“Technologies are exploding and conjoining like never before … historically value has never been created this quickly”.
Economics Nobel Laureate Daniel Kahneman
SUMMARY
You can’t see the road ahead through the rear-view mirror. The future development of the global knowledge economy over the coming decade will be of an order of magnitude greater than over the past, owing to the maturity, scale, and convergence of a whole swathe of technologies: AI, 5G, IoT, Cloud, Big Data, and 3D additive manufacturing.
Despite this positive potential outlook, the conventional wisdom about future global growth, both before (e.g., the debate around secular stagnation) and after Covid-19, has been distinctly negative. This paper argues that conventional wisdom is wrong. The world economy stands at the beginning of an upward turning point in global productivity and growth from the mid-2020s onwards: The Great Inflection[1].
Introduction
Perhaps the number one question in economics pre-Covid-19 was why is productivity growth so slow when technological change is so fast? There was a malaise akin to a ‘Sword of Damocles’ hanging over the world economy with much talk of secular stagnation. And the technology-productivity disconnect matters for individual and national prosperity. In the words of the Nobel Laureate Paul Krugman, “productivity growth isn’t everything, but in the long run, it’s almost everything.” If US productivity growth hadn’t slowed since 2005, US GDP would now be roughly $3+ trillion larger, and household income would be $25,000 higher, and GDP per capita $10,000 greater. So, do we now have the Solow Redux, in the wake of the Solow Paradox in the 1980s, when it was said you could see computers everywhere but in the productivity statistics? The McKinsey Global Institute [2] has stated: “We may be seeing a renewal of the Solow Paradox of the 1980s with the digital age around us but not yet in the productivity statistics.”
In contrast to Solow Redux’s arguments, this paper’s contrary view is that we shouldn’t have expected an upturn in productivity growth before now, but should from now on, in the mid-2020s onwards. The maturation, scale, and convergence of technologies such as AI, 5G, IoT, Cloud, Big Data, and 3D additive manufacturing throughout the 2020s are likely to be transformative for vast swathes of the world economy.
However, we shouldn’t be surprised that economists aren’t seeing this in advance. Too many economists are making the same mistake they always make, and fall foul of, by extrapolating the recent past into the future. They’re missing the fundamental point that you won’t see the road ahead in the 2020s through the rear-view mirror of the 2010s.
Weak growth yesterday and today doesn’t beget the same again tomorrow. There is a fragile economic relationship between productivity growth [3] in one decade and the next. Past productivity performance is no guide to the future. At the beginning of the 19th century, Thomas Malthus (asserting the conventional wisdom) made what was, without doubt, the worst forecast in economic history. He completely missed the Industrial Revolution’s impact and forecasted the continuation of a centuries-old stagnant economy, even though that a revolution was already underway. Failing to foresee The Great Inflection could yet become the second-worst economic forecast in history!
Technology optimism
While the economists have been fretting about secular stagnation, the technologists have been telling us that we are on the cusp of a technological revolution. Amazon founder, Jeff Bezos, says [4] that we stand at “the edge of a golden era.” We are told that the scale of potential change is simply staggering, with speculation – somewhat unrealistic – that we may see more technological change over the next 20 years than over the past 200. Eric Schmidt, former Executive Chairman of Google and Alphabet, also says [5] that “we are on the edge of a golden era.” Eric Brynjolfsson, one of the world’s leading digital economists, says [6] that “the IT revolution is poised to deliver transformative growth.” Brynjolfsson and McAfee have written [7] that “digital technologies are doing for human brainpower what the steam engine and related technologies did for human muscle power during the Industrial revolution … the world is in the throes of unprecedented technological change that will have a profound impact on the potential of the world economy.”
Technology futurists such as Patrick Dixon argue [8] that “we are only in the first hour of the first day of the Internet.” The bottom line is that even the most advanced high-tech economy, the US, is estimated to have only captured a small part of its digital potential so far, with one estimate a mere 18 percent. [9]
Economic pessimism
However, others are less optimistic. The eminent Northwestern University economist, Robert Gordon, argues [10] that the period 1870 to 1970 was a special century, never to be repeated, with the growth of electricity, gas, water, the telephone, mass production, the automobile and air transport. Gordon argues that if US productivity growth since 1970 had been as rapid as in the preceding 50 years back to 1920, then real GDP per person in the US would now be double its current value. In other words, on average, people would have twice the income they currently have. These are not dry and dusty numbers. They illustrate just how significant acceleration or deceleration in productivity growth can be.
Gordon argues that we aren’t experiencing a technological revolution at present and that any stimulus to productivity growth from the digital economy in the 21st century has been and gone, with the acceleration in US productivity growth experienced between 1995 and 2004 – which subsequently fell away. US total factor productivity growth (as measured by total factor productivity – TFP) growth averaged 2.1 percent per annum over the years of the halcyon post-war economic boom from 1947 to 1973. The growth rate of TFP then fell back to 0.5 percent per annum over the 1974 to 1995 period, accelerating to 1.5 percent per annum growth over the 1995 to 2007 period, decelerating to 0.5 percent per annum thereafter over the 2007-17 period. [11] Gordon argues that what we’re now seeing is the resumption of normal service after the aberration of the mid-1990s to mid-2000s. [12]
But these views are far from convincing. The previous decade is a very poor predictor of the next decade for TFP growth. Significant turning points are not signaled in advance, as the pre-eminent British economic historian Nicholas Crafts [13] has pointed out: “Sharp reversals of medium-term TFP growth performances are not identified in advance. Indeed, forecasting on this basis would have missed the productivity slowdown of the 1970s, the acceleration of the mid-1990s and the slowdown of recent years – in other words all the major episodes during the period!” Crafts concludes that: “Econometric estimates of trend TFP growth extracted from recent past performance should not be given much weight in projections of average TFP growth rates over the next 10 years.” This shouldn’t surprise us as GDP measures the impact of past technologies.
Reasons for the lack of inflection before the 2020s
I argue that there are three very practical explanations for the lack of upward inflection in productivity growth over the course of the technological revolution so far, which are easily understood:
(1) Economic history teaches that there are long lags on new technologies’ economic impact due to the time from invention to mass market penetration of the technologies and the time from mass-market penetration to new business models, products, and processes. One of the most compelling arguments why we shouldn’t have expected inflection in productivity before now is the nature of the technological change to date. According to McKinsey, [14] productivity booms tend to lag technology booms by around eight years. The reason is that it takes time for diffusion to creep across the landscape slowly. The full effects take time because of the need to build the stock of technology and discover and develop complementary waves of innovation that reconfigure both internal products and processes and supply and distribution chains.
If we look back to economic history, time lag influences were even greater. The first power station producing electricity was built in the US in 1881. By 1890 only 1 percent of all power used by companies was generated by electricity (steam or water were the main competitors). By 1900, the proportion had reached just 5 percent, rising to 25 percent in 1910, 53 percent in 1920, exploding to 76 percent in 1930. In its wake, electricity brought a fundamental redesign of factories to maximize productivity using technology (most notably in Henry Ford’s case in 1913), but this was more than 30 years after the invention of electricity generation.
The development of the motor vehicle was a general-purpose technology effect in its own right. Karl Benz produced the first ‘production’ automobile in 1885, but car registrations in the US were still at zero in 1900, before rising rapidly to 8 million in 1920 and 23 million in 1930. By 1927, Ford had produced 15 million Model T cars, but this was four decades after their invention. Moreover, early production of general-purpose technologies such as the car and electricity were characterized by an era of falling overall productivity growth (in the 1880s and 1890s).
Thus far, the digital economy has been largely confined to the digital sector. There is a wide divide between the frontier and non-frontier companies e.g., the very big productivity differential in favor of the FANGs, the four prominent American technology companies: Facebook, Amazon, Netflix, and Google.[15] From now on, we should expect a period of catch-up as the enormous mass of non-digital companies close the gap. Every business becomes digital in the wake of the technological revolution and the accelerator effect of Covid-19 on the speed with which work, shopping, and learning go online.
(2) Generational factors suggest the relative size of different generations more or less familiar with the new technologies needs to adjust before the full impact of new technologies is absorbed across the whole economy. As the less technologically savvy generations retire, those truly technologically literate assume control and introduce new business models.
Intergenerational differences are crucial to the nature of technological adoption. The Silent Generation (born 1920-1945) and Baby Boomers (born 1946-1965) were very weak or weak with regard to their technological dexterity. Generation X (born 1966-1979) has good technical knowledge. But it is Generation Y (born 1980-1995) and Generation Z (born since 1996) who have matured surrounded by technology and the Internet.
From a business perspective, over the past two decades, the Baby Boomers were still in charge and therefore ‘didn’t necessarily get it’ with regard to digital business models. Up to now, there has been a tech gap between C-suite and staff further down the hierarchy. All too often, business leaders in the 2010s were still thinking analog, not digital. They didn’t possess the ‘always on’ mentality of generation X or Y or tech-savvy Millennials living through their mobiles. These are the tech-savvy Millennial staff of the 2020s with extraordinary levels of motor response and tech familiarity, and their leisure activities and work skills have merged.
A report [16] from Samsung describes the differentiating characteristics of the ascending born disruptive generation of workers who grew up surrounded by technology and barely remember the time before the Internet. These next-generation achievers expect the best and the latest of everything and are likely to have a transformative impact on business models and strategy as they move higher up organizations.
According to an INSEAD survey, [17] only 40 percent of millennials think their organization’s digital capabilities are up to the mark. So, what causes this disconnect? Samsung states that “It’s a mixture of remoteness, complacency, lack of understanding and sometimes even arrogance on behalf of senior management.” In other words, the most senior staff are making key decisions not because they are completely up to date on technology but simply because they are in a position of authority. To compound the difficulties, next-generation achievers are far less likely to be impressed by hierarchy within organizations. The end result is that tech-savvy individuals avoid the most hierarchical organizations making it even less likely that these companies will pursue a frontier technology strategy. However, while this tendency may have predominated in the past, this is no guide to the future as the business generations shift.
(3) Cyberslacking and the attention deficit – evidence suggests significant time loss and wasting over the past decade as people have become more familiar with and addicted to new technologies. This adjustment period has seen weak productivity growth during what might be described as a transition period. Metcalfe’s Law says that a network’s usefulness increases with the square of the number of users, but it may have a dark side, with people overwhelmed with information. [18] In addition to being overwhelmed, according to one commentator, [19] we are living through a “cultural crisis of attention” and are distracted nearly 50 percent of our time, which is of great concern given that our life essentially amounts to what we pay attention to. Survey data [20] reveals that half of all Americans say they could not live without their smartphones.
The smartphone provides a level of distraction not seen before. In one device, it combines access to the full spectrum of mass media: telephone, texting, music, video, the Internet, video games, and voice-activated artificial intelligence. That is quite a distraction. A widely quoted study [21] found that we checked our mobile phones around 150 times every day. Another study [22] suggested people spend an average of 3 hours 15 minutes on their phones each day. Research by Deloitte [23] indicates that globally consumers check their phones over 80 billion times a day.
Obviously, some of this time will be for work purposes but even allowing for that, there is a distraction, which may explain why the explosion in smartphones over the past decade has run in parallel with weakening productivity growth. A Bank of England blog [24] cites two possible transmission mechanisms from distraction to lower productivity growth:
The crucial question is whether or not cyberslacking and attention problems are likely to recede in the future. The contention of this author is that advances in AI and 5G will begin to reverse cyberslacking and the attention deficit as they make us more personally and professionally efficient through our phones. Technology will start to transform time use for individuals through personal digital assistants and organizations through the Internet of Things, e.g., predictive maintenance. Therefore, we can see that there are persuasive explanations as to why we shouldn’t have expected an upward inflection in productivity growth before now. But what about from now on, beyond issues of lags, generational change, and attention deficits?
From now on
Brynjolfsson, Rock & Syverson [25] argue that, “… there really is good reason to be optimistic about the future productivity growth potential of new technologies, whilst at the same time recognizing that recent productivity growth has been low. The core of this story is that it takes a considerable time, – often more than is commonly appreciated – to be able to sufficiently harness new technologies … the more profound and far-reaching the potential restructuring, the longer the time lag between the initial invention of the technology and its full impact on the economy.”
Richard Baldwin, Professor of International Economics at the Graduate Institute Geneva [26] points out that it is only very recently, since 2016-17, that computers have become “as good or better than humans in some instinctual, unconscious mental tasks – things like recognizing speech, translating languages and identifying diseases from X-rays … Machine learning is giving computers – and the robots they run – new skills that are valuable in offices. Now they can mimic human thinking in tasks involving perception, mobility, and pattern recognition.”
Baldwin attributes this recent change to a breakthrough in machine-learning and argues that the result will be that automation will significantly impact office jobs, not just factory jobs. Moreover, the same new technologies and the legacy of Covid-19 could be outsourcing roles overseas through online work facilitated by genuinely digital business models.
But while these technological breakthroughs have only occurred very recently, we should not expect the lengthy delays seen in economic history – in terms of their economic impact. Baldwin states that: “Globalization after the Great Transformation started one century after automation started. Globalization during the Services Transformation started two decades after automation. In today’s Globotics Transformation, globalization and automation are taking off at the same time, and they are both advancing at an explosive pace.” [27]
Baldwin illustrates the scale of the recent improvement in computing with an illustration using the Apple iPhone, “… computer processing speeds … are doubling every 18 months or so. The iPhone 6S, which came out in 2015, processes information about 120 million times faster than the mainframe computer that guided Apollo 11 to the moon in 1969. That is amazing. But it gets more amazing. The iPhone X, which came out in 2017, is about three times faster than the iPhone 6S … Think about that. The increment in power in the two years after 2015 was twice as large as all the progress between 1969 and 2015. Twice as much progress in two years as there was in the 46 previous years.”
The recent acceleration in computing power is fundamental to future economic growth, because as Baldwin points out, training AI systems to recognize photos or understand spoken language requires astonishing amounts of computer horsepower: “For an algorithm that is looking at say hundreds of thousands of pixels, a single inversion involves millions of billions of calculations. That, in turn, is only feasible with processing power that used to be unthinkable for anything but the fastest supercomputers [the computational complexity of inverting an n-by-n matrix is of the order of n cubed]. Moore’s Law removed that limitation. Computer speeds that were out of reach in 2014 became run of the mill in 2016. The other reason this is happening now is that it is possible to collect, store, and transmit big data sets. Fast computing and big data are linked for a very simple reason. If computer capacity is machine-learning’s jet engine, data is the jet fuel. While Moore’s Law cranked up the engine power, Gilder’s Law kept the fuel pumping. The size of the data-sets being used is something that was thinkable but not doable just a few years ago.”
We are familiar with the sheer scale of the technological revolution underway, but we often forget how recent some of the biggest influences have been. The smartphone was first produced a little over a decade ago, with the introduction of the 3G iPhone. 4G mobile wireless penetration in Europe still only reached a quarter of the population in 2017-18. 4G penetration in North America was higher, but still only reached half the population in 2017-18. Mobile’s share of web traffic only exceeded 50 percent in 2017. As recently as 2014, it only accounted for a quarter of web traffic. Average global interconnection speeds have risen rapidly in recent years but within the past five years were less than 20 megabits per second (mbps).
Perhaps the greatest transformation in the new technological revolution is likely to be artificial intelligence (AI). Google already describes itself as “AI first.” The McKinsey Global Institute, commenting on the potential, has stated that AI is transforming society, and this is “happening 10 times faster and 300 times the scale, or roughly 3000 times the impact” of the Industrial Revolution. A recent edition of the Harvard Business Review [28] reported a survey of 250 executives familiar with the use of cognitive technology. It showed that three-quarters of them believed AI would substantially transform their company within three years. AI technology is only ‘developing’ at present because “most companies still don’t have enough people who know how to use cloud AI … once the cloud puts the technology within reach of almost everyone, the real AI revolution can begin.”
Over-optimism versus general-purpose technology effects
Amara’s Law teaches that we tend to overestimate the impact of technology in the short-term, during the hype phase of development, but substantially underestimate its effect in the long-term as new ideas, products, processes, and business models proliferate, which were not foreseeable at the outset. There is a wide spectrum of views regarding the impact of technology on the economy, from pessimists who assert that the digital economy boost to productivity growth has been and gone – in the decade before the financial crisis – to extreme technology optimists, such as Ray Kurzweil, who foresee a technological “singularity” with potential exponential economic growth. The singularity occurs when – sometime in the first half of the 21st century – machine intelligence finally exceeds human intelligence, with explosive consequences for economic growth and everything else as well. Another technology futurist, Peter Diamandis, foresees an economic singularity such that “within a generation, we will be able to provide goods and services, once reserved for the few, to any and all who need them, or desire them. Abundance for all is actually within reach.”
The argument of this paper is not that we are about to enter an age of abundance in this Diamandis utopian sense, but that we are set to experience a powerful General-Purpose Technology (GPT) effect, with a ‘digital ripple’ which will ultimately spread out across all sectors of the economy, much as was the case with electricity at the dawn of the 20th century. It’s important to note that there doesn’t need to be a singularity for a significant acceleration in productivity growth to occur. The historic parallels also extend to the legacy of the Spanish Flu in 1918-19, followed by ‘The Roaring Twenties’ (in the US) as the GPT impacts of electricity and the automobile transformed society.
Concluding thoughts
This paper argues that the trinity of lagged effects, generational change, and cyberslacking have prevented an acceleration in productivity growth until now but will not continue in the mid-2020s onwards. All three productivity dampener factors are likely to recede as new business models and every company becoming a digital company take hold. Technology is also expected to reduce cyberslacking as AI-driven personal digital assistant applications begin to take over many of the tasks we currently undertake ourselves – this is likely to boost our economically-unrecorded ‘personal’ productivity at home and thereby facilitate an improvement in our recorded ‘professional’ productivity. Technology is likely to facilitate a transformation in time use.
However, the case for an acceleration in productivity growth does not end here. Additional factors are likely to kick-in as the trinity effect recedes:
We stand at the cusp of a new era, but we risk missing the Great Inflection if we persist in looking at the road ahead through the rear-view mirror.
References:
[1] The author is writing a forthcoming book: The Great Inflection, on the potential upward turning point in productivity growth across the world economy in the 2020s.
[2] https://www.mckinsey.com/business-functions/mckinsey-digital/our-insights/is-the-solow-paradox-back
[3] As measured by total factor productivity growth (TFP).
[4] https://www.businessinsider.com/ap-amazons-jeff-bezos-golden-era-of-technology-is-coming-2016-6?r=US&IR=T
[5] The New Digital Age, Transforming Nations, Businesses and Our Lives, E. Schmidt & J. Cohen, Penguin Random House.
[6] https://www.nber.org/system/files/working_papers/w24001/w24001.pdf
[7] The Second Machine Age, Work, Progress and Prosperity in a Time of Brilliant Technologies, E. Brynjolfsson & A. McAfee.
[8] The future of almost everything, P. Dixon, 2015.
[9]https://www.mckinsey.com/~/media/mckinsey/business%20functions/mckinsey%20digital/our%20insights/twenty-five%20years%20of%20digitization%20ten%20insights%20into%20how%20to%20play%20it%20right/mgi-briefing-note-twenty-five-years-of-digitization-may-2019.ashx
[10] The Rise and Fall of American Growth, Robert J. Gordon, 2016.
[11] Need to update to 2018 figures. The acceleration and subsequent decline in productivity growth can also be seen in labor productivity. US output per hour worked rose 1.8 percent per annum over the 1990-1995 period, accelerating to just over 2 percent per annum over the 1995-2007 period. Since the economic recovery began, the growth rate of US output per hour worked has halved to just 1 percent.
[12] John Fernald, at INSEAD also argues that what is happening at present is simply a case of getting back to normal after the exceptional period of productivity growth over the 1995-2004 period.
[13] https://voxeu.org/article/slow-productivity-growth-may-not-be-new-normal-us
[14]https://www.mckinsey.com/~/media/mckinsey/featured%20insights/meeting%20societys%20expectations/solving%20the%20productivity%20puzzle/mg-solving-the-productivity-puzzle--report-february-2018.pdf
[15] FANGs are Facebook, Amazon, Netflix and Google.
[16] https://www.samsung.com/us/business/short-form/next-mobile-economy-workforce-enablement/
[17] https://flexsystems.co.uk/blog/mind-the-tech-gap
[18] Research by Bain & Company analyzed the approximate number of communications per executive per year. In the 1970s it was 1,000. In the 1980s, with the rise of voice mail, it rose to 4,000. In the 1990s with the emergence of e-mail it rose to 9,000. In the 2000s it surged to 25,000 and in the 2010s it has risen well above 30,000.
[19] https://www.theatlantic.com/magazine/archive/2017/09/has-the-smartphone-destroyed-a-generation/534198/
[20]https://www.huffingtonpost.co.uk/entry/pew-study-cell-phones_n_6985004 ri18n=true&guccounter=1&guce_referrer=aHR0cHM6Ly9jb25zZW50LnlhaG9vLmNvbS92Mi9jb2xsZWN0Q29uc2VudD9zZXNzaW9uSWQ9M19jYy1zZXNzaW9uXzdiOGIzODllLWY0MDctNGNmNS05ZmRmLWNiMmY2YjFkMGJmNg&guce_referrer_sig=AQAAAKC8Ssr2OLDbAVrurunA6tYaPpwll-GHAPOEWiOeE05uHQseyzR1kBz7XVlq1mh-B1_tn1_OIYXdS-bGZFonP7dw3v2VCtNzflyUpCTmTRdwXTN58JLewKIwX17QnPPAkkfdvm9NHH1Ju0hKI_5n6oHJeTWafR0gdCbzKiEWwqEi
[21] https://www.lifedatacorp.com/smartphone-prevalence/
[22] https://www.theguardian.com/lifeandstyle/2019/aug/21/cellphone-screen-time-average-habits
[23] https://www2.deloitte.com/qa/en/pages/about-deloitte/articles/Mobile-consumers-check-their-phones-over-80-billion-times-a-day.html
[24] https://bankunderground.co.uk/2017/11/24/is-the-economy-suffering-from-the-crisis-of-attention/
[25] Artificial Intelligence and the Modern Productivity Paradox: A Clash of Expectations and Statistics, NBER Dec. 2017 Brynjolfsson, Rock & Syverson.
[26] The Globotics Upheaval, Richard Baldwin, 2019.
[27] Baldwin (2019) refers to the original Industrial Revolution shift from an agrarian to an industrial society as The Great Transformation, and the shift from an industrial to services economy (roughly 1970 onwards) as The Services Transformation.
[28] Harvard Business Review, January-February 2018.
©2024 Trends Research & Advisory, All Rights Reserved.
Reviews (0)