The fact that our country's transactions are difficult to understand from the outside has been brought to light by the debate over keiretsu relationships. The word keiretsu, a broad concept in itself that cannot be grasped easily, is interpreted in various ways and has generated a great deal of debate, often involving political motives. Recently, Mr. Robert Lawrence of The Brookings Institute put forth an argument that Japanese corporations' keiretsu relationships are barriers to foreign imports. He argues that among keiretsu relations, the horizontal keiretsu -- the former zaibatsu -- are the biggest problem. 1 

Mr. Lawrence's statistical approach has been complemented by certain specialists as being clear and concise. The US media triumphantly stated that it is "the strongest proof yet of the US assertion that Japanese keiretsu represent a barrier to imports," and "it is enough evidence to make the Japanese take the defensive."2 Unfortunately, however, for ordinary people who are not used to the terminology of statistical analysis, his logic is not all that clear. It is, rather, extremely difficult to understand. Consequently, the reader is led only to his conclusion -- that the horizontal keiretsu formation is the biggest problem. Moreover, the report Lawrence made at the Subcommittee on International Trade of the US Congress included only his conclusions.3 

A particularly interesting point in Lawrence's debate is his assertion that although the vertical keiretsu structure does restrain imports, it is nonetheless practical to a certain degree. However, when discussing horizontal keiretsu by saying that it not only reduces imports but is also economically impractical, he develops the "logic of the liberator" conclusions unique to the SII, which leads one to conclude that horizontal keiretsu are the problem.4 

It is true that in general keiretsu structures present a closed image, and that they assume a rather fuzzy impression in this post-modern world. Thus, from now on we must increase the transparency of Japanese business practices so that foreign firms can understand them easily. However, we disagree with his conclusion that singles out horizontal keiretsu as the true evil. 

Lawrence's paper is unique in that it is an overall judgement based on a quantitative analysis of objective data. This means that it will not be effective to argue each of his theories with individual facts and data. This paper is set up to arrange Lawrence's logical arguments in a manner that makes them easy to understand and examine, while introducing comments about them given by Prof. Saxonhouse from University of Michigan.

In sum, I would say that Lawrence's analysis, showing that horizontal keiretsu is the problem, is not well-grounded. There are some problems in his methodology, and even if we did accept all the results of his analysis (the correlation between the keiretsu variable and both import penetration and the share of world exports, etc.) it is difficult to tie these results to the conclusion that horizontal keiretsu structures lack economic practicality.

II. Lawrence's development of logic

The summary of Lawrence's paper was also printed in the September issue of "Fair Trade(Kousei-Torihiki)."5 As I mentioned before, this paper's objective is to examine the logic he uses. There will be some overlap, but I would like to arrange it simply as the following.

1 Positioning of the various views on keiretsu

Lawrence distinguishes the positions of academics who have spoken so far on keiretsu into three groups, and attempts to objectively clarify which one is the correct position. 

The first is the benign neglect view, which claims that keiretsu relationships do not have important effects on Japanese economic performance. Thus, they assert that it is meaningless to discuss keiretsu at SII talks.

The second is the trust-busting view, which maintains that keiretsu structures operate as entry barriers to newcomers. This position also states that keiretsu work against Japanese consumers because they are economically inefficient. Proponents of this position find that it is necessary to focus on keiretsu at the SII talks in order to remove the trade imbalance between the US and Japan. 

The third view is the dilemma position, which concedes that the relationships have a negative impact on the export of foreign products to Japan, but also states that the keiretsu structure itself is practical and has increased efficiency in Japanese industry, and have enhanced Japan's overall industrial competitiveness. As a result, they say that it has become harder for foreign firms to export to Japan, although they are not disadvantageous to Japanese consumers. In this case, the Japanese are pressed to make a decision between efficiency and openness towards the outside. The options left to the US for resolving the trade imbalance would be either managed-trade (Clyde Prestowitz's opinion), to create a keiretsu of their own, or to use Japanese keiretsu to enter the Japanese market.

2 Which view is correct?

Determining which of these views is correct is an important judgement. However, keeping in mind that the views were not based on a fair analysis of objective data, Lawrence starts an analytical approach as follows.

First, in order to ascertain whether the benign neglect position is correct he investigates whether keiretsu forms hinder imports. He divides Japanese industry into 37 different industrial groups and measures the shares in sales for each according to the eight horizontal keiretsu groups, such as the former zaibatsu, and the nine vertical keiretsu groups such as Nippon Steel, Hitachi and Toyota etc. He uses these shares as a measurement for the degree of horizontal and vertical keiretsu orientation they have. Then he calculates the import penetration of that industry's product (the percentage that the imported product accounts for in domestic demand) and examines the statistical relationship between this number and the level of horizontal and vertical orientation.
 
 

If he could not determine a relationship between the two variables, the benign neglect view would be the correct one. However, Lawrence discovered a negative correlation between import penetration and the level of horizontal and vertical keiretsu orientation (he was able to observe that when the degree of vertical and horizontal keiretsu increased, import penetration decreased). Thus, he came to the conclusion that the benign-neglect view does not hold.

However, the question of why the amount of imports is small is left without clarification. Does it decrease because of closed transactions, or because keiretsu structures raise the industrial competitiveness of these keiretsu-related companies? Determining whether the trust-busting or the dilemma position is right may be determined according to this answer.

Next, it is necessary to analyze the correlation between the degree of keiretsu formation and the degree of industrial competitiveness, in order to determine whether keiretsu formation does in fact raise industrial competitiveness. The measurement for an industry's industrial competitiveness will be the shares of their products in the export market. (If it has strong industrial competitiveness, its shares in the export market should rise.) Examining the relationship between the degree of horizontal and vertical keiretsu formation and the shares in the export market showed that vertical keiretsu enhanced exports, but that horizontal keiretsu had no effect on exports.
 
 

These results show that vertical keiretsu formation impedes imports but at the same time is economically practical in the sense that it improves the international competitiveness of keiretsu corporations (the dilemma position). As for horizontal keiretsu formation, in addition to its negative impact on imports, it does not enhance the competitiveness of the industry concerned. Thus the conclusion is that internalized transactions of keiretsu is the factor that reduces imports.

Chart 1 is a relatively clear flow chart of Lawrence's logic. You can see that whatever data you input it will be sorted into one of the three positions.

III. Questions that arise with Lawrence's paper

While mixing in comments of people from within the firm and outside, I would like to put together the questions that arise concerning Lawrence's paper, as well as some points that did not convince me. 

1 Correlation and causality

This is a point that everyone comes up with immediately: is it correct to claim that causality exists between two numbers just because there is a correlation? It was easy to predict that if we conducted a regression analysis, a correlation would exist between the Japanese import-export performance and keiretsu because they are both unique characteristics to Japan. However, causality is a different matter.

For example, even if there is a negative correlation between the degree of keiretsu formation and import penetration, it is not clear whether the size of imports is small because of keiretsu or whether keiretsu are formed because of limited imports. It is not clear which is the cause and which is the effect. Naturally, if a product is mass-produced domestically the amount of importation should not be that large. As long as domestic mass-production is taking place it can be assumed that a long-term keiretsu relationship is formed already. In this case, a statistical correlation between keiretsu formation and low import penetration should be found, but this correlation is not enough to say that keiretsu formation impedes imports. 

2 Technical problems

Gary Saxonhouse -- a specialist on the subject -- has pointed out a technical problem. It is a criticism towards Lawrence's analysis using 1985's cross section data of each industry.6 The cross section analysis can explain the relative difference between the level of keiretsu formation in each industry and its import-export performance, but it cannot explain the relationship between the degree of keiretsu formation and the import-export performance of Japanese industry as a whole. The results of Lawrence's analysis simply show that manufactured goods of industries that have relatively low degrees of keiretsu formation have larger imports than products of other industries. This does not imply that if Japan's overall degree of keiretsu formation diminishes Japanese imports will rise altogether.

Further technical problems have been pointed out. Lawrence conducted a multiple regression analysis for the import-export performance of each industry with ten variables which include raw material intensity, capital intensity and technological intensity. However, a correlation already seems to exist within some of these variables. Therefore, there is some doubt about the precision of this overall regression analysis (multiple collinearity).7
 

Questions also arise concerning Lawrence's analysis based on import-export data of 1985. 1985 was the year in which a surge in the appreciation of the yen resulted temporarily in a huge restraint on exports from Japan.

Moreover, the most important technical problem seems to be what criteria Lawrence used to define horizontal and vertical keiretsu. Referring to the 17 corporate groups of the keiretsu almanac 1986-87 Industrial Groupings in Japan8 (from Dodwell Marketing Consultants, a foreign based consulting firm in Japan), Lawrence defined the eight bank-related groups as horizontal keiretsu and the other nine as vertical keiretsu. However, this selection process is extremely subjective.

For instance, when Dodwell published the 1990-91 version of the same almanac the number of keiretsu groups were expanded to 39. The horizontal keiretsu listed are the same as the 1986-87 version, while the vertical keiretsu not only increased in number but also changed significantly in terms of content. Tokyu, Seibu Railway and Seibu Saison, which were originally considered vertical keiretsu, were removed from the list. Meanwhile Sumitomo Metal Industries, Mitsubishi Heavy Industries, Kirin, NEC, Mitsubishi Electric and Mitsubishi Motors, which were categorized previously as horizontal keiretsu, were listed as vertical in the 1990-91 version.

There has never been a definite standard for defining horizontal or vertical keiretsu, so it is understandable that the grouping would become subjective. However, the point here is that even when a detailed statistical analysis is conducted, if the classification is subjective the results will inevitably become very rough figures. 

(The fact that the grouping of keiretsu is so fluid might represent the fact that the Japanese keiretsu relationships are not as fixed as the U.S. thinks, but in fact much more mobile.)

3 The evaluation process of industrial efficiency

Lawrence determines that even if keiretsu formation has a negative impact on imports, as long as it improves industrial competitiveness (efficiency) it is practical. I have already pointed out that in evaluating whether there was an improvement in industrial competitiveness he uses the industry export market shares. Are there really no problems in this method of measurement?

What I am trying to say is that among the factors that constitute a product's competitiveness, there must be something that is not price or quality but is unique to the Japanese market (one that is therefore impossible to export). Thus, the delicate delivery management services which corresponds to the production system called kanban plays an extremely important part for a good's competitiveness. Even if the close human relationships between the supplier and the dealer does not effect the price in the short-term, it is nonetheless an important advantage in the long-term because we can expect that it will pull down transaction costs. This is an important advantage. However, these types of services are hard to export. What we can export is the product "hardware" that limit this "software" to a minimum.

Incidentally, vertical keiretsu are normally the ones that fit the mass-production system. It can also be said that they are the ones that mass-produce the hardware at a low cost and sell on the global market. This is a simple and concise method that would be considered to be competitive according to Lawrence's logic. However, the software (the reduced transaction costs that arise from long-term transaction relations, or services such as information that the suppliers provide) is often viewed as important as the product itself in the transactions of traditional products, including steel or heavy chemical goods which are produced by corporations with a relatively high degree of horizontal keiretsu formation. These merits are appreciated in domestic transactions where the seller and the buyer are geographically close, but not in international transactions where the distance between the two is much greater.

Moreover, it is probably the case that these basic products tend to have a long history of being exported, and many are subjected to voluntary export restraint. Therefore, regardless of whether these products are competitive or not most of the time exports are kept at a low level. Thus, I think it is questionable to decide that an industry lacks competitiveness merely because its export levels are small.

4 What about evaluation in the long-run?

Even if we assume that keiretsu formation limits competition in the short-term it does not necessarily mean that the rule applies over the long-term. There was once a time when underrating the size of corporations -- caused by vertically classifying each horizontal keiretsu by each industry -- became a problem. This criticism stated that inefficient corporations were surviving because of keiretsu. However, if we look at this situation in the long-run, it is obvious that it was beneficial. The survival of small corporations thanks to keiretsu promoted competition in the long-run and strengthened the corporate structure. This argument has existed for a long time,9 but it is a historical fact that over the long-term horizontal keiretsu enhanced competitiveness within Japanese industries.

5 Would imports increase if keiretsu disappeared?

Granted that keiretsu are reducing imports, there is always the question left of whether imports will increase if we somehow get rid of keiretsu. Lawrence says that there is a possibility that the current 23 billion dollars worth of imports of the industries under survey would rise to 58 billion, which is more than twice as much. This argument is a very misleading. As I have mentioned before the results of industrial cross section analysis cannot be applied to one country's import-export performance. Moreover, a country's current-account balance is determined by the balance of savings and investments. Assuming this does not change, no matter how much micro-level control we impose (for instance, even if we add regulations to the industrial structure and break up the keiretsu), the current balance may shrink temporarily but with other parameters fluctuating (such as foreign exchange moving towards a weak yen) it would most likely return to the original level.10 I do not mean to say that there is no meaning in structural reform, but I think there is a problem in concluding "if keiretsu forms disappear imports will rise several billion dollars..." as the media already has. Such statements have given the statistics and numbers a life of their own.

6 The relationship with the distribution market

Supposing that the keiretsu form is inefficient and that it has been operating to limit imports, if the distribution market for final goods is completely open then inefficient keiretsu corporations could not survive competition with imported products. Thus, even if the keiretsu problem does exist, as long as the problem of distribution is settled it will naturally resolve itself. Chart 2 shows this relationship. The problem of imports not being able to enter should be resolved by reforming the distribution system. There is some doubt that the US is concentrating on the right subject by focusing on the keiretsu debate. 

The reason SII talks have steadily progressed is that within Japan there were many who agreed with the points brought up by the American side. However, there are obviously less Japanese citizens who view keiretsu as an issue compared to other topics. Graph1 is a survey conducted by the Nikkei Industrial Research Institute of 1000 executives all over Japan about the topics of SII. The results show that there is an obvious difference in the way they perceive the keiretsu issue compared to distribution or property issues. Ordinary Japanese people are not convinced that bringing the keiretsu issue to the negotiating table is useful. 

Graph 1 Survey on 1000 businessmen

(Nikkei Industrial consumer Research Institute June,1991)

Opinions on measures towards Japan(%)
1 2 3 4 5
 

Savings & investment pattern 47.8 14.4 2.3 25.1 9.3

Property policy 79.9 4.3 3.3 10.4 2.3

Distribution 71.1 6.1 7.6 11.6 2.9

Exclusionary transactions 62.6 10.3 3.9 16.9 6.0

Keiretsu formation 37.1 20.8 4.4 28.1 9.4

Price mechanism 72.9 5.1 4.0 12.6 5.1

(note) excluded unanswered

1: Should definitely be talked about

2: Slightly unreasonable to Japan

3: Permissive to Japan

4: Cannot judge at this stage

5: Do not know

Amidst the existing values that say it is good to produce things while cooperating and getting along with one another, I do not think that we can gain constructive results by bringing up keiretsu -- which is the essence of Japanese corporate society -- and viewing it as a problem. I have already stated that SII is based on the "logic of the liberator." Working through this logic naturally requires the existence of "the people to be liberated." However, in the keiretsu debate, these crucial people who are to be emancipated do not appear. 

IV. Conclusion

In conclusion, Lawrence's argument that the problem lies in "keiretsu not being efficient..." is not a well-grounded one. However, I do not want to blindly assert the efficiency of keiretsu. If one supposes that keiretsu formation is efficient in all situations, among the three positions that I have mentioned the dilemma position becomes dominant. This is because in actuality there are many people who think that keiretsu is a problem for US industry because it is efficient and strong.11 I have already stated that there is a possibility of the dilemma view leading to the opinion of managed-trade.
 

On October 6 and 7 The Washington Post ran a special series of related articles on Japanese keiretsu.12 While publishing statements from the top official of the Department of Commerce saying that keiretsu is a bedrock, the Post also objectively reported the complexity of the issue, stressing that US government is troubled over how to deal with it. After all, the reality of economic society must not be something that can be simplified into a model like Lawrence's logic, clearly defined in black and white terms. It will be desirable to have continuous talks based on individual issues rather than conceptual arguments. 

Chart 1 Lawrence's logic

Start-->Do keiretsu restrain imports?-NO-->Benign neglect
I
 

YES

I
Do keiretsu promote exports?-YES-->Dilemma position
 

I

NO

I

Trust busting

Chart 2 As long as the problem of distribution is resolved..
International exporters
 
 

Japanese intermediate manufacturers-->Japanese final manufacturers-->Japanese distribution market (distribution keiretsu)-->Japanese consumers