Morris DavisBy Morris A. Davis
Assistant Professor, Department of Real Estate

Since joining the Wisconsin School of Business, Assistant Professor Morris Davis has maintained three databases that he created (with coauthors) over the 2004-2006 period while serving at the Federal Reserve Board. He created these databases in order to provide estimates of data he viewed as key to understanding house prices that were not available elsewhere at the time.

The databases, now available at the Lincoln Institute of Land Policy’s website ( in Resources and Tools,  explore:

  1. Rent-Price Ratio: The ratio of rents to prices in the aggregate for U.S. owner-occupied homes.
  2. Aggregate U.S. Land Prices: The price and value of land and location in the aggregate for all U.S. homes (owner-occupied and rental).
  3. Metro Area Land Prices: The price and value of land for owner-occupied homes for 46 large U.S. metro areas.

The first data set I worked on involved the ratio of rents to prices for owner occupied homes, the so-called "rent-price" ratio. The rent-price ratio is a cap rate for owner-occupied homes. These data were not publicly available anywhere because the rent accruing to owner occupied homes must be estimated. My coauthors and I estimate these rents by using micro data on rental properties from the 1960, 1970, 1980, 1990, and 2000 Decennial Census of Housing. We compute cap-rates quarterly, and the data now track the 1960:1 – 2009:1 period.

As we teach our students, cap rates are useful metrics for gauging valuations, because, under certain conditions, it can be shown that the cap rate is equal to the required return on the asset (housing in this case) less the expected growth rate of rents. If we notice the rent-price ratio fall, it implies that either (a) required returns to housing fell, (b) the expected growth rate of rents increased, or (c) some combination of a and b.

The Price of Land

The second and third data sets are about the price of land. Honestly, I got the idea to study the price of land in a moment of panic. I was informed in August of 2003 that I had to brief former Fed Chairman Alan Greenspan in October of 2003. I wanted to brief Greenspan on whether or not I thought housing was "overvalued," but didn’t have anything new to say (thus, the panic). So, I sat down to think through what it meant for housing to be overvalued and I came up with the following conclusion: A house can be thought of as a bundle of land and structures. Structures are perfectly reproducible, so there should not be any "bubble" in structures prices, much in the same way there are rarely bubbles in used car prices – the price of a recent vintage of used car is pinned down by production costs on new cars. Thus, the only way for there to be a bubble in house prices (outside of temporary surges in construction costs due to shortages of labor or materials) is for there to be a bubble in land prices. If land prices were displaying some historically unusual trends, then this could be evidence of a "bubble" or mispricing or some such thing.

So, I figured I would brief Greenspan on the price of land.

The one problem with this is that there were no publicly available data set on the price of land (except for some data on farm land maintained by the USDA). This might seem like an insurmountable hurdle to some, since it was literally impossible to brief Greenspan on the price of land without data on land prices! So, my goal was to figure out how to measure land prices. This was the hard part, since no one had any clue how to do it (outside of studying data on raw land sales, and these data are essentially non-existent). I had an "AHA!" moment. I determined that it is possible to measure the price of land as long as it is assumed that the value of housing is equal to the exact sum of the replacement cost of the housing structure and the market value of the land (and location) underneath. Under this assumption, if you know (i) growth in house prices, (ii) growth in structures costs, and (iii) the share of the value of housing attributable to the replacement cost of structures, then you can measure growth in the price of land as a residual in an equation. It seems so obvious in retrospect, but no one had thought of this previously. Anyway, data on (i), (ii), and (iii) are publicly available, or could be quickly constructed, and this insight lead to my two quarterly data sets on land: One for the aggregate United States, now tracked over the 1975:1 – 2009:1 period (with Jonathan Heathcote, now at the Federal Reserve Bank of Minneapolis) and on for the 46 large metro areas in the US (with Michael Palumbo at the Federal Reserve Board), now tracked over the 1984:4 – 2009:1 period.

Irregularity in House Prices and Housing Returns

In retrospect, all these data sets pointed to an enormous nationwide irregularity in the pricing of housing over the 1997-2006 period. I could go on for a while about this, but I'll try to be brief. I'll list four points summarized from the three data sets that suggest there was some enormous nationwide irregularity in house prices and housing returns over the 2000-2006 period:

  1. From 1960 – 2000, the rent-price ratio in the United States was about 5 percent, give or take 0.5 percentage points. From 2000 – 2006, the rent-price ratio fell from 5 percent to 3 percent. This meant that either (a) the required return to housing fell by 40 percent ((5-3)/5), the expected real growth rate of rents tripled (from 1 percent per year to 3 percent per year), or (c) some combination of (a) and (b).
  2. In the aggregate, the share of housing attributable to the value of land was roughly stable at 33 percent from 1975 – 2000, give or take 4 percentage points. The share of land attributable to housing increased from about 33 percent in 2000 to nearly 50 percent by mid-year 2006.
  3. Over the 1975-1996 period, the real price of land in the aggregate increased by about 2 percent per year. Over the 1997-2006 period, the real price of land in the aggregate increased by 15 percent per year.
  4. Based on the land price data for the 46-MSA study, almost all cities in the analysis had an outsized run-up in the price of land. For the median metro area we study, land’s share of home value surged from 36.5% in 1995:4 to nearly 50 percent in 2006:2. In nominal terms, the price of land more than doubled in 32 of the 46 MSAs we study over the 1995:4 – 2006:2 period. In percentage terms, the biggest increases to the price of land were in Riverside (a 5.6 times increase), Los Angeles (5.1 times), Minneapolis (5.1 times), St. Louis (5.0 times), and Tampa (4.8 times).

(Of course, housing was not the only asset that experienced a surge in prices over the 2000-2006 period: Yields on Treasuries fell; Stock prices were high; Commercial real estate prices were high.)

Interestingly, my three data sets now suggest that, as of 2009:Q1, housing and land are either fairly valued or potentially undervalued. (This does not mean that house prices will not continue to fall; it just means that, according to some historical metrics, housing is priced to earn approximately its average return). Rent-price ratio: As of 2009:Q1, the rent-price ratio was at its historical norm of 5% — suggestive that housing is fairly priced at historical risk premia. Aggregate land data: Land’s share of home value in the aggregate is 25.7 percent, its lowest point in my data (the previous low was in 1982:Q3, 26.6 percent) – suggestive that land is underpriced, or that construction costs must fall more. 46 metro area land data: In 25 out of the 46 metro areas I study, land’s share of home value in 2009:1, 29.7%, is less than it was in 1995:4. Further, in 11 of the metro areas the nominal value of the land in 2009:1 is less than it was in 1995:4. For all intents and purposes, land in residential use is almost free in Atlanta, Cleveland, Detroit, and Minneapolis.

What I find most interesting about the data is the perspective: I estimate that a lot that was valued a $121 thousand in Los Angeles in 1995:4 was valued at $620 thousand as of 2006:2. My estimates suggest that the same lot is valued at $450 thousand as of 2009:1.

The Bottom Line

The bottom line is that I hope these data provide future researchers with the tools they need to warn us about the next time something this unusual is occurring! If you have comments etc., please drop me a line. My email is