Learning Curve

Introduction to Environmental Economics

Page 4:  Learning Curve

A learning curve (also called experience curve) shows a lowering cost of producing each unit of a product as more units are produced (see “economies of experience” in John Kenneth Galbraith, American Capitalism, p. 34). The lowering of cost does not always happen, and when it does happen it is usually temporary (cost per unit eventually levels out and does not keep going down as more units are produced).

“As the cumulative number of units produced becomes large, the learning effect is less noticeable.”
Krajewski et al., Operations Management, p. I-6

For example, when manufacturing an automobile, the first car off the assembly line cost more to produce than the second car, which cost more to produce than the third car, etc. Eventually, the ten thousandth car may cost about the same to manufacture as the next car to produce, and so on.


In graphs depicting a learning curve, the horizontal axis from left to right represents cumulative quantities and/or time, and the vertical axis upward represents cost per unit. A standard learning curve shows cost of production dropping as time increases, until cost levels out toward the right with increasing time.

An example learning curve is depicted as Figure I.1, in Supplement I of Lee J. Krajewski, Manoj K. Malhotra, Larry P. Ritzman, Operations Management: Processes and Supply Chains, Global Edition, 11/E. The portion of the graph in which cost per unit is dropping is labeled “Learning Period”, and the portion thereafter (toward the right) in which cost per unit is stabilizing is labeled “Standard Time”.

Learning Rate

An initial learning rate of 80 percent has been found to be common in many manufacturing industries, including the first industry in which learning curves were observed, which was the aircraft manufacturing industry. A learning rate of 80 percent indicates that cost per unit drops 20 percent upon doubling the total quantity ever produced (cumulative quantity).

PV manufacturing costs have been found to have the standard 80 percent learning curve, which reduced cost per watt by 20 percent for each doubling of the cumulative quantity of production.

However, PV manufacturing costs have started stabilizing, as commonly happens with the standard learning curve which PV manufacturing corresponds with. And doubling of cumulative quantities manufactured is not possible into the future for PV manufacturing, because so much electricity is required to manufacture photovoltaics that severe national electricity shortages would result if production was ramped up.

“To contain cost, many polycrystalline-Si and Si wafer producers place their factories near hydropower plants for cheap electricity. This strategy works only when the solar cell industry is small, since the availability of hydropower is limited.”
Meng Tao, Terawatt Solar Photovoltaics, p. 74

Lack of Breakthroughs

PV manufacturing has never had a major breakthrough. It has only had incremental improvements, not the kind of major breakthroughs that have been experienced in pioneering (revolutionary) industries such as the steel industry, semiconductors, etc.

Michael C. Jensen reviews some of the major breakthroughs of pioneering industries in A Theory of the Firm, p. 19. On page 23 he lists RISC processing as an example of a breakthrough in the information technology revolution:

“An example is the Reduced Instruction Set CPU (RISC) processor innovation in the computer workstation market. RISC processors bring about a tenfold increase in power, but they can be produced by adapting the current production technology.”
Jensen, A theory of the Firm, p. 23

RISC was a pioneering breakthrough developed from research at Berkeley and Stanford. It allowed processors to take up less semiconductor space, so that more processors could be manufactured in each batch (on the same amount of semiconductor material), and so that each processor could do the job that previously required multiple processors (allowing a single smaller processor to replace multiple larger processors).

“The Berkeley team concentrated on understanding the principles for achieving the most effective use of the area on a VLSI chip… the limited number of instructions required a relatively small amount of on-chip control logic… The Stanford team concentrated…on combining an optimizing compiler with…the pipeline technique to enable a number of instructions to be active at once.”
Silc, Robic, Ungerer, Processor Architecture, p. 3

That was just one of many pioneering breakthroughs in electronics. PV manufacturing has never experienced any such breakthrough, much less a string of such breakthroughs as the electronics industry has.

Not Moore’s Law

Media arguments in favor of ramping up photovoltaics manufacturing predict that PV manufacturing will correspond to the historic success of sharply reducing the manufacturing cost of pocket calculators and subsequent electronics, citing that PV and electronics both use semiconductor materials.

However, the semiconductor materials are used very differently (see preceding page of this report), hence the respective learning curves are very different. The following graph, which is page 5 of the slides (PDF) of Scott Elrod’s 2008 lecture at Stanford, shows the learning curve of electronics (called Moore’s law) compared to the learning curve of photovoltaics:

Moore’s law is much steeper (goes off-chart) while the PV learning curve corresponds to traditional learning curves that level out. And pretending that traditional learning curves do not level out is inaccurate. For example, PV manufacturers may not be able to secure further limited supplies of cheap electricity for manufacturing, further subsidies from an impatient public, etc.

“market or product changes can disrupt the expected benefits of increased production. For example, Douglas Aircraft management assumed that it could reduce the costs of its new jet aircraft by following a learning curve formula and committing to fixed delivery dates and prices. Continued engineering modification of its planes disrupted the learning curve, and the cost reductions were not realized. The resulting financial problems were so severe that Douglas Aircraft was forced to merge with McDonnell Company.”
Krajewski et al., Operations Management, p. I-2

Return to EnviroStudies.net

[+] Show Contents of this Report
Copyright © 2017 by 3D Software, All rights reserved
3D Software, P.O. Box 221190, Sacramento CA 95822 USA   3DSoftware.com   Contact us
Wednesday, 16-Jan-2019 05:01:57 GMT