Commercial real estate and nonlocal investors: price disparities on entry and exit Yu Liu Georgia State University Paul Gallimore University of Reading Jonathan A. Wiley Georgia State University Primary question • Do nonlocal investors pay more than local investors for the same real estate assets? Previous work o o o o o Turnbull and Sirmans (1993) Watkins (1998) Lambson, McQueen and Slade (2004) Clauretie and Thistle (2007) Ihlanfeldt and Mayock (2012) Motivation for this study Prior Research This Study Residential (single and multifamily) Office (CoStar COMPs® database) Single market 138 markets Buying only Buying and Selling Smaller sample size (generally) 10,971 in purchase sample 11,444 in sales sample Varied sample horizons 1996 through 2012 No sale conditions 36 individual sale conditions & combinations No investor clienteles 24 investor types No control for selection bias Propensity-score matching Nonlocal investors: 22% of purchase sample, 29% of sale sample Summary Statistics – Purchase Sample Variable Full Sample (n=10,971) Mean Std dev Nonlocal (n=2,383) Mean Std dev Local (n=8,588) Mean Std dev Price per square foot ($) 159 153 167 146 204 147 Land area (SF) 88,284 263,923 164,926 427,004 67,017 190,591 Building size (SF) 53,349 143,531 120,582 209,871 34,693 111,792 Property age (years) 40 32 31 28 43 32 Class A 0.10 0.30 0.26 0.44 0.05 0.23 Class B 0.47 0.50 0.53 0.50 0.46 0.50 Class C 0.43 0.50 0.21 0.41 0.49 0.50 Nonlocal buyer 0.22 0.41 1 0 0 0 Rent difference (%) 3.82 20.09 17.59 40.21 0 0 Buyer distance (miles) Multi-tenant building 133 0.70 375.95 0.46 597 0.77 612.38 0.42 4 0.69 8.49 0.46 Summary Statistics – Sales Sample Variable Full Sample (n=11,444) Mean Std dev Nonlocal (n=3,335) Mean Std dev Local (n=8,109) Mean Std dev Price per square foot ($) 136 96.90 134 99.60 137 95.76 Land area (SF) 87,400 248,765 119,907 231,627 74,030 254,290 Building size (SF) 48,283 128,806 78,106 155,196 36,018 113,996 Property age (years) 39 31.73 34 27.64 42 33.00 Class A 0.10 0.29 0.18 0.38 0.06 0.24 Class B 0.47 0.50 0.50 0.50 0.45 0.50 Class C 0.44 0.50 0.32 0.47 0.48 0.50 Nonlocal seller 0.29 0.45 1 0 0 0 Rent difference (%) 4.26 22.50 14.63 39.82 0 0 Seller distance (miles) Multi-tenant building Marketing duration 186 0.70 368 428.99 0.46 370.31 624 0.73 302 599.76 0.44 326.66 5 0.69 394 10.60 0.46 383.21 Method OLS Regression Model ln(Price/SF) = Controls + βN·I{Nonlocal} + ε Purchases Sales Expectation (+) (–) Controls: Property characteristics, investor types, calendar year, sale conditions, geographic markets Propensity-score matching approach price paid by nonlocals vs. price paid by local buyers for exact same properties price paid by nonlocals vs. price paid by local buyers for similar properties Exclude local transactions that look least like nonlocal transactions • Perform whole sample probit regression with binary dependent variable Nonlocal (independent variables same as main model) Pr{Nonlocal = 1} = Φ{β0 + βXX + βTT + βYY + βCC + βMM} • Use variable coefficients to produce estimate of probability that transaction involves nonlocal buyer Use this to match each nonlocal transaction with closest local transaction Local transactions: some post-match summary stats Purchases Full Sample Nonlocal Local (n=10,971) (n=2,383) (n=8,588) 2,283 Variable Mean Std dev Mean Std dev Mean Std dev Price per square foot ($) 159 153 204 167 147 169146 117,415 Land area (SF) 88,284 263,923 164,926 427,004 67,017 190,591 78,067 Building size (SF) 53,349 143,531 120,582 209,871 34,693 111,792 Sales Full Sample Nonlocal Local (n=11,444) (n=3,335) (n=8,109) 3,335 Variable Mean Std dev Mean Std dev Mean Std dev Price per square foot ($) 136 96.90 134 99.60 137 142 95.76 112,995 Land area (SF) 87,400 248,765 119,907 231,627 74,030 254,290 Building size (SF) 48,283 128,806 78,106 155,196 36,018 113,996 64,486 Results: ln(Price/SF) = Controls + βN·I{Nonlocal} + ε Estimated premium – nonlocal buyers Buyers, propensity-score matched sample Sellers, propensity-score matched sample Variable Constant ln(Land area) ln(Building size) ln(Property age) Class A Class B Multi-tenant building Coefficient 6.678 *** -0.046 ** -0.110 *** -0.168 *** 0.426 *** 0.110 *** -0.076 *** (t-stat) (51.34) (-1.84) (-4.33) (-17.66) (13.69) (4.39) (-3.88) Variable Constant ln(Land area) ln(Building size) ln(Property age) Class A Class B Multi-tenant building Coefficient 6.514 *** -0.039 ** -0.077 *** -0.164 *** 0.468 *** 0.095 *** -0.073 *** Nonlocal buyer 0.138 (8.00) Included [22 variables] Included [11 variables] Included [105 variables] Included [119 variables] 56.49% 4,766 Nonlocal seller Seller type indicators: Year indicators: Sale conditions: Market indicators: Adjusted R2: Observations: -0.070 *** (-4.75) Included [21 variables] Included [11 variables] Included [123 variables] Included [123 variables] 53.84% 6,670 Buyer type indicators: Year indicators: Sale conditions: Market indicators: Adjusted R2: Observations: *** (t-stat) (51.95) (-1.89) (-2.97) (-10.98) (18.31) (5.56) (-6.79) Results: Estimated premium/discount – nonlocal buyers Buyers, propensity-score matched sample Variable Constant ln(Land area) ln(Building size) ln(Property age) Class A Class B Multi-tenant building Coefficient 6.678 *** -0.046 ** -0.110 *** -0.168 *** 0.426 *** 0.110 *** -0.076 *** Nonlocal buyer 0.138 Buyer type indicators: Year indicators: Sale conditions: Market indicators: Adjusted R2: Observations: *** Sellers, propensity-score matched sample (t-stat) Variable (51.34) Constant (-1.84) ln(Land area) (-4.33) ln(Building size) Base case price effects (-17.66) ln(Property age) (13.69) Overpay byClass 13.8% A (4.39) Sell at discount Class Bof 7% (-3.88) Multi-tenant building (8.00) Included [22 variables] Included [11 variables] Included [105 variables] Included [119 variables] 56.49% 4,766 Nonlocal seller Seller type indicators: Year indicators: Sale conditions: Market indicators: Adjusted R2: Observations: Coefficient 6.514 *** -0.039 ** -0.077 *** -0.164 *** 0.468 *** 0.095 *** -0.073 *** -0.070 *** (t-stat) (51.95) (-1.89) (-2.97) (-10.98) (18.31) (5.56) (-6.79) (-4.75) Included [21 variables] Included [11 variables] Included [123 variables] Included [123 variables] 53.84% 6,670 What explains price differences? • Information Asymmetry – nonlocal investors less well-informed so get poorer deal when they both buy and sell Proxy: Distance • Market Anchoring – investors from higher value markets anchor valuations on those markets Means they overbid when they buy but they have to take the market price when they sell (unless they sell to another investor from a high-price market) Proxy: Rent difference What explains price differences? • Information Asymmetry – nonlocal investors less well-informed so get poorer deal when they both buy and sell Proxy: Distance • Market Anchoring – investors from higher value markets anchor valuations on those markets. Means they overbid when they buy but they have to take the market price when they sell (unless they sell to another investor from a high-price market) Proxy: Rent difference ln(Price/SF) =Controls + βN·I{Nonlocal}+ βS·Distance + βR·Rent diff + ε. Purchase sample: (+) (+) (+) Sales sample: (–) (–) (0) Results: Information asymmetry and anchoring effects Buyers, propensity-score matched sample Variable Constant ln(Land area) ln(Building size) ln(Property age) Class A Class B Multi-tenant building Coefficient 6.704 -0.046 -0.114 -0.166 0.431 0.112 -0.074 Nonlocal buyer Buyer distance Rent difference 0.091 0.00007 0.076 Buyer type indicators: Year indicators: Sale conditions: Market indicators: Adjusted R2: Observations: *** ** *** *** *** *** *** *** ** *** (t-stat) (50.10) (-2.04) (-3.63) (-16.08) (10.42) (4.89) (-4.02) (5.30) (2.25) (2.47) Included [22 variables] Included [11 variables] Included [105 variables] Included [119 variables] 56.63% 4,766 Sellers, propensity-score matched sample Variable Coefficient (t-stat) *** Constant 6.501 (50.37) ** ln(Land area) -0.038 (-1.86) *** ln(Building size) -0.076 (-2.92) Overpay by 9.1% *** ln(Property age) -0.163 (-10.75) *** • Overpayment increases with distance Class A 0.469 (18.56) *** • Overpayment increases with rent differential Class B 0.094 (5.57) *** Multi-tenant building -0.073 (-6.72) Nonlocal seller Seller distance Rent difference Seller type indicators: Year indicators: Sale conditions: Market indicators: Adjusted R2: Observations: -0.046 -0.00004 0.003 *** *** (-2.40) (-3.95) (0.06) Included [21 variables] Included [11 variables] Included [123 variables] Included [123 variables] 53.86% 6,670 e.g. Buyer located 600 miles away pays 600x0.007% = 4.2% more e.g. Buyer from market with rents 17.5% higher pays 17.5x7.6% = 1.3% more Results: Information asymmetry and anchoring effects Buyers, propensity-score matched sample Sellers, propensity-score matched sample Variable Constant ln(Land area) ln(Building size) ln(Property age) Class A Class B Multi-tenant building (t-stat) (50.10) (-2.04) (-3.63) (-16.08) (10.42) (4.89) (-4.02) Variable Constant ln(Land area) ln(Building size) ln(Property age) Class A Class B Multi-tenant building Coefficient 6.501 -0.038 -0.076 -0.163 0.469 0.094 -0.073 (5.30) Nonlocal seller Seller distance Rent difference -0.046 -0.00004 0.003 Nonlocal buyer Buyer distance Coefficient 6.704 *** -0.046 ** -0.114 *** -0.166 *** 0.431 *** 0.112 *** -0.074 *** 0.091 0.00007 *** ** (2.25) *** Rent difference 0.076 (2.47) Buyer type indicators: Included [22 variables] Nonlocal seller gets Included 1% less[11than Year indicators: variables] locals for every 250 miles Sale conditions: Includedaway [105 variables] Market indicators: Included [119 variables] from market 2 Adjusted R : 56.63% Observations: 4,766 Sell at discount of 4.6% • Discounting increases with distance • Unaffected by nonlocal rent differential Seller type indicators: Year indicators: Sale conditions: Market indicators: Adjusted R2: Observations: *** ** *** *** *** *** *** *** *** (t-stat) (50.37) (-1.86) (-2.92) (-10.75) (18.56) (5.57) (-6.72) (-2.40) (-3.95) (0.06) Included [21 variables] Included [11 variables] Included [123 variables] Included [123 variables] 53.86% 6,670 Information Asymmetry: Additional test • Distance may be less than perfect proxy for information asymmetry • If nonlocal investors informationally disadvantaged, prices in transactions between them should: – reflect smaller premiums than when they buy from locals – reflect smaller discounts than when they sell to locals • Test this....... Information Asymmetry: Additional test • Estimate “between nonlocals” effect, using first model ln(Price/SF) = Controls + βN·I{Nonlocal} + ε Now describes transaction type rather than investor • ......apply to pooled sub-sample (produced by propensity score matching ) 657 “between nonlocals” transactions matched with most similar “between locals” transactions Results: Nonlocal/Nonlocal vs. Local/Local transactions Variable Constant ln(Land area) ln(Building size) ln(Property age) Class A Class B Multi-tenant building Coefficient 7.607 *** -0.067 ** -0.081 ** -0.160 *** 0.382 *** 0.144 ** -0.119 *** (t-stat) (31.98) (-1.86) (-1.92) (-5.74) (2.80) (1.88) (-2.60) *** Nonlocal investors 0.063 (2.75) Buyer type indicators: Included [19 variables] Seller type indicators: Included [20 variables] Year indicators: Included [7 variables] Sale conditions: Included [62 variables] Market indicators: Included [85 variables] Adjusted R2: 61.21% Observations: 1,314 Overvalue by 6.3% when nonlocals buy from nonlocals (sell to nonlocals) Overpayment much smaller than when nonlocals buy from locals (13.8%) Discount, accepted when nonlocals sell to locals (7%), disappears Findings As compared to local investors, nonlocal investors........ Overpay on purchase by estimated 13.8%. Discount on sale by 7% Overpayment positively related to distance (information asymmetry) and rent differentials (anchoring) Discounting also increases with distance (information asymmetry) Pay smaller premiums when buying from other nonlocals and no discount when selling to other nonlocals