Revision of the mandate for the Government Pension Fund Global – changes to the regulation of real estate investments
14 October 2016
Norges Bank's letter to the Ministry of Finance, 10 October 2016
In Report to the Storting No. 23 (2015-2016), the Ministry of Finance announced planned changes to the regulation of real estate investments in the Government Pension Fund Global (GPFG). In line with Norges Bank’s recommendations in its letter of 26 November 2015, the fund’s real estate investments will no longer be included in the benchmark index for the GPFG, and all of the fund’s real estate investments will be included in the limit for expected tracking error (relative volatility). The Ministry also proposed capping the fund’s allocation to unlisted real estate at 7 percent. The new framework means that the level and composition of the fund’s real estate investments will be decided by Norges Bank. The Storting backed the Ministry’s proposed new framework in Recommendation 326 S (2015-2016).
The Ministry stated in the report that it would work further on the implementation of the new framework for real estate in the mandate for the GPFG in consultation with Norges Bank. In its letter of 24 June 2016, the Ministry asked for the Bank’s input on three key issues: the composition of the benchmark index, the calculation of relative risk in the management of the fund, and the revision of the mandate. The Bank’s input follows below.
We take as our starting point the proposals in the Bank’s letter of 26 November for a holistic framework that retains the key features of the current division of responsibility between the Ministry and the Bank. The Bank’s proposals of 26 November mean that the benchmark index does not only define the investment strategy but serves as a limit for overall market and currency risk in the fund.
Composition of the benchmark index
Investments in real estate are currently included in the fund’s benchmark index. Real estate investments affect overall market risk in the current benchmark index. In its letter of 24 June, the Ministry asked the Bank to advise on an equity share in the new benchmark index that would be consistent with keeping the level of market risk in the fund largely in line with the restrictions in the current mandate. The discussion below is based on a global unlisted real estate portfolio equivalent to 5 percent of the fund.
The market risk associated with unlisted real estate investments can be measured by looking at the degree to which the return on these investments covaries with movements in the stock market. The return on a global portfolio of unlisted real estate investments cannot be observed in the market. The market risk associated with these investments must therefore be calculated using the most representative time series possible. There are several possible approaches. One key distinction is between time series based on estimated market values for unlisted real estate investments and those for shares in listed real estate investment trusts (REITs).
MSCI/International Property Databank (IPD) produces global indices for unlisted real estate investments. The IPD indices are based on reported valuations. Studies have shown that valuations in one period are influenced by valuations in previous periods, which can lead to time series fluctuating less than actual property prices. When allowance is made for this, volatility in the time series will normally increase and approach “market pricing”. There are various methods for performing such an adjustment, and the estimate of market risk will vary with the choice of method. In our calculations, we have used the de-smoothing method employed by Ang, Brandt and Denison in their 2014 report to the Ministry.1
Another approach is to use investments in listed REITs to represent movements in the value of unlisted real estate investments. The properties in these funds will normally be leveraged. The GPFG’s management mandate requires the fund’s allocation to real estate to be calculated on the basis of net market value – in other words the gross value of the properties less any associated debt. We therefore need to adjust the time series for shares in listed REITs to eliminate the effect of leverage in order to obtain a representative time series. Such an adjustment does not take account of the credit spread a borrower must pay having a tendency to increase in periods of financial market turmoil. The increase in credit spreads is a possible reason why the estimate of market risk is higher for the period that includes the 2008 financial crisis.
The calculations of market risk presented in Table 1 in Enclosure 1 are sensitive to the choice of data, estimation period and frequency, and the estimate of market risk for unlisted real estate investments ranges from 0.2 to 0.8. There is thus no definitive answer to the question of how investments in unlisted real estate affect overall market risk in the benchmark index. The decision on adjusting the equity allocation in the benchmark index must therefore be to some extent a matter of judgement. Based on the review above, the Bank considers it reasonable to assume that a diversified global portfolio of unlisted real estate investments will have a market risk over time of around 0.5. This means that the return on unlisted real estate investments is expected to move in the same direction, but not as far, as the return on the broad equity market.
The new framework for real estate investments in the GPFG has similarities with that for comparable funds. The Bank is in regular contact with several of these funds. The Bank’s assessment of market risk in unlisted real estate investments ties in well with the approach chosen by our peers. The estimate of 0.5 is also in line with the findings of academic studies of covariance between the return on unlisted real estate investments and the return on equities.2 If the market risk in the GPFG is to be kept largely unchanged relative to the current mandate, and we assume a market risk of around 0.5, this indicates that the equity allocation in the benchmark index should be increased by around 2.5 percentage points.
The equity allocation in the actual benchmark index moves within a band of +/- 4 percentage points around the strategic target. If the Ministry decides to change the equity share in the strategic benchmark index, consideration will need to be given to how the equity share in the actual benchmark index is to be adjusted. The Bank assumes that we can return to the issue of how such an adjustment should be implemented.
The Bank believes that a strategic equity allocation of 62.5 percent will keep the market risk in a benchmark index comprising only equities and bonds at around the same level as in the current management mandate for the GPFG, given a global unlisted real estate portfolio equivalent to 5 percent of the fund.
Calculation of expected relative risk in the management of the fund
In its letter of 26 November 2015, the Bank assumed that the Ministry would still regulate deviations from the benchmark index through a limit for expected tracking error. The Ministry confirmed this in Report to the Storting No. 23 (2015-2016), which announced that expected tracking error will be kept as the key risk limit, partly because the Ministry finds it appropriate to use a measure of risk with which there is long operational experience.
In the new framework, the fund’s real estate investments are to be included in the calculation of expected tracking error. To do so, it will be necessary to calculate a representative time series for the fund’s unlisted real estate investments. We have assessed three different methods.
The first uses data from IPD for unlisted real estate investments, adjusted for autocorrelation. The return series from IPD can be broken down into the countries and sectors in which the fund is invested. The greatest challenge when using IPD data for calculating tracking error is that the time series are updated only quarterly or annually. The relative volatility of equities and bonds is currently calculated using weekly data, equally weighted, over a three-year period. Even an extension of the estimation period to ten years, for example, will yield relatively few observations if the calculations have to be performed on quarterly or annual data.
The second method uses data for shares in listed REITs, adjusted for leverage. The main benefit of using REITs over IPD data is the availability of observable daily prices. To be able to represent the fund’s unlisted real estate investments meaningfully, we need to select individual funds in the markets in which the fund is invested. Their leverage must also be adjusted to the same level as the “equivalent” investment in the fund. This selection process and adjustments to take account of differences in risk profile will to some extent need to be based on criteria that will be difficult for experts outside the bank to verify.
The third method is based on an external risk model developed by MSCI. The Bank has commissioned MSCI to compute a return series for an unlisted real estate portfolio that resembles the GPFG’s portfolio of unlisted real estate investments. An external risk model gives the Bank less insight into, and less control over, the parameters that influence the return series, but has the advantage of being calculated by an independent party.
Tables 2 and 3 in Enclosure 1 compare the three methods. The tables show the isolated effect of unlisted real estate investments on expected tracking error. We assume in the tables that the equity allocation in the strategic benchmark index is 62.5 and 60 percent respectively, and that investments in unlisted real estate do not affect the fund’s currency composition. We find that expected tracking error is lowest using the method with data for shares in REITs, highest using the method with data from IPD, and somewhere between the two using MSCI’s risk model. We see that the allocations to equities and bonds in the portfolio vary with the level of real estate investments, the equity allocation in the benchmark index and the method for calculating expected tracking error.
Expected tracking error in the GPFG is currently calculated on the basis of weekly observations over a three-year period. When choosing the estimation period and frequency, we need to weigh the tendency for short time series to contain limited information about potential market movements against the need to adjust longer time series for any shortcomings in the data series. We recommend continuing with weekly observations over a three-year period. This rules out the use of IPD data, which will at best be available on a quarterly basis. The Bank also considers that it may be appropriate to choose a method for estimating expected tracking error that can be verified by experts outside the bank. This favours the use of MSCI’s risk model. The Executive Board will decide on the method for estimating expected tracking error by the end of 2016 and, in line with the mandate, submit it to the Ministry for approval.
The method for estimating the expected tracking error of the fund’s unlisted real estate investments should be verifiable by experts outside the Bank. The Bank is therefore looking at introducing MSCI’s risk model (MSCI Barra PRE2).
Revision of the mandate in light of the new framework for real estate investments
When the Ministry first proposed allowing part of the fund to be invested in real estate in Report to the Storting No. 16 (2007-2008), importance was attached partly to the potential for investments in real estate to contribute to better risk diversification in the fund. Accordingly, the Bank has built up the real estate portfolio with a view to improving the fund’s diversification of risk. The Bank believes that this should remain the aim of investing parts of the fund in real estate. The Ministry has asked the Bank to advise which provisions in the mandate for the GPFG should be revised in light of the new framework for real estate. The Bank’s recommendations are presented in Enclosures 2 and 3. The most important changes that need to be made now are summarised below.
In the new framework, the strategic and actual benchmark indices will consist of just two asset classes: equities and bonds. Section 1-5 Strategic benchmark index and Section 1-6 Actual benchmark index must therefore be amended.
Going forward, purchases and sales of real estate will be funded by sales and purchases of both equities and bonds. The best way to carry out these transactions is to allow the securities sold to vary with the type of property. The sales can then be made in a way that eliminates currency risk and limits market risk. Thus, for example, the purchase of a property in the UK might be funded through the sale of a combination of British equities and British bonds. Changes in market conditions might bring a need to adjust funding along the way. An investment-specific funding solution of this kind will best serve the fund’s overall performance and risk, and underlies the Bank’s proposals for a holistic framework in its letter of 26 November 2015. An investment-specific funding solution will also be robust to changes in the composition of the actual benchmark index.
Assuming the choice of investment-specific funding, there will be a need to amend some of the provisions in the mandate so that we can continue to manage, measure and report on our management results in a relevant and consistent manner. In our proposed revised mandate, this can be seen partly in Section 3-1 Investment universe and Section 3-5 Management restrictions, where we propose that in future these limits are set for the investment portfolio as a whole rather than for the sub-portfolios of equities, bonds and real estate.
The Bank recommends that in Section 3-5 the Ministry sets intervals for how much of the investment portfolio may be invested in equities, bonds and unlisted real estate respectively. The constraints on the Bank’s active risk-taking are retained in the same form as today. We also recommend that the limit for investments in high-yield bonds is stated relative to the size of the fund as a whole rather than as a share of the bond portfolio, as the level and composition of bond investments will depend on real estate investments.
Section 3-6 Management restrictions to be set by the Bank currently contains limits to be set by Norges Bank. We recommend that the mandate specifies more precisely that these limits are to be set by the Executive Board. The Executive Board would not be able to delegate responsibility for setting these limits. The largest and most important change to Section 3-6 of the current mandate is the proposal that the Executive Board establishes reference portfolios for different parts of the investment portfolio. These changes are needed to allow an investment-specific funding solution for real estate investments.
The composition of the reference portfolios for equities and bonds will depend on both the level and the composition of the fund’s real estate investments. The reference portfolios will serve as a starting point for the operational management of the fund, make it possible for the Executive Board to set limits for risk-taking for different parts of the investment portfolio, and form the basis for relevant and consistent reporting of results in different areas. At present, the sub-indices for equities and bonds defined in the management mandate are used for these purposes. For the reference portfolios to serve as objective yardsticks for different parts of the fund’s management, they need to be verifiable. Norges Bank has developed effective systems over a number of years for establishing, tracking and documenting the composition of operational reference portfolios. These systems are well-suited to the job of establishing reference portfolios for the fund’s management.
The Bank is currently permitted to hold more than 10 percent of the voting shares in companies included in the real estate portfolio, whether the company is listed or unlisted. The Bank’s proposed revised mandate retains this option. All of the Bank’s unlisted real estate investments are organised as limited companies. The Bank must be able to hold more than 10 percent of the shares in these companies. Being permitted to invest in both unlisted and listed real estate gives us access to a broader set of investment opportunities, which can potentially improve the fund’s diversification. The option of holding more than 10 percent of the shares in listed real estate gives the Bank scope to consider different ways of working with other players. The size of investments in listed real estate companies will be limited and within the restrictions set by the Executive Board.
Section 4-3 Measurement and management of market risk in the present mandate requires the method for estimating expected tracking error to be determined by the Bank and approved by the Ministry. The Bank recommends that in future the method for estimating expected tracking error is to be determined by the Executive Board and submitted to the Ministry no later than four weeks before implementation. In this light, it is proposed that this provision is moved to a new Section 3-3 Management restrictions to be set by the Executive Board.
The change in the composition of the benchmark index and the decision to delegate responsibility for the size and composition of real estate investments to Norges Bank mean that the wording of Section 6-2 Reporting requirements needs to be amended. The proposed changes to the reporting requirements do not affect the stated objective of public reporting. The Bank's reporting is to contribute to the greatest possible transparency regarding the management of the fund. The public reporting is to provide a true and comprehensive overview of how the Bank executes its management duties. Our goal is for the Norwegian people and other stakeholders to find all the information they require about the fund and its investments, unless this information is market-sensitive or cannot be disclosed as a result of agreements we have entered into.
The public reporting on risks and returns in the real estate portfolio has been extended in 2016 with a separate real estate report in which we aim to paint the broadest possible picture of the drivers of returns on our real estate investments and the types of risk these investments are exposed to. We plan to develop this report further to ensure the greatest possible transparency regarding the management of unlisted real estate investments. In future reports, the return on unlisted real estate investments will be compared with the return on a broad set of alternatives, such as IPD, an index of REITs, the benchmark index for the GPFG and the funding portfolio, cf. Section 6-2(3)(e) in the Bank’s proposed revised mandate.
The changes now being recommended are in line with the principle that the management mandate should be general in nature and based on limits. The new framework for real estate in the GPFG does, however, highlight the need for a general review of the mandate. We can return to this in connection with the Bank’s input on the planned review of the fund’s management.
The management mandate for the GPFG should ensure a holistic management approach. The Bank’s recommendations for how this should be achieved are presented in Enclosures 2 and 3.
Øystein Olsen Yngve Slyngstad
2. The Bank’s proposed revised mandate
3. Comparison of existing and revised mandates
1 Ang, A., M. Brandt and D. Denison (2014): ‘Review of the Active Management of the Norwegian Government Pension Fund Global’, https://www0.gsb.columbia.edu/faculty/aang/papers/AngBrandtDenison.pdf.
2 See, for example, Ang, Brandt and Denison (2014), Appendix C, Equation (A.3).