Five lessons from the semiconductor shortage

BofA with the note, in the hope that Canberra gives a flying $%&#:

Autos, chips and Covid: five economic lessons

Global supply chains are in disarray. A series of pandemic-related shocks have disrupted production, while goods demand has soared as consumers have substituted away from high-touch services. Each item faces its own specific set of issues, but many of the underlying problems are similar. Here we look at these problems and draw five economic lessons by digging deeper into one of the most widely-discussed stories: the auto production disruption driven by the shortage of semiconductors. We argue:
• A supply chain is only as strong as its weakest link. Seemingly small disruptions can snowball into a big economic loss.
• There is a strong case for diversification / localization of supply chains, but it will be hard to reduce dependence on Asia.
• “Transitory” goods price increases might turn out to be uncomfortably sticky, weighing on the consumer outlook.

Semiconductor supply chains

Broadly, there are three steps in semiconductor production. The first is integrated chip design. The US is dominant in this space, with Europe and Japan playing smaller roles.

Next is chip manufacturing. The US is the world’s largest single producer (33%), although collectively Asia dominates, making almost 60% of chips. The last stage is packaging, assembly and testing, which involves larger numbers of lower-skill, lowerwage workers. The ASEAN countries, including Malaysia, Singapore, the Philippines and Vietnam, and some North Asian countries such as China, Korea and Taiwan, are the main players in this step of production.

Vivek Arya and team estimate that global semiconductor chip sales amounted to $440bn last year, and should increase by around 20% this year. For a deep dive into global semiconductor supply chains, we recommend a report from earlier this year by our Korea economists Kathleen Oh and Bum Ki Son: Asia Economic Viewpoint: Can Korea and Taiwan maintain supremacy in semiconductor manufacturing?

Semiconductor trade

As with many manufactured products, semiconductors have complex global supply chains. We use data from the International Trade Center (ITC) and focus on goods that fall under Harmonized System (HS) code 8542: “electronic integrated circuits; parts thereof”. HS codes are used to classify goods in international trade. “Integrated circuits” are semiconductor chips (Exhibit 1).

Note that the trade data do not account for the full universe of semiconductor chips, only those that are traded internationally. For example, if a chip is manufactured, assembled and tested in China, and then stuffed into a cell phone in China, it would not show up in the chip trade data. Despite this, the total value of international trade in semiconductor chips ($784bn) is nearly twice the above-mentioned estimate of global chip sales. This is because the trade data measure the value of goods traded rather than each country’s value added, so there is considerable “double counting”. Exhibit 2 illustrates this issue with a stylized example of a chip that is designed in the US, manufactured in Korea, assembled in Malaysia and then exported to some other country (e.g., China) for use in some consumer product.

The ASEAN region plays a major role in semiconductor trade (Exhibit 1). On paper, Singapore is the region’s largest exporter, accounting for 11.0% of global chip exports.

But a significant share of Singapore’s exports are likely trans-shipments of chips assembled in other countries.

Malaysia is the other major player in ASEAN: its global export share has risen from around 5% in 2016-17 to 6.3% in 2020. This figure is a lower bound for the industry’s dependence on Malaysia: chips likely do not pass through the country more than once, since it is only involved in the last stage of production. Exhibit 1 also estimates an upper bound by dividing each country’s 2020 exports by the above estimate of total global sales ($440bn). This gives 11.2%. The truth is most likely somewhere in the middle.

Semiconductors and autos: the missing link

All complex electronic devices today run on semiconductor chips. In terms of end markets, the largest ones are memory devices (27% of global semi revenues in 2020), computers and storage (24%) and wireless communications (i.e. smartphones and wireless infrastructure, 19%). Meanwhile the auto sector accounted for just 8% of global chip revenues last year.

Our equity analysts estimate that the global average cost of components in a vehicle in 2020 was $15,100, which included $460 worth of semiconductor chips. Therefore chips make up only 3% of the total content value of autos. By contrast, they account for 25- 30% of the sale value of consumer electronics. Nonetheless, semiconductors are indispensable in auto production because all autos that are manufactured today include electronic parts (although chip intensity varies considerably across the different makers).

The Malaysian connection

Malaysia appears to play a disproportionate role in semiconductor supplies for the auto industry. It has a large share of the North American market, accounting for 47%, 37% and 14% of chip imports by the US, Mexico and Canada, respectively. While we do not have the data to confirm the end market (autos, computers, cell phones, etc.) for these chips, Malaysia is very likely a key supplier of finished chips to the North American auto industry (Exhibit 3).

Things are a little more nuanced in Europe. There are five major auto manufacturing hubs: Germany, Sweden, Spain, the Czech Republic and Poland. Malaysia is the thirdlargest supplier to Germany, accounting for 14% of its chip imports, and the largest supplier to the Czech Republic (31%).

Meanwhile, Sweden and Poland import a large share of their chips from the Netherlands (26% and 43%, respectively). But the Netherlands is not a major chip manufacturer.

Rather, Rotterdam is a huge port, through which many goods enter the EU. And digging deeper, we find that Malaysia is by far the biggest chip exporter to the Netherlands.

Finally, Spain imports nearly half of its chips from Germany and the Netherlands: again a meaningful share of these chips were probably assembled in Malaysia. So in the EU as well, it appears that Malaysia plays a bigger role in chip supply for autos than for other end markets.

In Asia, despite the geographical proximity to Malaysia, auto manufacturers appear to be less reliant on Malaysia. The major hubs – Japan, Korea, China and Thailand – all import less than 10% of their chips from Malaysia.

Five lessons

What economic lessons can we learn from the semiconductor-driven squeeze on auto production? We think there are five.

1. A supply chain is as strong as its weakest link
The difference between trade volumes and sales estimates for semiconductors highlights the complexity and fragility of the supply chain. With the same chip crossing multiple borders during production, a disruption in any country would hurt supply. During the pandemic we have seen two types of supply disruptions. The major chip manufacturers (e.g. Korea and Taiwan) cut investment last year because they expected demand to collapse. Instead, demand for electronics and autos surged.

More recently, South East Asia has been hit hard by the Delta variant. The testing and packaging that is done in the region is the most labor-intensive stage of the chip supply chain. Therefore the potential impact of mobility restrictions in the region on chip production could be significant. Although industry-level data are not yet available, the collapse in PMIs across the region is a worrying signal (Exhibit 4). Malaysia in particular is still dealing with a massive Covid wave, which forced the country into a nationwide lockdown in early June. Three months on, reported cases are even higher than in the US on a per-capita basis and deaths are more than double. There is probably significant under-reporting with the national test positivity rate at over 13%, so the actual statistics are likely much worse. Daily confirmed cases are still on an upward trajectory in key industrial states such as Penang and Johor, although there has been a notable decline in a larger state, Selangor, in recent weeks.

Since the vaccination rate is rising quickly, the government has started to ease restrictions in Selangor as of today (Sep 10). This should help with the movement and availability of labor. Factories are allowed to have 80% manpower (vs. 60% during the strictest phase of restrictions). Special allowances are also being made for firms that have a large proportion of vaccinated employees, taking advantage of a targeted vaccine rollout for the industrial sector.

While the peak disruption from the outbreak is likely behind us, it may still take a few months for production to recover to normal levels. Given evidence of lower vaccine efficacy against the Delta strain, we think some containment measures may well remain in place until early 2022.

2. The snowball effect

The auto industry is a low-priority end market for chip makers for two reasons. First, as discussed above, it accounts for a relatively small share of chip manufacturers’ revenues.

Second, the chips used for autos tend to be lower-margin products. This is why the auto industry has had to bear the brunt of the ongoing chip shortages, while production of consumer electronics has continued with relatively little interruption.

The fact that chips make up less than 3% of the sale value of autos means that a seemingly minor disruption has snowballed into a major loss in auto production.

According to IHS, the global auto industry lost 4mm units of production in 1H 2021.

News reports suggest this could be worth $110bn. IHS also thinks an incremental 3- 5mm units are at risk in 2H. Therefore the total loss in sales this year could climb to at least $200mn. This amounts to 0.2% of global GDP, although the actual output loss will be smaller because some parts of unfinished autos will be accounted for in inventories.

But as discussed above, there is a risk that production disruptions could continue into next year. Moreover, the auto shortage has led to a spike in prices of not only new but also used cars, which should further dampen economic activity.

In value-add terms, the source of the shock is particularly small in the case of the Covid outbreak in South East Asia, since the testing and packaging that is done in the region is the lowest value-add stage of chip production. But the key issue is the availability of substitutes, not the share of value added to the final product. There are no substitutes for semiconductor chips in most electronics or for electronics in autos, and the testing and packaging process is not easy to replicate elsewhere in the world in the near term.

3. Strong case for supply chain diversification and localization

It follows that firms and the global economy as a whole would benefit from diversification of supply chains, in order to reduce the risks of over-dependence on a specific country / region. The other option is to localize production. Indeed, our house view even before the pandemic was that we are moving towards greater localization of supply chains.

However, we expect these changes to be “tectonic” in nature: they will likely play out over years, or even decades, rather than months. Different stages of production require different skill levels and labor intensity. Therefore it would be very challenging to replicate each stage in a different country / region of the world in an economically efficient manner.

Political considerations are also important. China has arguably been the region’s most reliable supplier after the initial lockdowns because of its ability to control the pandemic.

Does that mean firms will shift more production to China? US companies probably will not, because the geopolitical risks around US-China tensions show no signs of easing.

4. Difficult, and perhaps undesirable, to wean off Asia

Localization of production will be particularly challenging in the tech sector. China’s dominance in electronics production is the main premise of the US-China tech war. It is one thing for the US to reduce its dependence on China by shifting its electronics imports to other Asia-Pacific (APAC) countries such as Vietnam. This should be feasible to a degree, and it is already happening.

But it is another matter altogether to significantly reduce the US’s dependence on Asia in tech. APAC is heavily involved in not only electronics production but also, as discussed above, every stage of chip production after the initial design process. Labor costs in the
US are not competitive with those in APAC. While Latin America is geographically more proximate to the US, APAC is a faster-growing region with greater tech production capacity and experience. So shifting tech supply chains to Latin America will also not be
easy. Given the ubiquity of electronics, this also means the global and US economy will remain heavily dependent on APAC in the medium term.

Even if the US could significantly decouple from all of Asia, there are political reasons not to. We think the Biden Administration is looking to take a multilateral approach to confronting China. To this end, it would be helpful to maintain strong economic ties with other Asian countries instead of crippling their economies by reducing their access to the massive US consumer market.

5. Transitory price increases might be here to stay

In the nearer term, our takeaway is that the ongoing “transitory” increases in goods prices might prove to be uncomfortably sticky. Continued supply disruptions could delay the expected collapse in motor vehicle prices, which in the US are running 17% above their pre-Covid trend (Exhibit 5). And there are many other stories of supply issues beyond autos. The global shipping container shortage continues, with shipping prices (including for autos) still skyrocketing (Exhibit 6). Covid-related lockdowns in ASEAN are also impacting apparel production, most notably in Vietnam, which was remarkably successful at containing the initial Covid outbreak last year but is facing a much larger Delta wave this year.

If prices remain elevated but do not increase significantly further, we should see some cooling off in inflation. Even in this case, high price levels would probably weigh on real goods spending. With the Delta surge also slowing the services recovery, we see growing risks to the consumer outlook.

Houses and Holes
Latest posts by Houses and Holes (see all)

Comments are hidden for Membership Subscribers only.