Aug. 19 (Bloomberg) -- Where do productivity gains come from? Economists have known for decades that the broadest measure of efficiency -- known as total factor productivity -- drives long-term income growth and prosperity. Progress occurs as firms figure out how to boost output without having to hire more workers or install more capital.
If total factor productivity sounds like something of a black box -- a quasi-magical ingredient that creates output out of seemingly nothing -- well, it sort of is. Sure, economists have some ideas about its sources: adoption of new technologies, better management practices, and improvements in production chains. But a more detailed understanding is emerging.
In a recent study, we focused on a major automobile assembly plant and the evolution of its production-defect rate in the course of a year. We reviewed production records for almost 200,000 vehicles. That inquiry allowed us to dig deep into an oft-cited source of productivity growth: learning by doing.
As the name implies, learning by doing is achieved through production itself. Workers figure out how to do their jobs better, managers allocate capital more efficiently, and so on. Previous research focused on measuring how fast learning-related productivity improves. Thanks to extraordinarily detailed data, our study moves beyond showing that learning happens, and instead explores the specific mechanisms through which it takes place.
For each car, we observed what sorts of production problems, if any, occurred. We examined each of the several hundred operations involved in assembly. We don’t reveal the plant’s owner or location for confidentiality reasons.
We found clear evidence of learning by doing at the plant: average defects per car fell more than 80 percent during the production year. (The next time you buy a car, consider in which part of the production year it was made.) This pattern is consistent with earlier research documenting big productivity gains from learning.
What’s more interesting is how learning happened. For example, this quality improvement didn’t come from addressing just the most defect-prone processes, even though the most troublesome 20 percent of processes accounted for 90 percent of all defects. Instead, such rates fell about equally across the board.
When the plant increased production of a new model variant, the learning process started anew. This happened three times a year, as the plant annually built a series of models on a common chassis, staggering the start of each.
Initial defect rates on a new model were well above those on the designs already being produced. Quality improvements occurred on new models at the same rate as for the earlier ones, but some negative spillovers were involved. Defect rates on vehicles already being produced rose temporarily during the increase in production of new models, ramp-up, as resources were redirected toward solving problems associated with the later brands.
One of the clearest patterns in the data was that that most “know-how capital” brought by learning wasn’t not bottled up in the plant’s individual workers, but rather incorporated into its physical or organizational capital. Two key pieces of evidence point to this. When the plant’s second shift started several weeks after the first one had begun, the second-shift defect rates were no higher than those being experienced on the first shift, even though most second-shift workers hadn’t yet been on the line that year.
Similarly, while worker absences slightly raised defect rates, their impact was, practically speaking, very small. Something bigger than each worker’s experience was at play. Broader changes, such as altering the layout of tools at the workstations and regrouping the sequence of operations, made the difference. Individual workers” suggestions quickly turned into new institutional practices. Total factor productivity rose as the plant adopted fresh ways of doing things.
It is important to recognize that our results reflect the learning process at one plant, in one industry, at one time. Some aspects of learning may work differently in other settings. Nevertheless, we believe managers can draw insights from our work and extend them to other production settings.
In business, swifter development and process improvements represent key drivers of success. Opening the black box of productivity growth to understand the sources of learning by doing, rather than just assuming that learning happens by itself, provides managers with guidance on ways to gain operational efficiencies earlier when production is increased, in the ramp-up process and offers insights into the underlying sources of economic growth.
(Chad Syverson is a professor of economics at the University of Chicago’s Booth School of Business. Steven Levitt and John List are economics professors at the University of Chicago.)
To contact the writer of this article: Chad Syverson at email@example.com
To contact the editor responsible for this article: George Anders at firstname.lastname@example.org.