Monday, March 9, 2015

Historical case study: Japanese banks in the 1990s

Julian Robertson is one of the investors whom John Train profiled in Money Masters of Our Time. Robertson's hedge fund, Tiger Management, was short Japanese banks during the 1990s, and the appendix of Money Masters includes an internal memorandum that a Tiger analyst wrote describing the firm's rationale for shorting them.


Japanese vs. American banks

The analyst, Tim Schilt, wrote in 1995 that "Japanese banks have the largest market capitalizations and lowest measures of profitability of any banks in the world" and illustrated this by comparing Citicorp to Mitsubishi, which he called "Japan's best bank."

While the two banks had approximately the same equity, Mitsubishi's balance sheet was 87% larger. The quality of its assets didn't match their quantity, however: Citicorp's return on assets was 4.2%, while the "best bank" managed only a 1.1% RoA. In light of Citicorp’s better returns, one might expect it to have garnered a higher valuation, but the opposite was true: Citi traded at 5x pre-tax earnings versus 27x for Mitsubishi.

While stark, this disparity wasn’t new. Grant's Interest Rate Observer described a similar contrast between Japanese and American banks in 1987:

U.S. Trust Co. is the Rolex watch of the American trust business. Last year it earned 18.3 percent on its equity. Yasude Trust & Banking, a kind of Japanese Timex, earned only 13.5 percent on its equity.

However, while U.S. Trust fetches only 10 times last year's earnings on the New York Stock Exchange, Yasuda commands 133 times earnings on the Tokyo and Osaka exchanges...

[Yasuda] has accumulated a book of Third World loans equivalent to 100 percent of its estimated equity, and it owns a book of "poor or non-performing" loans, both domestic and foreign, equivalent to four times its estimated equity.


Loan losses

Officially, Japan's 21 largest banks had non-performing loans equal to 3.6% of total loans. But as Schilt wrote, the actual level of losses was much higher:

Excluded from the NPL category are loans to the housing loan companies where at least Y5 trillion is in serious arrears. Also excluded are the restructrued loans, where companies are being supported by reduced interest rate and principal payment terms.

Finally, loans that have been sold to the Co-operative Credit Purchasing Company (CCPC), the major avenue by which banks are taking bad-debt-related charge-offs, are no longer included in NPL's, a questionable practice, in my opinion. Since its inception in March of 1993 through July of this year, the banks have sold Y8.8 trillion face amount in loans to the CCPC for Y3.9 trillion, a 56% aggregate charge-off level. The banks provide the financing for the CCPC to purchase the loans from them, and then, believe it or not, start accruing interest and paying taxes on this interest as if the loans are current at the contractual rate.

The reasons why I question the validity of removing loans sold to the CCPC from the NPL category are twofold. Firstly, the ultimate charge-off to the bank is determined when the loan (in virtually all cases a real estate property) is actually sold by the CCPC, and secondly, the CCPC has disposed of only Y204 billion (5.2%) of the properties to date, a telling reflection that bid and asked prices are still way apart...

While reported NPL's declined by Y1 trillion last year and the loan loss reserve increased from Y4.5 trillion to Y5.5 trillion, total charge-off and reserving costs of Y4.9 trillion indicate that loans are still going bad at an alarming rate. [One Japanese bank analyst] estimates that the twenty-one major banks have Y45 trillion in NPL's, a staggering 12.9% of their loan book. My estimate of Y30.5 trillion (8.8% of loans) is more toward the lower end. Either figure is huge when measured against total shareholders' equity of Y21.4 trillion, loan loss reserves of Y5.5 trillion, and unrealized securities gains of Y9.0 trillion, a total of Y35.9 trillion.

The unrealized securities gains were less of a cushion against losses than one might have expected, since they were pro-cyclical: Japanese stock prices fell during the 1990s as bad loans proliferated.

While the CCPC let Japanese banks understate their loan losses, some banks used even more questionable methods to hide bad loans:

In Japan, companies were only required to reveal what was happening in their subsidiaries if they owned a significant stake in them... As the problems mounted up, LTCB started to take advantage of this principle. In December 1991, the bankers created a company called 'NR,' which was designed to act as a warehouse for some of the nonperforming loans... the bank started to 'sell' its assets to NR at book value...

This did not, of course, really improve the problem in the long term. Companies like NR were only able to purchase these risky loans because LTCB itself lent it the cash. Thus LTCB had simply replaced one set of bad loans with another. However, NR was such a tiny, obscure group that nobody outside the bank noticed the scheme. Better still, the bank was allowed to classify its loans to NR as 'healthy,' since NR was considered to have the support of LTCB.


Normalized earnings

Schilt estimated that Japanese banks would need to spend at least six years repairing their balance sheets before earnings normalized, which was twice as much time as most bank analysts expected.

But even assuming that losses normalized at lower levels, the banks weren't particularly cheap. With annual losses equal to .20% of risk assets, the 21 largest banks would trade between 27x and 140x earnings. At a .40% rate, several banks would be unprofitable and the rest would trade between 37x and 261x earnings.

Schilt argued that an emphasis on real-estate loans to the exclusion of higher-interest consumer loans kept Japanese banks' profit margins and returns lower than those of foreign banks.


Bank stocks defy gravity... for a while

From 1990-97, Japanese bank stocks significantly outperformed the Japanese stock market. Accordingly, Schilt wrote that "The Japanese banks have been a frustrating investment experience for Tiger."

Below is a chart comparing Mitsubishi Bank (gold line) with Japan's TOPIX index (black line) from 1990-2002:


By 1994, Mitsubishi had outperformed the broad Japanese market by 50% and exceeded its price at the end of 1989, when the Japanese market peaked.

It wasn't until 1997, two years after Schilt's memorandum, that Japanese bank stocks collapsed. Their collapse coincided with several things:

• A tax hike in April 1997 that led to a recession.

• The failure of three Japanese financial companies in November 1997: the fourth- and seventh-largest brokers and the tenth-largest bank. Their collapse led to fears that Japan's Ministry of Finance had withdrawn its longstanding support for loss-plagued financial institutions.

• The Asian Financial Crisis, which generated a new round of loan losses for Japanese banks.

• The peak of Japan's working-age population and the peak of Japan's CPI.

I'm not sure how much the declines in Japan's workforce and price level contributed to the banks' woes. Each experienced a modest initial decline that didn't accelerate until 2000.

On the eve of the Asian Financial Crisis, Japanese banks accounted for one-third of the international loans made to Indonesia, Malaysia, South Korea, and Thailand. One academic paper argues that Japanese banks were instigators as well as casualties of the Crisis:

Japanese banks were the critical actors who triggered the devaluation inadvertently when they reduced their exposure to Asia due to the need to avoid losses and to protect their capital base in advance of adoption of the BIS-mandated capital adequacy requirement in Japan.

The Japanese banks' losses from the Crisis were small relative to their domestic loan losses, however. The paper suggests that Schilt's predictions regarding credit quality were accurate:

The rise of domestic problem loans in Japan to Y29.2 trillion ($232 billion) was threatening the stability of the financial system there as well. These problem loans would eventually be re-valued upward based on a new definition of 'problem loans' to Y76.7 trillion ($610 billion), representing 12 percent of total Japanese loans outstanding and equal to twice the size of Australia's economy.

While Japanese bank stocks held up much longer than bears had expected them to, bad loans were so pervasive that it was probably inevitable they would fall.

No comments:

Post a Comment