Japan opened its corporate bond market in 1983, and the firms the main-bank literature was written about were the first to leave

A Japanese firm's main bank was its largest lender, usually a shareholder as well, often the supplier of a director or two, and by reputation the party that would organise a rescue if it got into difficulty. For two decades this was the standard example, in textbooks and in advice to developing countries, of a relationship doing cheaply what a capital market does expensively: monitoring a borrower, and rescuing it when rescue beats liquidation. The scholarship on it is unusually contradictory, and the contradiction is more useful than any of its verdicts.

The first camp found the relationship valuable to the firm. In 1991 Takeo Hoshi, Anil Kashyap and David Scharfstein compared Japanese firms with close main-bank ties against firms with weaker ties, and found investment more sensitive to the firm's own liquidity in the second group.¹ The measure is doing real work. A firm whose investment rises and falls with the cash it happens to hold cannot persuade an outsider to fund a project on its merits; it builds when flush and stops when not, whatever is worth building. A firm insensitive to its own cash flow has a lender willing to look at the project instead. On that reading the main bank was solving an information problem, and firms inside the relationship had cheaper capital than their prospects alone would have bought.

The second camp found the relationship valuable to the bank. In 1998 David Weinstein and Yishay Yafeh examined Japanese firms over 1977 to 1986 and reported that those with main-bank ties paid more for capital, were less profitable, and grew no faster.²³ The benefits, they concluded, were largely appropriated by the banks, which also steered clients towards cautious projects.²³ The third camp denied there was much there to argue about. In 2002 Yoshiro Miwa and J. Mark Ramseyer argued that Japanese main banks neither agreed in advance to rescue troubled debtors nor monitored firms on behalf of other creditors, and that membership of a corporate group was not a usable proxy for a main-bank tie, intra-group cross-shareholdings having been trivial even when the ties were supposedly strongest.⁴ Their critics reply, fairly, that showing a description wrong is not showing the thing never existed, and that main banks did rescue some large debtors.⁵

The first two camps are usually presented as disagreeing about whether the main bank was good for the firm. They are not really in conflict. A relationship can relax a borrowing constraint and be priced at the same time: a lender who lowers the cost of a project and then charges for it has both delivered a benefit and kept it. Whether the firm came out ahead depends on how the gain was divided, which is set by the parties' alternatives. So the question separating the camps is regulatory, not cultural. What else could a Japanese firm have done?

For most of the post-war period, very little, and the reason was a rule rather than a preference. Japanese bond issues operated under the collateral principle, which required security behind every bond. Unsecured straight and convertible issues became legally possible in 1979, but the Bond Issue Criteria were so strict that only two companies in the country qualified, Toyota and Matsushita Electric.⁶ The criteria were then relaxed by stages through the 1980s, until a firm carrying a rating could sidestep them altogether and, in 1996, the rules on bond issues were lifted entirely.⁷ Gower's count puts eligibility by the late 1980s at about 300 firms for straight bonds and 500 for convertibles, against the two of 1979.⁶

What happened when the gate opened is the test, and a clean one, because the firms said nothing. The largest and most creditworthy borrowers left first. Hoshi and Kashyap argue that the banks lost their best clients quickly, and that compressing liberalisation into roughly the six years from 1983 to 1989 all but guaranteed the banks a shock.⁷ Had the relationship been worth more to the firm than it cost, the firms with the best outside options would have been least eager to use them. They were the first to go, which for the firms that had a choice points to Weinstein and Yafeh.

It also identifies who the first camp was describing. The firms that stayed were those that could not meet the criteria, mainly smaller firms whose only collateral was land, and a bank with a captive borrower has market power whether or not any group apparatus exists.⁶ That is the population for whom the relationship genuinely relaxed a constraint, and the one the banks were left holding. Kang and Stulz found that firms with a larger share of bank loans in their 1989 debt performed worse and invested less over 1990 to 1993, when the banks were themselves in trouble.⁸ One institution loosened a constraint in one decade and transmitted a shock in the next, with no change in it and a large change in the borrower's alternatives.

Miwa and Ramseyer's challenge lands on a different claim and survives intact. Their target is commitment: the implicit promise to rescue, and the monitoring said to be done for other creditors. Nothing above needs either. A bank can price market power over a borrower with nowhere else to go without promising anything, and a captive borrower needs no cross-shareholding to be captive. So the described institution can be largely a fable while the pricing finding holds, because bargaining power needs no culture. What does not survive is the claim the literature was most often exported for, that a relationship-based system is a durable substitute for a capital market. It was a substitute for the capital market Japan had forbidden.

The strongest objection is that the exit evidence is confounded. The firms that left first were also the fastest growing, and would have outgrown bank finance anyway, so their departure may reveal scale rather than dissatisfaction. A second finding cuts at the framework: Tsuruta reports that Japanese firms whose distressed main banks failed switched banks, kept access to credit, and performed better afterwards.⁹ The confound is real, and it is why the argument rests on the ordering rather than a magnitude: the claim is not that bonds were cheaper by some amount but that the firms with choices exercised them first. Tsuruta's result strengthens the reading, since a relationship replaceable at no cost was not worth much to the firm, which is the second camp's conclusion reached from the other end.

The lesson for anyone holding up such a system as a model is not that relationships are worthless. The measured value of a relationship to a borrower is a statement about the borrower's alternatives, and alternatives are usually set by regulation rather than custom. A study finding relationship lending beneficial where firms may not issue bonds has not shown that relationships beat markets; it has shown that something beats nothing. The falsifying test sits in the issuance records: if the earliest Japanese bond issuers after each relaxation were the firms with the weakest prior bank ties rather than the strongest, the reading offered here fails.

Notes

1Takeo Hoshi, Anil Kashyap and David Scharfstein, "Corporate Structure, Liquidity, and Investment: Evidence from Japanese Industrial Groups," Quarterly Journal of Economics 106, no. 1 (1991), 33-60, for the finding that investment was more sensitive to liquidity among firms with weaker bank ties. RePEc record; earlier working paper at MIT

2David E. Weinstein and Yishay Yafeh, "On the Costs of a Bank-Centered Financial System: Evidence from the Changing Main Bank Relations in Japan," Journal of Finance 53, no. 2 (1998), 635-672. Semantic Scholar record; Hebrew University institutional record

3Research Institute of Economy, Trade and Industry discussion paper 01-E-004, "Bank Relationships and Firm Performance," summarising Weinstein and Yafeh's three findings on interest payments, profitability and growth, and giving their sample period as 1977 to 1986. The Journal of Finance text could not be opened from this environment, so no magnitudes from it are quoted here. RIETI discussion paper

4Yoshiro Miwa and J. Mark Ramseyer, "The Myth of the Main Bank: Japan and Comparative Corporate Governance," Law and Social Inquiry 27 (2002), 401-424, and the same authors' The Fable of the Keiretsu (University of Chicago Press, 2006), for the claims that main banks neither promised rescues in advance nor monitored for other creditors, and that group membership does not proxy for a main-bank tie. Working paper version, University of Tokyo CIRJE; SSRN record; book record

5Dan W. Puchniak, "A Skeptic's Guide to Miwa and Ramseyer's The Fable of the Keiretsu," for the argument that evidence of a flawed description is not proof of non-existence, and that main banks did rescue some failing firms. Author's copy

6Luke Gower, "Some Structural Causes of Japan's Banking Problems," Reserve Bank of Australia Research Discussion Paper 2000-03, for the collateral principle, the two qualifying companies in 1979, the attraction of foreign markets, the rise in eligibility to 300 firms for straight bonds and 500 for convertibles, and banks' remaining market power over firms that could not meet the criteria; with Hesna Genay, "Japan's Corporate Groups," Federal Reserve Bank of Chicago Economic Perspectives, January 1991, on the 1983 relaxation. RBA discussion paper; Chicago Fed article

7Takeo Hoshi and Anil Kashyap, "The Japanese Banking Crisis: Where Did It Come From and How Will It End?," NBER Macroeconomics Annual 14 (1999), pp. 129-212, for unsecured straight and convertible bonds becoming permissible in 1979 under issue criteria so stringent that only Toyota Auto and Matsushita Electric qualified, for the gradual relaxation of those criteria through the 1980s, for rated firms being able to avoid the issuance criteria by the late 1980s and all rules on bond issues being lifted in 1996, and for the authors' statement that part of the reason banks in Japan got into trouble is that they lost many of their best borrowers in a very short period of time, between 1983 and 1989. Read in both the chapter and the working paper. NBER chapter; NBER working paper 7250

8Jun-Koo Kang and Rene M. Stulz, "Do Banking Shocks Affect Borrowing Firm Performance? An Analysis of the Japanese Experience," Journal of Business 73, no. 1 (2000), 1-23, for the finding that firms with a larger share of bank loans in their 1989 debt performed worse and invested less over 1990 to 1993. RePEc record

9Daisuke Tsuruta, "Changing Banking Relationships and Client-Firm Performance: Evidence from Japan for the 1990s," Review of Financial Economics (2014), for the finding that firms whose distressed main banks failed switched banks, kept access to credit, and performed better afterwards. Publisher's page; working paper version