Automation and the working class

Submitted by martin on 14 January, 2020 - 7:13 Author: Martin Thomas
Difference engine

According to one account in 2013, 47% of jobs in the USA risk being automated away within “a decade or two”.[1] That prospect has been interpreted as utopia or as dystopia. The near future will be one of networked individuals freed from drudgery by automation, and able easily to get what they want to consume and to undermine all hierarchies. Or: only a techno-elite will retain employment and wages. The rest of us will be reduced to a new pauper class vegetating on “universal basic income” handouts.

Further research has queried the projections. Many tasks can be automated, but jobs involve more than those tasks. Capitalists may hope and even believe that they have reduced labour processes to routines laid down by the boss, but “the supposed simplicity and routine character of assembly work is not all that simple... assembly work is also packed with different aspects of non-routine tasks and the capabilities to cope with them”. Human labour is more likely to continue, complementing automated procedures, than be completely replaced by them.[2] Even when capitalists could in theory automate jobs, often (in other than a few high-volume contexts) they will find it too risky to make the huge investments required, or too expensive to do that at any foreseeable wage level. Especially so in current and foreseeable conditions of slow and erratic growth, and wage depression.[3]

Some sectors will see big job cuts, but, overall, outside of slumps, we’re likely to see more jobs working with new technologies, and more jobs overall. The statistics since 2013 show not the zooming-away that the techno-enthusiast writers suggest, but exceptionally slow growth in US labour productivity. The UK is even worse. For a whole era now across the richer capitalist countries, industrial investment has been low, the proportion of surplus-value siphoned to consumption high, and growth of productivity mediocre.[4]

More jobs have poor and insecure hours, wages, and conditions. Overall job numbers are not dropping. As of 2018 the USA had its lowest unemployment rate since 1969. Fast food bosses were complaining of labour shortages.[5] The UK’s unemployment rate is the lowest since the mid-70s. The UK’s labour-force participation rate is at record highs. The USA’s is higher than it was before the late 1970s. It is down from its late-1990s peak, not so much, it seems, because of anything to do with automation as because of the USA’s epidemic of prescription opioid use, and consequent ill-health.[6] After a long history, back to the 19th century, during which the average working week was slowly shortened, it has now been stuck around the same length in the USA since 1980,[7] and has been increasing in the UK since around 2009. Some economists have argued the contrary of the techno-enthusiast thesis: that for the next while at least the USA, and probably other richer capitalist countries, are set for more stagnant economics than at any time since the late 19th century.[8]

The techno-enthusiasts have counter-arguments[9]. The growth of microelectronic capacities is exponential, in line with Moore’s Law: computer processing powers double every two years. Big technological innovations generally feed through to increased productivity only with delay, through the detailed engineering follow-up. For example, the productivity advantages of shifting from steam engines to electric motors mostly came from the 1920s (in the USA). That was 40 years after the first electricity-generating power station and 30 years after electric motors went from experimental to commercial. That the new computer technologies have not yet boosted productivity much is just the usual delay. And now information technology and artificial intelligence are on the verge of qualitative new breakthroughs.

Information technology is expanding from tasks where we give the machine exact instructions in advance to tasks which the machine learns to improve on. The scope for machine learning increases with “big data”. Data volumes double roughly every two years.[10] As the variety of new technologies grows, also, the possible combinations of them grow even faster.

But it is over 70 years now since the first computers. Almost as long since commercial transistors. We have had 60 years of “artificial intelligence” and “machine learning”, with waves of optimism about their potential (as today) and waves of pessimism. Twenty-odd years since people started talking about “big data”. Nearly sixty years since the first industrial robots, and forty years since they started “taking off”. Over 40 years since the internet was developed, ditto mobile phones. 40 years since the first widely-used PC. Over 25 since the internet “took off”. 32 years since the economist Robert Solow commented “you can see the computer age everywhere but in the productivity statistics”.[11]

Capitalists have deployed new information technologies very widely. If manual typewriters are still used in India and Brazil, it is not for lack of internet: it is because the electric power grid is unreliable.[12] As capitalists have spread the technologies, they have also drastically changed areas of the workforce. We do not have to speculate about that. It has been happening for half a century or more, and often faster than now. There are decades of experience with the transformation of industries by information technologies. New technologies, by definition, are new, and will introduce new patterns. Yet we can and should learn from past “shocks of the new”.

The car industry uses about half of all industrial robots. But the story there is not just of workers being replaced by robots. The collapse of Detroit’s economy, in the USA, has been just as much about car production being moved to the anti-union southern states of the USA. There are not many fewer car workers in the USA, overall, than there were in 1979. Many old jobs have been destroyed, new ones generated.[13] Ports are (slowly) being automated, and that is changing port jobs and displacing labour. The displacement of labour is less drastic than from containerisation, which in UK ports reduced the dock labour force by about 90%. Traffic has increased. The total number of “logistics” jobs (using new information technologies for stock control etc.) has increased fast, and the number of “logistics” jobs at and around ports (for example, in areas like the warehouse hub in the “Inland Empire” of southern California) has probably not fallen at all. Old jobs have gone, new jobs have been generated. Sometimes, as in Britain, union strength has been shattered and wages have been forced down on the waterside; pretty much everywhere, the port unions have failed at organising the new jobs a bit inland. The story is more complicated than just new technologies cutting overall job counts, and certainly more complicated than just information technologies cutting overall job counts.[14]

With the steep fall in transport costs since containerisation, the clothing industry has largely disappeared from relatively high-wage economies and grown in Bangladesh, Vietnam, Cambodia, India, Sri Lanka. That is not a story of information technology wiping out jobs. The sewing-together of garments remains little-automated, relying on old technology (the sewing machine). Machines find it difficult to manipulate materials which are soft and flexible enough for clothing. Jonathan Zornow claims he has a new technology which will stiffen the material temporarily for it to be shaped and sewn automatically, and then return it to being soft and flexible. Maybe that will automate the industry. Maybe not.[15]

The ratio of “white-collar” to “blue-collar” jobs has risen, but the biggest eliminations of whole categories of worker by computer technologies have been in “white-collar” jobs. There were 0.4 million telephone operators in the USA as late as 1970. There are almost none now. “Typing pools” and the armies of clerical workers who did calculations by adding machines have disappeared. Some jobs introduced in the early microelectronics era (computer operator, data-entry keyer) have been wiped out by its later years.[16]

The technologies are still being developed, and will continue to change workforces. But “this changes everything, and makes the previous experience of capitalism inapplicable”? That remains to be shown, as it did in the 1950s, the early 1960s, or the mid 90s, all of which saw similar automation-hype.[17] South Korea and Singapore have much bigger densities of industrial robots than other countries. Yet the job categories found in other developed capitalist countries have not disappeared there. Unemployment in South Korea has varied between 3% and 4% since the start of the century, with spurts above 4% from time to time attributable to the usual capitalist mechanisms (slumps) rather than to technology. Labour force participation there continues to rise.[18]

The US Bureau of Labor Statistics produces projections of which jobs will expand between now and 2026, and which will contract. Some of the jobs which techno-hype writers name as about to be trashed by new technologies are listed by the BLS as expanding fast. Lawyers? Accountants? About to be displaced, they say, by better computerised searches on legal records, by computerised “smart contracts”, and by computerised accounts and analysis of accounts. Yet “lawyer” is named by the BLS as the job in which there will be most new openings for those with more than a first university degree. The BLS expects “paralegal” jobs, with lower qualifications, to grow faster again. “Accountant” is the fourth-highest new-openings job for those with a first degree. Techno-hype writers stress how technology will transform the health-care industry. The highest-ranked new-openings job for those with a first degree is nursing. The BLS predicts “health-care assistant” jobs, with fewer qualifications, will grow even faster than nursing. The job sector highest-ranked for new openings not needing a degree is “personal care services”. New technologies are developing in that sector, but there is no indication of of those technolodisplacing jobs entirely, rather than mostly creating new jobs complementary to new technologies. The area where most jobs will be lost, says the BLS, is the one which is already losing most — backroom office work. The BLS is extrapolating from what is happening now. Always guesswork; but that the past is no guide to the future remains to be proved. Already junior lawyers in big law firms use sophisticated computerised searches, and use the time saved from searching on drafting legal advice and contracts. Result: more legal business — and more lawyers despite the increased productivity.[19] And accountants? More financial data collated, more analyses of it, so more accountants despite more of the job being done by computer.

Retail work has been transformed by technology over recent decades more than most. Supermarkets. Malls. Barcodes (from the 1980s). Electronic tills and payments. Automated stock control and ordering. Self-check-outs. On-line shopping. The BLS expects retail jobs to grow much slower than the average to 2026, but even those not to decrease overall.

Karl Marx, in Capital volume 1 and in passages of Theories of Surplus Value, developed ideas on how capitalism changes labour.[20] Comparing those ideas with the facts of recent decades sheds light on the pattern behind those facts, and also shows that some aspects which Marx considered marginal have become important.

Capital, wrote Marx, started with the “formal subsumption of labour”, bringing already-formed labour processes under the sway of capital. Workers are brought together in a workshop (simple cooperation); and then the labour is divided into specialties. In that phase (so Marx quotes a capitalist ideologue) “the more skilful the workman, the more self-willed and intractable he is apt to become, and... the less fit a component of a mechanical system”. Factories based on systems of machinery, says Marx, allow the “real subsumption of labour”. The labour process is defined by the system of machinery, not by the various inherited skills of individual workers. “In the form of machinery, the implements of labour become automatic, things moving and working independent of the workman... The automaton, as capital, and because it is capital, is endowed, in the person of the capitalist, with intelligence and will; it is therefore animated by the longing to reduce to a minimum the resistance offered by that repellent yet elastic natural barrier, man”.

Meanwhile, “the means of communication and transport became gradually adapted to the mode of production of mechanical industry, by the creation of a system of river steamers, railways, ocean steamers, and telegraphs”.

In this new way of organising labour the capitalists push for constant improvement of the machinery. This cheapens production by replacing refractory specialised workers by machine-operators, more easily replaced, more easily trained, more apt for flexible redeployment, and to some degree organised and paced by “the automaton” itself, and by replacing some categories of workers altogether by machines. In one passage Marx suggests the capitalist criterion is simply cheapening. In fact, as Marx’s broader argument indicates, it is not. Crane-driving in container terminals, for example, is a specialised job. It takes time and aptitude to develop the skill. A more skilled crane-driver on a good day will move more containers than a less skilled one on a poor day. More than an automatic crane, too, and at lower cost.

Capitalists introduce automated cranes (where they do: the process is slow) because even if they are slower and more expensive, they work at the same predictable rate each hour, they don’t require breaks and shift changes, etc. In a high-volume tightly interlinked system between ship, wharf, and truck or train, control can trump cost. However (and this Marx noted), if wages are sufficiently pushed down, then new machines will not be introduced even though they allow production with less labour. “Hence the invention nowadays of machines in England that are employed only in North America” (where wages were higher).

Marx noted that the new factories, primarily textile mills, employed women and children. Work with a system of machinery required less muscle-power. Women and children would be “more pliant and docile”. In time capitalists found that factory women could be as stubborn and rebellious as any men. For work with systems of machinery, though, capitalists have generally preferred young workers. Even if not as strong, they are more agile, alert, flexible, and energetic.[21]

Marx saw this mode of production as levelling labour, from the hierarchy of specialties in the old workshop to a division only between the machine-operatives and the youngest workers who fetched and carried, cleaned up, etc. He noted another group of workers “some of them scientifically educated, others brought up to a trade” doing maintenance and repair. In his era that was “numerically unimportant” compared to the operatives. Marx mentioned “salesmen, messengers, warehousemen, packers, etc.”, suggesting that their numbers too were secondary. Elsewhere he gave more emphasis to the diversity of labour: “the product, is formed, one working... as a manager, engineer, or technician, etc., another as an overlooker, the third directly as a manual worker, or even a mere assistant... more and more of the functions of labour apacity are included under the direct concept of productive labour”.[22]

Also secondary in number were workers in “entirely new branches of production, creating new fields of labour”, as distinct from mechanised or automated versions of older productions like clothing. Marx cited “gas works, telegraphs, photography, steam navigation, and railways”. The workers there (in 1861, in England and Wales) totalled 94,000, as against 643,000 in the textile factories, 566,000 in coal and metal mines, and 397,000 in metal-working industries (mostly still workshops).

In some passages Marx seems to suggest that factory operatives would come to dominate numerically. He noted and even stressed, however, that they were far from doing so. The system-of-machinery workers in textile factories were only around 10% of the working class, and only slightly better than half as numerous as domestic servants. “The extraordinary productiveness of modern industry... allows of the unproductive employment of a larger and larger part of the working class”. “New ramifications of more or less unproductive branches of labour are continually being formed”. “Unproductive” here meant “unproductive” in capitalist terms, i.e. not directly producing surplus-value. (Marx’s illustration is that a teacher in a profit-making private school is “productive”, but one in a state school is not).

Capital would continually invest in new systems of machinery, to generate “relative surplus-value”, but worker-numbers would rise much less than proportionately to production as machines replaced workers. Capital could replace specialised labour by capital-dictated and parcellised routines of machine operation; it could replace some “automatised” labour by machines. Marx denounced the bourgeois economists who argued that capitalist displacement of labour by machinery was harmless, because in the growing economy the workers were sure to find employment elsewhere. “The labourers... thrown out of work in any branch of industry can no doubt seek for employment in some other branch... Even should they find employment, what a poor look-out is theirs! Crippled as they are by division of labour, these poor devils... cannot find admission into any industries, except a few of inferior kind, that are over-supplied with underpaid workmen...” Moreover, if “accumulation [of capital] increases the demand for labour, it increases on the other the supply of labourers by the ‘setting free’ of them, whilst at the same time the pressure of the unemployed compels those that are employed to furnish more labour”.

Thus the constant revolutionisation of machinery would constantly regenerate an “industrial reserve army”. “It is true that in the long run the labour that has been released together with the portion of revenue or capital that has been released, will find an opening in a new sphere of production or in the expansion of the old one. [But in the meantime] there is nothing to prevent a part of the money capital lying idle and without employment... while at the same time workers who have been displaced by machinery are starving”. That process would allow for the continuation alongside advanced industry of “industries of inferior kind”, smaller-scale, less mechanised, relying on low wages. For his time Marx mentioned particularly home-working and, as we’ve seen, domestic servants. Marx denounced conditions in the factories. But, he argued, “all the horrors of the factory system, without participating in any of the elements of social progress it contains” would be generated in the less-integrated, lesssystematised labour which continued alongside.

Marx argued that real wages were generally be higher in more developed capitalist economies. Only, the ratio of surplus-value to wages would also be higher. Thus he also tacitly expected real wages to rise. The common contention that Marx predicted an iron law of immiseration of labour is false.[23]

The 150 years since Marx wrote have confirmed his theory of capital driving to “automatise” labour – to strive to shape it and parcellise it and routinise it as ancillary to systems of machinery – and to automate production outright (displace labour).[24] Those trends were sharpened by the rise from the early 20th century of assembly-line production and “Taylorism” (managers working systematically to extract and codify knowledge of methods of production, and on that basis to take control over the training of new workers, which had previously been the passing-on of knowledge from worker to worker).

150 years on from Marx, workers in “entirely new branches of production, creating new fields of labour” are no longer numerically unimportant. The new branches generally create new specialised jobs. Capitalists then strive to automatise those jobs, but with varying success. Also, the category of jobs “some of them scientifically educated” doing maintenance, repair, installation (and design), and the “salesmen, messengers, warehousemen, packers, etc.” have often been little automatised, and have gained greater relative weight.

The production lines of mass-market, more-or-less standardised goods have been steadily automated over the decades, allowing for expanded production together with a not-much-bigger workforce and the elimination or drastic changing of whole categories and jobs. In the USA, “operatives and craft workers” (a wider category than factory production workers) reached their maximum at 34% of the workforce in 1950, and “operatives” alone at 20% the same year. By 2016, 13% of US workers were in “goods-producing” industries (manufacturing, mining, construction). Not all those 13% were production workers. The Bureau of Labor Statistics classifies just 6% of the workforce as production workers — a smaller share than “transport and materials-moving”, and only modestly more than “installation and repair”.

This trend is mostly not a matter of production moving to other countries. Most countries, including much poorer ones than the USA, including China, now have manufacturing production employment declining as a percentage of the whole.[25] The has been a general relative rise in “white-collar” labour, even in poorer countries. A number of distinct trends are behind that. At the time Marx was writing, the scope for detailed accounting and checking was limited by it all having to be done in handwriting and by hand calculations. Before cash registers and adding machines became widespread, for example (in the USA between the late 1880s and World War 1, in Europe by World War 2), most shopkeepers could not keep detailed records of sales. In that sphere, the introduction of machinery has displaced some categories of labour (hand calculation and recording), but, much more, created new jobs (of analysis of the much greater bulk of data which it is now workable to collect.) In 1880 there were only 2,300 accountants in the whole USA; by 1920, 126,000; in 2018, 1.26 million.

When Marx wrote of capitalistically-unproductive workers, he referred to domestic servants and to the much smaller category, then, of public-service workers (such as state-school teachers). These were workers who suffered under capitalist control. Badly so: “the young servant girls in the houses of the London lower middle class are in common parlance called ‘slaveys’” (Marx). But they did not produce profits. He also wrote of a species of competitive, profitgenerating “unproductive” workers, in retail and in finance for example. Their labour did not expand the social total of surplus value. And yet, by their labour, they enabled their bosses to draw profits from that social total.[26] Since Marx’s day there has been a big expansion of different varieties of capitalistically-unproductive labour. As a result of the efforts of the labour movement, and the reluctant agreement of capitalist classes that they will do better with healthier and more educated workers, there has been a growth of state-funded education and health-care, what could be called “welfare unproductive labour”. There is also a growth of spheres where capitalists can make profits which are not new surplus value, but a redirected share of wider surplus value: finance, real-estate, marketing.[27] When competition sharpens, it can be capitalistically rational for each competitor to spend more on marketing and financial manipulation, even though in the aggregate the costs of that marketing and financial manipulation are a deduction from surplus-value produced.[28]

The effects of automation on job structure intertwine with the rise of “services” in household consumption, which is partly a result of the rise in real wages, partly a result of the fact that capitalist marketing is able to expand consumption of “services” faster than it can expand consumption of food (or even of fridges, washing machines, smartphones: one is usually enough, even for the well-off). In the USA in 1869, 44% of household spending was on food and drink. The 24% on “services” was mostly rent (or “imputed rent” for owner-occupiers). By 1940, 22% was on food and 43% on services (of which 13% on housing); by 2013, 8% food, 67% services (of which 16% housing).[29]

Much of the writing on automation assumes that automation hits jobs which are by their nature “routine” and “middle-level”. “High-level” jobs (like writing techno-hype articles and books) are assumed to be not automatable. The historical record suggests the assumption is false. The processes of “automatising” jobs and automating them away have often started with specialised jobs in which training, experience, and skill make a big difference, and where workers may have relatively high wages. Automata can do jobs often assumed to be beyond their reach: journalism, personal counselling, many tasks of caring for frail elderly people. An automated artist has had work exhibited in galleries. Automata can do calculations more complex than any human could do. They can check mathematical proofs; in fact, proofs have been produced which no human can check without a computer.[30]

Service industries, and “high-level” jobs in service industries, can often be automated. Yet automation in those industries often leads to an expansion of “service” production and of new jobs working with the machines. A skilled typist had many skills other than basic dexterity (literacy, comprehension of what was being typed, knowledge of how to set out different sorts of documents), and would be much faster and more accurate than a less-skilled one. But the abolition of typing pools has not emptied offices. The introduction of cash registers, adding machines, calculators, mainframe computers, and then networks of desktop computers has led to a great expansion of employment in accounting, stock control, logistics, etc.

Among jobs required advanced higher education, the BLS predicts the biggest percentage increase in numbers for statisticians. Statisticians are automating their computations more and more[31]. But that means many more statistical analyses can be done, on data which is collected more and more cheaply.[32]

The new jobs generated as complementary to automation are often “real” jobs, producing real benefits. Others are generated as a result of the expansion of “competitive-unproductive” labour, in finance for example. And yet others are generated and sustained because capital’s drive for control over labour is fallible. Sometimes it produces only costly simulations and pretences of control, or “displacement activities” which information technologies also facilitate: boxticking, patching-up, writing of reports or collating of statistics which no-one reads.

It generates “bullshit jobs” (and “bullshit” work within jobs which are not fully “bullshit”).[33] After writing an article on the theme, David Graeber received 110,000 words of testimony about “bullshit jobs”[34] and a poll in which 37% of all respondents in the UK described their jobs as “bullshit”. His evidence remains mostly anecdotal, but he digs up figures showing an expansion of backroom jobs (such as marketing, PR, etc.) in US universities much faster than the expansion of student or teacher numbers. Graeber considers the “financial industry... a paradigm for bullshit job creation”. It uses high technology. Most of the transactions in many financial markets now are automated trading. But that means more transactions and more jobs.

Among Graeber’s categories of “bullshit jobs” some are “competitive-unproductive” — “lobbyists, PR specialists, telemarketers, and corporate lawyers” — often jobs which use advanced communications technologies and could not emerge on a comparable scale without those technologies. As Marx put it in Capital, pioneer capitalists often had an austere dedication to accumulation. Then “when a certain stage of development has been reached, a conventional degree of prodigality, which is also an exhibition of wealth, and consequently a source of credit, becomes a business necessity... Luxury enters into capital’s expenses of representation”. It is expressed not only in shiny offices, but in the creation of managerial retinues, often with bombastic job titles.[35] Those are at the top end of what Graeber calls “flunky jobs”. At another end, domestic-servant jobs, which were 8% of all jobs in the USA in 1870, and declined to become statistically insignificant by about 1990, are increasing again. An unofficial estimate is 1.4% of the workforce in the USA.

In the expansion of “box-ticking” labour, there is a factor additional to what Graeber calls “managerial feudalism”. Neoliberalism, despite its own claims, has not so much “deregulated” as created a new web of regulations intended to be market-facilitating or market-simulating.[36] They are often more complex, in ways which would not even be thought of without post-1980s information technology and its capacities for collating large amounts of data. A similar thing happens with out-sourcing: it means more contracts, more contractual provisions to be checked. With neoliberalism, then, information technology is facilitating an expansion of reports, submissions, and “compliance” and “box-ticking” labour. The BLS reports that the number of “compliance officer” jobs in the USA almost doubled between 1997 and 2018 (to 0.3 million).

In “welfare unproductive” sectors, also, new technology may well increase, not reduce, jobs. As late as the 1970s, there had been almost no new technologies in schools since the coming of printed books, other than the blackboard, the exercise book (made workable by paper becoming cheaper), and the pencil (all mid 19th century). The “spirit duplicator”, allowing maybe 30 or 40 copies of a handwritten sheet, had been invented in the 1920s; in the USA multiplechoice tests had been often used, with automatic marking, since the late 1930s; and ballpoint pens had been widely used from the 1960s. That was about it. Now there are computer networks, laptops, tablets, photocopiers, interactive whiteboards... There are sporadic drives to get more systematic managerial control over teaching — to “automatise” it — and from time to time techno-optimists talk of computerised learning replacing education workers. In fact, though, living labour has not been removed. Teachers work more with new technologies. Schools add new job categories such as IT workers. The BLS expects education admin worker numbers to grow faster than teacher numbers and faster than the overall workforce. New technologies also facilitate the accrual in schools of “bullshit work”.[37]

The fever of technological leaps may accentuate big slumps, with resulting high unemployment. The 1920s and 30s in the USA were a time of fast technological advance as well as of the Great Depression. But the unemployment will come through slumps, not as a direct or automatic consequence of the new technology. Capitalist deployments of new technologies have displaced large numbers of workers, leaving them destitute or unable to “find admission into any industries, except a few of inferior kind, that are oversupplied with underpaid workers...” Big new deployments are likely to continue to do that. Capital is restructuring labour, destroying old jobs and creating new ones. That will happen faster and more widely if unions increase our strength and force higher wages, but it will happen anyway.

In the last 40 years or so, trade unions have too often let themselves fall back into managing gentler and easier decline for “legacy” workforces in traditional bastions, and failed to organise new sectors and younger workers. Unions need to change that approach, to reach out wider, and to fight for a shorter working week. In large areas of the global South, “Industrial Revolution #2” has yet to arrive (clean water, electrification, good food supplies), even though “Industrial Revolution #3” may have made big advances in those same areas.

If microelectronics is “Industrial Revolution #3” (after steam engine #1, electrification #2), then India shows that in some ways #3 has already spread wider than #2. In India, mobile phone connection numbers are 84% of household numbers, 66% of households have TV, 24% have smartphones, 13% have PCs. But only 18% have piped water (and that often only some hours a day); although about 88% have some electricity supply, it is often unreliable; only 27% have fridges, and only 5% motor vehicles.[38] Such combinations generally produces the job-cutting effects of “Industrial Revolution #3” in the cities while “Industrial Revolution #2” is transforming agriculture and driving people off the land. The result is much higher unemployment and semi-employment than in countries with more thoroughly advanced technologies.

The climate emergency means that we need not so much more of the same, economically, but rather to reconstruct economic infrastructures, to build new infrastructures where they do not yet exist in the global South, and to expand public provision, while checking some expansions of private consumption. Those tasks will require many new jobs and the best use of new technologies to do the tasks fast enough.

What will not happen is technology, in and of itself, creating either a utopia of “fully automated luxury” for all, or a dystopia of everyone becoming unemployed apart from a few “high-end” workers controlling new technologies. The trends which the BLS extrapolates in fact show more incremental and diffuse job-displacement in coming years rather than more cases of the rapid wholesale abolition of job categories (typists, telephone operators, etc.) typical of the last few decades.

Trends may change. Some sudden new leap in microelectronics technology may emerge and get applied widely and very fast. The probable result will be faster and bigger displacement of some sections of workers, and great enrichment of the capitalists using and producing the new technologies. Despite expanding productivity, capitalism is far from reaching “peak stuff”.[39] It has not moved on from the competitive acquisitiveness described by Marx in his 1844 Manuscripts: “Private property has made us so stupid and one-sided that an object [or we might say, a service] is only ours when we have it” (i.e. as private property). The top 5% in the USA account for 38% of all household consumption.[40] Capital’s capacity to get them to desire more hotel and restaurant services, more luxury watches, bigger cars and houses, more expensive education for their children, etc. is still strong.[41] If a new technological leap gives them a boost, they will spend their riches on more consumer goods and services, on hiring bigger retinues, or on developing new businesses. As a result, businesses will need workers. The dislocations of the new technology will make it easier to get those workers cheap (including for jobs which at higher wage rates would be automated).

A stronger labour movement could instead intervene in the processes of new technology to begin reducing weekly working hours again. The outcome depends on class struggles, and not just on technology.

• Thanks for comments on and criticisms of earlier drafts of this article by Janet Burstall and Bruce Robinson, especially.

Aalbers, M, 2016, Regulated deregulation, in Springer, Birch, MacLeavy (eds), Handbook of neoliberalism

Acemoglu, D, and Restrepo, P, 2017, Robots and jobs: evidence from US labour markets, MIT Dept of Economics Working Paper 17-04

Arntz, M, Gregory, T, Zierahn, U, 2017, Revisiting the risk of automation, Economics Letters, Volume 159, October 2017

Autor, D, 2015, Why are there still so many jobs? Journal of Economic Perspectives 29:3

Autor, D, and Salomons, A, 2018, Is Automation Labour-Share-Displacing? Brookings Papers on Economic Activity

Brynjolfsson, E, and McAfee, A, 2014, The Second Machine Age

Bureau of Labor Statistics, USA:

El-Sahli, Z, Upward, R, Dockworkers and the introduction of containers in UK shipping in the late 1960s,

Ford, M, 2018, The Rise of the Robots, in Paus, E (ed), Confronting Dystopia

Frey, C B, and Osborne, M, 2013, The Future of Employment,

Friedman, B, 2015, Work and consumption in an era of unbalanced technological advance, Journal of Evolutionary Economics 2017

Gordon, R J, 2016, The rise and fall of American growth

Graeber, David, 2018, Bullshit Jobs

Husson, M, 2008, Un pur capitalisme

Lapavitsas, C, 2013, Profiting without producing

Marx, Karl, 1864. The Results of the Direct Production Process

Mohun, S, 2014, Unproductive Labor in the U.S. Economy 1964-2010, Review of Radical Political Economics 46:3

Moody, K, 2018, High Tech, Low Growth: Robots and the Future of Work, Historical Materialism 26:4

Moseley, F, 1988, The Increase of Unproductive Labor in the Postwar US Economy, Review of Radical Political Economics 20:2-3

Panitch, L, and Konings, M, 2009, Myths of neoliberal deregulation, New Left Review 57

Pfeiffer, S, 2016, Robots, Industry 4.0 and Humans, or Why Assembly Work is more than Routine Work, Societies 6:2

Rachel, L, and Summers, L, 2019, On falling neutral real rates, fiscal policy, and the risk of secular stagnation. Brookings Papers on Economic Activity

Rifkin, J, 1995, The End of Work

Rodrik, D, 2015, Premature Deindustrialization, NBER Working Paper 20935

Singleton, C J, 1992, Auto industry jobs in the 1980’s: a decade of transition, BLS Monthly Labor Review 115:2

Spence, M, 1973, Job Market Signaling, Quarterly Journal of Economics 87:3

Thomas, Molly, 2016, Smart contracts 101 — driverless deals, Proctor 36:11

West, D F, 2018, The Future of Work: Robots, AI, and Automation

1 Frey/ Osborne

2 Quote from Pfeiffer. Autor, and Autor/ Salomons, document the "complementary jobs" thesis in some detail; Arntz, the thesis that automating away tasks doesn't mean automating away jobs. Acemoglu/ Restrepo, in another large econometric study, find that a newly-installed robot in the USA displaces between 3 and 5.6 jobs, but new IT capital more broadly tends to augment job counts. Despite their headline claim about "47% of US jobs at risk", Frey/ Osborne (pp.42-44) are explicit that they "make no attempt to estimate how many jobs will actually be automated". They suggest that more and more will be automated because "labour is the scarce factor". Even a beginning of mass "technological unemployment" would make labour not "scarce" at all, and facilitate wage depression or stagnation (as is already happening in the USA since the 1970s, and the UK since 2009), thus slowing automation. For example, all restaurant cook jobs figure in Frey/ Osborne's list as among the most "at risk" of being automated, but what fast-food chains are in fact automating now is order-taking

3 Moody

4 Husson; Moody

5 New York Times 3 May 2018

6 BLS Monthly Labor Review March 2019 Life expectancy in the US has fallen since 2014:

7 For USA: Friedman. For UK: Break times within the working week have also been reduced in the USA: Moody p.11.

8 See Gordon and Rachel/ Summers for different versions

9 See Brynholffson/ McAfee; West; Ford


11 Gordon discusses the rejoinder sometimes made that the productivity boost from computers is underestimated because statistics do not adequately capture improvements in quality arising from IT. He argues that this underestimation bias was actually greater for previous new technologies.

12 52% of households in Brazil have computers, and 49% internet access (World Bank). Power cuts in Brazil:

13 BLS figures, Moody p.9, and Singleton 1992

14 El-Sahli and Upward show the number of dock jobs falling fast from the 1960s to 1991, but then levelling off. See Moody for logistics jobs more generally.


16 BLS

17; 1950s; Rifkin 1990s


19 Even with "smart contracts" (the concept dates to 1994) "the initial negotiation of the terms of a contract would still be reliant on traditional legal services" (Thomas)

20 Mostly Capital vol.1 chapter 15 and Theories of Surplus Value Part 2 ch. 18

21 On young workers in the electronics industry today, Maybe the elimination by automation of many jobs operating and feeding factory assembly lines (though by no means of whole workforces) is a push behind the influx of young workers into industries like fast food, and the higher rates of unemployment today among young workers (when in the 1930s unemployment was lower among young workers than older ones)

22 Marx 1864

23 In the 1960s and 70s, especially, many writers claimed that working-class affluence had proved Marx's whole theory wrong. But if the economist and "technosceptic" Robert J Gordon's shoal of evidence against the idea that "Industrial Revolution #3" is more revolutionising than "Industrial Revolutions #1 and #2" is right, and he presents a shoal of evidence, then most of that access to "affluence" of the working class, in the USA at least, took place not in the era when writers were claiming loudest that affluence had proved Marx wrong, but in the era of Marx, Engels, Lenin, Luxemburg, and Trotsky. If the bourgeois apologists did not have enough evidence to prove Marx, Trotsky, etc. wrong, then all the more they don't have now. The "entirely new branches of production" which Marx had identified machine-producing not the same things as of old more economically, but entirely new things – opened the way for US workers to get their homes, as Gordon puts it, "networked" as they had not been in 1870. They got connected to water, sewage, electricity, gas, phone, mail, radio, movies. They got fridges and washing machines. Thanks to paved roads and cars, they could move outside their small town or neighbourhood. They mostly had high-school education. They benefited from publicly-enforced food and public hygiene standards, which drastically reduced the impact of infectious diseases (as distinct from the still-unconquered chronic diseases). They had weekends and annual holidays. Income-per-head figures underestimate improvement (in productivity and in life standards) more in that era than in others. First, many of the new goods working-class households had improved fast in quality as they fell in price. Second, some of the "networks" had an equalising tendency. Before, for example, the women in poorer households spent hours per day carrying heavy loads of water, often unclean water. Richer households had servants to get their water, and their own safe wells. Now the piped water was the same for everyone. Many improvements have been won since World War 2 in the USA. More, of course, in Europe and later in some countries of Asia and Latin America, as they "caught up" partially or wholly with the USA. But Gordon argues that the "Industrial Revolution #3" improvements ballyhooed by the techno-enthusiasts have been narrower than those won following "Industrial Revolutions #1 and #2", mostly confined to the sphere of entertainment, communications, and information.

24 Automation does not always follow on automatisation, but may result from technological innovation bypassing the job. Fundamentally, the computer operator job was "automated" away by real-time processing replacing batch processing, and computers becoming technically more reliable

25 Rodrik

26 "The unpaid labour of these clerks, while it does not create surplus-value, enables [the boss] to appropriate surplus-value, which, in effect, amounts to the same thing with respect to his capital. It is, therefore, a source of profit for him" Marx, Capital volume 3, chapter 17

27 Finance capital also draws surplus value direct from households: see Lapavitsas

28 Moseley and Mohun have researched the rise of unproductive labour in the USA. Mohun has found that in recent times the proportion of unproductive workers has stagnated, but the proportion of income going to unproductive labour has risen, because of high pay for many finance workers and managers. The BLS, drawing on recent experience, expects job numbers for "market research analysts and marketing specialists" to grow 23.2% between 2016 and 2026

29 Gordon Table 2-2

30 For mathematical proofs, see Keith Devlin For automated counselling, For robots in care for the elderly in Japan, For automated journalism, Andreas Graefe, For automated art,

31 SPSS the Statistical Package for the Social Sciences was launched in 1970 and has been much developed since.

32 For some big caches of data, statistical analysis falls down, and in recent years other mathematical tools (topology) have been more fruitful

33 In the USA the job category "management analyst" (or "management consultant") increased from 127,000 employed in 1997 to 684,000 in 2018. Marx on some other categories of unproductive labour: "With given conditions of production, it is known exactly how many labourers are needed to make a table, how great the quantity of a particular kind of labour must be in order to make a particular product. With many 'immaterial products' this is not the case. The quantity of labour required to achieve a particular result is as conjectural as the result itself. Twenty priests together perhaps bring about the conversion that one fails to make... The number of soldiers required to protect a country, of police to establish order in it, of officials 'to govern it' well, etc. all these things are problematical... although how much spinning labour is needed to spin 1,000 lbs. of twist is known very exactly in England. As for other 'productive' labourers of this kind, the concept of them includes the fact that the utility which they produce depends only on their number, consists in their number itself. For example, lackeys, who should bear witness to their master's wealth and elegance. The greater the number of them, the greater the effect they are supposed to 'produce'." Marx also names doctors as among those where the supposed product has no clear relation to the labour. That is not convincing: after all, some tables need an elusive quantum of carpenter-labour. But what Marx writes about priests, officials, etc. surely applies to "managerial" labour, and what he writes about lackeys corresponds to Graeber's "managerial feudalism".

34 Graeber pp.28ff

35 “Chief future officer”, “chief joy officer”, "chief happiness officer”, “chief thinking officer”, “chief vision officer” (each of whom of course will have deputies, assistants, etc.): Izabelle Kaminska, Companies need fewer mystics and more critical thinkers, Financial Times 19/2/19

36 See Panitch/ Konings and Aalbers

37 Graeber writes about the expansion of bullshit work and bullshit jobs in universities, and shows it has been worse in private US universities than in public ones. Students are now advised to consider university education an "investment", but maybe in fact universities have gone through their own processes of unproductive-labour-inflation and become less adapted to the world of work than they were in days gone by when they were closely tied to training lawyers, doctors, teachers, clerics, and officials. As "credentialism" expands, more and more a university degree is taken by employers as a signal that a worker is generically adapted to jumping through hoops. Any degree. The particular knowledge retained (or not) from university studies is secondary. See on "degree inflation" and Spence on signalling. The universities themselves, as bourgeois institutions, have an interest in this "degree inflation" and in the proliferation of e.g. business degree courses with little intellectual content. "Business" accounts for 19% of all degrees in US universities, 69% more than the runnerup area, all health-related fields of study totalled: See also

38 Financial Express for TV; Unicef (and Hindustan Times for water; World Bank for PCs; for mobiles; (and Times of India for electricity; Pew Research for smartphones; The Hindu for fridges; Indian Census for cars


40 Ford

41 The watch industry was perhaps the first factory industry, one with a developed craft division of specialised skills (Marx, Capital vol.1 ch.14/3). Those specialised skills proved resistant to automatisation until, in the 1970s, they were bypassed by the invention of the quartz digital watch in place of the mechanical watch. By 1987 the number of jobs in the Swiss watch industry was reduced to one-third of what it had been in 1970. It has since then doubled again, with a boom in luxury mechanical watches, but with jobs at (by Swiss standards) low wages.

Add new comment

This website uses cookies, you can find out more and set your preferences here.
By continuing to use this website, you agree to our Privacy Policy and Terms & Conditions.