RFF Workshop on LBD in Energy Technologies. Marvin B. Lieberman

Size: px
Start display at page:

Download "RFF Workshop on LBD in Energy Technologies. Marvin B. Lieberman"

Transcription

1 RFF Workshop on LBD in Energy Technologies Presentation by: Marvin B. Lieberman UCLA Anderson Graduate School of Management June 17, 2003

2 Learning-by-Doing Learning Curves a pervasive empirical regularity Often applied at different levels of aggregation: - Learning Curve workers, tasks, or teams - Experience Curve firm-level, product-specific - Technological Progress Function industry level

3 Convenient Representation of Micro-level Processes Aggregation of many effects resulting series look smooth and regular when plotted in log-log space Strong empirical regularity, but lacking an underlying economic theory Some theoretical models based on search processes with decreasing returns

4 An Example of Industry-Level Learning Curves 37 industrial chemicals from Lieberman (1984) Time period: 1950s to 1972 Plot of market price vs. industry cumulated output

5 Unit Price vs. Industry Cumulative Output for 37 Industrial Chemicals Log (Deflated Market Price per Pound) Log (Cumulative Output)

6 Distribution of Estimated Learning Curve Slopes for the 37 Products Number of Products % 60% 70% 80% Estimated Slope 90% 100%

7 What Drives the Learning Curve? Some key questions: What is the appropriate horizontal axis? What determines the learning rate or slope? How consistent is this rate across products? How consistent is the rate over time? How does learning differ from R&D, scale economies, etc. Can a more complex/accurate learning function be characterized?

8 What Drives the Learning Curve? Industry-level learning: Is the driver cum. output, or cum. investment or time? Economies of scale vs. learning Technical change R&D vs. learning Role of spillovers or knowledge transfers among companies

9 Unit Price vs. Industry Cumulative Output for Anhydrous Ammonia Log (Deflated Market Price per Pound) Reciprocating compressors introduced 80% curve Log (Cumulative Output)

10 What Drives the Learning Curve? Firm-level learning: Learning rates vary among organizations Learning should not be viewed as a uniform or automatic process Recent studies on transfer across organizational units or teams Recent studies on organizational forgetting (knowledge decay)

11 Investment Component of Learning If more output today reduces cost in in the future, one can consider part of the current as as investment in cost reduction. Reason to subsidize part of current cost Private firms may subsidize under some conditions

12 Unit cost Investment Value of an Increase in Current Output Learning curve Cumulative volume

13 Unit cost Investment Value of an Increase in Current Output c 1 x 1 Cumulative volume

14 Unit cost Investment Value of an Increase in Current Output c 1 Final c 2 x 1 Cumulative volume x 2 Final period

15 Unit cost Investment Value of an Increase in Current Output c 1 Total cost = area under curve Final c 2 x 1 Cumulative volume x 2 Final period

16 Unit cost Investment Value of an Increase in Current Output c 1 If produce one more unit today at current c 1 Final c 2 x +1 1 Cumulative volume x 2 Final period

17 Unit cost Investment Value of an Increase in Current Output c 1 Total cost rises only by final cost c 2 Final c 2 x 1 x Cumulative volume 2 +1 Final period

18 Unit cost Investment Value of an Increase in Current Output c 1 So, the difference between current cost and final cost can be viewed as an investment Final c 2 x 1 x Cumulative volume 2 +1 Final period

19 Unit cost Investment Value of an Increase in Current Output c 1 Producing more today reduces the cost of units produced in the future Final c 2 x 1 x Cumulative volume 2 +1 Final period

20 Factors that Diminish the (Private) Investment Value Temporal discounting Uncertainty Inter-firm spillovers

21 Spillovers Have Mixed Effects If learning remains proprietary, firm obtains future cost reduction, relative to rivals If learning spills over to rivals, firm does not get this benefit of investment So, spillovers reduce the private incentive to subsidize learning But spillovers also increase the public benefits from learning get more widespread cost reduction in the industry Rationale for public subsidy

22 Some Business Implications BCG promoted aggressive learning-based strategies in 1970s; failed for many clients. Few firms formally estimate learning curves, although critical for cost forecasting in some industries (e.g., airframes). Contractors and consultants often play an important role in diffusing technology. R&D consortia promote spillovers in some industries. Benefits greater if diverse players bring complementary skills.

23 Some Business Implications Recent studies highlight factors that influence the rate of organizational learning; less research emphasis on industry-level learning. LBD vs. micro processes: To speed cost reduction, is it better to focus on expanding output or promoting specific types of improvement?