The research register for this journal is available at The current issue and full text archive of this

of this journal is available at

http://www.emeraldinsight.com/researchregisters http://www.emeraldinsight.com/1463-578X.htm

ACADEMIC PAPERS Academic papers:

Real time
Real time valuation valuation

Jeffrey D. Fisher
Dunn Professor of Real Estate, Director, Indiana University Center for 213
Real Estate Studies, Bloomington, USA
Keywords Valuations, Appraisal, Reliability, Property portfolio
Abstract The purpose of this paper is to stimulate thinking as to how we might produce timely
and more reliable estimates of changes in the value of portfolios, price indices based on a portfolio
of properties, and other aggregate measures of trends in property values. It is argued that a
traditional market value appraisal of each individual property may not be necessary or optimal
when the objective is to value portfolios or get a leading indicator of shifts in market value at an
aggregate level. Rather, it is more important to use a critical mass of current market data that
captures systematic movements in property values. Although a traditional market value appraisal
is always more likely to capture the unique unsystematic characteristics of an individual property,
automated valuation models using a database of valuation data may provide the best way to get
real time interim updates of real estate portfolios and create more timely real estate indices.

Introduction
Real estate is usually considered an ``asset class'' that should be included in a
multi-asset portfolio along with stocks and bonds. Academic research suggests
that real estate investments are not highly correlated with stocks and bonds
and offer diversification benefits. The size of the real estate asset class is such
that even a naõÈve diversification strategy of including assets in proportion to
their weight in the ``market portfolio'' would suggest institutional investors
such as pension funds should include a significant amount of real estate in their
portfolios. Yet most pension funds have a relatively small amount of real estate
in their portfolios compared to stocks and bonds.
There are several possible explanations for the low proportion of real estate
that institutional investors actually hold. One answer is certainly the lack of
liquidity of private real estate investments compared with stocks and bonds.
But to the extent that there is a liquidity risk premium in expected returns for
real estate investments, investors who do not necessarily need liquidity for the
real estate portion of their portfolio can benefit by earning this risk premium.
At a recent meeting of a Pension Fund Advisory Committee of the National
Council of Real Estate Investment Fiduciaries (NCREIF)[1] the participants
pointed out that one of the major problems with real estate is that information
regarding the performance of real estate is not available as quickly as it is for
stocks and bonds. Whereas the performance of publicly-traded stocks and
bonds can be tracked on a real-time basis, information on the performance of
real estate is at best on a quarterly basis when the property is appraised[2]. Journal of Property Investment &
Finance, Vol. 20 No. 3, 2002,
pp. 213-221. # MCB UP Limited,
This paper was originally prepared for presentation at the World Valuation Conference, 1463-578X
26 April 2001, Singapore. DOI 10.1108/14635780210433463
JPIF Furthermore, due to the nature of the appraisal practice, changes in appraised
20,3 values tend to lag changes in transaction prices. This results in the well known
``appraisal smoothing'' problem with published real estate indices such as the
NCREIF Property Index (NPI).

Causes of appraisal smoothing

214 One of the most common methods of appraisal for a non-traded property is the
use of comparable sales evidence (Brown and Matysiak, 2000). Such an
approach is backward-looking. For this reason, the estimate of the market
value of a property may not necessarily reflect the current market information.
Quan and Quigley (1991) provide a rational appraiser behavior ± that the
optimal updating strategy for appraisers is to use the weighted average of the
previous appraisal and the most recent comparable transaction prices.
Smoothing in valuations arises from the relative uncertainty in general market
prices and individual transactions. As the two components change, so do the
weights given to the previous appraisal and the most recent comparable
transaction prices.
The problem of valuation smoothing is perhaps more acute in the
construction of commercial property return indexes (Clayton et al., 2001). As
Clayton et al. (2001) point out, there are two sources of smoothing at the
aggregate level ± index construction and systematic errors in the appraisals of
individual properties. The problem of stale appraisal afflicts index construction
and was examined by Fisher et al. (1994). At the individual property level, there
are two sources of errors. The first is purely random and is unique to the
specific property. The second is a systematic error that is common across all
appraisals at any given time. Clayton et al. (2001) attribute the second source of
error to rational appraiser behavior (Quan and Quigley, 1991) and the
anchoring effect (Diaz and Wolverton, 1998), and find evidence to support the
second source of error.
A slightly different perspective is provided by Brown and Matysiak (1997).
Based on the earlier work by Working (1960), they contend that the
aggregation of random returns can result in serial correlation in the index. In
addition, Brown (2001) and Brown and Ong (2001) recognize that returns
between properties in a portfolio could be correlated at any point in time. This
is somewhat analogous the systematic error pointed out by Clayton et al. (2001).

Appraisal process for ``market value'' of individual properties

The lag in appraised values versus transaction prices in indices such as the NPI
does not mean that there is a problem with the traditional appraisal process as
a way of estimating the ``market value'' of an individual property at a given
point in time. Appraisal standards require a process that includes sufficient
evidence to have a relatively high degree of confidence in the estimated value of
an individual property that requires reconciliation of all applicable approaches
to value, e.g. cost, sales comparison and income approaches.
 The appraisal process is still largely based on gathering information from Academic papers:
comparable sales that are by nature historical, and using discounted cash flow Real time
techniques that require some degree of time to gather the necessary valuation
information about existing leases, market rents for lease renewals, vacancy
rates, expenses, appropriate discount rates, etc. Information obtained by
appraisers on comparable sales, rents from recent leases, etc. is always subject
to a certain degree of error for several reasons. For example, we know that any 215
given ``price'' from a comparable sale is not necessarily indicative of ``market
value'' because real estate trades in a relatively inefficient market and prices
resulting in a particular transaction depend on the relative negotiating strength
of the buyer and seller. Furthermore, properties are heterogeneous and despite
attempts to adjust for differences between the comp and the subject property, it
is impossible to account for every difference. Thus, when estimating the market
value of an individual property it is desirable to have information from several
comparable sales to reduce the noise. In other words, we do not want to rely
solely on the most recent comparable sale or most recent lease transaction. It
may indicate a systematic shift in the market or it may have resulted from
unsystematic noise due to the characteristics of the particular transaction.
Thus other comparables are selected in addition to the most recent comparable
sale.

Valuation of portfolios and construction of price indices

Portfolios are typically valued by simply aggregating the appraised value for
the individual properties in the portfolio. This is also true when the ``portfolio''
is the properties in an index like the NPI. As discussed above, the traditional
appraisal process attempts to reduce the noise in individual property
appraisals, e.g. by averaging over several comparable sales. But when
individual appraisals are aggregated for a portfolio, we have seen that there
will also be a reduction in random noise by averaging the individual property
appraisals. But if we are able to reduce error or noise in the valuation of a
portfolio by averaging value estimates for individual properties, it may not be
desirable to use individual property appraisals that are based on averages of
comparable sales or lease transactions that are not the most recent
transactions. That is, although we do not want to just rely on the most recent
comp when appraising an individual property, we may want to just rely on the
most recent comp when the purpose is to aggregate the value estimates for
individual properties. This suggests that we want to use a different process to
estimate the value of individual properties when the real purpose is to value a
portfolio or create an index based on a portfolio of properties.
When the objective is to have the best estimate of the value of a portfolio (or
index), we do not necessarily want to start with the ``market value'' of the
individual property as it is defined in the appraisal literature. What we want is
information related to that property that is most likely to indicate current shifts
in the market, i.e. only the most recent comparable sale and only the most
recent lease on the property. Although limiting information in this way could
JPIF result in an unreliable estimate of value for a single property, it is likely to give
20,3 a more current and unbiased estimate of the value of a portfolio of properties.
In fact, if the purpose is valuation of a portfolio, an estimate of the market
value of each individual property is not really necessary. Although we want a
lot of the same market data that would be used to estimate the market value of
an individual property, we want to select those data that would result in the
216 best indication of the value of the portfolio. For example, suppose for a
particular property the appraiser had two comparable sales, one that was last
week but was not located as close to the subject property as another comp that
sold a month ago. For purposes of portfolio valuation or index construction, we
might only want to use the comp that sold last week. Figure 1 shows a
simplified illustration of this concept, i.e. comps chosen for valuation of an
individual appraisal may differ from comps chosen for valuation of a portfolio
or construction of an index.

Trade-off between number of comps and appraisal lag

Geltner and Ling (2001) provide a framework to examine the trade-off between
the two types of errors in appraisals ± estimation error and temporal lag bias.
Estimation error refers to the purely random error in comparable sales that is
by definition unique to the specific property, as noted by Clayton et al. (2001).
Temporal lag bias refers to the change in market conditions since the last
transactions were observed. Geltner and Ling show that there is a trade-off
between reduction in random estimation error versus reduction in temporal lag
bias property value estimation. They analyze this trade-off in a diagram in the

Figure 1.
Selection of comps for
individual property
valuation versus
portfolio valuation
spirit of the production possibility frontier where the client's utility function Academic papers:
can be maximized by minimizing the appraisal error. Real time
Figure 2, panel A illustrates this trade-off. To reduce random estimation valuation
error, more empirical value observations (comps) are required. But to obtain
more comps, transactions must be taken from a longer span of history, i.e.
comps that transacted further back in time from the current valuation data.
The convex shape to the curve is due to the fact that random estimation error 217
decreases by a factor of the reciprocal of the square root of the sample size[3].

Figure 2.
Trade-off between
random estimator error
and temporal lag bias
JPIF Point A is the optimal individual property value estimation (an appraisal
20,3 maximizing the client's utility of the appraisal). It has a certain amount of
random error and a certain amount of temporal lag bias. However, in the value
estimation of aggregated properties, only the common element in the aggregate
needs to be duplicated in the comps sample, providing many more relevant
empirical value observations per unit of historical time. This means that there
218 will be less random error for a given historical lag or less historical lag for a
given random error.
Panel B in Figure 2 illustrates the trade-off between random error and
temporal lag for a portfolio. We see that simply adding comps to reduce
random error does not result in a utility maximizing solution. Random errors
``diversify out'' at the aggregate (index) level and this pushes the accuracy
trade-off frontier out at the aggregate level. The main implication is that simple
aggregation of value estimates that were optimized at the disaggregated
individual property level will not produce an estimate of value that is optimal
at the aggregate index or portfolio level. Simple aggregation leads to point B in
panel B, whereas point C maximizes utility.
The implication of this analysis is that when valuing an aggregation of
properties, e.g. a portfolio, it is optimal to find ways to take advantage of the
natural reduction in random noise in a portfolio in order to reduce temporal lag.
This has implications for how comps should be selected for valuation of a
portfolio versus for valuation of individual properties. It also has implications
for the benefits of using ``mass appraisal'' techniques to value commercial real
estate portfolios.

Measuring value trends

The above discussion suggests that the process of gathering market
information, whether comparable sales, lease transactions or other market data,
should be different when the purpose is to determine changes in the value of a
portfolio. The term ``portfolio'' can be viewed more broadly than a portfolio that
is owned by a particular investor. A portfolio can be the properties in an index
such as the NPI. A portfolio can also be all properties in a particular market or
sub-market, or all properties in a particular sector such as all office properties
in a sub-market. The point is that the same concept applies to determining the
best way to measure trends in market values. If, for example, we want to know
how market values are changing for Chicago office properties, we do not want
to aggregate appraisals for individual office properties in Chicago. We want
selected information that may be part of the data that would be used to
appraise the individual properties, but we will process it differently to measure
market trends or sub indices measuring changes in property values.

Appraisal smoothing and portfolio risk

It is well known that holding a diversified portfolio can reduce a lot of the
unsystematic (property specific) risk. So the risk faced by institutional
investors is that systematic shifts in a particular market affect their portfolio,
e.g. the unexpected decrease in market rents and property values for Silicon Academic papers:
Valley office space due to the technology fallout over the past year. How has Real time
the decrease in values affected the value of a particular portfolio? How has the valuation
decrease in values affected the contemporary loan-to-value ratio of loans that a
particular lender might have in Silicon Valley? It is important for investors and
lenders to get timely information about shifts in market value to make
adjustments in portfolio allocations or better service the loan portfolio (e.g. 219
looking for potential defaults). Again, obtaining appraisals on individual
properties and aggregating the results may be too time consuming to get the
information fast enough to make decisions, and even if it could be done quickly,
may not be the optimal way to measure the true systematic change in market
values, for the reasons discussed earlier.

Toward real time valuation of commercial real estate

The previous discussion suggests that for portfolio valuation, index
construction, identifying market value trends, and portfolio risk analysis (debt
and equity) we need a process that differs from aggregating the value of
individual appraisals because it is too time-consuming and not the optimal way
to measure aggregate value trends. What is needed is a process of gathering the
most recent, ideally real time information on comparable sales, lease
transactions, etc. and processing this information in a different way than we
would if we were estimating market value for an individual property.
We have seen that when measuring aggregate value trends it is not
necessarily desirable to first determine the value of each individual property
using all the data that would normally be used in an individual appraisal. This
was illustrated earlier by suggesting that only the most recent comparable
sales would be used for estimates of aggregate value trends. This would be in
the context of a sales comparison approach. Automated valuation models
(AVMs) are now commonly used to estimate the value of residential properties
when the lender plans to hold a portfolio of properties. Although the AVM may
not provide the most accurate estimate of the value of an individual property,
especially if it is somewhat unique, the AVM may provide a better indication of
the value of properties that are to be held in a portfolio where unsystematic
property specific factors tend to average out.
Similarly, for commercial real estate, an AVM that draws on a database that
captures the most current data on sales, market rents, etc. may be appropriate
to use when the ultimate objective is to capture trends in the value of a portfolio
or an index of property value trends, such as the NPI.
For commercial real estate, a discounted cash flow analysis (income
approach) is often relied on to estimate the value of a property because it
captures the existing lease structure of the property. All three approaches
(income, sales comparison and replacement cost) are generally used, when
applicable, for a complete appraisal of an individual property. But again, if the
purpose of the value estimate is to get the best indication of changes in market
values for properties that are in a portfolio, for a commercial loan portfolio, or
JPIF to construct an index, capturing the most recent changes in market conditions
20,3 that systematically affect property values may be more important than a
traditional market value appraisal of each property. This is not to say that
periodic appraisals of individual properties are not important for portfolios for
many reasons, e.g. periodic accounting of portfolio values to investors. But a
complete appraisal for every property in a portfolio is not feasible more than
220 quarterly. If the objective is to get an estimate of changes in value on more of a
real time basis as is possible with stocks and bonds, we must rely on an
automated valuation model.
Until recently the technology was not available to gather information
quickly that could be applied to a portfolio of properties[4]. The data to do this
kind of analysis have traditionally been hard to obtain since ``private
information'' was critical to many market participants. Furthermore, the data
may be in disparate database systems. But in recent years, with the advent of
the Internet, information has become more of a commodity, with many Web
sites providing comparable sales data. Also, newer technologies such as
extensible mark-up language (XML) have started to allow data to be shared
across different databases and otherwise proprietary programs[5]. The Data
Consortium (www.DataConsortium.org) has developed standards for transfer
of appraisal and other real estate data across different platforms. This should
ultimately allow us to capture large amounts of current valuation data that will
make the use of automated valuation models feasible for commercial real
estate. As we have seen, valuation of portfolios can rely on just the most recent
data and still have a minimal amount of temporal lag. Thus it should be
feasible to eventually estimate the value for real estate portfolios on a more
frequent if not a real time basis. As Brown and Matysiak (1996) state: ``it is now
time for the market to consider introducing a real-time commercial property
index.''

Conclusion
The purpose of this paper has been to stimulate thinking as to how we might
produce timely and more reliable estimates of changes in the value of
portfolios, price indices based on a portfolio of properties, and other aggregate
measures of trends in property values. It is argued that a traditional market
value appraisal of each individual property may not be necessary or optimal
when the objective is to value portfolios or get a leading indicator of shifts in
market value at an aggregate level. Rather, it is more important to use a critical
mass of current market data that captures systematic movements in property
values. Although a traditional market value appraisal is always more likely to
capture the unique unsystematic characteristics of an individual property,
automated valuation models using a database of valuation data may provide
the best way to get real-time interim updates of real estate portfolios and create
more timely real estate indices.
Notes Academic papers:
1. NCREIF tracks the performance of properties managed by institutional investors on behalf Real time
of tax-exempt investors. For more information go to www.NCREIF.org
2. Most real estate investments held by institutional investors are not actually appraised by
valuation
an external appraiser each quarter. It is often externally appraised only once per year and
internally appraised during the other quarters. Although internal appraisal can be as
accurate as external appraisals, often the value is simply updated for accounting-related
information such as capital expenditures. 221
3. This is the same as the relationship between the standard deviation of the error in a
sample taken from a population, i.e. the standard deviation of the population divided by
the square root of the sample size.
4. Note that ``portfolio'' is now broadly defined to include indices, trends in property sectors
and submarkets.
5. Go to www.DataConsortium.com to see how the commercial real estate industry has been
working on data standards to allow transferring data across disparate database systems.

References
Brown, G.R. (2001), ``A note on the effects of serial cross correlations'', Journal of Property
Research, Vol. 18 No. 3, pp. 249-57.
Brown, G.R. and Matysiak, G.A. (1996), ``A real-time property index'', Estates Gazette.
Brown, G.R. and Matysiak, G.A. (1997), ``Sticky valuations, aggregation effects and property
indices'', Journal of Real Estate Finance and Economics, Vol. 20 No. 1.
Brown, G.R. and Matysiak, G.A. (2000), Real Estate Investment: A Capital Market Approach,
Prentice-Hall, Englewood Cliffs, NJ.
Brown, G.R. and Ong, S.E. (2001), ``Estimating serial cross correlation in property returns'',
Managerial and Decision Economics, Vol. 22 No. 7, pp. 381-7.
Clayton, J., Geltner, D. and Hamilton, S. (2001), ``Smoothing in commercial property valuations:
evidence from individual appraisals'', Real Estate Economics, Vol. 29 No. 3, pp. 337-60.
Diaz, J. and Wolverton, M. (1998), ``A longitudinal examination of the appraisal smoothing
hypothesis'', Real Estate Economics, Vol. 26 No. 2, pp. 349-58.
Fisher, J., Geltner, D. and Webb, B. (1994), ``Value indices of commercial real estate: a comparison
of index construction methods'', Journal of Real Estate Finance and Economics, Vol. 9
No. 2, pp. 137-66.
Geltner, D. and Ling, D. (2001), ``Ideal research and benchmark indexes in private real estate:
some conclusions from the RERI/PREA technical report'', Real Estate Finance, Vol. 17
No. 4, Winter, pp. 17-28.
Quan, D. and Quigley, J. (1991), ``Price formation and the appraisal function in real estate
markets'', Journal of Real Estate Finance and Economics, Vol. 4 No. 2, pp. 127-46.
Working, H. (1960), ``Note on the correlation of the first differences of averages in a random
chain'', Econometrica, Vol. 28, pp. 916-18.