TESTING FOR MICRO‐EFFICIENCY IN THE HOUSING MARKET

• Date: 01 November 2018
• Author: André Kallåk Anundsen,Erling Røed Larsen
• DOI: http://doi.org/10.1111/iere.12332
• Published date: 01 November 2018

INTERNATIONAL ECONOMIC REVIEW

Vol. 59, No. 4, November 2018 DOI: 10.1111/iere.12332

TESTING FOR MICRO-EFFICIENCY IN THE HOUSING MARKET∗

BYANDR ´

EKALL ˚

AK ANUNDSEN AND ERLING RØED LARSEN1

Norges Bank, Norway; Eiendomsverdi, Norway, BI Norwegian Business School, Norway

Using highly granular transaction-level data for the Norwegian housing market over the period 2002–2014,

we investigate whether excessive prices persist or revert in repeat sales. Excessiveness in prices is detected by

comparing selling prices to predicted prices implied by a hedonic model, which includes a rich set of attributes.

Persistence is rejected and there is substantial reversion in excessive prices. Our results also show little scope for

profitable arbitrage by investing in apparently underpriced units. We suggest that excessive prices are related to

the stochastic arrival of interested purchasers at public showings, which we show is nonrepeatable.

1. INTRODUCTION

House price indices display time persistence. This has led several researchers to conclude

that returns contain predictable components. However, evidence based on aggregate indices

is only part of the story since the development of an index between two time periods reflects

movements in the aggregate, that is, between two scalars that each summarize thousands of

individual transaction prices. An index reveals little about relative prices, which are interesting

because economists believe markets coordinate and assimilate information through them, so

that people can differentiate between bargains and rip-offs. When people search for bargains

and seek to avoid rip-offs, the resulting prices incorporate these efforts, which, in turn, are

reflected in partial prices for housing attributes. This price-correcting capacity lies at the heart

of an efficient market. We are interested in how the housing market handles relative prices, and

this article asks one main question: When a house is sold at an excessively high or low price,

what happens to the price the next time the house is sold?

If there is persistence, a high first selling price relative to an expected price tends to be

followed by a high second price relative to an expected price. If there is no persistence but

reversion in the spread between selling and expected prices, an investor who paid more than

the expected price, whatever the reason, cannot expect to collect a similar premium upon selling

the unit. The return on his investment will be lower than the market return. Conversely, a buyer

who purchased at a price lower than the expected price can reasonably expect to sell at a price

that is closer to the expected price. Thus, the absence of persistence and presence of reversion

imply that the market punishes overpayments and rewards underpayments. At the same time,

if underpayments are rewarded, it could be possible to detect units that are underpriced ex ante

and make an ex post gain by investing in these units.

∗Manuscript received July 2016; revised January 2018.

1This article should not be reported as representing the views of Norges Bank. The views expressed are those

of the authors and do not necessarily reflect those of Norges Bank. We thank the editor, Holger Sieg, and three

anonymous reviewers for comments that have improved this article. The article was presented at the 2016 Annual

AEA Meeting, the 2016 NBRE Spring Meeting, the 2015 ENHR Workshop, and the 2015 Annual WEAI Conference.

We thank participants at research seminars at the University of Stavanger, Statistics Norway, and Norges Bank. We are

grateful to Farooq Akram, Benjamin Beckers, Lasse Eika, Saskia ter Ellen, Solveig Erlandsen, Sigurd Galaasen, Steffen

Grønneberg, Joe Gyourko, Mathias Hoffmann, Steinar Holden, Andreas Kostøl, Spencer Norman, Are Oust, Dagfinn

Rime, Asbjørn Rødseth, Dag Einar Sommervoll, Bernt Stigum, Kjetil Storesletten, Genaro Succarat, Leif Anders

Thorsrud, Paloma Taltavull de La Paz, and Robert Wassmer for stimulating comments and feedback that helped

improve this article. Please address correspondence to: Andr´

eKall˚

ak Anundsen, Norges Bank Research, Norges Bank,

Bankplassen 2, P.O. Box 1179 Sentrum, NO-0107 Oslo, Norway. E-mail: andre-kallak.anundsen@norgesbank.no.

2133

C

(2018) by the Economics Department of the University of Pennsylvania and the Osaka University Institute of Social

and Economic Research Association

2134 ANUNDSEN AND RØED LARSEN

Our exploration of housing market efficiency starts by documenting that Norwegian data

follow the international pattern of time persistence in aggregate house price indices. Exploiting

data on 469,127 transactions of owner-occupier units between 2002 and 2014, we do, however,

find that the housing market does not display evidence of micropersistence. To reach this

conclusion, we follow units across repeat sales. We detect a clear pattern. When the first selling

price is higher than the price prediction of a standard hedonic model, the price is much closer to

the model-predicted price when the unit is sold the next time. The only exception is when the

third selling price is higher than the hedonic model’s price prediction. Then, the second selling

price is also high. This demonstrates that the market discovers what the hedonic model does not,

namely, key omitted variables. In fact, using the ask price, which reflects the seller’s knowledge

of the unit (Benitez-Silva et al., 2015; Windsor et al., 2015), we find the same phenomenon.

Moreover, there is little sign of persistence when we consider a repeated cross-section model

in which we control for time-invariant unit fixed effects. Results are further strengthened when

studying a subsample, for which we have information on appraisal prices set by external and

independent inspectors. These appraisal prices allow us to control for time-varying attributes

of individual units, such as changes in the exterior or interior. Controlling for this, we reject

full persistence and find evidence of full reversion in excessive prices. Although our findings

suggest excess return predictability, we show that risk-adjusted excess returns from investing in

units that are underpriced relative to the hedonic model ex ante cannot be made. This leads to

the conclusion that the Norwegian housing market is micro-efficient.

To understand the mechanism generating reversion, we explore bidding-specific factors. For

this purpose, we acquired a unique data set on the auction process from 7,915 housing auctions,

containing the number of bids, the appraisal price, the selling price, a unit identifier, and the exact

transaction date. We find no statistically significant relationship between the number of bids the

second time a unit is sold and the number of bids the first time the unit is sold. This suggests that

the number of bids is unrelated to the unit and that a high number is nonrepeatable. However,

units that receive many bids experience a significant increase in the selling price relative to the

common value component. Thus, our results suggest that reversion is related to randomness on

the buy side.

Our contribution is threefold. First, we propose a simple framework to test for microper-

sistence in housing markets. Our framework builds on the persistence idea from macrotests.

In contrast to macrotests, our results show little micropersistence. Moreover, we find that it is

difficult to beat the market by systematically investing in units that are underpriced relative to

the price implied by a hedonic model. Thus, our findings support the notion that the Norwe-

gian housing market is semistrong efficient at the micro level. Second, we bring results from a

comprehensive data set. The data allow ultrafine time grids, since all transaction observations

are supplemented through real-time, same-day entries by realtors. Thus, we have access to the

actual sale date, that is, the date on which a bid is accepted, not the contract signature date or

the publicly registered date of title transfer. The data set also contains information on ask and

appraisal prices, in addition to a long list of attributes. Institutionally, Norway is a well-suited

country for studying micro- versus macropersistence, since Norwegian households transact

houses through speedy and transparent ascending-bid auctions after public showings on one or

two preannounced dates. In these auctions, the realtor mediates bids by phone or electronically

after potential buyers have volunteered their names, phone numbers, and e-mail addresses

upon visiting the showing of the unit. This institutional arrangement makes the transaction pro-

cess fast and transparent, almost a laboratory of housing auctions. As a third contribution, we

have acquired data on this auction process that allow us to investigate the mechanisms behind

reversion in excessive prices.

Our findings suggest that the housing market is an example of what Jung and Shiller (2005)

dub “Samuelson’s Dictum,” which ventures that the stock market is micro-efficient, but macro-

inefficient. The underlying idea of the dictum is that the stock market produces accurate and

unexploitable relative prices, but price levels that, to a certain extent, contain forecastable and

exploitable components. Our results indicate that the housing market may involve a similar