How Auctions Amplify House-Price Fluctuations

Transcription

1 How Auctions Amplify House-Price Fluctuations Alina Arefeva Johns Hopkins Carey Business School NASMES June,

2 The house price growth in the Los Angeles MSA percent, % months Notes: The monthly data on the house prices comes from Zillow. Zillow applies the Henderson Moving Average filter and then STL filter to produce the seasonally adjusted series of the house price growth, see 2

3 The house price growth in the Los Angeles MSA and benchmark model from the literature percent, % 2 0 percent, % months months (a) Los Angeles MSA data (b) Benchmark model Hypothesis: Nash bargaining price determination in the benchmark model has to be changed to auctions 3

4 Motivation House prices are volatile relative to fundamentals & fluctuate between booms and busts Buyers and sellers in the housing market search for each other Research idea The benchmark model of housing search produces less volatility than observed in the data This paper builds a search model that produces fluctuations between booms and busts and volatility as in the data Idea: volatility is produced by competition between buyers during the housing boom 4

5 Trading mechanism 5

6 A condo in Brookline, Massachusetts The Boston Globe, March 31, 2015: Three hundred people came through the open house, 25 made offers, and the bidding war lasted eight rounds and four days. 6

7 Auctions in the housing markets Those bidding wars oh, those bidding wars. When inventory is low, those bidding wars can escalate into a kamikazelike battle with 17 other buyers... Trulia, June 10, 2015 Bidding wars are no longer uncommon, particularly in affluent areas of northern New Jersey, Los Angeles, the San Francisco Bay area and Boston New York Times, August, 13, 1997 In San Francisco s Bidding Wars, Home Prices Go Ballistic Wall Street Journal, August 27,

8 Percent of offers, facing competition 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% Monthly data, not seasonally adjusted; Source: Redfin, the real estate brokerage firm (based in Seattle, 3% of market, serves 19 markets) 8

9 Contribution to the literature dynamic search model, accounts for option value to sell and buy Housing auctions: Han, Strange 2015, Handbook: The literature on real estate auctions is small, especially on the theoretical side Albrecht, Gautier, Vroman 2010, Albrecht Gautier Vroman 2014, Quan 2002: static,choice between auction and bargaining mechanisms, Genesove, Hansen 2014: empirical evidence on auction & bargaining prices, no search Housing search and matching: Wheaton 1990, Krainer 2001, Novy-Marx 2009, Piazzesi Schneider 2009, Anenberg, Bayer 2010, Caplin, Leahy 2010, Ngai Tenreyro 2014, Genesove Han 2012, Diaz, Jerez 2013, Burnside, Eichenbaum, Rebelo 2015, Moen, Nenov, Sniekers 2015, Piazzesi, Schneider, Stroebel 2015, Anenberg, Kung 2017 Competing mechanisms and search: Eeckhout Kirchker 2010, Shimer 1996, Shimer 2005, Moen

10 Setup of the models dynamic search model of the open city with search costs driving source of fluctuations: shocks to the inflow of buyers discrete time, interest rate r, discount factor β = 1/(1 + r) infinitely-lived risk neutral agents: buyers and sellers two goods: numeraire consumption and housing {0, 1} buyers rent at exogenous rental rate until they buy a house live in house forever after purchase provides services x 10

11 Random meetings buyers randomly choose one seller to visit per period tightness is # of active buyers per active seller number of buyers per seller Poisson with mean = tightness buyer draws match-specific value of housing services x match-specific, realized upon visit x iid over time buyer-seller pairs with cdf F (x) hazard rate f(x)/(1 F (x)) is weakly increasing constant after purchase 11

12 Price determination benchmark: Nash bargaining with random search: random meetings seller picks one buyer at random urn-ball meeting function Price splits joint surplus with weights α and 1 α α - bargaining power of seller auction with random search: random meeting seller considers all buyers runs an ascending bid auction with reservation price auction with directed search: seller posts the reservation price buyers observe this price and decide which sellers to visit seller considers all buyers runs an ascending bid auction with reservation price 12

13 Intuition price determination mechanism becomes important because house values are heterogenous with heterogenous values, the method of choosing the buyer among interested buyer becomes important for quantitative properties of the model Nash bargaining: buyer is chosen randomly among interested buyers average value determine prices auction: buyer is chosen by the maximum bid among interested buyers second highest value determines prices during the booms, the highest value increases more than the average value, making the sales prices more volatile 13

14 Amplification of shocks in the auction models Dependance of the prices on the probabilities of sale for three price determination mechanisms , nash bargaining with random search, auctions with random search, -, auctions with directed search , probability of sale 14

15 House price growth from simulations 8 random auction directed auction random Nash bargaining percent, % months Notes: This figure shows the house price growth from the simulation of the models, where random auction is the auction model with random search, directed search is the auction model with directed search, random Nash bargaining is the Nash bargaining with random search. In each model the distribution of house values x Exp, and each model is hit by the same series of shocks. The line data shows the house price growth from Zillow Los Angeles MSA. There is no correspondence between the month and actual events, this figure is for only for purposes of comparing the volatility. 15

16 House price growth from data and simulations 8 8 random auction directed auction random Nash bargaining percent, % 2 0 percent, % months months (a) Los Angeles MSA data (b) Models 16

17 Auction house prices are more volatile than Nash bargaining house prices Data RA DA/SP NB σ logp monthly ρ logp monthly σ logp quarterly, last ρ logp quarterly, last σ logp quarterly, average ρ logp quarterly, average σ logp annual, last ρ logp annual, last σ logp annual, average ρ logp, annual, average Notes: This table shows the moments based on Zillow house price growth data in column Data, average moments from 100 simulations of the random/directed auction model in column RA / DA/SP and random Nash bargaining model in column NB. The last column σ logp, ρ logp stand for standard deviation/autocorrelation of the change in log prices. x Exp 17

18 Conclusion quantify the importance of the price determination mechanism for residential real estate market build a tractable dynamic search model of the housing market with micro-founded price determination mechanism - auctions a model with auctions produces high volatility, observed in the data, improving on the benchmark model from the literature 18

19 Appendix 19

20 House prices depending on the state of the housing market Price The number of sellers, in thousands The number of buyers, in thousands 2000 House prices are in thousands of dollars. The graph shows the surface of the house prices in the auction model with random search, depending on the current state: the number of buyers, the number of sellers for a given influx of buyers. 20