The Return Volatility of Publicly and Privately Owned Real Estate
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
The Return Volatility
of Publicly and
Privately Owned
Real Estate
The marketless value THE D E B AT E ABOUT whether pri-
vately owned real estate has a lower return
of real estate. volatility than its publicly owned counter-
part has been going on for a long time. It
began when academics analyzed the then-
newly compiled National Council of Real
Estate Investment Fiduciaries (NCREIF)
return series for unlevered real estate prop-
erties in the 1980s. These studies found
that the returns derived from the
NCREIF index had extremely low volatil-
ity, as well as very low correlations with
stock and bond returns. This led some to
PETER LINNEMAN proclaim that privately owned real estate
REVIEW 5was a “can’t lose” investment, providing Instead, the NCREIF total return in 1990
portfolio diversification with almost no was 2.3 percent, with negative returns reg-
return volatility, and average returns not istered only in 1991 (-5.6 percent) and
much different from stocks. Efficient 1992 (-4.3 percent). These numbers were
investment frontier research seeped into in stark contrast to the reality experienced
investment practice, which suggested that by private property owners. Since the dis-
privately owned real estate should com- connect between NCREIF and the market
prise almost 100 percent of an “efficient” reality was too great to be explained by dif-
portfolio. ferences in property or management qual-
As the 1990s dawned, investors ity, savvy observers quickly realized that
assumed that high-quality privately owned the NCREIF data was at best problematic,
real estate could never fall substantially in and at worst bogus.
value. But the early 1990s demonstrated The early 1990s witnessed the emer-
that private real estate assets had substan- gence of major publicy owned REITs.
tial return volatility, with values eroding by Faced with the complete absence of debt,
20 percent to 50 percent during the first and plummeting property values, many
half of the decade. As the chairman of major private owners of real estate went
Rockefeller Center, one of the nation’s public in order to recapitalize their proper-
prime core assets in the mid-1990s, I dis- ties, paying off maturing debt with IPO
covered all too well the volatility of private proceeds. Overnight, high-quality real
real estate returns. Yet, the NCREIF data estate was exposed to public market scruti-
failed to reveal significant negative returns. ny and pricing, with many of the finest
Figure 1: NCREIF Quarterly Returns
6 ZELL/LURIE REAL ESTATE CENTERproperty portfolios trading on the NYSE. properties, with returns to managers cap-
These REITs were managed by real estate tured by the asset management companies.
professionals with expertise equal to, or Returns to REITs, as proxied by the
better than, NCREIF asset managers. National Association of Real Estate
Thus, any differential return performance Investment Trusts (NAREIT) Equity
between the NCREIF index and REITs Index, and NCREIF should be highly cor-
could not be attributed to either product related, as the key determinant of their
or management quality. returns is the profitability of core quality
A third investment vehicle, real estate properties. REITs offer institutional
private equity funds, evolved to help equi- investors a transparent and liquid real
tize real estate investments. These non- estate investment alternative to direct own-
traded limited partnerships, with seven-to- ership. Since the mid-1990s, funds have
ten-year investment lives, use high lever- flowed into REITs from traditional core
age, frequently own foreign and lesser- real estate managers, causing many core
quality properties, and often pursue devel- managers to go out of business.
opment/redevelopment strategies. While Institutional investors were not only disap-
the returns for these vehicles are not sys- pointed in the performance of real estate
tematically reported, it is not be surprising versus their expectations, but also angry
that the return history for these higher- with managers who hid how badly their
risk, lower-quality, value-add investment investments performed.
vehicles substantially diverges from that of Yet, based upon studies of NCREIF
either NCREIF or REITs, as their returns returns, many researchers, managers, and
do not generally reflect the performance of investors continue to believe that privately
core U.S. properties. owned real estate has almost no correlation
with the returns of REITs, stocks, or
bonds. For example, from the first quarter
NCREIF VS. NAREIT of 1990 through the first quarter of 2004,
the correlation of NCREIF’s return with
NCREIF index properties are unlevered, that of NAREIT was -0.04, while with
while REIT’s are approximately 50 percent S&P 500 it was 0.01, and with long bonds
levered. In addition, REITs own both the -0.10. In addition, the standard deviation
company’s real estate as well as the profit of quarterly returns for this same period
stream generated by management (net of was 3.4 percent for NCREIF, versus 10.5
executive compensation). In contrast, the percent for NAREIT, 11.3 percent for
NCREIF index reflects only returns on S&P 500, and 6.9 percent for long bonds.
REVIEW 7As a result, many observers argue that part of a REIT’s valuation), or as a result
institutional investors should own real of leverage, or because short-term capital
estate both publicly and privately, with movements are insufficient to arbitrage
publicly owned real estate providing liq- public versus private pricing. However,
uidity, and privately owned real estate pro- the return differences between NCREIF
viding return stability and diversification. and NAREIT are not small, temporary,
But these results cannot be correct, as or occasional. In five of the past 14 years,
buildings are inanimate objects, which the annual returns for NCREIF and
do not know whether they are publicly NAREIT are of opposite signs.
or privately owned. Further, most core Moreover, the average absolute difference
properties are managed by high-quality in the annual returns of these series is a
managers, whether the properties are staggering 1715 basis points, with this
publicly or privately owned. Therefore, difference being fewer than 600 basis
large return discrepancies between public points in only two years. For example,
and private real estate ownership are not 1998 REIT returns were shocked by the
theoretically credible. Of course, minor Russian ruble crisis, yielding a -17.5 per-
return differences between public and cent return, while the NCREIF return
private real estate can arise due to the val- was 16.2 percent, a gap of 3370 basis
uation of management teams (which is a points!
Figure 2: Annual Total Returns
8 ZELL/LURIE REAL ESTATE CENTERFigure 3: Return Disparity between NAREIT and NCREIF (Basis Points)
W H AT ’ S GOING ON?
cash streams. While anomalies can exist,
The best research on the differences they will be arbitraged, particularly given
between NCREIF and NAREIT returns the large number of opportunity funds
has been conducted by Joe Gyourko, first with the broad mandate to simply generate
in a paper with Don Keim, and more risk-adjusted real estate returns. If return
recently in a paper in the Wharton Real differences are consistently as divergent as
Estate Review. He finds that NCREIF these series indicate, there should be no
returns are predictable based upon historic shortage of “smart money” to arbitrage the
REIT returns. Specifically, today’s REIT differences. In addition, REITs’ property
returns foretell the NCREIF returns that acquisitions and dispositions would arbi-
will be registered roughly 18 months from trage large differences in “Wall Street vs.
now. As has been stated before, since Main Street” values. Yet, during the past
buildings are inanimate, and since their 14 years, in spite of the extraordinary dif-
quality of management is roughly similar, ferences in NCREIF and NAREIT
this relationship cannot be due to signifi- returns, few REITs were taken private,
cant differences in property level cash very few major positions in REITs were
flows, risk profiles, or management. taken by opportunity funds, and almost
One need not be a believer in perfectly no REITs liquidated their portfolios.
efficient markets to feel that it is incon- The primary reason why large return
ceivable that capital markets so inefficient- discrepancies between NCREIF and
ly value public versus private real estate NAREIT exist is simple: the data are
REVIEW 9wrong. This was vividly demonstrated Many have noted the so-called
during the Russian ruble crisis, when appraisal lag in valuing NCREIF proper-
REITs fared terribly, while NCREIF regis- ties. However, the NCREIF return meas-
tered returns well above average. Yet urement problem is much deeper, as
almost no public to private arbitrage took most observers fail to appreciate how the
place, even though the return data indi- appraisal process, even when well done,
cate that such activity would have been generates meaningless valuations for
highly profitable. No opportunity funds evaluating return volatility and correla-
took advantage of the option suggested by tions. In fact, the appraisal process guar-
the data. Nor did entrepreneurial REIT antees that NCREIF’s appraisal-driven
operators see an opportunity to go private. returns will have very low volatility. Since
Instead, the market clearly believed that the appraisal process, rather than private-
there was no significant return differential market real estate pricing, creates near-
between public and private real estate. zero volatility in measured returns, near-
Like Sherlock Holmes’ famous “dog that zero return correlations with REITs,
didn’t bark,” the market’s silence demon- stocks, and bonds are no surprise, as
strated that the return gap is fiction rather these assets have considerable return
than reality. volatility. Specifically, if the returns for
stocks, bonds, and publicly traded real
estate are essentially random walks
VA L UAT I O N ISSUES reflecting relatively efficient market pric-
ing, NCREIF’s near-zero appraisal-
NAREIT pricing and returns reflect mar- induced volatility will necessarily show
ket pricing by third parties investing in little return correlation.
publicly traded securities, and thus have How does the use of appraised prop-
no notable measurement error. It is also an erty values produce this result? The vast
investable index, with several index funds majority of NCREIF properties have a
readily available for investors. In contrast, value appraisal only in the fourth quarter
the NCREIF index is neither investable of each year. This contrasts dramatically
nor a market-priced index. Specifically, it is with private property markets, where
impossible to create a portfolio that con- properties are constantly valued (though
tains the NCREIF properties, and not appraised) by owners. Opportunity
NCREIF property prices are very rarely set funds, entrepreneurial owners, and high-
by third-party investors. Instead, they are wealth families constantly evaluate their
established by appraisals. property values. Anyone who has worked
10 ZELL/LURIE REAL ESTATE CENTERwith these owners knows that, over the without accurately measuring quarterly
course of a year, the values of privately value changes. But if the preponderance
owned properties rise and fall, depending of core properties is appraised only in the
on leasing, market sentiments, rumors fourth quarter, the return registered for
about new developments, macroeco- NCREIF is by definition basically the
nomic hopes and fears, and capital mar- same in the fourth, first, second, and third
ket animal instincts. Many private own- quarters, as NOI does not change appre-
ers exploit these value movements by ciably quarter-by-quarter. The fact that
either selling or refinancing their proper- the recorded returns are basically the same
ties at opportune times, or by holding over a four-quarter period provides no
their properties while waiting for better information about whether the actual
market pricing. This is the reality of pri- quarterly returns were the same, and
vate markets. Companies such as Eastdil, merely reflects that no attempt was made
Secured Capital, and Goldman Sachs to determine whether asset prices changed
make their livings from the volatility of quarter-to-quarter. This is the first source
private real estate markets. of NCREIF’s return smoothing.
Consider what happens to a real estate Imagine what would happen if
return series if the value of a core proper- NAREIT quarterly returns were meas-
ty is recorded only on the last day of each ured simply by dividing the quarterly div-
year. Since a core property’s Net idend by the fourth quarter cap rate.
Operating Income (NOI) generally varies Since NAREIT dividends change rela-
relatively little throughout the year, so too tively little quarter-to-quarter, this
will the measured return if the property approach would record little REIT return
price remains unchanged for four quar- volatility. Gyourko’s research notes that
ters. For example, if the property has a during the first three quarters of the year,
quarterly NOI growth rate of 1 percent NCREIF returns register little volatility
(4.06 percent annual rate), and an 8 per- for private real estate. But savvy private
cent initial cap rate, the registered quar- owners know this is not the case. A point
terly returns absent quarter-to-quarter of reference is provided by (incorrectly)
price changes over the four quarters of the calculating NAREIT returns as the quar-
year are 2.02 percent, 2.04 percent, 2.06 terly dividend plus appreciation, divided
percent, and 2.08 percent, respectively. by the closing year-end stock price.
Hence, for NCREIF’s large pool of core Figure 4 reveals that this exercise notably
assets, it is almost impossible to have reduces the volatility of NAREIT’s quar-
much quarter-to-quarter return volatility terly returns.
REVIEW 11Figure 4: NAREIT Quarterly Returns
In fact, the standard deviation of When an “unbiased” appraiser (they may,
NAREIT quarterly returns falls from like unicorns, exist somewhere) is engaged,
10.5 percent for actual NAREIT returns their methodology for a core, stabilized
to 7.0 percent when this method is property is to divide “stabilized” NOI (a
employed. Thus, if investors want lower second smoothing) by the cap rate of recent
REIT return volatility, and to look more transactions for comparable properties.
like NCREIF’s, they should only look at The period for which the appraiser seeks
the REIT stock prices on the last day of comparable sales transactions is typically 24
each year! months. Over this 24-month period, the
appraiser will generally find five to eight
comparable property sales. The cap rate
QUADRUPLE SMOOTHING selected by the appraiser is usually the
mean (or sometimes median) of the cap
The NCREIF measurement error story rates for these transactions. Rarely do
does not end here, as NCREIF’s fourth- appraisers give more weight to more recent
quarter value generally reflects appraised— transactions, or evaluate cap rate trends.
as opposed to market—property prices. To Thus, although each comparable property
see how the appraisal process even further sold at a specific cap rate, at a specific time,
undercuts return measurement efforts, one the appraisal cap rate is an average (a third
must understand the appraisal process. source of smoothing), which eliminates the
12 ZELL/LURIE REAL ESTATE CENTERhigh and low valuations that existed in the The nature of the appraisal process
market; that is, it eliminates pricing volatil- means that even in the fourth quarter, the
ity. The appraiser’s rationale for cap rate registered NCREIF value fails to reflect the
averaging is that the market conditions that prevailing market pricing, reflecting
existed when those properties were sold are instead the average of market conditions
better or worse than those that exist “in that prevailed over the preceding two
more typical markets.” Further, the years. In fact, what NCREIF records as
appraisal for a property this year will gener- “today’s cap rate” is actually the mean cap
ally re-use approximately half of the com- rate about 12 months earlier; that is, at the
parable sales transactions of the previous midpoint of the appraiser’s time period.
year’s appraisal. Thus, this year’s cap rate is Hence, NCREIF fourth-quarter valua-
mechanically linked to last year’s cap rate, tions effectively reflect “stabilized” NOI
introducing a fourth source of NCREIF divided by the average cap rate a year ago.
return smoothing. Quadruple-smoothed, with a lag.
Note that the appraisal methodology This appraisal smoothing and lag not
adopts the position that while higher and only reduces measured return volatility,
lower cap rates than average existed, they but also almost necessarily eliminates any
are of no relevance to a property’s correlation with market return series.
appraised value. In fact, every property This is because if actual returns follow a
was transacted at a cap rate higher or random walk, inducing a one-year lag,
lower than the average, indicative of then- the lagged series is uncorrelated with the
current market conditions. Stated bluntly, original series, as the lag wipes out all cor-
the appraisal process eliminates—not relation with all random series. Since
reveals—the truth about how properties stock, bond, and REIT returns have been
are priced in private markets. In effect, the shown to basically follow random walks,
appraised value, far from being the market even if true private real estate returns were
value, is, in effect, a marketless value. That highly correlated with these series, the
is, it is a value net of the vagaries of the NCREIF appraisal lag would wipe out
market. Tellingly, no property is ever the correlation. The impact of lagging is
bought or sold at the appraiser’s cap rate. vividly demonstrated by the fact that the
Yet many researchers use the NCREIF correlation between quarterly S&P
return as if the cap rates used to value its returns and the eight-quarter moving
properties reflect private property market average of S&P returns is a mere 0.16
prices. But by design, this is absolutely not from 1990 through the first quarter of
the case. 2004. Thus, a series, which by definition
REVIEW 13is perfectly correlated with itself, is basi- average NAREIT price for the preceding
cally uncorrelated with a NCREIF-like eight quarters. This smoothing and lag-
lagged version of itself. ging reduces the standard deviation of
The fact is that while NAREIT returns NAREIT quarterly returns from 10.5 per-
reflect actual returns for an investable pub- cent for actual returns, to 4.9 percent.
lic real estate portfolio, NCREIF returns Conducting the same analysis with
measure nothing remotely like actual NAREIT price calculated as the NAREIT
returns for a core private portfolio. The dividend divided by the eight-quarter
12-month valuation lag is accentuated NAREIT moving average cap rate results
over the subsequent three quarters, as in an estimated REIT quarterly return
NCREIF’s appraised values are generally standard deviation of 5.7 percent.
not changed during these quarters. Thus, We also recalculate quarterly NAREIT
the cap rate used in the appraisals is ini- returns, where for the fourth quarter the
tially four quarters out of date, falling to return is the actual NAREIT dividend
five, six, and seven quarters over the next divided by the moving average NAREIT
three quarters. As the property values are cap rate for the preceding eight quarters,
reappraised in the fourth quarter, the lag plus the percentage increase in that price
once again reverts to four quarters, and the over the similarly calculated price of a year
process repeats. It is hardly surprising that earlier, where for the first, second, and
Gyourko’s research consistently finds a third quarters there is no price change. The
roughly 18-month statistical relationship standard deviation of NAREIT quarterly
between REIT returns (actual market pric- returns for this approach is 7.7 percent.
ing) and NCREIF’s lagged returns. How do these results compare to
NCREIF? Recall that NCREIF is unlev-
ered, while NAREIT is roughly 50 percent
IT’S O N LY REAL E S TAT E levered. To adjust for the different leverage,
we calculated quarterly returns for a 50
To see how quadruple-smoothing-and-a- percent leveraged NCREIF, at an interest
lag mechanically affects measured returns, rate of three-year Treasury plus 150 basis
we calculate a variety of incorrectly meas- points. The standard deviation of this lev-
ured quarterly REIT returns. First, for ered NCREIF series is 5.2 percent (versus
each quarter, the quarterly return is calcu- 3.4 percent for unlevered NCREIF). This
lated as the sum of the actual NAREIT levered NCREIF return volatility com-
dividend plus percentage price apprecia- pares to 10.5 percent for actual NAREIT,
tion, where price is defined as the moving and 5.7 percent and 7.7 percent for the
14 ZELL/LURIE REAL ESTATE CENTERsmoothed NAREIT series. Plus, one must tively measured. Further, smoothed
remember that the use of actual, rather NAREIT returns have a much higher
than “stabilized,” NAREIT dividends in (though still low) correlation with
these calculations makes NAREIT returns NCREIF.
more volatile than levered NCREIF, which The best series to measure real estate
uses “stabilized” NOI. Thus, when com- returns is neither NCREIF nor the
pared on an apples-to-apples basis, the smoothed NAREIT series, but rather the
return volatilities of NCREIF are NARE- actual REIT return series. This is because
IT are basically the same. mark-to-market, contemporaneous, arm’s-
The correlation coefficients of the length, non-smoothed pricing is the reality
alternative quarterly return series are dis- of both public and private real estate. The
played in Figure 5. Note that smoothed truth is that modestly leveraged core real
and lagged NAREIT returns, like both estate has a low correlation with stocks and
NCREIF and unlevered NCREIF, show bonds, but displays notable return volatili-
little correlation with the S&P 500 or long ty, though somewhat less so than stocks.
bonds. This is because the non-volatility No one involved in private real estate
induced by quadruple-smoothing-and-a- markets will find these results surprising.
lag correlates to almost nothing. In con- After all, real estate ownership of core
trast, the mean returns for the various assets incorporates many of the dimen-
NAREIT returns are only slightly altered sions of high-quality bonds, with superior
by these smoothing calculations, because residual value protection because it is a real
the time shifting only slightly alters the (rather than nominal) asset. The result is
time period over which returns are effec- that non-residential real estate has less cash
Figure 5: Quarterly Return Correlation Coefficients
Quarterly Return Correlation Coefficients
Smoothed Smoothed Unlevered Levered S&P 500 Long Bonds
NAREIT1 NAREIT2 NCREIF NCREIF
Acutal NAREIT -.16 -.21 -.05 -.05 .40 .08
Smoothed NAREIT 1
.55 .14 .15 -.04 -.10
Smoothed NAREIT2 .14 .14 .12 -.24
Unlevered NCREIF .99 .01 -.10
Levered NCREIF .01 -.10
S&P 500 -.09
1 Based upon eight-quarter moving average NAREIT price.
2 Based upon eight-quarter moving average NAREIT price, with estimated price unchanged for four consecutive quarters.
REVIEW 15stream volatility than the equity and debt ship. Specifically, this volatility allows
claims on its tenants, since in good times, institutional owners of REITs to take
tenants expand their space demand more advantage of momentary mispricings of
slowly than their profits increase, while in their stocks by selling or shorting when
bad times the reverse is true. In addition, prices are “too high,” and incrementally
the supply and demand fundamentals of buying when prices are “too low.” Such
real estate follow unique patterns, further trading provides an additional margin for
diminishing the return correlations with well-capitalized institutional investors to
other assets. Similarly, residential real exploit temporary pricing anomalies. Of
estate follows its own time patterns, as course, just as is the case for private real
even in poor economic times, population estate, if one is convinced that REIT prices
increases and absorption generally occurs, are too high, one can sell one’s entire posi-
although at a slower rate. Also, supply and tion. Similarly, if one believes that prices
demand patterns for these properties move are too low, one can hold stocks until
differently from other asset categories. prices rise. Publicly owned real estate
allows investors to incrementally alter their
investment position when they believe
D AY- T O - D AY pricing is too low or too high. In fact,
R E I T VO L AT I L I T Y aggressive institutional investors may go so
far as to short assets when they believe
As is the case with a publicly traded secu- prices are too high. Thus, far from being a
rity of any company, one is struck by the negative aspect of public real estate invest-
fact that in any one-hour or one-day trad- ment, the presence of micro price volatili-
ing period, the price of a REIT (or any) ty can only benefit well-capitalized long-
stock can go up or down by several per- term investors. At worst, the institutional
centage points, for no apparent reason. It is investor can simply ignore such pricing
true that privately owned real estate does variability, and simply trade out of their
not have such minute-to-minute, hour-to- holdings on last day of each quarter, in
hour, or day-to-day price volatility, as deals which case they realize NAREIT returns.
are not struck in private markets on such
an instantaneous basis. The presence of
such price volatility for REITs means W H AT ’ S IT ALL MEAN?
investment opportunities are available via
the public ownership of private real estate There is no magical potion in the private
that are not present with private owner- ownership of core real estate that elimi-
16 ZELL/LURIE REAL ESTATE CENTERnates return volatility and correlation with
other assets. The fact that investors contin-
ue to believe this is the case reflects either a
fundamental misunderstanding of the
NCREIF data, or purposeful ignorance
about the realities of real estate markets.
The truth is that high-quality, stabilized
real estate should be a major part of insti-
tutional investors’ portfolios, and that
public ownership provides the same long-
term return patterns as private ownership,
with the enhanced advantages of exploit-
ing temporary mispricings and liquidity.
Core real estate provides solid long-
term returns, somewhat lower volatility
relative to stocks, and relatively modest
correlation with the returns on other
assets. However, the purported advantages
of private core real estate ownership are a
mirage. What matters are the quality of the
property and the ability of the manager to
execute a viable operating strategy,
whether public or private.
Private core real estate ownership for
many institutions is a narcotic that creates
the “comfortably numb” illusion of non-
volatility in a harsh and demanding mark-
to-market world. It is one of the few
remaining assets where you can pretend
that your assets have not changed in price,
even when they have. Hopefully, such illu-
sions will soon be a thing of the past.
REVIEW 17You can also read
