Too Much Bubbly at the Fed?:
The New York Federal Reserve Board's Analysis of the Run-Up in Home Prices
Dean
Baker1
Center for Economic and Policy Research
baker at cepr dot net
June 12, 2004
Executive Summary
Since the third quarter of 1995, home sale prices have risen by more than 33
percent after adjusting for inflation. In the regions with the largest increases
in prices, the real increase has been more than 50 percent over this period.
This run-up in home prices has created almost $4 trillion in housing wealth
compared to a situation in which home prices had simply kept pace with
inflation.
A new study from the Federal Reserve Bank of New York questions whether this
run-up in home prices can be attributed to a housing bubble. Like prior studies
it attributes part of the increase in housing prices to lower interest rates.
However, this view, even if accurate, supports the bubble argument, since it
means that housing prices will plunge when interest rates return to more normal
levels, as is widely expected.
The novel part of this new study is the effort to attribute the run-up in home
prices to quality improvements. The study relies on the Census Bureau's constant
quality new construction price index as a measure of inflation in housing
prices. This index shows only a slightly more rapid pace of inflation in housing
prices than the overall CPI. The Fed study argues that the difference between
the Census index and the more commonly used House Price Index (HPI) is
attributable to the failure of the HPI to pick up quality improvements.
This paper shows that the evidence clearly contradicts the main claims of the
Fed study.
1) If the huge run-up in home prices in recent years can be explained by a rapid
pace of home improvement, then spending on home improvements should have been
very high in recent years. In fact, the Census Department's data show that
spending on home improvements, relative to the value of the housing stock, has
actually been lower in the years of the run-up in home prices than it was in the
early nineties, when home prices were not rising relative to inflation. This
suggests that a more rapid pace of quality improvement cannot possibly explain
the run-up in housing prices shown in the HPI.
2) Annual spending on home improvements is equal to less than 1.0 percent of the
value of the stock of residential housing. Changes in the size of this spending
are an order of magnitude too small to explain increases in the value of the
housing stock that have averaged more than 8 percent annually over the last
eight years.
3) The areas that have experienced the most rapid rate of quality improvement,
according to the methodology in the Fed study, are not the areas that have seen
the largest increases in spending on improvements, according to the Census
Bureau's data.
4) The number of vacant units is now increasing at a 750,000 annual rate. This
is more than one-third of the rate of housing construction and more than 50
percent of the rate of annual absorption. This growth in vacant units can be
explained by a speculative bubble. It is not clear what alternative explanations
could be plausible.
In short, the Fed's study is clearly contradicted by the available data. It does
not provide any plausible reason to question the existence of a housing bubble.
Introduction
The Federal Reserve Bank of New York recently issued a new study that questioned
whether the recent run-up in home prices in large parts of the country
constituted a housing bubble2. Two main
substantive points that questioned the existence of a housing bubble were
raised:
1) lower nominal interest rates reduced the user cost of housing, and therefore
justified considerably higher home prices than we have seen in the past; and
2) the Office of Housing Enterprise Oversight's House Price Index (HPI), the
standard measure of house price appreciation, had a large upward bias, because
it did not take account of quality improvements.
This brief discussion will focus on the second issue. The point concerning lower
nominal interest rates is not new, and cannot effectively respond to the
evidence of a housing bubble. Put simply, it assumes that homebuyers cannot
distinguish between real and nominal interest rates - an assumption radically at
odds with standard economic theory, and which also implies dire consequences for
current homebuyers3. Furthermore, it implies a
sensitivity of home prices to interest rates that has not been borne out in the
past and, if true, suggests that housing prices will collapse if interest rates
rise in coming years, in line with most economic projections4.
The new contribution of the Fed article is the claim that the extraordinary rise
in the HPI since 1995 is being driven primarily by improvements in the quality
of housing, rather than an increase in the price of housing of the same quality.
If quality improvements can explain most of the run-up in home prices, then
there is little basis for concern about a speculative bubble, which may
eventually collapse.
Before examining the evidence for this view, it is worth restating the basic
case for a housing bubble. Table 1 shows the increase in housing prices as
measured by the HPI for the nation as a whole and in each region from the third
quarter of 1995 to the first quarter of 2004. The first column shows the nominal
price increase, the second column shows the real increase, using the CPI-U as
the deflator.
Table 1
The Run-Up in Home Prices 1995-2004
|
Nominal |
Real |
U.S. |
62.6% |
33.4% |
Northeast |
94.3% |
59.4% |
Mid-Atlantic |
65.1% |
35.4% |
East South Central |
39.9% |
14.7% |
West South Central |
41.5% |
16.1% |
South Atlantic |
60.7% |
31.8% |
East North Central |
50.5% |
23.5% |
West North Central |
60.8% |
31.9% |
Mountain States |
87.0% |
53.4% |
Pacific |
86.9% |
53.3% |
Source: Office of Federal Housing Enterprise Oversight 2004.
The table shows an extraordinary run-up in housing prices since 1995, with
prices rising by more than 33 percent nationally, after adjusting for inflation.
This increase is without precedent in the post-war period. In the past housing
prices had largely kept even with inflation5.
Figure 1 shows the path of real housing sale prices from 1953 to the present.
(The graph uses the home purchase index from the consumer price index for the
years prior to 1975, when the House Price Index was first introduced.) As can be
seen, while there is some variation in real housing prices over this period,
there is no clear upward trend, and the 1995 level adjusted for inflation is
virtually identical with the level at the start of the period in 1953. On the
other hand, real median family income, which is also shown in this figure, has a
clear upward trend, although it does undergo a long period of stagnation from
the late seventies until the mid-nineties. These data show that there is no
reason to expect that house prices will rise in step with median family income,
as many analysts argue.
Source: BLS, Census Bureau, OFHEO, and author’s calculations, see text.
The recent run-up in home prices has led to the creation of
more than $3.8 trillion in additional housing wealth compared to a situation in
which home prices had just kept even with the overall rate of inflation6.
This additional wealth is approximately half the size of the paper wealth
created by the stock market bubble; however, it may actually have larger
implications for the economy, because housing wealth is much more evenly
distributed than stock wealth.
Not only is the gap between home prices and inflation without precedent in the
post-war period, there is also a large gap between home sale prices and rental
prices. The shelter index of the CPI has increased by just 29.9 percent over
this period, indicating a gap of more than 30 percentage points. In the past,
home sale prices and rents had largely moved in tandem. The periods in which
they diverged were reversed fairly quickly. The logic for this is
straightforward. On the demand side, if rents fall relative to sale prices then
potential homebuyers opt to rent instead. On the supply side, owners of rental
units who find that they have large numbers of vacancies and are forced to
accept lower rents try to sell off their housing. While owning and renting are
not perfect substitutes, and there are costs associated with selling off rental
units, over a long enough period of time these two markets must maintain some
balance.
It is also worth noting that the run-up in home prices has been very different
across regions. The Northeast, Mountain, and Pacific states have all seen a real
increase in home prices of more than 50 percent over the last nine years. By
contrast, the real increase in home prices in the East South Central and West
South Central states has been approximately 15 percent over the same period.
Given past patterns for home prices over the post-war period, any increase in
excess of the rate of inflation can be seen as evidence of a speculative run-up.
However, there clearly have been large differences in the extent of this run-up
across regions.
This difference is helpful in providing a basis for testing the main claim of
the Fed study - that unmeasured improvements in housing quality explain the
run-up in home prices, as measured by the HPI. However, before examining the
importance of these regional differences in price appreciation, it is worth
examining the plausibility of the study's main claim - that unmeasured quality
improvements explain most of the real increase in housing sale prices over the
last nine years.
Housing Price Appreciation and Quality Improvements - the Nationwide Story
The study's main claim is that the HPI misses quality improvements because it
tracks re-sales of the same homes, but it does not adjust for any additions or
improvements that might have been added between sales. This means that if extra
rooms or improvements like central air conditioning were added to a house
between sales, the HPI would record the resulting increase in the sales price as
housing inflation, rather then recognizing it as an improvement in the quality
of houses being sold.
The Fed study instead uses what it considers a pure measure of inflation in
housing prices, the Census Bureau's constant quality housing price index. This
index is used to measure the increase in the price of newly constructed homes.
The index is constructed through the use of hedonic regressions that include
characteristics such as the numbers of bedrooms and bathrooms, central air
conditioning and various types of construction materials. The index also
includes the value of the land on which the house is located.
The treatment of land is central. If newly constructed housing is in less
desirable locations than existing housing - either because it is further from
central cities or it is located in less desirable neighborhoods within cities -
then this index will not be picking up the full rise in existing home prices. In
effect, it will be comparing the price of newly constructed housing 50 miles
from a central city in 2004 with the cost of newly constructed housing in 1994,
that might have been 30 miles from the central city. In principle, a true price
index would track the price of the same housing, adjusting for any quality
improvements or deterioration through time.
The Census Bureau's index shows a much lower rate of inflation than the HPI,
rising by 34.3 percent from 1995 to 2003, approximately 1.2 percentage points a
year faster than the 21.9 percent cumulative rate of inflation over this period.
The study notes this relatively small gap between the increase in housing prices
and the overall CPI to argue that there is little evidence of a housing bubble.
In short, the question of whether the Census Bureau's constant quality housing
construction index or the HPI provides a better measure of inflation in housing
prices depends on:
1) the extent to which the HPI overstates house price inflation by failing to
pick up quality improvements, and
2) the extent to which the Census index understates house price inflation by
failing to pick up the fact that new housing is built in less desirable areas
than existing housing.
The first and most obvious test of the claim that the run-up in housing prices
can be explained by a rapid rate of quality improvement is to examine the Census
Bureau's data on spending on home improvements. The study's explanation would
require a rapid increase in spending on home improvements. If quality
improvements explain the recent run-up in home values, then spending on home
improvements should be a large and rising share of the value of residential
housing stock.
Table 2 shows the total value of the residential housing stock at the end of
each year along with the Census Bureau's estimate of the spending on
improvements to single family homes.
Table 2
Value of the Housing Stock and Spending on Improvements, 1991-2002
(billions of current dollars)
Value of Residential Housing Stock |
Spending on Improvements |
Spending on Improvements as a Percent of the Value of the Housing Stock |
|
1991 |
$6,709.2 |
$62.6 |
0.9% |
1992 |
7018.8 |
72.9 |
1.0% |
1993 |
7248.3 |
77.6 |
1.1% |
1994 |
7405.7 |
85.9 |
1.2% |
1995 |
7870.3 |
79 |
1.0% |
1996 |
8194.6 |
84.5 |
1.0% |
1997 |
8652.2 |
90.7 |
1.0% |
1998 |
9406.8 |
96.2 |
1.0% |
1999 |
10250.1 |
95.8 |
0.9% |
2000 |
11268.3 |
100.2 |
0.9% |
2001 |
12362.3 |
106 |
0.9% |
2002 |
13573 |
116.2 |
0.9% |
As can be seen, spending on improvements actually fell relative to the value
of the housing stock over this period, clearly ruling out the possibility that
the run-up in housing prices can be explained by a more rapid pace of quality
improvements to existing homes, as the Fed study argued. The decline in spending
on home improvements relative to the value of the housing stock suggests, other
things equal, that the quality of existing housing has actually been improving
less rapidly in the years of rapid price appreciation than it had been
previously8.
It is also worth noting the size of spending on improvements relative to the
increase in the value of the housing stock. The increases in the value of the
housing stock average 8.1 percent for the years from 1995 to 2002. Even if
spending on improvements had doubled from the early nineties to the period of
rapid price appreciation, it could only explain an increase in the annual rate
of appreciation of the housing stock of approximately 1 percentage point. This
means that the plausible effects of a more rapid pace of quality improvements
are an order of magnitude less than the actual rate of house prices appreciation
experienced over the last nine years. It should have been apparent that it is
not possible to explain house price increases of the size experienced in recent
years by increases in housing quality. This would have required a boom in home
improvement spending far in excess of anything the country has ever witnessed.
(Of course, since spending on home improvements actually fell relative to the
value of the housing stock, none of the run-up in housing prices can be
explained this way.)
In short, the evidence indicates that the difference between the Census Bureau's
constant price index and the HPI is not explained by quality improvements to
existing homes that have been missed in the HPI. Rather, it seems the
differences are attributable to fact that newly built homes are located in less
desirable areas than existing homes9.
Housing Price Appreciation and Quality Improvements - the Regional Story
There is a second important, and easily testable, implication of the Fed study's
claim that the gap between the rise in the HPI and the Census Bureau's index is
attributable to quality improvements. It implies that the most rapid rise in
spending on home improvements should be in the regions that have the largest gap
between the HPI and the Census Bureau's index. Table 3 shows the rise in the
Census Bureau's constant quality index for the nation as a whole, and by region,
for the period from 1995 to 2003.
Table 3
Increase in Census Bureau Constant Quality Price Index 1995-2003
U.S. |
34.3 |
Northeast |
42.2 |
Mid-West |
27 |
South |
27.3 |
West |
49.5 |
Source: Bureau of the Census10.
Accepting the analysis in the Fed study, the data in
table 3 can be viewed as the pure inflation in home sales prices over this
period. This means that a measure of quality improvements by region can be
calculated by dividing the price increase shown by the HPI, by the pure
inflation measure in the Census Bureau index. Unfortunately, the regions used in
the surveys do not match up precisely, but using the inflation measure from the
larger Census Bureau regions for the HPI regions should still give a reasonable
approximation of inflation in each HPI region. Table 4 shows the implicit
measure of housing quality improvement for the country as a whole and for each
region, following the assumptions of the Fed study.
Table 4
Housing Quality Improvement 1995-2003
HPI/ Census Bureau Index |
|
U.S. |
21.1% |
Northeast |
36.6% |
Mid-Atlantic |
16.1% |
East South Central |
9.9% |
West South Central |
11.2% |
South Atlantic |
26.2% |
East North Central |
18.5% |
West North Central |
26.6% |
Mountain States |
25.1% |
Pacific |
25.0% |
Table 5
Spending on Housing Improvements by Region
|
Avg 91-95 |
Avg 98-02 |
Increase |
Northeast |
$17.2 |
$20.1 |
17.1% |
Mid-West |
$19.0 |
$26.5 |
39.7% |
South |
$22.0 |
$28.3 |
28.5% |
West |
$16.7 |
$29.7 |
78.4% |
Source: Census Bureau11.
period before housing prices began to rise rapidly is essential, because the
assumption of the Fed study is that the rapid rise in housing prices since 1995
is due to a rapid rate of improvement in housing quality. This means that the
regions with the largest increase in prices in the last eight years should have
seen the largest increase in spending on home improvements.
The data in table 5 do not seem to support this conclusion. The most rapid
increase in spending on home improvements does occur in the West, which is a
region of substantial quality improvement as indicated in table 4. However, the
Mid-West also has a large increase in spending on home improvements, yet its
rate of quality improvement would be close to the national average, according to
the data in table 4. On the other hand, the Northeast, which should be near the
top in its rate of quality improvement according to table 4, is shown to have
the smallest increase in spending on home improvement in the data from the
Census Bureau. The Census Bureau data certainly do not seem to support the
contention that differences in the rate of increase in housing prices, as shown
in the HPI, can be explained by differences in the rate of quality improvement.
If There is No Bubble, Then Why Does Supply Exceed Demand?
There is one final point worth noting concerning the evidence for a housing
bubble. It is easy to verify that construction is continuing at a far more rapid
pace than growth in the demand for housing. The Census Bureau reported that the
number of vacant housing units (mostly rental units) increased by more than
750,000 units from the first quarter of 2003 to the first quarter of 2004 [http://www.census.gov/hhes/www/housing/hvs/q104tab4.html]. This increase is
large relative to the current annual rate of construction of 2 million units. It
is even larger relative to the 1.3 million implied rate of absorption. Until
there is a downward adjustment in housing prices, this rate of overbuilding will
persist. It is difficult to explain how the presence of a large and rapidly
growing number of vacant units (either owner occupied or rental) will not place
substantial downward pressure on housing prices. Those who argue against the
existence of a housing bubble must find an alternative explanation for the
country's exploding vacancy rate..
Conclusion
The Fed study's claim that the run-up in home prices since 1995 is attributable
to improvements in housing quality is not supported by the data. Spending on
improvements has actually fallen relative to the value of the housing stock in
the last eight years, indicating that this spending cannot possibly explain a
more rapid rate of increase in home prices. Furthermore, annual spending on home
improvements is far too small to explain the sorts of increases in home prices
that we have experienced since 1995 in any case. While spending on improvements
has averaged approximately $100 billion in recent years, the annual increases in
the value of the housing stock have exceeded $1 trillion. For an increase in
spending from recent levels to explain this sort of run-up in home values, home
improvement spending would have to be at least an order of magnitude larger.
In addition, there is no obvious relationship between the rate of home quality
improvement by region implied by the explanation in the Fed study and actual
spending on home improvements. In other words, if the HPI is missing quality
improvements, then the Census Bureau appears to be missing them as well in its
construction data. Finally, the Fed study has no explanation as to why the
housing vacancy is rising at a record rate from already record highs. This
pattern is consistent with a housing bubble, but not with the argument presented
in the Fed study.
In short, the Fed study really cannot explain any of the main features of the
current housing market. It does not support the case that the recent run-up in
housing prices can be explained by fundamentals rather than a speculative
bubble.
Footnotes
1. Dean Baker is co-director of the Center for Economic and
Policy Research. Debayani Kar and Mark Weisbrot gave useful comments on earlier
drafts of this paper.
2. McCarthy, J. and R. Peach, 2004. "Are Home Prices
The Next 'Bubble'?" Forthcoming, Federal Reserve Bank of New York, Economic
Policy Review [http://www.newyorkfed.org/research/epr/forthcoming/mccarthy.pdf].
3. Baker, D, 2002. "The Housing Affordability Index: A
Case of Economic Malpractice," Center for Economic and Policy Research. [http://cepr.net/housing_affordability_index.htm].
4. The Bureau of Labor Statistics' Home Purchase Index,
which in principle is a quality adjusted home sale price index, largely tracked
the overall rate of inflation from 1953 until it stopped being used in 1983.
This is in spite of a very large increase in nominal interest rates from the
fifties through the sixties and seventies.
5. The Bureau of Labor Statistics' home purchase index rose
by 225.5 percent from 1953 to 1983 (when the CPI shifted to a rental equivalence
for owner-occupied housing), while the overall CPI (chaining the CPI-URS and the
CPI-UX1) rose by 240.7 percent. In the years from 1983 to 1995, the Office of
Federal Housing Enterprise Oversight's House Price Index rose by 58.1 percent,
while the CPI-URS rose by 46.2 percent. Chaining the home sale price indexes,
gives a 256.3 percent rise in home sale prices over the 42 year period, compared
to a 251.9 percent rise in the overall price level, which amounts to annual
difference in inflation rates of 0.03 percentage points.
6. This calculation relies on the estimate of the value of
residential real estate in the first quarter of 2004 in the Federal Reserve
Board's Flow of Funds Table B100, line 4.
7. The data on the value of the residential housing stock
are from the Federal Reserve Board's Flow of Funds Table B100, line 4. The data
are taken from the Census Bureau's Expenditures for Residential Improvements and
Repaid by Property Type (all residential property) [http://www.census.gov/const/C50/histtab2.pdf].
8. It is important to remember that some amount of spending
on improvements is necessary just to keep housing at a constant quality. The
quality of housing would deteriorate through time due to wear and tear, if there
were no offsetting improvements.
9. It is also worth noting that the South, the region which
has seen the smallest increase in housing prices, has the largest weight in the
Census Bureau index (0.4). This is attributable to the fact that more than 40
percent of new construction spending takes place in the South. Of course, an
index that is measuring price changes for existing housing should reflect the
value of the existing housing stock, which would assign a much smaller weight to
the South.
10. This data is taken from the "Price Indexes of New
One-Family Houses Sold Including Lots, [http://www.census.gov/const/price_indexes.pdf].
11. These data are taken from the Census Bureau's
Expenditures for Owner-Occupied One-Unit Properties by Region [www.census.gov/const/c50/table_3pdf].