At least as far back as Marshall, economists have suggested that organizations store and accumulate knowledge that affects their technology of production. The returns to this production knowledge are generally associated with the concept of learning by-doing. There is an extensive empirical literature that explores the relationship between production experience and plant productivity. Recent studies such as Bahk & Gort (1993), Irwin & Klenow (1994), Jarmin (1994), Benkard (2000), Thompson (2001), Thornton & Thompson (2001), and Cooper & Johri (2002) find that agents and organizations appear to become more productive as they gain experience at producing a particular product or service. The results from these studies suggest that a doubling of cumulative output generally leads to an approximately 20% reduction in unit costs of production. In contrast to these studies, which cover a diverse range of narrowly defined industries, this paper studies the importance of learning-by-doing for generating productivity dynamics within the manufacturing industry as a whole.
Several authors have recognized that learning-by-doing introduces an inter-temporal component to output decisions made by production units. Higher current production rates, which involve an increase in current costs with no immediate effect upon productivity, lead to a reduction in future costs of production through learning-by-doing. Consistent with this literature, this paper treats the stock of experience as a state variable under the control of the production unit, and explicitly models the decision of how much output to produce as a joint decision with how much production experience to accumulate.
Recent work has also emphasized the quantitative importance of ‘organizational forgetting’ where relatively distant production experience becomes less relevant over time. Argote et al. (1990) provide empirical evidence for this hypothesis of organizational forgetting’ associated with the construction of Liberty Ships during World War II. Similarly, Darr et al. (1995) provide evidence for this hypothesis for pizza franchises and Benkard (2000) provides evidence for ‘organizational forget-ting’ associated with the production of commercial aircraft. Under a slightly different specification for ‘organizational forgetting’, Irwin & Klenow (1994) provide evidence for spillovers of experience across successive generations of chips in the semi-conductor industry.
The key contribution of this paper is to provide estimates of learning-by-doing parameters using the first order conditions from a structural model (of plant behavior) that allows for the accumulation of production experience, consistent with the hypothesis of organizational forgetting. Importantly, the structural model allows a separation of the parameters controlling the productivity of experience (learning-by-doing) from those controlling the accumulation of experience (organizational forgetting). The particular nature of the structural model, and the log-linear functional forms used in the estimation, have the advantage that estimation does not require data on the stock of production experience. This considerably reduces the data requirements since there is no need to track production units since birth in order to construct a series for the stock of experience. This allows estimation of learning parameters in a much broader context than previous studies. Specifically, estimates of learning and forgetting rates are obtained using data on both four-digit manufacturing industries and a large sample of plant-level observations.
The results suggest that the dynamic structure implied by the structural model of learning by-doing is broadly consistent with the 4-digit manufacturing industry data. The results also indicate that estimated learning rates might be considerably lower than existing (aggregate) estimates, such as those in Cooper & Johri (2002). The results identify variation in estimated learning rates, even at the 4-digit industry level. This is a noteworthy result that has previously not been identified in the existing literature.
Download
Learning-by-Doing and Productivity Dynamics in Manufacturing Industries
