Recently, Yale professor Robert Shiller has developed a long-term index of real housing prices for the U.S.¹ This index is shown in Figure 1. In my prior discussion of trends in real estate values, I presented a variety of information providing evidence for an 18-year cycle in real estate/land values.² What was not available to me was national data concerning real estate values over a long period of time. The closest thing to this was an index constructed by economist Daniel McFadden,³ which is also shown in Figure 1. This index measures the rise in construction costs relative to overall prices for the national economy. Thus, it measures the value of buildings implicitly, in terms of construction costs. It does not consider the value of the land, only what is built on it.

Figure 1. The Shiller house index compared to the McFadden construction cost index

The two indices show a broadly similar rise over the long run, but they differ significantly in the shorter term. In particular, any cyclical structure shown by one index is not supported by the other, making the idea of semi-regular cycles in real estate (at least since 1890) problematic. Since I have not employed the idea of real-estate cycle much in my forecasts, the likelihood that real estate or Kuznets cycle doesn't really exist is not much of a problem. Since the Shiller index is an explicit index of house prices, which includes land, whereas the McFadden index only deals with construction costs as an implicit measure of housing values, I prefer the Shiller index and will work with it.

The National Association of Realtors reports data on the monthly median price of existing homes.⁴ The Census bureau reports data on the median price of new homes.⁵ Both of these sets of data are plotted along with the Shiller index in Figure 2. The data for existing homes tracks Shiller's index fairly well. Prices for existing homes were range-bound over the 1973-2000 period--only rising above this range in the last five years. The same thing is true for Shiller's index. In contrast, real prices for new homes showed a rising trend from the 1960's through the 1980's that was not shown by the Shiller index. This difference can be explained by the rising trend in new home size and quality over the past four decades. Part of the rising prices of new homes reflects rising quality and size of new homes and not price inflation per se. Shiller's index is adjusted to control for size and quality and so it is expected that the two measures should diverge.

Figure 2. The Shiller index compared to trends in prices of new and existing homes

All three plots show cyclical alignment; the housing booms of the late 1970's, late 1980's and today show up on all three measures. This result shows that the Shiller index correctly describes recent cycles, suggesting that it probably identified historical cyclical behavior correctly also. That is, there likely never was a regular Kuznets cycle.

The Shiller index shows a 50% rise from its mid-1990's low through 2004. Over the same period new home prices rose 33% and those of existing homes 38%. Adjustments for quality cannot explain why the Shiller index should have risen faster than new home prices in the last decade and lagged behind in the decades before. The more rapid rise of the Shiller index compared to median prices of existing homes is also hard to explain. Since the quality of existing homes should rise slowly with time, one would expect existing home prices to rise somewhat faster than the quality-adjusted Shiller index.

There is the issue of sampling. The NAR data simply record the median sales price for those home that actually sold at that particular time. Since houses (unlike shares or commodities) are not interchangeable, the sales of some homes at a particular time may not be a good reflection of the value of all the homes not on the market. One can get around this issue in various ways. For example, one can look at sales prices for the same home over time assuming it changes hands fairly frequently. It is not unlikely that a more carefully constructed index could show discrepancies from the trend shown by the NAR data. One would expect a fairly close correspondence between such an index and the existing home trend over the longer term though. Fortunately, this is the case. For this reason and because the Shiller index is the only continuous indicator of quality-adjusted real estate prices available, I prefer it to other indicators of housing values.

I would like to develop some valuation tool for the "housing index" that is analogous to my concept of price to resources (P/R) which I use for valuing the S&P500 stock index.⁶ The financial value of a business is fundamentally about its ability to generate profits. The idea behind R is that for the economy as a whole, the ability for business (in aggregate) to generate profits is dependent on the amount of resources with which they can work. R provides a way to measure these resources for the S&P500 and so P/R provides a fundamental type of valuation for this index.

The value of a house is fundamentally about its ability to provide a residence. A person who does not possess a house (but has money) can obtain a residence by renting an apartment or house. Thus, the cost of rent should represent the fundamental value of housing in terms of a stream of rent payments. The present value of this stream, what I call the use value (UV), should represent the fundamental value of a house.

Figure 3. Calculation of UV for 1913-2005

To calculate UV I needed several pieces of information in addition to the rent value. These additional data are the discount rate and discount period for the present value calculation. The discount rate would be the return that the money needed to buy a residence would earn if it were invested elsewhere. This alternate investment would have to be of comparable safety to the investment in housing. A logical choice would be investment in other people's housing, that is, mortgage loans. Thus, the discount rate I chose was mortgage rates.⁷

The discount period is the length of time one would be willing to tie up one's money in the alternate investment. That is, it is the duration of typical mortgages. Figure 3 shows how I calculated UV for American housing for the period 1913 to present. Shown is an index from the Bureau of Labor Statistics that tracks real rent for primary residence.⁸ Also shown are interest rates and the length of a typical mortgage. Interest rates come from the Federal Reserve of St. Louis.⁹ The lengths used were 7 years before 1943 and 30 years after 1980, which reflect common mortgage durations for those times. Fifteen year mortgages were introduced in the 1930's. I assumed these did not become popular until the post-war housing boom and I stepped up the length one year every year from 1943 to 1950. I kept the length at 15 years throughout the 1950's and then stepped it up one-half year per year over the 1960's. During the inflationary 1970's I stepped up length one year per year until it reached 30 years in 1980. This particular profile was chosen in an attempt to fit the shape of the UV profile to that of Shiller's index as closely as possible while still keeping the assumed mortgage duration reasonably consistent with the changes in duration that have been observed historically.

Figure 4. Real Price, Real Use Value and P/UV

UV was set equal to Shiller's index in 1913. Both indices are shown in Figure 4. Also shown is the ratio between Shiller's index (P) and UV. This ratio (P/UV) represents a "valuation" for housing in the same way as P/R does for stocks. Like P/R, P/UV is not suitable for short-term valuation, prices can remain high or low in terms of either for a long time. Rather it is a measure of the extent of long-term swings in price and can be used to identify extremes in the market that are representative of secular trends.

P/UV is different from P/R in that UV can be depressed temporarily because of macroeconomic factors unrelated to the fundamental value of a house. During inflationary periods, when interest rates are high (e.g. the early 1980's) UV will be artificially depressed and P/UV will be very high. This situation is analogous to P/E valuation for stocks, which can be artificially high during recessions when E is depressed. Just as stocks were not overvalued in 2002 despite sky-high P/Es, neither were houses overvalued in the early 1980's, when P/UV reached its all-time high. This fact is shown by the rise in stock and house prices in the years after these peak valuations.

Valuations today do not reflect a depressed UV. In fact UV is higher than it has ever been. To compare today's housing valuations to those of the past, one should ignore those periods when UV was depressed.

Table 1 shows P/UV in 2005 and past peak values. Also shown is the P/R value in 2000 and previous peak values. Based on the data in Table 1, housing prices are high today, just as stock prices were high in 2000. In 2000, P/R was 1.58 standard deviations above the mean. In 2005 P/UV was 1.48 standard deviations above the mean. That is, housing is about as overvalued today as stocks were in 2000. Real stock prices (on an annual average basis) declined by 37% after the 2000 peak, about the same as the 35% average decline following a peak in P/R. Real housing prices have declined by an average of 13% following previous peaks in P/UV. If the stock market example from 2000 is a valid guide we might expect an average decline following a 2005 peak in housing prices (assuming that the peak price recorded then was the top).

Table 1. Past valuation peaks in stocks versus housing

A 13% real decline over the next few years roughly translates to flat nominal prices. Thus, the most likely aftermath of the present housing bubble would be a cessation of rising prices. The impact of this on the economy at this time does not have to be severe. Business investment spending should compensate for slowing growth in consumer spending until the Juglar investment cycle heads into a downturn. The recession that results should be more severe than the last one because falling interest rates will be no more capable of sparking a rally in housing prices than they were with stocks during the last recession.

Should real house prices decline by 13% over the next few years while real rent remained at current levels and interest rates stayed low, P/R would fall below 1.4. Should the first two of these happen while interest rates continued to fall as expected for the Kondratiev downwave, P/R could fall further, perhaps all the way back to one and the housing bubble would deflate with no fall in nominal house prices.

Another possibility would be increased inflation while interest rates remained low like happened in late in the last downwave (the late 1940's). Suppose house prices, being in a secular bear market following the peak in P/UV, fail to keep up with inflation, while rent does. In this case P/UV could also decline despite modestly rising nominal house prices. Such an uncoupling of interest rates from inflation is an expected outcome of the Kondratiev downwave (see the discussion of interest rate regimes in reference 10).

A third scenario would have nominal house prices drop substantially, perhaps not now but in a future recession. Such an outcome would drop real prices to an even larger degree, directly decreasing P/R to "normal levels". This outcome would be expected were interest rates to rise substantially. Since secularly rising interest rates are a feature of Kondratiev upwaves, not downwaves, I hold this scenario to be less likely than the others.

Noticeably absent is a scenario in which nominal house prices continue to rise. There are two ways this could happen without P/R rising along with it. One would be a reduction in interest rates with no decline in real rent. Since rising house prices reflect rising demand for houses, one would think this would reduce demand for rental units, lowering real rent. Indeed, this has been the case in recent years. This scenario is not likely.

Figure 5. Post-1998 P/UV values for different duration assumptions

The other way is an increase in mortgage duration, raising the value of US for a given interest rate and rent value. Media reports about the rising popularity of "interest-only" mortgages or 40+ year mortgages suggest that this might indeed be happening. Figure 5 explores this scenario by plotting P/UV values for 1998 to 2005 for various assumed mortgage durations. By increasing mortgage term from 30 years to infinite duration (interest only) P/US values can be kept low to 2004. Even with an infinite duration P/UV in 2005 was still very high. This exercise suggests that increasing mortgage duration cannot serve a justification for high prices today.

For these reasons I do not believe that current prices levels are justified and a continuation of the housing rally is unlikely.

References:
1 Robert J. Shiller (2006) "Long-Term Perspectives on the Current Boom in Home Prices", The Economists' Voice: Vol. 3: No. 4, Article 4 (http://economistsview.typepad.com/economistsview/2006/03/ shiller_longter.html).
2 Alexander, Michael A., "Generations and Business Cycles", Safehaven, November 2002 (www.safehaven.com/archive-7.htm)
3 Daniel McFadden, "Demographics, the housing market, and the welfare of the elderly", in David A Wise ed. Studies in the economics of aging, Chicago IL, University of Chicago Press, 1994, p241.
4 National Association of Realtors (www.realtor.org/research/index.html)
5 Bureau of the Census (www.census.gov/const/uspricemon.pdf)
6 Alexander, Michael A., "Secular Market Trends" Safehaven, March 2001 (www.safehaven.com/archive-7.htm)
7 Before 1964, 85% of the Aaa corporate rate was used as a mortgage proxy. After 1964 the FHA rate (1964-71) and FHLMC Rate (after 1971) was used.
8 Bureau of Labor Statistics (http://data.bls.gov/cgi-bin/srgate) Series CUUR0000SEHA, Rent of primary residence, US city average, 1982-84 = 100.
9 Federal Reserve Bank of St. Lous (http://research.stlouisfed.org/fred2/categories/22)
10 Alexander, Michael A., "Inflation, Monetary Stimulation and Interest Rates - A Cycle Perspective" Safehaven, March 2005 (www.safehaven.com/archive-7.htm)