The NBER-Sloan Project on Productivity Change

Martin Feldstein *

The papers presented at this session of the American Economic Association's annual meeting are based on research done at the NBER on the sources of productivity change in industrial companies. The project, supported by the Sloan Foundation, began in 1994. Professor Adam Jaffe and I have been the directors of the project with Adam taking primary responsibility for the project's operational management.

This investigation of the sources of productivity change is particularly timely because of the surprisingly rapid increases in productivity that have occurred in the past several years in the United States, particularly in manufacturing industries. Although an exogenous technological shift in communication and computing capability may be an important source of this productivity gain, the greater extent to which productivity has increased in the United States than abroad suggests that there are other factors about the industrial environment that are influencing the rate of increase of productivity.

An unusual feature of the project has been our emphasis on visiting companies and factories in order to see the production process and to talk with managements about the sources of productivity change. Since the beginning of the project, the NBER has supported some 50 individual studies done by NBER researchers in a wide range of different industries. Each of these projects has involved site based research in which the researcher observed the company in action, developed specific research ideas, and discussed those ideas with the business managers. A primary focus of the project has been to encourage such direct observation and discussion as an integral part of economic research.

While the specific research projects each involved between one and three researchers, some of the general visits to factories involved groups of ten to fifteen researchers. We have referred to these as our "Pin Factory" visits in deference to Adam Smith and his famous description in The Wealth of Nations of the production process in a pin factory. Smith was able to see how the division of labor increased productivity and went on to theorize about the gains from increased market size, expanded trade, improved infrastructure, and other things on the basis of this visit. We hoped that observing would be fruitful for us as well. Professor Susan Helper was particularly helpful in organizing these Pin Factory visits and in preparing us for what we would be seeing.

We found the managements at the companies we visited were generally eager to show their facilities, to tell about their management practices, and to explain why they did things and how their practices have changed over time. They were also generally open about answering our questions.

We would discuss not just the production process itself but also things like the compensation systems used to motivate workers, the criteria on which incentives were based, the sources of ideas for new products and new processes.

We also talked about why productivity-increasing changes were introduced. Among the reasons we heard were pressure or requirements from industrial customers, the impact of a merger or of being acquired, and the pressures resulting from increased financial leverage or external competition. In short, the sources of change included a wide range of things that seemed obvious to us -- after someone told us.

The group visits - our pin factory trips - persuaded many of the participants of the potential value of looking and of asking questions. That led to a substantial number of individual studies - by single researchers or pairs of researchers -- who had never done site based research before. The papers in today's session and in this volume are about some of those studies. A full list of NBER Working Papers describing these studies appears on the NBER Web Site ( A reference on the NBER home page directs users to the Sloan Foundation Productivity Project. The papers and their abstracts can be read and downloaded.

When I have described this project to non-economists, they were invariably surprised that the process of visiting companies, looking at production and asking questions is an unusual part of economic research. It seems like such a natural thing to do. But as economists all know, it is unusual. We economists are generally accustomed to getting our insights by reading economic literature, going to seminars, and thinking hard about problems. We elaborate these insights in more or less formal models and then sometimes test these theories with aggregate statistics or micro data.

But we rarely go and look and ask. I think that is a pity. Looking and asking provide insights and suggest hypotheses -- and can shoot-down wrong ideas -- in ways that go beyond introspection and reading. Let me offer a personal example that has changed my thinking.

An example

When our pin factory groups visited a manufacturing plant - whether it was Gillette's facility for making razor blades or a Ford assembly plant or a steel mill - the most striking thing was that there were virtually no workers. Most of the workers in the Gillette plant (where billions of blades are made) are maintenance engineers. In the steel mill, the workers were mostly in control rooms guiding the process by computer. And in both cases, they were very few in number.

One of my favorite questions to managers on these visits was therefore: "What percent of your costs are represented by these production workers?" As economists, we tend to think of labor's share of costs as about two-thirds or three-fourths. But the typical answer would be, "Oh, about 10 percent."

Now further questioning would bring out that one reason for the low labor share is that a significant part of the cost is purchased material that embodies labor inputs in other companies. But that leads to the natural follow-up question: "Leaving out purchased inputs, what percent of costs are represented by these production workers?" The typical answer was about 20 percent, far less than the share of costs that we normally associate with labor.

Well that tells you a lot - and it raises a lot of questions. As we probed we learned that most of the labor costs in a manufacturing firm are for non-production workers like product designers, accountants, sales people, other managers, and so on. An important implication of this is that raising the productivity of the production workers does relatively little to raise overall productivity. The big scope for productivity gains is with the non-production workers. That may help to explain why the information technology revolution may be a major source of the recent productivity improvements.

The more general lesson that I have taken from the Pin Factory experience is that plant visits and discussions with management can be an interesting and useful source of hypotheses. I emphasize hypotheses - not definitive answers. The next step in an economic research project is to elaborate these ideas and test them with data.

Seeing a company and talking with managers is likely to lead to a more interesting and informed approach to the data. But testing is important to know that the specific example that you've seen or the answer that you've heard is not just a special case - or indeed a misleading answer. When questions go beyond a simple fact - like the percentage of payroll outlays in total costs - there is room for ambiguity and for outright deception - perhaps self-deception - on the part of the person answering.

So the researcher has to ask: Are they telling me the truth? Or are they trying to paint a favorable picture of the way the company works - a more rational picture of the innovation process or a more favorable story about the effectiveness of compensation incentives. That's why looking and asking must be seen as a source of insights and hypotheses and not of definitive results. Looking and asking is not an end in itself -- but it can be a very useful beginning.

Cambridge MA
January 2000

* Professor of Economics, Harvard University, and President of the National Bureau of Economic Research. These remarks were prepared for the session on the NBER-Sloan Productivity Project at the January 2000 meeting of the American Economic Association.