Modeling residential location choice and real estate prices with a bid-auction approach

Transcription

1 Modeling residenial locaion choice and real esae prices wih a bid-aucion approach Ricardo Hurubia Francisco Marínez Michel Bierlaire Sevenh Workshop on Discree Choice Models, Augus 25-27, 2011 Lausanne, Swizerland

2 Ouline 1. Moivaion 2. Bid approach for locaion choice 3. Proposed mehod 4. Brussels case sudy 5. Discussion 6. Furher research

3 Moivaion Evoluion of land use (locaion choice) models: Aggregaed Disaggregaed Equilibrium Dynamic microsimulaion Marke clearing / locaion disribuion: Bid-aucion Choice Bid approach: consisen wih economic heory. Usually implemened in equilibrium models (e.g. MUSSA) Choice approach: easier o implemen in a microsimulaion conex (e.g. UrbanSim). Requires hedonic rens/prices

4 Moivaion Bid-aucion approach applied o microsimulaion Price formaion problem consisency wih observed prices Reacion o marke condiions Dynamics (pseudo-equilibrium) Acive bidders in he aucion (choice se)

5 Bid approach for locaion choice Assumpions: Real esae goods (locaions) are raded in aucions Agens bid heir willingness o pay for each locaion (B hi ) For each locaion he bes bidder is seleced The amoun/value of he bes bid deermines he ren/price

6 Bid approach for locaion choice Probabiliy of agen h being he bes bidder for locaion i: h / i g H Expeced maximum bid (ren): P = exp( µ Bhi ) exp( µ B 1 r = i ln exp( µ B gi ) µ g H gi )

7 Bid approach for locaion choice Problems: Requires equilibrium beween supply and demand (or a leas demand > supply) In he case of supply surplus i no clear which locaions are no seleced Logsum ( ) doesn necessarily reproduce observed prices or rens

8 Proposed framework Bid based locaion choice model Assumpions: Goods (locaions) raded in aucions, period-wise Agens bid heir willingness o pay for each locaion Agens adjus he level of heir bids as a reacion o marke condiions (represened by observed prices) Agens are myopic regarding he oucome of fuure and presen aucions

9 Proposed framework Bid funcion: B = b + hi h b hi Bid adjusmen (uiliy level) Willingness o pay for aribues b hi (,, β ) 1 = f z x esimaed via max log-likelihood, assuming = 0 i h b h

10 Proposed framework Bid adjusmen: Bidding households aemp o ensure winning, on average, a leas one aucion: i P exp( bh + b exp( B ) ) / = hi h i = i gi g H 1 * h Bu households do no observe bids of oher households in he same period. They can only observe ransacion prices in previous periods * µ = 1 H : full choicese

11 Proposed framework Bid adjusmen: i S P exp( bh + b exp( B ) ) / = hi h i = 1 i S gi g H 1 g H S: full choice se of dwellings/locaions exp( B gi ) = exp( r 1 1 i )

12 Proposed framework In each period: b h ( = 1 ln exp bhi ri ) i S In he base year (calibraion year): b 0 h ( = 0 * ln exp bhi ri ) i S Observed prices a he base year

13 Price dynamics Simulaion of a supply surplus scenario wih synheic daa Supply shock (increase)

14 Brussels case sudy Daa colleced for he SusainCiy projec: Census 2000 (aggregaed daa by zone) Household survey 2000 (disaggregaed daa, ~1000 obs) average ransacion prices by commune and dwelling ype households, dwellings 157 communes 4975 zones 4 ypes of dwelling Deached houses Semi-deached houses Aached houses

15 Brussels case sudy Bid funcion specificaion:

16 Brussels case sudy Esimaion resuls wih PyhonBiogeme Likelihood raio es agains null model 219.4

17 Brussels case sudy Number of people by commune

18 Brussels case sudy Number of people wih universiy degree by commune

19 Brussels case sudy Logsums for each locaion

20 Brussels case sudy Logsum for each locaion afer adjusmen of b h

21 Discussion Framework allows for supply or demand surplus Changes in (aggregae) marke condiions are capured in he price Adjusmen of b h produces maximum expec bids close o observed prices Scale of prices Arbirary? (posiive or negaive b h ) Esimaion of µ? Should bid s be also adjused locaion-wise (b i )? Relaive imporance of b hi? (re-esimaion of beas?)

22 Furher research Acive bidders (choice se generaion) Price is affeced by who is compeing for he locaion Choice se generaion or imporance sampling? Relevance of he scale of he logsum Locaion assignmen Mone Carlo simulaion following max bid probabiliies? Simulaneous locaion assignmen?

23 Thanks

24 Choice approach for locaion choice Assumpions: Each agen selecs he locaion ha provides maximum uiliy Agens are price akers Prices (usually) defined as funcion of he locaion aribues

25 Choice approach for locaion choice Assumpion: consumer surplus is a proxy of uiliy: Probabiliy of locaion i providing maximum uiliy o agen h:

26 Choice approach for locaion choice Problems: Price-aker assumpion (no good for quasi-unique goods) Marke condiions usually no capured by hedonic rens Advanages: If prices are he oucome of an aucion, he locaion disribuion is he same for he bid and choice approaches