The Stock of Private Real Estate Capital in U.S. Metropolitan Areas

Article excerpt

Abstract

The stock of private real estate capital is estimated for each of 242 MSAs, annually, for 1982 through 1994. Three series are computed: (1) total private real estate capital (residential and nonresidential); (2) private single-family residential capital; and (3) private income property capital (multifamily housing plus nonresidential real estate, or (1) less (2)). The determinants of each series are modeled, and the results are used to predict the value of the capital stock for a larger set of 295 MSAs.

Introduction

The bulk of the capital stock in the United States, indeed the world, is real estate (see Exhibit 1; and Ibbotson, Siegel and Love, 1985). As a general proposition, the need for careful measurement of such fundamental data needs no elaboration. Many recent studies of real estate markets adopt the metropolitan area as the unit of observation! Conversations with developers, investors and others confirm that market participants often use the metropolitan area as a decision unit. Commercial data providers such as Torto-Wheaton also focus on the metropolitan area as the unit of observation. This is hardly surprising, given that real estate's locational fixity is in fact its defining characteristic. Despite this, hardly any data exists on the stock of real estate capital by metropolitan area.

Many specific examples can be given of potential uses for such data. For example, studies of portfolio allocation can be much improved with such data. In fact, this study was undertaken in order to construct the data needed for Malpezzi and Shilling (2000), which required estimates of real estate capital stock by metropolitan area (MSA). While flow data on real estate investment are readily available from building permits, stock estimates are not, in general, widely available.

This article describes the construction of estimates of private real estate capital for each of 242 MSAs, annually, for 1982 through 1994. Three series are computed: (1) total private real estate capital (residential and nonresidential); (2) private single-family residential capital; and (3) private income property capital (multifamily housing plus nonresidential real estate, or (1) less (2). The determinants of each series are then modeled, and the results used to predict the value of the capital stock for a larger set of 295 MSAs.

Previous Literature

Only two previous studies have attempted to measure the real estate capital stock in such a disaggregated fashion. Miles, Pittman, Hoesli, Bhatnager and Guilkey (1991) begin with county level tax assessments, and regress them against a set of demographic variables to construct an instrument for the value of real estate in thirty-six metropolitan areas. The estimates are broken out by retail, office and industrial property types. They do not include multifamily investment property, and are limited to a small subset of metropolitan areas.

Hartzell, Pittman and Downs (1994) apply a similar method to county level data. Sixty-seven counties in thirty-two metropolitan areas are used. The baseline data are also property tax assessments. They then collected a set of exogenous variables related to population, structure of employment, income and employment. They use these to construct instrumental variables for each property type. Their regression models are then used to estimate the value of real estate by property type for each U.S. county.

Methodology

The inflation rate is from the NIA GNP price deflators, taken from Economic Report of the President. Separate inflation rates are used for residential and nonresidential real estate.

Two separate depreciation rates are used-one for single family residential and one for income property. Hulten and Wykoff (1980) estimate depreciation rates for sixteen categories of income property (nonresidential real estate and apartments). These rates range from 2. 1 % for shopping centers to 4. …