Thursday, 16 February 2017

Tyler Cowan, Karl Marx and transition costs of automation.

Tyler Cowan writes-
The Western world managed the shift out of agricultural jobs into industry, and continued to see economic growth. So will not the jobs being displaced now by automation and artificial intelligence lead to new jobs elsewhere in a broadly similar and beneficial manner? Will not the former truck drivers, displaced by self-driving vehicles, find work caring for the elderly or maybe fixing or programming the new modes of transport?
As economics, that may well be correct, but as history it’s missing some central problems. The shift out of agricultural jobs, while eventually a boon for virtually all of humanity, brought significant problems along the way. This time probably won’t be different, and that’s exactly why we should be concerned.
Consider, for instance, the history of wages during the Industrial Revolution. Estimates vary, but it is common to treat the Industrial Revolution as starting around 1760, at least in Britain. If we consider estimates for private per capita consumption, from 1760 to 1831, that variable rose only by about 22 percent. That’s not much for a 71-year period. A lot of new wealth was being created, but economic turmoil and adjustment costs and war kept down the returns to labor. (If you’re wondering, “Don’t fight a major war” is the big policy lesson from this period, but also note that the setting for labor market adjustments is never ideal.)
By the estimates of Gregory Clark, economic historian at the University of California at Davis, English real wages may have fallen about 10 percent from 1770 to 1810, a 40-year period. Clark also estimates that it took 60 to 70 years of transition, after the onset of industrialization, for English workers to see sustained real wage gains at all.
If we imagine the contemporary U.S. experiencing similar wage patterns, most of us would expect political trouble, and hardly anyone would call that a successful transition. Yet that may be the track we are on. Median household income is down since 1999, and by some accounts median male wages were higher in 1969 than today. The more pessimistic of those estimates are the subject of contentious debate (are we really adjusting for inflation properly?), but the very fact that the numbers are capable of yielding such gloomy results suggests transition costs are higher than many economists like to think.
Industrialization, and the decline of the older jobs in agriculture and the crafts economy, also had some pernicious effects on social ideas. The early to mid-19th century saw the rise of socialist ideologies, largely as a response to economic disruptions. Whatever mistakes Karl Marx made, he was a keen observer of the Industrial Revolution, and there is a reason he became so influential. He failed to see the long-run ability of capitalism to raise living standards significantly, but he understood and vividly described the transition costs and the economic volatility.
Cowan is wrong about the history. The French Revolution and subsequent wars, changed how the British viewed a standing army. It was no longer seen as the tool of the 'Court party' (i.e. increasing the power of the monarch) or as potentially hostile to the Established Church and Property arrangements (as happened under Cromwell). Rather, the military- as represented by 'the Iron Duke'- became a bastion against internal subversion which could be used against working people- for e.g. at 'Peterloo'.

British industrialization failed to raise real wages because the political system became increasingly weighted towards big Landowners and certain vested commercial interests. This happened because the pattern of representation in the House of Commons did not reflect demographic shifts. There were 'rotten boroughs'- once thriving market towns- which contained only one or two voters. Meanwhile rapidly growing urban centres had little or no Parliamentary representation. Under these circumstances, purely political forces, not economic ones, conspired to worsen the lot of the working man.

The Corn Laws kept the price of bread high so that the aristocrats prospered. The Combination Laws criminalised Trade Union activity. The Poor Law was used by the wealthy to reduce their labour cost and shift the burden to the rate-payer. Thus, an independent weaver like George Eliot's 'Silas Marner', or a small farmer working the land with his own family members, was forced by law to subsidise the wage bill for the big manufacturer or large agricultural estate.

David Ricardo, representing the new rentier middle class and a section of the City of London, developed an Economic theory which stigmatised the 'unearned increment' enjoyed by the Landowners. According to his theory, stagnation was inevitable unless the Entrepreneur, not the Landlord, got to keep a bigger share of the Social Product. The movement for Parliamentary Reform gained an impetus from his theory, though the Rev. Thomas Malthus developed an effective 'under-consumption' argument in favour of the idle rich but for whose prodigality the working man would starve in yet greater numbers.

Ricardo died in 1823, at the height of reaction, but had he lived he would scarcely have felt vindicated. The Corn Laws did not disappear after the Reform Act of 1832- after all, the wealthy Manufacturer could invest  in, or hold mortgages on,  Agricultural Estates-  and so the 'Chartist' struggle turned in a more radical direction. However, the lesson of Revolutionary year of 1848 when 'History reached a turning point, but failed to turn', had already been learned by the 'Physical Force Chartists'. It was the State which possessed a monopoly of coercion and was prepared to use it in a wholly ruthless manner. 

Marx and Engels differentiated themselves from the 'Young Hegelians' on the Continent by immersing themselves in English language empirical studies of the 'Proletariat'. However, in making sense of the huge amount of data Early Victorian reformers produced, Marx neglected the salience of distortions introduced by the Legal/Legislative system preferring to develop an abstract 'essentialist' theory. Thus, though a Classical Economist like Smith and Ricardo, Marx's oeuvre was not directly linked to contemporary agitation against corrupt rent-seeking in high places. On the one hand, this meant that there were no 'Marxist' politicians who, once elected, did a deal with Vested interest groups- i.e. Marxism retained a sort of intellectual purity. On the other hand, precisely because this intellectual purity would brook no competition, British Marxists resisted the Ricardian or populist conclusion- viz. tax 'the unearned increment' represented by Rent and eliminate other distortions in the Justice and Legislative system which had been introduced by rent-seeking. 

One major problem with Marx's theory is that he assumed that 'the organic composition of capital' in agriculture and mining was different- much more labour intensive- than in manufacturing. Further, because his analysis assumes a free market steady state, 'absolute rent' would not exist if agriculture or mining became more capital intensive than the average.

We, of course, live in a very different world from Marx. When Kennedy and Johnson deported hundreds of thousands of Mexican farm-workers, real wages for agricultural labour did not go up but capital intensity did. Crops which could not be mechanically harvested were abandoned. Agriculture adjusted to the supply shock very quickly- within a year. More importantly, 'Agribusiness' used some of its profits for 'rent-seeking' behaviour- i.e. influencing political and legal decisions to protect its own interests.

This is not to say that Marx's world was kinder than ours. In Ireland and the Highlands of Scotland, the Victorians presided over a vast depopulation on a familiar English pattern- sheep devoured the peasants- though the transition was longer and much more painful than in the case of the Mexican braceros.
Tyler Cowan, in his article, thinks that the transition from Agriculture to Industry in England was in conformity with Economic laws, rather than Political and Legal distortions which created rents. He thinks that Marx observed the costs of this transition and thus gained salience. He writes-
Western economies later turned to variants of the social welfare state, but along the way the intellectual currents of the 19th century produced a lot of overreaction in other, more destructive directions. The ideas of Marx fed into the movements behind the Soviet Union, Communist China and the Khmer Rouge. Arguably, fascist doctrine also was in part a response to the disruptions of industrialization in the 19th and early 20th centuries.
Cowan, admittedly, is painting with a broad brush, but there is a serious error in the above. The fact is Marx chose, like Cowan, to ignore actual rent-seeking, in order to develop a 'pure' or 'essentialist' theory of absolute rent. However, this theory was understood- for e.g. by Lenin- to mean that it was bourgeois capitalism which benefited by nationalising land. In other words, the Soviet Union, Maoist China, the Khmer Rouge and so on were always disingenuous in their land policy. The subscribed to an abstract essentialist Economic philosophy which classified the peasant proprietor as no better than a capitalist and therefore a 'class enemy'. This pathology in Communist thought did not arise because 'Marx was a keen observer of the (transition costs) of the Industrial Revolution' but because Marx chose to be a theoretician, abstracting from actual rent-seeking in England which is what caused real wages to fall, in order to have salience as the propounder of 'universal economic laws' for thinkers on the Continent where the Legal/Legislative regime was wholly different.

Cowan thinks Marx's mistake was 'the iron law of wages'- i.e. the notion that real wages can't rise for Malthusian reasons. Actually, the Marxist Economic system says nothing about what the physical standard of living will be. If the proletariat won't have babies (which is what the word proletariat means) unless a material threshold is met, then that is the new 'natural price' of labour and everything has to be adjusted accordingly.
If Cowan was correct in his analysis every Marxist in the world since 1960 would be either a fool or a hypocrite. 

More seriously, Cowan by ignoring what Marx, at least the mature Marx, too ignored- viz. rent seeking as responsible for the high transition costs in English industrialisation- is vitiating his own analysis of the likely costs of further automation.

Within the Marxist fold, we may mention Ernest Mandel as having salience here, however it is the American, Henry George, whom Cowan praised as one of the finest advocates of free trade, who really carried forward the Ricardian program and seized this bull by the horns.

Interestingly Stiglitz has a 'Henry George theorem' that is relevant in this context. A Technological revolution is like a public good. If it is associated with localised externalities and network effects then rents go up and can be taxed. This means that either equitable Hicks-Kaldor redistribution or the creation of new jobs can occur.

The problem is that such 'Knowledge Revolutions' may be 'off-shorable'. If Capital too has gone off shore, what is to prevent Technological Unemployment from triggering urban collapse? Increased Agricultural productivity depopulated the countryside. Might not once great cities- e.g Detroit- suffer a similar fate?

Subsidies to agriculture may have some good 'regret minimizing' or external benefits- e.g. maybe farmers can manage the countryside in a ecologically worthwhile manner- but politically motivated subsidies to sunset industries are unlikely to have any such advantage. During the stagflation of the Seventies, State subsidies to manufacturing industry worsened incentives to rationalise and innovate.

A Public good is something which has a zero marginal cost- it is ‘non rival’ and ‘non excludable’- like everybody profiting without necessarily having to individually pay for the benefit from National Defence or the Justice system. We have a free-rider problem here- people who benefit may not want to pay. Stiglitz has a ‘Henry George Theorem’ which says that for a localised public good- e.g. good transport infrastructure- rents go up in a particular way and the Local Authority can tax those rents so as to cover the deficit associated with providing the public good. A new Technology could be localised- e.g. around a Lab or University dept, or a particular company’s R&D facility. Some agents in this local networks can’t be excluded from having this new knowledge and can innovate on that basis. Intellectual property regimes differ but even the most stringent doesn’t allow a general Scientific principle or paradigm to be copyrighted. Since some of these agents are free-riders, there is a danger that the ‘Knowledge’ public good will be underprovided because not everybody who benefits pays for it. However, if some entity paying for the Knowledge production can levy rents, or extract rents by some mechanism, on local properties owned by these ‘free-riding’ agents, then the problem is solved.

Take a Govt. which pays for R&D at a University. It can make some of the money back by taxing property in the technology hub. Still, if the Knowledge and associated Capital can be moved off-shore- e.g. factories and labs can be moved overseas- then rents overseas will rise and so Stiglitz’s theorem seems to be defeated. The irony here is that Stiglitz is a champion of pro-poor Globalistation. However the argument could work for Trump. If the advanced country- the US- triggers a bad overseas intellectual property regime by taking Protectionist measures- Knowledge revolutions might yield only local public goods. Innovators will be wary of opening factories or labs in faraway places where the locals might simply steal their ideas. So, maybe, they will do ‘capital deepening’- i.e. double down on innovation in their own country and region. Then the Local Govt. can tax the rise in property values to fund the innovation in a virtuous circle. They could also compensate people who lose their jobs.

 A Hicks Kaldor improvement is one where we know that some people are benefiting so much they could compensate all the losers and still come out ahead. This is unlike a ‘polluter’ who makes money by inflicting more cost to others than he gets in benefit. Localised externalities and network effects should lead to higher ‘economic rent’- i.e. bigger gap between what can be earned in the next best occupation- for all inelastic factors of production. Land is what Henry George focused on but we can generalise this to other resources of an arcane type like 4G Spectrum. In theory, we can tax this ‘rent’ without a disincentive effect because the alternative occupation is so much less rewarding. In practice, this analysis falls down because longer term everything is elastic and so incentives matter more and more.

 Artificial restrictions like ‘zoning’ or ‘educational credentials’ (sheepskin effect) will tend to distort things and impose bigger and bigger ‘allocative’ efficiency losses. This is the big argument for Free Trade. Long run, any artificial distortion creates perverse incentives. The problem is that for an advanced country with very rapid Technological change and fundamental Knowledge revolutions, it may be that only the short run matters because faster innovation changes the landscape so much. It could be that by taking ‘offshoring’ off the table, a lot of time and effort which goes into doing things where it is cheapest will go instead into doing things smarter right here. The difference is that local people can capture some ‘rent’ associated with this. In particular, people who lose their jobs in manufacturing or admin can move to well paid service jobs in the same area because even if more work is done by robots or computers still the profits remain localised and so spending on high value added services will go up. One final point. Tiebout sorting means agents have a choice as to which ‘town’ to migrate to. Each town has a different mix of taxes and local public goods. Competition between towns makes for efficiency. Now imagine that Towns compete for different types of Technology/Knowledge goods. If agents are risk averse, the Town can offer a sort of implicit contract that if automation cuts jobs on the production line, locals will get preference in re-employment. Some ‘Company Towns’ do have this philosophy already. Long term, this may be what a lot of voters want- a new type of Social Contract where Knowledge based disruption is mediated by some Henry George type mechanism whereby the winners indirectly compensate the losers. The problem is that this limits the benefits of Globalised free trade. Returning to the story of ‘transition costs’- just as the majority of Britons suffered far more than necessary during the transition to Manufacturing because of corrupt political rent-seeking causing massive distortions, so too might the transition costs of a new type of Globalisation, in which Technology could be almost frictionlessly transferred to low-wage countries, have been greatly exacerbated by all sorts of distortions introduced by lobbyists. However ‘property rights in jobs’ and Trade Union power also represent distortions. A better way forward might be a new ‘Social Contract’ where the needs and fears of ordinary people are better addressed at the local level.

Trump's economics might appear completely foolish. Yet it continues a line of thinking found in Ross Perot- a billionaire with a much better reputation. In Economics, there is always another side to any argument. Thus, if 'Knowledge Revolutions' can be made to behave like 'local public goods' by certain measures we think of as Protectionist- more especially for advanced countries- then it may be possible to increase Equity without too much of an Efficiency cost because a 'Henry George mechanism' would exist so as to prevent net job loss. With subsidiarity, we might see diverse Tiebout models based on different mechanisms. In this case, even if one's job disappears because of new Technology, another job oriented towards the same Knowledge Culture would become available and so no great trauma would be experienced.

Will automation impose heavy transition costs? Yes, unless both mobility and 'Henry George type' Tiebout model diversity increases more quickly. This is unlikely to happen as a result of State action because the knee-jerk reaction would be to focus on the worst affected area and to subsidise a sunset sector while pretending to invest in a new technology centre which, it will turn out, is actually already obsolescent. Labour mobility gets frozen. Rent seeking snowballs. Stupid bureaucrats back losers. Policy Space becomes multidimensional and McKelvey Chaos prevails. 


  1. 'It may be that 'Knowledge Revolutions' can be made to behave like 'local public goods' by certain measures we think of as Protectionist- more especially for advanced countries. In that case a 'Henry George mechanism' would exist so as to prevent net job loss. With subsidiarity, we might see diverse Tiebout models based on different mechanisms'

    I'm afraid I don't follow. Are you saying that a new technology- e.g advanced general purpose robots- is 'non excludable' and 'non rival'? How could this be so? There may be externalities associated with Robot manufacture- but that's not the same thing as there being a specific public good associated with the industry.
    I don't understand how there can be any 'Henry George mechanism'in the context or how this relates to 'Tiebout sorting'.

    I think you misunderstand Marx's theory of absolute rent. However that is not the main source of confusion in the above.

    1. More Advanced general purpose robots would embody a 'Knowledge Revolution'- i.e some paradigm shift of a fundamental kind. If the application of this new knowledge to robotics is free it is likely to be non rival and non excludable for some set of agents. If it isn't free for any set of agents, even going up the decision tree to whichever set of agents came up with a paradigm shift, then we aren't really talking about Knowledge but rather some type of non-duplicable expert cognition or autistic oracle or there is a 'rent-to-ability' or something of that sort.
      Thus it makes sense to treat a Knowledge Revolution as a public good for some set of agents. Now these agents may be distributed across the world or they may be localised. In the former case, the network could itself become the basis of an unconventional Tiebout model- it could be like a global guild- but in the latter case- which we can easily visualise as University or Govt. Lab which is the hub for innovation- clearly something like a 'local public good' exists. Now whether or not the State funds the public good, clearly there is a possibility of charging for network access or taxing relevant property or capturing rent in some other way such that at optimal size there is no fiscal deficit. This is in conformity with the literature on Stiglitz's theorem.
      The obvious problem is how the tax or network charge is to be set. If we have a lot of 'Tiebout diversity'- lots of different models competing then there is a discovery process. All available information will be used.

      The problem is that if Knowledge is 'off-shorable', you will have 'underdeveloped' Tiebout models willing to pay for the privilege of hosting superior technology. Weighing against this incentive is the risk to intellectual property by going off shore. If a very advanced country takes protectionist measures, it is likely that global intellectual property regimes would become less favourable or more uncertain. In this case, under subsidiarity- i.e. where Tiebout sorting is empowered through devolving of fiscal powers and toleration of hybrid local covenants (laws)- it is possible that you have a golden path whereby social costs of automation are compensated for as they occur. It may be that this creates better correlated equilibria in Knowledge production itself.

      I do find Marx's theory in this regard confusing. I suppose I am influenced by how people like Hyndman developed 'Marxist' ideas in connection with the Indian Land Tax.

    2. I don't think that's how intellectual property works. I know of no attempt to extend Stiglitz's theorem in any such direction.

      I am not familiar with 'Hyndman' & the Indian Land tax and can't comment. The Marxist theory of absolute rent is of course open to interpretation but I am not aware that your remarks accurately reflect recent scholarship in this regard.