Return to Hartley's Projects Page | Project Metadata | Keywords | ||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
Automation was heralded as a great improver of productivity and many benefits have been seen in the manufacturing sector. However, productivity improvements in white collar work, especially in the service industries, have been elusive. In fact, as computerization has proceeded from the 1960s, annual productivity increases have generally decreased, a correlation called the "productivity paradox" [Ryan, Hawaii, Centre].
Explaining the Productivity Paradox
Several explanations have been advanced for the productivity paradox; however, there is no consensus [Kamimura, Strassmann]. One possibility is that software systems have failed or cost more than expected. There are several notable examples [Englehart]. Sperry was hired to modernize the Internal Revenue Service (IRS) computer system and failed ($4 billions, 1980s). IBM was hired to modernize the Federal Aviation Administration's Air Traffic Control System, but failed ($1 billion, 1990). These were custom systems; however, the implementation of standardized systems has not been worry-free. The SAP accounting system was widely touted as the best new accounting system of the 1990s and cost millions of dollars per installation. Despite many successful installations, there have also been many customers who have de-installed SAP after spending the millions of dollars. Some have even sued [Diederich]. In fact, a 1994 IBM survey of software system installations by 24 leading companies found that 55% cost more than projected, 68% took longer than projected, and 88% had to be redesigned.
Clearly, major software systems are not guarantees of increased productivity; however, failures such as these are not regarded as a major component of the productivity paradox. It appears that there are four possibilities: automation is not effective in increasing productivity outside of the manufacturing industries; countervailing factors are reducing overall productivity, masking the contribution of automation to increased productivity; the productivity gains are simply lagging the implementation and will be seen eventually; or there is a problem in measuring productivity in the white collar and service areas, a definitional problem.
Defining Productivity
Productivity measurements are more complex than they first appear. National-level, overall figures are fairly easy to arrive at, simply divide the annual Gross National Product (GNP) by the population to get GNP per capita. If GNP per capita is rising, productivity is rising. Alternative definitions, such as GNP per worker yield different values; however, the size of US economy washes out many variations and GNP per worker trends will be nearly identical to GNP per capita trends.
Finer detail productivity measurements become more problematic. A downward trend in hours worked can mask true productivity gains. To account for this, one computes GNP per worker hour and three figures are required: GNP, number of workers, and hours worked per worker. For industries in which hourly paid workers dominate, collecting these data are possible. However, for predominantly white collar industries, the hours worked data are not collected by the companies and can only be guessed at. A trend toward more hours worked, whether on the job site or at home, can falsely indicate productivity gains. (Certainly, the businesses may be experiencing productivity gains by achieving more work per employee; however, this is not a scalable factor, as the number of hours in a day are limited.)
At an even finer detail, more problems appear, especially for government services [Abraham]. Most individuals contribute indirectly to the value earned by an organization. For the FBI as a whole, one might compute the number of criminals caught per year per employee. If the number rises from year to year, productivity is increasing. However, it is difficult to determine whether a data entry clerk's productivity is improving by this measure. Proxy measures are invented, e.g., number of transactions processed per shift. Proxy productivity measures are necessary, but not without problems. For example, definition changes make comparisons over time difficult. It is easy to see that calling two old transactions one new transaction will cut the recorded productivity numbers in half. Some such definitional changes can be accounted for easily. But some more subtle changes are very difficult to handle. For example, consider a shift in transaction definition that requires exactly the same material to be entered; however, in the new transaction an additional step is required, a spell check. Productivity, as measured by transactions per shift, will decline because the work required per transaction has increased.
Proxy measures embody another flaw. A worker may improve his or her performance of transactions over time and not improve the organizations' productivity because management has the worker doing the wrong job. Alternatively, management may improve the process, creating new transactions for the worker. If these transactions are identical in difficulty to the old transactions, the worker's productivity may show no change, despite an increased contribution to the organization's productivity.
Thus, there are two components to the desired productivity measure of useful work per unit of time, the amount of work per unit of time (worker contribution) and the value of the work (management contribution). Benchmarking the worker contribution is virtually impossible because of the definitional problem; however, a detailed time and motion study of a particular job can define and measure the quantity of work being performed. Note that the productivity gains possible here are limited. If a worker is wasting 50% of his time, his productivity can be doubled; however, an expensive time and motion study is not necessary to see this level of lost productivity. A 5-10% increase is a more likely maximum to be gained. Changing the value of the work performed is more likely to yield improved productivity (in the desired sense of the word).
Measuring Value
Unfortunately, productivity, as measured, may not be equivalent to productivity in the desired sense. This is the crux of the "definitional" explanation of the productivity paradox. For example, consider the process of delivering answers to whether a particular individual has a criminal record. (Fictitious numbers will be used for illustratitive purposes.) Suppose the old system required 300 people to handle 9000 requests per month and that the answers were delivered in two weeks (average). Productivity is 30 requests serviced per person month (9000/300). Now suppose a new system is instituted in which 300 people handle 9000 requests per month, delivering the answers in two hours (average). Productivity is still 30 requests per person month (9000/300). The speed of delivery does not enter into the productivity equation, despite the clearly higher value of the work. The volume of work per month has not changed and the worker input has not changed, so the productivity number is the same. What has changed is the value of the work. To measure the desired sense of productivity, a numerical comparison of the change in value is required. What is the decrease in response time worth?
BIBLIOGRAPHY
Abraham, Katharine G. Measuring State and Local Government Labor Productivity: Examples from Eleven Services. U.S. Dept of Labor, Bureau of Labor Statistics, Bulletin 2495. 1998.
Acre Wood Consulting. "Intro to SAP." http://www.acrewood.com/intro.htm. 2000.
Baiada, R. Michael. "FAA’s air traffic control is headed for a crash," Dallas Business Journal, Sep 22, 1997. http://dallas.bcentral.com/dallas/stories/1997/09/22/editorial2.html?t=printable. 1997.
Centre for the Study of Living Standards. Productivity: Key to Economic Success. 111 Sparks St, Suite 500, Ottawa, Ontario, Canada K1P 5B5. Prepared for the Atlantic Canada Opportunities Agency. 1998.
Diederich, Tom. "Bankrupt firm blames SAP for failure," Computerworld, Aug. 28, 1998. http://www.computerworld.com/cwi/Printer_Friendly_Version/0,1212,NAV47_STO26181-,00.html . 1998.
Englehart, Kevin. "Advanced Software Engineering." http://www.ee.unb.ca/kengleha/CMPE3213/IntroToSoftwareEng.htm . Fall 1999.
Hawaii. "U.S. Productivity Trends and Issues," Hawaii’s Economy, 4th Quarter, 1996. http://www.state.hi.us/dbedt/he4q/trends.html. 1996.
Kamimura, Gary. "Worker Productivity Trends in Washington, 1977-97." Washington State Employment Security, Labor Market and Economic Analysis Branch. 1999.
Ryan, Ciaran. "The Productivity Paradox," Tech Trends 1999. http://www.govtech.net/publications/TechTrends/July99TechTrends/Productivity_Paradox/Productivity_Paradox.phtml. 1999.
Strassmann, Paul A. "Facts and Fantasies about Productivity," excerpted from Information Productivity. http://www.strassmann.com/publs/fnf/factnfantasy.shtml. 1997.
If you arrived here using a keyword shortcut, you may use your browser's "back" key to return to the keyword distribution page.
Return to Hartley's Projects Page