By Sarah Watson —

Which sector — services or manufacturing — will drive India’s future growth? The basic outlines of the debate are clear: proponents of manufacturing point out that low- to medium-skill manufacturing best suits India’s comparative advantages of a large, young labor force and low prevailing wages. Those who believe India should aim primarily for growth in the services sector often respond that in the future manufacturing jobs will be increasingly thin on the ground due to the replacement of human workers by robots. Current trends suggest that India’s leaders should indeed be deeply concerned about the risks automation poses to its strategy of growth through manufacturing. But automation poses almost as serious a threat to India’s other major growth engine: the services sector.

Recent news stories occasionally make it seem that a shift from humans to robots on the factory floor or service desk is imminent. Homegrown textiles champion Raymond Inc. has announced that it is going to replace 10,000 jobs with robots over the next three years, while India’s business-process outsourcing sector is likely to lose 640,000 low-skilled jobs to automation in the next five years—albeit while adding jobs in more complex fields.

But proponents of the manufacturing path, such as Mihir Sharma, contend that fears of an all-robot workforce are overblown and that it will be “several decades” before the process of automation is fully complete, leaving India just enough time to launch itself into the ranks of developed nations. Indeed, doomsayers have been forecasting the end of human labor for years; in 1982, for instance, two U.S. economists asked whether the rise of the silicon chip would render workers “obsolete.” The computer industry created immense value and gave rise to entirely new industries — but it was also accompanied by a long-term secular decline in the cost of labor as a percentage of GDP in developed countries.

While the imminence of the automation threat is often exaggerated, at least one report predicts that we are at the tipping point for a true revolution in automation, with its greatest impact felt in manufacturing. While highly developed countries with high wages such as South Korea currently lead the world in use of industrial robots, it’s not just the world’s richest countries that are turning to automation. In China, rising wages are starting to converge with the falling costs and rising capacities of robots, making automation more feasible and attractive than ever before. The Financial Times estimates that in China the “payback period” — the time it takes for robots to pay for themselves in foregone labor costs — has shrunk from 5.3 years in 2010 to 1.3 years in 2016 and will continue to drop. The massive recent spike in purchases of robots by Chinese manufacturers — since 2013, it has been the world’s largest purchaser of industrial robots — reflects this trend.

Is India safe? Calculations for the payback period in China assume an annual employee cost of $9,000, much more than an Indian manufacturing worker could expect to earn. The Indian Labour Bureau’s Survey of Industries for the 2013-14 fiscal year found that Indian textile workers cost their employers on average $7 a day (including wages and social security contributions), while workers in the auto manufacturing industry cost their employers around $24 a day, or about $6,000 a year. Thus the payback period in India for a $50,000 robot that replaces three highly-skilled workers would be 2.75 years — about where China was three years ago. For a low-skilled worker, assuming the robot cost the same, the payback period would be well over nine years.

But while Indian wages remain extremely low, they are growing rapidly. Between the Annual Surveys of Industry for fiscal year 2009 and 2014, the average compensation of workers in the textile industry grew from $4 a day to $7 (a rate of 12 percent a year), while the cost of workers in the auto industry grew from $15.40 to $24 (9.5 percent per year). Assuming, as Boston Consulting Group projects, that the cost of robots will drop 20 percent by 2025 and that Indian workers’ wages will continue to rise at about 10 percent per year, by 2025 the payback period for a robot capable of replacing three Indian textile workers could have shrunk to 2.7 years. Utilizing the same assumptions, in the case of higher-earning workers the payback period would be reduced to10 months by 2025. Of course the decreasing cost of automation could put a low — and sinking — ceiling on labor costs. But stagnant wages defy the whole point of economic development.

While the calculations above focus on manufacturing, the mid-skill services industry is far from immune. As one study of automation in Britain found, routine, predictable jobs with middling pay, whether in manufacturing or services, are most likely to be lost to automation because they are easiest to replicate from a technological standpoint and sufficiently high paid to make doing so worthwhile. What’s left is what the study’s authors called “lovely” and “lousy” jobs: high-skilled jobs that require too much creativity and spontaneity to be performed by machines; and low-paid, low-skilled jobs that require the worker to perform tasks too irregular to be done by a machine and where wages are too low to justify purchasing expensive machinery to perform the work. Similarly, a McKinsey study found that while jobs involving “predictable physical work” — such as on an assembly line — were the most susceptible to automation, jobs that involve predictable mental work, like data processing were a near second. Ironically, the “job polarization” that could follow automation looks a lot like India’s current labor landscape, in which a relatively small number of high-end knowledge workers flourish alongside millions of underpaid casual laborers.

Of course humans will still be needed on the factory floor for the foreseeable future: to supervise the machines, perform tasks that machines still cannot, and intervene if something goes wrong. But even if automation cuts human inputs in manufacturing by 25 percent (a conservative estimate) it will be a major blow for India, which is struggling to prepare for the 150 million additional workers who will enter the job market by 2033. The toll could be similar in the services industry, but there’s an upside: one projection found that the number of low-skilled outsourced services jobs will shrink by 28 percent in the next six years, while the number of high-skilled outsourceable jobs will double in the same period. Industry shifts to ‘robotic process automation,’ or ‘cognitive automation,’ are already resulting in outsourcing cuts.

Can India learn anything from this rather depressing picture? The future of automation shows that the divide is not between ‘white collar’ and ‘blue collar’ jobs but between the middle and the two extremes of extremely low and extremely high skills. Mid-skilled, mid-priced workers, whether in the business-process outsourcing industry or on the assembly line, are at nearly equal risk of being replaced by machines. In this light, India’s strategy of seeking manufacturing investment while churning out millions of engineering graduates with minimal skills looks particularly misguided. India must prepare for the future by equipping today’s young children with the skills they need to compete in an era when the middle of the labor market is the riskiest place to be.

Ms. Sarah Watson is an associate fellow with the Wadhwani Chair in U.S.-India Policy Studies at CSIS. Follow her on twitter @SWatson_CSIS.