Market outcomes are contrasted for uninformed investors with and without broker representation to evaluate the information role, and for parties opposing the uninformed (with and without representation) to evaluate the bargaining role. The setting is a sample of 17,000 office building transactions in more than 100 US markets, and the identification strategy for uninformed investors is based on the nonlocal clientele effect. Nonlocal investors buy high and sell low, paying significant premiums in acquisitions and accepting discounted offers in divestitures. Employing a commercial broker is found to have virtually zero impact on this disparity. Moreover, when the opposite party has broker representation, the degree of overpayment by nonlocal buyers is even higher. These findings are at odds with the conventional notion that brokers possess a high degree of specialized market knowledge which can be used to offset informational disadvantages suffered by their clients.
Real estate brokers are widely assumed to possess specialized market knowledge which can be commissioned to offset disadvantages faced by uninformed market participants. In this study, we empirically dissect the efficacy of brokerage intermediation for uninformed investors. The setting is commercial real estate (CRE), where assets are held for investment purpose rather than consumption. The identification strategy for an uninformed investor is based on geographic origin. Nonlocal investor clienteles persistently underperform in CRE markets: overpaying for purchases and selling assets at significant discounts to local players. Whether hiring a CRE broker has any impact on the outcome for nonlocal investors is the central focus of this study.
Nonlocal investors are buyers or sellers whose address is located in a different geographic market than the investment property. CoStar data includes around 17,000 transactions of office property for which investor address and investor type are available, covering 138 US markets during the period 1991 to 2013. Nonlocal investors represent a nontrivial component of the office market with roughly 30 percent of buyers and sellers classified as such. Outcomes for nonlocal investors have heightened relevance today as domestic funds pursue geographic diversification in their CRE portfolios, and foreign capital continues to flock to US markets responding to perceptions of lower risk and long-term capital preservation relative to alternative investment opportunities. Yet, evidence for underperformance by nonlocal investors suggests practical limitations to the theoretically-optimal diversification strategy.
Perhaps brokerage intermediation can resolve these issues if a CRE broker is able to provide their clients with specialized market information and bargaining expertise. However, brokerage compensation contracts may fail to align incentives with that of obtaining the best possible price for their clients. Existing literature focuses heavily on the search role of brokers, yet advancements in technology have increased informational transparency for market participants. Lacking from the literature is the evaluation of brokers who represent informed versus uninformed investors—potentially illuminating the informational role of brokers. Correspondingly, the bargaining role is revealed in market outcomes when the opposite party to a broker is informed versus uninformed.
To conduct the analysis, a propensity-score matching procedure is executed for each comparison set in order to minimize issues with sample heterogeneity and selection bias in CRE data. Four matched sample comparison sets are constructed on each side of the transaction. For purchases, prices paid by local buyers with broker representation are compared to those paid by local buyers without representation—to identify baseline brokerage intermediation effects for informed buyers. Nonlocal buyers are compared to each other, with and without representation, to estimate the impact of brokerage involvement for uninformed buyers. Transactions of unrepresented nonlocal buyers are compared to those of local buyers, who are also unrepresented, to quantify the nonlocal clientele effect that exists in the absence of brokers. Finally, outcomes for nonlocal buyers who have broker representation are compared to outcomes for represented local buyers to determine whether the nonlocal clientele effect differs when brokers are involved. Similarly for asset sales, the symmetric sequence of four-dimension comparison is provided.
The next section provides background on the nonlocal clientele effect and brokerage intermediation in real estate markets. The third section explains the data used in this study and the empirical methodology. The fourth section discusses the empirical findings, and the final section provides concluding remarks.
A) Nonlocal Clienteles in Real Estate Markets
Nonlocal investors in real estate markets suffer from informational disadvantages that distort their ability to obtain accurate asset valuations. Impacts from nonlocal investment can be severe enough to even produce localized asset price bubbles, as in the case of Japanese buyers converging on the Waialae-Kahala, Hawaii housing market during the late-1980s [
Studies for nonlocal investor clienteles include forest parcel sales in Adirondack Park [
Apart from residential, evidence for premiums paid by nonlocal investors on CRE assets includes apartment complexes in Phoenix [
The nonlocal clientele effect in CRE presents an ideal context within which to investigate the ability of brokers to resolve information problems in heterogeneous asset markets. As an investment good, CRE is valued primarily for its ability to generate cash flow—rather than for its consumption value, as occurs for housing. Nonlocal investors remain in their nonlocal offices during the holding period—rather than becoming local residents following the purchase, as occurs for homeowners. These aspects allow us to evaluate informational disadvantages on both purchase and sale sides of the transaction so that the effectiveness of brokerage involvement can be clarified. Transactions among uninformed investors can be compared, with and without brokerage representation, to transactions among informed investors; or when informed investors are matched against the uninformed.
B) Brokerage Intermediation in Real Estate Markets
Real estate brokers offer a wide range of services depending on which party to the transaction they represent. Listing brokers generally provide direct representation for the seller and offer information about the local property market, advising the seller in setting or changing the listing price, marketing and showing the property, negotiating the sale price, and assisting the seller with the closing process. Buyer brokers—who contractually represent the buyer’s interest—may also provide informational, search, bargaining, and facilitating roles. Informational transparency has increased for real estate assets with widespread adoption of property search technology in recent years, such as the MLS provides for the housing sector. Yet, it remains a widely-held belief that real estate brokers possess a high degree of specialized market knowledge that can act to offset informational disadvantages faced by either party, when represented. As Han and Strange [
There are issues with incentive alignment and brokerage compensation for both the listing broker and the buyer broker. In the United States, the conventional brokerage fee is structured as a percentage commission rate (typically around 6 percent of the sale price), and the listing agreement defines the expiration date. The listing broker’s incentive is then to ensure that the property sells within the contract period [
Buyer brokers have incentives to minimize search costs per successful transaction. The first-stage objective is to search until a property is found that sufficiently satisfies buyer criteria. Once an acceptable match is found, the objective then becomes to close the deal. If a buyer broker can convince their client to increase the offer, it increases the probability of concluding the search and linearly increases brokerage compensation [
Under dual agency, the brokerage firm represents both parties to the transaction—with conflicting implications for the marketing outcome. Brokers involved in dual agency are uniquely positioned to possess privileged information about the reservation prices of both parties. On the one hand, such information could be exploited to maximize sale price and broker compensation. On the other hand, pressure toward a lower price increases the likelihood of successful agreement during the bargaining stage. Accordingly, dual agency heightens litigation risk for the participating brokerage firm in the event that either party feels unfairly treated after closing. Gardiner, Heisler, Kallberg and Liu [
In the brokerage intermediation literature, almost all studies are based on housing data. One exception is Hardin, Johnson and Wu [
In the present study, we challenge these notions. Not all CRE investors are highly sophisticated (e.g. see earlier discussion for nonlocal clienteles) and the assets involved have highly heterogeneous attribute sets, distorting the valuation process. Informational transparency is arguably no greater in the commercial sector than it is for houses. CRE assets are thinly traded and detailed information about transactions is primarily available through private vendors, such as CoStar and Real Capital Analytics, with subscriptions that can be cost-prohibitive for even some commercial brokers to obtain. Whereas housing data in the modern era are essentially freely available to the general public through online platforms such as Zillow and Trulia. As a consequence of noisy asset values and arguablyless informational transparency in the CRE sector, it is possible that even brokers themselves do not always possess a high degree of specialized market knowledge that could be used to offset informational disadvantages.
An alternative view is that the CRE broker possesses information primarily limited to the seller’s reservation price (when serving as the listing broker), the buyer’s reservation price (as the buyer broker), or both (under dual agency). The broker’s objective is to successfully close a deal within the specified contract period of the brokerage arrangement. Success is measured relative to the client’s expectation, rather than some unobservable market value. Uninformed investors generate expected values for the asset that are subject to greater variance and potentially biased relative to expectations formed by informed investors. As a consequence, a property is expected to sell at a higher price when an informed seller matches with an uninformed buyer; or at a lower price whenever an uninformed seller matches with an informed buyer. If brokers play an important information role, then brokerage intermediation should significantly reduce price disparity in mismatched transactions. If, instead, brokers primarily negotiate relative to their client’s expectation, then brokers bargaining capabilities are vulnerable. Successful bargaining by brokers should result in informed investors receiving better prices for their assets when matched against uninformed investors. Dual agency scenarios potentially mitigate the exposure for uninformed participants. If litigation risk is relevant to the broker’s expected cost function, then dual agency brokers—who are likely informed about both sets of reservation prices—might steer negotiations toward a price where neither party appears disadvantaged.
Greensboro/Winston-Salem, Greenville/Spartanburg, Hampton Roads, Hartford, Hawaii, Houston,
The central motivation in this study is to explore the interaction of brokerage intermediation when CRE investors are uninformed. The identification strategy for uninformed investors is based on informational disadvantages faced by nonlocal clienteles who tend to underperform in the office market. We evaluate the roles of both listing brokers and buyer brokers, as well as dual agency scenarios. The lens through which we examine these effects applies a propensity-score matching technique to limit sample selection bias and to provide this comparison for CRE assets that have similar characteristics. Brokerage intermediation effects are compared for properties bought and sold by informed market participants, as well as for properties bought and sold by uninformed participants. This is the first research study to evaluate the interaction of broker involvement for nonlocal clientele effects in CRE investment markets.
CRE transactions data used in this study are from the CoStar COMPs® database, which provides nationwide (US) coverage for commercial property transactions. The purchase samples used to examine brokerage intermediation and nonlocal investor effects consists of 17,157 office property transactions; the sales sample contains 16,996 observations1. Transaction dates range from 1991 through 2013, with observations taken from 138 US office markets2. In order to be included in the sample, each observation is required to have information available in data fields for investor identity and investor address, along with all other variables relevant to the analysis.
Huntington/Ashland, Huntsville, Indianapolis, Inland Empire (California), Jackson, Jacksonville (Florida), Kansas City, Killeen/Temple/Fort Hood, Kingsport/Bristol/Bristol, Knoxville, Lafayette, Las Vegas, Lexington/Fayette, Lincoln, Little Rock/N Little Rock, Long Island (New York), Los Angeles, Louisville, Lubbock, Marin/Sonoma, McAllen/Edinburg/Pharr, Memphis, Milwaukee/Madison, Minneapolis/St Paul, Mobile, Montgomery, Myrtle Beach/Conway, Nashville, New Orleans/Metairie/Kenner, New York City, Northern New Jersey, Ocala, Oklahoma City, Olympia, Omaha/Council Bluffs, Orange County (California), Orlando, Pensacola, Peoria, Philadelphia, Phoenix, Pittsburgh, Port St Lucie/Fort Pierce, Portland, Portland/South Portland, Providence, Raleigh/Durham, Reno/Sparks, Richmond VA, Roanoke, Rochester, Sacramento, Salinas, Salt Lake City, San Antonio, San Diego, San Francisco, San Luis Obispo/Paso Robles, Santa Barbara/Sta Maria/Goleta, Santa Cruz/Watsonville, Savannah, Seattle/Puget Sound, Shreveport/Bossier City, South Bay/San Jose, South Bend/Mishawaka, South Florida, Southwest Florida, Spokane, Springfield, St. Louis, Stockton/Modesto, Syracuse, Tallahassee, Tampa/St Petersburg, Toledo, Tucson, Tulsa, Utica/Rome, Visalia/Porterville, Washington, DC, West Michigan, Westchester/So Connecticut, Wichita, Wilmington, Yakima, and Youngstown/Warren/Boardman.
In order to provide a meaningful and accurate comparison, a propensity-score matched sample is constructed. A probit model is estimated for the probability that an office building selected by a local buyer will involve buyer representation. The probit model includes variables for land area, building size, property age, along with indicators for property class, multi-tenancy, buyer classification3, acquisition timing, geographic market, and sale conditions4. For each observation of a local buyer who purchased an office building with buyer representation, an observation for a local buyer who did not have buyer representation is collected.
Panel A | Local Buyers | Nonlocal Buyers | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Full sample | Represented | Unrepresented† | Full sample | Represented | Unrepresented† | |||||||
Observations | 12,103 | 4512 | 4512 | 5054 | 1309 | 1309 | ||||||
Variable | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) |
Sale price ($/SF) | 140 | (122) | 144 | (120) | 138 | (122) | 168 | (132) | 170 | (133) | 182 | (143) |
Land area (SF) | 87,382 | (379,183) | 77,505 | (185,781) | 78,054 | (214,476) | 193,458 | (523,516) | 150,690 | (261,126) | 171,540 | (320,831) |
Size (SF) | 33,930 | (102,854) | 26,135 | (66,099) | 30,628 | (95,586) | 97,639 | (176,333) | 66,752 | (132,351) | 82,396 | (145,767) |
Age (years) | 39 | (31) | 38 | (30) | 40 | (31) | 27 | (24) | 26 | (21) | 28 | (26) |
Class A | 0.06 | (0.2) | 0.04 | (0.2) | 0.04 | (0.2) | 0.25 | (0.4) | 0.16 | (0.4) | 0.18 | (0.4) |
Class B | 0.46 | (0.5) | 0.50 | (0.5) | 0.48 | (0.5) | 0.54 | (0.5) | 0.62 | (0.5) | 0.60 | (0.5) |
Class C | 0.48 | (0.5) | 0.46 | (0.5) | 0.48 | (0.5) | 0.21 | (0.4) | 0.22 | (0.4) | 0.22 | (0.4) |
Multi-tenant | 0.70 | (0.5) | 0.69 | (0.5) | 0.69 | (0.5) | 0.80 | (0.4) | 0.78 | (0.4) | 0.79 | (0.4) |
Market time | 382 | (379) | 386 | (376) | 386 | (392) | 252 | (290) | 281 | (304) | 245 | (283) |
Buyer rep | 0.37 | (0.5) | 1 | (0) | 0 | (0) | 0.26 | (0.4) | 1 | (0) | 0 | (0) |
Seller rep | 0.64 | (0.5) | 0.92 | (0.3) | 0.51 | (0.5) | 0.61 | (0.5) | 0.91 | (0.3) | 0.60 | (0.5) |
Same rep | 0.12 | (0.3) | 0.33 | (0.5) | 0 | (0) | 0.11 | (0.3) | 0.42 | (0.5) | 0 | (0) |
Nonlocal | 0 | (0) | 0 | (0) | 0 | (0) | 1 | (0) | 1 | (0) | 1 | (0) |
Panel B | Represented Buyers | Unrepresented Buyers | ||||||||||
Full sample | Nonlocal | Local† | Full sample | Nonlocal | Local† | |||||||
Observations | 5821 | 1309 | 1309 | 11,336 | 3745 | 3745 | ||||||
Variable | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) |
Sale price ($/SF) | 150 | (124) | 170 | (133) | 153 | (120) | 147 | (126) | 167 | (132) | 141 | (124) |
Land area (SF) | 93,962 | (207,388) | 150,690 | (261,126) | 129,059 | (253,407) | 131,296 | (506,624) | 208,407 | (587,543) | 136,281 | (618,932) |
Size (SF) | 35,269 | (87,239) | 66,752 | (132,351) | 48,743 | (92,339) | 61,646 | (149,293) | 108,435 | (188,134) | 59,530 | (142,737) |
Age (years) | 35 | (29) | 26 | (21) | 28 | (22) | 36 | (30) | 27 | (25) | 34 | (28) |
Class A | 0.07 | (0.3) | 0.16 | (0.4) | 0.10 | (0.3) | 0.14 | (0.4) | 0.28 | (0.5) | 0.13 | (0.3) |
Class B | 0.53 | (0.5) | 0.62 | (0.5) | 0.63 | (0.5) | 0.46 | (0.5) | 0.52 | (0.5) | 0.52 | (0.5) |
Class C | 0.40 | (0.5) | 0.22 | (0.4) | 0.27 | (0.4) | 0.39 | (0.5) | 0.20 | (0.4) | 0.35 | (0.5) |
Multi-tenant | 0.71 | (0.5) | 0.78 | (0.4) | 0.78 | (0.4) | 0.74 | (0.4) | 0.81 | (0.4) | 0.77 | (0.4) |
Market time | 364 | (365) | 281 | (304) | 350 | (380) | 252 | (290) | 281 | (304) | 353 | (381) |
Buyer rep | 1 | (0) | 1 | (0) | 1 | (0) | 0 | (0) | 0 | (0) | 0 | (0) |
Seller rep | 0.92 | (0.3) | 0.91 | (0.3) | 0.89 | (0.3) | 0.48 | (0.5) | 0.51 | (0.5) | 0.46 | (0.5) |
Same rep | 0.35 | (0.5) | 0.42 | (0.5) | 0.36 | (0.5) | 0 | (0) | 0 | (0) | 0 | (0) |
Nonlocal | 0.22 | (0.4) | 1 | (0) | 0 | (0) | 0.33 | (0.5) | 1 | (0) | 0 | (0) |
†Based on propensity-score matched sample. Notes: This table presents summary statistics for the purchase samples, including the sample mean (Mean) and standard deviation (Std dev) in parentheses. Panel A presents the comparison between Local Buyers and Nonlocal Buyers, partitioned into subsamples of buyers who were Represented by brokers or Unrepresented in the transaction. Unrepresented subsamples are generated from propensity-score matched samples. Panel B provides the comparison between Represented and Unrepresented Buyers, partitioned into subsamples of Nonlocal and Local buyers. Subsamples of Local buyers are produced from propensity-score matched samples. Variable definitions: Sale price is the transaction price for the office property, in US dollars, divided by Size. Land area is the gross square footage (SF) of the lot. Size is the rentable building area, measured in square feet (SF). Age is structure age, measured in years relative to the transaction date. Class A, Class B, and Class C are indicator variables taking on a value of one for the respective property class and zero otherwise. Multi-tenant is an indicator variable for properties that are not designed to be fully-occupied by a single tenant. Market time is the listing duration, in calendar days. Buyer rep is an indicator variable for buyers who had broker representation in the transaction. Seller rep is an indicator for use of a listing agent. Same rep is an indicator variable for Buyer rep and Seller rep from the same brokerage company for the transaction. Nonlocal is an indicator variable for buyers having a company address in a different geographic market than the asset.
The matching process is carried out based on a one-to-one, nearest-neighbor propensity score matching with replacement. Summary statistics for the matched sample of unrepresented local buyers are also shown in
The analogous sequence is executed to produce matched samples for unrepresented nonlocal buyers in Panel A, as well as for represented buyers with local addresses (matched to transactions that have similar attributes to those purchased by represented buyers with nonlocal addresses), and for unrepresented buyers with local addresses in Panel B of
By constructing the comparison samples for the empirical test in this manner, the results in this study will reveal 1) whether broker representation has an impact on purchase prices for local buyers, 2) whether representation impacts prices paid by nonlocal buyers, 3) whether nonlocal origin impacts prices paid by buyers with broker representation, and 4) whether nonlocal origin impacts price for unrepresented buyers. In addition to purchases, this study considers the impact on exit price obtained by local and nonlocal sellers, depending on whether broker representation is involved or not. Summary statistics for the sales samples are provided in
Once the matched sampling procedures are applied, the next step in the empirical analysis estimates a hedonic regression with transaction price per square foot (Sale price), logged, as the dependent variable. The model is written as:
ln ( Saleprice ) = β 0 + β 1 ⋅ ln ( Landarea ) + β 2 ⋅ ln ( Size ) + β 3 ⋅ ln ( Age ) + β 4 ⋅ ClassA + β 5 ⋅ ClassB + β 6 ⋅ Multi-tenant + β 7 ⋅ ln ( Markettime ) + β 8 ⋅ Buyerrep + β 9 ⋅ Samerep + β 10 ⋅ Sellerrep + β 11 ⋅ Nonlocal + γ ⋅ X + δ ⋅ T + η ⋅ C + θ ⋅ M + ε . (1)
Land area, Size, Age, and Market time are logged. Class A and Class B are indicator variables for property class. Multi-tenant is an indicator variable for multi-tenant office assets. Buyer rep and Seller rep are indicator variables for broker representation of the respective parties. Same rep indicates that both parties were represented by brokers from the same firm. Nonlocal is an indicator variable for investors whose address is located in a different geographic market than the asset. Nonlocal identifies buyers in the purchase samples and sellers in the sales samples. X is a matrix of control variables for investor classification. T is a matrix of indicator variables for calendar year of the transaction date. C is a matrix of control variables for unique sale conditions. M is a matrix of geographic market indicator variables. γ, δ, η, and θ are vectors of parameters for
Panel A | Local Sellers | Nonlocal Sellers | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Full sample | Represented | Unrepresented† | Full sample | Represented | Unrepresented† | |||||||
Observations | 11,540 | 4053 | 4053 | 5456 | 1581 | 1581 | ||||||
Variable | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) |
Sale price ($/SF) | 147 | (124) | 148 | (119) | 151 | (129) | 141 | (119) | 148 | (120) | 151 | (129) |
Land area (SF) | 112,937 | (1,619,026) | 92,792 | (266,899) | 169,364 | (2,714,743) | 156,631 | (350,495) | 149,793 | (241,252) | 167,219 | (309,256) |
Size (SF) | 39,566 | (122,076) | 40,422 | (133,067) | 43,862 | (128,947) | 79,324 | (145,773) | 75,786 | (129,097) | 92,317 | (173,109) |
Age (years) | 39 | (32) | 39 | (32) | 38 | (31) | 30 | (26) | 31 | (26) | 31 | (27) |
Class A | 0.08 | (0.3) | 0.08 | (0.3) | 0.09 | (0.3) | 0.20 | (0.4) | 0.20 | (0.4) | 0.21 | (0.4) |
Class B | 0.46 | (0.5) | 0.46 | (0.5) | 0.47 | (0.5) | 0.52 | (0.5) | 0.51 | (0.5) | 0.50 | (0.5) |
Class C | 0.46 | (0.5) | 0.46 | (0.5) | 0.44 | (0.5) | 0.27 | (0.5) | 0.29 | (0.5) | 0.29 | (0.5) |
Multi-tenant | 0.71 | (0.5) | 0.72 | (0.5) | 0.73 | (0.4) | 0.77 | (0.4) | 0.76 | (0.4) | 0.77 | (0.4) |
Market time | 376 | (374) | 380 | (366) | 379 | (419) | 287 | (318) | 308 | (336) | 243 | (279) |
Buyer rep | 0.35 | (0.5) | 0.48 | (1) | 0.08 | (0) | 0.35 | (0.5) | 0.43 | (0.5) | 0.09 | (0.3) |
Seller rep | 0.65 | (0.5) | 1 | (0) | 0 | (0) | 0.71 | (0.5) | 1 | (0) | 0 | (0) |
Same rep | 0.12 | (0.3) | 0.18 | (0.4) | 0 | (0) | 0.13 | (0.3) | 0.17 | (0.4) | 0 | (0) |
Nonlocal | 0 | (0) | 0 | (0) | 0 | (0) | 1 | (0) | 1 | (0) | 1 | (0) |
Panel B | Represented Sellers | Unrepresented Sellers | ||||||||||
Full sample | Nonlocal | Local† | Full sample | Nonlocal | Local† | |||||||
Observations: | 11,362 | 3875 | 3875 | 5634 | 1581 | 1581 | ||||||
Variable | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) | Mean | (Std dev) |
Sale price ($/SF) | 142 | (119) | 137 | (114) | 151 | (121) | 151 | (129) | 151 | (129) | 151 | (123) |
Land area (SF) | 106,237 | (280,731) | 152,311 | (365,938) | 117,277 | (285,215) | 168,762 | (2,308,286) | 167,219 | (309,256) | 196,450 | (1,199,055) |
Size (SF) | 49,785 | (124,508) | 74,023 | (132,677) | 57,608 | (153,733) | 57,459 | (144,361) | 92,317 | (173,109) | 73,317 | (178,073) |
Age (years) | 36 | (30) | 30 | (25) | 33 | (28) | 36 | (30) | 31 | (27) | 34 | (29) |
Class A | 0.12 | (0.3) | 0.20 | (0.4) | 0.13 | (0.3) | 0.12 | (0.3) | 0.21 | (0.4) | 0.16 | (0.4) |
Class B | 0.48 | (0.5) | 0.53 | (0.5) | 0.52 | (0.5) | 0.48 | (0.5) | 0.50 | (0.5) | 0.51 | (0.5) |
Class C | 0.40 | (0.5) | 0.27 | (0.4) | 0.35 | (0.5) | 0.40 | (0.5) | 0.29 | (0.5) | 0.34 | (0.5) |
Multi-tenant | 0.72 | (0.5) | 0.77 | (0.4) | 0.73 | (0.4) | 0.74 | (0.4) | 0.77 | (0.4) | 0.77 | (0.4) |
Market time | 348 | (355) | 290 | (320) | 342 | (350) | 345 | (394) | 243 | (279) | 362 | (466) |
Buyer rep | 0.49 | (0.5) | 0.46 | (0.5) | 0.50 | (0.5) | 0.08 | (0.3) | 0.09 | (0.3) | 0.08 | (0.3) |
Seller rep | 1 | (0) | 1 | (0) | 1 | (0) | 0 | (0) | 0 | (0) | 0 | (0) |
Same rep | 0.18 | (0.4) | 0.18 | (0.4) | 0.19 | (0.4) | 0 | (0) | 0 | (0) | 0 | (0) |
Nonlocal | 0.34 | (0.5) | 1 | (0) | 0 | (0) | 0.28 | (0.5) | 1 | (0) | 0 | (0) |
†Based on propensity-score matched sample. Notes: This table presents summary statistics for the sales samples, including the sample mean (Mean) and standard deviation (Std dev) in parentheses. Panel A presents the comparison between Local Sellers and Nonlocal Sellers, partitioned into subsamples of sellers who were represented by brokers or Unrepresented in the transaction. Unrepresented subsamples are generated from propensity-score matched samples. Panel B provides the comparison between Represented and Unrepresented Sellers, partitioned into subsamples of Nonlocal and Local sellers. Subsamples of Local sellers are produced from propensity-score matched samples. All variables are defined in the notes to
fixed effects. Empirical results for the estimation of Equation (1) for four purchase samples and four sales samples are discussed in the next section.
Sample | Local Buyers | Nonlocal Buyers | Represented | Unrepresented | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Variable | Coefficient | t-stat | Coefficient | t-stat | Coefficient | t-stat | Coefficient | t-stat | ||||
Constant | 6.525 | *** | 33.2 | 6.884 | *** | 16.1 | 6.290 | *** | 23.3 | 6.884 | *** | 26.6 |
ln(Land area) | −0.007 | −0.9 | −0.088 | *** | −5.5 | −0.076 | *** | −5.6 | −0.075 | *** | −6.5 | |
ln(Size) | −0.200 | *** | −19.9 | −0.049 | *** | −2.7 | −0.079 | *** | −4.7 | −0.068 | *** | −5.0 |
ln(Age) | −0.146 | *** | −15.2 | −0.153 | *** | −7.9 | −0.162 | *** | −10.0 | −0.179 | *** | −11.8 |
Class A | 0.532 | *** | 10.5 | 0.435 | *** | 6.3 | 0.421 | *** | 7.0 | 0.477 | *** | 9.5 |
Class B | 0.150 | *** | 8.5 | 0.129 | *** | 3.0 | 0.115 | *** | 3.6 | 0.112 | *** | 3.6 |
Multi-tenant | −0.037 | ** | −2.3 | −0.154 | *** | −4.0 | −0.109 | *** | −3.6 | −0.123 | *** | −4.4 |
ln(Market time) | −0.032 | *** | −4.7 | −0.051 | *** | −3.4 | −0.059 | *** | −4.8 | −0.051 | *** | −4.9 |
Seller rep | −0.027 | −1.0 | 0.183 | ** | 2.5 | 0.055 | 0.8 | 0.073 | ** | 2.2 | ||
Same rep | −0.107 | *** | −5.3 | −0.081 | ** | −2.0 | −0.066 | ** | −2.6 | |||
Buyer rep | 0.107 | *** | 6.5 | 0.020 | 0.6 | |||||||
Nonlocal | 0.071 | *** | 3.0 | 0.078 | *** | 3.1 | ||||||
Buyer type | 21 indicators | 20 indicators | 20 indicators | 21 indicators | ||||||||
Calendar year | 6 indicators | 6 indicators | 6 indicators | 7 indicators | ||||||||
Sale conditions | 117 indicators | 61 indicators | 62 indicators | 79 indicators | ||||||||
Market | 127 indicators | 92 indicators | 137 indicators | 118 indicators | ||||||||
Adjusted R2 | 56.7% | 56.5% | 62.2% | 53.5% | ||||||||
Observations | 9024 | 2618 | 2618 | 7490 |
Notes: This table presents the results from the estimation of Sale price ($/SF), logged, for the purchase samples. Propensity-score matched samples are used in each case, consisting of equal number observations for Represented and Unrepresented Buyers (in the Local and Nonlocal subsamples), or equal number of observations for Local and Nonlocal Buyers (in the Represented and Unrepresented subsamples). The variables Land area, Size, Age, and Market time are each logged. The variable name is listed in the first column, the estimated coefficient in the second, and the t-statistic (t-stat) in italics is in the third column. The respective t-statistic and reported significance level are based on standard errors clustered by market and calendar year. All variables are defined in the notes to
There are at least two possible explanations. The first suggests that brokers have a perverse incentive to maximize transaction price because their compensation structure, as a percentage commission rate, is linearly increasing with price. Perhaps financial incentives for buyer brokers are misaligned with the buyer’s objective of minimizing the acquisition price. The second possibility is that selection bias exists for buyers who opt for broker representation. In the CRE investment market, experienced and sophisticated investors may be more likely to forgo the use of a buyer broker. Supplementary evidence in the sample appears to support this explanation. Buyers who use brokers the least (10 percent or less of the time) include listed funds, pension funds, sovereign funds, real estate investment trusts, and real estate operating companies—all relatively sophisticated institutional investors. Buyers who use brokers the most frequently (40 percent of the time or more) are trusts, corporates, educational funds, endowments, and individual buyers. Non-representation thereby functions as a proxy for the pre-existing competitive edge, such as firms with dedicated analysts and in-house CRE research divisions.
While buyer representation appears to convey a disadvantage, use of a broker from the same firm as the listing brokerage exactly offsets this outcome for local buyers. The estimated transaction price is reduced by 10.7 percent when using brokers from the same firm (coefficient for Same rep), subtracted from the 10.7 percent premium for having used a buyer broker in the first place nets to zero. Dual agency has the potential for efficiency gains such as when an internal client is identified as a potential match for the listing—also called internalized transactions. Yet, dual agency also creates the potential for incentive misalignments that stem from informational asymmetries inherent in CRE transactions. Brokers may possess private knowledge regarding the reservation price, motivation, urgency, or constraints of either party. While use of this knowledge may facilitate a more expeditious screening of potential matches, it also exposes the brokerage firm to increased risk of litigation if either party feels treated unfairly after the closing. The perception of heightened litigation risk can serve as a mechanism that encourages brokers who are involved in dual agency transactions to limit the degree to which they might allow a buyer to overpay.
Continuing with the results in
When faced with a seller who has broker representation, the most impactful decision a nonlocal buyer can make is to elect for dual agency and seek out a buyer broker from the same firm of the listing broker. The estimated price impact from dual agency reduces the degree of overpayment for nonlocals by 8.1 percent (coefficient of Same rep)—netting out to 10.2 percent overpayment, versus 18.3 percent in the absence of dual agency. Thus, while the optimal path for a nonlocal buyer is to pursue assets that are marketed by unrepresented sellers, when a seller has broker representation, nonlocal buyers should attempt to hire a broker to represent their own interest from the same firm. If not, the degree of overpayment by nonlocal buyers transacting with a represented seller is 18.3 percent—independent of whether buyer representation exists.
Whereas the first and second sets of results in
The third and fourth sets of results in
Sample | Local Sellers | Nonlocal Sellers | Represented | Unrepresented | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Variable | Coefficient | t-stat | Coefficient | t-stat | Coefficient | t-stat | Coefficient | t-stat | ||||
Constant | 6.612 | *** | 32.6 | 6.769 | *** | 19.1 | 6.904 | *** | 39.5 | 6.439 | *** | 10.1 |
ln(Land area) | −0.004 | −0.4 | −0.048 | *** | −2.9 | −0.049 | *** | −6.5 | −0.055 | −1.5 | ||
ln(Size) | −0.160 | *** | −13.0 | −0.076 | *** | −4.1 | −0.071 | *** | −7.9 | −0.109 | *** | −2.8 |
ln(Age) | −0.126 | *** | −10.5 | −0.170 | *** | −8.3 | −0.166 | *** | −18.3 | −0.129 | *** | −2.9 |
Class A | 0.564 | *** | 9.7 | 0.660 | *** | 9.0 | 0.571 | *** | 16.2 | 0.492 | *** | 3.2 |
Class B | 0.141 | *** | 6.2 | 0.131 | *** | 3.1 | 0.095 | *** | 5.0 | 0.099 | 1.1 | |
Multi-tenant | −0.076 | ** | −3.6 | −0.077 | ** | −2.1 | −0.084 | *** | −4.9 | −0.101 | −1.3 | |
ln(Market time) | −0.041 | *** | −4.7 | −0.042 | *** | −2.7 | −0.051 | *** | −7.2 | −0.053 | * | −2.0 |
Buyer rep | 0.101 | *** | 4.4 | 0.002 | 0.1 | 0.048 | *** | 2.9 | 0.056 | 0.5 | ||
Same rep | −0.079 | *** | −2.6 | −0.051 | −1.0 | −0.064 | *** | −3.1 | ||||
Seller rep | −0.029 | −1.1 | −0.006 | −0.1 | ||||||||
Nonlocal | −0.081 | *** | −5.5 | −0.120 | * | −1.7 | ||||||
Seller type | 19 indicators | 20 indicators | 20 indicators | 14 indicators | ||||||||
Calendar year | 9 indicators | 6 indicators | 7 indicators | 8 indicators | ||||||||
Sale conditions | 81 indicators | 87 indicators | 112 indicators | 38 indicators | ||||||||
Market | 115 indicators | 107 indicators | 121 indicators | 85 indicators | ||||||||
Adjusted R2 | 50.7% | 59.3% | 59.5% | 50.9% | ||||||||
Observations | 8106 | 3162 | 7750 | 3162 |
Notes: This table presents the results from the estimation of Sale price ($/SF), logged, for the sales samples. Propensity-score matched samples are used in each case, consisting of equal number observations for Represented and Unrepresented Sellers (in the Local and Nonlocal subsamples), or equal number of observations for Local and Nonlocal Sellers (in the Represented and Unrepresented subsamples). The variables Land area, Size, Age, and Market time are each logged. The variable name is listed in the first column, the estimated coefficient in the second, and the t-statistic (t-stat) in italics is in the third column. The respective t-statistic and reported significance level are based on standard errors clustered by market and calendar year. All variables are defined in the notes to
Sample | Local Sellers | Nonlocal Sellers | Represented | Unrepresented | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Variable | Coefficient | t-stat | Coefficient | t-stat | Coefficient | t-stat | Coefficient | t-stat | ||||
Constant | 5.286 | *** | 12.3 | 6.138 | *** | 8.2 | 6.277 | *** | 17.5 | 6.225 | *** | 4.2 |
ln(Land area) | 0.026 | 1.3 | 0.034 | 1.1 | 0.007 | 0.5 | 0.039 | 0.5 | ||||
ln(Size) | −0.010 | −0.4 | −0.005 | −0.2 | −0.020 | −1.2 | −0.030 | −0.4 | ||||
ln(Age) | −0.126 | *** | −5.5 | −0.031 | −0.8 | −0.120 | *** | −6.8 | −0.090 | −0.9 | ||
Class A | −0.464 | *** | −4.2 | −0.509 | *** | −3.6 | −0.370 | *** | −5.5 | 0.076 | 0.2 | |
Class B | 0.055 | 1.3 | 0.106 | 1.3 | 0.052 | 1.5 | 0.227 | 1.2 | ||||
Multi-tenant | 0.052 | 1.3 | −0.091 | −1.3 | −0.070 | ** | −2.2 | 0.168 | 1.0 | |||
ln(Sale price) | −0.148 | *** | −4.7 | −0.151 | *** | −2.7 | −0.180 | *** | −7.2 | −0.230 | * | −2.0 |
Buyer rep | 0.162 | *** | 3.7 | 0.174 | ** | 2.4 | 0.121 | *** | 3.9 | 0.111 | 0.5 | |
Same rep | −0.242 | *** | −4.2 | −0.088 | −0.9 | −0.150 | *** | −4.0 | ||||
Seller rep | 0.183 | *** | 3.7 | 0.245 | *** | 2.9 | ||||||
Nonlocal | −0.040 | −1.6 | 0.048 | 0.3 | ||||||||
Seller type | 19 indicators | 20 indicators | 20 indicators | 14 indicators | ||||||||
Calendar year | 9 indicators | 6 indicators | 7 indicators | 8 indicators | ||||||||
Sale conditions | 81 indicators | 87 indicators | 112 indicators | 38 indicators | ||||||||
Market | 115 indicators | 107 indicators | 121 indicators | 85 indicators | ||||||||
Adjusted R2 | 14.6% | 25.8% | 15.6% | 34.4% | ||||||||
Observations | 8106 | 3162 | 7750 | 3162 |
Notes: This table presents the results from the estimation of Market time, logged, for the sales samples. Propensity-score matched samples are used in each case, consisting of equal number observations for Represented and Unrepresented Sellers (in the Local and Nonlocal subsamples), or equal number of observations for Local and Nonlocal Sellers (in the Represented and Unrepresented subsamples). The variables Land area, Size, Age, and Sale price are each logged. The variable name is listed in the first column, the estimated coefficient in the second, and the t-statistic (t-stat) in italics is in the third column. The respective t-statistic and reported significance level are based on standard errors clustered by market and calendar year. All variables are defined in the notes to
with unrepresented buyers (represents 92 percent of transactions in the unrepresented seller sample). In a related manner, dual agency is occasionally referred to as “internalized” transactions, due to the increased likelihood that a buyer is matched from the established network of the listing brokerage firm. Dual agency appears to significantly reduce marketing duration for local sellers.
Brokers fail to play the intended role, at least on behalf of uninformed parties. Buyer representation has no detectable impact on prices paid by nonlocal investors—the proxy group for informationally-disadvantaged. Nonlocal buyers pay significant premiums, in the magnitude of 7 to 8 percent, regardless of broker representation. Seller representation also has no meaningful impact on price. Sale prices attained by nonlocals are not significantly different when a listing broker is involved, even though marketing durations are considerably longer. These findings confirm a pronounced clientele effect for nonlocal investors in CRE markets—one that is largely unaffected by brokerage intermediation. Either brokers do not possess specialized market information that might be used to offset the informational disadvantages faced by their clients, or they neglect to effectively deploy such knowledge. Impactful information for a particular transaction may be elusive due to the highly heterogeneous nature of CRE assets. Withholding such information may occur when brokerage incentives are misaligned with the principal.
Whereas the informational role appears limited when brokers act on behalf of uninformed investors, the bargaining role appears with distinction when brokers represent the party opposite to uninformed investors. The degree of overpayment is significantly higher for nonlocal buyers when the seller is represented by a broker. Yet, when opposing local buyers, a broker-represented seller is unable to have a significant impact on selling price. Thus, brokers are able to extract willingness-to-pay surplus from uninformed buyers only.
The mitigating circumstance to all of the above detrimental outcomes for nonlocal investors occurs under dual agency, where the same brokerage firm represents both buyer and seller and litigation risk is amplified. Under dual agency, the degree of overpayment is significantly reduced for all disadvantaged buyers. This result suggests that brokers are likely to possess privileged information about investor reservation prices, rather than market values. Knowing both informed and uninformed investor expectations, dual agency brokers are able to facilitate more favorable outcomes for uninformed investors.
Nonetheless, the belief that brokers possess a high degree of specialized market knowledge that can act to offset informational disadvantages faced by either party is widely endorsed. We find little evidence to support this notion. Nonlocal investors are persistent underperformers in the markets for CRE investments, with and without brokers, suggesting that there are practical limitations to theoretically-optimal geographic diversification within CRE portfolios.
The authors declare no conflicts of interest regarding the publication of this paper.
Wiley, J.A., Liu, Y., Guo, L. and Gallimore, P. (2019) The Information and Bargaining Roles of Commercial Brokers When Investors Are Uninformed. Theoretical Economics Letters, 9, 1651-1668. https://doi.org/10.4236/tel.2019.95105
Panel A. Purchases | Local Buyers | Nonlocal Buyers | Represented | Unrepresented | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Dependent | Buyer rep | Buyer rep | Nonlocal buyer | Nonlocal buyer | ||||||||
Variable | Coefficient | (Wald Χ2) | Coefficient | (Wald Χ2) | Coefficient | (Wald Χ2) | Coefficient | (Wald Χ2) | ||||
Constant | −0.629 | ** | (5.7) | 0.264 | (0.5) | −2.640 | *** | (39.1) | −2.850 | *** | (123.7) | |
ln(Land area) | 0.004 | (0.1) | 0.017 | (0.5) | 0.061 | ** | (5.8) | 0.047 | *** | (9.0) | ||
ln(Size) | 0.041 | ** | (6.0) | −0.030 | (1.5) | 0.213 | *** | (46.9) | 0.177 | *** | (92.5) | |
ln(Age) | −0.039 | ** | (5.3) | −0.060 | * | (3.7) | −0.070 | ** | (5.2) | 0.019 | (0.9) | |
Class A | −0.227 | *** | (9.8) | −0.170 | (2.6) | 0.016 | (0.0) | 0.042 | (0.4) | |||
Class B | 0.077 | ** | (6.0) | 0.029 | (0.2) | 0.010 | (0.0) | 0.030 | (0.5) | |||
Multi-tenant | −0.060 | ** | (4.2) | 0.038 | (0.4) | −0.080 | (1.9) | −0.100 | *** | (7.2) | ||
Buyer type | 22 indicators | 22 indicators | 21 indicators | 23 indicators | ||||||||
Calendar year | 17 indicators | 14 indicators | 11 indicators | 16 indicators | ||||||||
Sale conditions | 231 indicators | 143 indicators | 161 indicators | 216 indicators | ||||||||
Market | 135 indicators | 126 indicators | 137 indicators | 136 indicators | ||||||||
Pseudo-R2 | 13.6% | 25.5% | 32.6% | 37.4% | ||||||||
Observations | 12,103 | 5,054 | 5,821 | 11,336 | ||||||||
Panel B. Sales | Local Sellers | Nonlocal Sellers | Represented | Unrepresented | ||||||||
Dependent | Seller rep | Seller rep | Nonlocal seller | Nonlocal seller | ||||||||
Variable | Coefficient | (Wald Χ2) | Coefficient | (Wald Χ2) | Coefficient | (Wald Χ2) | Coefficient | (Wald Χ2) | ||||
Constant | 0.362 | (1.7) | 1.158 | *** | (10.7) | −2.634 | *** | (70.9) | 1.158 | *** | (44.0) | |
ln(Land area) | −0.049 | *** | (12.7) | −0.006 | (0.1) | 0.073 | *** | (23.0) | −0.006 | (0.1) | ||
ln(Size) | −0.050 | *** | (8.9) | −0.065 | *** | (6.8) | 0.151 | *** | (68.2) | −0.065 | *** | (52.4) |
ln(Age) | 0.016 | (0.9) | 0.010 | (0.1) | 0.086 | *** | (21.9) | 0.010 | ** | (4.5) | ||
Class A | 0.190 | *** | (7.3) | 0.273 | *** | (8.7) | 0.091 | (1.8) | 0.273 | (0.0) | ||
Class B | 0.026 | (0.6) | 0.102 | * | (2.9) | 0.073 | * | (3.5) | 0.102 | (0.1) | ||
Multi-tenant | −0.059 | * | (3.7) | −0.024 | (0.2) | −0.072 | ** | (4.3) | −0.024 | ** | (5.3) | |
Seller type | 23 indicators | 22 indicators | 21 indicators | 23 indicators | ||||||||
Calendar year | 14 indicators | 16 indicators | 11 indicators | 16 indicators | ||||||||
Sale conditions | 222 indicators | 172 indicators | 220 indicators | 162 indicators | ||||||||
Market | 134 indicators | 128 indicators | 136 indicators | 129 indicators | ||||||||
Pseudo-R2 | 13.8% | 20.5% | 28.5% | 29.3% | ||||||||
Observations | 11,540 | 5456 | 11,362 | 5634 |
Notes: This table presents the results from the probit estimation for investor identity according to subsamples for local investors (dependent variable is Buyer rep or Seller rep), nonlocal investors (dependent variable is Buyer rep or Seller rep), represented investors (dependent variable is Nonlocal), and unrepresented investors (dependent variable is Nonlocal). Panel A provides results for the purchase samples; Panel B for the sales samples. The variables Land area, Size, and Age are each logged. The variable name is listed in the first column, the estimated Coefficient in the second, and the Wald Χ2 test statistic (in parentheses) is in the third column. All variables are defined in the notes to