United States: Selected Issues

International Monetary Fund

doi:10.5089/9781475504910.002.A002

Jump to Content Jump to Main Navigation

United States: Selected Issues

Author: International Monetary Fund

Publication Date:: 02 Aug 2012

eISBN:: 9781475504910

Language:: English

Keywords:: ISCR; CR; recession; mortgage; housing stock; household formation; ownership market; construction activity; housing cycle; Housing; Housing prices; Stocks; Employment

Download PDF (1 MB)

This Selected Issues paper for the United States discusses the microeconomics of the country—household wealth and savings. Households’ consumption-saving decisions have an important bearing on the U.S. economic outlook. This paper demonstrates how households with consistently lower income, which have shown growth in the years prior to the crisis, experienced larger declines in their saving rates and a larger rise in their indebtedness before the crisis, contributing significantly to the dynamics of the mean saving rate.

Abstract

II. The Residential Construction Sector: When Will it Emerge From its Rubble?¹

Despite some recent improvement, activity in the U.S. construction sector remains depressed four years after the start of the Great Recession, and the sector’s share in GDP is at an all-time low. This paper shows that while the number of vacant homes in 2011 is well above its equilibrium level, the implied excess stock of houses is smaller because of the subdued formation of households since 2007. Using an empirical model and our forecasts on employment and financial conditions, we project household formation to steadily increase over the next few years. As a result, housing starts would increase gradually and return to their 1990s’ average by 2016, with risks being tilted to the upside.

A. Introduction

1. Instead of powering the economy as it has done after past recessions, the U.S. housing market has remained depressed since the Great Recession. The ratio of construction to GDP, which reached 6.3 percent in Q4 2005, has continuously declined to reach 2.2 percent of GDP in Q1 2011. The anemic level of construction activity has also contributed to the current high unemployment rate, with implications for both consumption and real activity. The residential sector has seen some steady improvement since the second half of 2011, supported by an increase in multifamily housing starts. Nevertheless, the housing market remains very weak relative to historical levels.

2. While excess construction during the pre-crisis boom year meant that an adjustment period was unavoidable, depressed household formation has also contributed to the weakness of construction activity. Since 2007 household formation has been nearly at half of its average during the 1990s, reflecting high unemployment rates, a tightening of lending standards, and house price uncertainty.

3. The outlook for construction activity depends on both the current number of excess vacant units and the strength of household formation going forward. The higher the stock of excess vacant units, the longer it would take for construction to rebound, as the demand for housing will be largely met with the existing vacant stock. The speed of recovery in construction will also depend crucially on the future pace of demand, which in turn depends on the extent to which the number of households is currently below equilibrium.

4. This paper uses a stock-flow model approach to shed light on the outlook for residential investment. The paper proceeds in three steps. First, the paper assesses the extent to which the two components of the housing stock (occupied and vacant units) are above “equilibrium” levels. Second, the paper estimates an error-correction model (ECM) to forecast a path for household formation. Finally, these elements are put together to project a path for housing starts. The paper also discusses other factors that could affect the recovery of construction activity in the United States, such as the decline in homeownership.

5. The paper suggests that construction activity is likely to recover slowly in the coming years, with housing starts returning to their 1990s levels only by 2016. The stock of excess vacant units has declined since 2009 but remains elevated at around 3.7 million units. In contrast, the number of households is estimated to be currently below equilibrium. The ECM suggests that household formation could average slightly over 1.1 million over the next five years. Based on this prediction, housing starts will likely reach 1.35 million by 2016.

B. A Stock-Flow Model and Estimates of Excess in Housing

6. The dynamics of the housing stock could be described using a stylized stock-flow model. For simplicity, this first section will assume housing units to be homogenous, and therefore will not differentiate between the rental and for-sale markets (see Section E). The stock of existing housing (H_t) is the sum of occupied units (O_t), or equivalently the number of households, and vacant units (V_t):²

\begin{matrix} H_{t} = O_{t} + V_{t} & (1) \end{matrix}

The evolution of the stock of occupied units can be written as follows:

\begin{matrix} O_{t} = O_{t - 1} + {lN}_{t} - {OUT}_{t} & (2) \end{matrix}

where IN_t is the number of newly occupied units, or gross household formation, and 0UT_t the newly vacated units. The term IN_t—OUT_t captures therefore the net increases in occupied units, or equivalently, net household formation, at time t. Similarly, we could describe the evolution of the stock of vacant units as follows:

\begin{matrix} V_{t} = V_{t - 1} - ({lN}_{t} - {OUT}_{t}) + {NEW}_{t} - {DEM}_{t} & (3) \end{matrix}

where NEW_t is the number of newly built units, and DEM_t is the number of demolished units. These simple equations describe the evolution of the housing stock, and will be helpful in analyzing equilibrium levels and in making forecasts of future housing flows and stocks.

7. This section estimates equilibrium levels for the stocks of occupied and vacant units, in order to provide estimates of the overall excess in the stock of housing. The excess vacancy $\begin{matrix} (V_{t} - V_{t}^{*}) \end{matrix}$ is a measure commonly used to gauge the potential for construction activity. However, this measure alone does not take into account the potential demand for housing. In other words, excess vacancy could be due to depressed levels of occupancy (low household formation). Therefore, it is important to examine also the excess occupancy $\begin{matrix} (O_{t} - O_{t}^{*}) \end{matrix}$ . For a given excess vacancy $\begin{matrix} (V_{t} - V_{t}^{*}) \end{matrix}$ , one would expect a lower excess occupancy to be associated with a faster recovery in construction (more on this in section E). Hence, $\begin{matrix} (V_{t} - V_{t}^{*}) \end{matrix}$ could be usefully supplemented with a measure of the overall excess in the housing stock:

\begin{matrix} H_{t} - H_{t}^{*} = (O_{t} - O_{t}^{*}) + (V_{t} - V_{t}^{*}) & (4) . \end{matrix}

Equilibrium Occupancy Ratio

8. The number of occupied units has grown rapidly during the boom years, but is currently estimated to be below equilibrium. The demand for housing boomed throughout the years 2003–06, when the number of households increased by around 4.5 million. In comparison, household formation averaged slightly more than 1 million a year during the 1990s. While some of this increase was due to demographic changes, the increase in the occupancy ratio (the ratio of the number of households to overall population) also reflects economic factors. There are two common methods to estimate the equilibrium occupancy ratio. The first method is based on its historical average, while the second computes the equilibrium occupancy by age group (commonly called “headship rate”). Thus the equilibrium based on headship rates takes into account changes in the composition of the population. Figure 1 shows that, according to both methods, the stock of occupied units was above its equilibrium level during the 2004–07 period, but it has fallen below equilibrium with the crisis and remains “excessively” low at end-2011.

Equilibrium Vacancy Ratio

9. The increase in the stock of vacant units during the boom was even more dramatic than the increase in occupancy, and has left a legacy of a substantial excess. Fueled by buoyant construction activity, the stock of vacant units grew strongly between 2001 and 2008, by around 650 thousand units a year, three times higher than the yearly average in the 1990s. Since 2010, the stock of vacant units has been on the decline, particularly the stock of for-sale/for-rent units and seasonal units. Units held-off of the market, on the other hand, have continued to increase between 2007 and 2010, likely owing to the decline in house prices (which could have led some sellers to wait for the recovery).

Currently, the stock of vacant homes stands at around 18.7 million.³ The “vacancy ratio” (the ratio of the number of vacant units to the overall population) has increased rapidly between 1960s and the 1980s, likely due to structural demographic and economic changes. In the 1990s, however, it has hovered at around 5 percent before it started increasing rapidly as of 2001. Therefore, its value in 2000 (which is roughly equal to its average in the 1990s) is used as an estimate of the equilibrium level. Based on this estimate, the number of excess vacant units has peaked at around 4.1 million in 2009 but has since been declining, and it is estimated at around 3.7 million as of end-2011 (Figure 2).⁴

The Overall Excess Housing Stock

10. Adding the excess stocks in occupied and vacant homes yields an overall excess housing stock of around 2.2 million as of end 2011. As discussed earlier, excess vacant housing might be large due to depressed household formation. The excess in the overall housing stock also takes into account shortages or excesses in the stock of occupied units, and hence provide a better indication of the potential speed of recovery for construction activity.⁵ Based on the estimated excess vacancy and excess occupancy, the total excess housing stock is estimated at around 2.2 million units as of end-2011 (Figure 3), down significantly from its peak in 2007.⁶

Factoring in the Shadow Vacancy

11. The ongoing foreclosure crisis is expected to slow down the absorption of the excess stock of vacant houses. While foreclosure rates have declined since their 2009 peak, they remain elevated and the stock of foreclosure inventory is near a record high. This implies that a share of currently occupied homes will continue to progress toward foreclosure and vacancy. While most households who experienced a foreclosure are likely to move to the rental market, some will instead relocate with their relatives. Therefore, even in a scenario where gross household formation recovers to levels in line with underlying demographic factors, net household formation is likely to be weakened by foreclosures. The shadow vacancy is currently estimated at around 4½ million.⁷ The paper will henceforth assume that this shadow vacancy will subtract an annual ¼ million from household formation over the next two years.⁸

C. Recovery in Household Formation

12. This section estimates a model of household formation, both at the national and at state level. A projection for household formation is a crucial input for any projection of housing starts. For that purpose, an empirical model is estimated to examine the determinants of household formation. A natural formulation is an error-correction model (ECM) in which the change in household formation is determined by its long term equilibrium (based on population) while short-term variations are explained by changes in a series of variables, such as employment growth. The benchmark specification is of the following form:

\begin{matrix} Δln (\frac{O_{t}}{P_{t}}) = α + β {\ln (\frac{O_{t - 1}}{P_{t - 1}}) - K} + γΔln (\frac{O_{t - 1}}{P_{t - 1}}) + {θZ}_{t - 1} + ɛ_{t} \end{matrix}

where ln $(\frac{O_{t}}{P_{t}})$ is the logarithm of the occupancy ratio (the change in which approximates the difference between the growth rate of occupancy, O_t, and population, P_t), and K is a constant capturing the long term equilibrium of $(\frac{O_{t - 1}}{P_{t - 1}})$ .⁹ Z_t-1 are factors capturing other short-term determinants of $\begin{matrix} Δln (\frac{O_{t}}{P_{t}}) \end{matrix}$ , specifically, employment growth, the mortgage rate, and the Senior Loan Officer Opinion Survey (SLOOS) index of banks’ mortgage lending standards. The equation is estimated using three different specifications: Model A is estimated at the national level without Z_t-1 Model B is estimated with the full specification at the national level; and Model C is estimated with the full specification at the state-level.¹⁰ The results are shown in Table 1. The error-correction term has a negative and significant coefficient in all models. The lagged dependent variable is only significant in Model A and is thus not included in the other models.

Table 1:

Determinants of Household Formation

Table 1:

Determinants of Household Formation

Dependent variable: Δln(O_t/P_t)
	Model A	Model B	Model C
Lag Δ ln(O_t/P_t)	0.6921
Lag Δ ln(O_t/P_t)	[10.02]
Lag ln(O_t/P_t)	-0.0130	-0.3080	-0.6381
Lag ln(O_t/P_t)	[2.08]	[2.22]	[9.79]
Lag Δ ln(employment)		0.0759	0.0706
Lag Δ ln(employment)		[2.64]	[2.59]
FRB Sr Loan Survey		-0.0007	-0.0003
FRB Sr Loan Survey		[2.02]	[0.42]
Real mortgage rate		-0.2387	-0.4991
Real mortgage rate		[3.08]	[5.87]
Constant	-0.0056	-0.1253	-0.2588
Constant	[1.93]	[2.12]	[9.00]
Model type	Long term equilibrium	Full model	Full model
Level	National	National	State
N	54	34	1038
R-squared	0.5916	0.5662	0.1687
The state-level model is estimated with state fixed effects. When Lag Δ ln(O_t/P_t) is not included, it is not significant.

Table 1:

Determinants of Household Formation

Dependent variable: Δln(O_t/P_t)
	Model A	Model B	Model C
Lag Δ ln(O_t/P_t)	0.6921
Lag Δ ln(O_t/P_t)	[10.02]
Lag ln(O_t/P_t)	-0.0130	-0.3080	-0.6381
Lag ln(O_t/P_t)	[2.08]	[2.22]	[9.79]
Lag Δ ln(employment)		0.0759	0.0706
Lag Δ ln(employment)		[2.64]	[2.59]
FRB Sr Loan Survey		-0.0007	-0.0003
FRB Sr Loan Survey		[2.02]	[0.42]
Real mortgage rate		-0.2387	-0.4991
Real mortgage rate		[3.08]	[5.87]
Constant	-0.0056	-0.1253	-0.2588
Constant	[1.93]	[2.12]	[9.00]
Model type	Long term equilibrium	Full model	Full model
Level	National	National	State
N	54	34	1038
R-squared	0.5916	0.5662	0.1687
The state-level model is estimated with state fixed effects. When Lag Δ ln(O_t/P_t) is not included, it is not significant.

13. Based on the estimation results, the growth rate of household formation is projected to pick up in the near term. Projections for employment growth assume a continued recovery in employment growth (averaging 1.5 percent between 2012 and 2017), consistent with the U. S. projections in the IMF April 2012 WEO. We also assume a continued easing of lending standards (which declines by nearly 40 percent in 2017 from its peak in 2009), and a path for mortgage rates consistent with the consensus forecast.¹¹ Figure 4 shows the projection based on Model A for national data and Model C with state-level data. Model A, which explains household formation with its lag and the error-correction term only, predicts a smooth rebound over the two years before converging to the equilibrium level. The stronger rebound predicted by the state-level model (Model C) is due to the projected improvement in employment and easing of financial conditions, in addition to the pull-back effect from the error-correction term. Projected household formation averages around 1.2 million a year, between 2012 and 2017 in both models. These results do not take into account the shadow vacancy, which will be incorporated in the forecast in Section D.

14. The projection of household formation is consistent with the evidence from previous housing boom-bust episodes at the state level. While housing cycles have been modest at a national-level, 23 states have experienced significant real boom-bust episodes over the period 1977–2001.¹² Figure 5 shows that, on average across these episodes, household formation has rebounded strongly following years of depressed levels during the bust period. The rebound of household formation has lagged the recovery in employment, and has coincided with the trough in house prices. Our projections for household formation are qualitatively similar to these patterns, as they predict a rebound by 2013, a year in which house prices are expected to reach their trough and two years after the start of the recovery in employment.

D. Putting the Pieces Together

15. This section projects housing starts over the next five years, using household formation from the ECM model as an input, and assuming that the excess vacancy will be eliminated in five years. The forecast for household formation is taken from Model A, and does not take into account the impact of the units in the shadow vacancy stock. These units are therefore subtracted from the projection, leading to a lower level in household formation in 2012 and 2013. We then assume that the stock of vacancies will gradually return to their equilibrium level in 2017. Based on equation (3), having projected a path for household formation and assumed a path for the stock of vacancy units we can now project housing starts. In particular, we obtain that they will return to their 1990s average as of 2016, at a pace of around 1.35 million units a year (Figure 6). If housing starts were to continue increasing at their current depressed pace of around 0.7 million a year, equation (3) implies that there will be a shortage of vacant units of around 1.2 million units in 2017.¹³

E. Other Considerations

Geographical Heterogeneity

16. The excess vacancy rate varies significantly across states. Figure 7 shows the distribution of excess vacancy at the state level (excess vacancy is computed as the difference between the vacancy rate as of 2010 and the average vacancy rate in the years 1990 and 2000).¹⁴ The figure also shows the share of each state housing stock in the national stock, as well as the share of its excess vacancy in the total excess vacancy. A comparison of these two measures indicates whether the state contributes to the overall excess stock above what is implied by the size of its housing stock. For example, the contribution of Texas to the national excess supply is less than half its share in the national house stock, while the contributions of Florida, Ohio and Michigan exceed their share in the national stock.

17. The heterogeneity in excess vacancy suggests that the recovery in construction activity could also vary widely across states. If employment were to rebound in a relatively homogeneous way across states, the hardest-hit states would take much longer to clear their excess housing inventory, and construction would be driven mainly by the less affected states. Figure 9 shows a scatter of the excess vacancy in 2010 relative to the level in 1990, and the change in permits in 2011 as a percentage increase from their average 1990s level. There is a strong negative relation between these two variables, supporting the view that a high excess inventory is associated with lower construction levels.¹⁵ Whether the heterogeneity in vacancy across states is a positive for aggregate construction (in comparison to what is implied by the benchmark national-level model) is uncertain. Conceptually, if the decreasing relation between excess vacancy and construction is convex (as anecdotally suggested by Figure 8) then this heterogenity would lead to a higher aggregate construction level than what is implied by the national average. ¹⁶

Differentiating Between the Ownership and Rental Markets

18. The boom in homeownership that took place during the bubble period is currently unwinding, and this process is likely to continue for some time due to the large shadow inventory. The homeownership rate has risen dramatically from the mid-90s to the mid-2000s, from 64 to 69 percent. Since 2005, homeownership has been on the decline, and is currently (as of Q1 2012) at 65.4 percent. This decline is the result of an adjustment to excess household formation in the ownership market. The transition of a large number of households from rental to ownership during the boom was fueled by excessive loosening of lending standards.¹⁷ Many of these same households were later unable to make their mortgage payments and went back to a more affordable rental arrangement. The large shadow inventory indicates that the process is far from over, and is likely to continue for some time. Figure 9 shows a projection in which the homeownership rate settles to slightly above its 1990s average by 2017.

19. The increased demand for rental units is expected to be positive news for the construction sector given the segmentation in the real estate market. Only a share of the units in the vacancy (and shadow vacancy) stock can be converted for the purpose of rental. Meanwhile, as more households are choosing rental over ownership, the vacancy rate of rental units is declining and rental prices are increasing. This has already led to an increase in multi-family housing starts and a continuation of this trend, together with a potential recovery in the labor market, is likely to further support construction in this sector. The model in section B is used to assess both the rental and the ownership markets, taking into account movements between the two. Figure 10 compares the implied path for housing starts under the model that takes into account the segmentation of the market with that implied by the benchmark model. Differentiating between the two markets leads to a higher projection of housing starts (by a cumulative 0.35 million between 2013 and 2015) than under the benchmark model.

F. Conclusion

20. This paper estimated the excess stock of housing in the United States and assessed the determinants of household formation in order to shed light on the outlook for residential investment. The main conclusion is that, conditional on a continued improvement in overall employment and lending conditions, household formation is likely to rebound over the next few years and absorb the remaining part of the already shrinking excess housing supply. A recovery path for household formation based on a state-level empirical model suggests steady, but moderate, gains in housing starts, which are likely to return to their mid-1990s level of around 1.35 million units by 2016. The increased demand for rental units could lead to higher levels of construction than what is predicted under the benchmark scenario. A better than expected improvement in employment is an upside risks, while setbacks to house prices and to the economic environment remain significant downside risks.

Prepared by Jihad Dagher and Julien Reynaud with research assistance from David Reichsfeld.

The stock of vacant housing is defined as the sum of units that are for rent or sale, those that are held off the market for occasional use or other reasons, and units that are used on a seasonal basis.

Foreclosed homes that are vacant could enter the vacant stock in the various subcategories.

A regression analysis in which the vacancy ratio is regressed on various demographic and economic factors yields similar results.

As can be seen from the above equations, H_t = H_t-1 + NEW_t - DEM_t.

Henceforth, the measure of excess occupancy is based on the approach that takes into account compositional changes in the population (see paragraph 8).

The stock of non-listed homes at serious risk of foreclosure is often referred to as the shadow inventory. What is relevant for the above calculations (and for construction prospects in general) is a variation on the shadow inventory measure that takes into account listed but non-vacated homes (“shadow vacancy”). This stock includes a share of (i) currently delinquent loans (at least 60 days past due), (ii) re-performing loans, (iii) underwater mortgages, and (iv) non-vacated homes that are in the process of foreclosure.

In other words, of the 4½ million households that are likely to vacate their owner-occupied units, this paper assumes that ½ million will return to live with parents or relatives while the rest will move into rental units, thus not affecting the net change in vacant homes.

This model assumes, for simplicity, that the occupancy ratio is constant at the equilibrium and thus assumes that the equilibrium level of O_t to be fully determined by the number of population. The constant K will be absorbed by the constant in the regression.

State-level fixed effects and cluster are used for the estimation of Model C.

Controlling for lagged population growth (including migration) does not change the main results.

The annual FHFA house price index is used to identify the real housing cycle at the state level over the period 1977–2011. 23 episodes of boom-bust are identified, defined as peak-to-trough changes in real prices of at least 25 percent.

A shortage in vacancies means that vacancies are below their equilibrium level.

Given data limitation at the state-level, the most reliable sources of information are the decennial CENSUS.

A simple linear regression suggests that a reduction of one standard deviation in excess vacancy is associated with an increase by 15 percent in construction activity.

This is based on a highly stylized model. Assuming that construction c is a function of excess vacancy: c = F(v) then if $\begin{matrix} v_{A} \neq v_{B}, \frac{F (v_{A}) + F (v_{B})}{2} > F (\frac{v_{A} + v_{B}}{2}) \end{matrix}$ if F(.) is strictly convex.

This is consistent with the findings of the CBO, Background paper “The Outlook for Housing Starts, 2009 to 2012”, November 2008.

Other Publishers

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://www.elibrary.imf.org/view/journals/002/2012/214/article-A002-en.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully

United States: Selected Issues

Author: International Monetary Fund

Volume/Issue:: Volume 2012: Issue 214

Publisher:: International Monetary Fund

ISBN:: 9781475504910

ISSN:: 1934-7685

Pages:: 119

DOI:: https://doi.org/10.5089/9781475504910.002

Keywords
II. The Residential Construction Sector: When Will it Emerge From its Rubble?1

View raw image

Actual and Equilibrium Number of Occupied Units

View raw image

Actual and Equilibrium Number of Vacant Units

View raw image

Overall Excess Inventory

View raw image

Scenario for Household Formation

View raw image

Household Formation and Employment During Past Housing Cycles

View raw image

Housing Starts Projections

View raw image

The Heterogeneity of Excess Vacancy

View raw image

Excess Vacancy and Construction

View raw image

Homeownership Rate

View raw image

Projections for Housing Starts

Table 1:

Determinants of Household Formation

Dependent variable: Δln(O_t/P_t)
	Model A	Model B	Model C
Lag Δ ln(O_t/P_t)	0.6921
Lag Δ ln(O_t/P_t)	[10.02]
Lag ln(O_t/P_t)	-0.0130	-0.3080	-0.6381
Lag ln(O_t/P_t)	[2.08]	[2.22]	[9.79]
Lag Δ ln(employment)		0.0759	0.0706
Lag Δ ln(employment)		[2.64]	[2.59]
FRB Sr Loan Survey		-0.0007	-0.0003
FRB Sr Loan Survey		[2.02]	[0.42]
Real mortgage rate		-0.2387	-0.4991
Real mortgage rate		[3.08]	[5.87]
Constant	-0.0056	-0.1253	-0.2588
Constant	[1.93]	[2.12]	[9.00]
Model type	Long term equilibrium	Full model	Full model
Level	National	National	State
N	54	34	1038
R-squared	0.5916	0.5662	0.1687
The state-level model is estimated with state fixed effects. When Lag Δ ln(O_t/P_t) is not included, it is not significant.

Topics

Countries

Series

About

Resources

Abstract

II. The Residential Construction Sector: When Will it Emerge From its Rubble?1

A. Introduction

B. A Stock-Flow Model and Estimates of Excess in Housing

Equilibrium Occupancy Ratio

Equilibrium Vacancy Ratio

The Overall Excess Housing Stock

Factoring in the Shadow Vacancy

C. Recovery in Household Formation

D. Putting the Pieces Together

E. Other Considerations

Geographical Heterogeneity

Differentiating Between the Ownership and Rental Markets

F. Conclusion

Same Series

Other IMF Content

Other Publishers

Asian Development Bank

Inter-American Development Bank

Nordic Council of Ministers

The World Bank

Share Link

Table of Contents

Headings

Figures

Metrics

II. The Residential Construction Sector: When Will it Emerge From its Rubble?¹