The Journal of Real Estate Finance and Economics

, Volume 36, Issue 4, pp 367–404

Avoiding Taxes at Any Cost: The Economics of Tax-Deferred Real Estate Exchanges

Authors

    • Department of Finance, Insurance and Real Estate, Warrington College of BusinessUniversity of Florida
  • Milena Petrova
    • Department of Finance, Whitman School of ManagementSyracuse University
Article

DOI: 10.1007/s11146-007-9099-6

Cite this article as:
Ling, D.C. & Petrova, M. J Real Estate Finan Econ (2008) 36: 367. doi:10.1007/s11146-007-9099-6

Abstract

This study examines the role tax-deferred exchanges play in the determination of reservation and transaction prices in U.S. commercial real estate markets. Taxpayers face significant time constraints when seeking to complete a delayed tax-deferred exchange. In a perfectly competitive market, a weakened bargaining position would not affect the transaction price. However, in illiquid, highly segmented commercial real estate markets, the exchanger may be required to pay a premium for the acquired property relative to its fair market value. Using a unique and rich dataset of commercial property transactions, we find that tax-motivated exchange buyers pay significantly more, on average, than non-exchange investors for their apartment and office properties, all else equal. Moreover, these average price premiums generally exceed the tax deferral benefits investors obtain by the use of a tax-deferred exchange. This result is robust to a number of alternative specifications. Thus, for many investors the pursuit of tax avoidance comes at a steep price.

Keywords

Commercial real estateTax-deferred exchangesTransaction price

Introduction

In asset markets characterized by perfect competition and investment value revelation, all transactions take place at the true market value of the asset, as determined by the marginal buyer and seller. In such markets, there is no need for buyers and sellers to search for the “true” market value of an asset; it is continuously revealed by the transaction prices of perfect substitutes. Moreover, the heterogeneous investment motivations of potential buyers and sellers and their relative negotiation abilities have no role in the price formation process.

Real property markets, however, are far from perfect and differences across buyers and sellers in information availability, search costs, motivations, and bargaining power are thought to be pervasive. A growing literature examines the extent to which these differences are priced in the market for owner-occupied homes.1 However, the pricing of these imperfections in commercial real estate markets has received relatively little attention in the literature.2

This study examines the extent to which income tax deferral benefits available to some market participants affect transaction prices. More specifically, we analyze both conceptually and empirically the role tax-deferred exchanges play in the determination of reservation prices and observed transaction prices in U.S. commercial real estate markets. Tax-deferred exchanges are transactions in which a taxpayer is able to defer payment of some, or all, of the federal income taxes associated with the disposition of real property by acquiring another property (or properties) of “like kind.” A taxpayer who initiates a delayed tax-deferred exchange has up to 45 days after the disposition of the “relinquished” property to identify a “replacement” property and 180 days (135 beyond the 45-day period) to complete the delayed exchange by acquiring the replacement property. If a taxpayer is successful in completing an exchange, the realized tax liability will be deferred until the replacement property is subsequently disposed of in a fully taxable sale.3 The present value of the income tax deferral benefit is therefore a function of the magnitude of the deferred capital gain, the expected holding period of the replacement property, and the applicable discount rate.

Taxpayers face significant compliance risk when seeking to complete the second leg of a tax-deferred exchange by identifying and purchasing a replacement property within the 45- and 180-day time limits. Moreover, the exchanger may compromise his or her bargaining position with potential sellers of replacement properties. In a perfectly competitive market where the law of one price obtains, a weakened bargaining position would not affect the transaction price. However, in illiquid, highly segmented commercial real estate markets, the exchanging taxpayer may be required to pay a premium for the replacement property relative to its fair market value.

The use of tax-deferred exchanges has grown significantly during the last decade.4 For example, in 2004 an estimated 80% of commercial real estate transactions on the West Coast involved the use of an exchange (McLinden 2004). Moreover, according to Weller and Halfacre (2004), “the prevailing wisdom is that 1031 buyers pay more, do less due diligence and accept more risk relating to their property purchases than buyers not motivated by tax savings.” Nevertheless, very little empirical evidence exists about the effects of tax-deferred exchanges on observed transaction prices and whether the price effects have varied over time, across geographic markets, or across property types. In fact, only a handful of papers have investigated empirically the pricing effects of Section 1031 exchanges (Holmes and Slade 2001; Lambson et al. 2004). Moreover, a major weakness of prior studies is that they examine just one property type (apartments) in one geographic market (Phoenix). The results are therefore difficult to generalize.

The paper proceeds as follows. The “The Mechanics of Tax-Deferred Exchanges” section presents an overview of tax-deferred exchanges and discusses their potential advantages and disadvantages. The “Estimating the Value of Capital Gain Tax Deferral” section develops a numerical model for quantifying the tax-deferral benefits of Section 1031 exchanges. In the “Data” section, we present the empirical dataset and summary statistics. The “Empirical Methodology” and “Empirical Results” sections discuss the regression methodology, present the empirical results, and compare the magnitude of the price premiums estimated by regression to the value of tax deferral obtained from the numerical model. The “Conclusion” section summarize the analysis and offer some concluding comments.

The Mechanics of Tax-Deferred Exchanges

Realized gains from the sale of real property must generally be recognized for federal income purposes in the year of sale.5 In general, the realized gain is equal to the net selling price of the property minus the adjusted tax basis.6 However, under Section 1031 of the IRC, real estate owners who dispose of their investment, rental, or vacation property and reinvest the net proceeds in other “like kind” property are able to defer recognition of some or all of the capital gain realized on the sale of the relinquished property.7 The taxpayer’s basis in the replacement property is set equal to the transaction price of the replacement property, minus the gain deferred on the disposition of the relinquished property. Therefore, when (if) the replacement property is subsequently disposed of in a fully taxable sale, the realized gain will be larger to the extent of the deferred gain. If the subsequent disposition of the replacement property is also structured in the form of a Section 1031 exchange, the realized gain can again be deferred.8

In order for the exchanging taxpayer to avoid completely the immediate recognition of the accrued taxable gain, he or she must acquire a like-kind property of equal or greater value than the relinquished property. In addition, the taxpayer must use all of the net cash proceeds generated from the sale of the relinquished property. The transaction is taxable to the extent that (1) the value of the replacement property is less than the value of the relinquished property and (2) there is cash left over after the purchase of the replacement property.

Most Section 1031 transactions are “delayed” exchanges that involve the use of a qualified intermediary (QI).9 In a delayed exchange, ownership of the relinquished property is transferred to the buyer. However, the buyer of the relinquished property transfers the agreed upon cash amount to the QI, not the selling taxpayer. This first phase of the delayed exchange, often referred to as the taxpayer’s “down-leg,” is depicted in the top portion of Fig. 1. The cash paid by the buyer of the relinquished property is “parked” with the QI until the taxpayer is able to identify and close on a replacement property. Typically, the taxpayer has not yet identified a replacement property when title to the relinquished property is transferred to the buyer of the relinquished property.
https://static-content.springer.com/image/art%3A10.1007%2Fs11146-007-9099-6/MediaObjects/11146_2007_9099_Fig1_HTML.gif
Fig. 1

Delayed exchange with intermediary

Within 45 days of the sale of the relinquished property, the taxpayer must formally identify the replacement property. To allow for the possibility that the taxpayer may not be able to come to terms with the owner of the potential replacement property, the taxpayer may designate more than one replacement property.10 The taxpayer must acquire one or more of the identified replacement properties within 180 days of the date of the closing of the relinquished property; that is, the 45 and 180 day periods run concurrently. There are no exceptions to these time limits and failure to comply will convert the transaction to a fully taxable sale. At the closing of the replacement property, the QI transfers cash to the seller of the replacement property and the seller transfers ownership to the taxpayer. This second phase of the delayed exchange, often referred to as the taxpayer’s “up-leg,” is depicted in the bottom portion of Fig. 1. Delayed exchanges are often referred to as “Starker” exchanges, which is a reference to the 1979 U.S. Supreme Court case that established the legal basis for non-simultaneous, tax-deferred exchanges.11

In general, both real and personal property can qualify for tax-deferred treatment. However, some types of property are specifically disqualified; for example, stocks, bonds, notes, and ownership interests in a limited partnership or multi-member limited liability company.12 Both the relinquished property and the replacement property must be held for productive use in trade or business or held as a “long-term investment.” Thus, personal residences and property held for sale to consumers (i.e., “dealer” property) cannot be part of a Section 1031 exchange.13 A holding period greater than 1 year is commonly assumed to qualify the relinquished property as a long-term investment for the purposes of implementing a tax-deferred exchange; however, the 1-year rule of thumb has no basis in statutory or case law.

The tax literature and popular press point to several motivations for use of Section 1031 exchanges. First, exchanges serve as an effective shelter from taxes, thereby preserving investment capital. In addition, exchanges can be used to upgrade portfolios (Fickes 2003). By deferring taxes, the taxpayer can also leverage appreciation and afford to acquire a larger/higher priced replacement property. Section 1031 exchanges can also be used to consolidate or diversify properties, exchange low-return properties for high-return properties, or to substitute depreciable property for nondepreciable property (Wayner 2005a, b).

Despite the potential advantages of tax-deferral, Section 1031 exchanges have several drawbacks. First, the taxpayer’s basis in the replacement property is set equal to the market value of the replacement property, minus the deferred gain. Thus, the larger the amount of tax-deferral, the smaller is the depreciable basis in the replacement property and, therefore, the smaller is the allowable annual deduction for depreciation. Moreover, the larger the amount of tax-deferral, the larger will be the realized gain if and when the replacement property is subsequently disposed of in a fully taxable sale.

A second disadvantage is that the transaction costs (both monetary and non-monetary) associated with initiating and completing an exchange will likely exceed the costs of a fully taxable sale. The additional costs may include settlement fees, intermediary fees, and attorney preparation fees (Wayner 2005b). These disadvantages are explicitly considered in the numerical tax-deferral valuation model presented in the next section.14

Estimating the Value of Capital Gain Tax Deferral

Assume a taxpayer who owns an income producing property has decided that the risk–return characteristics of her portfolio would be enhanced by disposing of the asset and reinvesting his (her) equity into a replacement property located in a market with more growth potential. Assume also that the replacement property has already been identified. The first strategy available to the taxpayer is to dispose of the existing property in a fully taxable sale and then use the net sale proceeds, along with additional equity capital, to acquire the replacement property. The second option is to take advantage of Section 1031 of the IRC and exchange out of the existing property and into the replacement property. The second strategy would allow the taxpayer to defer recognition of the accrued taxable gain.

The net present value of the sale–purchase strategy, NPVSALEt, assuming all-equity financing, can be represented as
$${\text{NPVSALE}}_{t\,} = \,\left( {{\text{ATSP}}_t^1 \, - \mathop P\nolimits_t^2 } \right)\,\, + \,\sum\limits_{i = 1}^n {\frac{{\left( {1\, - \,\tau _o } \right)I_i \, + \,\tau _o {\text{DEP}}_i^{2,s} }}{{\left( {1\, + \,k} \right)^i }}\, + \,\frac{{P_{t + n}^2 \, - \,{\text{SC}}_{t + n}^2 \, - \,\tau _{{\text{cg}}} {\text{CG}}_{t + n}^{2,s} \, - \,\tau _{{\text{dr}}} {\text{RECAP}}_{t + n}^{2,s} }}{{\left( {1\, + \,k} \right)^n }}} $$
(1)
where:
\({\text{ATSP}}_t^1 \)

the net after-tax proceeds from the sale of the existing property at time t;

\(\mathop P\nolimits_t^2 \)

the acquisition price of the replacement property at time t;

\(\tau _o \)

the taxpayer’s marginal tax rate on ordinary income;

Ii

the expected net cash flow of the replacement property in year i of the expected n-year holding period;

\({\text{DEP}}_i^{2,s} \)

allowable depreciation on the replacement property in year i, conditional on a sale–purchase strategy;

k

the required after-tax rate of return on unlevered equity;

\(P_{t + n}^2 \)

the expected price of the replacement property in year t+n;

\({\text{SC}}_{t + n}^2 \)

expected selling costs on the disposition of the replacement property in year t+n;

\(\tau _{{\text{cg}}} \)

the tax rate on capital gain income;

\({\text{CG}}_{t + n}^{2,s} \)

expected capital gain income on the sale of the replacement property in year t+n, conditional on a sale–purchase strategy;

\(\tau _{{\text{dr}}} \)

the tax rate on depreciation recapture income; and

\({\text{RECAP}}_{t + n}^{2,s} \)

depreciation recapture income on the sale of the replacement property in year t +n, conditional on a n-year sale–purchase strategy.

The first term on the right-hand-side of Eq. 1 represents the additional equity capital that must be invested at time t under the sale–purchase strategy, and is equal to the after-tax proceeds from a fully taxable sale minus the acquisition price of the replacement property. Note that \({\text{ATSP}}_t^1 = P_t^1 - {\text{SC}}_t^1 \, - \,{\kern 1pt} \,{\text{TDS}}_t^1 \), where \({\text{SC}}_t^1 \) and \({\text{TDS}}_t^1 \) represent sale costs and taxes due on sale, respectively. Therefore, if the price of the replacement property is equal to the price of the existing property, then \({\text{ATSP}}_t^1 - P_t^2 \) is equal to total taxes due on the sale of the existing property, plus total selling costs.
The second term on the right-hand-side of Eq. 1 represents the cumulative present value of the replacement property’s net cash flows from annual operations, plus the present value of the annual tax savings generated by depreciation. Annual depreciation, \({\text{DEP}}_i^{2,s} \), is equal to
$${\text{DEP}}_i^{2,s} \, = \,\frac{{\left( {1\, - \,L_t^2 } \right)P_t^2 }}{{{\text{RECPER}}}}$$
(2)
where \(P_t^2 \) is the acquisition price of the replacement property, \(L_t^2 \) is the percentage of \(P_t^2 \) that represents non-depreciable land, and RECPER is the allowable cost recovery period for the replacement property.15 Because the replacement property is purchased with the proceeds from a fully taxable sale, the original tax basis of the replacement property is “stepped up” to equal the total acquisition price, \(P_t^2 \) , thereby maximizing allowable depreciation deductions over the expected n-year holding period.

The third and final term on the right-hand-side of Eq. 1 represents the expected after-tax cash proceeds from the sale of the replacement property at the end of the assumed n-year holding period. Deducted from the expected selling price of the replacement property at time t + n are the following: expected selling costs \(\left( {{\text{SC}}_{t + n}^2 } \right)\), the expected capital gain tax liability \(\tau _{{\text{cg}}} {\text{CG}}_{t + n}^{2,s} \), and the expected depreciation recapture tax \(\left( {\tau _{{\text{dr}}} {\text{RECAP}}_{t + n}^{2,s} } \right)\). Capital gain income, under the tax rules in place in 2007, is subject to a maximum 15% tax rate.16 In contrast, the maximum federal statutory rate on ordinary income is 35%.

The second disposition option available to the taxpayer is to take advantage of Section 1031 of the IRC and exchange into the replacement property. The net present value of the exchange strategy, assuming all-equity financing, can be represented as
$${\text{NPVEX}}_t = P_t^1 - {\text{EC}}_t - P_t^2 + \sum\limits_{i = 1}^n {\frac{{\left( {1 - \tau _o } \right)I_i + \tau _o {\text{DEP}}_i^{2,e} }}{{\left( {1 + k} \right)^i }} + \frac{{P_{t + n}^2 - SC_{t + n}^2 - \tau _{{\text{cg}}} {\text{CG}}_{t + n}^{2,e} - \tau _{{\text{dr}}} {\text{RECAP}}_{t + n}^{2,e} }}{{\left( {1 + k} \right)^n }}} $$
(3)
where:
\(P_t^1 \)

the selling price of the relinquished property;

\({\text{EC}}_t^{} \)

the total transaction costs of exchanging out of the relinquished property and into the replacement property at time t;

\({\text{DEP}}_i^{2,e} \)

depreciation on the replacement property in year i, conditional on the use of an exchange at time t;

\({\text{CG}}_{t + n}^{2,e} \)

the expected capital gain income on the sale of the replacement property in year t+n, conditional on an exchange strategy; and

\({\text{RECAP}}_{t + n}^{2,e} \)

depreciation recapture income on the sale of the replacement property in n years assuming an exchange at time t.

All other variables are as previously defined. The taxpayer should exchange into the replacement property if the net present value of the exchange strategy exceeds the net present value of the sale–purchase strategy. Subtraction of Eq. 1 from Eq. 3 produces the following expression for the incremental NPV of the exchange strategy:
$$\begin{array}{*{20}l} {{\text{INCNPV}}_t \, = \,\left[ {{\text{SC}}_t^1 \, - \,{\kern 1pt} {\text{EC}}_t \, + \,{\text{TDS}}_t^1 } \right]{\kern 1pt} - {\kern 1pt} \,\sum\limits_{i = 1}^n {\frac{{\tau _o \left( {{\text{DEP}}_i^{2,s} \, - \,{\text{DEP}}_i^{2,e} } \right)}}{{\left( {1\, + \,k} \right)^i }}\,\,{\kern 1pt} + \,{\kern 1pt} \frac{{\tau _{{\text{dr}}} {\kern 1pt} \left( {{\text{RECAP}}_{t + n}^{2,s} \, - \,{\text{RECAP}}_{t + n}^{2,e} } \right)}}{{\left( {1\, + \,k} \right)^n }}} \,} \hfill \\ {\quad \quad \quad \quad - \,\,\frac{{\tau _{{\text{cg}}} {\kern 1pt} \left( {{\text{CG}}_{t + n}^{2,e} \, - \,{\text{CG}}_{t + n}^{2,s} } \right)}}{{\left( {1\, + \,k} \right)^n }}\;.\;} \hfill \\ \end{array} $$
(4)

The first term in Eq. 4, \(\left[ {{\text{SC}}_t^1 \, - \,{\kern 1pt} {\text{EC}}_t \, + \,{\text{TDS}}_t^1 } \right]\), captures the immediate net benefit of tax deferral. Note that if the time t selling costs associated with the sale–purchase strategy and exchange strategy are equal, the immediate advantage of the exchange is equal to \({\text{TDS}}_t^1 \), the deferred tax liability. To the extent exchanges are more expensive to execute than fully taxable sales, \({\text{SC}}_t^1 \, - \,{\kern 1pt} {\text{EC}}_t \) will be negative and this incremental outflow will be netted against the positive tax deferral benefits.

The second term in Eq. 4 captures the cumulative present value of the foregone depreciation deductions over the n-year holding period. As noted above, the tax basis of the replacement property is reduced by the amount of the taxable gain deferred by the exchange, which insures that \({\text{DEP}}_i^{2,s} {\text{ $>$ DEP}}_i^{2,e} \). To the extent annual depreciation deductions are reduced by an exchange, the amount of depreciation recaptured when the replacement property is sold in year t+n is lower. The present value of the reduced depreciation recapture taxes is reflected in the third term in Eq. 4.

Finally, because the tax deferral associated with an exchange reduces the tax basis in the replacement property, the taxable capital gain due on the sale of the replacement property will be larger with an exchange. The negative effect of the increased capital gain tax liability on the incremental NPV of an exchange is captured by the fourth term in Eq. 4.

Estimating the Magnitude of Tax Deferral Benefits

Equation 4 is used to estimate the magnitude of INCNPVt under a number of plausible assumptions. Simulated values of INCNPVt are then divided by the price of the replacement property at time t to determine the percentage price effect. These simulations are intended to quantify the range of maximum price premiums exchange motivated buyers can afford to pay for replacement properties. To solve Eq. 4 numerically, the following base-case parameter assumptions are employed:
  • Price of relinquished and replacement property: \(\mathop P\nolimits_t^1 = \mathop P\nolimits_t^2 \)

  • Cost of a fully taxable sale (\({\text{SC}}_t^1 \) and \({\text{SC}}_{t + n}^2 \)): 3% of sale price

  • Exchange costs (ECt): equal to \({\text{SC}}_t^1 \)

  • Ordinary income tax rate (τo): 35%

  • Capital gain tax rate (τcg): 15%

  • Depreciation recapture tax rate (τdr): 25%

  • After-tax discount rate (k): 8%

  • Non-depreciable portion of original tax basis (\(L_{t - h}^1 \) and \(L_t^2 \)): 20%

We assume \(\mathop P\nolimits_t^1 = \mathop P\nolimits_t^2 \) to abstract from any effects unequal equity positions would have on time t inflows and outflows and future depreciation deductions. Other key assumptions include the number of years since acquisition of the relinquished property (HOLD1), the annualized rate of price appreciation since acquisition of the relinquished property (π1), and the expected holding period of the replacement property (HOLD2).

Table 1 presents the numerical results for residential income producing property. Panel A contains the base case simulation results. One pattern is especially noteworthy: the incremental value of an exchange is unambiguously positively related to HOLD1. For example, assuming HOLD1 = 5, HOLD2 = 8, and π1 = 6%, INCNPVt is equal to 2.48% of \(\mathop P\nolimits_t^2 \). This implies the taxpayer could afford to pay up to a 2.48% premium for the replacement property without decreasing his wealth, all else equal. As HOLD1 increases to 10, the value of tax deferral rises from 2.48% to 4.03%. Assuming HOLD1 = 20, the value of tax deferral increases further to 5.33%. In short, the relative attractiveness of the exchange strategy is unambiguously positively related to the magnitude of the accumulated gain on the relinquished property.
Table 1

Incremental NPV of apartment exchange as a percent of replacement property value

 

Holding period of replacement property

Holding period of replacement property

(HOLD1)

(π1) (%)

4

8

12

16

20

24

28

4

8

12

16

20

24

28

Panel A: τcg = 15%, k = 8%, \({\text{EC}}_t = {\text{SC}}_t^1 \)

Panel B: τcg = 15%, k = 10%, \({\text{EC}}_t = {\text{SC}}_t^1 \)

5

2

1.85

2.14

2.27

2.30

2.28

2.23

2.17

2.01

2.35

2.51

2.55

2.55

2.51

2.47

5

6

2.00

2.48

2.69

2.75

2.71

2.63

2.53

2.25

2.84

3.10

3.18

3.16

3.11

3.04

5

10

2.11

2.74

3.03

3.10

3.06

2.95

2.81

2.45

3.23

3.57

3.67

3.66

3.58

3.49

5

20

2.30

3.21

3.61

3.72

3.66

3.51

3.31

2.79

3.90

4.38

4.53

4.51

4.41

4.27

10

2

3.45

4.01

4.26

4.33

4.29

4.20

4.07

3.75

4.44

4.74

4.83

4.82

4.76

4.67

10

6

3.20

4.03

4.40

4.50

4.44

4.30

4.12

3.64

4.66

5.11

5.24

5.22

5.13

5.00

10

10

3.03

4.04

4.49

4.61

4.54

4.37

4.15

3.57

4.80

5.35

5.51

5.49

5.37

5.22

10

20

2.81

4.05

4.61

4.75

4.67

4.46

4.19

2.79

3.90

4.38

4.53

4.51

4.41

4.27

20

2

5.80

6.80

7.25

7.36

7.29

7.13

6.91

6.34

7.55

8.09

8.25

8.23

8.12

7.97

20

6

4.12

5.33

5.87

6.02

5.93

5.73

5.46

4.76

6.25

6.91

7.11

7.07

6.94

6.75

20

10

3.35

4.67

5.26

5.41

5.32

5.10

4.81

4.06

5.66

6.37

6.59

6.55

6.40

6.20

28

2

7.04

8.29

8.85

9.00

8.91

8.71

8.43

7.71

9.24

9.91

10.12

10.09

9.95

9.76

28

6

4.15

5.50

6.11

6.27

6.17

5.95

5.65

4.87

6.53

7.26

7.48

7.44

7.29

7.08

28

10

3.19

4.57

5.19

5.36

5.26

5.03

4.72

3.93

5.62

6.37

6.60

6.56

6.40

6.19

Panel C: τcg = 15%, k = 8%, \({\text{EC}}_t = 1.2 \times {\text{SC}}_t^1 \)

Panel D: τcg = 15%, k = 10%, \({\text{EC}}_t = 1.2 \times {\text{SC}}_t^1 \)

5

2

1.33

1.60

1.73

1.76

1.74

1.69

1.63

1.48

1.81

1.96

2.00

2.00

1.97

1.92

5

6

1.47

1.94

2.15

2.21

2.17

2.10

1.99

1.72

2.30

2.55

2.63

2.62

2.56

2.49

5

10

1.58

2.21

2.49

2.56

2.52

2.42

2.28

1.92

2.68

3.02

3.12

3.11

3.04

2.94

5

20

1.78

2.67

3.08

3.18

3.12

2.97

2.77

2.26

3.35

3.84

3.99

3.96

3.86

3.72

10

2

2.92

3.48

3.73

3.79

3.75

3.66

3.54

3.22

3.90

4.19

4.29

4.27

4.21

4.12

10

6

2.67

3.49

3.86

3.96

3.90

3.77

3.58

3.11

4.12

4.56

4.69

4.67

4.58

4.46

10

10

2.50

3.50

3.95

4.07

4.00

3.83

3.61

3.04

4.26

4.80

4.96

4.94

4.83

4.67

10

20

2.29

3.52

4.07

4.21

4.13

3.92

3.65

2.95

4.45

5.11

5.31

5.28

5.14

4.96

20

2

5.28

6.26

6.71

6.82

6.75

6.59

6.37

5.80

7.01

7.54

7.71

7.68

7.57

7.42

20

6

3.59

4.80

5.34

5.48

5.39

5.19

4.93

4.23

5.71

6.36

6.56

6.53

6.39

6.21

20

10

2.83

4.13

4.72

4.87

4.78

4.56

4.27

3.52

5.12

5.82

6.04

6.00

5.86

5.66

28

2

6.52

7.76

8.32

8.46

8.38

8.17

7.89

7.18

8.70

9.37

9.57

9.54

9.40

9.21

28

6

3.63

4.97

5.57

5.73

5.63

5.41

5.11

4.34

5.99

6.71

6.93

6.89

6.74

6.54

28

10

2.66

4.04

4.65

4.82

4.72

4.49

4.19

3.40

5.08

5.82

6.05

6.01

5.86

5.65

Panel E: τcg = 20%, k = 8%, \({\text{EC}}_t = {\text{SC}}_t^1 \)

Panel F: τcg = 20%, k = 10%, \({\text{EC}}_t = {\text{SC}}_t^1 \)

5

2

1.46

1.93

2.20

2.33

2.39

2.40

2.38

1.66

2.22

2.52

2.66

2.72

2.74

2.72

5

6

1.89

2.69

3.15

3.38

3.47

3.49

3.45

2.23

3.18

3.68

3.93

4.03

4.05

4.03

5

10

2.24

3.30

3.90

4.21

4.33

4.35

4.30

2.69

3.95

4.61

4.94

5.07

5.11

5.08

5

20

2.84

4.36

5.21

5.65

5.83

5.86

5.78

3.49

5.27

6.23

6.69

6.88

6.93

6.89

10

2

2.77

3.71

4.24

4.51

4.62

4.64

4.59

3.17

4.28

4.87

5.16

5.28

5.31

5.28

10

6

3.15

4.54

5.32

5.72

5.88

5.91

5.84

3.74

5.38

6.25

6.68

6.85

6.89

6.86

10

10

3.39

5.08

6.03

6.51

6.71

6.75

6.66

4.11

6.10

7.16

7.68

7.89

7.94

7.90

10

20

3.71

5.78

6.95

7.55

7.79

7.83

7.73

4.60

7.04

8.34

8.97

9.24

9.30

9.24

20

2

4.76

6.42

7.36

7.84

8.04

8.07

7.99

5.47

7.43

8.48

8.99

9.20

9.25

9.21

20

6

4.32

6.35

7.50

8.08

8.32

8.36

8.26

5.19

7.58

8.86

9.48

9.74

9.80

9.75

20

10

4.12

6.32

7.56

8.19

8.45

8.49

8.38

5.06

7.65

9.03

9.70

9.98

10.05

9.99

28

2

5.85

7.94

9.12

9.72

9.97

10.01

9.90

6.74

9.21

10.52

11.16

11.42

11.49

11.43

28

6

4.59

6.85

8.13

8.78

9.04

9.09

8.97

5.56

8.22

9.64

10.33

10.62

10.69

10.63

28

10

4.17

6.49

7.80

8.46

8.74

8.78

8.66

5.16

7.89

9.35

10.06

10.35

10.42

10.36

HOLD1 is the number of years since acquisition of the relinquished property, π1 is the annualized rate of price appreciation since acquisition of the relinquished property, τcg is the capital gain tax rate, k is the after-tax discount rate, ECt is the up-front cost of a tax-deferred exchange, and \(SC_t^1 \) is the up-front cost of a fully taxable sale

The relation between INCNPVt and π1, however, for a given HOLD1 is less clear. For example, assuming HOLD1 = 5, increased price appreciation prior to time t produces slight increases in INCNPVt. However, with HOLD1 = 20, higher values of π1 produce lower values of INCNPVt. With HOLD1 = 10, the relation between π1 and INCNPVt is sensitive to the assumed value of HOLD2.

All else equal, the value of tax deferral increases with the expected holding period of the replacement property. However, Panel A of Table 1 indicates that INCNPVt increases with HOLD2, but at a decreasing rate. For expected holding periods longer than eight to ten years, INCNPVt is largely unaffected by increases in HOLD2 and, in fact, for holding periods in excess of 16 years the value of INCNPVt begins to decrease. Overall, the benefits of tax deferral in Panel A range from 1.85% to 9.00% of value.

The tax deferral values displayed in Panel B of Table 1 are based on an increase in the assumed after-tax equity discount rate (k) to 10% from 8%. All other variables remain at their base case levels. Comparison of Panel A and Panel B demonstrates that a higher discount rate unambiguously increases the incremental value of the exchange option. This is because the value of tax deferral produced by the exchange is immediate. In contrast, the foregone depreciation deductions and the increased future capital gain tax liability at sale that results from the decreased tax basis both occur in subsequent years. Thus, the present value of these future cash outflows is reduced by a higher discount rate. The present value of tax deferral in this panel ranges from 2.01% to 10.12% of property value.

The tax deferral values reported in Panel C of Table 1 assume the discount rate has been reset to 8%, but the dollar costs of executing an exchange (ECt) are 20% higher than the costs of a fully taxable sale \(\left( {{\text{SC}}_t^1 } \right)\). As expected, higher up-front exchange costs reduce the value of an exchange (relative to the base case in Panel A). However, the decreases are modest, averaging approximately one-half of a percentage point across varying assumptions for HOLD1, HOLD2, and π1. The value of tax deferral in this panel ranges from 1.33% to 8.46%.

In Panel D, the dollar costs of executing an exchange (ECt) are assumed to be 20% higher than the costs of a fully taxable sale \(\left( {{\text{SC}}_t^1 } \right)\) and the discount rate is set to 10%. With these assumptions, the value of tax deferral ranges from 1.48% to 9.57%. Panel E reports the results assuming a tax rate on capital gain income of 20% (the maximum statutory τcg from 1999 to 2003). Clearly, the immediate value of tax deferral is larger the higher is τcg. However, the simulated tax deferral values in Panel E are not uniformly higher than those reported in Panel A. This is because a higher capital gain tax rate will also increase the tax liability that results from the eventual sale of the replacement property.

The longer the expected holding period of the replacement property, the more likely it is that the immediate tax deferral benefits associated with a higher τcg will exceed the present value of the increased taxes due on the subsequent sale of the replacement property. This anticipated result is confirmed in panel E. That is, increasing τcg from 15% to 20% produces larger values of INCNPVt, except in some cases where the magnitude of the deferred gain is small (i.e., when HOLD1 and π1 are small) or when the expected holding period of the replacement property is relatively short. The value of tax deferral in Panel E ranges from 1.46% to 10.01%. Finally, Panel F of Table 1 reports the results assuming a tax rate on capital gain income of 20% and after tax discount rate of 10%. The maximum value of tax deferral in this panel is 11.49 (HOLD1 = 28, HOLD2 = 24 and π1 = 2%).

Overall, the maximum value of tax deferral reported in Table 1 range from 8.50% to 11.49% across the six panels. A corresponding set of simulations were performed for non-residential real estate with its 39-year cost recovery period. Maximum tax deferral values provided by an exchange to an owner of non-residential property ranged from 9.3% to 13.3% (see Table 8 in the Appendix).

Data

Property level transaction data were obtained from CoStar Group, Inc. The CoStar Comps Professional database includes historical information on commercial real estate transactions in more than 50 major U.S. markets dating back to 1999.17 To assure reliability of the data, CoStar requires agents to physically inspect the site and record a variety of property characteristics and transaction details. After the deletion of observations with missing data, with sale prices less than $250,000, or not confirmed by CoStar, our initial 1999–2005 sample contains 124,830 transactions in five property markets segments: office, industrial, apartment, retail and hotel/motel. To keep the analysis manageable, we focus on the apartment and office market; thus, the 63,799 retail, hotel and industrial property transactions were dropped from the sample. In addition, we exclude 8,943 transactions associated with a CoStar delineated “special condition;” for example, sales that are part of an auction or bankruptcy, or sales that involve building contamination, natural disaster damage, or the threat of contamination. In detailed notes, the CoStar database also contains descriptive information on the type of exchange employed by the buyer, seller, or both (if applicable). Based on manual inspection of these notes, 1,553 additional observations were deleted because it could not be determined what type of exchange was involved or the identified type of exchange was something other than a delayed (Starker) exchange.18 These deletions produced a final usable sample of 50,555 apartment and office transactions.

Finally, to ensure our sample includes a number of delayed exchanges sufficient to generate meaningful statistical results, we limit our regression analysis to the 15 metropolitan markets that contain the largest number of apartment and office property transactions. This final sample contains 31,533 transactions over the 1999–2005 period. Of these, 23,153 are apartment transactions and 8,300 are office transactions.

Table 2 displays the number of exchange and non-exchange transactions by sample year and property type. Of the 23,153 apartment transactions, 7,482 (approximately 32%) involved the use of a delayed exchange. Moreover, this percentage has remained remarkably stable over the sample period. In our office sample, delayed exchanges account for 20% of all transactions.
Table 2

Distribution of sample exchange and non-exchange sales by year

 

Apartments

Offices

Exchangea

Non-exchange

All

Exchange

Non-exchange

All

1999

383

660

1,043

60

258

318

2000

1,310

2,634

3,944

295

1,116

1,411

2001

1,220

2,687

3,907

286

1,104

1,390

2002

1,464

2,926

4,390

305

1,201

1,506

2003

1,429

2,882

4,311

314

1,250

1,564

2004

1,263

3,009

4,272

317

1,349

1,666

June 2005

413

873

1,286

129

396

525

Total

7,482

15,671

23,153

1,706

6,674

8,380

Property level transaction data were obtained from CoStar Group, Inc. To assure reliability of the data, CoStar requires agents to physically inspect the site and record a variety of property characteristics and transaction details. After the deletion of observations with missing data, with sale prices less than $250,000, or that were not confirmed by CoStar, our initial 1999–2005 sample contains 124,830 transactions in five property markets segments: office, industrial, apartment, retail and hotel/motel. The 63,799 retail, hotel and industrial property transactions were dropped from the sample. In addition, we exclude 8,943 transactions associated with a CoStar delineated “special condition;” 1,553 additional observations were deleted because it could not be determined what type of exchange was involved or the identified type of exchange was something other than a delayed (Starker) exchange. Finally, we limit our regression analysis to the 15 metropolitan markets that contain the largest number of apartment and office property transactions. This final sample contains 31,533 transactions over the 1999–2005 period. Of these, 23,153 are apartment transactions and 8,300 are office transactions

aExchange properties include only those transactions that involved the use Section 1031 delayed exchanges

Apartments

Table 3 disaggregates our apartment and office samples by metropolitan statistical area (MSA). The table’s top panel reveals that the most active apartment market has been Los Angeles, which represents 33% of all transactions in our apartment sample. New York City is the second most active apartment market with 17% of all transactions. Table 3 clearly reveals that the use of delayed exchanges varies substantially across the major metropolitan markets. Delayed exchanges represent the dominant form of property transactions in four of our 15 apartment markets (San Diego, San Francisco, Portland and Sacramento). Interestingly, all of these markets are located in the Western U.S., and three are in California. Delayed apartment exchanges have been much less prominent in the other major markets tracked by CoStar. For example, over the 1999–2005 time period, only 1.3% of apartment sales in New York involved an exchange.
Table 3

Description of sample property sales by metropolitan market: 1999–2005

#

Name

All

% of total

Exch.

% Exch.

EXREPL

EXRELQ

RELQ_REPL

Apartment properties

1

Los Angeles

7,748

33.5

2,838

36.6

1,020

1,149

669

2

New York City

3,922

16.9

50

1.3

20

25

5

3

San Diego

1,981

8.6

1,143

57.7

409

417

317

4

Chicago

1,532

6.6

238

15.5

122

83

33

5

Seattle

1,270

5.5

606

47.7

189

315

102

6

Phoenix

1,241

5.4

109

8.8

29

63

17

7

Oakland

1,173

5.1

536

45.7

220

170

146

8

San Francisco

765

3.3

401

52.4

220

103

78

9

Denver

730

3.2

342

46.8

108

145

89

10

Riverside/San Bernardino

578

2.5

226

39.1

72

96

58

11

Tucson

526

2.3

177

33.7

29

102

46

12

Portland

488

2.1

316

64.8

71

137

108

13

San Jose

470

2.0

225

47.9

106

74

45

14

Sacramento

370

1.6

230

62.2

66

96

68

15

Boston

359

1.6

45

12.5

14

30

1

 

Total

23,153

100

7,482

32.3

2,695

3,005

1,782

Office properties

1

Los Angeles

1,444

17.2

323

22.4

122

156

45

2

Phoenix

966

11.5

86

8.9

10

69

7

3

Washington D.C.

804

9.6

63

7.8

19

40

4

4

Chicago

708

8.4

72

10.2

19

49

4

5

Seattle

629

7.5

206

32.8

76

103

27

6

San Diego

582

6.9

202

34.7

47

118

37

7

Denver

549

6.6

160

29.1

47

89

24

8

Tampa

510

6.1

32

6.3

11

19

2

9

Ft. Lauderdale

426

5.1

32

7.5

7

23

2

10

Oakland

358

4.3

132

36.9

38

60

34

11

Dallas/Fort Worth

303

3.6

33

10.9

8

21

4

12

Las Vegas

298

3.6

94

31.5

14

68

12

13

Sacramento

287

3.4

126

43.9

20

72

34

14

Riverside/San Bernadino

268

3.2

104

38.8

17

69

18

15

Tucson

248

3.0

41

16.5

5

30

6

 

Total

8,380

100

1,706

20.4

460

986

260

Transaction data were obtained from CoStar Group, Inc. We delete observations with missing data, with sale prices less than $250,000, or that were not confirmed by CoStar. In addition, we exclude transactions associated with a CoStar delineated “special condition” as well as observations where it could not be determined what type of exchange was involved or the identified type of exchange was something other than a delayed (Starker) exchange. We limit our regression analysis to the 15 metropolitan markets that contain the largest number of apartment and office property transactions

EXREPL is a sale transaction that involves the purchase of a replacement property, EXRELQ is a transaction that involves the sale of a relinquished property, RELQ_REPL is a transaction that involves both the sale of a relinquished property and purchase of a replacement property

Aggregate summary statistics for the variables in our apartment regression dataset are presented in the first two columns of Table 4. The average apartment property sold for $2,075,996 (PRICE), is 49 years old, contains 21,902 ft2 of improvements (SQFT), is built on 38,186 ft2 of land (LANDSQFT), and has two floors (FLOORS), 28 parking spaces (PARKING), and 27 units (UNITS). Table 4 also reveals that 13% of apartment transactions involve the purchase of a replacement property to finalize a delayed exchange (EXREPL); 12% involve the sale of a relinquished property to initiate an exchange (EXRELQ); and 8% involve the sale of a property being used as both the seller’s relinquished property and the buyer’s replacement property (RELQ_REPL). Seven percent of apartment properties are classified as being in below average condition (COND_BA), while 14% are deemed to be in above average condition (COND_AA). The remaining 79% are classified as being in average condition (COND_A). Five percent of the apartment transactions occurred in 1999; 6% in 2005. The percentage of sale transactions that occurred in 2000–2004 ranges from 17% to 19%. Seven percent of apartment property buyers reside out-of-state (BUYOUT). Finally, 99% of the transaction involved traditional multifamily properties (MULTI); the remaining transactions involved government subsidized (SUBSIDY), or senior oriented (SENIOR) properties or properties that included one or more condominium units (CONDO).
Table 4

Aggregate summary statistics by property type

 

Apartment

Office

N = 23,153

N = 8,380

Mean

SD

Mean

SD

PRICE

$2,075,996

$4,438,127

$5,194,210

$9,606,126

AGE

49

26

29

22

SQFT

21,902

45,316

40,131

72,688

LANDSQFT

38,186

110,530

74,999

133,999

FLOORS

2

1

2

2

PARKING

28

77

97

166

UNITS

27

55

  

Binary variables

EXREPL

0.13

0.34

0.12

0.32

EXRELQ

0.12

0.32

0.05

0.23

RELQ_REPL

0.08

0.27

0.03

0.19

EXCH

0.32

0.47

0.20

0.40

COND_BA

0.07

0.26

0.03

0.16

COND_A

0.79

0.41

0.66

0.47

COND_AA

0.14

0.34

0.31

0.46

BUYOUT

0.07

0.26

0.15

0.36

YR1999

0.05

0.21

0.04

0.19

YR2000

0.17

0.38

0.17

0.37

YR2001

0.17

0.37

0.17

0.37

YR2002

0.19

0.39

0.18

0.38

YR2003

0.19

0.39

0.19

0.39

YR2004

0.18

0.39

0.20

0.40

YR2005

0.06

0.23

0.06

0.24

Apartment sub-property binary variables

MULTI

0.99

0.12

  

SENIOR

0.00

0.06

  

SUBSIDY

0.00

0.06

  

CONDO

0.01

0.09

  

Office sub-property binary variables

MULTI_LRISE

  

0.40

0.49

SNGL_LRISE

  

0.27

0.44

MED_DENT

  

0.10

0.31

OTHER

  

0.10

0.30

MIDRISE

  

0.08

0.26

HIRISE

  

0.05

0.21

These statistics combine, respectively, the 15 apartment and office metropolitan markets

PRICE is the nominal sale price; AGE is age of the property in years; SQFT is total square footage of improvements; LANDSQFT is land square footage; FLOORS is number of floors; PARKING is the number of parking spaces; UNITS is the number of apartment units; EXREPL is a binary variable set equal to one if transaction represents the purchase of a replacement property; EXRELQ is a binary variable set equal to one if the transaction represents sale of a relinquished property; RELQ_REPL is a binary variable set equal to one if transaction represents both sale of a relinquished property and purchase of a replacement property; COND_BA, COND_A, and COND_AA are binary variables that indicate the property is in below-average, average, or above average condition, respectively. BUYOUT is a binary variable indicating the buyer is an out-of-state resident; YRi are indicator variables for each year; MULTI, SENIOR, SUBSIDY, and CONDO are binary variables set equal to one if the property is: a traditional multifamily property, primarily a seniors oriented property, a property that contains units subsidized by one or more governmental programs, or a property that contained one or more owner-occupied (i.e., condo) units when acquired. MULTI_LRISE, SNGL_LRISE, MED_DENT, MIDRISE, and HIRISE, are binary variables that, respectively, indicate that the office property is a multi-tenant, low rise property, a single-tenant, low rise property, a medical–dental office property, a mid-rise property, or a high rise property. We classify the remaining seven properties types identified by CoStar as OTHER

Office Properties

The bottom panel of Table 3 documents that the most active office market has been Los Angeles, which accounts for 17% of our office transactions. Phoenix is the second most active market with 11% of all transactions, while Washington, DC is the third most active office market. The office sample differs from the apartment sample in several respects. First, there is less concentration of transactions in the most active markets. Second, exchanges do not represent the dominant form of property transactions in any office market, although exchange markets in the Western U.S. have relatively more activity. Replacement exchanges are more frequently observed than relinquished exchanges in the office sample. For example, replacement exchanges in Phoenix outnumber relinquished exchanges by a factor of seven, while in Las Vegas replacement exchanges are five times more frequent than relinquished property sales.

Aggregate summary statistics for the variables in our office data set are displayed in columns three and four of Table 4. The mean sales price of the 8,380 office transactions is $5,194,210 with a standard deviation of $9,606,126. The average office property is 29 years old, has 40,131 ft2 of improvements on two floors, is built on 74,999 ft2 of land, and has 97 parking spaces. Twelve percent of the office transactions involve the purchase of a replacement property to complete an exchange; 5% involve the sale of a relinquished property; and 3% represent both a sale of relinquished property and purchase of a replacement property. A large portion of the office properties are classified as being in excellent condition (31%), 66% are in good condition, while only 3% are categorized as being in below average condition. Fifteen percent of office property buyers reside out-of-state.

CoStar classifies office properties into 12 subcategories. Forty percent of the assets in our office sample are multitenant, low rise properties (MULTI_LRISE) and 27% are single-tenant, low rise properties (SNGL_LRISE). Medical–dental office assets (MED_DENT) constitute 10% of the sample. Mid-rise (MIDRISE) and high rise (HIRISE) properties account for 8% and 5% of the sample, respectively. The remaining seven properties types (OTHER) constitute 10% of the office sample.19

Empirical Methodology

We use hedonic regression to identify empirically the impact of delayed exchanges on observed transaction prices. Griliches (1971), Rosen (1974), and Epple (1987) have developed the traditional hedonic framework for modeling the prices of heterogeneous assets, which views a property as a bundle of utility-generating characteristics, such as size, age, amenities, and location. The shadow prices of these hedonic characteristics are, in theory, revealed through the sale of properties with differing characteristic bundles. According to the hedonic valuation model, the value of a property is the sum of the values of each component in the bundle.

More specifically, following Harding et al. (2003b), let i denote a commercial property defined by a bundle of characteristics, Ci. The vector of shadow prices corresponding to Ci is defined as s. The market value, Pe, of property i is simply a linear combination of Ci and s:
$$\ln \left( {P_i^e } \right)\; = \;s\prime C_i .$$
(5)
Note that there is no role for differential buyer and seller motivations, search costs, or bargaining power in Eq. 5. Rather, the model predicts that transactions will occur at a price equal to the market value of the property, net of adjustments for the physical and locational characteristics of the property.

In practice, it is difficult for potential buyers and sellers to observe the relevant vector of characteristic prices because commercial real estate assets trade in thin, informationally inefficient markets. As a result, search costs increase and some market participants may gain a degree of market power which, in turn, creates incentives for bargaining.20 In the context of the current paper, tax-motivated and time-constrained exchange buyers may be forced to share at least a portion of their expected tax deferral benefits with the seller in the form of a transaction price that exceeds the true market value of the property.

Assume bargaining power does not influence the underlying shadow prices of the physical and locational characteristics. Rather, bargaining increases or decreases the transaction price by a fixed percentage relative to Pe. Then, for property i we can write
$$\ln \left( {P_i^e } \right)\; = \;s\prime C_i \; + B_i \,,$$
(6)
where Bi denotes the impact of bargaining on the observed transaction price for property i. Positive values of Bi may obtain, for example, when a weak buyer negotiates with a strong seller.

We employ three proxies for Bi in our hedonic regressions. The first, EXREPL, is a binary variable that indicates the sale of a property being used by the buyer to complete a deferred Section 1031 exchange. As discussed above, such buyers are at a clear bargaining disadvantage relative to less constrained buyers. A positive coefficient on EXREPL would therefore suggest that, on average, constrained buyers pay for this inferior bargaining position in the form of an increased purchase price. Our second proxy for Bi, EXRELQ, is a binary variable that indicates the sale of a property being used by the seller to begin a delayed exchange. Note that such sellers are not at a competitive disadvantage relative to other sellers because in a delayed exchange they are under no time constraint to complete the sale of the relinquished property.21 We therefore posit that the coefficient on EXRELQ is zero for sellers of relinquished properties. Our final bargaining power proxy, RELQ_REPL, is a binary variable that indicates the property is being used both by the buyer to complete a delayed exchange and by the seller to begin a separate delayed exchange.

Because the economics of commercial real estate markets are decidedly local in nature, we separately estimate our apartment regression model for each of our 15 metropolitan areas, while controlling for the location of the property within a submarket of the MSA. The complete specification of our semi-log hedonic apartment regression model is as follows:
$$\begin{array}{*{20}c} {{\text{LNPRICE}}_m = \alpha _0 + \alpha _1 {\text{EXREPL}} + \alpha _2 {\text{EXRELQ}} + \alpha _3 {\text{RELQ\_REPL}} + \alpha _4 {\text{AGE}} + \alpha _5 {\text{AGE}}2 + \alpha _6 {\text{SQFT}}} \\ { + \alpha _7 {\text{SQFT}}2 + \alpha _8 {\text{LANDSQFT}} + \alpha _9 {\text{LANDSQFT}}2 + \alpha _{10} {\text{PARKING}} + \alpha _{11} {\text{FLOORS}} + \alpha _{12} {\text{UNITS}}} \\ \begin{aligned} & + \sum\limits_{i = 2}^3 {\beta _i {\text{CONDITION}}_i } + \alpha _{13} {\text{BUYEROUT}} + \alpha _{14} {\text{SENIOR}} + \alpha _{15} {\text{SUBSIDIZED}} + \alpha _{16} {\text{CONDO}} \\ & \quad \\ \end{aligned} \\ { + \sum\limits_{n = 2000}^{2005} {\chi _n {\text{YR}}_n } + \sum\limits_{s = 2}^T {\delta _s {\text{SMDUM}}_s } + \varepsilon _{\text{m}} } \\ \end{array} $$
(7)
Similarly, the specification of our office regression model, estimated separately for the 15 most active office MSAs, is:
$$\begin{array}{*{20}c} {{\text{LNPRICE}}_m = \alpha _0 + \alpha _1 {\text{EXREPL}} + \alpha _2 {\text{EXRELQ}} + \alpha _3 {\text{RELQ\_REPL}} + \alpha _4 {\text{AGE}} + \alpha _5 {\text{AGE}}2 + \alpha _6 {\text{SQFT}}} \\ { + \alpha _7 {\text{SQFT}}2 + \alpha _8 {\text{LANDSQFT}} + \alpha _9 {\text{LANDSQFT}}2 + \alpha _{10} {\text{PARKING}} + \alpha _{11} {\text{FLOORS}} + \sum\limits_{i = 2}^3 {\beta _i {\text{CONDITION}}_i } {\text{ }}} \\ { + \alpha _{12} {\text{BUYEROUT}} + \alpha _{13} {\text{SNGL\_LRISE}} + \alpha _{14} {\text{MED\_DENT}} + \alpha _{15} {\text{MIDRISE}} + \alpha _{16} {\text{HIRISE}} + \alpha _{17} {\text{OTHER}}} \\ { + \sum\limits_{n = 2000}^{2005} {\chi _n {\text{YR}}_n } + \sum\limits_{s = 2}^T {\delta _s {\text{SMDUM}}_s } + \varepsilon _{\text{m}} } \\ \end{array} $$
(8)
where:
LNPRICE

is the natural logarithm of the sale price;

α0

is a constant term;

EXRELQ

is a binary variable set equal to one if the transaction represents the sale of a relinquished property;

EXREPL

is a binary variable set equal to one if the transaction represents the purchase of a replacement property;

RELQ_REPL

is a binary variable set equal to one if the transaction represents both the sale of a relinquished property and purchase of a replacement property;

AGE

is age of the building(s) in years;

SQFT

is total square footage of improvements in thousands;

LANDSQFT

is the land square footage in thousands;

FLOORS

is number of floors in the structure;

UNITS

is the number of apartment units;

CONDi

is the physical condition of the property based on inspection. The categories include below average, average, and above average. The omitted category is average.

BUYOUT

is a binary variable indicating the buyer is an out-of-state resident;

PARKING

is the number of parking spaces;

SUBSIDY

is a binary variable set equal to one if the use of the apartment property is subsidized multi-family;

SENIOR

is a binary variable set equal to one if the use of the apartment property is senior multi-family;

CONDO

is a binary variable set equal to one if the use of the apartment property is multi-family condominium;

SNGL_LRISE, MED_DENT, MIDRISE, HIRISE, and OTHER

are binary variables that, respectively, indicate the office property is a single-tenant, low rise property, a medical–dental office property, a mid-rise property, a high rise property, or one of the remaining seven office properties types identified by CoStar. Multitenant low rise is the omitted category.

YRi

is a binary variable indicating the year of the transaction with 1999 suppressed; and

SMDUMi

is a binary variable indicating in which of the T submarkets within the metropolitan area the property is located.

An advantage of using the log of sale price as the dependent variable is that less weight is given to extreme values than when using untransformed prices. With this semi-log functional form, unit sale price per unit change in the characteristic is obtained by multiplying the estimated coefficient by the observed selling price.22

We expect a negative relation between property age and sale price and a positive coefficient on age squared (AGE2) in both our apartment and office models. This expectation reflects the frequently observed quadratic relation between price and age, although a positive “vintage” effect is sometimes observed for older properties. The estimated coefficients on SQFT, LANDSQFT, PARKING, FLOORS and UNITS are expected to be positive.

The variable CONDi controls for CoStar’s qualitative assessment of building condition. Buildings in average condition are used as the control group. We expect a positive (negative) relation between selling price and buildings deemed to be in above (below) average condition. Indicator variables for each year (YRi) are included to control for the effects of time; 1999 is used as the base year. SUBSIDY is a binary variable set equal to one if the property contains units that are subsidized by one or more governmental programs. SENIOR indicates whether the property is classified by Costar as a senior-oriented property. Finally, CONDO indicates whether the property contained one or more owner-occupied (i.e., condo) units when acquired. The comparison group is standard multifamily use.

We control for location in our MSA-level regressions by including standard submarket dummy variables. The submarkets are defined by CoStar and are based upon discussions with local brokers. For example, CoStar identifies 42 distinct apartment submarkets in Los Angeles, 30 in New York, and 21 in Sacramento. As a robustness check (discussed in a later section), we also implement the interactive specification developed by Fik et al. (2003) designed to address the potential limitations of using submarket dummy variables to control for location within a metropolitan area.

All continuous dependent variables in the regressions are winsorized at the top and bottom 1% of the distribution in order to avoid the effect of outliers. The winsorizing procedure takes the nonmissing values of a continuous variable sorted in ascending order and replaces its 1% highest and lowest values by the next value counting inwards from the extremes. The only exceptions are FLOORS, PARKING and UNITS, which we winsorize at the top and bottom 0.5% of the distribution, to account for the narrower distribution of these variables.

To preserve degrees of freedom, backward stepwise regression estimation is employed to select the submarket dummy variables to include in each of the final MSA-level models. The stepwise procedure begins with all submarkets included and then selectively removes submarket dummies that are not significant at the 10% level or that do not influence the fit of the model. We use a robust estimation method to account for potential heteroskedasticity; therefore all reported p-values are based on adjusted standard errors.

Equation 8 is estimated for our 15 office markets with SNGL_LRISE, MED_DENT, MIDRISE, HIRISE, and OTHER included. MULTI_LRISE is the omitted property type subcategory. CoStar delineates 488 distinct submarkets in our 15 office markets, with 51 present in the Los Angeles Metropolitan areas, 35 in Phoenix, and six in Tucson, the least active of our 15 office markets.

Empirical Results

Apartment Properties

Table 5 presents summary results for the 15 MSA-level apartment regressions; the individual MSA results are included in Tables 9 and 10 in the Appendix. Table 5 reveals that the estimated coefficients on the structural attributes are of the predicted sign and statistically significant in most of the regressions. For example, the estimated coefficient on AGE is negative and statistically significant in 10 of 15 apartment regression models. The estimated coefficient on SQFT is positive and significant in all 15 MSA-level regressions; however, the coefficient on SQFT2 is negative and significant in all 15 regressions. Somewhat surprisingly, the estimated coefficient on LANDSQFT is positive and significant in only one of the MSA-level regressions. We attribute this result to the significant correlation between the square footage of improvements and land. The estimated coefficient on PARKING displays little consistency across the 15 MSA estimations. Price tends to be positively related to the number of floors in the building and to the number of units, even after controlling for the square footage of the land and improvements.
Table 5

Summary of regression results for 15 apartment markets

Variable

Average coefficienta

Number of positive coefficientsa

Number of negative coefficients

Average P value

Number of positive and significant coefficients

Number of negative and significant coefficients

EXREPL

0.08

14

1

0.13

11

0

EXRELQ

0.04

13

2

0.29

7

1

RELQ_REPL

0.15

13

2

0.09

13

0

AGE

−0.01

1

14

0.20

0

10

AGE2

0.00

12

3

0.30

3

0

SQFT

0.04

15

0

0.00

15

0

SQFT2

0.00

0

15

0.00

0

15

LANDSQFT

0.00

10

5

0.56

1

0

LANDSQFT2

0.00

8

7

0.34

0

3

PARKING

0.00

6

9

0.34

2

2

FLOORS

0.06

12

3

0.20

8

0

UNITS

0.00

15

0

0.17

10

0

COND_BA

−0.12

0

15

0.15

0

11

COND_AA

0.11

14

1

0.16

9

0

BUYOUT

0.02

8

7

0.49

2

0

SENIOR

−0.03

8

7

0.31

4

4

SUBSIDY

−0.21

6

9

0.21

2

5

CONDO

0.48

12

3

0.20

10

1

YR2000

0.06

10

5

0.26

7

0

YR2001

0.18

13

2

0.21

10

0

YR2002

0.28

14

1

0.14

11

0

YR2003

0.39

15

0

0.05

13

0

YR2004

0.50

15

0

0.02

14

0

YR2005

0.61

15

0

0.00

15

0

SMDUMi (not reported)

      

CONST

13.29

15

0

0.00

15

0

Adjusted R-squared

0.87

     

aOut of a total of 15 separate MSA regressions. The dependent variable in each regression is the log of the sale price

EXREPL is a binary variable set equal to one if the transaction represents the purchase of a replacement property; EXRELQ is a binary variable set equal to one if transaction represents sale of a relinquished property; RELQ_REPL is a binary variable set equal to one if transaction represents both sale of a relinquished property and purchase of a replacement property. AGE is age of the building(s) in years; AGE2 is age squared; SQFT is total square footage of improvements in thousands; SQFT2 is SQFT squared divided by 1,000; LANDSQFT is the land square footage in thousands; LANDSQFT2 is LANDSQFT squared divided by 1,000; PARKING is the number of parking spaces; FLOORS is number of floors; UNITS is the number of apartment units. COND_BA and COND_AA are binary variables set equal to one if the property is in below-average or above average condition, respectively. The omitted condition category is average. BUYOUT is a binary variable indicating the buyer is an out-of-state resident; SENIOR, SUBSIDY, and CONDO are binary variables set equal to one if the property is, respectively, a senior oriented property, a property that contains units subsidized by one or more governmental programs, or a property that contained one or more owner-occupied (i.e., condo) units when acquired. The omitted property type category is traditional multifamily. YRi are dummy variables for each year of the sample (1999 is suppressed); SMDUMi are dummy variables indicating the CoStar delineated apartment submarket in which the property is located. CONST is a constant

As expected, apartment properties in below average condition transact at lower prices than properties in average condition; properties in above average condition sell at prices that are statistically significantly higher in 9 of 15 MSA-level regressions. Properties that contain subsidized units sell at lower prices, on average, and properties that contain condominium units sell at higher prices, all else equal. The estimated year dummies, with 1999 as the omitted year, are generally positive and significant, and their magnitude reveals substantial nominal price appreciation over the 7-year study period. Finally, although the estimated coefficients on the submarket dummy variables are not reported in Table 5, many are statistically significant and model fits are improved substantially by their use.

The estimated coefficient on the primary variable of interest, EXREPL, is positive and statistically significant in 11 of 15 MSA regressions. The average magnitude of the coefficient on EXREPL is 0.08. These results suggest that buyers of replacement properties are paying statistically significant price premiums in exchange for tax deferral benefits.

The coefficient estimates for EXRELQ are positive and significant in 7 of the 15 MSA-level regressions. What explains this unexpected result? Recall that the vector of important property characteristics, Ci, is known to the buyer and seller of the ith property, but only partially observable in the data (Harding et al. 2003b). Furthermore, the unobserved characteristics may be correlated with one or more characteristics of the buyer or seller. In the context of the current paper, if an important motivation of the seller is to postpone capital gain taxes, the relinquished properties in our sample may have experienced more rapid price appreciation than otherwise comparable properties in the market. To the extent we are unable to fully control for the determinants of relative price appreciation within a market, our empirical model may suffer from an omitted variable bias. This issue is addressed later in the paper.

The coefficient on RELQ_REPL is also positive and significant in 13 of 15 MSA-level regressions. As discussed, this coefficient represents the combined price effect of relinquished and replacement exchange motivation.

Office Properties

Table 6 reports summary regression results for the office sample. MSA-level results are contained in Tables 11 and 12 in the Appendix. The estimated coefficients on the structural attributes are again of the predicted sign and statistically significant in most MSA-level regressions. The estimated coefficient on AGE is negative in all 15 office regressions and statistically significant in all but one. The coefficient on AGE2 is positive and significant in 11 of 15 regressions. The estimated coefficient on SQFT is positive and significant in all MSA-level regressions; the coefficient on SQFT2 is negative and significant in all models. The positive relation between lot size and price is more pronounced than in the apartment sample with 11 of 15 coefficients on LANDSQFT displaying positive and statistically significant coefficients. The nonlinearity of the relation between lot size and price is also substantially more pronounced than in the apartment sample, as nine of the estimated LANDSQFT2 coefficients are negative and significant.
Table 6

Summary of regression results for 15 office markets

Variable

Average coefficienta

Number of positive coefficientsa

Number of negative coefficients

Average P value

Number of positive and significant coefficients

Number of negative and significant coefficients

EXREPL

0.21

15

0

0.02

14

0

EXRELQ

0.07

13

2

0.41

3

1

RELQ_REPL

0.23

13

2

0.21

9

0

AGE

−0.02

0

15

0.01

0

14

AGE2

0.00

15

0

0.08

11

0

SQFT

0.03

15

0

0.00

15

0

SQFT2

−0.06

0

15

0.00

0

15

LANDSQFT

0.00

15

0

0.10

11

0

LANDSQFT2

0.00

0

15

0.18

0

9

PARKING

0.00

6

9

0.34

1

3

FLOORS

0.11

15

0

0.04

13

0

COND_BA

−0.06

5

10

0.52

0

4

COND_AA

0.09

15

0

0.22

7

0

BUYOUT

0.12

12

3

0.19

6

0

YR2000

−0.03

9

6

0.50

0

1

YR2001

0.04

10

5

0.36

2

1

YR2002

0.12

12

3

0.27

8

0

YR2003

0.18

12

3

0.16

8

0

YR2004

0.32

13

2

0.13

13

0

YR2005

0.43

15

0

0.09

12

0

SNGL_LRISE

−0.14

0

15

0.15

0

9

MED_DENT

0.06

10

5

0.31

4

0

MIDRISE

−0.15

3

12

0.34

1

5

HIRISE

−0.57

2

13

0.19

0

9

OTHER

−0.13

1

14

0.24

0

7

CONST

13.27

15

0

0.00

15

0

SMDUMi (not reported)

      

Adjusted R-squared

0.89

     

aOut of a total of 15 separate MSA regressions. The dependent variable in each regression is the log of the sale price

EXREPL is a binary variable set equal to one if the transaction represents the purchase of a replacement property; EXRELQ is a binary variable set equal to one if transaction represents sale of a relinquished property; RELQ_REPL is a binary variable set equal to one if transaction represents both sale of a relinquished property and purchase of a replacement property. AGE is age of the building(s) in years; AGE2 is age squared; SQFT is total square footage of improvements in thousands; SQFT2 is SQFT squared divided by 1,000; LANDSQFT is the land square footage in thousands; LANDSQFT2 is LANDSQFT squared divided by 1,000; PARKING is the number of parking spaces; FLOORS is number of floors; COND_BA and COND_AA are binary variables set equal to one if the property is in below-average or above average condition, respectively. The omitted condition category is average. BUYOUT is a binary variable indicating the buyer is an out-of-state resident; YRi are dummy variables for each year of the sample (1999 is suppressed); SNGL_LRISE, MED_DENT, MIDRISE, and HIRISE are binary variables that, respectively, indicate the office property is, a single-tenant, low rise property, a medical–dental office property, a mid-rise property, or a high rise property. We classify the remaining seven properties types identified by CoStar as OTHER. The omitted office type subcategory is multi-tenant, low rise properties. CONST is a constant. SMDUMi are dummy variables indicating the CoStar delineated office submarket in which the property is located

The estimated coefficient on FLOORS is positive and statistically significant in 13 of the MSA regressions. The estimated coefficient on CONDITION_BA carries the expected negative sign, but is significant in just 4 of 15 MSA regressions. In contrast, the coefficient on CONDITION_AA is positive and significant in seven of 15 MSA models. The nominal sale prices of office properties in 2000 and 2001 very little from 1999, the base year. However, 2002 marks the beginning of a substantial price appreciation trend which continues into 2005.23 The inclusion of submarket dummy variables again significantly improves the fit of the models.

Turning to the variables of primary interest, Table 6 reveals that the estimated coefficient on EXREPL averages 0.21, almost three times the magnitude of the average for the apartment sample. Moreover, the EXREPL coefficient is positive and significant in 14 of the MSA-level regressions. This strongly suggests that buyers of replacement office properties are paying statistically significant price premiums. The coefficient estimates on EXRELQ are positive and significant in just three of the 15 regressions. Moreover, the coefficient estimates on EXRELQ are, on average, one-third the magnitude of the coefficient estimates on EXREPL. Finally, the estimated coefficient on RELQ_REPL is positive and statistically significant in nine of 15 MSA regressions.

Robustness Checks

Location Fixed Effects

As a robustness check for the importance of our location control variables, we estimated a pooled regression with fixed effects for our 15 MSAs. This alternative specification for both our apartment and office samples forces the estimated coefficients on all explanatory variables to be constant across the 15 MSAs. In the pooled apartment regression, the magnitude of the coefficients on the structural characteristics and year dummies, as well as their statistical significance, remain virtually unchanged. However, the estimated coefficients on EXREPL, EXRELQ, and RELQ_REPL are 0.11, 0.09, and 0.16, respectively, all of which are somewhat larger than the average coefficients from the MSA-level regression reported in Table 5. In the pooled office regression, the coefficient on EXREPL is 0.21 (t-statistic = 11.7). The estimated coefficient on EXRELQ is 0.11 (t-statistics = 4.6), while the coefficient on RELQ_REPL is 0.26 (with t-statistics of 9.0). These results are very similar to the average coefficients from the MSA-level regressions.

We also estimated pooled regression models with submarket fixed effects. The results for both the apartment and office models are similar to the MSA fixed-effects model, although the magnitude of the estimated coefficients on EXREPL, EXRELQ, and RELQ_REPL are reduced slightly.

Interactive Specifications

To examine the stability of the exchange coefficients over time, we estimated a model for both property type samples with MSA fixed effects in which we interact EXREPL, EXRELQ, and RELQ_REPL with the year dummies. The estimated exchange coefficients in both the apartment and office estimations are largely unaffected by the addition of the interaction variables. Moreover, none of the interaction variables are significantly different from zero in the apartment model, suggesting that exchange related price premiums have been constant over time. In the office model, the estimated coefficients on the variables created by interacting EXRELQ and the year dummies are not statistically significant. However, all EXREPL and year interaction variables are positive and significant, suggesting that the average price premiums paid by buyers of replacement office properties have increased over time.

Segmentation by Price Level

To examine whether our primary results vary by transaction price levels, we create two sub-samples from both the original apartment and office samples based on whether the selling price is less than or greater than $1 million (in nominal terms).24 When the apartment regression sample is restricted to properties with nominal sale prices less than or equal to than $1 million, the estimated coefficients on EXREPL, EXRELQ, and RELQ_REPL using our MSA fixed-effects model are 0.06, 0.06, and 0.10, respectively, and are all significant at the 1% level. However, when the apartment regression sample is restricted to properties with nominal sale prices greater than $1 million, the estimated coefficients on EXREPL, EXRELQ, and RELQ_REPL cannot be distinguished from zero. Thus, we conclude that exchange motivated apartment investors are paying significant price premiums, on average, only for relatively inexpensive properties.25

Using the office subsample of transactions with prices less than $1 million, we again obtain positive and significant coefficient estimates for EXREPL, EXRELQ, and RELQ_REPL. However, when the office regression sample is restricted to properties with nominal sale prices in excess of $1 million, the estimated coefficients on the three exchange variables remain positive and significant at the 5% level or better. Thus, unlike with the apartment sample, we observe exchange motivated price premiums for large, as well as small, properties.

Alternative Location Controls

The traditional approach to controlling for location in residential and commercial transaction price models is to include location variables that explicitly depict a spatial pattern (e.g., distance gradients from a central urban node or potential externality) and/or binary variables that indicate property location by submarket or other areal designation. The use of binary location variables assumes that parcels in a given locality (census tract, zip code, CoStar designated submarket, etc.) are subject to the same generalized localized conditions and therefore share the same situational characteristics and externalities. This assumption, however, may be limiting for at least two reasons. First, the size of the areal units may vary over space. Second, the processes which determine the intensity and impact of specific external effects can be highly variable even over short distances.

In contrast to areal units, distance gradients express an explicit spatial pattern. In fact, the use of a distance gradient forces the estimation of a uniform slope in all directions from the node. However, the effects of employment centers and externalities on property values may vary with direction from a given node. In addition, many spatial effects on property prices have an influence over only small geographic areas, suggesting the effects may diminish within relatively short distances. In short, distance-based measures are inadequate representations of composite external effects on property values.

Several recent papers have developed estimation techniques designed to address the shortcomings of submarket dummies and distance variables. For example, Clapp (2003) develops a semi-parametric method for valuing residential location and includes latitude and longitude as explanatory variables. Case et al. (2004) use a second order latitude–longitude expansion to control for location, as well as a number of demographic characteristics. Finally, Fik et al. (2003) use a variable interactive approach to model the log of sale price as a function of structural characteristics, discrete location dummies, and {x, y} coordinates interacted with structural characteristics and with the submarket location dummies. This interactive specification allows Fik et al. (2003) to effectively estimate separate price surfaces for identified submarkets, rather than constrain the estimated coefficients on structural characteristics to be constant across submarkets with the price surface shifted up or down by location dummies only. In addition, interacting submarket dummies with absolute location allows the researcher to capture discontinuities or structural shifts that occur as the price surface crosses submarket boundaries.

We estimate a modified version of Eqs. 7 and 8 in which we implement the interactive specification developed by Fik et al. (2003) in each of our 15 apartment (office) submarket regressions. Despite the more complete modeling of location, the estimated coefficients on EXRELQ, EXREPL, and RELQ_REPL are virtually unaffected. To conserve space we do not report these results.

Endogeneity of Exchange Variable

There is a concern that the relinquished exchange variable, EXRELQ, is not exogenous with respect to selling price. In particular, we are concerned that an omitted variable is correlated with the transaction price and with the probability that the property is being used as a relinquished property in a delayed exchange. This bias could produce an observed statistical relation between EXRELQ and price that is due to their mutual dependence on the omitted variable.

To address this concern, we perform a Durbin–Wu–Hausman (DWH) test for endogeneity (see Davidson and MacKinnon 1993, pp. 237–242). More specifically, we estimate our OLS model using two-stage least squares regressions, where EXRELQ is specified as an endogenous variable. The DWH test first estimates the endogenous variable as a function of all exogenous variables. In the second stage it regresses the dependent variable (in this case, natural log of price) on all variables and includes the residuals from the first stage estimation. The second stage coefficient estimates are not different from the coefficients estimated in the original model. Also, the DWH test shows that there is no significant bias in the OLS model estimates. Therefore, the results from the OLS model appear to be robust to concerns about the endogeneity of the exchange variables.

Economic Significance of Exchange Variables

The regression results presented in Tables 5 and 6 strongly suggest that many exchange motivated buyers pay significant price premiums to acquire replacement properties. In this section we examine whether the statistically significant price premiums reported above are economically significant.

To quantify the economic significance of our empirical findings, we transform our estimated regression coefficients into percentage (i.e., price) effects. More specifically, following Halvorsen and Palmquist (1980), we calculate percentage price changes for the estimated coefficients on EXREPL, EXRELQ, and RELQ_REPL that are statistically significant at the 10% level or greater in our MSA-level regressions using the following formula:
$${\text{Percentage price effect}} = 100 \times g = 100 \times \left\{ {\exp \left( x \right) - 1} \right\}{\text{.}}$$
g is the estimated effect on sale price of an exchange (EXREPL, EXRELQ, or RELQ_REPL) and x is the form of the exchange. These marginal price effects by property type and MSA are reported in Table 7.
Table 7

Marginal effects of exchange variables

Market

Obs

EXREPL

EXRELQ

RELQ_REPL

Apartments

Boston

357

n.s.

*

*

Chicago

1,532

11.0%

16.2%

29.3%

Denver

726

20.2%

16.7%

19.9%

Los Angeles

7,656

7.6%

5.0%

10.1%

New York City

3,915

n.s.

n.s.

n.s.

Oakland

1,172

7.4%

5.2%

11.9%

Phoenix

1,238

n.s.

n.s.

n.s.

Portland

486

7.3%

n.s.

10.5%

Riverside/San Bernardino

578

8.6%

11.8%

25.6%

Sacramento

367

16.8%

20.0%

19.7%

San Diego

1,944

7.3%

n.s.

7.8%

San Francisco

763

8.1%

6.9%

12.7%

San Jose

469

10.8%

n.s.

9.2%

Seattle

1,249

4.8%

n.s.

9.2%

Tucson

526

n.s.

n.s.

14.5%

Office properties

Chicago

706

25.3%

n.s.

n.s.

Dallas/Ft. Worth

302

34.7%

n.s.

n.s.

Denver

546

26.1%

n.s.

*

Ft. Lauderdale

422

31.8%

*

*

Las Vegas

297

25.7%

n.s.

29.7%

Los Angeles

1,426

22.4%

8.6%

29.4%

Oakland

357

n.s.

n.s.

30.7%

Phoenix

963

17.7%

*

n.s.

Riverside/San Bernardino

266

16.9%

n.s.

n.s.

Sacramento

287

21.4%

n.s.

42.5%

San Diego

571

27.7%

15.5%

30.2%

Seattle

627

14.8%

n.s.

*

Tampa

493

35.2%

n.s.

*

Tucson

248

29.6%

n.s.

n.s.

Washington

802

18.0%

n.s.

n.s.

n.s. indicates the estimated coefficient is not significant at the 10% level; therefore, the economic significance of the coefficient in not calculated. The asterisk indicates that the number of sale observations that involved the use of one of the three exchange types was less than 20. We calculate percentage price changes for the exchange coefficients that are statistically significant at the 10% level or greater in our MSA-level regressions using the following formula: Percentage price effect = 100 × g = 100 × {exp(x) − 1} where g is the estimated effect on sale price of an exchange and x is the form of the exchange

EXREPL is a binary variable set equal to one if transaction represents purchase of a replacement property; EXRELQ is a binary variable set equal to one if transaction represents sale of a relinquished property; RELQ_REPL is a binary variable set equal to one if transaction represents both sale of a relinquished property and purchase of a replacement property

The results for the apartment sample are displayed in the top section of the table. The percentage price premiums for replacement property exchanges range from 5% in Seattle to 20% in Denver. In apartment markets where the estimated coefficient on EXRELQ is statistically significant, the marginal price effects range from 5% in Los Angeles to 20% in Sacramento. The price premiums associated with an apartment sale being part of both a replacement and relinquished exchange range from 8% in San Diego to 29% in Chicago.

How do these empirical price premiums compare to the maximum tax deferral values estimated with Eq. 4 and reported in Table 1 and Table 8 in the Appendix? In our numerical analysis, we estimate that the maximum value of tax-deferral ranges from 5% to 10%, depending on other key variable assumptions. These results suggest that if buyers of replacement properties pay a price premium, relative to market value, in excess of 5% to 10%, the price premiums more than offset the value of tax deferral obtained with the use of a delayed exchange. This is especially true the smaller is the differed gain and the shorter is the expected holding period of the replacement property. As discussed previously, wealth destroying exchanges have been more commonly observed in transactions with sale prices less than $1 million.

The estimated marginal price premiums for our office sample are reported in the bottom panel of Table 7. Percentage price effects associated with replacement exchanges are statistically and economically significant in all but one of the studied markets (Oakland) and range from 15% in Seattle to 35% in Tampa. These empirical price premiums far exceed the range of maximum tax deferral benefits available from a commercial property exchange (see Table 8 in the Appendix), suggesting that the use of an exchange strategy has been wealth reducing for a significant number of office market investors.

Conclusion

This study examines the extent to which tax deferral benefits available to some market participants affect transaction prices. More specifically, we analyze the role tax-deferred exchanges play in the determination of reservation prices and sale prices in U.S. commercial real estate markets. Tax-deferred exchanges are transactions in which a taxpayer is able to defer payment of some, or all, of the federal income taxes associated with the disposition of real property by acquiring another property (or properties) of “like kind.” If a taxpayer is successful in completing an exchange, the realized tax liability is deferred until the replacement property is subsequently disposed of in a fully taxable sale.

However, taxpayers seeking to complete the second-leg of a delayed exchange face significant time constraints. As a result, the difficulty and cost of continuing to search for a replacement property can be extreme and the reservation prices of such tax motivated and time constrained buyers should be higher, all else equal, than other potential acquirers. Moreover, the exchanger may have compromised his or her bargaining position with potential sellers of replacement properties. In illiquid, highly segmented commercial real estate markets, the exchanging taxpayer may be required to pay a premium for the replacement property relative to its fair market value. If this price premium exceeds the present value of tax deferral, the use of a tax-deferred exchange will destroy, not enhance, wealth.

To evaluate the economics of tax-deferred exchanges, we first develop a numerical model for quantifying the present value of the tax-deferral benefits associated with Section 1031 exchanges. We then use hedonic regression and a unique dataset to examine whether exchange motivated investors pay more, on average, for replacement properties than other investors. We then compare the price premiums estimated by the regression model to the estimated value of tax deferral obtained from the numerical model.

In our apartment sample, we find strong evidence that replacement exchanges are associated with significant price premiums. Our robustness analysis reveals that these price premiums primarily involve properties valued at $1 million or less; properties with sale prices greater than $1 million are not associated with an exchange price premiums. Thus, we conclude that exchange motivated apartment investors are paying significant price premiums, on average, only for relatively inexpensive replacement properties.

We find strong and consistent evidence that replacement exchanges in office markets are associated with price premiums that are statistically and economically significant. This effect is robust across the 15 metropolitan markets in our office sample. The magnitude of the empirical price premiums varies from 15% to 35% across markets and far exceed estimates of the value of tax-deferral obtained via an exchange, especially if the exchanger expects to hold the replacement property for a relatively short period of time. Thus, it appears that many office market investors are paying price premiums that may offset, in whole or in part, the gain from the deferment of taxes. For these investors, the pursuit of tax avoidance comes at a steep price.

Footnotes
1

See, for example, Glower et al. (1998), Turnbull and Sirmans (1993), Harding et al. (2003a, b).

 
2

Exceptions include papers by Colwell and Munneke (2006), Holmes and Slade (2001), Shilling et al. (1990), Forgey et al. (1994), Hardin and Wolverton (1996), Lambson et al. (2004), and Hardin and Wolverton (1999).

 
3

A portion of the realized gain will be recognized in the tax year in which the exchange occurs to the extent that the value of the relinquished property exceeds the value of the replacement property.

 
4

With the development of IRS regulations concerning Section 1031, tax-deferred exchanges are also used to trade lines of business, such as such as television and radio stations, newspapers, distributorships, and franchises, including, among others, sports teams, beer distributorships, and professional service practices (McBurney 2004). Because a line of business includes multiple classes of assets—real, personal and intangible property—an exchange for each class needs to be completed (McBurney and Boshkov 2003; McBurney 2004).

 
5

Internal Revenue Code, Title 26, Section 1001(c) (2006).

 
6

The adjusted tax basis is equal to the original cost basis (including the value of the land), plus the cost of any capital improvements undertaken since acquisition of the property, minus cumulative depreciation.

 
7

Like kind means “similar in nature or character.” In fact, virtually any real estate is like-kind to any other real estate. However, real property is not like-kind to personal property. Therefore, for example, a warehouse cannot be exchanged for jewelry. In addition, foreign property cannot be exchanged for U.S. property. See IRC 2006, Title 26, Section 1031.

 
8

Tax deferral turns into permanent tax savings upon the death of the taxpayer because the basis of the property is “stepped-up” to its current fair market value. Thus, the taxpayer’s heirs can dispose of the property in a fully taxable sale and not have to pay taxes on gains deferred through the prior use of one or more Section 1031 exchanges.

 
9

The qualified intermediary is an independent agent who facilitates the exchange. The qualified intermediary takes an assignment of rights in the sale of the relinquished property and the purchase contract for the replacement property. In short, the QI buys and then resells the two properties for a fee.

 
10

More specifically, the taxpayer can (1) identify up to three properties of any value or (2) identify more than three properties so long as their combined values do not exceed 200 percent of the value of the relinquished property.

 
11

Starker vs. United States, 602 F. 2d 1341 (9th cir., 1979)

 
12

Since 2003, a percentage ownership interest as a tenant-in-common (TIC) is qualified property for the purposes of a Section 1031 exchange. The taxpayer, however, must be careful that the TIC has been structured to avoid its recharacterization by the IRS as a partnership for federal income tax purposes.

 
13

Vacation homes will only qualify if they have been rented out the majority of the year.

 
14

An additional disadvantage is that Section 1031 exchanges do not allow for the recognition of a loss for tax purposes. Thus, taxpayers will avoid using exchanges if they have not realized positive capital gain, which introduces a potential selection bias as to who participates in the exchange market.

 
15

Congressional legislation has repeatedly altered the period of time over which rental real estate may be depreciated. As of 2007, residential real property (e.g., apartments) may be depreciated over no less than 27 and 1/2 years. The cost recovery period for nonresidential real property (e.g., shopping centers, industrial warehouses, and office buildings) is 39 years.

 
16

From 1997 to May 6, 2003, the maximum capital gain tax rate was 20%.

 
18

Other exchange types include direct (simultaneous) exchanges, reverse exchanges, and exchanges involving tenancy-in-common properties.

 
19

The remaining seven property types are office/residential, office with street level retail, office park, office condominium, office planned unit development, and veterinary hospital/clinic.

 
20

Discussions of the impact of search costs on buyer behavior can be found in Haurin (1988), Miceli (1989), Forgey et al. (1996) and Arnold (1999).

 
21

In a “reverse” exchange, the replacement property is purchased before the sale of the relinquished property. Thus, the seller of the relinquished property is negotiating at a competitive disadvantage relative to other sellers. Our dataset does not include a large enough sample of reverse exchanges to examine empirically how the pricing of such exchanges varies from delayed exchanges and fully taxable sales. We therefore do not discuss reverse exchanges here.

 
22

The choice of functional form is extremely important in hedonic regression and little theory exists to guide the choice. Weirick and Ingram (1990) argue that the linear form has serious deficiencies from a market theory standpoint because the marginal contribution to value of variables such as square footage and lot size is not likely to be constant. The semi-log and log-linear models, in contrast, assume a nonlinear relation between sale price and the explanatory values.

 
23

One notable exception to this general trend in price appreciation after 2002 is Dallas/Forth Worth, which is the only office market in which no price appreciation is observed.

 
24

About 56% of the apartment properties have a sales price less than one million; for the office sample this percentage is 42%.

 
25

This result is robust to our treatment of location.

 

Acknowledgements

We thank CoStar Group, Inc. for providing the data used in this study. We also thank Lynn Fisher, Wayne Archer, Andy Naranjo, seminar participants at Syracuse University, and participants in the 2006 Maastricht-Cambridge-MIT International Real Estate Finance & Economics Symposium for the helpful comments and suggestions.

Copyright information

© Springer Science+Business Media, LLC 2007