The Empirical Literature

Almost all of the underlying data predate the 2017 Act, when exchanges of vehicles and equipment were still common. This chapter organizes the literature by question, identifies who funded what, and marks the questions the studies leave open.

63.1 Transaction-level studies of exchange incidence

Two kinds of evidence. How often exchanges occur, and in what property, can be measured in two ways: from tax returns, where every exchange should be reported on Form 8824, and from commercial transaction databases that flag sales involving an exchange. Each source has blind spots, discussed in 63.4, and the two answer slightly different questions.

Tax return aggregates. Economists in Treasury's Office of Tax Analysis reported that total deferred gain rose from $25.7 billion in 1998 to $102.8 billion in 2006, fell to $33.8 billion in 2009 during the real estate collapse, and recovered to about $57 billion by 2012.¹ For 2007, which an Office of Tax Analysis paper describes as the most recent year with consistent data for every type of taxpayer, the total divides among C corporations ($25.8 billion), partnerships ($35.6 billion), and individuals ($21.2 billion). The groups used exchanges differently. Firms in the vehicle industry accounted for 51.8 percent of C corporation deferred gain, while real estate accounted for 88.2 percent of partnership deferred gain and 92.8 percent of individual deferred gain.² Adding the real estate categories across the three groups gives $54.3 billion of $82.6 billion, about two-thirds; Ling and Petrova used the same two-thirds share as an upper bound for real estate in their 2015 study.³

By 2012 the balance had moved toward corporations. C corporations deferred $33.5 billion, 69.7 percent of it in firms classified in vehicles and transportation, yet only 17,249 C corporations reported exchanges, about 1 percent of all C corporations. Partnerships deferred $15.6 billion, 51.4 percent of it in real estate firms, and only 10,551 partnerships reported exchanges.⁴

Industry is not asset type. Those figures classify gain by the filer's industry, not by the property exchanged. When the Treasury economists read the property descriptions on 2009 electronically filed partnership returns, real estate accounted for 76.2 percent of deferred gain, but by number of transactions farm equipment (33 percent) and cars and trucks (30 percent) outnumbered real estate (18 percent).⁵ Individuals show the same pattern. In 2011 they reported 213,305 exchange transactions and $3.3 billion of deferred gain. Real estate of various kinds produced about 80 percent of the dollars, while vehicle trade-ins made up 123,038 of the transactions and, in the aggregate, a deferred loss of $239 million. The loss figure is a reminder that section 1031 was mandatory and applied to losses as well as gains: a vehicle traded in for less than its adjusted basis produced a loss the exchange rule did not allow.⁶

Commercial transaction databases. Ling and Petrova's 2015 study, funded by a real estate industry coalition, used the CoStar database of 1,609,711 confirmed commercial sales from 1997 through 2014, with total volume of $4.8 trillion. Sales flagged as involving an exchange by the buyer, the seller, or both numbered 81,104, or 5 percent of sales and 6 percent of dollar volume. The authors cautioned that the figures understate incidence, because CoStar flags a sale only when it can verify the exchange and because its coverage grew rapidly during 2007 through 2010, largely through purchases of other data companies whose tracking of exchanges is unknown.⁷ Incidence within particular segments can be far higher. In their earlier peer-reviewed study, as the authors later summarized it, 32 percent of apartment sales and 20 percent of office sales in their 1999 through 2005 data involved an exchange.⁸ Holmes and Slade built a data set documenting nondirect exchanges in the Phoenix apartment market.⁹ For farmland, Williamson, Brady, and Durst presented the first national analysis of federal tax data on farmland exchanges, covering 1999 through 2005, and Dillard and coauthors studied 3,580 farm real estate sales in 55 Indiana counties from 2003 through 2006.¹⁰

What the incidence evidence shows. Exchanges are a small minority of commercial sales nationally, a substantial share in some property types and markets, and highly sensitive to the real estate cycle. Before 2018, vehicle and equipment trade-ins dominated the count of exchanges, while real estate and corporate vehicle fleets dominated the dollars. The published tax series examined here ends in 2012, and this chapter did not locate a comparable published tabulation for the years after the 2017 Act, when only real property remains eligible.

63.2 Evidence on holding periods and subsequent taxable sales

The two claims. Defenders of section 1031 make two empirical claims about time. Exchanges shorten holding periods, which improves liquidity in illiquid markets, and exchange property is eventually sold in a taxable sale, so the provision defers tax rather than forgiving it. The main evidence for both comes from one research program.

Holding periods. Ling and Petrova's 2015 study examined 336,572 properties that sold at least twice in the CoStar data between 1997 and 2014. Properties disposed of through an exchange had been held an average of 3.49 years, against 3.98 years for other dispositions. In a sample matched on observable characteristics the averages were 3.38 and 3.66 years. Both differences were statistically significant.¹¹ The pattern was not uniform: of the ten states reported separately, exchange holding periods were longer in Arizona, Colorado, and Florida.¹² The authors' 2020 peer-reviewed article likewise associates exchanges with shorter holding periods, along with increased investment and reduced leverage.¹³

Subsequent taxable sales. In the same repeat-sales sample, about 4 percent of the properties had been acquired in an exchange, and 11 percent of those were later disposed of through another exchange. The authors report that in 88 percent of cases the replacement property was disposed of in a taxable sale. They also found that rollover exchanges by the same investors declined after 2006 and were virtually zero after 2009, and they estimated that the taxes paid when an exchange is followed by a taxable sale average 19 percent more than when an ordinary sale is followed by another ordinary sale.¹⁴

What the 88 percent measures. The figure is conditional on a second observed sale within an eighteen-year window. A replacement property still held at the end of 2014, or held until the owner's death, cannot appear as a disposition in a repeat-sales sample, and the sample's short average holding periods reflect the same construction. The authors themselves note that their holding period difference is likely understated because it rests on repeat sales within a limited window.¹⁵ The finding shows how exchange property that is resold within a few years leaves the chain. It does not estimate the share of exchange-deferred gain that is ever taxed, and it should not be read as answering the concern about basis step-up at death discussed in 62.2.

Survey expectations. A study by Ernst & Young, prepared on behalf of an industry coalition, surveyed members of nine trade associations; 221 businesses that conducted, brokered, or advised on exchanges responded. Weighting responses by the value of property exchanged, respondents expected repeal to lengthen the average holding period of real estate from 8 years 6 months to 11 years 7 months, and of personal property from 2 years 4 months to 3 years 4 months.¹⁶ These are expectations of interested parties, used as inputs to an economic model, not observed behavior. They are also not comparable to the repeat-sales averages, which describe only properties sold twice within a window.

The contrary estimate. Johnson's argument that most gain deferred in exchanges is never taxed rested on estimates that between 20 and 46 percent of unrecognized gain is eventually taxed, drawn from research on capital gains and wealth generally, including estimates that most wealth is held until death.¹⁷ That research concerns capital assets as a whole. Neither side of the debate has published exchange-specific data following replacement properties to their final disposition.

63.3 Effects on leverage, capital improvement, and price

Leverage and the size of the investment. Ling and Petrova found that replacement properties acquired in exchanges carried median loan-to-value ratios of 63 to 64 percent, compared with 70 percent for properties acquired without an exchange.¹⁸ They also found that an investor acquiring replacement property through an exchange invested about $305,000 more, roughly 33 percent of value, than the same investor did when buying after a taxable sale, and that exchange properties tended to be larger, newer, and less vacant.¹⁹ Their 2020 article reports the same direction of effect on investment and leverage.²⁰ The finding fits the mechanics. An exchanger keeps the deferred tax in the deal as equity, and the boot and liability rules in Chapters 33 and 34 push an exchanger who wants full deferral toward property at least equal in value to what was sold. The comparison group, however, is not randomly assigned, a point taken up in 63.4.

Capital improvements. The same study reported that capital expenditures on replacement properties were higher by about $0.27 to $0.40 per square foot, and building improvements by $0.18 to $0.24. The authors' own footnote states that the difference in capital expenditures is not statistically significant at conventional levels.²¹ Statements that exchanges drive improvement spending rest on this result and should carry that qualification.

Price premiums paid by exchange buyers. The most consistent finding in the peer-reviewed literature is that exchange buyers pay more. Holmes and Slade found that exchange participants in the Phoenix apartment market paid an economically significant premium to acquire replacement property, consistent with theories of price pressure and tax capitalization.²² Ling and Petrova's 2008 study found that exchange buyers paid significantly more than other investors for apartment and office properties, and that the average premium generally exceeded the value of the tax deferral itself. They attributed the premium to the exchanger's weakened bargaining position under the time limits of a deferred exchange in illiquid, segmented markets.²³ For Indiana farmland, Dillard and coauthors estimated a premium of 1.32 percent for properties acquired in exchanges.²⁴ Colwell and Dehring's model of farmland on the urban periphery explains the observed pattern from the seller's side: owners who exchange agricultural land tend to receive replacement properties of lower market value.²⁵ For practice, the premium studies give empirical content to the timing rules in Chapters 25 and 26. The 45-day and 180-day deadlines have a price, and for some buyers it exceeds the tax deferred.

Modeled effects of repeal. Everything else in this area is a model rather than a measurement. Barker, Ling, and Petrova estimate the incremental present value of an exchange at under 1 percent to about 8 percent of property value and consider whether that value would be capitalized into market prices if exchanges ended. Treating it as a price effect requires assuming that the marginal buyer is a taxable investor completing an exchange and that no offsetting market effects occur.²⁶ Ling and Petrova's 2015 typical project model concluded that prices of commercial properties in moderately taxed markets would need to fall 8 to 12 percent, or rents rise 8 to 13 percent over the longer run, to offset repeal, with larger effects in high-tax states.²⁷ The Ernst & Young study estimated that repeal paired with a revenue-neutral cut in the corporate rate would reduce long-run GDP by $8.1 billion a year (0.04 percent), investment by $7.0 billion, and labor income by $1.4 billion, with a larger GDP decline ($13.1 billion) if the revenue financed spending and a smaller one ($6.1 billion) if it financed rate cuts for both corporations and pass-through businesses.²⁸ These are the outputs of assumptions about behavior, some of them drawn from the survey described in 63.2.

63.4 Methodological limits and data constraints

Finding the exchanges. Transaction databases rely on flags the vendor can verify, and they miss exchanges it cannot. Coverage also changes over time, which makes trends hard to read.²⁹ Tax data capture every reported exchange, but in the published tabulations the type of property is often unknown. Many businesses answer the property description on their capital gains forms with "Available upon request," so corporate and partnership figures are classified by the filer's industry, and the individual figures come from periodic samples of capital asset sales rather than from every return.³⁰ The 2014 Treasury paper described 2007 as the most recent year with consistent data for all types of taxpayers.³¹

Truncated windows. Repeat-sales designs observe only properties sold twice within the study period. Owners who hold longer, or until death, are invisible. The truncation shortens measured holding periods and conditions the taxable sale percentage on a resale, as 63.2 explains.

Selection. Exchange buyers and their properties differ from other buyers and properties; in the 2015 study, exchange properties were larger, newer, and less vacant.³² Matching on observed characteristics narrows the difference but cannot remove differences that are not observed, such as the sophistication of the buyer or qualities of a building that the data do not capture. The price premium studies face a related difficulty in separating deadline pressure from unmeasured quality.

Place and time. The peer-reviewed price studies examine a single metropolitan apartment market, farmland in 55 Indiana counties during a rising market, and national commercial samples ending no later than 2014. Nearly all the data predate the 2017 Act, so they describe a regime in which vehicles and equipment were eligible and tax rates were different.

Models and measurements. Price, rent, and GDP effects of repeal come from models whose conclusions follow from their assumptions: the identity of the marginal buyer, the discount rate, expected holding periods, and the use of any revenue raised. Some of those inputs came from surveys of industry participants.

Sponsorship. The 2015 Ling and Petrova study acknowledges financial support from the Real Estate Like-Kind Exchange Coalition, the 2017 Barker, Ling, and Petrova working paper acknowledges partial support from the same coalition, and the Ernst & Young study was prepared on behalf of the Section 1031 Like-Kind Exchange Coalition.³³ The case for repeal has its own advocacy framing: Johnson's article appeared as a proposal of the Shelf Project, a collaboration developing revenue-raising options for Congress.³⁴ Sponsorship does not make a finding wrong. It is a reason to separate what a study measured from what its authors concluded, and to give more weight to results that passed peer review: the 2001 articles by Holmes and Slade and by Colwell and Dehring, the 2008 Ling and Petrova article, the 2010 Williamson, Brady, and Durst article, the 2013 Dillard article, and the 2020 articles by Ling and Petrova and by Barker, Ling, and Petrova.

63.5 What the literature has not established

Whether deferred gain is ever taxed. No study located for this chapter follows exchange replacement property to its final disposition, including disposition at death. The 88 percent taxable sale figure is conditional on a resale within the study window, and the estimates that most unrecognized gain escapes tax are drawn from capital assets generally. The present-value cost of section 1031 turns on this question (62.2), and it remains unmeasured.

The revenue effect of repeal. The Joint Committee's revenue estimates incorporate behavior, but the published tables for repeal and restriction proposals report figures without the behavioral assumptions behind them.³⁵ The industry studies' revenue conclusions rest on static present values adjusted qualitatively for behavior, or on revenue feedback from modeled GDP effects; the Ernst & Young study, for example, derives a decline of about $1.6 billion a year in federal revenue from its GDP estimate.³⁶ An independent estimate of what repeal would raise, with its assumptions disclosed, has not been published.

Who captures the benefit. The premium studies show that exchange buyers pay more, which suggests that part of the value of deferral passes to sellers of replacement property. Whether tenants share in it through lower rents, as the 2015 study argues, has been modeled but not measured.

What happened after 2017. The 2017 Act ended exchanges of personal property, a change that could serve as a natural experiment for equipment, vehicle, and collectibles markets. This chapter did not locate a published study of its effects. The pre-Act evidence that expensing substitutes for exchange deferral, including the Hertz example in 62.3, suggests that any effect depends heavily on the depreciation rules in force.

Distribution. Neither estimator publishes a distribution of the like-kind item by income class. The Joint Committee's distributional table covers selected individual tax expenditures, and the like-kind item is not among them.³⁷ Arguments about the fairness of section 1031, which feature prominently in the policy debate, therefore proceed without official distributional data.³⁸

Compliance. Section 1031 depends on basis carried across a chain of properties, sometimes for decades. Johnson questioned whether basis rules are enforced well enough to preserve deferred gain.³⁹ This chapter did not locate a published study of the accuracy of basis reporting for exchange property.