Chapter 8:

Economic and Social Significance of Information Technologies


Impacts of IT on the Economy




Diffusion of IT has had significant effects on business activity. Computer-integrated manufacturing, for example, enables automated model changes on the production line as well as fully integrated design and manufacture. Resulting shortened cycle times and the declining significance of economies of scale have led to a competitive environment that focuses on quality, customization, and timeliness of delivery. Firm-level IT networks ("intranets") integrate finance, manufacturing, R&D, operations, and marketing, and have fostered the rise of strategic management in industry. The IT-based integration of producer-supplier and wholesaler-retailer networks enables responsiveness to daily changes in customer demand and a fundamental revolution in inventory management. Advanced telecommunications technologies have integrated international capital markets and literally created a global financial industry. In short, IT has moved economic markets and business be! havior far closer to "real-time" mode than has ever existed in the past.

Yet in almost all instances, the precise economic impacts of these effects cannot be quantified, and there is often contradictory evidence about the role of IT. For example, research (reviewed below) shows that IT has contributed to both deskilling and skill upgrading in the workplace, although the trend appears to be toward upgrading. Until very recently, the empirical record demonstrated that, in spite of the enthusiastic adoption of IT by business, IT has had little observable impact on productivity growth in the United States (a paradox explored further below).

This section summarizes quantitative indicators of the economic effects of IT in three core areas of interest:

The findings indicate that IT has diffused unevenly throughout the economy and that the net impacts on employment and productivity are uncertain—at least as traditionally measured.

Economic Growth and the Service Economy top

IT contributes to macroeconomic output in a variety of ways. For example, IT can create better ways of generating goods and services, improve production efficiencies, and increase both labor and multifactor productivity. Growth accounting[6] studies confirm IT's positive impact on total U.S. economic output; estimates of the total contribution of IT to the real U.S. growth rate range from 0.16 to 0.52 percent (Jorgenson and Stiroh 1995, Oliner and Sichel 1994, and Sichel 1997).

IT is commonly credited as being a key reason for the structural shift from manufacturing to services in the U.S. economy. Rapid growth in existing services, such as banking, and the creation of new industries, such as software engineering, are attributed to the widespread diffusion of IT in the service sector infrastructure (NRC 1994a, and Link and Scott 1998). From 1959 to 1994, the service sector grew from 49 to 62 percent of U.S. gross domestic product (GDP), while manufacturing declined from 28 to 17 percent.[7]  (See figure 8-6 and appendix table 8-3.) In the past three decades, growth in services has, on balan! ce, exceeded growth in every oth er industrial sector-agriculture, mining, construction, and manufacturing.

The expansion of the service sector has been driven entirely by industries that are often classified as "knowledge" industries (see Machlup 1962)—finance, insurance, and real estate (FIRE)—as well as a number of professional services, such as health and education. The share of GDP accounted for by wholesale and retail trade actually declined from 1959 to 1994, while personal services and transportation and utilities remained essentially unchanged. (See figure 8-7.) In contrast, FIRE's share of GDP grew by 4.8 percentage points, while that of professional services increased by 7.1 percentage points. Employment data reflect the same structural shift in the! economy as GDP data. From 1960 to 1990, employment in the service sector grew from one-half to two-thirds of total U.S. employment, with growth strongest in producer services (FIRE and professional services) and social services, particularly health care. (See figure 8-8.)

IT has not, however, been empirically linked in any definitive way to the expansion of the service sector. In a detailed study of several key service industries (banking, insurance, air transport, and telecommunications), the National Research Council concluded that although the benefits of IT for individual industries could be qualitatively described, IT could not be causally linked to gross product output of the individual industry for methodological reasons (NRC 1994a).[8]  Two observations are worth making, however. First, based on case study evidence and expert reviews, it is unlikely that the expansion of the air transport, banking, finance, and trade industries would have been as significant in the absence of IT (NRC 1994a). In this sense, IT acted as a technological precondition for growth in many service industries.

Second, IT is unevenly distributed throughout the economy and is particularly concentrated in the service industries that have experienced rapid expansion. This suggests that IT is instrumental to the delivery of many services, and that growth in services fuels demand for IT (and vice versa). For example, only 14 percent of workers use computers in agriculture, but 85 percent do so in banking and finance. (See figure 8-9.) Investments in IT similarly vary among industries. For example, the communications industry invests five times as much in IT as would be expected given the size of this sector relative to overall GDP. (See figure 8-10.) The disparity in the relative presence of IT among industries indicates that IT is clearly more critical for some types of business activities than others, and thus may be said to be responsible-in part-for the growth of those industries.

IT and Employment top

IT has demonstrable benefits for employment and skill levels, although not unequivocally so. Evidence indicates that IT contributes to growth in demand for labor, as well as an overall skill upgrading in the workplace. Computerization of the workplace appears to have enlarged the wage gap between workers with a college education and those with a high school education or less. With respect to the impact of IT on individual workers' health and emotional well-being, the record is mixed. While the number of IT-related health disorders is clearly on the rise, trends may be in part a socio-psychological phenomenon. Computerized surveillance and monitoring of employees may lead to greater stress and alienation in the workplace, but not necessarily: evidence suggests that IT may increase workers' sense of worth, accomplishment, and job autonomy.

Aggregate Employment top

Establishing the net effect of IT on aggregate U.S. employment is difficult for one primary reason: IT is both labor-creating and labor-saving. As new jobs are created in some industries and occupational classes, they are lost in others. For example, banking employment has declined by 100,000 workers since its peak in 1990; analysts attribute this trend in part to the growing use of ATMs (Morisi 1996). IT-driven employment losses are, however, also offset by employment expansion in new industries such as computer and data processing services. Isolating the employment effects of IT from other factors—such as business cycles, industry conditions, and labor mobility—is problematic.

In an evaluation of the research on employment impacts of technology, the National Academy of Sciences concluded that the displacement effects of IT were indeterminate, and depended heavily on conditions in individual firms and industries. Because the nature of the research was so varied and the findings often contradictory, the Academy concluded that

the contrasting results of these studies...illustrate the sensitivity of empirical estimates of the employment impacts of [IT] to detailed assumptions concerning diffusion rates, technological improvement, and the organization of manufacturing and production processes (Cyert and Mowery 1987, p. 292).

Employment trends in key IT-related sectors further illustrate the difficulty of establishing the overall employment effects of a new technology. Employment in IT-producing industries is projected to nearly double from 1986 to 2006. (See text table 8-3.) Yet this trend is driven almost exclusively by growth in computer and data processing services (including prepackaged software), the third fastest growing industry in terms of employment. Employment in two of the three IT-producing industries has been declining rather steadily since the early 1980s. Thus, trends in one sector mask patterns in another, much the way that the expansionary effects of IT could mask displacement effects within specific industries or occupations (and vice versa). For a discussion of trends in IT occupations see chapter 3, "Science and Engineering Workforce"

Skill Impacts and Wages top

Assumptions about the information society and post-industrial economy suggest that the development of IT should increase the demand for workers who manipulate and analyze information relative to the demand for non-knowledge workers or those who simply enter and collate data. Yet there is a persistent popular fear of the deskilling effects of IT, a fear that automation will reduce the demands on an individual's conceptual talents and facility with machinery, equipment, and tools. Individual case studies of specific industries, occupations, and information technologies clearly illustrate that deskilling and skill upgrading take place simultaneously (for reviews, see Attewell and Rule 1994, and Cyert and Mowery 1987). On balance, however, several studies—using different data sets and methodologies—suggest that no overall lessening of skills is occurring in the workforce, and that upgrading may b! e widespread.

For example, Castells (1996) finds that employment in managerial, professional, and technical classes has been expanding at a rate faster than in non- and semi-skilled occupations. After an extensive review of trends in occupational categories, he concludes that:

The widespread argument concerning the increasing polarization of the occupational structure of the information society does not seem to fit with this data set...I am objecting to the popular image of the information economy as providing an increasing number of low-level service jobs at a disproportionately higher rate than the rate of increase in share of the professional/technical component of the labor force (p. 219).

Howell and Wolff (1993) conclude much the same. Using detailed data on the cognitive and motor skills required for specific occupations from 1959 to 1990, they found that skill restructuring (principally upgrading) in the labor force began in the 1970s and continued in the 1980s in patterns that "are broadly consistent with what one might expect from the rapid expansion of new [information] technology" (p.12). They also found that the demand for the most cognitively skilled information occupations grew more rapidly than for other occupations during some periods. Analyzing data from the Annual Survey of Manufacturers, Berman, Bound, and Griliches (1994) document a significant skill upgrading throughout the manufacturing sector over the 1980s; they attribute the trend in part to computerization of the workplace. Their findings indicate a distinct shift in the demand for labor from less skilled to more highly (cognitively)! skilled labor in the United Sta tes, a shift that has been linked theoretically and empirically to the diffusion of IT.

Autor, Katz, and Krueger (1997) similarly find evidence that computerization of the workplace may explain from 30 to 50 percent of the additional growth in demand for labor from 1970 to 1995, compared to growth from 1940 to 1970. They find that the increase in the rate of growth for skilled labor since 1970 is driven by rapid skill upgrading in industries that are the most computer-intensive (e.g., those that have the highest levels of computer investment per worker and the largest growth in the proportion of employees who use computers, and those in which computers account for a larger share of total investment). This study finds that those industries that experienced the largest growth in computer use also tended to shift their employee mix from administrative and support workers toward managers and professionals (a finding consistent with Castells 1996). Nonetheless, more systematic insight into which jobs are upgra! ded (or deskilled), and what hap pens to individuals whose jobs are deskilled, can provide a better sense of the organizational dynamics surrounding IT and their ultimate employment impacts.

Assumptions about the IT-skill upgrading relationship extend one more step, and also associate wage gaps with computerization in the workplace. Higher wages are attributed to the higher demand for computer-skilled labor, and lower wages are thought to reflect the absence of computer skills (see Bresnahan 1997 for a discussion of this literature). Autor, Katz, and Krueger (1997) support this thesis; as do Berman, Bound, and Griliches (1994).

However, Howell (1997) refutes the argument that skill mismatch is responsible for wage stagnation among less skilled workers by identifying a crucial anomaly in labor market behavior: employment in low-skill occupations is declining relative to more highly skilled jobs, but the proportion of low-wage workers is actually increasing. Bresnahan (1997) also provides an important critique of the research and empirical evidence on the impact of IT on the demand for skilled labor and wage gaps. He reviews alternative research that indicates that the actual use of IT (particularly PCs) on the job is inconsistent with assumptions about job enrichment, and concludes that "there is little complementarity between highly skilled workers and PC use, certainly not enough to affect skill demand."

IT and the Worker top

While IT may affect the individual worker in any number of ways, two particular effects are worth attention because of their negative physical and psychological aspects: the health hazards associated with the use of IT, and the emotional and behavioral consequences of workplace surveillance and monitoring.

IT is particularly associated with repetitive motion injury, even though a variety of other negative health effects are common, including eyestrain and a complex of musculoskeletal disorders (Huff and Finholt 1994). IT-based repetitive motions include barcode scanning, data entry and keying, and keyboard typing, all of which can lead to carpal tunnel syndrome and tendonitis—sometimes to the point of permanent disability. Data from the Bureau of Labor Statistics (BLS) indicate that the incidence rate of repeated trauma disorder rose from 6.4 per 10,000 FTE (full-time equivalents) in 1986 to 41.1 per 10,000 in 1994.[9]  Although the manufacturing sector still accounts for the vast majority of these repetitive motion injuries, the number of repeated trauma disorders increased more than fivefold in the service sector between 1988 and 1992. Grocery stores, newspaper publishing, hospitals, and casualty insurance industries now rank among the 20 sectors with the highest incidence of the disorder, and BLS indicates several other service industries are "poised to enter the list," including airline scheduling, department stores, and mail order retailers (U.S. BLS 1994). The intensive use of IT is clearly an occupational hazard for individuals prone to repetitive motion disorder, but some researchers have found that a number of social and organizational factors can influence both the incidence of IT-related repetitive motion trauma and its severity (Kiesler and Finholt 1994, and Rowe 1994).

Workplace surveillance and monitoring also raise issues concerning workers' psychological health. The U.S. Office of Technology Assessment defined electronic workplace monitoring as "the computerized collection, storage, analysis, and reporting of information about employees' productive activities" (1987, p. 27); and it includes such measures as keystrokes typed per minute, length of time on a phone call, and length of time away from a computer terminal. Workplace monitoring is estimated to have doubled from 20 percent of all office workers in the early 1980s to 40 percent in the early 1990s, and spending on monitoring software is believed to exceed $1 billion (Aiello 1993).

The effect of workplace monitoring on the individual's well-being and work performance is unclear. One study (Grant, Higgins, and Irving 1994) found that monitored customer service employees believed that good work performance was quantity based, while nonmonitored employees focused on the quality of service and teamwork. Another analyst observed workers disconnecting phone calls if it appeared the caller would exceed the 22-second maximum time allotted by the firm to each call (Aiello 1993).

Overall, studies show that workplace monitoring may both increase and decrease productivity, and may or may not lead to greater stress, anxiety, isolation, and diminished work motivation. Actual outcomes depend on a variety of moderating factors in the workplace, including worksite, supervisor style, type and frequency of feedback, and the individual's sense of control over the monitoring itself.[10]  Indeed, one study (on the impacts of IT on quality of worklife) concluded that although IT could intensify work pressures, it also enhances workers' sense of worth, accomplishment, and autonomy (Danziger and Kraemer 1986). Van Alstyne (1997) nonetheless regards surveillance and monitoring with suspicion, and concludes that there is good reason to expect that "those suffering reduced autonomy due to IT will seek ways to subvert the system, for example, through sabotage,! disuse, delay, use of alternati ve procedures, supplying inaccurate data, or sticking to the letter but disregarding the intent of the system" (p. 40).

IT and the Productivity Paradox top

One of the most debated issues about the impact of IT on the economy is that of the "productivity paradox"—the inability to find a statistical association between IT investments and productivity in the private sector. Despite compelling reasoning and evidence about the highly positive effects of IT on competitiveness and cost reduction,[11]  traditional econometric analyses fail to find any productivity benefits for IT, and some studies identify negative productivity impacts for IT investments. The meaning of these findings is subject to considerable debate, with most experts advising caution in interpretation of the data. Problems with measurement and organizational learning lags are two explanations commonly offered to make sense of the counterintuitive empirical findings. However, the most current research on the IT productivity paradox suggests that it may have "disappeared" in the early 1990s; s! ome analysts argue that the para dox is primarily the result of overly optimistic expectations about IT's economic effects.

The Empirical Studies top

The IT productivity paradox was revealed by over 20 econometric analyses conducted and published between 1980 and 1990 (for detailed reviews, see Brynjolfsson and Yang 1996, and NRC 1994a). Regardless of the level of analysis chosen—the macroeconomy or specific industries and sectors—these studies demonstrated that there was no statistically significant, or even measurable, association between investments in IT and productivity.

The findings were troublesome not only because they contradicted strong expectations about positive effects, but also because productivity impacts apparently failed to materialize anywhere (not in services or manufacturing), by any measure (a variety of data sets and methods were used), or at any time (the studies collectively covered the late 1960s to the late 1980s). Findings of positive effects are reported in the literature, but this research represents one-time-only case studies of a single industry or small set of firms. The preponderance of the IT productivity research—which incorporates large and relatively comprehensive data sets at the firm, industry, and macro levels—consistently fails to demonstrate a significant positive impact by IT on productivity, regardless of sector or industry. Indeed, one widely cited study finds a negative correlation between investments in IT and multifactor productivity. Furthermore, this study identifies yet another anomaly: industries ! that are IT-intensive are more p rofitable than others; but within industries, such intensity is negatively associated with profitability (Morrison and Berndt 1990, and Berndt and Morrison 1995).

Two recent avenues of empirical analysis are, however, notable. Oliner and Sichel (1994) and Sichel (1997) report a small but positive association between IT and productivity using a growth accounting approach. Brynjolfsson and Hitt (1995 and 1996) find large and significant contributions by IT to productivity using a new firm-level database. Both sets of findings are highly suggestive about the nature of the IT productivity paradox. Sichel (1997) argues that it is primarily our expectations that are out of line with the long-term historical trends regarding both IT diffusion and the overall level of IT capital in the economy. Brynjolfsson and Hitt note that a full 50 percent of the variation in IT's contribution to marginal product can be accounted for by firm-level variables. This suggests that aggregate data are not likely to detect patterns in IT impacts, and! that the effective use of IT is highly contingent upon the context of its use at the organizational level.

To elaborate, Oliner and Sichel find small but real contributions of computers to the economy. From 1970 to 1992, computer hardware contributed 0.15 percentage points to the total U.S. output growth rate of 2.8 percent. When software and computer-related labor are included, this contribution doubles to 0.31 percentage points for the period 1987 to 1993 (or 11 percent of total growth).[12]  Other capital and labor inputs, as well as multifactor productivity gains, account for about 90 percent of the growth in U.S. output during this period.[13]  The authors explain the small contribution of computers by observing that computing-related inputs are a very small portion of total capital and labor, and have only recently grown large enough to have a measurable impact. They conclude that "computing equipment can be productive at the firm level and yet make little contribution to agg! regate growth, precisely because computers remain a relatively minor factor of production" (Oliner and Sichel, p. 286). Sichel (1997) expands on this argument by reflecting on trends in the diffusion of a variety of information technologies. He concludes that computing technologies are part of a 150-year trend toward greater information intensity in the United States, and that we should not expect the effects of computers to be large and sudden, but modest and part of a historical continuum.

Brynjolfsson and Hitt (1996) analyzed the impact of IT on marginal output using a new firm-level database and found large contributions of IT to marginal product for the firms in their study. Every additional dollar of computer capital stock was associated with an increase in marginal output of 81 cents, and every additional dollar spent on IT-related labor was associated with an increase in marginal output of $2.62. Their earlier work also demonstrates that firm-level factors account for half of the variability in IT's marginal product contributions (Brynjolfsson and Hitt 1995). In contrast, previous studies indicate that increases in IT are not associated with increases in marginal output; Morrison and Berndt (1990) found a negative relationship between IT spending and marginal output.

Several factors may explain the dramatically different findings of Brynjolfsson and Hitt relative to the earlier productivity studies. The later time period of their study (1987-91); the use of a larger data set; more detailed, firm-level[14]  data; and the inclusion of IT-related labor (note that IT capital expenses are typically a small fraction of a firm's total IT-related costs) are all reasons why their findings are more positive than those resulting from earlier research. Using similar data and methods, other analysts have also found significant positive rates of return at the firm level, including Lichtenberg (1995) and Link and Scott (1998).

The studies by Oliner and Sichel, and Bryjolfsson and Hitt highlight the complexity of research into the effects of IT on productivity. Both sets of findings suggest that IT does have measurable payoffs for economic productivity, but the orders of magnitude are quite different. Macroeconomic impacts may be quite modest at best (as measured by Oliner and Sichel), whereas firm-level benefits may be more substantial (as measured by Brynjolfsson and Hitt). While they do not indicate that the productivity paradox has been resolved, these findings do suggest that the relationship between IT and productivity may be changing. Explanations for the paradox and the lagged benefits of IT therefore require further exploration.

Explanations for the Paradox top

There are a number of interpretations of the productivity paradox, most falling into one of three categories:

This third interpretation is not explored in this chapter, since the weight of evidence suggests that there are meaningful impacts of IT, challenging measurement problems, and very real social lags.

Excluding disagreements about the quality of various data used in the IT productivity studies (which has implications for sources of error in the findings), there are still a number of core measurement issues.[15]  The first is, what constitutes IT? Is it capital investments only, or does it include labor, which represents the bulk of IT operating costs? Do IT capital investments include more than computers, and if so, what? The choices of what to count as an IT equipment expense include computing hardware and software, communications equipment, and a variety of office machines (such as photocopiers and some instruments). At present, there is little consistency among studies, and sources of IT investment data vary from aggregate government data to private survey-based firm data. One fundamental measurement issue is simply standardizing the definition of IT itself (labor, capital, and types of capital): standardized definition! s can facilitate data collection , comparability across data sets, and cumulation of findings.

A second key measurement issue is how to assign dollar values to IT as a factor input. IT can be measured as a flow (annual expenses or purchases) or as a stock (the cumulation of equipment over time). In both instances, price deflators are required to compare stocks or flows over time by converting them to "real" dollars. IT equipment is especially problematic for establishing reliable deflators. For example, not only has the sales price of computing equipment been falling rapidly, but because quality has increased exponentially, existing computing stock becomes obsolete very quickly. The pace of technological change in information technologies greatly complicates analysts' abilities to construct quality-adjusted price deflators[16]  and appropriate depreciation rates; distortions in time-series data can significantly affect research outcomes by over- or undervaluing expenses and stocks in different periods.

A third measurement concern relates to output-specifically, how to measure the output of information processing. IT is used extensively for "activities" that do not result in tangible market outputs (e.g., accounting, scheduling, reporting).[17]  Consequently, it is difficult to assign a dollar value to the output of IT-a measurement that is crucial to accurate productivity analysis. This measurement challenge is exacerbated in the service sector, where output measures must also capture qualitative differences in services (Mark 1982 and Noyelle 1990); the problem is sufficiently severe that BLS does not report labor productivity for the software industry, a core IT sector (Goodman 1996). The potential for mismeasurement of services and information processing outputs, as well as IT as a factor input, is so troublesome that mismeasurement is u! sually cited as the primary expl anation for the productivity paradox.

A fourth measurement issue deserves attention and has less to do with mismeasurement of a specific indicator (such as factor inputs and product outputs) than of measuring the wrong indicator to begin with. Studies of the applications and use of IT repeatedly demonstrate that IT benefits do not show up as classical efficiency gains, but as cost savings, improved inventory management, and qualitative improvements in customer service. These improvements reflect such dimensions as enhanced timeliness, performance, functionality, flexibility, accuracy, precision, customization, cycle times, variety, and responsiveness regardless of whether the output is a product or service or the consumer is an original equipment manufacturer, a distributor, or an end user (NRC 1994a; Byrne 1996; and Bradley, Hausman, and Nolan 1993). These qualitative dimensions are much more likely to show up as downstream ben! efits to the consumer (Bresnahan 1986) or as greater competitiveness for a firm-an outcome known as a "distributional effect" (Banker and Kauffman 1988, Baily and Chakrabarti 1988, Brynjolfsson 1993, and Porter and Millar 1985). In addition, Weill (1992) has found that the type of information processing (transaction) matters. In a study of valve manufacturers, data processing could be associated with productivity gains, but general business systems (like sales and marketing support) could not.

Institutional Lags top

Another compelling explanation of the productivity paradox argues that it is a real but temporary phenomenon. Sociologists and economic historians have long argued (very cogently) that society's ability to fully exploit a new technology lag—often by decades—the introduction of the technology itself (Ogburn 1964 and Perez 1983). Similarly, in organizational change scholarship, analyzing institutional resistance to change (technological or otherwise) is the coin of the disciplinary realm. In theory and in practice then, as humans and their institutions become more accustomed to IT, productivity and other aspects of performance should improve.

There is a good deal of evidence to support this argument. Important technological analogies for IT are electric generators and the electric power infrastructure. David (1989) found that it took nearly 20 years for the electric generator—an invention comparable to IT in scope and consequence—to have a measurable effect on industrial productivity; Friedlander (1997) found that historically it has been difficult to measure the benefits of most infrastructure technologies. With respect to IT specifically, firm-level performance can vary considerably, and the effective use of IT is apparently contingent upon a number of moderating variables at the organizational level—including strategy, leadership, attitudes, organizational structure, appropriate task and process reengineering, individual and organizational learning, and managerial style and decisionmaking (Cron and Sobol 1983; Curley and Pyburn 19! 82; Graham 1976; Thurow 1987; Landauer 1995; Tapscott 1996; Danziger and Kraemer 1986; Khosrowpour 1994; Banker, Kauffman, and Mahmood 1993; and Allen and Morton 1994). Other analysts argue that information technologies themselves are the cause of low productivity, since they are not necessarily user-friendly or well-designed. In this respect, evidence suggests that technological adaptation to social need has a lagged effect as well (Eason 1988, Landauer 1995, and Forester 1989).

The productivity paradox may thus be partially explained, but it does not dispel the observation that even as IT is radically changing the nature of some business activity, that activity does not necessarily get translated into greater efficiency or economic welfare. The banking industry is a good case study of the complexity of the paradox, and also of the possibility that the paradox may have "vanished" in the early 1990s for some sectors.

IT and the Banking Industry  top

The banking industry is one of the oldest users of information technologies—in the early 1950s, Bank of America was the first commercial user of mainframe technology (Morisi 1996). Yet the banking industry reflects most of the empirical dilemmas associated with measuring the impacts of IT: heavy investments in IT; slow (or no) visible improvements in productivity until relatively recently; and impacts that reflect quality improvements, rapid product diversification, and substantial growth in volume of commercial transactions.

IT has clearly changed both the structure and service quality of banking, and appears finally to have a positive impact on cost reduction. But it has taken decades to achieve these results, and traditional productivity analyses still do not detect positive associations between IT investments and productivity in the commercial banking sector.

Banking industry investments in IT increased substantially from the late 1960s to the late 1980s. (See figure 8-11.) Annual investments in IT (in constant 1982 dollars) grew from $0.1 billion in 1969 to $1.6 billion in 1980 to $13.8 billion in 1989. By 1989, the banking industry was annually investing more funds in IT than were all of the other major service industries except telecommunications. The banking industry invested more in IT relative to its gross product output than the insurance, health care, air transport, telecommunications, wholesale trade, and retail trade industries. (See text table 8-4.)

IT uses are diverse in the banking sector. Initial applications included accounts management and check processing via magnetic ink character recognition. Automated clearinghouses, which enabled electronic funds transfer (EFT), were introduced in the early 1970s and ATMs in the late 1970s. EFT, ATM, and telephone transaction capabilities have replaced a wide variety of paper and in-person transactions in banking, including account deposit and withdrawals, accounts management, credit applications and approvals, cash dispensing, funds transfers, point-of-sale transactions, credit card payments, and consolidation of banking operations.

Impacts on Productivity top

Reviews of the traditional econometric productivity literature indicate that IT investments by the banking industry do not systematically result in measurable, positive productivity impacts. Major cross-sector studies (see Brynjolfsson and Yang 1996 for reviews) do not detect positive productivity returns for IT in the banking industry, and Franke's (1989) study of the financial sector (insurance and banking combined) suggests that IT is associated with negative productivity impacts. However, Brand and Duke (1982) do find productivity growth of 1.3 percent per year attributable to computers. Using qualitative evidence and interviews with chief executive officers, the National Research Council attributed the lack of productivity impact to a variety of factors. One is the ever present measurement issue: measures of output in the banking industry are extrapolated from ! employment data by the U.S. Bure au of Economic Analysis and estimated from indices of financial transactions (loans, deposits, and so forth) by BLS. Neither procedure fully accounts for the volume of banking transactions or wider variety of financial services; the inherent difficulty of measuring commercial banking output seriously qualifies productivity analysis using aggregate data sets.

Note, however, that labor productivity has been steadily improving in the banking industry. (See figure 8-12.) Morisi reports that "during the 1973-93 period, commercial banks had the highest long-term growth in productivity than any of the measured finance and service industries" (1996, p. 30). The difficulty is in empirically linking these improvements to IT.

A second reason for the apparent lack of IT-led productivity growth in this industry relates to problems with early generations of information technologies and organizational adaptation. The National Research Council study reported that:

early applications of IT proved to be costly and cumbersome. Software and equipment had to be updated and replaced frequently...IT systems required large amounts of tailoring, training, upgrading, and updating. Cost control, management skills, and productivity tracking systems lagged behind the new technologies in a rapidly changing competitive marketplace...The result was that tangible paybacks from IT investments were delayed (NRC 1994a, pp. 80-81).

Other Business Impacts of IT top

The significance of IT emerges in areas of business impact other than conventionally measured productivity gains. Three types of effects are worth particular note: the expansion of banking products and services, time and cost savings, and competitive positioning.

Banking products and services have proliferated with the use of EFT, ATM, telephone transactions, and automated credit and loan procedures. Banks thus process billions of transactions a year—everything from clearing individual checks, to ATM cash dispersal, to account inquiries, to loan approvals—a volume of interactions that would simply not be possible without automation. For example, the Clearinghouse for Interbank Payment Systems was processing nearly $2 trillion worth of transactions per day by the late 1980s, and Visa's capacity for authorizing credit card transactions increased from 30,000 per day in 1978 to 1.4 million per day in 1991 (NRC 1994a, pp. 83-84). Bresnahan (1986) estimates that the benefits to consumers of the use of mainframe computers for financial services was five times greater than the investments in the computers themselves.

The qualitative improvement in customer convenience, ease, and scope of access to financial resources is reflected in the overall growth of electronic transactions. Figure 8-13 illustrates the expansion of electronic (ATM and point-of-sale) transactions in the United States; the number of electronic cash transactions and payments for goods and services was more than 10 billion in 1995, compared to just over 5 billion in 1989.

Time and cost savings for the industry are also notable. For example, Mellon Bank reduced the average processing time of customer complaints by 20 days when it installed an integrated document system; Visa reduced its processing time for electronic credit card authorizations from 5 minutes in 1973 to 1.1 seconds in 1991; and the Bank of Boston reduced its staff requirements by 17 percent and increased its transaction volume by 80 percent when IT allowed the bank to consolidate its mainframe operations (NRC 1994a, pp. 83-84). The American Bankers Association estimates that ATM transactions cost 27 cents compared to $1.07 for a human teller, and telephone transactions cost about $0.35 compared to $1.82 for a phone call processed by bank personnel (Morisi 1996). In a study of 759 banks, Alpar and Kim (1991) found that a 10 percent increase in IT expenses led to a 1.9 percent decrease in total ba! nk costs.

Although the productivity measures do not find a link between banking industry output and IT investments, it is important to note that while the volume of financial transactions has been increasing at a dramatic rate, employment in the sector has been falling. By 1996, employment in the commercial banking industry was 100,000 employees below its historic peak in 1990. During the same period, the number of ATM transactions doubled to more than 10.5 billion.

IT is of value to the banking industry not only for time savings, cost reductions, and customer services, but for the ability to give individual banks a competitive advantage or the ability to maintain a competitive position. Deregulation of the industry in 1980 led to intense rivalry among institutions, and expanding automated services was one way of attracting depositors and customers. Thus Banker and Kauffman's (1988) study of 508 branch banks found that ATMs were essential to maintaining market share and customer base—not necessarily to reducing costs.

Implications for IT Metrics top

The banking industry illustrates many of the issues involved with establishing useful metrics for analyzing the economic impacts of IT. Not only are there problems with measuring the output of this industry in a meaningful way (productivity estimates require output estimates), but there is the issue of what to measure in the first place. IT clearly provides "value added" in a range of consumer and producer activities that are not captured by productivity analysis, such as convenience, scope of services, access, time savings, transaction volume, and transaction cost reductions. The challenge is to select one or two representative measures of impact and track their performance over time.

The industry may have experienced a long learning curve in terms of adaptation to new information technologies. Insight into how banks reengineered their organizations, management strategies, and work tasks could inform IT strategies in other industries and shorten the lag between the time a technology is introduced and the time it begins to measurably enhance business performance.

green bar

Footnotes

[6] Most output and productivity studies use what is known as a "production function" model. The resulting statistics are typically least-squares correlations and estimates based on a log-linear regression. Growth accounting, a technique developed by Denison (1985), principally uses an arithmetic/algebraic procedure on national income accounts data. Robert Solow received the Nobel Prize in economics for his estimates of the contribution of technical change to aggregate productivity using a production function model (Solow 1957). For more detail on these models, see NSB (1996), chapter 8.

[7] Note that these figures differ somewhat from those frequently published; this is because the U.S. Bureau of Economic Analysis recently revised its methodology for calculating the contribution of specific industries to GDP. See U.S. BEA (1996).

[8] Specifically, IT investment impacts in the 1980s cannot be isolated from the effects of many market, industry, and economic factors such as the deregulation of banking, telecommunications, and air transport.

[9] These data are from U.S. BLS's Survey of Occupational Injuries and Illnesses and may be accessed from the Occupational Health and Safety Agency Web site http://www.osha-slc.gov/ergo/chart3.html.

[10] For a good overview of these issues and findings, see Aiello (1993).

[11] See Bender (1986); Benjamin et al. (1984); Harris and Katz (1991); Malone, Yates, and Benjamin (1987); Porter and Millar (1985); Bradley, Hausman, and Nolan (1993); NRC (1994a); and Byrne (1996).

[12] Note that Jorgenson and Stiroh (1995), who also use a growth accounting approach, find an appreciably higher level of contribution by computing hardware to macroeconomic output. These authors estimate that computer hardware contributed 0.38 percentage points to the 2.49 percent growth rate from 1985 to 1992—more than double the 0.15 estimate provided by Oliner and Sichel. Differences are due in large part to the different time periods of the studies and to differing assumptions about depreciation rates. As with other economic analyses, assumptions can have a substantial impact on empirical estimates.

[13] Sichel (1997) asserts that there is no additional contribution of IT hidden in the multifactor productivity (MFP) estimate. MFP is a residual element that reflects technical and organizational changes that improve the efficiency of converting inputs into outputs, hence IT could contribute to gains that are captured by MFP. However, given the nature of growth accounting techniques, IT inputs would have to have a "supernormal" rate of return, and Sichel argues that there is no compelling evidence for such an assumption.

[14] Findings are based on a data set of 367 firms generating $1.8 trillion in aggregate sales in 1991.

[15] The measurement problems are substantial and are discussed in detail elsewhere (Bryjolfsson 1993, Baily and Chakrabarti 1988, Griliches 1997, NRC 1994a, and Oliner and Sichel 1994).

[16] Note that the issues surrounding the measurement of services and their impacts are comparable to the methodological problems of measuring services and their impacts. Outputs are often intangible, quality is difficult to account for, and constructing R&D-specific price deflators is a complicated task. For more on R&D measurement issues, see NSB (1996), chapter 8.

[17] "Activities" are defined as repetitive and structured sets of work tasks; see NRC (1994a).


Previous | Top | Next

Indicators Home | Contents | Help | Comments