Review of Mendenhall and Sutter Results for STRS
- Rudy Fichtenbaum
- 3 hours ago
- 9 min read
By Rudy Fichtenbaum
Recently Allen Mendenhall & Dan Sutter published an article “Retirement at Risk: The Political Economy of Public Pension Governance” in the Journal of New Finance comparing STRS’s and OPERS’s reported returns and returns calculated using audited data. What follows is my review of the Mendenhall and Sutter article as it pertains to STRS.
Let me start my review by pointing out what I believe to be a problem with the Mendenhall & Sutter article. The problem I see with their article is that they used gross returns for what they term STRS reported returns. Then they compare those returns to returns they calculated using audited data. The problem is the audited data report net investment income so the return being calculated is a net return. Therefore, I believe Mendenhall and Sutter should have used STRS reported net returns.
Using gross returns will overstate the difference between the investment returns that STRS reports and the investment returns that you get when you use the audited data. The reason for the overstatement is that audited returns report net investment income.
There are two things at play in looking at the differences between reported investment returns and returns calculated using audited data. Investment returns that STRS publishes in the ACFR each year are calculated using a method called Modified Dietz. Without getting into the weeds, Modified Dietz returns are calculated monthly and use daily weighted cash flows. This is a standard calculation that most pensions make. As STRS senior staff are fond of saying, it is the method that is GIPS compliant.
These monthly calculations produce gross returns that are for each asset class and the total fund. To get a net return, STRS backs out expense data after calculating the gross return. The Mendenhall and Sutter article deals exclusively with the returns for the total fund which are a weighted average of the asset class returns, although the weighting is complicated. (If anyone is interested in how you calculate the weights you can write to me and I will send you the formula for the weights.)
Returns calculated using audited data, which you can also find in the STRS Actuarial Valuation Report are calculated using the Simple Dietz method (as opposed to Modified Dietz); the Simple Dietz method uses annual data making use of beginning and ending year assets, net investment income, and cash flows. The formula used in my calculations and presumably being used in the Mendenhall and Sutter article is the same formula you will find on p. 23 of Cheiron’s 2025 Actuarial Valuation Report.
The main finding of the Mendenhall article is that when you compare STRS’s reported investment returns to the returns using audited data and you do the same thing for OPERS, STRS’s reported investment returns are nearly always higher than returns calculated using audited data. But for OPERS, the differences between reported investment returns, which are presumably calculated using the same method as STRS (Modified Dietz), and returns using audited data (Simple Dietz) generally cancel each other out over time.
Mendenhall and Sutter note that the staff calculated investment returns are self-reported and are not reviewed by a CPA during the annual audit. Now of course STRS will claim that the investment returns are reviewed by ACA. However, ACA is not verifying that the inputs are correct. Instead, ACA is simply verifying that the method being used is correct. But ACA is not auditing the beginning and ending monthly assets nor are they auditing the monthly cash flows.
The Mendenhall & Sutter article concludes STRS’s reported calculated using self-reported data overstate its investment returns and that using those self-reported returns for PBI is a conflict of interest, which results in over paying investment staff receiving PBI. Therefore, they recommend using total fund returns derived from the audited statements for paying PBI.
Previously when I looked at the monthly data for 2021 and calculated the rate of return for the pension using the modified Dietz method with data provided to me by the staff, I was able to replicate the return reported by the staff. However, I did notice that the assets on July 1, 2020 and the assets on June 30, 2021 in the data the staff provided to me did not match the assets reported in the audited statements. In addition, the cash flows for the year in the data used to calculate the return differed from the cash flow that is derived from the audited financial statements. At the time, I was told this was due to timing differences. If the data used to calculate the return was audited data, there would be no timing issues, and the Board members and our members would have confidence that when PBI was paid, it was paid because the fund had more money with which it could pay benefits.
I believe Mendenhall and Sutter’s conclusion is fundamentally sound, although as I already said, I disagree with their use of gross investment returns. The STRS reported returns in the article are clearly gross returns. In fairness to Mendenhall and Sutter, before 2020 the investment returns reported by STRS in its ACFR were gross returns. It was only in 2020 that STRS started showing both gross and net returns for the total fund in the ACFR. Before 2020, there was generally a footnote to the effect that most of the returns shown in the ACFR were gross returns, the exceptions being returns for real estate and alternatives.
The problem with using gross returns and comparing them with the returns calculated using audited data, is that investment returns in the audited data are net investment returns. What I believe Mendenhall and Sutter should have done is to take the STRS reported returns, which were mostly gross returns, and subtract expenses, associated with the gross portion of the total fund return, which were generally in the area of 12 basis points (.12%). That would have given them a self-reported net return which could have been compared with the net return they calculated using audited data.
Table 1 shows the audited data. Net income (the numerator in the Simple Dietz Return aka “audited return) is calculated taking end of year assets and subtracting beginning year assets and cash flows. Average assets (denominator in the Simple Dietz Return) are beginning year assets plus 0.5 times cash flow. When cash flows are negative, as is the case for STRS, this makes the denominator smaller and raises the rate of return. Usually, when people calculate a rate of return, they take the difference between the ending value and beginning value and divide by the beginning value. But in a pension where there are negative cash flows, we add those back into the numerator (subtracting a negative number makes it positive) because cash flows should not add to or subtract from investment income. Adding .5 times the cash flows to the denominator assumes that cash flows occur at the midpoint of the year or are spread evenly across the year. When cash flows are negative including half of the flows in the denominator also increases the rate of return. The idea behind this adjustment is that when you have negative cash flows the assets you begin with are smaller, but that is not because of poor investment performance.
I have replicated Mendenhall and Sutter’s results for STRS using data from Table 2. Then I adjust STRS reported returns for expenses subtracting 0.12% from STRS reported gross returns to obtain STRS reported net returns. Finally, in the last column of Table 2, I subtracted the Mendenhall-Sutter Simple Dietz calculation, what they call the audited rate of return, from my calculation of the Simple Dietz return. When you subtract the two calculations, they are the identical out to two decimal places and there is a plus or minus 1 difference at three decimal places. Clearly, the difference my calculations and Mendenhall and Sutter’s calculations is that they rounded all of their audited rates of return, as evidenced by the fact that they all end in zero. So, I believe their calculations are correct. To repeat what I said earlier the only difference between our calculations is that I adjusted STRS’s reported returns by subtracting 0.12% from their reported gross return to obtain an estimate of a net return.
In Mendenhall and Sutter’s study they found that in 19 of 20 years (2003-2022) the audited returns were significantly lower than reported gross returns. In Table 2 you will see that the audited returns (which are net) are lower than reported net returns (gross returns minus expenses) in 14 of 20 years. The years where they are not lower are highlighted in red.
How much of a difference does this make to the pension? If you use Mendenhall and Sutter’s returns, the dollar difference is $4.5 billion. If instead you use reported net returns, the difference is still $2.95 billion. In other words, if you accept Mendenhall and Sutter’s results, STRS overestimated its investment earnings by $4.5 billion over a 20-year period; but my results, using net returns, show that STRS overestimated its investment earnings by “only” $2.95 billion over 20 years. Even the latter figure is real money.
At the end of the day the thing that matters to our members is amount of money the pension has to pay benefits. Using the audited data and calculating a total fund return that would be used in a PBI calculation would ensure that the return being used to award PBI reflected the actual dollars added to the pension’s assets. Let me add that I believe the Board also needs to look at how returns are being calculated at the asset class level, which actually plays a more important role in awarding PBI than the total fund return, for most PBI that is awarded. What assurances does the Board have that the data being used to calculate returns at the asset class level, which is a subset of the data used to calculate to total fund return, is accurate? This is all the more important given the fact that until 2024 gross returns were being used at the asset class level to award PBI. In fact, at one point the Board was told that there was no choice but to use gross returns because expenses were not available at the asset class level. The fact that the Board changed the policy and required the use of net returns at the asset class level clearly shows that the Board was not being told the truth.
Now I am sure that the STRS staff will point to ORC 3307.15 and claim that they are required to use GIPS returns, which are “reported returns” calculated using Modified Deitz. So, let’s look at exactly what 3307.15 say in regards to returns. It says “When reporting on the performance of investments, [my emphasis] the board shall comply with the performance presentation standards established by the CFA institute.” Reporting on performance and paying PBI are two different actions. The law only deals with reporting on performance. No one has suggested that STRS stop “reporting investment performance” using Modified Deitz returns (aka GIPS returns) or what Mendenhall and Sutter call self-reported returns.
The fact is STRS reports two sets of returns. It reports GIPS (Modified Deitz) returns in the ACFR and it reports returns using audited data (Simple Deitz) returns every year, in the Actuarial Valuation Report (AVR). The AVR is important because it determines how much money the pension has to pay benefits.
What is being suggested is that STRS stop using the GIPS returns in calculating PBI, since they do not reflect the actual money, the pension has to pay benefits. The Board is not required by ORC to pay PBI. However, it is allowed to use a system that rewards performance if it believes it is in the best interest of the members. But the system it uses and how it measures the performance of employees is totally at the discretion of the Board. PBI is not addressed in ORC. So clearly the Board can use whatever performance measure it deems to be in the interest of members. What is being suggested is that using the actuarial return, which is based on audited numbers, and reflects how much money the pension has to pay benefits, is what is in the members’ interest.
Let me end by saying that I recognize that reducing PBI is not going to give the pension the money it needs to pay a permanent ongoing COLA and move the age for an unreduced retirement to 32 years permanently. To accomplish those goals the pension needs additional funding in the form of an increase in the employer contribution. Nevertheless, ensuring that every dollar that is spent by the pension is spent ultimately for the benefit of members is the primary responsibility of the Board as fiduciaries.
If returns being used to pay PBI, do not accurately reflect the amount of money the pension has to pay benefits, then the link between pay and performance is broken. While fixing this problem may not get members a COLA or 32 years, it is nevertheless the duty of the Board to fix it. Board members are constantly told if you make any changes to PBI, investment staff will leave and then asked how will you pay benefits. But if you are paying people more than they should be paid because returns are overstated, the public and even Ohio’s political leaders will lose confidence in the Board and the System, and that is also a threat to the pension.
Rudy Fichtenbaum stood up for Ohio's teachers. Now it's our turn to stand up for him. Please donate to his GoFundMe campaign to help him pay his legal fees.
Thank you for your support!
Comments