The Other Side of Eight Mile * Suburban Housing Supply Allen C. Goodman Wayne State University September 2004 Presented at AREUEA Meetings, Philadelphia PA January 2005 Housing Supply • Estimates have been all over the map. • Depends on whether it is new housing or existing housing. • For central cities stock, Goodman (2004) finds: – +0 to +0.10 in negative direction – about +1.00 in the positive direction Value Positive Direction More Elastic Vo Negative Direction Less Elastic Qo Quantity Direct Estimates of Change Populationt = (Dwel. Units)t (Occupancy Rate)t (HH Size/Occupied Dwel. Unit)t and: Pt = Ut Ot St Populationt+1 = (Dwel. Units)t+1 (Occupancy Rate)t+1 (HH Size/Occupied Dwel. Unit)t+1 Pt+1 = Ut+1 Ot+1 St+1 Population = Pt+1 - Pt = and: U O ( S t 1 S t ) S O (U t 1 U t ) U S (Ot 1 Ot ) % Population = Pt 1 Pt S t 1 S t U t 1 U t Ot 1 Ot P S U O Supply and Demand Model Housing Services Demand: ln QtD a ln Yt b ln Rt d ln N t e tD (3) Supply of Housing Stock: ln QtS g ln Vt h k Gtk e tS (4) k Product Market Equilibrium ln QtS ln QtD (5) Capital Market Equilibrium ln Rt ln Vt ln r t (6) Solving for Q and V yields: ln Vt a g b ln Yt b g b ln r t d g b ln N t ln Vt J1 ln Yt J 2 ln r t J3 ln N t Jk Gtk k ln Qt g ln Vt h k Gtk . k k hk g b Gtk ,or (7) (7´) (8) Instrument for user cost r D %r s %rc 0 s r s c rc k Gk (10) k • This follows the expectations implicit in value-rent ratios. An initially high rs (low suburban value/rent ratio) would be expected to predict a decrease (s < 0) in D. • Similarly an initially high central city rc would predict a central city user cost decrease relative to the CC, or a rise (c > 0) through the decade in D. • Predicted value from equation (10) is then used as an alternative measure of user cost in the supply-demand regressions Instrumental Estimate – Equation 10 1970s 1980s 1990s -0.0629 0.2471 0.0764 0.0520 0.0499 0.0370 -61.4445 -209.9906 -156.9625 7.3276 9.8411 10.7593 36.8553 179.6729 110.7284 6.4661 6.5060 14.1687 0.0492 -0.0679 0.1622 0.0224 0.0223 0.0276 -0.0342 -0.0770 0.1117 0.0220 0.0212 0.0289 -0.0320 -0.0763 0.1468 0.0245 0.0225 0.0289 0.0885 -0.1092 0.1267 0.0232 0.0269 0.0290 0.1275 0.1266 0.1554 0.3330 0.7593 0.6387 Dependent Var: Pct. rs - Pct. rc Constant Initial Suburban rs Initial Central City rc South Midwest Southwest Mountain/West SER R2 Table 6 1970-1980 Instruments for r Demand Supply Elasticities Variable Coefficient Std. Error. Constant 0.2488 0.0151 16.53 -0.0961 0.0499 -1.93 % Sub Income 0.0200 0.0165 1.21 % Metro Pop 0.6993 0.0584 11.97 Std. Error 0.1488 % Sub r Variable Coefficient Std. Error. Constant -0.1424 0.0500 Pct. Sub Value 1.3662 Std. Error 0.2238 Supply 1.3662 Demand Price Demand Income Demand Pop -0.1453 0.0302 1.0225 t-ratio t-ratio -2.85 0.1310 10.43 Three Decade Means Three Decades – 3SLS Estimators Mean 1.2585 -0.0547 Median 1.3662 -0.0697 Demand Income 0.1311 0.1280 Demand Pop 0.9893 1.0225 Supply Price Demand Price Regional Supply Elasticity Estimates B. Regions with Shift Terms Number Northeast/North Central South/Southwest/ MW Column Weighted Mean 144 173 19701980 19801990 19902000 Row Mean Row Median 1.5983 0.6252 0.4468 0.8901 0.6252 0.3572 0.1113 0.2651 1.7872 1.5352 2.2663 1.8629 1.7872 0.3645 0.2863 0.7083 1.7014 1.1218 1.4398 1.4210 1.2594 Metropolitan Elasticities Conclusions • Direct method to estimate housing stock elasticity. • Results are plausible. – – – – – Elasticity (Central City – decreasing) Elasticity (Central City – increasing) Elasticity (Suburbs) Northeast quadrant Other regions +0.0 - +0.1 +1.0 - +1.1 +1.3 - +1.5 approx. +0.9 approx. +1.9. • Further directions – Compare older and newer suburbs. – Decompose changes in values into changes in quantities and changes in prices Where is the Speculative Bubble in US House Prices? Allen C. Goodman – Wayne State University Thomas G. Thibodeau – University of Colorado AREUEA Meetings – Chicago January 2007 © A.C. Goodman, T. Thibodeau, 2007 Questions to Address • How much real appreciation in house prices is justified by the economic fundamentals of local housing markets? • How much real appreciation is attributable to speculation?’ © A.C. Goodman, T. Thibodeau, 2007 What’s Our Approach? 1. We examine real house price appreciation using a simple simulation of long-run housing market behavior. The simulation model demonstrates that the key explanation for the observed spatial variation in house price appreciation rates is spatial variation in supply elasticities. 2. The empirical model of the paper attempts to estimate supply elasticities for 133 metropolitan areas across the US. We then use the estimated elasticities to estimate how much of each metropolitan area’s appreciation can be attributed to economic fundamentals and, by inference, how much is attributable to speculation. © A.C. Goodman, T. Thibodeau, 2007 Simulation Model – 2 Questions • Over the 2000-2005 period what shift in aggregate demand was required for owneroccupied housing to observe a 12.7% increase in the number of owner-occupied housing units in the US over this period? • What was the corresponding increase in the equilibrium house price? © A.C. Goodman, T. Thibodeau, 2007 Evaluate Supply and Demand Shifts • What shifts must occur for quantity to increase by 12.7%? P D S Po Especially when it is clear that the Supply curve is indicating higher costs © A.C. Goodman, T. Thibodeau, 2007 Qo Qox 1.127 Q Table 1: Increases in Real House Prices Necessary to Achieve 12.7% Increase in the Number of Owner-Occupied Housing Units for Alternative Housing Supply Elasticities (ED = -0.8) ES 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.5 2.0 5.0 10.0 Demand Shift Quantity Price 63.50% 127.00% 35.28% 63.50% 25.87% 42.33% 21.17% 31.75% 18.34% 25.40% 16.46% 21.17% 15.12% 18.14% 14.11% 15.88% 13.33% 14.11% 12.70% 12.70% 10.82% 8.47% 9.88% 6.35% 8.18% 2.54% 7.62% 1.27% D+S Shift Price 151.00% 87.50% 66.33% 55.75% 49.40% 45.17% 42.14% 39.88% 38.11% 36.70% 32.47% 30.35% 26.54% 25.27% © A.C. Goodman, T. Thibodeau, 2007 Empirical Model Demand for Housing Units: ln QtD a ln Yt b ln Rt d ln H t e tD Supply of Housing Units: ln Q g ln Vt h j G jt e tS jJ S t j 1 Capital Market Equilibrium: User Cost: Rt Vt [i d tr E{ p}] Product Market Equilibrium ln QtS ln QtD © A.C. Goodman, T. Thibodeau, 2007 ln Rt ln Vt ln r t Data • HUD’s State of the Cities Database augmented by, • Location (latitude and longitude) obtained from the 1990 Census; • Metropolitan area construction costs from RS Means; • Agricultural land prices obtained from the US Department of Agriculture; • BLS data on the CPI. © A.C. Goodman, T. Thibodeau, 2007 Table 2: Descriptive Measures Variable Place Information Central City Dummy Density/square kilometer Distance to CBD (in kilometers) Number of Places in MSA Number of gov’t per capita Name N Mean Std Dev CC density distance nplaces Numgov 9180 9180 9180 9180 9180 5.90% 974 27.92 83.21 0.0243 23.57% 1283 42.48 83.78 0.0425 Decadal Changes Change in Population Change in Total Units Change in Occupied Units Change in Owner-occupied Units Change in Occupancy Rate Change in Household Size Change in Minority Households Change in Median Rents Change in Median Values Change in Median Incomes Change in User Cost popch totunch occunch ownoccch occratch hhsizech minoritych medrntch medvalch medincch rhoch 9180 9180 9180 9179 9180 9175 9180 9150 9146 9179 9117 12.36% 13.90% 14.67% 16.35% 0.81% -2.33% 0.41% 0.59% 5.01% 4.96% -7.36% 24.22% 22.78% 23.33% 26.59% 4.80% 6.65% 0.57% 15.79% 23.27% 12.94% 22.17% Table 5 - Elasticities Within and Among Metropolitan Areas Mean Median Supply Price (all) Supply Price (+ only) Supply Price (neg. set to 0) 0.3457 0.6181 0.3050 0.5960 0.4508 0.3050 Demand Price Demand Income -0.4430 0.3559 -0.4030 0.3237 Pct. Pct Correct Significant Sign 10% Sig. Within Metropolitan Areas 71.40% 63.2 Among Metropolitan Areas Supply Price 0.3457 Demand Price Demand Income -0.2193 0.4250 © A.C. Goodman, T. Thibodeau, 2007 Prices HIGHER than Expected © A.C. Goodman, T. Thibodeau, 2007 Expected nominal appreciation Fort Myers Sacramento Riverside San Diego Orange Los Angeles Monmouth NJ Oakland Las Vegas Santa Rosa Atlantic City Washington DC Fresno Nassau-Suffolk Orlando Tampa Phoenix Middlesex NJ Miami Poughkeepsie Honolulu CDP Baltimore Newburgh 54.19% 57.64% 66.21% 53.56% 62.73% 73.20% 68.16% 66.32% 49.95% 63.48% 59.29% 78.30% 100.98% 66.21% 58.56% 66.21% 59.27% 67.73% 102.00% 69.23% 66.21% 66.21% 68.11% Observed appreciation Observed expected 151.69% 154.17% 160.76% 147.72% 149.66% 151.32% 135.94% 133.27% 115.31% 127.68% 118.04% 136.49% 155.68% 118.90% 110.29% 113.37% 106.41% 114.71% 146.01% 111.73% 108.37% 107.49% 106.38% 97.49% 96.53% 94.55% 94.16% 86.93% 78.13% 67.78% 66.96% 65.36% 64.20% 58.76% 58.19% 54.70% 52.69% 51.73% 47.16% 47.14% 46.98% 44.01% 42.50% 42.16% 41.28% 38.28% Prices LOWER than Expected © A.C. Goodman, T. Thibodeau, 2007 Exp nominal appreciation Seattle Madison Syracuse Austin Nashville-Davidson Portland OR Houston Birmingham McAllen Dallas Memphis Kansas City Springfield MA Raleigh Lancaster Rochester NY Chicago Columbus OH Ann Arbor Charlotte Hartford Greensboro Denver Fort Worth Salt Lake City Fort Wayne Dayton Rockford Appleton Indianapolis Atlanta Bergen-Passaic Tacoma Providence Omaha Louisville Detroit 83.74% 70.88% 66.21% 58.90% 58.69% 87.31% 59.17% 66.21% 57.01% 60.07% 54.59% 70.28% 114.09% 56.80% 84.84% 66.21% 100.41% 69.10% 74.68% 66.21% 111.97% 70.47% 90.78% 76.25% 92.62% 79.36% 82.17% 94.57% 100.22% 93.60% 115.59% 203.03% 187.17% 245.43% 157.32% 247.92% 286.22% Observed Observed appreciation Expected 63.46% 49.64% 43.96% 33.03% 31.76% 59.52% 31.12% 36.22% 24.89% 27.44% 21.57% 37.22% 80.59% 22.37% 48.84% 28.05% 61.42% 29.73% 34.67% 25.01% 68.81% 23.12% 41.68% 26.98% 33.38% 19.83% 22.10% 32.42% 34.84% 24.41% 35.99% 97.67% 73.24% 117.93% 29.26% 30.46% 29.47% -20.28% -21.24% -22.25% -25.87% -26.93% -27.79% -28.05% -29.99% -32.12% -32.63% -33.02% -33.06% -33.50% -34.43% -36.00% -38.17% -38.99% -39.37% -40.01% -41.20% -43.15% -47.36% -49.09% -49.27% -59.24% -59.52% -60.07% -62.15% -65.37% -69.18% -79.60% -105.36% -113.93% -127.50% -128.07% -217.46% -256.74% Conclusions – 1 • We attempt to identify how much of the recent appreciation in house prices can be attributable to economic fundamentals and how much can be attributed to speculation. • After reviewing the relevant literature, we investigate the relationship between house price appreciation rates and supply elasticities using a simulation model of the housing market. • The model illustrates that the expected rate of appreciation in house prices is very sensitive to the assumed supply elasticity. © A.C. Goodman, T. Thibodeau, 2007 Conclusions – 2 • We then produce estimates of metropolitan area supply elasticities using cross-sectional place data obtained from HUD’s State of the Cities Data System. • Our empirical analyses yield statistically significant supply elasticities for 84 MSAs. We then compute expected rates of appreciation for these places and compare the expected appreciation rates to the rates observed over the 2000-2005 period. • We find that speculation has driven house prices well above levels that can be justified by economic fundamentals in less than half of the areas examined. © A.C. Goodman, T. Thibodeau, 2007 Conclusions – 3 • Establishing “20% over the expected increase” as a housing bubble threshold, we find that only 23 of the 84 metropolitan areas with positive supply elasticities exceed this threshold. • Moreover, with the exception of Las Vegas, Phoenix, and Honolulu, every single one of these areas is either within 50 miles of the Atlantic coast or California’s Pacific coast. • This suggests that extreme speculative activity, so prominently publicized, has been extraordinarily localized. © A.C. Goodman, T. Thibodeau, 2007