Use It or Lose It: Learning on the Line

Research by C. Lanier Benkard, Assistant Professor of Economics, Stanford Graduate School of Business

Learning is probably one of the most important concepts in business because it drives changes in technology and productivity. It is how companies improve. Many economists have studied "learning-by-doing," the idea that the more a firm produces, the more its production costs decrease. However, people have had the idea that when companies learn, they never forget.

That's not the case, says C. Lanier Benkard, assistant professor of economics at the Graduate School of Business. Benkard has shown that in certain industries there is evidence of organizational forgetting-that is, production experience actually depreciates over time, and knowledge gained from building one product doesn't necessarily spill over to the next generation.

Benkard draws on unique data from the commercial aircraft industry. Because units of production are huge in the industry, learning can be measurably observed as a firm goes from one unit to the next. The same is true in many other manufacturing industries and even in services.

Benkard also identifies characteristics that are likely to lead industries to higher rates of forgetting. They include products that are labor intensive, learning that is thought to be important at the individual worker level, and high turnover.

Benkard was able to get access to once highly proprietary data-the production records of 250 Lockheed L-1011 Tri-Star passenger jets-because Lockheed has withdrawn from the commercial jet market. First, he looked at man-hours. He found that the first L-1011 jet Lockheed produced required 1.5 million man-hours to build, but the 100th plane four years later took only 220,000 hours, a savings of 85 percent. That's strong evidence of learning-by-doing.

But he also found that the dynamics of production are much more complex than previously thought. Unlike military aircraft production, which has been studied extensively, commercial manufacturing is subject to many uncertainties, such as swift recessionary drops in market demand. In the fourth year of production, for example, Lockheed made more than 40 Tri-Stars. In the seventh year, they made only seven. In the 10th year, they built 30. This variance in production rates does not show up in military production.

However, by looking at commercial data, Benkard was able to prove a hunch-that learning rates vary a great deal because of the forgetting phenomenon. He found that the forgetting rate was high in commercial aircraft construction. From one year to the next, the firm retains only 60 percent of its experience. Turnover plays a key role.

"If anything, forgetting increases during times of low production," says Benkard. "Often lower production rates are accompanied by inactivity and layoffs."

There are also different kinds of turnover. If workers change jobs within the company, they have to be retrained, as do the workers who take over their old jobs. It's wise to promote people but keep them on the same team so their experience is not lost.

Benkard also examined how much learning spills over from one product to another. The data contained information for several different models of airplanes produced at the same time in the same plants and on the same production line. Benkard found that 70 percent of experience spilled over across aircraft generations, which is low given the similarity between the aircraft.

Benkard's findings about organizational forgetting have implications for many areas of industrial organization, including pricing, industrial structure, and industrial policy. At the macroeconomic level, forgetting implies that recessions may lead to a reduction in productivity that lasts well into the next economic boom. If the timing is bad, a firm may be least productive at the peak of a boom, which is what happened when Boeing laid off workers in the early 1990s. When the next boom hit in the late 1990s, it was unable to ramp up production and had to temporarily shut down two production lines. In terms of the incomplete spillover of knowledge from one product to another, competition often forces firms to produce many variants of their products in order to make sales.

There is reason to believe, says Benkard, that the number of products may be too high. If so, that would also generate more support for a restrained antitrust policy. Benkard concludes that the rate of forgetting is under a firm's control. For example, avoiding layoffs may be in the firm's best interest even if workers are underutilized. If layoffs are unavoidable, the firm may want to make a provision to rehire workers in a future period to avoid the wasteful loss of experience.

November 1999

Learning and Forgetting: The Dynamics of Aircraft Production, C. Lanier Benkard May 1999

Click here to view the original article.