Investment and Capital Improvements in Commercial Real Estate: The Case of REITs

Abstract

Land, capital, and labor constitute essential components of economic production. This axiom holds particular significance in commercial real estate, where an incorporation of prime land, meticulously executed capital improvements in building structures, and adept professionals significantly influence performance and returns. This study illuminates the critical role of capital investment intensity in synergy with strategic location choices (land) and proficient management. Drawing upon a dataset encompassing U.S. equity Real Estate Investment Trusts (REITs) spanning 1995–2022, the analysis uncovers a compelling correlation between a higher capital allocation towards property improvements and augmented market valuations. This association maintains its potency when utilizing an instrumental variable approach to mitigate potential endogeneity caveats. It remains robust even when the sample is streamlined to focus on a REIT's core property type. Further, by estimating a firm-level production function that accounts for endogenous input choices, the study reveals that initial and continuing investments in building capital constitute nearly half of a REIT's output. This underscores the pivotal role of initial and recurrent capital investments in building structures in generating commercial property revenue and overall productivity. The findings emphasize the indispensability of capital improvements that foster optimal property utilization and intensity in driving performance and returns.

Appendices

Appendix 1: Discussion on Land Value and Building Investment

REITs' acquisitions on properties are accounted for in accordance with the authoritative accounting guidance on acquisitions and business combinations. Generally, REITs' methodology of allocating the cost of acquisitions to assets acquired and liabilities assumed is based on estimated fair values, replacement cost, and appraised values. When REITs acquire operating real estate properties, the purchase price is allocated to land and buildings, intangibles such as in-place leases, and to current assets and liabilities acquired, if any.

The building value [S&P Global KeyField: 132108, BUILDINGS_EQUIP] includes all buildings, equipment, and capital improvements gross of real estate depreciation on operating properties. The properties under the development or construction in progress are also excluded in the construction of building value.

The land value [S&P Global KeyField: 132107, LAND] is defined as the value of all land underlying operational properties. If the land is currently vacant or in the preparation for the development, the property is tagged under the Development Pipeline. It is reported separately by the company and does not include the value of the land. The properties under the development or construction in progress are also excluded in land value.

The two variables are primarily constructed based on REITs' company fillings (10 K, Schedule III —Consolidated Real Estate and Accumulated Depreciation). Schedule III generally narrowed down REITs' property value into two: Land and Building & Improvements, under two categories: Initial Cost and Gross Carrying Amount. The S&P Global Market Intelligence defines the Initial Cost as the historical cost basis assigned to the property by the acquirer. In the financial context, this is the value currently on the financial statements, which may differ from the cost reported at the time of purchase. The Gross Carrying Amount on a specific year-end date represents the latest value of the property as reported from the company filings. Moreover, S&P Global tracks the property's Net Book Value to identify how the property value changes over the years in terms of tax or changes on the physical aspect. S&P Global defines this item as the carrying value of the property, reflecting the historical cost of the property and improvements, and net of accumulated depreciation.

Some typical examples of the land value and building value in the sample in 2017 are as follows:

Institution Name	Ticker	Property Type	Building or Land	Value ($000)	Decomposition (Item Lines)	Value ($000)
Acadia Realty Trust	AKR	Shopping Center	Building	2,615,303	Buildings and improvements	2,406,488
					Tenant improvements	131,850
					Properties under capital lease	76,965
			Land	658,835	Land	658,835
Alexander & Baldwin, Inc	ALEX	Diversified	Building	546,300	Buildings	471,600
					Machinery and equipment	74,700
			Land	613,300	Land	613,300
Alexander's, Inc	ALX	Diversified	Building	988,846	Buildings and leasehold improvements	988,846
			Land	44,971	Land	44,971
Alexandria Real Estate Equities, Inc	ARE	Office	Building	9,780,743	Buildings and building improvements	9,000,626
					Other improvements	780,117
			Land	1,312,072	Land (related to rental properties)	1,312,072
Americold Realty Trust	COLD	Industrial	Building	2,448,528	Buildings and improvements	16,827
					Capitalized leases: Machinery and equipment	59,389
					Property, plant, and equipment: Buildings and improvements	1,819,635
					Property, plant, and equipment: Machinery and equipment	552,677
			Land	389,443	Land	389,443
Apple Hospitality REIT, Inc	APLE	Hotel	Building	4,791,663	Buildings and improvements	4,362,929
					Furniture, fixtures and equipment	428,734
			Land	720,465	Land	720,465
Boston Properties, Inc	BXP	Office	Building	14,541,700	Buildings and improvements	12,284,164
					Tenant improvements	2,219,608
					Furniture, fixtures and equipment	37,928
			Land	5,080,679	Land	5,080,679
Brixmor Property Group Inc	BRX	Shopping Center	Building	8,063,871	Buildings and tenant improvements	8,145,085
					Less Construction in progress	81,214
			Land	1,984,309	Land	1,984,309
CareTrust REIT, Inc	CTRE	Health care	Building	1,195,334	Buildings and improvements	1,114,605
					Integral equipment, furniture and fixtures	80,729
			Land	151,879	Land	151,879
Cedar Realty Trust, Inc	CDR	Shopping Center	Building	1,217,966	Buildings and tenant improvements	1,230,362
					Less Construction in progress	12,396
			Land	304,237	Land	304,237
CIM Commercial Trust Corporation	CMCT	Office	Building	980,088	Buildings and improvements	847,849
					Furniture, fixtures, and equipment	3,363
					Tenant improvements	128,876
			Land	239,530	Land	221,785
					Land improvements	17,745
Digital Realty Trust, Inc	DLR	Specialty	Building	14,368,761	Buildings and improvements	15,215,405
					Tenant improvements	553,040
					Less Total development construction in progress	1,399,684
			Land	783,935	Land	1,136,341
					Less Land inventory	352,406
Education Realty Trust, Inc	EDR	Specialty	Building	2,554,926	Furniture, fixtures and equipment	108,754
					Leasehold improvements	74
					Buildings and improvements	2,446,098
			Land	316,307	Land	247,259
					Land improvements	69,048
Equity LifeStyle Properties, Inc	ELS	Manufactured Home	Building	572,517	Buildings and other depreciable property	649,217
					Less In-place lease intangible assets	76,700
			Land	4,266,596	Land	1,221,375
					Land improvements	3,045,221
Forest City Realty Trust, Inc	FCE-A	Diversified	Building	6,518,490	Total real estate: Buildings and improvements	7,009,272
					Less Construction and development: Buildings and improvements	490,782
			Land	636,117	Total real estate: Land and improvements	771,183
					Less Construction and development: Land and improvements	77,770
					Less Land inventory	57,296
Kimco Realty Corporation	KIM	Shopping Center	Building	8,500,290	Building and improvements	9,231,644
					Less In-place leases and tenant relationships	577,870
					Less Above-market leases	153,484
			Land	3,019,284	Land	3,019,284
Life Storage, Inc	LSI	Self-Storage	Building	3,520,399	Building, equipment, and construction in progress	3,534,782
					Less Construction in progress	14,383
			Land	786,628	Land	786,628
Mack-Cali Realty Corporation	LI	Diversified	Building	3,668,939	Total before real estate depreciation and amortization	5,102,844
					Less Land and leasehold interests	414,502
					Less Land held for development	483,432
					Less Development and construction in progress	535,971
			Land	414,502	Land and leasehold interests	414,502
Prologis, Inc	PLD	Industrial	Building	16,849,349	Buildings and improvements	16,849,349
			Land	5,735,978	Improved land	5,735,978
Regency Centers Corporation	REG	Shopping Center	Building	5,787,114	Buildings	4,999,378
					Building and tenant improvements	787,880
					Less Land held for future development (building & improvements)	144
			Land	4,729,111	Land	4,235,032
					Land improvements	556,140
					Less Land held for future development	62,061
Simon Property Group, Inc	SPG	Regional Mall	Building	32,254,456	Buildings and improvements	32,379,190
					Furniture, fixtures and equipment	378,958
					Less Construction in progress	503,692
			Land	3,635,316	Land	3,635,316
UDR, Inc	UDR	Multifamily	Building	7,614,568	Building, improvements, and furniture, fixtures and equipment	7,614,568
			Land	1,970,148	Land	1,780,229
					Land improvements	189,919
Whitestone REIT	WSR	Shopping Center	Building	789,553	Building and improvements	825,359
					Less Total—Development portfolio: Building and improvements	35,361
					Less Total—Land held for development: Building and improvements	445
			Land	298,768	Land	324,095
					Less Total—Development portfolio: Land	8,731
					Less Total—Land held for development: Land	16,596

Appendix 2: Production Function Estimates

The article also estimates a firm-level production function correcting for endogenous input choices and then assesses the contributions of land, building, and labor to the overall production of REITs. The output for REIT $i$ at period $t$ (${Y}_{i,t}$) is model as a Cobb–Douglas production function whose inputs are land capital (${LK}_{i,t}$), building capital ($B{K}_{i,t}$), labor (${L}_{i,t}$), and an unobservable term, ${\epsilon }_{i,t}$. The following parametric log-linear form of the firm-level production function is the starting point for the analysis of REIT production.

$${y}_{i,t}={\beta }_{lk}{lk}_{i,t}+{\beta }_{bk}{bk}_{i,t}+{\beta }_{l}{l}_{i,t}+{\epsilon }_{i,t}$$

(4)

Lowercase variables are used to denote logarithms of output and inputs of REIT $i$ at year $t$, respectively. The term ${\epsilon }_{i,t}$ represents the information that REIT managers possess, which may be used for input selection. ${\beta }_{lk}$, ${\beta }_{bk}$, and ${\beta }_{l}$, measure contributions to output from land, building, and labor, respectively.

The major concern on the estimation of the above firm-level production function is that these inputs are typically the choice variables of the firm. These input choices are made to maximize a firm's profit or shareholders' wealth. The productivity or efficiency levels may be known to firm managers when they decide their input utilizations but are unobservable to researchers. For instance, when there are negative productivity shocks, profit-maximizing firms may cut part of their inputs (e.g., fewer capital expenditures or tenant improvements in the case of real estate). That is, a firm may possess information on ${\epsilon }_{i,t}$ when deciding its inputs. Due to the correlation between production inputs and the error term, the production parameter estimates from a traditional ordinary least squares regression (OLS) are likely to be biased (Griliches & Mareisse, 1998).

The dynamic panel data (DPD) approach, which is stemmed from Arellano and Bond (1991), Arellano and Bover (1995), and Blundell and Bond (1998, 2000), is widely adopted to tackle the aforementioned simultaneity issue in recent empirically works on production function parameter estimation. The main advantage of using the DPD approach in REIT production function estimation is that the DPD allows a time-invariant fixed-effect in the evolution of unobserved productivity. REITs have many characteristics (e.g., the property locations, land sizes, zoning regulations, and even building structures, to some extent) that are time-invariant, whereas other aspects of REIT productivity (e.g., management, operation, and marketing) evolve over time. Hence, the DPD framework fits the REIT setting and provides internally consistent sources of variation that can be used to identify the parameters. In the DPD approach for production function estimation, the last term, ${\epsilon }_{i,t}$, in Eq. (4) can also be further decomposed into four components:

$${\epsilon }_{i,t}={a}_{i}+{\gamma }_{t}+{\omega }_{i,t}+{\eta }_{i,t}$$

(5)

where ${a}_{i}$ is a time-invariant firm fixed effect and ${\gamma }_{t}$ is a time-varying productivity shock. These factors are likely to be related to the observed inputs. ${\omega }_{i,t}$ is an unobserved productivity term, which might be correlated with the observed inputs. It evolves as an autoregressive process, ${\omega }_{i,t}=\rho {\omega }_{i,t-1}+{\xi }_{i,t}$, where $\left|\rho \right|<1$ and ${\xi }_{i,t}$ is a pure stochastic component. The innovation on unobserved productivity,${\omega }_{i,t}$, is assumed to be uncorrelated with the observed inputs. The last term, ${\eta }_{i,t}$, reflects a productivity shock, which might be correlated with the observed inputs and might evolve as a moving average process. The impact in time for ${\eta }_{i,t}$ might also last for a long period.

Even if the term ${\epsilon }_{i,t}$ in Eqs. (4) and (5) consists of a firm fixed effect and a component of the evolving productivity, it is still likely to be correlated with the observed inputs. By solving for ${\omega }_{i,t-1}$ and substituting it into Eq. (4), a dynamic form is generated as follows:

$${y}_{i,t}=\rho {y}_{i,t-1}+{\beta }_{lk}{lk}_{i,t}-\rho {\beta }_{lk}{lk}_{i,t-1}+{\beta }_{bk}{bk}_{i,t}-\rho {\beta }_{bk}b{k}_{i,t-1}+{\beta }_{l}{l}_{i,t}-\rho {\beta }_{l}{l}_{i,t-1}+{a}_{i}-\rho {a}_{i}+{\gamma }_{t}-\rho {\gamma }_{t-1}+{\omega }_{i,t}+{\epsilon }_{i,t}$$

(6)

After renaming the respective coefficients and grouping the error components, an unrestricted dynamic representation of Eq. (6) that can be estimated is obtained, as follows:

$${y}_{i,t}={\delta }_{1}{y}_{i,t-1}+{\delta }_{2}{lk}_{i,t}+{\delta }_{3}{lk}_{i,t-1}+{\delta }_{4}{bk}_{i,t}+{\delta }_{5}{bk}_{i,t-1}+{\delta }_{6}{l}_{i,t}+{\delta }_{7}{l}_{i,t-1}+{{a}_{i}}^{*}+{{\gamma }_{t}}^{*}+{{\omega }_{i,t}}^{*}$$

(7)

The common factor restrictions are ${\delta }_{3}=-{\delta }_{1}*{\delta }_{2}$, ${\delta }_{5}=-{\delta }_{1}*{\delta }_{4}$ and ${\delta }_{7}=-{\delta }_{1}*{\delta }_{6}$, with ${{a}_{i}}^{*}={a}_{i}(1-\rho )$, ${{\gamma }_{t}}^{*}={\gamma }_{t}(1-\rho )$ and ${{\omega }_{i,t}}^{*}={\omega }_{i,t}+{\epsilon }_{i,t}$. Assuming all the common factor restrictions hold, consistent parameters can be obtained in the firm fixed-effect (FE) model if $E\left({\omega }_{i,t}{x}_{i,t}\right)=0$ and $E\left({\epsilon }_{i,t}{x}_{i,t}\right)=0$.

Estimation of Eq. (7) requires valid instruments that are correlated with the endogenous regressors but uncorrelated with the error term. According to Wooldridge (2009), the lags of capital and labor inputs could be a possible set of instruments for the output in time $t-1$ and the changes in capital and labor inputs in time $t-1$. Theoretically, the optimal choice of inputs across firms varies. Hence, it is justified to use lagged inputs as identifying instruments for the values of current inputs in firm-level production function estimations.

The DPD approach provides consistent parameters under less restrictive assumptions than the FE approach. The system generalized method of moments (GMM) estimators are designed for dynamic "small-T, large-N" panels that may contain fixed effect and idiosyncratic errors that are heteroskedastic and correlated within but not across firms. In GMM, a linear functional relationship between the variable of interest and the regressors is assumed. It also allows the dependent variable to have a dynamic structure depending on its own past realizations. Thus, given the high degree of persistence in unobserved productivity, this article adopts a system GMM framework that simultaneously estimates the production function using both levels and differences specifications where appropriate lags of the levels and differenced variables can be used as instruments. Lagged levels are used as instruments for the differences equation, while lagged differences are used as instruments for the levels equation. Besides, the system GMM estimation provides an over-identification test for the validity of instruments.

Appendix 3

Fig. 3

Number of REITs in Each Property Type. This figure presents the numbers of REITs in in each property type in our sample