Effects of re-invention on industry growth and productivity: evidence from steel refining technology in Japan, 1957–1968

Abstract

This paper examines the economic impact of re-invention – the degree to which an innovation is modified by user – on industry growth and productivity. The paper focuses on two re-inventions made by a Japanese steel company; these inventions improved the productive efficiency of Austrian-made refining technology, namely basic oxygen furnace (BOF). Results obtained from the plant-level production function estimation indicate that re-inventions account for approximately 30% of the total factor productivity of the BOF, substantially promoting the dissemination of the BOF technology. Our simulation analysis indeed reveals that re-inventions contributed to steel output growth by about 14%. This paper also documents that innovating companies played the role of a ‘lead user’ in developing and disseminating their re-invented technologies.

Keywords:

• re-invention
• lead user
• total factor productivity
• steel

Acknowledgements

We thank Hiroyuki Odagiri, Arnold Picot, Eric von-Hippel and seminar and conference participants at the Japanese Economic Association meetings, LMU and the HBS-MIT User Innovation Workshop for helpful comments.

Notes

The process of technological improvements is sometimes called by other terms, including follow-on innovations, accumulated improvements, or incremental innovations. In this paper, we collectively call re-invention, following Rogers (2003).

While it was freely disclosed in the domestic market, Yawata licensed its re-invented technologies to foreign competitors under royalty agreements.

Lerner and Tirole (2002) attempt to explain this behaviour in the context of open source software development.

We assume that MHL _{i, t} (or OG _{i, t} ) takes a value equal to the proportion of the furnaces equipped with the MHL (or the OG) systems in plant i in year t. Our estimation results discussed in this section are quantitatively unaltered under the alternative assumption that the variable takes the value of 0.5, when some but not all furnaces in plant i adopted the corresponding re-inventions.

The stability of market share is often observed in other industries in Japan. See Sutton (2007) for details.

An alternative method to control for unobserved productivity is to create a proxy for u _it by introducing an input demand equation from outside the production function framework. A previous version of this paper attempted to apply this method and reports that the infrequency of investment fails to use the Olley and Pakes (1996) method and that the use of material input (pig iron and scrap in our case), as per the idea adopted from Levinsohn and Petrin (2003), generates unreasonable productivity estimates. The Levinsohn–Petrin approach has also been recently criticized by Ackerberg, Caves, and Frazer (2005). Based on these findings in the previous version, this paper does not employ these methods to control for unobserved productivity.

While Arellano and Bond (1991) propose an alternative method to address serial correlation, the method is known to have poor performance with small sample size, the property of the data set which our study would likely belong to.

Our data set is unsuitable for testing a hypothesis related to wage premium and human capital. The purpose of the discussion in this paper is to illustrate the importance of controlling for self-selection in the choice of technology.

Yawata installed the OG system for two BOFs out of a total of seven furnaces in 1962; thus takes the value of 0.286. The value of 12.6 is obtained by multiplying 0.286 by the estimate of .

Our simulation exercises do not allow for plant entry and exit. It is probably unreasonable to consider that the absence of re-invented technologies triggers a plant's entry, which is a decision that involves large sunk costs.

Alternatively, we could assume that the firm maximizes its profits by solving its allocation problem across plants. Although this alternative approach may be more realistic, modelling the multi-plant feature requires complex computational issues, which are beyond the scope of this paper.

The steel production process converts pig iron and scrap into crude steel. Thus, our price measure p _it is the price of crude steel, netted out of the sum of the pig iron and scrap prices.

Note that we do not provide confidence interval for , which is complicated non-linear combination of estimates.