Do enrollment rewards enhance behavioral loyalty in a frequent flyer program?

(1)

Do enrollment rewards enhance behavioral loyalty

in a frequent flyer program?

The predictive role of self-selection and self-determination.

(2)

Do enrollment rewards enhance behavioral loyalty

in a frequent flyer program?

The predictive role of self-selection and self-determination.

Master thesis

Marlijn van Opzeeland June 30th, 2014

Faculty of Economics and Business Department of Marketing Msc. Marketing Management Sanderij 19 7491 GX Delden 06-45006019 marlijn@vanopzeeland.com Student number 1905961

(3)

Management Summary

This master thesis is focused around the concepts of self-selection and self-determination concerning a frequent flyer program. The main research question is: To what extent do enrollment rewards enhance behavioral loyalty in a frequent flyer program?

Firms need to find the right balance between customer acquisition and customer retention to maximize customer value (Blattberg and Deighton 1996) and firm profits. It should not be a firm’s objective to attract the most customers but rather to persuade the profitable customers with high potential (Zeithaml 2000). Loyalty programs can assist this process, since firms can influence which customers can join their program and can therefore be selective. Loyalty programs can be used as an instrument to reward and encourage loyal behavior and to establish long-term relationships (Sharp and Sharp 1997). Loyal customers are valuable to firms since it is argued that loyal customers are less price sensitive, behave more loyal in terms of share-of-wallet and buying frequency and are less sensitive to competitors’ offers (Demoulin and Zidda, 2007). Despite the ongoing debate about whether loyalty programs are affective tools to create real long-term customer loyalty, Nako (1992), Kopalle and Neslin (2003) and Liu and Yang (2009) argue that in the aviation industry, frequent flyer programs do enhance behavioral loyalty and are proven to be effective.

(4)

Self-determination. Members who enroll on own initiative (intrinsic motivation) are referred to as self-initiated members. Members who believe they enrolled through a firm’s marketing initiative (extrinsic motivation) are referred to as incentive-based members (Dholakia 2006). Customers, who join on their own initiative, will show more behavioral loyalty in terms of expenditure levels, higher purchase frequency and higher future behavioral intentions than members enrolled through incentive awards. (Dholakia 2006; Steffes et al. 2008).

An exploratory research is designed to test the effectiveness of enrollment rewards on behavioral loyalty in a frequent flyer program accounting for the self-selection bias. This research is based on one specific marketing enrollment campaign by one of Europe’s leading airlines, executed in the Dutch market. Overall the self-selection bias did not hold in this thesis. This might be influenced and explained by the data-source used that only possesses part of the total bookings. Only members have higher initial levels of transactions than non-members, which should be taken into account to prevent overestimating the effect of enrollment. Based on the self-determination, it was found that frequent flyer members showed higher purchase frequency, total revenue and revenue per transaction after 18 months in comparison to non-members. Thus, enrollment rewards do enhance behavioral loyalty of members compared to non-members. A deeper analysis accounted for the differences between incentive-based members, who were extrinsically motivated to participate in the program and self-initiated members, who were intrinsically motivated to enroll. It was found that customers that are intrinsically motivated to participate in the loyalty program show higher behavioral loyalty than the extrinsically motivated members based on flown behavior registered on members’ loyalty accounts. Therefore, intrinsic motivation enhances behavioral loyalty stronger than acquisition incentives.

(5)

Introduction

‘Enroll now and receive 5000 award miles!’ Consumers are constantly exposed to commercial offers. Discounts, gift-cards, coupons and other rewards are utilized to trigger potential and existing customers. Firms implement such pricing mechanisms to boost brand consideration and ultimately try to convince consumers into buying from their company. Through these commercial actions companies attempt to establish and improve long-term relationships with their customers and to enhance customer loyalty.

Why is loyalty important? The corporate perspective.

(7)

The link between retention and profits.

Anderson and Mittal (2000) scrutinize the Satisfaction-Profit-Chain concept by closely analyzing the linkages within the chain. They argue that rather than linear relations, the linkages are asymmetric and nonlinear. Firms need to consider the asymmetric nature of the linkages to strengthen all aspects within the chain to not over or underestimate the outcomes of firm actions. The complexity within the links should be kept in mind in marketing actions and actively monitored at all times (Anderson and Sullivan 1993). Additionally, the satisfaction-retention relationships as well as the retention-profitability relationship are industry and segment specific. The competitive environment affects the relationship between satisfaction and loyalty (Jones and Sasser 1995). In the airline industry, customers incur switching costs that are relatively high compared to other industries, such as the retail industry, especially if they are enrolled in a frequent flyer program. Therefore, they are less likely to switch airlines even though they might not have high satisfaction levels. Anderson and Mittal (2000) found that there is an optimal customer retention rate in the link between retention and firm profits. These findings suggest that a firm should focus on enrolling the valuable customers and not just focus on maximizing the customer base. Firms need to find the right balance between customer acquisition and customer retention to maximize customer value (Blattberg and Deighton 1996) and firm profits. Loyalty programs can assist this process, since firms can influence which customers can join their program and can therefore be selective. Though firms that assume the link between retention and profits is linear will never reach this optimum level since they are exclusively focused on reaching the highest retention rates rather than on the quality of the customers. Therefore, it should not be a firm’s objective to attract the most customers but rather to persuade the profitable customers with high potential (Zeithaml 2000). Leenheer et al. (2007) stress the importance of carefully evaluating customers’ potential for future behavioral loyalty in order to identify the valuable customers upfront. Especially effects on long-term changes in behavior can give insights into how profitable customers really are for the firm.

The impact of loyalty programs

(8)

recognize that loyalty programs can add true value to both customers and the firm (Berman 2006; Wansink 2003), while others doubt the real effectiveness (e.g. Sharp and Sharp 1997). The effectiveness of the loyalty program is also dependent on the design of the program. Continuous programs seem to be more affective on the long-term than short-term loyalty programs (Liu 2007), since customer experience long-term rewarded behavior effects (Taylor and Neslin 2005). Short-term point pressure might also positively affect loyal behavior, though it results in lower long-term loyalty since it requires less involvement and commitment (Taylor and Neslin 2005). The effect on customer behavior likewise varies in findings. Behavioral loyalty describes actual purchase behavior, whereas attitudinal loyalty is about customers’ perceptions and attitudes (Dick and Basu 1994). A positive impact from

loyalty membership (Lal and Bell 2003; Liu 2007; Taylor and Neslin 2003) and none or limited impact (e.g. Sharp and Sharp 1997) on behavior has been discussed throughout time. Sharp and Sharp (1997) observed weak level of excess loyalty, both non-members of the loyalty program as well as members overall showed the same weak level of excess loyalty. Kopalle and Neslin (2003) and Nako (1992) argue that in the aviation industry, frequent flyer programs do enhance behavioral loyalty and are proven to be effective. Frequent flyer programs are therefore found to positively influence consumer’s frequency of flying (Liu and Yang 2009). Liu and Yang (2009) found that in the airline industry, membership increases purchase frequency with 4.24%. These findings emphasize that loyalty programs in the airline industry are an effective tool to create long-term commitment, and to create valuable relationships with customers. The mixed findings on the effectiveness of loyalty programs are related to the diverse research methods that are being used by authors, which make it hard to compare and generalize the outcomes.

Research direction

(9)

frequency of flying (Liu and Yang 2009), so far no research linked different marketing acquisition tools to the effectiveness of the frequent flyer program in terms of behavioral loyalty. Additionally, Bolton, Lemon, and Verhoef (2004) argue that there is more research needed to investigate to what extent the gap between members and non-members’ loyalty is determined by self-selection. The self-selection bias explains that customers who are initially already loyal are more likely to show loyal behavior after enrollment than customers who were not loyal prior to enrollment. Only few studies take the self-selection bias into account (Meyer-Waarden and Benavent 2008; Leenheer et al. 2007), all conducted in the retail industry.

This thesis will provide understanding in the effects of incentives as an acquisition instrument on customers’ long-term behavioral loyalty. Therefore, this thesis will seek to analyze the effectiveness of the usage of incentives to get customers motivated to sign up for a frequent flyer program. It will provide insights into whether this marketing instrument generates more revenue and transactions and it will show under which conditions such an investment is profitable. This will be realized by comparing self-initiated enrolled members, who deliberately decide to join, to members enrolled through the use of the incentive award offer. As a control group, customers of the airline who are not enrolled in the program will be monitored simultaneously during the same period of time to objectively analyze the real effectiveness and profitability of the program. This comparison makes it possible to study the general effects of joining a frequent flyer program on long-term behavioral loyalty in addition to the differences between self-initiated and incentive-based enrollments. Furthermore, this study serves as an approachable thesis to give directions into how initial flying and buying behavior can be used as guidance for the right segmentation for incentive enrollment email campaigns, taking the self-selection bias into account.

Therefore, the central research question of this thesis is:

To what extent do enrollment rewards enhance behavioral loyalty in a frequent flyer program?

Furthermore, this thesis seeks to answer the following additional questions:

(10)

Can initial behavior be used as a predictor for behavioral loyalty in a frequent flyer program, and can it serve as a future directive for segmentation of ‘incentive-based’ acquisition mailings?

Scope

This research is based on one specific marketing enrollment campaign by one of Europe’s leading airlines, executed in the Dutch market. Thus, only customers based in the Netherlands were in the target group. Furthermore, the research is industry specific; therefore this paper is only representative for the airline industry. Throughout this thesis only firm accessible data will be used as input, no use of external and competitive data, since external and competitive information is often incomplete, outdated or hard to obtain for firms. Therefore, this study aims to present a pragmatic and realistic examination of behavioral loyalty only.

(11)

Theoretical Framework

Loyalty from a corporate perspective

Loyalty programs

Companies started introducing loyalty programs as part of relationship marketing as a tool to create customer loyalty and to encourage repeat behavior (Kumar and Reinartz 2002). Loyalty programs can be defined as programs that serve as an instrument to reward and encourage loyal behavior (Sharp and Sharp 1997). Incentives and benefits are offered to program members exclusively to reward and stimulate repurchase behavior and to establish long-term relationships with the customers. One of the first companies that introduced a loyalty program was American Airlines in 1981 with their AAdvantage program. Members accumulate points for the flights they make with American Airlines. Customers that earn sufficient amounts of award miles can reimburse the awards for free flights and upgrades. Today, it remains the world’s largest frequent flyer program. After AAdvantage proved itself successful many other firms followed their lead across different sectors. For example credit card companies, supermarkets, gasoline stations and hotel chains. Especially in the retail industry, loyalty management gained popularity. Berman (2006) states that by 2003 approximately 37% of all retail chains have a loyalty program in America and approximately 90% of all American customers were enrolled in loyalty programs. Bijmolt et al. (2003) find that loyalty programs are also popular in the Netherlands, where approximately 80-90% of customers own a loyalty card, and around 35% own 4 or more cards.

(12)

best. Prior to enrollment customers compare potential costs and benefits (Bell and Lal 2003). Customers who evaluate the potential benefits more substantial than the associate costs and risks are more likely to adopt such a program (Demoulin and Zidda 2009; Leenheer et al. 2007). Therefore, Kivetz and Simonson (2003) argue that customers’ assessment of personal effort that must be invested in order to be able to obtain expected benefits explain the probability of joining. This will be further elaborated on in the self-selection section and the intrinsic and extrinsic motivation section. The rewards and benefits are driven by repeat purchases, which strengthens the long-term customer lock-in (Sharp and Sharp 1997). So, if customers perceive the benefits from a frequent flyer program more positive than the costs they might consider enrolling.

Additionally, firms benefit from enrolling customers in a loyalty program since it enables firms to use data tracking. These data tracking systems are member-based. Therefore, firms can analyze the behavior of members within loyalty programs. Analysts can recognize trends within the loyal customer base and make future prediction based on past behavior. Customer analyses helps to gather data-driven knowledge that will help firms to better understand their customers. With this better understanding of behavior, customers can be served more effectively. Ideally, companies can use the available customer data to tailor its marketing offers. Customers can receive preferential treatment and personalized offers to fit personal preferences.

Loyalty from a customers perspective

Self-selection bias.

Little research has been done on the causal relationship between past behavior and loyalty program enrollment. Customers have individual differences, which can be taken into account in marketing acquisition selections for loyalty programs (Lewis 2004). The individual factors that increase the likelihood of enrollment are called ‘self-selection’. Lewis (2004) for example

finds that customers’ demographics, the physical distance to a supermarket is positively

(13)

Simonson (2003) find that consumers’ usage level can drive the perceived effort advantage.

The heavy buyers also referred to as early adopters (Meyer-Waarden and Benavent 2009), obtain the benefits without having to substantially increase their effort or purchase behavior (Demoulin and Zidda 2009; Dholakia 2006; Kivetz and Simonson 2003). Based on this perceived effort advantage, heavy buyers are found to be more likely to enroll in a loyalty program (Kitvetz and Simonson 2003). Additionally, heavy buyers perceive lower costs (Mauri 2003). Whereas, for light and moderate buyers is it not as easy to receive rewards based on their behavior prior to enrollment. These two groups of customers need to significantly increase their spending and frequency in order to obtain the program’s benefits.

While the heavy buyers obtain the program’s benefits without having to increase their

spending and frequency.

(14)

Thus, the self-selection bias explains that already loyal customers may have a higher likelihood to enroll into the program (Leenheer et al. 2007). Heavy users that continue their behavior within the loyalty program will be rewarded earlier than light and moderate users that need to increase their buying behavior to receive the same awards as the heavy users receive. Therefore, loyalty programs are often more attractive for heavy buyers, as they perceive the program as adding more value compared to light buyers. For light buyers it is less attractive to enroll. Based on existing theory, the first hypothesis is formulated.

H1: Frequent flyer members will show higher purchase frequency, total revenue and revenue/transaction prior to (non) enrollment than non-members.

Ceiling effect.

Lal and Bell (2003) and Liu (2007) are the first to examine the moderating effect of consumers’ initial usage levels on consumers’ loyalty in the retail industry. Meyer-Waarden

and Benavent (2009) and Liu (2007) found that after enrollment customers with initial lower levels of behavior will change their behavior relatively more compared to the initially heavy buyers. Liu (2007) segments customers’ behavior in three categories according to their total

spending in their first month of enrollment in a convenience store chain loyalty program. A separation is made between light, moderate and heavy buyers. Liu’s research (2007) therefore

does not fully cover the self-selection bias since customers might already change their behavior patterns at the moment of enrollment. The classification is therefore not based on actual previous behavior. However, Liu (2007) finds that customers’ usage levels at the

beginning of a loyalty program can be used to predict future usage levels and their exclusive loyalty to the firm. The study examines the convenience stores’ loyalty programs. Heavy

(15)

This can be explained by the fact that heavy users might already spend their maximum amount of money, which might make increasing purchase frequency and spending impossible. Whereas light and moderate buyers still have room in their budget to change their spending patterns. This so-called ceiling effect might explain why customers with relative low purchase levels can change their behavior more over time (Lal and Bell 2003). The latter findings explain the ceiling effect in the retail industry (e.g. supermarkets). The airline industry works differently. In the retail industry there is room to increase spending patterns for low spenders, since the prices are relatively low compared to flight ticket prices and the buying frequency is a lot higher for supermarkets compared to buying flight tickets. Therefore, in the airline industry it is less likely that customers can increase their spending patterns. Though, it might be possible for low spenders in the airline industry to increase their spending and frequency if they do not yet buy all their flights from the focal airline.

Translating the ceiling effect into the airline industry, customers’ flying behavior is limited by their personal budget and time available. Dowling and Uncles (1997) argue that personal restrictions can limit customers’ demand. For certain people one flight per year might be the maximum possible, whereas for more wealthy people this maximum can be a lot higher. Differences can originate from money available as well as time available to travel. This study aims to better understand this ceiling effect by comparing the changes in behavior of members with the behavior of non-members before and after (non) enrollment. Based on personal restrictions, overall, customers with high initial levels of transactions are assumed to remain high in flight frequency after (non) enrollment and customers with low initial levels of transactions are assumed to remain low in flight frequency after (non) enrollment. It can be assumed that both customers with low and customers with high initial levels of transactions are limited in their buying behavior due to the ceiling effect. Therefore, the ceiling effect can explain that, overall, customers with low initial levels of transactions will remain low after (non) enrollment or even become inactive (no purchases) due to personal restrictions. Also, overall, customers with high initial levels of transactions will remain high after (non) enrollment, assuming their share of wallet is already high. Consequently, the enrolled members with high initial levels of transactions are more likely reach higher tier levels. From this theory the following hypotheses can be derived:

(16)

H3: Members with high initial levels of transactions are more likely to end up in higher tier levels than members with initial low levels of transactions.

Intrinsic and extrinsic motivation

Internal motivation from customer self-enrollment.

Customers who join a firm, more specifically, who join a loyalty program, can be split into two groups, incentive-based and self-initiated. This division is based on the underlying reason that drives the decision to enroll in a loyalty program. This decision can either be on own initiative (intrinsic motivation) or through marketing initiatives (extrinsic motivation). The two groups of customers react differently to marketing initiatives and show different behavior after joining a firm. When the decision to enroll is initiated by the consumer, the customer is referred to as self-determined (Dholakia 2006). Whereas people who believe they enroll through a firm’s marketing initiative are referred to as firm-determined (Dholakia 2006). This separation originates from the self-determination theory (Deci 1971). This psychology theory implies that customers’ motivation to engage in behavioral loyalty is context dependent and varies upon the type of reward (Deci 1971). People who are intrinsically motivated engage with a firm purely for themselves, since it provides an internal reward (Meyer-Waarden 2013). Their buying behavior is led by and based on their internal gratification. Customers want to interact with the firm. Extrinsic motivation is the result of an external offer award, for example an economic benefit (Deci et al. 1999). The reward is obtained after customers’ action, for example enrollment in a loyalty program, in return. In this study the self-determined customers are the self-initiated customers who enroll in the frequent flyer program without the use of a direct incentive reward. The firm-determined customers in this study are the incentive-based customers who enrolled through the direct incentive reward, extrinsically motivated. Dholakia (2006) and Steffes et al. (2008) both argue that compared to customers who join a loyalty program through an enrollment incentive, customers who join on their own initiative, intrinsically motivated, show more behavioral loyalty in terms of expenditure levels, higher purchase frequency and higher future behavioral intentions.

(17)

and loyalty. The study indicates that only for budget-optimizing oriented customers an economic award is perceived intrinsically motivated, whereas for uncertainty avoiding, hedonist, social-relational and functional purchase oriented customers an economic reward is not intrinsically motivated and does not increase loyalty and purchase behavior (Meyer-Waarden 2013). It can be assumed that an award does not create behavioral loyalty unless the motivation to engage in expected behavior is in line with customers’ intrinsic motivation. Though, the intrinsic motivation can be lead by the self-selection bias. Before testing the behavioral loyalty of the incentive-based members, self-initiated members and non-members, the self-selection bias needs to taken into account. Customers who are intrinsically motivated could have higher initial levels of behavioral loyalty due to the self-selection bias. After having controlling for the self-selection bias the differences in behavioral loyalty between members and non-members can be tested objectively. Therefore, the following hypothesis is formulated.

H4: Self-initiated enrolled members will show higher purchase frequency, total revenue and revenue/transaction prior to enrollment than incentive-based enrolled members.

External motivation from corporate incentives.

(18)

no long-term effect on sales, since it is an extrinsic motivation, which is often not in line with customers’ long-term goals. So, single-shot promotions are not in line with the main concept of a loyalty program, which is to create a long-term relationship with valuable customers (Yi and Jeon 2003). Therefore it is assumed that customers who are intrinsically motivated members will establish more valuable long-term relationships with the airline that extrinsically motivated members. From this theory the following hypothesis can be derived:

H5: Self-initiated enrolled members are more likely to be in higher current tier levels than incentive-based enrolled members.

Additionally, extrinsically motivated customers will be less active after enrollment, since the incentives do not enhance long-term commitment. Before comparing the self-initiated members with the incentive-based members, H6 first aims to verify the positive effect of membership in a frequent flyer program on behavioral loyalty. Thus,

H6: Non-members will be more inactive after 18 months than frequent flyer members.

H7: Self-initiated enrolled members will be more active after 18 months than incentive-based enrolled members.

Delayed rewards, like receiving a gift after making your first purchase are in line with the objective of a loyalty program, since it is intended to increase loyalty on a long-term perspective and to retain customers (Taylor and Neslin 2005). Therefore, delayed rewards do create switching costs since customers get rewarded for their purchase behavior. In contrast, direct rewards, such as a direct gift, do not result in any switching costs, since customers do not have to perform any actions in return to receive the reward. This non-existence of switching costs does create customer lock-in. Therefore, Leenheer et al. (2007) advice to rather use delayed rewards over direct rewards since it will have a stronger effect on customers’ decision for enrollment.

(19)

is conducted in a US financial services firm, where customers with high account balances can enroll in the firm’s loyalty program. Dholakia’s (2006) findings show that one-year after enrollment, the self-determined customers spend $182, whereas the firm-determined customers induced by an incentive or sales-pitch only spend $155. Thus, customers who feel they were induced to join the company, on average buy less frequent and spend less money compared to the customers who feel like they joined on their own initiative. The self-determined customers do significantly increase their relational behaviors. Therefore, Dholakia (2006) argues that the direct one time awards to persuade customers might not be the most effective tools to create long-lasting profitable relationships with customers. Additionally, Dholakia’s (2006) theory of self-determination is supported by academic research in psychology. Williams et al. (1996) find that self-determined people enrolled in a weight-loss program show more overall weight-loss and have higher attendance. Another psychology paper (Vallerand, Foltier and Guay 1997) argues that student with high self-determined motivation to attend school show lower levels of drop out in the following years.

The latter findings by Dholakia (2006) in combination with findings that suggest that single-shot promotions especially tend to attract the less involved customers (Cortinas et al. 2008; van Heerde and Bijmolt 2005; Taylor and Neslin 2003) would suggest that customers who are firm-determined are more prone to be less involved. Consequently, the low involved customers are more likely to join the frequent flyer program through the direct incentive. Therefore it can be assumed that these low involved customers, who are firm-determined, will show less behavioral loyalty on the long-term. Whereas customers who join the frequent flyer program based on self-determination will show more long-term behavioral loyalty. In this study the airlines offers customers a direct incentive of 5000 award miles for direct enrollment in the frequent flyer program. This offer is not directly related to customer behavior, since no relational behavior like a first purchase is required to be able to receive the award. The offer undermines the long-term design of a loyalty program and simultaneously does not create any switching costs.

(20)

relationships. Thus, this study predicts that self-initiated enrolled members will shower higher levels of behavioral loyalty than incentive-based enrolled members. Before comparing the self-initiated members with the incentive-based members, the general effect behavioral loyalty in the loyalty program needs to be analyzed by comparing members with non-members. Thus,

H8: Frequent flyer members will show higher purchase frequency, total revenue and revenue/transaction after 18 months than non-members

H9: Self-initiated enrolled members will show higher purchase frequency, total revenue and revenue/transaction after 18 months than incentive-based enrolled members.

The combined effect of self-selection and self-determination

In conclusion, there are still mixed findings about members’ behavioral loyalty after enrollment. It is proven that the self-selection bias does enhance the likelihood of enrolling in a loyalty program (Liu 2007). Therefore, it assumed that the higher the initial behavioral loyalty is, the higher the likelihood of enrollment. Though the self-selection bias says little about how members will behave after enrollment due to potential ceiling effects (Lal and Bell 2003). This study’s additional value is that it considers the differences between self-initiated enrollment and incentive-based enrollment. Simultaneously this research tries to control for the ceiling-effect by using a control group of non-enrolled members. It is assumed that the self-determination explains that customers with low initial levels of transactions are less likely to not enroll and consequently show less behavioral loyalty. This is in line with the self-selection bias, which assumes that customers with low initial levels of transactions are less likely to enroll. Additionally it is assumed that customers with high initial levels of transactions are more likely to enroll, where self-initiated members will have higher levels of initial transactions than incentive-based members. Based on latter theories discussed throughout this chapter, we assume the following hypothesis:

(21)

Conceptual model

The theoretical framework and the associated hypotheses are conceptualized in figure 1.

H2, H3

H1, H4 H10

H5, H6, H7, H8, H9

Figure 1. Conceptual model

Methods

Research design

The real-life, internal data provided by one of Europe’s leading airlines is used to design a quantitative research. The latter defined hypotheses will be tested in a quantitative approach where theory can be tested in practice. Explanatory research can be explained as “a type of research design, which has as its primary objective the provision of insights into and comprehension of the problem situation” (Malhotra 2009). The research aims to better understand the causal relationships between the independent variables: initial loyal behavior, the test units: (non) enrollment in a loyalty program (where enrollment is split up in two types: incentive-based and self-initiated) and the dependent variables: current behavioral SELF DETERMINATION 1. Incentive-based enrollment 2. Self-initiated enrollment 3. Non- enrollment 4. BEHAVIORAL LOYALTY - # Transactions - Total revenue

- Revenue per transaction - Active/ inactive - Tier level INITIAL BEHAVIORAL LOYALTY - # Transactions - Total revenue

(22)

loyalty, 18 months after (non) enrollment. This enables to explain to what extent different levels of former purchase behavior drives the likelihood of joining a loyalty program. The major objective of causal research is to obtain evidence regarding causal-effect relationships (Malhotra 2009). This research is based on real customer data; therefore this research will provide a realistic analysis of the theory in practice concerning this sample. No empirical research yet combines the behavioral data before and after (non) enrollment for comparing two groups of differently enrolled members and a non-enrolled control group.

Sample

The target segment that received the offer was selected based on their flying frequency prior to (non) enrollment, expressed in transactions. These transactions are assumed to be return flights. The range was set between 1 to 10 transactions over the last 18 months before the incentive offer email was send. On April 18th 2012, 7330 people, who all satisfied the criteria, received the incentive offer in their email inbox. The offer consisted of receiving 5000 award miles directly for enrolling in the frequent flyer program. The offer was valid till the 2nd of May. The campaign resulted in 233 enrolled customers who made use of this direct offer. This is a conversion rate of 3.18%. Throughout this research, this group of 233 incentive-based enrolled members will be referred to as group 1, the self-initiated members as group 2 and the non-enrolled customers as group 3. The following table provides an overview of group size, male/female ratio and age range for all 3 groups. It can therefore be assumed that all groups have relatively even distribution in female-male ratio. The members’ age appears to be normal for both the incentive-based and self-initiated members according to the histogram with a slight positive skew and the normal Q-Q plot, which shows little data points strayed from the line. Both groups show a peak around the age of 30 and both have a median of 43, which means that half of the members’ age is below 43. Though, the Shapiro-Wilk test for normality performed for both groups (p= .000), results in a non normal distribution concerning members’ age.

Group 1 (Incentive-based) Group 2 (Self-initiated) Group 3 (Non-enrollment) Total N 233 194 196 623 N male 105 87 99 291 N female 128 107 97 332 % male 45,1% 44,8% 50,5 46,7% % female 54,9% 55,2% 49,5% 53,3%

Age range 21-82 22-78 unknown 21-82

Mean age 45 45 unknown 45

(23)

The three distinct different groups are selected based on their (non) enrollment and for the enrolled members, whether they enrolled through an incentive offer or self-initiated without any incentive offer. The group of incentive-based enrollments, group 1, (n=233) is a fixed number of enrollments. The number of customers in this group is established by counting the number to enrollments in the offer period who made use of a special ‘incentive code 5000’ in the enrollment process. Thus it is not a random sample. This study is a non-probability sample since the group is selected based on specific criteria (Malhotra 2009). First is the type enrollment; incentive-based, self-initiated or non-enrollment and secondly, it is based on the fixed criteria of 1-10 transactions prior to enrollment. Thus, all groups are purposely chosen, resulting in purposive sampling (Malhotra 2009). This implies that non-probability selective method might be subject to error and bias (Malhotra 2009). Additionally, it needs to be addressed that this research is based on only one campaign, performed in the aviation industry. Therefore, this study might not be generalizable for other industries that also make us of loyalty marketing campaigns offering an enrollment incentive. The sample used in this paper is therefore not representative for the entire population. Though this research is representative for this specific case and will help to understand the effects of this particular situation. It will provide solid basic insights into the effectiveness of similar loyalty management campaigns in the aviation industry. Additionally it accounts for the self-selection bias by analyzing the effects of previous loyalty behavior on current behavior after (non) enrollment.

Data Collection

(24)

Customers who book a flight online use their email-accounts to receive their flight confirmations and electronic tickets. Within this booking process customers are given the option to subscribe for the airline’s online newsletter. This option can be ticked in a box at the bottom of the website in the booking process. The customers who sign up for this newsletter do automatically give permission to the airline to be contacted for commercial purposes. Optional in this sign-up process is to share limited personal information, such as customers’ gender and birthday. This option was not mandatory at the time the customers in this sample enrolled. Since the enrollment offer of 5000 miles can be identified as a commercial offer, only customers who actively enrolled for the newsletter were approached with this offer by email. Unfortunately, customer information prior to enrollment is limited. Customers who are not (yet) enrolled in the loyalty program can only be tracked by their email-account as the identification key. Ideally the registered name should be used, though last name and first name cannot be used as a unique identification key since too many people share the same name. As a result only booking information can be deducted from the Ebt (Electronic Billing Tool), which represents direct online bookings at the airline’s homepage. As a result behavioral loyalty prior to the email offer can exclusively be measured by online buying behavior. This Ebt-data is available for all 3 groups, for both the 18th months before (non) enrollment, and the 18 months after (non) enrollment. It is assumed that of all bookings, Ebt data account for approximately 30% of all the airline’s ticket sales. Tickets that are not sold through Ebt are for example offline sales and sales through travel agencies.

(25)

Ebt cannot be traced back to an individual (email). Therefore, the complete available data for group 1 and 2, data per FFN, cannot be used for this comparison, since the behavior of these two groups will be heavily overstated in comparison with group 3. The first database therefore consists of Ebt data only. For all 3 groups this type of data is used for measuring initial loyal behavior and behavior after (non) enrollment. A second research design is introduced in order to use the full available flown data from the enrolled members. It is designed to account for all behavioral loyalty of group 1 and 2, based on their flying behavior registered on their frequent flyer account. The second design is introduced to solve the problem of different types of data. This enables to compare incentive-based members to self-initiated members, which will be further explained in the following sections.

Group 1 consists of the 233 enrolled members, who are identified by their enrollment code. The initial mailing list, 7097 (excluding the incentive-enrolled members) was used to select the people for group 2 and 3. The mailing list was imported in a customer-tooling program that ran a query to identify which customers are currently enrolled. Criteria were that they enrolled after the incentive offer was finished. The query resulted in two lists, one with enrolled members and one with non-enrolled members. The first list was used to select group 2. From the date that the offer became invalid customers were selected until the group consisted of 200 members. Resulting in enrollments between the 3rd of May and 27th of August 2012. After removing duplicates 194 remained. Group 3 was randomly selected from the list of non-enrolled customers, under the condition that gender was known. After removing duplicates 196 remained.

Measures

(26)

per individual was accumulated, representing total initial revenue over 18 months. Lastly, the total revenue was divided by the flying frequency to calculate the average amount of euros spends per transaction, referred to as initial revenue per transaction. These 3 measurements are used as the initial behavioral loyalty variables. The same procedure was applied to set up the after (non) enrollment behavioral loyalty variables, based on Ebt data. Thus, flying frequency Ebt, total revenue Ebt and revenue per transaction Ebt were computed into variables. Additionally, the same variables were subtracted from the loyalty program information system. The frequent flyer numbers of the people from group 1 and 2 were entered in this system and a query was run to calculate Flying Frequency, total revenue and revenue per transaction per loyalty program member. Note: transactions in the online loyalty system are measured in segments. Every part, single flight, of the journey is counted as a segment. Therefore, the transactions in the LP tool are overstated compared to the Ebt transactions. Consequently, it is not possible to align and compare the transactions directly. Consequently also the revenues do not measure the same value. The following table provides an overview of all variables used throughout this thesis and shows which variables are available for which group. The variables that are not yet covered in the methods section will be explained in the following data analysis section.

Table 2. Overview variables

* 1= incentive-based, 2= self-initiated, and 3= non-enrollment

** Ebt= bookings direct at airline’s webpage, ACCOUNT= bookings per FFN

Variable Unit Available Data-base

Initial flying frequency Transactions: retour flights Group 1, 2 & 3* Ebt** Total initial revenue Euro Group 1, 2 & 3* Ebt** Initial revenue per

transaction

Euro Group 1, 2 & 3* Ebt**

Initial flying frequency categories

Low, mediate, and high Group 1, 2 & 3* Ebt** Flight activity Ebt Active & inactive Group 1, 2 & 3* Ebt** Flying frequency Ebt Transactions: retour flights Group 1, 2 & 3* Ebt** Total revenue Ebt Euro Group 1, 2 & 3* Ebt** Revenue per transaction

Ebt

Euro Group 1, 2 & 3* Ebt**

Flying frequency Transactions: segments Group 1 & 2* ACCOUNT**

Total revenue Euro Group 1 & 2* ACCOUNT**

(27)

Data analysis

Data Processing: The initial selection was based on initial behavior with a range of 1-10

transactions (number of flights). The frequency distribution of initial flight frequency can be found in Appendix A. The variables were checked for extreme values. In the account-data there are several extreme values in transactions and in total revenue (see table 4). Though these values represent real people with high purchase frequency and can therefore not be disregarded. Thus, the extreme values will not be deleted.

Ebt-data Group Initial Transaction Initial total revenue Initial revenue/ transaction Transactions Total revenue Revenue/ transaction 1 Incentive-based N 233 233 233 233 233 233 Median 3,00 699,22 211,38 1,00 244,46 145,45 Mean 3,14 1028,97 336,12 1,87 583,70 219,31 SD 1,34 911,11 260,92 2,24 834,70 266,91 Minimum 1 105,95 88,37 0 0,00 0,00 Maximum 10 6971,37 1599,93 12 5408,62 1313,57 2 Self-initiated N 194 194 194 194 194 194 Median 3,00 724,34 234,50 1,00 359,36 174,49 Mean 3,19 1088,34 344,99 2,40 845,38 254,07 SD 1,36 949,00 255,54 2,87 1254,52 288,88 Minimum 1,00 108,37 99,53 0,00 0,00 0,00 Maximum 8,00 4778,04 1615,50 16,00 7303,67 1460,73 3 Non- Enrollment N 196 196 196 196 196 196 Median 3,00 743,65 278,63 0,00 0,00 0,00 Mean 2,77 1009,29 367,56 1,01 392,10 155,12 SD 1,29 834,21 262,81 1,68 820,65 252,43 Minimum 1 104,82 103,58 0 0,00 0,00 Maximum 8 4165,10 1268,69 10 5784,02 1192,36 Total N 623 623 623 623 623 623 Median 3,00 714,68 233,75 1,00 215,60 143,14 Mean 3,04 1041,26 348,77 1,77 604,91 209,94 SD 1,34 899,06 259,78 2,37 995,90 272,08 Minimum 1 104,82 88,37 0 0,00 0,00 Maximum 10 6971,37 1615,50 16 7303,67 1460,73 Table 3. Descriptives variables - Ebt-data

(28)

which is the result of many extreme values in a data set. Resulting in the mean and median on the right of the peak value. All variables look not normal on the normal Q-Q plot, since the data points strayed from the line. Therefore the variables are non-normal distributions. This is confirmed by Shapiro-Wilk tests of normality, with p=.000 for all three independent variables. Therefore the null hypothesis (the samples come from a normal distribution) was rejected for all three variables resulting in three non-normal distributions. In order to use the ANOVA test or the Pearson correlation, a normal distribution is assumed. The dependent variables were transformed into natural logarithm variables to correct the distribution. The Log. variables showed a more normal distribution when assessing the histogram and QQ-plot (Appendix B). The Shapiro-Wilk tests of normality indicated that the distribution is still not normally distributed (p=.000). Though, for the ANNOVA test it can be assumed that the outcome is reliable using the transformed variables.

Account data Group Initial Transaction (retour flights) Initial total revenue Initial revenue/ transaction Transactions (segments) Total revenue Revenue/ transaction 1 Incentive-based N 233 233 233 233 233 233 Median 3,00 699,22 211,38 2,00 145,00 45,00 Mean 3,14 1028,97 336,12 3,63 478,08 94,48 SD 1,34 911,11 260,92 5,24 1106,64 135,74 Minimum 1 105,95 88,37 0 0,0 0,00 Maximum 10 6971,37 1599,93 50 13851,00 847,00 2 Self-initiated N 194 194 194 194 194 194 Median 3,00 724,34 234,50 4,00 355,50 73,44 Mean 3,19 1088,34 344,99 6,26 955,65 118,66 SD 1,36 949,00 255,54 9,81 1697,16 133,16 Minimum 1 108,37 99,53 0 0,00 0,00 Maximum 8 4778,04 1615,50 102 13353,00 608,67 Total N 427 427 427 427 427 427 Median 3,00 710,52 217,71 3,00 230,00 55,83 Mean 3,16 1055,94 340,15 4,82 695,05 105,47 SD 1,35 927,89 258,23 7,76 1424,28 134,96 Minimum 1 105,95 88,37 0 0,00 0,00 Maximum 10 6971,37 1615,50 102 13851,00 847,00 Table 4. Descriptives variables - account-data

(29)

indicates that the number of transactions, total revenue and revenue per transaction are all non-normal distributions.

Correlation independent variables: The natural logarithm computed variables were used.

There was a strong, positive correlation between total revenue and revenue per transaction, which was statistically significant (r .816, n = 623, p = .000). This can be explained by the fact that if a customer spends more per transaction, his/her total revenue will increase accordingly. Also, there was a moderate positive correlation between total revenue and transactions, which was statistically significant (r .545, n = 623, p = .000). If customers make more transactions the total revenue will increase along. There was no significant correlation between transactions and revenue/transaction. A scatterplot showed a weak downhill negative linear relationship. Customers’ revenue per transaction is independent from the number of transactions. Therefore, revenue per transaction was not used as a predictor (independent) variable to test behavioral loyalty after (non) enrollment. Additionally, since total revenue increases along the number of transactions, there is no need to use both variables. In line with the airline’s segmentation strategy, the number of transactions was used to measure initial behavioral loyalty for comparisons with after (non) enrollment behavioral loyalty.

Data Recoding: To be able to measure the differences between low mediate and high initial

behavioral loyalty, the variable transactions was split into 3 groups, using percentile split. Low (n=110) 1-2 transactions, moderate (n=191) 3 transactions, high (n=126) 4-10 transactions. Most customers made 3 transactions, which is the same value as the median. Therefore, a split into 2 groups was not desirable, since that would result in approximately 25%-75% members in the 2 groups. To measure the differences between active and inactive members a dummy variable was computed where customers who did not show flight activity after (non) enrollment was coded inactive, the rest was coded active. Additionally, to be able to measure the differences between tier levels, another dummy variable was computed. Members in the entry-level tier were coded ivory. The members in one of the upper 3 tiers were coded elite. The upper 3 tiers are not coded separately due to low numbers of members per tier, which makes statistical tests not reliable.

Nonparametric tests. First, to test associations between variables the Chi-square was used. It

(30)

(31)

Results

Self-selection bias.

A Mann-Whitney U test is performed to test whether the self-selection bias can predict why customers enroll in a loyalty program. A 3x3 matrix is designed, group (1,2 and 3) and initial behavioral loyalty variables (transactions, total revenue and revenue/transaction). It is hypothesized that frequent flyer members (group 1 and 2) will show higher initial purchase frequency, total revenue and revenue per transaction than non-members (group 3). Additionally, it is hypothesized that self-initiated members will score higher on all three initial behavioral loyalty variables than incentive-based members. The following table shows the significant relationships only.

Mann-Whitney Test, Initial behavioral loyalty

Table 5. Test results. Significance at α = 0,05

Only 3 from the initial 9 relations (3x3 matrix) were significant. The test showed that there was no significant difference in initial total revenue between the 3 groups. Furthermore, the revenue per transaction was insignificantly different between group 1 and 2, and group 2 and 3. Remarkable finding is that the revenue per transaction is initially significantly higher for group 3, the non-members than the incentive-based members from group 1. The purchase frequency is found insignificantly different between group 1 and 2. The hypothesis only holds for the part that the frequent flyer members indeed show significant higher initial purchase frequency compared to non-members. Though they do not show significantly more total revenue and

(32)

revenue/transaction prior to (non) enrollment than non-members. Therefore, H1 is partly rejected. Furthermore, there is no significant difference in initial purchase frequency between group 1 and 2, though the mean rank of group 1 is higher. The same accounts for the total revenue variable and the revenue per transaction variable. Both had a higher mean rank for group 2 compared to group 1, though the difference was not found to be significant. Therefore, H4 is rejected. Thus, the self-selection bias only explained a higher initial amount of transactions for customers who decided to enroll compared to the non-enrolled customers. It did not show differences concerning total initial revenue and initial revenue per transaction between members and non-members. In addition, it does not hold a predictive role concerning the assumed higher initial behavioral loyalty for self-initiated members (group 2) compared to incentive-based members (group 1).

Ceiling effect: actives versus non-actives. A 2x3 design Pearson Chi-Square test was conducted

to test whether there was a significant relationship between initial purchase frequency (low, moderate and high) and activity (active or inactive). Theory predicts that customers with low initial levels of purchases are more likely to become inactive than customers with high initial levels of purchases, due to personal restrictions (e.g. time and budget). This test was performed while controlling for the three different groups. Over all three the groups the test was significant, Chi-Square X2 (2) = 52,470, n=623, p= .000. In all three groups, the active-members showed a negative standardized residual for the low initial number of transactions, which indicates that the number of active members was below the expected count. Thus, customers with low initial purchase frequency are more likely to become inactive than customers with high initial purchase frequency. In addition, in all groups for the active-members, the standardized residual was found positive for the high initial number of transactions. This indicated that there were more actives members in the high initial category than the expected count. Thus, active members are more likely to have a higher level of initial transactions than inactive members. Therefore, H2 is supported.

Current tier level by initial frequency. To determine whether current tier level is related to the

(33)

levels of transactions are currently in the elite tiers, whereas 7 Elite members had a moderate levels of transactions, and of the customers with initial high levels of transactions, 11 members are currently in the Elite tier levels. There was a significant relationship between tier level and initial levels of transactions, X2 (2) = 7,088, n=427, p= .029. Thus, members with initial high levels of transactions are more likely to end up in higher tier levels than members with initial low levels of transactions, which confirms the ceiling effect. Thus H3 is supported.

Self-determination.

Current tier level by group. To determine whether current tier level is related to the way through

which people enrolled into the loyalty program a Pearson Chi-Square test was conducted using a 2x2 design, group (1 and 2) versus current tier level (ivory and elite) was set up. Of all the enrolled members, 5 members from incentive-based group are currently in the elite tiers, whereas 15 of the self-initiated customers are currently part of the elite tier. There was a significant relationship between tier level and enrollment type, X2 (1) = 7,399, n=427, p= .007. Self-initiated enrollment was associated with more members currently in the elite tiers than incentive-based enrollment. The standardized residual from the elite tier for group 1 (Std. Residual=-1.8) indicated that for group 1 the number of elite members was below the expected count. The standardized residual from the elite tier for group 2 (Std. Residual = 2.0) indicated that for group 2 there were more elite members than the expected count. Therefore, H5 is supported. Self-initiated enrolled members are more likely to be in higher current tier levels than incentive-based enrolled members.

Groups: actives versus non-actives. In order to analyze whether the three groups differ

(34)

marginally significant, Chi-Square: X2 (2) = 3,650, n=427, p= .059. From this it can be concluded that self-initiated members (group 2) are more active (73.7%) than incentive-based members (group 1), (65.2%), though this difference is only marginally significant. Thus, H6 is supported, since members are more active than non-members. H7 is rejected, since self-initiated members are not significant more active than incentive-based members.

The assumption was that the flown data (active only) registered per individual loyalty member would include the Ebt data. To verify this assumption both databases were linked to test the overlap. As shown in Figure 2, the overlap of the Ebt database and the database presenting flown record per frequent flyer is 233. This indicates that 55 frequent flyers are present in the Ebt-base but not in the account-base. These 55 members book flights directly at the airline’s website but do not admit their frequent flyer number. Consequently they do not receive any reward miles for their purchases. Contrary, there are 62 members that are actively using their account but do not make any bookings directly on the airline’s website. These customers do not show in the Ebt-data. Though they do they received reward miles for their flight purchases. A complete overview of flight activity can be found in appendix C.

Figure 2. Members’ flight activity overview after enrollment

To examine whether hypothesis 8 and 9 hold, the relationship between the three groups and the three dependent behavioral loyalty variables were compared by the Mann-Whitney test.

(35)

Group 1-2 (Ebt). The Z-score for the relationship between group 1 and 2 regarding number of

transaction equaled -1.620. This indicated that the two groups differed -1,620 times the expected standard deviation of the mean value on the number of transactions variable. The hypothesis assumed that self-initiated members would show significantly higher number of transactions compared to the incentive-based members. The p-value had a value of .105, which is above the critical significance α of .05. Thus, following this assessment it can be concluded that group 2 did not show a significantly higher number of transactions than group 1. A marginally significant difference was recorded between group 1 and 2 concerning the total revenue (Z= -1.773 p =.076) and the revenue per transaction (Z= -1.747 p = .081). Thus,

(36)

group 1 and 2 showed insignificant differences for all 3 behavioral loyalty variables, though total revenue and revenue per transaction were found marginally significant.

transaction equaled -4.961. The hypothesis assumed that incentive-based members would show significantly higher number of transactions compared to the incentive-initiated members (p=.000). Thus, this indicates that group 1 did show a significantly higher number of transactions than group 3. The same significant differences are found between group 1 and 3 concerning the total revenue (Z= -4.457, p =.000) and the revenue per transaction (Z= -3.717, p= .000). Thus, group 1 and 3 showed significant differences for all 3 behavioral loyalty variables.

transaction equals -6.096. The hypothesis assumes that self-initiated members will show significantly higher number of transactions compared to the incentive-initiated members, which was confirmed (p=.000). Therefore, group 2 does show a significantly higher number of transactions than group 3. The same significant differences are found between group 2 and 3 concerning the total revenue (Z= 5.725, p=.000) and the revenue per transaction (Z= -4.923, p= .000). Since all 3 variables had a p-value below the critical significance α of .05 the differences between the groups can be considered significant on all three variables.

Therefore, H8 is supported. 18 months after (non) enrollment, both group 1 and 2, the enrolled members show significant higher purchase frequency, total revenue and revenue per transaction than group 3, the non-enrolled customers.

Account-data. The same design for analysis is executed for the behavioral loyalty variables

(37)

Mann-Whitney Test (Account) Dependent Variable Group N Mean Rank Sum of Ranks Mann-Whitney U Z score p-value (2-tailed) Hypothesis Transactions 1 233 195.99 45666.00 18405.00 -3.364 .001 Accepted 2 194 237.50 45712.00 Total Revenue 1 233 194.43 45302.50 18041.50 -3.645 .000 Accepted 2 194 225.24 46075.50 Revenue/ Transaction 1 233 199.50 46484.00 19223.00 -2.701 .007 Accepted 2 194 231.41 44894.00

The hypotheses assume that self-initiated members will show significantly higher number of transactions, total revenue and revenue per transaction compared to the incentive-initiated members. With the significance at α = .05, a p-value of .001 was found for the number of transaction. A p-value of .000 was found for the total revenue and a p-value .001 was found for revenue per transaction. Based on these results, the differences between the groups can be considered significant on all three behavioral loyalty variables.

For both the Ebt-data and the account-data the mean-rank of all loyalty variables is higher for group 2 compared to group 1. The self-initiated enrolled members show higher purchase frequency, total revenue and revenue/transaction after 18 months than incentive-based enrolled members based on Ebt-data. Though these differences are not found to be significant. The account-data analysis does show a significant difference for all 3 variables. The self-initiated enrolled members show a significant higher purchase frequency, total revenue and revenue/transaction after 18 months than incentive-based enrolled members based. Therefore, H9 is only supported for the loyalty behavior registered on the individual frequent flyer accounts. H9 is not supported when accounting for members’ bookings only at the Airline’s website.

Combined effect of self-selection and self-determination.

(38)

variable, second with the natural logarithm calculated variable for total revenue, and lastly, the natural logarithm calculated variable for revenue per transaction after (non) enrollment. It is assumed that after the log transformation the variables are close enough to a normal distribution to use the Two-Way ANOVA test to get reliable outcomes. Theory suggested that both the self-selection and self-determination could explain the behavioral loyalty. This was found significant. Findings show that the independent variables ‘individually’ all have a significant p-value for all three dependent variables. Thus, there is a statistically significant difference in the mean number of transactions between the groups. Also, there is a statistically significant difference in the mean total revenue and the mean revenue per transaction between the groups. With the initial behavioral loyalty as independent variable, there is a statistically significant difference in the mean number of the three levels of transactions (and total revenue and revenue/transaction) between the initial levels of transactions. Though when assessing the interaction effects of the independent variables on the dependent variable, all interaction effects were insignificant (see table 8 below). Thus, the two independent variables combined have no significant effect on the behavioral loyalty variables after (non) enrollment. This means that although the variables individually are significantly different, the two independent variables combined do not explain the dependent variable in the same way. The variables both have a different relationship with the dependent variables. Therefore H10 is rejected. Members’ behavioral loyalty after (non) enrollment cannot be explained by the combined effect of the self-selection and the customers’ self-determination for enrollment. The ‘separate’ effects, of the self-determination and self-selection, can only explain the relationship with the behavioral loyalty.

Variable Dependent Variable Sig. Hypothesis

Group Transactions .000 Supported

Initial levels of transactions Transactions .001 Supported Group *

initial levels of transactions Transactions .471 Rejected

Group Total revenue .000 Supported

Initial levels of transactions Total revenue .000 Supported Group *

initial levels of transactions Total revenue .333 Rejected

Group Revenue per transactions .000 Supported

Initial levels of transactions Revenue per transactions .000 Supported Group *

(39)

Table 8. Two-Way Anova (Dependent Variable: Log. Transactions Ebt)

To understand the specific groups’ differences within the independent variables, a post-hoc test, a multiple comparisons table was used. The Tukey post-hoc test shows whether the relationships within independent variables (groups and initial level of transactions) are significant for the 3 dependent behavioral loyalty variables (see table 9 below)

This overlaps with the findings from the latter hypothesis H8 and H9. The p-values from the Tukey test slightly differ from the p-values from the Mann-Whitney test, since for the latter Tukey test the natural logarithm variables were just, whereas the Mann-Whitney test used the original variables. The same outcome was found, group 1 (incentive-based) and 2 (self-initiated) showed insignificant differences for all 3 dependent behavioral loyalty variables, though total revenue and revenue per transaction were found marginally significant.

Concerning the interactions within the independent variable initial level of transactions (low-moderate, moderate-high, low-high) all interactions were significant for the 3 different dependent behavioral loyalty variables. Thus, low, moderate and high levels of initial transaction differ significantly when analyzing the dependent behavioral loyalty variables. All mean ranks were constantly higher for the highest initial purchase frequency. This supports the ceiling effect.

Dependent variable

Groups Sig. Different

Transactions Incentive-based self-initiated .080 No non-enrollment .000 Yes Transactions self-initiated Incentive-based .080 No non-enrollment .000 Yes Transactions

non-enrollment Incentive-based .000 Yes

self-initiated .000 Yes

Dependent variable

Initial transactions Sig. Different

Transactions

low initial level of transactions moderate initial frequency .000 Yes high initial frequency .000 Yes Transactions

moderate initial level of transactions low initial frequency .000 Yes high initial frequency .000 Yes Transactions

Do enrollment rewards enhance behavioral loyalty in a frequent flyer program?

Do enrollment rewards enhance behavioral loyalty

in a frequent flyer program?

The predictive role of self-selection and self-determination.

Do enrollment rewards enhance behavioral loyalty

in a frequent flyer program?

The predictive role of self-selection and self-determination.

Management Summary

Contents

Introduction

Theoretical Framework

Conceptual model

Methods

Results