[Federal Register Volume 80, Number 101 (Wednesday, May 27, 2015)]
[Pages 30237-30246]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 2015-12781]



[No. 2015-N-03]

Notice of Establishment of Housing Price Index

AGENCY: Federal Housing Finance Agency.

ACTION: Notice and Request for Input.


SUMMARY: The Federal Housing Finance Agency (FHFA) is establishing and 
shall maintain a method for assessing the national average single-
family house price for use in adjusting the conforming loan limits of 
Fannie Mae and Freddie Mac (the ``Enterprises''). For these purposes, 
FHFA has considered a number of different measures, including the House 
Price Index maintained by the Office of Federal Housing Enterprise 
Oversight (OFHEO) of the Department of Housing and Urban Development 
before the effective date of the Federal Housing Finance Regulatory 
Reform Act of 2008.\1\ FHFA also considered house price indexes of the 
Bureau of the Census of the Department of Commerce as well as other 
privately-produced indexes.\2\

    \1\ Division A of the Housing and Economic Recovery Act of 2008, 
Pub. L. No 110-289, 122 Stat. 2654, 2659 (2008). Note that OFHEO was 
one of the predecessor agencies to FHFA.
    \2\ The S&P/Case-Shiller and CoreLogic house prices indexes, for 
instance, were considered.

    FHFA intends to use the FHFA ``expanded-data'' house price index 
(HPI)--an index it publishes on a quarterly basis--to adjust the 
conforming loan limit. This Notice solicits public input. Once public 
input is reviewed, another Notice will be published describing FHFA's 
final determination.

DATES: FHFA will accept input on the Notice on or before July 27, 2015. 
For additional information, see SUPPLEMENTARY INFORMATION.

ADDRESSES: You may submit your input on the Notice, identified by 
``Notice No. 2015-N-03,'' by any of the following methods:
     Agency Web site: https://www.fhfa.gov/AboutUs/Contact/Pages/Request-for-Information-Form.aspx.
     Hand Delivery/Courier to: Alfred M. Pollard, General 
Counsel, Attention: Input/Notice No. 2015-N-03, Federal Housing Finance 
Agency, Constitution Center, 400 Seventh Street SW., Eighth Floor, 
Washington, DC 20024. Deliver the package to the Seventh Street 
Entrance Guard Desk, First Floor, on business days between 9 a.m. and 3 
     U.S. Mail Service, United Parcel Service, Federal Express, 
or other commercial delivery service to: Alfred M. Pollard, General 
Counsel, Attention: Input/Notice No. 2015-N-03, Federal Housing Finance 
Agency, Constitution Center, 400 Seventh Street SW., Eighth Floor, 
Washington, DC 20024.

FOR FURTHER INFORMATION CONTACT: Andrew Leventis, Principal Economist, 
202-649-3199, [email protected], or Jamie Schwing, Associate

[[Page 30238]]

General Counsel, 202-649-3085, [email protected], (not toll-free 
numbers), Federal Housing Finance Agency, 400 Seventh Street SW., 
Washington, DC 20024.


Table of Contents

I. Input
II. Statutory and Regulatory Background
III. House Price Index for Loan Limit Adjustments
    A. Summary
    B. Background
    1. Safety and Soundness Act Section 1322
    2. Evaluating Existing Measures of Price Changes
    i. Available Measures
    ii. Evaluation Criteria
    C. Basics of the Proposed Methodology
    D. Other Measures of Home Prices
    E. Implementation Issues--Details
    F. Empirical Estimates of Price Changes: Expanded-Data HPI vs. 
Other Measures
IV. Conclusion

I. Input

    FHFA invites input on all aspects of the Notice and will take all 
relevant input into consideration. A final Notice will be published 
after FHFA considers public feedback.
    Copies of all submissions received will be posted without change, 
including any personal information you provide such as your name, 
address, email address and phone number, on the FHFA internet Web site, 
http://www.fhfa.gov. In addition, copies of all submissions received 
will be available for examination by the public on business days 
between the hours of 10 a.m. and 3 p.m. at the Federal Housing Finance 
Agency, Constitution Center, 400 Seventh Street SW., Washington, DC 
20024. To make an appointment to inspect submissions, please call the 
Office of General Counsel at (202) 649-3804.

II. Statutory and Regulatory Background

    The Housing and Economic Recovery Act of 2008 (HERA), Public Law 
110-289, 122 Stat. 2654 (July 30, 2008), amended the Federal Housing 
Enterprises Financial Safety and Soundness Act of 1992 (12 U.S.C. 4501 
et seq.) (Safety and Soundness Act) to establish FHFA as an independent 
agency of the Federal Government.\3\ Pursuant to section 1322 (12 
U.S.C. 4542) of the Safety and Soundness Act, as amended by section 
1124(d) of HERA, 122 Stat. 2693,\4\ FHFA is required to establish and 
maintain a House Price Index for use in adjusting the conforming loan 
limits of the Enterprises.\5\ A number existing metrics, including 
those identified in section 1322, could serve this purpose. Also, HERA 
sections 1124(a) and (b), 122 Stat. 2691-2692, amended sections 
302(b)(2) of the Federal National Mortgage Association Charter Act (12 
U.S.C. 1717(b)(2), and 305(a)(2) of the Federal Home Loan Mortgage 
Corporation Act (12 U.S.C. 1454(a)(2) (together, the Charter Acts), to 
specify that the baseline national loan limit should be changed 
annually by the percentage change in the established index.

    \3\ Division A of HERA titled, the Federal Housing Finance 
Regulatory Reform Act of 2008, established FHFA to oversee the 
operations of the Federal National Mortgage Association, the Federal 
Home Loan Mortgage Corporation (collectively, Enterprises), and the 
Federal Home Loan Banks (Banks) (collectively, regulated entities). 
FHFA is to ensure that the regulated entities operate in a safe and 
sound manner including being capitalized adequately; that their 
operations foster liquid, efficient, competitive and resilient 
national housing finance markets; that they comply with the Safety 
and Soundness Act and their authorizing statutes, and with rules, 
regulations, guidelines and orders issued under those statutes; that 
they carry out their missions through activities authorized and 
consistent with the Safety and Soundness Act and their authorizing 
statutes; and that the activities and operations of the entities are 
consistent with the public interest. See 122 Stat. 2659, 2663-2664 
    \4\ Original section 1322 was repealed by section 1121(2) of 
HERA, (122 Stat. 2689).
    \5\ Section 1322 states in relevant part that ``the Director 
shall take into consideration the monthly survey of all major 
lenders conducted by the Federal Housing Finance Agency to determine 
the national average 1-family house price, the House Price Index 
maintained by the Office of Federal Housing Enterprise Oversight of 
the Department of Housing and Urban Development before the effective 
date of the Federal Housing Finance Regulatory Reform Act of 2008, 
any appropriate house price indexes of the Bureau of the Census of 
the Department of Commerce, and any other indexes or measures that 
the Director considers appropriate.''

III. House Price Index for Loan Limit Adjustments

A. Summary

    Section 1322 of the Safety and Soundness Act requires that FHFA 
``establish and maintain a method of assessing the national average 1-
family house price for use in adjusting the conforming loan 
limitations.'' 12 U.S.C. 4542. The conforming loan limit is the maximum 
size of mortgage that the Enterprises are allowed to acquire in a given 
year. With some exceptions, the Safety and Soundness Act requires that 
FHFA annually adjust the maximum loan size by the percentage change in 
the index over the preceding year.
    After reviewing the landscape of available measures and analyzing 
candidate new methodologies, FHFA has chosen its ``expanded-data'' HPI 
for tracking average home values and adjusting the conforming loan 
limit. The index, which is already produced by FHFA on a quarterly 
basis, uses data from a number of different sources and employs the 
well-established ``repeat-transactions'' methodology for measuring 
price changes. A number of privately-produced indexes in fact use the 
same fundamental methodology, but have not been selected. The expanded-
data index is deemed to be relatively attractive because of the lengthy 
publication track record of the FHFA (and OFHEO) price indexes and the 
methodological control that production of the relied-upon index allows.
    Public input is sought on the relative merits of the selected 
index. Feedback is also desired on technical implementation matters 
addressed in this Notice.

B. Background

1. Safety and Soundness Act Section 1322
    Under section 1322 of the Safety and Soundness Act, the FHFA 
Director is required to ``establish and maintain'' a measure of average 
U.S. home prices. In doing so, the Safety and Soundness Act requires 
that FHFA ``take into consideration'' various measures of home prices 
when developing the index. The reference measures include the FHFA 
HPI,\6\ data from the Census Bureau, information from a contemplated 
FHFA survey of national lenders, and ``any other indexes or measures 
that the Director considers appropriate.'' 12 U.S.C 4542.

    \6\ The Safety and Soundness Act describes the FHFA HPI as ``the 
House Price Index maintained by the Office of Federal Housing 
Enterprise Oversight of the Department of Housing and Urban 
Development before the effective date of the Federal Housing Finance 
Regulatory Reform Act of 2008.''

    In the context of the Safety and Soundness Act, the purpose of the 
established index is to adjust the conforming loan limit. Specifically, 
it is used to adjust the baseline loan limit that applies in most of 
the country. This limit applies everywhere except for areas where 
median home values are high or are otherwise designated as ``high-
cost'' areas. Loan limits in high-cost areas will be addressed later in 
this Notice.
    Sections 302(b)(2) and 305(a)(2) of the Charter Acts specify that 
the baseline national loan limit should be changed annually by the 
percentage change in the established index. The change in the baseline 
limit is constrained when price declines occur, however. Specifically, 
the national loan limit is not permitted to decline when the national 
average price declines. Also, after a period of price declines, when 
the national

[[Page 30239]]

average home value finally does increase, the loan limit cannot 
increase until prices regain all of their prior losses.
    Prior to and immediately following the enactment of HERA, the 
national average home price declined significantly. FHFA's house price 
indexes and all other reliable measures of home price movements 
evidenced substantial declines. FHFA's expanded-data house price index, 
for instance, declined by more than twenty percent between the third 
quarter of 2007 and the third quarter of 2011. Given the Safety and 
Soundness Act's prohibition against declines in the baseline loan 
limit, declining U.S. home prices meant that the selection of a 
specific index for adjusting the loan limit under the Safety and 
Soundness Act was of little practical import; the baseline loan limit 
would be the same irrespective of the index used. With each year's 
publication of the conforming loan limits for the following year, FHFA 
noted this and kept the baseline loan limit the same ($417,000 for one-
unit properties in most of the country).\7\

    \7\ The announcement for 2015, for example, can be found on 
FHFA's Web site at http://www.fhfa.gov/Media/PublicAffairs/Pages/FHFA-Announces-2015-Conforming-Loan-Limits-Unchanged-in-Most-of-the-U-S.aspx. See, in particular, the second page of the Addendum to the 
release: http://www.fhfa.gov/DataTools/Downloads/Documents/Conforming-Loan-Limits/CLLAddendum_CY2015.pdf.

    Housing markets have improved substantially over the last few years 
and home values are getter closer to where they were just before HERA's 
enactment. Indeed, FHFA's expanded-data house price index is within a 
few percentage points of its level in 2007.\8\ Given the rising prices, 
it is now important that FHFA formally establish the specific 
methodology it will use for tracking prices and adjusting the baseline 
loan limit.

    \8\ As of the fourth quarter of 2014, the seasonally adjusted 
version of the index was about 7.3 percent below the 2007Q3 level.

    It should be noted that sections 302(b)(2) and 305(a)(2) of the 
Charter Acts specify that in locations where the 115 percent of the 
local median home value is above the baseline loan limit (``high-cost'' 
areas) the local limit is set at 115 percent of the median value. In no 
case, however, can the local loan limit be more than 150 percent of the 
baseline limit. The baseline loan limit thus acts as both a ``floor'' 
on loan limits and as a determinant of a ``ceiling'' on loan limits. 
The methodology for adjusting the baseline loan limits thus plays an 
indirect role in setting limits in these areas.
    The adjustment process for setting the baseline loan limit is also 
important to certain statutorily-defined areas. Legislation enacted 
prior to HERA set out Alaska, Hawaii, Guam, and the U.S. Virgin Islands 
as areas with higher loan limits.\9\ In these statutorily-defined 
areas, the local ``floor'' on loan limits is 150 percent of the 
baseline loan limit in the rest of the country. If area median home 
values are sufficiently high in these areas, the local limit can be 
even higher, as it can rise to a maximum of 150 percent of the ceiling 
in the rest of the country (which in turn is 150 of the baseline loan 
limit). Today, the highest possible loan limit for one-unit properties 
in the statutorily defined areas is $938,250 (i.e., 225 percent of the 
baseline loan limit of $417,000). The baseline loan limit establishes 
the floor and ceiling limits in these statutorily-defined areas and 
thus the index used for adjusting the baseline plays a role in 
determining limits in the statutorily-defined areas.
2. Evaluating Existing Measures of Price Changes

    \9\ The higher limit in the U.S. Virgin Islands, for example, 
was established in PL 102-550.

i. Available Measures
    A significant number of home price measures are available and could 
be used for adjusting the baseline conforming loan limit. Available 
metrics include:
    [ssquf] Any of FHFA's existing price indexes, including the 
purchase-only HPI, the all-transactions HPI, and the expanded-data HPI;
    [ssquf] The Census Bureau's Constant Quality House Price Index;
    [ssquf] The CoreLogic HPI;
    [ssquf] The S&P/Case-Shiller Indexes; and
    [ssquf] The National Association of Realtors' Average or Median 
Home Prices.
    The first two of these are specifically identified in section 1322. 
The other listed measures are produced by private data suppliers. When 
deciding which metric to be used for measuring price changes, FHFA 
considered all of the measures above.
    In 2010, FHFA published a Research Paper titled ``An Approach for 
Calculating Reliable State and National House Price Statistics.'' The 
paper, which is available for download on the FHFA Web site,\10\ 
described a methodology that might be used for measuring the national 
average home price. The methodology will generally produce estimates of 
average price changes that are similar to those estimated by FHFA's 
expanded-data HPI, but involves the addition of supplemental data. This 
more-complicated methodology may be considered as an option in the 
future, but is not considered here.

    \10\ The paper, authored by Andrew Leventis, is available at: 

ii. Evaluation Criteria
    In evaluating various measures of home prices changes that might be 
used for section 1322, FHFA considered a number of factors. The most 
important factor was whether price changes reflected in the measure 
would correlate closely with changes in the U.S. average home price. 
The purpose of the index referenced in the Safety and Soundness Act is 
to adjust the conforming loan limit, and thus the reliable measurement 
of price changes is of the highest importance. As closely as possible, 
changes in the selected index should reflect changes in the average 
value of homes.
    Section 1322 indicates that the measure should ``assess'' average 
U.S. home prices. Whether or not the measure needs to show the actual 
level of the average U.S. home prices is of little practical import for 
the Safety and Soundness Act's purposes. The critical use of the metric 
is to measure the price change and for FHFA to adjust the loan limit 

    \11\ The Safety and Soundness Act implicitly recognizes that 
primacy of the change estimate by describing the measure as an index 
as opposed to merely the average value.

    The absence of any real need to measure the level of prices is 
notable because many existing house price measures do not actually 
report statistics on the absolute level of home prices; rather, they 
report indexes that can be used for measuring changes. No average or 
median house prices are currently published for the FHFA HPI, for 
instance. Similarly, other measures (e.g. the S&P/Case-Shiller index, 
the CoreLogic index) are not generally accompanied by level estimates. 
All of these measures, despite the absence of the estimated level of 
home prices, thus can act as reasonable candidates for the index to be 
used for loan limit adjustment.
    Before the next evaluation criteria is discussed, it is important 
to briefly address the target of the index--the ``average'' price. 
Interestingly, the Safety and Soundness Act references the average 
price in the context of measuring changes in national home price and 
adjusting the baseline conforming loan limit, but references median 
home values in the setting of loan limits in high-cost areas.
    Ultimately, the practical impact of the average-median distinction 
is modest:

[[Page 30240]]

the long-term growth rates in average and median home prices are very 
similar and thus the choice of the target statistic (average vs. 
median) likely will have only a minimal impact on long-term loan 
limits. Even in the shorter term--during the recent housing bust--there 
was no dramatic difference in the measured declines for the median and 
mean U.S. prices.\12\ The index FHFA intends to use for loan limit 
adjustment tracks the geometric average U.S. home price--a measure that 
tends to correlate closely with median and average home prices.\13\

    \12\ According to estimates from the National Association of 
Realtors' Existing Home Sales series, for instance, the decline 
between September of 2007 and September of 2011 was roughly 20.7 
percent for average prices and 16.9 percent for median prices.
    \13\ The geometric mean of N numbers is computed as the product 
of the numbers taken to the 1/N root.

    Aside from the issue of the relevance of the statistic and the 
target (the average vs. median), the methodological transparency is 
also deemed to be a key attribute for evaluating various alternatives 
for the index. Details concerning how the statistics are constructed 
are important, as is information about methodological changes that 
might be made over time. In the landscape of available home prices, 
FHFA found vast differences in the amount of background information 
    Beyond relevance and transparency, FHFA also values reliability and 
control. The selected index should have a historical ``track record'' 
to minimize the risk that the relied-upon metric would be discontinued.
    Agency production of the index also is important, not only because 
it would ensure continued publication of the important statistic, but 
also because production of the index enables the agency to make 
appropriate enhancements. The scope of available house price 
information has expanded sharply over the last several years and new 
developments may soon make more and better transactions information 
available. Agency production of the index will mean that new 
information can be added in a way that improves the precision of 
estimates, while not being disruptive to the setting of loan limits.
    Finally, cost considerations were taken into account when 
evaluating candidate measures. While use of the expanded-data HPI and a 
number of externally-produced indexes would entail no incremental cost, 
one option would be for FHFA to develop and maintain a new index (for 
example, the one considered in the 2010 FHFA Research Paper). Efforts 
spent on maintaining a new measure, which would be yet another variant 
of FHFA's already-expansive suite of available price indexes, would 
entail a substantial expenditure of resources. The benefits of any 
increased precision of the estimates would need to be weighed against 
these costs.

C. Basics of the Proposed Methodology

    FHFA intends to use the ``expanded-data'' HPI for the purpose of 
tracking average U.S. home prices as contemplated in section 1322. 
While any of a number of existing measures might produce similar 
results, FHFA's expanded-data HPI for the U.S. is found to be 
particularly attractive under the evaluation criteria discussed above.
    The index, which has been published by FHFA since August of 2011, 
is constructed using the same ``repeat-transactions'' methodology as is 
used to construct the traditional FHFA HPI. The basic approach has been 
used by FHFA and OFHEO, one of FHFA's predecessor agencies, since 1996 
when the HPI was first publicly released. The details on how the index 
is constructed are found in a technical primer available on FHFA's Web 
site.14 15

    \14\ See Charles Calhoun, ``OFHEO House Price Indexes: HPI 
Technical Description,'' available at
     http://www.fhfa.gov/PolicyProgramsResearch/Research/PaperDocuments/1996-03_HPI_TechDescription_N508.pdf. Hereafter, this 
paper is referred to as the HPI Technical Primer.
    \15\ Other publicly-available measures, including notably the 
S&P/Case-Shiller and the CoreLogic suite of indexes, employ the same 
basic methodology, although some details concerning their 
construction are not publicly available. The methodologies used in 
forming those indexes and decisions related to the release of the 
measures are not within FHFA's control.

    The technical elements of the methodology are not detailed in this 
Notice, but the basic statistical model was first developed in the 
1960s and was refined by Karl Case and Robert Shiller more than twenty 
years ago. The fundamental approach entails finding homes that have 
been sold two or more times in the past and calibrating a set of 
numbers--index values--to broadly reflect changes in value observed for 
such homes. Using millions of historical real estate transactions, the 
model begins by creating transaction ``pairs,'' where each pair 
reflects the price growth (or decline) that occurred for a given 
property over a specific interval of time. For example, if a 
hypothetical home was sold two times in the past--once for $100,000 in 
the first quarter of 2001 and again for $225,000 in the fourth quarter 
of 2014--then a pair would be created showing appreciation of 125 
percent between 2001Q1 and 2014Q4.\16\ Using this pair and millions of 
other pairs for other properties, the basic model entails estimating a 
regression model \17\ that ``explains'' observed price changes using 
only information about when the individual property transactions 
occurred. The statistical model attempts to explain price changes (as 
opposed to price levels), a feature that makes it less susceptible to 
certain biases when measuring overall price movements in the 
marketplace.\18\ The output of the model is a series of index values 
whose changes broadly mimic the price changes observed for the millions 
of properties in the dataset.

    \16\ A home with three historical sales will produce two pairs. 
The first pair will reflect the price change between the first and 
second transactions and the second pair will show the change in 
selling price between the second and third transactions.
    \17\ A regression model is a well-established method for showing 
the statistical relationship between variables.
    \18\ For instance, if a large number of expensive homes transact 
in any given quarter, then the average and median transaction values 
will rise for a given area, even if there is no underlying home 
price appreciation. The repeat-transactions index, by contrast, will 
generally not reflect spurious price ``increases' in such 

    The FHFA expanded-data HPI uses the repeat-transaction model for 
estimating price changes in individual cities, all 50 states (and 
Washington, DC), and in the U.S. as a whole. Consistent with the way 
other FHFA indexes, for example the ``purchase-only'' and ``all-
transactions'' indexes, are formed, the change in the expanded-data 
U.S. index is constructed to reflect the weighted average changes 
across the 50 states and Washington, DC. This ensures that changes in 
relative real estate volumes across states do not bias the measurement 
of the change in U.S. prices. If the expanded-data U.S. index was 
estimated by simply pooling transactions data from all states together 
and directly estimating it, the measured price change would be 
susceptible to biases when relative transaction volumes shift across 
states. In an environment in which prices are rising and transaction 
activity increases dramatically in those states with the most extreme 
price increases, for instance, the weighting ensures that the volume 
shifts do not inflate the measured price measure for the U.S. as a 

    \19\ During market downturns (when transaction volumes tend to 
shrink in areas with the most extreme price declines), the constant 
weighting approach prevents the index from reporting undersized 
price declines.

    Although the expanded-data HPI employs the same basic methodology 
as is used for forming FHFA's two Enterprise-only datasets (the ``all-
transactions'' and ``purchase-only'' indexes), it uses slightly 
different historical transactions data. Like

[[Page 30241]]

FHFA's other measures, the expanded-data index incorporates sales price 
information for homes with Enterprise-purchased mortgages. Unlike 
FHFA's ``all-transactions'' index, however, appraisal values from 
refinance mortgages are not used in the data sample. Also, importantly, 
unlike both of the other two measures, the expanded-data indexes 
incorporate transaction prices for homes with FHA-endorsed loans and 
homes whose transactions have been recorded at various county recorder 
offices through the country. FHFA works with an outside data vendor--
currently CoreLogic--to obtain the county records data from hundreds of 
counties throughout the country.
    The addition of the two supplemental data sources (FHA and 
CoreLogic) to the Enterprise data provides for a better estimate of the 
overall change in the U.S. average home price than is available from 
the other indexes. To be sure, price changes reported in FHFA's other 
datasets will often closely resemble those reported by the expanded-
data index. However, as has been discussed in prior OFHEO and FHFA 
publications, trends in home values sometimes have been demonstrably 
different for homeowners with different types of financing.\20\ The 
expanded-data HPI is well-suited for capturing and incorporating those 
trends into its estimate of aggregate home price movements, unlike the 
other FHFA indexes.

    \20\ See, for example, ``Recent Trends in Home Prices: 
Differences across Mortgage and Borrower Characteristics,'' August 
2008, available at http://www.fhfa.gov/PolicyProgramsResearch/Research/PaperDocuments/20080825_RP_RecentTrendsHomePrices_N508.pdf.

    Changes in the expanded-data HPI do not perfectly measure changes 
in the average or median U.S. home prices, to be sure. As discussed in 
the technical primer that details the FHFA methodology \21\ and in the 
academic literature on the subject of price indexes,\22\ FHFA's basic 
methodology tracks the geometric average home price. In most cases, 
however, the index will very closely correlate with any index that 
would specifically track the median (and often the average) price.

    \21\ See the HPI Technical Primer available at http://www.fhfa.gov/PolicyProgramsResearch/Research/PaperDocuments/1996-03_HPI_TechDescription_N508.pdf.
    \22\ For a lengthy discussion, see Shiller, Robert, ``Arithmetic 
Repeat Sales Price Estimators'' Journal of Housing Economics 1, pp. 
110-125, 1991.

    In the context of the estimation of house price indexes, a robust 
debate has occurred over the last several years regarding whether 
``distressed sales'' should be included in the calibration data sample. 
Distressed sales, which include sales of bank Real Estate Owned (REO) 
properties as well as short sales,\23\ tend to have lower prices than 
other transactions. These lower prices generally result from two 
factors: poor property condition and greater-than-average seller 

    \23\ Short sales are transaction for which: (a) The homeowner 
was in financial distress and (b) the transaction price was an 
amount lower than the loan balance. In such situations, to avoid the 
costs associated with foreclosure, lenders allow the distressed 
homeowner to sell the property for less than the loan amount.

    Like other FHFA indexes and house price metrics produced by many 
others, FHFA's expanded-data HPI incorporates price data from 
distressed sales. As with all transactions, the distressed sales are 
included in the calibration of the expanded-data HPI as long as the 
buyer obtained an Enterprise or FHA loan or the property is in one of 
the counties for which FHFA has licensed county recorder information.
    The primary justification for including such distress transactions 
is that they provide indications of value in situations where, without 
such data, price declines may be understated. It is well established 
that, during housing market downturns, sellers commonly pull their 
properties from the market, preferring to ``wait out'' declines rather 
than selling at a loss. In such environments, transaction volumes may 
shrink dramatically and the few observed transactions that do occur may 
show relatively limited price declines.\24\

    \24\ Another reason for including the transactions is pragmatic: 
it is often difficult to identify distressed sales using available 
data. FHFA has done so in the past and it does produce a set of 
``distress-free'' indexes for select cities. The distress-free 
indexes take advantage of a unique dataset that aids in the 
identification of distress only in select cities, however.

    One final note about the expanded-data HPI is important: as new 
opportunities arise for the addition of transactions data to the 
modeling dataset, FHFA may take advantage of those to improve the 
index. Since the inaugural release of the expanded-data HPI in 2011, 
the term ``expanded'' has referred to the addition of FHA and county 
recorder data to the standard Enterprise dataset. There is no reason 
that additional data sources may not be included into the calibration 
dataset in the future. For instance, transaction prices embedded within 
property appraisal data \25\ might supplement the existing data 
sources. As with all significant changes in FHFA indexes, FHFA would 
notify the public of any such data enhancements.

    \25\ To be clear--this would not entail the inclusion of 
appraisal values, but rather property sales prices (e.g., sales 
prices for ``comparable'' properties) found in electronic appraisal 

D. Other Measures of Home Prices

    While other existing (and potential) measures had some attractive 
qualities, given the criteria used, FHFA believes that the expanded-
data HPI is the best option for the purpose of adjusting the loan 
    The data sources that the Safety and Soundness Act explicitly 
requires the Director to consider are the FHFA's ``monthly survey of 
all major lenders'' and any ``appropriate house price indexes'' 
published by the Census Bureau. Viable options for measuring 
appropriate price changes are not available from either. In the case of 
the monthly survey, the requisite data fields are currently under 
development, and therefore FHFA has not yet conducted the survey. 
Statistics from the Census Bureau are comprehensive for tracking the 
prices of new homes that are sold, but generally do not show price 
changes for existing homes. Price trends for new homes can differ 
substantially from price trends for existing homes, and thus the new 
home focus of the Census Bureau data is deemed to be a significant 
drawback in this context.
    In theory, one might track changes in the average or median U.S. 
home prices by looking at statistics published monthly by the National 
Association of Realtors (NAR). The NAR's estimates focus on prices for 
existing homes, as direct estimates of the average and median 
transactions prices are reported using data from a large number of 
local Multiple Listing Services. NAR's estimates are attractive in 
their simplicity (no statistical models are employed in their 
derivation) and in the fact that the statistics have been published 
consistently for decades. The major problem with their use, however, is 
that--like all summary statistics--they are susceptible to short-term 
biases caused by fluctuations in the types of properties that transact 
in any given quarter. If a substantial number of expensive homes 
transact in any given quarter, for instance, the reported average and 
median home values will tend to rise even if no real market 
appreciation was present. If the ``quality'' of transacting homes is 
not held constant from quarter to quarter, the resulting statistic can 
produce volatile measures and may bias estimates of price changes 
(particularly in the short run). As has been discussed at length in 
academic and practitioner literature, other indexes--for example those 
that rely on the repeat-transaction methodology (e.g., the expanded-

[[Page 30242]]

HPI)--are less susceptible to these biases.\26\

    \26\ The repeat-transactions statistical model is sometimes 
described as producing a ``constant-quality'' index.

    U.S. house price indexes published by S&P/Case-Shiller and 
CoreLogic use the repeat-transactions approach for measuring price 
changes and thus would not be susceptible to these biases. Use of 
either of these indexes--or other external measures of house price 
movements--in the context of setting loan limits would entail 
substantial operational risks, however. The external measures do not 
generally have track records that rival the lengthy publication history 
of the FHFA HPI. Reliance on an external measure would mean that FHFA 
would be dependent on its continued publication and on the 
methodological decisions made by the producer. If the producer opted to 
discontinue publication or to make undesirable methodological changes, 
significant complications would arise, and the publication of the 
conforming loan limits ultimately could be disrupted. Separately, 
ignoring the issue of continued publication risks, details concerning 
the methodology employed in the production of external indexes are not 
always publicly available and, therefore, have less transparency than 
FHFA's indexes. The prospect that FHFA would rely on an index having 
little public descriptive material for the important function of 
setting loan limits is not appealing to the agency.

E. Implementation Issues--Details

    While it will be enlightening to compare price trends for the 
expanded-data HPI to trends for other measures, it is useful to first 
address details concerning implementation timing. In particular, this 
section describes the ``when'' and ``how'' of loan limit changes under 
the use of the expanded-data HPI.
    The Safety and Soundness Act requires that loan limits be 
``adjusted'' each year and that the newly adjusted limits apply 
beginning in January. Since the passage of HERA--and in years prior 
(when OFHEO was setting the loan limit)--annual adjustments have been 
announced in the latter part of November. Under the terms of the 
Charter Acts, adjustments are to reflect the percentage price change in 
the index over the ``most recent'' 12-month or 4-quarter period.\27\ 
Given the large price changes that occurred and the Safety and 
Soundness Act's prohibition on declines in the baseline loan limit, it 
has not been necessary for FHFA to formally designate the reference 
period: i.e., whether price changes will be measured on a 4-quarter or 
12-month basis and the specific comparison interval (e.g., July vs. 
July of the preceding year or Q3 vs. Q3).

    \27\ See Charter Acts sections 302(b)(2) (12 U.S.C. 1717(b)(2) 
and 305(a)(2) (12 U.S.C. 1454(a)(2).

    Given the existing publication schedule for the expanded-data HPI, 
when setting loan limits on a go-forward basis, FHFA anticipates 
measuring price changes between the third quarter and the third quarter 
of the preceding year. As always, FHFA will produce its suite of house 
price indexes (including the expanded-data HPI) in November using data 
through the most recent quarter--the third quarter. Then, using the 
measured price increase in the expanded-data HPI between the third 
quarter of the prior year and the third quarter of the present year, 
FHFA will compute the new baseline loan limit. The new loan limit will 
be announced toward the end of November at roughly the same time as the 
HPI report is published.\28\

    \28\ FHFA's third quarter HPI for 2015 is set to be released on 
November 25, 2015.

    The proposed focus on third quarter prices means that, in the 
current situation in which average prices are below levels prevalent 
prior to the passage of HERA, the third quarter of 2007 represents the 
relevant reference period for determining when the baseline loan limits 
can rise again. The baseline conforming loan limit was first set in 
late 2008 and, as such, the first interval for assessing price changes 
was 2007Q3 to 2008Q3. Under the expanded-data index (and other 
measures), that 2007Q3-2008Q3 change was a price decline, thus 
triggering the prescriptive terms of the Safety and Soundness Act 
requiring that prices rise to the 2007Q3 level before the baseline loan 
limit can be increased. In successive years of setting loan limits, the 
expanded-data HPI found further declines--and then a partial recovery--
in U.S. average home prices. As shown in the next section, the latest 
expanded-data index value for the U.S. (for 2014Q4) shows that prices 
are still 7.9 percent below the 2007Q3 level. When the conforming loan 
limit is set for 2016 later this year, the index will generally have to 
exceed the 2007Q3 level for there to be an increase in the baseline 
loan limit.
    One final technical note must be made about historical values of 
the expanded-data HPI. Under the basic repeat-transactions indexing 
model used for producing the index (and other repeat-transactions 
measures), all historical values of the index are unconstrained, 
meaning that they are revised in each period.\29\ Unlike other types of 
price indexes, where an index value for a given period may be initially 
revised once or twice and then will be fixed forever, the repeat-
transactions house price index produces index values that are 
constantly in flux. That is--values for all historical quarters, even 
distant quarters, are modified slightly each period to account for new 
historical data. To be sure, most values are revised only slightly 
(e.g., the index value for a quarter in the late 1990s might change 
from 175.02 to 175.04 between one quarter and the next). Changes are 
constantly made, however.

    \29\ Other publicly available measures deviate somewhat from the 
basic repeat-transactions model and sometimes constrain historical 
price levels.

    FHFA's measurement of price changes for the setting of loan limits 
will use the most recently released index values as of the third 
quarter and will ignore prior vintages. For example, in setting 2016 
loan limits, FHFA will rely on the most recent time series of index 
values for comparing price levels. The 2015Q3-vintage estimates of the 
relevant historical values will be compared. To illustrate--although 
the most recent HPI publication showed that the expanded-data index 
estimate for 2007Q3 was 215.19,\30\ when determining whether prices 
have risen for loan-limit setting purposes in November, FHFA will use 
the 2007Q3 value published in November. If the 2015Q3 index value 
exceeds the index value for 2007Q3 (as determined in the 2015Q3 index 
vintage), then the baseline loan limit will be increased.\31\

    \30\ This value was the seasonally adjusted index estimate for 
the U.S. published on February 26, 2015. FHFA anticipates using 
seasonally adjusted index values in evaluating price changes. 
Because all annual price comparisons are made relative to the same 
(third) quarter in prior years, however, this choice has little 
practical effect.
    \31\ Note that, as indicated earlier, the loan limit will only 
increase by the net percentage increase since 2007Q3. In general, in 
market environments where prior price declines do not need to be 
overcome, the increase percentage will be the proportionate increase 
between the third quarter of the prior year and the third quarter of 
the contemporary year.

F. Empirical Estimates of Price Changes: Expanded-Data HPI vs. Other 

    Using the expanded-data HPI and several other commonly-cited 
measures of home prices changes, Figure 1 and Table 1 compare price 
trends calculated by the expanded-data HPI and other estimates of price 
change. Figure 1 indicates that all of the indexes report a very 
similar evolution of prices since 2007. The metrics generally show 
significant price declines between 2007 and sometime in 2011 and then a 
robust recovery. The measures show that the

[[Page 30243]]

most recent price level is still somewhat below the 2007Q3 level.
    Reconciling the small short-run differences in the price trends 
reflected in the various measures is complicated and even an in-depth 
analysis would likely conclude with much of the differences remaining 
unexplainable.\32\ In general, however, the variations are a function 
of differences in the underlying datasets, differences in the 
methodology employed, and variations in the weighting of sub-areas. 
Over the long-term, however, all of the indexes show similar patterns. 
Even the NAR median price, which is constructed using the most 
simplistic approach, trends similarly to the other measures. The NAR 
figure is notably volatile, likely a function of the fact that it is 
susceptible to certain short-term biases the repeat-transactions-based 
measures are immune to. Over the time frame shown and even over a more 
extended period, however, its evolution is similar to that of the 

    \32\ In a series of OFHEO papers published in 2007 and 2008, 
Andrew Leventis attempted to reconcile differences between the OFHEO 
HPI and the S&P/Case-Shiller indexes. See, for instance, http://www.fhfa.gov/PolicyProgramsResearch/Research/PaperDocuments/20080115_RP_RevisitingDifferencesOFHEOSPCaseShillerHPI_N508.pdf. The 
analysis, which just focused on the indexes produced by the two 
providers, explained some but not all of the variations in measured 
price changes.
    \33\ Observers will notice that Figure 1 reports the S&P/Case-
Shiller ``20-City Composite'' index as opposed to a pure national 
measure. Although the S&P/Case-Shiller suite of indexes includes a 
``U.S.'' measure, that measure is published under a timeline that 
would make it inconvenient for use in adjusting conforming loan 
limits. In particular, the S&P/Case-Shiller U.S. index is published 
quarterly and the third quarter estimate would not be available to 
FHFA until late in November. The absence of (even preliminary) 
information about price changes before the end of November would 
mean that, were FHFA to rely on it, year-ahead loan limits could not 
be published until early December. The S&P/Case-Shiller 20-City 
composite index is published on a monthly basis, by contrast. If 
FHFA were to rely on that measure, it could use the August-to-August 
price change estimate, which would be available in late October 
(meaning that a late-November release of loan limits would be 

    Table 1 provides estimates of the overall price deficit--the change 
in prices between 2007Q3 and the most recent data reading--for the 
various measures. As of the fourth quarter of 2014, the expanded-data 
HPI estimates that the average U.S. price was roughly 7.3 percent below 
its 2007Q3 level. This deficit is slightly below the midpoint of the 
two extreme values in the table: The S&P/Case-Shiller 20-City Composite 
(down 12.0 percent) and the FHFA purchase-only HPI (down 1.2 percent).

IV. Conclusion

    A very significant number of methodological and implementation 
options exist for satisfying section 1322. This Notice has described 
FHFA's use of the expanded-data index as the preferred option for 
annually setting loan limits under the procedure outlined (e.g., 
comparing third-quarter prices to third-quarter prices when evaluating 
the most recent year's price change). FHFA recognizes that other 
methodological and implementation decisions could be made. Given the 
material impact on the Enterprises and in light of the significant 
number of market participants affected by the level of the conforming 
loan limit, FHFA has released this Notice and Request for Input to 
ensure that public input is widely solicited.
    FHFA encourages submitters to address any theoretical or practical 
issues deemed to be important in this context. Once all submissions are 
received, they will be reviewed by FHFA staff and a final Notice will 
be published in the Federal Register. The final Notice will communicate 
FHFA's ultimate determination and may address some of the submissions 
received in response to this Notice.
    FHFA intends to publish a final determination in the Federal 
Register by the time the Enterprise 2016 conforming loan limits must be 
published (i.e., by late November 2015). As in the past, the conforming 
loan limit release will be published on FHFA's Web site.

[[Page 30244]]


[[Page 30245]]


[[Page 30246]]

    Dated: May 18, 2015.
Melvin L. Watt,
Director, Federal Housing Finance Agency.
[FR Doc. 2015-12781 Filed 5-26-15; 8:45 am]