Behavioral Beijing

8/10/2019 Behavioral Beijing

1/37

- 1 -

Christopher K. Hsee

Jean-Pierre Dub

Yan Zhang

*Christopher K. Hsee is Theodore O. Yntema professor of behavioral science and marketing at

the Graduate School of Business, University of Chicago, 5807 South Woodlawn Ave., Chicago

IL 60637, Email: [email protected]. Jean-Pierre Dub is a professor of marketing

and Neubauer Faculty Fellow at the Graduate School of Business, University of Chicago, 5807

South Woodlawn Ave., Chicago, IL 60637, Email:[email protected]. Yan Zhang is a

Ph.D candidate at the Graduate School of Business, University of Chicago, 5807 South

Woodlawn Ave., Chicago IL 60637, Email: [email protected].


2/37

- 2 -

A Behavioral Analysis of Shanghai Real Estate Prices

Abstract

In a field study, we identified an intriguing inconsistency between real estate developers desired

sales pattern (selling all apartments at the same rate) and the actual sales pattern (selling good

apartments faster). We explained this inconsistency with Tversky, Sattath and Slovic (1988)s

prominence principle, according to which buyers, who were in a choice mode, weighed the

desirability of floors more heavily than developers, who were in a matching mode when setting

prices. We corroborated our explanation with controlled experiments using potential Shanghai

homebuyers and actual Shanghai real estate developers as participants. To the best of our

knowledge, this is the first research to relate the prominence principle to a real-world issue in

one of the worlds most active markets.


3/37

- 3 -

A Behavioral Analysis of Shanghai Real Estate Prices

Since the new millennium, China has emerged as one of the worlds most formidable

consumer markets. In 2004, the size of the consumer market was estimated at 36 million urban

Chinese and, every year, roughly 20 million Chinese turn 18 (Grant, 2006). The impact of this

consumer boom has been particularly apparent in the Chinese real estate market, one of the most

active real estate markets in the world. During the late 1990s, the Chinese government enacted

several radical changes in the regulation of state-provided housing to stimulate economic

growth. In 1999, the state lifted a ban on the resale of privatized housing. These measures led to

an explosion in the Chinese housing market, which has subsequently been estimated to have

been contributeding 1.5% of annual GDP growth (Economist 2001). Rapidly rising housing

pricesin Shanghai, prices have nearly doubled between 2000 and 2004 (Economist 2005)

have fueled billions of dollars in housing construction. Since 1996, 67,333 residential properties

were sold in Shanghai. In 2000 alone, about 7.5 million square meters (80.73 million square feet)

of existing houses were sold in Shanghai, with a total transaction value of roughly RMB 65.6

billion (about U.S. $8 billion) (Ye, 2004). At a more micro level, real estate developers are

increasingly relying on careful pricing practices to avoid missing profitable opportunities in this

fast-paced, but otherwise relatively young housing market.

We begin with a field study that involves real estate developers in Shanghai. From the

study, we learn that most developers strategically set their prices to obtain a sales pattern

whereby all their units sell at the same rate over time. Yet despite their best efforts to price

apartments commensurate to their perceived quality by consumers, in practice sellers routinely

find themselves stuck with the less-desirable units selling at a much slower pace. Specifically,

apartments on the lowest floors, the least desirable floors in the Chinese market, are routinely


4/37

- 4 -

slower to sell. We confirm this good-faster-than-bad pattern empirically by looking at actual

price and sales data for specific Shanghai apartment complexes.

It should be noted at the onset that in this article we do not make any claims as to which

sales pattern is normatively optimal. The answer to this normative question would depend on

many additional factors including developers budgetary constraints and their rate of time

preference. Instead we are only interested in the discrepancy between developers desired sales

pattern and the actual sales pattern.

We explain the discrepancy between the desired and the actual sales patterns using

Tversky, Sattath, and Slovic (1988)s prominence principle. According to that principle, when

individuals choose among alternatives with no clearly dominant choice (i.e. an alternative is

superior on one attribute but inferior on another attribute), the option that is superior on the more

prominent attribute is more likely to be preferred in choice even though the two options are

equally attractive in matching. An attribute is defined as more prominent than another attribute if

laypeople perceive it to be more important than the other attribute, regardless of whether it is

more important in the regression sense. For example, most people would consider life as more

important than money, even though in the regression sense whether life or money is more

important depends on the ranges and the distributions of the two attributes.

In the context of the real estate development market, developers are in a matching task

when they set prices. Namely, they first set a base price for an average-quality unit in a building

and then make adjustments to set the prices for the other units so that all the units would are be

equally attractive to homebuyers. On the other hand, homebuyers are in a choice task, and as our

data show, they perceive floor to be more important (prominent) than price. Failure by

developers to account for the relative prominence between floor and price in the buyers choice


5/37

- 5 -

process may lead the former to under-price the more desirable floors and/or over-price the less

desirable floors.

To show that the inconsistency between developers intentions and buyers choices we

found in the field study was indeed due to the prominence principle, we conducted several

additional studies. First, we ran a survey designed (a) to show that the floors we assumed to be

undesirable in the field study were indeed considered undesirable by Shanghai homebuyers; and

(b) to show that floors were indeed perceived to be more important than prices in home-purchase

decisions.

We then conducted two controlled experiments to replicate the inconsistency found in the

field data. The first experiment consisted of two between-subjects conditionsone choice and

one matchingfor a pool of participants from Shanghai who had revealed the income potential

to buy a new apartment. In the matching condition, participants were asked to state prices that

would equalize their perceived values between two hypothetical apartments on different floors.

In the choice condition, we used the same two apartments and used the equalized prices from

the matching condition. Participants were then asked to choose one of the apartments. We

observed a striking inconsistency between price-setters and choosers. Participants in the choice

condition overwhelmingly chose the apartment on the better floor, despite the use of prices that

were revealed to equalize the attractiveness of the apartments, on average, in the matching

condition. In the second controlled experiment, we asked actual developers to set prices. Again,

we replicated the inconsistency between price setters and choosers. The results of these

controlled experiments were remarkably parallel to the results of the field study, thus reinforcing

our belief that the prominence principle was the underlying reason.


6/37

- 6 -

Previous research has documented laboratory evidence for the prominence effect, namely,

inconsistencies between matchers and choosers. To the best of our knowledge, we are the first to

make a connection between this behavior and real outcomes in the market place. Our findings

also shed light on a peculiar pricing anomaly in one of the worlds most turbulent and rapidly-

growing marketsthe Shanghai real estate market.

The remainder of the paper is organized as follows. In the first study, we summarize the

interviews of Shanghai developers. In the second study, we examine actual field sales data for

representative Shanghai developments. In the third study, we report a survey of potential

Shanghai home-buyers to show their floor preferences. We then discuss related literature in

judgment and decision-making to suggest an explanation for our stylized fact. In the fourth and

fifth studies, we report two controlled experiments to replicate the inconsistency observed in the

field. Finally, we conclude with a discussion of possible remedies.

Study 1: Interview of Realtors Regarding Their Desired Sales Pattern

The main purpose of this study was to examine developers desired sales patterns. In

addition, the study also explored why developers wanted the sales pattern as they indicated, and

how they set prices to achieve their desired sales pattern.

Method

We conducted extensive interviews with 47 experienced price-setters who closely

collaborated with an online website doing business in apartment trading and financial mortgage.

Their qualifications ranged from general managers of real estate companies, marketing managers


7/37


8/37

- 8 -

Developers also provided several explanations for why they would want apartments on

less desirable floors to sell at least at the same rate as good apartments. A very common

explanation was that good apartments were important for attracting potential consumers who

need to expend time and effort to visit a development in the first place. If most of the units on

good floors had been sold out, a building had more difficulty attracting potential consumers.

Some developers speculated that if units on good floors were unavailable, consumers would not

even bother to come to visit a new building. In general, if there were only bad floors available,

developers would have to resort to other means of attracting consumers, such as aggressive price

reductions that often lowered the developers profit.

Developers also provided cultural reasons why they would prefer not to sell the

apartments on the best floors first. In China, it was common for people to live close to their

relatives (including parents and children) and friends. Therefore, friends and relatives often

bought apartments in the same building. For social reasons, individuals would not want to

purchase apartments on inferior floors. A developer would prefer to have early purchasers buy

apartments on relatively undesirable floors because it would be much easier subsequently to sell

to friends and family on more desirable floors. In contrast, if early buyers purchased on the better

floors, it was much more difficult to sell units subsequently to their friends and family on a less

desirable floor. For these social reasons, developers preferred to sell the less desirable floors

first.

The interviews also provided accounts of the pricing strategy used to achieve the desired

sales pattern. To determine the list price, developers first chose an average quality unit in a

building. By dividing the total desired profit from the building by the total square meters, they

obtained the base price per square meter for this average unit (i.e. they used cost-plus pricing).


9/37

- 9 -

They adjusted the prices of the remaining units up or down depending on their quality relative to

the average unit. For the floor of an apartment, they typically used a constant price adjustment,

assuming that higher floors were more desirable. For instance, they might pick an average

quality floor such as the 5th

floor and set the base price at 6000 RMB per square meter. They

would then add 300 RMB per square meter for every floor above the 5th

, and deduct 300 RMB

per square meter for every floor below the 5th. In addition to assigning an upward price

adjustment to higher floors, developers might make further adjustments, e.g., assign a downward

adjustment to the price of the top floor, as it may get very hot during the summer. Floor was not

the only price differentiator. Developers might also make price adjustments based on other

factors such as exposure and floor plan.

To determine the price adjustment between floors in a building, developers typically took

the following steps. First, they surveyed potential consumers with questions such as: if the 5 th

floor is worth 6000 RMB per square meter, then how much do you think the 6thfloor is worth?

These types of stated-preference data generated preliminary estimates of consumer willingness-

to-pay for floors. To gauge the external validity of these estimates, some developers visited

buildings in the surrounding area to measure the competitive prices. Developers might also use

crude checks, such as comparing results with their own personal views of what they would be

willing to pay for a 6thfloor apartment given the availability of a 6000 RMB 5thfloor unit. In

general, however, the price adjustment for floors was obtained via a matchingstrategy whereby

floors were priced according to their perceived value relative to a base level.

Based on the interviews, we provided quantitative support for the selling strategy of

experienced developers who preferred to sell units on less desirable floors at least as fast as those

on good floors. We also provided anecdotal accounts of the pricing strategy used to obtain this


10/37

- 10 -

outcome. Finally, we provided anecdotal accounts of the actual sales patterns of apartments,

which were inconsistent with the intended patterns. In the next section, we examine actual field

sales data to provide quantitative support for the list pricing practices and the actual sales

patterns described in the interviews.

Study 2: Field Sales Data

Study 1 showed that most developers intended to sell all units in a building at the same

rate. We now examine actual sales patterns of several representative buildings in Shanghai to

verify whether they are consistent with the desired sales pattern.

Method

We selectively used developments from three major residential areas in Shanghai to obtain

a roughly representative account of the real estate market. The three properties contained both

mid-rises and high-rises targeted to consumer groups ranging from middle-income employees to

relatively rich individuals. In each of the properties studied, all units were listed on the market

from their launch dates. Hence, developers of these properties did not tactically withhold good

apartments to manipulate the sales rate.

Results and Discussion

Dataset 1: Four 6- or 7- floor Midrises

Our first dataset came from a gated community consisting of two six-story and two seven-

story mid-rises. Each floor in each building had two units with an average unit size of 140 square

meters. All the units faced south. The four buildings were initially offered in November 1999. In


11/37

- 11 -

Figure 1, we report the average price per square meter for each floor across these four buildings.

Consistent with our interviews with developers, prices per square meter increased with the floor

on which an apartment was situated, with the exception being the top floor.

To assess the sales pattern across the four buildings, we report the average number of

months for a unit to sell by floor in Figure 2. We will provide a quantitative analysis of the sales

pattern in the next section. Here, we just want to make a qualitative observation. Contrary to the

intentions of most developers, apartments on different floors seemed to sell at different rates. For

example, units on the 4thfloor took only 15.1 days to clear the market, but units on the 1stfloor

took 20.8 days.

Dataset 2: Seven 11-floor Midrises

Our second dataset came from seven mid-rises located in a newly-developed

neighborhood in Shanghai. Each building had 11 floors, with two units on each floor; the

average size of each unit was 150 square meters. All seven buildings were initially offered in

September 2001. In Figure 3, we report the average list price for an apartment by floor.

Again, contrary to most developers intentions, units on different floors sold at different

speeds. In Figure 4, we report the average number of months an apartment was on the market by

floor. The sales duration for apartments in the seven residential buildings again seemed quite

different.

Dataset 3: A 30-floor Highrise

In the two previous field datasets, we examined the sales pattern of mid-rises. The third

dataset corresponded to a thirty-story high-rise building with twenty units on each floor. With


12/37

- 12 -

the exception of floors 28 and 39 which had larger units, each floor had the same floor plan.

Floors 1 to 3 were for commercial use and not for sale. The unit sizes ranged from 25 square

meters to 105 square meters. Figure 5 shows the average list price for each floor.

Figure 6 shows the duration in months that an apartment was on the market since initial

offering. Again, we observe systematically different sales speeds for units on different floors.

In summary, all three datasets suggested that contrary to the intentions of most developers,

units on different floors did not sell at the same speed. Furthermore, the sales pattern across

floors did not seem random: On average, apartments on low floors appeared to sell more slowly

than apartments on other floors.

Study 3: Survey of Potential Shanghai Apartment Buyers for Their Attitudes about Floors

The graphs in the previous section suggested that units on different floors sold at

systematically different speeds: slow-selling units were generally on low floors and fast-selling

units were generally on middle or high floors.

In China, particularly in Shanghai, low floors are considered undesirable because units on

low floors usually are damp, have poor views, and are susceptible to mosquitoes. Thus, we

hypothesize that slow-selling units are generally on undesirable floors and fast-selling units are

generally on desirable floors.

To test this hypothesis, we need a more fine-tuned definition of good and bad floors.

Instead of defining good and bad floors ourselves, Study 3 asked potential homebuyers to tell us

which floors were good and which floors were bad. We then treated their judgments as the

operational definition of good and bad floors and reanalyzed the three datasets in Study 2.


13/37

- 13 -

Method

We conducted a survey of 84 participants in a short-term non-degree executive program in

a large university in Shanghai. The ages of the participants ranged from 28 to 55, with most of

the participants in their mid thirties. Most (93.5%) of the participants had purchased homes in

Shanghai.

In the questionnaire, they indicated whether floor or price was more important to them

when considering buying an apartment in Shanghai. They were then asked to consider buying an

apartment in a hypothetical 7-story building with identical floor plans for each apartment. For

this building, participants were asked to indicate which floor(s) they would consider as bad

floor(s). Similar exercises were then carried-out in the context of a hypothetical 11-story

building and then a 30-story building for which the first 3 floors were not for sale. We chose

these three hypothetical buildings to mimic the three types of buildings in our field study.


The results from the survey confirmed the strong role of floor. Seventy two percent of the

respondents indicated that floor was more important than price.

The survey also clarified which floors buyers considered bad. We define a floor as bad if more

than 50% of respondents judged it bad. In the latter part of this section, we will use this

characterization of good versus bad floors to reanalyze the field datasets we presented in study 2

and test whether bad floors sell significantly slower than good floors.

In the 7-story building, the 1stand the 2ndfloors were considered bad. In the 11-story

building, the 1stand the 2ndfloor were considered bad. In the 30-story building, the 4th, the 5th,

and the 30thfloors were considered bad.


14/37

- 14 -

Interestingly, the relative position of a floor in a building seemed to dictate whether it was

perceived as bad or good as opposed to the absolute floor number. For example, in the 30-

story building, the relatively low 4th

and 5th

floors were considered bad even though this was not

the case in the smaller 7 and 11 story buildings. In the 30-story building, the highest floor was

also considered a bad floor.

Analysis of the Field Sales Data Based on Buyers Attitudes about Floors

We now revisit the field sales data (Study 2). Recall from Figures 2, 4 and 6 that we

observed different selling rates for different floors. Using the survey (Study 3) results to define

good and bad floors, we could empirically test whether bad floors in the field study (Study 2)

indeed sold more slowly than good floors.

The results supported our hypothesis that units on bad floors sell slower than units on

good floors. In Dataset 1 the bad floors stayed on the market for 21.00 months whereas the good

floors for only 15.00 months (t=2.19,p


15/37

- 15 -

Theory: Pricing versus Choice

We have now confirmed that consumers indeed considered the floor on which an

apartment was situated to be more important than its price. Apparently, developers were aware

of the importance of floor. They attempted to gauge consumer willingness-to-pay for floors

when adjusting prices of units on different floors. The observed inconsistency between the actual

sales pattern and the developers desired sales pattern seems perplexing. Developers appeared to

be systematically failing to set the relative prices of apartments on good and bad floors to

achieve the desired sales pattern. We now describe theory that could potentially explain the

underlying source of this bias in pricing.

When homebuyers decide which unit to purchase, they are in a choice mode: they make

comparisons between units and choose the one they like most. In the choice mode, a homebuyer

selects a unit from an offered set of two or more units after comparing different attributes such as

price and floor. However, when developers set apartment prices, they are in a matching mode:

they anchor on a base price and then adjust the prices to make all the units equally attractive to

consumers. Normatively speaking, consumers should be indifferent between apartments that,

through matching, have been priced to appear equally attractive.

However, established research in judgment and decision-making demonstrates that people

do not have well-defined preferences (e.g., Liu & Soman, 2006; Payne, Bettman and Schkade,

1999; Slovic, 1995). Different preference elicitation procedures may highlight different aspects

of options, suggest alternative judgment strategies, and reverse or alter preferences.

A classic example of preference reversal is the p-bet/$-bet reversal offered by

Lichtenstein and Slovic (1971). Consider the following pair of lotteries:

A: .95 Win $2.50; .05 Lose $.75


16/37

- 16 -

B: .40 Win $8.50; .60 Lose $1.50.

Lottery A (called a P-bet) gives a large probability of winning a small amount of money;

lottery B (called a $-bet) gives a smaller probability of winning a larger amount. Participants

were either asked to pick the lottery they wanted to play (the choice condition) or to price the

two lotteries (the pricing condition). Normatively, these two elicitation methods should yield the

same preference order. But Lichtenstein and Slovic documented a robust preference-reversal. In

the choice condition, most participants chose the P-bet, but in the pricing condition, most

assigned a higher price to the $-bet. The choice-pricing preference-reversal has been replicated

by many other researchers (e.g., Grether & Plott, 1979; Pommerehne, Schneider & Zweifel,

1982; Reilly, 1982), and also replicated in casinos with real gamblers and real money (e.g.,

Lichtenstein and Slovic, 1973).

Extending the original p-bet/$-bet preference reversal, Slovic (1975) and Tversky, Sattath

and Slovic (1988) identified another robust reversal. Participants were presented with

information about two hypothetical job candidates applying for a production engineer position

who differed on two attributes, technical knowledge and human relations, as follows:

Technical Knowledge Human relations

Candidate A 86 76

Candidate B 78 91

Both attributes were rated on a 40 (very weak) to 100 (superb) scale. The participants were

told that technical knowledge was more important than human relations. One group of

participants (the choice condition) was asked to choose between the two candidates. Another

group of participants (the matching condition) was presented with the same two alternatives,

with one of the four scores missing, and was asked to fill in that missing score so that the two


17/37

- 17 -

candidates were equally attractive. In choice, most people chose Candidate A (the one with the

better technical score), but in matching, the score most people filled in suggested that they would

have preferred Candidate B if that score had not been missing and had been the same as

presented in the choice condition. This finding suggested aprominence effect-- that the more

important attribute in a choice set (e.g., technical knowledge in the study) receives more weight

in choice than in matching. This finding has been replicated in various forms by other

researchers (e.g., Auh & Johnson, 2005; Carman & Simonson, 1998; Chernev, 2005; Fischer &

Hawkins, 1993; Nowlis & Simonson, 1997).

Tversky et al. (1988) proposed the compatibility principle to explain the choice-matching

reversal (see also Slovic, Griffin, & Tversky, 1990). The weight of a stimulus attribute is

enhanced by its compatibility with the task. Generally speaking, qualitative information about

the ordering of the dimensions weighs more in the ordinal method of choice than in the cardinal

method of matching; whereas quantitative information weighs more in matching than in choice.

Our survey of potential apartment buyers indicated that the floor on which an apartment was

situated figured prominently in consumer preferences. Given the compatibility rule, in choice

mode we expected the floor of an apartment, the qualitative attribute, to receive more weight

than in matching mode. When developers used matching to determine the price adjustment for

apartments on different floors of a building, they assigned relatively less weight to floor. In

essence, the developers use of matching led to a bias in their measure of consumer willingness-

to-pay for floor because the actual demand for apartments was generated by choice.

In the next section, we present two controlled studies to replicate the inconsistency

between price-setters and buyers, as we observed in the field data.


18/37

- 18 -

Study 4: Consumer Pricing versus Consumer Choice

To make a connection between the experiments and the Shanghai real estate market, we

used potential home-buyers from Shanghai as participants. In the pricing condition, we asked

potential consumers to set the prices for two units in a hypothetical building with the objective of

making the two units equally attractive to them. We then used the prices generated from the

participants in the pricing condition to set the prices in the choice condition. In the choice

condition, we used the same two hypothetical units with prices set at the average levels from the

pricing condition. A separate group of participants was asked to choose among the two units.

Method

We conducted this study using 140 respondents who were either students in a large

business school in Shanghai or employees in two companies in Shanghai who had worked in the

city for three to five years.

Participants were assigned to either a pricing or a choice condition. In both conditions,

participants were asked to imagine that after graduation, they planned to buy a two-bedroom

unit. They were told that they were interested in a newly-built, 15-story building in which only

two units were left. The two units had the same 100 square meter floor plan. The only difference

was their floor location: Unit 101 was on the 1stfloor and Unit 801 on the 8thfloor.

In the pricing condition, participants completed two tasks in which they were told the

price of one unit and were asked to state the price for the other unit that would make the two

units equally attractive to them. In the first task, they were told that Unit 101, on the first floor,

had a price of 500,000 RMB and then asked to state the indifference price for Unit 801, an

identical unit on the eighth floor. The second task was identical except that they were told the


19/37


20/37

- 20 -

500,000 RMB and Unit 801 at 560,000 RMB. We inferred that any participant in the pricing

condition who stated a matched price for Unit 801 above 560,000 RMB when Unit 101 was

500,000 RMB would have chosen the former. In the second choice task, Unit 101 was priced at

440,000 RMB and Unit 801 was priced at 500,000 RMB. We inferred that anyone in the pricing

condition who stated a matched price for Unit 101 above 440,000 RMB when Unit 801 was

500,000 RMB would have chosen the former.

As expected, floor played a more prominent role in the choice condition than in the pricing

condition. Under choice, participants presented with an offer of Unit 101 at 500,000 RMB and

Unit 801 at 560,000 RMB chose the latter 85.1% of the time. Participants in the pricing

condition were inferred to choose Unit 801 only 30.3% of the time. Similarly, under choice,

participants presented with an offer of Unit 101 at 4400,000 RMB and Unit 801 at 500,000 RMB

chose the latter 78.4% of the time. Participants in the pricing condition were inferred to choose

Unit 801 only 33.3% of the time (85.1% versus 30.3%, two-way chi-squared = 43.45,p


21/37

- 21 -

units in a hypothetical building instead of only two units. Second, the unit prices in the choice

task were set by professional developers rather than consumers.

Method

As in study 4, study 5 had two between-subjects conditions: one pricing condition in

which developers were asked to set prices for the units in a hypothetical building and one choice

condition where consumers were asked to rank the units according to their preference orders.

Unlike study 4, which used only two apartments, study 5 used an entire building with 10 units to

make the price-setting process more realistic for the developers who participated in the study.

In the pricing condition, the respondents were the same 47 developers surveyed earlier in

our field study, who generally desired a flat or a bad-before-good sales pattern. Their task was to

set unit prices for a hypothetical building before the building was listed on the market. They

were asked to imagine that the building was located in a nice residential neighborhood in

downtown Shanghai. The building had 10 floors and each floor had one 100 square meter unit.

Each unit had a living room and two bedrooms, with the living room and one bedroom facing

south, and the other bedroom facing north. The average per-square meter cost of the building

was 5,000 RMB and, thus, the average per-apartment cost was 500,000 RMB.

In the choice condition, the respondents consisted of 43 potential homebuyers recruited

from a population of participants in a non-degree executive program in a large university in

Shanghai. Virtually all of them either already owned an apartment or could afford one in

Shanghai. These respondents were told that they were going to buy a unit in a newly developed

building in which they were interested. The description of the hypothetical building was the

same as in the pricing condition, except that the prices of each unit, instead of the costs, were


22/37

- 22 -

presented to the participants. We set the price of each apartment in the choice condition at the

average recommended price across the developers. As in study 4, we felt the use of the average

prices from the pricing condition would homogenize the relative attractiveness of apartments to

some extent, making the design more conservative. Participants were first asked to assume that

all the units were available and to choose the unit they liked the most. They were then asked to

assume that the unit they liked the most was no longer available and to choose the one they liked

the most among the remaining units. They repeated this procedure until all the floors were

exhausted. This method provided us with each consumers ranking scheme for the 10 apartments.

The procedure also mimicked what typically happens during the sales of a building in real life:

when more desirable units are sold, potential buyers could only choose among the remaining

units.


The results not only replicated the findings of Study 4 but also closely resembled the

findings of the field study.

In Figure 7, we report the prices set by the developers. As expected, we observed an

almost linear upward adjustment for units on successively higher floors, with a slight downward

adjustment for the top floor. This pattern was consistent with the price adjustment process

described by developers in study 1. Furthermore, the relatively low pricing on lower floors and

the downward adjustment on the top floor were consistent with our findings in studies 1 and 3

whereby both consumers and developers associated these floors as less desirable.

As in study 4, we again expected floor to weigh more heavily in the choice condition than

in the pricing condition. To construct the test, we used the results from the 11-story building in


23/37

- 23 -

study 3 to define the first two floors as bad, and the remaining floors as good. A participant was

deemed to have chosen a good floor (bad floor) if her top-ranked floor was a good floor (bad

floor). Under choice, we simply used the observed ranks. Under pricing, we used each

developers reported price-differential between each pair of floors to determine her

corresponding floor price premium. Hence, if a developer set a 500 RMB price difference

between the 3rdand the 6thfloors, we concluded that she perceived 500 RMB to be the monetary

equivalent of the difference in value between those two floors (i.e. 500 RMB was interpreted as

the compensating differential for the difference in floor). We then used these differentials to infer

each developers hypothetical ranking scheme had they been in the choice condition. For

instance, under choice, the unit on the 3rdfloor was 6280 RMB and the unit on the 6thfloor was

6780 RMB. For any developer who had set a price-differential between the 3rdand 6thfloors of at

least 500 RMB, we would infer a higher ranking for the 6thfloor. For the few instances where a

developer would have been indifferent (i.e. actual price difference is equal to their matched price

difference), we assigned each floor an equal rank.

As expected, floor played a more prominent role in choice than in pricing. We first

compare choosers and price-setters best choice (rank one). Choosers assigned the rank of one to

a good floor 97.7% of the time (only one participant assigned a rank of one to a bad floor). In

contrast, price-setters (developers) were inferred to have assigned a rank of one to a good floor

only 36.2% of the time. (two-way chi-squared = 37.62,p


24/37

- 24 -

floors separately for choice (upper panel) and pricing (lower panel). Interestingly, if we assume

that differences in floor ranks correspond to differences in the expected sales rates, we observe

roughly a good-faster-than-bad pattern for participants in the choice condition. In contrast, we

observe roughly a bad-faster-than-good pattern for developers in the pricing condition. Recall

that the prices under choice were simply the mean prices reported by the developers. Hence,

roughly speaking, the pricing strategies used by the developers appear to have generated the

same sales pattern amongst potential consumers as we observed in the field sales data in study 2.

However, the sales pattern we observed among the developers when using the means of their

stated prices mimics the desired sales patterns described by price-setters in study 1.

Study 5 confirmed the potential role of the prominence principle in housing demand in

Shanghai. Using actual potential buyers from Shanghai as well as actual price-setters, our

findings show the inconsistency between choice (buyers) and matching (sellers pricing

strategies). Furthermore, the use of 10-story building allowed us to replicate the inconsistency

between the desired bad-faster-than-good pattern of sellers and the actual good-faster-than-bad

pattern observed in the field.

Conclusion

Decades of judgment and decision-making research has generated a vast number of

intriguing anomalies. One of the most celebrated anomalies is the preference reversal between

choice and matching. Most anomalies in the judgment-and-decision-making literature are

demonstrated only using hypothetical scenarios or using real outcomes with very low stakes. The

present research relates the choice-matching preference reversal to a real-world problem

involving billions of dollars in one of world most active real estate markets.


25/37

- 25 -

Specifically, we have identified an inconsistency between Shanghai real estate developers

desired sales pattern and the actual sales pattern that emerged in the market place. Interviews

with developers showed that developers wanted to achieve a flat or a bad-faster-than-good sales

pattern. Furthermore, developers routinely used stated-preference data to learn consumer

willingness-to-pay for different floors and used this information to price in a manner that should

have achieved the desired sales pattern. Nevertheless, anecdotal accounts from developers and

actual sales data from selected Shanghai real estate developments clearly demonstrated a good-

faster-than-bad sales pattern.

To provide a potential explanation for the inconsistency, we conducted two controlled

laboratory experiments. The results supported our explanation, indicating that the difference

between choice mode and matching mode could generate the inconsistency between the desired

and actual sales patterns of apartments on different floors of a building. That is, the inconsistency

could arise from the incompatibility between the matching mode used by developers to measure

willingness-to-pay for floors and the choice mode used by consumers to buy apartments on

different floors.

To the extent that our explanation drove the observed inconsistency, developers were

systematically setting prices using a biased estimate of demand. According to the compatibility

principle, developers could potentially reduce this inconsistency by simulating the consumer

choice condition in the determination of prices. For instance, developers could ask potential

buyers to hypothetically choose between a unit on a desired floor and one on a less desired floor

in repeated iterations, and through these iterations, systematically vary the price difference

between the two units. For example, they could start with two extreme cases: Would you buy

Unit 201 or Unit 801 if 201 is $300,000 and 801 is also $300,000? Would you buy Unit 201 or


26/37

- 26 -

Unit 801 if Unit 201 is $100,000 and Unit 801 is $500,000? Assuming that respondents choose

Unit 801 in the first case and Unit 201 in the second, then the developers can gradually increase

the difference in price from the first scenario and reduce the difference in price from the second.

Through this procedure they could find a point where consumers are indifferent between the two

units.

Our findings for the context of real estate pricing are interesting in their own right given

the sheer magnitude of the Shanghai market. Nevertheless, we do not expect the good-faster-

than-bad sales pattern we identified in this research to be limited to the real estate market in

Shanghai. Almost all products involve a tradeoff between quality and price, and consumers often

consider quality as more important than price (Simonson & Tversky, 1992) and quality may

frequently take more weight than price in consumers choices (Hardie, Johnson, & Fader, 1993).

To the extent that sellers rely on matching to set prices, we expect to observe the same pricing

and sales-speed phenomena in other product categories.


27/37

- 27 -

7100

7200

7300

7400

7500

7600

7700

7800

1 2 3 4 5 6 7

Floor

Pricepersquaremeter(RMB)

Figure 1: Average price of units on different floors in the four mid-rises in Study 2 Dataset 1.


28/37

- 28 -

0.0

5.0

10.0

15.0

20.0

25.0

1 2 3 4 5 6 7

Floor

Months

Figure 2: Average market-clearing duration of units on different floors in the four mid-rises in

Study 2 Dataset 1.


29/37

- 29 -

4000

4200

4400

4600

4800

5000

5200

5400

5600

5800

6000

1 2 3 4 5 6 7 8 9 10 11

Floor


Figure 3: Average price of units on different floors in the seven mid-rises in Study 2 Dataset 2


30/37

- 30 -

0

6

12

18

24

30

36

42

1 2 3 4 5 6 7 8 9 10 11

Floor

Days

Figure 4: Average marketing-clearing duration of units on different floors in the seven mid-rises

in Study 2 Dataset 2.


31/37

- 31 -

10000

10500

11000

11500

12000

12500

13000

4 5 6 7 8 9 10 11 12 13 1 4 15 16 17 1 8 19 20 2 1 22 23 24 25 26 2 7 28 2 9 30

Floor


Figure 5: Average price of units on different floors in the high-rise in Study 2 Dataset 3.


32/37

- 32 -

0

3

6

9

12

15

4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Floor

Months

Figure 6: Average marketing-clearing duration of units on different floors in the high-rise in

Study 2 Dataset 3.


33/37

- 33 -

5200

5600

6000

6400

6800

7200

7600

1 2 3 4 5 6 7 8 9 10

Floor


Figure 7: Average price of units on different floors in the 10-story building in Study 5.


34/37

- 34 -

0

2

4

6

8

10

1 2 3 4 5 6 7 8 9 10

Floor

Rank

0

2

4

6

8

10

1 2 3 4 5 6 7 8 9 10

Floor

Rank

Figure 8: Average choice rank for units on different floors in the 10-story building in Study 5.

The upper panel is for the choice condition and the lower panel is for the pricing condition.


35/37

- 35 -

References

Auh, Seigyoung and Michael D. Johnson (2005), Compatibility effects in evaluations of

satisfaction and loyalty, Journal of Economic Psychology, 26(1), 35-57.

Carman, Ziv and Itamar Simonson (1998), Price-Quality Trade-Offs in Choice Versus

Matching: New Insights Into the Prominence Effect,Journal of Consumer Psychology,

7(4), 323-343.

Chernev, Alexander (2005), Feature Complementarity and Assortment in Choice,Journal of

Consumer Research, 31 (4), 748-759.

Economist (2001), Sticky as Treacle, Finance and Economics, June 14.

Economist (2005), Sky-High Shanghai, Finance and Economics, March 23.

Fisher, Gregory W. and Scott A. Hawkins (1993), Strategy Compatibility, Scale Compatibility,

and the Prominence Effect, Journal of Experimental Psychology, 19(3), 580-597.

Grant, Andrew (2006), The New Chinese Consumer, The McKinsey Quarterly, Special

Edition, 4-6.

Grether, David M., and Charles R. Plott (1979), Economic theory of choice and the preference

reversal phenomenon,American Economic Review, 69, 623-638.

Hardie, Bruce G.S., Eric J. Johnson, and Peter S. Fader (1993), Modeling loss aversion and

reference dependence effects on brand choice,Marketing Science, 12, 378-394.

Lichtenstein, Sharman, and Paul Slovic (1971), Reversals of preference between bids and

choices in gambling decisions,Journal of Experimental Psychology, 89, 46-55.

Lichtenstein, Sharman, and Paul Slovic (1973), Response-induced reversals of preference in

gambling: An extended replication in Las Vegas,Journal of Experimental Psychology,

1O1, 16-20.


36/37

- 36 -

Liu, Wenjing and Dilip Soman (Forthcoming), Behavioral pricing, in theHandbook of

Consumer psychology, Frank R. Kardes, Curtis P. Haugtvedt, and Paul Herr eds.

Mahwah, NJ: Erlbaum.

Nowlis, Stephen M. and Itamar Simonson (1997), Attribute-task compatibility as a determinant

of consumer preference reversals,Journal of Marketing Research, 34(2), 205-218.

Payne, John W., James R. Bettman, David A. Schkade (1999), Measuring constructed

preferences: Towards a building code,Journal of Risk and Uncertainty, 19 (1-3): 243-

270.

Prommerehne, Werner, W., Friedrich Schneider, and Peter Zweifel (1982), Economic theory of

choice and preference reversal phenomenon: A reexamination,American Economic

Review, 72(3), 569-574.

Reilly, Robert J. (1982), Preference reversalFurther evidence and some suggested

modifications in experimental-design,American Economic Review, 72 (3), 576-584.

Simonson, Itamar and Amos Tverskey (1992), Choice in context: Tradeoff contrast and

extremeness aversion,Journal of Marketing Research, 29, 281-295.

Slovic, Paul. (1975). Choice between equally valued alternatives. Journal of Experimental

Psychology: Human Perception and Performance, 1, 280-287.

Slovic, Paul (1995), The construction of preference,American Psychologist, 50(5), 364-371.

Slovic, Paul, Dale Griffin, and Amos Tversky (1990), Compatibility effects in judgment and

choice, inInsights in decision making: A tribute to Hillel J. Einhorn,Robin M. Hogarth

Ed. Chicago: University of Chicago.

Tverskey, Amos, Shmuel Sattach, and Paul Slovic (1988), Contingent Weighting in Judgment

and Choice, Psychological Review, 95(3), 371-384.


37/37

Ye, Xiannian (2004), China Real Estate Market, http://www.china-

window.com/china_market/china_real_estate/china-real-estate-market--5.shtml.

Behavioral Beijing

Documents

Transcript of Behavioral Beijing