chapter 27: real estate investment management part ii ... · chapter 27 outline: introduction 27.1...

Chapter 27:Chapter 27:Real Estate Investment Management Part II: Real Estate Investment Management Part II:

Performance Attribution & EvaluationPerformance Attribution & Evaluation

• What is meant by investment “performance attribution” at both the “macro-property level” and the portfolio level;

• How to quantify segment “allocation” versus asset “selection” effects in a portfolio’s differential performance relative to an appropriate benchmark;

• What is meant by formal, quantitative investment performance evaluation, and the role of this function in the relationship between investment managers and their investor clients;

• The nature of manager “custom benchmarking” in the private real estate asset class, and how this differs from corresponding practices in the public securities investment industry.

Chapter 27 Learning ObjectivesChapter 27 Learning Objectives

Chapter 27 Outline:Introduction27.1 Macro-level Investment Performance Attribution

27.1.1 Macro property level performance attribution27.1.2 Portfolio level performance attribution27.1.3 The use of a benchmark in performance attrib.27.1.4 The case for using mgr alloc wts in the bnchmk

27.2 Investment Performance Evaluation & Benchmarking27.2.1 The basic idea...27.2.2 Benchmarks in the private R.E. asset class27.2.3 Matching evaluation, responsibility, & auth.27.2.4 The problem of statistical significance27.2.5 Implications of the lack of stat. significance27.2.6 Adjusting for risk

27.3 Conclusions

MacroMacro--Level Investment Policy Analysis:Level Investment Policy Analysis:•• StrategicStrategic Investment Decision Making:Investment Decision Making:

-- LongLong--run objectives & plan:run objectives & plan:-- Chapter 21 (Chapter 21 (PortfPortf Theory) Theory) Basic methodologyBasic methodology..-- Chapter 22 (Chapter 22 (EquilibrEquilibr Models) Models) Rational expectations to apply Rational expectations to apply MPT.MPT.

•• TacticalTactical Investment Policy:Investment Policy:-- Short/medium term actions (opportunistic):Short/medium term actions (opportunistic):-- Chapter 22 (Chapter 22 (EquilibrEquilibr Models) Models) Find Find mismis--priced assets.priced assets.

•• Policy Policy ImplementationImplementation::-- Investment management:Investment management:-- Chapters 26 & 27 (Chapters 26 & 27 (PerfPerf. Attrib., Benchmarking).. Attrib., Benchmarking).-- Chapter 22 (Chapter 22 (EquilibrEquilibr Models) Models) RiskRisk--adjustment for adjustment for perfperf. Attrib.. Attrib.-- Most Most releventrelevent for private direct investment for private direct investment -- Also relevant for REIT mgt to consider, as Also relevant for REIT mgt to consider, as REITsREITs can be viewed as can be viewed as mgrs of direct mgrs of direct privpriv R.E. investments on behalf of stockholders.R.E. investments on behalf of stockholders.

And, of course, Chapters 23, 24 & 25 (macro-level valuation & return data issues) are basic to the underlying quantitative analysis of all three of these macro-level investment policy analysis functions.

DEFINITIONDEFINITION: : The formal (quantitative) The formal (quantitative) comparisoncomparison of a given of a given investment agent’s (or portfolio’s) investment investment agent’s (or portfolio’s) investment performance with that of a suitably defined index or peer performance with that of a suitably defined index or peer group, over a specified time period.group, over a specified time period.

EXAMPLEEXAMPLE::•• A largeA large--cap stock manager is “benchmarked against the cap stock manager is “benchmarked against the S&P500”…S&P500”…•• Manager averages a 10% return over a 3Manager averages a 10% return over a 3--yr mgt yr mgt contract period…contract period…•• S&P500 averages 12% over the same period…S&P500 averages 12% over the same period…•• Manager has some “Manager has some “explainingexplaining” to do.” to do.

What isWhat is““BenchmarkingBenchmarking”” ??

Same idea in real estate… An “institutional core” manager might be benchmarked against the NCREIF Index, similar to a large-cap stock manager being benchmarked against the S&P500.

DEFINITIONDEFINITION: : The The decompositiondecomposition of the total investment performance of the total investment performance into additive componentsinto additive components so as to “attribute” the total so as to “attribute” the total performance to sources that may reflect various performance to sources that may reflect various investment management investment management functionsfunctions. .

EXAMPLEEXAMPLE::•• Manager 10% Manager 10% vsvs NCREIF 12%, but…NCREIF 12%, but…•• If mgr had allocated among different property types in If mgr had allocated among different property types in same proportion as NCREIF Index, then mgr would have same proportion as NCREIF Index, then mgr would have earned 13%:earned 13%:

Mgr relative good at individual asset Mgr relative good at individual asset ““selectionselection””;;Source of poor relative performance must be Source of poor relative performance must be

“allocation”“allocation” across property type market segments.across property type market segments.

What isWhat is““Performance AttributionPerformance Attribution”” ??

Performance attribution is a useful extension and complement to simple benchmarking of the total return performance. Performance attribution can help diagnose the source or cause of the relative performance differential between the subject manager and his benchmark.

Some historical Perspective…Some historical Perspective…•• MPT (1950s & 60s)MPT (1950s & 60s)

•• CAPM (1960s & 70s)CAPM (1960s & 70s)InvstmtInvstmt Mgt Mgt PerfPerf..EvalEval: : ““RiskRisk--adjusted returnsadjusted returns””::“Jensen’s Alpha”, “Sharpe Ratio”, ““Jensen’s Alpha”, “Sharpe Ratio”, “TreynorTreynor Ratio”.Ratio”.

•• Disillusionment with CAPM (1980s & 90s):Disillusionment with CAPM (1980s & 90s):Going beyond “beta” to explain expected returns: Going beyond “beta” to explain expected returns: Size, Size, B/M Ratio,…B/M Ratio,…

•• ““StyleStyle” oriented ” oriented invstmtinvstmt strategy (1980s & 90s):strategy (1980s & 90s):StyleStyle--based benchmarking as an based benchmarking as an invstmtinvstmt mgt tool, mgt tool,

consistent with styleconsistent with style--based portfolio strategy.based portfolio strategy.

•• Property “Crash” of early 1990s:Property “Crash” of early 1990s:Benchmarking expands from equities to R.E. (1990s).Benchmarking expands from equities to R.E. (1990s).

Modern finance really begins with Markowitz Portfolio Theory (aka “Mean-Variance Portfolio Theory”, or “Modern Portfolio Theory” – MPT for short). Developed in the 1950s, spread to widespread application during the 1960s. The need for a more practical method for applying MPT within the stock market led to the development of the Sharpe-Lintner “Capital Asset Pricing Model” (CAPM), by simplifying MPT and embedding it within the broader and more rigorous economic framework of asset market equilibrium pricing theory. Initially developed in the 1960s, the CAPM was widely applied in practice during the 1970s. Helped to stimulate the development of the mutual fund industry, especially the advent of passive “index” funds. CAPM carried implications for quantitative investment performance evaluation. A money manager’s total return performance could be adjusted for risk, using “beta” as the risk measure that determines the capital market’s expected returns. E.g., compare managers or portfolios on the basis of their “Treynor Ratios” rather than simple unadjusted total returns. But, while helpful, the CAPM turned out to be too simplistic, too “blunt” a tool for practical investment manager performance evaluation. From an academic perspective, other asset characteristics besides beta were found to be important in explaining expected returns, most prominently size and book/market ratio (e.g., Fama-French studies), and current yield (e.g., Campbell, Shiller studies). From a practical industry perspective, risk-adjusted returns never really gained dominance over unadjusted returns as a widely-applied performance evaluation metric. Practical investment strategy in the equity industry evolved toward a “Style” oriented perspective, with the styles largely reflecting different risk and return characteristics, such as size and value-vs-growth characteristics (e.g., large,mid,&small cap; value,growth,&neutral orientation – the latter often related to B/M ratio &/or yield). This development occurred in the 1980s. The greater practicality of style-based investment strategy made formal, quantitative benchmarking a more useful investment performance evaluation and management tool. Equity managers with a common style and specialization could be logically compared with each other and with a common benchmark reflective of their style and specialization. Growth in the use & importance of benchmarking in the equities industry took off in the 1980s & 90s. Style-based benchmarking tends to be a less “academic” exercise than earlier risk-adjusted return oriented techniques, based more on common sense than on highly sophisticated academic theory. (No one is likely to win a Nobel Prize for developing benchmarking theory.) But this is part of the reason why benchmarking has become such a widely-used tool. Professional portfolio management in the private (direct) real estate investment industry really only began in a big way in the 1970s. At first (in the 70s & 80s), there was an attempt to fit real estate investment strategy and management into the MPT/CAPM mold. But this didn’t work very well, even as the equity industry was evolving toward the style-based benchmarking model of investment strategy and mgt. The property “crash” of the early 1990s increased the need for credibility and accountability in the R.E. investment management industry, and led to the growth of incentive contracting and formal benchmarking in the private R.E. investment industry during the 1990s.

•• 27.1. Performance Attribution27.1. Performance Attribution

•• 27.2. Basics of Benchmarking27.2. Basics of Benchmarking

•• 27.2 (cont.). Issues in Benchmarking 27.2 (cont.). Issues in Benchmarking Private Real EstatePrivate Real Estate

Lecture Outline:Lecture Outline:

27.1. Performance Attribution27.1. Performance Attribution

•• Definition & Overview.Definition & Overview.

•• 27.1.2. Portfolio27.1.2. Portfolio--Level Performance Attribution.Level Performance Attribution.

•• 27.1.1 Property27.1.1 Property--Level Performance Attribution.Level Performance Attribution.

DEFINITION: DEFINITION: The The decompositiondecomposition (or “breaking down”, or “parsing”) of (or “breaking down”, or “parsing”) of the total investment return of a subject property or the total investment return of a subject property or portfolio of properties (or an investment manager).portfolio of properties (or an investment manager).

PURPOSE:PURPOSE:To assist with the To assist with the diagnosisdiagnosis and understanding of what and understanding of what caused the given investment performance.caused the given investment performance.

USAGE:USAGE:By investment managers (agents) and their clients By investment managers (agents) and their clients (principals).(principals).REIT managers also could use.REIT managers also could use.

Performance Attribution Overview Performance Attribution Overview . . .. . .

•• Portfolio levelPortfolio levelPertains to dynamic portfolios or investment manager Pertains to dynamic portfolios or investment manager (or fund) level.(or fund) level.

•• Property levelProperty levelPertains to individual properties or static portfolios of Pertains to individual properties or static portfolios of multiple properties.multiple properties.


Two levelsTwo levelsat which performance attribution is performed:at which performance attribution is performed:

Portfolio level performance attribution was developed in (and for) the public equity investment industry, and is carried into private real estate pretty much as it is applied in that industry. It focuses on the functions of active portfolio managers, those who try to “beat the market” essentially by doing one or both of two functions: (i) Allocating (and re-allocating) the overall portfolio among different sectors or segments of the stock market (e.g., energy, technology, services, producer goods, consumer staples, consumer cyclicals, alcohol & tobacco, health-care, financial services, utilities, REITs, etc.); (ii) Picking individual stocks within each categories (relative “winners” vs “losers”). Active portfolio management of this type involves relatively high turnover, relatively short average holding periods for individual assets, as compared to “passive” portfolio management such as index funds, which simply try to mimic the holdings of a stated market index. Investors pay for active fund management in higher fees. It makes sense to compare active fund manager performance to an appropriate passive market index, and for diagnostic purposes to break down the sources of the relative performance of the fund manager into allocation and asset-selection components corresponding to these two basic functions of active portfolio management. Private real estate investment management is also “active”, in that the investment manager does not simply duplicate the holdings of a passive index, and strives to add value by active management of the investments. Indeed, it is not possible to have truly “passive” management in private real estate investment for at least two reasons: (i) There is currently no passive market index that is investable and broadly representative of the entire relevant asset market (as noted in Lecture 3, such an index would have to be based on a statistical sampling type of procedure, and would still not be truly “investable” at the individual asset level, because whole assets are traded by individual owners at their discretion); and (ii) More fundamentally and importantly, private real estate investment brings with it the inescabable responsibility (and the important profit-making opportunity) for direct operational management of the individual assets that are being invested in. This is a different type of “active” management, not involving rapid turnover of assets held, but very important in affecting long-term total return performance nonetheless. It suggests that portfolio-level performance attribution as it is practiced in the equities industry is less meaningful and less complete, for helping to understand private real estate investment performance, although the portfolio level functions of allocation and selection are still important and of interest in private real estate. But direct real estate investment is a long-term, income-oriented type of investment (particularly for “core” investments made in stabilized fully-operational property with little leverage). How well the properties are managed can have a big impact on the investment total return performance. This suggests the need for “property level ” performance attribution, a second level of performance attribution that has been developed particularly for the private real estate investment industry, reflecting the fact that direct real estate investment carries with it the need for the investors to actually directly manage or operate the assets they are investing in, something equity money managers do not have to do (indeed, can’t do). Property level performance attribution is important because individual assets are typically held for long periods of time (in part due to high transaction costs) Property level performance attribution is a way to try to evaluate and understand the operational management level of real estate investment performance. I believe much of the credit for developing property-level performance attribution as a formal methodology integrated into the pre-existing portfolio-level exercise goes to the real estate investment industry in England in the 1980s, and in particular the Investment Property Databank (IPD) firm that was established there during that period.

Major Major attributesattributes (return components):(return components):

At the PORTFOLIO LEVEL:At the PORTFOLIO LEVEL:AllocationAllocationSelectionSelection

InteractionInteraction

At the PROPERTY LEVEL:At the PROPERTY LEVEL:Initial Cash YieldInitial Cash Yield

Cash Flow ChangeCash Flow ChangeYield ChangeYield Change


(Aside: Note that the break-down I am suggesting here at the property level is slightly different than the traditional attribution supplied by IPD and the Australian PCA in their standard performance attribution reports. In particular, I am stressing the identification and distinction of the initial cash yield, that is, the yield at the time when the property was acquired (the “acquisition cash yield”, similar to the “going-in cap rate”). I feel that, to be useful as a performance diagnostic tool, property-level performance attribution is inherently a multi-period exercise, corresponding to the period over which the (subject) property (or portfolio) has been held by the manager. One of the important functions a professional real estate investment manager is hired to perform is property “acquisition”, which involves more than just what the active stock market money manager does in stock “picking”. The private real estate investment manager does not have a highly liquid and informationally efficient asset market to protect against “over-paying” for an asset. And the private real estate investment manager must negotiate a purchase price in a private deal negotiation that can often be quite complex. Handling this acquisition process can be quite complicated, and requires skill and expertise to execute effectively. It is definitely a source of private real estate investment relative performance among managers, for good or bad. The initial (at acquisition) cash yield (relative to that of a common inception-date cohort of similar properties or managers), can provide some prima facie evidence of how well this function has been performed for the subject property or portfolio.

PortfPortf Tot.Return Tot.Return –– BnchmkBnchmk Tot.ReturnTot.Return

AllocationAllocation SelectionSelection InteractionInteraction

Portfolio Level:Portfolio Level:

Prop.IRR Prop.IRR –– BnchmkBnchmk Cohort IRRCohort IRR

Init.YieldInit.Yield CF GrowthCF Growth Yield Yield ChgeChge

Property Level:Property Level:


•• At both levels, At both levels, diagnosticdiagnostic purpose is facilitated by purpose is facilitated by comparison between comparison between subjectsubject portfolio or mgr with an portfolio or mgr with an appropriateappropriate benchmarkbenchmark..•• Relative Relative (or (or differentialdifferential) performance ) performance betwbetw subjsubj vsvsbenchmrkbenchmrk is quantified, in total & w/in components.is quantified, in total & w/in components.

If the data is available, it is conceptually possible to “drill down” deeper into the sources of property-level performance. In particular, the initial yield and cash flow growth components can be attributed to gross revenue generation, operating expense management, and/or capital expenditures management (the latter can also affect the yield-change component). However, at these deeper levels, it probably typically makes more sense to compare relevant performance measures directly with the benchmark on a dollars per square foot basis rather than to attempt to quantify them as components of a multi-period total return percent measure. For example, compare rental revenue per square foot, and operating expenses per square foot, (and multi-year average annual capital expenditures per square foot, or per dollar of net operating income), between the subject property or portfolio and the benchmark (inception date cohort) portfolio of properties.

AllocAlloc Strategy Strategy Across SegmentsAcross Segments

Asset Picking & Asset Picking & OperOper.Mgt.Mgt

AllocAlloc & Selection & Selection SynergySynergy

AllocationAllocation SelectionSelection InteractionInteraction

For example, at the For example, at the portfolioportfolio level…level…

Performance Attribution Overview Performance Attribution Overview . . .. . .Basic idea is that components of the performance Basic idea is that components of the performance differential between the subject & the benchmark reflect differential between the subject & the benchmark reflect performance of various distinct performance of various distinct functionsfunctions of investment of investment management.management.

Note: In private R.E., the asset “Note: In private R.E., the asset “SelectionSelection” ” function also includes asset function also includes asset Operational Operational Management.Management.

Note that the “selection” attribute at the portfolio level for private real estate, as distinct from public equities, inherently includes not only the “asset picking” function as it is understood in the equity industry (picking properties that are “bargains”, or that will tend to do particularly well in the future, from an investment perspective), but also involves the function of operational management of the properties that are invested in, as the private real estate investor has direct responsibility for and control over operational management of the assets invested in. It is impossible to distinguish the “asset picking” performance (per se, as it is understood in the equities industry) from the “operational management” performance at the portfolio attribution level. However, by going to the next level, to property-level performance attribution, some insight can be gained on the manager’s operational management performance.

Init.YieldInit.Yield CF ChangeCF Change Yield Yield ChgeChge

AcquisitionAcquisition

SelectionSelection

OperationalOperational

MgtMgt

Mgt & Mgt & DispDisp

SelectionSelection


For example, at the For example, at the propertyproperty level…level…

Basic idea is that components of the performance Basic idea is that components of the performance differential between the subject & the benchmark reflect differential between the subject & the benchmark reflect performance of various distinct performance of various distinct functionsfunctions of investment of investment management.management.

The relationship between the three performance attributes identified here and the corresponding management functions is always a bit fuzzy and ambiguous in practice.

A basic problem with all performance A basic problem with all performance attribution analyses:attribution analyses:

It is generally impossible to define unique, It is generally impossible to define unique, unambiguous breakunambiguous break--downs or components of the downs or components of the total return that correspond to clear causal total return that correspond to clear causal determinants or investment management determinants or investment management functions.functions.

Nevertheless,Nevertheless,Some useful insights can be obtained from Some useful insights can be obtained from

performance attribution, provided results are performance attribution, provided results are used carefully.used carefully.


For example: The selection component at the portfolio level includes also the effect of asset operational management, and does not include the synergistic “interaction” component that reflects the combined effects of allocation and selection. The allocation component at the portfolio level is “incomplete” in the sense that it does not include allocation’s role in the synergy between allocation and selection, as represented by the “interaction” component. The interaction component at the portfolio level is difficult to interpret in terms of specific portfolio-level investment management functions, because it is a combination resulting from all functions interacting together. The property-level IRR-based performance does not aggregate up exactly to the selection component of the TWRR-based portfolio-level performance. The property-level yield-change component may be due to any or a combination of: (i) asset operational management performance, (ii) acquisition and/or disposition asset selection performance, or (iii) changes beyond the manager’s control in the micro-level market for the specific subject type of property. The property-level initial yield component may reflect asset acquisition selection performance, or simply the relative quality of the property and its existing leases (e.g., the presence of over-market leases that are about to expire could cause a high initial yield without implying any sort of “bargain” purchase. The property-level cash flow change component may reflect asset operational management performance, or simply the effect of the pre-existing lease expirations (e.g., expiration of pre-existing below-market leases would naturally lead to a jump in cash flow). In summary, be careful in interpreting the meaning of performance attribution measures. By themselves, these measures may be more useful for raising questions than providing answers. Simply by trying to answer the questions raised by such analysis, managers may be stimulated to think about and discover things about how they are performing their jobs.

•• Performance attribution lacks significance for any one Performance attribution lacks significance for any one short holding period (e.g., quarter or year).short holding period (e.g., quarter or year).

•• When aggregating across periods:When aggregating across periods:•• At the At the portfolio levelportfolio level (i.e., for (i.e., for allocationallocation and and selectionselection attribution), it is generally best to use the attribution), it is generally best to use the arithmetic TWRRarithmetic TWRR of the attribution components (for of the attribution components (for component component additivityadditivity).).•• At the At the property levelproperty level (i.e., for (i.e., for initial yieldinitial yield, , cash flow cash flow changechange, and , and yield changeyield change attribution), it is probably attribution), it is probably more appropriate to use the more appropriate to use the IRRIRR, because capital , because capital flow timing is controlled by the property flow timing is controlled by the property (investment) manager at the property level.(investment) manager at the property level.


MultiMulti--period Attributionperiod Attribution

Simple Numerical Example:Simple Numerical Example:•• Bob & Sue are competing property investment managers Bob & Sue are competing property investment managers specializing in office and industrial property.specializing in office and industrial property.•• Over the same time period they made the following segment Over the same time period they made the following segment allocations and achieved the following returns…allocations and achieved the following returns…

Exhibit 27-2: Bob & Sue's returns realized for clients:Weights: Bob: Sue:Industr 90% 10%Office 10% 90%Returns: Bob: Sue:Total Portfolio 9.20% 9.70%Industrial Properties 9.00% 7.00%Office Properties 11.00% 10.00%

••Sue beat Bob by 50 basis points, 9.70% to 9.20%, in total portfoSue beat Bob by 50 basis points, 9.70% to 9.20%, in total portfolio lio performance, which is what counts.performance, which is what counts.•• But how can we But how can we attributeattribute this total performance differential between this total performance differential between the two major functions an investment manager performs @ the two major functions an investment manager performs @ portfportflevel: segment level: segment allocationallocation, and asset , and asset selectionselection?…?…

27.1.2 Portfolio27.1.2 Portfolio--LevelLevel Performance Attribution Performance Attribution . . .. . .

This example is taken from Geltner-Miller, Chapter 27,pp.721-726, where it is described in more depth.

Suppose (to simplify the illustration) that Bob’s performance isSuppose (to simplify the illustration) that Bob’s performance isidentical to whatever benchmark index would be an appropriate identical to whatever benchmark index would be an appropriate reference for Sue’s performance. Then we can attribute Sue’s reference for Sue’s performance. Then we can attribute Sue’s differential performance, with respect to her benchmark, to allodifferential performance, with respect to her benchmark, to allocation cation ((AASS––AABB) and selection (S) and selection (SSS––SSBB)) as follows:as follows:

SSSS––SSBB = = wwBIBI((rrSISI –– rrBIBI) + ) + wwBOBO((rrSOSO –– rrBOBO))

= 0.9(7%= 0.9(7%--9%) + 0.1(10%9%) + 0.1(10%--11%) 11%)

= = --1.8% 1.8% -- 0.1% = 0.1% = --1.9%1.9%

27.1.3. Use of a Benchmark in Performance Attribution 27.1.3. Use of a Benchmark in Performance Attribution . . .. . .The most logical way to perform this type of comparative attribuThe most logical way to perform this type of comparative attribution tion analysis is to define a common analysis is to define a common benchmarkbenchmark that is an appropriate that is an appropriate reference point for both Bob’s and Sue’s performance.reference point for both Bob’s and Sue’s performance.

AASS––AABB = = rrBIBI((wwSISI –– wwBIBI) + ) + rrBOBO((wwSOSO –– wwBOBO))

= 9%(0.1= 9%(0.1--0.9) + 11%(0.90.9) + 11%(0.9--0.1) 0.1)

= = --7.2% + 8.8% = +1.6%7.2% + 8.8% = +1.6%

Sue's returns vs her benchmark:Weights: Benchmark Sue:Industr 90% 10%Office 10% 90%Returns: Benchmark: Sue:Total Portfolio 9.20% 9.70%Industrial Properties 9.00% 7.00%Office Properties 11.00% 10.00%

Clearly, Sue outperformed her benchmark (9.70% Clearly, Sue outperformed her benchmark (9.70% vsvs9.20%, a +50 9.20%, a +50 bpbp differential) because of her superior differential) because of her superior allocation (90% to the office segment that did better than allocation (90% to the office segment that did better than industrial property), not because of her selection (which industrial property), not because of her selection (which actually hurt her performance, as her properties did worse actually hurt her performance, as her properties did worse than the benchmarks within each segment).than the benchmarks within each segment).

The The purepure effect of her differential performance in the effect of her differential performance in the selection and allocation functions is quantified as follows…selection and allocation functions is quantified as follows…

PortfolioPortfolio--Level Performance Attribution Level Performance Attribution . . .. . .

The pure effect of Sue’s allocation choices, relative to The pure effect of Sue’s allocation choices, relative to her benchmark, is quantified as the sum across all her benchmark, is quantified as the sum across all segments, of the difference between Sue’s allocation segments, of the difference between Sue’s allocation and the benchmark’s allocation multiplied by the and the benchmark’s allocation multiplied by the benchmark’sbenchmark’s return in each segment.return in each segment.



The pure effect of Sue’s asset selection, relative to her The pure effect of Sue’s asset selection, relative to her benchmark, is quantified as the sum across all segments benchmark, is quantified as the sum across all segments of the difference between Sue’s withinof the difference between Sue’s within--segment return segment return and the benchmark’s withinand the benchmark’s within--segment return, weighted by segment return, weighted by the the benchmark’sbenchmark’s allocation to each segment.allocation to each segment.


Thus, the pure effect of Sue’s selection performance relative Thus, the pure effect of Sue’s selection performance relative to her benchmark was:to her benchmark was:


= 0.9(7%= 0.9(7%--9%) + 0.1(10%9%) + 0.1(10%--11%) 11%)

= = --1.8% 1.8% -- 0.1% = 0.1% = --1.9%1.9%

She lost 1.9% relative to her benchmark because her She lost 1.9% relative to her benchmark because her industrial properties did 2% worse than the benchmark’s industrial properties did 2% worse than the benchmark’s industrial properties and her office properties did 1% worse industrial properties and her office properties did 1% worse than the benchmark’s office properties, and her benchmark than the benchmark’s office properties, and her benchmark allocation was 90% to industrial.allocation was 90% to industrial.

Note that in order to avoid mixing the effect of Sue’s Note that in order to avoid mixing the effect of Sue’s allocation in with the effect of her selection, we quantify her allocation in with the effect of her selection, we quantify her selection effect using the selection effect using the benchmark’sbenchmark’s allocation.allocation.


Similarly, the pure effect of Sue’s allocation performance Similarly, the pure effect of Sue’s allocation performance relative to her benchmark was:relative to her benchmark was:


= 9%(0.1= 9%(0.1--0.9) + 11%(0.90.9) + 11%(0.9--0.1) 0.1)

= = --7.2% + 8.8% = +1.6%7.2% + 8.8% = +1.6%

She gained 1.6% relative to her benchmark because she put She gained 1.6% relative to her benchmark because she put 80% more weight than her benchmark into office 80% more weight than her benchmark into office properties, and office properties performed 2% better than properties, and office properties performed 2% better than industrial properties within the benchmark.industrial properties within the benchmark.

Note that in order to avoid mixing the effect of Sue’s Note that in order to avoid mixing the effect of Sue’s selection performance in with the effect of her allocation, we selection performance in with the effect of her allocation, we quantify her allocation effect using the quantify her allocation effect using the benchmark’sbenchmark’sselection performance.selection performance.


The sum of these two pure effects of Sue’s selection and allocatThe sum of these two pure effects of Sue’s selection and allocation ion performance does not equal her total return performance differenperformance does not equal her total return performance differential tial with respect to her benchmark.with respect to her benchmark.

The remaining performance differential is due to the The remaining performance differential is due to the combined effectcombined effectof Sue’s selection and allocation performances of Sue’s selection and allocation performances interactinginteracting together. together. (There is no meaningful way to disentangle this interaction effe(There is no meaningful way to disentangle this interaction effect and ct and allocate it to either one of the two pure effects.)allocate it to either one of the two pure effects.)

Sue vs her benchmark:Attribute:Allocation 1.60%Selection -1.90%Interaction 0.80%Total differential 0.50%


Sue's returns vs her benchmark:Weights: Benchmark Sue:Industr 90% 10%Office 10% 90%Returns: Benchmark: Sue:Total Portfolio 9.20% 9.70%Industrial Properties 9.00% 7.00%Office Properties 11.00% 10.00%

Sue vs her benchmark:Attribute:Allocation 1.60%Selection -1.90%Interaction 0.80%Total differential 0.50%

The interaction effect is sometimes called the “crossThe interaction effect is sometimes called the “cross--product”. product”.

Some analysts suggest that it reflects the investment manager’s Some analysts suggest that it reflects the investment manager’s “specialization” ability, her degree of success in over“specialization” ability, her degree of success in over--weighting segments in weighting segments in which the manager has relatively better (or less bad) selection which the manager has relatively better (or less bad) selection skills even skills even though her overall selection ability may be weak and such segmenthough her overall selection ability may be weak and such segments may ts may not be strategically superior from an overall allocation perspecnot be strategically superior from an overall allocation perspective. tive.

e.g., Sue overe.g., Sue over--weighted office, where her selection was “less bad” than in weighted office, where her selection was “less bad” than in industrial, so she achieved a positive interaction (specializatiindustrial, so she achieved a positive interaction (specialization) effect.on) effect.


This attempt at giving a separate and substantive management functional interpretation to the interaction component seems to me to be a bit “heroic”. I just can’t quite see it, personally.

Some analysts use the manager’s weights to compute the “selectioSome analysts use the manager’s weights to compute the “selection” effect:n” effect:

SSSS––SSBB = = wwSISI((rrSISI –– rrBIBI) + ) + wwSOSO((rrSOSO –– rrBOBO))

= 0.1(7%= 0.1(7%--9%) + 0.9(10%9%) + 0.9(10%--11%) 11%)

= = --0.2% 0.2% -- 0.9% = 0.9% = --1.1%1.1%

This allocates the entire performance differential to just selecThis allocates the entire performance differential to just selection and tion and allocation, eliminating the interaction effect:allocation, eliminating the interaction effect:

AllocAlloc (Sue (Sue –– BnchmrkBnchmrk) = ) = +1.6%+1.6%+ Select(Sue+ Select(Sue--BnchmrkBnchmrk) = ) = -- 1.1%1.1%

--------------Total Total +0.5%+0.5%

27.1.4. Mgr Allocation Weights in the Benchmark?...27.1.4. Mgr Allocation Weights in the Benchmark?...

But it is misleading to define the selection effect this way andBut it is misleading to define the selection effect this way and think of it as a think of it as a pure selection effect. It would be more meaningful to explicitlypure selection effect. It would be more meaningful to explicitly combine the combine the interaction effect with selection and call it “Selection&Allocatinteraction effect with selection and call it “Selection&Allocation”.ion”.

Or, more completely: call it the “Selection Plus Synergy Between Allocation & Selection Effect”.

The The meaningmeaning of portfolioof portfolio--level attribution in private level attribution in private real estate investment (as distinct from public securities real estate investment (as distinct from public securities investment)…investment)…

•• The The allocationallocation component has a very similar meaning in private component has a very similar meaning in private real estate as compared to public securities (corresponding to treal estate as compared to public securities (corresponding to the he segment weight allocation decision at the portfolio level), but…segment weight allocation decision at the portfolio level), but…

•• The The selectionselection component in private real estate encompasses both component in private real estate encompasses both the traditional “asset picking” function of securities investmenthe traditional “asset picking” function of securities investment t management (picking good individual assets), and also (unlike wimanagement (picking good individual assets), and also (unlike with th public securities) the public securities) the acquisition acquisition andand operational managementoperational managementfunction associated with negotiating private asset deals and lonfunction associated with negotiating private asset deals and longg--term holding of a real property asset that must be managed by itterm holding of a real property asset that must be managed by its s investor/owner. (i.e., purely “investor/owner. (i.e., purely “passivepassive” investment is not possible in ” investment is not possible in the private real estate investment industry.)the private real estate investment industry.)


(See notes to earlier slides, describing this second point in more depth.)

It is very difficult to break down the overall It is very difficult to break down the overall selectionselectionperformance between the contributions provided by performance between the contributions provided by the “traditional selection” (asset picking) function and the “traditional selection” (asset picking) function and that provided by the that provided by the acquisitionacquisition and and operational operational managementmanagement functions.functions.

Some insight may be obtained by analysis of Some insight may be obtained by analysis of propertyproperty--levellevel performance attribution. In particular:performance attribution. In particular:

•• Initial yieldInitial yield relates to the relates to the traditional selectiontraditional selection & & acquisitionacquisitionfunctions;functions;•• Cash flow changeCash flow change relates largely to the relates largely to the operational operational managementmanagement function (but not purely or necessarily, e.g., it function (but not purely or necessarily, e.g., it may reflect expiration of vintage leases);may reflect expiration of vintage leases);•• Yield changeYield change may reflect either a may reflect either a traditional selectiontraditional selection (asset (asset picking) effect or an picking) effect or an operational managementoperational management effect, or both.effect, or both.


27.1.1 Property27.1.1 Property--LevelLevel Performance Attribution Performance Attribution . . .. . .

Property level performance attribution focuses on “Property level performance attribution focuses on “property property levellevel” investment performance, i.e., the total return achieved ” investment performance, i.e., the total return achieved within a given property or a static (fixed) portfolio of within a given property or a static (fixed) portfolio of properties (that is, apart from the effect of investment properties (that is, apart from the effect of investment allocation decisions, as if holding allocation among categories allocation decisions, as if holding allocation among categories constant).constant).Property level attribution should be designed to break out Property level attribution should be designed to break out the property level total return performance in a manner the property level total return performance in a manner useful for shedding light on the four major property level useful for shedding light on the four major property level investment management functions:investment management functions:•• Property selectionProperty selection (picking “good” properties as found);(picking “good” properties as found);•• Acquisition transaction executionAcquisition transaction execution;;•• Operational managementOperational management during the holding period (e.g., during the holding period (e.g., marketing, leasing, expense mgt, capital expenditure mgt);marketing, leasing, expense mgt, capital expenditure mgt);•• Disposition transaction executionDisposition transaction execution..

PropertyProperty--Level Performance Attribution Level Performance Attribution . . .. . .

These propertyThese property--level management functions are related level management functions are related generally to three attributes (components) of the propertygenerally to three attributes (components) of the property--level level sincesince--acquisition IRRacquisition IRR, essentially as indicated below…, essentially as indicated below…

Initial Yield

(IY)

Cash Flow Change

(CFC)

Yield Change

(YC)

Property Selection

Acquisition Transaction Execution

Operational Management

Disposition Transaction Execution

Three equations, and four unknown variables. (Darn!)

Conventional property level performance attribution is Conventional property level performance attribution is based on periodic returns, or on timebased on periodic returns, or on time--weighted multiweighted multi--period period returns (returns (TWRRsTWRRs).).

But But IRRIRR--based performance attribution is arguably more based performance attribution is arguably more useful for property level management diagnostic purposes, useful for property level management diagnostic purposes, because:because:

•• At the property level, the investment manager is typically At the property level, the investment manager is typically responsible for the major cash flow responsible for the major cash flow timingtiming decisions that can decisions that can significantly effect property level (static portfolio) returns, significantly effect property level (static portfolio) returns, e.g., e.g., leasing decisions, capital expenditure decisions.leasing decisions, capital expenditure decisions.•• The IRR is sensitive to the effect of cash flow timing, the TWRThe IRR is sensitive to the effect of cash flow timing, the TWRR is R is not.not.•• The The IRR is cash flow basedIRR is cash flow based (net of capital improvement (net of capital improvement expenditures), therefore, more accurately reflecting the investmexpenditures), therefore, more accurately reflecting the investment ent return effect of capital improvement decisions.return effect of capital improvement decisions.


Useful IRRUseful IRR--Based property level performance attribution Based property level performance attribution benchmarking requires the use of:benchmarking requires the use of:

SinceSince--acquisition IRRacquisition IRR

•• IRR is computed IRR is computed since acquisitionsince acquisition of property (or portfolio):of property (or portfolio):•• In order to reflect investment operational performance In order to reflect investment operational performance during entire holding period since acquisition;during entire holding period since acquisition;

•• Property investment holding periods are typically multiProperty investment holding periods are typically multi--year (single period or periodic returns do not reflect effectiveyear (single period or periodic returns do not reflect effectiveinvestment management holding period).investment management holding period).

•• IRR is computed for appropriate IRR is computed for appropriate benchmark cohortbenchmark cohort, , defined as universe of similar investments by competing defined as universe of similar investments by competing managers, measured from same inception date (equal to managers, measured from same inception date (equal to property acquisition date).property acquisition date).


Also, to quantify asset selection and acquisition transaction execution performance, the initial yield at the time of acquisition must be considered, which requires use of since-acquisition IRR. (See subsequent slides for further elaboration.)

Simple Numerical Example:Simple Numerical Example:•• Property (or static portfolio) bought at initial cash Property (or static portfolio) bought at initial cash yield of 9%.yield of 9%.•• Net cash flow grew at 2% per year.Net cash flow grew at 2% per year.•• Property (or properties) sold (or appraised) after 10 Property (or properties) sold (or appraised) after 10 years at a terminal yield of 10%, based on yr.11 years at a terminal yield of 10%, based on yr.11 projected cash flow (also 2% more than yr.10).projected cash flow (also 2% more than yr.10).•• IRR is 10.30%.IRR is 10.30%.•• How much of this IRR is due to 3 components: How much of this IRR is due to 3 components: Initial Yield (Initial Yield (IYIY), Cash Flow Change (), Cash Flow Change (CFCCFC), and ), and Yield Change (Yield Change (YCYC)?…)?…

Recall Ch.10 Appendix: Recall Ch.10 Appendix: PropertyProperty--Level Performance Attribution Level Performance Attribution . . .. . .

This example is taken from Geltner-Miller, Chapter 10 Appendix (pp.232-235). There is also a Study Question (10.24) at the end of that chapter, with solution on p.858.

Yr IRRs: 0 1 2 3 4 5 6 7 8 9 10 11(1) Actual Oper.CF 1.0000 1.0200 1.0404 1.0612 1.0824 1.1041 1.1262 1.1487 1.1717 1.1951 1.2190(2) Actual Capital CF -11.1111 12.1899(3) Actual Total CF (=1+2) 10.30% -11.1111 1.0000 1.0200 1.0404 1.0612 1.0824 1.1041 1.1262 1.1487 1.1717 13.3850(4) Init.Oper.CF constant 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000(5) Capital CF @ Init.Yld.on(4) -11.1111 11.1111(6) Init.CF @ Init.Yld (=4+5) 9.00% -11.1111 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 12.1111(7) Capital CF @ Init.Yld.on(1) -11.1111 13.5444(8) Actual Oper. CF @ Init.Yld (=1+7) 11.00% -11.1111 1.0000 1.0200 1.0404 1.0612 1.0824 1.1041 1.1262 1.1487 1.1717 14.7395(9) Capital CF @ ActualYld.on(4) -11.1111 10.0000(10) Init.CF @ Actual Yld (=4+9) 8.32% -11.1111 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 11.0000

There are several ways one might answer this question. The approThere are several ways one might answer this question. The approach that ach that seems most intuitively related to the 4 basic mgt seems most intuitively related to the 4 basic mgt fcnsfcns is presented here…is presented here…

•• Initial yield = 9.00%,Initial yield = 9.00%, computed from line (6) IRR.computed from line (6) IRR.•• Cash flow change component = 2.00%Cash flow change component = 2.00% = 11%= 11%--9%, computed as the line (8) IRR 9%, computed as the line (8) IRR less the line (6) IRR: = IRR with actual CF less the line (6) IRR: = IRR with actual CF –– IRR with no CF growth, (with constant IRR with no CF growth, (with constant yldyld at initial rate).at initial rate).•• Yield change component = Yield change component = --0.68%0.68% = 8.32%= 8.32%--9.00%, computed as the line (10) IRR 9.00%, computed as the line (10) IRR less the line (6) IRR: = IRR with actual less the line (6) IRR: = IRR with actual yldyld chg chg –– IRR with no IRR with no yldyld chg, (with constant chg, (with constant CF at initial level).CF at initial level).•• InterractionInterraction effect = effect = --0.02%0.02%, the difference , the difference bertweenbertween the line (3) overall IRR and the line (3) overall IRR and the sum of the three other attributes [10.3%the sum of the three other attributes [10.3%--(9%+2%(9%+2%--0.68%)].0.68%)].

Other ways to define the break-down of the three components would be, for example, to define the components on the basis of the actual or the terminal cash flows rather than the initial cash flow as done here. No one method of basing the break-down is theoretically uniquely correct. I think basing the break-down on the initial cash flow as done here gives results that are the most intuitive for identifying the substantive property-level management functions of acquisition and operational management, and potentially on the disposition function as well (this last if the terminal value is based on a disposition price) .


Subject Property: IRR & Component Breakout

-4%-2%0%2%4%6%8%

10%12%14%

IRR InitYld CFchg YldChg Interaction

IRR & Components

Here is a graphical presentation of the IRRHere is a graphical presentation of the IRR--Based propertyBased property--level level performance attribution we just performed:performance attribution we just performed:

Suppose we computed the same type of IRR component breakdown Suppose we computed the same type of IRR component breakdown for an appropriate benchmark, that is, a NCREIF subfor an appropriate benchmark, that is, a NCREIF sub--index cohort index cohort spanning the same period of time…spanning the same period of time…

PropertyProperty--Level Performance Attribution Level Performance Attribution . . .. . .We could compare our subject performance with that achieved by aWe could compare our subject performance with that achieved by apeer universe of managers, for similar properties (e.g., Calif. peer universe of managers, for similar properties (e.g., Calif. IndustrIndustr. . bldgsbldgs):):

Subject vs NCREIF Cohort Performance Comparison: IRR & Component Breakout

-5%

0%

5%

10%

15%


IRR & Components

Subject NPI Cohort

The NCREIF Index (or any sub-index thereof) can be converted from periodic returns to an IRR over any specified holding period, using the income and appreciation return components given in the index. (To do this properly, the index components should be defined based on removing the capital expenditures from the income return component and not removing them from the appreciation return component, the “cash flow based” version of the NPI that can be downloaded from the NCREIF web site.) Set the value of the investment outflow at the beginning of the holding period equal to negative unity (-1) at time zero = “t-1”. Then the value of the net cash flow in the first period (“t”) of the holding period is given by the periodic income return that period: y(t) = CF(t)/V(t-1). In general, the periodic income return in any period “t” is: y[t] = CF[t]/V[t-1]. The periodic appreciation return asset value multiple in period “t” is: (1+g[t]) = V[t]/V[t-1], where g[t] is the appreciation return. Thus: (y[t]/y[t-1])(1+g[t-1]) = (CF[t]/V[t-1])/(CF[t-1]/V[t-2])(V[t-1]/V[t-2]) = CF[t]/CF[t-]). Use the constructed CF[t]/CF[t-1] series to construct the periodic net cash flow series as a fraction of the initial time zero investment, throughout the holding period. Compute the terminal asset value at the end of the holding period (period “T”), per dollar of initial investment, by compounding the appreciation returns: V[T] = V[t-1](1+g[t])(1+g[t+1])(1+g[t+2])…(1+g[T]). Given the initial cash outflow of -1, the subsequent stream of: CF[t], CF[t+1],…,CF[T-1],(CF[T]+V[T]), the IRR for an implicit investment buying the NCREIF Index at time t-1 and selling it at time T can be computed. This can serve as the benchmark cohort IRR.

PropertyProperty--Level Performance Attribution Level Performance Attribution . . .. . .Here is the Here is the relative performancerelative performance, the , the differencedifference between our subject between our subject property and its benchmark, by attribute:property and its benchmark, by attribute:

The above pattern could be plausibly interpreted as tentative evThe above pattern could be plausibly interpreted as tentative evidence for idence for the following hypothesis: Subject performed relatively poorly duthe following hypothesis: Subject performed relatively poorly due largely to e largely to some combination of poor selection, acquisition, and operationalsome combination of poor selection, acquisition, and operational mgt, mgt, partially offset by some combination of good disposition executipartially offset by some combination of good disposition execution (or on (or optimistic terminal appraisal), futureoptimistic terminal appraisal), future--oriented capital improvements, &/or oriented capital improvements, &/or market movements during the holding period.market movements during the holding period.

Subject - NCREIF Cohort Relative Performance: IRR & Component Breakout

-2.50%-2.00%-1.50%-1.00%-0.50%0.00%0.50%1.00%1.50%


IRR & Components


These propertyThese property--level management functions are related level management functions are related generally to three attributes (components) of the propertygenerally to three attributes (components) of the property--level level sincesince--acquisition IRRacquisition IRR, essentially as indicated below…, essentially as indicated below…

Initial Yield

(IY)

Cash Flow Change

(CFC)

Yield Change

(YC)

Property Selection

Acquisition Transaction Execution

Operational Management

Disposition Transaction Execution

Three equations, and four unknown variables. (Darn!)

PropertyProperty--Level Performance Attribution Level Performance Attribution . . .. . .Here is the Here is the relative performancerelative performance, the , the differencedifference between our subject between our subject property and its benchmark, by attribute:property and its benchmark, by attribute:

Now suppose we computed these relative performance differentialsNow suppose we computed these relative performance differentials across a across a number of different properties (or portfolios) we have invested number of different properties (or portfolios) we have invested in…in…

Subject - NCREIF Cohort Relative Performance: IRR & Component Breakout

-2.50%-2.00%-1.50%-1.00%-0.50%0.00%0.50%1.00%1.50%


IRR & Components

PropertyProperty--Level Performance Attribution Level Performance Attribution . . .. . .We might gain some insights about our propertyWe might gain some insights about our property--level investment and level investment and management performance:management performance:

In this case Subject Properties #1 & 3 have similarly poor perfoIn this case Subject Properties #1 & 3 have similarly poor performance (rmance (relrel to to benchmkbenchmk), due to poor initial yield & poor CF change, suggesting poor a), due to poor initial yield & poor CF change, suggesting poor acquisition & cquisition & poor operational mgt. Property #2 did better, with good poor operational mgt. Property #2 did better, with good InitYldInitYld & & CFchgCFchg, but poor , but poor YldChgYldChg (suggesting good acquisition, but poor disposition or mgt actio(suggesting good acquisition, but poor disposition or mgt actions that hurt ns that hurt future outlook (e.g., inadequate Cap.Improvement). future outlook (e.g., inadequate Cap.Improvement). MktMkt movements can also affect movements can also affect these results (less so the longer the holding period).these results (less so the longer the holding period).

Three Properties Comparison:Subject - NCREIF Cohort Relative Performance

-5.00%-4.00%-3.00%-2.00%-1.00%0.00%1.00%2.00%3.00%


Subject 1 Subject 2 Subject 3

Such comparisons of relative performance across properties can sometimes be useful even if the properties have different acquisition dates. (Of course, the IRR-cohort benchmark against which the relative performance is computed must be “custom” for each property, corresponding to the holding period for each property.) All such formal quantitative exercises must be “taken with a grain of salt”, applied in an “investigative” spirit. They often bring insights by the questions they raise rather than by providing any definitive answers. They get property and asset managers thinking about how they did and what they did in specific cases, and what impact their actions had in the overall performance. These types of results should never be used mechanistically, without broader and more qualitative analysis, and they should generally not be used for evaluative purposes (in my opinion), but rather for diagnostic (I.e., “learning”) purposes.

27.2 Investment Performance 27.2 Investment Performance Evaluation & BenchmarkingEvaluation & Benchmarking

•• How Benchmarking Is DoneHow Benchmarking Is Done

•• Purpose & Role of BenchmarkingPurpose & Role of Benchmarking

•• Criteria of an Ideal BenchmarkCriteria of an Ideal Benchmark

How Benchmarking Is Done in the Securities Industry. . .How Benchmarking Is Done in the Securities Industry. . .

3-year

Time-Wtd

Return

Investment Mgt “Style” or Asset Class

Style A

Compute ex post Compute ex post total TWRR of all total TWRR of all competing competing managers in an managers in an active mgr “active mgr “peer peer universeuniverse” (same ” (same “style”), over a “style”), over a common historical common historical period (indicated by period (indicated by this this rangerange).).

As noted before, an investment style in the equities industry typically refers to size and value vs growth characteristics, e.g., small-cap growth, large-cap value, etc. Benchmarking may also consider specialization, such as if the portfolio manager specializes in public utility stocks, REITs, international stocks, and so forth.


3-year

Time-Wtd

Return


Style A

Compute interCompute inter--quartile range of quartile range of the peer universethe peer universe

Box indicates TWRR of 25th and 75th percentile active manager in peer universe, over the common historical time period of the analysis.


3-year

Time-Wtd

Return


Style A

Compute median Compute median TWRR of the peer TWRR of the peer universeuniverse


3-year

Time-Wtd

Return


Style A

Index AIndex A

Compute TWRR of Compute TWRR of a a passive indexpassive indexrepresentative of representative of style.style.

A “passive index” in this context refers to an index that has relatively static allocation across assets, and is broadly representative of the entire market of assets of reflective of the manager’s style and specialization. For example, for the large-cap stock manager style, the S&P500 is a widely used representative passive index. For a REIT manager, the NAREIT Index or the Morgan-Stanley or Wilshire Indices are widely used.


3-year

Time-Wtd

Return


Style A

Index AIndex A

Mgr AMgr A

Compute TWRR of Compute TWRR of the subject active the subject active manager or manager or activelyactively--managed managed portfolio.portfolio.


3-year

Time-Wtd

Return


Style A

Index AIndex A

Mgr AMgr ACompare Compare subjsubj mgr’s mgr’s performance to that of:performance to that of:

•• The passive indexThe passive index

•• The peer universeThe peer universe

This is the focal point of the benchmarking exercise. The subject manager’s performance is compared either to that of the representative passive index, or to the median (and distribution) of his peer universe, or both, as appropriate. The comparison may be done using a risk-adjusted return measure, but often more attention is focused simply on the unadjusted time-wtd average of the total returns. As noted, risk-adjustments are imperfect, and if the comparison is limited to a common style & specialization (i.e., carefully defined peer universe and appropriately selected passive market index), then there may be little difference in risk across the comparison. (However, differences in use of leverage may need to be considered and controlled for.)


3-year

Time-Wtd

Return


Style A

Index AIndex A

Mgr AMgr A

Style B

Mgr BMgr B

Index BIndex B

Do the same Do the same withinwithin each styleeach style


The Importance of StyleThe Importance of Style--Purity in the Benchmark:Purity in the Benchmark:

•• Not appropriate to compare performance of one style of Not appropriate to compare performance of one style of investment against that of another for evaluating a investment against that of another for evaluating a manager who is specialized in (and hired for the purpose manager who is specialized in (and hired for the purpose of) investing in one particular style. of) investing in one particular style.

•• MultiMulti--style managers and portfolios can be benchmarked style managers and portfolios can be benchmarked using composite indices constructed from an appropriate using composite indices constructed from an appropriate combination of stylecombination of style--pure indices. pure indices.

•• Comparative analysis and diagnostics regarding the Comparative analysis and diagnostics regarding the differential performance of different types (styles) of differential performance of different types (styles) of investments is facilitated by the availability of styleinvestments is facilitated by the availability of style--pure pure indices.indices.

1.1. Communication (Client Communication (Client Manager).Manager).

2.2. InterestInterest--Alignment (Client Alignment (Client Manager).Manager).

3.3. Weed out Weed out inferiorinferior managers.managers.

The Purpose of Benchmarking . . .The Purpose of Benchmarking . . .

Ex AnteEx Ante Purposes (most important):Purposes (most important):

Ex Post Ex Post Purpose (less important or less feasible):Purpose (less important or less feasible):

4. Identify 4. Identify superiorsuperior managers (in combination managers (in combination with qualitative information).with qualitative information).


Ex Post Ex Post Purpose (less important or less feasible):Purpose (less important or less feasible):4. Identify 4. Identify superiorsuperior managers (in combination managers (in combination

with qualitative information).with qualitative information).

Why is the identification of superior managers Why is the identification of superior managers difficult in practice?difficult in practice?

•• “Superior mgr” = One who can “Superior mgr” = One who can consistentlyconsistently beat mkt.beat mkt.

•• “Superior mgr” = One who can “Superior mgr” = One who can consistentlyconsistently beat peers.beat peers.

•• This is This is rarerare, and normally only , and normally only marginalmarginal..

•• Moving targetMoving target: : MktMkt (& peer universe) learns & improves (& peer universe) learns & improves over time.over time.


Why is the identification of superior Why is the identification of superior managers difficult in practice?managers difficult in practice?

•• “Superior mgr” = One who can “Superior mgr” = One who can consistentlyconsistently beat mkt.beat mkt.

•• “Superior mgr” = One who can “Superior mgr” = One who can consistentlyconsistently beat peers.beat peers.

•• This is This is rarerare, and normally only , and normally only marginalmarginal..

•• Moving targetMoving target: : MktMkt (& peer universe) learns & improves (& peer universe) learns & improves over time.over time.

•• But there is also a But there is also a statistical reason . . .statistical reason . . .

27.2.4: 27.2.4: The Problem ofThe Problem of Statistical SignificanceStatistical Significance in Performance in Performance Measurement . . .Measurement . . .

•• “Noise” (or “Noise” (or randomnessrandomness) in ex post returns obscures ) in ex post returns obscures our ability to draw rigorous conclusions about our ability to draw rigorous conclusions about managers’ abilities to managers’ abilities to consistentlyconsistently beat a benchmark as beat a benchmark as a result of a result of skillskill (& effort) rather than (& effort) rather than luckluck..

•• There are three sources of randomness in empirically There are three sources of randomness in empirically observable real estate returns:observable real estate returns:

1.1. CrossCross--sectionalsectional ((individindivid. prop. heterogeneity).. prop. heterogeneity).

2.2. LongitudinalLongitudinal randomness in true randomness in true mktmkt (volatility).(volatility).

3.3. Measurement errorMeasurement error (in R.E. returns).(in R.E. returns).

•• Any (& all) of these sources of “noise” obscures the Any (& all) of these sources of “noise” obscures the “signal” about mgrs’ true relative abilities.“signal” about mgrs’ true relative abilities.

There might be an argument that source 1 (property heterogeneity) is an aspect of performance the manager is hired for, and so is not “noise” in performance measurement. But it is unreasonable to expect a manager to completely forecast future idiosyncratic returns of any asset.

Numerical Example:Numerical Example:•• Quarterly returns, Quarterly returns, 33--yryr comparison period.comparison period.•• 12 12 obsobs drawn from drawn from true true difference between mgr(i) & difference between mgr(i) & benchmark(B) return: benchmark(B) return: EE[ [ rrii –– rrBB ].].•• Suppose favorable circumstances for making inference:Suppose favorable circumstances for making inference:

•• EE[ [ rrii –– rrBB ] constant] constant over time.over time.•• VolatilityVolatility of of rrii & & rrBB = = 5%/qtr5%/qtr (each).(each).•• CORRCORR [ [ rrii , , rrBB ] = ] = +90%.+90%.

•• Then mgr TWRR would have to beat benchmark by Then mgr TWRR would have to beat benchmark by >500 >500 basisbasis--pointspoints per annum before we could conclude with per annum before we could conclude with statistical significance that the central tendency of the mgr’s statistical significance that the central tendency of the mgr’s return exceeds that of the benchmark return (i.e., rigorous return exceeds that of the benchmark return (i.e., rigorous case for case for skillskill as opposed to as opposed to luckluck).).


27.2.4: 27.2.4: The Problem ofThe Problem of Statistical SignificanceStatistical Significance in Performance in Performance Measurement . . .Measurement . . .Calculations:Calculations:•• Standard error of quarterly difference in returns:Standard error of quarterly difference in returns:

( )( )

[ ]( )

%24.2)05.0)(05.0)(9.0(205.005.0

2

2122

2122

=−+=

−+=− BiBiBi C σσσσσ

•• Standard error of observed Standard error of observed qtrlyqtrly avgavg difference =difference =

( )( )[ ] %675.0112%24.21 21

=−=−= − NBiσ

•• X 2 std.errs for 95% statistical confidence: 2(0.675%) = X 2 std.errs for 95% statistical confidence: 2(0.675%) = 1.35%.1.35%.•• Annualize: (1.0135)Annualize: (1.0135)44 –– 1 = 1 = 5.51%5.51% /yr. difference required./yr. difference required.

This example assumes independent periodic returns (I.e., zero serial correlation), as if there were no inertia in real estate returns. With inertia, the std.err of the avg difference would be larger than indicated here (i.e., effectively less than 12 independent observations of the difference). On the other hand, inertia probably causes the quarterly volatility to be smaller than the 5% used in this example.


Implications:Implications:

•• You cannot use purely quantitative benchmarking You cannot use purely quantitative benchmarking comparisons to rigorously identify superior managers ex comparisons to rigorously identify superior managers ex post.post.

•• Don’t apply such benchmarking Don’t apply such benchmarking mechanisticallymechanistically or in or in isolation.isolation.

•• Include broader analyses and qualitative information in Include broader analyses and qualitative information in conjunction with quantitative benchmarking:conjunction with quantitative benchmarking:

•• Performance attribution analysis may be useful in this Performance attribution analysis may be useful in this regard.regard.

•• And remember . . .And remember . . .


Implications:Implications:

Ex AnteEx Ante Purposes (most important):Purposes (most important):

•• Communication (Client Communication (Client Manager).Manager).

•• InterestInterest--Alignment (Client Alignment (Client Manager).Manager).

•• Weed out Weed out inferiorinferior managers.managers.

•• The The ex anteex ante role of benchmarking is not negated by the role of benchmarking is not negated by the problem of lack of statistical significance:problem of lack of statistical significance:

•• Even with noise, a better mgr who works harder for the client iEven with noise, a better mgr who works harder for the client is s more likely to beat the benchmark.more likely to beat the benchmark.•• Communication Communication betwbetw mgr & client is improved by identification of mgr & client is improved by identification of benchmark at outset of contract.benchmark at outset of contract.

27.2.2. Criteria for an 27.2.2. Criteria for an Ideal Benchmark IndexIdeal Benchmark Index . . .. . .

3-year

Time-Wtd

Return


Style A

Mgr AMgr A

Index AIndex A

Recall that two different Recall that two different types of benchmarks are types of benchmarks are often used in the often used in the securities industry:securities industry:

•• Passive Passive mktmkt indicesindices, &, &

•• Peer universesPeer universes..

Which of these is better Which of these is better as a benchmark?as a benchmark?


In the securities industry, there is some argument that peer In the securities industry, there is some argument that peer universes are not as good as passive market indices for universes are not as good as passive market indices for benchmarking purposes:benchmarking purposes:

•• Peer universes suffer from “survivor bias”.Peer universes suffer from “survivor bias”.

•• Peer universes don’t meet all of the “Peer universes don’t meet all of the “Bailey CriteriaBailey Criteria”:”:

The “Bailey Criteria”:

• Unambiguous

• Investable

• Measurable

• Specified in advance

• Appropriate to Mgr

• Reflective of Mgr

27.2.2. Criteria for an 27.2.2. Criteria for an Ideal Benchmark IndexIdeal Benchmark Index . . .. . .Alas, in private real estate, we don’t really have a Alas, in private real estate, we don’t really have a passive passive market indexmarket index, that is, an index that:, that is, an index that:•• Covers the entire commercial property market; &Covers the entire commercial property market; &•• Maintains a constant set of assets.Maintains a constant set of assets.

Private real estate benchmark indices are Private real estate benchmark indices are peer universe peer universe indicesindices::•• Indices that represent the performance of (almost) the Indices that represent the performance of (almost) the entire populationentire population of properties held by the universe of of properties held by the universe of investment managers competing for the business of the investment managers competing for the business of the institutional “institutional “corecore” style of private real estate investment.” style of private real estate investment.•• “Active”“Active” mgt is mgt is necessarynecessary in private R.E. investment in private R.E. investment (operational control of assets).(operational control of assets).•• The NCREIF Index is a peer universe index in the US.The NCREIF Index is a peer universe index in the US.•• The IPD Index is a peer universe index in the UK.The IPD Index is a peer universe index in the UK.


How serious a problem is the lack of a passive market index How serious a problem is the lack of a passive market index for the private real estate investment industry? . . .for the private real estate investment industry? . . .

•• Such an index would be very useful for Such an index would be very useful for researchresearch purposes.purposes.

•• But, in contrast to the securities industry, the argument for But, in contrast to the securities industry, the argument for the superiority of passive market indices for the superiority of passive market indices for benchmarkingbenchmarkingpurposes purposes does not carry over to the private real estate does not carry over to the private real estate investment industryinvestment industry::

•• There is not much of a survivorship problem in propertyThere is not much of a survivorship problem in property--level level indices such as NCREIF.indices such as NCREIF.•• Investment turnover is not rapid in R.E. indices.Investment turnover is not rapid in R.E. indices.•• Some of the “Bailey Criteria” do not make sense for the privateSome of the “Bailey Criteria” do not make sense for the private real real estate investment industry.estate investment industry.

e.g., perfect index e.g., perfect index investabilityinvestability is not possible in an asset class where is not possible in an asset class where unique, whole assets are traded.unique, whole assets are traded.


Fundamental Criteria of Benchmark Indices:Fundamental Criteria of Benchmark Indices:1.1. Measurability: Measurability: QUANTITATIVE COMPARISON.QUANTITATIVE COMPARISON.

2. Client 2. Client InvestabilityInvestability: :

MGR HIRED FOR MGR HIRED FOR INCREMENTALINCREMENTAL CONTRIBUTION CONTRIBUTION OVER WHAT CLIENT COULD DO.OVER WHAT CLIENT COULD DO.

3. Manager No Forced Bet: 3. Manager No Forced Bet: MGT PRINCIPLE: MGT PRINCIPLE: EVALUATION = RESPONSIBILITY = AUTHORITY.EVALUATION = RESPONSIBILITY = AUTHORITY.

4. Appropriateness (Mgr Style & Spec.): 4. Appropriateness (Mgr Style & Spec.): USEFULNESS FOR USEFULNESS FOR CLIENT STRATEGY IMPLEMENTATION, FAIRNESS FOR MGR.CLIENT STRATEGY IMPLEMENTATION, FAIRNESS FOR MGR.

5. Non5. Non--manipulatabilitymanipulatability:: CREDIBILITY & CREDIBILITY & EX ANTEEX ANTE ROLE.ROLE.

6. Agreement in Advance: 6. Agreement in Advance: COMMUNICATION & FAIRNESS.COMMUNICATION & FAIRNESS.

27.2.2. Criteria for an 27.2.2. Criteria for an Ideal Benchmark IndexIdeal Benchmark Index . . .. . .Peer universe indices (like the NCREIF Index) implement Peer universe indices (like the NCREIF Index) implement these criteria reasonably well for the private real estate these criteria reasonably well for the private real estate investment industry. E.g.,:investment industry. E.g.,:

Client Client InvestabilityInvestability:: Client could have hired any other competing Client could have hired any other competing mgr from peer universe, instead of subj.mgr., mgr from peer universe, instead of subj.mgr., so peer universe is “so peer universe is “ex ante ex ante investableinvestable”.”.

Manager No Forced Bet:Manager No Forced Bet: Mgr can duplicate Mgr can duplicate typestypes of properties in of properties in peer universe index.peer universe index.

••Peer universe benchmark indices are like a Peer universe benchmark indices are like a ““Consumer ReportsConsumer Reports”” of of the investment industry.the investment industry.••You compare a given investment manager’s performance with that You compare a given investment manager’s performance with that of all of his relevant competitors (“peers” of all of his relevant competitors (“peers” –– managers the investor managers the investor could have hired instead of the given manager: i.e., those with could have hired instead of the given manager: i.e., those with the the same style & specialization).same style & specialization).

Peer universe indices are Peer universe indices are reasonablereasonable and and prudentprudent measures for measures for benchmarking private real estate performance.benchmarking private real estate performance.


Some practical implications of ideal benchmark Some practical implications of ideal benchmark criteria:criteria:

CRITERIA:CRITERIA:3. Manager No Forced Bet3. Manager No Forced Bet

&&4. Appropriateness4. Appropriateness

•• Fixed real return targets Fixed real return targets are are NOTNOT good Benchmarks.good Benchmarks.

•• Direct (private) real estate Direct (private) real estate investment mgrs or investment mgrs or portfolios should portfolios should NOTNOT be be benchmarked against REIT benchmarked against REIT Indices.Indices.

27.2 (cont.):27.2 (cont.):

Issues in Benchmarking Private Real EstateIssues in Benchmarking Private Real Estate

•• Population Population vsvs sample implications.sample implications.

•• Controlling for risk.Controlling for risk.

•• Segment weighting in the benchmark index.Segment weighting in the benchmark index.

•• Valuation methodology.Valuation methodology.

•• IRR IRR vsvs TWRR benchmarking.TWRR benchmarking.

27.2.2 (cont.): Population 27.2.2 (cont.): Population vsvs Sample Implications . . .Sample Implications . . .

Is the ideal benchmark index an entire Is the ideal benchmark index an entire populationpopulation, , or is it a or is it a samplesample??

What is the difference?, and why does it matter?What is the difference?, and why does it matter?

Census of entire population Census of entire population No need for statistical No need for statistical inference.inference.Sampling Sampling CanCan’’t avoid statistical inference issues.t avoid statistical inference issues.Ideal Ideal benchmarkbenchmark index = Entire peer universe index = Entire peer universe populationpopulation::

•• Practical feasibility to include entire peer universe (almost).Practical feasibility to include entire peer universe (almost).•• Legalistic function, Avoid “error” in comparisons.Legalistic function, Avoid “error” in comparisons.

Ideal asset class Ideal asset class researchresearch index = Scientific index = Scientific samplesample::•• Asset class too large, Census impractical.Asset class too large, Census impractical.•• But peer universe is not a “But peer universe is not a “scientificscientific” sample.” sample.

The first issue, the question of whether the benchmark index is ideally a “population” or a “sample”, we have really already broached. We said that the real estate benchmark index is inevitably a “peer universe index”. In fact, this has implications for the population vs sample question. But what is the difference between a population and a sample, and why does this distinction matter?…

Implications for appropriate index construction Implications for appropriate index construction methodology…methodology…

∑=

=N

itit r

Nr

1,

1Index as Sample Index as Sample EqualEqual--weighted returns (EW):weighted returns (EW):

Index as Population Index as Population ValueValue--weighted returns (VW): weighted returns (VW):

∑= −

−=N

iti

t

tit r

VV

r1

,1

1,

27.2.2 (cont.): Population 27.2.2 (cont.): Population vsvs Sample Implications . . .Sample Implications . . .

rasriv

The sample vs population issue carries a very basic implication regarding the appropriate way to construct the index as a cross-sectional aggregation of individual asset returns each period. The value-wtd index, which is appropriate for the index viewed as a population, gives the return which would have been earned each period by all of the constituent assets in the index as if they were a single portfolio. The VW index tells you what your return would have been each period if you owned all the properties in the index. Each individual property’s return is weighted by its share in the total aggregate value as of the beginning of the period. Larger assets will have a bigger impact on the index return. Most securities benchmarks are value-weighted, such as the S&P500 Index and the NAREIT Index. The equal-wtd index, which is more appropriate for the index viewed as a sample, gives the best estimate of the true underlying population return if each asset in the index is viewed as an equally valid representative of that population. Each individual property’s return is weighted equally in computing the index return, each period. EW indices are often used in academic studies where the index is used as a sample for inference purposes. In principle (viewed as a sample), an EW index will have less random error or “noise” than a VW index.

The Problem:The Problem:

•• The Capital Market naturally provides higher returns The Capital Market naturally provides higher returns (on (on avgavg over LR) for “more risky” assets.over LR) for “more risky” assets.

•• Manager must not be able outperform benchmark Manager must not be able outperform benchmark simplysimply by investing in more risky assets or using by investing in more risky assets or using leverage.leverage.

•• Hence, we need to Hence, we need to CONTROL FOR RISKCONTROL FOR RISK..

27.2.6. Controlling for Risk . . .27.2.6. Controlling for Risk . . .

Controlling for risk is a crucial part of good benchmarking. The fundamental ex ante purposes of benchmarking (the communication and incentive alignment role) would be under-cut if managers could beat their bogeys (ex ante) simply by investing in more risky assets. More broadly, in evaluating the performance of investment managers, the concept of “good” or “superior” performance is not meant to include simply taking on more risk.. (At least not without controlling for timing.) Therefore, we need to be able to control for risk in performance evaluation.

Two ways to control for risk:Two ways to control for risk:

•• QuantitativelyQuantitatively, using “, using “riskrisk--adjusted returnsadjusted returns”, by ”, by adjusting the mgr’s ex post return to reflect the adjusting the mgr’s ex post return to reflect the market’s “price of risk”.market’s “price of risk”.

•• NonNon--quantitativelyquantitatively, using “, using “stylestyle--based benchmarkingbased benchmarking”, ”, by constraining manager to same style (similarby constraining manager to same style (similar--risk) risk) assets, same style & risk as his benchmark.assets, same style & risk as his benchmark.


There are two ways to control for risk, in principle: 1) We can try to adjust the manager’s return for the amount of risk in his portfolio. This involves using “risk-adjusted return measures”. If we can do this well enough, then the manager’s benchmark need not have exactly the same risk as the manager’s own portfolio. We might call this “quantitative” control for risk. 2) The second way is we could constrain the manager to investing only in assets that are of essentially equal risk, which is also the risk of his custom benchmark. Then we do not need to quantitatively adjust for risk. We can directly compare the unadjusted returns. This might be referred to as “non-quantitative” risk adjustment, and it is the basic idea behind “style-based benchmarking”, a key aspect of the way benchmarking is applied today.

i

fii

rrTR

β−

=

BetaBeta = Amount of “Risk” = Amount of “Risk” AS IT AS IT MATTERS TO THE CAPITAL MARKETMATTERS TO THE CAPITAL MARKET..

This enables Portfolio “i”’s ex ante excess This enables Portfolio “i”’s ex ante excess return, E[return, E[rrii--rrff], to be ], to be ADJUSTED FOR ADJUSTED FOR RISKRISK..

27.2.6. Controlling for Risk . . .27.2.6. Controlling for Risk . . .The The Quantitative Quantitative Approach: e.g., The Approach: e.g., The TreynorTreynor RatioRatio......

First consider the “quantitative” or risk-adjusted return method of controlling for risk. There are several measures spoken of in investments textbooks, and used to some extent in practice in the securities investments industry. These include the “Sharpe Ratio”, the “M-squared” (or Modigliani & Modigliani – father &daughter) Measure, the “Jensen’s Alpha”, and the Treynor Ratio. The Treynor Ratio would make the most sense for real estate, in theory, because real estate investment managers are generally not responsible for the investor’s entire wealth portfolio (as it does not make sense to allocate the entire portfolio just to the private real estate asset class). But in applying the Treynor Ratio, it is crucial that the “beta” must reflect the amount of risk as it matters to the capital market. Only in this way can we estimate the market’s price of risk, that is, the way the market would adjust the expected return risk premium, to reflect the amount of risk in the portfolio.

The The Quantitative Quantitative Approach: e.g., The Approach: e.g., The TreynorTreynor RatioRatio......

Avg. Excess Return

Beta

SML

11

rM - rf

0

ri - rf

TRTRii

ββii


Here is a “picture” of the “Treynor Ratio” method of controlling for risk by use of this risk-adjusted return measure. A manager’s or portfolio’s Treynor Ratio is its excess return over T-bills divided by its “beta”. In the example pictured here, the manager’s excess return (ri-rf) is adjusted downward to reflect the effect of the manager taking on greater risk. The manager’s benchmark could be similarly adjusted. Then the manager and his benchmark could be compared on an equal-risk, “apples-to-apples” basis, using their Treynor Ratios.

27.2.6. Controlling for Risk . . .27.2.6. Controlling for Risk . . .Within the private real estate investment industry, we must Within the private real estate investment industry, we must rely primarily on the “Nonrely primarily on the “Non--Quantitative” (StyleQuantitative” (Style--Based) Based) Approach to controlling for risk. Approach to controlling for risk. Why?Why?

Alas, within the private real estate asset class, we cannot use this quantitative approach. We can only use the “Non-Quantitative” Approach to controlling for risk. Why? . . .

27.2.6. Controlling for Risk . . .27.2.6. Controlling for Risk . . .Within the private real estate investment industry, we must Within the private real estate investment industry, we must rely primarily on the “Nonrely primarily on the “Non--Quantitative” (StyleQuantitative” (Style--Based) Based) Approach to controlling for risk. Approach to controlling for risk.

•• We cannot effectively quantify risk in the private We cannot effectively quantify risk in the private real estate asset class real estate asset class in the way that it matters to the in the way that it matters to the capital marketcapital market, so we cannot quantify , so we cannot quantify riskrisk--adjusted adjusted return measuresreturn measures with sufficient credibility for use in with sufficient credibility for use in benchmarking.benchmarking.

•• The property market cannot distinguish between the The property market cannot distinguish between the amount of risk in different types or locations of amount of risk in different types or locations of properties (so long as they are all “institutionalproperties (so long as they are all “institutional--quality”).quality”).

Here’s why: We don’t have a good “CAPM”, a simple quantitative equilibrium model of the way the capital market perceives and prices risk in expected returns within the private real estate asset market. Part of the problem is that the property market cannot distinguish between the amount of risk in different types or locations of properties (at least so long as they are all “institutional-quality” or “core” properties).

NCREIF Division/Type Portfolios: Returns vs NWP Factor Risk

-1.5%

-1.0%

-0.5%

0.0%

0.5%

1.0%

1.5%

-0.5 0.0 0.5 1.0 1.5

Beta w rt NWP=(1/3)St+(1/3)Bn+(1/3)RE

Avg

Exc

ess

Retu

rn (o

vr T

-bills

)


Horiz. Axis is Beta wrt“National Wealth Portf”, but that doesn’t matter.

Within the private real estate investment industry, we must Within the private real estate investment industry, we must rely primarily on the “Nonrely primarily on the “Non--Quantitative” (StyleQuantitative” (Style--Based) Based) Approach to controlling for risk. Approach to controlling for risk. Recall Ch.22 (sect.22.3)…Recall Ch.22 (sect.22.3)…

The asset market cannot distinguish between the amount of risk in the different real estate market segments within the institutional real estate asset class. There is no risk/return relationship evident empirically ex post within this asset class, for any reasonable, practical measure of risk (such as volatility or “beta”). (However, see Section 22.3.3: Perhaps more sophisticated models will enable more sophisticated risk analysis in the future.) In this plot, institutional real estate portfolio returns (actually, NCREIF sub-index returns) are plotted against beta computed with respect to the “National Wealth Portfolio” (which consists stylistically of (1/3) stocks, (1/3) bonds, and (1/3) real estate). But it does not matter how we compute the “beta”. (For example, we can compute it with respect to the stock market, the NCREIF Index, National Consumption, whatever…) There is no ex post relationship between risk and return. This means that there is no significant distinction in ex ante real estate risk premia for different types or locations of (institutional quality) commercial property, within the private property market. In other words, the cross-section of expected returns across property types & locations is essentially FLAT (within the institutional quality asset class). (See Geltner-Miller Chapter 22, section 22.3, pp.573-583.)

At this level of detail, At this level of detail, WITHIN A WITHIN A GIVEN STYLEGIVEN STYLE ofof the (the (instlinstl) R.E. ) R.E. asset class, the market cannot asset class, the market cannot distinguish stable, reliable distinguish stable, reliable differences at the property level in differences at the property level in ““betabeta”, the risk that matters.”, the risk that matters.


Therefore, we not only Therefore, we not only cannotcannot, but we , but we need notneed notadjust for risk, as long as the manager & his adjust for risk, as long as the manager & his benchmark are both confined benchmark are both confined WITHIN A GIVEN WITHIN A GIVEN STYLESTYLE of the institutional RE asset class.of the institutional RE asset class.

Within the private real estate investment industry, we must Within the private real estate investment industry, we must rely primarily on the “Nonrely primarily on the “Non--Quantitative” (StyleQuantitative” (Style--Based) Based) Approach to controlling for risk. Approach to controlling for risk.

Perhaps this lack of a risk/return relationship is because, at this fine level of detail, within the institutional property asset class, stable and reliable differences are not perceptible in the risk that matters to the property market.

Real Estate “Styles” are a bit vaguely defined:Real Estate “Styles” are a bit vaguely defined:•• Domestic “Core”Domestic “Core”•• ValueValue--added/Opportunisticadded/Opportunistic•• InternationalInternational•• etc (NCREIF currently working on etc (NCREIF currently working on standardized style definitions).standardized style definitions).


Discrete style/risk categories defined by:Discrete style/risk categories defined by:•• Property size limits;Property size limits;•• Leverage limits;Leverage limits;•• Development project exposure limits (occupancy).Development project exposure limits (occupancy).

In practice, this approach requires considerable In practice, this approach requires considerable specification & elaboration specification & elaboration betwbetw agent & principal at outset agent & principal at outset of mgt contract, & perhaps onof mgt contract, & perhaps on--going. going.

27.2.6. Controlling for Risk . . .27.2.6. Controlling for Risk . . .Example:Example:•• Fund restricted to fullyFund restricted to fully--stabilized core property stabilized core property

investments, but investments, but allows use of high leverageallows use of high leverage (up to 75%, (up to 75%, at discretion of fund mgr).at discretion of fund mgr).

•• Appropriate benchmark & risk control could be either:Appropriate benchmark & risk control could be either:1.1. Peer universe of all (& only) other such funds, orPeer universe of all (& only) other such funds, or2.2. NCREIF Index adjusted for leverage using a WACC NCREIF Index adjusted for leverage using a WACC

formula, with a preformula, with a pre--specified target leverage ratio:specified target leverage ratio:rrEE == rrDD + (+ (rrPP –– rrDD))LRLR

Where: Where: rrEE == Levered equity return,Levered equity return,rrDD == Debt return,Debt return,rrPP == Property (Property (unleveredunlevered) return (NPI),) return (NPI),LRLR = = Leverage ratio (Leverage ratio (V/EV/E), pre), pre--specified.specified.

27.2.6. Controlling for Risk . . .27.2.6. Controlling for Risk . . .Some implications of styleSome implications of style--based benchmarking:based benchmarking:•• Fixed real return targets are not good benchmarks.Fixed real return targets are not good benchmarks.

•• Direct property investment should not generally be Direct property investment should not generally be benchmarked against a REIT Index, benchmarked against a REIT Index, howeverhowever::

•• Property operational level performance comparisons Property operational level performance comparisons & attribution analyses may be of interest for & attribution analyses may be of interest for diagnostic purposes at the propertydiagnostic purposes at the property--level across all level across all properties of a given type, including different types properties of a given type, including different types (or styles) of investor/owners.(or styles) of investor/owners.

•• Comparisons Comparisons acrossacross styles or peer groups are of styles or peer groups are of interest at a broader level, relevant for strategic interest at a broader level, relevant for strategic investment and asset class investment and asset class researchresearch (as opposed to (as opposed to agent performance evaluation benchmarking, per se).agent performance evaluation benchmarking, per se).

The previous conclusion about inability to use quantitative The previous conclusion about inability to use quantitative (risk(risk--adjusted returnadjusted return--based) control for risk does not based) control for risk does not necessarily hold for portfolios (or mgrs) that span necessarily hold for portfolios (or mgrs) that span ACROSSACROSSmultiple asset class quadrants. .multiple asset class quadrants. . ..

•• REITsREITs

•• CMBSCMBS

•• Whole LoansWhole Loans

•• Private R.E. EquityPrivate R.E. Equity

•• etc. . .etc. . .

Controlling for Risk at a Broader Level (Ch.22, Sect.22.2). . .Controlling for Risk at a Broader Level (Ch.22, Sect.22.2). . .

The next few slides are a bit of a digression or review of material in Chapter 22. Our previous conclusion about inability to use quantitative (risk-adjusted return based) control for risk does not extend to managers or portfolios that span ACROSS multiple asset classes or “quadrants”... The capital markets are often divided into “4 quadrants”: private & publicly-traded, debt & equity products. In addition to private real estate equity, what about REITs, CMBS, Commercial mortgage whole loans?... Suppose the manager can allocate across more than one of these quadrants?...

Look at this plot of average returns and betas Look at this plot of average returns and betas ACROSSACROSSmultiple quadrants. . .multiple quadrants. . .

Ex Post CAPM on Mkt=(1/3)RE+(1/3)Bonds+(1/3)Stocks

0.0%

1.0%

2.0%

3.0%

0 0.5 1 1.5 2 2.5

Beta* RE betas = sum of 8qtrs lagged coeffs

Avg

Exc

ess

Retu

rn (p

er q

tr, o

ver T

bills

)

NCREIF

LTBond

SP500

HOUSCMORT

REIT

SMALST


Look at this regression of asset class ex post returns onto their betas with respect to the National Wealth Portfolio, as it was previously stylistically defined (equal shares of stocks, bonds and real estate). This scatter plot looks very different from the one we previously looked at within the private equity quadrant.

Regression statistics for historical returns Regression statistics for historical returns ACROSSACROSSmultiple quadrants. . .multiple quadrants. . .


0.0%

1.0%

2.0%

3.0%

0 0.5 1 1.5 2 2.5


Avg

Exc

ess

Retu

rn (p

er q

tr, o

ver T

bills

)

NCREIF

LTBond

SP500

HOUSCMORT

REIT

SMALST

•Adj. R2 = 73%

• Intercept is Insignif.

•Coeff on Beta is Pos & Signif.


Here, at this ACROSS-QUADRANT level, the historical empirical evidence is quite strong. Adjusted R-square is a very robust 73%. The “intercept” is insignificant (which is exactly what we expect when we have a good model of risk and return). The coefficient on beta is positive & significant. [See Geltner-Miller Chapter 22, section 22.2, especially pp.568-572, Exh.22-2b on p.571.]

““BetaBeta” ” wrtwrt ““National WealthNational Wealth” is priced ” is priced ACROSSACROSS the the quadrants quadrants . . .. . .


0.0%

1.0%

2.0%

3.0%

0 0.5 1 1.5 2 2.5


Avg

Exc

ess

Retu

rn (p

er q

tr, o

ver T

bills

)

NCREIF

LTBond

SP500

HOUSCMORT

REIT

SMALST

•Adj. R2 = 73%

• Intercept is Insignif.

•Coeff on Beta is Pos & Signif.

“CAPM works...”“CAPM works...”


In other words, “Beta” with respect to the “National Wealth” is priced ACROSS the quadrants. A National Wealth-based CAPM “works” pretty well at this broad-brush level. (Ch.22, Sect.22.2.2.)

The Capital Market does perceive (and price) risk differences The Capital Market does perceive (and price) risk differences ACROSSACROSS quadrants. . .quadrants. . .

National Wealth BETABETA

Pub.Pub.EqEq

Pub.DbPub.Db

PriPri.Db.Db

PriPri..EqEq

E[r]


Apparently, the capital market does perceive and price risk differences ACROSS the quadrants. PRIV EQ Quad is the low-beta quad. PUBL EQ Quad is the high-beta quad. The DEBT Quads lie in between.

Asset Class Ex Post Betas and Risk Asset Class Ex Post Betas and Risk PremiaPremia (Per (Per Annum, over TAnnum, over T--bills, 1981bills, 1981--98)...98)...

Asset Class:Excess

Return: Beta:Small Stocks 8.48% 1.94

S&P500 10.48% 1.72

REITs 4.32% 1.22

LT Bonds 6.24% 1.07

Com.Mortgs 4.15% 0.66

NCREIF 1.15% 0.34

Houses 3.59% 0.23


Here are the historical annual excess returns (over T-bills, aka “ex post risk premia”) of seven asset classes spanning the four quadrants, including four real estate asset classes. (CMBS are not included because we do not have a long enough history for them.) And here in the right-hand column are the historical ex post “betas” with respect to the National Wealth Portfolio. The beta estimates for private real estate are corrected for smoothing bias. Even so, private real estate equity is a “low-beta” asset class. It also has received a low excess return, historically. [Additional relevant discussion in Geltner-Miller Chapter 19, sect. 19.2.2, especially pp.487-488.]

The The TreynorTreynor Ratio Ratio (or something like it) should be applied to (or something like it) should be applied to multimulti--quadrant managers (portfolios)...quadrant managers (portfolios)...

Avg. Excess Return

Beta

SML

11

rM - rf

0

ri - rf

TRTRii

ββii


Because of these risk differences, the Treynor Ratio (or something like it) should be applied to multi-quadrant managers or portfolios, in order to adjust their performance measurement for risk. Note: “Sharpe Ratio” probably not most appropriate measure, because real estate is usually only a small part of the investor’s total portfolio, so the real estate share’s volatility is not what matters to the investor (and not what is “priced” in the capital market’s expected return for real estate).

The The Beta Beta can be estimated based on the can be estimated based on the ““National Wealth PortfolioNational Wealth Portfolio”” (( = = (1/3)Stocks + (1/3)Bonds + (1/3)RE (1/3)Stocks + (1/3)Bonds + (1/3)RE ) as the multi) as the multi--quadrant “Risk Benchmark”quadrant “Risk Benchmark”..

Avg. Excess Return

Beta

SML

11

rM - rf

0

ri - rf

TRTRii

ββiiBased on “National Wealth Portfolio”“National Wealth Portfolio”


The Beta can be estimated based on the “National Wealth Portfolio” ( = (1/3)Stocks + (1/3)Bonds + (1/3)RE ) . In other words, we might consider using a simplified version of the National Wealth Portfolio could serve as a multi-quadrant “Risk Benchmark”. In any case, when benchmarking for performance evaluation “across the quadrants”, we cannot ignore the need to adjust for risk in some logical manner, at least in principle… Of course, in practice there is still the problem of short time samples and lack of statistical significance (see later). This makes it difficult to distinguish results due to skill, from random outcomes, ex post. But, as noted previously, benchmarking can still “do its job” in the ex ante purpose of incentive alignment and communication.

27.2.3. Segment weighting in the benchmark index. . .27.2.3. Segment weighting in the benchmark index. . .

Definition:Definition:Real estate asset market “Real estate asset market “segmentsegment””

= Property type & geographic location.= Property type & geographic location.

•• Suppose manager has discretion to allocate capital Suppose manager has discretion to allocate capital across more than one segment.across more than one segment.

•• Then benchmark should include all the segments Then benchmark should include all the segments manager could choose among.manager could choose among.

•• But what should be the allocation weights across But what should be the allocation weights across the segments in the benchmark index? . . .the segments in the benchmark index? . . .

Getting back to our focus on benchmarking within the private real estate investment industry, another important topic is that of the segment weighting in the benchmark index. Should the allocation across the different property types and geographic regions in the benchmark be whatever it is in the aggregate NCREIF Property Index (the NPI)? Or should it be some custom-weighting more appropriate for the given subject manager or portfolio being evaluated?…


•• No single general rule for benchmark segment allocation No single general rule for benchmark segment allocation weights.weights.•• Weights should be “Weights should be “appropriateappropriate”. Should reflect:”. Should reflect:

Client’s objectivesClient’s objectives for the manager;for the manager;Manager’s specializationManager’s specialization and;and;Manager’s range of Manager’s range of discretiondiscretion over allocation policy.over allocation policy.

•• General management principle involved is:General management principle involved is:

RESPONSIBILITYRESPONSIBILITY AUTHORITYAUTHORITY

EVALUATIONEVALUATION

Equate responsibility to authority, & evaluate accordingly.Equate responsibility to authority, & evaluate accordingly.


General implications:General implications:•• Peer universe subPeer universe sub--indexes should often be used to indexes should often be used to construct customconstruct custom--weighted benchmark, with weighted benchmark, with weights fixed weights fixed in advancein advance (i.e., NPI used only for benchmark (i.e., NPI used only for benchmark withinwithin--segmentsegment return performance).return performance).•• Example: Example:

•• Mgr hired to place capital into industrial & office properties Mgr hired to place capital into industrial & office properties at at her discretion:her discretion:•• Benchmark = 50% NPI Office + 50% NPI Industrial.Benchmark = 50% NPI Office + 50% NPI Industrial.

•• In absence of clear reason otherwise, simple equalIn absence of clear reason otherwise, simple equal--weighting may make sense.weighting may make sense.•• Typical mgr will avoid bet against benchmark, so Typical mgr will avoid bet against benchmark, so benchmark weight should reflect client’s allocation target benchmark weight should reflect client’s allocation target objective. (e.g., skew weights accordingly).objective. (e.g., skew weights accordingly).

Always remember that risk-aversion tends to lead many if not most managers to try to mimic the weights in their benchmark, at least to a considerable extent. An unintended consequence of widespread benchmarking, using the same standard benchmark index, can thus be to exacerbate “herd behavior” in the private real estate investment industry. Such a result would not be beneficial for the industry in the long run.

Note:Note:Previous general point about fixing benchmark weights Previous general point about fixing benchmark weights in advance to reflect client’s target objective for the mgr in advance to reflect client’s target objective for the mgr holds not only for market segment weighting in the holds not only for market segment weighting in the benchmark, but also for other characteristics of the benchmark, but also for other characteristics of the benchmark, such as the degree of leverage, the benchmark, such as the degree of leverage, the allocation across different styles (e.g., development allocation across different styles (e.g., development exposure), etc. (e.g., for benchmarking a multiexposure), etc. (e.g., for benchmarking a multi--style style manager).manager).


For example, if a manager has discretion to use leverage, then the benchmark might be calculated with a previously agreed upon target leverage ratio. For example, if the idea is to be aggressive (use substantial leverage), a 60% target LTV ratio might be established for the manager (or fund). If otherwise the fund is to invest in all the four NCREIF core property types and regions, then the 16 NCREIF property type/region sub-indices might be used to create the benchmark (either with equal-weighting, or some other appropriate weighting reflecting the client’s expectations and the manager’s expertise), only geared up to a leverage ratio of 2.5 [= 1/(1-0.6)]. If the manager decides to use a different leverage ratio than this target, then his performance differential wrt his benchmark will reflect the ex post effect of his decision in this regard. The WACC formula and a bond (or mortgage) periodic return (HPR) index can be used to simulate the effect of leverage in the benchmark. WACC formula: rP = (D/V)rD + (E/V)rE , where rP = Property return (unlevered, as from NCREIF Index), rD = Debt return (as from an HPR mortgage index, like Giliberto-Levy Mortgage Price Index), rE = Equity return (levered), (D/V) = Debt value as fraction of underlying asset gross property holdings value (loan/value ratio, LTV), (E/V) = Equity value as fraction of underlying asset gross property holdings value (1 – LTVratio). Just invert the formula and solve it for rE each period: rE = [rP – (D/V)rD]/(E/V)


•• A reasonable (& popular) alternative is to use the peer A reasonable (& popular) alternative is to use the peer universe weights (e.g., NPI weights for the segments the universe weights (e.g., NPI weights for the segments the mgr has discretion to invest in).mgr has discretion to invest in).

•• Rationale is “Rationale is “client client investabilityinvestability” criterion:” criterion:-- It’s what client would have gotten (ex ante) if they had It’s what client would have gotten (ex ante) if they had

chosen a “typical” alternative (competing) mgr.chosen a “typical” alternative (competing) mgr.

Other alternatives for benchmark segment weighting:Other alternatives for benchmark segment weighting:


•• Theoretical alternative: Use Theoretical alternative: Use market market weights weights –– Share of Share of each segment in total asset class market value.each segment in total asset class market value.

•• Advantage of this approach:Advantage of this approach:•• May reflect mgr’s relative ability to find available propertiesMay reflect mgr’s relative ability to find available properties..•• May reflect relative ability of market to absorb new capital.May reflect relative ability of market to absorb new capital.

•• Problems with this approach:Problems with this approach:•• Since CAPM doesn’t work well for R.E., we don’t know if Since CAPM doesn’t work well for R.E., we don’t know if market weights are theoretically optimal from a portfolio stratemarket weights are theoretically optimal from a portfolio strategy gy perspective.perspective.•• Empirically it is difficult to precisely determine the market Empirically it is difficult to precisely determine the market weights.weights.

Other alternatives (cont.):Other alternatives (cont.):

Ch.23 & Ch.25: Valuation Methodology. . .Ch.23 & Ch.25: Valuation Methodology. . .

How asset & portfolio values are measured or estimated in How asset & portfolio values are measured or estimated in constructing the benchmark index (and also in computing constructing the benchmark index (and also in computing the subject manager’s performance record).the subject manager’s performance record).

•• Fundamental issue relevant not just for benchmarking.Fundamental issue relevant not just for benchmarking.

•• Issue that is unique and particular to private real estate.Issue that is unique and particular to private real estate.

Fundamental source of the problem is that in private real Fundamental source of the problem is that in private real estate markets:estate markets:

•• uniqueunique, , whole assetswhole assets are tradedare traded•• infrequently and irregularlyinfrequently and irregularly through time, through time, •• in deals that are privately negotiated between in deals that are privately negotiated between one one buyer and one sellerbuyer and one seller..(All three of these characteristics differ from securities (All three of these characteristics differ from securities mktsmkts.).)

See Chapters 23 & 25.

IRR IRR vsvs TWRR BenchmarkingTWRR Benchmarking. . .. . .

Benchmarking is an exercise in comparison of multiBenchmarking is an exercise in comparison of multi--period period returns.returns.

Recall from Chapter 9 that there are two types of multiRecall from Chapter 9 that there are two types of multi--period return measures (TWRR & IRR), and that:period return measures (TWRR & IRR), and that:

•• TWRR is the standard performance measure in the TWRR is the standard performance measure in the securities industry and in traditional (“core”) real estate securities industry and in traditional (“core”) real estate investment.investment.

•• TWRR is neutral with respect to the TWRR is neutral with respect to the timingtiming of capital flow in of capital flow in and out of the investment.and out of the investment.

•• IRR reflects such timing, and IRR reflects such timing, and

•• IRR can be computed without regular periodic “IRR can be computed without regular periodic “markingmarking--toto--marketmarket” (appraisals).” (appraisals).

See Ch.9 & earlier discussion of property-level performance attribution.

IRR benchmarking is the standard practice in the private IRR benchmarking is the standard practice in the private capital (e.g., “venture capital”) investment industry.capital (e.g., “venture capital”) investment industry.

It is also appropriate for performance evaluation in at least It is also appropriate for performance evaluation in at least two applications in the real estate investment industry:two applications in the real estate investment industry:

•• Benchmarking certain types of Benchmarking certain types of “opportunistic”“opportunistic” real real estate investments, most notably estate investments, most notably development projectsdevelopment projects..•• PropertyProperty--level performance attributionlevel performance attribution of the type of the type described earlier in this lecture.described earlier in this lecture.

Because, in both cases:Because, in both cases:•• The subject investments are often The subject investments are often not regularly markednot regularly marked--toto--marketmarket, , andand•• The responsible investment managers (whose performance is beingThe responsible investment managers (whose performance is beingevaluated) are largely evaluated) are largely responsible for cash flow responsible for cash flow timingtiming decisions decisions that can significantly effect returns.that can significantly effect returns.


IRR IRR vsvs TWRR BenchmarkingTWRR Benchmarking. . .. . .IRRIRR--Based benchmarking requires the use of:Based benchmarking requires the use of:

InceptionInception--date Cohortsdate Cohorts(or “Acquisition(or “Acquisition--date cohorts”).date cohorts”).

•• IRR is computed IRR is computed since inceptionsince inception (or acquisition) of (or acquisition) of portfolio (or property).portfolio (or property).

•• IRR is computed for appropriate benchmark IRR is computed for appropriate benchmark cohortcohort, , defined as similar investments with same inception (or defined as similar investments with same inception (or acquisition) date, over same period of time as subject.acquisition) date, over same period of time as subject.

•• Benchmark cohort IRR normally “Benchmark cohort IRR normally “pooledpooled” (all cash flows ” (all cash flows aggregated together each period).aggregated together each period).

•• Terminal (current) value must be estimated for all assets Terminal (current) value must be estimated for all assets remaining in pool at end of evaluation period, based on remaining in pool at end of evaluation period, based on appraisalappraisal at terminal point in time.at terminal point in time.


In real estate, IRRIn real estate, IRR--based benchmarking will often require based benchmarking will often require the use of the use of simulatedsimulated IRR cohortsIRR cohorts as the benchmark, as the benchmark, because of lack of sufficient data.because of lack of sufficient data.

See: J.Fisher & D.Geltner, “Property Level Benchmarking of Real Estate Development Investments”, Real Estate Finance 19(1): 71-87, available in the course Supplemental Reading packet.

chapter 27: real estate investment management part ii ... · chapter 27 outline: introduction 27.1...

Documents