Reindustrialisation
June 2024
The decline of British and western manufacturing is buttressed by two broad problems. The first of these is an aging workforce, one which is spread across a fragmented industry. Production will soon be hampered by the former, while the transfer of skills will be limited by the latter. The clock is ticking: UK welders having an average age of 55, machinists 53, and the overall manufacturing workforce 49. Small shops will close as owners retire without successors. 2022 gave us a strenuous insight into what supply-side inflationary shocks look like. A repeat of circumstances with even weaker domestic manufacturing capacity will give us something far uglier, more so when you add the risk of direct war at an industrial scale.
The second has been the decline of domestic manufacturing capacity and its subsequent outsourcing elsewhere. This process is a product of political complacency. Shifting production abroad was the ‘efficient’ but ultimately lazy move. In the short run, governments can cause disinflationary prices with minimal effort by outsourcing production to lower cost countries. In the process UK manufacturing output as a share of the total economy halved since 1990. Over the long run, it is the countries that make the effort to maintain their industrial competitiveness that will prosper. A country’s comparative advantage in manufacturing (and other sectors) will evolve as a function of time, input cost differentials and global economic trends, so it holds true that opportunities remain for the countries that are committed. In the last two decades, China’s industrial output rose 6-fold, staying at ~30% of GDP, despite real manufacturing wages rising 13-fold to levels far above regional competitors. South Korea, whose people are now richer than those in the UK, has maintained manufacturing at 25% of GDP since the late 1980s despite their wages rising 9-fold and the UK’s 3-fold over that time period. While the West rested on the laurels of cheap credit and stable global supply-chains, it outsourced steelmaking to countries that decades later now export the world’s leading semiconductors and solar panels back to us.
These problems, deep but not irreversible, beg the question of what is to be done. The consolidation of machining and tooling shops is inevitable. Large incumbents and new startups alike can agglomerate those leaving the market. Bringing them together offers economies of scale: localised knowledge can be introduced systematically across a company to expand technical offerings, whilst also being taught to a critical mass of new employees to sustain and grow skill sets internally.
In the UK, an obvious bottleneck for manufacturing – perhaps the critical one – is the lack of industrial automation. The UK has just 111 robots per 100,000 workers (aka robot density). For South Korea this is 1,012, for Germany 415, and for Japan 397. China increased their robot density 4-fold over the last five years to 397. Even among less advanced manufacturing countries the UK still lags behind: Canada, Italy, France have approximately double the robot density. Ideas for policy reforms are numerous – more generous R&D tax credits above 13%/230% model, loosening write down allowances to 25%+, an HVM Catapult dedicated to automation – but British industry must take the lead in admittance that much of the failure to automate is driven by a cultural reluctance. With strong execution, British firms that offer automation services will be doing an important job with large opportunities for profit (£1.5bn/£3bn per annum of extra sales if we can reach the robot density of western/northern Europe).
Automating larger segments of British manufacturing will require newer products targeted to the appropriate sub-sectors. Incremental sales can be made to industries like automotive manufacturing, food and beverage processing, and electronics manufacturing, which are already saturated with automation. The highest marginal returns lie in sub-sectors ripe for disruption like construction, agriculture, and logistics which remain largely unpenetrated. Unlocking this will require more radical product offerings in the hardware and, especially, the software space.
From a cost perspective, hardware components are undergoing structural deflation with the prices of conventional industrial robots falling 25% from 2010-20. This is a positive trend, yet Amdahl’s law means that the gains from optimising only one part of a system will lead to diminishing returns at the system level. Hardware parts becoming cheaper will become increasingly irrelevant as they represent an ever-smaller percentage of overall costs. Despite this hardware deflation, all-in producer prices for industrial robot installation rose 3-fold over that same decade due to other constraints such as labour, programming costs, and control systems. New machine learning software has the potential to be a driver of broader cost efficiency: first-principles safety systems free of cumbersome constraints, IoT diagnostics with more intelligent predictive powers that reduce downtime, enhanced visual recognition lowering the need for human-robot interaction (further reducing risks to safety); the opportunities are boundless. Importantly, intelligent software systems will allow more to be done with the same amount of hardware, loosening the grip of Amdahl’s law and allowing hardware deflation to drive system-wide price competitiveness.
That’s the plan. Agglomerate smaller businesses and modularise hardware solutions with cheaper software to drive economies of scale. Iterate the process within and across firms to optimise software with more training data, and to drive a learning curve effect with cheaper hardware installation. Let’s increase that robot density.