Training

1
An Extreme Value Reference Price Approach
Sanjoy Ghose and Oded Lowengart
January 19, 2005
2
Effect of Price on Choice

• Price Only models
• Inclusion of Reference Price

3
Reference Price Categories

• External Reference Price
• Internal Reference Price

4
Internal Reference Prices

• Many different operationalizations
• Issue of appropriateness

5
Logic Forms

• Decaying memory of past occurrences
• Last Price paid (Winer, 1986 Mayhew Winer,
  1992)
• Variation of past average prices
• Weighted log-mean average (Kalwani et al., 1990)
• Exponentially weighted average (Obermiller, 1990)

6
Event Recall

• Hasties theory on memory
• Srulls experiments
• Incongruence vs Congruence of Information
• Effect on recall

7
Price Information Congruence

• Let Pexp Expected price of consumers
• Let price at time t Pt
• If Pt is similar to Pexp then Pt is congruent
  information
• If Pt gtgt (or ltlt) than Pexp, then Pt is
  incongruent information

8
Price Information Congruence

• The greater the degree of deviation of Pt from
  Pexp, the greater the incongruency of
  information.
• The greatest incongruency should occur with the
  maximum and minimum prices faced by consumers
  from t0 to tt.

9
Price Information Congruence

• Such maximum and minimum prices should be most
  easily recalled
• We hypothesize that these prices would be used as
  reference points in price evaluations.

10
Other Related Literature

• Monroe (1979)
• Range Theory (Volkmann, 1951)
• Applications to Pricing in the Mktg. lit.
• Experimental studies
• Janiszewski and Lichtenstein, 1999
• Niedrich, Sharma, and Wedell, 2001
• price attractiveness
• recommends that it was important for future
  research to consider range in the
  operationalization of reference prices in choice
  models.

11
Let V be the Utility
Similar to Rajendran Tellis (1994)..
12
(2)
(3)
Substituting (2) and (3) into (1),
(4)
Where,
13
Extreme Values of Reference Price

• Consumers would utilize the maximum and minimum
  prices they have paid in their previous shopping
  trips as reference prices.
• This should be reflected in superior performance
  of a model based on the EVRP approach.

14
Range Theory A stimulus range is based on its
extreme points Relative judgment and anchoring
effects
Human Association Memory A new incongruent
stimulus leads to a larger associative memory
network Different memory retrieval for
Incongruent information
Anchoring Points - Product Line A new extreme
stimulus is more noticeable than other stimulus
Behavioral Theory Individuals can be happy and
sad at the same time
Price Implications A price range is related to
the extreme price levels Price attractiveness is
relative to the extreme prices New extreme prices
change the range
Price Implications Both maximum and minimum
prices can be simultaneously used in evaluating
new prices
Price Implications A new extreme (high/low) price
has more memory associations than an expected new
price New extreme prices retrieved better from
memory than regular prices
Price Implications A new extreme
(maximum/minimum) price is more noticeable
Internal Reference Price Conceptualization Consume
rs use both high and low extreme points (price)
in their evaluations of a new price at the same
time Consumers can recall better extreme values
(price) as compared with regular prices
(expected) they paid previously Consumers use
extreme points (price) to decide about the
attractiveness of the offer Consumers use maximum
and minimum prices as anchoring
Choice/Purchase Quantity Implications Focus of
the Current Research Consumers use two internal
reference prices to evaluate current price -
comparing current price against the two,
simultaneously in a brand choice/purchase
quantity situation A maximum paid price - high
anchoring - creates gains A minimum paid price -
low anchoring - creates losses
Theoretical Framework
15
Hypotheses

• 1) For the aggregate sample, the EVRP approach
  for modeling consumer choice can serve as a
  better representation of internal reference price
  as compared to a last price paid formulation.

16
Hypotheses

• 2) For the aggregate sample, the EVRP approach
  for modeling consumer choice can serve as a
  better representation of internal reference price
  as compared to an average price paid formulation.

17
EVRP Segments

• Ratio of incongruent congruent Info (Srull,
  1981)
• Number of price points faced by consumer
• Purchase frequency

18
Hypotheses

• 3) The EVRP approach for modeling consumer choice
  can serve as a better representation of internal
  reference price in the high frequency segment
  than in the low frequency segment.

19
Hypotheses

• 4) For each of the two buyer frequency segments,
  the EVRP approach for modeling consumer choice
  can serve as a better representation of internal
  reference price as compared to a last price paid
  formulation.

20
Hypotheses

• 5) For each of the two buyer frequency segments,
  the EVRP approach for modeling consumers choice
  can serve as a better representation of internal
  reference price as compared to an average price
  paid formulation.

21
Gains Losses

• Consumers evaluate losses gains differently
  (Kahneman Tversky, 1979)
• We believe On any given purchase occasion, a
  consumer is always evaluating a loss as well as a
  gain

22
Model
23
EVRP Model
24
LPP Model
25
APP Model
26
Data

• A.C. Nielsen company scanner panel data set of
  laundry detergents Sioux Falls market
• Seven leading brands of liquid detergents
• Tide 128 oz, Tide 96 oz, Tide 64 oz, Wisk 64 oz,
  Wisk 32 oz, Surf 64 oz, and Surf 32 oz.

27
Variables

• Minimum Price - the lowest price paid or observed
  by consumer i for choice alternative j in
  previous purchase occasions
• Maximum Price - the highest price paid or
  observed by consumer i for choice alternative j
  in previous purchase occasions

28
Description of Conceptual Approach
Price
Subject Node
5.95
5.95 ... Max
Max....
5.12
4.56
4.50
4.12
3.95
4.01
3.95
3.24
Min...
3.12
3.12 ... Min
t1
t2
t4
t5
t6
t7
t8
t10
Time
t3
t9
29
(No Transcript)
30
Results

• EVRP model Significant gain and loss parameters
• Losses loom larger than gains consistent with
  Prospect Theory
• Less face validity for LPP and APP models
  especially for loss parameters

31
(No Transcript)
32
Results

• EVRP model provides superior fit based on the
  four different measures in Table 2
• Supporting hypotheses 1 and 2

33
(No Transcript)
34
Results

• EVRP gave better hit rate predictions than LPP or
  APP
• Superiority similar to other works in marketing
  literature (e.g., Manchanda et al, 1999 Mktg Sci
  Heilman et al., 2000 JMR)
• Further support to hypotheses 1 2

35
Segmentation

• To test hypotheses 3 to 5
• High low frequency of purchase
• Checked segmentation scheme
• LL test (Gensch, 1985)

36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
Segment level findings Tables 5 and 6

• EVRP parameter signs are generally consistent
  with expectations
• losses loom larger than gains
• model has face validity
• signs significances of gain loss parameters
  show less face validity for LPP and APP models.

40
(No Transcript)
41
(No Transcript)
42
Segment level findings

• EVRP has the best fit (Table 7)
• Also has the best holdout sample predictive
  accuracy (Table 8)
• True for both high purchase frequency and low
  purchase frequency segments
• Supports hypotheses 4 and 5

43
Results

• EVRP (High Freq. Segment) McFaddens R-sq.
  .550 and Hit Rate 65
• EVRP (Low Freq. Segment) McFaddens R-sq. .408
  and Hit Rate 56
• EVRP provides better data representation for high
  vs low freq segment Supports Hypothesis 3

44
Quantity Analysis
Table 9 Regression Results Aggregate Level
45
Quantity Analysis
Table 10 Regression Results High Frequency
Purchasing Segment
46
Quantity Analysis
Table 11 Regression Results Low Frequency
Purchasing Segment
47
Results

• Extreme value points model consistent with
  expectations ? both gains and losses are
  statistically significant
• A larger effect for gains than losses for the low
  frequency segment
• The high frequency segment show a larger effect
  for losses than gains

48
Summary

• Reference Price based choice models have always
  done better than price-only models
• Internal Reference Price models have been mainly
  driven by the decaying memory concept

49
Summary

• Instead, incorporating the incongruency of
  information approach together with the range
  theory concept
• Recent work (2001) suggest the attractiveness of
  range theory approach for price attractiveness
  judgments

50
Summary

• Niedrich et al (2001) say it is important to
  consider range in the operationalization of
  choice models
• EVRP --- a first step in that direction

51
Summary

• Past studies on Internal reference price ---
  either a gain or a loss on a given purchase
  occasion
• Our concept consumers maybe experiencing a gain
  and a loss on each purchase occasion

52
Managerial Implications

• While a price promotion strategy might have a
  short-run positive impact on sales, the lowered
  price may result in the installation of a new
  lower minimum price in consumers' memories
• may lead to a negative effect on market shares in
  the medium and long terms
• Managers may want to consider non-price forms for
  promotion if the goal is to increase short-term
  sales

53
Managerial Implications

• While a price increase may have an immediate
  adverse effect on sales, the possible higher
  maximum price level can help future market share
  values in the form of positive effect of gains
• Similar logic for choice of skimming vs.
  penetration strategies for new product
  introductions.

54
(No Transcript)