Upstairs, Downstairs: Computers and Skills on Two Floors of a Large Bank

Many empirical studies document a positive correlation between workplace computerization and the employment of skilled labor in production.Does this mean that computers necessarily substitute for the tasks performed by less educated workers and complement the tasks performed by more educated workers?We explore this question by positing that computerization leads to the automation of tasks that can be fully described in terms of procedural or 'rules-based' logic.This process typically leaves many tasks to be performed by humans.Management decisions play a key role -at least in the short run -in determining how these tasks are organized into jobs, with potentially significant implications for skill demands.We illustrate how this conceptual framework helps to interpret the consequences of the introduction of digital check imaging in two back office departments of a large bank.We argue that the model has applicability to many organizations and helps to reconcile differences between the approaches economists and sociologists typically take to studying the consequences of technological changes.Two recent trends have rekindled interest in the questions of how technological change impacts the skills that workers use at their jobs and the way in which these skills are remunerated.The first is the increase in earnings inequality.Since 1980, the earnings of highly educated workers have increased relative to those with less education, a phenomenon that economists attribute primarily to large changes in relative demands for the skills supplied by these groups of workers.The second trend is the remarkable proliferation of computers and information technology, beginning with the spread of mainframe applications during the 1970s, moving to greater use of personal computers in the 1980s, and to enormous growth in applications of networked computers in the 1990s.A number of studies have documented positive correlations between the use of computers and the use of more educated labor in production across detailed industries and across plants within industries in the U.S. Similar relationships are present in data from other industrialized countries.Some analysts cite these correlations as evidence that computers embody skill-biased technical change, meaning that computers substitute for less educated workers in performing some tasks and complement more educated workers in performing other tasks.Other observers reject this conclusion as unduly deterministic.Based on analyses of case studies, they argue that equating computers with skill-based technical change ignores management's role in job and organizational design and relies on simplistic definitions of skill.In this paper, we argue that the introduction of computer-based technology creates strong economic pressure to substitute machinery for people in carrying out tasks that can be fully described in terms of procedural or 'rules-based' logic and hence performed by a computer.This process typically leaves many tasks to be performed by humans, and management decisions play a key roleat least in the short runin determining how these tasks are organized into jobs, with potentially significant implications for skill demands.We show how this model helps to interpret how the introduction of digital check imaging in two back office departments of a large bank, one (downstairs) department processing deposits, the other (upstairs) department handling exceptions (e.g.overdrafts or stop check requests).The introduction of check imaging led to the computerization of certain tasks in both departments.However, the re-organization of the remaining tasks differed across the two departments.In deposit processing, check imaging led to greater specialization in jobs; in exceptions processing, it led to the integration of tasks, with a resulting increase in the demand for problem-solving skills.We use ideas articulated by Autor, Levy and Murnane (2001), Lindbeck and Snower (2000), and Osterman (1994) to explain why the introduction of check imaging led managers to reorganize work so differently in the two departments.We begin with a brief discussion of the literature that bears on our case.We then turn to a discussion of what computers actually dothe execution of rules-based logic.This provides the background for describing and interpreting the evidence from our case study.Though we present only one case, we believe that the model -computers displacing humans in some tasks with management decisions reshaping othersoffers a potentially valuable framework for future studies of computerization's skill impacts. Previous ResearchThe explanation favored by many economists for the positive relationship between computer use and the demand for educated labor is computer-skill complementarity or skill-biased technological change.1The essence of this hypothesis is that technological change involving computers increases the productivity of highly educated workers more than it increases the productivity of less-educated workers.An alternative explanation popularized in books such as The End of Work (Rifkin, 1995) is computer-labor substitution: computers substitute for low skilled labor in carrying out a variety of tasks.Both of these explanations imply an increase in relative demand for highly skilled workers.Some social scientists, especially those trained in disciplines other than economics, find the concept of skill-biased technological change troubling.For example, in a thought-provoking article Paul Attewell (1990) pointed out that there are many different ways to think about the concept of skill.In one tradition (positivism), skill is treated as an attribute ofjobs, and jobs that are substantively complex are viewed as skilled.A difficulty here is the arbitrariness of ranking in a single dimension of complexity jobs that are very different, for example, conducting biological research versus managing a large organization.Perhaps all that these highly skilled jobs have in common is that both require a great deal of conscious thought.A different research tradition, ethnomethodology, sees a variety of human activities such as walking across a crowded room or carrying on a conversation with many voices in the background as extremely skilled tasks that humans learn to do without conscious thought, an insight validated by decades of research in artificial intelligence (Pinker, 1997).Consistent with this view, many of these "simple" activities cannot be performed by computers, a point we return to below.These very different conceptions of skill raise questions about exactly what economists mean when they refer to "skill-biased technological change" and what the predictions are about