Printer Friendly

Internet banking: gold mine or money pit?


This paper explores the impact that implementing internet banking has on a bank's profits, costs and revenues. A production function approach utilizing maximum likelihood estimation procedures is used. We find that internet banking improves bank profits, and that the extent of this impact depends on the number of customers the bank succeeds in getting to adopt it. At lower levels of consumer adoption, profit gains are driven by increased revenues. These revenue improvements, however, taper off until profits are largely unaffected by further consumer adoption. Finally, at a sufficiently high level of consumer adoption, lower costs drive increased profits.


Firms in myriad industries spend millions, even billions, of dollars on Information Technology (IT). In 2001, IT spending topped $527 billion in the US (CIO Magazine, November 8, 2002.) and $1.2 trillion worldwide (CIO Magazine, April 12, 2002). The banking industry, for example, invests heavily in IT--increasing investment from 6% of revenues in 2000 to 6.5% in 2001 (Parton, C. & J. Glaser, 2002). However, many bank CEOs are unsure of IT's impact (Vrechopoulos, A., & G. Siomkos, 2002), while others claim that these investments lead to increased services but decreased profits (Ross, J. & P. Weill, 2002). A specific technology--Internet Banking (IB)--has been heralded as both a competitive advantage and a competitive necessity (Dynamicnet, 2001; Vrechopoulos, A., & G. Siomkos, 2002; Retail Banking Research, Ltd., 1996). Accordingly, banks spent 15% of their IT dollars on internet banking in 2002 (Olazabal, N, 2003). Yet, many banks question its profitability (Hoffmann, K, 2003, March). In fact, half of all U.S. banks have no plans to offer internet banking (Orr, B, 2001), with the most common reason given being an unclear return on investment (Parton, C. & J. Glaser, 2002).

This paper utilizes firm level data gleaned from public and proprietary sources to take a detailed econometric look at the profit implications to banks of implementing internet banking. Three questions concerning internet banking are addressed. 1) Does offering IB improve bank profits and are these improvements derived from lower costs and/or increased revenues? 2) Does bank management influence the value of IB through either the decision to be an early-mover or by its ability to get its customers to adopt IB? 3) Does the value of IB change over time?

We find that internet banking does improve bank profits, and that the extent of this impact depends on the number of customers the banks succeeds in getting to adopt it. At lower levels of consumer adoption profit gains are driven by increased revenues and not by lower costs. These revenue improvements, however, diminish as additional consumers adopt until profits are largely unaffected. Then, after a sufficiently high level of consumer adoption is reached, lower costs do indeed drive increased profits. That is, IB can generate both revenue and cost advantages for those banks that implement it. We also find no evidence of an early-mover advantage, nor do we see that the impact of IB has softened over time.

The paper proceeds as follows. We begin with a discussion of how internet banking provides value to both a bank and its customers. This discussion produces four hypotheses for investigation. Next, the regression models needed to address our research questions are developed. The data then are described and our empirical findings presented. We conclude with a summary that highlights managerial and academic implications.


Internet banking is defined as a transaction-oriented system that enables a bank's customers to engage in online banking activities. IB enhances a bank's offerings by providing its customers with 24/7 access to many of its services. The services available through IB can vary but typically include informational account access (view balances and past transactions), funds transfers among accounts at the providing institution, bill payment, bill presentment, and loan application and approval (including credit cards, mortgages and other personal loans).

For the individual consumer or small business, this additional channel offers improved service and convenience. A customer can access their accounts and conduct transactions from home or work at any time, thus saving the time and effort of going to a branch location. For some group of customers, this added convenience will make IB their preferred means of transacting. In such, a bank's implementation of IB provides a reason to switch to that particular bank or remain and, perhaps, do additional business with it.

IB, as with all investments, is expected to increase bank profits by either increasing revenues more than costs or by decreasing costs. Simply put, market forces would not permit continued investment in IB if it did not pay off. Our expectation of higher profits is buoyed by the fact that internet banking was initially heralded as a competitive advantage. As such, banks that implemented IB effectively should have higher profits than those that did not. Later, IB was proclaimed a competitive necessity (Vrechopoulos, A., & G. Siomkos, 2002; Retail Banking Research, Ltd., 1996). If IB is indeed a competitive necessity, then implementing IB is still done to increase profits relative to what would be earned without IB. In either case, banks that have not adopted internet banking should be at a competitive disadvantage (i.e., be less profitable) relative to those that have adopted it.

Bank profits increase through increased revenues and/or decreased costs. Revenues come from two primary sources--interest on loans and fees on services. Offering IB may increase loan revenue by attracting new consumers to the bank who then go on to use the bank's loan products, such as credit cards, mortgages and installment loans. Additionally, IB may entice existing customers to deepen their relationship with the bank and use more of the banks loan products. In a similar fashion, IB may increase fee revenue by attracting new customers and by encouraging existing customers to use more fee-based services. Banks may even charge for some or all of their IB services. In sum, whether due to increased fees or loans, increased revenues due to IB are linked to level of IB adoption by the bank's customers.

Previous findings support the notion that consumers who use IB are more lucrative. Woodford (Woodford, R, 2001) reports that customers acquired due to implementing IB are likely to generate above average revenues. Hitt and Frei (Hitt, L. & F. X. Frei, 2002) find that customers who use PC banking (which provides similar functionality and benefits to IB) are more profitable than those who do not. In particular, IB customers have significantly higher mortgage and loan balances, and they use twice the number of other financial services products (Woodford, R, 2001). IB users also reflect the demographics coveted by bankers (they have almost twice the household income of people not using IB ($86,000 vs. $47,000)) and such "attractive demographics mean attractive economics"(Olazabal, N, 2003). On the other hand, this reported appeal of IB and PC banking customers is likely enhanced by the time period when the studies were performed. In particular, a diffusion of innovations argument likely is relevant (Horsky, D, 1990; Rogers, E, 1995). Early adopters of an innovation typically have the most interest and willingness to pay for the innovation. Later adopters are typically less inclined toward the innovation. Relative to IB, this implies that customers who adopt IB early on are likely to be the most profitable, and as more and more people adopt IB, the incremental revenue per customer will diminish.

Previous findings also support the notion that IB users are likely to come largely from the bank's installed customer base (Hitt, L. & F. X. Frei, 2002). Furthermore, these customers tend to deepen their relationship with the bank by shifting a greater share of their financial business to the bank. In particular, existing customers who adopt online banking acquire products at a faster rate and maintain higher loan and deposit balances.

Bank costs are driven by two primary factors--the cost of money (the interest paid on deposits and other funds) and operating costs (what is paid for IT, buildings, labor, etc). Although IB is unlikely to affect a bank's price of funds, the amount of funds on which it pays interest will increase if IB brings in new customers or leads existing customers to deepen their relationship with the bank by increasing the number of deposit accounts.

IB's impact on operating costs is more complex. As with all IT, there are fixed costs related to the purchase, implementation and maintenance of IB. On the other hand, both total variable cost and fixed cost reductions related to the replacement of the "earlier" technology are tied to how much consumers use IB. Due to the "paper-less" nature of IB and its use of the internet infrastructure, IB transactions likely cost less to perform than do transactions done through an ATM or teller. In fact, a Booze, Allen and Hamilton study (Dynamicnet, 2001; Goldfinger, C., 2003) claims IB has an expense rate that is 1/3 that of traditional banking, with a per transaction cost of $.01 for IB versus $.27 for an ATM transaction and $1.07 for a teller-facilitated transaction. However, despite a lower per transaction cost, total variable costs may increase if IB, either through more transactions per customer or more customers, results in a significant increase in the total number of transactions. IB also may result in lower fixed infrastructure costs related to, say, the number of tellers or branch banks, if the substitution of IB transactions for offline transactions is sufficiently high. However, incomplete substitution of the earlier technology fits with Sullivan's (Sullivan, R, 2000) finding that banks with IB have higher non-interest expenses.

What this means is that simply installing IB is likely to raise bank operating costs, since fixed IB costs are incurred and no fixed infrastructure or total variable cost reductions are realized. Whether costs increase with increased usage of IB is less clear. As IB transactions replace offline transactions, transaction (variable unit) cost reductions are almost surely realized and infrastructure cost reductions are possible. It follows that at some level of IB usage, a lower total cost is incurred since total variable and infrastructure cost reductions will exceed the fixed costs associated with IB. However, this logic holds the number of bank transactions constant. Increased demand for bank services due to IB means more bank transactions are performed. If this increase is sufficiently large, total variable costs actually rise rather than fall since the cost associated with the increased number of transactions exceeds the cost reduction due to using the less costly IB transactions.

Increased operating expenses are further indicated because the banks we analyze use an IB service provider. Consequently, their variable transaction cost structure is a bit different from that for banks that develop their own IB system, and, therefore, different from what the Booze, Allen and Hamilton report predicts. To implement IB, there is still a fixed, but small, one-time cost for the system in the $20,000 to $40,000 range. Once the system is installed, there is a nominal fixed monthly fee (less than $1000) plus a fee for each IB customer as well as fees for various transactions. The total fee for each customer depends on the services the customer is signed up for (e.g., bill-pay) and how much historical information the bank retains (i.e., it costs the bank more if the bank lets its customers view deposits and withdrawals from two years ago than if the maximum a customer can go back is six months). This fee can range from $2 to $15 per customer per month. Additional fees also are charged for most transactions, including signing up a new IB customer and for a customer making a payment. Inquiries are free. This cost structure makes most of the bank's costs variable. This lowers the risk of implementing IB, but also makes lowering the bank's overall operating costs more difficult.

In sum, given the above discussion, we expect revenues to increase as consumer usage of IB increases, with this positive impact gradually petering out. Due to the use of an ASP (rather than incurring large fixed costs), total costs are expected to rise initially as consumer IB usage increases, with this increase gradually subsiding until eventually scale economies allow costs to decrease. Profits thus are likely to take on an "S-shape" with respect to consumer IB usage. This discussion of how IB likely influences bank costs and revenues, and, hence, profits, leads to two null hypotheses:

H1: Simply implementing IB has no impact on a bank's costs, revenues or profits.

H2: Greater customer adoption of IB has no impact on bank costs, revenues or profits.

The decision of when to implement IB also may be important to banks given that Porter (Porter, M, 1998) identifies switching costs and reputation as key determinants of whether or not an early-mover advantage exists for technology investments. Each is especially relevant since the time required for the consumer to set up and use an IB account is significant and the perceived risk inherent to anything associated with money is high. Case-based work by Reich and Benbasat (Reich, B. & I. Benbasat, 1990) provides empirical support. They find that being an early-mover significantly impacts the bottom-line of customer-oriented strategic systems. Correspondingly, we propose a third null hypothesis:

H3: Banks that adopt IB early have no cost, revenue or profit advantage relative to banks that adopt IB later.

A related issue is whether or not the value a bank derives from IB changes over time. A reduction may occur over time as the value that IB generates increasingly goes to consumers in the form of improved quality (more convenience) rather than to the firm as increased revenues or lowered costs (Landuaer, T, 1995; Licht, G. & D. Moch, 1999). This rationale focuses on competitive pressures--as the number of competitors who utilize a technology increases a greater share of the value generated by the technology goes to consumers. A fourth null hypothesis follows:

H4: The impact of IB on costs, revenues and profits does not change over time.


To represent the core relationships of profits, revenues and costs with a set of input-output variables, we follow the IS productivity literature and use a Cobb-Douglas functional form with a standard set of input-output variables (Brynjolfsson, E. & L. Hitt, 1999; Lichtenberg, F, 1995). The model's multiplicative nature explicitly allows the influence of each explanatory variable to depend on the levels of the other explanatory variables. More complex functional forms such as the translog (Alpar, P & M. Kim, 1990) and Fourier-flexible (Berger, A. & L. Mester, 1997) also have been used. These forms require the estimation of numerous inter-related parameters that make discussion of how a particular explanatory variable influences the dependent variable quite difficult. The translog form, for example, involves additional parameters for the squares of each explanatory variable as well as all possible cross-products of these variables. For our model, this would require fifteen additional parameters. The Fourier-flexible form expands the translog model by adding Fourier trigonometric terms as well. Because we have data on only 275 banks, we use the simpler Cobb-Douglas form.

What input-output variables are appropriate depends on the focus of the analysis. The banking efficiency literature focuses on cost and profit (Berger, A. & D. Humphrey, 1997) while the IT productivity literature looks at revenue. We follow the banking literature which predominantly views banks as financial intermediaries that enable the matching of borrowers and lenders. A standard set of input-output variables is used in this literature (Berger, A. & L. Mester, 1997; Rice, T, 2003) and we follow this standard. Hence, the input-output variables are: Quantity or Price of Labor, Quantity or Price of Funds, Quantity or Price of Loans, Quantity or Price of Securities, Physical Capital and Financial Equity Capital. Whether quantities or prices are used depends on whether the cost, revenue or profit function is evaluated (further details are provided in the data section).

Since our interest lies in whether the profit, revenue and cost functions of a bank are influenced by internet banking, we include IB measures related to each of our research questions as explanatory "shift" variables that move the three functions in or out depending upon whether the IB measure has a positive or negative effect.

The influences of potential temporal effects (due to the use of time series data) and bank heterogeneity (due to the use of cross-sectional data) are modeled explicitly through the use of time and market size covariates. Heterogeneity also is dealt with functionally through the use of a random effects model. (A fixed effects model generates results similar in nature to those of the random effects model). The random effects model views the intercept as a random variable with each particular bank's intercept (which remains constant over time) being a random draw from a typically normal distribution (Greene, W, 2003; Hsiao, C, 2003; Wooldridge, J, 2002).

In sum, the profits, revenues or costs of a particular bank i at time t depend on an intercept term (that is bank-specific due to the random effects formulation), a standard set of input-output variables, a set of IB-related shift variables, time and market size covariates and an random error term. In particular, the dependent variable (profit, revenue or cost) is expressed as:


Where [int.sub.i] is distributed normal with mean [] and standard deviation []. An exponential form is used for the IB-related variables since nearly all of them are binary (0/1) in nature and the empirical analysis utilizes the natural log of equation (1).

After a log transformation of (1), the average relationship between the dependent variable and the explanatory variables is:


We estimate (2) using standard maximum likelihood procedures with heteroskedasticity-robust standard errors (Wooldridge, J, 2002). It follows that our IB-related research questions are tested by finding out whether or not the [[alpha].sub.m] parameters of the IB shift variables for the profit, revenue and cost functions are statistically significantly different from zero.

Note that a number of previous banking productivity papers (Berger, A. & D. Humphrey, 1997; Berger, A. & L. Mester, 1999; Rice, T, 2003) investigate an extremal or frontier relationship rather than the "average" relationship developed above. Through Data Envelopment Analysis or Stochastic Frontier Analysis, these papers typically focus on identifying the relative efficiencies of decision making units (e.g., banks). For our data, the stochastic frontier parameter estimates are quite similar to those reported in this paper. Because identifying the efficiencies of particular banks is not the focus of this paper, we describe and estimate the simpler "average" relationship discussed above.


To investigate our questions of interest we utilize a combination of public and proprietary data for 275 banks over a fourteen quarter time span (the first quarter of 1998 to the second quarter of 2001). The proprietary data for these banks was obtained from an application service provider offering a broad "product line" of banking services that included IB. These data consist of quarterly measures pertaining to whether or not a particular bank offered IB at that time and, if so, how many IB accounts it had. The publicly available data for these banks are taken from the Reports of Income and Condition (Call Reports) that all banks are required to submit to the FDIC. They are available at the FDIC website. These data include quarterly profit, revenue and cost figures for each bank. The input-output variables are also financial in nature and gleaned from this publicly available source as is the market size covariate. Descriptive statistics are provided in Tables 1 and 2.

The 275 banks analyzed were chosen because they all were clients of the same application service provider that provided the IB data. While not randomly selected, these banks provide a reasonably representative sample from the approximately 7,000 small banks and 9,000 credit unions in the US. The banks analyzed are located in 39 states and cover the full gamut of locales--urban (e.g., New York City), rural (Mebane, North Carolina), small towns (Danville, Virginia) and suburbs (Fairfax, Virginia). Bank size is small with assets ranging from approximately $13 million to $10 billion and the number of direct deposit accounts (DDAs) and savings accounts ranging from 440 to 540,000. Only two of the banks are savings banks--offering only savings accounts and no checking services. While only 35% (97) of the banks in our sample offer IB, all that do utilize IB use the same system, acquired from one of the leading IB service providers. This precludes a potentially key confounding factor--variations in the impact of IB across banks due to functionality differences. No pure e-banks are analyzed.

While the banks analyzed are small, together small financial institutions hold over a trillion dollars in assets and, hence, constitute a significant business. In addition, to stay competitive with the major money center banks, they must focus on their customer base and offer a competitive set of services. Due to their small size, they also cannot easily absorb losses from poor investments. In such, they provide a window on how other small and medium sized service businesses can compete with larger institutions.

Quarterly data act to both help and hinder estimation. On one hand, statistical significance is improved since the number of observations is increased. More importantly, quarterly data allow more precise measurement of when a bank adopts IB and how quickly customers adopt it. Due to the rapid adoption and change rates insipient to all things internet-related, a good deal of information would be lost with yearly data. On the other hand, the quarterly data are inherently more volatile than annual data causing model fit to degrade and making statistically significant findings concerning IB less likely. A standard technique with such time series data is to account for potential time effects through the inclusion of time-related variables. We allow maximum flexibility through the use of thirteen binary quarter variables--one for each quarter (Q2 through Q14). A binary variable for the first quarter of 1998 (Q1) is excluded to avoid identification problems.

The cross-sectional nature of our data causes additional "noise" that must be accounted for. This bank heterogeneity necessitates using a random effects model rather than a simple regression model. Furthermore, potential business environment differences due to market size differences are explicitly modeled using a covariate. Market size is measured using a Metropolitan Statistical Area (MSA) binary variable. This is part of the FDIC demographic data on banks and classifies a bank as being located in a metropolitan area (MSA=1) or not (MSA=0).

All fourteen quarters of data are available for 154 of the banks. The other banks have shorter time periods due to mergers, failures and start-ups during the 3 1/2 year period analyzed. Since the estimation techniques used are appropriate for unbalanced panel data, we use the entire data set.

We follow the banking literature (see (Berger, A. & D. Humphrey, 1997) for a survey) in defining quarterly bank profits, revenues and costs as well as the input-output variables particular to each. Cost equals operating cost plus interest expense (including the cost of purchase funds and deposits). Revenue includes income from loans and securities plus fee (non-interest) income. Profit is revenue minus cost. Since the profit, revenue and cost data cover a significant time span, we adjust for changes in the value of the money (inflation). We use the Consumer Price Index (CPI) for this. These definitions view the bank as a financial intermediary. In such, the bank's primary role is as a broker between fund providers and borrowers. Labor and any account that provides funds (e.g., deposits and purchased funds) are considered inputs. Any account that uses funds (e.g., loans and purchased securities) is considered an output (Berger, A. & D. Humphrey, 1997; Hancock, D, 1985).

Correspondingly, the standard input-output variables below are used (Berger, A. & L. Mester, 1997; Rice, T, 2003):

Profit is a function of: Price of Labor, Price of Funds, Price of Loans, Financial Equity Capital and Physical Capital;

Revenue is a function of: Amount of Labor, Amount of Funds, Price of Loans, Financial Equity Capital and Physical Capital;

Cost is a function of: Price of Labor, Price of Funds, Amount of Loans, Financial Equity Capital and Physical Capital.

The specific definition of each variable follows the banking efficiency literature (Berger, A. & L. Mester, 1997; Rice, T, 2003). The Price of Funds is defined as the expense for funds divided by the Amount of Funds, which includes core deposits, domestic transaction accounts, time and savings accounts plus the expense for purchased funds including jumbo CDs, foreign deposits, federal funds purchased and all other financial liabilities. The Price of Labor is salary expense divided by the Amount of Labor, which is the number of full-time equivalents. The Price of Loans is the income from loans and securities divided by the Amount of Loans, which includes business loans, consumer loans, real estate loans and securities. Financial Equity Capital is stock and retained earnings. Physical Capital includes buildings and other fixed assets. These variables also are adjusted by the CPI to make them comparable over time.

Eleven IB-related variables are utilized to capture the possible impact of internet banking on profits, revenues or costs. The first variable is a binary variable identifying simply whether or not bank i offers IB during quarter t. If a bank has an IB system in place, then [] = 1; otherwise it equals zero.

The next three IB measures are used to assess the potentially non-linear relationship between a bank's profits, revenues and costs and the number of bank customers using internet banking at time t ([]). [] is defined as the number of IB accounts in existence at time t. For modeling simplicity we estimate a piecewise rather than continuously differentiable relationship. This piecewise estimation style is commonly utilized. Piecewise relationships between the dependent and independent variables where the data dictates the breakpoints are central to all Data Envelopment Analysis procedures (Horsky, D. & P. Nelson, 1996; Seiford, L. & R. Thrall, 1990). Data-dictated breakpoints are also utilized to identify when scale economies kick in (White, L, 1971; White, L, 1982).

A grid search along with the maximum likelihood estimation was used to identify the levels of consumer adoption at which one line segment ended and the next line segment began. This follows work done concerning scale economies in the automotive industry in (White, L, 1971; White, L, 1982). In such, we utilize three variables to reflect the impact of increased consumer adoption of IB on bank profits, revenues and costs. The first variable [[] <=X] equals simply [] when the number of adopters is less than or equal to X and equals X if the number of adopters exceeds this amount. The second variable [X<[] <=Y] equals 0 if [] <=X, equals [] - X if the number of adopters exceeds X and is less than or equal to Y, and equals Y if the number of adopters exceeds Y. The third variable [[] >Y] equals [] - Y if the number of adopters exceeds Y and zero otherwise. Note that the breakpoints X and Y need not be the same for revenues, costs and profits.

The fifth through tenth IB variables are interactions of the above three adoption level variables with time. These variables act to make the impact of [[] <=X], [X<[] <=Y] and [[] > Y] on profit, revenue or cost differ depending upon the time period analyzed. In such, they determine if the impact of IB changes over time. Given that the first implementation of IB occurred in Q4 (between October and December of 1998) and quarterly time dummies are already modeled in order to account for "random" temporal variation, "fiscal year" breakdowns FY1, FY2 and FY3 are utilized. Defining [] =1 if 7<=t<=10, and zero otherwise, and []=1 if 11<=t<=14, and zero otherwise, our six interaction terms are: [[] <=X*[]], [X<[] <=Y*[]], [[] > Y*[]], [I[] <=X*[]], [X<[] <=Y*[]], and [[] > Y*[]]. For identification purposes, FY1 (Q3-Q6) is not included.

The final IB variable is used to examine whether an early-mover advantage exists. For this, we measure when a bank implemented internet banking relative to its peers. A binary [] variable equals one for those banks that installed internet banking prior to March 31, 1999 (prior to Q6), and zero otherwise. Seven banks installed IB in Q4 and twelve more installed IB in Q5. For these innovator banks, the innovator variable equals one only for those quarters when the bank has IB. These 19 innovator banks are among the first five percent of banks to adopt IB in the nation and the first seven percent of our sample.

Given the above definitions, our profit function is explicitly (cost and revenue functions are similar):


This model assumes all banks have the same basic profit (cost, revenue) function. That is, the basic structure--the impact of the price of labor, the price of funds, the price of loans (or the amount of labor, funds and loans) and the amount of financial equity capital and physical capital--is the same for all banks. Implicitly, banks choose the level of each variable to optimize profit (cost, revenue). However, for quasi-fixed variables such as physical capital there is a cost associated with changing their usage rate. Hence, a bank does not adjust them continuously. Hitt and Brynjolfsson (Hitt, L. & E. Brynjolfsson, 2000) note that if these variables are not in equilibrium, their parameter estimates may be biased. However, the function shifts over time (due to the quarterly time variables) and for each bank (due to the bank-specific intercept terms). Our interest lies in the shifts in the model that occur due to the IB variables.


The equation (3) random effects model results for profit, revenue and cost are presented in Table 3. To simplify presentation, only the IB variables that are significant at the 5% level or better are included. For each of the three models the fit is statistically superior at better than the 1% level to a model that excludes the IB variables. Face validity also is strong as the signs of the five input-output parameters are consistent with those found in the banking productivity literature. This section discusses each of our four hypotheses in turn. It closes with a discussion of possible alternative explanations.

H1: Installation of Internet Banking

Since the [[alpha].sub.HIB] parameter is statistically insignificant in all three models, the simple existence of IB at a bank has no impact on profits, revenues or costs (i.e., we cannot reject H1). For revenues this is unsurprising since revenues only accrue through customer usage. The use of an IB service provider minimizes the fixed costs associated with offering IB, so costs also are not significantly impacted. It follows that no measurable effect on profit is found.

H2: Consumer Adoption of Internet Banking

Profits, revenues and costs, however, are significantly affected by the number of IB users a bank succeeds in acquiring (thus rejecting H2). The stylized piecewise linear relationship is found for each of the three models. The piecewise function's breakpoints X and Y are 3,300 and 6,000 for profits and 3,100 and 6,000 for costs. For revenues, a single breakpoint occurs at 3,500. This piecewise function approximates a continuously differentiable function. Consequently, true profits, for example, do not increase strictly linearly with each additional consumer adopter until 3,300 consumers adopt, and the 3,301st adopter does not have exactly zero effect on profit. Similarly, the levels of X and Y are based on statistical criteria. Hence, the cost and revenue saturation breakpoints X vary a bit from that for profit. However, there is little statistical or economic difference between the models reported and ones that restrict the breakpoint X for all three models to be the same. Finally, the dollar impacts that we report are to show reasonableness. We do not expect that these are the exact impacts that banks have seen or will see from implementing IB and getting their customers to adopt IB. What is important is the direction of the changes and the relative shapes of the curves.

The positive sign for the [[] <=X] parameter [[alpha].sub.<=X] in all three models indicates that an increase in the level of consumer adoption increases revenues, costs and profits until their particular "saturation" level X is achieved. The statistical insignificance of the [X<[] <=Y] parameter [[alpha].sub.<X<=Y] in all three models implies that additional adopters in excess of the respective saturation level X do not add further to bank profit, revenue or cost until the second breakpoint Y is possibly reached. The cost and profit models experience a change at this second breakpoint but revenue does not. In the cost model, the negative value for the [[] > Y] parameter [alpha] [[alpha].sub.>Y] indicates that after enough customers adopt IB (6,000), a bank's costs begin to decrease. Correspondingly, profits trend upwards after this point.

Note that reaching these IB user adoption level breakpoints constitutes a major undertaking for the vast majority of studied banks. For 25% of the banks, reaching 3,100 adopters requires over half of their customers to adopt IB. For another 25%, this requires a customer adoption level exceeding 25%. In fact, only eight banks reach the 3,100 adopter level during the time span analyzed, and they take, on average, 6.67 quarters to do so. Four banks reach an adoption level greater than 6,000 customers. In fact, for almost half (125) of the banks this second breakpoint exceeds their customer base. For twenty of the banks, however, this is less than 20% of their customer base. Of these "larger banks," only eleven implemented IB, so we have a balance of larger banks that did and did not adopt IB. (These percentages use the number of direct deposit accounts to approximate the number of bank customers). This relatively high saturation level coupled with the distribution of the banks based on size points towards a possibility that the larger banks in our sample drive the estimation results. However, the discussion at the end of this section provides evidence against this possibility.

The cost estimation results imply that for a median sized bank quarterly costs increase between $16 and $164 (with a mean of $91) for each customer who adopts IB until the saturation level of adopters X is met. (This range results from inputting the median quarterly cost of $3,037,931 and the endpoints of the 95% confidence interval for [[alpha].sub.<=X] into the partial derivative of the cost function with respect to [[] <=X] and setting it equal to [[alpha].sub.<=X] time Cost.) This means that for the median bank with 1,109 IB users (the average number of adopters in Q14) costs are between $17,735 and $181,930 higher per quarter than if the bank did not offer IB, with an average cost increase of $100,070. Note, however, that this impact of IB on a bank's total costs is relatively small. This increase corresponds to just 3.3% of the median bank's total cost. Also note that the increased cost per IB account is not tied solely to the fees charged the bank by the IB service provider. This additional cost also includes additional operating expenses such as IB marketing and increased customer service, as well as added interest payments on new or enlarged DDAs. For those banks that exceed the 6,000 adopter level Y, costs start to decrease at slower rate than at which they had earlier increased. This decrease for a median bank is between $13 and $91 (with an average of $52) per additional IB user per quarter.

The initial positive impact that consumer adoption of IB has on costs arises because all four of the key cost drivers point towards higher costs. First, there is a small fixed cost to operating the IB system. Second, as discussed previously, the basic wholesale pricing structure for IB services includes a base fee, a fee per customer and transaction fees for certain types of transactions (e.g., bill payment). Each bank negotiates the fee structure with the IB service provider, but given this fee structure, the expected cost savings from significantly lower per transaction costs (Dynamicnet, 2001; Goldfinger, C., 2003) are unlikely to materialize. Third, banks that implement IB perform significantly more transactions. For example, their growth rate in the number of accounts is 14% higher for DDAs and 32.8% higher for savings accounts. In addition, IB is associated with a larger average deposit size per account of about $800 at a median bank. Coupled with the work by Hitt and Frei (Hitt, L. & F. X. Frei, 2002) and findings reported by Woodford (Woodford, R, 2001), this indicates that while new customers are drawn to the bank, existing customers also are doing more business with their bank. This increase in bank activities necessarily increases costs. Clearly previous bank customers who utilize IB increase their use of bank services and/or IB attracts new customers to the bank. Furthermore, not all these bank services are transacted through IB. Last, despite this increase in transactions, the actual volume of transactions performed using IB relative to more traditional transactions is typically quite small. For example, on average, in our sample, IB bill payment transactions are only two tenths of one percent of the DDA transactions. Correspondingly, reductions are unlikely in the fixed infrastructure costs related to traditional transactions (e.g., the number of tellers employed or ATMs operated) unless a large number of customers adopt IB.

The decrease in costs after the 6,000 user adoption level likely results from changes that occur in the last two of these four cost drivers as customer adoption of IB becomes substantial. With a significant number of IB transactions relative to traditional transactions the likelihood that IB services replace rather than augment ATMs, tellers and so on increases. A simple check indicates that the number of accounts per full-time equivalent employee is larger when the number of IB adopters exceeds 6,000. Additionally, the diffusion of innovations concept discussed earlier kicks in. That is, later consumer adopters of IB (the early and late majorities and the laggards) are more likely to use fewer IB services for which the bank is charged and non-IB services which the bank must provide than are the "consumer innovators" that constitute the early consumer adopters (Horsky, D, 1990).

The revenue and profit results strongly reflect the diffusion of innovations argument. The first consumer adopters of IB do appear to be more profitable. The plateaus at 3,300 adopters for profit and 3,500 for revenue reflect that incremental gains in these measures dissipate as the number of adopters swells. Until these adopter breakpoints, with each additional IB consumer the median bank experiences a quarterly revenue increase of between $4 and $185 with an average of $93. The corresponding profit increase ranges from $44 to $124 with an average of $83. Note that the estimated incremental profit figures fall within the combined range of the estimated incremental revenues minus the incremental costs.

The additional increase in profits related to adoption levels exceeding the second 6,000 adopter breakpoint results from the decreased costs discussed above. In fact, the estimated cost reduction and profit increase are very similar. In particular, at the median bank each adopter in excess of 6,000 corresponds to a per quarter profit increase of between $31 and $99 with an average of $65.

As with cost, it is important to note that the impact of IB usage on revenue and profit is relatively small. For example, the median quarterly revenue per DDA is $765 ($3,855,548/5,040 DDAs), the median cost per DDA is $603 and the median profit per quarter per DDA is $162. Additionally, at the Q14 average level of IB adoption (1,109) revenues are on average increased by $102,619, which account for just 2.7% of median revenues. To put this in perspective, note that a single IB adopter taking out a $200,000 mortgage at 6% interest increases the bank's revenues by $3,000 per quarter. Hence, only a few additional such customers are needed to drive the revenue numbers. As for profits, in approximately 60% of the bank-quarters where IB was implemented the bank that adopted IB has an expected profit increase of less than 5%. In fact, for the few banks that do achieve a user base large enough to generate a 10% increase in quarterly profits, it took, on average, 5.46 quarters to reach this level. On the other hand, while the profit impact is small in relative size, in concordance with Hitt and Frei's (Hitt, L. & F. X. Frei, 2002) findings, IB users are significantly more profitable than the average bank customer.

Four additional models are used to take a deeper look at the sources of the revenue and cost increases. These models, reported in Table 4, use fee revenue, interest revenue, operating expenses and interest expenses as their dependent variables and the same independent variables as in the original revenue or cost models. For both the fee revenue and operating cost, we tried models that included variables for the number of accounts. Since these variables did not affect the results, we have not included a detailed discussion of them. As shown in Table 4, the significant [[alpha].sub.<=X] parameter implies that interest revenue, like total revenue, increases up to a saturation level X = 3,200 after which additional IB adopters have no measurable impact on interest revenue. Specifically, for the first 3,200 adopters at the median bank, interest revenue increases between $15 and $104 per quarter (with an average of $59) with each adopter.

This large increase in interest revenue is coupled with a much smaller increase in fee revenue. While no saturation level exists for fee revenue, the incremental impact of consumer adopters on fees is minimal (the [[alpha].sub.<=X] parameter is very small but statistically significant and we found no level X where this relationship changed). The [[alpha].sub.<=X] parameter value implies that, for the median firm, each adopter is associated with between $2 and $12 of additional fee revenue per quarter (with an average of $7). This figure constitutes less than 10% of the estimated impact on revenue and in combination with interest revenue roughly coincides with the total revenue increase estimates. Because the impact of fee revenue is so small, after the total revenue saturation level of 3,500 users it likely becomes "lost" in the random noise when aggregated into the total revenues for banks with high adoption rates.

The two additional cost models reported in Table 4 show a meaningful relationship between IB consumer adoption and both higher interest expenses and higher operating expenses. For interest expense, costs increase up to 2,900 IB users. The [[alpha].sub.<=X] parameter estimate implies that interest costs increase between $6 and $88 (with an average of $47) per IB user. This is due to an increase in interest bearing accounts and account balances. As noted earlier, banks with IB have higher deposits per customer, and a IB adopter putting just an additional $10,000 into a bank account with 3% interest generates $75 in additional interest expenses per quarter.

Operating expenses also show an initial increase of between 50[cents] and $46 per quarter (with a mean of $23) per IB user up to 6,000 users. This operating expense estimate appears reasonable given that the average total fee per user actually paid by a bank to the IB service provider is $21.61 per quarter, which does not include operating costs associated with account management, loan processing or ATM fees. A key finding is that these additional operating costs appear to decrease when a high number of customers (more than 6,000) adopt IB. The rate of this operating cost decrease per IB user is between $36 and $107 (with an average of $71) for a median bank. This is consistent with the rate of decrease found for the overall cost, as is the level at which the costs start to decrease.

In sum, all else equal, banks offering internet banking are more profitable than those that do not. However, when weak marketing efforts result in limited consumer adoption of IB, very limited gains are realized relative to overall bank profits. Improved profits initially result from increased revenues despite increased costs. These improved revenues are the result of an increase in the bank's interest income through soliciting new customers and getting more business from previous customers as well as more fee-based bank services being sold. Furthermore, at high consumer adoption levels banks do begin to realize some substitution of less expensive internet-based services for more traditional ATM and teller-based services. The resulting reduction in operating costs drives a revived profit upswing.

H3: Internet Banking Early Movers

Our empirical results do not refute H3--being an innovator has no impact on bank profits, revenues or costs (the [[alpha].sub.innovator] parameter is not statistically significant in any model). The sourcing approach--a vendor supplied and managed IB system--provides a likely partial explanation. The higher costs and operational risks typically associated with innovation are mitigated by the use of a vendor, and banks that use this service are unable to physically distinguish themselves from other banks using the same service.

The key explanation for finding no early-mover advantage is likely that the standard definition of an early-mover is ill-suited to this industry. Being in roughly the first 5% of banks to adopt IB nationally is not really what is relevant when thinking about an early-mover advantage. Rather, most banks, especially the small banks analyzed, are regional in nature and so what is really relevant is how early they adopt IB relative to the other banks that do business in their service area. The insignificance of the [[alpha].sub.innovator] parameter means that this competitive adoption issue has not yet become an issue. In fact, the statistics previously reported in both this paper and Hitt and Frei (Hitt, L. & F. X. Frei, 2002)that point toward increased revenues being largely due to the bank acquiring a larger share of their previous consumers' financial business indicate that, perhaps, this competitive pressure may never become strong.

H4: Internet Banking Impact Over Time

Similar to our findings concerning banks that are early adopters, the impact of IB on profits, revenues and costs is not found to change over time (H4 cannot be rejected). First, in the profit, revenue and cost models, the six [fiscal year * number of consumer adopters] interaction term parameters [[alpha].sub.<=XFY2], [[alpha].sub.<=XFY3], [[alpha].sub.<X<=YFY2], [[alpha].sub.<X<=YFY3], [[alpha].sub.>YFY2,] and [[alpha].sub.>YFY3], are not statistically significant. Second, time alone (other than a subset of the quarterly dummy variables Q2-Q14) has no statistically significant impact even though it is inherently correlated with the number of IB users since the cumulative number of IB customers invariably must increase. Even when the three IB user variables (e.g., [[] <=X]) and the six interaction terms are replaced simply by the [] and [] binary variables, these time variables remain statistically insignificant.

This finding that revenues, costs and profits per IB user do not change over time implies that increased competition over time is not forcing banks to pass on more benefits to customers and adds further support for a lack of an early-mover advantage. This finding is good news for the banks that are considering whether or not to offer IB--a profit opportunity exists.


A key question with this research is whether the results presented are due to IB's impact or to some other factor that is correlated with IB. One possibility is that IB implementation and use are correlated with bank size and that larger banks are more efficient (i.e., have higher profits, higher revenues and lower costs). Two separate analyses show this is not the case. To investigate potential spurious correlation, we normalized profits, revenues, costs, labor, funds, financial equity capital and physical capital by dividing them by bank assets and reran the analyses. Results similar to those reported in the paper arise (with slight differences in when the breakpoints occur).

As an alternative approach to investigating the impact of bank size, we limited the analysis to only those banks in the top half of the sample in terms of their assets. Again, very similar results are found with the primary difference being that the cost increase at low levels of IB adoption is significant at only the 9% level. When we limit the analysis to only the smaller half of the banks we find that these banks don't attract enough IB customers to reach a breakpoint. Correspondingly, we find profits, revenues and costs increase with more IB use but never reach a point where they level off. Finally, we performed the analysis after removing the ten largest banks. Again, the results are similar to those reported in the paper. Taken together, these sub-sample results indicate that the full sample results presented earlier are valid--the overall structure of the impact of implementing IB and getting customers to adopt it appears robust.

Another possibility is that banks with richer more lucrative customers offer IB and this not IB causes these banks to be more profitable, have higher revenues and have lower costs. Simple correlation analyses disprove this notion. To examine this clientele explanation, we looked at assets per deposit account (ApDA) across banks. Two methods were used to define ApDA. The first divided total assets by the total number of DDA and savings accounts. Since a customer may have multiple accounts, this implicitly understates ApDA. Hence, the second method defined ApDA as assets divided by just the number of DDAs. For the two banks that did not have DDAs, we used their number of savings accounts. For each measure, its correlation with both IB adoption and the number of IB users is small and negative (between 0.00 and -.02).

As with all research, there are limitations to this study. The implications of this analysis are limited by the fact that our sample banks are small. The benefits of IB to large institutions like Wells Fargo and Bank of America are likely to differ in magnitude but not in direction. The smaller size of the banks we analyze points towards an adoption level of around 3,500 customers where additional IB customers become fairly "average" in their profit appeal. For larger banks, given their installed customer base and greater mass appeal, this saturation level is likely to be higher. In addition, incremental costs are still likely to rise at first. The increased number of transactions and incomplete substitution of traditional transaction-based fixed costs remain appropriate, and while these larger banks are likely to witness variable transactions cost reductions, they are also more prone to build their own IB system, thereby incurring greater fixed costs. As with our smaller banks, at higher IB adoption levels more substitution from ATM and teller servicing is likely and, hence, costs should decrease and profits increase. In sum, larger banks with greater fixed costs and a larger customer bases also are likely to exhibit a nonlinear relationship between profits (costs or revenues) and the number of their customers using IB. The "breakpoints" in this relationship are likely, however, occur at higher usage levels.

The time horizon of the study also is limited to the first few years of IB availability. It is possible that as more banks adopt IB, competition for the lucrative IB customers will drive the revenue and profit advantages of IB downwards. Alternatively, it also is possible that with enough consumer adoption and with time to adjust how banks deliver services, the often forecasted decrease in costs may be realized. On the other hand, the finding that most of the incremental revenues and costs associated with IB are due to a deepening of the bank's business relationship with its installed customer base mitigate concerns over this point.


This paper explores whether and how implementing internet banking impacts bank profits. By focusing on a single type of IT system within an industry with well documented and standardized profit, revenue, cost and input-output variable measures, we were able to perform a detailed econometric analysis of the impact of internet banking and the drivers behind increased profits. The level of analysis is most appropriate for managers, since their IT investment decisions are by nature industry and technology specific. In such, our results provide insights into the impact of other customer-oriented information systems on firm performance, especially in an environment where small and medium sized business must compete aggressively with much larger players.

The paper's key finding is that IB is a desirable opportunity for banks and that the key to success is customer adoption. At first, the benefits of internet banking do not come from lower costs as is often predicted with information technologies, but, rather, come from higher revenues driven by growth from the more lucrative customer demographic profiles. Since the fixed costs associated with internet banking through a service provider are low, our results show even low levels of customer adoption allow an expected profit increase. Cost reduction does, however, have a significant impact on profits at sufficiently high levels of consumer adoption. Once the level of internet-based transactions reaches a sufficiently large level substitution away from more costly ATM and teller-based transactions becomes feasible. Hence, profits originally increase as consumer adoption of IB grows. Profits then flatten as these adopters become less lucrative until a relatively large amount of IB usage takes place allowing operating costs to decrease and profits again to increase. It appears that internet banking, even if not well marketed to consumers, does NOT resemble a money pit. It can, however, become a small gold mine if properly and aggressively marketed.

Two secondary findings also are reported. First, whether or not a bank is an early-mover when it comes to implementing IB has no impact on its profits other than that the increased profits derived from IB are reaped over a longer period. The regional nature of most banks implies that a bank should be compared to its regional rather than national competitors when thinking about such an issue. Furthermore, since much of the benefit derived from IB is due to a deepening of the bank's relationships with its installed customer base, the importance of an early-mover advantage is secondary. Similarly, the benefits derived from IB are not found to dissipate over time. The lack of competitive pressure just mentioned again drives this finding.


US IT spending grows in 2003-2004, CIO Magazine, November 8, 2002.

IT spending grows .2 percent in 01, CIO Magazine, April 12, 2002.

Transforming Consumer Banking Through Internet Technology, Dynamicnet, White Paper, (2001).

Alpar, P & M. Kim (1990). Microeconomic Approach to the Measurement of Information Technology Value. Journal of Management Information System, 7, 55-69.

Barua, A., C. Kriebel, & T. Mukhopadhyay (1995). Information Technologies and Business Value: An Analytic and Empirical Investigation. Information Systems Research, 6, 3-23.

Berger, A. & D. Humphrey (1997). Efficiency of Financial Institutions: International Survey and Directions for Future Research. The Wharton Financial Institutions Center Working Paper 97-05, Philadelphia, PA 19104.

Berger, A. & L. Mester (1997). Inside the black box: What explains difference in the efficiencies of financial institutions? Journal of Banking & Finance, 21, 895-947.

Berger, A. & L. Mester (1999). What Explains the Dramatic Changes in Cost and Profit Performance of the U.S. Banking Industry, The Wharton Financial Institutions Center, University of Pennsylvania, Working Paper No. 99. (1999), Philadelphia, PA 19104.

Brynjolfsson, E. & L. Hitt (1999) Paradox Lost? Firm-Level Evidence on the Returns to Information Systems Spending. In L. P. Willcocks, and S. Lester (Eds.), Beyond the IT Productivity Paradox, John Wiley & Sons.

Greene, W (2003). Econometric Analysis (Fifth Ed.). Upper Saddle River, New Jersey: Prentice Hall.

Goldfinger, C. (2003). Internet Banking Issues Paper, Financial Internet Working Group, White Paper. Hancock, D (1985). The Financial Firm: Production with Monetary and Nonmonetary Goods. Journal of Political Economy, 93, 859-880.

Hitt, L. & E. Brynjolfsson (2000). Computing Productivity: Firm-Level Evidence, Working Paper.

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

Horsky, D (1990). Diffusion Model Incorporating Product Benefits, Price, Income and Information. Marketing Science, 9, 342-365.

Horsky, D. & P. Nelson (1996). Evaluation of Sales Force Size and Productivity Through Efficient Frontier Benchmarking. Marketing Science, 15, 301-320.

Hoffmann, K (2003, March). Online Banking Aligns Practices. Banking Technology News.

Hsiao, C (2003). Analysis of Panel Data (Second Ed.). Cambridge, UK: Cambridge University Press.

Landuaer, T (1995). The Trouble with Computers: Usefulness, Usability and Productivity. Cambridge, Massachusetts: MIT Press.

Licht, G. & D. Moch (1999). Innovation and information technology in services. Canadian Journal of Economics, 32(2), 363-383.

Lichtenberg, F (1995). The Output Contribution of Computer Equipment and Personnel: A Firm-level Analysis. Economics of Innovation and New Technology, 3, 201-217.

Olazabal, N (2003). Banking, the IT Paradox. McKinsey Quarterly, 1.

Orr, B (2001). E-Banking What Next? American Bankers Association ABA Banking Journal, 93, 12, 40-43.

Parton, C. & J. Glaser (2002). Myths About IT Spending. Healthcare-informatics.

Porter, M (1998). Competitive Advantage, New York, NY; Free Press.

Reich, B. & I. Benbasat (1990). An Empirical Investigation of Factors Influencing the Success of Customer-Oriented Strategic Systems. Information Systems Research, 1, 325-347.

Rice, T (2003). The Importance of Payments-Driven Revenues to Franchise Value in Estimating Bank Performance, Working Paper, Federal Reserve Bank of Chicago.

Rogers, E (1995). Diffusion of Innovations, 4th ed. New York, NY: The Free Press.

Ross, J. & P. Weill (2002). Six IT Decisions Your IT People Shouldn't Make. Harvard Business Review, November 2002, R0211F.

Seiford, L. & R. Thrall (1990). Recent Developments in DEA: The Mathematical Programming Approach to Frontier Analysis. Journal of Econometrics, 46, 7-38.

Steiner, T. & D. Teixeira (1990). Technology in Banking: Creating Value and Destroying Profits. Homewood, Illinois: Business One Irwin.

Sullivan, R (2000). How Has the Adoption of Internet Banking Affected Performance and Risk in Banks? Financial Industry Perspectives, 1-16.

Vrechopoulos, A., & G. Siomkos (2002). The Adoption of Internet Shopping by Electronic Retail Consumers in Greece: Some Preliminary Findings. Journal of Internet Banking and Commerce.

Banking in the on-line World. Retail Banking Research, Ltd., White Paper, (1996).

White, L (1971). The Automobile Industry Since 1945. Cambridge, Massachusetts: Harvard University Press.

White, L (1982). The Automobile Industry in The Structure of American Industry, (W. Adams ed.) New York, New York: MacMillan Publishing Co, Inc.

Woodford, R (2001). Popularity of Internet Banking is on the Rise. Birmingham Business Journal, June 4, 2001.

Wooldridge, J (2002). Econometric Analysis of Cross Section and Panel Data, Cambridge, Massachusetts: The MIT Press.

Paul Nelson, University of Rochester

William Richmond, Western Carolina University
Table 1: Quarterly Summary Statistics: Public Data for the 275 Sampled

Variable Minimum 1st Quartile Median

Profit $2,071 $373,606 $809, 881
Revenue $229,083 $1,974,532 $3,855,548
Cost $198,973 $1,560,186 $3,037,931
Price of Labor $3,358 $10,800 $12,835
Amount of Labor 3 31 54
Price of Funds 0.4% 0.91% 1.05%
Amount of Funds $10,697,180 $71,026,220 $135,524,000
Price of Loans 0.2% 1.9% 2.0%
Amount of Loans $9,527,000 $88,733,500 $188,576,600
Financial Eq. Cap. $1,067,000 $10,898,160 $19,560,830
Physical Capital $59,000 $1,362,691 $2,957,528
Num of DDAs 1 2,013 5,040

Variable 3rd Quartile Maximum

Profit $1,712,056 $63,831,700
Revenue $8,030,471 $242,237,400
Cost $6,242,663 $190,320,000
Price of Labor $14,847 $67,071
Amount of Labor 123 2446
Price of Funds 1.19% 3.2%
Amount of Funds $273,948,700 $9,635,581,000
Price of Loans 2.2% 5.5%
Amount of Loans $388,672,600 $9,524,253,000
Financial Eq. Cap. $42,516,090 $719,323,000
Physical Capital $7,129,025 $172,556,000
Num of DDAs 10,429 249,155

Table 2: Quarterly Summary Statistics
Proprietary Data for the 275 Sampled Banks

Number of Banks by Customer Adoption Level and Quarter

Number of
IB Adopters Q1 Q2 Q3 Q4 Q5 Q6 Q7

1-99 0 0 0 7 12 25 18
100-199 0 0 0 0 4 2 5
200-299 0 0 0 0 1 1 6
300-399 0 0 0 0 2 1 2
400-499 0 0 0 0 1 0 4
500-749 0 0 0 0 0 1 3
750-999 0 0 0 0 0 1 3
1000-1249 0 0 0 0 0 0 0
1250-1499 0 0 0 0 0 0 0
1500-1999 0 0 0 0 0 0 0
2000-2499 0 0 0 0 0 0 0
2500-3299 0 0 0 0 0 0 0
3300-4999 0 0 0 0 0 0 0
5000-9999 0 0 0 0 0 0 0
10000+ 0 0 0 0 0 0 0

Total Num-
ber of Banks
with IB 0 0 0 7 20 31 41

Number of
IB Accounts
at Banks
with IB na na na 22 105 96 239

Number of
Banks with
or without
IB 210 206 206 198 194 188 187

Number of
IB Adopters Q8 Q9 Q10 Q11 Q12 Q13 Q14

1-99 19 21 19 17 15 9 19
100-199 6 8 10 9 9 11 10
200-299 3 4 8 9 5 4 4
300-399 6 2 5 4 10 9 8
400-499 3 5 4 8 5 5 10
500-749 6 6 8 10 13 13 14
750-999 0 5 7 6 8 5 6
1000-1249 3 1 1 5 5 8 3
1250-1499 1 3 2 2 3 2 9
1500-1999 1 0 2 4 6 5 1
2000-2499 0 2 0 0 1 2 3
2500-3299 0 0 0 0 1 4 4
3300-4999 0 0 3 1 0 1 1
5000-9999 0 0 0 2 3 2 4
10000+ 0 0 0 0 0 1 1

Total Num-
ber of Banks
with IB 48 57 69 77 84 81 97

Number of
IB Accounts
at Banks
with IB 333 412 519 677 819 1082 1109

Number of
Banks with
or without
IB 188 190 193 192 187 188 190

Table 3: Profit, Revenue and Cost Empirical Results

Variable Parameter Std Error t-value Pr > t

 [less than or
 equal to] 3,300 0.000103 0 4.1 <.0001
 > 3,300 0.000081 0 3.69 0
Intercept (mean) -2.0576 0.5384 -3.82 0
Price of Labor -0.6677 0.1058 -6.31 <.0001
Price of Funds -7.5754 17.2419 -0.44 0.6604
Price of Loans 25.4933 9.2870 2.75 0.0061
Financial Equity Capital 0.9583 0.0567 16.81 <.0001
Physical Capital 0.0474 0.0383 1.24 0.2169
MSA 0.1418 0.0702 2.02 0.0436

Q3, Q4, Q11, Q12 are significant and negative; [chi square] test
relative to a model without the IB variables = 19.7 (p < .005)


 [less than or
 equal to] 3,500 0.000024 0 2.05 0.0408
Intercept (mean)t -2.3133 0.4105 -5.64 <.0001
Labor in FTEs 0.0976 0.0354 2.76 0.0059
Total Funds 0.483 0.1284 3.76 0
Price of Loans 13.7711 3.2747 4.21 <.0001
Financial Equity Capital 0.3305 0.0997 3.31 0.001
Physical Capital 0.0689 0.0325 2.12 0.034
MSA 0.0071 0.0274 0.26 0.7957

Q9-Q14 are significant and positive; [chi square] test relative to a
model without the IB variables = 22.8 (p < .005)

 [less than or
 equal to] 3,100 0.00003 0 2.38 0.0173
 > 6,000 -0.000017 0 -2.62 0.0088
Intercept (mean) -3.3855 0.2669 -12.69 <.0001
Price of Labor 0.2199 0.0321 6.85 <.0001
Price of Funds 32.0702 6.1497 5.21 <.0001
Loans 0.6912 0.0479 14.44 <.0001
Financial Equity Capital 0.1207 0.0361 3.34 0.001
Physical Capital 0.1187 0.0182 6.5 <.0001
MSA -0.0195 0.0232 -0.84 0.4003

Q2-Q4, Q7, Q8, Q10-Q12 are significant and positive; [chi square] test
relative to a model without the IB variables = 40.2 (p<.005)

Table 4: Particular Revenue and Cost Empirical Results

Variable Estimate Std Error t-value Pr > t

Interest Revenue

 [less than or equal
 to] 3,200 0 0 2.59 0.01
Intercept (mean) -3.1488 0.1394 -22.59 <.0001
Labor in FTEs 0.0564 0.0337 1.67 0.095
Total Funds 0.7061 0.0467 15.13 <.0001
Price of Loans 13.0547 3.0878 4.23 <.0001
Financial Equity Capital 0.2068 0.0484 4.27 <.0001
Physical Capital 0.0229 0.0169 1.36 0.1754
MSA -0.0061 0.0136 -0.45 0.6547

Q3-Q6 and Q9-Q14 are significant. Q3-Q6 are negative, Q10-Q14 are
positive [chi square] test relative to a model without the IB
variables = 14.5 (p < .005)

Fee Revenue

 [less than or equal
 to] [infinity] 0 0 3 0
Intercept (mean) -5.6876 1.084 -5.25 <.0001
Labor in FTEs 0.5904 0.1344 4.39 <.0001
Total Funds 0.7791 0.2233 3.49 0
Price of Loans -4.0377 6.5141 -0.62 0.5354
Financial Equity Capital -0.3773 0.175 -2.16 0.031
Physical Capital 0.1638 0.0693 2.36 0.018
MSA -0.5431 0.1133 -4.79 <.0001

Q3, Q4 Q9, Q13 and Q14 are significant. Q9 is negative, Q3, Q4, Q13
and Q14 are positive [chi square] test relative to a model without the
IB variables = 98.8 (p < .005)

Interest Expense

 [less than or equal
 to] 2,900 0 0 2.26 0.024
Intercept (mean) -4.4493 0.2209 -20.14 <.0001
Price of Labor 0.0863 0.0267 3.24 0
Price of Funds 40.7503 7.3281 5.56 <.0001
Amount of Loans 0.8581 0.0484 17.72 <.0001
Financial Equity Capital 0.0433 0.0384 1.13 0.2607
Physical Capital 0.0769 0.0163 4.7 <.0001
MSA -0.0315 0.0203 -1.55 0.1219

Q3, Q5, Q9-Q14 are significant. Q5 is negative, Q3 and Q9-Q14 are
positive [chi square] test relative to a model without the IB
variables = 33.9 (p < .005)

Operating Expense

 [less than or equal
 to] 6,000 0 0 2 0.045
[] >
 6,000 0 0 -4.34 <.0001
Intercept (mean) -3.7268 0.3893 -9.57 <.0001
Price of Labor 0.5359 0.0559 9.59 <.0001
Price of Funds 10.1248 7.8898 1.28 0.1995
Amount of Loans 0.4421 0.0533 8.29 <.0001
Financial Equity Capital 0.2077 0.0451 4.61 <.0001
Physical Capital 0.2178 0.0266 8.17 <.0001
MSA -0.0361 0.0444 -0.81 0.4158

Q2-Q6 and Q10-Q14 are significant. Q2-Q6 are positive, Q10-Q14 are
negative [chi square] test relative to a model without the IB
variables = 15.8 (p < .005)
COPYRIGHT 2007 The DreamCatchers Group, LLC
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2007 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Nelson, Paul; William, Richmond
Publication:Academy of Banking Studies Journal
Geographic Code:1USA
Date:Jan 1, 2007
Previous Article:Letter from the editors.
Next Article:Derivative use by banks in India.

Related Articles
Production expected to begin at Taylor Mine.
Top operating mines in Northern Ontario.
Mining trouble.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters