In collaboration with Jack Hoy and Robert Perina
According to a report by Reuters, housing starts in the United States have soared as of July 2015; new residential construction is at one of its highest levels in the last 8 years. While this is could be a good sign for the economy, many may wonder why this is happening. Clearly, there are a variety of factors that go into an increase in housing starts, but what are the most important ones? Christopher Mayer and Tsuriel Somerville present and analyze a model that helps answer this question, which leads to valuable conclusions about the implications of this economic signal.
Mayer and Somerville (2000) explain new housing construction as a function of changes in house prices and costs. To develop their model, they review previous empirical studies of housing supply, which consider residential development like other types of investments. In this paper, however, they try to present a model that incorporates how land differs from typical investments because supply of land is inelastic.
In order to develop their model, Mayer and Somerville used several sources of data to come to their conclusions. They first looked at figures showing the differences between housing starts and price levels in the United States. The paper, however, suggests that using this data limits explaining house starts and price levels because of its inconsistency. Instead, Mayer and Somerville use data that shows the relationship between housing starts and the changes in housing prices.
Mayer and Somerville (2000) rely on the previous findings regarding the relationship between starts and house prices to find how they can create a model that most accurately represents new housing construction. Specifically, they look at the two approaches used by existing empirical studies. The first approach concludes that the supply curve for the housing starts is perfectly elastic, while the second approach estimates housing starts as a function of the level of house prices and cost shifters. Mayer and Somerville specifically describe the models created by Topel and Rosen (1988) and DiPasquale and Wheaton (1994), but they conclude that these models do not capture the role land plays in a model for housing starts.
After developing a model that depicts housing supply with consideration to the land development process in a single city, Mayer and Somerville use data to apply their model to the nation. Specifically, they use the Freddie Mac repeat sales index to measure house price movements to allow their supply model to be applicable to the United States. The authors do note that there can be problems with using sales indexes; however, the Freddie Mac repeat sales index matches up well with the model.
The paper argues that as a city’s population increases, so does the amount of land it occupies, which increases housing prices and stock. The city then needs a greater number of housing starts to fill its existing stock. Therefore, housing starts and prices will be higher after an increase in population.
Mayer and Somerville use the data to develop a model for supply of new housing that is consistent with the land development process, which previous empirical models failed to accomplish. A main point in its model is that Mayer and Somerville realize that there are delays in developing land from nonurban use to residential use, and these delays must be accounted for in predicting demand.
Their model for housing starts accounts for the lag in price and cost changes, which depends on “the length of time required to obtain developed land, acquire housing permits, and builders’ expectations about changes in future house prices.”
Mayer and Somerville (2000) also present a model that shows that housing starts are a function of changes in construction costs and prices. They argue that the size of a city is directly correlated to the price and cost of housing. In their equations, new housing construction can be explained and estimated by changes in house prices and construction costs. The model, however, is best used for a single city.
Mayer and Somerville used data from the Freddie Mac repeat sales index, which shows quarterly changes in house prices from 1975 to 1974. In order to use this data effectively and compare it to existing models, they converted the Freddie Mac index to house price levels in 1991. Based on that data, Mayer and Somerville concluded that prices do not fully adjust to market equilibrium in the short run.
After running the regression, Mayer and Somerville found that the most significant variable was changes in house prices; a $943 increase led to a quarterly increase in housing starts of 18,300 units. They also found that changes in interest rates were statistically significant. However, these interest rate changes were not quite as impactful as the housing price changes. For instance, in the regression they ran, a 1.3-percent increase in interest rates decreased new housing starts by about 12,000 units. Furthermore, the authors found that one variable, material prices, did not end up being significant. Finally, Mayer and Somerville note that the variable of median time-to-sale, which takes into account market conditions, was significant and does impact new housing starts. More specifically, it is negatively correlated with construction. For instance, if one increases the median time-to-sale variable by one standard deviation, housing starts decrease by 16,300 units. Based on its significance, the authors determine that nonprice variables can also impact housing starts.
Mayer and Somerville conclude that changes in house prices and construction costs (lagged and current) are significant factors in determining new housing starts. They use their model to estimate that when house prices increase by 10 percent, housing stock increases by 0.8 percent. It is important to note that the variables used are changes in the level of prices and costs, not simply the levels of them. This is one of the reasons their model does well when compared to other notable models.
Mayer and Somerville also conclude that it only takes about 1 year for stock to adjust to a shock in demand, which is completely different from DiPasquale and Wheaton’s (1994) estimate of 35 years. Finally, they determined that although an increase in house prices leads to more new construction, this increase is only temporary, not permanent. Overall, Mayer and Somerville create a model to show what factors impact new construction of houses and analyze the impact of their findings.
This article can be easily related to a variety of other real estate literature; in particular, the discussion of development in the book “Real Estate Principles: A Value Approach” seemed to be relevant to the article. In this article, the authors, David Ling and Wayne Archer, discuss what it is like to be a developer, noting that the process can be very unpredictable and complicated. This reflects what the article implies: there are numerous factors that go into construction of new housing. The authors of the article did an excellent job of creating a model to help predict housing starts, but it is important to note that the unpredictability of development in general could make the actual number of housing starts significantly different from what the model predicts.
Furthermore, Ling and Archer’s book also discusses various types of construction costs. This is relatable to the article, as changes in construction costs were found to be a significant factor in housing starts. Some of the costs listed in the text include soft costs such as inspections and insurance and hard costs such as permits/fees. It is important to know all of the costs that go into the one construction cost variable.
Although the paper presents a very sound model to predict the construction of new houses, there is some potential for improvement in future research. First of all, much of the empirical analysis that Mayer and Somerville do is based on national data. However, their model is for an individual city. This mismatch forces them to rely on two assumptions: “First, that an urban form framework is applicable to national data; second, that there is a single national housing market.” By having to rely on these assumptions, it increases the likelihood that this model may fail when actually used in practice. In addition, individual cities are very different from one another, which means that the model may be ineffective in one city but effective in another. Finally, when testing their model against other notable models, they use a small sample size that could create skewed results.
In the future, there are also other multiple concepts that the authors could continue to research. For instance, their article deals with the housing supply in one market. As the authors note, they would like to apply the model to different markets. By studying the supply across markets, they could introduce new factors that would affect new construction, such as governmental regulations.
Another idea would be to study new construction for nonresidential buildings. This would allow the researchers to potentially find any similarities between the variables that impact the supply of commercial and residential buildings, which could provide valuable information for the real estate industry.
Multiple Choice Questions
-
1. Which of the following is not a reason presented by Mayer and Somerville (2000) for a lag in price and cost changes?
a. Length of time required to obtain developed land
b. Acquiring housing permits
c. Builders’ expectations about changes in future house prices
d. Zoning changes required for urban development
Explanation: Mayer and Somerville (2000) incorporate a lag in price and cost changes due to the length of time required to obtain developed land, acquiring housing permits, and builders’ expectations about changes in future prices. Mayer and Somerville do not mention zoning changes as a reason for a lag in price and cost changes. They do, however, incorporate a lag for the other reasons because these delays often occur in developing land from nonurban uses to residential units. If the delay was not included in the model, the timing of housing starts would inaccurately be represented.
-
2. According to Mayer and Somerville (2000), after an increase in population, housing starts and prices will _______?
a. Both increase
b. Housing starts will decrease but prices will decrease
c. Housing starts will increase but prices will increase
d. Both decrease
Explanation: Housing starts and prices will both increase with an increase in population. This holds true because as a city’s population increase so does its amount of land it occupies, which increases housing prices and stock. The city then needs a greater number of housing starts to fill its existing stock. A city with a higher population will then have a higher construction costs and housing starts. Housing starts, however, will only increase as needed to adjust to the increase in population.
-
3. According to Mayer and Somerville (2000), changes in _______ have the strongest effect on housing starts?
a. Construction costs
b. Housing prices
c. Interest rates
d. High unemployment rates
Explanation: In the regression in the paper, Mayer and Somerville show that housing prices have the strongest effect on housing starts. A $943 increase (which equates to one standard deviation) in housing prices, or one standard deviation, increases starts by 53,800 units in a year. Although interest rates have an effect on housing starts, but it is not nearly as impactful as an increase in housing prices. For comparison, a 1.3-percent increase (one standard deviation) in interest rates leads to 12,000 unit decrease in housing starts. Clearly, it is not as impactful as changes in housing prices. Interest rates affect the housing market through the demand instead of the supply. Higher interest rates and housing costs would cause a drop-off in housing start because they are negatively correlated.
References
DiPasquale, D. and W. C. Wheaton (1994), “Housing Market Dynamics and the Future of Housing Prices,” Journal of Urban Economics 35, 1–28.
Ling, D. C. and W. R. Archer (2013), “Real Estate Principles: A Value Approach,” 4th ed. US: McGraw-Hill Education, 625–628.
Mayer, C. J. and C. T. Somerville (2000), “Residential Construction: Using the Urban Growth Model to Estimate Housing Supply,” Journal of Urban Economics 48, 85–109.
Mutikani, L. (2015), “U.S. Housing Starts Approach Eight-Year High in July,” US: Thomson Reuters.
Topel, R. and S. Rosen (1988), “Housing Investment in the United States,” Journal of Political Economy 96, 718–740.