[Federal Register Volume 80, Number 204 (Thursday, October 22, 2015)]
[Notices]
[Pages 63996-64000]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 2015-26778]
=======================================================================
-----------------------------------------------------------------------
FEDERAL HOUSING FINANCE AGENCY
[No. 2015-N-10]
Notice of Establishment of Housing Price Index
AGENCY: Federal Housing Finance Agency.
ACTION: Final notice.
-----------------------------------------------------------------------
SUMMARY: On May 27, 2015, the Federal Housing Finance Agency (FHFA)
published a Notice and Request for Input (Notice) describing a method
for assessing the national average single-family house price for use in
adjusting the maximum conforming loan limits of Fannie Mae and Freddie
Mac (the ``Enterprises''). The Notice responded to section 1322 of the
Federal Housing Enterprises Financial Safety and Soundness Act of 1992
(12 U.S.C. 4501 et seq.) (``Safety and Soundness Act'') which required
FHFA to ``establish and maintain a method of assessing the national
average 1-family house price for use in adjusting the conforming loan
limitations.'' The Notice indicated that FHFA intends to use its
existing ``expanded-data'' house price index (HPI) for such purpose and
invited public feedback.
In line with the proposal in the original Notice, after reviewing
the public feedback, FHFA has decided to use the expanded-data HPI for
annual loan-limit adjustment. Specifically, FHFA will use the
seasonally adjusted, expanded-data HPI for the United States.
DATES: Effective Date: October 22, 2015.
FOR FURTHER INFORMATION CONTACT: Questions about the expanded-data HPI
and the implementation of the conforming loan limit rules can be
addressed to Andrew Leventis, Principal Economist, 202-649-3199,
[email protected], or Jamie Schwing, Associate General Counsel,
202-649-3085, [email protected], (not toll-free numbers), Federal
Housing Finance Agency, 400 Seventh Street SW., Washington, DC 20024.
SUPPLEMENTARY INFORMATION:
A. Background
The ``Notice of the Establishment of Housing Price Index'' that
FHFA issued in May \1\ announced that the agency intended to use its
expanded-data HPI for the purpose of satisfying section 1322 (12 U.S.C.
4542) of the Safety and Soundness Act.\2\ Section 1322 requires FHFA to
``establish and maintain'' a house price index that tracks the average
U.S. home price. May's Notice detailed FHFA's rationale for the choice
of the expanded-data index over other measures. The Notice discussed
the advantages and disadvantages of several metrics and outlined the
various considerations FHFA found most compelling in choosing the
index. Identifying the seasonally adjusted, expanded-data HPI for the
U.S. as the selected index, the Notice invited public input and
provided for an input period that extended through July 27, 2015. This
Final Notice summarizes the input submissions received and responds to
questions and concerns that were raised in the submissions.
---------------------------------------------------------------------------
\1\ See 80 FR 30237 (May 27, 2015).
\2\ Section 1124(d) of the Housing and Economic Recovery Act of
2008 (HERA), 122 Stat. 2693, amended the Safety and Soundness Act to
include this section.
---------------------------------------------------------------------------
B. Overview of Input Submissions Received
FHFA received a total of 20 submissions in response to the Notice.
Submissions were received from private citizens, trade associations, a
think tank, and one private company. Twelve of the submissions did not
address the issue on which input had been requested: the
appropriateness of the chosen home price measure. In most cases, these
submissions opined on the desirability of having higher conforming loan
limits, rather than FHFA's choice of index.
In general, the eight responsive submissions were favorable to
FHFA's proposed use of its expanded-data index for loan limit
adjustment. Most submissions supported the basic underlying methodology
used in the index construction and appreciated the breadth of the data
sample used in forming the index. More generally, submitters agreed
that reliance on an agency-produced measure (as opposed to a privately
produced index) would be beneficial in that it would ensure continued
publication of the reference index. They also concurred with FHFA's
belief that its control over the reference index would ensure that
undesirable modifications to methodology would not be made (as might
happen if the agency relied on an external measure of home prices).
Five of the eight responsive submissions were generally supportive
of the use of the expanded-data index as-is. The remaining three did
not object to the use of the expanded-data index, but suggested
modifications to the process or augmentations. In particular, the
proposed adjustments recommended the use of multiple price indexes and,
in one case, the consideration of other mortgage market factors.
For the purpose of summarizing and addressing the responsive
submissions received, this Final Notice divides them into two groups:
``Supportive'' and ``Other.'' This classification is for convenience;
as will be clear in the discussion, responses in both categories were
not uniform. For instance, in some cases, the ``Supportive''
submissions included questions or expressed modest concerns. Meanwhile,
the ``Other''
[[Page 63997]]
submissions often included strong praise for certain characteristics of
FHFA's proposal.
C. Discussion of the Five Responsive ``Supportive'' Submissions
1. Summary
Three of the five ``supportive'' submissions were wholly in
agreement with the proposed use of the expanded-data index for tracking
the average U.S. home price. None of the three, which were all
submitted by trade associations, provided any material criticism. They
expressed strong support for FHFA's choice and, to varying degrees, the
principles FHFA used in evaluating measures.
The remaining two ``supportive'' submissions--one from a trade
association and one from a private company--provided supplementary
recommendations. The submissions addressed the following issues.
a. Data Inputs
Submissions urged FHFA to incorporate as much transaction data as
possible in the formation of the expanded-data index.
b. Distressed Sales and Gaps between House Price Indexes
Submissions asked that FHFA track the impact of distressed sales
\3\ on index estimates over time, while also monitoring divergences
between the FHFA index and other home price measures.
---------------------------------------------------------------------------
\3\ ``Distressed sales'' include short sales and sales of
properties that have gone through foreclosure.
---------------------------------------------------------------------------
c. Transparency and Data Releases
The submitter recommendation was that FHFA publish additional
details about the underlying data used for index construction.
d. Constraints on Historical Index Values
One submission asked FHFA to consider constraining the historical
index series. That is--the request was that FHFA consider not
permitting revisions in prior index estimates. Like all of FHFA
indexes, the expanded-data HPI has historical values that are regularly
updated to account for new data.
e. Geometric vs. Arithmetic Index
Without veering from its support of the expanded-data index, one
submission also noted a theoretical bias in the expanded-data index's
measurement of trends in average home prices. In particular, the
submitter stated that the underlying methodology used in forming the
expanded-data index will create indexes that track the geometric
average home price as opposed to the arithmetic average home price.\4\
In doing so, as a theoretical matter, the index reportedly would grow
somewhat more slowly over time than would an arithmetic index. The
letter conceded that the differences will be small over the short term
(e.g., on an annual bias), but worried about long-term compounding
effects. The letter noted that the CoreLogic-produced indexes track
arithmetic average home prices and thus are not susceptible to this
bias.
---------------------------------------------------------------------------
\4\ The geometric average of a set of numbers is computed by
multiplying the numbers together and then raising the product to the
power of one divided by the number of observations. Although not
necessarily the case, the geometric average can be close to the
median value. The arithmetic average is formed by adding numbers
together and dividing by the number of observations.
Although the ``arithmetic'' average is probably the most common
interpretation of the term ``average,'' it is not the only
recognized meaning of the term, and the statutory text does not make
explicit which type of ``average'' the index is supposed to track.
Which type of average to use is thus left to the judgment of FHFA,
as the agency charged with administering and interpreting the
statute.
---------------------------------------------------------------------------
2. FHFA Response
a. Data Inputs
With respect to the submitter interest in having FHFA increase the
amount of data used in calibrating the expanded-data index: FHFA agrees
that this is a desirable goal. In the context of tracking overall home
values across the country, more data will tend to provide more precise
estimates of price changes. While the database currently used is
extensive and incorporates a wide array of transaction data, FHFA will
continue exploring opportunities for increasing the sample size.\5\ As
stated in the Notice, to the extent that new data become available and
are incorporated, FHFA will communicate the effects of those changes to
the public.
---------------------------------------------------------------------------
\5\ For instance, opportunities may exist for supplementing the
existing data sample with sales data from Multiple Listing Services
and electronic appraisal data.
---------------------------------------------------------------------------
b. Distressed Sales and Gaps between House Price Indexes
With respect to monitoring of distressed sales and divergences
between the FHFA index and other metrics: FHFA concurs that these are
reasonable activities. FHFA, in fact, has been doing this type of
monitoring for many years and has published a number of papers showing
the results of its work.\6\ Also, FHFA publishes ``distress-free''
house price indexes for twelve large cities so that it and the general
public can review the localized impact of distressed sales on price
measurement. FHFA plans to continue such releases and, more generally,
will continue evaluating price movements across multiple measurements.
---------------------------------------------------------------------------
\6\ See, for instance, Andrew Leventis, ``Revisiting the
Differences between the OFHEO and S&P/Case-Shiller Housing Price
Indexes: New Explanations'' OFHEO Research Paper, January 2008,
available at http://www.fhfa.gov/PolicyProgramsResearch/Research/PaperDocuments/20080115_RP_RevisitingDifferencesOFHEOSPCaseShillerHPI_N508.pdf;
Andrew Leventis, ``The Impact of Distressed Sales on Repeat-
Transactions House Price Indexes,'' FHFA Research Paper, May 27,
2009, available at http://www.fhfa.gov/PolicyProgramsResearch/Research/PaperDocuments/20090527_RP_ImpactDistressedSalesHPI_RP_508.pdf; and Will Doerner
and Andrew Leventis, ``Working Paper 13-1: Distressed Sales and the
FHFA House Price Index,'' FHFA Working Paper, August 2013, available
at http://www.fhfa.gov/PolicyProgramsResearch/Research/PaperDocuments/2013-08_WorkingPaper_13-1_508.pdf.
---------------------------------------------------------------------------
c. Transparency and Data Releases
A longstanding tradition in HPI production has been to communicate
relevant summary data about the data sample to the public. Accordingly,
FHFA appreciates the submitter interest in maximizing the transparency
of the data used in index calibration. FHFA regularly publishes
information about the share of the overall data sample comprising
refinance loans and, for the expanded-data index, identifies index
estimates that have been calibrated with limited county recorder
data.\7\ For the purchase-only indexes, flags identify states having
small sample sizes. Highlights articles and Technical Notes in the past
have provided information about the data samples as well. Aside from
the FHFA-provided data, relevant information is also available from the
Enterprises. Because few data filters are applied to the data sample
before the indexes are estimated, index users seeking information about
the Enterprise portion of the expanded-data transactions can benefit
from reviewing loan-level summary statistics regularly published by the
Enterprises.\8\
---------------------------------------------------------------------------
\7\ See the downloadable expanded-data HPI estimates and the
``loan type'' table at http://www.fhfa.gov/DataTools/Downloads/Pages/House-Price-Index-Datasets.aspx.
\8\ See, for instance, Fannie Mae's quarterly ``Credit
Supplement'' and Freddie Mac's quarterly ``Financial Results
Supplement.''
---------------------------------------------------------------------------
Although a great deal of information is already available, FHFA
will continue evaluating opportunities for enhancing its release of
summary data. In reviewing those opportunities, FHFA will weigh the
likely value of the additional detail against the required resource
demands. It must also consider whether the release of more data would
violate the terms of any applicable data
[[Page 63998]]
licenses or would inappropriately release confidential data.
d. Constraints on Historical Index Values
The suggestion that FHFA should contemplate constraining historical
values of the expanded-data HPI is motivated by a concern that
historical index revisions might cause confusion among some index
users. The submitter recognizes that the entire historical index series
is revised with each new index release, but it expresses concern that
such revisions will make it difficult for the public to evaluate price
changes.
Although FHFA understands the argument, it does not believe that
artificial constraints on historical values are warranted. The
suggestion, which was not a matter of particular stress in the
submitter's letter, would entail a significant departure from the basic
repeat-transactions indexing model and would require a significant re-
tooling of the programming code. Furthermore, historical index
revisions tend to be relatively small, particularly over short periods
of time. The index constraints would also necessarily reduce the
accuracy of the index estimates. Finally, many--if not most--users of
FHFA's suite of public indexes are already accustomed to the fact that
historical index values are always subject to revision.
e. Geometric vs. Arithmetic Index
Regarding the theoretical biases associated with FHFA's use of an
index that tracks the geometric average home value: FHFA appreciates
the feedback and understands the issue. As a geometric index, FHFA's
expanded-data measure will tend to correlate somewhat more closely with
changes in median home values as opposed to arithmetic-average home
values and, in theory, will grow slightly more slowly than an
arithmetic-based price index would. Recognizing the theoretical issue,
FHFA notes that growth rate differences will likely be small and
increases in a geometric index in practice can actually exceed
increases for an arithmetic measure.\9\ A conversion to an arithmetic-
average index would also inconvenience those index users who find the
existing FHFA methodology superior for their applications. Coupled with
the fact that a conversion to an arithmetic-average index would require
a significant expenditure of internal resources (to change programming
code and perform model validation), these considerations lead FHFA to
believe that continuing with the existing methodology is appropriate.
---------------------------------------------------------------------------
\9\ See page 118 of Robert Shiller, ``Arithmetic Repeat Sales
Price Estimators'' Journal of Housing Economics 1, 1991, pages 110-
126.
---------------------------------------------------------------------------
D. Discussion of the Three Responsive ``Other'' Submissions
1. Summary
As mentioned above, the three ``other'' responsive submissions
suggested various modifications to the proposal described in the
initial Notice. None of them expressed outright disapproval of the use
of the expanded-data HPI and, indeed, incorporated it into their
proposals. Submitters felt that adjustments were necessary to address
perceived shortfalls, however.
The first of the ``other'' submissions expressed support for the
use of the expanded-data index, but worried that the index does not
adequately reflect price trends for new homes. It noted that the
underlying repeat-transactions approach used in forming the index is
calibrated using homes that have had two or more historical sales. The
upshot of reliance on homes with multiple transactions is that price
trends for brand new homes will not be incorporated into the index.
To mitigate the perceived problem, the submitter suggested that
FHFA form a weighted index that incorporates the expanded-data measure
as well as the price index for new homes published by the Census
Bureau--the Constant Quality House Price Index (CQHPI). The change in
the new combined index would be calculated as the weighted average of
the changes in the FHFA expanded-data HPI and the change in the CQHPI,
where the weights would be the relative shares of existing-vs-new home
sales. So, for instance, if 15 percent of all property sales in a year
were sales of new homes, then the growth in the combined index would be
85 percent times the change in the expanded-data index plus 15 percent
times the change in the CQHPI.
The second of the ``other'' submissions expressed no concerns about
the absence of new homes in the data sample, but rather was troubled by
the potential effects of distressed sales on index estimates. The
submitter was concerned that variations in the volumes of distressed
sales across geographic areas could inappropriately bias index
estimates. To mitigate this problem, the letter recommended that FHFA
use both the expanded-data index and its traditional ``purchase-only''
index, which is calibrated using only Enterprise data. Specifically, it
suggests that FHFA use the higher of the two appreciation rates--the
rates reflected in expanded-data and purchase-only indexes--when
adjusting the conforming loan limit. No indication is provided as to
why a ``higher-of'' rule is better than some other type of rule (e.g.,
a simple averaging of the two numbers).
The same submitter asked that FHFA ``explain and justify'' its use
of indexes that reflect changes in the geometric average home price.
While endorsing the use of the expanded-data and purchase-only indexes,
both of which rely on the geometric approach, the letter broadly
worries about the same bias as was addressed earlier.
The third of the ``other'' submissions did not address the issue of
the geometric index bias, but was otherwise similar in that it
suggested the same ``higher-of'' rule for estimating price changes. It
contends that a ``superior alternative'' to the use of the expanded-
data index would be for FHFA to adjust the loan limits by the higher of
the annual appreciation rates observed in the expanded-data and
purchase-only index. FHFA has had difficulty following the
justification set forth in the letter, but the rationale appears to
rest on the assumption that, because of tightened credit availability,
homes outside of the conforming market (e.g., expensive homes) will
evidence relatively anemic price growth in the early stages of economic
recoveries. By including homes financed with non-Enterprise loans, the
expanded-data HPI reportedly will tend to exhibit lackluster price
growth during recoveries.\10\ The ``higher-of'' rule would ensure that
the conforming loan limit grows by a reasonable rate during recoveries.
---------------------------------------------------------------------------
\10\ The submitter showed that the expanded-data HPI grew more
slowly than the purchase-only series in the latest recovery.
---------------------------------------------------------------------------
The same submitter presented the idea that a ``more sophisticated''
approach to loan limit adjustment might be taken. The alternative
approach would take into account market factors beyond home prices when
adjusting loan limits. Measures of loan ``access,'' for instance, might
be incorporated. The letter also suggests that in lieu of this ``more
sophisticated'' measure, FHFA might simply use its purchase-only
index--either in its existing form or in a value-weighted form.\11\ As
justification for the use of the purchase-only index, the letter simply
indicates that ``it grows faster during market's expansion through the
housing cycle.''
---------------------------------------------------------------------------
\11\ The value-weighted index would track the arithmetic average
home price.
---------------------------------------------------------------------------
2. FHFA Response
None of the three ``other'' submissions expressed particularly
strong sentiment against the use expanded-data HPI and, in evaluating
the rationale for the
[[Page 63999]]
proposed modifications, FHFA does not find the arguments to be
particularly persuasive. In general, the suggested adjustments have
limited support from both a statutory and statistical perspective.
a. Price Trends for New Homes
In assessing the criticism that FHFA's index--like other repeat-
transactions indexes--does not specifically incorporate information
about price trends for brand new homes, FHFA agrees that this may be a
theoretical shortfall. However, FHFA does not believe that this will be
a particularly significant problem in practice. First, while not
capturing price trends for brand new home sales, the repeat-transaction
model will reflect price changes for relatively new homes. This is
because the underlying calibration dataset includes cases in which new
homes were sold and then sold again within a relatively short period of
time. The price change for these ``young'' homes will presumably be
quite similar to price trends for brand new homes.
In weighting by the share of sales for new homes, the submitter's
proposal assumes that the index of interest should reflect price trends
for homes that have recently sold. FHFA does not agree that this is
appropriate in this context. FHFA's expanded-data index, like its other
indexes, aims to track average home prices for all U.S. properties--the
overall housing stock--and not just values for homes that were sold. To
implement the right weighting, FHFA forms the national index by taking
a housing-stock-weighted average of outcomes in the respective states.
To be sure, price changes in the individual states necessarily must
be calculated using recent transaction prices. However, as evidenced by
the fact that FHFA uses housing stock estimates when forming the
national index, FHFA's goal is to reflect price trends for the overall
housing stock.
Given this goal, the relevant statistic for evaluating the
importance of new homes is the share of the housing stock that such
homes comprise. New homes represent a very small proportion of the
overall housing stock and thus the submitter's concern about the
exclusion of new homes is not particularly problematic. Although new
home sales constitute a reasonable share of transactions in a given
year (according to the submitted letter, they have averaged about 17
percent of sales over the past 15 years), new homes are a very small
proportion of the housing stock. In 2014, for instance, about 620,000
one-unit new homes were built.\12\ For comparison purposes, estimates
from the Census Bureau indicate that there were more than 89 million
one-unit properties in the country in the preceding year.\13\ New homes
thus were substantially less than one percent of the housing stock in
2014. A stock-weighted combined index thus would place more than 99
percent of its weight on the price change reported by the expanded-data
index.
---------------------------------------------------------------------------
\12\ See new private new home ``completions'' in Table 5 of the
New Residential Construction report (available at http://www.census.gov/construction/nrc/pdf/newresconst.pdf).
\13\ See http://factfinder.census.gov/faces/tableservices/jsf/pages/productview.xhtml?pid=ACS_13_1YR_B25024&prodType=table.
---------------------------------------------------------------------------
b. Distressed Sales and Housing Cycles
With respect to the argument that distressed sales can distort home
price measurements: As detailed in the Notice and noted by another
submitter, there are advantages and disadvantages associated with the
inclusion of such sales in the data sample. Such transactions can
provide valuable information about price trends in cases where non-
distressed sales volumes are modest, for instance. Also, even if
removing distressed sales was deemed to be desirable after balancing
the various considerations, given current available data sources, it is
difficult to clearly identify such sales and remove them from the data
sample.\14\
---------------------------------------------------------------------------
\14\ FHFA currently publishes distress-free measures for 12
metropolitan areas, but such measures make use of a special, third-
party-sourced dataset to identify distressed transactions.
---------------------------------------------------------------------------
The submission that raises concerns about the expanded-data index
showing relatively limited price growth during market recoveries
provides no evidence that the (anticipated) slow growth would
misrepresent actual appreciation in the market. Tracking of home prices
is the key statutory requirement and, accordingly, the relevant issue
for FHFA is not whether certain market factors may influence lending
and home prices during market cycles; rather, the key issue for FHFA is
the reliability and accuracy of price measurement. The plain language
of the statute does not ask FHFA to evaluate market conditions (as the
submitter would have done using a ``more sophisticated measure'') or to
somehow account for likely market factors when selecting the
appropriate index. It also does not ask FHFA to select an index that
maximizes measured price appreciation during certain parts of the
housing cycle.
Given the basic goal of tracking the average home value over time,
the ``higher-of '' rule suggested by submitters is not well aligned
with the statutory language. By construction, the higher-of rule will
clearly inflate estimates of home price appreciation (and minimize
measured price declines) and thus would tend to lead to artificial
growth in conforming loan limits.
One of the two submitters that advances the ``higher-of '' rule
does so to mitigate the effect of distressed sales on index estimates.
Even assuming that the inclusion of distressed sales is problematic--an
issue addressed above--it is not clear why the maximum of the two price
change estimates would be superior over the long term to use of the
midpoint (or some other function of the two). Also, although there may
be some differences, the two indexes generally will be affected
similarly by changes in the volumes of distressed sales.
E. Conclusion
While not unanimous, the submissions received in response to the
Notice were, on balance, quite positive. All submitters seemed to agree
that an FHFA-produced measure was appropriate. The only matter of some
(limited) debate seemed to be whether small adjustments were necessary.
In some cases, the contemplated adjustments would have a limited
influence on index estimates. In other cases, FHFA believes that the
adjustments are not supported by the statutory language.
FHFA will begin using the seasonally adjusted, expanded-data HPI
for the U.S. for the purpose of adjusting the baseline conforming loan
limit.\15\ Consistent with the usual timing of loan-limit releases, the
first use of the index will be in late November of this year when FHFA
announces the 2016 Enterprise loan limits. As in prior years, FHFA will
publish actual loan limits as well as detailed information about the
relevant calculations. Given that the expanded-data index is now the
reference index, the relevant discussion will include an evaluation of
changes in the expanded-data index.
---------------------------------------------------------------------------
\15\ As discussed in the prior Notice, because index values will
be compared for the same quarter over time, only the most trivial
difference will exist between the selected seasonally adjusted index
and an unadjusted index.
---------------------------------------------------------------------------
As detailed at length in the Notice, certain loan-limit provisions
in the Enterprise Charters require that, after a period of home price
declines, the baseline loan limit cannot rise again until home prices
exceed their pre-decline levels.\16\ In accordance with this
[[Page 64000]]
requirement and as discussed in the prior Notice, when determining the
2016 baseline conforming loan limit this November, the third quarter
2015 price level will be compared to the price level in the third-
quarter of 2007--the base period for the recent price decline. As the
expanded-data HPI has now been selected as the reference index, market
participants can expect that the net price change (positive or
negative) will be computed over that interval using the expanded-data
HPI.
---------------------------------------------------------------------------
\16\ See Section 302(b)(2) (12 U.S.C. 17179b)(2)) of the Fannie
Mae Charter and Section 305(a)(2) (12 U.S.C. 1454(a)(2)) of the
Freddie Charter. These sections were amended by HERA sections
1124(a) and (b), 122 Stat. 2691-2692.
Dated: October 15, 2015.
Melvin L. Watt,
Director, Federal Housing Finance Agency.
[FR Doc. 2015-26778 Filed 10-21-15; 8:45 am]
BILLING CODE 8070-01-P