Over the last couple of months I have been banging our drum about the importance of flexibility in manufacturing, the value of flexibility in manufacturing, and why human-machine collaboration is one way to provide this flexibility.
It all comes down to the fact that manufacturing as currently practiced — with poor human-machine collaboration — can be suboptimal and not sufficiently responsive to the long-term trends of shorter product life cycles and increasing product diversity. In other words, the inflexibility of manufacturing today is costly.
The current incremental model of manufacturing productivity improvement (for example, kaizen1) is not enough to generate value and long-term growth. Productivity in manufacturing has hit a wall and jumping over this wall will require novel approaches2 to introducing flexibility in factories and workshops.
Practically, we would expect some evidence of this inflexibility to show up in the numbers — in the macro, as proxied by productivity statistics, and in the micro, with poor factory-level economics. In the discussion that follows we explore some high-level findings on how productivity growth has stalled across the economy—particularly in manufacturing—before we move on to a discussion at the micro level in a subsequent post.
Productivity Growth: A General Macro View
Measurement of productivity and productivity growth relies on the classical economist’s “growth accounting” toolbox, first developed by Robert Solow in the 1950s. In its simplest formulation, GDP growth (or growth in a component of GDP) can come from:
- Increases in the quantity of labor;
- Increases in the quantity of capital;
- An increase in growth not explained by increases in the quantities of labor or capital, called “total factor productivity,” or TFP. This “measure of our ignorance” is a metric that captures the interaction between the other components of labor productivity not directly measured or measurable.
This TFP (also known as Multifactor Productivity, MFP) describes the efficiency gains (or losses) associated with growth (or decline) in output that is not a result of changes in measured inputs. It relates the change in an industry’s real output to changes in the combined inputs used in producing that output. Increases in TFP can result from improvements in technology (e.g., better software or more effective automation), improvements in managerial practices, a more educated workforce, reallocation of resources from sectors that are less productive to those that are more productive, and other unmeasured (or unmeasurable) factors.
TFP and Productivity
Northwestern University professor Robert Gordon, in his monumental 2016 book “The Rise and Fall of American Growth: The US Standard of Living Since the Civil War,”3 postulates that TFP is the key to the puzzle of low productivity growth.4 The other drivers of growth are exogenous and will vary with hours worked, investment rates, and business cycles. Only TFP growth is a clear indicator that labor productivity is rising permanently and independently of its drivers.
Gordon asserts that the surges of TFP between 1920–1970 and 1994–2004 are anomalous and that TFP levels between 2004–2014 are a better indication of the future (Figure 1). The 1920–1970 numbers can be attributed to the Second Industrial Revolution’s general purpose technologies (GPTs): electricity and the internal combustion engine,5 and the other bump in 1994–2004 was driven by “an unprecedented and never-repeated rate of decline in the price of computer speed and memory and a never-since-matched surge in the share of GDP devoted to investment in information and communication technology (ICT).”6
It is true that these two periods experienced drastic technological changes that are largely responsible for spikes in TFP. But, counter to Professor Gordon’s stance, the fact that those trends were unprecedented does not preclude something similar happening in the future. It is entirely possible that we will experience another technological renaissance that produces a spike in TFP. Consider, for example, Thomas J. Watson’s or even Bill Gates’ pronouncements on how large the computer market could be.
Professor Gordon does not dispute the frenetic pace of innovation in digital technology, robots, automation, and AI. However, he notes that the fast pace of innovation does not necessarily mean these innovations are having an impact on the growth rates of labor productivity and TFP. If we are to trust Professor Gordon, productivity growth in the overall economy has stalled, it has hit a wall. How does that look for manufacturing?
When inflexible automation collides with manufacturing trends requiring more flexibility, productivity at the macro aggregate suffers.
Manufacturing Productivity Growth—Hitting a Wall
TFP growth in the US manufacturing sector has been challenged over the last few decades. Figure 2 from the US Bureau of Labor Statistics (BLS) shows YoY changes in TFP, with a noticeable slowdown in the post-2004 period.7 TFP in the manufacturing sector grew by an average of 2.0% per year from 1992 to 2004, with manufacturers increasing their production of goods with relatively fewer inputs. From 2004 through 2016, however, manufacturing TFP declined by an average of 0.3% per year. The reasons for this slowdown in TFP growth have been widely studied and have been attributed by many fine economists (including Professor Gordon above) to the “one off” semiconductor and IT revolution driven by (now-waning) Moore’s Law.8
In a different cut, the team at the BLS disaggregate the components of manufacturing TFP at the industry subsector level. They conclude that, since 2004, semiconductors, electronic components, and computer and peripheral equipment manufacturing have been the biggest contributors9 to TFP slowdown, consistent with the “end of Moore’s Law” thesis.
However, looking at the average TFPs for a select number of manufacturing subsectors, namely, durable goods such as automobiles, white goods, and other capital equipment (Figure 3), we can see that there has been a generalized productivity slowdown relatively independent of that in computers and semiconductors. Clearly, the decline in manufacturing TFP is not driven only by computer and semiconductor manufacturing TFP, so we must dig a bit deeper.
Figure 3: Average TFP 1992-2004 and 2004-2016 for US durable goods manufacturing.
Fortunately, the BLS research team thought the same. They carried out an additional piece of valuable analysis looking at the economic factors of production used in manufacturing—capital, labor, materials, purchased services, and energy.
Factors of Prodcution
Under the hypothesis that interaction among the factors of production affects TFP, they compare the shares of the factors of production between 1992–2004 and 2004–2016 (Figure 4). And what they find is that the share of capital grew from 21% to 24%, while the share of labor dropped from 32% to 25%, which is consistent with increased (but not necessarily more efficient) automation. Additionally, the share of materials and purchased services combined grew from 45% to 49%, which is consistent with increased outsourcing and external purchasing of materials and services.
According to the BLS team, these changes in factor shares were not driven by only a few large industries. Of the 86 manufacturing industry groups analyzed, only 14 industries reported growth in the average share of labor in production between the two periods. More importantly, changes in factor shares appear to be related to changes in TFP growth. The average annual rate of 2004–2016 TFP growth was lower than during 1992–2004 in 62 out of the 86 manufacturing industries. Of these 62 industries, 53 had decreases in the average factor-share weight of labor, which is likely a result of higher automation and outsourcing.10
So we can infer that the replacement of capital for labor in manufacturing (i.e., lower labor intensity in production) is a culprit in the TFP slowdown of the recent decades. The BLS team points out that if TFP growth is a result of innovation, and if more of the production process happens without the input of workers, the result is less manufacturing innovation.11
But this shift toward more capital and less labor is also indicative of higher automation levels — and, as we like to repeat, too much automation, or the wrong kind of automation, is also inflexible. Only when this additional automation combines with labor to generate higher productivity (that is, higher TFP) will automation be a net gain.12 When inflexible automation collides with manufacturing trends requiring more flexibility, productivity at the macro aggregate suffers.
In Part 2 of this article, I will revisit some of the micro evidence we have previously presented and examine it through the lens of productivity.
Editors Note: Robotics Business Review would like to thank Alberto Moel for permission to publish this piece (lightly edited). The original can be found HERE. All views, thoughts, and opinions expressed therein belong solely to the author.
1 Approaches, of course, such as the Veo FreeMove system.
2 Professor Gordon’s book is extremely well-articulated and provides an outstanding survey of technology innovation over the last two centuries, but often extrapolates historical incidents to future dynamics—which is to say, take its predictions with a grain of salt.
3 And possibly rising inequality, but a discussion for another day.
4 At an annualized growth rate of 1.89%, TFP rose 2.5x in the period 1920–1970.
5 The Second Industrial Revolution created a surge in productivity growth that lasted for 50 years, while the advent of widespread computing had a weaker and more delayed effect, lasting only about 10 years. This delay was well-noted by Robert Solow in 1987 with his classic quip that “you can see the computer age everywhere but in the productivity statistics.” It took a while, but they were indeed reflected.
6 Michael Brill, Brian Chansky, and Jennifer Kim, “Multifactor productivity slowdown in U.S. manufacturing,” Monthly Labor Review, U.S. Bureau of Labor Statistics, July 2018. The data and some of the analyses in this section are taken from this paper.
7 See, for example, David Byrne, Stephen Oliner, and Daniel Sichel, “Is the information technology revolution over?” International Productivity Monitor, no. 25, Spring 2013, pp. 20–36.
8 The team at the BLS also found that pharmaceuticals, medicines, and petroleum and coal products were important factors in the TFP slowdown. Thankfully, they provided a helpful appendix showing average TFP growth rates for specific subsectors.
9 A team from the Brookings Institution also identified outsourcing as a causal factor in the TFP slowdown: Susan Helper, Timothy Krueger, and Howard Wial, “Why does manufacturing matter? Which manufacturing matters? A policy framework,” Brookings Institute, February 2012. Several other researchers from Cornell University’s ILR School credited manufacturers’ outsourcing to staffing services (now more or less complete) as having contributed as much as or even more than IT to manufacturing labor productivity growth in the 1990s to 2006: Matthew Dey, Susan N. Houseman, and Anne E. Polivka, “Manufacturers’ outsourcing to staffing services,” ILR Review, vol. 65, no. 3, 2012, pp. 533–59.
10 There is empirical evidence for this. For example, Baptist and Hepburn find that higher labor input (and lower intermediate inputs such as materials and energy) is positively (and possibly causally) related to TFP growth (Simon Baptist and Cameron Hepburn, “Intermediate inputs and economic productivity,” Philosophical Transactions of the Royal Society, January 28, 2013). This argument is also made by David Mindell in his 2015 book “Our Robots, Ourselves,” which argues against “lights-out” factories—if no humans are involved in the production process, there can be no manufacturing innovation and improvements.
11 As this is the Veo Robotics blog, we’d like to remind our kind readers that the Veo FreeMove system is a form of capital investment that is designed to make human-machine collaboration more productive. In other words, to make better use of capital and labor to increase TFP.
Alberto Moel is the VP of Partnerships and Strategy at Veo Robotics, where he is responsible for industry partnerships, intellectual property, market and competitive analysis, and company strategy. Prior to Veo Robotics, Moel was a Senior Research Analyst at Sanford C. Bernstein in Hong Kong, where he covered Asian high technology companies in the automation, robotics, and manufacturing technology sectors.
Previously as a principal in the Hong Kong office and co-leader of the Tokyo office of Monitor Group, Moel specialized in technology strategy and corporate finance for Asian high-tech companies. In addition, he has been a Professor and Lecturer at the Hong Kong University of Science and Technology and at Harvard Business School, co-founded a successful Brazilian hedge fund, and worked at JP Morgan (New York and Mexico City), Toshiba Corporation in Japan, and IBM Corporation in New York, Cambridge, and Zurich. Moel has SB, SM, and ScD degrees from MIT, and an MBA from Harvard Business School.
- China Source of 50% of Collaborative Robotics Shipments by 2023
- Veo Robotics’ FreeMove Turns Any Industrial Robot Arm Into a Collaborative Robot
- Collaborative Robots and Lean Integrators Bring Automation to Small and Medium-Size Enterprises
- Yaskawa Motoman Releases the HC20XP, the First Food-grade Collaborative Robot
- Universal Robots Launching America’s Largest Virtual Cobot Expo