Servicification of Manufacturing: Facts and reflections on policy implications

My paper is based on a paper titled, "Unveiling the Evolving Sources of Value Added in Exports," and is part of an Institute of Developing Economies-Japan External Trade Organization (IDE-JETRO) project. The basic idea is to find the statistical determinants of the servicification of exports. The paper I presented here last year looked at facts; this one looks at determinants with regression analysis. But first I'd like to put today's talk in a more general setting and explain servicification and the "Smile Curve."

Servicification and the Smile Curve are related, and, in some sense, servicification is one of the less complex and better defined aspects of the Smile Curve. Servicification specifically focuses on the idea that value added from the service sector is more important in manufacturing output than it had been in the past. I believe that such change is due to changes in globalization or in the ways in which globalization has changed production: value added has shifted away from fabrication toward service value added.

The argument that globalization changed around 1990 can be seen empirically. The G7's world gross domestic product (GDP) share, world trade share, and manufacturing share all started to fall rapidly around 1990. Using Asian data, we can see that this happened all around the world at essentially the same time. Globalization should be thought of as two processes, not the traditional one where trade and communication costs come down, national economies come together, and globalization proceeds as each nation specializes in its own comparative advantage sector. The second process is qualitatively different and perhaps more important recently.

Three types of costs separate national economies. The first is transportation cost. Clustering occurs at the micro-level (factories) when transportation costs were greatly reduced by the Steam Revolution and postwar trade liberalization. The second cost is communication costs. The third cost is the face-to-face costs, i.e., the cost of moving people to hold meetings or fix machinery. In short, these are the costs of moving goods, ideas, and people. Globalization has not been a gradual reduction of all of these barriers at the same pace. The costs fell at different times, so the world came up against different constraints and acted differently as each constraint being relaxed.

In the pre-globalized world, all three costs were very high, causing almost the entire world economy to resemble village economies, in which most people consumed items that were made very near where they lived. The first big phase shift was the Steam Revolution, which radically lowered the cost of moving goods. This was the first unbundling: production and consumption were unbundled, so nations could specialize their production, focusing on their comparative advantage sectors. Countries started producing things that they wouldn't consume with the excess exported, and people were consuming things that weren't produced locally, so trade boomed.

Globalization as seen through the lens of trade theory attempts to argue that when the cost of trade declines, free trade results, economies specialize, and consumption patterns converge, but production patterns diverge as people specialize in their comparative advantage sectors.

However, removing the cost of moving goods did not make the world flat. In fact, it led to increased clustering within nations. Instead of goods being produced in cottage industries, after this change, industrial districts and the like emerged. The world came to face the communication and face-to-face constraints, causing production to cluster in very small areas to reduce the impact of these remaining constraints.

Stages of production inside factories were coordinated by the two-way flows of goods, services, information, investment, training, quality checking, and so on. The stages needed to be conducted within an area of tight geographical proximity so that there could be coordination. This is the same reason that within companies and government offices and universities, people involved in the same processes tend to be physically close, say on the same wing and floor of a building.

The next phase transition was the information and communications technology (ICT) revolution, which radically lowered the cost of moving ideas. This better and cheaper communications made it possible for companies to physically separate some of these stages. This happened within nations as well as across nations. Detroit, for example, was the hub of auto factories in the United States, but now many auto part factories are spread out across the U.S. South to take advantage of lower wages and because they can use ICT to coordinate over distance.

This is the "second unbundling" where factories unbundled stage by stage, much as economies unbundled sector by sector in the 19th and 20th centuries. This allows firms to put some of the stages in low-wage countries, but in order to keep the process unified, the offshore factories also receive the headquarter economy's technology. Offshoring is not just moving a stage geographically and using the destination's technology and know-how. It involves moving the technology and know-how as well and using the destination's labor, capital, etc. This intrinsically brings know-how, etc. to the destination country. The types of flows that once happened only inside factories in rich countries are now crossing national borders, so the nature of trade changes radically.

There's one remaining high cost: face-to-face cost. Again, we lowered another cost and things spread out more, but it is still not a perfectly flat world. Offshoring is very regionalized. German and French companies do most of their offshoring in Europe, while those from Japan, Korea, and others do most of it in East Asia. That's because face-to-face cost is still expensive, e.g., to move skilled technicians or managers to the stages that have been offshored. So, there is one more unbundling left. When we get much better virtual presence technology, it's possible that offshoring will spread further.

There are three basic differences in the second unbundling. The first is that not just goods but also factories are crossing borders. If the Trans-Pacific Partnership (TPP) only lowers the cost of moving goods, then Japanese factories will have better access to foreign markets and production will increase in Japan. Thinking about the TPP in the first globalization process, there's no particular reason for servicification or offshoring. To understand the TPP, you need to understand that the second unbundling is also happening: factories, ideas, know-how, capital, and people move.

Know-how is making radical changes in the global economy. China went from 3% of global manufacturing to 18% in 20 years by combining G7 know-how with Chinese labor and Chinese institutions, which revolutionized manufacturing and created a combination of high tech and low wages.

In essence, the technological boundaries are redrawn. We are very used to thinking of Japan as a high-tech nation and China as less so, for example. When flows all occurred within a factory or country, this was true. Now, with global value chains, companies' know-how is crossing borders or, at a minimum, flowing within a global value chain. Technological boundaries are specific to global value chains rather than to countries.

The second major change is denationalization of comparative advantage or the competitive advantage of nations. A nation's competitiveness depends on a mix of different nations' sources of excellence. As an example, modern products from Apple Inc. depend on a combination of the company's know-how, Chinese labor, and Taiwanese management.

The third, another new phenomenon, is that globalization is hitting with a finer degree of resolution, so the global impact is more sudden, individual, unpredictable, and uncontrollable.

To focus on the Smile Curve and the servicification of manufacturing, we look to see whether the second unbundling is associated with total servicification in general and foreign sourcing of service value added in particular.

The Smile Curve stylizes the production process into three major stages: pre-fabrication services, fabrication, and post-fabrication services. Fabrication is specifically the physical assembly or transformation of things. Apple Inc. in the United States is classified as a wholesale firm rather than as a manufacturing firm, and is not an exporter in the U.S. statistics. Uniqlo Co. Ltd. is a similar example in Japan--namely, a winning firm in the manufacturing sector whose success relies on service inputs, not fabrication inputs.

RIETI did one study in Japanese on this, but the three stages are mostly unstudied. Post-1990, the value went into the corners of the smile--services--at least in part because fabrication was commoditized. Instead of being done by high-wage workers in high-tech countries, it was done by high-tech, low-wage workers in low-wage countries, and value added was a cost accounting exercise. Economy-wide, such data are collected by sector rather than at the value chain stage, which is why a global Smile Curve hasn't been plotted out. The real problem is that one company's downstream is another's upstream, so this concept isn't very well defined in an economy. Our intuition with regard to global value chains and the Smile Curve has come from business studies focused on firms. We increasingly try to apply this firm-level intuition to an economy, which doesn't really work, as there is no well-defined chain of production at the economy-wide level.

Looking at the economy-wide Smile Curve, we focus on sectoral value added from the primary sectors, manufacturing sectors and service sectors, and focus on exports rather than production. The international dimension of the Smile Curve is one of its sub-themes: the fabrication stage goes to developing country factories and pre- and post-fabrication stages go to cities in developed countries.

If we consider a dollar of exports then, as a matter of an accounting identity, the dollar equals the cost of the value added, which is basically labor, capital, intellectual property, profit margin, etc., plus the cost of intermediate inputs. Then these intermediate inputs are themselves subject to the same accounting identity, so their dollar value can be broken down in the same way, between value added and intermediate inputs. Repeatedly applying this process and keeping track of which sector and country add the value, we can get a picture of where a dollar's worth of exports were really made. Ultimately, export value can be broken down into where something is made, in which sector and which country, and can then be totaled.

As an example, an output like an electric fan involves several input sectors--primary inputs, inputs from the manufacturing sector and from the service sector--so in essence, the fan is conceived as the sum of its inputs. In 1985, Japanese exports were mostly manufactured goods, and, of those, about 80% of the value added came from the manufacturing sector from all around the world but mostly in Japan. About 13% came from the service sector and 7% from the primary sector. However, in 2005, services share of value added had already more than doubled. This is the servicification of Japanese exports. It's not so much that the exported goods have changed but that the contribution of services to their value added content has increased. Translating that into differences, from 1985 to 2005, services increased by about 13%, manufacturing declined by 10%-15%, and primary declined as well, causing the graph of this to resemble a smile.

This is an extremely broad pattern. Using the Association of Southeast Asian Nations (ASEAN), for all countries from 1995 to 2005, the same smile pattern in both rich and poor countries can be seen. This Smile Curve phenomena is fairly recent and is likely related to the second unbundling.

The trend also looks the same pretty much everywhere when examined by sector instead of by country. Importantly, the value added to exports from the manufacturing sector has fallen while the value added from services has risen, for all products and all exporters.

The main point of this paper was to determine why services were changing. We have data by sector, country and year, on how much servicification has occurred, and we try to identify with what it is statistically associated.

Four possible reasons for this have been identified. The first is simple reclassification with outsourcing. The cost of a manufacturing company's non-manufacturing workers would be classified as manufacturing, but outsourcing their work would move the cost from the manufacturing sector to the service sector. Since we measure value added by cost, this would resemble servicification. For example, outsourcing general office work would cause this this change. This reclassification is, in some sense, not real--it was only misclassified before. I suspect this is a huge factor but it is very hard to measure.

The second reason is that more services are embedded in the good itself; more design and technology. Uniqlo's HEATTECH involves a lot more technology than a typical t-shirt, and since it costs something and is rewarded with a markup, this counts as a service input into t-shirts. As products become more high-tech, they involve more design and technology, and their quality adjusted prices increase, which resemble servicification. There is a lot more software in automobiles now, from the service sector, etc.

Third, more services are involved in the production process. With domestic and foreign sourcing/outsourcing, more coordination and transportation services are required--more IT and more services. Production robots require far more services than human workers. Medical costs aren't included for humans, but robot repair is included in services.

Fourth, a change in relative prices could be at work, i.e., commoditization of fabrication. Cheaper fabrication would reduce the share of value added coming from manufacturing and increase it everywhere else, as the shares have to add up to 100. A change in the relative cost of fabrication versus services would look like servicification because we are doing it in value terms. This is also a major factor, as we are combining high tech and low wages in nations like China, the cost of fabrication--and thus the value added stemming from fabrication--is reduced. As offshoring has affected fabrication more than services to date, the relative price change between fabrication and pre- and post-fabrication services is contributing to servicification.

Which service sectors are most important in providing export value added? The share of value added coming from transportation services has risen quite sharply in all nine covered Asian countries. Wholesale, retail, and other services also have come up often. But many others--for example, finance and insurance, unclassified services, telecoms, building, and government services--haven't changed much. Increases in transportation services, wholesale and retail, and other services have been seen in other countries as well.

Collapsing all of the service sectors together, we can see where the increase in servicification originates. Where does the additional services value added come from? Over time, the U.S. input of services into Asian manufacturing and exports has risen substantially, as has China's. Japan's input has risen slightly. The standouts, though, are Singapore, having to do with transportation, China, and the United States.

The question is what determines servicification. We will focus on three possible determinants: reclassification; change in the nature of production, especially global value chain production; and change in the nature of goods. The results are not highly satisfying, but this is a work in progress.

Is global value chain participation associated with more servicification, by product, by country, by year? We have to measure global value chain participation. We have three broad measures. The first is "importing to produce" which is the share of inputs used as intermediate goods. The second is "importing to export"--the share of inputs used in goods of the importing nation's exports. And the third, related to the second but on the selling side, is "exporting to re-export."

We looked at the change in the service value added share by product, by country, by time, and then we compared it to changes in global value participation, plus controls, in order to try to find gross correlations between servicification and global value chain participation, using ordinarily least squares, the difference-in-differences approach. When we looked for total servicification, we couldn't find anything significant. I expect this is because it is dominated by reclassification.

Instead, we looked at foreign servicification, the purchase of foreign services in exports, which is slightly finer. Here, there is an interesting fact: there is no correlation between foreign and domestic servicification. This will be useful as they may have very different causes.

The first thing we found that seems to make sense is that the changes in "import to produce" in developing countries differ from those in developed countries. It looks like a Smile Curve. Developing countries that are particularly involved in global value chains tend to buy foreign service-inputs.

The interaction between outsourcing and services is statistically significant especially in plastic products and metal products but not so in some of the classic global value chains such as machinery, motor vehicles, and electronics. For other global value chain measures, it works even more effectively. Why should foreign service-value be associated with global value participation for these products? Perhaps because metal and plastic products are fairly simple, the foreign service is their design and marketing.

We have begun analyzing what foreign servicification is, and it does seem to be associated with global value chain participation, as the Smile Curve theory would suggest.

Fabrication workers in Japan are competing with robots at home and abroad in China. Abundant jobs will probably never again exist for workers with low levels of education. Good manufacturing jobs will be in services, not fabrication, going forward. Therefore, excellent services will be a new source of comparative advantage in manufacturing.

Service excellence and diversities in cities are creating a new source of comparative advantage. In this world, a great diversity of services that can be recombined quickly is needed. You can find such services in rich countries' cities. In this sense, cities are the 21st century factories or industrial districts; that is, in the value added sense. Urban policy should be considered a part of industrial policy.

We should change the way we think about trade policies, development, and job creation. The analytic underpinnings are different: everything is crossing borders, not just ships. Low-skill jobs inevitably are going abroad.

Q1. We see the Smile Curve in developed countries, but why don't we see an inverse Smile Curve in developing countries?

Richard BALDWIN
That is a puzzle. We expected an inverse Smile Curve when we first did this. Some of the more general trends, changes in relative prices, may explain that. The amount of services in all countries' exports in all sectors has gone up, so that would have to be accounted for with more general trends. That may be why the total change in servicification investigation did not work.

Q2. I'd like to pick up on something you skipped: trade policy. I agree with you that opening the market to foreign services is key to making the manufacturing industry competitive. Services are sensitive so we need to be sector-specific. In Japan, we need to know which sectors. The classification is not sufficiently fine-tuned. Could you contribute to improve this situation?

Richard BALDWIN
In these international input-output tables, the value added is presented in a rather aggregated level. The general trend is to make the value added more disaggregated, and ultimately to go to the firm level, then maybe we could help. In the meantime, perhaps you could use the service trade data to make a statistical connection between this aggregate, other services, and movement in the other aggregates, and assume that it stays the same and construct data. I will look for a correlation where those countries that started importing foreign services most exported more. That is most surely true.

Q3. You said that services are a very important factor in trade for G7 countries, including Japan. The creation of services is accomplished only by a small number of workers. We fear the effect on incomes and job distribution after the servicification society comes into existence. Do you have some ideas on that?

Richard BALDWIN
It is now superstar economics where a few people make all of the designs. That's a challenge for governments. I don't have a solution, but we have to recognize the changes, including servicification, which will put more stress on income equality. There are good jobs in other unrelated sectors as well. Interestingly, it is not high- versus low-skilled workers. In many countries, the middle-skilled workers are being hurt the most. High-skilled workers are helped by the fact that they can sell their skills around the world. Manufacturing jobs have been declining for 30-40 years, and those people have found jobs elsewhere, many of them in services. The responsibility should be allocated to governments, which are responsible for justice, and companies, which are responsible for efficiency.

Q4. Japanese people in general tend to be not good at English and are shy. Do you think that is an advantage or disadvantage in servicification?

Richard BALDWIN
Many services embodied in these exports are embodied inside a Japanese process. The Japanese design and service are component before it leaves the country. The classic service export, like IT export, may require more English, but that's not really what I'm talking about--services embedded in manufactured goods. The effect of language differences is also being reduced by technology. Things get done in a Japanese office, but there also need to be a few people to communicate outside in English.

Q5. You mentioned that with increased servicification, cities and urban policy are important. What does that mean for the rural economy?

Richard BALDWIN
This is a source of problems for rural regions. The wrong thing to do would be to try to spread out the services so that they don't recombine. Rural-urban inequalities need to be addressed in other ways. Tourism is the key in rural Switzerland.

Q6. Many people working on productivity argue that intangible investment or capital is important for productivity and competitiveness. However, in current input-output tables, they are treated as intermediate inputs. Calculation would then overestimate the importance of services.

Richard BALDWIN
That is an acknowledged problem that they are trying to fix. On the one hand, the good thing is that where the value is added can be figured out, but to do that, aggregating and mixing things up are necessary. Where the intangibles are added or who is doing them aren't known, except through anecdotes. With the right data, it could be constructed, but it is very difficult.

Q7. What do you think about immigration? The working population of Japan is declining. With servicification, robots and China take on low-skilled jobs. Is there any need then for immigration?

Richard BALDWIN
I can't say much about immigration from my analysis today. I don't think you need many immigrants to work in Japanese factories. Employment and manufacturing will change in nature. Immigrants may be important in the service sector. All Japanese people should be utilized. Japanese women and older people should be utilized first.

Q8. I agree that the city is the 21st century factory. The Shinzo Abe administration, however, says that rural cities are important for Japan to maintain sustainable growth, and it has a policy to revitalize such cities. Is that a contradiction?

Richard BALDWIN
In England and the Netherlands, efforts are being made to boost up secondary and tertiary cities, so it is not about just one major city. But it shouldn't be pretended that cities with a population of under 10,000 people are where the export excellence is going to come from.