End in sight for housing troubles?

D. L. Chertok November 5, 2012

Abstract A historical relationship between home prices and family income is examined based on more than 40 years of data. A new home aordability ratio based on the average home price, family income and mortgage rates is analyzed in the historical context. This indicator is used to gauge the current state of the residential housing market in the United States. Historical data points to an imminent but slow recovery in the housing market over the next few years.

Three years after the Great Recession ocially ended, the state of the US housing market remains an area of signicant concern and continues to attract a lot of attention from academics and practitioners. In a recent article Mizuno and Tabner (2011) discuss housing price trends in the residential markets of the US, UK and Japan using the Housing Price Index (HPI) deated by gross domestic product (GDP) per capita as a measure of home aordability. Beracha and Hirschey (2009) examine the trend in housing aordability in the United States based on the Housing Price Index (HPI) and average income per capita. While the authors present compelling arguments for the use of the aforementioned indicators, it is important to note that the indices in question are not free from limitations. First, HPI data is available only from 1975 onward. Second, it covers only those sales and renancings that have been purchased or securitized by Fannie Mae (Federal National Mortgage Association) or Freddie Mac (Federal Home Loan Mortgage Corporation). Third, even though per capita income and per capita GDP are valid proxies for housing aordability, family income (DeNavasWalt, C., Proctor, B. and Smith, J. (2012)) reects the purchasing power more accurately when it comes to securing a mortgage. Finally, it is easier to reduce
D. L. Chertok, Ph. D., CFA, (daniel chertok@hotmail.com) is a quantitative investment professional. This is a revision of an article presented at the 2011 Meeting of the Midwest Econometrics Group http://faculty.chicagobooth.edu/midwest.econometrics/Papers/MEGChertok.pdf

the number of possible price outliers by using the median, rather than average values1 . A popular alternative is the S&P/Case-Shiller Home Price Index, a widely followed benchmark of residential real estate prices in the United States based on repeat sales of existing homes. Data is available since 1987 (see, e.g., Standard and Poors (2012)) and is not ination-adjusted. Another publicly available index is published annually by the U.S. Census Bureau and reects nationwide prices of new homes dating back to 1963. This index covers a longer historical period and is less volatile than its two other counterparts; in addition, it includes sales nanced through sources other than Fannie Mae and Freddie Mac. Despite those dierences, as can be seen from Fig. 12 , qualitatively all three indices reect the same developments in the U.S. housing market over the last 20 years3 .
Comparison of New Home Price, S&P/Case-Shiller and HPI indices, 1987 = 1
New Home S&P/Case-Shiller HPI

1.75

normalized value

1.5

1.25

0.75
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

year

Figure 1: Comparison of the New Home Price, Housing Price and S&P/CaseShiller Indices, 1987 - 2011. Shaded areas represent recessions. For the purpose of this paper, the New Home Price (NHP) index is the most representative measure. It illustrates the evolution of the U.S. residential real estate market valuations and reects the sector of this market that arguably has the most impact on other industries, especially construction. In Fig. 1, all three indices are normalized to their 1987 values (set to 1) and adjusted for ination using the CPI-U-RS as the consumer price index4 , also 2

normalized to its 1987 value. All three peaked in 2005 - 2006 and began their rapid decline in 2007. HPI and NHP followed each other very closely until 2000 and began to diverge in 2005. In a reversal of this trend, S&P/Case-Shiller converged to HPI and NHP in 2011. Per capita income and median family income for 1967-2011 are compared in Fig. 2. Both are ination-adjusted and normalized to 1967. Evidently,
CPI-adjusted income per capita and median family income, 1967 - 2011, 1967 = 1
"income per capita" Linear ("income per capita") median family income

1.95

y = 0.0226x + 1.0848 R2 = 0.9385

1.75

normalized value

1.55 y = 0.0055x + 1.039 R2 = 0.7698

1.35

1.15

0.95

0.75 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011

year

Figure 2: Comparison of income per capita and median family income, 1967 2011. Shaded areas represent recessions. per capita income grew faster than family income, which may be explained by While both indices exhibit the same an overall reduction in family size5 . qualitative behavior6 , median family income (DeNavas-Walt, C., Proctor, B. and Smith, J. (2012)) is more relevant to the issue of housing aordability and is less volatile. Using the family income, NHP and mortgage rate, an aordability ratio ( Chertok (2009) ) can be constructed to gauge the state of the housing market7 . Historical data for this ratio suggests that, while housing aordability improved dramatically since 2005 indicating a possible bottom in residential real estate, recovery in this market over the next several years is likely to occur at a gradual pace.

Historical average aordability of a new home in the U.S.

During the period from 1963 to 2011 median new home prices in the US grew at an average rate of 2.25% per year while median household income grew at an average rate of 0.58% per year adjusted for ination; the corresponding statistics is presented in Table 1. Fig. 3 depicts the growth rates of median new home prices and median household income. Consistent with conventional wisdom,
Inflation-adjusted housing prices and median family income 1967 - 2011, 1967 = 1
CPI-adjusted housing prices normalized income

2.5

2 Mean house price

y = 0.0218x + 0.9326 R2 = 0.8658

1.5

1 y = 0.0055x + 1.039 R2 = 0.7698 0.5

0
1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011

year

Figure 3: Evolution of the median new home price and median household income in the US, 1967 - 2011. Shaded areas represent recessions. real median family income rises and falls with economic cycles. Over the last 40 years, its peaks preceded subsequent recessions, and its troughs coincided with or trailed the recessions (with the notable exception of the 2001 recession, due to the coincident easing of credit availability and proliferation of adjustable rate mortgages). The same holds true for the real new home price. It is evident from Fig. 4 that the dierential between the rate of growth of new home prices and the rate of growth of the median family income reaches a trough during a recession (the trough corresponding to the 2001 recession is delayed by two years). After attaining its all-time high in 2005, this spread contracted very signicantly reaching its lowest level in 38 years in 2008 before bouncing back

Table 1: US new home price and income statistics, 1963 - 2010.

Median income, CPI-adjusted to 2011, $ Median price, CPI-adjusted to 2011, $ Aordable monthly payment, $ 30-year xed mortgage rate, % Mortgage payment, $, 30-yr. xed, 20% down Affordability ratio

Year

1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 38,771 40,442 41,945 41,620 41,215 42,980 43,848 42,459 41,348 42,034 42,300 43,937 43,814 42,429 41,724 41,613 41,322 42,605 43,402 44,939 45,502 45,852 46,670 46,049 44,726 44,359 44,143 44,636 46,034 46,704 47,665 49,397 50,641 50,557 49,455 48,878 48,835 48,665 49,202 49,568 50,233 50,112 49,777 49,445 50,054 7.38 8.04 9.19 9.04 8.86 8.84 9.63 11.19 13.77 16.63 16.08 13.23 13.87 12.42 10.18 10.20 10.34 10.32 10.13 9.25 8.40 7.33 8.35 7.95 7.80 7.60 6.94 7.43 8.06 6.97 6.54 5.82 5.84 5.86 6.41 6.34 6.03 5.04 4.69 4.45 981 1,024 1,062 1,053 1,043 1,088 1,110 1,075 1,047 1,064 1,071 1,112 1,109 1,074 1,056 1,053 1,046 1,078 1,099 1,137 1,152 1,161 1,181 1,166 1,132 1,123 1,117 1,130 1,165 1,182 1,206 1,250 1,282 1,280 1,252 1,237 1,236 1,232 1,245 1,255 1,271 1,226 1,218 1,186 1,168 734 866 967 965 1,011 1,047 1,200 1,404 1,567 1,825 1,676 1,453 1,548 1,426 1,284 1,413 1,484 1,515 1,455 1,272 1,163 1,067 1,185 1,147 1,152 1,152 1,112 1,207 1,303 1,181 1,191 1,122 1,241 1,313 1,377 1,337 1,168 980 948 916

Median new home price, non-adjusted, $ 18,000 18,900 20,000 21,400 22,700 24,700 25,600 23,400 25,200 27,600 32,500 35,900 39,300 44,200 48,800 55,700 62,900 64,600 68,900 69,300 75,300 79,900 84,300 92,000 104,500 112,500 120,000 122,900 120,000 121,500 126,500 130,000 133,900 140,000 146,000 152,500 161,000 169,000 175,200 187,600 195,000 221,000 240,900 246,500 247,900 232,100 216,700 221,800 227,200 115,669 119,821 124,877 129,934 133,651 139,938 138,845 120,955 124,791 132,697 147,051 147,668 149,376 158,860 164,981 176,223 181,607 167,879 163,489 155,080 161,609 164,738 168,040 180,128 197,915 205,524 210,159 205,020 193,252 190,830 193,888 195,089 196,216 199,836 203,992 210,141 217,320 220,722 222,571 234,536 238,462 263,145 277,535 274,971 268,904 242,450 227,225 228,796 227,200 1.48 1.28 1.11 1.08 1.05 1.02 0.93 0.79 0.69 0.58 0.63 0.72 0.70 0.77 0.89 0.82 0.78 0.78 0.80 0.89 0.97 1.05 0.95 1.02 1.03 1.05 1.12 1.06 0.98 1.06 1.04 1.10 0.99 0.95 0.91 0.95 1.05 1.24 1.25 1.28

CPI (Dec 1977 = 100) 51.4 52.1 52.9 54.4 56.1 58.3 60.9 63.9 66.7 68.7 73.0 80.3 86.9 91.9 97.7 104.4 114.4 127.1 139.2 147.6 153.9 160.2 165.7 168.7 174.4 180.8 188.6 198.0 205.1 210.3 215.5 220.1 225.4 231.4 236.4 239.7 244.7 252.9 260.0 264.2 270.1 277.4 286.7 296.1 304.5 316.2 315.0 320.2 330.3

Difference between growth rates of housing prices and median family income 1967 - 2011, 1967 = 1
0.15

0.1 difference in growth rates

0.05

-0.05

-0.1

-0.15
1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

year

Figure 4: Relative dierence in ination-adjusted growth rates between the median new home price and median family income, 1968 - 2011. Shaded areas represent recessions. into positive territory in 2010 and settling back near zero in 2011. A substantial correction in the residential real estate market appears to have passed through its most dramatic stage by 2010. The latest available annual data is historically consistent with continuing economic expansion during which home price growth rate should return to a historically more sustainable level relative to family income. The third important ingredient in the housing aordability mix, besides prices and income, is the availability of nancing, i.e., the average mortgage rate8 . The rise (and fall) of mortgage securitization played a crucial role in improving the accessibility of credit for the average American family. Increased liquidity, thanks to wider availability of credit through securitization, has been instrumental in keeping mortgage rates low. From their heights in 1982, mortgage rates have steadily trended down in the last 30 years helping make the American Dream a reality for the majority of the population (see Fig. 5). Based on the above-mentioned three factors, an aordability ratio can be

Average national mortgage rates, 1972 - 2011

16.2%

14.2%

12.2% rate

10.2%

8.2%

6.2%

4.2%
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

year

Figure 5: Average national 30-year conventional mortgage rates 1972 - 2011. Shaded areas represent recessions. constructed as follows9 : pa pr where: pa pr I H rm aordable monthly payment based on income, required monthly payment based on mortgage amount, tenor and mortgage rate (see, e.g., Fabozzi (2005)), annual family income, home purchase price, mortgage rate. = = = pa , pr 0.28I , 12 0.8H
rm 12

(1) (2) 1+
rm 12 rm 12 360 360

1+

(3)

The preceding equations assumed a conventional 30-year xed-rate mortgage with monthly advance payments of interest and no principal curtailment, a conservative underwriting ratio of principal and interest to total before-tax income of 0.28 and a down payment of 20%. The eects of property taxes and utility costs are omitted for the ease of exposition. As can be seen from Fig. 6, in 2011 historical housing aordability reached levels not seen since 1973, thanks to rapidly falling home prices and mortgage rates at their historical lows. One could reasonably conclude that, as home
Affordability ratio, 1972 - 2011

1.60 1.40 affordable to required payment 1.20 1.00 0.80 0.60 0.40 0.20 0.00
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

year

Figure 6: Historical ination-adjusted home aordability ratio, 1972 - 2011. Shaded areas represent recessions. aordability continues to climb, the housing market will improve. It is not obvious, however, whether this improvement will be partially oset by declining family incomes and rising mortgage rates, should ination spawned by government spending become a reality. Factor sensitivity of home aordability is discussed in the next section.

What matters more: price, income or mortgage rates?

As follows from (1) - (3), home aordability is linearly dependent on the family income. It is also inversely proportional to the home price since the required monthly payment is a fraction of the total home price. As a rst approximation, aordability is inversely proportional to the mortgage rate. The sensitivities of home aordability are computed below: I H rm , I = , H pr = pr rm = (4) (5) (6)

Of most interest to us are the sensitivities of the aordability ratio to relative changes in home prices , income and mortgage rates. Computing the elasticities of home aordability with respect to these parameters, we obtain:
I I H H rm rm

1,

(7) (8)
rm 361 1 + 361 rm 12 12 rm rm 360 1 + 12 1 12

= 1, = 1+ 1+ .

(9)

As follows from (7) - (9) the elasticity of the aordability ratio with respect to income by absolute value is exactly the same as its sensitivity to the housing price. At the end of 2011, the average conventional 30-year xed mortgage rate was 4.45%. Substituting this into (9), we obtain
rm rm

= 0.52 .

(10)

In relative terms, home price and income currently have more inuence on home aordability than mortgage rates. This gap has been widening for the last 30 years ever since the mortgage rates reached their peak as can be seen in Fig. 7. As expected, the absolute value of elasticity of the aordability ratio with respect to the mortgage rate in 2011 was at a historical low since the mortgage rate itself was at a historical low. An expected rise in mortgage rates, however, would lead to an increase in the corresponding elasticity of the aordability ratio (although this elasticity will not reach 1 unless mortgage rates exceed their levels of early 1980s).

1 0.95 0.9 0.85 0.8 elasticity 0.75 0.7 0.65 0.6 0.55 0.5 1972 1974 1976 1978

Absolute value of elasticity of the affordability ratio with respect to mortgage rates, 1972 - 2011

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

year

Figure 7: Historical absolute value of elasticity of the aordability ratio with respect to mortgage rate, 1972 - 2011. Shaded areas represent recessions. Another important consideration is the relative probability of a signicant relative change in the underlying variables behind the elasticity ratio. As can be seen from Fig. 8, in the past 35 years year-over-year relative rates of change of the mortgage rate were substantially higher than those of the median home price and family income. As evident from Table 210 , the rate of change of the mortgage rate exhibits more than double the volatility of the rate of change of home prices and almost ve times the volatility of the rate of change of family income. Table 3 indicates that a 20.8% increase in the mortgage rates (amounting to 93 b.p. at the 2011 level) would cause the aordability ratio to decrease by approximately 9.9%. At the same time, a 9.8% decrease in the real home price, which historically is almost equally as likely, would cause the aordability ratio to increase by 10.9%. A similarly historically likely 4.3% increase in the real family income would raise the aordability ratio by only 4.3%. From the probabilistic point of view, the change in mortgage rates has almost the same eect on the aordability ratio as the change in home prices.

10

2010

Rates of change of mortgage rates, income and home price, y-o-y, 1972 - 2011
Income change, y-o-y 30% Home price change, y-o-y mortgage rate change, y-o-y

25%

20%

15%

10% rate of change, %

5%

0%

-5%

-10%

-15%

-20%

-25%
1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011

year

Figure 8: Rates of change of mortgage rates, median family income and home price, 1973 - 2011. Shaded areas represent recessions. Table 2: Mean and standard deviation of year-over-year rates of change of the median home price, family income and mortgage rates, 1973 - 2011.
Statistic Rate of change of median home price, y-o-y, % 1.5 4.8 -9.8 10.4 Rate of change of median family income, y-o-y, % 0.4 2.2 -3.2 4.3 Rate of change of mortgage rates, y-o-y, % -0.8 10.1 -17.7 20.8 Rate of change of aordability ratio, % 0.0 9.0 -14.9 15.0

Mean Standard deviation 5th percentile 95th percentile

What lies ahead?

Judging from historical data, three base scenarios can be realized with diering degrees of likelihood. The still water scenario reects at housing prices, low income growth and a small increase in mortgage rates. The full steam ahead scenario corresponds to a substantial uptick in housing prices, moderate income growth and lower mortgage rates. The doom-and-gloom scenario is a combina-

11

Table 3: The eects of home price, family income and mortgage rate changes on the aordability ratio. Scenario Variable change, Rate of change of aordability % ratio, % Home prices down, -9.8 10.9 5th percentile Family income up, 4.3 4.3 95th percentile Mortgage rates -17.7 9.9 down, 5th percentile

tion of falling housing prices and family incomes exacerbated by rising mortgage rates. The resulting aordability ratios are presented in Table 4. The still waTable 4: The still water, full steam Rate of SceRate of Menario change dian change of median of home family median price, income, $ home y-o-y, % price, y-o-y, % base n/a 227,200 n/a (2010) still 0.0 221,800 0.5 water 5.0 238,560 5.0 full steam ahead doom-5.0 215,840 -5.0 andgloom ahead and doom-and-gloom scenarios. Fam- Rate of Mort- Aor- Rate of change dchange gage ily of in- of mort- rate, ability aordra% gage come, ability tio rates, $ ratio , y-o-y , % % 50,054 n/a 4.45 1.28 n/a 50,304 5.0 4.67 1.25 -2.1

52,557

-10.0

4.01

1.35

5.4

47,551

10.0

4.90

1.11

-5.0

ter scenario results in a modest decline in the aordability ratio, the full steam ahead scenario leads to a substantial increase in the aordability ratio, and the doom-and-gloom scenario leads to a substantial decrease in the aordability ratio. Historical data presented in Fig. 9 points to the rst scenario as the most likely one to be realized. A local peak in the rate of growth of the aordability ratio tends to lead (or coincide with) a local peak in the rate of growth of housing prices. By historical standards, a substantial further increase in the aordability ratio appears unlikely. If this pattern were to hold, housing prices could be expected to increase within the next two years but would subsequently 12

Rates of change of home price and affordability ratio, y-o-y, 1972 - 2011
Home price change, y-o-y affordability ratio, y-o-y

25% 20% 15% 10%

rate of change, %

5% 0% -5% -10% -15% -20%

1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011

year

Figure 9: Rates of change of home price and aordability ratio, y-o-y, 1972 2011. Shaded areas represent recessions. grow at a slower rate. An increase in mortgage rates spawned by an uptick in ination and a decline in family income caused by an economic downturn could hamper the recovery in the housing market even further.

Conclusion
An unprecedented recent drop in home prices has increased the aordability of the average house for the average family in the US as measured by the aordability ratio to historically high levels. This increase can be reasonably expected to increase demand and, as a result, lead to higher home prices. Historical data implies that such an increase may be gradual with a possibility of it being hampered by rising mortgage rates and falling incomes. Since historically housing prices tend to overreact to market dislocations, it may take several years for the market to return to sustained growth.

13

Notes
1. Consider, for example, a situation where one $1,000,000 earner lives in a $1,000,000 house, one $300,000 earner lives in a $400,000 house, and two $100,000 earners live in two $200,000 houses. The average home price in this example is $450,000, and the average income is $375,000, or 5/6 of the average price. The median price is $300,000 and the median income is $200,000, or 2/3 of the average price. The median here paints a more representative picture of housing aordability, at least from an underwriters point of view. 2. This and subsequent gures are superimposed over the time frame of recession periods as determined by the National Bureau of Economic Research (2012). 3. In fact, pairwise correlation coecients between the three indices exceed 95%. 4. CPI-U-RS index published by the Bureau of Labor Statistics (Bureau of Labor Statistics (2012)). Its use is recommended for historical research whenever the current rental-equivalence method of measuring the cost of homeownership needs to be extrapolated to periods prior to 1983. 5. As an example, consider a family of four that went from making $50,000 a year to $100,000 a year. The increase in per capita income is 100%. If, during the same period, the family split into two families earning $60,000 and $40,000 with two members each, median family income would remain at $50,000. 6. They are 96% correlated. 7. For a similar quantity that does not factor in mortgage aordability see Haines and Rosen (2007) 8. Contract interest rates on commitments for 30-year xed-rate conventional rst mortgages. Source: Primary Mortgage Market Survey R data provided by Freddie Mac. 9. The calculation assumes end-of-month payments for simplicity. 10. All variables are ination-adjusted; statistics are calculated for all avaliable histories for each variable.

References
Beracha, E. and M. Hirschey (2009). When Will Housing Recover? Financial Analysts Journal 65 (2), 3647. Bureau of Labor Statistics (2012). Updated CPI-U-RS, All Items and All Items Less Food and Energy, 1978-2011 . http://www.bls.gov/cpi/ cpiursai1978_2011.pdf. Chertok, D. (2009). Safe house: The housing market and the end of the recession. The Investment Professional 2 (3), 1820. DeNavas-Walt, C., Proctor, B. and Smith, J. (2012). Income, Poverty, and Health Insurance Coverage in the United States: 2011. US Census Bureau. http://www.census.gov/prod/2012pubs/p60-243.pdf. 14

Fabozzi, F. (2005). McGraw-Hill.

The Handbook of Fixed Income Securities (7th ed.).

Haines, C. and R. Rosen (2007). Bubble, Bubble, Toil, and Trouble. Economic Perspectives. Federal Reserve Bank of Chicago 31 (1), 1635. http://www.chicagofed.org/webpages/publications/economic_ perspectives/2007/1qtr2007_part2_haines_rosen.cfm. Mizuno, M. and I. Tabner (2011). The margin of safety and turning points in house prices: Observations from three developed markets. Financial Analysts Journal 67 (3), 7693. National Bureau of Economic Research (2012). Business Cycle Expansions and Contractions. http://www.nber.org/cycles.html. Standard and Poors (2012). S&P/Case-Shiller Home Price Indices. http://us.spindices.com/indices/real-estate/ sp-case-shiller-20-city-composite-home-price-index.

15