To What Extent is Promotional Effectiveness Moderated by Type of Price Frame and Temporal Patterns

(1)

To What Extent is Promotional

Effectiveness Moderated by Type of

Price Frame and Temporal Patterns

Carolyne Saunders

s1655272

Supervisor:

(2)

Abstract

Existant research about the effect of price presentations is limited to studying consumer perceptions in an experimental setting. In this study, I extend research on the effect of price presentations by examining its influence on actual consumer purchase behaviour. I use store-level scanner data on daily unit sales of cotton wool pads for a period of over one year. The results show that the price presentation “3+1 free” generate more sales than the price presentation “was €__, now €__”. Hence I reach the conclusion that price presentations for multiple products are more effective than the price presentations for a single product. In addition, I find substantial variation in promotional effectiveness during the promotion period.

(3)

1: Introduction

Price promotions are ubiquitous. Equally widespread are the different ways in which the price promotions are presented. “Buy one get one free”, “50% off” and “two for the price of one” are synonymous offers, but it has been suggested by many academics (Della Bitta et al., 1981; Chen et al., 1998 and Hardesty and Bearden, 2003 to name a few) that consumers react differently to different price frames. Another differentiating feature of promotions is the length of time for which the promotion is held. Although there is much extant literature on lead and lag effects of promotions (see for example Doyle and Saunders, 1985 or Neslin, 2002), little attention has been paid to the most effective length of a promotion.

This study aims to answer two questions: “How is a price promotion best communicated?” and “For what length of time should a price promotion be offered?”. Both questions are important from a managerial point of view. Operating a promotion is a costly exercise, so retailers are likely to want to maximise their returns using the most efficient price frame and promotion length.

Literature on price framing to date has generated conflicting outcomes. A meta-analysis by Krishna et al. (2002), which incorporated 20 studies, 345 observations, and 6 “deal frames”, drew the conclusion that no deal frame performed particularly better than any other. Conversely, studies by Sinha and Smith (2000), Estelami (2003) and Della Bitta et al. (1981) among many others have found price frame to have significant effect on perceived value of a promotion. A more extensive review of the literature will be presented in Section 2.

(5)

pay special attention to changes in promotional effectiveness over the promotional period. Such an investigation is lacking in promotions research.

This study will take the following format. In the next section literature concerning promotion frame will be reviewed. Then, in section 3, a pricing framework will be discussed that enables the classification of different price frames. Section 4 will provide a brief introduction to the type of data we will use in this study. In section 5 and 6 empirical analysis of two different products will be applied. The paper ends with the conclusion and a discussion of further research avenues and limitations of the study.

2: Literature Review

Literature on price framing to date has generated conflicting outcomes. A meta-analysis by Krishna et al. (2002), which incorporated 20 studies, 345 observations, and 6 “deal frames”, drew the conclusion that no deal frame performed particularly better than any other. Conversely, experimental studies by Sinha and Smith (2000), Estelami (2003) and Della Bitta et al. (1981) among many others have found price frame to have significant effect on perceived value of a promotion. In this paper, I endeavour to give an overview of price framing literature to date, with the aim of demonstrating the diversity of results and resultant lack of empirical generalizations in this field. Second, I will propose a framework which might be used as a preliminary step in classifying price frames and hence enabling more comparable research to be undertaken.

Why is Price Frame Important?

(6)

to both retailers and manufacturers”. Precisely for this reason, price framing research is of interest.

When a consumer is presented with a price promotion, the information within the promotion must be evaluated and processed. Because different promotions are evaluated and processed in different ways, the perceived value of a promotional offer, and hence effectiveness of the promotion, tends to differ across promotion frames (Chen, Monroe and Lou, 1998; Grewel et al., 1996). More generally, Thaler (1985) and Monroe and Chapman (1987) present the idea that different cues, in the form of advertisements, coupons, rebates and discounts, moderate product evaluation and hence willingness to purchase. This perceived moderating effect explains why research into the impact price frame on promotion effectiveness is important.

Mental Accounting

(7)

Percentile Discounts – Better or Worse?

Findings that a percentile discount is more difficult to process and hence may lead to greater processing and so greater perceived value is supported by the mental accounting principles. Heath, Chatterjee and France (1995) use this argument to present the heuristic that observation of a high percentage discount indicates a good deal (as indicated by price perception) and go on to conclude that if percentage discounts are large, then such percentiles stated as promotions framed in this way give higher perceived value. Thaler (1999) also used the mental accounting argument, finding that integrating price into a single amount is perceived to have lower disutility than segregating prices into their individual components. This finding might be especially relevant if a promotion frame using bundles of products is being considered.

(8)

Having overviewed only a sample of the literature on promotion frames, it is already becoming apparent where the problem concerning comparability of the studies arises. Let us take for example the Barnes (1992) paper and that by Sinha and Smith (2000). As mentioned above, the “50% off frame” was found to have the highest perceived value by Sinha and Smith (2000). However, how can this outcome be compared to the outcome by Barnes (1992)? What is noticeable in this paper is that the true dollar amount is presented in all price frames, hence highlighting the fact that papers are hard to compare as they all investigate different combinations of pricing frames. Therefore, although the frame that begins in a percentile is outperformed in this case, can this outcome be directly compared to that of Sinha and Smith (2000)?

Furthermore, there is evidence that experimental research which takes the price frame as the only differential feature between promotions on products is not very realistic. Grendall, Hoek and Pope (2005) attempt to create a more realistic experimental setting by allowing both brand and price frame to vary. It was found that the depth of the price promotion moderated the effectiveness of different frames. Specifically, they found that a price discount framed in dollars or cents to be more effective for high-priced items, but a percent discount was more effective at lower-prices. Additionally, they found brand to moderate effectiveness. However, regardless of the extensions this research makes, it is still questionable how it can be compared to prior research, where different frames were used and conflicting results can be found.

Financial Implication of Purchase

As was briefly mentioned above in discussion of the Das (1992) paper, the financial implication of the deal has in some cases been found to moderate the effectiveness of particular frames.

(9)

discount, consumers do not process information fully as the monetary gain is less and at high discount levels, there is less uncertainty about the potential monetary gain, so consumers feel less of a need to process information in it entirety. The conclusion is met that at these low and high discount levels, percentile offers are beneficial. This intuitively stems back to the mental accounting rationale that was discussed earlier. Hardesty and Bearden (2003) extend on this research, finding that price discounts and bonus packs are valued similarly for both low and moderate promotional benefit levels, while price discounts are preferred when promotional level is high.

Gains vs. Losses

In promotions literature, a distinction might be made between a gain or a loss frame. Idson et al. (2000) define a gain as getting a discount and a loss as avoiding payment. Literature concerning gains and losses tends to follow this definition closely. The idea of gains vs. loses is not isolated within price frame literature. Brendl et al. (1995) look at whether hypothetical consumers of airline tickets would view saving extra money (i.e. a gain) as positively as not having to spend extra money (i.e. making a loss) or vice versa.

Research concerning gains vs. loss frames in promotion literature is relatively new. Kim and Kramer (2007) is one of the first papers to look at gains vs. losses in a promotion framework. Specifically, they investigate the effect of framing a coupon from a gains perspective, “get 20 percent off the final price”, versus a loss perspective, “don‟t pay 20% of the final price”. They conclude that in a promotions setting, a gains frame is preferred. This finding is attributed to a processing fluency mechanism. In their own words, “consumers‟ deal perceptions are more favourable, their perceived savings are greater, and their intentions to take advantage of the price promotion are higher, when coupons highlight gains to be obtained, as compared to normatively equivalent losses to be avoided”.

(10)

Novelty

One issue that Kramer and Kim (2007) wish to highlight is that novelty induced deal elaboration does not occur. For the case of their paper, this claim is justifiable. If novelty induced deal elaboration did occur, they should find that the loss frame leads to higher perceived value of a coupon, where in fact the antithesis is found. However, there is research in the price framing field that contradicts these findings. Gupta and Cooper (1992) claim that novelty leads to the need for higher processing power than the equivalent common discount presentation and go on to claim that therefore novel price discounts are often underestimated. Conversely, Aaker and Williams (1998) find that novelty leads to increased elaboration of the advertised message, as was found in an experiment incorporating cultural differences in promotion presentation. They attribute what might be called a cultural mismatch to promotion frame effectiveness.

The novelty argument is furthermore supported by Kim and Kramer (2006) who find that if an individual processes promotion information systematically as opposed to heuristically they are more likely to process the information correctly and furthermore, more likely to purchase. Since it is believed that deeper processing of information will occur when a more novel promotion is encountered, one might conclude that novelty leads to higher perceived value of a promotions and also higher purchase probability. Earlier research suggested that perceived value of a frame is affected by the level of systematic processing (Wegener et al., 1993).

However, the processing fluency theory remains a strong counter argument to the novelty theory. Lee and Aaker (2003) claim that a less novel frame makes a promotion easier to process, so leading to increased persuasive power of said promotion.

Bringing the Literature Together

(11)

1. the topic being analyzed is well defined,

2. there are at least three articles by at least three different authors in which empirical research has been conducted in the specific area (by different authors) and,

3. the empirical evidence is consistent (i.e. the sign of the effect is the same in each study)

Taking this definition into account, as can be seen from the above reviewed literature, it is impossible to draw any empirical generalizations in promotions field where price framing is concerned. The results above indicate that the price frame of a promotion does in fact affect perceived value. However, where exactly this effect lies cannot be specified. It seems that no two studies investigate exactly the same frames and if similar frames are considered, results are often conflicting.

Further evidence of the lack of empirical generalizations in this area comes from a meta analysis by Krishna et al. (2002). This meta-analysis Krishna incorporated 20 studies, 345 observations, and 6 “deal frames”, drew the conclusion that no deal frame performed particularly better than any other.

What must now be asked is how can research on price frame proceed with the final aim of being able to find empirical generalizations. In the final sections of this paper, a framework is introduced which enables the classification of different types of promotions. If price frames are to be successfully compared, they must first be clearly classified. The below proposes one possible way in which such classifications might be made. If a coherent classification system can be instilled, the foundations for building a more comparable shelf of research are strengthened.

3: Classification

Step 1: Types of Sales Promotion

(12)

A sales promotion (as defined in Kotler, Wong, Saunders and Armstrong, 2005) is said to consist of “short-term incentives…to encourage the purchase or sales of a product or service”.

A sales promotion may follow one of three “flows”: manufacturer  retailer, manufacturer  consumer or retailer  consumer. These are called trade promotions, consumer promotions and retailer promotions respectively (Blattberg and Neslin, 1993). The flows can be visualised in Figure 1 below.

Figure 1: Sales Promotion Types and Flows

Most studies concerning the effectiveness or the perceived value of different price frames are looking at retailer promotions. This can be logically induced – when a promotion is being framed, it is often done at the retailer‟s discretion, hence explaining the different price frames that can be found over time within retailer and between retailers at any one time. However, it is important that the different possible types of promotions are distinguished between if any studies are to be compared and contrasted.

Step 2: Price Framing

The second step in classification is the specific classification of the price frame of the promotion. Here I introduce the method proposed by Dijk (2006). She says that the

(13)

type of price frame can be identified in a stepwise process as laid out in figure 2 below.

The first step is to identify whether there is a comparative reference price. The reference price may be defined as a price against which any discount is compared. In the case of there being no comparative reference price, it could be that there is no promotion. Alternatively, there may be a price discount but it may not be advertised to the consumer. In the case where there is a reference price, one must decide whether the price comparison is made within or between stores. If the reference price is equal to the retailer‟s regular price, then the comparative pricing is made within stores. If the reference price is comparing prices between, a comparison is either being made to the price offered by other retailers or to the manufacturer‟s suggested price.

What must then be determined is whether the promotion (made with or without a comparative price) involves one product (e.g. “50% off”) or multiple products (e.g. “3 for the price of 2”). Then, the way in which the promotion is presented to the consumer must be determined. Regardless of whether the promotion involves one or multiple products, the deal might be presented in a dollar or percentage form. Alternatively, if the promotion involves multiple products, it could be presented in a “one free with several” format. Finally, it must be determined whether the promotion is tensile or non-tensile. If a promotion is tensile, it is objectively framed and the deal is unambiguous, such as 30% off. Alternatively, if a deal is non-tensile it is more ambiguous, such as “up to 50% off”. If the promotion is tensile, it might give a maximum level of discount, a minimum level of discount or a range of a discount.

(14)

4: Data

Data Handling

The data used for this study came from a large department store chain in the Netherlands. A distinguishing feature of this store is that it sells all of its products under its own brand name. Therefore, there is no competition between brands within any of the stores.

Two products will be modeled in this study – first cotton wool pads and secondly rookworst – a type of dried sasauge sold in the Netherlands. These products were chosen because they were identified as having a relatively large number of price promotions during the sampling period and there were different price frames used to present these price cuts.”

Prior to the datasets being used to carry out any modelling, it had to be examined and any outliers removed. The following steps were taken to identify possible outliers that were subsequently removed:

1. Identify outlying store:

a) stores which are not open throughout the sampling period

b) stores where the product is not sold over the majority of the sampling period

c) stores where the product was only sold inside a promotion period d) stores where the price is reduced outside of the promotion period e) stores where there are no sales outside of the promotion period

f) stores whose regular price is not approximately equal to that of other stores belonging to the retailer.

2. Identify outlying days:

a) days which are at the very beginning of the sampling period and have a reduced price as it is therefore not possible to identify when the promotion started and promotion day etc.

(15)

a) observations outside of a promotion where the regular price is notably different from the average regular price,

b) observations where the discount is greater than 50% of the regular price,

c) negative sales/prices (if this was observed, observation were not removed, but the values were set to zero).

4. Identify extremely large unit sales

a) are there extremely large unit sales that lie outside of the promotional period? If this was found observations were not omitted, but the unit sales value was replaced with a rounded off mean of sales on non-promotion days for that particular store.

Data on two separate products was organised in the way described above and the results of this data analysis will be presented below. First we shall present the results of modelling sales of cotton wool pads and then the results using data on sales of rookworst.

5: Empirical Investigation 1 – Cotton Wool Pads

The first product that will be modelled in this study is cotton wool pads, which are sold in packs of four. Note that cotton wool pads were also sold in single unit packs, and this substitute will be accounted for in the model building process.

Promotions Data

(16)

Figure 3: Promotions Offered

The first promotion is a “buy 3 get one free” offer. Using the price framing outlined in Figure 2, this promotion can be identified as follows:

1. comparative pricing 2. within store

3. regular price 4. multiple products 5. one free with several 6. non-tensile

This promotion occurs 3 times over our sample as a store wide promotion.

The second promotion is a “was €3.20, now €2.40”. Using the price framing outlined in Figure 2, this promotion can be identified as follows:

(17)

3. regular price

4. one product (note that the 4 pack of cotton pads are sold as a single product) 5. euro amount

6. non-tensile

What must be noted, however, is that in the case of the second promotion, the 4 bundle of cotton wool pads are never sold at the advertised regular price of €3.20 and are only ever sold at the advertised reduced price of €2.40. €3.20 is the cost of buying 4 single tubes of cotton wool pads were there no discount.

The department store in question only begen to offer the promotion “buy 3 get one free” from week 40 of 2005. Prior to this date this promotion did not exist. Therefore, the data used in this research will begin in week 40 of 2005. Observations are obtainable up until week 6 of 2007.

Over the period inestigated there were 6 store wide promotions. These are introduced in Table 1 below. Three promotions used the price frame “3+1 free” and three the frame “Was__, Now__”.

Table 1: Price Promotion Timetable

Start End

3 + 1 free 6-03-06 Monday 19-03-06 Sunday Two weeks no promotion

3 + 1 free 3-04-06 Monday 17-04-06 Sunday Six weeks no promotion

New price 29-05-06 Monday 11-06-06 Sunday Twelve weeks no promotion

3 + 1 free 4-09-06 Monday 17-09-06 Sunday Four weeks no promotion

New price 16-10-06 Monday 29-10-06 Sunday Ten weeks no promotion

New price 8-01-07 Monday 21-01-07 Sunday Two weeks no promotion

Sales Data

(18)

Table 2: Average unit sales

New Price 3+1 free No promotion All observation Mean 7.148 7.583 2.189 4.442

Standard Deviation 7.807 7.498 3.603 6.201

As can be seen, there is clearly a positive effect of sales promotions on unit sales. Looking at mean data, it seems that the “3+1 free” promotion generates slightly higher mean unit sales that the “was €__, now €__” or new price promotion. A test for the difference between means (see Appendix 1), shows that there is a significant difference between the mean unit sales in the new price promotion versus the 3+1 free promotion.

Figure 4: Unit Sales and Promotion Periods

(19)

free” promotion and the lower bars indicate s “new price” promotion. By observation of the graphs a clear correlation between promotion period and increased unit sales can be seen.

What can also be observed in the graphs is that unit sales tend to peak at some point in the promotion period instead of being at an equal height throughout the period. This is evidence that there may be an optimal promotion length. Also, it can be seen that unit sales are elevated either side of the promotions, giving evidence for lead and lag effect.

Model Specification

Before proceeding, let us once again make reference to the main research question:

“To What Extent is Promotional Effectiveness Moderated by Type of Price Frame and Temporal Patterns”

I therefore endeavour to develop a model which will enable these questions to be answered.

The model that will be used is a hierarchical linear model that enables the incorporation of the store level involved in the data and also the longitudinal nature of the data. Such a model allows for the incorporation of random effects. In other words, it allows for parameter estimates to vary over stores and thus accounts for possible correlation of observations between stores. This may lead to a better model specification.

(20)

for i=1,…I stores and t=1…T days, where

it

S

= adjusted sales, which equals unit sales in store i on day t (Sit) divided by the

average sales in store i on non-promotion days (S ):i Sit Sit/Si.

WDpit = a dummy for the day of the week in store i on day t; p=1 for Tuesday, ...,

p=6 for Sunday.

Drieit , NPit = dummies which equal 1 if there is a promotion on day t in store i and

the price presentation is “3+1 free” or “new price,-” respectively, and 0 otherwise.

PDhit = a dummy which equals 1 if day t in store i is the hth promotion day, 0

otherwise.

Seassit = a dummy for different seasons; s=1 for autumn 2004… s=4 for summer

2005.

AV81it= availability of single tubes of cotton pads

Disc81it = a dummy, =1 if there is a discount on the single-unit product

it

S~₈₁ = adjusted sales of single unit product (calculated similarly to adjusted sales of 4-unit product as described above)

Lag_Drielit = a dummy for the lth day after a promotion in store i on day t if the

promotion is of the frame “3+1 free”, l=1,…, L.

Lag_NPkit = a dummy for the lth day after a promotion in store i on day t if the

promotion is of the frame “New price”, k=1,…, K.

Leaddit = a dummy for the rth day before a promotion in store i on day t, q=1,…,Q.

it

=

_error term.

i = random effect of discount across stores.

(21)

Restricted iterated generalized least squared (RIGLS) methods will be used to estimate the model. Parameter estimates generated using RIGLS are equivalent to restricted maximum likelihood estimates (REML), which produce unbiased estimates when a multivariate normal distribution is assumed for the random effects (Goldstein, 1989).

Before the results of running the model are presented, a brief discussion of the chosen explanatory dependent variables will be undertaken.

Week Day

Since daily data is being modelled, it is appropriate to model for variation in sales given the day of the week. One might expect that sales would be higher on weekend days, Friday and Saturday, for example, as opposed to days at the beginning of the week. This may also explain why promotional effectiveness seems to fluctuate over the promotional period. To account for different effects of days of the weeks, dummy variables are created for days of the week Tuesday until Sunday, Monday being used as the control day. Any pattern found to exist over days of the week may help answer the question of temporal patterns over the promotion period.

Price Frame

By including dummy variables indicating the use of different price frames during different promotional periods, one is able to investigate one of the main research questions of whether different price frames encourage different promotion effectiveness.

Promotion Day

(22)

Single Unit Cotton Wool Pads

The only logical substitute in store for the 4-unit cotton wool pads is the single unit cotton wool pads. Therefore, it is appropriate to account for possible substitution effects between the two products. For this reason, variables indicating whether the 1-unit pack was available in store, whether the 1-1-unit pack was on offer and the adjusted sales for the 1-unit pack in the same time period are included in the model.

Lead and Lag Effects

As was discussed in the literature review, there is evidence that promotions lead to the deference and acceleration of purchases. Therefore, it is important to incorporate these possible effects into the model. To do so, dummy variables are including indicating a number of days before and after the promotion. To account for possible differences in effects for different price frames, the lead and lag variables are interacted with the price frame dummy variables.

Choosing a Model

As was mentioned above, RIGLS methods will be used to estimate the model, which generate results identical to REML estimation methods. This method is preferred over maximum likelihood methods as maximum likelihood methods tend to underestimate the variance components of the random effects and the error term (Pinheiro and Bates, 2000).

Model Building

Although the general model has been specified, to arrive at the optimal model, certain decisions have to be made about which exact variables should be included in the specification and not. This was done is a stepwise process.

(23)

between Monday and the first promotion day etc. Therefore, although looking at the effect the day of the week has on promotion effectiveness would be interesting to investigate, it must be removed from the model.

Table 3: Models with different lead and lag effects and different random effects

Model No. parameter estimates

Log restricted likelihood

1. All possible variables 119 -86085.108

2. Eliminate days of the week because of collinearity with promotion day

113 -86099.653

3. with promotion day, without lead effects, 6 weeks of lags, without days of the week, random intercept

106 -86277.902

4. with promotion days, without lead effects, 5 weeks of lags, without days of the week, random intercept

92 -86412.582

78 -86447.968

64 -86497.929

7. 7 with promotion day, without lead effects, 2 weeks of lags, without days of the week, random intercept

50 -86650.072

8. with promotion day, with lead effect, 5 weeks of lags, random intercept

99 -86179.78

9. 8 with promotion day, with lead effects, 6 weeks of lags, random effects for promotion variables.

114 -82692.24

In order to decide on the optimal length of leads and lags to include in the model, a series of equations were run with different lag and lead lengths. To decide on a preferred equation, the likelihood ration test was used (see Appendix 2). A selection of the models run are listed in Table 3 below, followed by the results of some of the likelihood ratio tests in Table 4.

Table 4: Likelihood Ratio Tests Nested model

(regression no.)

Non-nested model (regression no.)

Accept H0 – nested model the

(24)

Table 4 only includes a selection of the likelihood ratio tests of models thought to be of most importance in this study.

Following these tests, it can be concluded that the preferred model include 6 weeks of lags, the 7 lead dummies and no random effects on any explanatory variables, but instead having a random intercept only.

Final Model

The results to the regressions will be presented in sets of parameters to make analysis more clear and accessible. Tests on the residuals confirmed that all assumptions underlying multilevel analysis (Maas and Hox, 2004) held.

First, I shall look at the promotion frame parameters. Both parameters are significant1 at the 10% level, being 1.0440 for the price frame “3+1 free” and 0.4025 for “New price”. As can been seen, the bundle promotion appears to have a greater effect on relative sales than the new price offer. A significant difference is confirmed by a Chi-test, testing the null that the two parameter estimates are equal to each other. This null is rejected at the 5% level ( 2 256).

I will now investigate the parameter estimates on promotional days. First note that the dummy variables on promotional days 14 and 15 were dropped because of collinearity problems. The parameter estimate on the 8th promotional day is insignificant. All other parameter estimates for the promotional day dummies are significant and are plotted in Figure 5 below.

1

(25)

Figure 5: Effect of Promotion Day 1.00 2.00 3.00 4.00 5.00 6.00 7.00 9.00 10.00 11.00 12.00 13.00 PromDay 0.50 1.00 1.50 2.00 2.50 3.00 3.50 Mean Effect

As can be seen in Figure 5, the effect of a promotion is much higher in the first week than the second week. It peaks in the weekend of the first week and then falls dramatically at the beginning of the second week. However, it rises slightly at the end of the promotion period. In addition, a Chi-test was used to confirm that there is a significant difference between the promotion day parameter estimates ( 3368). Therefore, the conclusion can be drawn that the promotional effectiveness does differ across the promotional period.

All season variables are negative and significant at the 5% level. They are 0.1664, -0.4334, and -0.1543 for spring, summer and autumn respectively. This implies that relative unit sales in winter are greater than in any other season.

(26)

product. As expected, the relative sales of the single unit pack of cotton wool pads has a negative and significant (at the 5% level) impact on relative sales of 4-unit packs of cotton wool pad (-0.1172, s.e.=0.0149).

The final variables of interest are the lead and lag effects. Lead effects were considered for only 7 days prior to the promotion. These effects are presented in Table 5 below.

The first interesting observation that can be drawn from these parameter estimates on the lead effects is that most of the estimates are positive, which contradicts the theory that consumers defer purchase in anticipation of a promotion. The lead effects for 1 day before a promotion and 6 days for a promotion are insignificant at the 10% level. The only lead effect that shows the anticipated sign is that for 7 days before the promotion commences and this is only significant at the 10% level. All other parameter estimates are positive and significant at the 5% significance level.

Table 5: Lead Effects

Parameter Coefficient p-value Lead1 0.3262 0.123 Lead2 1.4267 0.000 Lead3 0.8670 0.000 Lead4 0.2997 0.005 Lead5 0.5833 0.000 Lead6 -0.0516 0.641 Lead7 -0.2580 0.062

The final parameters I will look at are the lagged effects. Several of the lagged parameters were insignificant. See Appendix 4 for the output and a list of insignificant lagged effects. The parameter estimates are shown in Figure 6 below.

Note in Figure 6 that the bundle line refers to lagged effects of “3+1 free” promotions and the single line refers to lagged effects of the “Was €__, now €__” promotion.

(27)

occur over the later lags of the “3+1 free” promotion and between the 15th

and 20th lag of the “New Price” promotion. One possible explanation is that there are other forces within the market influences sales in these lagged phases that are not accounted for in out model.

Figure 6: Plot of Lag Effects

2.00 7.00 15.00 19.00 20.00 22.00 23.00 24.00 25.00 28.00 30.00 32.00 33.00 36.00 38.00 39.00 40.00 41.00 42.00 lag -1.00 -0.50 0.00 0.50 1.00 Mean bundle single

Over the earlier lags, it seems that the stockpiling effect of promotions is greater in the “3+1 free” promotions. However, after the 22nd

day of lags, the “new price” promotion seems to have a greater negative effect on sales. Overall, due to a substantial number of lags being found insignificant and the unusual positive lagged effects found, it is hard to draw any solid conclusions concerning lag effects in this model.

(28)

iwas estimated to be normally distributed with mean zero and variance 2.0575 (s.e. – 0.0073), meaning there is evidence that there is a significant difference between relative sales over stores.

6: Empirical Investigation 2 – Rookworst

Promotion Data

The second product to be investigated is rookworst. Rookworst is a kind of smoked sausage that originated from the Netherlands. Two main competitors to this product were identified – half rookworst and diet rookworst.

Over the sample there are three types of promotion offered: 20% off, 2 for €3 and €1.50 each. These promotions can be categorized using the price frame framework as laid out in Table 6 and these categorizations are outlined below:

The first promotion is a “20% off ” offer. Using the price framing outlined above, this promotion can be identified as follows:

1. comparative pricing 2. within store 3. regular price 4. single product 5. percentage discount 6. non-tensile

The second promotion is a “2 for €3”. Using the price framing outlined above, this promotion can be identified as follows:

(29)

6. non-tensile

Table 6: Price Promotion Timetable for Rookworst

Promotion In brochure? Start Day no End Day no

1 Now 20% off Yes 14-11-05 43 Monday 27-11-05 56 Sunday 2 weeks no promotion

2 Now 2 for €3 No 12-12-05 71 Monday 18-12-05 77 Sunday 3 week s no promotion

3 Now 2 for €3 Yes 09-01-06 99 Monday 22-01-06 112 Sunday 2 weeks no promotion

4 Now €1.50 Yes 06-02-06 127 Monday 19-02-06 140 Sunday 5 Now €1.50 No 20-02-06 141 Monday 26-02-06 147 Sunday 3 weeks no promotion

7 Now €1.50 No 21-08-06 323 Monday 03-09-06 336 Sunday 2 weeks no promotion

10 Now 2 for €3 Yes 13-11-06 407 Monday 05-12-06 429 Tuesday 1.5 weeks no promotion

11 Now €1.50 No 15-12-06 439 Friday 17-12-06 441 Sunday 3 weeks no promotion

12 Now 2 for €3 No 08-01-07 463 Monday 21-01-07 476 Sunday

The third promotion is a “€1,50 each”. Using the price framing outlined above, this promotion can be identified as follows:

(30)

5. euro amount 6. non-tensile

The data used in this research will begin in week 40 of 2005. Observations are obtainable up until week 6 of 2007 420 daily observations for each store

Over the period investigated there were 12 store wide promotions. These are introduced in Table 6 below.

Sales Data

In Table 7below some descriptive statistics for the unit sales data can be seen.

Table 7: Average unit sales*

20% Off 2 for €3 €1.50 each No promotion All observation Mean 127 139 85 59 80 Standard Deviation 98.55 118.60 76.40 61.24 74.40 *To the nearest whole SKU.

As can be seen, there is, once again, a positive effect of sales promotions on unit sales. Looking at the mean data, it can on be seen that there appears to be a positive impact of promotion on sales in SKU‟s. Also, it is noticeable that the promotions “20% off” and “2 for €3” appear to generate a higher mean unit sales than the promotion “€1.50 each”. More formally, the test for differences between means shows that all means are significantly different from each other.

Model Specification

As with the cotton wool pads, hierarchical linear modelling will be used to investigate the impact of promotion frame and temporal pattern on weighted sales of rookworst.

(31)

for i=1,…I stores and t=1…T days, where

it

S

_{= adjusted sales of rookworst, which equals unit sales in store i on day t (}Sit)

divided by the average sales in store i on non-promotion days (Si):

S

_it

S S

_it

/

_i.

Percentit, Bundleit, Singleit = dummy variables which equal 1 if there is a promotion

on day t in store i and the price presentation is “20% off” , “2 for €3”, or “€1.50” respectively, and 0 otherwise.

PDhit = a dummy which equals 1 if day t in store i is the hth promotion day, 0

otherwise.

Monthsit = a dummy for different months; s=2 for February. s=3 for March etc.

Disc02it = a dummy =1 if there is a discount on the competiting product – diet

rookworst

, = adjusted sales of diet rookworst and half rookworst (calculated similarly to adjusted sales of regular rookworst as described above)

Lag_Percentlit = a dummy for the lth day after a promotion in store i on day t if the

promotion is of the frame “20% off”, l=1,…, L.

Lag_Bundlekit = a dummy for the lth day after a promotion in store i on day t if the

promotion is of the frame “2 for €3”, k=1,…, K.

Lag_Singlekit = a dummy for the lth day after a promotion in store i on day t if the

promotion is of the frame “€1.50 each”, k=1,…, K.

Leaddit = a dummy for the rth day before a promotion in store i on day t, q=1,…,Q.

it

= error

term.

i = random effect of discount across stores.

(32)

Model Building

The means by which a final model was arrived at differs slightly from the approach used for the cotton wool pads. The model specification was highly driven by the observations of the data made prior to the model being run. Looking at the patterns in the data, two issues arrive.

First, it is apparent that there is often no more than two weeks between successive promotions. Therefore, it was thought unreasonable to include lag effects of more than two weeks in the model. Including greater lags would be likely to induce collinearity in the model. Even if collinearity were not present, interpretation of any significant parameters would be difficult as any new promotions might hide or confound the effects of lagged promotions and vice versa.

Second, from observation of the plot of sales and promotion against time, it was made clear that little pattern from the promotions could be seen due to what appeared to be a strong annual cycle. To take out this cycle, monthly dummies were included in the model.

Final Model

Once again, the results to the estimated HLM model will be presented in sets of parameters to aid clear understanding of the results.

First, we will examine the promotion frame parameter estimates. The parameter estimate for “20% off” and “2 for €3” are significant, but the parameter estimate on the third promotion frame – “€1.50 each” is very small and in fact insignificant at the 10% level (although note should be made that this promotion frame intercepted with lags does produce significant effects – this will be discussed more later). Referring back to the two significant promotion frame parameter estimates, the parameter estimate on the “2 for €3” is significantly greater (0.825) than that for “20% off” (0.258)2.

2

(33)

Now turning to promotion days, note must first be made that at the 10% level, the parameter estimates on the dummy variables representing promotion day 8, 14, 20 and 24 are insignificant at the 10% level. Days 14 and 20 are both Sundays, which may contribute to the insignificance as many shops are closed on Sunday, so there is likely to be fewer shoppers in general and so less interest in promotions. The insignificance for day 8 might be explained because of the loss of interest in the promotion after the first weekend of the promotions where sales peak massively. However, it is hard to draw any reason as to why day 24 of the promotion is insignificant, apart from the fact that very few promotions run for 24 days, so there may simply not have been enough data to get a significant effect. All other parameter estimates for promotion day were significant at the 5% level and are plotted in Figure 7 below. As can be seen, sales are noticeably higher in the fist week of the promotion

Figure 7: Effect of Promotion Day for rookworst

(34)

pointed out that most promotions only lasted for two weeks. Overall, it can be concluded that promotional effectiveness does vary over the promotional period, and generates higher sales during the first week of the promotion and that there tends to be peaks towards the end of each promotion week.

All month variables are significant at the 5% level. The coefficient estimates can be seen in Table 8. Noticeable is that the sales of rookworst in the winter months is greater than in the Summer months. These findings correspond with the observations made in the plot of sales against time, and suggest that we were correct to include month dummies. Table 8 Month Coefficient P>z February 0.2201034 0 March -0.4997332 0 April -0.6317114 0 May -0.8061267 0 June -0.9714322 0 July -1.047575 0 August -0.8365844 0 September -0.6702312 0 October 0.1994851 0 November 0.3272373 0 December 0.4903381 0

The two main competitors to regular rookworst are diet rookworst and half rookworst. Positive weighted sales of diet and half rookworst have a small, but positive and significant effect on sales of regular rookworst, being 0.024 and 0.005 respectively. This is surprising as one would expect the effect of sales of competing products to be negative. However, this positive effect might be attributed to the fact that purchasing the competing products draws consumers to notice the regular rookworst on the shelf as well. As would be expected, however, the existence of a promotion on the diet rookworst has a negative and positive effect on weightes sales of regular rookworst (coeff. = -0.227, s.e = 0.008).

(35)

similar observation can be made here as was made with the lead effects in the cotton wool pads example – namely that the parameter estimates for most of the lead effects are positive. Especially of interest are the 2 and 3 day lead effects, where the effect is quite large. Note must also be made that the lead effect for 4 days is insignificant at the 10% level. Only the 1st and the 7th day lead effects show the anticipated (negative) sign.

Table 9: Lead Effects for rookworst

Parameter Coefficient p-value

Lead1 -0.3954435 0 Lead2 0.2313999 0 Lead3 0.1089945 0 Lead4 0.0217503 0.136 Lead5 0.0456326 0.003 Lead6 0.0309093 0.035 Lead7 -0.1843369 0

(36)

Finally, we will look at the lagged effects. Several of the lagged parameters were insignificant. The parameter estimates are plotted in Figure 8 below. What is interesting is that the expected sign of the lagged effects is negative, indicating stockpiling during promotions. However, as can be seen in below, many of the lagged effects for the promotions “20% off” and “€1.50” are in fact positive. It would be interesting to investigate the cause of this observed effect further. It is possible that the advertising of a promotions draws consumers to purchase a product even if the promotion has not yet been implemented.

What is also of interest is that the only promotion that has a sizeable number of lagged effects with negative values is the promotion frame that boosts unit sales the most – “2 for €3”. A possible reason for this is that the anticipation for these promotions by a consumer is greater and so the consumer is more encouraged to stockpile when the promotion is on.

Finally, we will pay note to the estimated intercept. A fixed intercept was chosen for this model, with an estimated value of 0.7817 (s.e, = 0.0985), as there was no significant difference was found between relative sales between stores.

7: Conclusion

Two issues of main interest in this piece of research were whether temporal patterns and pricing frames influenced promotional effectiveness.

(37)

promotion – being able to buy in bulk provides increased shopping convenience which appeals to the consumer (Chandon et al., 2000)

In addition, it was found that there are distinct temporal patterns surrounding a price promotion. Specifically, sales were much higher in the first week of the promotion as compared to the second week and subsequent weeks of the promotion. From the study of cotton wool pads, there is also evidence that sales in units will also peak that the end of a promotional period. From both studies it might also be concluded at unit sales over a promotional period are cyclical, peaking towards the end of each promotional week.

These two findings imply that when initiating a promotion, it is advisable that attention be paid to which price frame is used and the length of time for which a promotion should be held if a manager wants to maximise his returns from the promotion.

Finally, it was found that lead effects were positive, which is a surprising result. Findings concerning lag effects were ambiguous as some lag parameters were also found to be positive.

8: Limitations and Further Research

(38)

A second limitation is that for the products investigated here it was only possible to investigate five price frames in total. As evidenced by the price frame framework, there are in fact many more price frames that may be used which were not included in this study. Further research might incorporate different price frames and replication studies of the price frames considered in this study might be a step towards being able to make empirical generalizations. For example, it might be investigated whether the depth of a promotion or affects sales. Or is there a difference in effectiveness of a promotion framed in percentage rather than monetary unit form.

A third limitation is that although I looked at the promotional effectiveness over the promotions that run for full weeks and for 2, 3 or 4 weeks. Not all promotions are of these set lengths. Therefore, it is hard to make generalizations about, for example, what happens at the end of the promotion period as it is not possible to identify the end of such a period for every store in this output.

(39)

9: References

Aaker, J.L. and Williams, P. (1998) Empathy versus pride: the influence of emotional appeals across cultures. Journal of Consumer Research, 25, 241-261

Ailawadi, K.L. and Neslin, S.A. (1998) The effect of promotion consumption: Buying more and consuming it faster. Journal of Marketing Research, 35(3), 390-398

Allenby, J.M., Broniarczyk, S.M., Shimp, T.A., and Urbany, J.E. (1994). The influence of prior beliefs; frequency cues and magnitude cues on consumers‟

perception of comparative price data. Journal of Consumer Research, 21 (September), 219-235

Barnes, J.G. (1975) Factors influencing consumer reaction to retail newspaper “sale” advertising, Proceedings, Fall Educator‟s Conference, Chicago, American Marketing Association, 417-477

Beach, L.R. and Mitchell, T. (1978) A contingency model for the selection of decision strategies. Academy of Management Review, 3, 439-449

Bemmaor, A.C. and Rossi, P.E. (1991). Quality perceptions and asymmetric switching between brands. Marketing Science, 10(3), 185-235

Bettman, J.R., Johnson, E.J. and Payne, J.W. (1990) A componential analysis of cognitive effort in choice. Organizational Behaviour and Human Decision

Processes, 45, 111-139

Bijmolt, T.H.A, Van Heerde, H.J. and Pieters, R.G.M. (2005) New empirical

generalizations on the determinants of price elasticity. Journal of Marketing Research XLII, 141-156

Blattberg, R.C. and Neslin, S.A. (1990). Sales Promotions Concepts, Methods, and Strategies, Englewood Cliffs, NJ: Prentice-Hall

_____ and Wisniewski, K.J. (1989). Price-induced patterns of competition.

Marketing Science, 8(4), 291-309

Blattberg, R.C. and Neslin, S.A. (1993). Sales promotion models. Handbook in

Operations Research and Management Science 5, Elsevier Science Publishers B.V.,

Amsterdam, 553-609.

Blattberg, R.C., Breisch, R. and Fox, E.J. (1995). How promotions work. Marketing

Science, 14, 122-132

(40)

Chaiken, S. (1980) Heuristics versus systematic information processing and the use of source versus message cues in persuasion. Journal of Personality and Social

Psychology, 39(5), 752-766

Chandon, P., Wansink, B., Laurent, G. (2000) A Benefit Conruency Framework of Sales Promotion Effectiveness, The Journal of Marketing, 64 (4), 65-81

Chandon, P. and Wansink. B.(2002) When are stockpiled products consumed faster? A convenience-salience framework of postpurchase consumption incidence and quantity. Journal of Marketing Research, 39(3), 321-335

Chen, S.-F.S., Monroe, K.B., and Yung-Chein Lou (1998). The effects of framing price promotion messages on consumers‟ perceptions and purchase intentions.

Journal of Retailing, 74(3), 353-372

Chevalier, M. and Curhan, R.C. (1976). Retail promotions as a function of trade promotions: A descriptive analysis. Sloan Management Review, Fall, 19-32

Cooper, L.V. (1988). Competitive maps: the structure underlying asymmetric cross elasticities. Management Science, 34 (6), 707-723

Curhan, R.C. and Kopp, R.J. (1986). Factors influencing grocery retailers‟ support of trade promotions. report no. 86-114, Cambridge, MA: Marketing Science Institute, July.

Della Bitta A. K., Monroe, K.B. and McGinnis, J.M. (1981). Consumer perceptions of comparative price advertisements. Journal of Marketing Research, 18 (November), 416-427

Diamond, W.D. (1992) Just what is a „dollar‟s worth‟? Consumer reactions to price discounts vs. extra product promotions. Journal of Retailing, 68(3), 254-270 Diamond, W.D., and Campbell, L. (1989) The framing of sales promotions: Effects on reference price change. In S.Thomas (Ed.), Advances in Consumer Research, 16, 241-247

Diamond, W.D. and Johnson, R.R. (1990) The framing of sales promotions: An approach to classification. In M.E. Goldberg, G. Gorn and R.W. Pollay (Eds),

Advances in Consumer Research, 17, 494-500

Dijk, A. (2006). To what extent is promotional effectiveness moderated by price framing and temporal patterns. Masters Thesis, Rijksuniversiteit Groningen.

Doyle, P. and Saunders, J. (1985). The lead effect of marketing decisions. Journal of

Marketing Research, 22 (February), 54-65

(41)

Foekens, E.W., Leeflang, S.H. and Wittink, D.R.(1999). Varying parameter models to accommodate dynamic promotion effects. Journal of Econometrics, 89 (1-2), 249-268 Folkes, V.S., Martinf, I.M., and Gupta, K. (1993) When to say when: effects of

supply on usage. Journal of Consumer Research, 20, 467

Grewel, D., Marmorstein, H and Sharma, A (1996) Communicating price information through semantic cues: the moderating effects of the situation and discount size.

Journal of Consumer Research, 23, 148-155

Grover, R. and Srnivasan, V. (1992). Evaluating the multiple effects of retail promotions on brand loyal and brand switching segments. Journal of Marketing

Research, 29 (February), 76-89

Gupta, Sunil (1988) The impact of sales promotions on when, what and how much to buy. Journal of Marketing Research, 25(4), 342-355

Hardesty, D.M. and Bearden, W.O. (2003). Consumer evaluations of different promotion types and price presentations: the moderating role of promotional benefit level. Journal of Retailing, 79, 17-25

Heath, T.B., Chatterjee, S. and France, K.R. (1995) Mental accounting and change in price: The frame dependence of reference dependence. Journal of Consumer

Research, 22, 90-97

Kalwani, M.U., Rinne, H.J., Sugita, Y. and Yim, C.K. (1990). A price expectations model of customer brand choice. Journal of Marketing Research, 27 (August), 251-262

_____and Yim, C.K. (1992). Consumer price and promotion expectations: and Experimental study. Journal of Marketing Research, 29(February), 90-100

Kamakura, W.A. and Russell, G.J. (1989). A probabilistic choice model for market segmentation and elasticity structure. Journal of Marketing Research, 26 (November), 379-390

Kim, H.M. and Kramer, T (2006)The moderating effects of need for cognition and cognitive effort on responses to multi-dimensional prices. Marketing Letters, 17, 193-203

Kotler, P., Wong, V., Saunders, J., Armstrong, G. (2005). Principles of Marketing,

Fourth European Edition. Pearson Eduction Limited, Harlow, United Kingdom

(42)

_______ and ______ (1991). An empirical analysis of the relationship between brand loyalty and consumer price elasticity. Marketing Science, 10(2), 172-183

Kumar, V. and Leone, R.P. (1988). Measuring the effect of retail store promotions on brand and store substitution. Journal of Marketing Research, 25 (May), 178-183 Lattin, J.M. and Bucklin, R.E. (1989). Reference effects of price and promotion on brand choice behaviour. Journal of Marketing Research, 26 (August), 299-310 Leeflang, P.S.H. and Wittink, D.R.(1996) Competitive reaction versus consumer response: Do managers overreact. International Journal of Research in Marketing, 13(2), 103-119

Mace, S, and Neslin, S.A. (2004). The determinants of pre- and post-promotion dips in sales of frequently purchased goods, Journal of Marketing Research, XLI, 339-350 Mayhew, G.E. and Winer, R. (1992). An empirical analysis of internal and external reference prices using scanner data. Journal of Consumer Research, 19(June), 62-70 Moriarty, M.M. (1985). Retail promotional effects on intra- and interbrand sales performance. Journal of Retailing, 61(3), 27-47

Morwitz, V.G., Greanleaf, E.A. and Johnson, E.J. (1998) Divide and Prosper: consumers‟ reactions to partitioned prices. Journal of Marketing Research, 35, 453-463

Mulhern, F.J. and Leone, R.P. (1991). Implicit price bundling of retail products: a multiproduct approach to maximising store profitability. Journal of Marketing, 55 (October), 63-76

Neslin, S.A. (2002). Sales Promotion, Marketing Science Institue, Cambridge Massachusetts.

Nijs, V.R., Dekimpe, M.G., Steemkamp, J.E.M. and Hanssens, D.M. (2001) The category-demand effects of price promotions. Marketing Science, 20, 1-22

Parreno Selva, J. (2005). Modeling cross-promotion effects, Universidad de Alicante, Spain.

Payne, J.W. (1982) Contingent decision behaviour. Psychological Bulletin, 92, 382-4-2

Petty, R.E., Cacioppo, J.T. amd Schumann, D. (1983) Central and peripheral routes to advertising effectiveness: the moderating role of involvement. Journal of Consumer

Research, 10, 135-146

(43)

Raju, J.S. (1992). The effect of price promotions on variability in product category sales. Marketing Science, 11(3), 207-220

Shoemaker, R.W. (1979). An analysis of consumer reactions to product promotions.

Proceedings of the American Marketing Association, 244-248

Sinha, I., Smith, M.F. (2000). Consumers perceptions of promotional framing of price. Psychology and Marketing, 17, 257-275

Thaler, R.H. (1999) Mental accounting matters. Journal of Behavioural Decision

Making, 12, 183-206

Thompson, P.A. and Noordewier, T. (1992). Estimating the effects of consumer incentive programs on domestic automobile sales, Journal of Business and Economic

Statistics, 10(4), 409-417

Van Heerde, J.V., Leeflang, P.S.H. and Wttink, D.R. (2001) Semiparametric analysis to estimate the deal effect curve. Journal of Marketing Research, 38, 198-220

Van Heerde, H.J., Leeflang, P.S.H. and Wittink, D.R. (2004). How promotions work: SCAN*PRO-Based Evolutionary Model Building, Schmalenbach Business Review, 54, 198-220

Van Heerde, J.V., Leeflang, P.S.H., Wittink, D.R. (2004b) Decomposing the sales promotion bump with store data. Marketing Science, 23, 317-334

Vilcassim, N.J. and Chintagunta, P.K. (1992). Investigating retailer pricing strategies from household scanner panel data. Working Paper, February.

Walter, R.G. (1991) Assessing the impact of retail price promotions on product substitution, complementary purchase, and interstore sales discplacement. Journal of

Marketing, 55(April), 17-28

Wansink, B. and Deshpande, R. (2005) “Out of sight, out of mind”: Pantry stockpiling and brand-usage frequency. Marketing Letters, 5(1), 91-100

______ and MacKenzie, S.B. (1988). A structural equations analysis of the impact of price promotions on store performance. Journal of Marketing Research, 62(3), 237-266

Winer, R. (1986). A reference price model of brand choice for frequently purchased products. Journal of Consumer Research, 13(September), 250-256

Wittink, D., Addona, R.M., Hawkes, W. and Porter, J.C. (1988). SCAN*PRO: the estimation, validation, and use of promotional effects based on scanner data. Working

Paper, Johnson Graduate School of Management, Cornell Univesity, Ithaca, NY.

To What Extent is Promotional Effectiveness Moderated by Type of Price Frame and Temporal Patterns