Daniel l. Rubinfeld

236 Part 2 Producers, Consumers, and Competitive Markets
:h, ... "jr",~ 7 The Cost of Production 237
firm looking only at the high initial labor requirement -will obtain an m"erly pessimistic
vie-w of the business. Suppose the firm plans to be in bus~ness for "a long
time, producing 10 units per year. Suppose the total labor reqUlrement tor the
first year's production is 10. In the first year of production, the firm's cost v"ill be
high as it learns the business. But once the learning effect has taken place, production
costs will falL After 8 years, the labor required to produce 10 units will
be only 5.1, and per-unit cost will be roughly half what it was in the first year of
production_ Thus the learning curve can be important for a firm deciding
whether it is profitable to enter an industry.
100
Production
Hours
per Aircraft 80
60
JO
Awrage for First 100 Aircraft
20
Suppose that as the manager of a firm that has just entered the chemical processing
indush,)" you face the following problem: Should you produce a relatively
Im'\' level of output and sell at a high price, or should you price your
product lower and increase your rate of sales The second alternative is appealing
if there is a learning CUIye in this industry In that case, the increased volume
will lower your average production costs over time and increase the firm's
profitability.
To decide what to do, you can examine the available statistical evidence that
distinguishes the components of the learning curve (learning new processes by
labor, engineering improvements, etc.) from increasing returns to scale. For
example, a study of 37 chemical products reveals that cost reductions in the
chemical processing industry are directly tied to the growth of cumulative
indush'Y output, to inveshnent in improved capital equipment, and, to a lesser
extent, to economies of scale ll In fact, for the entire sample of chemical products,
average costs of production fall at 5.5 percent per yeaL The Shldy reveals
that for each doubling of plant scale, the average cost of production falls by
11 percent For each doubling of cumulative output, however, the average cost
of production falls by 27 percent The evidence shows clearly that learning
effects are more important than economies of scale in the chemical processing
industryY
The learning curve has also been shown to be important in the semiconductor
indushy A study of seven generations of dynamic random-access memory
(DRAM) semiconductors from 1974 to 1992 fOlmd that the learning rates averaged
about 20 percent; thus a 10-percent increase in cumulative production
OL-______ L-____ ~ ______ -L ______ ~ ____ ~ __ _
o 100 200 300 400 500
Number of Aircraft Produced
The learning curve relates the labor requirement per aircraft to the cumulative number
of aircraft produced. As the production process becomes better organized and
workers familiarity with their jobs, labor requirements fall dramatically
would lead to a 2-percent decrease in cost,u The study also compared learning
bv firms in Japan to firms in the United States and found that there was no dis­
~guishable difference in the speed of learning.
Another example is the aircraft industry, where studies have fOlmd learning
rates that are as high as 40 percent This is illustrated in Figure 7_14, which
shows the labor requirements for production of aircraft by Airbus Industrie.
Obsen"e that the first 10 or 20 airplanes require far more labor to produce than
the hundredth or two hundredth airplane. Also note how the learning cun'e
flattens out after a certain point; in this case, nearly all learning is complete
after 200 airplanes have been built.
Learning curve effects can be important in determining the shape of longrun
cost curves and can thus help guide managem_ent decisions. Managers can
use learning CUITe information to decide whether a production operation is
profitable and, if it is, how to plan how large the plant operation and the volume
of cumulative output need be to generate a positive cash flow.
11 The studv was conducted by MaITin Lieberman, "The Learning Cun"e and Pricing in the Chemical
Processing"Industries," RAND JOllmal of Economics 15 (198J): 213-28_
12 The author used the average cost AC of the chemical products, the cumulati\:e industry
output X, and the a\-erage scale of a production plant Z. He then estimated the relatlOnshlp log
(AC) - 0387 log (X) - OJ73 log (Z). The - 0387 coefficient on cumulative output tells us that f~~
e\"ery I-percent increase in cumulative output, average cost decreases 0387 percent. The - 0.1/~
coefficient on plant size tells us that for e\-ery I-percent increase in plant size, cost decreases 0.17J
percent
By interpreting the two coefficients in lio-ht of the output and plant-size variables, we can allocate
about 15 percent of the cost reduction to i~creases in the average scale of plants and 85 percent to
increases in cumulative industry output. Suppose plant scale doubled while cumulati\-e output
increased by a factor of 5 during the study. In that case, costs would fall by 11 percent from the
increased scale and by 62 percent from the increase in cumulati\"e output
A business that is expanding or contracting its operation must predict how costs
will change as output changes. Estimates of future costs can be obtained from a
cost function, which relates the cost of production to the level of output and
other variables that the firm can control.
13The study \,-as conducted by D A. Irwin and P J Kleno,,", "Learning-by-Doing Spillo\-ers in the
Semiconductor Industry," Journal of Political Eco1l01I1I/102 (December 199J): 1200-27
cost function Function relating
cost of production to le\'el
of output and other variables
that the firm can controL