What's Happening with the Pension Fund? -- Part 9

by Charles Schwartz, Professor Emeritus, University of California, Berkeley
schwartz@physics.berkeley.edu                                      May 18, 2001

>> This series is available on the Internet at http://socrates.berkeley.edu/~schwrtz

UC Admits an Error by Wilshire - We Explore How Bad It Is

A top UC official now admits that Wilshire Associates erred in their earlier calculations for the University's pension fund. We are left to figure out the magnitude of these mistakes (about $19 Billion in Assets and $39 Billion in Liabilities for one year); and we also discuss the implications of these errors.

Letter from UC's Senior Vice President for Business and Finance

                                                                                                       University of California
                                                                                                       Office of the President
...                                                                                                   April 16, 2001
Dear Professor Schwartz:

I am following up on your concern about the differences between Wilshire Associates and Towers Perrin in comparing respective asset and liability projections for the University Retirement System. The differences in funded level projections relate to differences in assumptions and not to discrepancies in the models. Given the same inputs both the Wilshire and Towers Perrin models produce nearly identical results. The important differences are:

1. Lump Sum Distributions: Wilshire assumed no lump sum distributions while Towers Perrin assumed a 20% lump sum election. The effect of lump sums is to reduce assets and liabilities in equal dollar amounts and, as a result, Towers Perrin's liability projections will be lower than Wilshire's and its ratio of assets-to-liabilities will be higher. This difference in assumption explains about half of the difference in projected liabilities in year 2015 and almost all of the difference in projected assets.

2. New Entrant Age and [Salary]: The remaining difference in liability projections is explained by assumptions made for the average age and salary in new entrants. This became a particularly important input given the assumed 2.5% growth in the active workforce. Wilshire assumed new entrants would come in at a significantly higher age and salary than did Towers Perrin, a difference that caused a more rapid growth in Wilshire's liability projections. This age and salary assumption difference fully explains the remainder of the difference in 2015 liability projections.

One purpose for the Towers Perrin liability study was to bring greater accuracy to the liability projections and to use their findings to further examine asset allocation for the retirement plan. Wilshire reports that although the Towers Perrin study shows a stronger projected funded position for the retirement plan compared to their study, their asset allocation recommendation at the time would not have changed.

Thank you for your interest in this topic. I hope this addresses your concerns.

                                                                            Joseph P. Mullinix
                                                                            Senior Vice President -- Business and Finance

When that cryptic letter arrived (10 days later) I wrote back:

Dear Vice President Mullinix;                                                                April 26, 2001

Today I received your letter, dated April 16, concerning the differences between Wilshire Associates and Towers Perrin in their projections of UCRP asset/liability ratios. I have been trying - so far unsuccessfully - to make mathematical sense of the explanations you present.

I suggest that we meet to go over the details of this matter and try to reach some common intelligent understanding. (I have made such suggestions in the past and never gotten a response.)

I also request the following detailed data (which you refer to in your letter but do not provide):

1) From the Towers Perrin calculations - separate values of their projected Asset and Liability dollar values, at each of the five-year markers.

2) From the Wilshire Associates calculations - the dollar amount of the corrections, for the Assets and for the Liabilities at each five-year marker, which you refer to in paragraphs numbered "1." and "2." of your letter.

I never got any answer, although I sent this email a second time. So it is up to us to try to solve the riddle of Mullinix' letter and figure out how big Wilshire's error really is.

We Solve the Riddle

In Table 1 below we have data from Wilshire's original calculations (Exhibit 9 of their March 16, 2000 Study) in the first three columns. Column 1 shows their projections of UCRP's Accrued Liability at each indicated year; Column 2 shows their projected Market Value of the Assets, using their Policy C for asset allocations; and Column 3 shows the critical Ratio of Assets to Liabilities. These data are given by Wilshire as Expected Values, ignoring the statistical uncertainties in such calculations.

Column 4 in Table 1 shows the Asset/Liability Ratios as calculated by Towers Perrin earlier this year; and we see the major discrepancy between these numbers and those of Wilshire in Column 3 - a discrepancy which grows larger in later years. Wilshire projected a declining ratio while Towers Perrin projected an increasing ratio.
Table 1.   Wilshire's Original Projection of UCRP Assets and Liabilities
($ Billions)
     1      2      3      4
Accrued Liability Assets - Policy C Ratio: Assets/
Accrued Liability
Ratio Calculated
by Towers Perrin**
1999 $22.2 $38.1   1.72
2000   23.9   40.7   1.70   1.54
2005   36.1   56.6   1.56   1.81
2010   55.5   78.8   1.40   1.85
2015   84.8 108.4   1.26   1.93
2020 126.6 150.4*   1.18   2.07

* Includes a "contribution" of $2.7 B.
** I have interpolated the original Towers Perrin numbers to fit the years used by Wilshire; so the Towers Perrin ratios shown here differ slightly from those I gave in Part 6.

Data Sources for Table 1:
Wilshire report: http://www.ucop.edu/bencom/news/revisedasset.html (page 10)
Towers Perrin report: http://www.ucop.edu/regents/regmeet/jan01/502att.pdf (page 13)

Now we are ready to use the information provided by SVP Mullinix in his letter.

1) Lump Sum Distributions. This mistake by Wilshire means that we must reduce both Assets and Liabilities by some amount - call it X.
2) New Entrant Age and Salary. This mistake by Wilshire means that we must reduce the Liabilities by an additional amount - call it Y.
3) For the year 2015, the amounts X and Y are approximately equal.

We can now write some algebraic equations (this is like taking the SAT) for the year 2015, using the original Wilshire numbers from Table 1:

Correct Liability = $84.8B - X - Y,
Correct Assets = $108.4B - X,
X = Y,
(Correct Assets)/(Correct Liability) = 1.93,
where the last equation comes from using Column 4 in Table 1. The numerical solution of these equations is: X=Y= $19.3B. So we deduce that Wilshire's error in calculating the UCRP Liability for the year 2015 (X+Y) was about $39 Billion.

That is a huge mistake; it boggles the mind that those experts could goof by so much. Maybe, when I some day get the requested data from Vice President Mullinix it will turn out that the exact number is somewhat different. In any case, it is certainly not something to be lightly brushed aside.

How big is the error for other years? Wilshire's original calculations went to 2029 and their original value for the Liability in that year was about twice the value for 2015. Was their error here also twice as big, more, less? Mullinix keeps us in the dark.

Implications - 1

Maybe there is some reason to excuse Wilshire Associates for these mistakes - after all, Vice President Mullinix doesn't seem to be bothered by them.

Rereading the Wilshire report's section titled "Analysis of the UCRP Liabilities" (pages 4-5 of their March 16, 2000, Study) we see that they present themselves as experts in this field:

"Wilshire uses a pension model it pioneered 25 years ago to project a retirement plan's assets and liabilities into the future." (page 5)
We also see that they are aware of the importance of getting the assumptions right:
"The liability numbers presented here rely heavily upon the underlying actuarial assumptions used to calculate them. Small differences in assumptions are known to create quite different liability values." (page 4).
And they acknowledge getting their inputs from the authoritative sources:
"Wilshire's analysis of the Plan is based upon the June 30, 1999 actuarial report prepared by Towers Perrin" (page 3) ... "The workforce growth rate is among the most important assumptions in the projections and this information was derived from discussions with Towers Perrin and the [UC] President's Office." (page 10).
Thus it appears difficult to ascribe Wilshire's mistakes - the ones now acknowledged in the letter from Mullinix - to either their lack of experience in this particular subject or to some other form of simple incompetence. Nor is there any hint, from Mullinix' letter, that anyone but Wilshire bears responsibility for these mistakes.

Implications - 2

Near the end of Mullinix' letter we read,

"Wilshire reports that although the Towers Perrin study shows a stronger projected funded position [i.e., a larger ratio of assets to liabilities] for the retirement plan compared to their study, their asset allocation recommendation at the time would not have changed." (emphasis added)
I am amazed at the audacity of this statement. The primary purpose of Wilshire's Investment Strategy Study was to ascertain the future financial requirements of the University's pension plan and from that data to fashion an appropriate investment policy in terms of asset allocations. How can such a drastic change in the former (the now revised projections of UCRP Liability) not have a major impact upon the latter?

To see this logical agenda in explicit detail, all you have to do is read the Wilshire report as delivered to The Regents. I'll excerpt the most salient parts:

"Wilshire Associates Inc. was retained by the UC Board of Regents to conduct an Asset Allocation Study for the UC Retirement Plan ("UCRP"), ...The Study's purpose is to recommend investment strategies, policies, and guidelines consistent with the long-term fiscal requirements of the UC Retirement Plan ..." (page 1)
"The most striking characteristic of the UCRP is its exceedingly strong fiscal status with assets comfortably exceeding the Plan's benefit obligations or liability. The June 30, 1999 actuarial value of the Plan's liability was $22.2 billion. The $38.1 billion in market value of assets in the UCRP exceeds the $22.2 billion actuarial liability by $15.9 billion." (page 4)
"[T]he Plan has a current sufficiency of assets to provide for its current and future obligations to members and have a reserve left over that exceeds $9 billion! ... Even the more conservative $32.1 billion actuarial value of assets [as distinct from the market value] is sufficient to cover the larger $28.8 billion liability [including future obligations]." (page 4) (emphases in original)
"In particular, The Regents should be aware of the assumptions the actuary makes for the performance of assets because they can have a very heavy influence upon the selection of how much risk is appropriate in the investment portfolio." (page 4) (emphasis added)
"Referring back to the Plan's envious level of assets in relation to liabilities, it is worth noting that today's circumstances were created because a 15 year bull market in stocks and bonds, combined with the Treasurer's Office management, produced investment returns of 15% annually, double the 7.5% interest rate assumption. Looking ahead, Wilshire estimates that the $38.1 billion investment portfolio has to earn only a 6.1% annual return to fully provide for the higher $28.8 billion liability, the calculation that includes current and future benefits to active members and retirees in the Plan. At first glance, that might appear to be a goal that is easily achieved by investing the entire $38.1 billion in assets in Treasury bonds, since that is what bonds are currently yield[ing]." (page 5) (emphasis added)
"There are several reasons for not declaring victory and selecting an asset allocation of 100% Treasury bonds. The first is future additional employees. ... [U]nlike most defined benefit plans that are projecting no change or even a declining employee population, the UC system is forecasting the possibility of a 2.5% annual employment growth through 2011 and 1.5% annual employment growth thereafter. At a 6.1% return, new Plan contributions will be required sometime in the future under this high employment growth scenario." (page 5) (emphasis added.)
"The purpose of an asset allocation study is to find an asset mix that best fits the Plan's expected growth in employees and uncertain future economic events. ... The UCRP's unique liability parameters were used as inputs for Wilshire's pension model in order to provide projections of assets and liabilities over a 20-year forecasting period." (page 5)

Let me take that narrative and give it numerical form. In Table 2 below I repeat part of the data from Table 1 above and I also add the projections for an alternative asset allocation policy - what Wilshire calls "Policy A" - which is 100% investment in bonds, as suggested in the paragraphs quoted above. (Wilshire did not give these numbers for Policy A; I have calculated them myself using data provided in the Wilshire report.)
Table 2.  Comparative UCRP Projections for two of Wilshire's Alternative Policies     ($ Billions)
     1      2      3
Accrued Liability Assets - Policy C Assets - Policy A
1999 $22.2 $38.1 $38.1
2000   23.9   40.7   40.0
2005   36.1   56.6   50.8
2010   55.5   78.8   64.5
2015   84.8 108.4   81.9
2020 126.6 150.4* 104.0
   * Includes a "contribution" of $2.7 B.

We see in Table 2 that the projected values of Assets following Policy A grow substantially slower than those projected for Policy C; this is just what we would expect from investing everything in bonds versus a portfolio heavily weighted with stocks. The suggested attraction of bonds is their lower level of risk: Wilshire gives the Expected Risk of Policy A as 7.0%, much lower than the 12.3% Expected Risk of Policy C.

The main issue, however, as discussed in Wilshire's narrative quoted above, is whether this much safer Policy A would be able to meet the growing needs of the University's Retirement Plan: we must compare the Assets in Column 3 with the Liabilities in Column 1. We see, from the data in Table 2, that Policy A would fail to cover the UCRP obligations by the year 2015. This numerical comparison confirms the conclusion stated in Wilshire's narrative, above, that an "all bonds" investment strategy would not succeed in covering UC's needs because of the growing employment level expected in the years ahead.

But that conclusion followed from the old calculations that Wilshire did for the projected value of UCRP liability; and we now have, from the Mullinix letter, admission that Wilshire's calculations were wrong in exactly this place. The numerical size of that error - as we determined it earlier for the year 2015 - amounts to about $19 Billion reduction in liability, relative to assets. With this new correction taken into account, we see that Policy A could very easily meet that future obligation of the UCRP!

Now you see how that previously quoted sentence from Mullinix' letter,

"Wilshire reports that although ...[they erred in the earlier calculation], their asset allocation recommendation at the time would not have changed."
is totally contradicted by the logic of Wilshire's own stated methodology.


I want to say, as I have said before, that I do not present my critiques as arguing for or against any particular asset allocation policy for UCRP. Some of my expert correspondents have expressed the view that Wilshire's recommended asset allocation policy is very reasonable; and I have no argument with such opinions.

What I have focused upon is the uncovering of flaws - mistakes, omissions, contradictions - that can be found in the official documents, from Wilshire Associates and from the UC President's office. You can call this Intellectual Quality Control; I also call it whistleblowing. There appear to be quite a large number of these flaws; and while a few that I have complained about may be seen as nit-picking by some readers, some of these faults - if my critiques are valid - are of major significance.

Most of my whistleblowing complaints (from the earlier Parts 1-5 of this series) are still waiting for the comprehensive review and response that was promised by the University President and his top staff.

Two previous rebuttals delivered by Wilshire Associates have been shown to be false (see Part 6.) Some new data and analysis (in Parts 7 and 8) raise additional pointed questions about the professional competence and integrity of Wilshire. And in a number of places I have suggested that some UC officials are implicated as well.

In this paper we studied the first instance in which Wilshire has admitted a mistake, one that I had first detected and brought to light in January. It was a whopper of a mistake and its implications are very serious indeed. Yet, it appears from Vice President Mullinix' letter that both Wilshire and UCOP are not only trying to diminish or dismiss this error but are continuing a cover-up to conceal the full set of facts.

The question which The Regents must face is this: How high must this tide of criticism rise before you decide that Wilshire Associates does not deserve the confidence required to serve as the University's investment consultant? Fiduciary responsibility means, among other things, that you should move expeditiously to acknowledge and clean up your mistakes, regardless of whom that might embarrass.