The McKinsey study and the BoAML study

In late 2017, McKinsey Global Institute published a very detailed paper on the expected impact of automation on both employment, productivity growth and GDP growth. In the paper (called “Jobs lost, jobs gained: Workforce transitions in a time of automation”) they concluded that up to 375 million people worldwide could be displaced between now and 2030 and would need to be re-trained.

Under the assumption that all those who are displaced re-join the workforce and maintain their productivity level in the new job, TFP should increase by 0.8-1.4% annually as a result of automation, according to McKinsey. In most countries, that will more than offset the fall in GDP caused by a shrinking workforce, i.e. GDP growth should remain positive almost everywhere, if McKinsey’s calculations are correct.

In another ground-breaking research paper from November 2015, the research team at BoAML concluded that;

the adoption of robots and AI could boost productivity by 30% in many industries, while cutting manufacturing labour costs by 18-33%.

In other words, although the range of the estimated impact is quite wide, the expected impact of automation can only be described as dramatic. Having said that, I should point out that the net impact will vary substantially from country to country.

In that context, I note that the UK is falling desperately behind the curve as far as automating its industry is concerned (Exhibit 1). The countries on the left-hand side of Exhibit 1 are far more likely to see GDP growth re-accelerate as the result of increased automation.

Going back to the McKinsey study for a moment, a similar conclusion was reached, i.e. that the impact from automation will vary a great deal from country to country. As you can see from Exhibit 2 below, almost 30% of Japan’s workforce could be displaced between now and 2030 with the corresponding number in many EM countries ‘only’ around 10-15%.

Furthermore, just like I concluded earlier (on the back of Exhibit 1) that the UK is falling behind the curve, McKinsey reached exactly the same conclusion in their 2017 paper.

Adapting to new circumstances?

I am often confronted with the view that automation is simply too big a story not to affect the employment market dramatically. No way can all those who have been displaced be absorbed by other industries, the critics claim. Allow me to make a few comments on that.

First and foremost, people have been displaced for hundreds of years. In 1850, almost 60% of the US workforce worked in the agricultural industry. Today, the number is no more than 3-4% (Exhibit 3). The human race has always had an amazing ability to adapt to changing circumstances.

Secondly, according to the United States Census Bureau, 11.1 million people are currently employed in manufacturing over there. That corresponds to just over 7% of the US workforce. If you look at Exhibit 3 again, you’ll see that the manufacturing workforce is not particularly big when measured as a percentage of the entire workforce. If people employed in retailing can find another job after having been displaced by online retailers (and, so far, they have been able to), why can’t those who work in manufacturing?

A key challenge will be to establish a re-training programme that will allow those workers who are displaced to re-join the workforce relatively quickly. Another challenge for the Americans is the fact that the US workforce will continue to grow unlike in most other countries but more about that later.

Penetration of advanced robotics by industry

The penetration of advanced robotics varies not only from country to country but also from sector to sector and from industry to industry. The industry having been penetrated the most so far is the automotive industry (Exhibit 4), but that will probably change over time. Likewise, going forward, robots will no longer be limited to manufacturing industries but will increasingly find their way into various service industries as well.

Up to this point, the decision to go for robots over humans has almost always been a function of cost, and I don’t think that will change any time soon, unless the lack of qualified labour becomes an even bigger problem than it already is.

According to BofAML, the automotive industry was the first industry to pass the critical point of robots being cheaper than humans, but many more industries are getting closer and closer to that point, meaning that many more jobs will be lost to robots over the next few years.

The estimated impact on employment

The introduction of advanced robotics will most definitely have negative implications for job security in the short term. Re-training takes time, it may not always be available to the individuals concerned, or they may not want to be re-trained. Having said that, over the longer term, the employment market has always absorbed those who have been displaced, and that would probably also happen this time.

Bruegel did a study recently on displacements in the EU caused by robots. Not surprisingly, the only industry that has benefitted meaningfully from automation is the tech industry, and the most negatively affected workers are those who work as machine operators or assemblers (Exhibit 5).

Going forward, we will need fewer workers to do the jobs robots can do but, at the same time, we will need more workers to look after those machines. There can be no doubt that the net effect on employment will be negative, at least in the short term, but that’s where the demographic outlook enters the frame.

The winners and losers

Countries like China with access to enormous amounts of relatively cheap labour, and the US with a benign demographic outlook, don’t have the same incentive to automate as quickly as an ageing Europe does, where the working age population will shrink dramatically in the years to come. Could that be why China and the US are both falling behind Europe on the robotics implementation curve (Exhibit 6)?

Meanwhile, countries like Germany, Japan and Korea, where the workforce will shrink significantly for many years to come, are in a very different position. In those countries, robots cannot be installed quickly enough, assuming employers want to ensure that manufacturing is unaffected by a retiring workforce. Let me share some numbers with you.

Let’s assume that McKinsey is spot on when arguing that 23% of all American workers will be displaced between now and 2030. The US working age population (those aged 20-64) add up to just under 194 million and the active US workforce about 160 million. If 23% of that workforce will be displaced by robots over the next ten years, assuming the displacement is linear, almost four million people will lose their jobs to robots every year. That’s an awful lot of people to re-train!

On top of that, the US workforce will continue to grow by 1.5-2 million people every year. Consequently, it may prove a great deal more difficult than it normally would for new job market entrants to land their first job. All in all, this provides a cocktail for serious socioeconomic problems that could take years to sort out.

Meanwhile, the German working age population adds up to about 50 million people with 41 million of those being active in the workforce. Again, if we assume McKinsey is correct, 24% of those will be displaced by robots over the next ten years - almost exactly one million Germans every year between now and 2030. However, the German workforce will shrink by almost half a million people ever year over the next ten years, i.e. half of those displaced will retire anyway. In other words, the Germans (and you can add many other countries to that list) will have a much more manageable job at their hands than the Americans will.

In this context, I should point out the negative impact advanced robotics may have on many EM economies. Quite a few large manufacturing companies in the OECD have, over the last ten years, moved much of their manufacturing to EC countries for cost reasons. With advanced robotics knocking on the door, I would be quite surprised if much of this manufacturing is not moved back again, leading to a rise in the rate of unemployment in many EM countries.

The UK productivity outlook

The UK workforce is expected to grow but only marginally in the years to come. The 20-64 year old age-group will grow 0.03% annually between now and 2050.

You may recall from Exhibit 1 that the UK doesn’t do particularly well in the automation league tables. One could argue that, with a (marginally) growing workforce, the UK doesn’t face the same need to automate that Germany does, and that is indeed correct if the sole purpose is to ensure manufacturing is kept alive. However, if you want to be competitive, you don’t really have a choice but to ramp up automation.

Bank of England is seemingly very concerned about recent trends. Over the last few years, TFP in the UK has quite simply collapsed (Exhibit 7). Although I would never argue that poor productivity is only because of an outdated British industry, I would certainly argue that more automation would most definitely boost productivity in the UK.

In an international context, the picture is no less troublesome. Since the Global Financial Crisis in 2008, labour productivity across the OECD has improved by approximately 9%. In the UK, it has improved by less than 2% (Exhibit 8).

Even more worryingly, the gap has risen since the Brexit referendum in June 2016 with the UK falling further and further behind the rest of the OECD. One obvious reason for this is the limited appetite for making new investments amongst UK businesses since 2016.

Investment implications

Automation is not constrained by national borders. Just because the UK economy (and equity market) as a whole may underperform as a consequence of poor productivity growth, or that the US may face significant socioeconomic problems as the employment market struggles to pick up all those who have been displaced by robots, doesn’t mean that the German equity market will necessarily do better.

As I see it, the best way to invest in automation, and in the Digital Revolution more broadly, is to invest in various disruptive technologies and, in that respect, the US and China stand out. Those two countries are clearly at the forefront of technological innovation with most of the corporate elite in IT being either US or Chinese. Having said that, it is indeed already being recognised by investors who have enjoyed a veritable feast in listed US and Chinese IT companies since the Global Financial Crisis (Exhibit 9).

Going forward, I firmly believe investors need to think differently when investing. The days of investing primarily in asset classes (bonds, equities, commodities, etc.) and secondarily in countries are long gone.

To achieve a respectable return in these difficult times, you must invest thematically, and one of the first themes you should explore is disruption. Automation is very much part of this theme, and so are other disruptive technologies such as smartphones, AI, driverless cars, IoT and blockchain (which I will cover in part III).

Active or passive?

The first question to address when investing in disruption is whether to invest actively or passively. Numerous passive investment vehicles which adopt the disruptive technology theme have emerged over the past few years. According to Morningstar, about €7bn is now ‘managed’ in European thematic ETFs, the vast majority of which are disruptive technology funds (Exhibit 10).

One passive approach to disruption deserves a special mention, and that is the family of five disruptive ETFs which was established by Goldman Sachs Asset Management (GSAM) earlier this year. In at least two aspects, those ETFs are unique when compared to other disruptive technology ETFs:

1. The team at Goldman Sachs take the view that technology is no longer a single industry but that technology is permeating all industries.

2. Unlike virtually all other ETFs, GSAM’s ETFs are not weighted by market capitalisation.

In other words, when investing in one of GSAM’s ETFs, you don’t just invest in tech but also in other industries – in those companies that are deemed to benefit from the disruptive factor in question. Secondly, by not adopting the market cap principle, GSAM can, at least in principle, better capture tomorrow’s winners rather than yesterday’s.

GSAM have identified a number of indicators that point them to tomorrow’s winners. One such example is the number of patent applications on file. Consequently, one could argue that GSAM’s approach is smart beta investing rather than archetypical passive investing, but does that matter?

Before going any further, I must point out that we have not (yet) conducted due diligence on any of the investment opportunities I mention (by name) in this paper – passive or active – i.e. you should not regard any of this as a recommendation to invest.

In the active space, there is also much to choose from – in fact so much that a thorough review of the opportunity set would amount to writing another book! Based on the work I have done so far, names like Robeco, Pictet and BlackRock all show up on my radar screen. Take for example an actively managed fund called the Pictet-Robotics Fund, which invests in companies that contribute to and/or profit from the value chain in robotics and enabling technologies. Pictet has a long history of thematic investing and, in general, a very respectable track record.

At the end of the day, I have no strong views whether you should go active or passive when investing in disruption but believe the final decision should be dictated by the opportunity set at hand. Sometimes, no superior talent exists but, every now and then, the higher fees you pay when investing actively can be justified.

On that note, we have recently come across a very talented hedge fund investment manager who runs a very successful long/short disruption strategy, and he is willing to carve out the (extraordinarily well performing) long-part of his portfolio for a management fee you would normally only be offered when investing passively. Please call us if you want to learn more.

Niels C. Jensen

10 October 2018