Predicting Tomorrow's Land Values Today

Article Image: 

The most popular discussion today about home prices is one-dimensional and incomplete for it focuses only on demand. Looking only at whether the "median-priced household" can afford the "median-priced home" does not consider the adequacy of the housing supply. Yet prices are set by supply and demand.

To redress this shortcoming, Hearthstone, an investment firm devoted to single-family home building, and Rosen Consulting, which specializes in real estate economics, developed a simple shorthand measure that assesses the supply and demand for housing at the same time. This measure is the gross metropolitan product (GMP) per residential lot.

On the demand side, our simple proxy is GMP, a measure of a region's production. Because income equals output, GMP is a rough measure of a region's ability to pay for land. The more productive the region, the more it can afford (i.e., demand) land.

On the supply side, our proxy for the availability of residential land is the existing number of residential lots. Thus, GMP per lot is a shorthand measure for determining the balance between demand and supply.

Like any simple measure, our approach has shortcomings. For example, it does not take into account the elasticity of the land supply - meaning how difficult and expensive it is in each metro area to add to the supply. In areas such as the San Francisco Bay area and northern New Jersey, it is extremely difficult to add new residential lots to the existing supply. In areas such as Houston, it's easy.

In addition, our model is based upon data that are necessarily imprecise. The U.S. government and the private sector do not collect data on residential land values, so we had to estimate land values using other types of data (new home prices, construction costs, etc.). Imperfect input necessarily means less-than-perfect output. Our model should not be viewed as a finely tuned instrument for determining land prices with precision. Rather, it is a tool for assessing relative orders of magnitude.

Our statistical analysis shows that GMP per lot is a very good predictor of land values. The higher the GMP per lot, the higher the imputed value of land in a residential area. GMP per lot is three times or more higher in New York City and San Francisco (about $601,000 and $331,000, respectively) than in most other metro areas (e.g., only $137,000 in Atlanta). After taking the GMP per lot into account, the notoriously high residential land values in San Francisco are in line with values in other parts of the country.

In other words, home prices in the Bay area are more understandable - if, alas, not more affordable.

The Background
While much has been written about skyrocketing home prices, very little makes economic sense. Invariably the discussion focuses on housing affordability, or the lack thereof, to the "median-income household." Analysts compare house price inflation with growth of personal incomes. Home price increases should never be larger than personal income gains, they warn, or we're headed for a housing crash.

Balderdash. What the typical home buyer can pay for a house is only half the analysis. Like a one-armed paperhanger, it cannot do the job alone. What the average buyer can pay helps explain the demand side of the equation. But prices are set by supply as well as demand.

While home prices are growing faster than incomes in many markets, and households with median incomes may not be able to afford a median-priced home, supply is so constrained in many markets that there simply aren't enough homes to go around. So even if the median-income household cannot afford the median-priced home, there are enough higher-income households in the market relative to the supply of available homes to support existing price levels. Thus, simply saying that housing is not affordable for the median-income household tells us nothing about whether home prices are too high, too low or just right.

Everyone knows that house prices vary significantly across markets. A three-bedroom, two-bath home on the San Francisco peninsula can cost more than $1 million while the same house in Albuquerque, N.M., costs one-tenth that amount. Prices vary because of supply and demand.

The importance of considering supply as well as demand was amply demonstrated in a recent study in the Harvard Economic Review. In much of the country, the authors observed, housing is priced near the cost of construction, but in a number of markets, mostly large metro areas, home prices far exceed replacement cost. For example, in virtually all of the Bay area, construction costs represent less than 70% of the final cost of housing. The study presents convincing evidence that government-imposed development limitations cause this wide discrepancy in house prices.

In another article, in The Journal of Law and Economics, the authors come to a similar conclusion after assessing growth controls. Their empirical work shows that house prices are 17% to 38% higher in San Francisco-area communities that have growth controls or growth moratoria than in those that do not.

What both studies show conclusively is that limitations on housing supply (principally, in the cases studied, in the form of growth controls) inevitably raise the price of housing. But when have home prices risen too high? Neither study addresses that.

Hearthstone and Rosen Consulting collaborated to take the analysis one step further - to try to develop a simple measure of supply and demand that can help assess whether residential properties are overpriced.

Because the cost of construction is relatively stable from market to market, it is generally agreed that the main reason house prices vary is differences in land values. We therefore focused our analysis on a search for the determinants of land value.

Valuing Land
We believe that the value of urban land is based ultimately on the value of production that takes place on the land. Where production per unit of land (per square mile or per developable lot) is high, the value of the land should be high if supply is constrained. Indeed, the value will be high regardless of whether the land is used for residential or commercial development because even land devoted to residential uses must reflect the high production value in order for residential developers to bid the land away from commercial uses (assuming that, in the long term at least, land can be rezoned depending upon its highest and best use). We therefore set out to determine whether the level of production (as measured by GMP) per unit of land could explain land values.

To test our belief, we developed data on the value of residential lots. We did this by calculating the implicit value of a finished residential lot in 20 markets by taking the finished value of a new home and subtracting from it the cost of construction to arrive at a residual value for the land.

Tables showing these results can be found here. The lowest imputed value for a finished residential lot among the 20 markets is in Albuquerque ($21,521), and the highest is in San Francisco ($484,984). For almost all 20 markets, the rate of appreciation of imputed lot values was greater than the rate of appreciation of the median-priced house.

These results are consistent with our real-world understanding of residential development. When house prices are rising, lot prices rise even faster, and when house prices are falling, lot prices fall faster. This is because construction costs tend to be more stable than home prices, so a rise or fall in home prices tends to affect mostly land value.

Calculating Land
Assuming that land is an "input" in the production process, the total amount of goods and services produced in an area will be critical in determining the value of land within that area. A measure of the average productivity of land can be obtained by dividing the total production of a metro area by the number of square miles of land within the area. (While the land itself does not command the entire value of production, capital and labor are critical inputs. We assumed that other factors are constant across metro areas.) The table below contains estimates of the GMP for each of 20 metro areas for 2001, as well as GMP per square mile.

It is clear that there is a huge variation in the "productivity" of land across these 20 metro areas. In New York, the top market, each square mile generated $379.2 million of goods and services, whereas in Albuquerque each square mile in the metro area generated $3.8 million of goods and services.

Using the "productivity" data presented in the table, we conducted an analysis using land values and GMP per square mile. The results indicated a statistically significant relationship between land values and GMP per square mile.

To refine the analysis further, we used the concept of a residential lot as the unit of measurement for determining the amount of useable residential land in a metropolitan area. Using imputed lot value and GMP per lot, we obtained a dramatic improvement in the fit compared with using GMP per square mile.

To bid away a unit of land for residential use, residential developers must compete against commercial users who can lease or sell their land to users who generate a very high GMP per lot. Our analysis shows that this high productivity of useable land in San Francisco, for example, goes a long way in explaining the area's high housing costs.