Paper presented at the ERES 2009 conference Stockholm

Paper presented at the ERES 2009 conference, Stockholm June 24 -27 2009 Improved price index for condominiums by Han-Suck Songa and Mats Wilhelmssonb a) Department of Real Estate Economics, Royal Institute of Technology (KTH), Stockholm, Sweden han-suck. song@abe. kth. se b) Centre for Banking and Finance (CEFIN), Royal Institute of Technology (KTH), Stockholm, Sweden mats. wilhelmsson@abe. kth. se

Who presents the correct price information? DN: March 17, 2009 DN: March 12, 2009 2

Why better apartment price index? – Economic indicator. – ”Underlying asset” in derivatives and insurance contracts. – More… 3

Construction of real estate price indexes Example of Index construction methods: • Average prices • Repeated-Sales (Case & Shiller) • Hedonic price index (different alternatives) • Date of sale: Contract date or deed date? 19 mars 2018 Confidential - Valueguard AB © 2009 All rights reserved 4

Date of Sale: contract date vs. deed date On average 55 days Open house Advertisement Deed date Contract date KTH/Valueguard, Mäklarstatistik Statistic Sweden S&P 5

Case – Stockholm apartment (condominium) market Variable Definition Price Living area Rooms Fee Balc First Top Byear 1 Byear 2 Byear 3 Byear 4 Byear 5 Byear 6 New SEK Square meters Number of rooms Monthly fee: SEK Dummy: Balcony Dummy: First floor Dummy: Top floor Dummy: Before 1900 Dummy: 1900 -1939 Dummy: 1940 -1959 Dummy: 1960 -1975 Dummy: 1976 -1990 Dummy: After 1990 Dummy: New building 2168827 60. 96 2. 25 2976. 62 0. 1200 0. 1986 0. 2973 0. 0580 0. 3812 0. 2321 0. 0699 0. 0496 0. 2092 0. 0175 1287835 25. 62 1. 00 1268. 56 0. 3249 0. 3990 0. 4571 0. 2337 0. 4857 0. 4222 0. 2550 0. 2171 0. 4067 0. 1310 256250 14 1 1 0 0 0 0 0 9975000 225 9 11966 1 1 1 1 1 Elev Distance Dummy: Elevator Distance: Meters to city 0. 5071 3882. 35 0. 4924 2727. 43 0 272. 62 1 12987. 20 NE NW SW No. of observations Dummy: Northeast Dummy: Northwest Dummy: Southwest 0. 1998 0. 3151 0. 1786 32, 380 0. 3999 0. 4646 0. 3830 0 1 1 1 January 2005 to January 2009 Average Standard deviation Min Max 6

X and Y Coordinates • • Distance to the “middle” Direction – Quadrants NW NE SW SE 7

Estimation procedure I Single hedonic price equtation Vs. Multiple hedonic price equation (Moving Windows Regression)

Moving Window Regression 9

Estimation procedure II Spatial hedonic price index Vs. Multiple hedonic price equation (Moving Windows Regression) – There is a presence of spatial dependency in our hedonic models. – However this seems not to spill over to the price index.

ols sar /0 7 M ay /0 7 Ju l/0 7 Se p/ 07 No v/ 07 Ja n/ 08 M ar /0 8 M ay /0 8 Ju l/0 8 Se p/ 08 No v/ 08 Ja n/ 09 7 06 n/ 0 M ar Ja 6 06 v/ No p/ Se 6 6 /0 l/0 Ju M ay /0 6 05 n/ 0 M ar Ja 5 05 v/ No p/ Se 5 5 /0 l/0 Ju M ay /0 5 n/ 0 M ar Ja Spatial hedonic models 170 160 150 140 130 120 110 100 sem 11

Result: Single hedonic price equation Variable Constant Coefficient t-value VIF -136. 1965 -7. 59 Living area 0. 9035 89. 27 6. 17 Rooms 0. 1243 25. 15 4. 69 Fee -0. 2045 -16. 62 3. 12 Balc 0. 0122 3. 07 1. 46 First -0. 0047 -1. 35 1. 66 Top 0. 0228 7. 09 1. 79 Byear 1 0. 0591 11. 48 1. 60 Byear 2 0. 0186 5. 09 2. 57 Byear 3 -0. 0140 -3. 22 2. 29 Byear 4 -0. 1398 -21. 33 1. 46 Byear 5 -0. 1413 -17. 38 1. 47 New 0. 0222 2. 52 1. 13 Elev -0. 0117 -3. 71 1. 96 Dist -0. 2887 -28. 30 37. 47 Dist*NE -0. 0178 -1. 91 1055. 74 Dist*NW 0. 0839 5. 59 2390. 66 Dist*SW 0. 0525 3. 75 1320. 92 NE 0. 1719 2. 49 925. 99 NW -0. 4283 -3. 73 2180. 63 SW -0. 2462 -2. 26 1257. 40 No. of observations . 8780 • Reasonable magnitude • Statistical significant • High explanatory power 31390 R 2 -adj • Expected signs 12

Average price vs. Hedonic Index 13

Date of Sale: Contract date vs. Deed date The difference in index as compared to figures above is due to different samples; not all observations with deed date have contract dates, and vice versa. 14

Conclusion – Date of sale important (contract date vs. deed date). – Among the many alternatives to estimate an apartment price index, estimating a single hedonic price eqationn seems sufficient. *Important to test for parameter heterogenity and spatial dependency. – Further work to improve quality of apartment price index: *Most important is to improve input of quality of inputs and number of attributes.

Specification Dependent variables: ln(price) Independent variables: – – – – ln(area), ln(room), ln(fee) First floor, top floor, floor level, no. of floors Elevator, balcony 6 periods ln(distance), geographical direction Administrative parish … Time dummies

Periods Six periods – – – Before 1900 -1940 -1960 -1975 -1990 After 1990 1800 and older Before World War II After post-war ”Million programe” Construction subsidies Abolishment of the subsidy system

Specification Dependent variables: ln(price) Independent variables: – – – – ln(area), ln(room), ln(fee) First floor, top floor, floor level, no. of floors Elevator, balcony 6 periods ln(distance), geographical direction Administrative parish … Time dummies