2 � WWW.CFAINSTITUTE.ORG

and valuing unlisted infrastructure investments, whichmay also apply to other unlisted asset classes. Finally, weoffer specific suggestions to build on current practices inorder to refine institutional investors’ approach to infra-structure benchmarking and valuation. The article endswith a discussion of implications of the proposed changesfrom the standpoint of the chief investment officer, infra-structure team, risk group, and performance group.

To refine its approach to infrastructure benchmark-ing and valuation, Alberta Investment ManagementCorporation (AIMCo) conducted a three-week brain-storming, critiquing, and idea refining session. The moti-vation for this article was drawn from discussions betweenthe authors and others at this session. The analysis isinformed by interviews with nine institutions that main-tain dedicated infrastructure investment allocations, aswell as a review of scholarly literature.

The intended contribution is to improve insight intothe complex interplay between benchmark selection andimplementation decisions and other interdependentissues within an institutional investment organization,including asset allocation, investment selection, valua-tion, performance measurement, value-added compensa-tion, risk management, and hedging.

APPROACHES USED IN PRACTICE TODAY• As the unlisted infrastructure asset class arose, insti-

tutional investors experimented with a hodgepodgeof approaches to benchmark their portfolios. No stan-dard has emerged yet. Our poll of nine institutionalinvestors showed a diversity of distinct approaches:

• British Columbia Investment Management Corpo-ration (bcIMC): 8% absolute return with adjustmentsfor asset, country, and currency risks

• Borealis Infrastructure: Absolute return set at thebeginning of the year based on operating plan

• California Public Employees’ Retirement System(CalPERS): CPI + 5%

• Caisse du Depot: 50% S&P 500/TSX + 25% S&P500 + 25% MSCI EAFE Index

• CPP Investment Board (CPPIB): Calculated on aninvestment-by-investment basis:(w1) (CDN equity return) +(w2) (CDN bond return) +(w3) (Equity return of asset currency/country) +

(w4) (Bond return of asset currency/country) +Inflation sensitivity + Leverage sensitivity

• Municipal Employees’ Retirement System (MERS)of Michigan: Barclays Aggregate Bond Index

• OPSEU Pension Trust (OP Trust): CPI + 5%• Ontario Teachers’ Pension Plan (OTPP): CPI + 4%

+ Sovereign spread (where CPI is based on countryand currency of investment)

• PSP Investments: CPI + Bond return + EquitypremiumOverall, two institutions use absolute real return

approaches. Two use simple CPI-plus approaches. Twouse CPI plus bond and equity premiums. One uses a fixedincome index. Finally, three institutions use hybridapproaches, although each with rather different compo-nents, weightings, and construction. Our discussionswith the various institutions suggest that few are partic-ularly satisfied with their benchmark choice.

WHY THE HODGEPODGE?The diversity of benchmark approaches reflects a numberof factors, including the newness and heterogeneity of theinfrastructure asset class, variations in risk–return expec-tations across institutions, and a lack of widely cited time-series performance data for infrastructure. Several institu-tions noted a desire for greater benchmark stability whenselecting real return benchmarks. Some hybrid approachesmay represent an attempt to integrate more of the desiredfeatures of infrastructure investing (low volatility,inflation-linkage, cash yield, etc.) into a single composite.

The diversity also reflects the fact that infrastructureinvestments play different roles in different investors’portfolios. For example, a recent poll found that fewerthan 40 percent of pensions have a unique “infrastructure”allocation.2 Some plans group infrastructure in with “realassets,”3 emphasizing tangible physical assets believed tohave an intrinsic value, such as gold and commodities,real estate, and equipment; some with “inflation-linked,”4

emphasizing long-term asset and liability matching; andsome with “absolute return,”5 emphasizing the impor-tance of achieving a net return above other asset classes,usually global equities. Other investors view infrastruc-ture as a subset of “fixed income,” which recognizes thepotential to provide predictable income distributionsfrom day one.6 Yet others have added it in alongside

©2012 CFA INSTITUTE � 3

general private equity, alternatives, and real estate alloca-tions. Unsurprisingly, different strategic approachesnecessitate different benchmarks.

Despite the different allocations, the conventionalwisdom among institutional investors is that on the risk–return continuum, infrastructure investments fit some-where between regular equities and fixed income, asFigure 1 shows.

An investment’s characteristics determine its loca-tion on the curve. It is this line of reasoning that ledAIMCo in 2009 to move toward adopting a hybridbenchmark of 50 percent real return bond (DEX RealReturn Bond [RRB] Index) and 50 percent internationalequities (MSCI World Index). At AIMCo, the view isthat “infrastructure shouldn’t just be another form ofequity. It should be a somewhat higher return, somewhathigher risk substitute for real return bonds.”The emphasisis on finding the relatively small subset of “unrisky”projects that generate predictable, long-term, inflation-linked cash flows with low volatility. These so-called“core” infrastructure investments offer a sort of holy grail,

generating equity-like returns with bond-like risks andserving as a first-order proxy for long-dated liabilities.

BENCHMARK PERFORMANCEFigure 2 presents the cumulative performance of four ofthe benchmark families discussed above. The figure illus-trates that the straight DEX RRB and the CPI + 6%indices generated the lowest volatility and the highesthurdle over the 10-year period ending August 2009, withannualized returns of approximately 8.2 percent and 8.1percent, respectively. The 50% MSCI World/50% DEXRRB and the UBS Global Utilities and InfrastructureIndex generated lower annualized returns, with 4.1 per-cent and 3.1 percent, respectively. The lower returns areexplained by two significant market corrections over the10-year period, in 2001 and 2008. In addition to gener-ating the lowest hurdle, the UBS Global Utilities andInfrastructure Index also exhibited the highest volatility.7

This analysis shows that different benchmarks producequite different return expectations and volatility levels. �

Figure 1. Risk–Return Continuum for Unlisted Infrastructure Assets

Source: Based on authors’ analysis.

4 � WWW.CFAINSTITUTE.ORG

Figure 2. Cumulative Historical Performance of Benchmark Families (1999–2009)

©2012 CFA INSTITUTE � 5

NOTES

1. For a more detailed discussion of these topics, see Weisdorf (2007).2. A survey by Probitas Partners of more than 200 executives from

global institutional funds found that nearly 40 percent of investorsuse dedicated infrastructure allocations. See “Investing in Infra-structure,” summer 2009.

3. The Teacher Retirement System of Texas program has a “real asset”allocation with an intended objective “to contribute favorably to diver-sification of the Total Fund through exposure to real assets’ low ornegative correlation to the Public Markets." See www.trs.state.tx.us/investments/documents/investment_policy_statement.pdf andwww.trs.state.tx.us/about/documents/infrastructure_investing_opportunities.pdf.

4. CalPERS has assigned a benchmark of CPI + 4% for the Inflation-Linked Asset Class and CPI + 5% for the infrastructure component.See: www.calpers.ca.gov/eip-docs/investments/policies/inv-asset-classes/fixed-income/ILAC.pdf.

5. California State Teachers’ Retirement System (CalSTRS) createdsuch a program in 2009, with the stated goal of achieving a targetnominal return of 6.5 percent and 6.1 percent standard deviationthrough allocations to infrastructure and Treasury Inflation-Protected Securities (TIPS).

6. MERS of Michigan created a sub-asset class for infrastructure in itsfixed income portfolio.

7. Although we do not present the data, our analysis also found thatthe UBS Global Infrastructure & Utilities Index is highly corre-lated with the S&P Global Infrastructure Index, Macquarie GlobalInfrastructure 100, and Dow Jones Brookfield Global Infrastruc-ture Index. Hence, we chose not to include any of these others.

REFERENCES

Bailey, J.V. 1992. “Evaluating Benchmark Quality.” Financial Analysts Journal, vol. 48, no. 3 (May/June):33–39.

Blake, D., and Timmermann, A. 2002. “Performance Benchmarks for Institutional Investors: Measuring, Monitoring, and Modifying Investment Behavior.” The Pensions Institute PI-0106 Discussion Paper:108–141.

Ljungqvist, A., and Richardson, M. 2003. “The Cash Flow, Return and Risk Characteristics of Private Equity.” New York University Working Paper:1–41.

McIntosh, W. 1997. “Real Estate Portfolio Benchmarking.” Journal of Real Estate Portfolio Management, vol. 3, no. 1:75–77.

Meketa Investment Group. 2009. “Infrastructure.” Research documents (February).

Nesbitt, S.L., and Reynolds, H.W. 1997. “Benchmarks for Private Market Investments.” Journal of Portfolio Management, vol. 25, no. 4:85–90.

Northern Trust. 2007. “How the Liability Benchmark Is Developed and Used in Practice.” Company research report.

Vineburg, S. 2007. “A Look at Wholesale Infrastructure Fund Benchmarks.” CFS Infrastructure Research Note.

Weisdorf, M.A. 2007. “Infrastructure: A Growing Real Return Asset Class.” CFA Institute Conference Proceedings Quarterly, September 2007.

Jagdeep Singh Bachher is deputy chief investment officer, changemanagement, at Alberta Investment Management Corporation.

Ryan J. Orr is executive director at Stanford University’s Collabo-ratory for Research on Global Projects.

Daniel Settel is co-founder and vice president of operations atZanbato Group, LLC.

6 � WWW.CFAINSTITUTE.ORG

Table 2. Strengths and Weaknesses from the Perspectives of Various Organizational Groups

Benchmark Family Strengths Weaknesses

Absolute Return PG—Unambiguous and easy to implement.IT—Sets a stable performance assessment target and makes it easy to assess attractiveness of investment opportunities.

PG—Overall value is subjectively determined.RG—No relation to risk present in the portfolio.CIO—Cannot be used to measure the opportu-nity cost of not investing in public markets.

Inflation + Margin PG—Unambiguous and easy to implement.IT—Sets a stable performance assessment target and makes it easy to assess attractiveness of investment opportunities.

PG—Margin value is subjectively determined.RG—No relation to risk present in the portfolio.CIO—Cannot be used to measure the opportu-nity cost of not investing in public markets and is not investable.

Fixed Income + Margin CIO—Captures the opportunity cost of not investing in a fixed-income portfolio.IT—Sets a relatively stable performance assessment target and makes it easy to assess attractiveness of investment opportunities.

PG—Margin is subjectively determined.RG—No relation to risk present in the portfolio.

Equity Return + Margin CIO—Captures the opportunity cost of not investing in a fixed-income portfolio.

PG—Margin is subjectively determined.RG—Little relation to risk present in the portfolio.

Hybrid Return (e.g., 50% MSCI World/50% DEX RRB)

CIO—Captures the opportunity cost of not investing in a combined public debt and equity portfolio.

PG and IT—Does not capture unlisted assets; therefore, it cannot be used to measure the value-add contribution of private investment managers. RG—Little relation to risk present in the portfolio.

Custom Portfolio Benchmark IT—Captures some of the securities that the IT has available within the private opportunity set.RG—A well-constructed portfolio benchmark (or a whole set of tiny customized baskets mirroring each asset) is perhaps the best mirror of risk contained within the portfolio.

PG and IT—Does not capture unlisted assets; therefore, it cannot be used to measure the value-add contribution of private investment managers. Moreover, it has chaotic short-term volatility typical of all benchmarks made up of listed securities.

Peer Group PG and IT—Can capture the unlisted universe and opportunity set; therefore, it can be used to measure the value-add contribution of private investment managers.

CIO—Cannot be used to measure the opportu-nity cost of not investing in public markets and is not investable.PG—Obtaining the data is difficult.IT—Peer group data may originate from inconsistent valuation methodologies, and the return numbers may be skewed upward by survivorship bias.RG—No relation to risk present in the portfolio.

Liability Based CIO—Aligns investment decisions with long-term obligations and matches cash realizations from investments with distributions to pensioners.

CIO—Cannot be used to measure the opportunity cost of not investing in public markets and is not investable.IT—Benchmark does not create a useful tool for assessing manager value-add.

Legend:CIO = Chief investment officer.PG = Performance group.RG = Risk group.IT = Infrastructure team.

©2012 CFA INSTITUTE � 7

A Farming Analogy. Analogies can make complexproblems more tractable. Perhaps a good way to thinkabout measuring performance of the infrastructure port-folio is with the aid of a farming analogy:

When a farmer plants his fields with wheat, potatoes,and Christmas trees, he doesn’t know what the marketprice will be at harvest; commodity prices tend to bouncearound quite a bit during the growing season. A lot ofmoney gets ploughed into the ground when the field isfirst planted. Over the course of many months (or years,in the case of Christmas trees), green shoots sprout upand seedlings grow stronger. The farmer doesn’t commita lot of time and money to measuring with a ruler theheights of new shoots and seedlings but instead directshis efforts towards more productive activities: fertilizing,watering, and weeding. Occasionally, there will be a hailstorm or a bug infestation and crops will be damaged.When a negative event occurs, at that point the farmerwill likely assess the financial implications and recalculatethe values of his crops. But generally, if crops are growingaccording to plan, there is no need to revalue on a quar-terly or more frequent basis. Trying to track the value ofthe crops based on daily commodity movements wouldbe a greater degree of precision than warranted.

Investing in illiquid infrastructure assets with a buy-and-hold strategy over long time horizons is a lot likemanaging a multicrop farm. Once capital has been com-mitted, it is difficult to redeploy. Returns cannot beknown precisely until after the harvest, when the actualperformance is tallied.

General Principles. The following are eight gen-eral principles for selecting and implementing bench-marks for unlisted assets.1. Separate the “opportunity cost” and “manager value-

add” problems. There are two very distinct problemswhen selecting and implementing a benchmark. Thefirst is the CIO’s initial asset allocation problem—how to assess the opportunity cost of whetherunlisted infrastructure investments are attractive rel-ative to other opportunities to deploy capital. Thesecond is the ongoing manager value-add problem—how to determine whether or not the investmentteam’s performance is beating the performance ofother investment teams who are also in the same assetmarket at the same point in time. Both of thesequestions are critical and lead in the direction of a

two-tiered benchmarking system. The first problemcalls for an investable equity or bond or a customportfolio benchmark. The second problem calls foreither a peer group or, in its absence, perhaps acustom portfolio benchmark. It may not always bepossible to use one benchmark to achieve both pur-poses. The benchmark for the CIO’s purposes mustbe investable so that if attractive infrastructureopportunities are limited in supply in a given year,then the capital can be allocated to the listed bench-mark alternative.

2. Take the long view. A long-term investment strategyrequires a long-term approach to benchmarking andperformance evaluation. This principle has a numberof ramifications. The first ramification is that thebenchmark must have long-term staying power; itmust survive as a fair yardstick of opportunity costand manager value-add for more than 20 years,including times of high inflation, stock market bub-bles, and severe recessions. Staying power is lesscertain for custom portfolio benchmarks than forbroad-based indices because of inevitable changes incomponents and politics surrounding periodic rebal-ancing. Changing a benchmark can create all kindsof political problems within the organization, includ-ing potentially complex issues of retroactive compen-sation, and it can be demoralizing and frustrating forthe investment team, who has spent several yearsbuilding up a portfolio to perform well against agiven benchmark only to have to move the originalgoalposts. The second ramification is that the bench-mark should be relatively stable and reflect long-termaverages. If it is highly volatile in the short andmedium run (e.g., the equity markets are volatile ona daily basis), the infrastructure portfolio looks likeit is gaining and losing value relative to the bench-mark when in reality, the portfolio value hasn’tchanged in value at all. The third ramification is thatthe investment team should not be penalized fortaking long-term positions (i.e., plowing money intothe ground, recalling our farming analogy, and notrecovering it for a few years, if this indeed is part ofthe intended strategy). What is important from aperformance monitoring standpoint is to analyzecash flows associated with each investment relativeto cash-generation forecasts, a topic that we discussin more detail later.

8 � WWW.CFAINSTITUTE.ORG

3. Create the right incentives. The interplay betweenbenchmark choice, the institution’s policy for valu-ing assets on the book, and the institution’s policyfor awarding performance compensation creates aset of incentives that influences the decisions andbehavior of the investment team, who select, man-age, and exit investments. For example, if a valuationsystem is implemented that holds assets at cost anddoes not recognize capital appreciation until sale,then the investment team will be incentivized to“clip coupons” (i.e., to buy assets that return entirelyin the form of dividend yield) or, from time to time,to sell down the choicest assets in the portfolio thathave generated the greatest total return in order torecognize the gain. Neither of these outcomes wouldbe desirable. On the contrary, if a valuation systempermits aggressive revaluation of assets on the bookbased on mark-to-model assumptions, then theinvestment team will be more likely to hunt downassets with aggressive growth and operationalassumptions and higher levels of uncertainty in rev-enue and EBITDA forecasts. In order to create amore balanced incentive, it would be necessary toincrementally recognize economic value-add for thepurpose of awarding performance compensation—but on a conservative basis, to avoid overcompensat-ing the manager before the harvest is in. Ultimately,the incentive system must reinforce the long-terminvestment strategy and must recognize long-termperformance with long-term compensation.

4. Match benchmark and vintage period. When calculat-ing and comparing returns relative to a benchmark,it is crucial to always match the period over whichthe benchmark returns are calculated to the periodover which the asset/portfolio returns are calculated.This principle is intended to avoid calculating bench-mark and asset/portfolio returns over asymmetrictime periods. For example, if the asset/portfolioreturns are calculated over 2010–2015, then thebenchmark returns should be calculated over thatsame time period and not over, say, 2005–2010.Comparing returns over asymmetric time periods ispotentially misleading because each time period ischaracterized by a unique set of inflation, interestrate, growth, and leverage parameters. If privateinvestments are intended to exceed the public invest-ment alternatives by an illiquidity premium, then the

only way to objectively assess whether or not theyhave done so is to compare truly symmetric timeperiods. We refer to this concept as the “principle ofvintage matching.”

5. Match benchmark and valuation frequency. With aportfolio of illiquid assets, investors are well advisednot to publish new performance results relative to thebenchmark more frequently than new valuationpoints are actually available for the assets. Publishingdaily or even weekly updates to the benchmark cre-ates a lot of meaningless noise that is potentiallydistracting and confusing. Most infrastructure fundsand companies provide third-party audited financialstatements once every year. Some provide statementsquarterly or on a six-month basis, but the commondenominator is once annually. Any attempt tobenchmark the portfolio more frequently is not agood investment of time and effort, especially if thebenchmark consists of an equity market componentwith a lot of short-term market volatility. We callthis idea the “principle of frequency matching” or“the principle of valuation point matching.”

6. Contemplate cost–benefit trade-offs. In establishing asystem of benchmarks and valuation protocols forunlisted assets, it is possible to create a process that ishighly time and resource intensive, with more fre-quent points of measurement, or alternatively, onethat is less burdensome and more straightforward.The general rule of thumb when considering variousoptions is to assess the marginal benefit of addingadditional features and complexity. For example,based on our first principle, it can sometimes be desir-able to have two separate benchmarks—one for deter-mining opportunity cost when making allocationdecisions and the other for assessing manager value-add. Of course, going to a two-layer system increasesthe cost and complexity of implementation. Thisprinciple also builds on the idea that in a portfolio thatresembles a farming operation, with deep cash outlays,a long-term hold period, and illiquidity, there is neg-ligible benefit to overmeasuring in the short run. Wecall this the “principle of operational efficiency.”

7. Track the business plan. It is difficult to know how tointerpret early performance for a long-term invest-ment, just as the farmer cannot forecast his harvestfrom the first shoots and seedlings. After cash goesin, it may take several years before cash distributions

©2012 CFA INSTITUTE � 9

accumulate to the point of ensuring success. Ratherthan spending time measuring shoots and seedlings,it would make more sense to invest time in watering,fertilizing, and weeding. One solution is to put moreemphasis on tracking the performance of the invest-ment relative to the initial business plan (i.e., track-ing operational targets, performance milestones, anddeviations from base-case financial model projec-tions). It would represent a paradigm shift for theindustry, but the investment team compensationcould conceivably be indexed to the achievement ofthe initial business plan and cash flow projections,basically tracking the performance-based compensa-tion for senior executives in each of the underlyingbusinesses. The benefit of tracking business develop-ments more closely is that principal direct investorswould be better prepared to utilize board seats andvoting rights to provide strategic input and holdmanagement to account. We refer to this as the“principle of endogenous benchmarking,” which is alabel that communicates the idea of wanting to trackthe success of an investment against the specific goalsand objectives of each business rather than against anexogenous basket of unrelated securities.

8. Revalue conservatively. There are two risks if illiquidasset values are marked up too quickly on the bookbefore performance is unequivocally borne out. Thefirst risk relates to overcompensating managers.3

The second risk involves positioning the organiza-tion for negative press coverage and board scrutiny ifwrite-downs occur as a result of overambitiousmarks. As such, the revaluation process should err onthe side of conservatism, which is a common princi-ple in cost accounting. Markdowns should occurimmediately as value is impaired; markups shouldonly occur if multiple years of consistently highercash flows have been reported.

Future research should examine how the infrastructurebenchmark can be adjusted with appropriate premiumsfor leverage risk, country risk, capital expenditure risk,and other factors. �

NOTES

1. The few available indicative data sources on the performance ofunlisted infrastructure investments over history are as follows:Benjamin Esty, “Returns on Project-Financed Investments: Evo-

lution and Managerial Implications,” Journal of Applied CorporateFinance, vol. 15, no. 1 (Spring 2002):71–86 and Sophie Sirtaine,Maria Elena Pinglo, J. Luis Guasch, and Vivien Foster, “HowProfitable Are Infrastructure Concessions in Latin America?Empirical Evidence and Regulatory Implications,” World BankGroup (August 2004):1–83.

2. There is a peer group index offered by the International PropertyDatabank (IPD) that solves this problem for the real estate industry.It pools data from seven or more of the Canadian pension funds thathave portfolios larger than $1.5 billion and own relatively compara-ble-sized assets. Each of the institutions pays a base fee of approxi-mately $20,000 and an additional fee for data customization.

3. In private equity, it is not uncommon to have a “clawback mecha-nism” to provide recourse in the event that the general partner iscompensated in the short term for what ultimately turns out to belousy performance in the long term.

BIBLIOGRAPHY

Bailey, Jeffery V. 1992a. “Are Manager Universes Acceptable Performance Benchmarks?” Journal of Portfolio Management, vol. 18, no. 3 (Spring):9–13.

———. 1992b. “Evaluating Benchmark Quality.” Financial Analysts Journal, vol. 48, no. 3 (May/June):33–39.

Blake, David, and Allan Timmermann. (2002) “Performance Benchmarks for Institutional Investors: Measuring, Monitoring, and Modifying Investment Behavior.” Pensions Institute Discussion Paper PI-0106:108–141.

Ljungqvist, Alexander, and Matthew P. Richardson. 2003. “The Cash Flow, Return and Risk Characteristics of Private Equity.” Finance Working Paper No. 03-001, New York University (January):1–41.

McIntosh, William. 1997. “Real Estate Portfolio Benchmarking.” Journal of Real Estate Portfolio Management, vol. 3, no. 1:75–77.

Meketa Investment Group. 2009. “Infrastructure.” Meketa Investment Group research document (February): www.meketagroup.com/documents/InfrastructureWP_001.pdf.

Nesbitt, Stephen L., and Hal W. Reynolds. 1997. “Benchmarks for Private Market Investments.” Journal of Portfolio Management, vol. 23, no. 4 (Summer):85–90.

Northern Trust. 2007. “How the Liability Benchmark Is Developed and Used in Practice.” Northern Trust research report.

Vineburg, S. 2007. “Infrastructure Research Note: A Look at Wholesale Infrastructure Fund Benchmarks.” Colonial First State Global Asset Management research (12 March).

Weisdorf, Mark A. 2007. “Infrastructure: A Growing Real Return Asset Class.” CFA Institute Conference Proceedings Quarterly, vol. 24, no. 3 (September):17–27.

Jagdeep Singh Bachher is deputy chief investment officer atAlberta Investment Management Corporation.

Ryan J. Orr is executive director at Stanford University’s Collabo-ratory for Research on Global Projects.

Daniel Settel is co-founder and vice president of operations atZanbato Group, LLC.