The Structure of Intra-Group Ties: Innovation in Taiwanese Business Groups

The paper I am going to present today is part of my research interest in business groups and innovation. How do we actually define what a business group is? In Japan for example, you have the keiretsu; in Singapore, government-owned business groups, like Temasek Holdings; in Malaysia Sime Darby; in Korea, the chaebols such as Samsung, and Hyundai; in India, there are Tata, Reliance, etc. Basically, business groups are everywhere. Although the definition of groups varies across countries, the broad idea is that business groups are sets of legally independent companies with activities in multiple industries that are linked as affiliates through persistent informal and formal relationships. Formal ties could be equity ties when there is cross-shareholding, like in Japanese keiretsu sometimes. It could be director ties when they share a common director. Or it could be operating ties when two or three firms within the same group act as buyers and suppliers - very common, again, in the keiretsu system. The important point is the legally independent part, because that allows them to rank as independent. So there are limited liabilities. But they are different from independent firms; in an independent firm, sometimes ties are formed and then broken. Here they are more persistent.

Why do we see these kinds of business groups all over the world? One explanation is that in developing countries at the early stages of development, when you do not have well-developed market institutions, groups operating in multiple businesses act as a de facto market, filling up the "institutional voids." Because groups are big, and because they contribute a huge part of the national GDP, groups tend to have a lot of control over the economies in which they operate. On the one hand, groups have the critical mass and the scope of multiple lines, which can allow them to get external knowledge from other countries or companies and diffuse it within the group. On the other hand, the same kind of resources could also allow groups to create entry barriers for non-group firms which do not have this kind of financial muscle. Overall, we have a relatively good sense of why groups exist but do not really know whether groups facilitate or inhibit technological dynamism. Given the dominance of groups in many economies including Korea and Taiwan, the role groups play in these economies will determine if these countries will move forward or stagnate. Thus, it is important to understand when and how business groups affect innovation.

The role of group may vary across institutions. The weaker the institutions are, generally speaking, the higher the benefits of group affiliation are. If I need access to money, in developed countries I can seek venture capital (VCs); but many emerging economies do not really have VCs. So it is important to have somebody within the group who is able to lend the credit when one needs to do something quite risky. Without that, one would not be able to innovate. My paper with Sea-Jin Chang of Korea University and Chi-nien Chung, my colleague at the National University of Singapore (NUS), which is forthcoming in Organization Science, finds that business groups are more important in Korea than in Taiwan, because Korea is behind Taiwan in terms of innovation supporting infrastructures such as venture capitals, or government funded research institutes.

Also, the type of industry matters. For example, if I am in the chemical industry where basic R&D is really risky, without somebody allowing me to diversify some of my risk I would not be able to do R&D. So in my paper with Will Mitchell at the Fuqua School, Duke University, which is published in Management Science in October 2004 that the most innovative industries are the ones where there are a combination of business group-affiliated firms and a combination of independent firms.

However, we often observe that within the same institution setups as well similar industrial settings, there is significant variation across groups in terms of innovativeness. For example, Daewoo and Samsung, before 1997, were in similar sets of industries but Samsung was much more innovative than Daewoo. Even within the same business group, not all firms are equally innovative. For example, in the TSMC Group TSMC comes out with lots of patents and Vanguard also has lots of patents, but not to the extent of TSMC, producing similar kinds of things. How do we explain this variation? A central insight of my research is that intra-group ties may hold an important key to understanding why some groups are more innovative than others and why some affiliates within the same group are more innovative than others.

Why should ties matter? The general public mind has gone through a kind of reversal. In 1992 Business Week talked about how NEC could do great things because Sumitomo Bank was able to lend money credit to NEC that American companies would not be able to. But in late 2003, Carlos Ghosn was talking about how Nissan has to dismantle its style with all these keiretsu members, which were actually pushing it down. At the same time, Ghosn was not saying that they need to get rid of all the ties; they needed to get rid of some ties but maintain others. It is not that ties themselves are always bad, but what kind of ties? The fact is: the ties which are still alive and kicking. If you look at Korea, for example, SK Telecom, when it was having problems, had to rely on SK Global to bail it out.

How do ties matter for innovation? What is the effect of these ties that the different affiliates have with each other? How does it affect the innovativeness of the individual affiliates? Does having more ties make an individual affiliate more innovative or less innovative? The other side of the puzzle is: as a group if I have more ties with all my affiliates, if it is a very dense group and if everybody has ties with each other, does it make my group as a whole more innovative than other groups? We are interested in the innovativeness of the firm and of the group as a whole.

If you look at the technology studies, each type of ties - equity ties, director ties, operating or the buyer-supplier ties - can benefit or inhibit innovation. Equity ties or investment ties may provide access to money which can allow me to diversify some of the risk when I do not really have VCs. Also, people say in the United States, company managers are sometimes driven by stock markets which sometimes have short-termism. So, to an extent, a business group-affiliated firm does not need to worry about business failure because it is on more solid ground; it can take long-term bets. The negative side is about the relationship among affiliates within the group. Sometimes families, for example, use these ties to channel resources from the more profitable businesses to the less profitable ones, a practice sometimes known as "tunneling." How about director interlocks? The benefit here is the access to information. Haunschild and Davis talk about the situation: when somebody from HP sits at the board of IBM, he knows what kind of things HP is doing, or what kind of things IBM is doing. It could act as a source of information, especially in the context of an emerging economy which does not really have alternative sources of information - analysts, newspapers, and trade journals. Similarly, buyer-supplier ties give you access to complementary resources. In developed countries, you may have independent firms which can provide better quality products than firms within the same group, but you may not have that in an emerging economy. The problem is you might be forced to buy things from each other, even if it is poor quality, or the price is not good. It is especially problematic if you happen to be a superiorly performing affiliate within the group, because then you might be subsidizing the inferior ones. You may have less incentive to invest in R&D because if you succeed, somebody will squeeze all the profits from you anyway; if you do not succeed your survival might still be guaranteed, because somebody else is subsidizing you.

The group literature tells us that ties are important but it does not tell us how ties actually affect innovation. Where do you look if you want some answers? One place could be the works about diversification and multi-business firms. But this research talks about diversification and does not really focus on the ties. The only set of research recently which talks about this is the network literature, which says that ties are important, but does not really focus on which types of ties are important. However, it also shows some aspects on the benefits and costs of ties that depend not only on the types of ties but also on the structure of ties.

What do we mean by "structure of ties?" This will be one of my most important variables. There are two concepts. The first one is centrality. Centrality is a firm level concept. It measures how many ties one affiliate has with other affiliates within the same group. On the one hand, being central is good because you can actually get resources from everybody in the group. But it can also mean you are spending too much time talking to everybody else and have less time focusing on innovation. It can also mean that you have fewer reasons to go beyond this room, because everything that you need, you have. But innovation sometimes needs ideas out of the box; you need to go outside. How about density? Unlike centrality which is a firm level construct, density is a group level concept; it measures how connected firms within a group are to each other. Is it good to have density of ties? In the network research, some people say that density is good because it leads to trust. If you see people all the time, you trust each other. You often talk about things which are unrelated to what you really need, but these unrelated things eventually lead to something you did not expect, and new ideas happen. An equally compelling view is that density is not good because of too much inbreeding; also, a lot of duplication. If you talk to somebody in the same room, after two or three people, maybe you are not adding anything new because you are not going outside. But an important point to remember is that centrality is a concept which is used at the individual firm level and density is a concept used at the group level.

Because we were stuck with a theory which does not provide us with a set of definitive predictions, what we did was something of a new cutting-edge frontier research. The scope of the study: we have three kinds of ties, we have structure of ties, innovation, and I used patents as a dependent variable. In innovation studies, you have different streams of research. People focus on individual companies or individual industries; I work on large samples. Working with large samples has its own weaknesses, because we would not be able to go into depth in certain areas. Some people use R&D as a measure of innovation. But in many countries research and development could be operating expenses that are done on the shop floor, so you do not really know how reliable this data is.

On the one hand, a patent is something where everybody can say "these are normalized." Normally in my research when I compare across countries, I use U.S. patents, because they are useful for normalization purposes. Here I would be using domestic patents. Why does a company patent? Maybe to create entry barriers, but if I can use non-patent measure to stop people from entering, I do not need to patent. One could argue that these business groups, because they are so big, can be innovating but do not really take domestic patents; they do not need to. So if I look at domestic patents, maybe I will be under-representing business groups, because they do not patent does not mean they do not innovate. On the other hand, if I look at U.S. patents, I would be over-representing firms which may be big and which have a reason to export to the United States. But smaller groups which would be innovating do not compete in the U.S. market, so they do not patent in the U.S. Often what I found in Taiwanese data is those who patent locally also patent in the U.S. So I used domestic patents.

The level of analysis comprises two stages. In the first stage, we examine the question of why one affiliate is more innovative compared to another? The second stage is about group-level innovation: why is one group more innovative than others? Why do I care about Taiwan? What I found out in a chapter that I wrote for the Oxford University Press which just came out this year, is that the groups in Taiwan are actually becoming stronger. There is a prevailing view that you have small independent firms which are dominating and you also have big groups in Taiwan like Formosa Plastics that is actually becoming bigger. Why do I care about innovation and why do we care about groups? Almost 40% of the U.S. patents awarded in Taiwan between 1990 and 1999 went to group-affiliated firms. If you look at seven of the top 10 Taiwan-based recipients of the U.S. patents, they were business group-affiliated firms. I found out in my other research that in Korea, group affiliation helps much more than in Taiwan, because in Taiwan you have the Industrial Technology Research Institute (ITRI). ITRI is helping to do a lot of the innovation for small firms; you do not need to be a big group-affiliated firm. On the other hand, if you look at groups such as United Microelectronics Corporation (UMC) Group, Taiwanese Semiconductor (TSMC) Group and the National Science Council - again, a think-tank or research institute - Vanguard, part of TSMC, seven out of the 10 are business group affiliates. They are actually important forces if I want to study innovation in Taiwan.

There are also some other reasons why I care about Taiwan. We need to know the boundaries. Why do I not use Japan as a context? That is because it is not always so clear where the boundaries of keiretsu are. In Taiwan we know which groups are a part of which business groups. On the one hand, with chaebol you have very centralized systems. You have precedence with a chairman, a patriarchal system. Then on the other extreme, the Japanese system of keiretsu, you have a precedence council, where different members of firms meet but everybody is a kind of independent. They are not so strong. Taiwan is in the middle, in a way.

This is an example of Hon Hai Group. Hon Hai Precision is now the largest OEM contractor for LCD displays, connectors, enclosures, cabling stuff, and all of this. They are big OEM suppliers for Cisco, Dell, Nokia, Sony, Apple, IBM, and everybody. Often it works under the name of Foxconn, which makes wireless mobile telephone sets. It gets the LCD panels from Innolux and it gets the thermal casts from another firm, maybe Foxconn Technology. The business model for the Hon Hai Group combines centralization and autonomy. The group locates R&D and design, production process testing, and sample production facilities within business units close to its key customers, while also retaining a central development facility within HHPIC. The group uses a global ERP production management system to connect manufacturing facilities across its major market regions in Asia, North America, and Europe. At the same time, executives in the group note that each subsidiary has a particular mission, such as making specific components, and that achieving its mission is the primary focus of a subsidiary.

Ties can help as well as hurt. For example, one of the daughters of Mr. Wang (founder of Taiwan's Formosa Group), Charlene, heads First International Computers FIC, in 1997 the largest producer of motherboards in the world. The other sister, Cher, in 1997 started the company VIA, making CPUs. FIC and VIA share equity, director, and operating ties within FPG. When VIA began to produce central processing units (CPUs), most motherboard manufacturers in Taiwan were reluctant to use its CPUs because they feared being punished by Intel, their major CPU provider. FIC used VIA's CPUs in its LEO computers, however, which helped VIA become established while FIC gained high quality, lower cost CPUs. The ties between FIC and VIA have contributed to ongoing innovation at the two firms, as the companies have co-developed a series of motherboards for the Intel Pentium 4 platform. For instance, FIC introduced the P4-800T ATX motherboard in 2003, based on the VIA PT800 chipset. Nonetheless, the same ties that facilitate innovation sometimes also create constraints. In 1997, for example, Cher Wang and her husband established Xander Inc., as another affiliate of FPG. Since Cher Wang sat on the boards of both FIC and Xander, as well as her own VIA affiliate, she was able to use her interlocking directorships to transfer customers from FIC to Xander. This hurt the profitability of FIC and its ability to fund innovative activities, fuelling wide spread resentment. An additional constraint emerged in 1999, when Intel sued both FIC and VIA, claiming that VIA had violated Intel's chipset technology intellectual property. Intel included FIC in the suit because of its group ties to VIA, while ignoring other companies that used VIA's chipsets. The suit diverted time and money that would otherwise have been available for development activities at FIC. The core point is that intra-group ties can create both benefits and constraints on innovative activity.

Some data - centrality of firm A: it has four ties with other members, divided by the total number of firms - four - so centrality of firm A would be simply one over four. So with centrality of one firm within the network of firms, it goes from zero to one. Density is simply the actual number of ties divided by the number of potential ties. For each type of tie that we talk about - equity ties, or directorate ties - you have one centrality measure which is at the firm level and you have a density measure which is at the group level. Why? Because if I am interested in why one firm is more innovative than the other, I need to have a firm-level measure; if I am interested in why one group is more innovative than another, I must have a group-level measure. What we find is that Taiwanese groups are becoming more independent in terms of buyer-supplier ties over time, because over time they are becoming more diversified. So there is less to buy and supply to. On the other hand, the director density and equity density is slightly increasing as they become more diversified. You need other ways to control them.

I am trying to explain innovation using patents. My main explanatory variable is centrality. If I am interested in affiliated innovation, my main variable is affiliate centrality. I have three types: equity, operating and director. Controls: access to capital, technological opportunity, industry. Methods: a patent is "zero, one, two, three," those kinds of things. The problem is: you may have lots of zeros in this case. Only 24% of firms in this sample have any patenting at all. So if I see a firm which does not do any patenting, I do not know whether the firm actually patented at some time but for some reason does not patent now or if it is a firm which never patents for some reason. In econometric terms, it could be unobserved heterogeneity. The model to use, estimation technique, would be zero-inflated negative binomial.

The results: from the individual firm's point of view, it is good to have lots of equity ties. However, the benefits are maximum when a firm is part of a group where there are few ties. It is good to have many equity ties (high equity centrality) as long as there are no indirect ways of getting the same resource. If there are other ways, having direct ties with everybody is actually a waste of time; we are duplicating. While equity ties might be good, a high-director centrality is bad; having directors who sit on the boards of multiple firms is a bad thing for innovation. Directors are managers who are managing different things. They have limited cognitive ability to understand and compare things. Sometimes they focus on something easy, like the return on assets, not the technology which goes on; something which gives them immediate return. So having competing demands for the cognitive ability of top managers is not a good thing. This has implications for matrix management, when we see one vice president responsible for five, six, seven, versus another organization with one vice president per division. I would say having multiple alliances within one vice president is not a good thing. How about centrality for operating ties? Up to some level it is good to have many operating ties - you are combining ideas and resources - but with too many buyer-supplier ties you become entrenched.

The group level reinforces the results. At the group level, we see that director density is a bad thing. Somehow it had no negative or positive effects on equity density. If density is so bad, why do we see so many groups with high density of inter-group ties? Maybe they do not care about innovation; maybe there is a tradeoff between innovation and profitability. What I found out is, maybe density always has a negative effect on innovation, but for performance you have some positive effects before it becomes negative.

The last thing is: it has implications for network research and multi-business theory. It also has implications for the type of structure which is best for innovation: you need a combination of a central firm, which can actually act as an integrator for getting resources from each other, and lots of scouts. These have only one tie each, so they will go outside and get the ideas. When the ideas come, they will pass them on to this guy, because without the ideas, they cannot do anything.

The affiliates have fertile opportunities for innovative activity if they have broad access to financial resources within otherwise sparse financial networks, access to a moderate degree of operating knowledge from other members of the group, and autonomy from inter-affiliate director interlocks and over-embedded buyer-supplier ties that would impose strategic constraints. By contrast, business groups with dense networks of inter-locking directorships and operating ties often constrain innovative activity. More generally, the study suggests that multi-business firms that require only limited overall operating and strategic interdependence among their units will place few constraints on the innovative activities of their subsidiary units. In turn, rather than simply act as a conglomerate holding company, multi-business firms will facilitate innovation if they have a few central units that receive corporate financial support and maintain several operating linkages with other more peripheral units. The peripheral units can experiment with technologies and markets, and then pass ideas to the central units through their operating linkages. Managers of multi-business firms can draw on implications of the results to help shape their organizational designs to enhance innovativeness.

Q: I usually set one stage for a new idea chain, and then if we make one design for a new commodity, this kind of design is to be materialized for the mass production. So we have two levels of chains in one company. In Formosa, those two chains are interrelated. So what do you think about these two types of chain in your study?

A: I have not differentiated between types of innovation. Scouts are very important when we are trying to get ideas which are really creative, out-of-the-box thinking; on the other hand, when you have more incremental types, maybe what you need is more recombination of existing ideas. My feeling is that these scouts, this structure network is more important for innovation which is truly innovative.

Q: In Japan, information on what consumers want recently is very important; people think you need many business ties or money to buy that. In your idea, how do these scouts afford to get information? Do they have huge financial resources?

A: Because they are peripheral they have no choice other than to look beyond. Often, I think about leaders of countries which are in empires; often the leaders come from the edge, because they have no ties. They are the ones that have to come up with something new and push through the ideas. If I am a central firm I am already very comfortable, and I do not have time, because all my energy and resources are spent on maintaining the ties I have. But you need some ties, so that whatever the scouts get, you have an efficient way for that to be transferred to the integrator.

Q: You mentioned that the Taiwanese business groups are less hierarchically organized than the Korean chaebol, but more coordinated than the Japanese keiretsu. Is there a trend? You give figures and from there I have the feeling there is no big trend; the equity tie density is more or less constant, directorate ties density constant, operating ties a little down, but static. Is there anything changing in the structure of Japanese and Taiwanese business? I have the impression that in Japanese keiretsu the links are getting weaker; the networks are in place, you can count on each other but the group members are less willing to do it if it is against their own interest. Is the Taiwanese situation static or is there also a trend somewhere?

A: The changes are not so drastic, but the operating ties are becoming less permanent. In Korea there is a lot of effort to make them smaller, but in Taiwan some of these business groups are getting bigger; they have informal ways to make the control work and they are realizing that director ties and equity ties work quite well. So they are in a way like Japanese keiretsu. The difference: Japanese keiretsu are not family-owned, but in Taiwan family control is very important. They are moving towards the chaebol in some sense. As direct buyer-supplier ties are going down, the others seem to be going up.

C: Evaluation: great study, rigorous study, and one of the first studies to deal with the very complex issues in a single study. It is really a comprehensive study that sheds light on the underlying conditions of particular types and natures of networks, and how density and centrality affect innovation in a differentiated way. I think this paper could be much more applicable and refined if you can provide further elaborations and interpretations of the empirical results. On the one hand, family ownership prevents the investors from getting into a risk investment, but if the owner agrees, it encourages bold decision-making; Samsung and old Korean chaebol are famous for bold decision-making, relative to the Japanese companies. Further development possibilities: this study could go beyond single business group studies, especially for the affiliate companies. Secondly, it might be interesting to go into more organizational dimensions like quality of management and leadership issues. This paper is mostly focused on multi-business context, but it could go beyond this to include multinationals. This framework is really relevant when talking about why certain subsidiaries within multinational companies are more innovative. I think different types of innovations can be considered; different types of network ties influence innovation, so this is the main focus of this study. It would be even more interesting if the author could include different types of innovations; quantitative and qualitative measures. Lastly, it was on Taiwan, but if you could include comparative studies to include other business groups in other countries, that would be very interesting.

C: Innovation includes many things, not only patents but also new materials, investments, and solutions. Your analysis should include the new thinking or new ideas.

A: It is not because I like patents; I just do not really have a good alternative for research for large sample properties. If I wanted to get the kind of data for different types of innovation, I would have to do surveys for each company. What I gained is in generalization, but I lose in richness. That is the tradeoff that I have chosen.

These firms have internal versus external linkages, so it is definitely important to control for this. At the group level, I find that business operating density is bad, but the negative effect goes away when you have lots of external joint ventures. If you have lots of outside ties, that may neutralize some of the negative effects. The joint ventures are like scouts.

There are other limitations. I am using the count of ties. So what passes through these ties? How do I know, other than a detailed study? It is almost impossible. Almost every single network study that I know is a single network study. Here, because each group is a network, we can gain in comparison, but we lose in richness.