs, typically mutual funds) by most plans, with that large portfolio share dropping precipitously among plans in the highest asset categories (top one or two deciles), which show a corresponding surge in MTIA investments (Figures 3 and 4). The drop in RIC usage by high-asset plans survives linking first-tier DFE investments; mid-range DB plans (fifth and sixth deciles) that use DFEs hold about 25% of their assets in RIC shares, while the overall weighted average is only 7%. Although important, RIC shares do not dominate MTIA portfolios (see Figure 10), so attributing MTIA assets to the high-asset plans that invest through MTIAs does not counteract the drop in direct ownership of mutual funds. That finding poses another question: what actually substitutes for the drop in RIC holdings by high-asset plans? Due to the high matching failure rate for DB plans, interests in MTIAs (meaning mostly unmatched MTIAs) still constitute substantially larger shares of the portfolios of top-decile than mid-range plans after matching. For DB plans, however, linking does reveal an increase in utilization with plan asset level of preferred corporate debt securities and partnership/joint venture interests. For DC plans, examination for variation in post-linking portfolio composition by plan asset level shows a substantial fall in RIC investments only at the very top of the asset spectrum (roughly, the top 5% of DC plans), and that drop in RIC holdings is counteracted by notable growth in the share of assets that the largest DC plans invest in employer securities (ESOPs, presumably), and CCT interests (largely second-tier CCTs).
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||II. Background D. Pension Plan Utilization of Indirect Investment Vehicles through IV. Results A. DB and DC Plan Allocations, p. 626-666|
|Author:||Wiedenbeck, Peter J.; Hinkle, Rachael K.; Martin, Andrew D.|
|Publication:||Virginia Tax Review|
|Date:||Mar 22, 2013|
|Previous Article:||Invisible pension investments.|
|Next Article:||Invisible pension investments.|