RATZ IRR: PERFORMANCE MEASUREMENT IN THE ABSENCE OF CASH FLOW DATA

Transcription

1 RATZ IRR: PERFORMANCE MEASUREMENT IN THE ABSENCE OF CASH FLOW DATA Oxford, December 14 th 2016 John Renkema Senior Portfolio Manager Private Equity APG AM

2 CONTENT RATZ IRR: Reconstructed Average Time Zero Internal Rate of Return Quality of RATZ IRR estimation Application: US Pension Plans portfolio performance Future work & other applications 2

3 RATZ IRR RATZ IRR: Reconstructed Average Time Zero Internal Rate of Return. In the absence of cash flow data we cannot estimate the IRR of a portfolio if we only know the IRR & MM of the portfolio constituents. If we would average over IRR s, we would not take into account the effect of Time Shift, Investment Pace and Duration of the constituents. There are huge datasets available that include the performance of fund investments in terms of IRR and MM. We have developed a methodology to estimate the portfolio IRR if we have the IRR and MM for the underlying constituents. This allows us to use the data provided for different investments that have reported IRR and MM for portfolio construction analysis. 3

4 EXAMPLE 1 - TIME SHIFT EFFECT Year Cashflow 1 Cashflow 2 Portfolio Year Cashflow 1 Cashflow 2 Portfolio Year Cashflow 1 Cashflow 2 Portfolio IRR 5.00% 15.00% 10.45% IRR 5.00% 15.00% 9.34% IRR 5.00% 15.00% 11.76% MM MM MM Log(MM;(1+IRR) Log(MM;(1+IRR) Log(MM;(1+IRR) If we calculate the IRR of a portfolio of 2 investments, there is an effect on the portfolio IRR if one investment starts later than the other investment 4

5 EXAMPLE 2 INVESTMENT PACE EFFECT Year Cashflow 1 Cashflow 2 Portfolio Year Cashflow 1 Cashflow 2 Portfolio Year Cashflow 1 Cashflow 2 Portfolio IRR 5.00% 15.00% 10.99% IRR 5.00% 15.00% 10.63% IRR 5.00% 15.00% 11.36% MM MM MM Log(MM;(1+IRR) Log(MM;(1+IRR) Log(MM;(1+IRR) If we calculate the IRR of a portfolio of 2 investments, there is an effect on the portfolio IRR if one investment has a lower investment pace than the other investment 5

6 EXAMPLE 3 DURATION EFFECT Year Cashflow 1 Cashflow 2 Portfolio Year Cashflow 1 Cashflow 2 Portfolio Year Cashflow 1 Cashflow 2 Portfolio IRR 5.00% 15.00% 10.45% IRR 5.00% 15.00% 12.18% IRR 5.00% 15.00% 8.96% MM MM MM Log(MM;(1+IRR) Log(MM;(1+IRR) Log(MM;(1+IRR) If we calculate the IRR of a portfolio of 2 investments, there is an effect on the portfolio IRR if one investment has a longer duration or holding period than the other investment 6

7 RECONSTRUCTED AVERAGE TIME ZERO IRR What we do, when we have only IRR & MM is: we can reconstruct a very simple cash flow pattern for a fund i, being the cash flow in as the called capital of the fund and the cash flow out being the MM times the called capital; the moment of the cash flow in; we assume to be T 0 and the moment of the cash flow out we define as T 0 +H i, where H i is the log of the MM basis (1+IRR) (If the IRR is not equal to 0). In such a way we reconstruct cash flow patterns for each constituent of a portfolio. We can then aggregate the reconstructed cash flows and calculate a new IRR of the aggregated cash flows. Fund 1 IRR = 5% MM = 1.28 H 1 =5 Fund 2 IRR = 15% MM = 4.05 H 2 =10 Portfolio (Fund1+Fund2) RATZ IRR = % MM = 2.66 H RATZ = 8.51 Cf in at T 0 =0 Cf in at T 0 =0 Cf in at T 0 =0 Cf in = -100 Cf in = -100 Cf in = Cf out at T 0 +h 1 = 5 years Cf out =128 Cf out at T 0 +h 2 = 10 years Cf out =405 Cf out1 at T 0 +h 1 = 5 Cf out1 =128 Cf out2 at T 0 +h 2 = 10 Cf out2 =405

8 QUALITY OF RATZ IRR ESTIMATION There is a deviation in the estimation of the portfolio IRR, as we do not have the cash flow data. We can analytically express the difference between the RATZ IRR and the portfolio IRR, with a Taylor expansion. We get the following expression for the difference between the RATZ IRR and the portfolio IRR: Where MM p is the money multiple of the portfolio. The RATZ IRR methodology addresses the Duration issue of the estimation of the portfolio IRR, but not the Time shift and the Investment Pace issue. As investors do not control the start or the pace of the investments, in some cases we argue, that the RATZ IRR is a better performance measure to analyze for investors than the portfolio IRR. 8

9 QUALITY OF RATZ IRR ESTIMATION We have a large dataset of cash flow data of funds from the investment portfolio of APG executed by AlpInvest on our clients behalf. It consists of about fund investments, which we can group in different regions and investment styles, as such creating groups which do not share constituents. We have empirically checked the quality of the RATZ estimation through a portfolio of fund investments from the period If we take the full period together and compare RATZ IRR to Portfolio IRR and RATZ IRR to TZ IRR, we see the following: 9

10 QUALITY OF RATZ IRR ESTIMATION If we do the same but now look at each vintage year separately and compare RATZ IRR to Portfolio IRR and RATZ IRR to TZ IRR, we see the following: 10

11 APPLICATION: US PENSION PLANS PORTFOLIO PERFORMANCE 11 We have applied the RATZ methodology to calculate the primary fund performance of US public pension schemes. We have downloaded the portfolio constituents of the US public pension funds from the Preqin data base for the vintage years Data is as of 31/12/2014 We have downloaded the data for the 16 Largest US Public Pension funds in terms of PE program. (Massachusetts PRIMB did not have sufficient quality of data to be adopted in the results). The IRR and MM are reflected in local currency. For those Funds for which Preqin did not report a commitment level for a US Public pension fund we estimated the commitment level to be the average partition of a fund that the US Public Pension on average takes. If this was bigger than the largest commitment that was reported for the Pension Fund, we estimated the Pensions Fund commitment to be the same as the largest commitment made to an other fund. We excluded funds for which Preqin did not report either the IRR or the MM. We did not include any Secondaries or Co-Investments in the analysis, so we can not say anything about the total quality of the investment programs, only about the Primary fund investment part.

12 APPLICATION: US PENSION PLANS PORTFOLIO PERFORMANCE This is the total level of commitments from we included in the analysis from each of the US Public Pension plans: 12

13 APPLICATION: US PENSION PLANS PORTFOLIO PERFORMANCE This table represents the number of funds that the US Public Pension plans committed to in the period : 13

14 APPLICATION: US PENSION PLANS PORTFOLIO PERFORMANCE This table represents the RATZ IRR of the US Public Pension plans portfolio Money Multiples: 14

15 APPLICATION: US PENSION PLANS PORTFOLIO PERFORMANCE This table represents the RATZ IRR of the US Public Pension plans portfolio fully brought back to one RATZ IRR, not respecting the different vintage years in the RATZ IRR: 15

16 FUTURE WORK & OTHER APPLICATIONS We still would like to establish better the quality and characteristics of the RATZ IRR estimation method. We can now slice and dice the performances of a portfolio, of which we only have constituent IRRs and MMs, in any way we like. We can now construct benchmarks for portfolios in such a way that they adhere to characteristics we would like to incorporate, exact similar investments amounts in regions styles and fund sizes, etc. We could calculate the portfolio return if all funds were equally weighted or if all vintage years were equally weighted (see next slides) We have made calculations to randomly draw from the data set of Preqin and construct different portfolios to analyze the diversification effects (see next slides) 16

17 CYCLICALITY IN PERFORMANCE We can observe a cyclicality in the commitment programs of the large US pension funds. Below the relative commitments per vintage years by the pension, relative to the total commitments over the period

18 RETURN ATTRIBUTION 18 We can attribute return to allocation decisions and yearly relative commitment decisions, by running the portfolios as is and comparing it to the same runs with equal fund commitments and equal yearly allocations. As such we can attribute returns to fund commitment strategy and yearly allocations. As can be seen in the table below Eq V & 1 Com Eq V & Actual Com RATZ IRR Actual V & 1 com Actual & Actual Comm. Strategy Attribution Vintage allocation Total effect CalPERS 9.6% 11.2% 8.8% 9.5% 0.7% -1.7% -0.1% CalSTRS 10.8% 13.1% 9.1% 9.4% 0.2% -3.8% -1.5% Florida BA 12.4% 11.9% 12.0% 10.8% -1.2% -1.1% -1.5% Michigan DT 10.4% 11.2% 9.2% 9.9% 0.7% -1.4% -0.5% New Jersey SIC 7.9% 8.0% 8.1% 8.7% 0.7% 0.7% 0.9% NY State CRF 12.9% 13.7% 11.1% 12.2% 1.1% -1.5% -0.7% NY State TRS 13.2% 14.2% 11.1% 12.3% 1.2% -1.9% -0.9% Ohio PERS 10.7% 12.2% 9.2% 9.5% 0.4% -2.7% -1.2% Oregon ST 14.6% 14.4% 12.3% 11.2% -1.2% -3.3% -3.5% Pennsylvania PSERS 12.4% 14.4% 9.8% 10.4% 0.6% -4.0% -2.0% Pennsylvania SERS 10.4% 11.8% 9.9% 10.6% 0.7% -1.2% 0.2% Wisconsin IB 12.2% 12.4% 11.4% 11.4% 0.0% -1.0% -0.9% Texas Teachers 14.0% 14.8% 12.3% 11.0% -1.2% -3.7% -3.0% Virginia RS 11.1% 12.8% 9.9% 10.9% 1.0% -1.9% -0.2% Washington SIB 10.2% 12.3% 9.0% 10.0% 1.0% -2.2% -0.2% Median 11.7% 12.6% 9.9% 10.5% 0.6% -1.9% -0.9% Average 11.7% 12.6% 10.4% 10.5% 0.1% -2.1% -1.2% Standard deviation 1.9% 1.7% 1.5% 1.0% 1.1% 1.3% 1.5%

19 FUTURE WORK & OTHER APPLICATIONS We were able to construct some sort of Efficient Market Frontier. But we are still unclear about the risk parameter defining the portfolio risk here. 19

20 20 QUESTIONS