Forecasting methods and uses for demand deposits of U.S. commercial banks.
During the past decade, the banking industry has witnessed a multitude of dramatic changes, such as the deregulation of the financial sector, competition from other financial institutions, and new information technology such as the Internet. All of these changes have produced a combined effect, leading to the unprecedented present day competitive market environment. In order to survive in this highly volatile industry, competent forecasting and planning have become vital activities for banks. The managers of these institutions require timely and accurate forecasts of variables such as deposits, loans, exchange rates, and interest rates, in order that they might fulfill their planning and control responsibilities in an effective manner. In essence, all of the major budgeting practices of these institutions are dependent upon the forecasting function.
Given the critical role of the forecasting function, insights into current demand deposit forecasting practices and the possible success of these practices should be of major value to bank management. However, most past studies on forecasting have focused on a cross-sectional analysis (Dalrymple, 1987, 1975; Mentzer & Cox, 1984; Sanders, 1992, 1994). Sanders (1997) has noted that since the management of service organizations is in many ways different from that of manufacturing companies, combining information on forecasting practices in manufacturing and service firms can only lead to diffused generalizations and is not helpful in understanding practices in a specific industry segment.
Unfortunately, detailed studies of the forecasting methods employed in the banking industry have not been undertaken, although other aspects of the forecasting function, such as developing forecasting models and comparing their accuracies, have been assessed (Ellis, 1995).
In this study the focus is on U.S. commercial banks. It assesses current bank demand deposit forecasting practices and probes into problems that are specific to this environment. The specific objectives are: (1) to explore the uses of demand deposit forecasts; (2) to evaluate forecasting methods and forecasting time parameters; and (3) to examine the criteria used for evaluating forecasting effectiveness and the measures used for forecasting accuracy.
In this study a mail survey was utilized to obtain information about demand deposit forecasting practices in commercial banks. An initial mail questionnaire was developed, based upon questionnaires utilized in previous studies (Dalrymple, 1987, 1975; Giroux, 1980; Mentzer & Cox, 1984; Peterson & Jun, 1999). This preliminary measuring instrument was reviewed by two practitioners from the banking industry and several alterations were produced, based upon their inputs. The final questionnaire was designed and formulized to collect data which could be of value to bank managers.
The survey was forwarded to the presidents of a sample of U.S. banks, requesting them to forward the survey questionnaire to the manager who is responsible for preparing demand deposit forecasts. A total of 400 banks were randomly selected from the Thomson Bank Directory (Thomson Financial Publishing, 1999). Of the responses received, 83 questionnaires were usable. This results in a response rate of 20.8%, which is comparable to similar surveys and can be regarded as an acceptable rate considering the length (seven-page) of the questionnaire.
Table 1 summarizes the characteristics of respondents regarding the approximate size of annual demand deposits, the number of years respondents have been employed with the banks, and the approximate ages of the banks classified by company size: large and small size. In this inquiry a large bank was defined as one with more than $500 million of demand deposits and a small firm as one with less than or equal to $500 million.
The bulk of the respondents were executives whose job titles included chief executive officer, president, vice-president of branch management, forecasting manager, controller, and director of management information systems.
This section considers the key results of the survey. The findings are discussed in order--preparation of demand deposit forecasts, preparation of other types of forecasts, uses of demand deposit forecasts, forecasting methods used, forecast time parameters, criteria for evaluating forecasting effectiveness, and measures of forecasting accuracy.
Preparation of Demand Deposit Forecasts
Table 2 summarizes data on the generation of demand deposit forecasts, based on bank size--large and small. The table indicates that most of the large banks (84.6%) prepare demand deposit forecasts whereas approximately two-thirds of the small banks (63.2%) develop the forecasts. This is not unexpected since, compared with small banks, larger institutions command more of the financial, technical, and human resources necessary to engage in forecasting programs and are more likely to integrate systematic forecasting systems with their formal planning processes.
Preparation of Other Types of Forecasts
The respondents were requested to identify the forecasts they prepared (with the exception of demand deposit forecasts which was addressed in the previous question) from the listing of forecasts set forth in the questionnaire. As shown in Table 3, for large banks, most of the responding companies are developing forecasts on time deposits (92.3%), commercial loans (92.3%), and consumer loans (88.5%). These were followed by short-term interest rate (69.2%) and long term interest rate (57.7%). On the other hand, for small banks, about two-thirds of the banks generate four types of forecasts: in descending order of frequency, commercial loans (77.2%), short term interest rate (66.7%), time deposits (68.4%), and consumer loans (64.9%). From Tables 2 & 3, it is evident that demand deposit forecasts are one of the primary categories for both bank groups.
Uses of Demand Deposit Forecasts
The respondents were asked to list in order of relative importance the managerial processes in which the forecasts were employed. A total of six categories were elicited: cash budgeting, profit planning, capital budgets preparation, strategic planning, market planning, and personnel planning.
Table 4 presents the frequencies of response to this question for both large and small banks. It indicates that those in each of the two groupings utilize demand deposit forecasts most often, as input for profit plans. This was followed by, in order of descending frequency, profit planning, strategic planning, cash budgeting, and market planning. In addition, three respondents indicated that the forecasts were utilized as input for liquidity management, asset liability management, and loan growth planning, respectively.
The respondents were asked to identify the major forecasting methods which they employed. They were provided with a listing of 14 forecasting methods. The description of each of the techniques was provided in the questionnaire to prevent respondents from making classification errors, based on nomenclature alone. Those techniques include jury of executive opinion, sales force composite, customer expectations, decomposition, exponential smoothing, moving average, regression analysis, simulation, straight-line projection, Box-Jenkins time series models, expert systems, neural networks, trend line analysis, and life cycle analysis.
In order to gain detailed insights on the degree of usage of the techniques, they were asked to indicate which of these were used regularly, occasionally, or never/no longer used. Respondents were then requested to identify the forecasting time horizons for each of the forecasting methods used, from three categories: less than 3 months, 3 months to 2 years, or over 2 years, and to check their satisfaction levels for the forecasting techniques identified previously from three alternatives: satisfied, neutral, or dissatisfied.
Table 5 presents the frequencies regarding the usage rates of forecasting methods. The bankers commonly used the following six forecasting techniques on a regular basis for generating demand deposit forecasts: in descending order of frequency, jury of executive opinion (70.6%), straight line projection (43.1%), sales force composite (37.3%), decomposition (27.5%), simulation (23.5%), and moving average (21.6%). In terms of the extent of satisfaction, all those six methods received "satisfied" from more than 50% of the managers who used those techniques. As for the time horizon, all of the six techniques mentioned above were primarily used for developing medium range forecasts with a time horizon of from 3 months to two years, all of which received over 80% of frequency. On the other hand, techniques such as life cycle analysis, expert system, and neural networks are rarely used and Box-Jenkins time series were never utilized by the bankers.
It is evident that the jury of executive opinion is the most popular forecasting technique with the highest satisfaction level (88.6%). This result is consistent with Giroux (1980), in which the dominant forecasting technique used by commercial banks was found to be the "judgmental only" forecasting technique.
Among the quantitative forecasting methods used, straight line projection (43.1%) is the most widely cited, followed by decomposition, simulation, and moving average. These findings are somewhat inconsistent with those of Giroux (1980). In his study, the most widely utilized quantitative technique was multiple regression, followed by ties of multiple equation models and simulation.
With respect to the effect of firm size on forecasting methods used on a regular basis, both the large and small bank groups manifest similar patterns in relation to the forecasting techniques used, the degree of their satisfaction levels, and the forecast time horizons for which each of the techniques are used. Large banks most often deployed, in descending order of frequency, jury of executive opinion (61.1%), sales force composite (39.9), decomposition (38.9%), straight line projection (38.9%), and simulation (27.8%), whereas small banks predominantly employed jury of executive opinion (75.8%), followed by straight line projection (45.5%), sales force composite (36.4%), moving average (24.2%), decomposition (21.2%), and simulation (21.2%) (see Table 6).
Forecast Time Parameters
The questionnaire requested that the respondents specify the forecast time parameters--the horizon (time period covered), the interval (time periods for which data were inputted into the forecast model), and the frequency of preparation for their firms. Table 7 presents the results for the large and small banks. Large institutions use from four to 12 months (40.9%) and from 13 to 24 months (40.9%) of time horizons the most frequently and next from four to five years (18.2%). In the case of small banks, the most frequently employed forecast horizon is from four to 12 months (52.8%) and the second less than one month (33.3%). Hence, it appears that large banks tend to have longer time horizons than their counterparts. This is not unexpected since larger firms tend to formally develop long term forecasts and strategic plans to greater extent than do small banks.
Concerning the time interval, as shown in Table 7, monthly demand deposit figures were the most often used by both of the two groups (for small banks, 55.6%; for large banks, 90.9%) and yearly demand figures were the next most frequently used (for small banks, 30.6%; for large banks, 9.1%). As for forecasting frequency, the largest proportion of the large bank group developed the forecasts monthly (40.9%), and the second largest annually (27.3%) whereas the majority of the small bank group prepared the forecast yearly (61.1%), and next monthly (16.7%) (see Table 7).
Criteria for Evaluating Forecasting Effectiveness
The respondents were asked to rank the following five criteria in evaluating the effectiveness of demand deposit forecasts: accuracy, credibility, ease of use, customer service performance, and amount of data required. As shown in Table 8, there is virtually no difference between large and small banks in the relative importance of those evaluative criteria. In turn, the criteria most commonly cited by most of the firms were (in terms of the first and second ranks combined) accuracy and credibility. For the small bank group these were followed by ease of use, customer satisfaction, and amount of data, and for the large bank group, followed by customer satisfaction performance.
Measures of Forecasting Accuracy
The questionnaire asked respondents to indicate which forecasting error measurements they used. A total of seven accuracy measurements were listed: mean absolute percentage error, mean absolute deviation, mean squired error, deviation, percentage error, forecast ratio, and standard deviation.
As shown in Table 9, large banks primarily employed percentage error (54.5%), forecast ratio (31.8%), and deviation (22.7%), while small banks most frequently employed mean absolute percentage error (33.8%), deviation (30.6%), percentage error (19.4%), and forecast ratio (19.4%). Conversely, mean absolute deviation, mean squared error, and standard deviation, were seldom used by large or small banks.
SUMMARY AND CONCLUSIONS
Effective management of funds is essential to the success of financial institutions including commercial banks. Particularly, to optimally control the flow of demand deposits and associated cost structure, and in turn to increase the efficiency in managing both the asset and liability sides of the balance sheet, commercial banks need to forecast their demand deposits accurately.
This inquiry into the demand deposit forecasting practices of the banking industry derived a number of major findings. Demand deposit forecasts are developed by most of the banks surveyed for a variety of important plans such as cash and capital budgets, profit plans, and marketing plans. The most popular forecasting method is jury of executive opinion. This was followed by straight line projection and sales force composite. The majority of the responding managers are satisfied with the performances of those techniques and used them for generating particularly medium range forecasts. Both of the two groups--large and small banks-show similar usage of forecasting methods on a regular basis, but the large banks develop demand deposit forecasts more frequently and for longer time horizon than do the small banks. The majority of the banks most frequently utilize two criteria--accuracy and credibility--for evaluating forecasting effectiveness, and also commonly employ percentage error and forecast ratio for accuracy measurements.
The findings of this study are illuminating in terms of revealing forecasting practices of one specific industry and differences by size of firm. It is recommended that further research be conducted for other forecasting types such as time deposit and interest rates, in an effort to assess the prevalence of the results set forth herein.
Dalrymple, D. J. (1975). Sales forecasting methods and accuracy. Business Horizons, 18, 69-73.
Dalrymple, D. J. (1987). Sales forecasting practices--results from a United States survey. International Journal of Forecasting, Summer, 379-391.
Ellis, D. F. (1995). Consensus forecasts of financial institutions. Journal of Business Forecasting, 14 (3), 34-37.
Giroux, G. A. (1980). A survey of forecasting techniques used by commercial banks. Journal of Bank Research, Spring, 51-53.
Mentzer, J. & J. Cox (1984). Familiarity, application and performance of sales forecasting techniques. Journal of Forecasting, Summer, 27-36.
Peterson, R. T. & M. Jun (1999). Forecasting sales in wholesale industry. The Journal of Business Forecasting, 18(2), 15-18.
Sanders, N. R. (1992). Corporate forecasting practices in the manufacturing industry. Production and Inventory Management Journal, 33(2). 54-57.
Sanders, N. R. & K. B. Manrodt (1994). Forecasting practices in US corporations: survey results. Interfaces, 24(2), 92-100.
Sanders, N. R. (1997). The status of forecasting in manufacturing firms. Production and Inventory Management Journal, Second Quarter, 32-35.
Thomson Financial Publishing (1999). Thomson Bank Directory, (1999). Skokie, IL: Thomson Financial Publishing.
Minjoon Jun, New Mexico State University
Robin T. Peterson, New Mexico State University
Table 1: Sample Characteristics Volume of Annual Demand Deposit Frequency (%) Below $100 mil. 26 (31.3) $100 mil. - $500 mil. 31 (37.3) $500 mil. - $1 bil. 9 (10.8) $1 bil. - $10 bil. 17 (20.5) More than $10 bil. 1 (1.2) Total 83 (100.0) Years of Respondents Employed Frequency (%) with the Banks Less than 1 1 (1.2) 1 - 3 14 (16.9) 4 - 6 15 (18.1) 7 - 10 11 (13.3) More than 10 42 (50.6) Total 83 (100.0) Age of Banks (years) Large Small Frequency (%) Frequency (%) 1 - 10 1 (3.9) 0 (0.0) 11 - 30 1 (3.9) 14 (24.6) 31 - 60 6 (23.1) 5 (8.8) 61 - 100 6 (23.1) 23 (40.4) Over 100 12 (46.4) 15 (26.3) Total 26 100.0) 57 (100.0) Table 2: Preparation of Demand Deposit Forecasts Large Banks Small Banks Frequency % Frequency % Yes 22 84.6 36 63.2 No 4 15.4 21 36.8 TOTAL 26 100.0 57 100.0 Table 3: Preparation of Other Types of Forecasts Types of Forecasts Large Banks (n=26) Frequency % Time Deposits 24 92.3 (1) Commercial Loan Forecasts 24 92.3 Consumer Loan Forecasts 23 88.5 Short-term Interest Rate Forecasts 18 69.2 Long-term Interest Rate Forecasts 15 57.7 Others 5 19.2 Types of Forecasts Small Banks (n=57) Frequency % Time Deposits 39 68.4 Commercial Loan Forecasts 44 77.2 Consumer Loan Forecasts 37 64.9 Short-term Interest Rate Forecasts 38 66.7 Long-term Interest Rate Forecasts 18 31.6 Others 11 19.3 Note: 1. Percentages do not add to 100 % because of multiple answers given. Table 4: Uses of Demand Deposit Forecasts Uses of Forecasts Large Banks (n=22) Rank 1 Rank 2 Total Profit Plans 15 3 18 Strategic Plans 3 8 11 Cash Budget 3 3 6 Market Planning 1 2 3 Capital Budgets 0 2 2 Personnel Planning 0 0 0 Uses of Forecasts Small Banks (n=36) Rank 1 Rank 2 Total Profit Plans 17 9 26 Strategic Plans 10 3 14 Cash Budget 7 5 13 Market Planning 1 6 7 Capital Budgets 0 4 4 Personnel Planning 1 1 2 Table 5: Uses of Forecasting Methods Forecasting Techniques N=51 (1) Regularly Rank Judgemental Methods Jury of Executive Opinion 36 (2) (70.6) (3) 1 Sales Force Composite 19 (37.3) 3 Customer Expectations 7 (13.7) 7 Quantitative Methods Straight Line Projection 22 (43.1) 2 Decomposition 14 (27.5) 4 Simulation 12 (23.5) 5 Moving Average 11 (21.6) 6 Trend Line Analysis 6 (11.8) 8 Exponential Smoothing 5 (9.8) 9 Regression 3 (5.9) 10 Neural Networks 3 (5.9) 10 Expert Systems 1 (2.0) 12 Life Cycle Analysis 1 (2.0) 12 Box-Jenkins Time Series 0 (0.0) 14 Forecasting Techniques Occasionally Never/No Longer used Judgemental Methods Jury of Executive Opinion 8 (15.9) 7 (13.7) Sales Force Composite 11 (21.6) 21 (41.2) Customer Expectations 23 (45.1) 21 (41.2) Quantitative Methods Straight Line Projection 7 (13.7) 22 (43.1) Decomposition 9 (17.6) 28 (54.9) Simulation 9 (17.6) 30 (58.8) Moving Average 10 (19.6) 30 (58.8) Trend Line Analysis 12 (23.5) 33 (64.7) Exponential Smoothing 1 (2.0) 45 (88.2) Regression 8 (15.7) 40 (78.4) Neural Networks 5 (9.8) 43 (84.3) Expert Systems 4 (7.8) 46 (90.2) Life Cycle Analysis 10 (19.6) 40 (78.4) Box-Jenkins Time Series 0 (0.0) 51 (100.0) Note: (1): Total number of respondents (2): Frequency (3): Percentage (Percentages do not add to 100 % because of multiple answers given.) Table 6: Regular Usage of Forecasting Methods by Bank Size Forecasting Techniques Large Banks (n = 18) Percentage Rank Judgemental Methods Jury of Executive Opinion 61.1 (1) 1 Sales Force Composite 38.9 2 Customer Expectations 5.6 9 Quantitative Methods Straight Line Projection 38.9 2 Decomposition 38.9 2 Simulation 27.8 5 Trend Line Analysis 22.2 6 Exponential Smoothing 16.7 7 Moving Average 16.7 7 Regression 5.6 9 Expert Systems 5.6 9 Neural Networks 5.6 9 Life Cycle Analysis 0.0 13 Box-Jenkins Time Series 0.0 13 Forecasting Techniques Small Banks (n = 33) Percentage Rank Judgemental Methods Jury of Executive Opinion 75.8 1 Sales Force Composite 36.4 3 Customer Expectations 18.2 7 Quantitative Methods Straight Line Projection 45.5 2 Decomposition 21.2 5 Simulation 21.2 5 Trend Line Analysis 6.0 8 Exponential Smoothing 6.0 8 Moving Average 24.2 4 Regression 6.0 8 Expert Systems 0.0 13 Neural Networks 6.0 8 Life Cycle Analysis 3.0 12 Box-Jenkins Time Series 0.0 14 Note: (1.) Percentages do not add to 100 % because of multiple answers given. Table 7: Forecast Time Parameters Time Horizon Large Banks (n=22) Small Banks (n=36) Frequency % Frequency % Less than 1 month 1 4.5 12 33.3 2 - 3 months 1 4.5 3 8.3 4 - 12 months 9 40.9 19 52.8 13 - 24 months 9 40.9 2 5.6 25 - 36 months 0 0.0 0 0.0 4 - 5 years 4 18.2 1 2.8 6 - 10 years 1 4.5 1 2.8 Over 10 years 0 0.0 0 0.0 Time Interval Large Banks (n=22) Small Banks (n=36) Frequency % Frequency % Weekly 0 0.0 0 0.0 Monthly 20 90.9 20 55.6 Quarterly 1 4.5 4 11.1 6 months 0 0.0 1 2.8 Annually 2 9.1 11 30.6 Forecasting Large Banks (n=22) Small Banks (n=36) Frequency Frequency % Frequency % Weekly 0 0.0 0 0.0 Monthly 9 40.9 6 16.7 Quarterly 4 18.2 5 13.9 Semi-annually 3 13.6 3 8.3 Annually 6 27.3 22 61.1 As Needed 1 4.5 0 0.0 Note: Percentages do not add to 100 % because of multiple answers given. Table 8: Criteria for Evaluating Demand Deposit Forecasting Effectiveness Criteria Large Banks (n=21) Rank 1 Rank 2 Total Accuracy 11 6 17 Credibility 8 8 16 Customer Satisfaction 2 4 6 Performance Ease of Use 0 0 0 Amount of Data Required 0 0 0 Criteria Small Banks (n=35) Rank 1 Rank Total 2 Accuracy 21 7 28 Credibility 10 13 23 Customer Satisfaction 3 2 5 Performance Ease of Use 0 8 8 Amount of Data Required 1 2 3 Table 9: Measures of Forecasting Accuracy Accuracy Measurements Large Banks (n=22) Frequency % Percentage Error 12 54.5 (1) Forecast Ratio 7 31.8 Deviation 5 22.7 Standard Deviation 1 4.5 Mean Absolute Percentage Error 1 4.5 Mean Absolute Deviation 0 0.0 Mean Squared Error 0 0.0 None 2 9.1 Accuracy Measurements Small Banks (n=36) Frequency % Percentage Error 7 19.4 Forecast Ratio 7 19.4 Deviation 11 30.6 Standard Deviation 2 5.6 Mean Absolute Percentage Error 12 33.3 Mean Absolute Deviation 1 2.8 Mean Squared Error 1 2.8 None 2 5.6 Note: 1. Percentages do not add to 100 % because of multiple answers given.
|Printer friendly Cite/link Email Feedback|
|Author:||Jun, Minjoon; Peterson, Robin T.|
|Publication:||Academy of Banking Studies Journal|
|Date:||Jan 1, 2003|
|Previous Article:||Restructuring commercial banks in the Republic of Uzbekistan.|
|Next Article:||Insider trading around bank failures.|