Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Green Urban Development: The Impact Investment Strategy of Canadian Pension Funds

Green Urban Development: The Impact Investment Strategy of Canadian Pension Funds JOURNAL OF SUSTAINABLE REAL ESTATE 2022, VOL. 14, NO. 1, 75–94 ARES https://doi.org/10.1080/19498276.2022.2125203 American Real Estate Society Green Urban Development: The Impact Investment Strategy of Canadian Pension Funds a b a c a Alexander D. Beath , Sebastien Betermier , Maaike Van Bragt , Yuedan Liu and Quentin Spehner a b c CEM Benchmarking, Toronto, Canada; Desautels Faculty of Management, McGill University, Montreal, Canada; Public Sector Pension Investment Board, Montreal, Canada KEYWORDS ABSTRACT Active management; impact This paper investigates the investment strategy that large Canadian pension funds imple- investing; pension funds; ment in the private real estate market. Even though they manage just 6% of global pension real estate; sustainable assets in our data, Canadian pension funds are responsible for 60% of the total value of pri- finance vate real estate deals that directly involve a pension fund. A key component of their strat- egy consists of internally developing and greening urban properties in core downtown areas. Using a common benchmarking methodology across funds, we show that this impact strategy delivers superior performance net of fees and drives the green development of major city centers. Introduction funds. In light of this evidence, we ask three ques- tions. Have large Canadian pension funds been able The growth of interest in sustainable investing over to generate high-risk-adjusted returns in private real the past decade raises hard questions for asset own- estate? Have they been able to combine financial ers. Will the cost of investing green result in lower profitability with positive environmental impact? If portfolio returns in the long-run? Or are there win- win strategies where green investments can gener- so, can asset owners around the World successfully ate high returns alongside positive environmental implement this strategy moving forward? impact? These questions are particularly relevant in Our analysis consists of three parts. In the first the real estate space since green real estate devel- part, we evaluate the historical performance of opment has been shown to generate significant Canadian pension funds’ real estate portfolios. Using economic value (Eichholtz et al., 2010; Fuerst & data from CEM Benchmarking, a global benchmark- McAllister, 2010; Miller et al., 2008). ing firm, we compare the annual returns of 241 real The present paper investigates the investment estate portfolios of Canadian and non-Canadian strategy that large Canadian pension funds imple- pension funds from 2005 to 2019. The sample ment in the private real estate market. These funds includes 36 funds (eight Canadian) with more than are interesting to study because they actively invest USD 50 billion of AUM. We henceforth refer to these in green buildings, such as the Deloitte Tower in funds as large funds. Importantly, the dataset Montreal and the Texas Tower in Houston, and they includes for each fund the proportion of the real allocate an unusually large share of their portfolio to estate portfolio that is private, as well as the pro- real assets. A recent study by Beath et al. (2021)finds portion that is directly managed, either through dir- that, among pension funds with over USD 50bn ect property ownership or through a wholly owned assets under management (AUM), real assets account operating subsidiary. There are 27 funds (14 for 23% of the portfolios of Canadian funds on aver- Canadian) that manage more than 50% of their real age but just 10% of the portfolios of non-Canadian CONTACT Sebastien Betermier sebastien.betermier@mcgill.ca Desautels Faculty of Management, McGill University, 1001 Sherbrooke Street West, Montreal, QC H3A 1G5, Canada 2022 The Author(s). Published with license by Taylor & Francis Group, LLC This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 76 A. D. BEATH ET AL. estate AUM directly. Among them are seven of the investments. Even though they manage 6% of glo- eight large Canadian pension funds. bal pension assets in our data, they are responsible Our performance evaluation methodology specif- for 60% of the total value of real estate deals that ically addresses the lack of consistent benchmarking directly involve a pension fund. across real estate portfolios. Many funds adjust their We then combine the information on deals from benchmark by adding an inflation index, a private Preqin with data on green building certification of risk premium, or a lag. Moreover, while some funds newly-built and retrofitted projects by Leadership in use listed REIT indices as benchmarks, such as the Energy and Environmental Design (LEED), a widely- FTSE Nareit U.S. Real Estate Index, other funds rely used green certification program. We show that on peer-based benchmarks, such as the National large Canadian pension funds systematically pursue Property Index (NPI). The use of different types of an impact strategy that consists of internally devel- benchmarks is problematic because the returns of oping, managing, and greening urban properties. REIT benchmarks are marked-to-market whereas the Specifically, they initiate and complete LEED proj- returns of peer-based benchmarks are based on ects on 27% of their portfolio of direct acquisitions appraisal prices, which are often smoothed and reported in Preqin, whereas non-Canadian funds lagged (Geltner, 1989). The lack of consistency only do so for 14% of their portfolio. The gap across benchmarks makes net value-added compari- increases once we focus on direct acquisitions of sons across funds open to interpretation. U.S. office properties where LEED certification is We address this issue by employing a common most common. Furthermore, Canadian pension benchmarking approach. For each fund, we con- funds are more likely to purchase real estate assets, struct a benchmark that consists of a fund-specific activate and complete the LEED certification pro- geographic blend of listed equity REITs, appropri- cess, and then re-sell the assets. ately (de-) levered, smoothed, and lagged to match To further investigate the green urban develop- the properties of the real estate portfolio’s reported ment strategy in large Canadian cities where returns. Our method delivers benchmarks that Canadian pension funds are most involved, we aug- closely track the reported portfolio returns, with an ment the dataset by manually collecting hundreds average correlation of 93%. Using this consistent of local deals made by the top nine Canadian pen- benchmarking methodology, we measure the per- sion funds from their annual reports and news formance of a real estate portfolio in terms of its releases. We then combine this information with net return in excess of the benchmark. We refer to data on land parcels and buildings and construct 3- this net return as net value added (NVA). D maps of the funds’ LEED value-added activity in Our analysis reveals that portfolios of large Toronto, Montreal, and Vancouver. In each city, we Canadian pension funds with a high proportion of show that the city center has been largely greened direct real estate investments outperform their by Canadian pension funds. The patterns are most benchmarks. On average, large Canadian pension striking in Toronto, which is Canada’s largest city funds that directly manage more than 50% of their and the home to the head offices of five of the real estate AUM generated an NVA of 148 basis nine largest Canadian pension funds. The area points (bps) between 2005 and 2019. We verify that located inside and surrounding Toronto’s financial this strong performance is not driven by the funds’ district is owned and greened predominantly by exposure to the local real estate market, and it is these funds. Moreover, 17 of the 26 Toronto build- not just the result of managing assets directly. ings that obtained LEED Platinum—the highest level In the second part of the analysis, we look into of certification—correspond to Canadian pension the real estate investment strategy pursued by large fund projects. Canadian pension funds. We first use deal-level data We find that Canadian pension funds do more from Preqin and compare hundreds of real estate than obtain LEED certification on existing buildings. deals made by large Canadian and non-Canadian They also undertake large greenfield projects where pension funds from 2010 to 2018. The data confirm they purchase the land, develop and green the that large Canadian pension funds are uniquely properties from the ground up. These projects tend active in the market of direct real estate to be clustered in up-and-coming districts near the JOURNAL OF SUSTAINABLE REAL ESTATE 77 city centers, such as the new Quartier des Gares in which provides multiple operational advantages Montreal and the South Core district in Toronto. including experience in developing and managing The combined impact of these projects on the cit- green projects, access to a large network of suppliers ies’ development is large. In Toronto, two of the 10 and retailers that lease the space, protection from tallest office towers currently under construction are legal risks, additional flexibility regarding employee being developed by Canadian pension funds. compensation, and the reduction of agency issues. In the third part of the analysis, we discuss the First, we describe the paper’s contribution to dif- benefits and risks of the Canadian pension funds’ ferent strands of the research literature. Next, we real estate investment strategy. We view green assess the performance of large Canadian pension urban development as an impact strategy because funds’ real estate portfolios. Then, we identify green (i) it combines financial profitability alongside posi- urban development as a key driver of the Canadian tive environmental impact and (ii) it is a strategy funds’ strategy. Finally, we discuss the benefits and where the parent investors—Canadian pension risks of this strategy for asset owners and conclude. funds—have a direct influence on the projects’ exe- cution. The benefits of pursuing this strategy intern- Contribution to the Literature ally include full control of the assets, direct Our paper contributes to several research literature communication between owners and managers, low on sustainable investing, impact investing, private fees due to the elimination of financial intermedia- investing, and the Canadian pension fund model. In ries, lower cost of project financing, and the ability this section, we briefly review these literatures and to extract the full economic profit from the real summarize the paper’s main contributions. estate ventures. Of course, there are legal and operational risks associated with the development and management Sustainable Investing of real assets. There are also liquidity risks related to A growing literature in finance shows that sustain- the long-term nature of development projects, and able investing leads to low expected returns. For risks associated with the real estate industry as evi- example, Pastor et al. (2020) show that green assets denced by the covid-19 pandemic (Ling et al., have low expected returns in equilibrium because 2020). We discuss in the paper several strategies investors enjoy holding them and because green that asset owners can take to mitigate these risks. assets hedge climate risk (see also Zerbib, 2022). One risk-mitigating strategy is the application of Consistent with this analysis, Betermier et al. (2022) green urban development to a specific set of proj- develop a supply and demand approach to capital ects that have well-defined clienteles, economies of markets and confirm that any ESG factor that scale, and a risk profile that naturally matches the expands the supply of capital to a productive asset long-term bond profile of the funds’ pension liabil- naturally results in a lower expected return for the ities. A good example is the construction of a green asset. The size of these effects remains a source of office tower in a major city center. Such a tower debate (Berk & van Binsbergen, 2022). caters to the need of corporate clientele for sustain- In this paper, we show that, in spite of the low able office space. It can also be replicated in multiple risk-return trade-off of sustainable investments, it is cities and provides a stream of rental income that still possible for investors to create a win-win for shares similarities with a stream of pension payouts. high returns on green investments by pursuing an A second risk-mitigating strategy is the concentra- internal development strategy, such as green urban tion of development projects in particular cities and development. The high return comes from the neighborhoods. Concentration allows a fund to work “alpha” side of the investment. By bringing the in an environment that it knows well and thus better development function in-house, Canadian pension manages the operational risks (Huffman, 2002). Concentration also allows the fund to create positive funds capture the developer’s economic profit asso- housing externalities (Rossi-Hansberg et al., 2010). ciated with the productive investments in green A third risk-mitigating strategy is the use of a assets. This economic profit results in high risk- wholly-owned real estate operating subsidiary, adjusted returns for the investors. 78 A. D. BEATH ET AL. Impact Investing allocating more capital to strategic assets, including real estate and infrastructure. These assets not only Impact investing is a type of sustainable investing diversify the portfolio but also align the risk of their where investors exercise their right of control to liability portfolio to their assets. The study finds that change corporate policy to generate both a financial this business model has resulted in a strong perform- return and additional environmental and social gains. ance for the pension funds both in terms of financial Previous research finds that impact funds tend to performance and liability hedging. Building on this generate lower returns than traditional funds (Barber literature, we take the analysis down to the asset et al., 2021), partly because they cater to a clientele level and identify green urban development as a key of investors who derive nonpecuniary utility from component of the active strategy that Canadian pen- investing and are thus willing to sacrifice returns. By sion funds internally pursue in real estate. comparison, our analysis shows that the green urban development strategy implemented by Canadian pension funds results in positive risk-adjusted returns. Performance of the Canadian Pension Funds’ The high returns are consistent with the evidence Real Estate Portfolios that green real estate development generates posi- We begin the analysis by assessing the performance tive economic profits (Eichholtz et al., 2010;Fuerst& of Canadian pension funds’ real estate portfolios. McAllister, 2010;Miller et al., 2008). The next section introduces a common benchmark- ing method across funds. The section on Data Private Investing describes the dataset used for the performance ana- lysis. Finally, the section on Performance of Our paper also contributes to the growing body of Canadian Pension Funds presents the results. research on the profitability of private investments. Because private assets are less liquid and more complex to operate and value than public assets, it Benchmarking Method and Performance Metric is important to understand whether and how invest- The major impediment to comparing the performance ors can benefit from investing in private markets. of real estate portfolios is a lack of consistent bench- Recent research finds that investments in private marking. We illustrate this issue in Table 1 by summa- equity and infrastructure funds may not be as prof- rizing the properties of the benchmarks used by the itable as commonly believed (Andonov et al., 2021; 241 real estate portfolios in our sample (described in Beath & Flynn, 2020; Gupta & Van Nieuwerburgh, the next Section). While some funds use listed REIT 2021; Phalippou & Gottschalg, 2009). Our analysis indices, fixed hurdle rates, inflation rates, and interest confirms that private real estate portfolios that are rates-based benchmarks, most funds (around 70%) externally managed underperform their benchmarks. use peer-based benchmarks which rely on the However, we show that private real estate portfolios appraised returns of private real estate reported by that are internally managed outperform their bench- asset managers or asset owners (e.g., NPI, IPD Canada, marks, consistent with Andonov et al. (2015). MSCI PREA). Peer-based benchmarks are often custom- ized in a variety of ways through the use of lags, cur- Canadian Pension Fund Model rency adjustments, leverage adjustments, or risk premia. The clear lack of consistency leaves perform- Finally, the paper contributes to the literature on the ance comparisons across funds open to interpretation. Canadian pension fund model. Large Canadian pen- We address this issue by employing a common sion funds are known to be highly involved in direct benchmarking method. For each fund in our sam- investments and in private markets (Ambachtsheer, ple, we construct a benchmark portfolio of local 2016, 2021;Bedart-Page et al., 2016; Lipshitz & and foreign REIT indices that is appropriately (de-) Walter, 2020). Beathetal. (2021) study the business levered, smoothed, and lagged to match the prop- model of these funds and show that the use of in- house teams allows Canadian pension funds to erties of the fund’s reported return. We then calcu- reduce costs and in turn redeploy resources by (i) late the net value added (NVA) of the fund’s growing the funds’ internal capabilities and (ii) portfolio in relation to its inferred benchmark. JOURNAL OF SUSTAINABLE REAL ESTATE 79 Table 1. Properties of real estate benchmark portfolios. benchmark return R , where D is the number B, i, tl=D Range (bps) of trading days during the year. Count Benchmark type Min Max We then solve for the combination of the four 177 Property index parameters ðg, b, s, lÞ that meets two conditions: (i) 9 Premium 120 300 the fund’s reported return R has a maximal correl- 37 Lag i, t 30 Customized blend s ation with the benchmark return R , and (ii) B, i, tl=D 17 Inflation linked 17 Premium 300 550 the fund’s reported return R has a beta of one with i, t 2 Rolling N year respect to R : The details of the optimization B, i, tl=D 7 Public market 1 Stock are provided in Appendix A. The first condition 4 Bond ensures that the benchmark is comparable to the 4 REIT 3 Premium 100 700 fund’s real estate portfolio in terms of its geograph- 3 Lag ical mix, smoothing, and lag properties. The second 40 Other 13 Own benchmark condition ensures that the benchmark has the same 1 Stock volatility as the fund’s real estate portfolio. 6 Bond 7 Property Under these conditions, the difference between 5 Inflation 6 Premium 150 600 the fund’s reported return R and the benchmark i, t 7 Absolute 600 900 return R corresponds to the net value added B, i, tl=D 1 Country premium 12 Unidentified of the fund’s real estate portfolio in year t : This table summarizes the properties of real estate benchmark portfolios NVA ¼ R  R : i, t i, t reported by pension funds. The sample includes 241 funds in the CEM B, i, t Benchmarking database. We categorize benchmarks into different types and include the range of premia (in basis points) reported by the funds. We calculate the fund’s average NVA over the sample period as Let R be the real estate return net of fees i, t reported by fund i 2f1, :::, Ig in year t 2ft , t þ 0 0 NVA ¼ NVA i i, t 1, :::, t g: The fund’s benchmark portfolio B contains n i t¼2005 a local REIT index, a global REIT index, and a local and use it as our performance metric. A-rated corporate bond index. These indices have Our method brings several advantages. It is easy returns R , R , and R : We denote LOCAL, i, t GLOBAL, i, t BOND, i, t to implement, consistently applicable to real estate by g the proportion of the REIT portfolio invested in portfolios around the World and relies on listed the global REIT index, and by b the proportion of benchmarks that are investible and give real-time the total benchmark portfolio invested in the bond information about the risks present in these port- index. The return of the benchmark portfolio is folios (Gyourko & Keim, 1992). It also provides a reli- ðÞ able way to estimate the average NVA for a group R ¼ b  R þ 1  b B, i, t BOND, i, t of funds, whereas peer-based benchmarks imply (by ðÞ ðÞ 1  g R þ g  R : LOCAL, i, t GLOBAL, i, t definition) an average NVA of zero for the peer The parameter g governs the geographic mix of group. This feature is important in the context of the benchmark and the parameter b determines its our study because our purpose is to estimate the NVA for the group of large Canadian pension funds. leverage. A positive (negative) value of b indicates Another advantage of our method is that it works that the benchmark has lower (higher) leverage well even in the absence of information about the than the REIT portfolio. geographic mix, leverage, smoothing, and lag prop- To estimate the appraisal-based returns of private erties of a fund’s real estate portfolio. real estate, we smooth and lag the (de-)levered mar- One possible limitation of our approach is the ket returns of listed equity REITs. The smoothed return risk of over-fitting the data due to the multi-dimen- R of the benchmark portfolio is estimated as B, i, t sional nature of the optimization. To reduce this s s R ¼ sR þð1  sÞ R , B, i, t B, i, t B, i, t1 risk, we restrict the number of values that the where s is the smoothing parameter. A low value parameters g, b, s, and l can take. The proportion of of s indicates a high degree of smoothing. Given a global REITs g is allowed to vary in 0.05 increments lag parameter l (in trading days), we obtain the from 0 to 1. The smoothing parameter s can take 80 A. D. BEATH ET AL. four possible values: 0.5 (maximum smoothing), correlation between the adjusted REIT and the NPI 0.67, 0.75, and 1 (no smoothing). The lag param- remains low (9%). eter l varies from 0 to 520 trading days. The weight In Figure 1C, we smooth the adjusted REIT index. in the bond portfolio b is set such that the fund’s The model suggests a value of s ¼ 0:67, which is in reported return R has a beta of one with respect the range of estimates by Barkham and Geltner i, t to R for a given combination of parameters (1995) and Brown and Matysiak (2000). The correl- B, i, tl=D g, s, and l: In the next Section, we verify that the ation between the smoothed REIT index and the inferred parameter values are consistent with survey NPI increases to 41%. results on the funds’ actual investments. In Figure 1D, we lag the smoothed series by using Figure 1 illustrates the ability of a de-levered, themodel-predicted valueof l ¼ 272 trading days. smoothed, and lagged REIT index to match the The returns of the smoothed and lagged benchmark return properties of a private real estate portfolio. look nearly identical to those of the NPI. Indeed, the We consider the NCREIF National Property Index correlation between the lagged index and the NPI is (NPI), which tracks a portfolio of U.S. properties, and equal to 93%, which is large. The lag value predicted construct the benchmark portfolio from the S&P U.S. by the model is consistent with Gyourko and Keim REIT index using the methodology described above. (1992) who find that fourth-quarter lags of REIT returns Figure 1A plots the annual return of both indices work best in predicting the returns of the NPI index. from 2005 to 2019. The returns of the S&P REIT index Altogether, this exercise confirms that using REIT (dotted grey) are twice as volatile as those of the NPI benchmarks and adjusting for their geographical in black (19 vs. 9%), and the correlation between mix, leverage, smoothing, and lag structure makes it both times-series is low (15%). Without adjustments, possible to evaluate the performance of private real the REIT index is a poor benchmark for the NPI. estate portfolios across funds. In Figure 1B, we adjust the leverage ratio of the REIT index based on the results of our estimation. Data The model suggests a value of b ¼ 0:52: The result is a significant de-levering of the REIT index, which We use data from CEM Benchmarking, a Toronto- reduces its volatility to 11%. However, the based global benchmarking company that has 40% 40% 20% 20% 0% 0% 2005 2007 2009 2011 2013 2015 2017 2019 2005 2007 2009 2011 2013 2015 2017 2019 -20% -20% -40% -40% B. De-levering the REIT Index A. REIT Index (Dotted) vs. NPI 40% 40% 20% 20% 0% 0% 2005 2007 2009 2011 2013 2015 2017 2019 2005 2007 2009 2011 2013 2015 2017 2019 -20% -20% -40% -40% C. + Smoothing the REIT D. + Lagging the REIT Index Figure 1. Application of the benchmarking method to the NPI index. This figure displays the evolution of the National Property Index (NPI, solid black) and its REITs-based benchmark (dotted grey) between 2005 and 2019. We construct the benchmark by de- levering, smoothing, and lagging the S&P U.S. REIT index in a way that maximizes the correlation between the benchmark and the NPI. (A) Reports the REIT index unmodified. (B–D) Report the incremental impacts of the leverage, smoothing, and lag adjust- ments on the benchmark. For leverage, the benchmark corresponds to a mix with a 48% weight in the REIT index and a 52% weight in the AAA-corporate bond index. The smoothing parameter is 2/3. The lag is 272 trading days. JOURNAL OF SUSTAINABLE REAL ESTATE 81 collected detailed annual cost and performance large funds. In total, there are 241 funds from eight data from more than 1000 pension, endowment, countries (55 Canadian), including 36 large funds (8 and sovereign wealth funds in 18 countries. We Canadian). We group other funds in the category retain funds that report a minimum of five years of “Rest-of-World” (RoW hereafter). The majority of real estate returns from 2005 to 2019. The choice of RoW funds (140) are U.S. funds. We also study the 2005 as the starting date comes from the fact that subsets of Canadian and RoW funds that directly i) we require several years of REIT data to apply the manage more than 50% of the real estate portfolio. Table 2A reports the average proportion of assets smoothing procedure, and ii) the S&P REIT indices we use become consistently available across coun- invested in real estate for Canadian and RoW funds tries in the early 2000s. We eliminate the 10% of in 2019. Large Canadian pension funds invest sig- nificantly more in real estate (14%) than their peers funds with the lowest return correlation between the benchmark and the funds’ real estate portfolio, (8.5% for RoW funds). This result is in line with as well as funds for which the inferred leverage Beath et al. (2021), who find that Canadian pension coefficient is either negative or above 5. All remain- funds rank first globally in terms of the proportion ing funds have a benchmark correlation above 74%, of AUM invested in real assets. and the average benchmark correlation is 93%. The table also reports the proportion of real estate We categorize a fund as large if it manages more assets managed directly. The gap between Canadian than $50 bn in 2018. Throughout the paper, dollar and RoW funds is considerable. Large Canadian funds values are expressed in USD. Table 2 reports sum- invest 81% of their real estate portfolio directly, mary statistics for the full sample and the subset of whereas RoW funds invest 16% of their real estate Table 2. Summary statistics of pension fund real estate portfolios. (A) Summary statistics Number of funds Average AUM % Real estate % Direct Canadian funds All 55 26,747 0.10 0.27 Large 8 143,985 0.14 0.81 Direct (50%þ) 14 80,481 0.12 0.86 Large and direct (50%þ) 7 150,474 0.14 0.90 R.o.W funds All 186 27,845 0.07 0.07 Large 28 118,384 0.08 0.16 Direct (50%þ) 13 39,919 0.09 0.81 Large and direct (50%þ) 4 73,182 0.08 0.77 (B) Inferred benchmark parameters Lag (l) Smoothing (s) Bond (b) Gross/net Canadian funds All 220 0.62 0.22 1.30 Large 217 0.58 0.23 1.28 Direct (50%þ) 204 0.58 0.24 1.26 Large and Direct (50%þ) 238 0.60 0.24 1.24 R.o.W funds All 196 0.63 0.13 1.24 Large 199 0.57 0.12 1.24 Direct (50%þ) 159 0.58 0.05 1.39 Large and Direct (50%þ) 92 0.58 0.04 1.37 (C) Inferred geographic mix of benchmark Canada U.S. Europe Other Canadian funds 0.79 0.14 0.02 0.05 U.S. funds 0.01 0.79 0.06 0.14 Europe/UK funds 0.00 0.18 0.74 0.07 This table displays summary statistics on the real estate portfolios of Canadian and Rest-of-World (RoW) funds. The sample includes 241 funds in the CEM database. For each group of funds, we report in (A) the average value of the funds’ AUM (in USD millions), the proportion of assets invested in real estate, and the proportion of the real estate portfolio that is managed directly as of 2018. (B) Reports average parameter values of the funds’ REITs-based benchmark portfolios over the 2005–2019 period. The parameters include the lag (in trading days), degree of smoothing, and proportion of the benchmark invested in a local bond index. The fifth column reports the benchmarks’ corresponding gross asset-to-net asset leverage ratio. (C) Reports the geographic mix of the benchmark portfolios. 82 A. D. BEATH ET AL. directly. Moreover, seven of the eight large Canadian In Table 2C, we report the inferred geographic mix funds invest more than 50% of their real estate port- for the real estate portfolios of Canadian, U.S., folio directly, while only four of 28 large RoW funds European, and UK pension funds. The inferred mix do so. This significant difference is again consistent takes into account the geographic mix of the underly- with Beath et al. (2021) who find that Canadian pen- ing REIT indices. According to our estimation, the sion funds rank first globally in terms of the propor- averageexposureto local markets (1-g) was 74% for tion of total assets managed in-house. funds from Europe and UK and 79% for U.S. funds Together, these statistics indicate that Canadian from 2005 to 2019. These values are again close to pension funds are significantly more active than the results from the 2020 CEM Global Leaders Survey. their peers in the direct real estate market. For each According to the survey, the average exposure of group of funds, we estimate the proportion of real European and UK funds to local markets was 70% estate that is directly managed as the product of (i) and that of US funds was 82% at the end of 2019. the average proportion of AUM that is invested in For Canadian pension funds, the average expos- real estate and (ii) the average proportion of real ure to the local market (79%) exceeds the average estate that is invested directly. This proportion is local exposure reported in the CEM Global Leaders 0:14  0:81 ¼ 12% for large Canadian pension funds survey (57% at the end of 2019). The survey esti- but only 1.2% for large RoW funds. One Canadian mate is lower for two reasons. First, funds in the pension fund is therefore equivalent to ten RoW survey are larger than the typical Canadian fund funds of the same size in this market. and invest a greater proportion of their assets Table 2B reports the average inferred values of abroad than smaller funds. Second, the foreign real the l, s, b, and g parameters of the benchmark estate exposure of Canadian pension funds has portfolios. The values are close to the inferred increased since 2005. parameters of the NPI benchmark and are stable across groups of funds, which confirms that the Performance of Canadian Pension Funds benchmark methodology works well. For example, We now compare the net value added of Canadian the average smoothing parameter is s ¼ 0.62 for and RoW funds. For each subset of these funds, we Canadian funds and s ¼ 0.63 for RoW funds. report the average, median, and standard deviation We note that heterogeneity in the bond param- of the NVA across funds in Table 3A. eter b does not necessarily indicate heterogeneity in We generally observe that externally-managed the gross asset-to-net asset leverage ratio of the real estate portfolios perform worse than internally- real estate portfolios. The reason for the gap is that managed portfolios, consistent with the findings of the underlying REIT indices have varying leverage Andonov et al. (2015). This pattern is visible both ratios. For example, the average Canadian fund has inside and outside Canada. For example, RoW funds, a lower bond weight than the average RoW fund which mostly include funds that manage the major- (b ¼ 0.22 vs. 0.13) but a slightly higher leverage ity of their assets externally, have an average NVA ratio (1.3 vs. 1.24) due to the high leverage of equal to negative 215 bps. A likely explanation is Canadian REITs. The low leverage ratio (1.24) of the that these funds incur high fees due to the high seven large Canadian funds with high direct real share of external management. This finding is con- estate management can be explained by the fact sistent with the evidence that external investments that a subset of them use leverage at the total fund in private markets tend to underperform (Andonov level. The overall leverage is used in part to fund real estate investments at a low cost. et al., 2021; Beath & Flynn, 2020; Gupta & Van We verify that the leverage ratios inferred from Nieuwerburgh, 2021; Phalippou & Gottschalg, 2009). our estimation are consistent with the 2020 CEM Quite strikingly, large Canadian pension funds that Global Leaders survey of 26 of the largest institu- invest the majority of their assets directly have the tional investors in the CEM database. As of 2019, the highest NVA among all groups of funds in our sam- surveyed leverage ratios ranged from 1.07 to 1.79 ple. Their average NVA of 148bps is statistically sig- with an average value of 1.47. This is in line with the nificant and economically large. Their median NVA of range of leverage ratios reported in Panel B. 295bps is also large, confirming that the large JOURNAL OF SUSTAINABLE REAL ESTATE 83 Table 3. Net value added and expected return. Count Average Median Standard deviation Standard error (A) Net value added (basis points) Canadian funds All 55 35 79 361 49 Large 8 122 149 248 88 Direct (50%þ) 14 138 176 239 64 Large and direct (50%þ) 7 148 295 255 96 R.o.W funds All 186 215 227 427 31 Large 28 68 35 324 61 Direct (50%þ)13 45 83 693 192 Large and direct (50%þ)4 167 139 287 143 (B) Average return Canadian funds All 55 0.085 0.091 0.036 0.005 Large 8 0.095 0.094 0.020 0.007 Direct (50%þ) 14 0.096 0.097 0.020 0.005 Large and direct (50%þ) 7 0.097 0.102 0.021 0.008 R.o.W funds All 186 0.043 0.045 0.041 0.003 Large 28 0.058 0.060 0.029 0.006 Direct (50%þ) 13 0.069 0.053 0.067 0.018 Large and direct (50%þ) 4 0.053 0.060 0.019 0.009 This table displays statistics of the net value added (NVA) for Canadian and Rest-of-World (RoW) funds. The sample includes 241 funds in the CEM database. For each group of funds, we report in (A) the cross-sectional average, median, standard deviation, and standard error of the funds’ average NVA in basis points between 2005 and 2019, and in (B) the cross-sectional average, median, standard deviation, and standard error of the real estate portfolio’s average return net of costs. For each fund, the average return is calculated as the sum of the benchmark portfolio average return and the fund’s average NVA. average is not driven by outliers. These results indi- 100bps lower than that of comparable Canadian funds. cate that the real estate strategy pursued by these Moreover, for the small sample of RoW funds that are funds delivers superior performance net of fees. both large and invest a high proportion of assets dir- The strong performance of Canadian pension ectly, the average NVA was negative 167bps. These sta- funds is not restricted to the seven largest funds tistics confirm that the investment strategy used by managing more than $50 bn of real estate invest- large Canadian pension funds is distinct. ments. The average NVA of the 14 Canadian funds Average Return that directly manage a majority of their real estate In Table 3B, we report the average return of real investments was 138 bps over the sample period. estate portfolios across the different groups of funds These funds make up 25.5% of Canadian funds and from 2005 to 2019. For each fund, the average together fully explain the average NVA of Canadian annual return is estimated in local currency as the funds (0.255 138 ¼ 35 bps). sum of (i) the average return of the corresponding We emphasize that this strong performance is not driven by the funds’ exposure to the local real benchmark portfolio over the full period and (ii) the fund’s average NVA. This calculation allows us to estate market. As explained in the previous section, make use of the full sample period for each fund our benchmarking method controls for the param- eter g that governs the geographic mix of the and thus avoid a situation in which we compare funds’ average returns over short and different sam- benchmark. Moreover, we find that the other Canadian funds that manage fewer than $50 billion ple periods. There is a high level of heterogeneity in the and a minority of their real estate investments dir- funds’ return performance. Part of the performance ectly have an average NVA that is close to zero. We also verify that the value added to Canadian gap is driven by the performance of local real estate pension funds’ real estate portfolios is not just the result markets. For example, between 2005 and 2019 Canadian pension funds generated an average of managing assets directly. Using RoW funds as a con- trol group, we find that RoW funds that invest the return of 8.5% while RoW funds generated an aver- majority of their real estate portfolio directly outper- age return of 4.3%. This return differential is partly formed their benchmark by 45bps on average between due to the booming real estate sector in Canada, 2005 and 2019. While this NVA value is positive, it is both in the residential and commercial spaces. 84 A. D. BEATH ET AL. The other driver of the heterogeneity in return December 31, 2018, according to WillisTower performance is the funds’ NVA. Consider for Watson. We include in this list the wholly-owned example the heterogeneity in the NVA within the real estate operating subsidiaries of these funds. group of Canadian pension funds, which all invest Table 4A reports statistics of these large pension heavily in Canada. The average return of the seven funds. Six of the 88 pension funds in the list are Canadian. Approximately 50% of RoW funds are U.S. large Canadian pension funds with a high share of funds. The six Canadian pension funds manage about direct investments was 9.7% from 2005 to 2019. 6% of total pension fund assets. Their median fund This is 1.2% greater than the average return of all size ($93.35bn) exceeds that of RoW funds ($81.19bn). Canadian funds. Such outperformance is significant We also report the subset of funds that have considering that the return time-series for the seven made direct real estate deals as reported by Preqin. large funds have a lower implied leverage ratio than All six Canadian funds do so but fewer than half of that of the average Canadian fund. the RoW funds (36) do. Unsurprisingly, funds that Overall, our performance analysis confirms that directly invest in real estate tend to be larger. The the real estate strategy implemented by large median size of RoW funds is $93.33bn, which is Canadian pension funds has created significant similar to the median size of the large Canadian value in 2005–2019 net of fees. pension funds. Table 4B presents statistics on the real estate Green Urban Development transactions made by these funds. There are 139 We now take the analysis down to the asset level buy transactions made by Canadian funds and 146 and investigate the real estate strategy imple- buy transactions made by RoW funds between 2010 mented by large Canadian pension funds. The next and 2018. A typical transaction value ranges section reviews data on real estate transactions between $100 and 110 mn. In total, these transac- made by these funds. Then we identify green urban tions amount to 2.4% of all real estate transactions development as a key component of their strategy. in the sample of Preqin properties. If we also incorp- The section on Green Urban Development in Large orate transactions where sellers include one of these Canadian Cities quantifies and illustrates the size of pension funds, then the total value of deals that dir- green urban development in large Canadian cities. ectly involve a large pension fund corresponds to about 4% of the aggregate value of deals reported Finally, in the section on Beyond LEED-VA Development: Greenfield Projects we describe the in Preqin. funds’ involvement in greenfield projects. Table 4B illustrates the extent to which Canadian pension funds are involved in the private real estate market. Between 2010 and 2018, Canadian pension Data funds directly invested $32.24bn in real estate prop- We use data from Preqin, a data provider that col- erties, while RoW funds directly invested $22.38bn. In lects detailed information on thousands of real total, Canadian pension funds were, therefore, estate transactions around the world. The informa- responsible for 60% of the total value of real estate tion includes deal size, date, location, sector, asset acquisitions involving a pension fund. This proportion type, and buyer and seller identification. The dataset is 10 times greater than the share of total pension goes from 1988 to 2018; however, because 99.9% of fund assets they manage (6% as reported above). reported transactions at the property level took Sell transactions reveal a similar pattern. Canadian place after 2010, we primarily study the 2010–2018 pension funds directly sold a total of $20.21bn of period. In total, real estate transactions reported in real estate properties between 2010 and 2018, Preqin during this period amounted to USD 2.44 tril- whereas RoW pension funds directly sold $13.8bn. lion, 61% of which consisted of U.S. properties. In addition to being more active than their peers To analyze the deals that directly involve a large in the private real estate market, Canadian pension pension fund, we restrict the sample to deals where funds also invest more abroad. We count the total buyers include a fund that manages assets for a number of foreign countries where pension funds pension plan that is worth more than $50bn as of make direct real estate deals that are reported in JOURNAL OF SUSTAINABLE REAL ESTATE 85 Table 4. Summary statistics of real estate deals that directly and CAGBC websites. The LEED certification brings involve pension funds. several advantages to our analysis. It is global, Canadian funds RoW funds widely used in Canada and North America, and cov- (A) Pension funds ers a wide range of properties, including office, resi- All funds Number of funds 6 82 dential, mixed-used, and industrial. Moreover, the Median AUM (USD Mn) 93,352 81,192 LEED certification focuses on environmental initia- % Of total AUM 0.06 0.94 Funds that directly invest in RE tives that add value to an asset, as confirmed by Number of funds 6 36 Median AUM (USD Mn) 93,352 93,329 Eichholtz et al. (2010), Fuerst and McAllister (2010), (B) Real estate deals and Miller et al. (2008). According to the U.S. Green Buy deals Number of deals 139 146 Building Council (USGBC) website, LEED certification Median size (USD Mn) 111 105 provides: (i) instant recognition for the building; (ii) Total (USD Mn) 33,236 22,380 Sell deals faster lease-up rates; (iii) higher resale value; (iv) Number of deals 102 127 healthier indoor space; (v) lower use of energy, water, Median size (USD Mn) 166 83 Total (USD Mn) 20,217 13,796 and other resources; and (vi) brand enhancement. Geography The capital investment that is required to obtain a Number of foreign countries 10 2 Sector (%) LEED certification usually ranges from 2.5 to 9.4% of Residential 0.19 0.09 Office 0.41 0.46 the property’s value depending on the certification Retail 0.03 0.18 level (Kats, 2003; Matthiessen and Morris, 2004; Mixed use 0.31 0.13 Other 0.05 0.15 Nyikos et al., 2012), which is significant. This table displays summary statistics of direct real estate deals made by Another advantage of working with the LEED cer- Canadian and Rest-of-World (RoW) funds that manage assets for a pen- tification is that the information made available by sion plan with more than USD 50 bn of AUM in 2018. (A) Presents sum- mary statistics about (i) all funds, and (ii) funds that directly invest in real USGBC and CAGBC includes the street address, level estate. Statistics include the total number of funds, the median AUM, and the proportion of aggregate AUM managed by each group of funds. (B) of certification achieved (Platinum, Gold, Silver, Presents statistics about the direct real estate deals made by these funds Certified), and the dates at which the certification between 2010 and 2018, where the data are obtained from Preqin. The first two sets of rows correspond to buy and sell transactions, respect- was requested and approved for every certified ively. Deal size is reported at the asset level. The third and fourth sets of building. We can therefore track the greenness of rows report the geography and sector distribution of buy deals. The num- ber of foreign countries corresponds to the average number of foreign every building in a pension fund’s real estate port- countries where deals were made by all funds of a particular country. folio and tease apart whether a pension fund pur- chases buildings that are already LEED-certified or Preqin. There are 10 foreign countries where deals actively seeks to obtain the LEED certification after are made by Canadian pension funds. By contrast, the acquisition. there are on average 2.2 foreign countries where To this end, we introduce the following termin- deals are directly made by funds from an ology. A “LEED value-added (VA)” investment corre- RoW country. sponds to a real estate investment where the fund In terms of industry sectors, Canadian pension actively seeks some form of LEED certification after funds invest in a balanced mix of assets. About 41% having acquired the property and before selling it. of their real estate portfolio is invested in office This action could either be the registration of a new space, 31% is invested in mixed-use properties, and project, the certification of an existing project, or 19% is invested in residential properties. In compari- both. A “premium” investment is one where the son, RoW funds invest the majority of their portfolio fund purchases an asset already LEED-certified and in office properties (46%). does not make LEED upgrades to it. A “basic” investment is one that does not have any LEED cer- tification throughout the sample. LEED Value-Added Activity In the first set of rows of Table 5A, we report the The greenness of a real estate investment can be portfolio allocation in basic, premium, and LEED measured according to a variety of certification pro- value-added properties. Canadian funds invest grams, including LEED, BREEAM, Energy Star, and 27% of their assets in LEED-VA strategies. In com- others. In this study, we use data from the LEED cer- parison, RoW funds invest only 14% of their assets tification program that is available on the USGBC in LEED-VA strategies. The majority of assets held by 86 A. D. BEATH ET AL. Table 5. Comparison of green value-added activity across funds and then sold again between 2010 and 2018. pension funds. The table confirms the dominance of Canadian pen- Canadian funds RoW funds sion funds in this market, which are responsible for (A) Buy deals 24 of the 39 transactions. All properties Basic 0.60 0.74 Table 5B further shows that Canadian pension Premium 0.13 0.13 funds are more likely to directly purchase proper- LEED-VA 0.27 0.14 US office properties ties, increase their value through LEED-certification, Basic 0.23 0.46 Premium 0.29 0.34 and then re-sell the properties afterward. Buy-and- LEED-VA 0.48 0.20 sell transactions indeed represent a greater fraction (B) Buy-and-sell deals Buy-and-sell deals of total transaction volume for Canadian funds Number of deals 24 15 (17%) than for RoW funds (5%), which indicates Total (USD Mn) 5,745.44 1,069.64 % Of buy deals 0.17 0.05 greater turnover. Moreover, these transactions are All properties mostly LEED-VA transactions for the Canadian funds Basic 0.36 0.96 Premium 0.03 0.04 (61%) whereas virtually all transactions made by LEED-VA 0.61 0.00 RoW funds are either basic or premium. (C) Geography breakdown for Canadian funds In the final part of our analysis based on Preqin Investment type data, we focus on the geographical breakdown of Basic Premium LEED-VA basic, premium, and LEED-VA assets for all buy deals Geography (%) made by Canadian funds. Table 5C shows that Canada 0.59 0.00 0.41 U.S. 0.48 0.17 0.35 LEED-VA assets amount to 41% of their investments Other 1.00 0.00 0.00 in Canadian properties, 35% of their investments in This table displays the degree of LEED value-added activity by Canadian and Rest-of-World (RoW) funds between 2010 and 2018. (A) Focuses on U.S. properties, and 0% of their investments in other buy deals. The first (second) set of rows reports the percentages of all foreign properties. Therefore, Canadian funds are (US office) properties that are basic, premium, and LEED-VA. (B) Focuses on properties that were bought by Canadian and Rest-of-World (RoW) much more likely to conduct the internal green funds and then sold again between 2010 and 2018. The first set of rows value-added activity for real assets close to home. reports the number of these deals, their total value, and the proportion of their total value relative to the total value of buy deals. The second set of rows provides the breakdown of these deals into basic, premium, and LEED-VA investments. (C) Zooms in on the LEED value-added activity by Green Urban Development in Large Canadian funds and provides the breakdown for properties located in Canada, U.S., and other countries. For each geography, we report the per- Canadian Cities centages of all properties that are basic, premium, and LEED-VA. To further investigate the green urban development RoW funds are invested in basic strategies (74%). strategy implemented by Canadian pension funds in The evidence thus reveals that Canadian pension large Canadian cities where the funds are most funds are significantly more involved in green urban involved, we augment the dataset by manually col- development than RoW funds as measured by lecting hundreds of real estate transactions from LEED activity. the funds’ annual reports and press releases for the Once we account for the fact that LEED certifica- nine largest Canadian pension funds. These funds tion is most common in U.S. office space, the gap consist of Alberta Investment Management between Canadian funds and RoW funds’ involve- Corporation (AIMCo), British Columbia Investment ment in LEED-VA strategies increases. In the second Management Corporation (BCI), La Caisse des set of rows of Table 5A, we report the composition D epots et Placements du Quebec (CDPQ), Canada of basic, premium, and LEED-VA assets within the Pension Plan Investment Board (CPPIB), Healthcare portfolio of U.S. office properties. Canadian funds of Ontario Pension Plan (HOOPP), Investment invest approximately half (48%) of their portfolio of Management Corporation of Ontario (IMCO), Ontario direct U.S. office acquisitions into LEED-VA assets, Municipal Employees’ Retirement System (OMERS), whereas RoW funds only invest 20% of their port- Ontario Teachers Pension Plan (OTPP), and Public folio in these assets. Sector Pension Investment (PSP). The information Table 5B provides additional insights into the we collect includes the street address, building use, nature of these LEED-VA assets. We zoom in on the and the date of the development or acquisition of 39 properties that were bought by the pension the property. JOURNAL OF SUSTAINABLE REAL ESTATE 87 Figure 2. Canadian pension fund ownership and green development of Toronto. This figure displays the real estate properties dir- ectly owned and greened by the top nine Canadian pension funds in Toronto’s downtown area. Blue properties are not LEED certi- fied (“basic”). Green properties have obtained LEED certification by the pension funds (“value-added”). The red boundary delineates Toronto’s financial district. The orange boundary delineates Toronto’s South Core district. To capture building size, we obtain mapping data presents a map of the Toronto downtown area on land parcels and buildings. This information is where the pension fund activity is most pro- available online for Toronto, Vancouver, and nounced. The blue and green colors, respectively Montreal—Canada’s three largest cities. We convert indicate basic and LEED-VA investments. The red addresses into geocodes, geocodes into land parcels, boundaries delineate Toronto’s financial district. and then land parcels into buildings. The use of land The map illustrates the magnitude of green parcels as an intermediary step allows us to address urban development by the nine large Canadian cases where the geocode does not directly intersect pension funds. Almost all buildings in Toronto’s with building data. These cases arise for buildings that financial district are directly owned by these only take up a fraction of the land parcel and build- funds. The vast majority of these buildings have ings with varying sections of elevation. Our spatial obtained some form of LEED certification under the matching procedure ensures that all buildings owned funds’ ownership, as evidenced by the dominance by Canadian pension funds are properly captured. of green-colored buildings. Once we move away Using our data on deals and LEED certification, from the city center, the proportion of buildings we construct 3-D maps of the properties developed owned by pension funds rapidly declines. and/or acquired by the top nine Canadian pension Figures 3 and 4 present equivalent 3-D maps for funds in Toronto, Montreal, and Vancouver. Figure 2 Montreal and Vancouver. The dark green color on 88 A. D. BEATH ET AL. Figure 3. Canadian pension fund ownership and green development of Montreal. This figure displays the real estate properties directly owned and greened by the top nine Canadian pension funds in Montreal’s downtown area. Blue properties are not LEED certified (“basic”). Green properties have obtained LEED certification by the pension funds (“value-added”). The dark green property obtained LEED certification before being acquired by pension funds (“premium”). The red boundary delineates Montreal’s financial district. The orange boundary delineates Montreal’s Quartier des Gares. the Montreal map indicates a premium investment. An additional statistic of interest is the proportion Both maps show the same pattern. Canadian pen- of buildings that obtained LEED Platinum, the high- sion funds own and green a large proportion of est level of LEED certification. Only 4% of all properties in the downtown areas, and their activity Canadian projects have obtained Platinum certifica- rapidly declines outside these areas. Altogether, tion according to the CAGBC database. In Toronto, these figures reveal a systematic pattern of LEED- 26 buildings have obtained it. Seventeen of these VA activity. buildings (65%) correspond to LEED-VA projects by The maps likely understate the true level of five of the top nine Canadian pension funds. These LEED-VA activity by Canadian pension funds for sev- results show that Canadian pension funds have eral reasons. It may be that we have not identified played a significant role in bringing Toronto’s prop- the full list of properties owned by the top nine erties to the highest LEED standards. pension funds, as only the largest deals are made publicly available. It may also be that some of the Beyond LEED-VA Development: Greenfield Projects remaining downtown properties are owned by Canadian pension funds outside the top nine. LEED-VA activity on existing buildings only represents Additionally, several of the non-LEED-certified build- a fraction of the Canadian pension funds’ real estate ings in blue correspond to buildings under con- strategy. We also find that these funds undertake struction that are expected to obtain LEED large greenfield projects where they purchase the certification in the future. Finally, the building maps land and develop the green properties themselves. for Montreal and Vancouver are from 2016 and The maps in Figures 2–4 show that Canadian pen- 2009, respectively, so any building developed after sionfunds arehighly active indeveloping up-and-com- these dates is missing from the maps. ing districts near the city centers. In Toronto, Canadian JOURNAL OF SUSTAINABLE REAL ESTATE 89 Figure 4. Canadian pension fund ownership and green development of Vancouver. This figure displays the real estate properties directly owned and greened by the top nine Canadian pension funds in Vancouver’s downtown area. Blue properties are not LEED certified (“basic”). Green properties have obtained LEED certification by the pension funds (“value-added”). The red boundary delin- eates Vancouver’s financial district. pension funds developed the majority of towers in We find that the proportion of buildings devel- Toronto’s emerging South Core district near the water- oped by Canadian pension funds is large. Two of the front, as delineated by the orange boundaries on the tallest 10 towers in Toronto that are under construc- map. These towers include the Telus Tower (HOOPP), tion (CIBC Square II and 160 Front St West) are RBC WaterPark place (OMERS), and CIBC Square Canadian pension fund projects. Moreover, the tow- (CDPQ). In Montreal, pension fund activity is highest in ers developed by these funds are more likely to the Quartier des Gares which was established in 2015 obtain Platinum certification. For example, if we focus (again delineated by the orange boundaries). Examples on the 50 tallest towers in Toronto that are either of development projects in this district include the completed or under construction, two of the five Deloitte Tower and Tours des Canadiens (OTPP). towers already developed by Canadian pension funds To quantify local greenfield activity by Canadian have obtained LEED Platinum. In contrast, only one pension funds, we use data from the Council on Tall of the 11 towers developed by non-pension funds Buildings and Urban Habitat (CTBUH) which pro- has obtained it. That tower, Scotia Plaza, obtained vides information on the developers of skyscrapers LEED Platinum in 2020 after the Canadian pension for each major city. We focus on Toronto because fund AIMCo became a major shareholder in 2016. the information available is the most complete. Toronto ranks among the world’s top cities in terms Discussion of the number of skyscrapers taller than 150 m that are completed (#19 with 80 buildings as of July Our findings highlight a possible investment strat- 2022) and under construction (#8 with 27 buildings egy for other large asset owners to follow. In this as of February 2022). Section, we discuss the benefits and risks of this 90 A. D. BEATH ET AL. strategy and conclude with several questions for that the green buildings developed by large future analysis. Canadian pension funds are mostly limited to stand- ard multi-use properties located in prime downtown locations of major cities with diversified industries. Distinct Features of the Green Urban A second risk-mitigating strategy is the concen- Development Strategy tration of projects. Focusing on a set of projects in The green urban development strategy pursued by a particular neighborhood brings multiple advan- Canadian funds has several distinct features. First, it tages. It allows the fund to work in an environment is done mainly in-house. The benefits of internal that it knows well and thus better manages the management are low fees due to the elimination of operational risks (Huffman, 2002). In other words, financial intermediaries. As a result, Canadian funds concentration acts as a risk mitigant from an oper- have the ability to extract the full economic profit ational perspective. Working on local projects also from the real estate ventures. In comparison, most allows the fund to align the risk of its pension liabil- institutional investors outsource the majority of their ities to its assets since local properties correlate private investments to external fund managers and more with the fund’s liabilities than foreign proper- consequently incur large fees that frequently offset ties do. Finally, developing a complete neighbor- the active gains on the portfolio. Additional hood allows the fund to generate housing benefits of pursuing the green urban development externalities (Rossi-Hansberg et al., 2010). internally include full control of the assets, direct In line with this logic, we find that Canadian pen- communication between owners and managers, and sion funds do the majority of green urban develop- a lower cost of project financing since any debt ment in highly localized markets. If we define issuance can be done at the total portfolio level Montreal as the home market of CDPQ and PSP and instead of the project level. Toronto as the home market of CPPIB, OTPP, The green urban development strategy is also dis- OMERS, and HOOPP, we find that, for each fund, tinct in the way it departs from the well-known the home market represents at least 75% of its LEED-VA projects that we have collected across all forms of pension fund activism (Barber, 2020;Del three cities. Guercio & Hawkins, 1998;Wahal, 1996). Pension We further find that, inside a particular city, funds typically aim to engage with large public firms Canadian pension funds tend to concentrate their where they own a sizeable ownership stake. By con- investments in a few locations. A good example is trast, Canadian pension funds share the ownership of real estate projects with few investors and focus on OTPP’s recent investments in Montreal’s Quartier des Gares (Windsor Station, Deloitte Tower, and the project execution, by concentrating their efforts on multiple Tours des Canadiens) which are all located developing and greening urban properties. within a few steps of the central train station. Another example is OMERS’s large investment in Risk-Mitigation Quartier DIX30, a 2.3 million square feet multi-use Developing and managing real assets involves sev- complex in Brossard, which is a municipality part of eral legal and operational risks. There are also the Greater Montreal area that will soon serve as a liquidity risks related to the long-term nature of station for the Reseau express metropolitain development projects, and risks associated with the (REM) network. real estate industry (Ling et al., 2020). A third risk-mitigating strategy is the use of a We discuss several strategies asset owners can wholly-owned real estate operating subsidiary. use to mitigate these risks. One risk-mitigating strat- Examples of subsidiaries in Canada include Ivanhoe egy is the application of green urban development Cambridge (CDPQ), Cadillac Fairview (OTPP), Oxford to a specific set of projects which are well defined Properties (OMERS), QuadReal (BCI), HOOPP Realty (i.e. construction of a downtown tower), can be (HOOPP), and AIMCo Realty (AIMCo). The use of a replicated in multiple cities, and have a risk profile subsidiary provides several advantages. One advan- that naturally matches the long-term bond profile tage is experience. Subsidiaries, such as Cadillac of the funds’ liabilities. Our analysis indeed confirms Fairview, which was purchased by OTPP in 2000, JOURNAL OF SUSTAINABLE REAL ESTATE 91 have acquired decades of experience in developing the growing integration of ESG into corporate busi- and managing hundreds of real estate projects. ness models may encourage demand for green Experience not only helps to identify valuable devel- urban development outside city centers. For opment opportunities and manage operational risks example, condominium units or office buildings that but also to establish long-term relationships with are located either in smaller cities or outside the suppliers and retailers that lease the space. major city centers could become prime targets for Another advantage of using a subsidiary has to green retrofitting moving forward. A notable do with protection from legal risks. Because subsid- example is Downsview Airport in Toronto, which was iaries are limited-liability entities, they do not purchased by PSP in 2018 and is currently being con- engage the funds’ investments outside the subsid- verted into a mixed-use urban community. Finally, can smaller pension funds successfully iary. Additionally, a subsidiary provides flexibility implement green urban development? The strategy regarding employee compensation. Compensation we have uncovered requires scale and the ability to contracts in limited partnership funds tend to be conduct operations in-house. A scaled-down version long, complex, and different from contracts of buy- could potentially be achieved either through co- side institutional funds (Phalippou, 2009). Having a investments or through direct investments in subsidiary can make it easier for pension funds to smaller buildings. Here again, the current ESG wave set up these types of contracts while keeping the could make such a strategy possible for smaller investments in-house. pension funds. We leave these questions for Finally, by fully owning the subsidiary, asset own- future research. ers can mitigate agency conflicts that arise when ownership is fragmented. Typical conflicts include expropriation or inefficient use of funds by manag- Notes ers when investors have limited ability to control the firm’s assets. Having large shareholders can miti- 1. This form of alpha resulting from value-added improvement is also referred to as the gamma of the gate these problems because they have both a gen- expected return of a real estate portfolio (Kaiser, 2005). eral interest in profit maximization and sufficient 2. Because time-series of S&P REIT indices we use only control over the assets of the firm (Shleifer & become consistently available across countries in the Vishny, 1997). early 2000s, we use three return values of the indices to estimate the smoothed return of the benchmark portfolios (i.e. RB,i,t, RB,i,t-1, RB,i,t-2). Their weights are Expanding Green Urban Development? then re-scaled so that they sum up to 100%, and earlier index returns are assigned a weight of zero. Our analysis shows that green urban development 3. We set a lower bound of 0.5 for the smoothing can be a successful investment strategy for asset parameter s, consistent with the estimates of Barkham owners that can carefully manage the risks and gen- and Geltner (1995). If we use values of s below 0.5, the erate economies of scale. The asset owners we have volatility of the smoothed benchmark portfolio studied in this paper are large pension funds with becomes so small that it is no longer possible to match sizable in-house teams. the volatility of the fund’s reported return unless we apply an unreasonably high degree of leverage. Several questions arise from our analysis. Is LEED- 4. For a small proportion of funds the information on VA the only form of green urban development? We 2018 is unavailable. So their AUM value corresponds to have used the LEED certification for the purpose of the latest available year. our analysis, but it may be that other certification 5. According to the annual reports published by large programs provide similar benefits to investors in Canadian pension funds, the proportion of their real terms of value-added gains and environmen- estate assets invested in Canada has generally decreased from 2009 to 2017. For example, Public tal benefits. Sector Pension Investment (PSP Investments)’s share Has the existing market for green urban develop- decreased from 51 to 21%, La Caisse des Depots et ment reached a point of saturation? On the one Placements du Quebec (CDPQ)’s share decreased from hand, the 3-D map of Toronto shows that the major- 43 to 32%, Alberta Investment Management ity of downtown buildings have already been retrofit- Corporation (AIMCo)’s share decreased from 97 to 71%, ted to existing LEED standards. On the other hand, Canada Pension Plan Investment Board (CPPIB)’s share 92 A. D. BEATH ET AL. decreased from 56 to 15%, and Healthcare of Ontario the east, Front St to the south, and University Ave to Pension Plan (HOOPP)’s share decreased from 99 the west. to 69%. 17. For Montreal’s and Vancouver’s financial districts, we 6. The performance analysis in Andonov et al. (2015) uses set approximate boundaries based on publically benchmark returns from the funds’ reported real estate available information. benchmarks. Using our common benchmarking 18. The data is available at www.skyscrapercenter.com. approach, we also find that direct real estate 19. This information is available at https://open.toronto.ca. management generally leads to superior performance. Statistics on the number of buildings come from www. 7. Another possible explanation is that several large U.S. skyscrapercenter.com. funds suffered large losses during the 2008 Great Financial crisis due to highly leveraged positions and Acknowledgments subsequently de-levered their portfolio as part of the Omar Mohsen, Oriane Pacic, and Frederic Sfeir provided out- re-structuring. For example, California Public Employees’ standing research assistance. We thank Jonathan Critchley Retirement System (CalPERS) posted a negative total for help with the 3D-mapping and Chris Flynn, Eric Fontaine, return of 48.7% in the first nine months of 2009 Mario Lefebvre, Luc McSween, Mark Staley, Eduard van according to “CalPERS considers real estate clearout” by Gelderen, and seminar participants at the Bank of Canada, Richard Lowe, IPE Real Assets, December 14, 2009. KPMG, and the 2021 National Pension Hub Insight Series for However, we note that the median NVA of RoW funds insightful comments. The views and opinions expressed in is also significantly negative, which confirms that the this paper are solely those of the authors. They do not negative average NVA of these funds is not driven by necessarily reflect the views or opinions of the Public Sector the actions of a few funds. Pension Investment Board. All errors are our own. 8. WillisTower Watson Thinking Ahead Institute, Pensions & Investments, World 300. In total there are 90 funds that manage more than USD 50 billion. Two of these Funding funds, however, are managed by CDPQ. We exclude This research is funded by the National Pension Hub of the Norway’s Government Pension Fund as it mainly Global Risk Institute. represents the country’s sovereign wealth fund. This leaves a total of 88 funds. 9. In many cases, transactions consist of portfolios involving multiple properties. We approximate the value of each References property as the total value of the deal divided by the number of properties mentioned in the deal. Ambachtsheer, K. (2016). The future of pension fund manage- 10. The information is available at https://new.usgbc.org/ ment. Wiley Finance Editions. leed and at www.cagbc.org. Ambachtsheer, K. (2021). The Canadian pension model: Past, 11. For deals where we only observe the year of acquisition present, and future. The Journal of Portfolio Management, instead of the exact date, we assume that a LEED 47(5), 150–158. https://doi.org/10.3905/jpm.2021.1.216 action that takes place in the same year counts as a Andonov, A., Eichholtz, P., & Kok, N. (2015). Intermediated value-added activity. investment management in private markets: Evidence 12. A “basic” investment should not interpreted as being from pension fund investments in real estate. Journal of not green. Even though the building has no LEED Financial Markets, 22,77–103. https://doi.org/10.1016/j.fin- certification, it may have another green certification. mar.2014.11.002 13. In the Appendix, webreak down thedifferent types of € Andonov, A., Kraussl, R., & Raugh, J. (2021). Institutional LEED value-added activity: LEED registration only, LEED investors and infrastructure investing. The Review of certification only, and the combination of registration and Financial Studies, 34, 3880–3934. https://doi.org/10.1093/ certification. Across all types of funds, the bulk of value- rfs/hhab048 added projects involves the full LEED certification process. Barber, B. (2020). Chapter 15: Pension fund activism: A dou- 14. This information is available at https://open.toronto.ca, ble-edged sword. In O. Mitchell and G. Anderson (Eds.), https://opendata.vancouver.ca and http://donnees.ville. The future of public employee retirement systems. Oxford montreal.qc.ca. University Press. 15. Buildings with varying elevation are categorized as a Barber, B., Morse, A., & Yasuda, A. (2021). Impact investing. collection of sub-buildings. Therefore, without the use Journal of Financial Economics, 139(1), 162–185. https://doi. of land parcels as an intermediary step, we would only org/10.1016/j.jfineco.2020.07.008 capture the part of the building that intersects with the Barkham, R., & Geltner, D. (1995). Price discovery in American geocode data. and British property markets. Real Estate Economics, 23(1), 16. We use the boundaries provided by the Wikipedia page 21–44. https://en.wikipedia.org/wiki/Financial_District,_Toronto. Beath, A., Betermier, S., & Spehner, Q. (2021). The Canadian The boundaries are Queen St W to the north, Yonge St to pension fund model: A quantitative portrait. The Journal of JOURNAL OF SUSTAINABLE REAL ESTATE 93 Portfolio Management, 47(5), 159–177. https://doi.org/10. Matthiessen, L. F., Morris, P. (2004). Costing green: A compre- 3905/jpm.2021.1.226 hensive cost database and budgeting methodology. https:// Beath, A., & Flynn, C. (2020). Benchmarking the performance www.usgbc.org/resources/costing-green-comprehensive- of private equity porfolios of the world’s largest institu- cost-database-and-budgeting-methodology tional investors: A view from CEM benchmarking. The Miller, M., Spivey, J., & Florance, A. (2008). Does green pay Journal of Investing, 30(1), 67–87. https://doi.org/10.3905/ off? Journal of Real Estate Portfolio Management, 14(4), joi.2020.1.155 385–400. https://doi.org/10.1080/10835547.2008.12089822 Bedart-Page, G., Demers, A., Tuer, E., & Tremblay, M. (2016). Nyikos, D., Thal, A., Hicks, M., & Leach, S. (2012). To LEED or Large Canadian public pension funds: A financial system not to LEED: Analysis of cost premiums associated with perspective. In Financial system review (pp. 33–38). Bank of sustainable facility design. Engineering Management Canada. Journal, 24(4), 50–62. https://doi.org/10.1080/10429247. Berk, J., & van Binsbergen, J. H. (2022). The impact of impact 2012.11431955 investing. Working paper. Phalippou, L. (2009). Beware of venturing into private equity. Betermier, S., Calvet, L., & Jo, E. (2022). A supply and demand Journal of Economic Perspectives, 23(1), 147–166. https:// approach to capital markets. Working paper. doi.org/10.1257/jep.23.1.147 Brown, G. R., & Matysiak, G. A. (2000). Sticky valuations, Phalippou, L., & Gottschalg, O. (2009). The performance of aggregation effects, and property indices. Journal of Real private equity funds. Review of Financial Studies, 22(4), Estate Finance and Economics, 20(1), 49–66. 1747–1776. https://doi.org/10.1093/rfs/hhn014 Del Guercio, D., & Hawkins, J. (1998). The motivation and Rossi-Hansberg, E., Sarte, P., & Owens, R. (2010). Housing impact of pension fund activism. Journal of Financial externalities. Journal of Political Economy, 118(3), 485–535. https://doi.org/10.1086/653138 Economics, 52, 293–340. https://doi.org/10.1016/S0304- Shleifer, A., & Vishny, R. (1997). A survey of corporate govern- 405X(99)00011-2 ance. The Journal of Finance, 52(2), 737–783. https://doi. Eichholtz, P., Kok, N., & Quigley, J. (2010). Doing well by org/10.1111/j.1540-6261.1997.tb04820.x doing good? Green office buildings. American Economic Stambaugh, R. F., & Taylor, L. A. (2020). Sustainable investing Review, 100(5), 2492–2509. https://doi.org/10.1257/aer.100. in equilibrium. Journal of Financial Economics, 142(2), 5.2492 550–571. https://doi.org/10.1016/j.jfineco.2020.12.011 Fuerst, F., & McAllister, P. (2010). Green noise or green value? Wahal, S. (1996). Pension fund activism and firm perform- Measuring the effects of environmental certification on ance. The Journal of Financial and Quantitative Analysis, office values. Real Estate Economics, 39(1), 45–69. https:// 31(1), 1–23. https://doi.org/10.2307/2331384 doi.org/10.1111/j.1540-6229.2010.00286.x Zerbib, O. (2022). A sustainable capital asset pricing model Geltner, D. (1989). Estimating real estate’s systematic risk (S-CAPM): Evidence from green investing and sin stock from aggregate level appraisal-based returns. Real Estate exclusion. Review of Finance. https://doi.org/10.1093/rof/ Economics, 17(4), 463–481. https://doi.org/10.1111/1540- rfac045 6229.00504 Gyourko, J., & Keim, D. B. (1992). What does the stock market tell us about real estate returns? Journal of the American Appendix A Real Estate and Urban Economics Association, 20(3), 457–485. This Appendix presents additional empirical results and pro- Gupta, A., & Van Nieuwerburgh, S. (2021). Valuing private vides more details about the benchmark estimation method. equity investments strip by strip. The Journal of Finance, 76(6), 3255–3307. https://doi.org/10.1111/jofi.13073 Estimation of Benchmarking Method Huffman, F. (2002). Corporate real estate risk management In the estimation of our benchmarking methodology, the and assessment. Journal of Corporate Real Estate, 5(1), weight in the bond portfolio b is set such that the fund’s 31–41. https://doi.org/10.1108/14630010310811984 reported return R has a beta of one with respect to the i, t Kaiser, R. W. (2005). Analyzing real estate portfolio returns. benchmark return R for a given combination of B, i, tl=D The Journal of Portfolio Management, 31(5), 134–142. parameters g, s, and l: https://doi.org/10.3905/jpm.2005.593896 We define b as the univariate beta of the fund’s real Kats, G. H. (2003). Green building costs and financial benefits. estate portfolio with respect to the benchmark. By definition, Massachusetts Technology Cooperative. b satisfies Ling, D., Wang, C., & Zhou, T. (2020). A first look at the impact of COVID-19 on commercial real estate prices: covðR , R Þ i, t B, i, tl=D b ¼ Asset-level evidence. The Review of Asset Pricing Studies, varðR Þ B, i, tl=D 10, 669–704. https://doi.org/10.1093/rapstu/raaa014 s s N ER , R ER Þ E ðR i, t i, t Lipshitz, C., & Walter, I. (2020). Public pension reform and the B, i, tl=D B, i, tl=D 49th parallel: Lessons from Canada for the U.S. Financial 2 2 s s N ER ER B, i, tl=D B, i, tl=D Markets, Institutions & Instruments, 29(4), 121–162. https:// doi.org/10.1111/fmii.12133 94 A. D. BEATH ET AL. where N corresponds to the number of annual observations Table A.1. Green value-added activity by pension funds. ðÞ of the return R and EX is the expectation operator. i, t (A) LEED breakdown of all Preqin deals Solving for b ¼ 1 implies Investment type s s N ER , R  E ðR Þ E ðR Þ i, t i, t Basic Premium LEED-VA B, i, tl=D B, i, tl=D Sector (%) s 2 s 2 ¼ ER ER : Hotel 1.00 0.00 0.00 B, i, tl=D B, i, tl=D Industrial 0.95 0.00 0.05 Land 1.00 0.00 0.00 For a given combination of parameters g, s, and l, solving Mixed use 0.56 0.10 0.34 for the remaining parameter b amounts to solving the quad- Niche 1.00 0.00 0.00 ratic equation above. If no solutions are found, there is no Office 0.51 0.22 0.27 Residential 0.94 0.03 0.03 benchmark such that the fund has a beta of one for the Retail 0.80 0.07 0.13 combination of parameters g, s, and l: When one or more Geography (%) solutions are found, we select the values of the parameters Canada 0.59 0.00 0.41 US 0.52 0.18 0.30 b, g, s, l that maximize the correlation between the fund Other 0.95 0.02 0.03 return R and the benchmark return R : i, t B, i, tl=D (B) Geography distribution of all LEED projects Canada US Other LEED Activity around the World Weighting of projects (%) Number 0.02 0.81 0.17 In the main text, we use LEED certification activity as a proxy Square-footage 0.01 0.66 0.33 for green value-added investments. We now show that LEED This table reports aggregate statistics about LEED green building certifica- certification varies by sector and geography in our sample of tion. (A) Shows the LEED breakdown of all deals in Preqin made by large pension funds between 2010 and 2018. For each industry sector and Preqin deals involving a pension fund. geography, we report the percentages of properties that are not LEED Table A.1A reports statistics about LEED green building certified (“basic”), properties that were already LEED-certified prior to the certification on Preqin deals made by all large pension funds acquisition by a pension fund and did not receive additional certification afterwards (“premium”), and properties where pension funds actively between 2010 and 2018. For each industry sector and geog- sought some form of LEED certification after having acquired the property raphy, we report the percentages of properties that are not and before disposing it (“LEED-VA”). (B) Reports the geography distribu- LEED certified (“basic”), properties that were already LEED- tion of all projects in the USGBC database. The distribution is weighted either by the number of projects of their square footage. certified prior to the acquisition by pension funds and did not receive additional certification afterwards (“premium”), and properties where pension funds actively sought some Table A.2. Types of value-added activity. form of LEED certification after having acquired the property Canadian funds RoW funds and before disposing it (“LEED-VA”). We see that mixed-use Type of LEED-VA (%) and office-space properties in Canada and the U.S. are the Registration only 0.17 0.08 Certification only 0.41 0.58 most likely to qualify as premium or LEED-VA. Reg and Cert 0.42 0.35 Table A.1B reports the geography distribution of all LEED This table reports the different types of LEED value-added activity by projects reported in the USGBC database. The distribution is Canadian and Rest-of-World (RoW) funds between 2010 and 2018. Types weighted either by the number of projects or their square include LEED registration only, LEED certification only, and LEED registra- tion and certification. footage. We see that LEED projects are predominantly based in the U.S. certification. For both Canadian and RoW funds, the bulk of In Table A.2, we break down the different types of LEED value-added projects goes beyond registration and generally value-added activity: LEED registration only, LEED certifica- involves the full LEED certification process. tion only, and the combination of registration and http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Sustainable Real Estate Taylor & Francis

Green Urban Development: The Impact Investment Strategy of Canadian Pension Funds

Green Urban Development: The Impact Investment Strategy of Canadian Pension Funds

Abstract

Abstract This paper investigates the investment strategy that large Canadian pension funds implement in the private real estate market. Even though they manage just 6% of global pension assets in our data, Canadian pension funds are responsible for 60% of the total value of private real estate deals that directly involve a pension fund. A key component of their strategy consists of internally developing and greening urban properties in core downtown areas. Using a common benchmarking...
Loading next page...
 
/lp/taylor-francis/green-urban-development-the-impact-investment-strategy-of-canadian-9BOdwKT74T
Publisher
Taylor & Francis
Copyright
© 2022 The Author(s). Published with license by Taylor & Francis Group, LLC
ISSN
1949-8284
DOI
10.1080/19498276.2022.2125203
Publisher site
See Article on Publisher Site

Abstract

JOURNAL OF SUSTAINABLE REAL ESTATE 2022, VOL. 14, NO. 1, 75–94 ARES https://doi.org/10.1080/19498276.2022.2125203 American Real Estate Society Green Urban Development: The Impact Investment Strategy of Canadian Pension Funds a b a c a Alexander D. Beath , Sebastien Betermier , Maaike Van Bragt , Yuedan Liu and Quentin Spehner a b c CEM Benchmarking, Toronto, Canada; Desautels Faculty of Management, McGill University, Montreal, Canada; Public Sector Pension Investment Board, Montreal, Canada KEYWORDS ABSTRACT Active management; impact This paper investigates the investment strategy that large Canadian pension funds imple- investing; pension funds; ment in the private real estate market. Even though they manage just 6% of global pension real estate; sustainable assets in our data, Canadian pension funds are responsible for 60% of the total value of pri- finance vate real estate deals that directly involve a pension fund. A key component of their strat- egy consists of internally developing and greening urban properties in core downtown areas. Using a common benchmarking methodology across funds, we show that this impact strategy delivers superior performance net of fees and drives the green development of major city centers. Introduction funds. In light of this evidence, we ask three ques- tions. Have large Canadian pension funds been able The growth of interest in sustainable investing over to generate high-risk-adjusted returns in private real the past decade raises hard questions for asset own- estate? Have they been able to combine financial ers. Will the cost of investing green result in lower profitability with positive environmental impact? If portfolio returns in the long-run? Or are there win- win strategies where green investments can gener- so, can asset owners around the World successfully ate high returns alongside positive environmental implement this strategy moving forward? impact? These questions are particularly relevant in Our analysis consists of three parts. In the first the real estate space since green real estate devel- part, we evaluate the historical performance of opment has been shown to generate significant Canadian pension funds’ real estate portfolios. Using economic value (Eichholtz et al., 2010; Fuerst & data from CEM Benchmarking, a global benchmark- McAllister, 2010; Miller et al., 2008). ing firm, we compare the annual returns of 241 real The present paper investigates the investment estate portfolios of Canadian and non-Canadian strategy that large Canadian pension funds imple- pension funds from 2005 to 2019. The sample ment in the private real estate market. These funds includes 36 funds (eight Canadian) with more than are interesting to study because they actively invest USD 50 billion of AUM. We henceforth refer to these in green buildings, such as the Deloitte Tower in funds as large funds. Importantly, the dataset Montreal and the Texas Tower in Houston, and they includes for each fund the proportion of the real allocate an unusually large share of their portfolio to estate portfolio that is private, as well as the pro- real assets. A recent study by Beath et al. (2021)finds portion that is directly managed, either through dir- that, among pension funds with over USD 50bn ect property ownership or through a wholly owned assets under management (AUM), real assets account operating subsidiary. There are 27 funds (14 for 23% of the portfolios of Canadian funds on aver- Canadian) that manage more than 50% of their real age but just 10% of the portfolios of non-Canadian CONTACT Sebastien Betermier sebastien.betermier@mcgill.ca Desautels Faculty of Management, McGill University, 1001 Sherbrooke Street West, Montreal, QC H3A 1G5, Canada 2022 The Author(s). Published with license by Taylor & Francis Group, LLC This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 76 A. D. BEATH ET AL. estate AUM directly. Among them are seven of the investments. Even though they manage 6% of glo- eight large Canadian pension funds. bal pension assets in our data, they are responsible Our performance evaluation methodology specif- for 60% of the total value of real estate deals that ically addresses the lack of consistent benchmarking directly involve a pension fund. across real estate portfolios. Many funds adjust their We then combine the information on deals from benchmark by adding an inflation index, a private Preqin with data on green building certification of risk premium, or a lag. Moreover, while some funds newly-built and retrofitted projects by Leadership in use listed REIT indices as benchmarks, such as the Energy and Environmental Design (LEED), a widely- FTSE Nareit U.S. Real Estate Index, other funds rely used green certification program. We show that on peer-based benchmarks, such as the National large Canadian pension funds systematically pursue Property Index (NPI). The use of different types of an impact strategy that consists of internally devel- benchmarks is problematic because the returns of oping, managing, and greening urban properties. REIT benchmarks are marked-to-market whereas the Specifically, they initiate and complete LEED proj- returns of peer-based benchmarks are based on ects on 27% of their portfolio of direct acquisitions appraisal prices, which are often smoothed and reported in Preqin, whereas non-Canadian funds lagged (Geltner, 1989). The lack of consistency only do so for 14% of their portfolio. The gap across benchmarks makes net value-added compari- increases once we focus on direct acquisitions of sons across funds open to interpretation. U.S. office properties where LEED certification is We address this issue by employing a common most common. Furthermore, Canadian pension benchmarking approach. For each fund, we con- funds are more likely to purchase real estate assets, struct a benchmark that consists of a fund-specific activate and complete the LEED certification pro- geographic blend of listed equity REITs, appropri- cess, and then re-sell the assets. ately (de-) levered, smoothed, and lagged to match To further investigate the green urban develop- the properties of the real estate portfolio’s reported ment strategy in large Canadian cities where returns. Our method delivers benchmarks that Canadian pension funds are most involved, we aug- closely track the reported portfolio returns, with an ment the dataset by manually collecting hundreds average correlation of 93%. Using this consistent of local deals made by the top nine Canadian pen- benchmarking methodology, we measure the per- sion funds from their annual reports and news formance of a real estate portfolio in terms of its releases. We then combine this information with net return in excess of the benchmark. We refer to data on land parcels and buildings and construct 3- this net return as net value added (NVA). D maps of the funds’ LEED value-added activity in Our analysis reveals that portfolios of large Toronto, Montreal, and Vancouver. In each city, we Canadian pension funds with a high proportion of show that the city center has been largely greened direct real estate investments outperform their by Canadian pension funds. The patterns are most benchmarks. On average, large Canadian pension striking in Toronto, which is Canada’s largest city funds that directly manage more than 50% of their and the home to the head offices of five of the real estate AUM generated an NVA of 148 basis nine largest Canadian pension funds. The area points (bps) between 2005 and 2019. We verify that located inside and surrounding Toronto’s financial this strong performance is not driven by the funds’ district is owned and greened predominantly by exposure to the local real estate market, and it is these funds. Moreover, 17 of the 26 Toronto build- not just the result of managing assets directly. ings that obtained LEED Platinum—the highest level In the second part of the analysis, we look into of certification—correspond to Canadian pension the real estate investment strategy pursued by large fund projects. Canadian pension funds. We first use deal-level data We find that Canadian pension funds do more from Preqin and compare hundreds of real estate than obtain LEED certification on existing buildings. deals made by large Canadian and non-Canadian They also undertake large greenfield projects where pension funds from 2010 to 2018. The data confirm they purchase the land, develop and green the that large Canadian pension funds are uniquely properties from the ground up. These projects tend active in the market of direct real estate to be clustered in up-and-coming districts near the JOURNAL OF SUSTAINABLE REAL ESTATE 77 city centers, such as the new Quartier des Gares in which provides multiple operational advantages Montreal and the South Core district in Toronto. including experience in developing and managing The combined impact of these projects on the cit- green projects, access to a large network of suppliers ies’ development is large. In Toronto, two of the 10 and retailers that lease the space, protection from tallest office towers currently under construction are legal risks, additional flexibility regarding employee being developed by Canadian pension funds. compensation, and the reduction of agency issues. In the third part of the analysis, we discuss the First, we describe the paper’s contribution to dif- benefits and risks of the Canadian pension funds’ ferent strands of the research literature. Next, we real estate investment strategy. We view green assess the performance of large Canadian pension urban development as an impact strategy because funds’ real estate portfolios. Then, we identify green (i) it combines financial profitability alongside posi- urban development as a key driver of the Canadian tive environmental impact and (ii) it is a strategy funds’ strategy. Finally, we discuss the benefits and where the parent investors—Canadian pension risks of this strategy for asset owners and conclude. funds—have a direct influence on the projects’ exe- cution. The benefits of pursuing this strategy intern- Contribution to the Literature ally include full control of the assets, direct Our paper contributes to several research literature communication between owners and managers, low on sustainable investing, impact investing, private fees due to the elimination of financial intermedia- investing, and the Canadian pension fund model. In ries, lower cost of project financing, and the ability this section, we briefly review these literatures and to extract the full economic profit from the real summarize the paper’s main contributions. estate ventures. Of course, there are legal and operational risks associated with the development and management Sustainable Investing of real assets. There are also liquidity risks related to A growing literature in finance shows that sustain- the long-term nature of development projects, and able investing leads to low expected returns. For risks associated with the real estate industry as evi- example, Pastor et al. (2020) show that green assets denced by the covid-19 pandemic (Ling et al., have low expected returns in equilibrium because 2020). We discuss in the paper several strategies investors enjoy holding them and because green that asset owners can take to mitigate these risks. assets hedge climate risk (see also Zerbib, 2022). One risk-mitigating strategy is the application of Consistent with this analysis, Betermier et al. (2022) green urban development to a specific set of proj- develop a supply and demand approach to capital ects that have well-defined clienteles, economies of markets and confirm that any ESG factor that scale, and a risk profile that naturally matches the expands the supply of capital to a productive asset long-term bond profile of the funds’ pension liabil- naturally results in a lower expected return for the ities. A good example is the construction of a green asset. The size of these effects remains a source of office tower in a major city center. Such a tower debate (Berk & van Binsbergen, 2022). caters to the need of corporate clientele for sustain- In this paper, we show that, in spite of the low able office space. It can also be replicated in multiple risk-return trade-off of sustainable investments, it is cities and provides a stream of rental income that still possible for investors to create a win-win for shares similarities with a stream of pension payouts. high returns on green investments by pursuing an A second risk-mitigating strategy is the concentra- internal development strategy, such as green urban tion of development projects in particular cities and development. The high return comes from the neighborhoods. Concentration allows a fund to work “alpha” side of the investment. By bringing the in an environment that it knows well and thus better development function in-house, Canadian pension manages the operational risks (Huffman, 2002). Concentration also allows the fund to create positive funds capture the developer’s economic profit asso- housing externalities (Rossi-Hansberg et al., 2010). ciated with the productive investments in green A third risk-mitigating strategy is the use of a assets. This economic profit results in high risk- wholly-owned real estate operating subsidiary, adjusted returns for the investors. 78 A. D. BEATH ET AL. Impact Investing allocating more capital to strategic assets, including real estate and infrastructure. These assets not only Impact investing is a type of sustainable investing diversify the portfolio but also align the risk of their where investors exercise their right of control to liability portfolio to their assets. The study finds that change corporate policy to generate both a financial this business model has resulted in a strong perform- return and additional environmental and social gains. ance for the pension funds both in terms of financial Previous research finds that impact funds tend to performance and liability hedging. Building on this generate lower returns than traditional funds (Barber literature, we take the analysis down to the asset et al., 2021), partly because they cater to a clientele level and identify green urban development as a key of investors who derive nonpecuniary utility from component of the active strategy that Canadian pen- investing and are thus willing to sacrifice returns. By sion funds internally pursue in real estate. comparison, our analysis shows that the green urban development strategy implemented by Canadian pension funds results in positive risk-adjusted returns. Performance of the Canadian Pension Funds’ The high returns are consistent with the evidence Real Estate Portfolios that green real estate development generates posi- We begin the analysis by assessing the performance tive economic profits (Eichholtz et al., 2010;Fuerst& of Canadian pension funds’ real estate portfolios. McAllister, 2010;Miller et al., 2008). The next section introduces a common benchmark- ing method across funds. The section on Data Private Investing describes the dataset used for the performance ana- lysis. Finally, the section on Performance of Our paper also contributes to the growing body of Canadian Pension Funds presents the results. research on the profitability of private investments. Because private assets are less liquid and more complex to operate and value than public assets, it Benchmarking Method and Performance Metric is important to understand whether and how invest- The major impediment to comparing the performance ors can benefit from investing in private markets. of real estate portfolios is a lack of consistent bench- Recent research finds that investments in private marking. We illustrate this issue in Table 1 by summa- equity and infrastructure funds may not be as prof- rizing the properties of the benchmarks used by the itable as commonly believed (Andonov et al., 2021; 241 real estate portfolios in our sample (described in Beath & Flynn, 2020; Gupta & Van Nieuwerburgh, the next Section). While some funds use listed REIT 2021; Phalippou & Gottschalg, 2009). Our analysis indices, fixed hurdle rates, inflation rates, and interest confirms that private real estate portfolios that are rates-based benchmarks, most funds (around 70%) externally managed underperform their benchmarks. use peer-based benchmarks which rely on the However, we show that private real estate portfolios appraised returns of private real estate reported by that are internally managed outperform their bench- asset managers or asset owners (e.g., NPI, IPD Canada, marks, consistent with Andonov et al. (2015). MSCI PREA). Peer-based benchmarks are often custom- ized in a variety of ways through the use of lags, cur- Canadian Pension Fund Model rency adjustments, leverage adjustments, or risk premia. The clear lack of consistency leaves perform- Finally, the paper contributes to the literature on the ance comparisons across funds open to interpretation. Canadian pension fund model. Large Canadian pen- We address this issue by employing a common sion funds are known to be highly involved in direct benchmarking method. For each fund in our sam- investments and in private markets (Ambachtsheer, ple, we construct a benchmark portfolio of local 2016, 2021;Bedart-Page et al., 2016; Lipshitz & and foreign REIT indices that is appropriately (de-) Walter, 2020). Beathetal. (2021) study the business levered, smoothed, and lagged to match the prop- model of these funds and show that the use of in- house teams allows Canadian pension funds to erties of the fund’s reported return. We then calcu- reduce costs and in turn redeploy resources by (i) late the net value added (NVA) of the fund’s growing the funds’ internal capabilities and (ii) portfolio in relation to its inferred benchmark. JOURNAL OF SUSTAINABLE REAL ESTATE 79 Table 1. Properties of real estate benchmark portfolios. benchmark return R , where D is the number B, i, tl=D Range (bps) of trading days during the year. Count Benchmark type Min Max We then solve for the combination of the four 177 Property index parameters ðg, b, s, lÞ that meets two conditions: (i) 9 Premium 120 300 the fund’s reported return R has a maximal correl- 37 Lag i, t 30 Customized blend s ation with the benchmark return R , and (ii) B, i, tl=D 17 Inflation linked 17 Premium 300 550 the fund’s reported return R has a beta of one with i, t 2 Rolling N year respect to R : The details of the optimization B, i, tl=D 7 Public market 1 Stock are provided in Appendix A. The first condition 4 Bond ensures that the benchmark is comparable to the 4 REIT 3 Premium 100 700 fund’s real estate portfolio in terms of its geograph- 3 Lag ical mix, smoothing, and lag properties. The second 40 Other 13 Own benchmark condition ensures that the benchmark has the same 1 Stock volatility as the fund’s real estate portfolio. 6 Bond 7 Property Under these conditions, the difference between 5 Inflation 6 Premium 150 600 the fund’s reported return R and the benchmark i, t 7 Absolute 600 900 return R corresponds to the net value added B, i, tl=D 1 Country premium 12 Unidentified of the fund’s real estate portfolio in year t : This table summarizes the properties of real estate benchmark portfolios NVA ¼ R  R : i, t i, t reported by pension funds. The sample includes 241 funds in the CEM B, i, t Benchmarking database. We categorize benchmarks into different types and include the range of premia (in basis points) reported by the funds. We calculate the fund’s average NVA over the sample period as Let R be the real estate return net of fees i, t reported by fund i 2f1, :::, Ig in year t 2ft , t þ 0 0 NVA ¼ NVA i i, t 1, :::, t g: The fund’s benchmark portfolio B contains n i t¼2005 a local REIT index, a global REIT index, and a local and use it as our performance metric. A-rated corporate bond index. These indices have Our method brings several advantages. It is easy returns R , R , and R : We denote LOCAL, i, t GLOBAL, i, t BOND, i, t to implement, consistently applicable to real estate by g the proportion of the REIT portfolio invested in portfolios around the World and relies on listed the global REIT index, and by b the proportion of benchmarks that are investible and give real-time the total benchmark portfolio invested in the bond information about the risks present in these port- index. The return of the benchmark portfolio is folios (Gyourko & Keim, 1992). It also provides a reli- ðÞ able way to estimate the average NVA for a group R ¼ b  R þ 1  b B, i, t BOND, i, t of funds, whereas peer-based benchmarks imply (by ðÞ ðÞ 1  g R þ g  R : LOCAL, i, t GLOBAL, i, t definition) an average NVA of zero for the peer The parameter g governs the geographic mix of group. This feature is important in the context of the benchmark and the parameter b determines its our study because our purpose is to estimate the NVA for the group of large Canadian pension funds. leverage. A positive (negative) value of b indicates Another advantage of our method is that it works that the benchmark has lower (higher) leverage well even in the absence of information about the than the REIT portfolio. geographic mix, leverage, smoothing, and lag prop- To estimate the appraisal-based returns of private erties of a fund’s real estate portfolio. real estate, we smooth and lag the (de-)levered mar- One possible limitation of our approach is the ket returns of listed equity REITs. The smoothed return risk of over-fitting the data due to the multi-dimen- R of the benchmark portfolio is estimated as B, i, t sional nature of the optimization. To reduce this s s R ¼ sR þð1  sÞ R , B, i, t B, i, t B, i, t1 risk, we restrict the number of values that the where s is the smoothing parameter. A low value parameters g, b, s, and l can take. The proportion of of s indicates a high degree of smoothing. Given a global REITs g is allowed to vary in 0.05 increments lag parameter l (in trading days), we obtain the from 0 to 1. The smoothing parameter s can take 80 A. D. BEATH ET AL. four possible values: 0.5 (maximum smoothing), correlation between the adjusted REIT and the NPI 0.67, 0.75, and 1 (no smoothing). The lag param- remains low (9%). eter l varies from 0 to 520 trading days. The weight In Figure 1C, we smooth the adjusted REIT index. in the bond portfolio b is set such that the fund’s The model suggests a value of s ¼ 0:67, which is in reported return R has a beta of one with respect the range of estimates by Barkham and Geltner i, t to R for a given combination of parameters (1995) and Brown and Matysiak (2000). The correl- B, i, tl=D g, s, and l: In the next Section, we verify that the ation between the smoothed REIT index and the inferred parameter values are consistent with survey NPI increases to 41%. results on the funds’ actual investments. In Figure 1D, we lag the smoothed series by using Figure 1 illustrates the ability of a de-levered, themodel-predicted valueof l ¼ 272 trading days. smoothed, and lagged REIT index to match the The returns of the smoothed and lagged benchmark return properties of a private real estate portfolio. look nearly identical to those of the NPI. Indeed, the We consider the NCREIF National Property Index correlation between the lagged index and the NPI is (NPI), which tracks a portfolio of U.S. properties, and equal to 93%, which is large. The lag value predicted construct the benchmark portfolio from the S&P U.S. by the model is consistent with Gyourko and Keim REIT index using the methodology described above. (1992) who find that fourth-quarter lags of REIT returns Figure 1A plots the annual return of both indices work best in predicting the returns of the NPI index. from 2005 to 2019. The returns of the S&P REIT index Altogether, this exercise confirms that using REIT (dotted grey) are twice as volatile as those of the NPI benchmarks and adjusting for their geographical in black (19 vs. 9%), and the correlation between mix, leverage, smoothing, and lag structure makes it both times-series is low (15%). Without adjustments, possible to evaluate the performance of private real the REIT index is a poor benchmark for the NPI. estate portfolios across funds. In Figure 1B, we adjust the leverage ratio of the REIT index based on the results of our estimation. Data The model suggests a value of b ¼ 0:52: The result is a significant de-levering of the REIT index, which We use data from CEM Benchmarking, a Toronto- reduces its volatility to 11%. However, the based global benchmarking company that has 40% 40% 20% 20% 0% 0% 2005 2007 2009 2011 2013 2015 2017 2019 2005 2007 2009 2011 2013 2015 2017 2019 -20% -20% -40% -40% B. De-levering the REIT Index A. REIT Index (Dotted) vs. NPI 40% 40% 20% 20% 0% 0% 2005 2007 2009 2011 2013 2015 2017 2019 2005 2007 2009 2011 2013 2015 2017 2019 -20% -20% -40% -40% C. + Smoothing the REIT D. + Lagging the REIT Index Figure 1. Application of the benchmarking method to the NPI index. This figure displays the evolution of the National Property Index (NPI, solid black) and its REITs-based benchmark (dotted grey) between 2005 and 2019. We construct the benchmark by de- levering, smoothing, and lagging the S&P U.S. REIT index in a way that maximizes the correlation between the benchmark and the NPI. (A) Reports the REIT index unmodified. (B–D) Report the incremental impacts of the leverage, smoothing, and lag adjust- ments on the benchmark. For leverage, the benchmark corresponds to a mix with a 48% weight in the REIT index and a 52% weight in the AAA-corporate bond index. The smoothing parameter is 2/3. The lag is 272 trading days. JOURNAL OF SUSTAINABLE REAL ESTATE 81 collected detailed annual cost and performance large funds. In total, there are 241 funds from eight data from more than 1000 pension, endowment, countries (55 Canadian), including 36 large funds (8 and sovereign wealth funds in 18 countries. We Canadian). We group other funds in the category retain funds that report a minimum of five years of “Rest-of-World” (RoW hereafter). The majority of real estate returns from 2005 to 2019. The choice of RoW funds (140) are U.S. funds. We also study the 2005 as the starting date comes from the fact that subsets of Canadian and RoW funds that directly i) we require several years of REIT data to apply the manage more than 50% of the real estate portfolio. Table 2A reports the average proportion of assets smoothing procedure, and ii) the S&P REIT indices we use become consistently available across coun- invested in real estate for Canadian and RoW funds tries in the early 2000s. We eliminate the 10% of in 2019. Large Canadian pension funds invest sig- nificantly more in real estate (14%) than their peers funds with the lowest return correlation between the benchmark and the funds’ real estate portfolio, (8.5% for RoW funds). This result is in line with as well as funds for which the inferred leverage Beath et al. (2021), who find that Canadian pension coefficient is either negative or above 5. All remain- funds rank first globally in terms of the proportion ing funds have a benchmark correlation above 74%, of AUM invested in real assets. and the average benchmark correlation is 93%. The table also reports the proportion of real estate We categorize a fund as large if it manages more assets managed directly. The gap between Canadian than $50 bn in 2018. Throughout the paper, dollar and RoW funds is considerable. Large Canadian funds values are expressed in USD. Table 2 reports sum- invest 81% of their real estate portfolio directly, mary statistics for the full sample and the subset of whereas RoW funds invest 16% of their real estate Table 2. Summary statistics of pension fund real estate portfolios. (A) Summary statistics Number of funds Average AUM % Real estate % Direct Canadian funds All 55 26,747 0.10 0.27 Large 8 143,985 0.14 0.81 Direct (50%þ) 14 80,481 0.12 0.86 Large and direct (50%þ) 7 150,474 0.14 0.90 R.o.W funds All 186 27,845 0.07 0.07 Large 28 118,384 0.08 0.16 Direct (50%þ) 13 39,919 0.09 0.81 Large and direct (50%þ) 4 73,182 0.08 0.77 (B) Inferred benchmark parameters Lag (l) Smoothing (s) Bond (b) Gross/net Canadian funds All 220 0.62 0.22 1.30 Large 217 0.58 0.23 1.28 Direct (50%þ) 204 0.58 0.24 1.26 Large and Direct (50%þ) 238 0.60 0.24 1.24 R.o.W funds All 196 0.63 0.13 1.24 Large 199 0.57 0.12 1.24 Direct (50%þ) 159 0.58 0.05 1.39 Large and Direct (50%þ) 92 0.58 0.04 1.37 (C) Inferred geographic mix of benchmark Canada U.S. Europe Other Canadian funds 0.79 0.14 0.02 0.05 U.S. funds 0.01 0.79 0.06 0.14 Europe/UK funds 0.00 0.18 0.74 0.07 This table displays summary statistics on the real estate portfolios of Canadian and Rest-of-World (RoW) funds. The sample includes 241 funds in the CEM database. For each group of funds, we report in (A) the average value of the funds’ AUM (in USD millions), the proportion of assets invested in real estate, and the proportion of the real estate portfolio that is managed directly as of 2018. (B) Reports average parameter values of the funds’ REITs-based benchmark portfolios over the 2005–2019 period. The parameters include the lag (in trading days), degree of smoothing, and proportion of the benchmark invested in a local bond index. The fifth column reports the benchmarks’ corresponding gross asset-to-net asset leverage ratio. (C) Reports the geographic mix of the benchmark portfolios. 82 A. D. BEATH ET AL. directly. Moreover, seven of the eight large Canadian In Table 2C, we report the inferred geographic mix funds invest more than 50% of their real estate port- for the real estate portfolios of Canadian, U.S., folio directly, while only four of 28 large RoW funds European, and UK pension funds. The inferred mix do so. This significant difference is again consistent takes into account the geographic mix of the underly- with Beath et al. (2021) who find that Canadian pen- ing REIT indices. According to our estimation, the sion funds rank first globally in terms of the propor- averageexposureto local markets (1-g) was 74% for tion of total assets managed in-house. funds from Europe and UK and 79% for U.S. funds Together, these statistics indicate that Canadian from 2005 to 2019. These values are again close to pension funds are significantly more active than the results from the 2020 CEM Global Leaders Survey. their peers in the direct real estate market. For each According to the survey, the average exposure of group of funds, we estimate the proportion of real European and UK funds to local markets was 70% estate that is directly managed as the product of (i) and that of US funds was 82% at the end of 2019. the average proportion of AUM that is invested in For Canadian pension funds, the average expos- real estate and (ii) the average proportion of real ure to the local market (79%) exceeds the average estate that is invested directly. This proportion is local exposure reported in the CEM Global Leaders 0:14  0:81 ¼ 12% for large Canadian pension funds survey (57% at the end of 2019). The survey esti- but only 1.2% for large RoW funds. One Canadian mate is lower for two reasons. First, funds in the pension fund is therefore equivalent to ten RoW survey are larger than the typical Canadian fund funds of the same size in this market. and invest a greater proportion of their assets Table 2B reports the average inferred values of abroad than smaller funds. Second, the foreign real the l, s, b, and g parameters of the benchmark estate exposure of Canadian pension funds has portfolios. The values are close to the inferred increased since 2005. parameters of the NPI benchmark and are stable across groups of funds, which confirms that the Performance of Canadian Pension Funds benchmark methodology works well. For example, We now compare the net value added of Canadian the average smoothing parameter is s ¼ 0.62 for and RoW funds. For each subset of these funds, we Canadian funds and s ¼ 0.63 for RoW funds. report the average, median, and standard deviation We note that heterogeneity in the bond param- of the NVA across funds in Table 3A. eter b does not necessarily indicate heterogeneity in We generally observe that externally-managed the gross asset-to-net asset leverage ratio of the real estate portfolios perform worse than internally- real estate portfolios. The reason for the gap is that managed portfolios, consistent with the findings of the underlying REIT indices have varying leverage Andonov et al. (2015). This pattern is visible both ratios. For example, the average Canadian fund has inside and outside Canada. For example, RoW funds, a lower bond weight than the average RoW fund which mostly include funds that manage the major- (b ¼ 0.22 vs. 0.13) but a slightly higher leverage ity of their assets externally, have an average NVA ratio (1.3 vs. 1.24) due to the high leverage of equal to negative 215 bps. A likely explanation is Canadian REITs. The low leverage ratio (1.24) of the that these funds incur high fees due to the high seven large Canadian funds with high direct real share of external management. This finding is con- estate management can be explained by the fact sistent with the evidence that external investments that a subset of them use leverage at the total fund in private markets tend to underperform (Andonov level. The overall leverage is used in part to fund real estate investments at a low cost. et al., 2021; Beath & Flynn, 2020; Gupta & Van We verify that the leverage ratios inferred from Nieuwerburgh, 2021; Phalippou & Gottschalg, 2009). our estimation are consistent with the 2020 CEM Quite strikingly, large Canadian pension funds that Global Leaders survey of 26 of the largest institu- invest the majority of their assets directly have the tional investors in the CEM database. As of 2019, the highest NVA among all groups of funds in our sam- surveyed leverage ratios ranged from 1.07 to 1.79 ple. Their average NVA of 148bps is statistically sig- with an average value of 1.47. This is in line with the nificant and economically large. Their median NVA of range of leverage ratios reported in Panel B. 295bps is also large, confirming that the large JOURNAL OF SUSTAINABLE REAL ESTATE 83 Table 3. Net value added and expected return. Count Average Median Standard deviation Standard error (A) Net value added (basis points) Canadian funds All 55 35 79 361 49 Large 8 122 149 248 88 Direct (50%þ) 14 138 176 239 64 Large and direct (50%þ) 7 148 295 255 96 R.o.W funds All 186 215 227 427 31 Large 28 68 35 324 61 Direct (50%þ)13 45 83 693 192 Large and direct (50%þ)4 167 139 287 143 (B) Average return Canadian funds All 55 0.085 0.091 0.036 0.005 Large 8 0.095 0.094 0.020 0.007 Direct (50%þ) 14 0.096 0.097 0.020 0.005 Large and direct (50%þ) 7 0.097 0.102 0.021 0.008 R.o.W funds All 186 0.043 0.045 0.041 0.003 Large 28 0.058 0.060 0.029 0.006 Direct (50%þ) 13 0.069 0.053 0.067 0.018 Large and direct (50%þ) 4 0.053 0.060 0.019 0.009 This table displays statistics of the net value added (NVA) for Canadian and Rest-of-World (RoW) funds. The sample includes 241 funds in the CEM database. For each group of funds, we report in (A) the cross-sectional average, median, standard deviation, and standard error of the funds’ average NVA in basis points between 2005 and 2019, and in (B) the cross-sectional average, median, standard deviation, and standard error of the real estate portfolio’s average return net of costs. For each fund, the average return is calculated as the sum of the benchmark portfolio average return and the fund’s average NVA. average is not driven by outliers. These results indi- 100bps lower than that of comparable Canadian funds. cate that the real estate strategy pursued by these Moreover, for the small sample of RoW funds that are funds delivers superior performance net of fees. both large and invest a high proportion of assets dir- The strong performance of Canadian pension ectly, the average NVA was negative 167bps. These sta- funds is not restricted to the seven largest funds tistics confirm that the investment strategy used by managing more than $50 bn of real estate invest- large Canadian pension funds is distinct. ments. The average NVA of the 14 Canadian funds Average Return that directly manage a majority of their real estate In Table 3B, we report the average return of real investments was 138 bps over the sample period. estate portfolios across the different groups of funds These funds make up 25.5% of Canadian funds and from 2005 to 2019. For each fund, the average together fully explain the average NVA of Canadian annual return is estimated in local currency as the funds (0.255 138 ¼ 35 bps). sum of (i) the average return of the corresponding We emphasize that this strong performance is not driven by the funds’ exposure to the local real benchmark portfolio over the full period and (ii) the fund’s average NVA. This calculation allows us to estate market. As explained in the previous section, make use of the full sample period for each fund our benchmarking method controls for the param- eter g that governs the geographic mix of the and thus avoid a situation in which we compare funds’ average returns over short and different sam- benchmark. Moreover, we find that the other Canadian funds that manage fewer than $50 billion ple periods. There is a high level of heterogeneity in the and a minority of their real estate investments dir- funds’ return performance. Part of the performance ectly have an average NVA that is close to zero. We also verify that the value added to Canadian gap is driven by the performance of local real estate pension funds’ real estate portfolios is not just the result markets. For example, between 2005 and 2019 Canadian pension funds generated an average of managing assets directly. Using RoW funds as a con- trol group, we find that RoW funds that invest the return of 8.5% while RoW funds generated an aver- majority of their real estate portfolio directly outper- age return of 4.3%. This return differential is partly formed their benchmark by 45bps on average between due to the booming real estate sector in Canada, 2005 and 2019. While this NVA value is positive, it is both in the residential and commercial spaces. 84 A. D. BEATH ET AL. The other driver of the heterogeneity in return December 31, 2018, according to WillisTower performance is the funds’ NVA. Consider for Watson. We include in this list the wholly-owned example the heterogeneity in the NVA within the real estate operating subsidiaries of these funds. group of Canadian pension funds, which all invest Table 4A reports statistics of these large pension heavily in Canada. The average return of the seven funds. Six of the 88 pension funds in the list are Canadian. Approximately 50% of RoW funds are U.S. large Canadian pension funds with a high share of funds. The six Canadian pension funds manage about direct investments was 9.7% from 2005 to 2019. 6% of total pension fund assets. Their median fund This is 1.2% greater than the average return of all size ($93.35bn) exceeds that of RoW funds ($81.19bn). Canadian funds. Such outperformance is significant We also report the subset of funds that have considering that the return time-series for the seven made direct real estate deals as reported by Preqin. large funds have a lower implied leverage ratio than All six Canadian funds do so but fewer than half of that of the average Canadian fund. the RoW funds (36) do. Unsurprisingly, funds that Overall, our performance analysis confirms that directly invest in real estate tend to be larger. The the real estate strategy implemented by large median size of RoW funds is $93.33bn, which is Canadian pension funds has created significant similar to the median size of the large Canadian value in 2005–2019 net of fees. pension funds. Table 4B presents statistics on the real estate Green Urban Development transactions made by these funds. There are 139 We now take the analysis down to the asset level buy transactions made by Canadian funds and 146 and investigate the real estate strategy imple- buy transactions made by RoW funds between 2010 mented by large Canadian pension funds. The next and 2018. A typical transaction value ranges section reviews data on real estate transactions between $100 and 110 mn. In total, these transac- made by these funds. Then we identify green urban tions amount to 2.4% of all real estate transactions development as a key component of their strategy. in the sample of Preqin properties. If we also incorp- The section on Green Urban Development in Large orate transactions where sellers include one of these Canadian Cities quantifies and illustrates the size of pension funds, then the total value of deals that dir- green urban development in large Canadian cities. ectly involve a large pension fund corresponds to about 4% of the aggregate value of deals reported Finally, in the section on Beyond LEED-VA Development: Greenfield Projects we describe the in Preqin. funds’ involvement in greenfield projects. Table 4B illustrates the extent to which Canadian pension funds are involved in the private real estate market. Between 2010 and 2018, Canadian pension Data funds directly invested $32.24bn in real estate prop- We use data from Preqin, a data provider that col- erties, while RoW funds directly invested $22.38bn. In lects detailed information on thousands of real total, Canadian pension funds were, therefore, estate transactions around the world. The informa- responsible for 60% of the total value of real estate tion includes deal size, date, location, sector, asset acquisitions involving a pension fund. This proportion type, and buyer and seller identification. The dataset is 10 times greater than the share of total pension goes from 1988 to 2018; however, because 99.9% of fund assets they manage (6% as reported above). reported transactions at the property level took Sell transactions reveal a similar pattern. Canadian place after 2010, we primarily study the 2010–2018 pension funds directly sold a total of $20.21bn of period. In total, real estate transactions reported in real estate properties between 2010 and 2018, Preqin during this period amounted to USD 2.44 tril- whereas RoW pension funds directly sold $13.8bn. lion, 61% of which consisted of U.S. properties. In addition to being more active than their peers To analyze the deals that directly involve a large in the private real estate market, Canadian pension pension fund, we restrict the sample to deals where funds also invest more abroad. We count the total buyers include a fund that manages assets for a number of foreign countries where pension funds pension plan that is worth more than $50bn as of make direct real estate deals that are reported in JOURNAL OF SUSTAINABLE REAL ESTATE 85 Table 4. Summary statistics of real estate deals that directly and CAGBC websites. The LEED certification brings involve pension funds. several advantages to our analysis. It is global, Canadian funds RoW funds widely used in Canada and North America, and cov- (A) Pension funds ers a wide range of properties, including office, resi- All funds Number of funds 6 82 dential, mixed-used, and industrial. Moreover, the Median AUM (USD Mn) 93,352 81,192 LEED certification focuses on environmental initia- % Of total AUM 0.06 0.94 Funds that directly invest in RE tives that add value to an asset, as confirmed by Number of funds 6 36 Median AUM (USD Mn) 93,352 93,329 Eichholtz et al. (2010), Fuerst and McAllister (2010), (B) Real estate deals and Miller et al. (2008). According to the U.S. Green Buy deals Number of deals 139 146 Building Council (USGBC) website, LEED certification Median size (USD Mn) 111 105 provides: (i) instant recognition for the building; (ii) Total (USD Mn) 33,236 22,380 Sell deals faster lease-up rates; (iii) higher resale value; (iv) Number of deals 102 127 healthier indoor space; (v) lower use of energy, water, Median size (USD Mn) 166 83 Total (USD Mn) 20,217 13,796 and other resources; and (vi) brand enhancement. Geography The capital investment that is required to obtain a Number of foreign countries 10 2 Sector (%) LEED certification usually ranges from 2.5 to 9.4% of Residential 0.19 0.09 Office 0.41 0.46 the property’s value depending on the certification Retail 0.03 0.18 level (Kats, 2003; Matthiessen and Morris, 2004; Mixed use 0.31 0.13 Other 0.05 0.15 Nyikos et al., 2012), which is significant. This table displays summary statistics of direct real estate deals made by Another advantage of working with the LEED cer- Canadian and Rest-of-World (RoW) funds that manage assets for a pen- tification is that the information made available by sion plan with more than USD 50 bn of AUM in 2018. (A) Presents sum- mary statistics about (i) all funds, and (ii) funds that directly invest in real USGBC and CAGBC includes the street address, level estate. Statistics include the total number of funds, the median AUM, and the proportion of aggregate AUM managed by each group of funds. (B) of certification achieved (Platinum, Gold, Silver, Presents statistics about the direct real estate deals made by these funds Certified), and the dates at which the certification between 2010 and 2018, where the data are obtained from Preqin. The first two sets of rows correspond to buy and sell transactions, respect- was requested and approved for every certified ively. Deal size is reported at the asset level. The third and fourth sets of building. We can therefore track the greenness of rows report the geography and sector distribution of buy deals. The num- ber of foreign countries corresponds to the average number of foreign every building in a pension fund’s real estate port- countries where deals were made by all funds of a particular country. folio and tease apart whether a pension fund pur- chases buildings that are already LEED-certified or Preqin. There are 10 foreign countries where deals actively seeks to obtain the LEED certification after are made by Canadian pension funds. By contrast, the acquisition. there are on average 2.2 foreign countries where To this end, we introduce the following termin- deals are directly made by funds from an ology. A “LEED value-added (VA)” investment corre- RoW country. sponds to a real estate investment where the fund In terms of industry sectors, Canadian pension actively seeks some form of LEED certification after funds invest in a balanced mix of assets. About 41% having acquired the property and before selling it. of their real estate portfolio is invested in office This action could either be the registration of a new space, 31% is invested in mixed-use properties, and project, the certification of an existing project, or 19% is invested in residential properties. In compari- both. A “premium” investment is one where the son, RoW funds invest the majority of their portfolio fund purchases an asset already LEED-certified and in office properties (46%). does not make LEED upgrades to it. A “basic” investment is one that does not have any LEED cer- tification throughout the sample. LEED Value-Added Activity In the first set of rows of Table 5A, we report the The greenness of a real estate investment can be portfolio allocation in basic, premium, and LEED measured according to a variety of certification pro- value-added properties. Canadian funds invest grams, including LEED, BREEAM, Energy Star, and 27% of their assets in LEED-VA strategies. In com- others. In this study, we use data from the LEED cer- parison, RoW funds invest only 14% of their assets tification program that is available on the USGBC in LEED-VA strategies. The majority of assets held by 86 A. D. BEATH ET AL. Table 5. Comparison of green value-added activity across funds and then sold again between 2010 and 2018. pension funds. The table confirms the dominance of Canadian pen- Canadian funds RoW funds sion funds in this market, which are responsible for (A) Buy deals 24 of the 39 transactions. All properties Basic 0.60 0.74 Table 5B further shows that Canadian pension Premium 0.13 0.13 funds are more likely to directly purchase proper- LEED-VA 0.27 0.14 US office properties ties, increase their value through LEED-certification, Basic 0.23 0.46 Premium 0.29 0.34 and then re-sell the properties afterward. Buy-and- LEED-VA 0.48 0.20 sell transactions indeed represent a greater fraction (B) Buy-and-sell deals Buy-and-sell deals of total transaction volume for Canadian funds Number of deals 24 15 (17%) than for RoW funds (5%), which indicates Total (USD Mn) 5,745.44 1,069.64 % Of buy deals 0.17 0.05 greater turnover. Moreover, these transactions are All properties mostly LEED-VA transactions for the Canadian funds Basic 0.36 0.96 Premium 0.03 0.04 (61%) whereas virtually all transactions made by LEED-VA 0.61 0.00 RoW funds are either basic or premium. (C) Geography breakdown for Canadian funds In the final part of our analysis based on Preqin Investment type data, we focus on the geographical breakdown of Basic Premium LEED-VA basic, premium, and LEED-VA assets for all buy deals Geography (%) made by Canadian funds. Table 5C shows that Canada 0.59 0.00 0.41 U.S. 0.48 0.17 0.35 LEED-VA assets amount to 41% of their investments Other 1.00 0.00 0.00 in Canadian properties, 35% of their investments in This table displays the degree of LEED value-added activity by Canadian and Rest-of-World (RoW) funds between 2010 and 2018. (A) Focuses on U.S. properties, and 0% of their investments in other buy deals. The first (second) set of rows reports the percentages of all foreign properties. Therefore, Canadian funds are (US office) properties that are basic, premium, and LEED-VA. (B) Focuses on properties that were bought by Canadian and Rest-of-World (RoW) much more likely to conduct the internal green funds and then sold again between 2010 and 2018. The first set of rows value-added activity for real assets close to home. reports the number of these deals, their total value, and the proportion of their total value relative to the total value of buy deals. The second set of rows provides the breakdown of these deals into basic, premium, and LEED-VA investments. (C) Zooms in on the LEED value-added activity by Green Urban Development in Large Canadian funds and provides the breakdown for properties located in Canada, U.S., and other countries. For each geography, we report the per- Canadian Cities centages of all properties that are basic, premium, and LEED-VA. To further investigate the green urban development RoW funds are invested in basic strategies (74%). strategy implemented by Canadian pension funds in The evidence thus reveals that Canadian pension large Canadian cities where the funds are most funds are significantly more involved in green urban involved, we augment the dataset by manually col- development than RoW funds as measured by lecting hundreds of real estate transactions from LEED activity. the funds’ annual reports and press releases for the Once we account for the fact that LEED certifica- nine largest Canadian pension funds. These funds tion is most common in U.S. office space, the gap consist of Alberta Investment Management between Canadian funds and RoW funds’ involve- Corporation (AIMCo), British Columbia Investment ment in LEED-VA strategies increases. In the second Management Corporation (BCI), La Caisse des set of rows of Table 5A, we report the composition D epots et Placements du Quebec (CDPQ), Canada of basic, premium, and LEED-VA assets within the Pension Plan Investment Board (CPPIB), Healthcare portfolio of U.S. office properties. Canadian funds of Ontario Pension Plan (HOOPP), Investment invest approximately half (48%) of their portfolio of Management Corporation of Ontario (IMCO), Ontario direct U.S. office acquisitions into LEED-VA assets, Municipal Employees’ Retirement System (OMERS), whereas RoW funds only invest 20% of their port- Ontario Teachers Pension Plan (OTPP), and Public folio in these assets. Sector Pension Investment (PSP). The information Table 5B provides additional insights into the we collect includes the street address, building use, nature of these LEED-VA assets. We zoom in on the and the date of the development or acquisition of 39 properties that were bought by the pension the property. JOURNAL OF SUSTAINABLE REAL ESTATE 87 Figure 2. Canadian pension fund ownership and green development of Toronto. This figure displays the real estate properties dir- ectly owned and greened by the top nine Canadian pension funds in Toronto’s downtown area. Blue properties are not LEED certi- fied (“basic”). Green properties have obtained LEED certification by the pension funds (“value-added”). The red boundary delineates Toronto’s financial district. The orange boundary delineates Toronto’s South Core district. To capture building size, we obtain mapping data presents a map of the Toronto downtown area on land parcels and buildings. This information is where the pension fund activity is most pro- available online for Toronto, Vancouver, and nounced. The blue and green colors, respectively Montreal—Canada’s three largest cities. We convert indicate basic and LEED-VA investments. The red addresses into geocodes, geocodes into land parcels, boundaries delineate Toronto’s financial district. and then land parcels into buildings. The use of land The map illustrates the magnitude of green parcels as an intermediary step allows us to address urban development by the nine large Canadian cases where the geocode does not directly intersect pension funds. Almost all buildings in Toronto’s with building data. These cases arise for buildings that financial district are directly owned by these only take up a fraction of the land parcel and build- funds. The vast majority of these buildings have ings with varying sections of elevation. Our spatial obtained some form of LEED certification under the matching procedure ensures that all buildings owned funds’ ownership, as evidenced by the dominance by Canadian pension funds are properly captured. of green-colored buildings. Once we move away Using our data on deals and LEED certification, from the city center, the proportion of buildings we construct 3-D maps of the properties developed owned by pension funds rapidly declines. and/or acquired by the top nine Canadian pension Figures 3 and 4 present equivalent 3-D maps for funds in Toronto, Montreal, and Vancouver. Figure 2 Montreal and Vancouver. The dark green color on 88 A. D. BEATH ET AL. Figure 3. Canadian pension fund ownership and green development of Montreal. This figure displays the real estate properties directly owned and greened by the top nine Canadian pension funds in Montreal’s downtown area. Blue properties are not LEED certified (“basic”). Green properties have obtained LEED certification by the pension funds (“value-added”). The dark green property obtained LEED certification before being acquired by pension funds (“premium”). The red boundary delineates Montreal’s financial district. The orange boundary delineates Montreal’s Quartier des Gares. the Montreal map indicates a premium investment. An additional statistic of interest is the proportion Both maps show the same pattern. Canadian pen- of buildings that obtained LEED Platinum, the high- sion funds own and green a large proportion of est level of LEED certification. Only 4% of all properties in the downtown areas, and their activity Canadian projects have obtained Platinum certifica- rapidly declines outside these areas. Altogether, tion according to the CAGBC database. In Toronto, these figures reveal a systematic pattern of LEED- 26 buildings have obtained it. Seventeen of these VA activity. buildings (65%) correspond to LEED-VA projects by The maps likely understate the true level of five of the top nine Canadian pension funds. These LEED-VA activity by Canadian pension funds for sev- results show that Canadian pension funds have eral reasons. It may be that we have not identified played a significant role in bringing Toronto’s prop- the full list of properties owned by the top nine erties to the highest LEED standards. pension funds, as only the largest deals are made publicly available. It may also be that some of the Beyond LEED-VA Development: Greenfield Projects remaining downtown properties are owned by Canadian pension funds outside the top nine. LEED-VA activity on existing buildings only represents Additionally, several of the non-LEED-certified build- a fraction of the Canadian pension funds’ real estate ings in blue correspond to buildings under con- strategy. We also find that these funds undertake struction that are expected to obtain LEED large greenfield projects where they purchase the certification in the future. Finally, the building maps land and develop the green properties themselves. for Montreal and Vancouver are from 2016 and The maps in Figures 2–4 show that Canadian pen- 2009, respectively, so any building developed after sionfunds arehighly active indeveloping up-and-com- these dates is missing from the maps. ing districts near the city centers. In Toronto, Canadian JOURNAL OF SUSTAINABLE REAL ESTATE 89 Figure 4. Canadian pension fund ownership and green development of Vancouver. This figure displays the real estate properties directly owned and greened by the top nine Canadian pension funds in Vancouver’s downtown area. Blue properties are not LEED certified (“basic”). Green properties have obtained LEED certification by the pension funds (“value-added”). The red boundary delin- eates Vancouver’s financial district. pension funds developed the majority of towers in We find that the proportion of buildings devel- Toronto’s emerging South Core district near the water- oped by Canadian pension funds is large. Two of the front, as delineated by the orange boundaries on the tallest 10 towers in Toronto that are under construc- map. These towers include the Telus Tower (HOOPP), tion (CIBC Square II and 160 Front St West) are RBC WaterPark place (OMERS), and CIBC Square Canadian pension fund projects. Moreover, the tow- (CDPQ). In Montreal, pension fund activity is highest in ers developed by these funds are more likely to the Quartier des Gares which was established in 2015 obtain Platinum certification. For example, if we focus (again delineated by the orange boundaries). Examples on the 50 tallest towers in Toronto that are either of development projects in this district include the completed or under construction, two of the five Deloitte Tower and Tours des Canadiens (OTPP). towers already developed by Canadian pension funds To quantify local greenfield activity by Canadian have obtained LEED Platinum. In contrast, only one pension funds, we use data from the Council on Tall of the 11 towers developed by non-pension funds Buildings and Urban Habitat (CTBUH) which pro- has obtained it. That tower, Scotia Plaza, obtained vides information on the developers of skyscrapers LEED Platinum in 2020 after the Canadian pension for each major city. We focus on Toronto because fund AIMCo became a major shareholder in 2016. the information available is the most complete. Toronto ranks among the world’s top cities in terms Discussion of the number of skyscrapers taller than 150 m that are completed (#19 with 80 buildings as of July Our findings highlight a possible investment strat- 2022) and under construction (#8 with 27 buildings egy for other large asset owners to follow. In this as of February 2022). Section, we discuss the benefits and risks of this 90 A. D. BEATH ET AL. strategy and conclude with several questions for that the green buildings developed by large future analysis. Canadian pension funds are mostly limited to stand- ard multi-use properties located in prime downtown locations of major cities with diversified industries. Distinct Features of the Green Urban A second risk-mitigating strategy is the concen- Development Strategy tration of projects. Focusing on a set of projects in The green urban development strategy pursued by a particular neighborhood brings multiple advan- Canadian funds has several distinct features. First, it tages. It allows the fund to work in an environment is done mainly in-house. The benefits of internal that it knows well and thus better manages the management are low fees due to the elimination of operational risks (Huffman, 2002). In other words, financial intermediaries. As a result, Canadian funds concentration acts as a risk mitigant from an oper- have the ability to extract the full economic profit ational perspective. Working on local projects also from the real estate ventures. In comparison, most allows the fund to align the risk of its pension liabil- institutional investors outsource the majority of their ities to its assets since local properties correlate private investments to external fund managers and more with the fund’s liabilities than foreign proper- consequently incur large fees that frequently offset ties do. Finally, developing a complete neighbor- the active gains on the portfolio. Additional hood allows the fund to generate housing benefits of pursuing the green urban development externalities (Rossi-Hansberg et al., 2010). internally include full control of the assets, direct In line with this logic, we find that Canadian pen- communication between owners and managers, and sion funds do the majority of green urban develop- a lower cost of project financing since any debt ment in highly localized markets. If we define issuance can be done at the total portfolio level Montreal as the home market of CDPQ and PSP and instead of the project level. Toronto as the home market of CPPIB, OTPP, The green urban development strategy is also dis- OMERS, and HOOPP, we find that, for each fund, tinct in the way it departs from the well-known the home market represents at least 75% of its LEED-VA projects that we have collected across all forms of pension fund activism (Barber, 2020;Del three cities. Guercio & Hawkins, 1998;Wahal, 1996). Pension We further find that, inside a particular city, funds typically aim to engage with large public firms Canadian pension funds tend to concentrate their where they own a sizeable ownership stake. By con- investments in a few locations. A good example is trast, Canadian pension funds share the ownership of real estate projects with few investors and focus on OTPP’s recent investments in Montreal’s Quartier des Gares (Windsor Station, Deloitte Tower, and the project execution, by concentrating their efforts on multiple Tours des Canadiens) which are all located developing and greening urban properties. within a few steps of the central train station. Another example is OMERS’s large investment in Risk-Mitigation Quartier DIX30, a 2.3 million square feet multi-use Developing and managing real assets involves sev- complex in Brossard, which is a municipality part of eral legal and operational risks. There are also the Greater Montreal area that will soon serve as a liquidity risks related to the long-term nature of station for the Reseau express metropolitain development projects, and risks associated with the (REM) network. real estate industry (Ling et al., 2020). A third risk-mitigating strategy is the use of a We discuss several strategies asset owners can wholly-owned real estate operating subsidiary. use to mitigate these risks. One risk-mitigating strat- Examples of subsidiaries in Canada include Ivanhoe egy is the application of green urban development Cambridge (CDPQ), Cadillac Fairview (OTPP), Oxford to a specific set of projects which are well defined Properties (OMERS), QuadReal (BCI), HOOPP Realty (i.e. construction of a downtown tower), can be (HOOPP), and AIMCo Realty (AIMCo). The use of a replicated in multiple cities, and have a risk profile subsidiary provides several advantages. One advan- that naturally matches the long-term bond profile tage is experience. Subsidiaries, such as Cadillac of the funds’ liabilities. Our analysis indeed confirms Fairview, which was purchased by OTPP in 2000, JOURNAL OF SUSTAINABLE REAL ESTATE 91 have acquired decades of experience in developing the growing integration of ESG into corporate busi- and managing hundreds of real estate projects. ness models may encourage demand for green Experience not only helps to identify valuable devel- urban development outside city centers. For opment opportunities and manage operational risks example, condominium units or office buildings that but also to establish long-term relationships with are located either in smaller cities or outside the suppliers and retailers that lease the space. major city centers could become prime targets for Another advantage of using a subsidiary has to green retrofitting moving forward. A notable do with protection from legal risks. Because subsid- example is Downsview Airport in Toronto, which was iaries are limited-liability entities, they do not purchased by PSP in 2018 and is currently being con- engage the funds’ investments outside the subsid- verted into a mixed-use urban community. Finally, can smaller pension funds successfully iary. Additionally, a subsidiary provides flexibility implement green urban development? The strategy regarding employee compensation. Compensation we have uncovered requires scale and the ability to contracts in limited partnership funds tend to be conduct operations in-house. A scaled-down version long, complex, and different from contracts of buy- could potentially be achieved either through co- side institutional funds (Phalippou, 2009). Having a investments or through direct investments in subsidiary can make it easier for pension funds to smaller buildings. Here again, the current ESG wave set up these types of contracts while keeping the could make such a strategy possible for smaller investments in-house. pension funds. We leave these questions for Finally, by fully owning the subsidiary, asset own- future research. ers can mitigate agency conflicts that arise when ownership is fragmented. Typical conflicts include expropriation or inefficient use of funds by manag- Notes ers when investors have limited ability to control the firm’s assets. Having large shareholders can miti- 1. This form of alpha resulting from value-added improvement is also referred to as the gamma of the gate these problems because they have both a gen- expected return of a real estate portfolio (Kaiser, 2005). eral interest in profit maximization and sufficient 2. Because time-series of S&P REIT indices we use only control over the assets of the firm (Shleifer & become consistently available across countries in the Vishny, 1997). early 2000s, we use three return values of the indices to estimate the smoothed return of the benchmark portfolios (i.e. RB,i,t, RB,i,t-1, RB,i,t-2). Their weights are Expanding Green Urban Development? then re-scaled so that they sum up to 100%, and earlier index returns are assigned a weight of zero. Our analysis shows that green urban development 3. We set a lower bound of 0.5 for the smoothing can be a successful investment strategy for asset parameter s, consistent with the estimates of Barkham owners that can carefully manage the risks and gen- and Geltner (1995). If we use values of s below 0.5, the erate economies of scale. The asset owners we have volatility of the smoothed benchmark portfolio studied in this paper are large pension funds with becomes so small that it is no longer possible to match sizable in-house teams. the volatility of the fund’s reported return unless we apply an unreasonably high degree of leverage. Several questions arise from our analysis. Is LEED- 4. For a small proportion of funds the information on VA the only form of green urban development? We 2018 is unavailable. So their AUM value corresponds to have used the LEED certification for the purpose of the latest available year. our analysis, but it may be that other certification 5. According to the annual reports published by large programs provide similar benefits to investors in Canadian pension funds, the proportion of their real terms of value-added gains and environmen- estate assets invested in Canada has generally decreased from 2009 to 2017. For example, Public tal benefits. Sector Pension Investment (PSP Investments)’s share Has the existing market for green urban develop- decreased from 51 to 21%, La Caisse des Depots et ment reached a point of saturation? On the one Placements du Quebec (CDPQ)’s share decreased from hand, the 3-D map of Toronto shows that the major- 43 to 32%, Alberta Investment Management ity of downtown buildings have already been retrofit- Corporation (AIMCo)’s share decreased from 97 to 71%, ted to existing LEED standards. On the other hand, Canada Pension Plan Investment Board (CPPIB)’s share 92 A. D. BEATH ET AL. decreased from 56 to 15%, and Healthcare of Ontario the east, Front St to the south, and University Ave to Pension Plan (HOOPP)’s share decreased from 99 the west. to 69%. 17. For Montreal’s and Vancouver’s financial districts, we 6. The performance analysis in Andonov et al. (2015) uses set approximate boundaries based on publically benchmark returns from the funds’ reported real estate available information. benchmarks. Using our common benchmarking 18. The data is available at www.skyscrapercenter.com. approach, we also find that direct real estate 19. This information is available at https://open.toronto.ca. management generally leads to superior performance. Statistics on the number of buildings come from www. 7. Another possible explanation is that several large U.S. skyscrapercenter.com. funds suffered large losses during the 2008 Great Financial crisis due to highly leveraged positions and Acknowledgments subsequently de-levered their portfolio as part of the Omar Mohsen, Oriane Pacic, and Frederic Sfeir provided out- re-structuring. For example, California Public Employees’ standing research assistance. We thank Jonathan Critchley Retirement System (CalPERS) posted a negative total for help with the 3D-mapping and Chris Flynn, Eric Fontaine, return of 48.7% in the first nine months of 2009 Mario Lefebvre, Luc McSween, Mark Staley, Eduard van according to “CalPERS considers real estate clearout” by Gelderen, and seminar participants at the Bank of Canada, Richard Lowe, IPE Real Assets, December 14, 2009. KPMG, and the 2021 National Pension Hub Insight Series for However, we note that the median NVA of RoW funds insightful comments. The views and opinions expressed in is also significantly negative, which confirms that the this paper are solely those of the authors. They do not negative average NVA of these funds is not driven by necessarily reflect the views or opinions of the Public Sector the actions of a few funds. Pension Investment Board. All errors are our own. 8. WillisTower Watson Thinking Ahead Institute, Pensions & Investments, World 300. In total there are 90 funds that manage more than USD 50 billion. Two of these Funding funds, however, are managed by CDPQ. We exclude This research is funded by the National Pension Hub of the Norway’s Government Pension Fund as it mainly Global Risk Institute. represents the country’s sovereign wealth fund. This leaves a total of 88 funds. 9. In many cases, transactions consist of portfolios involving multiple properties. We approximate the value of each References property as the total value of the deal divided by the number of properties mentioned in the deal. Ambachtsheer, K. (2016). The future of pension fund manage- 10. The information is available at https://new.usgbc.org/ ment. Wiley Finance Editions. leed and at www.cagbc.org. Ambachtsheer, K. (2021). The Canadian pension model: Past, 11. For deals where we only observe the year of acquisition present, and future. The Journal of Portfolio Management, instead of the exact date, we assume that a LEED 47(5), 150–158. https://doi.org/10.3905/jpm.2021.1.216 action that takes place in the same year counts as a Andonov, A., Eichholtz, P., & Kok, N. (2015). Intermediated value-added activity. investment management in private markets: Evidence 12. A “basic” investment should not interpreted as being from pension fund investments in real estate. Journal of not green. Even though the building has no LEED Financial Markets, 22,77–103. https://doi.org/10.1016/j.fin- certification, it may have another green certification. mar.2014.11.002 13. In the Appendix, webreak down thedifferent types of € Andonov, A., Kraussl, R., & Raugh, J. (2021). Institutional LEED value-added activity: LEED registration only, LEED investors and infrastructure investing. The Review of certification only, and the combination of registration and Financial Studies, 34, 3880–3934. https://doi.org/10.1093/ certification. Across all types of funds, the bulk of value- rfs/hhab048 added projects involves the full LEED certification process. Barber, B. (2020). Chapter 15: Pension fund activism: A dou- 14. This information is available at https://open.toronto.ca, ble-edged sword. In O. Mitchell and G. Anderson (Eds.), https://opendata.vancouver.ca and http://donnees.ville. The future of public employee retirement systems. Oxford montreal.qc.ca. University Press. 15. Buildings with varying elevation are categorized as a Barber, B., Morse, A., & Yasuda, A. (2021). Impact investing. collection of sub-buildings. Therefore, without the use Journal of Financial Economics, 139(1), 162–185. https://doi. of land parcels as an intermediary step, we would only org/10.1016/j.jfineco.2020.07.008 capture the part of the building that intersects with the Barkham, R., & Geltner, D. (1995). Price discovery in American geocode data. and British property markets. Real Estate Economics, 23(1), 16. We use the boundaries provided by the Wikipedia page 21–44. https://en.wikipedia.org/wiki/Financial_District,_Toronto. Beath, A., Betermier, S., & Spehner, Q. (2021). The Canadian The boundaries are Queen St W to the north, Yonge St to pension fund model: A quantitative portrait. The Journal of JOURNAL OF SUSTAINABLE REAL ESTATE 93 Portfolio Management, 47(5), 159–177. https://doi.org/10. Matthiessen, L. F., Morris, P. (2004). Costing green: A compre- 3905/jpm.2021.1.226 hensive cost database and budgeting methodology. https:// Beath, A., & Flynn, C. (2020). Benchmarking the performance www.usgbc.org/resources/costing-green-comprehensive- of private equity porfolios of the world’s largest institu- cost-database-and-budgeting-methodology tional investors: A view from CEM benchmarking. The Miller, M., Spivey, J., & Florance, A. (2008). Does green pay Journal of Investing, 30(1), 67–87. https://doi.org/10.3905/ off? Journal of Real Estate Portfolio Management, 14(4), joi.2020.1.155 385–400. https://doi.org/10.1080/10835547.2008.12089822 Bedart-Page, G., Demers, A., Tuer, E., & Tremblay, M. (2016). Nyikos, D., Thal, A., Hicks, M., & Leach, S. (2012). To LEED or Large Canadian public pension funds: A financial system not to LEED: Analysis of cost premiums associated with perspective. In Financial system review (pp. 33–38). Bank of sustainable facility design. Engineering Management Canada. Journal, 24(4), 50–62. https://doi.org/10.1080/10429247. Berk, J., & van Binsbergen, J. H. (2022). The impact of impact 2012.11431955 investing. Working paper. Phalippou, L. (2009). Beware of venturing into private equity. Betermier, S., Calvet, L., & Jo, E. (2022). A supply and demand Journal of Economic Perspectives, 23(1), 147–166. https:// approach to capital markets. Working paper. doi.org/10.1257/jep.23.1.147 Brown, G. R., & Matysiak, G. A. (2000). Sticky valuations, Phalippou, L., & Gottschalg, O. (2009). The performance of aggregation effects, and property indices. Journal of Real private equity funds. Review of Financial Studies, 22(4), Estate Finance and Economics, 20(1), 49–66. 1747–1776. https://doi.org/10.1093/rfs/hhn014 Del Guercio, D., & Hawkins, J. (1998). The motivation and Rossi-Hansberg, E., Sarte, P., & Owens, R. (2010). Housing impact of pension fund activism. Journal of Financial externalities. Journal of Political Economy, 118(3), 485–535. https://doi.org/10.1086/653138 Economics, 52, 293–340. https://doi.org/10.1016/S0304- Shleifer, A., & Vishny, R. (1997). A survey of corporate govern- 405X(99)00011-2 ance. The Journal of Finance, 52(2), 737–783. https://doi. Eichholtz, P., Kok, N., & Quigley, J. (2010). Doing well by org/10.1111/j.1540-6261.1997.tb04820.x doing good? Green office buildings. American Economic Stambaugh, R. F., & Taylor, L. A. (2020). Sustainable investing Review, 100(5), 2492–2509. https://doi.org/10.1257/aer.100. in equilibrium. Journal of Financial Economics, 142(2), 5.2492 550–571. https://doi.org/10.1016/j.jfineco.2020.12.011 Fuerst, F., & McAllister, P. (2010). Green noise or green value? Wahal, S. (1996). Pension fund activism and firm perform- Measuring the effects of environmental certification on ance. The Journal of Financial and Quantitative Analysis, office values. Real Estate Economics, 39(1), 45–69. https:// 31(1), 1–23. https://doi.org/10.2307/2331384 doi.org/10.1111/j.1540-6229.2010.00286.x Zerbib, O. (2022). A sustainable capital asset pricing model Geltner, D. (1989). Estimating real estate’s systematic risk (S-CAPM): Evidence from green investing and sin stock from aggregate level appraisal-based returns. Real Estate exclusion. Review of Finance. https://doi.org/10.1093/rof/ Economics, 17(4), 463–481. https://doi.org/10.1111/1540- rfac045 6229.00504 Gyourko, J., & Keim, D. B. (1992). What does the stock market tell us about real estate returns? Journal of the American Appendix A Real Estate and Urban Economics Association, 20(3), 457–485. This Appendix presents additional empirical results and pro- Gupta, A., & Van Nieuwerburgh, S. (2021). Valuing private vides more details about the benchmark estimation method. equity investments strip by strip. The Journal of Finance, 76(6), 3255–3307. https://doi.org/10.1111/jofi.13073 Estimation of Benchmarking Method Huffman, F. (2002). Corporate real estate risk management In the estimation of our benchmarking methodology, the and assessment. Journal of Corporate Real Estate, 5(1), weight in the bond portfolio b is set such that the fund’s 31–41. https://doi.org/10.1108/14630010310811984 reported return R has a beta of one with respect to the i, t Kaiser, R. W. (2005). Analyzing real estate portfolio returns. benchmark return R for a given combination of B, i, tl=D The Journal of Portfolio Management, 31(5), 134–142. parameters g, s, and l: https://doi.org/10.3905/jpm.2005.593896 We define b as the univariate beta of the fund’s real Kats, G. H. (2003). Green building costs and financial benefits. estate portfolio with respect to the benchmark. By definition, Massachusetts Technology Cooperative. b satisfies Ling, D., Wang, C., & Zhou, T. (2020). A first look at the impact of COVID-19 on commercial real estate prices: covðR , R Þ i, t B, i, tl=D b ¼ Asset-level evidence. The Review of Asset Pricing Studies, varðR Þ B, i, tl=D 10, 669–704. https://doi.org/10.1093/rapstu/raaa014 s s N ER , R ER Þ E ðR i, t i, t Lipshitz, C., & Walter, I. (2020). Public pension reform and the B, i, tl=D B, i, tl=D 49th parallel: Lessons from Canada for the U.S. Financial 2 2 s s N ER ER B, i, tl=D B, i, tl=D Markets, Institutions & Instruments, 29(4), 121–162. https:// doi.org/10.1111/fmii.12133 94 A. D. BEATH ET AL. where N corresponds to the number of annual observations Table A.1. Green value-added activity by pension funds. ðÞ of the return R and EX is the expectation operator. i, t (A) LEED breakdown of all Preqin deals Solving for b ¼ 1 implies Investment type s s N ER , R  E ðR Þ E ðR Þ i, t i, t Basic Premium LEED-VA B, i, tl=D B, i, tl=D Sector (%) s 2 s 2 ¼ ER ER : Hotel 1.00 0.00 0.00 B, i, tl=D B, i, tl=D Industrial 0.95 0.00 0.05 Land 1.00 0.00 0.00 For a given combination of parameters g, s, and l, solving Mixed use 0.56 0.10 0.34 for the remaining parameter b amounts to solving the quad- Niche 1.00 0.00 0.00 ratic equation above. If no solutions are found, there is no Office 0.51 0.22 0.27 Residential 0.94 0.03 0.03 benchmark such that the fund has a beta of one for the Retail 0.80 0.07 0.13 combination of parameters g, s, and l: When one or more Geography (%) solutions are found, we select the values of the parameters Canada 0.59 0.00 0.41 US 0.52 0.18 0.30 b, g, s, l that maximize the correlation between the fund Other 0.95 0.02 0.03 return R and the benchmark return R : i, t B, i, tl=D (B) Geography distribution of all LEED projects Canada US Other LEED Activity around the World Weighting of projects (%) Number 0.02 0.81 0.17 In the main text, we use LEED certification activity as a proxy Square-footage 0.01 0.66 0.33 for green value-added investments. We now show that LEED This table reports aggregate statistics about LEED green building certifica- certification varies by sector and geography in our sample of tion. (A) Shows the LEED breakdown of all deals in Preqin made by large pension funds between 2010 and 2018. For each industry sector and Preqin deals involving a pension fund. geography, we report the percentages of properties that are not LEED Table A.1A reports statistics about LEED green building certified (“basic”), properties that were already LEED-certified prior to the certification on Preqin deals made by all large pension funds acquisition by a pension fund and did not receive additional certification afterwards (“premium”), and properties where pension funds actively between 2010 and 2018. For each industry sector and geog- sought some form of LEED certification after having acquired the property raphy, we report the percentages of properties that are not and before disposing it (“LEED-VA”). (B) Reports the geography distribu- LEED certified (“basic”), properties that were already LEED- tion of all projects in the USGBC database. The distribution is weighted either by the number of projects of their square footage. certified prior to the acquisition by pension funds and did not receive additional certification afterwards (“premium”), and properties where pension funds actively sought some Table A.2. Types of value-added activity. form of LEED certification after having acquired the property Canadian funds RoW funds and before disposing it (“LEED-VA”). We see that mixed-use Type of LEED-VA (%) and office-space properties in Canada and the U.S. are the Registration only 0.17 0.08 Certification only 0.41 0.58 most likely to qualify as premium or LEED-VA. Reg and Cert 0.42 0.35 Table A.1B reports the geography distribution of all LEED This table reports the different types of LEED value-added activity by projects reported in the USGBC database. The distribution is Canadian and Rest-of-World (RoW) funds between 2010 and 2018. Types weighted either by the number of projects or their square include LEED registration only, LEED certification only, and LEED registra- tion and certification. footage. We see that LEED projects are predominantly based in the U.S. certification. For both Canadian and RoW funds, the bulk of In Table A.2, we break down the different types of LEED value-added projects goes beyond registration and generally value-added activity: LEED registration only, LEED certifica- involves the full LEED certification process. tion only, and the combination of registration and

Journal

Journal of Sustainable Real EstateTaylor & Francis

Published: Dec 31, 2022

Keywords: Active management; impact investing; pension funds; real estate; sustainable finance

References