considering the self-selection effect

26
Effect of Online-Banking Usage on CLV Considering the Self-Selection Effect E-Finance Lab Jour Fixe, 03. May 2004 Dr. Sonja Gensler

Upload: others

Post on 03-Feb-2022

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Considering the Self-Selection Effect

Effect of Online-Banking Usage on CLVConsidering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004

Dr. Sonja Gensler

Page 2: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 2

Introduction• Value-Based Customer Management• Multi-Channel Management in Retailbanking

Evaluation of Multi-Channel PerformanceResults of Empirical StudyConclusions and Future Research

Outline of Presentation

Page 3: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 3

Financial institutions should derive benefits from operationalizing customer orientationCustomer focused performance measures are crucial for value-based customer managementCustomer Lifetime Value (CLV) measures the discounted profit streams of a customer across the entire customer life cycle

Value-based Customer ManagementTodays Challenges in Retailbanking

( )( )

'', , ' , '

' 0 ' 00

0

1 i I

1

+∞= =

=

⋅ ⋅ += − ∀ ∈

+

∏ ∏∑

t t tti t i t i t

t ti t

t

m r wCLV a

d

mi,t: margin of the i-th customer in the t-th period,ri,t: retention rate of the i-th customer in the t-th period,wi,t: growth rate of the i-th customer in the t-th period,d: discount rate,a0: acquisition cost.

Page 4: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 4

Multi-Channel Management constitutes the integrated management of channels with the aim to maximize a firm’s customer equityContrasting to a multi-channel strategy is a multiple channel strategy, where several sales channels exist independently from each otherSpecifics of Multi-Channel Management in retailbanking: retail banks own all sales channelsChallenges of Multi-Channel Management in retailbanking: value-based integrated management of all sales channels to enhance customer equity

Multi-Channel Management in RetailbankingProliferation of Multiple Sales Channels

Page 5: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 5

Value-Based Multi-Channel ManagementMission Statement of Cluster III

Customer Orientation

How to keep customers happy and loyal?It is crucial today for financial institutions to understand what the customers want and to provide it.

Satisfying Customer Needs

Value-BasedCustomer Management

Financial institutions are very interested in beingcustomer-centric, but the shackles of product-centricityare very difficult to break.

Financial institutions should derive benefits from operationalizing customer orientation.Customer oriented performance measures are crucial for value-based management.

Customer Management in a Multi-Channel Environment

"Optimizing Customer Equity through Multi-Channel Management." [Mission Statement of Clusters III of the EFL]

Page 6: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 6

Introduction• Value-Based Customer Management• Multi-Channel Management in Retailbanking

Evaluation of Multi-Channel PerformanceResults of Empirical StudyConclusions and Future Research

Outline of Presentation

Page 7: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 7

Evaluation of Multi-Channel PerformanceFindings from Practice

“Die Online-Kunden weisen um 40% höhere Zahlungseingänge und eine um 13% höhere Cross-Selling Quote auf.” [Postbank, 2003]“Multi-Channel Bankkunden sind um 25-50% profitabler als reine Filialkunden.”[Bachem, 2003]“Der Ertrag eines Online-Kunden pro Jahr ist um 50€ höher als der eines Offline-Kunden.” [LBBW, 2002]“… banks report that online-banking customers are more profitable, retain higher bank balances and are the highest household income bracket.” [Bernstel, 2003]

However, just comparing the mean is not sufficient to evaluatemulti-channel performance.

Page 8: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 8

Evaluation of Multi-Channel PerformanceCustomer-oriented Performance Measures

Customer Satisfaction

Revenue Contribution

Cost Contribution

Customer‘s Retention

Rate

Customer Lifetime Value

Customer‘s Profit Contribution

Page 9: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 9

Self-Selection Effect Online-banking usage identifies the

profitable customersOnline-banking customers have always been the more profitable customers

Channel effect and self-selection effect determine the profitability difference of online-banking and offline-banking customers.

Multi-channel usage changes customer behavior and customers become more profitable (e.g. increased cross-sellingrate)

Channel EffectOnline-banking usage introduces

behavioural change

Evaluation of Multi-Channel PerformanceDifferentiation between Self-Selection Effect and Channel Effect

Page 10: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 10

Evaluation of Multi-Channel PerformanceConsideration of Self-Selection Effect is necessary

Self-selection effect is present if online-banking customers have always been the more profitable customersChannel effect arises if online-banking customers are more profitable because they started to use the online channelDistinction between those two effects allows to evaluate the impact of the online channel on customer profitability

• If self-selection effect exists a simple comparison between the profitability of online and offline customers is misleading

• No adequate implications can be derived for value-based multi-channel management

Page 11: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 11

Evaluation of Multi-Channel PerformanceBasic Idea of Matching Method

One way to account for selection biases is the use of matching algorithmsMatching is a well-known method derived from the economics fieldMatching approach aims to build matched pairs of comparable individuals (twin building) from the group of online-banking customers and offline-banking customersBuilding comparable twins is ought to reduce any observable difference between online-banking and offline-banking customers Selection biases considered in the marketing literature

• Degeratu/Rangaswamy/Wu (2000), „Consumer choice behavior in online and traditionalsupermarkets:The effects of brand name, price, and other search attributes“, IJRM, 17, 55-78.

• Hitt, L./Frei, F. (2002), „Do Better Customers Utilize Electronic Distribution Channels? The Case of PC Banking“, Management Science, 48, 6, 732-748.

• Danaher/Wilson/Davis (2003), „A Comparison of Online and Offline Consumer Brand Loyalty“, Marketing Science, 22, 4 (Fall), 461-476.

• Busse/Silva-Risso/Zettelmeyer (2004), „$1000 Cash Back: Asymmetric Information in Auto Manufacturer Promotions“, Working Paper University of California, Berkeley.

Page 12: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 12

Evaluation of Multi-Channel PerformanceSelected Literature: Matching Method

Dehejia, R./Wahba, S. (2002), „Propensity Score-Matching Methods for Nonexperimental Causal Studies“, The Review of Economics and Statistics, 84, 1, 151-161.Dehejia, R./Wahba, S. (1999), „Causal Effects in Nonexperimental Studies: Reevaluating the Evaluation of Training Programs“, JASA, 94, 1053-1062. Hujer, J./Caliendo, M./Radic, D. (2003), „Methods and Limitations of Evaluation and Impact Research“,Working Paper, Faculty of Economics and Business Administration, Frankfurt University.Humphreys, K./Phibbs, C./Moos, R. (1996), „Addressing Self-Selection Effects in Evaluation of Mutual Help Groups and Professional Mental Health Services: An Introduction to Two-Stage Sample Selection Models“, Evaluation and Program Planning, 19, 4, 301-308.Lee, L. (2000), „Self-Selection“, in: Baltagi, B. (Ed.), „A Companion to Theoretical Econometrics“, Blackwell Publishers. Roy, A. (1951), „Some Thoughts on the Distribution of Earnings“, Oxford Economic Papers, 3, 2, 135-146. Rubin, D. (1974), „Estimating Causal Effects of Treatments in Randomized and Nonrandomized Studies“, Journal of Educational Psychology, 66, 688-701 Singer, B. (1986), „Self-Selection and Performance Based Ratings: A Case Study in Program Evaluation“, in: Wainer, H. (Ed.), „Drawing Inferences from Self-Selected Samples“, New York.

Page 13: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 13

Observable and relevant customer characteristics: age, ownership of stocks, incomeComparing the mean: 14 – 8,67 = 5,33Result: online-banking customers have a higher CLV than offline-banking customers

1

2

3

4

5

6

20

26

48

37

34

59

37,4

1

0

0

1

1

0

0,5

2

1

4

5

4

3

3,17

1

7

12

6

10

16

8,67

age stocks income CLV

20

34

45

33

1

1

0

0,67

3

4

6

4,3

age stocks income CLV

Online-banking customers

8

14

20

14∅

Offline-banking customers

Evaluation of Multi-Channel PerformanceMatching Method for Considering the Self-Selection Effect

Page 14: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 14

Only online-banking customer „2“ and offline-customer „5“ can be matchedWhen not considering the variable „income“ the number of matched pairsincreases from one to two (loss of information)Trade-Off: The more of the relevant variables are considered, the better thecontrol of observable selection bias, but the harder to find matching partners

Evaluation of Multi-Channel PerformanceMatching Method for Considering the Self-Selection Effect

1

2

3

4

5

6

20

26

48

37

34

59

1

0

0

1

1

0

2

1

4

5

4

3

1

7

12

6

10

16

age stocks income CLV

20

34

45

1

1

0

3

4

6

age stocks income CLV

Online-banking customers

8

14

20

Offline-banking customers

Page 15: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 15

Basic idea of matching by Balancing Scores• To overcome the problem of not finding identical twins Balancing Scores are

introduced• Compute a balancing score for every customer based on the relevant characteristics• A multi-dimensional problem is reduced to a one-dimensional problem

Propensity Score as Balancing Score• Propensity Score represents the probability of online-banking usage given the

observed customer characteristics (multinomial Logit or Probit model)• Propensity Score ranges from 0 to 1• Under ideal conditions corresponds the distribution of the propensity score in both

groups (online-banking and offline-banking customers)

Evaluation of Multi-Channel PerformanceSolving the Problem of Having Limited Number of Matched Twins

Page 16: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 16

Matching without replacement: every offline-customer can be used for matching just once

Matching with replacement: every offline-customer can be used several times for matching purposes

Trade-off: bias versus variance• Matching with replacement reduces the bias, but increases the variance of

estimates

Matching methods• Nearest-Neighbour Matching• Caliper and Radius Matching• Stratification/Interval Matching• Kernel Matching

Evaluation of Multi-Channel PerformanceAlternative Procedures for Matching

Page 17: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 17

Introduction• Value-Based Customer Management• Multi-Channel Management in Retailbanking

Evaluation of Multi-Channel PerformanceResults of Empirical StudyConclusions and Future Research

Outline of Presentation

Page 18: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 18

ResultsSimple mean comparison of online-banking and offline-banking customers

For confidentiality reasons offline-banking customers have been set to 100%.All differences are significant at the 5% level.

103.80%

518.48%

69.66%

76.87%

66.60%

139.02%

75.50%Duration of relationship

Number of products

Number of transactions

Value of current fondportfolio

Value of current stockportfolio

Balance of saving account

Balance of current account

Offline-banking customers Online-banking customers

Page 19: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 19

Matching by propensity score (logit model)• Dependent variable: Usage of online-banking• Independent variables: (duration of relationship), number of saving accounts, number

of current account, usage of call center, ownership of a credit, ownership of a credit card, lifecycle segment, sex

Matching without replacement

Imposition of common support: online-banking customers whose propensity score is higher than the maximum or less than the minimum propensity score of the offline-banking customers are dropped

5% trim level: imposes common support by dropping 5 percent of the online-banking customers at which the propensity score density of the offline-banking customers is the lowest

Approach results in about 90% per cent reduction in bias for every consideredindependent variable

ResultsDifferences between online-banking and offline-banking customers

Page 20: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 20

ResultsComparison of online-banking and offline-banking customers (matched)

For confidentiality reasons offline-banking customers have been set to 100%.

100.00%

94.84%

87.78%

93.30%

125.16%

89.19%Duration of relationship

Number of products

Number of transactions

Value of current fondportfolio

Value of current stockportfolio

Balance of saving account

Balance of current account

Offline-banking customer Online-banking customer

1183.77%

Page 21: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 21

ResultsSummary of preliminary results

For confidentiality reasons offline-banking customers have been set to 100%.

518.48%

139.02%

103.80%

69.66%

76.87%

66.60%

75.50%

125.16%

100.00%

94.84%

87.78%

93.30%

89.19%Duration of relationship

Number of products

Number of transactions

Value of current fondportfolio

Value of current stockportfolio

Balance of saving account

Balance of current account

Online-banking customer (mean) Offline-banking customer Online-banking customer (matched)

1183.77%

Page 22: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 22

ResultsEffect of Online-Banking Usage on Customer Lifetime Value

Hypothesis that online-banking customers have a lower balance of accounts can be approved

Hypothesis that online-banking customers have a higher value of current stockportfolio can be approved

Hypothesis that online-banking customers have a higher value of current fondportfolio can not be approved

Hypothesis that online-banking customers have a higher number of transactions can be approved

Hypothesis that online-banking customers own more products can not be approved

Effect of online-banking usage on Customer Lifetime Value is twofold.

Page 23: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 23

Introduction• Value-Based Customer Management• Multi-Channel Management in Retailbanking

Evaluation of Multi-Channel PerformanceResults of Empirical StudyConclusions and Future Research

Outline of Presentation

Page 24: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 24

Evaluating the channel effect enables value-based Multi-Channel Management

ConclusionsImplications for Multi-Channel Management

Offline-banking

customers

Online-banking

customers Channel Effect Balance of current account 100.00% 94.84% -5.16% Balance of saving account 100.00% 87.78% -12.22% Value of current stock portfolio 100.00% 1183.77% +1083.77% Value of current fond portfolio 100.00% 93.30% -6.70% Number of transactions 100.00% 125.16% +25.16% Number of products 100.00% 100.00% 0.00% Duration of relationship 100.00% 89.19% -10.81%

Customer channel migration seems to some extent reasonabledue to the fact that channel effect exists.

Page 25: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 25

Effect of online-banking usage on customer retention and customer loyalty

Differences between online-banking and offline-banking customers over timeCompute average change in Customer Lifetime Value (CLVi,online vs. CLVi,offline)

• Different retention rates for online-banking and offline-banking customers result from previous analyses

• Multiplication over all customers allows for analyzing changes in revenues• Separate analysis of cost allows for evaluation of multi-channel investments

Future ResearchA more detailed analysis and evaluating multi-channel investments

( ) ( )( )

+∞= =

=

∆ = − ⋅ ⋅ + − ⋅ ⋅ + =

+

∏ ∏∑

, ,

' '' ', , , ', , ', , , , ', , ',

' 0 ' 0

0

1 1

1

i i online i offlinet tt tt t

i t online i t online i t online i t offline i t offline i t offlinet t

tt

CLV CLV CLV

m r w m r w

d

Page 26: Considering the Self-Selection Effect

E-Finance Lab Jour Fixe, 03. May 2004 26

Thank you for your attention.Any questions?

Dr. Sonja GenslerTel.: +49 69 4272 [email protected]

Department of Electronic Commerce

www.ecommerce.wiwi.uni-frankfurt.dewww.efinancelab.de