IN THE early 20th century the future seemed bright for horse employment. Within 50 years cars and tractors made short work of equine livelihoods. Some futurists see a cautionary tale for humanity in the fate of the horse: it was economically indispensable until it wasn’t. The common retort to such concerns is that humans are far more cognitively adaptable than beasts of burden. Yet as robots grow more nimble, humans look increasingly vulnerable. A new working paper concludes that, between 1990 and 2007, each industrial robot added per thousand workers reduced employment in America by nearly six workers. Humanity may not be sent out to pasture, but the parallel with horses is still uncomfortably close.

Robots are just one small part of the technological wave squeezing people. The International Federation of Robotics defines industrial robots as machines that are automatically controlled and re-programmable; single-purpose equipment does not count. The worldwide population of such creatures is below 2m; America has slightly fewer than two robots per 1,000 workers (Europe has a bit more than two). But their numbers are growing, as is the range of tasks they can tackle, so findings of robot-driven job loss are worth taking seriously.

The paper’s authors, Daron Acemoglu of the Massachusetts Institute of Technology (MIT) and Pascual Restrepo of Boston University, are careful to exclude confounding causes as best they can. Their results are not driven by a few robot-intensive regions or industries, and are distinct from the effect of trade with China, or offshoring in general. Increased robot density does not seem to raise employment among any group of workers, even those with university education. Since relatively few industrial robots are in use in the American economy, the total job loss from robotisation has been modest: between 360,000 and 670,000. By comparison, analysis published in 2016 found that trade with China between 1999 and 2011 may have left America with 2m fewer jobs than it would otherwise have had. Yet, if the China trade shock has largely run its course, the robot era is dawning.

Economically speaking, this should not be a problem. Automation should yield savings to firms or consumers which can be spent on other goods or services. Labour liberated by technology should gravitate toward tasks and jobs in which humans retain an advantage. Yet that should also have been true of horses. The use of tractors in agriculture rose sharply from the 1910s to the 1950s, and horses were displaced in vast numbers. But some useful horse-work remained (as indeed it does today). The difficulty facing horses was in reallocating the huge numbers displaced by technology to places where they could still be of use.

The market worked to ease the transition. As demand for traditional horse-work fell, so did horse prices, by about 80% between 1910 and 1950. This drop slowed the pace of mechanisation in agriculture, but only by a little. Even at lower costs, too few new niches appeared to absorb the workless ungulates. Lower prices eventually made it uneconomical for many owners to keep them. Horses, so to speak, left the labour force, in some cases through sale to meat or glue factories. As the numbers of working horses and mules in America fell from about 21m in 1918 to only 3m or so in 1960, the decline was mirrored in the overall horse population.

The analogy with horses can clearly be taken too far. Yet the experience is instructive. Automation is reducing human wages; Messrs Acemoglu and Restrepo reckon that one additional industrial robot per thousand workers reduces wages across the economy by 0.5%. Real wage growth in many rich economies has been disappointing for much of the past two decades. Low wages are enabling some reallocation of workers. An overwhelming share of the growth in employment in rich economies over the past few decades has been in services, nearly half in low-paying fields like retailing and hospitality. Employment in such areas has been able to grow, in part, because of an abundance of cheap labour.

Yet low pay leads to policies that complicate the labour-market adjustment. Instead of bumping off excess labour, rich economies provide some social support: unemployment benefits, social security or disability payments, and assistance with housing and food. When the jobs on offer are poor, that cushion, though meagre, can be enough to draw people out of the labour force into indolence—particularly if families offer extra help.

The horses of instruction

Horses might have fared better had savings from mechanisation stayed in rural areas. Instead, soaring agricultural productivity led to falling food prices, lining the pockets of urban workers with more appetite for a new suit (or car) than anything four-legged. Similarly, the financial returns to automation flow to profitable firms and their shareholders, who not only usually live apart from the factories being automated but who save at high rates, contributing to weak demand across the economy as a whole. Indeed, roughly half of job losses from robotisation (as from exposure to Chinese imports) are attributable to the knock-on effect from reduced demand rather than direct displacement.

Today’s horses are not entirely without work. Some still find gainful employment; a few are very valuable indeed. For people to fare better, and retain more than a rump of work reserved for those of exceptional ability, they must prove a better match for clever machines than horses were for mechanical equipment. And societies should perhaps respond with more determination and care than horse-owners did a century ago.