WHAT IS THE RELATIONSHIP BETWEEN THE CUSTOMER EQUITY CURRENT AND FUTURE COMPONENT AND FIRM VALUE

(1)

FUTURE COMPONENT AND FIRM VALUE

&

HOW IS CUSTOMER EQUITY STRATEGY RELATED TO FIRM VALUE AND CUSTOMER BASE

Master thesis, MScBA, specialization Marketing University of Groningen, Faculty of Economics and Business

December 2, 2010

DANIELA NAYDENOVA Student number: 1747959 Address: Padangstraat 43A Place: 9715CL, Groningen tel.: +31 624884633 e-mail: d.a.naydenova@student.rug.nl Supervisors Dr. Thorsten Wiesel prof. dr. J.C. Hoekstra

Acknowledgment: I would like to express special thanks to my supervisor Dr. Thorsten Wiesel for

the great support and patience throughout the whole process of this research. Also, I would like to thank to Martinus van der Wal and Meng Janine Lui for their input in the dataset preparation and the fruitful discussions on the research problems.

Keywords: customer equity, firm value, observable marketing metrics, churn rate, acquisition rate,

(2)

The present master thesis belongs to the stream of research on the marketing finance interface, specifically firm valuation based on observable marketing metrics. The focus is on analyzing the effects on firm value of the marketing asset customer equity. Customer equity is decomposed into customer equity of retained customers and customer equity of newly added customers and their effects are studied separately. Firm value is measured by the ratio between the market and book value of the company - Tobin’s Q. Further, a customer equity strategy classification is developed on the basis of the change in the customer metrics, components of customer equity: average revenue per user, churn and acquisition rate. This classification is applied to study the association of eight inferred customer equity strategies with firm value and the customer base. Of specific interest is the customer equity maximization strategy, which based on literature is expected to have a strong positive association with both firm value and the customer base. The goal of the research is to provide a relevant contribution in terms of finding strategy implications related to customer equity and firm value and also in terms of the application of appropriate empirical methods.

A time-series dataset with quarterly marketing and financial data of 133 international wireless companies has been compiled for the analysis, which contains quarterly data for the period of ten years.

The first finding from this research is that the customer equity of the newly added customers has a significant positive relationship to Tobin’s Q, while the relationship between the customer equity of the retained customers and Tobin’s Q does not turn out to be significant. Further, the customer equity strategies that are significant and positively associated with Tobin’s Q are not the expected ones, i.e. customer equity maximization is not one of them. However, what makes the outcomes quite interesting is that the positively associated with Tobin’s Q strategies are characterized by their specific focus, i.e. customer base expansion or customer revenue maximization only, but not both as implied by the customer equity maximization objective.

(4)

INTRODUCTION

Over the past decade, a growing body of research linking the disciplines of marketing and finance has aimed to establish the impact of marketing on firm value. A number of studies investigate the relationships between important marketing metrics and firm profitability, risk and shareholder value (Aksoy, Cooil, Groening, Keiningham, & Yalcin, 2008; Luo, Homburg, & Wieseke, 2010; O'Sullivan, Hutchinson, & O'Connell, 2009; Rego, Billett, & Morgan, 2009; Srinivasan & Hanssens, 2009). Relevant analytical models for marketing and firm valuation have been developed, raising intense discussions on the appropriate methodology for financial markets research in marketing (Mizik & Jacobson, 2009; Srinivasan & Hanssens, 2009).

There are several reasons which justify the need for ongoing research on the marketing-finance interface. First, from the perspective of the marketing profession, one of the major drivers of the influence of the marketing department within the firm is marketing accountability (Verhoef & Leeflang, 2009). Nowadays it has become critical for the marketing professionals to employ effective methods in order to account for marketing resource allocation and translate the consequences from marketing performance into financial performance. Second, marketing metrics provide forward-looking information that is interesting for investors and stock analysts (Wiesel, Skiera, & Villanueva, 2008). From the investors’ perspective, information that assists in decision making by explaining trends and factors impacting the development, performance and business position of a firm should be included in financial reporting (Wiesel, et al., 2008). Third, the development of the marketing finance interface should lead to aligning marketing goals (related to increasing customer value) with corporate goals (related to increasing shareholder value). The discrepancies between marketing and corporate goals are reflected by the short-term focus on assessing management strategy based on immediate financial outcomes and the failure of recognizing the long-term returns on investment in marketing. According to Srinivasan & Hanssens (2009) marketing has to convince the management and shareholders to adapt an investment perspective on its spending, comparable to the practices with respect to research and development (R&D).

In recent years, considerable marketing research has been devoted on the link between brand equity and brand-related intangible assets and firm value. Brands are viewed as a strong marketing asset that generates future cash flows (Rao, Agarwal, & Dahlhoff, 2004). The discussion whether brand equity should be included in financial reporting has been brought up already in the past (Clement, Barth, Kasznik, & Foster, 1998) and there is a good potential for interdisciplinary research to prove the relevance of the marketing metrics in finance (Wiesel, Kräussl, & Srivastava, 2010). Madden, Fehle & Fournier (2006) find out that strong brands (according to the brand classification of Interbrand) deliver greater stock returns, moreover with a lower risk. In line with that, the results of a study by (Rego, et al., 2009) show that brand equity has a negative relationship with upside and downside systematic and unsystematic firm risk. According to Srinivasan & Hanssens (2009) it can be generalized on the basis of prior findings that improvements in brand equity have a significant, positive impact on firm valuation.

(5)

metrics, where different linkages exist between those constructs. The unobservable metrics are related to customer perceptions, attitude and intentions, while the observable metrics involve actual behaviors of customers, such as customer attrition.

Customer satisfaction is the unobservable metrics which impact on financial performance has been researched profoundly in recent years. Gupta & Zeitahml (2006) have formulated the marketing generalization that improvement in customer satisfaction has a significant and positive impact on the firms' financial performance, where the strength of the link varies across industries and across companies within an industry. Different metrics for measuring financial performance have been employed such as: profit, stock prices, Tobin’s Q (the ratio of market value of the firm to the replacement cost of its tangible assets) (Anderson, 2004; Gruca & Rego, 2005), return on investment (Anderson & Mittal, 2000), market value of the equity (Fornell, Mithas, Morgeson Iii, & Krishnan, 2006), abnormal returns (Aksoy, et al., 2008; Gruca & Rego, 2005; Jacobson & Mizik, 2009; O'Sullivan, et al., 2009), cash flow (Gruca & Rego, 2005). In addition, Tuli & Bharadwaj (2008) find out that there is a significant negative effect of customer satisfaction on the firm’s systematic and idiosyncratic risk.

The present research adds to the studies on the link between observed customer metrics and firm value. The need for contributing to this stream of research is motivated by the notion that a financial approach to valuation helps to make marketing more financially relevant and accountable (Srinivasan & Hanssens, 2009).Gupta et al. (2006) also point out that the growing interest in this concept is due to the increased pressure on companies to make marketing accountable, whereas traditional marketing metrics such as market share have failed to show return on marketing investment. For example, in some cases improved sales or increased market share may harm the long run profitability as Hanssens & Yoo (2005) find from their study on the luxury automobile market.

A number of studies find that marketing decisions based on improving observed customer metrics are positively linked to the firm's financial performance (Gupta & Zeitahml, 2006). Such a metric is customer lifetime value (CLV), which measures customer value taking into account the expected future cash flows per customer, discounted by a discount factor and the rate of customer retention. The marketing asset customer equity (CE) stands for the value of the firm’s customer base, which according to the most basic model is the product of CLV and the number of customers. Therefore, churn rate and acquisition rate are the observable metrics that are direct inputs for calculating customer equity. Churn rate measures customer attrition as the percentage of lost customers from the average customer base over a specified reporting period (i.e. monthly churn rate). Acquisition rate is a measure of the gained new customers as a percentage of the average customer base over a specified reporting period.

(6)

several billion dollars, which is in line with the findings of Gupta, Stuart & Lehmann (2004) that the value based on customers can be a strong determinant of firm value. According to Bonacchi, Kolev, & Lev (2010) companies employing subscription models, such as the wireless companies, must focus on strategies to acquire customers at the lowest possible cost, to increase the monthly average profit margin per user, and to retain existing customers. Further, Bonacchi, et al. (2010) recognize that the economics of subscription models are driven by four key factors: average revenue per user (ARPU), cost per acquisition, cost of service, and churn. The question is whether in practice it is possible to focus on improving all factors, i.e. maximizing total customer equity. As Villanueva & Hanssens (2007) point out, it is unclear whether firms competing for the same pool of customers will be better-off by maximizing CE. Indeed, CE maximization may not the best strategy for all firms, but a firm’s performance can be enhanced by focusing on an achievable strategy matching the firm’s resources and market position.

RESEARCH OBJECTIVES

Blattberg, Getz, & Thomas (2001) point out that customer equity needs to be measured, managed and maximized in order to optimize firm performance. However, there is lack of research studying the effect of CE strategy on firm value in the long run, where there may be trade-offs between focusing only on the profitable customers and reducing the customer base or expanding the customer base to reduce firm risk. This is stated by Srinivasan & Hanssens (2009) as an area in need of further research.

The first goal of this research is to find out what is the association between customer equity and Tobin’s Q. Further, in relation to the research direction suggested by Srinivasan & Hanssens (2009), the main contribution of the present research is that a customer equity strategy classification is developed to study further the relationship between customer equity and firm value, and customer equity and a firm’s customer base. The reasoning behind the CE strategy classification is that firms exhibit different CE strategies that can be inferred by the changes in their customer metrics over a specified time window. For example, if ARPU and acquisition rate have increased and churn has decreased, compared to the previous year levels for the same quarter, it can be inferred that the firm is maximizing its customer equity. Of course, if transactional data for individual customers are available, it is possible to value each customer separately and directly check the firm’s strategy (i.e. if a firm is focusing on the profitable customers or on expanding its customer base). Therefore, the proposed classification is most useful when only aggregated customer metrics are available, as in the dataset used for this study. The following research questions have been formulated:

- How is the marketing asset customer equity related to firm value (Tobin’s Q)? - What methodology to use for developing a CE strategy classification using as an

input the components of customer equity (ARPU, churn and acquisition rate)? - What is the relationship between the CE strategies from the developed

classification and firm value (Tobin’s Q)?

(7)

CONCEPTUAL FRAMEWORK

Customer Equity, Firm Value and Customer Base

In their survey of methods and metrics pertaining to marketing and firm value Srinivasan & Hanssens (2009) point out that customer equity and market valuation are intrinsically related because they are both using the principle of the present value of a stream of expected future cash flows. For example, Gupta, et al. (2004) find in their study of five internet companies that for three of them customer equity moves in parallel with the market value. Moreover, the findings show that the other two companies from the reseacrh are potentially mispriced. Gupta, et al. (2004) conclude that improvements in customer equity are significantly related to firm value.

On the other hand, Srinivasan & Hanssens (2009) point out that maximizing customer equity can imply a narrowing of the customer base, because the firm focuses on retaining only the most profitable customers. As a consequence, in the longer run the firm’s risk may be increased and firm value may also be negatively affected. In relation to that, there is an important distinction pointed out by Dreze & Bonfrer (2009), namely that CLV maximization and CE maximization are two different strategies for customer equity management, where the transition from CLV maximization to CE maximization is not straightforward. Their major finding is that a firm following a CLV maximization strategy may end up with a smaller and less profitable customer base than one that follows a CE maximization strategy. This is because the CE maximization strategy focuses not only on increasing the average revenue per customer, but also on expanding the customer base. In line with the idea of Srinivasan & Hanssens (2009) about the reverse relation between customer base and firm risk in the long run, Dreze & Bonfrer (2009) find that CLV maximization strategy is suitable only for companies with a short or ending lifespan. Otherwise, their findings show that CE maximization is the most profitable strategy.

Of course, it can also be argued that CLV maximization, when referring to the individual customers’ CLV, is the preferred strategy for managing the existing customer base, profiling potential profitable customers and acquiring them. In relation to that, Ryals & Knox (2005) propose the use of risk-adjusted CLV, which is termed the economic value (EV) of a customer, to assess both customer profitability and shareholder value gains. The practical implications from this research are that selective customer retention through CLV analysis and a risk-adjustment process may be the means for developing successful relationship marketing strategies to increase firm value. This analysis pertains to the business-to-business relationships in the financial service industry where individual data on each customer can be used for valuing CLV. However, the distinction that Dreze & Bonfrer (2009) make between CLV maximization and CE maximization is based on valuing the customer base by using average aggregate customer metrics, when individual customer data is not available, which is also the case in the present research.

(8)

Since no individual customer data is available, the eight customer equity strategies are inferred from the aggregated customer metrics data, specifically the direction of change in churn rate, acquisition rate and average revenue per user.

Firm value represents the collective discounted future cash flows to the firm’s equity investors and bondholders (Rao, et al., 2004), where the risk of cash flows is negatively related to firm value. According to Srivastava, Shervani, & Fahey (1998) the cash flows and their risk are affected in part by the management of market-based assets. Therefore, the reasoning behind the conceptual model is that CE strategy affects firm value due to the impact of the CE strategy on the future cash flows and their risk.

Two of the customer metrics used to define the strategies - ARPU and churn rate are components of CLV, whereas the third customer metric - acquisition rate is related to the CE calculation which includes current and newly added customers. Hence, it can be assumed that CLV is a proxy for the future cash flows and CE - for the cash flows and their risk, since the latter depends on the customer base and its change.

Considering the importance of the customer base for the firm value, the direct relationship between the customer equity strategies and the customer base is also of interest. Checking this relationship will show whether the best strategies with respect to firm value are also most beneficial for the number of subscribers. It is interesting to compare the significance and direction of the relationships between customer equity strategy and firm value and customer equity strategy and customer base. In case the directions of those relationships are the same, this will also confirm the assumption of Srinivasan & Hanssens (2009) that firm risk is related to the size of the customer base, where a smaller customer base can result in higher risk which in the long run can have a negative effect on the value of a firm. However, in this model the issue of endogeneity is expected because customer strategy is not exogenous to customer base, since it is derived from churn rate and acquisition rate, which are directly related to the customer base.

(9)

There are three types of control variables considered to influence the studied relationships. The industry variables include country penetration rate and industry concentration Herfindahl-Hirschmann index (HHI); the firm financial variables include firm leverage and EBITDA (earnings before interest, taxes and depreciation) margin; the firm other variables include firm market share and age. The financial control variables are not expected to be associated with customer base so they are not included in this model.

Measure of Firm Value

The choice of Tobin’s Q as a measure for firm value is due to its forward looking nature - it provides market-based views of investor expectations of the firms future profit potential (Rao, et al., 2004). Rao, et al. (2004) use in their research of the relationship of branding strategy to firm value the Tobin’s Q simplified formula proposed by Chung & Pruitt (1994). The same formula is adopted for the present research.

Tobin’s Q = (MVE + PS + DEBT)/ TA (1)

where:

Market value of the equity (MVE) = share price * number of common stock outstanding

Preferred shares (PS) = the value of the preferred shares of the firm

Debt = (current liabilities – current assets) + book value of the long term debt Total assets (TA) = book value of total assets

Customer Equity Model and Hypotheses

Kumar & George (2007) discuss the best way to measure the value of a customer asset. According to them the value of a customer asset can be computed in a similar manner as the value of a financial asset – this involves calculating the associated cash flow and then applying a discount factor to arrive at the present value of the asset. A firm can either calculate the total worth of its customer base from aggregate financial measures or compute the value of each customer individually based on buying characteristics and purchasing history (Kumar & George, 2007).

In the present research, an aggregate level approach to calculating CLV is adopted because data for calculating the individual CLV of wireless customers are not available. The wireless dataset contains no data on marketing expenditures and therefore the model of Gupta & Lehmann (2003) for calculating CLV using publicly available firm level data, retention rate and average contribution margin is most suitable.

) 1 ( i r r m CLV    (2) where:

(10)

This model assumes an infinite projections period and therefore avoids overestimation of the CLV. The other assumption is that the profit margin and retention rate are constant over time. Combining the customer metrics with the appropriate discount rate (which according to Gupta & Lehmann (2003) can be between 8% and 16%) provides a calculation of the net present value of an average customer’s cash flows. This is the CLV of an average customer before marketing expenditures, which when multiplied by the number of existing and newly added customers will provide the current customer equity and the future customer equity, respectively. Therefore with the available data, the customer equity can be decomposed according to kinds of customers (retained and new) but only with one value component. The limitation is that in the absence of marketing expenditures data, the customer equity cannot be further decomposed according to all its three value components: net present value of customer cash flows, retention expenditures, and acquisition expenditures.

According to Hogan et al. (2002) and Kumar & Shah (2009) the total customer equity is the sum of the lifetime value of the existing and new customers of the firm.

For the purpose of the present research we will look at the two separate components of customer equity given in the following two equations, issuing (2) as an input.

CEretained = CLV*retained customers (3)

CEadded = CLV*newly acquired customers (4)

In line with the results of Bonacchi, et al.(2010), who document a positive and significant association between market value of the equity and CE of the current customers (when controlling for the book value of the equity and net income), the following hypothesis are formulated:

H1: Customer equity of the retained customers is positively associated with firm Tobin’s Q

H2: Customer equity of the newly added customers is positively associated with Tobin’s Q

Customer Equity Strategies and Hypotheses

(11)

trend of churn and decreasing trend of acquisition (see Appendices: Figure 2). The trends can be explained by the fact that in 1999 the wireless industry was still in an early phase of its development and the country penetration rates were very low. As a lot of new customers were gradually entering the market during the last 10 years, acquisition rates were rising. At the same time, with the appearance of more players on the market churn rates were gradually increasing as well, due to increased competition.

Table 1 provides an overview of the eight strategies that have been formulated.

Table 1 Customer Equity Strategies Overview

Customer Equity Strategy Dummies

ARPU Churn Acquisition Interpretation Hypothesis w.r.t. Tobin’Q Hypothesis w.r.t. number of Subscribers 1 Low revenue & Customer

retention

(ARPU↓ Churn↓ Acquisition↓) price decrease for current customers

- +/-

2 Customer base expansion focus

(ARPU↓ Churn↓ Acquisition↑) price decrease for all customers

- +

3

CLV maximization

(ARPU↑ Churn↓ Acquisition↓) focus on short-term profitability

+ +/-

4

CE maximization

(ARPU↑ Churn↓ Acquisition↑) focus on long-term profitability

+ +

5 Opposite to CE maximization

(ARPU↓ Churn↑ Acquisition↓) failure to execute a strategy

- -

6 Low revenue & Customer acquisition

(ARPU↓ Churn↑ Acquisition↑) price decrease for new customers

- +/-

7 High revenue customers focus

(ARPU↑ Churn↑ Acquisition↓) removing low revenue customers

+ -

8 High revenue & Customer acquisition

(ARPU↑ Churn↑ Acquisition↑) removing low revenue customers/ attracting high revenue customers

+ +/-

Further, the possible advantages and disadvantages of the customer equity strategies are explained and the related hypotheses are stated. With respect to the model examining the relationship between customer equity strategy and number of subscribers, a hypothesis cannot be stated for every strategy. For strategies 1, 3, 6 and 8 there is no clear cut expectation about the association with number of subscribers because it depends on the levels of change in churn and acquisition which are not reflected in the proposed classification. Nevertheless, the customer base model tests those associations and the results may be interesting for the analysis.

Low revenue & customer retention: this strategy is characterized by a decrease in ARPU,

churn and acquisition rate. It can be inferred that the firm decreases prices to keep its customers and that the price decrease is not directed towards acquiring new customers, since the acquisition rate decreases. An advantage of this strategy is low firm risk because the customer base is retained. However, the disadvantage of this strategy is that it leads to a decreased revenue which has a negative effect on CLV. According to Reinartz & Kumar (2000) the revenue drives CLV and not the duration of a customer’s tenure. Further, Reinartz & Kumar (2002) find only a weak to moderate correlation between customer's tenure and profitability across four data sets. Moreover, when also acquisition is decreased the customer equity of future customers decreases, which should have a negative impact on the total customer equity. This leads to the following hypothesis:

(12)

Customer base expansion focus: this strategy is characterized by a decrease in churn rate

and APRU and an increase in acquisition rate. It can be inferred that the firm decreases prices for all customers in order to expand its customer base. On one hand, a decrease in churn and an increase in acquisition have a positive impact on the total customer equity. Therefore, the advantage of this strategy is that it leads to an expansion of the customer base, which may decrease risk in the long run and increase firm value as a result. However, as with the previous strategy, the major disadvantage of this strategy is the decrease in revenue which has a negative impact on CLV. Further, price decrease is a bad sign for investors, which is likely to result in a decreased firm value.

H4: Strategy 2 “customer base expansion focus” is negatively associated with Tobin’s Q

H5: Strategy 2 “customer base expansion focus” is positively associated with number of subscribers

CLV maximization: this strategy is characterized by an increase in ARPU and a decrease

in churn rate and acquisition rate. It can be reasoned that this strategy is short-term oriented because since acquisition rate is decreased it is not directed towards maximization of the customer equity. Dreze & Bonfrer (2009) argue that CLV maximization is a profitable strategy in the short-run but not preferable for the long run. However, a positive association with Tobin’s Q is expected because increase in revenue is a strong driver of profitability. Also, this strategy has a negative effect only on the customer equity of future customers but still a positive effect on the customer equity of current customers.

H6: Strategy 3 “CLV maximization” is positively associated with Tobin’s Q CE maximization: this strategy is characterized by an increase in ARPU, a decrease in

churn and an increase in acquisition rate. According to Dreze & Bonfrer (2009) this is the best strategy in the long run. It is better than CLV maximization because, besides directed towards revenue increase, it is a long-term oriented strategy that reduces risk through expansion of the customer base.

H7: Strategy 4 “CE maximization” is positively associated with Tobin’s Q

H8: Strategy 4 “CE maximization” is positively associated with the number of subscribers

Opposite to CE maximization: no clear strategy can be inferred from a decrease in ARPU,

an increase in churn and a decrease in acquisition. It is unlikely that a firm would intentionally follow a strategy that is clearly decreasing its value, because both the customer base and the revenue decrease. Therefore this case most likely results from a failure to execute a firm’s strategy.

H9: Strategy 5 “opposite to CE maximization” is negatively associates with Tobin’s Q

H10: Strategy 5 “opposite to CE maximization” is negatively associates with number of subscribers

Low revenue & Customer acquisition: this strategy is characterized by a decrease in

(13)

is focused on customer acquisition and therefore decreases prices for the new customers. This strategy involves higher risk because the newly acquired customers attracted by a price reduction are more price-sensitive and have a lower probability to be retained in the future. They usually exhibit frequent switching behavior because their priority is the lowest price and therefore it is difficult to built loyalty in this group. In line with this, Lewis (2005) found that for new customers, price sensitivity increases with time lapsed, whereas for current customers, it decreases with time.

H11: Strategy 6 “low revenue & customer acquisition” is negatively associated with Tobin’s Q

High revenue focus: this strategy is characterized by an increase in ARPU, an increase in

churn and a decrease in acquisition. It can be inferred that the firm tries to keep its high revenue customers and get rid of its low revenue customers. This strategy leads to a reduction of the customer base which is likely to increase the firm risk in the long run (Srinivasan & Hanssens, 2009). However, due to the clear focus on high revenue, short term profitability will be increased and it can be expected that firm value will also increase. Especially firms that incur high retention costs from unprofitable customers will benefit from this strategy.

H12: Strategy 7 “high revenue customer focus” is positively associated with Tobin’s Q

H13: Strategy 7 “high revenue customer focus” is negatively associated with number of subscribers

High revenue & Customer acquisition: this strategy is characterized by an increase in

ARPU, churn and acquisition. It can be inferred that the firm aims to attract high revenue customers and get rid of low revenue customers. This strategy, as the previous strategy, is directed towards increased profitability and selection of a more profitable customer base. Therefore it is likely that the firm value will be increased.

H14: Strategy 8 “high revenue & customer acquisition” is positively associated with Tobin’s Q.

Control variables

Herfindahl - Hirschmann index

The HHI is included in both - the firm value and customer base models, in order to capture some effects of competition. The HHI is an industry control variable measured as the sum of squares of all the market shares of the firms in a particular market. According to Curry & George (1983), it is the best measure of market concentration. Low HHI shows high competitive intensity while high HHI shows low competitive intensity. For the parent companies operating in different countries, the concentration index of the market of the subsidiary with the greatest EBITDA is taken.

(14)

the HHI in the Tobin’s Q models. The expected sign of the HHI in the customer base model is negative because higher concentration stands for high market power of one firm and therefore less growth opportunity.

Market Penetration

The market penetration rate is used as an industry control variable, where high concentration rate is characteristic for more mature, developed markets. As with HHI, the market penetration rate of the mother company is the penetration rate of its subsidiary that has the greatest EBITDA. There is no clear cut expectation for the sign of penetration rate in the Tobin’s Q models. The expected sign of penetration in the Subscribers model is negative because the customer base growth potential is less in markets where not many new customers enter the market and customers have to be taken from competition.

Leverage

The firm financial control variable leverage is the ratio of the long-term debt to the total assets of the firm. The expected sign of leverage in the Tobin’s Q models is positive because firms with higher leverage have a tax benefit meaning that they can deduct their interest costs, which results in greater cash flow (Rao, et al., 2004).

EBITDA Margin

The EBITDA margin is the ratio between the firm total EBITDA and total revenue. The expected sign of EBITDA margin in the Tobin’s Q models is positive because it triggers expectations among the investors for higher earnings (Rao, et al., 2004).

Market Share

Market share is a firm control variable which is expected to have a positive effect in the Tobin’s Q models because it stands for market power. Also, the expected sign of market share in the customer base model is positive because higher market share corresponds to a higher number of subscribers.

Company age

The longer a company has been public the more information investors should have, which should lead to valuing it closer to its true potential (Rao, et al., 2004). It can be expected that age may have a negative effect on Tobin’s Q because newer companies in the technological sectors may be considered by the investors as having greater growth potential. In the customer base model a positive effect of company age is expected because older companies usually have accumulated greater number of subscribers.

DATASET

A panel dataset containing data from 133 wireless telecommunication firms from 46 countries was constructed for the quantitative analysis. The data covers 39 quarters from the first quarter of 1999 to the third quarter of 2008.

(15)

in order to match the quarterly wireless firms’ data with the correct quarterly stock prices data. The financial market data were obtained from the financial data source Compustat, provided by Wharton Research Data Services.

The firms used for the analysis are the firms directly listed on the stock exchange and the firms that have one or two parents listed on the stock exchange. Firms that have multiple parents are not suitable for the purpose of the research and are excluded from the dataset because it is difficult to find a reliable source for the ownership structure, which also may change throughout the reporting period. Therefore it is difficult to trace the impact of such daughter firms on the financial performance of the parent firms.

Table 2 gives an overview of the companies included in the wireless dataset.

Table 2 Wireless Dataset Companies

Subsidiaries with parent/s listed 122

Directly listed companies 11

Parent companies 33

Total publicly listed companies 44

Countries of operation 46

The total publicly listed companies for which it was possible to obtain financial data from Compustat are 44. From those 44 companies, 33 are parent companies of 122 subsidiaries in total. It is very important to point out here that these 122 subsidiaries are represented at the stock exchange by their parent companies, which means that only parent companies have stock prices data. Since stock prices are used to measure firm value (in Tobin’s Q calculation), it is necessary to aggregate all the subsidiary data to the parent company level. Therefore, the data from the 122 subsidiaries belonging to the 33 parent companies is aggregated in order to calculate the quarterly data for each parent company. For example, the churn rate per parent company is calculated from the total sum of quarterly subscribers and the total sum of lost subscribers of all the subsidiaries belonging to that parent company. When a subsidiary is owned by two listed companies the subsidiary data is aggregated to the parent companies data, weighted by the percentage of ownership. In cases of missing data for a subsidiary, the subsidiary is omitted from the total calculations for the parent company for the quarter for which the data are not available. Also, not all of the wireless firms have existed throughout the whole period of reporting and therefore the data points per reporting period vary. Because of this and cases of missing data that was not available from the data sources used, the dataset is unbalanced. Further, when constructing the dataset it was important to keep track of the firms that have changed parents but should remain with the same ID. When wireless firms merge, a new firm with a new firm ID is entered in the dataset and the former two are discontinued. In cases when only the brand name of a firm changes but the stock listing is unchanged, the firm ID in the dataset remains unchanged.

(16)

The subscribers’ data in the wireless dataset include number of customers per quarter, net subscriber additions per quarter and the average monthly churn rate. This makes it possible to calculate the average monthly acquisition rate. The market data include firm market share, penetration rate of the wireless industry per country and the Herfindahl-Hirschman index, which is an indicator of the competitive intensity on the wireless market of each country. Firm financial data include monthly ARPU, quarterly EBITDA and EBITDA margin. In the wireless dataset all the data provided as monthly average data are multiplied by three to convert it to quarterly data. Also, for some countries the reporting was in local currency and therefore the financial data are converted to USD using the quarterly average exchange rate of the European central bank. Besides quarterly stock prices, additional financial data necessary for the analysis are obtained from Compustat: firms’ current and total assets, short-term and long-term debt, number of outstanding shares, preferred shares value, date of IPO (initial public offering), and country of headquarters.

An important clarification with respect to the data is that the wireless penetration rate, HHI and market share pertain to the market/country of operation of each subsidiary. Parent companies operate in a number of countries via their subsidiaries but it is not possible to aggregate these variables to parent company level because they are market/country specific. In order to use these variables in the analysis we have decided to take the quarterly data from the subsidiary with the greatest quarterly EBITDA, because we assume it has the highest weight with respect to the firm value of the parent company.

EMPIRICAL ANALYSIS

Customer equity and Tobin’s Q

The relationship between customer equity and Tobin’s Q is estimated using a panel data OLS regression model that controls for the firm-specific and industry variables listed in the previous section. Since customer equity is decomposed into retained and added customers, two models are formulated to study the association between each customer equity type and Tobin’s Q.

The specification of Model 1 is as follows:

Tobin’sQit = β0+ β1it CEretained + β2itEBITDA_Margin + β 3itPenetration + (5)

+ β4itHHI + β5itMarketShare + β6it Leverage+ β7itCompanyAge + ε

where: i = 1,...,N (firms) and t = 1,…,T (quarters);

CEretained is the estimate of the value of the retained customers’ equity which effect on Tobin’s Q is measured by the coefficient β1it. The coefficients β2it to β7it measure the effects of the control variables on Tobin’s Q.

The specification of Model 2 is as follows:

Tobin’sQit = β0 + β1it CEadded + β2itEBITDA_Margin + β 3itPenetration + (6)

(17)

CEadded is the estimate of the value of the newly acquired customers’ equity which effect on Tobin’s Q is measured by the coefficient β1it. The coefficients β2it to β7it measure the effects of the control variables on Tobin’s Q.

Logarithmic Transformation

After looking into the descriptive statistics (see Appendices, Table 9) and making residuals diagnostics, it was concluded that it is better to transform the dependent variable. First, the mean of Tobin’s Q in the whole sample is 1.40 and the median is 1.15. This means that the distribution of Tobin’s Q is positively skewed to the right, which can also be seen from the histogram in the Appendices (Figure 3 Tobin’s Q histogram). The plots of the regression standardized residuals to the predicted value of both Model 1 and Model 2 show that there is heteroscedasticity, because there is an increasing variance of the error terms (see Appendices - Figure 5 & Figure 12). This means that the variance of the errors is not the same across all levels of Tobin’s Q. According to Berry & Feldman (1985) and Tabachnick & Fidell (2001) slight heteroscedasticity has little effect on the significance tests, but when heteroscedasticity is marked it can lead to distortion of the findings and seriously weaken the analysis, thus increasing the possibility of a Type I error. A common way of stabilizing the variance is to apply a logarithmic transformation which in this case will be applied to the dependent variable Tobin’s Q.

After taking the natural logarithm of Tobin’s Q a symmetric distribution is obtained – the skewness decreases from 5.25 to 0.45 (see Appendices - Table 9).

The second specification of Model 1 is:

lnTobin’sQit = β0+ β1it CEretained + β2itEBITDA_Margin + β 3itPenetration + (7)

CEretained is the estimate of the value of the retained customers’ equity which effect on Tobin’s Q is measured by the coefficient β1it. The coefficients β2it to β7it measure the effects of the control variables on Tobin’s Q. Due to the logarithmic transformation of the dependent variable every unit change in the independent variables is expected to multiply the original dependent variable by e on the power of the estimated β coefficient.

The second specification of Model 2 is:

lnTobin’sQit = β0 + β1it CEadded + β2itEBITDA_Margin + β 3itPenetration + (8)

(18)

lnTobin’s Q does not have the pattern that was seen before the transformation of Tobin’s Q (see Appendices - Figure 8 & Figure 15).

Next, to check the error terms for normality, the histogram of the residuals and a normal probability plot of the residuals are studied for both models (see Appendices - Figure 7 & Figure 14). In both cases the normal probability plot is nearly linear, indicating that the error terms are normally distributed, which can also be seen from the histograms (Appendices -Figure 6 & Figure 13).

First Differences Specifications

Further, it is important to address the issue of autocorrelation. The Durbin-Watson statistics of Model 1 is d1 = 0.414 and of Model 2 is d2 = 0.422. Small positive values of d (d < 1) indicate that successive error terms are, on average, close in value to one another, or positively correlated (Leeflang, Wittink, Wedel, & Naert, 2000). This is a typical problem for time series OLS models because the subsequent measures in time belonging to the same firm are not independently and identically distributed. For both models the lnTobin’s Q lag plot of the residuals indicates autocorrelation of the measurements, i.e. there is a time trend in the data (see Appendices - Figure 9 & Figure 16).

Mizik & Jacobson (2009) point out that a number of marketing and other studies across disciplines attempt to assess the financial market implications of marketing variables by estimating levels models that link a highly autocorrelated financial performance metric (i.e. market value, Tobin’s Q, market-to-book ratio) to explanatory factors that are also autocorrelated (i.e. book value, earnings or intangible asset measures, such as customer satisfaction or a brand attribute). The problem is that when both the error term and the independent variable are positively autocorrelated, least squares estimates of the standard errors are biased and understate the true standard errors. The conventional t-statistic does not have a standard normal limiting distribution, which invalidates the use of t-distribution to test the hypothesis of statistical significance (Mizik & Jacobson, 2009). In order to address this issue a first differences specification of the models is made to remove time-constant company specific effects (Landsman & Magliolo, 1988).

The third specification of Model 1 taking first differences is:

ΔlnTobin’sQit = β0 + β1it ΔCEretained + β2itΔEBITDA_Margin + (9)

+ β 3itΔPenetration + β4iΔHHI+ β5itΔMarketShare +

+ β6it ΔLeverage+ β 7itΔCompanyAge + ε

where: i=1,...,N (firms) and t=1,…,T (quarters);

ΔlnTobin’sQ is the quarterly change in Tobin’s Q; ΔCEretained is the quarterly change in CEretained; the Δ control variables are the quarterly changes in the control variables.

The third specification of Model 2 taking first differences is:

ΔlnTobin’sQit = β0 + β1it ΔCEadded + β2itΔEBITDA_Margin + (10)

+ β 3itΔPenetration + β4iΔHHI + β5itΔMarketShare +

+ β6it ΔLeverage + β 7itΔCompanyAge + ε

(19)

ΔlnTobin’sQ is the quarterly change in Tobin’s Q; ΔCEadded is the quarterly change in CEadded; the Δ control variables are the quarterly changes in the control variables.

As a result of the first difference specifications the Durbin-Watson statistics increased to d1 = 2.11 for Model 1 and d2 = 2.12 for Model 2. From the residuals plots it can also be seen that there is an improvement with respect to autocorrelation (see Appendices - Figure 10 & Figure 17). The lag plots of the first differences specifications for both models show that the time trend is removed (Appendices - Figure 11 & Figure 18).

In Table 3 the estimates from the three specifications of Model 1 are presented.

Table 3 Estimates of the Three Specifications of Model 1

Expected Sign of Coefficient M 1A Tobin's Q M 1B ln Tobin's Q M1C ln Tobin's Q first dif

Constant .730** -.276* 2.030

(3.614) (-2.319) (1.517)

Customer Equity retained customers + .062 -.026 .004

(1.070) (-.465) (.053) Control variables: Market share + .140* .185** .021 (2.501) (3.374) (.312) HHI +/- .034 .038 .002 (.604) (.695) (.032) Penetration + .090 .142 .118 (1.695) (2.721) (1.788) Leverage + .201** .210** .136* (3.979) (4.228) (2.067) EBITDA Margin + .184** .192** -.110 (3.716) (3.944) (-1.569) Company age - -.143* -.102 -.103 (-2.306) (-1.672) (-1.532) ( t-values in brackets) Adjusted R square .107 .135 .031 ** p value<.001; * p value <.05

The first hypothesis that there is a positive association between CEretained and Tobin’s Q cannot be supported because in the three model specifications the parameters are not significant. With respect to the control variables in M1A only industry penetration rate and HHI are not significant, all the other control variables are significant and with the expected signs. In M1B only market share, EBITDA margin and leverage are significant and with the expected signs. The only significant parameter in M1C is leverage which is positively associated with Tobin’s Q.

Further, the adjusted R2 of the M1B is the highest but still only .135. This means that the model specification can be improved through inclusion of more explanatory variables. It has to be pointed out that stock prices are explained to a great extent by their value in the previous period and therefore the R2 of the model can be significantly increased by the inclusion of the lag term of Tobin’s Q. However, this is not of interest here because we are studying the direction of the relationships.

(20)

Table 4 Estimates of the Three Specifications of Model 2 Expected Sign of Coefficient M 2A Tobin's Q M 2B ln Tobin's Q M2C ln Tobin's Q first dif Constant .763** -.261* 2.543 (3.895) (-2.213) (1.899) Customer Equity added customers + .265** .138* .132* (5.006) (2.580) (1.988) Control variables: Market share + .081 .138* .017 (1.499) (2.546) (.260) HHI +/- .051 .057 .002 (.956) (1.046) (.023) Penetration + .142** .178** .103 (2.723) (3.387) (1.565) Leverage + .195** .209** .134* (3.993) (4.244) (2.057) EBITDA margin + .163** .177** -.109 (3.391) (3.662) (-1.636) Company age - -.230** -.181** -.128* (-4.085) (-3.188) (-1.913) ( t-values in brackets) Adjusted R square .160 .150 .048 ** p value<.001 * p value <.05

The second hypothesis that there is a positive association between CEadded and Tobin’s Q is supported by the three model specifications.

With respect to the control variables in M2A only market share and HHI are not significant, the other control variables are significant and with the expected signs. In M2B only HHI is not significant, the other control variables are significant and with the expected signs. The only significant control variables in M2C are leverage, which is positively associated with Tobin’s Q, and age, which is negatively associated with Tobin’s Q.

M1A explains 16% of the variance in the dependent variable, M1B 15% and M2C only 5%. As already explained for the previous model, adding a Tobin’s Q lagged term will significantly improve the R2 but the aim here is to study the association with CEadded.

Customer Equity Strategy and Tobin’s Q

The third model specification is for estimating the effects of the eight customer equity strategies on Tobin’s. The model includes eight strategy dummies. The control variables are the same as in the previous two models. Again the logarithm of Tobin’s Q is applied.

lnTobin’sQit = β0 + β1itPenetration + β2itHHI + β3itMarketShare + (11)

+ β4it Leverage + β5it*EBITDA_Margin + β6it CompanyAge +

(21)

The coefficients β1it to β6it measure the effects of the control variables on lnTobin’s Q. The coefficients β7it to β14it measure the average impacts of customer equity strategy on firm value for the subset of companies that follow the same strategy, after accounting for the effects of the control variables.

As with Model 1 and Model 2, autocorrelation persists as a problem also in Model 3 (see Appendices - Figure 19). For the model with strategic dummies instead of taking the first differences using OLS the cluster robust standard errors method to correct the significance of the parameters is applied using GLS1.

Cluster-robust standard errors are obtained by relaxing the assumption of error independence and allowing for correlation within a “cluster” (i.e. observations for the same company but from different quarters) (Gow, Ormazabal, & Taylor, 2010). They are based on estimates of the residuals covariance within a cluster (instead of assuming zero correlation as in OLS). The use of robust standard errors does not change the estimates of the coefficient, but it affects the standard errors and therefore the significance of the coefficients.

The errors in model (11) have been clustered by time because it allows observations to be cross-sectionally correlated, but assumes independence over time. This method is chosen because it makes the model robust to time-series dependence, which is the source of autocorrelation in the model. Alternatively, clustering by company would allow observations to be serially correlated, but assumes independence across companies.

1

(22)

In Table 5 the estimates from the two specifications of Model 3 are presented.

Table 5 Estimates of the Two Specifications of Model 3

Expected sign of Coefficient M3 A lnTobin's Q M3 B lnTobin's Q robust errors Constant .466* .466* 2.011 1.922 Control variables: Market Share + .139* .139* (2.557) (2.893) HHI +/- .034 .034 (.596) (.699) Penetration + .050 .050 (.883) (.786) Leverage + .213** .213* (4.119) (2.776) EBITDA margin + .244** .244 (4.877) (1.664) Company Age - -.119* -.119* (-2.199) (-2.820) Strategy

dummies: Strategy characteristics

Low revenue &

Customer retention (ARPU↓ Churn↓ Acquisition↓) - .141* .141*

(2.071) (2.600)

Customer base

expansion focus (ARPU↓ Churn↓ Acquisition↑) - .170* .170*

(2.776) (2.009)

CLV maximization (ARPU↑ Churn↓ Acquisition↓) + .021 .021

(.316) .(353)

CE maximization (ARPU↑ Churn↓ Acquisition↑) + .036 .036

(.565) .(492)

Opposite to CE

maximization (ARPU↓ Churn↑ Acquisition↓) - .041 .041

(.571) .(625)

Low revenue & Customer acquisition

(ARPU↓ Churn↑ Acquisition↑) - .087 .087

(1.111) (1.675)

High revenue

customers focus (ARPU↑ Churn↑ Acquisition↓) + .201* .201*

(2.993) (2.242)

High revenue & Customer acquisition

(ARPU↑ Churn↑ Acquisition↑) + .104 .104

(1.362) (1.196) ( t-values in brackets) Adjusted R square .157 .157 ** p value<.001 * p value <.05

(23)

related to Model 3 only H12 is supported, namely that strategy 7 “high revenue customer focus” is positively associated with Tobin’s Q. H1 and H2 are rejected because instead of negative, their association with Tobin’s Q is positive. This means that “customer retention focus” and “customer base expansion focus” show a positive effect on firm value. With respect to model fit, the adjusted R2 is .16.

Customer Equity Strategy and Customer Base

The fourth model specification is for estimating the effects of the customer equity strategies on the customer base, measured by the number of active subscribers at the end of each quarter.

Subscribersit = β0 + β1itMarketShare + β2itHHI + β3itPenetration + (12)

+ β4itCompanyAge + β5it to β12it Customer strategy dummiesit +ε

where i = 1,...,N (firms) and t = 1,…,T (quarters).

(24)

In Table 6 the estimates from the two specifications of Model 4 are presented.

Table 6 Subscribers: Regression Analysis Estimates

Expected sign of Coefficient M4 A Subscribers M4 B Subscribers Error Robust Constant -24651.9* -24651.9* (-1.952) (-2.586) Control variables Penetration + -.252** -.252** (-6.805) (-7.258) HHI - -.033 -.033* (-.939) (-2.204) Market share + .285** .285** (8.460) (9.291) Company age + .482** .482** (13.889) (19.395)

CE Strategy dummies Strategy characteristics Low revenue & Customer

retention (ARPU↓ Churn↓ Acquisition↓) -/+ .083 .083

(1.476) (1.166)

Customer base expansion

focus (ARPU↓ Churn↓ Acquisition↑) + .024 .024

(.531) (.543)

CLV maximization (ARPU↑ Churn↓ Acquisition↓) -/+ .084 .084

(1.561) (1.354)

CE maximization (ARPU↑ Churn↓ Acquisition↑) + .064 .064

(1.377) (1.568)

Opposite to CE

maximization (ARPU↓ Churn↑ Acquisition↓) - .095 .095

(1.618) (1.877)

Low revenue & Customer

acquisition (ARPU↓ Churn↑ Acquisition↑) -/+ .141* .141*

(2.397) (1.961)

High revenue customers

focus (ARPU↑ Churn↑ Acquisition↓) - .070 .070

(1.390) (1.830)

High revenue & Customer

acquisition (ARPU↑ Churn↑ Acquisition↑) -/+ .093 .093

(1.681) (1.429)

( t-values in brackets)

Adjusted R square .275 .275

** p value<.001

* p value <.05

(25)

subscribers, as well as company age, as explained in the section about the control variables.

DISCUSSION

Major Findings

An unexpected outcome from the analysis is that hypothesis 1, stating that there is a positive association between the customer equity of retained customers and Tobin’s Q, was not supported. Based on previous theory, old customers pay higher prices and it is cheaper to retain a customer rather than to acquire a new one (Villanueva & Hanssens, 2007). Reichheld and Sasser Jr (1990) suggest that long-life customers are more profitable than first-time customers. They find that a 5% increase in customer retention could increase firm profitability from 25% to 85%. However, further research challenges this finding. Reinartz & Kumar (2000) demonstrate that long-life customers are not necessarily more profitable than first-time customers but their findings are in a non-contractual setting (catalogue customers). Bonacchi et al. (2010) reason that in a contractual setting a current customer is more likely to stay with the company through “good times and bad times”, whereas the new customer acquisition may be strongly impacted by macro and micro economic factors. Further, according to Lewis (2005) new customers are more price sensitive than old customers.

When put in the perspective of the wireless industry the reasoning supported from the literature is valid but perhaps there are other factors that moderate the relationship between the customer equity of retained customers and firm value. Since investors are influenced by forward looking information (Wiesel, et al., 2008) it is possible that they are more interested in valuing the future growth opportunities of the firm in terms of newly acquired customers. The retained customer base is already valued and it is likely that its customer equity has an effect on firm risk but not an effect on future growth. Therefore it is interesting to consider whether company age could be a moderator of the relationship between the customer equity of retained customers and firm value. This would mean that with firm age the association between the customer equity of retained customer and firm value may be decreasing.

Further, it is interesting to consider that contractual relationships have barriers to exit and sometimes high switching costs. Therefore, not all the retained customers stay as a result of a successful retention strategy and the ones that are trapped may simply reduce using the service to a minimum, which reduces profitability.

On the other hand, the support for hypothesis 2, stating that there is a positive association between the customer equity of newly added customers and Tobin’s Q, implies that there might be a signaling effect that investors see in acquisition rate, which can be directly related to stock prices. This means that acquisition rate may be a signal of the long-term growth outlook of the firm and therefore be evaluated by the investors. This effect can be measured by the unanticipated change in the acquisition rate or the changes relative to the competitors on the market, which is an interesting aspect to model further.

(26)

Table 7 Customer Equity Strategies: Hypotheses and Findings

Customer Equity Strategy Dummies

ARPU Churn Acquisition Hypothesis w.r.t. Tobin’Q Findings w.r.t. Tobin’Q Hypothesis w.r.t. number of Subscribers Findings w.r.t. number of Subscribers 1 Low revenue &

Customer retention

(ARPU↓ Churn↓ Acquisition↓) - + +/- NS

2 Customer base expansion focus

(ARPU↓ Churn↓ Acquisition↑) - + + NS

3

CLV maximization

(ARPU↑ Churn↓ Acquisition↓) + NS +/- NS

4

CE maximization

(ARPU↑ Churn↓ Acquisition↑) + NS + NS

5 Opposite to CE maximization

(ARPU↓ Churn↑ Acquisition↓) - NS - NS

6 Low revenue & Customer acquisition

(ARPU↓ Churn↑ Acquisition↑) - NS +/- +

7 High revenue customers focus

(ARPU↑ Churn↑ Acquisition↓) + + - NS

8 High revenue & Customer acquisition

(ARPU↑ Churn↑ Acquisition↑) + NS +/- NS

The outcomes of the third model relating customer equity strategies to Tobin’s Q show only three significant strategies: strategy 1 - “low revenue & customer retention”, strategy 2 - “customer base expansion focus” and strategy 7 - “high-revenue customers focus”. The size of the positive association is comparative across the three strategies. However, the size of the association is not interpreted due to the methodology of deriving the strategies, because we were interested in the direction of the strategy and not the magnitude of the change in the customer metrics. Nevertheless, the results can propose some insights that can be further studied with an improved methodology. Only one hypothesis has been supported (H12) - about the positive association between strategy 7 “high revenue customer focus” and Tobin’s Q. It is interesting to note that strategy 2 “customer base expansion focus” and strategy 7 “high-revenue customers focus” are two opposing strategies. However, they both have one common characteristic - one-sided focus. Strategy 2 focuses on expanding the customer base, while strategy 7 is most likely to lead to a reduction of the customer base but it focuses on maximizing the revenue per customer. Strategy 1 is also characterized by a strong focus - on retaining the existing customers.

(27)

Hanssens, 2009), time of entrance to the market, also being a first mover or a follower (Villanueva & Hanssens, 2007). This means that the present research can be improved by looking for interactions between the formulated strategies and additional firm specific and industry specific factors that would explain further the relationship to firm value. For example, in the wireless industry the prices have been generally decreasing since the introduction of the service with the increase of the country penetration rates. This means that if one of the CE optimization goals is to increase revenue per user, the wireless company may not be able to expand its customer base and compete for customers to comply with the second goal. Especially in the past several years, when the penetration rates have reached even above 100% in some countries, it is possible that a goal of increasing average revenue per user simply does not match the market conditions and the competitive environment. Since quality of the service is comparable between wireless providers they tend to compete primarily on prices and promotional packages. Therefore, it is very important to capture the effect of the competitors marketing efforts on customer attraction and retention when assessing the CE strategies. In order to capture some competitive effects the industry concentration HH index is used, but it is not significant in the model linking customer equity strategy to Tobin’s Q. It is possible that HHI interacts with the relationship between customer equity strategy and Tobin’s Q, which needs to be studied further.

To continue to the last studied model that relates CE strategy to the customer base, the only significant strategy is strategy 6 - “low revenue& customer acquisition”, which has a positive association with the number of subscribers. Besides an increase in acquisition rate, this strategy is also characterized with an increase in churn, which means that it is off-set by the increasing acquisition rate. It is rather difficult to interpret the outcomes of this model because the customer equity strategies do not reflect the magnitude of the change in churn and acquisition but only the direction. One implication of this strategy is that it is effective for gaining customers from competition, since especially in the developed markets there is no influx of new customers. Therefore, in markets with high penetration rate the only way to acquire new customers is to win them from competition. This means that HHI and penetration rate most probably interact with the relationship between customer equity strategy and number of subscribers. Again this is an area of further research.

Limitations

In this section some of the limitations of the present research are discussed. The first limitation is the lack of data on marketing expenditures which excludes the possibility for calculating customer equity of the retained customers and customer equity of the added customers with a more comprehensive CE equity model, i.e. as the model proposed by Kumar & Shah (2009). Therefore, two value elements are missing in the model used for this research - the net present value of the acquisition costs and the retention costs.