APPLICATION

data mining: opportunities and challenges
Chapter XVII - Data Mining in Information Technology and Banking Performance
Data Mining: Opportunities and Challenges
by John Wang (ed) 
Idea Group Publishing 2003
Brought to you by Team-Fly

Consider a set of 15 banks obtained from the Functional Cost and Profit Analysis data set collected by the Federal Reserve Bank under its Functional Cost Analysis (FCA) program. The data for this study is based on the FCA Plus data set from 1997. Table 2 presents the data. In the first stage, we have three inputs: Transactional IT investment, Strategic IT investment, and Labor expenses, and two outputs: number of Accounts and number of Transactions. In the second stage, we have two outputs: Revenue and Equity.

Table 2: Data for 15 banks

Bank

Tran IT

Strategic IT

Labor

Accounts

Transactions

Revenue

Equity


1

68,374,883

27,207,296

96,644,431

28,430.63

109,486,141

4,561,844,525

507,245,000

2

43,674,296

20,728,716

82,211,383

4,020.105

149,772,606

819,640,950

1,697,263,000

3

36,240,480

22,876,333

41,677,506

1,417,653

42,281,949

705,076,794

740,666,344

4

10,974,089

3,604,030

23,386,094

4,801.735

41,624,391

252,167,755

250,842,739

5

8,165,838

4,598,047

11,074,813

5,777.526

14,023,708

137,014,849

120,597,000

6

5,826,344

1,383,431

10,723,845

9,335

1,982,414

160,695,007

218,540,000

7

3,039,890

2,701,362

17,225,000

16,539.81

45,748,993

353,726,000

451,221,000

8

2,631,636

891,139.9

7,585,518

4,245.225

10,054,751

50,543,312

81,451,129

9

2,445,674

531,648.3

7,537,537

1,668.805

8,294,216

59,544,335

55,444,711

10

1,481,997

459,703.1

370,169

3,801

433,717

28,100,210

23,699,857

11

1,471,033

1,070,595

6,378,768

2,230.61

5,915,536

70,210,212

65,192,518

12

1,321,121

863,879.5

3,718,510

16,935.8

7,093,301

37,509,806

37,123,155

13

1,287,557

687,890.1

5,609,930

3,121.808

6,223,975

41,187,111

36,704,000

14

1,164,952

762,892.3

3,449,934

3,822.564

4,476,600

28,526,966

25,739,404

15

1,121,468

391,687.7

3,185,637

4,939.944

8,530,050

41,220,844

43,067,174

The IT investment measures are constructed using data from the following FCA defined expense categories: Vendor Data Processing (all expenses for data processing performed by outside banks, e.g., check processing centers); Telephone and Electronic Access (total expense for communications, including fees for telephone lines, online charges, software installation, and modification); Network Expense Fees (ATM) (all membership and participation fees charged by ATM networks); EFT/ACH Cost Center expense (all expenses related to electronic banking delivery systems other than ATMs; ATM Cost Center expenses (all expenses related to maintenance and support of all ATM transactions on ATMs either owned or leased by the bank); Proof & Transit Cost Center expense (all expenses related to check processing, such as encoding, check sorting, and generating account balances); and Data Processing Cost Center expense (all expenses related to internal data processing, i.e., services provided by the bank's own data processing staff, maintenance and support of institution's software, operating systems, PCs, mainframes, etc.).

Transactional IT refers to IT investment aimed at automation of routine and repetitive tasks to reduce processing costs. Thus, the Transactional IT measure is constructed by adding the total expenses for Vendor Data Processing, the Proof & Transit Cost Center, and the Data Processing Cost Center. On the other hand, Strategic IT refers to IT investment aimed at increasing market share or generating revenues. Thus, the Strategic IT measure is constructed by adding all expenses related to Electronic Access and Automation of customer interface, total expenses for Telephone and Electronic Access, ATM Network Fees, the EFT/ACH Cost Center, and the ATM Cost Center. The labor input is the sum of the salary plus benefits costs of full-time equivalent personnel.

Table 3 reports the efficiency based upon models (2) and (3) in the first and second stage, respectively. The last column reports the overall efficiency that is calculated using the model (2) with Transactional IT, Strategic IT, and Labor as the inputs and Revenue and Equity as the outputs. (i.e., we ignore the intermediate measures.) It can be seen that overall efficient banks do not necessarily indicate efficient performance in the two stages (see, e.g., banks 1, 7, 17, and 15). Also, if a bank is efficient in stages 1 and 2, this bank may be identified as inefficient in overall efficiency. For example, bank 3. This indicates that we need to use model (6) to correctly characterize the banking performance.

Table 3: Banking performance
 

DEA

 

Bank

θ*

Φ*

Overall


1

0.59

1.00

1.00

2

1.00

1.00

1.00

3

1.00

1.00

0.88

4

0.68

1.83

0.52

5

0.47

1.82

0.44

6

0.33

1.00

1.00

7

1.00

1.40

1.00

8

0.55

2.04

0.68

9

0.74

1.00

0.97

10

1.00

1.00

1.00

11

0.76

1.00

0.88

12

1.00

6.89

0.85

13

0.87

2.40

0.87

14

0.96

3.23

0.96

15

1.00

3.68

1.00

Table 4 reports the results from model (6) with different weight combinations. When w1 = w2 = 1, we have, given the optimal intermediate measures of Accounts and Transactions, (i) two banks (2 and 10) that achieve 100% efficiency in both stage 1 and stage 2; (ii) 10 banks that achieve 100% efficiency in the IT-related activity (stage 1) without achieving 100% efficiency in stage 2; (iii) two banks (1 and 3) that do not achieve 100% efficiency in the IT-related activity while achieving 100% efficiency in stage 2; and (iv) two banks (4 and 6) that do not achieve 100% efficiency in both stages.

Table 4: IT and banking performance
 

w1 = w2 =1

w1 = 5, w2 = 1

w1 = 1, w2 = 5


Bank

α*

β*

α*

β*

α*

β*


1

0.59

1

0.59

1

0.59

1

2

1

1

1

1

1

1

3

0.62

1

0.62

1

1

1.31

4

0.75

2.42

0.13

1.05

1

2.61

5

1

3.72

0.18

1.89

1

3.72

6

0.86

1.84

0.33

1.29

1

1.98

7

1

1.4

0.68

1

1

1.4

8

1

4.43

0.51

3.5

1

4.43

9

1

4.98

0.88

4.8

1

4.98

10

1

1

1

1

1

1

11

1

4.82

0.82

3.99

1

4.82

12

1

7.51

0.96

7.41

1

7.51

13

1

7.5

1

7.5

1

7.5

14

1

8.17

1

8.17

1

8.17

15

1

3.68

1

3.68

1

3.68

When w1 = 5 and w2 = 1, i.e., we are more interested in the potential saving on the IT investment. The efficiency of banks 1, 3, 10, 13, 14, and 15 stays the same. The efficiency of some other banks changes because of the tradeoff between the two stages. For example, bank 5 eliminates some inefficiency in its second stage in exchange of input savings in the first stage. If we use w1 = 1 and w2 = 5, (i.e., we are more interested in the potential improvement on the financial performance in the second stage), we obtain a set of different results as shown in the last two columns of the Table 4.

Table 4 only presents the efficiency scores. In fact, model (6) also yields benchmarks for the inefficient banks. For example, consider bank 3 with w1 = w2 = 1. In the first stage, we have λ*2 = 0.14 (bank 2) and λ*7 = 0.86 (bank 7), indicating that banks 2 and 7 are used as benchmarks. In the second stage, we have μ*1 = 0.07 (bank1), μ*2= 0.34 (bank 2), and μ*6 = 0.59 (bank 3), indicating that banks 1, 2, and 6 are used as benchmarks.

Model (6) provides optimized values on the intermediate measures of Accounts and Transactions. Table 5 reports optimal values of Accounts and Transactions. Consider bank 3 when w1 = w2 =1. Model (6) indicates that bank 3 should increase its values on Accounts and Transactions. Such results are very important because they provide banks with valuable information in terms of how their business operations objectives should be.

Table 5: Optimal accounts and transactions
 

w1 = w2 =1

w1 = 5, w2 = 1

w1 = 1, w2 = 5

Bank

Accounts

Transactions

Accounts

Transactions

Accounts

Transactions

1

28,431

109,486,141

28,431

109,486,141

28,431

109,486,141

2

4,020

149,772,606

4,020

149,772,606

4,020

149,772,606

3

14,845

59,830,119

14,845

59,830,119

11,829

84,890,085

4

16,540

45,748,993

9,571

7,936,096

15,913

50,957,680

5

11,458

29,444,588

9,745

4,664,650

11,458

2,9444,588

6

8,926

21,318,248

9,194

8,631,691

9,921

24,511,377

7

16,540

45,748,993

10,653

26,860,582

16,540

45,748,993

8

8,823

16,231,013

9,096

8,629,344

8,823

16,231,013

9

9,647

9,773,165

9,654

8,642,667

9,647

9,773,165

10

3,801

433,717

3,801

433,717

3,801

433,717

11

9,628

14,045,011

9,661

7,964,574

9,628

14,045,011

12

9,562

9,735,907

9,566

9,322,128

9,562

9,735,907

13

9,710

9,898,598

9,710

9,898,598

9,710

9,898,598

14

7,553

8,217,129

7,553

8,217,129

7,553

8,217,129

15

4,940

8,530,050

4,940

8,530,050

4,940

8,530,050

Brought to you by Team-Fly


Data Mining(c) Opportunities and Challenges
Data Mining: Opportunities and Challenges
ISBN: 1591400511
EAN: 2147483647
Year: 2003
Pages: 194
Authors: John Wang

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net