The traditional neoclassical real business cycle model assumes that technology arrives as an exogenous process, after which labor productivity immediately responds positively until the economy eventually converges to the new steady state where labor productivity is permanently higher. However, David (1990), Rogers (1995), and Hall (2004), among others, have provided evidence that technology diffuses slowly throughout the economy. This means that a new technology is adopted by agents over time and that all agents do not adopt the technology immediately. This process of adoption and diffusion of technology takes the form of an S-shaped curve. That is, the technology initially diffuses slowly, followed by a period of rapid diffusion until the speed decreases when the technology has been absorbed in the economy. This view of slow diffusion therefore challenges the notion that technology shocks have immediate and positive effects on the economy. Furthermore, Robert Solow’s statement: “You can see the computer age everywhere but in the productivity statistics” clearly states how the literature lacks economic understanding of how productivity is affected by the arrival of new technology.
This paper will show, through use of vector auto regressions and more than a century of U.S. data, that labor productivity may respond negatively in the short run to a technology shock. This case can arise if the arrival of a new technology initiates high installation costs or a learning stage for the productive labor. During this stage labor productivity does not necessarily increase as assumed by the standard neoclassical models. Rather, labor productivity can actually fall below trend temporarily. After a time lag from when the technology was invented, the technology eventually becomes adopted in the economy and the inflection point of the S-shaped diffusion curve is reached. Inputs can then once again be active in the production process and it is likely that labor productivity will increase above trend.
The existing empirical literature which has focused on technological progress and subsequent productivity slowdown has mainly relied on simple graphical analysis. This paper therefore provides formal statistical evidence that the arrival of new technology can lead to a temporary slowdown in productivity using both an actual measure of technological progress and a long time-series as is important when studying productivity growth. Furthermore, the paper compares differences between the response of labor productivity to technological progress in the Electrification period and the more recent period, when the computer and the internet became widely adopted. The results from the post-World War II (post-WWII) period have important implications for understanding whether technological progress is the main reason for the productivity slowdown observed after 1973.
The focus of this paper is the response of labor productivity and other macroeconomic variables following the initial arrival of new technology. In this paper, technology is measured as new inventions that have experienced a patent application. Because of long diffusion lags, the main focus of this paper is not on long-run impacts on productivity but instead on possible adverse effects in the short run. Therefore, the paper does not argue that there are no positive effects on productivity from new inventions but instead argues that the positive effects may not arrive immediately after the invention of new technology. The contribution of this paper to the macroeconomic literature is therefore to supply empirical and statistical evidence for how aggregate variables historically have responded to technology shocks in the short run.
As explained in the next section of the paper, much theoretical research has addressed the subject of possible contractionary effects of technology shocks. Further, empirical methods have been employed within the applied microeconomics literature, but this question has not been adequately addressed with long macroeconomic time-series data. The time-series literature therefore lacks direct empirical evidence on the effects of changes in technology.
Contents
I. Introduction
II. Existing Literature
III. Data
IV. Methodology
V. Empirical Evidence
- A. Benchmark Specification
B. Robustness of the Results
VI. Pre- and Post-WWII
- A. VAR Estimation on Two Sample Periods
B. Post-WWII VECM
C. Variance Decomposition
VII. Theoretical Implications of the Results
VIII. Conclusion
Download
PDF Ebook Do Technology Shocks Lead to Productivity Slowdowns? Evidence from Patent Data
