A Quote

Learning About Computers: Learning About Computers: An Analysis of Information Search An Analysis of Information Search

and Technology Choice and Technology Choice

Tülin ErdemTülin Erdem University of California, BerkeleyUniversity of California, Berkeley

Michael KeaneMichael KeaneYale University Yale University Sabri Sabri ÖncüÖncü

Stanford UniversityStanford UniversityJudi StrebelJudi Strebel

San Francisco State UniversitySan Francisco State University

IO workshop, DukeIO workshop, Duke University University October 26, 2004October 26, 2004

2

A QuoteA Quote““How should we allocate resources How should we allocate resources between channels? We really don’t between channels? We really don’t know exactly how consumers learn know exactly how consumers learn about technology” about technology”

Marketing manager for a major computer manufacturerMarketing manager for a major computer manufacturer

3

Motivation Motivation • Issues for a major computer manufacturer in Issues for a major computer manufacturer in

1995/96:1995/96:

1) What is the process consumers use to search 1) What is the process consumers use to search for information? What factors affect this process?for information? What factors affect this process?• Information sources utilized in technology choiceInformation sources utilized in technology choice• How do information sources interactHow do information sources interact• How does accuracy/cost of information provided affect How does accuracy/cost of information provided affect

search and technology processsearch and technology process

2) How do consumers decide which technology to 2) How do consumers decide which technology to choose and when to buychoose and when to buy• Role of expectationsRole of expectations

4

Objective Objective • Develop and estimate a model of consumer information search Develop and estimate a model of consumer information search

and choice behavior in high-tech durable goods markets and choice behavior in high-tech durable goods markets characterized by three key features: characterized by three key features:

1) there are two or more technological alternatives, 1) there are two or more technological alternatives,

2) consumers have uncertainty about the quality of 2) consumers have uncertainty about the quality of each alternative (and/or its suitability to their each alternative (and/or its suitability to their particular needs), and particular needs), and

3) there is a rapid pace of technological 3) there is a rapid pace of technological improvement, reflected in a rapid rate of price improvement, reflected in a rapid rate of price decline for any given level of quality. decline for any given level of quality.

Our application is to the computer market.Our application is to the computer market.

5

Contribution Contribution • Develop and estimate a structural dynamic model of Develop and estimate a structural dynamic model of

INFORMATION SEARCH (ACTIVE LEARNING) and INFORMATION SEARCH (ACTIVE LEARNING) and TECHNOLOGY CHOICETECHNOLOGY CHOICE

• To understand consumer information search and To understand consumer information search and technology behavior in high-tech durable good markets technology behavior in high-tech durable good markets

• To investigate the managerial/ consumer welfare To investigate the managerial/ consumer welfare implications of changes in cost and variability of implications of changes in cost and variability of information, pricing policy, consumer expectation information, pricing policy, consumer expectation formation processesformation processes

• Use data on price expectations to estimate consumer Use data on price expectations to estimate consumer future price expectations (Manski 2003), as well as future price expectations (Manski 2003), as well as use data on quality expectations, which facilitate use data on quality expectations, which facilitate identification (McFadden 1989)identification (McFadden 1989)

• Individual level panel data on “real” consumers Individual level panel data on “real” consumers

6

Relevant Previous WorkRelevant Previous Work

• Information search: Information search: • Antecedents (Srinivasan and Ratchford 91)Antecedents (Srinivasan and Ratchford 91)• Patterns (Meyer 82, Hagerty and Aaker 84)Patterns (Meyer 82, Hagerty and Aaker 84)

• High-Tech DurablesHigh-Tech Durables• Diffusion of innovation (Bass 69), prelaunch models Diffusion of innovation (Bass 69), prelaunch models

(Urban, Hauser, Roberts 90), market share models (Urban, Hauser, Roberts 90), market share models (Bridges et. al. 95), expectations (Sultan and Winer 93)(Bridges et. al. 95), expectations (Sultan and Winer 93)

• Consumer Choice Consumer Choice • Learning models (Roberts and Urban 88, Erdem 98, Learning models (Roberts and Urban 88, Erdem 98,

Anand and Shachar 00, Ackerberg 02, Ching 02, Anand and Shachar 00, Ackerberg 02, Ching 02, Crawford and Shum 02, Erdem and Keane 96)Crawford and Shum 02, Erdem and Keane 96)

• Forward-Looking expectations and consumer trade-offs: Forward-Looking expectations and consumer trade-offs: Gönül and Srinivasan 96, Erdem, Imai and Keane 03, Gönül and Srinivasan 96, Erdem, Imai and Keane 03, Melinkov 2000, Song and Chintagunta 2003) Melinkov 2000, Song and Chintagunta 2003)

7

A Dynamic Structural Model of A Dynamic Structural Model of Information Search and Information Search and

Technology ChoiceTechnology Choice

• Each period the consumer decides Each period the consumer decides • whether or not to obtain information from whether or not to obtain information from

several sourcesseveral sources• whether and what to buywhether and what to buy

• Consumers are uncertain about quality Consumers are uncertain about quality levels of alternatives, and changes in levels of alternatives, and changes in pricesprices

• over time, they learn about quality levels, over time, they learn about quality levels, • they form expectations about pricesthey form expectations about prices

8

Dynamics arise due to...Dynamics arise due to...

• The trade-off btw buying today versus The trade-off btw buying today versus delaying the purchasedelaying the purchase Computer prices tend to drop over time for a Computer prices tend to drop over time for a

given configuration and uncertainty about quality given configuration and uncertainty about quality levels decreases over time levels decreases over time reasons for waiting reasons for waiting

Opportunity costs that arises from not having a Opportunity costs that arises from not having a new computer during the period of delay new computer during the period of delay incentive to buy now incentive to buy now

9

• Computer Purchase Panel: n = 345; Computer Purchase Panel: n = 345; Six waves of surveys Six waves of surveys

• By t=6, 98 consumers bought, 102 By t=6, 98 consumers bought, 102 consumers did not buy, attrition: 145 (42%)consumers did not buy, attrition: 145 (42%)

• Sample CharacteristicsSample Characteristics• Expertise: 34% Novice, 52% Intermediate, 14% ExpertExpertise: 34% Novice, 52% Intermediate, 14% Expert• Past Purchase: 45% First time buyerPast Purchase: 45% First time buyer• Gender: 62% MaleGender: 62% Male• Income: 50% over 50KIncome: 50% over 50K• Age: Average 39Age: Average 39• Education 58 % college or graduate school degreeEducation 58 % college or graduate school degree

Data Data

10

Panel Data on...Panel Data on...

Search and Choice InformationSearch and Choice Information• Information Channels visited each periodInformation Channels visited each period

– Retail StoresRetail Stores– Computer SourcesComputer Sources– General Sources: General Sources: – AdvertisementsAdvertisements– WOMWOM

• Whether s/he had bought a PC (if so, its Whether s/he had bought a PC (if so, its description/cost)description/cost)

• Ratings about perceived quality of each Ratings about perceived quality of each technologytechnology

• The individual’s perceived price of the type of The individual’s perceived price of the type of configuration considered a) at the present time, configuration considered a) at the present time, b) six months earlier, c) price forecastsb) six months earlier, c) price forecasts

11

Perceived Quality Perceived Quality Construct Construct

• Will meet my needs for a long time to comeWill meet my needs for a long time to come

• User friendlyUser friendly

• PowerfulPowerful

• A large number of software titlesA large number of software titles

• All components operate together without any All components operate together without any problems (Hardware, software, peripherals)problems (Hardware, software, peripherals)

12

A few words on the Data A few words on the Data

• Search Search • Individuals search the most intensively in the first Individuals search the most intensively in the first

period and then search activity declines for all period and then search activity declines for all channels; no large differences across channelschannels; no large differences across channels

• Demographics do not affect the search patterns over Demographics do not affect the search patterns over time but they affect slightly the amount of search in time but they affect slightly the amount of search in each period (“Intermediate” expertise individuals each period (“Intermediate” expertise individuals search the most in each period than experts and search the most in each period than experts and novices) novices)

• Price Expectations Price Expectations • Similar expectations for Apple versus IBM/Compatible Similar expectations for Apple versus IBM/Compatible

technologytechnology• Demographics seem to affect expectations very little Demographics seem to affect expectations very little

(experts expect bigger declines than the rest)(experts expect bigger declines than the rest)

15

The Model: The The Model: The Preliminaries Preliminaries

• Let Let UUijrtijrt denote the utility to person i from purchase of denote the utility to person i from purchase of technology technology jj, where , where jj=Apple (platform), =Apple (platform), IBM/compatible (Windows Platform), in dollar amount IBM/compatible (Windows Platform), in dollar amount PPrr at time at time tt=1,=1,TT. Let . Let PPrr for for rr=1,=1,RR be a set of discrete be a set of discrete dollar amounts that the consumer may choose to dollar amounts that the consumer may choose to spend on a computer. spend on a computer.

• Assume that consumers have a utility function defined Assume that consumers have a utility function defined over the efficiency units of computer capabilities they over the efficiency units of computer capabilities they possess, possess, GG, and consumption of an outside good, , and consumption of an outside good, CC. .

• If a consumer sends If a consumer sends PPrr dollars on a computer, then dollars on a computer, then his/her consumption of the outside good is his/her consumption of the outside good is CCirir = = IIii – – PPrr, , where where IIii is the consumer’s income. is the consumer’s income.

16

Preliminaries continuedPreliminaries continuedGGijrtijrt = = jtjtPPrrQQjj

jtjt is an is an indexindex of the efficiency units of computer of the efficiency units of computer capabilities that one can purchase by spending one capabilities that one can purchase by spending one dollar on technology j at time t dollar on technology j at time t can be thought as can be thought as inverse price indices. inverse price indices.

• QQjj is the per dollar quality of technology j is the per dollar quality of technology jjtjt Q Qj j is the efficiency units of computer is the efficiency units of computer

capabilities that one can purchase by spending one capabilities that one can purchase by spending one dollar on technology j at time tdollar on technology j at time t

GGijrtijrt : : the efficiency units of computer the efficiency units of computer capabilities that one can purchase by capabilities that one can purchase by

spending Pspending Prr dollars on technology j at dollars on technology j at time ttime t

17

Utility SpecificationUtility Specification

i indexes individual; j indexes technology; t indexes time.i indexes individual; j indexes technology; t indexes time.

Uijrt: the utility to person i from purchase of technology j in dollar amount Pr at time t.

: risk aversion parameter

i : individual specific utility weight

: price coefficient

jt: an an indexindex of the efficiency units of computer capabilities that one can purchase by of the efficiency units of computer capabilities that one can purchase by spending one dollar on technology j at time t spending one dollar on technology j at time t

Qj : per dollar quality level of technology j.

Pr for r=1,..R is a set of discrete dollar amounts the consumer may choose to spend on a computer

ijrtriijrjtiijrt PQPU )}exp(1{

ijrtijrtijrtiijrt CGU })exp{1(

18

ij,t+1ij,t+1: the inverse of the consumer’s report of his/her : the inverse of the consumer’s report of his/her expectation of the price decline from t to t+1.expectation of the price decline from t to t+1.

E denotes the Expectation operator.E denotes the Expectation operator.

Forecasting Future Forecasting Future PricesPrices

),0(~

]|)/([lnln2

1,

1,1,1,

N

IE

tij

tijitjttjtij

19

Price Expectations Price Expectations ProcessProcess

)/ln()/ln(

]|)/([ln

2,21,10

1,

tjjttjjt

itjttjt IE

If 0=2=0 and 1=1, then consumers simply extrapolate the most recent one period (inverse) price change into the future. If 0=0, 1 > 2|2|, 2 < 0, then consumers expect that whatever acceleration or deceleration that occurred from t-2 to t will continue in the future. If 00, 0 < 1 < 1, 0 < 2 < 1, then the consumer expects the rate of price change to revert to some “natural” rate.

20

Distribution of Future Distribution of Future PricesPrices

)],|[ln(~ln 21,1, ittjtj IEN

Iit : information set that individual i has at time t

Point estimates of expected future prices are not sufficient to solve the consumer’s dynamic choice problem We need to specify the expected distribution of future prices.

Assume that agents expect that the distribution of future prices be:

21

Learning about Quality Learning about Quality

• Consumers lean about quality of each technology Consumers lean about quality of each technology through information channel (source) signals in a through information channel (source) signals in a Bayesian fashion; for each signal we estimate the Bayesian fashion; for each signal we estimate the variability of the information source variability of the information source

• Over time, their mean expectations will evolve and Over time, their mean expectations will evolve and the variance of their quality beliefs will shrink the variance of their quality beliefs will shrink (provided that the signals are not very noisy)(provided that the signals are not very noisy)

• We estimate mean quality perceptions for each We estimate mean quality perceptions for each technology for two segments (discrete mass approach technology for two segments (discrete mass approach to capture unobserved heterogeneity) but we also use to capture unobserved heterogeneity) but we also use consumer self-reports of quality expectationsconsumer self-reports of quality expectations

22

Learning about Learning about Quality of TechnologyQuality of Technology

Consumer Priors about Quality LevelsConsumer Priors about Quality Levels

),(~ 2ojojj QNQ

Qj : quality of technology j.Qoj : consumer’s prior expectation of the quality of computers type j2

oj: consumer’s prior uncertainty about about computers of type j

23

Information Channel Information Channel SignalsSignals

),(~ 2kjjkt QNS

k indexes the information source, Sjk : Signal from source k about technology type j. k

2 : Variance of the information provided by the k-th signal (inverse of the precision of information provided by k) [accuracy/precision of information]

24

Expected Quality of TechnologyExpected Quality of Technology

z is the expectation errorz is the expectation error

is the variance of consumer beliefs of quality of is the variance of consumer beliefs of quality of

technologytechnology

I is the information setI is the information set

,),0(~,]|[ 2

ijtijtijtjitj NzzQIQE ,

}|])|[{( 22ititjijt IIQEQjE

.

)],|[(~ 2ijtitjj IQENQ

2ijt

25

Information obtained from Information obtained from

Information ChannelsInformation Channels

S denotes the signal received about the quality of technologyS denotes the signal received about the quality of technology

x is the random term (noise) associated with the information receivedx is the random term (noise) associated with the information received

),0(~ 2kijktijktjijkt NxxQS

26

Expectations of Technology Expectations of Technology Quality Quality

& Technology Quality Ratings& Technology Quality Ratings

(8) ,

),0(~,]|[ 2ijtijtijtjitj NzzQIQE

jLitjijt IQEifLq ]|[

jHitjjLijt IQEifMq ]|[

.

jHitjijt IQEifHq ]|[

27

Ordered Probit to Estimate the Ordered Probit to Estimate the Cut-off Points for the Signal Levels Cut-off Points for the Signal Levels

of L, M and H for Quality of of L, M and H for Quality of TechnologyTechnology

• is the cumulative normal distribution function with is the cumulative normal distribution function with

mean 0 and variance mean 0 and variance

, )()(Pr jjLijt QLqobijt

)()()(Pr jjLjjHijt QQMqobijtijt

,

)(1)(Pr jjHijt QHqobijt

ijt

2ijt

28

Consumer’s Dynamic Choice Consumer’s Dynamic Choice ProblemProblem

m indexes the combination of information sources (the m=1,..,32 search options, where 32=25); V denotes “Value function”; P denotes “Purchase”; NP denotes “No=Purchase”Jkm: an indicator for whether source k is included in combination m.ck: the costs of obtaining information from each source kimt: an i.i.d. stochastic shock to the cost of search option m at time t.VP captures the maximum over all possible technology and price choices {j,r} of the expected utilities associated with choices Ui0 : per period utility from the current computer (if applicable)

imtitNitit

Pitkkm

kitimt mIVmIVEcJIV

)},(),,({max)(

5

1

],|[max},{

mIUEV itijrtrj

Pit

ittitimm

iNit eIVEUV ][max 1,1,0

29

Choice Probability for Choice Probability for Search OptionsSearch Options

Pr(MPr(Mimtimt): Probability of search option m for person i at time t): Probability of search option m for person i at time t

ΘΘ: : the set of parametersthe set of parameters

z: expectation errors z: expectation errors

ττ: : latent class typelatent class type

32

1

1,

))(exp(

))(exp(),,|1Pr(

litilt

itimttiimt

IV

IVzM

30

Choice Probabilities for Purchase Choice Probabilities for Purchase and No-Purchaseand No-Purchase

• DD stands for probability, stands for probability, θθ for the set of for the set of parameters, parameters, zz for expectation errors, and for expectation errors, and ττ for for latent class typelatent class type

2

1 1

)]|[exp()exp(

)]|[exp(),,|1Pr(

l

R

qitilqt

Nit

itijrtitijrt

IUEV

IUEzD

2

1 1

0

)]|[exp()exp(

)exp(),,|1Pr(

l

R

qitilqt

Nit

Nit

itti

IUEV

VzD

31

Likelihood FunctionLikelihood Function

• L1L1 corresponds to information choices corresponds to information choices

• L2 L2 corresponds tocorresponds to purchase decisionspurchase decisions

• L3 L3 corresponds tocorresponds to reported quality ratingsreported quality ratings

• L4 L4 corresponds tocorresponds to reported price expectationsreported price expectations

)(),,(),,(),,(),,,( 4321,11, ititititittiittiitit LzLzLzLzzL

32

Likelihood Function IILikelihood Function II

2

1)(),,()(

ii

iziii dzzfzLL

33

Parameter EstimatesParameter EstimatesNo Purchase Utility Coefficients

0 0.041 (0.012) Constant 3.917 (1.258)

1 1.239 (0.163) Ownership of Computer 0.149 (0.061)

2 -0.958 (0.194) Experience 0.009 (0.049) 0.088 (0.013) Age 0.644 (0.278)

Education -0.019 (0.089) Gender 0.543 (0.245)

0.076 (0.043) Income -0.294 (0.124)

Information Search Process Parameters

Price Coefficent 0.478 (0.193) Prior Std.Dev. of Quality PerceptionIBM 0.927 (0.160)

1.000 - Apple 2.071 (0.350)

0.995 - Variability of Information

Store Visit 1.977 (0.659) Ads and Articles 4.741 (1.976)

Constant 1.000 - Word of Mouth 3.128 (1.325) Experience 0.479 (0.238) Age 0.207 (0.103) Information CostsEducation -0.734 (0.257) Store Visits 1.415 (0.123) Gender 0.639 (0.283) General Articles 0.642 (0.088) Income -0.670 (0.399) Computer Articles 1.137 (0.175)

Advertising 0.379 (0.052) Quality Coefficients Word of Mouth 0.318 (0.057)

Mean Quality( IBM)a 0.891 (0.397)

Mean Quality( Apple)a -0.891 -

Mean Quality( IBM)b -0.275 (0.125) IBM-Left -1.741 (0.267)

Mean Quality( Apple)b 0.275 - IBM-Right 2.867 (0.419) Apple-Left -3.217 (0.569) Apple-Right 2.433 (0.388)

1st Latent Class 0.879 (0.155) 2nd Latent Class 0.121 -

a 1st Latent Classb 2nd Latent Class

Note. Values enclosed in parentheses represent standard errors.

Utility Weight

Risk Aversion Coefficient

Latent Class Probabilities

Discount Factor

Quality Perception Interval Coefficients

Parameter Estimates

Price Process Parameters

Utility Parameters

Utility Parameters (continued)

Price Process Coefficents

Measurement Error Std. Dev.

34

Parameter Estimates: Parameter Estimates: Price Process Parameters Price Process Parameters

Price process parametersPrice process parameters

00=-0.041 , =-0.041 , 11= 1.239 , = 1.239 , 22= - 0.958, = - 0.958, = 0.088= 0.088

22 is negative and larger in absolute value than is negative and larger in absolute value than 11+ + 22

consumers expect mean reversion in price consumers expect mean reversion in price declinesdeclines

If price stayed constant from t-2 to t then consumers If price stayed constant from t-2 to t then consumers would expect a 4.1% price decline from t to t+1.would expect a 4.1% price decline from t to t+1.

A steady state expected rate of price decline of roughly A steady state expected rate of price decline of roughly 2.5% per two-month period. 2.5% per two-month period.

Substantial measurement error in consumer’s reports of Substantial measurement error in consumer’s reports of their own expectations. their own expectations.

35

Parameter Estimates IIParameter Estimates II

• Prior uncertainty is substantial. It is higher for Apple. Prior uncertainty is substantial. It is higher for Apple.

• Two segments of consumers, first segment’s mean expected Two segments of consumers, first segment’s mean expected quality is higher for IBM/Compatible than Apple; the opposite quality is higher for IBM/Compatible than Apple; the opposite holds for the second segment. The first segment constitutes holds for the second segment. The first segment constitutes 88% of the individuals. 88% of the individuals.

• No-purchase utility is higher for individualsNo-purchase utility is higher for individuals• who 1) already own a computer, 2) older people, 3) women who 1) already own a computer, 2) older people, 3) women

and 4) lower income people. Education and experience do and 4) lower income people. Education and experience do not have a statistically significant effect on No-Purchase not have a statistically significant effect on No-Purchase Utility.Utility.

• Utility weight is higher Utility weight is higher • 1) the more experienced they are with computers, 2) the 1) the more experienced they are with computers, 2) the

older they are, 3) the less educated they are, and 4) if they older they are, 3) the less educated they are, and 4) if they are male.are male.

36

Parameter Estimates III: Parameter Estimates III: Variability and Cost of the information Variability and Cost of the information

SourcesSources

• Computer magazines, and general sources Computer magazines, and general sources and advertising provide the noisiest and advertising provide the noisiest information, whereas store visits provide information, whereas store visits provide the most precise information.the most precise information.

• Reading computer articles, followed by Reading computer articles, followed by store visits, seem to be the most costly store visits, seem to be the most costly sources, whereas obtaining word of mouth sources, whereas obtaining word of mouth information seems to be the least costly information seems to be the least costly one, followed by reading advertisements.one, followed by reading advertisements.

37

Model Fit: Model Fit: Comparing sample frequencies Comparing sample frequencies

of search and purchase behavior with of search and purchase behavior with baseline simulations based on parameter baseline simulations based on parameter estimatesestimates

• The model fits the data quite well. The model fits the data quite well. It slightly overstates the No-Purchase rate in wave It slightly overstates the No-Purchase rate in wave

6, and overpredicts purchases in wave 1.6, and overpredicts purchases in wave 1. It slightly under predicts the extent of search.It slightly under predicts the extent of search. It accurately predicts the relative utilization of It accurately predicts the relative utilization of

each source, as well as the time-path of utilization. each source, as well as the time-path of utilization. Percentage of people using each information Percentage of people using each information

source declines over time both in the data and source declines over time both in the data and according to the model. Evidence for duration according to the model. Evidence for duration dependence.dependence.

38

Model Fit: Table 4Model Fit: Table 4

Wave Sample Size No Purchase IBM Apple No Purchase IBM Apple1 281 281 0 0 100.0 0.0 0.02 281 261 19 1 92.9 6.8 0.43 245 225 18 2 91.8 7.3 0.84 202 179 22 1 88.6 10.9 0.55 162 145 17 0 89.5 10.5 0.06 120 102 17 1 85.0 14.2 0.8

Wave Sample Size No Purchase IBM Apple No Purchase IBM Apple1 281 281 0 0 100.0 0.0 0.02 281 261 19 1 92.9 6.8 0.43 261 240 19 2 91.8 7.3 0.84 240 213 26 1 88.6 10.9 0.55 213 191 22 0 89.5 10.5 0.06 191 162 27 2 85.0 14.2 0.8

Total 281 162 113 6 57.7 40.2 2.1

TABLE 4

Sample Frequencies vs. Model Simulation: Purchase Behavior

Sample frequencies

Number of Purchases Percentage of Purchases

Attrition Adjusted Sample Frequencies


39

Table 4 (con’t)Table 4 (con’t)

Wave Sample Size No Purchase IBM Apple No Purchase IBM Apple1 281.0 265.1 13.3 2.6 94.3 4.7 0.92 265.1 246.6 17.3 1.2 93.0 6.5 0.53 246.6 226.1 19.8 0.6 91.7 8.0 0.34 226.1 204.5 21.2 0.4 90.5 9.4 0.25 204.5 183.5 20.9 0.2 89.7 10.2 0.16 183.5 163.6 19.7 0.2 89.2 10.7 0.1

Out of Sample7 163.6 142.4 19.5 1.8 87.0 11.9 1.18 142.4 120.7 20.3 1.4 84.8 14.2 1.09 120.7 100.9 18.7 1.1 83.6 15.5 0.9

10 100.9 81.1 19.0 0.9 80.3 18.8 0.811 81.1 64.8 15.6 0.6 80.0 19.3 0.812 64.8 51.8 12.6 0.5 79.9 19.4 0.8

First 6 281 163.6 112.1 5.2 58.2 39.9 1.9Total 281 51.8 217.8 11.5 18.4 77.5 4.1


Model Simulation - Baseline

40

Model Fit II: Table 5Model Fit II: Table 5

Sample Frequencies

Wave Sample SizeStore Visits

General Articles

Computer Articles Advertising

Word of Mouth

Store Visits

General Articles


Word of Mouth

1 281 181 164 148 188 247 64.4 58.4 52.7 66.9 87.92 281 115 116 98 163 204 40.9 41.3 34.9 58.0 72.63 245 82 101 87 132 136 33.5 41.2 35.5 53.9 55.54 202 77 87 83 121 116 38.1 43.1 41.1 59.9 57.45 162 46 64 53 77 94 28.4 39.5 32.7 47.5 58.06 120 35 50 41 59 76 29.2 41.7 34.2 49.2 63.3


General Articles


Word of Mouth

Store Visits

General Articles


Word of Mouth

1 281 181 164 148 188 247 64.4 58.4 52.7 66.9 87.92 281 115 116 98 163 204 40.9 41.3 34.9 58.0 72.63 261 87 107 92 140 144 33.5 41.2 35.5 53.9 55.54 240 91 103 98 143 137 38.1 43.1 41.1 59.9 57.45 213 60 84 69 101 123 28.4 39.5 32.7 47.5 58.06 191 55 79 65 93 120 29.2 41.7 34.2 49.2 63.3

Total 1467 589 653 570 828 975 40.1 44.5 38.9 56.4 66.5

Sample Frequencies vs. Model Simulation: Search Behavior

Number of Consumers Collecting Information Percentage of Consumers Collecting Information

Attrition Adjusted Sample Frequencies


41

Table 5 (con’t)Table 5 (con’t)Model Simulation - Baseline


General Articles


Word of Mouth

Store Visits

General Articles


Word of Mouth

1 281 170 164 139 189 231 60.5 58.4 49.5 67.3 82.22 265 88 103 81 155 179 33.2 38.9 30.6 58.5 67.53 247 71 92 83 129 126 28.7 37.2 33.6 52.2 51.04 226 83 84 78 125 123 36.7 37.2 34.5 55.3 54.45 205 51 72 57 89 106 24.9 35.1 27.8 43.4 51.76 184 47 65 55 81 109 25.5 35.3 29.9 44.0 59.2

Out of Sample7 164 42 59 46 70 98 25.6 36.0 28.0 42.7 59.88 143 35 47 40 58 76 24.5 32.9 28.0 40.6 53.19 122 29 39 34 46 62 23.8 32.0 27.9 37.7 50.810 102 24 36 27 41 50 23.5 35.3 26.5 40.2 49.011 82 19 27 23 31 40 23.2 32.9 28.0 37.8 48.812 65 16 21 19 26 32 24.6 32.3 29.2 40.0 49.2

First 6 1408 510 580 493 768 874 36.2 41.2 35.0 54.5 62.1

Total 2086 675 809 682 1040 1232 32.4 38.8 32.7 49.9 59.1


42

Policy Experiments I: Effects Policy Experiments I: Effects of Expected of Expected Price Declines and Price Declines and Learning Learning on Purchase on Purchase DelayDelay

• No expectations of price declineNo expectations of price decline• Relatively small effect of expected price Relatively small effect of expected price

declinesdeclines– Acceleration of purchases Acceleration of purchases – However, total purchases are almost However, total purchases are almost

unchanged at T=12unchanged at T=12

• Substantial increase in search costsSubstantial increase in search costs• Lower computer sales overallLower computer sales overall• Decreases positive duration dependenceDecreases positive duration dependence

43

Policy Experiments II: Policy Experiments II: Effects of Price Expectations Effects of Price Expectations

on on Price Elasticities of Demand Price Elasticities of Demand

• Temporary %20 price cutTemporary %20 price cut Expectations adjust Expectations adjust versusversus They are fixedThey are fixed

• Elasticity of demand wrt transitory price cur is Elasticity of demand wrt transitory price cur is almost 50% greater if we allow price expectations almost 50% greater if we allow price expectations to adjust rather than holding them fixed to adjust rather than holding them fixed

• Acceleration of purchases and total Acceleration of purchases and total purchasespurchases Incremental sales due to the price cut only 4.6% at Incremental sales due to the price cut only 4.6% at

T=12 but 80% at T=6.T=12 but 80% at T=6. Price elasticity higher for ApplePrice elasticity higher for Apple

44

Policy Experiments IIIPolicy Experiments III

• EFFECT OF DECERASE IN INFORMATION EFFECT OF DECERASE IN INFORMATION COSTS ON PUCRHASE BEHAVIORCOSTS ON PUCRHASE BEHAVIOR

Decreasing information costs leads to Decreasing information costs leads to acceleration of purchasesacceleration of purchases

Decreasing information costs affects Decreasing information costs affects Apple sales more than IBM/Compatible Apple sales more than IBM/Compatible salessales

Advertising has the least effects Advertising has the least effects

45

Policy Experiments IVPolicy Experiments IV• EFFECTS OF INCREASING THE PRECISION OF EFFECTS OF INCREASING THE PRECISION OF

INFORMATION ON SEARCH AND PURCHASE BEHAVIOR INFORMATION ON SEARCH AND PURCHASE BEHAVIOR Decreasing the variability of information sources Decreasing the variability of information sources

(increasing the precision) encourages consumers (increasing the precision) encourages consumers to search slightly more (own variability effects are to search slightly more (own variability effects are positive) and accelerates purchases positive) and accelerates purchases

Most cross-effects are positive (except for negative Most cross-effects are positive (except for negative effect of store visits precision on advertising and effect of store visits precision on advertising and general sources)general sources)

Increased precision helps Apple relatively moreIncreased precision helps Apple relatively more

46

Conclusions & Future Conclusions & Future ResearchResearch

• Consumers are forward-looking Consumers are forward-looking

• Both learning and price expectations affect purchase Both learning and price expectations affect purchase timing and purchase decisiontiming and purchase decision

• Ignoring price expectations severely bias price elasticitiesIgnoring price expectations severely bias price elasticities

• Cost and accuracy of information sources affect Cost and accuracy of information sources affect consumer search behaviorconsumer search behavior

• Availability of survey data about expectations facilitate Availability of survey data about expectations facilitate identificationidentification

• Future research: Future research: Information search: Learning to learn?Information search: Learning to learn? Brand choice, consideration set formation Brand choice, consideration set formation

47

AppendixAppendix

48

Table 6Table 6 Wave Sample Size No Purchase IBM Apple No Purchase IBM Apple1 281.0 263.7 14.4 2.8 93.8 5.1 1.02 263.7 241.5 20.1 2.2 91.6 7.6 0.83 241.5 217.2 22.9 1.3 90.0 9.5 0.54 217.2 193.6 22.8 0.8 89.1 10.5 0.45 193.6 171.2 21.9 0.5 88.4 11.3 0.36 171.2 150.9 20.0 0.3 88.1 11.7 0.2

Out of Sample7 150.9 131.4 18.1 1.4 87.1 12.0 1.08 131.4 111.2 19.0 1.1 84.6 14.5 0.99 111.2 92.2 17.7 1.3 82.9 15.9 1.2

10 92.2 75.8 15.8 0.7 82.2 17.1 0.711 75.8 62.1 13.1 0.5 82.0 17.3 0.712 62.1 50.9 10.9 0.3 81.9 17.5 0.6

First 6 281 150.9 122.2 7.9 53.7 43.5 2.8Total 281 50.9 216.7 13.4 18.1 77.1 4.8


Out of Sample7 172.8 154.4 17.5 0.9 89.4 10.1 0.58 154.4 135.7 17.9 0.8 87.9 11.6 0.59 135.7 117.6 17.5 0.7 86.6 12.9 0.5

10 117.6 100.8 16.3 0.5 85.7 13.8 0.511 100.8 86.4 14.0 0.4 85.7 13.9 0.412 86.4 74.0 12.2 0.3 85.6 14.1 0.3

First 6 281 172.8 104.5 3.7 61.5 37.2 1.3Total 281 74.0 199.8 7.3 26.3 71.1 2.6

B. Increase Search Cost 60%Number of Purchases Percentage of Purchases

Simulated Effects of Changing Price Expectations and Raising Search Costs

A. Consumers Expect No Downward Trend in PricesNumber of Purchases Percentage of Purchases

49


Out of Sample7 153.7 133.4 18.5 1.8 86.8 12.0 1.28 133.4 112.8 19.3 1.4 84.5 14.4 1.19 112.8 93.9 17.8 1.1 83.2 15.8 1.0

10 93.9 75.0 18.1 0.8 79.9 19.2 0.911 75.0 59.5 14.9 0.6 79.4 19.8 0.812 59.5 47.1 11.9 0.5 79.1 20.1 0.8

First 6 281 153.7 120.0 7.3 54.7 42.7 2.6Total 281 47.1 220.4 13.5 16.8 78.4 4.8


Out of Sample7 157.5 137.1 18.8 1.7 87.0 11.9 1.18 137.1 116.2 19.6 1.3 84.8 14.3 1.09 116.2 97.1 18.1 1.0 83.5 15.6 0.9

10 97.1 77.9 18.3 0.8 80.3 18.9 0.911 77.9 62.2 15.1 0.6 79.8 19.4 0.812 62.2 49.6 12.1 0.5 79.7 19.5 0.8

First 6 281 157.5 117.3 6.2 56.1 41.7 2.2Total 281 49.6 219.2 12.2 17.6 78.0 4.3


A. 20% Price Decline in Wave 2 - Total EffectNumber of Purchases Percentage of Purchases

Price Cut Simulations

B. 20% Price Decline in Wave 2 - Expected Future Price Changes Held Fixed

50

Table 8Table 8


Out of Sample7 145.8 125.1 18.9 1.9 85.8 12.9 1.38 125.1 104.8 18.8 1.5 83.8 15.0 1.29 104.8 86.1 17.5 1.2 82.1 16.7 1.1

10 86.1 68.8 16.4 0.9 79.9 19.0 1.111 68.8 54.6 13.6 0.7 79.3 19.7 1.012 54.6 42.4 11.7 0.4 77.7 21.5 0.8

First 6 281 145.8 124.1 11.0 51.9 44.2 3.9

Total 281 42.4 221.0 17.6 15.1 78.7 6.2

Simulated Effects of Decreasing Search Costs

A. 20% Decrease in the Cost of a Store Visit


51



Out of Sample7 151.8 130.8 19.4 1.6 86.1 12.8 1.18 130.8 110.0 19.5 1.3 84.1 14.9 1.09 110.0 91.5 17.6 1.0 83.1 16.0 0.9

10 91.5 73.5 17.2 0.8 80.3 18.8 0.811 73.5 58.9 14.0 0.6 80.2 19.0 0.812 58.9 47.3 11.2 0.4 80.3 19.0 0.7

First 6 281 151.8 121.5 7.7 54.0 43.2 2.7Total 281 47.3 220.4 13.3 16.8 78.4 4.7

B. 20% Decrease in the Cost of Gathering Information from Advertising


52


Out of Sample7 141.7 121.1 18.7 1.9 85.4 13.2 1.38 121.1 101.0 18.6 1.5 83.4 15.3 1.29 101.0 83.3 16.5 1.2 82.5 16.4 1.2

10 83.3 66.2 16.1 0.9 79.5 19.4 1.111 66.2 52.4 13.2 0.7 79.1 19.9 1.012 52.4 41.4 10.6 0.4 79.0 20.2 0.8

First 6 281 141.7 126.2 13.1 50.4 44.9 4.7

Total 281 41.4 219.9 19.7 14.7 78.3 7.0


Out of Sample7 136.5 116.3 18.5 1.7 85.2 13.6 1.28 116.3 96.8 18.2 1.3 83.3 15.6 1.19 96.8 78.6 17.2 1.0 81.2 17.7 1.1

10 78.6 62.3 15.6 0.8 79.2 19.8 1.011 62.3 49.2 12.5 0.6 79.0 20.1 0.912 49.2 38.2 10.5 0.5 77.6 21.4 1.1

First 6 281 136.5 130.5 14.0 48.6 46.4 5.0

Total 281 38.2 222.9 19.9 13.6 79.3 7.1

A. Store Visit - 20% Increase in Precision


B. Advertising, Articles in General and Computer Publications - 20% Increase in Precision


Simulated Effects of Making Information Sources More Accurate

53



Out of Sample7 142.3 121.5 19.3 1.5 85.4 13.6 1.08 121.5 101.5 18.9 1.1 83.5 15.5 0.99 101.5 84.0 16.6 0.9 82.7 16.4 0.9

10 84.0 68.1 15.3 0.7 81.0 18.2 0.811 68.1 55.0 12.5 0.5 80.8 18.4 0.712 55.0 44.6 10.0 0.4 81.1 18.2 0.7

First 6 281 142.3 129.5 9.2 50.6 46.1 3.3Total 281 44.6 222.1 14.3 15.9 79.0 5.1


C. Word-of-Mouth - 20% Increase in Precision

A Quote

Documents

Transcript of A Quote