M A R K E T IN G
A study on the effect of quality on advertising and promotional effectiveness
by
Kimberly Ollivierre
University of Groningen Faculty of Economics and Business
MSc. Marketing Intelligence June 20th 2016 Bloemstraat 47-23 9712LC Groningen (06) 39562922 k.s.ollivierre@student.rug.nl student number 1873598
ABSTRACT
Marketing has been losing steam and power in companies. In an effort to regain its previous position, managers are being held more accountable for its marketing investments made. This movement is apparent in the recent emphasis on exploring marketing effectiveness and various qualitative as well as quantitative interaction effect. To contribute to the expanding literature on marketing effectiveness, this study looks at a mass service provider. A SUR model is applied on aggregated data collected over 30 months on two different customer groups, single-purchasers and relational customers. The results reveal effect on the all three interaction models in various ways. For the relational customers, own promotion effectiveness on sales is reduced when objective and perspective quality decreases. For the single-purchase customers, objective quality trend increases own promotion effect but decreases retail promotion effects on sales. In addition, the results show the preceding elements affect relational-customers more and differently than single-purchasers.
PREFACE
First I want to thank Mr. Gijsenberg for his very valuable insight and support. More importantly, for his patience while navigating the hectic process of writing my thesis.
Secondly I would like to thank my Marketing Dynamics group for helping with doubts and overall support, especially those close days to deadline.
Last, but not least I want to thank my friends for listening to me complain and panic when things do not go my way. They served as a listening board when I have ideas, giving me feedback if things “sound weird” and just for being there for me and showing endless support.
Thank you!
TABLE OF CONTENTS
ABSTRACT ... 2 PREFACE ... 3 1 INTRODUCTION ... 5 2 THEORETICAL FRAMEWORK ... 7 2.1 Direct Effects ... 7 2.1.1 Marketing activities ... 7 2.1.2 Past performance ... 9 2.1.3 Perceived quality ... 92.2 Moderation On Marketing Activities ... 10
2.2.1 Perceived quality trend ... 11
2.2.2 Objective quality trend ... 12
2.2.3 Crisis ... 13 2.3 Conceptual Model ... 14 3 DATA ... 16 3.1 Description Of Data ... 16 3.2 Model Development ... 20 3.2.1 Multicollinearity ... 22 3.2.2 Correlated disturbances ... 22 3.2.3 Heteroscedasticity ... 22 3.2.4 Nonnormal errors ... 23 3.2.5 Stationary ... 23 3.3 Analysis Plan ... 23 4 RESULTS ... 24 4.1 Model Diagnostics ... 24
4.2 Main Effects of Promotion and Advertising ... 24
4.3 Main Effect of Preceding Occurrences ... 25
4.4 Moderating Role of Quality On Marketing Effectiveness ... 27
4.5 Robustness Check ... 28
5 DISCUSSION ... 29
REFERENCES ... 31
APPENDICES ... 36
A1. Appendix A: Data ... 36
1 INTRODUCTION
“Firms are under increasing pressure to justify their marketing expenditures.”
(Van Heerde, et al., 2013, p. 177)
“Marketing practitioners are under increasing pressure to demonstrate their contribution to firm performance.” (O'Sullivan & Abela, 2007, p. 79)
“The general conclusion has been that in many companies, the marketing function is in steep decline.” (Verhoef & Leeflang, 2009, p. 14)
Moorman and Rust (1999) found that marketing influence within a firm to be strongly related to accountability. Linking activities to actions has become essential to increase the importance of marketing departments as managers prefer activities with clear return of investment (ROI) measurements (Moorman & Rust, 1999; Kumar, 2004). The inability to deliver clear ROI has led to contradicting answers on the effect of advertising and promotion which in its turn has made marketing to be looked at as unproductive and inefficient. This research answers the call for an increase in attention for accountability and marketing metrics (Farris, 2006).
The rise of the digital world facilitates the collection of big data, and with it the development of better metrics for accountability. Since, what is the point of having all this information and still not getting the most bang out of your marketing buck? (Leeflang & Wittink, 2000). Marketing performance measurements have been shown to improve firm performance (Moorman & Rust, 1999; Morgan, et al., 2002). The increase in available data has presented the opportunity to analyze various effects on marketing effectiveness, e.g. Gijsenberg (2014) found that before and during major sports events own promotion effectiveness decreases and advertising increases around focused, single-sport events. Van Heerde, et al., (2013) compared marketing effectiveness during economic downturns and expansions and found long-term price sensitivity to decrease and advertising elasticities to increase during expansions. This study seeks to add to this body of work by looking at the moderating effects of quality on marketing effectiveness.
failure in transportation service. The Dutch railway NS recorded an increase of delays in 2015 and a decrease in customer satisfaction for that same year (NS, 2015). American airlines experienced customers defecting to competitors in 2012 when they had an increase in cancelled and delayed flights (Isidore, 2012).
These examples are prime indications of the importance for service providers, more specifically mass service providers. They show how behaviour and attitude of customers change after a service failure, causing defecting customers and a decrease in their satisfaction. In this case, creating a dire need to return to the pre-crisis stage. This need to recover can be seen in the extensive body of work on crisis management (e.g. Powell, 1995). A more limited work is available on the consequences beyond these behavioural and attitudinal changes (Van Heerde, et al., 2007).
Academic and managerial contributions are made by taking the interaction between quality and marketing activities into account. First, it adds additional metrics for marketing accountability by revealing key insights for marketing effectiveness. Second, when generalized, the results contribute to the less popular (and a bit neglected) mass service industry. Third, it provides new insights to managers for combating quality declines and/or failures. It also provides a new perspective into how service quality can improve firm performance or hinder performance improvements in combination with marketing activities. It also offers further insight into possibilities for marketing activities not performing as expected. This way making a contribution to the fight to increase the importance of marketing.
Using a Seemingly Unrelated Regression (SUR) model, marketing effectiveness on firm performance for two customer types are analysed three ways accounting for quality scenario. This is done with data from a major European railway company (mass service provider). The following research questions are answered:
1. What is the effect of promotional and advertising marketing activities on firm performance? 2. How are promotional and advertising elasticities affected by product crisis, change in
perceived quality and change in objective quality?
3. Are there differences between the two customer types, relational and single purchasers? In answering these questions, the main question can be answered: How marketing activities are moderated by objective and subjective quality for mass service providers?
2 THEORETICAL FRAMEWORK
There are a large number of studies on the marketing effect on goods (O'Sullivan & Abela, 2007; Van Heerde, et al., 2007; Ataman, 2010). However, since the focus of this study is on a services, (physical) goods will be referred to when emphasizing comparison between goods and services. Services differ from goods in that with services consumers are affected immediately by the product unlike goods where there is a buffer time between releasing the product to the market and consumption (Gijsenberg, et al., 2015; Huang and Rust, 2013; Zeithaml, et al., 2012). Furthermore, public transport differs from ordinary, one-on-one, services because it affects multiple people at the same time. This also means that the consequences of mass service failures are greater (Gijsenberg, et al., 2015). As for railway service, it holds the properties of both mass market products and traditional services. Railway service lacks the personal connection compared to traditional services, but has the effect of mass market products. It affects numerous people at the same time and there is no face to assign blame to if something goes wrong or compliment with exceptional service.
Moreover, the performance of the company is indicated by km-travelled, i.e. usage. Usage in previous studies has been used as a metric for firm performance, measured by monetary and count values as well as ratings additionally to outcomes of e.g. elasticity analysis (Leeflang, et al., 2015). For this reason, this study assumes kilometres travelled to be equal to units sold because in the railway service customers do not pay per trip, but per distance or kilometres travelled. There are two different customer type observed for performance, ticketholders and cardholders. Ticketholders are those who pay per km travelled and cardholder buy a season pass for unlimited travel. That is, a cardholder would not buy tickets and a ticketholder would not be a cardholder. But, these travellers are still related in the sense that they use the same service and extrinsic forces (such as weather) would influence both customer type. Although they are related, it is expected that they react differently to marketing activities, e.g. cardholders as relational customers are expected to be less sensitive to promotion than a ticketholder who are single-purchase customers (Garbarino & Johnson, 1999).
2.1 Direct Effects
For this study three direct effects are taken into account, marketing activities, past firm performance and perceived quality.
2.1.1 Marketing activities
Promotion and advertising are part of the marketing mix, which Ataman (2010) has shown to have a positive short-term effect on sales. Allenby and Hanssens (2004) and Dekimpe & Hanssens (1999) found an advertising elasticity of .01. Advertising has the same effect in long-run. During economic downturn advertising has a stronger effect, creates more financial benefits, than during economic expansions. The indirect effect of advertising is through its effect on brand image, awareness, differentiation, and brand equity (Dekimpe & Hanssens, 1999; Aaker, 1996; Van Heerde, et al., 2013). Advertising is defined as “any paid form of non-personal communication about and organization, product, service, or idea by an identified sponsor” (Belch & Belch, 2003, p. 16). In the short-term, as mentioned by Belch and Belch (2003), it is expected that advertising activities would lead to a sales increase. On average sales-to-advertising elasticities are .10 (Sethuraman et al., 1991; Tellis, 2003). The impact of advertising is dependent on the product category, durable goods have higher elasticities and advertising on experience are more effective (Sethuraman et al., 1991; Vaktrasas et al., 1999). The latter would lead to an expectation of higher advertising elasticity for mass service providers.
Sales promotion are “those marketing activities that provide extra value or incentives to the sales force, the distributors, or the ultimate consumer and can stimulate immediate sales” (Belch & Belch, 2003, p. 21). Rather, promotion temporarily cut prices. Customers are known to be value oriented, getting the most results for minimal costs (Heskett et al. 1994). The prospect theory, which suggests that outcomes that are probable get underweighted in comparison to those that are obtained with certainty, has been linked to this behaviour of cost aversion (Köbberling & Wakker, 2005; Kahneman & Tversky,1979), which would explain the short-term effect of promotion. Promotions have been shown to have short-term positive effect on sales, and no effect in the long-term, thus creating sales spikes (Jedidi, et al., 1999).
The term “promotional dip” has also been used to describe the decreasing sales that happen after the spike. Ataman (2010) found that marketing mix still has a positive direct effect after controlling for these sales spikes. In the long-term promotion has the opposite effect. Over time people have become more price sensitive, the average price elasticity has decreased over time (becoming more negative from -1.76 to -2.62) (Tellis, 1988; Bijmolt, et al., 2005). As promotion increases, price would decrease which in its turn would increase sales.
The literature supports that both promotional and marketing activities have a positive relationship with firm performance, therefore:
2.1.2 Past performance
That past performance influences current performance is expected. There are statistical and theoretical reasons behind this assumption. Theoretically, based on (Aaker, et al., 1982), current performances respond to past performance decisions. This is done through “good will “gained in the previous period, which is linked to brand loyalty created and marketing efforts made. Statistically, adding the lagged dependant variable decreases the chance of serial correlation and guarantees that the residuals hold little to no information. This is because the lagged dependant variable “acts as a proxy picking up some of the unmeasurable variables’ effect” (Achen, 2000, p. 7).
Because the customer groups are mutually exclusive, both performances of the preceding month are taken into account. These account for routine in customers’ behaviour and confound variable that carry over to the current month. Meaning, that if the sales in the previous month are lower because ticketholders switched to cards, this effect would be captured. If previous marketing decisions are made to increase loyalty or sales, this affects current customer decision making.
In addition, we see that marketing has become more than just satisfied exchanges, the focus has shifted to creating relationships (Belch & Belch, 2003). Relationships are built with those who return to make repeat purchases. This goodwill created with past sales and relationship building efforts leads to the belief that relational purchase customers will be more affected by past performances than single-purchase customers.
H2: Previous performance has a positive effect on current performance.
2.1.3 Perceived quality
As quality perception gains importance so does the knowledge surrounding it (Bolton & Drew, 1991; Carman, 1990; Zeithaml et al., 1990). Srinivasan, et al. (2010) include mind-set metrics in their sales response model and account for the marketing mix short- and long-term effects. They found that mind-set metrics can be used as warning signals, allowing ample time for managerial action before market performance itself is affected and that specific action effect specific mind-set metrics. That is, previous mind-set can determine current performance.
as a performance indicator. This comes back to the argument that customers recall previous performance, and make decisions based on those experiences (Carmen, 1990; Verhoef, et al., 2009). Marketing indirectly effects performance through attitudes. Liking metrics affects high sales conversion for all the participating product categories, liking to sales conversion of 0.48 on average. Showing a positive effect Hanssens et al., 2014). This expectation can also explain the relationship between satisfaction and perceived quality. Customers have a performance forecast that is (dis)confirmed by the current experience. Above this threshold indicates a positive reaction, satisfaction. The opposite reflects dissatisfaction, indicating quality as the predictor of satisfaction. There is a lack of study on service providers in this area, most studies deal with service as an additional offering and not as the core product. Service (the additional service) has been shown to be essential in success strategies and to creating competitive advantage (Sheth et al., 2009; Anderson & Sullivan, 1993). Ford shows above average profit for high service quality scores (Ford 1990). A study by Koska (1990) shows that this statement also holds when service is the core product. Koska states that a strong link between perceived quality and profitability was found by The Hospital Corporation of America (Anderson & Sullivan, 1990; Cronin & Taylor (1992). In conclusion, it can be expected that since satisfaction and perceived quality are closely related, they can both affect trust and commitment which in turn predict customer intentions (Cronin & Taylor, 1992). In the same way Rego et al. (2008) found service-focused firms to have a less negative correlation between customer satisfaction and market share than the average correlation of -.18. This negative effect is mostly due to the heterogeneity that accompanies wider customer base. In the transport industry this heterogeneity is expected to have a non-existent effect because of its mass service setting. Leading to the following hypothesis:
H3: Previous perceived quality has a positive effect on performance.
Furthermore, it should be noted that different strength is expected for relational and single purchase customers. Garbarino and Johnson’s (1999) study on customers of the New York off-Broadway repertory theatre company demonstrated that low-relational customers are more driven by satisfaction than high-relational customers. In this case, low-relational would be the ticketholders and high-relational cardholders.
2.2 Moderation On Marketing Activities
more responsive of marketing activities compared to those in a foul mood (Owalbi, 2009). And because people often base their judgement and behaviour on feelings/emotions, the implication of their action are expected to be consistent with that feeling (Owalbi, 2009). Studies on the intangible measurement and the importance of the indirect effect of marketing activities have been gaining importance (Bruce et al., 2012; Hanssens et al., 2014; Srinivasan et al., 2010). Failures and changes in the service affect the reference point, so memory of consumer, and expectations. This way changing the way following events are perceived (Mela et al., 1998).
It is expected that objective quality and perceived quality will perform differently. External service value drives customer satisfaction which indirectly drives revenue growth and profitability (Heskett et al., 1994). This link between actual quality and customer satisfaction can be further explained by Oude Ophuis and Van Trijp (1995) and Steenkamp (1990), who found intrinsic and extrinsic quality cues to be used in forming abstract beliefs about perceived product quality. Objective quality can be categorised and quality cues, this way connecting objective and perceived quality.
2.2.1 Perceived quality trend
Trend is defined as “a general direction in which something is developing or changing” (Oxford University Press, 2010). An increase in perceived quality trend would indicate an improvement in perceived quality and a decrease in trend would indicate a worsening perceived quality. Trends are important to take into account because as resulted by Mowen’s (1979) study, previous failure to meet the standard influences current service perception.
quality and marketing activities, but also that an interaction between the two would improve marketing effectiveness.
Perceived quality has been linked to service quality in the definition of satisfaction as proposed by (Parasuraman et al., 1988, p. 1): “the degree and direction of discrepancy between consumers’ perception and expectations”. If marketing increases expectation, and the service perceptually meets these expectations, customers are satisfied which leads to increasing performance. From the company’s point of view: if expectations are increased through marketing but they fail to deliver, there will be dissatisfaction which creates a bad mood and reduces responsiveness to marketing. This would indicate a positive relationship between perceived quality trend and firm performance which is similar to the relationship between marketing and performance. Taking into account the willingness to pay price premiums, the following hypotheses are formulated:
H4a: Perceived quality trend strengthened the effect of advertising activities on firm performance.
H4b: Perceived quality trend weakens the effect of promotional activities on firm performance.
2.2.2 Objective quality trend
Objective quality is the actual quality that is delivered. It is not the feeling of the customer but a quantifiable variable. A change in objective quality can affects customer perception until six years after. Decreases in quality are more noticeable than increases in quality (Gijsenberg, et al., 2015; Mitra, et al., 2006).
consequences of this delay and his or her current mood. Customers who experienced no recent service problems (positive objective quality) with a company have significantly better service quality perceptions than customers who experienced recent service problem (negative objective quality) (Zeithaml et al., 1990). There is a positive relationship between customer experience and its overall perception of the service. And since marketing activities can be used to manage expectations, the relationship between objective quality and marketing activities is positive.
H5: Objective quality trend strengthens the effect of marketing activities on performance.
Ticketholders are likely to react stronger to objective quality because they evaluate each purchase separately, while cardholders are likely to react less since they are in a relationship (a binding contract) with the provider, making them more likely to be more flexible to service failure because they are “already paying either way”.
2.2.3 Crisis
Crisis in service is comparable to recalls of good. Recalls are directly linked to low quality perception (Crafton et al. 1981). But recalls on themselves are a failure of the product, which usually occurs in production (e.g. KitKat having plastic in its bars)
advertising campaign, consumers would be more likely to be less responsive, because they are less trusting towards the company and because crisis is likely to create foul moods.
H6: Crisis weakens the effect of marketing activities on performance.
Sheth et al. (2009) defines quality as “doing it right the first time”, retaining customer satisfaction includes reliability and quality assurance. Meaning, customers are more likely to choose products that have a stable performance. A sudden change in quality would fail to meet expectations. Prior trend in objectives influence service performance (Gijsenberg, et al., 2015; Verhoef et al., 2009). Companies also rely on their brand equity to get them out of a crisis, stronger brands are less affected by crisis than smaller/weaker brands (Cleeren, et al., 2008). Awareness, perceived quality, loyalty, and associations are all part of brand equity (Aaker, 1996). Awareness and association are linked to advertising expenditures, loyalty and perceived quality to satisfaction. This creates and indirect path to an increase in customer usage. Customer relationship correlates strongly (.52) with customer loyalty and business performance (.35) (Palmatier, et al, 2006). This would indicate that relational customers would be less affected by crisis than single purchase customers.
2.3 Conceptual Model
Figure 1 Conceptual Model
Figure 2 Conceptual Model
H5 H3 H2 H4 H1 Marketing Activities (t) Promotion Advertising Performance (t) Perceived Quality (t-1)
Individual ticket sales Perceived
quality trend (t)
Crisis (t) Objective
qualitytrend (t)
Season ticket sales
Performance (t-1)
Individual ticket sales Season ticket sales
3 DATA
This research is conducted using time series data of a European railway company during 30 months, from April 2007 to September 2009. The railway company has around 1.2 million passengers daily. The data has been aggregated to monthly values. Each month is assigned a code, e.g. M0706 (3) indicates June 2007 (t=3). All tests are conducted in Eviews 9.5.
3.1 Description Of Data
Sales (S) is measured by the total amount of kilometres travelled by passengers. There are two
types of tickets, subscription and individual tickets. Subscription holders (called seasonal cardholders by the company) pay a fee depending on the subscription type allowing them to travel for free or with a discount during the subscription period using their card, hence they are termed cardholders. Individual ticketholders are termed ticketholders. Individual tickets regard a one-time purchase for a single trip. On average cardholders travelled about 257 million kilometres, the least km travelled was in M0908(29) and the most in M0909(30) (table 1). Ticket holders travelled about 2.5 times more than card holders, 647 million kilometres, least in M080(11) and most in M0810(19). Figure 2 illustrates both travelled km during the 30 months.
Figure 2 Sales in Total Km
Perceived quality (PQ) is measured by the score given by customers when asked how satisfied they
were with the service. As mentioned was discussed customer satisfaction is comparable to perceived quality. The score is based on a ten-point scale (1 = could not be worse, 10 = excellent) surveyed by more than 6,000 randomly selected customers. The available data indicates the aggregated average per month. Alternatively, perceived quality can be measured by the aggregated percentage of customers that gave a rating higher than seven (in that month). Although Gijsenberg et al. (2015) uses the average score as measurement, for this study percentage measurement is used to avoid multicollinearity (discussed afterwards) in the model. Table 2 shows the descriptive of these measurements. Their combined illustration is given in figure in appendix. It clearly shows that these two measurement are highly correlated (r=.99).
Table 2 Perceived Satisfaction
Descriptive Statistics
Average Satisfaction 7+ Satisfaction %
Mean 6.99 75.19%
Standard Deviation 0.09 3.45%
Minimum 6.80 67.16%
Maximum 7.10 80.41%
On average 75% of the passengers gave a rating higher than a 7. The lowest satisfaction was in months M0704(1). The highest in M0808 (29).
Perceived quality trend (PQT) is calculated based on the perceived quality measurement, aggregated
from last month, lower than 0 a decrease from last month. The trend is calculated until the 30th month (M0909, t = 30). Figure 3 illustrates the perceived quality and its trend.
Figure 3 Perceived Quality & Trend
Advertising (ADV) is measured by the total euros spent on advertising per month. The average
advertising during the 30 months were €316 (M advertising = 315.83, SD = 469.48). In M0710 the most was spent on advertising, €1,500.
Promotion (OP & RP) is measured by two dummy variables indicating where the promotions took
place, own (railway) store or retail store. Indicated by 1 = promotion and 0 = no promotion. Figure 4 illustrates advertising and promotion over time.
Figure 4 Advertising Activities
Objective quality (OQ) is measured by the amount of achieved connections. That is, the trains which the transfer duration is less than the time between the arrived train and the connection’s departure. The percentage indicates the trains that meet these criteria. If a passenger needs to make a connection, this is an objective percentage indicating whether the connection is made. A different measure is available for this variable. The percentage of trains that were on time or less than three minutes late. Gijsenberg et al. (2015) chose the connection measurement based on expert opinion that it is the critical measure. Missing a connection disrupts the transport chain of customers. In addition, it is possible to have a delay and not miss a connection, but it is not possible to miss a connection and not have a delay, the connection data is the most appropriate for measuring quality objectively. The punctuality measurement is used to estimate the model as a robustness check. Both objective quality descriptive can be found in table 3. The measurements are illustrated in figure Appendix A. These two measurements are highly correlated, r=.96.
Table 3 Objective Quality
Descriptive Statistics
Achieved Connection % Punctual Arrival %
Mean 92.30% 87.14%
Standard Deviation 1.69% 3.27%
Minimum 87.36% 77.21%
Maximum 95.28% 92.11%
On average there are 92% of the trains are achieved for a connection (SD = 1.69). In M0901 the lowest percentage was successful (87%) and the highest was in M0908 (95%).
Objective quality trend (OQT) is calculated similarly to the perceived quality trend, subtracting the
Figure 5 Objective Quality & Trend
Crisis (C) is calculated using the objective quality measurement, achieved connections. Using a
dummy variable, a value of 1 is assigned to the periods where the success was below the first quartile, Q1 = 91.32%. As is illustrated in figure 5, there are seven crisis months. These can be group in three crisis periods: The first lasts one month, M0705 (t = 2); the second, two months, M0711 (t= 8) and M0712 (t = 9); the third, five months M0811 (t = 20) – M0902 (t = 23). Figure 5 illustrates these points.
Since observed quality has an alternative measurement, crisis would also have an alternative measurement of based on the alternative OQ. Calculated the same way, there are seven crisis points. T=9 to t=10 and t=20 to t=24.
3.2 Model Development
The seemingly unrelated regressions (SUR) model proposed by Zellner (1962) captures individual relationships of which their disturbance term correlate. In other words, the individual equations are related to each other. The SUR model offers additional information as opposed to when the equations are considered separately, because it takes into consideration that the error terms correlate. If the assumption of cross-correlation does not hold, then the SUR model estimation holds no difference to the OLS model and the models are in fact unrelated.
The multiplicative model is:
Equation 1 ) Multiplicative SUR Performance Model
Where Sht is the observation in month t on sales of holder type h (T= ticketholder, C= cardholder); OP, RP, ADV, SCt-1, STt-1 and PQt-1 is the observation in month t appearing in the holder type’s equation. Bh is the coefficient associated with the explanatory variable t each observation; εht is the value of the random disturbance term associated with the holder type’s equation of the model.
More specifically, SCt is the total sales of the cardholders at time t; STt is the total sales of the ticketholder at time t. OPt and RPt are dummy variables representing own promotion and retail promotion, respectively at time t; ADVt indicates the advertising expenditure at time t. SCt-1 is the total cardholder sales of the previous month (t-1). STt-1 is the total sales of ticket holders in the previous month (t-1). PQt-1 is the perceived quality of the previous month (t-1). Mt represents the trends PQTt or OQTt or the crisis period Ct, which are interchangeable but are not combined in the same model1. εCt / εTt is the (unobserved) value of the disturbance term.
The model is linearized and skewness reduced using the logarithm transformation2.
Equation 2 Linearized SUR Model
Where ; ; ; and so on3. The coefficient of these log-log
models are indicated by: β0 is the intercept. β1, β2, β3 is the direct effect of marketing activities (respectively own promotion, retail promotion, advertising expenditure) on performance. β4 and Β5 are the effect of the lagged sales (cardholder and ticketholder). Β6 is the effect of the lagged perceived quality. Β7 indicates the direct effect of the moderators. Β8, Β9, Β10 indicate the moderating effect on marketing activities. The log-log nature of the model allows for these results to be
1 If M
t = Ct, .
2 To avoid undefined values, +1 is added to the ADV, PQT, OQT data when the models are linearized. 3 If M
t = Ct,
interpreted as elasticities, i.e. β3 is the advertising elasticity. Next, the model is validated to assess the quality of the model outcomes. The assumptions required for OLS application are tested as proposed by (Leeflang, et al., 2015).
3.2.1 Multicollinearity
Correlation between the independent variables create unreliable parameter estimates. Multicollinearity is detected with the VIF and Tolerance statistics ( ). A VIF of 10 is the maximum recommended, higher VIFs indicate multicollinearity (e.g. Hair et al, 1995). Multicollinearity is detected for the PQT extended model for PQT and its interaction with and RP (see appendix for the detailed solution).
Collinearity is eliminated by eliminating the RP interaction in the PQT-model. This solution is only applied to the PQT-model. Because of this, the PQT interactive model does not account for the RPM interaction. The full multicollinearity table with the VIF scores are presented in table 4.
Table 4 Vif Scores
VIF score Basic Model Mt = Ct M t = PQTt Mt = OQTt Solved β1C OPt 1.28 3.54 1.47 1.47 1.34 β2C RPt 1.74 2.74 2.66 1.82 2.36 β3C ADVt 1.21 1.62 1.30 1.29 1.28 β4C SCt-1 1.08 1.72 1.22 1.22 1.19 β5C STt-1 1.92 3.43 2.45 2.34 2.37 β6C PQt-1 1.49 2.47 1.86 1.76 1.95 β7C Mt 7.92 15.71 8.39 3.24 β8C OPt x Mt 3.41 9.92 8.32 1.65 β9C RPt x Mt 5.04 15.17 2.59 β10C ADVt x Mt 2.30 2.42 2.42 2.16
3.2.2 Correlated disturbances
Autocorrelation or serial correlation indicate that the residuals have a pattern and thus contain some information, which causes biased parameter estimates (Leeflang, et al., 2015). This test is done by plotting the residuals over time using, which show that there is no serial correlation in the residuals (Appendix).
3.2.3 Heteroscedasticity
3.2.4 Nonnormal errors
To test if the error terms follow a normal distribution the Jarque-Bera test is used. The p-value >.05 which does not reject the null hypothesis of normality.
3.2.5 Stationary
Next, the performance variables are tested for stationary using the Augmented Dickey-Fuller (ADF) Unit Root Test using an intercept and trend as exogenous variables. H0 = variable is not stationary (unit root), H1 = variable is stationary (no unit root). If the variable has unit root it is evolving. As a second test Phillips and Perron’s test (P&P) using an intercept and trend as exogenous variables is used to confirm the results. The tests are performed on all the log-transformed variables. If the absolute test statistic is more than the critical value, p < 0.05, then H0 (unit root) is rejected. p>0.05 indicates that there is insufficient evidence to support H1 and so H0 is not rejected. Due to the low mount of observations (n=30), ADF and P&P might not be strong enough to accept H1, a third test is conducted if both results indicate unit root. The Kwiatkowski-Phillips-Schmidt-Shin (KPSS) unit root test is performed using an intercept and trend as exogenous variables. Importantly, KPSS indicates unit root (H1) when p < 0.05, conversely to the previous two tests. If the KPSS contradicts P&P, the KPSS is considered the determinant result. The stationary assessment is shown in table 5.
Table 5 Stationary Assessment (P-Value)
ADF P&P KPSS Result Cardholder sales <0.01 <0.01 Stationary
Ticketholder sales 0.13 0.12 0.06 Stationary
Advertising .01 .01 Stationary Perceived quality 0.03 0.16 0.10 Stationary
Objective quality <0.01 <0.01 Stationary
Perceived quality trend <0.02 <0.02 Stationary Objective quality trend 0.06 0.46 0.06 Stationary
3.3 Analysis Plan
In conclusion the final two models (formula) are analysed with the SUR model which uses OLS as all the assumptions are met.
To analyse the rest of the hypothesis different versions of the model is used. The basic model, without interaction, is used for H1, H2 and H3. The extended model, includes the three different interaction separately, is used for the analysis of H4, H5 and H6.
4 RESULTS
Before interpreting the results, the model diagnostics are presented to elaborate on the quality of the models. The basic model is briefly discussed followed by a more extensive presentation of extensive models.
4.1 Model Diagnostics
The (adjusted) coefficient of determination (R-squared) is used to evaluate the fit of the model (table6). The best fit model is the PQT model.
Table 6 Model Fit
R2 Adj. R2
Cardholder Ticketholder Cardholder Ticketholder Basic 0.22 0.75 0.004 0.68
Crisis 0.48 0.77 0.18 0.65
PQT 0.62 0.79 0.44 0.68
OQT 0.73 0.78 0.58 0.65
The extended perceived quality trend model R2 show that the ticketholder model has a better fit than the cardholder model. That is, the explanatory variables account for 62% of the variation in cardholder sales and for 79% of the variance in ticketholder sales. Additionally, the PQT moderation effects improves the model fit significantly for cardholders compared to the basic model.
4.2 Main Effects of Promotion and Advertising
Table 7 reports the parameter estimates for the basic model. Table 8 reports the parameters of the extended models. The results of the basic model are briefly presented before continuing with the main focus of the results, the three extended models.
For the basic model a significant long-term perceived quality (β = .91, p = .02) effect is found on cardholder sales. On ticketholder sales a significant main effect is found for promotion (βown= .05, βretail= .04, p<.001) and long-term perceived quality (β=.21, p = .05) ticketholder sales.
Table 7 Basic model
Basic Model parameter estimates
Cardholder sales (SCt) Parameter estimate S.E. t-Statistic p-Value
β0C Constant 40.49 10.77 3.76 0.00 β1C OPt 0.00 0.04 0.10 0.92 β2C RPt 0.01 0.05 0.25 0.81 β3C ADVt 0.00 0.01 0.14 0.89 β4C SCt-1 -0.28 0.18 -1.55 0.13 β5C STt-1 -0.76 0.47 -1.62 0.11 β6C PQt-1 0.91 0.39 2.34 0.02 R2 0.22 R2 adj 0.004 Ticketholder sales (STt) β0T Constant 18.01 2.98 6.05 0.00 β1T OPt 0.05 0.01 4.31 0.00 β2T RPt 0.04 0.01 3.32 0.00 β3T ADVt 0.00 0.00 -0.14 0.89 β4T SCt-1 0.08 0.05 1.63 0.11 β5T STt-1 0.03 0.13 0.26 0.79 β6T PQt-1 0.21 0.11 1.99 0.05 R2 0.75 R2 adj 0.68
4.3 Main Effect of Preceding Occurrences
Table 8 Extended model
Extended Model parameter estimates
Mt = Ct Mt = PQTt Mt = OQTt Cardholder sales (SCt) Parameter estimate SE p-Value Parameter estimate SE p-Value Parameter estimate SE p-Value β0C Constant 58.83 2.52 0.00 38.08 8.39 0.00 45.94 6.49 0.00 β1C OPt -0.06 0.05 0.26 0.05 0.03 0.09 -0.02 0.03 0.48 β2C RPt 0.05 0.05 0.36 0.05 0.03 0.19 0.02 0.03 0.46 β3C ADVt 0.00 0.00 0.48 -0.00 0.00 0.46 0.01 0.00 0.04 β4C SCt-1 -0.54 0.19 0.01 -0.17 0.13 0.21 -0.42 0.11 0.00 β5C STt-1 -1.41 0.51 0.01 -0.76 0.36 0.04 -0.89 0.28 0.00 β6C PQt-1 1.65 0.41 0.00 0.28 0.30 0.35 1.02 0.24 0.00 β7C Mt -0.04 0.08 0.58 3.50 1.58 0.03 1.63 1.43 0.26 β8C OPt x Mt 0.25 0.07 0.00 -7.59 1.73 0.00 -6.43 1.51 0.00 β9C RPt x Mt 0.00 0.07 0.96 -1.33 1.59 0.41 β10C ADVt x Mt -0.01 0.01 0.64 0.24 0.18 0.19 0.21 0.25 0.41 R2 0.48 0.62 0.73 R2 adj 0.18 0.44 0.58 Ticketholder sales (STt) β0T Constant 13.46 3 .99 0.00 16.02 3.05 0.00 17.86 3.57 0.00 β1T OPt 0.05 0.02 0.01 0.05 0.01 0.00 0.05 0.02 0.01 β2T RPt 0.03 0.02 0.03 0.04 0.01 0.00 0.05 0.01 0.00 β3T ADVt -0.00 0.00 0.60 0.00 0.00 0.88 -0.00 0.00 0.38 β4T SCt-1 0.15 0.06 0.02 0.10 0.05 0.04 0.11 0.06 0.08 β5T STt-1 0.19 0.16 0.24 0.11 0.13 0.40 0.01 0.14 0.92 β6T PQt-1 0.11 0.13 0.39 0.28 0.11 0.01 0.19 0.11 0.10 β7T Mt -0.03 0.03 0.24 0.77 0.57 0.19 -0.01 0.02 0.75 β8T OPt x Mt -0.02 0.02 0.46 -0.21 0.63 0.74 1.28 0.64 0.05 β9T RPt x Mt 0.02 0.02 0.50 -1.07 0.62 0.09 β10T ADVt x Mt 0.00 0.00 0.20 -0.02 0.06 0.78 -0.00 0.10 0.99 R2 0.77 0.79 0.78 R2adj 0.65 0.68 0.65
Preceding cardholder sales have a positive effect on ticketholder sales in all models, Crisis (βcardholder= .15, p=.02), PQT (βcardholder= .10, p=.04), and PQT (βcardholder= .11, p=.08). This could be due to the cardholders’ desire to have more flexibility when it comes to travelling by train. Which would indicate a diminishing loyalty. In other words, the bad experience created by overcrowding creates the need for relational customers to break off their relationship and become single-purchasers in order to lower switching costs.
to no surprise the perceived quality which is indicated by satisfaction affects relational customers stronger than ticketholders.
4.4 Moderating Role of Quality On Marketing Effectiveness
For the Crisis-model the result shows clear evidence that crisis affects the impact of promotion on cardholder sales. More specifically, the interaction between own promotion and crisis is positive (β = .25, p <.001). This implies that when there is service failure, cardholders react more to own promotions. This effect is not found for ticketholder sales. This would support Cleerem et al. (2008) that weaker brands are more affected by crisis. Compared to relational-customers, single purchasers can be considered to have a lower equity. During product failure consumers become less trustworthy of the company. Relational-customers are able to neutralise this effect due to the strong correlation between loyalty and customer relationship. Returning to the cross aversion theory, own promotion can be used to lessen the effect of crisis on cardholder performance by lowering the monetary damage (costs) created by the product failure. Own promotion has no main effect, but during crisis periods it has a strong positive effect on km travelled.
For the PQT-model a significant interaction is found between own promotion and perceived quality trend. The effect of own promotion is weakened by a positive trend (β = -7.59, p <.001). Trend indicates the increase or decrease of perceived quality from the previous period. In order to facilitate interpretation, it can be stated that a decreasing trend would indicate dissatisfaction (current satisfaction is less than what was expected, which is based on the previous PQ). When relational-customers are satisfied they are less affected by own promotion and own promotion weakens the effect satisfaction. Simply put, if happy customers are willing to pay a price premium, they are less price sensitive which makes price cuts not as attractive.
surpasses expectations customers are less responsive to retail promotions, this could be because they would rather respond to own promotions or become cardholders.
These Significant promotional interactions effects are illustrated in figure 6.
Figure 6 Significant Promotional Effects
4.5 Robustness Check
(KPSS=0.10), PQT (P KPSS=0.13), OQ (KPSS=0.08) and OQT (KPSS=0.06) are all stationary. The full results of the alternative measurements can be seen in appendix [C. Table 9 is extended to include the alternative measurements and show that the chosen model outperforms the alternative specifications.
Table 9 Robustness Check
R2 Adj. R2
5 DISCUSSION
Marketing managers recently are being pressured to account for their expenditures. With the increase in market transparency, offering simple promotions or spending on advertising are not enough. Indirect effects on marketing activities has become essential. Managers should not only take their goals into account but also the current situation of the firm and the mindset that customers are in. By taking a closer look at these factors this study generates insight for a mass service company. Having a mass service as a core product shares characteristics with both the traditional (one-on-one-) service and physical goods (mass production). This study fills the gap in the literature. In addition, in this study more connected loyal customers are analysed separate from one-transactional customers. Cardholders are deemed as being in a fixed relationship with the service provider. I found that for these marketing activities have no effect on their consumption. These customers are less sensitive because a price reduction does not influence them to travel more nor does advertising since with a monthly fee they are already getting the most use out of their dollar. Single ticket users steadily showed to be positively influenced by promotional activities. This could imply that marketing programs need to focus on trust and commitment and not price for relational customers and focus the promotions on the single purchasers.
This is not the case for own promotion during a prices, when cardholders’ consumption increases with promotional activities. I.e. the worst the service becomes, the more effective the own promotions are. If there is a crisis, cardholder will be influenced by the promotions of the company. If customer satisfaction has increased since the previous month, this will reduce the effect of promotion on cardholders. If people are already happy hey will not be as influence by own promotion if they were disappointed.
Crisis does not influence the effect of marketing activities for ticketholders. If the company improves its services by being on time or often or offering more trains to make connections, there are different consequences. Promotional effect would decrease for cardholders, but own promotion would increase for ticketholders. This makes sense, better service at a lower price would attract more customers. It also shows that dissatisfied customers react less to promotions.
One of the most important aspect is perceived quality from the previous period. Previous perception showed to have a positive impact on consumption for both groups.
of both groups. Previous single purchases and subscriptions have a negative effect on current subscriptions. This might indicate that traveling is done mostly for utilitarian reasons and adding the hedonic effect every other month increases the amount of travelling. Noticeably, previous cardholder consumption does lead to a current increase in ticketholder consumption.
REFERENCES
Aaker, D. A., 1996. Measuring Brand Equity Across Products and Markets. California Management Review, 38(3), pp. 102-120.
Aaker, D. A., Carman, J. M. & Jacobson, R., 1982. Modeling Advertising-Sales Relationships Involving Feedback: A Time Series Analysis of Six. Journal of Marketing Research, 19(1), pp. 116-125.
Ahn, J.-H. H. S.-P. L. Y.-S., 2006. Customer churn analysis: Churn determinants and mediation effects of partial defection in the Korean mobile telecommunications service industry. Telecommunications Policy,, 30(10-11), p. 552–568.
Achen, C.H. 2000. “Why Lagged Dependent Variables Can Suppress the Explanatory Power of Other Variables.” Ann Arbor 1001:48106–1248.
Anderson, Eugene W. and Mary W. Sullivan (1990), "Customer Satisfaction and Retention Across Firms," presentation at the TIMS College of Marketing Special Interest Conference on Services Marketing, Nashville, TN, (September).
Anderson, E. and Sullivan, M. (1993), “The antecedents and consequences of customer satisfaction for firms”, Marketing Science, Vol. 12 No. 2, pp. 125-43.
Ataman, M. B., Van Heerde, H. J. & Mela, C. F., 2010. The Long-Term Effect of Marketing Strategy on Brand Sales. Journal Of Marketing Research, 47(5), pp. 866-882.
Allenby, G. M. a. D. M. H., 2004. Advertising Response, Cambridge, MA: Marketing Science Institute: MSI Special Report 05-200a.
Belch, G. & Belch, B., 2003. Advertising and Promotion: an integrated marekting communications persepctive. 6th ed. s.l.:The McGraw-Hill Companies.
Berger, Jonah, Michaela Draganska, and Itamar Simonson (2007), “The Influence of Product Variety on Brand Perception and Choice,” Marketing Science, 26 (July–August), 460–72.
Bijmolt, T.H.A., van Heerde, H.J., Pieters, R.G.M., 2005. New empirical generalizations on the determinants of price elasticity. J. Mark. Res. 42, 141–156
Bolton, R.N. and Drew, J.H. (1991), "A multistage Model of Customers' Assesement of Service Quality and Value," Journal of Consumer Research, 17 (March), 375-384
Bredahla, L., Grunert, K. & Fertinb, C., 1998. Relating consumer perceptions of pork quality to physical product characteristics. Food Quality and Preference, 9(4), pp. 273-281.
Bruce, N. I., Peters, K., & Naik, P. A. (2012). Discovering how advertising grows sales and builds brands. Journal of Marketing Research, 49(6), 793-806.
Choi, Y., & Lin, Y-H. (2009). Consumer responses to Mattel product recalls posted on online bulletin boards: Exploring two types of emotion. Journal of Public Relations Research, 21(2), 198–207.
Cronin, J. Joseph, Jr. and Steven A. Taylor (1992), "Measuring Ser- vice Quality: A Reexamination and Extension," Journal of Marketing, 56 (July), 55-68.
Cleeren, K., Dekimpe, M. G. & Helsen, K., 2008. Weathering product-harm crises. Journal of the Academy of Marketing Science, 36(2), pp. 262-270.
Cohn, E. G. & Rotton, J., 1997. Assault as a function of time and temperature: A moderator-variable time-series analysis. Journal of Personality and Social Psychology, 72(6), pp. 1322-1334.
Cronin, J. and Taylor, S. (1992), “Measuring service quality: a re-examination and extension”, Journal of Marketing, Vol. 56, pp. 55-68.
Dekimpe, M. G. & Hanssens, D. M., 1995. The persistence of marketing effects on sales. Marketing Science, 14(1), pp. 1-21.
Dekimpe, M. G. & Hanssens, D. M., 1999. Sustained Spending and Persistent Response: A New Look at Long-Term Marketing Profitability. ournal of Marketing Research, 36(November), pp. 397-412.
Dibbo, A., 1993. Measuring Marketing Performance. Engineering Management Journal, 3(6), pp. 255-258.
Farris, P. B. N. P. P. R. D., 2006. Marketing Metrics: 50+ Metrics Every Executive Should Master. Upper Saddle River: Wharton School Publishing.
Ford Motor Company (1990), Memorandum to dealers, (October 3).
Gama, A. P. d., 2011. An expanded model of marketing performanc. Marketing Intelligence & Planning, 29(7), pp. 643 - 661.
Garbarino, E. & Johnson, M., 1999. The Different Roles of Satisfaction, Trust and Commitment for Relational and Transactional Consumers. Journal of Marketing, 63(2), pp. 70-87.
Gijsenberg, M. J., 2014. Going for gold: Investigating the (non)sense of increased advertising around major sports events. International Journal of Research Marketing, 31(1), pp. 2-15.
Gijsenberg, M. J., Van Heerde, H. J. & Verhoef, P. C., 2015. Losses Loom Longer Than Gains: Modeling the Impact of Service Crises on Perceived Service Quality over Time. Journal of Marketing Research, Volume 2, p. 642–656.
Hanssens, D., Pauwels, K., Srinivasan, S., Vanhuele, M., & Yildirim, G. (2014). Consumer attitude metrics for guiding marketing Mix decisions. Marketing Science, 33(4), 534–550
Hanssens, D. M. (2009). Advertising impact generalizations in a marketing mix context. Journal of Advertising Research, 49(2), 127.
Heskett, J. L., & Schlesinger, L. A. (1994). Putting the service-profit chain to work. Harvard business review, 72(2), 164-174.
Huang, Ming-Hui and Roland T. Rust (2013), “IT-Related Service: A Multidisciplinary Perspective,” Journal of Service Research, 16 (3), 251–58.
Hunneman, P.C. Verhoef & L.M. Sloot, 2015, "The Impact of Consumer Confidence on Store Satisfaction and Share of Wallet Formation". Journal of Retailing, Vol. 91, Issue 3, pp. 516-532.
Isidore, C., 2012. American Airlines is losing important customers. [Online] Available at:
http://money.cnn.com/2012/10/08/news/companies/american-airlines-customers/ [Accessed 13 March 2016].
Jedidi, K., Mela, F. & Gupta, S., 1999. Managing Advertising and Promotion for Long-Run Profitability. Marketing Science, 18(1), pp. 1-22.
Joby, J. (1992). “Patient Satisfaction; the impact of past experience”. Journal of healthcare marketing. 12(3), pp. 56-64.
Kahneman, D. & Tversky, A., 1979. Prospect Theory: An Analysis of Decision under Risk. Econometrica, 47(2), pp. 263-292.
Koska, Mary T. (1990) "High-Quality Care and Hospital Profits: Is There a link?" Hospitals, 64 (March 5), 62-63.
Köbberling, V. & Wakker, P., 2005. An index of loss aversion. Journal of Economic Theory, 122(1), pp. 119-131.
Kumar, N., 2004. Marketing as Strategy. Boston : Harvard Business School Press.
Leeflang, P. S. & Wittink, D. R., 2000. Building Models for Marketing Decsions: Past, Present and Future. International Journal of Research in Marketing, 17(2–3), p. 105–126.
Leeflang, P., Wieringa, J., Bijmolt, T. & Pauwels, K., 2015. Modeling Markets: Analyzing Marketing. s.l.:s.n.
Marsh, T. L., T. C. Schroeder, J. Mintert. 2004. Impacts of meat product recalls on consumer demand in the USA. Appl. Econom. 36(9) 897–910.
Mela, Carl F., Sunil Gupta, and Kamel Jedidi (1998), “Assessing Long-Term Promotional Influences on Market Structure,” International Journal of Research in Marketing, 15 (2), 89–107.
Moorman, C. & Rust, R. T., 1999. The Role of Marketing. Journal of Marketing, 63(Special Issue), p. 180– 197.
Morgan, N. A., Clark, B. H. & Goonera, R., 2002. Marketing productivity, marketing audits, and systems for marketing performance assessment: integrating multiple perspectives. Journal of Business Research, 55(5), p. 363–375.
NS, 2015. NS Jaarverslag, Utrecht: NS.
Netemeyer, R.G., Krishnan, B., Pullig, C., Wang, G., Yagci, M., Dean, D., Ricks, J. and Wirth, F. (2004), “Developing and validating measures of facets of consumer-based brand equity”, Journal of Business Research, Vol. 57 No. 2, pp. 209-24.
Oude Ophuis, P. A. & Van Trijp, H. C., 1995. Perceived quality: A market driven and consumer oriented approach. Food Quality and Preference, 6(3), pp. 177-183.
Owalbi, A. B., 2009. Effect of Consumers Mood on Advertising Effectiveness. Europe’s Journal of Psychology, Volume 4, pp. 118-127.
O'Sullivan, D. & Abela, A. V., 2007. Marketing Performance Measurement Ability and Firm Performance. Journal of Marketing, 71(2), pp. 79-93.
Oxford University Press, 2010. Oxford Dictionary, s.l.: Oxford University Press.
Palmatier, RobertW., Rajiv P. Dant, Dhruv Grewal, and Kenneth R. Evans (2006), “Factors Influencing the Effectiveness of Relationship Marketing: AMeta-Analysis.” Journal of Marketing 70 (4), 136–53
Parasuraman, A., Valarie Zeithaml, and Leonard Berry (1988), "SERVQUAL: A Multiple-Item Scale for Measuring Consumer Perceptions of Service Quality," Journal of Retailing, 64 (Spring), 12-40.
Powell, T., 1995. Total quality management as competitive advantage: a review and empirical study. Strategic Management Journal , 16(1), p. 15–37.
Richards, C., 2015. WhatsApp down: Messaging app fails leaving users unable to send or receive messages. [Online] Available at:
http://www.mirror.co.uk/news/technology-science/technology/whatsapp-down-messaging-app-fails-7099744 [Accessed 13 03 2016].
Telegram Messenger, 2014. Twitter @Telegram. [Online] Available at:
https://twitter.com/telegram/statuses/437313902058426368 [Accessed 13 March 2016].
Sethuraman, Raj, and Gerard J. Tellis (1991), “An Analysis of the Tradeoff Between Advertising and Pricing.” Journal of Marketing Research 31 (2), 160–74
Sheth, J., Sisodia, R. and Sharma, A. (2009), “The challenges of improving marketing productivity and performance”, pp. 1‐35
Shapiro, C. (1983) “Premiums for high quality products as returns to reputations,” Quarterly Journal of Economics, vol. 98, pp. 659–680
Srinivasan, S., Vanhuele, M., & Pauwels, K. (2010). Mind-set metrics in market response models: An integrative approach. Journal of Marketing Research, 47(4), 672-684.
Steenkamp, J.-B. E., 1990. Conceptual Model of the Quality Perception Process. Journal of Business Research, 21(1), pp. 309-333.
Tellis, Gerard J. (1988), "The Price Elasticity of Selective Demand: A Meta-Analysis of Econometric Models of Sales,” Journal of Marketing Research, 25 (November), 331-4 1
Van Heerde, H., Helse, K. & Dekimpe, M. G., 2007. The Impact of a Product-Harm Crisis on Marketing Effectiveness. Marketing Science, 2(26), pp. 230-245.
Vakratsas, Demetrios, and Tim Ambler (1999), “How Advertising Works:What DoWe Really Know?” Journal of Marketing 63 (1), 26–43
Van Doorn, J. & Verhoef P. (2008). Critical Incidents and the Impact of Satisfaction on Customer Share. Journal of Marketing. 72(40): 123-42.
Van Heerde, H. J., Gijsenberg, M. J., Dekimpe, M. G., & Steenkamp, J. B. E. (2013). Price and Advertising Effectiveness Over the Business Cycle. Journal of Marketing Research, 50(2), 177-193.
Van Heerde, H., Helsen, K. & Dekimpe, M. G., 2005. The Impact of a Product-Harm Crisis on Marketing Effectiveness. Marketing Science, 26(2), pp. 230-245.
Verhoef, P. C. et al., 2009. Customer Experience Creation: Determinants, Dynamics and Management Strategies. Journal of Retailing, 1(85), pp. 31-41.
Yoo, B., Donthu, N. and Lee, S. (2000), “An examination of selected marketing mix elements and brand equity”, Journal of the Academy of Marketing Science, Vol. 28 No. 2, pp. 195-211.
Zeithaml, Valarie A., Leonard L. Berry, and A. Parasuraman (1996), “The Behavioral Consequences of Service Quality,” Journal of Marketing, 60 (April), 31–46.
Zellner, A. (1962) An e• cient method of estimating seemingly unrelated regressions and tests for aggregate bias, Journal of the American Statistical Association, 57, 348- 68.
Zomerdijk, Leonieke G. and Voss Christopher A. (2010), “Service-Design for Experience-Centric Services,” Journal of Service Research, 13 (1), 67–82
Zeithaml, Valarie A., A. Parasuraman, and Leonard L. Berry (1990), Delivering Quality Service: Balancing Customer Perceptions and Expectations. New York: The Free Press. Reprint No.
APPENDICES
A1. Appendix A: Data
Perceived satisfaction measurements
Objective quality measurements
1) Multicollinearity
(Leeflang, et al., 2015, p. 140) suggests the following remedies: 1) Obtain more data: Not a possibility for this study.
2) Reformulate the model with the specific objective to decrease multicollinearity: This implies combining two predictor variables that correlate. In this case it would be combining PQTT and RPT, which is not appropriate since the two “capture different phenomena”.
4) Apply estimation methods specifically developed for cases with severe multicollinearity: Applying the interaction one at a time, thus estimating the interaction separately, fails to eliminate collinearity. Estimating the model in terms of changes over time (SCH,t-SCH,t-1) also does not eliminate the problem.
5) Eliminate predictor variable with a t-ratio close to zero: This is the “last resort”. Eliminating the interaction with OP does not solve the problem, so the second furthest is eliminated, RP this way eliminating multicollinearity.
2) Autocorrelation
Basic
Crisis
PQT
3) Heteroscedasticity
HETEROSCEDASTICITY TEST: WHITE
F-statistic
BASIC Card 1.49 Prob. F(6,22) 0.23
Ticket 0.51 Prob. F(6,22) 0.79
CRISIS Card 0.82 Prob. F(10,18) 0.62
Ticket 0.57 Prob. F(10,18) 0.82
PQT Card 0.65 Prob. F(9,19) 0.74 Ticket 0.35 Prob. F(9,19) 0.95
OQT Card 0.76 Prob. F(10,18) 0.66 Ticket 0.28 Prob. F(10,18) 0.98
4) Nonnormal errors
SC on the left, ST on the rightBasic
Crisis
OQT
A2. Appendix B: Alternative Results
Basic Model parameter estimates
Cardholder sales (SCt) Parameter estimate S.E. t-Statistic p-Value
