VoxEU Column

Measuring robot quality: Slowing improvement is a possibility

Faculty Fellow, RIETI

PhD student, Pennsylvania State University

For KIMOTO Ryo's full bio,

Professor, Faculty of Economics, Keio University

For SHIRATSUKA Shigenori's full bio,

SHIROTA Toyoichiro
Associate Professor, Aoyama Gakuin University

For SHIROTA Toyoichiro's full bio,

A large literature explores the implications of the introduction of robots into the workplace for the labour market, yet few studies exist which measure robot quality over time. This column uses data from the Japan Robot Association and the Bank of Japan to document a significant decline in the improvement in robot quality in Japan in the last decade. The differences in the growth rates of robot quality between the 2000s and 2010s are large, at around -3 percentage points per year.

As the introduction of robots into the workplace increases, there is a growing concern over whether robots will cause human jobs to disappear. In response to this societal fear, academics have tackled this issue from both theoretical and empirical angles (e.g. Acemoglu and Restrepo 2017, Baldwin 2019, Dauth et al. 2017, Michaels and Graetz 2015).

However, to date, no study has specifically investigated the rate of technological progress, namely, the quality improvement of robots. For any attempt to predict how robots will affect the macroeconomy, in recognition of society’s existing anxiety, it is vital to understand the progress of robot production and the quality improvement path of robots. If the pace of quality improvement in robots slows down or has already diminished, fear regarding robots taking human jobs away may dissipate. In a new paper (Fujiwara et al. 2021), we aim to fill in this gap.

Our study uses two novel datasets – Production and Shipments of Manipulators and Robots collected by the Japan Robot Association and the Corporate Goods Price Index from the Bank of Japan – to measure the amount of progress made in improving robot quality in Japan between 1990 and 2018. First, we construct quality-unadjusted robot price indices using the Production and Shipments of Manipulators and Robots dataset and three techniques: index number, stochastic, and structural approaches. We then measure quality per robot by dividing this quality unadjusted price index by the Corporate Goods Price Index, an industrial robot price index that is quality-adjusted.

Figure 1 showsa the evolution of quality per robot estimated using the three approaches. Despite different approaches being used, there is no significant difference in trends. The pace of quality improvement per robot has slowed or decreased significantly since 2010. The rate of quality improvement per robot in the 2010s was around three percentage points per annum lower than in the 2000s.

Figure 1
a) Index number approach
Figure 1 a) Index number approach
b) Stochastic approach
Figure 1 b) Stochastic approach
c) Structural approach
Figure 1 c) Structural approach
Note: All measures are in logarithmic scale and normalised to zero in 1990.

The result of the decline in the rate of quality improvement of robots may be in line with the findings of the recent studies by economists at the IMF and Federal Reserve such as Byrne and Pinto (2015) and Lian et al. (2019), which point to a decline in investment-specific technological progress, i.e. a slowdown in the pace of decline in the relative price of capital goods to consumer goods. The main conclusion also implies that the hypothesis that ‘ideas are getting harder to find’, advocated by Bloom et al. (2020), may apply to robot production.

As the estimates are based on various assumptions, the results should be treated with a certain degree of caution. Micro-level data for prices and product characteristics for individual robots are needed for more rigorous quality adjustments. Furthermore, this analysis does not capture the expansion of the range of robot applications due to advances in software, including algorithms and other factors. Measuring service flows from such intangible capital remains an issue for future studies.

This article first appeared on www.VoxEU.org on May 19, 2022. Reproduced with permission.


May 23, 2022

Article(s) by this author