INFORMATION FRICTIONS AND HOUSING MARKET DYNAMICS

• DOI: http://doi.org/10.1111/iere.12204
• Author: Elliot Anenberg
• Published date: 01 November 2016
• Date: 01 November 2016

INTERNATIONAL ECONOMIC REVIEW

Vol. 57, No. 4, November 2016

INFORMATION FRICTIONS AND HOUSING MARKET DYNAMICS∗

BYELLIOT ANENBERG1

Federal Reserve Board, U.S.A.

I examine the effects of seller uncertainty over their home value on the housing market. Using evidence from

home listings and transactions data, I first show that sellers do not have full information about current period demand

conditions for their homes. I incorporate this type of uncertainty into a dynamic microsearch model of the home

selling problem with Bayesian learning. The estimated model highlights how information frictions help to explain

the microdecisions of sellers and how these microdecisions affect aggregate market dynamics. The model generates a

significant microfounded momentum effect in short-run aggregate price appreciation rates.

1. INTRODUCTION

Since the seminal work of Stigler (1961), economists have long recognized the importance of

imperfect information in explaining the workings of a variety of markets. Surprisingly, given its

importance to the macroeconomy, little work has focused on the effects of imperfect information

in the housing market.2The housing market is a classic example of a market affected by imperfect

information. Each house is a unique, differentiated asset; trading volume of comparable homes

tends to be thin due to high transaction costs; and market conditions are volatile over time.

These features of the housing market make it difficult for sellers to determine their home values

at any point in time.

In this article, I model the effect of this type of seller uncertainty on the housing market. The

model adds a framework for seller uncertainty and Bayesian learning in the spirit of Lazear

(1986) to the typical features of the dynamic microsearch models in the housing literature

(Salant, 1991; Horowitz, 1992; Carrillo, 2012). I estimate the model and use it to test whether

uncertainty is important for explaining several key stylized facts about housing market dy-

namics that have attracted much attention in the literature, in part, because some of them are

inconsistent with the predictions of standard asset pricing models.

One key fact is that price appreciation rates display predictability in the short run. In their

seminal papers, Case and Shiller (1989) and Cutler et al. (1991) find that a 1% increase in real

annual house prices is associated with a 0.2% increase the next year, adjusting for changes in the

nominal interest rate.3Figure 1 illustrates the persistence in house prices during the recent U.S.

recession, which, as I describe below, is the time period that my sample covers. Whereas the

∗Manuscript received July 2013; revised October 2014.

1This is a revised version of my job market paper. I am very grateful to my advisor, Pat Bayer, and committee

members Jimmy Roberts, Andrew Sweeting, and Chris Timmins for comments. I also thank Peter Arcidiacono,

Ed Kung, Jon James, Steve Laufer, Robert McMillan, Guido Menzio, Karen Pence, and Jessica Stahl. An earlier

version of this article was circulated under the title “Uncertainty, Learning, and the Value of Information in the

Residential Real Estate Market.” The analysis and conclusions set forth are those of the author and do not indicate

concurrence by other members of the research staff or the Board of Governors. Please address correspondence to:

Elliot Anenberg, Federal Reserve Board, 20th and C St. NW, Washington, DC 20551, U.S.A. Phone: 202-452-2581.

E-mail: elliot.anenberg@frb.gov.

2Levitt and Syverson (2008) and Taylor (1999) are examples of studies that focus on the effect of information

asymmetries on microfeatures of the data, but less is understood about the broader effects of information frictions on

housing market dynamics.

3Numerous other studies have also documented this persistence. See Cho (1996) for a survey of the literature on

house price dynamics.

1449

C

(2016) by the Economics Department of the University of Pennsylvania and the Osaka University Institute of Social

and Economic Research Association

1450 ANENBERG

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69

Months From Peak

Monthly Case Shiller House Price Index (Peaked in Apr 2006)

Monthly S&P 500 Stock Price Index (Peaked in Oct 2007)

NOTES: Stock price index is a three-month moving average to be consistent with the construction of the house price

index. Both house and stock prices are normalized to one in their peak month.

FIGURE 1

HOUSE PRICES AND STOCK PRICES DURING THE RECENT U.S.RECESSION

stock market took less than two years to reach bottom, house prices fell at a relatively slower

pace and consistently for over half a decade.4This persistence in house price changes, which

occurs throughout the house price cycle, has led some to question the efficiency of the housing

market because it cannot be explained by fundamentals (e.g., Case and Shiller, 1989; Glaeser

et al., 2014). Thus, an important question is whether the amount of short-run momentum found

in the data is consistent with a rational model of the housing market.

At a more microlevel, the literature has also documented a set of stylized facts about the

behavior of individual sellers (Merlo and Ortalo-Magne, 2004).5For example, sellers tend to

adjust their list prices downward, even when market conditions do not change, and sales prices

for observationally equivalent homes depend on time on market (TOM). These empirical

patterns are inconsistent with the predictions of existing search, matching, and bargaining

models of housing transactions, which are stationary models and thus do not accommodate

duration dependence in seller behavior (see Horowitz, 1992; Chen and Rosenthal, 1996; Novy-

Marx, 2009; Carrillo, 2012). In addition to explaining several macro stylized facts, I will show that

uncertainty and the gradual acquisition of information during the listing period is an explanation

for these and many other dynamic features of the microdata.

The existing literature has modeled the home selling problem in a stationary framework in

part because existing microdata sets on home listings and transactions are limited.6In order to

identify the parameters of a model where optimal seller strategies vary over the selling horizon

as learning occurs, I compile a new microdata set with more detail on the dynamic decisions of

sellers. For close to the universe of single-family homes listed for sale with a realtor in the two

major California metropolitan areas from 2007 to 2009, the combined data set describes the

precise location of each home listed for sale, the list prices each week that the home is listed for

sale, TOM, and sales prices (for listings that result in sales).

4The house price declines were interrupted briefly by the effects of the Obama administration’s first-time homebuyer

tax credit.

5Descriptions and models of buyer behavior are rare because available data sets typically summarize data on seller

behavior (e.g., list price choice, how long to keep the home on the market, etc.).

6A notable exception is contemporaneous work by Merlo et al. (2015), which I discuss in further detail in Section

4.1.