International OR
Bubbles, Bad Bets & Bankruptcy
"Gaijin" professor's models predicted market crash but advice discarded as Japan's fourth largest security firm goes under
By William T. Ziemba
The phone rang. Would I be interested in going to Japan as the first Yamaichi Visiting Professor of Finance at the University of Tsukuba some 60 kilometers northeast of Tokyo in the city of the Japan's science center?
It was 1988 and two decades of Japanese export marketing expertise had made Japan the world's largest creditor nation. They exported cars, stereos, TVs, cameras, printers and the like and imported collateral IOUs which they reinvested primarily in their own land and stocks. These financial assets were at extraordinary levels and the Japanese stock market accounted for 44 percent of the world's equity.
At the time, my hot academic topics included stock market anomalies and financial market crashes. I was invited to go to Japan and make a proposal to the sponsors, the Yamaichi Research Institute. They accepted my plan, which was to teach investments at the University Monday through Wednesday and then go to Tokyo on Thursdays and Fridays as a consultant to YRI. It was a full load of teaching, but I was able to teach courses I was interested in.
The lectures in investments, futures, options and anomalies set the stage for the Thursday and Friday research. My group of students who trained or bussed up to Tsukuba were in my Anomalies and Crash study groups later in the week. I would lecture on the U.S. evidence and they would work with me on new research. Finance was not taught in Japanese universities; stockbrokers were shunned in society and rated somewhat below metermaids.
We all worked hard. There was a reason why the Japanese economy was cooking. A high yen caused the inflation rate to be negative for wholesale prices and the resulting low nominal and real interest rates fueled the boom.
My family was treated with enormous respect. My wife, Sandra Schwartz, was teaching economics at the University and our daughter, a nine-year old redhead in a sea of black hair, went right into Japanese public school in Japan — studying the BC correspondence curriculum on her own and joining the Japanese lessons as she improved her Japanese. Rachel became quite fluent in the language through after-school play with others in her classes set up by the most friendly and helpful mothers.
At YRI, Mr. Okada, the vice chair, took a great interest in my work. We talked endlessly about financial markets. Borrowing from my racetrack anomaly and portfolio theory research, I quickly came up with a crash model. Shigeru Iishi, a Yale MBA in Finance, helped me with the calculations. The idea was simple: let stocks and bonds compete for the money. When interest rates were low, stocks would get the money and would rise. When interest rates rose, investors would move out of stocks and into bonds. If this occurred quickly there could be a crash.
The model predicted the 1987 U.S. and worldwide crash perfectly. (See sidebar, "When to Pull the Trigger" and Table 2 taken from "Invest Japan" which Sandra and I wrote for Probus in 1991.) I asked Iishi to use the same model on the Japanese market, looking at the variable D equal to the difference between the long bond rate and the earnings yield on stocks (the reciprocal of the trailing PE ratio).
We defined a crash to be a drop of 10 percent or more in a year-long period. From 1948 to 1988, there were 20 such crashes in the Japanese market. That was a lot for a stock market that was up 553 times in dollars and 221 times in yen. The model did not predict all the crashes; no model will. But we found that whenever the model was in the danger zone — that is, so many standard deviations above normal times (the past) — there eventually would be a crash with no misses.
|
Throughout 1988, there was talk of a Japanese asset price bubble. In mid-1988, there were all sorts of analyses justifying the high prices of Japanese stocks. Land was astronomical. It could drop 90 percent and still be a lot more expensive than Vancouver. These analyses assumed high growth rates (real growth rates of 2 percent above the United States forever) or low interest rates so that the high PE ratios could be justified. The Cabinet did not like the situation and started to raise interest rates in late 1988 and more aggressively in 1989.
Iishi and I checked the model; it was way into the danger zone. I thought YRI and Yamaichi should benefit from their paid research. All along, I had been giving a number of well-publicized lectures to groups of 100 or more people from all over Japan invited by YRI. The lectures were enjoyable; we professors like to talk. The people listened. We were socially accepted both in Tsukuba and Tokyo.
I became very interested in Japanese golf course memberships as an investment subject. Golf course membership prices constitute one of the best real estate data sets — weekly data of bid-ask spreads from market makers in various areas of Japan for a homogeneous product. These prices were in the sky. One membership cost $5 million just for the right to pay to play golf. Later I would work with Zari Rachev of U.C. Santa Barbara and Doug Stone of Frank Russell Company to study this further (see Journal of Economic Perspectives,1993).
While most land/housing data is complex because it includes location, non-homogenous product, etc., this was close to clean data. The distributions were very fat-tailed. This meant that the probability of a large rise or fall in any given time interval was very large. They invited me for a golf game. Of course, the etiquette was that the gaijin professor was supposed to come in last. That was easy to accomplish.
All this social and academic success led me to a false security regarding how much the YRI heads would listen to our findings. They were later willing to have me work a bit more in the fall of 1989 on an anomalies factor model. The idea was to rank all the stocks using factors (see Keim and Ziemba, eds, "Worldwide Security Market Imperfections," Cambridge University Press, 2000). So after the golf game, which included the YRI head, I briefed Iishi and sent him upstairs to explain our results in Japanese to the head of YRI.
I thought that would be the most effective way of communicating that an enormous crash had a good chance of occurring as the model predicted. Unfortunately, it was a time of up, up and away for Japanese stocks, and the research of a gaijin professor and his young assistants was discarded. Later Yamaichi, the fourth largest security firm in Japan and one of the top 20 in the world, would declare bankruptcy. The model was simple, but it would have saved Yamaichi had they hedged, even partially, in late 1989 or even in early 1990 as the stock market began to fall.
Figure 1, drawn in May 1990, shows the spread between bond and stock yields in the Japanese stock market from 1980 to the end of May 1990. Each time the spread exceeded the 4.23 cutoff (which has higher than 95 percent confidence) there was a correction. By the end of 1989, the stock market was way into the danger zone and the lower confidence level indicated a bottom of about 17,000. In fact, the market fell to 14,300 in 1992 and later fell as low as 11,000. The model also indicated that the valuation was still high as of May 29, 1990, at 4.88.
Figure 1: Bond and stock yield differential model for Japan, 1980-90.
Source: Yamaichi Research Institute
|
Date/level |
Spread |
NSA |
|
May 29, 1990 |
4.88% |
32,818 |
|
Mean |
3.79% |
20,022 |
|
Upper Limit |
4.23% |
23,754 |
|
Lower Limit |
3.35% |
17,303 |
Table 1: Value of NSA for various spread values.
I will finish up with two final topics: the stock market spread in the United States in 2000 and 2001 and more on the Japanese economic woes of 1990 to 2001.
There were dangerous conditions throughout 2000 in the U.S. stock market according to the spread. In December 2000 and January 2001, the market was finally out of the danger zone so 2001 looks better in its declining interest rate environment. The S&P 500 returns in January are another powerful signal. Chris Hensel and I found, and wrote in two papers, that if January is positive then the rest of the year is positive about 80 percent or more of the time for the United States and major foreign stock markets. However, if January is negative then the rest of the year is noise — that is, the chances are about 50-50 for a gain or drop. Moreover, if January is positive and the signal fails, the resulting drop is lower on average, and if the signal works, then the average gains are larger than average. The reverse is true for negative Januarys. January 2000 was negative and we know that 2000 was a rough stock market year with negative returns for the S&P 500 and other indices.
In January 2001, the S&P 500 returns were positive which is another signal that the odds favor a rising stock market in 2001. It's clear that the Greenspan policy of high real interest rates in 2000 helped exacerbate the dramatic slowdown we are in now. The two half-point cuts helped the economy, but were a bit late; more is needed. Since the stock market predicts six to nine months ahead, the March-to-December decline predicted the current slowdown, and current stock market action is looking toward the fall. The odds favor a rise in stock prices, but it will be a bumpy ride with much volatility as earnings disappointments show up and the guessing of the Fed's intentions will continue.
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Stochastic programmers, such as myself, model all this with scenario-dependent correlations. For example, in InnoALM, a model I designed that has been implemented for the Austrian employees of Siemen's Pension plan, we use three such matrices (aptly called the good, bad and ugly). Developing aggregations, sampling such scenarios (especially in the tails) and using them in models is the seminal research problem in applied stochastic programming research. Size of models, the culprit of the 1970s and 1980s, is still an active research area but of less practical relevance, as we can solve big models now with modern PCs.
As for Japan, 2001 marks the 11th year of decline since the Nikkei peaked at 38,916 in December 1989. As of Feb. 7, it was down to 13,366. The stock market and the economy are in as bad shape as ever, and the future looks especially bleak. Some of the key problems and happenings in Japan and related areas:
- The interest rate was cut from 0.50 percent to 0.35 percent on Feb. 9.
- In the last six years, the U.S. market has returned 190 percent versus 37 percent for the rest of the world (of which Japan is a large but declining percentage), and the correlation based on daily movements is 0.765 according to Merrill Lynch. Not surprisingly, closed-end country funds are declining in popularity and are trading at large discounts.
- Japanese direct foreign investment in the United States as a percent of all sources of DFI rose from 5 percent in 1985 to a peak of 30 percent in 1990. It has steadily declined since then and in now about 3 percent.
- There is a worldwide trend by U.S. investors to invest less in equity abroad. Much higher correlations with the U.S. markets means there is less diversification benefit, especially in crises when these correlations approach one.
- Japanese land prices have fallen about 60 percent since peaking in 1991 (they lag the stock market which peaked at the end of 1989) and speculative land such as golf course membership prices have fallen even more. Land prices are now back to 1985 levels, the time period before the huge run up in asset prices until the 1990-91 decline. Land prices have fallen nine consecutive years.
- Japanese banks have non-performing loans — largely backed by land and somewhat by stocks and other collateral worth much less than the loans and fairly illiquid — in the $1 trillion range. These banks are kept afloat by the government which has been unwilling to free the market and let the losers go bankrupt and start again. Such banks naturally must pay a premium when they obtain funds outside Japan. This premium in London is about 20 basis points versus 100 in the 1997-98 Asia currency crisis.
- In April 2001, banks will need to mark stocks at market as opposed to the current book values and this will create another round of problems. This will lock in losses unless the market rallies which seems unlikely. The changing rules on cross holdings, which must be cut back by the banks and other companies, will cause further problems.
- The economy has much excess capacity; production costs are too high. What is needed is to estimate actual market demand and produce for that, and move away from governmental industrial policies, and weak consumer spending and business confidence.
- The consumers are loaded with money, but they prefer to save it rather than spend it — as they have lost wealth. Japanese still save about 30 percent of their net income. Some $2 trillion is parked in postal savings at very low interest rates and is not taxed.
- There is simply too much levered debt in Japan and too little equity; too much savings at fixed rates and not enough at risk.
Unfortunately, Japanese officialdom has spent the past decade running away from the hard decisions needed to revitalize the economy. They need to address structural problems and change tax and pension rules. But the political will is not there. There have been some half-hearted attempts to prop-up the stock market called price-keeping operations. This has had periodic minor success in the past, and there have been some runs in the stock market. One was in 1999. In a talk to Canadian Fund managers at Lake Louise in March 1999 organized by UBC's Bureau of Asset Management, my research pointed to an undervalued Japanese stock market, especially since the countries in the 1997-98 Asian currency crisis were in the process of rebounding. So, at 13,000, the call was up, and it did rally to the 18,000 range that summer. But the old troubles returned in 2000 (the U.S. slowdown in late 2000 did not help) and the outlook is even worse now.
William T. Ziemba is the Alumni Professor of Financial Modeling and Stochastic Optimization, University of British Columbia.
OR/MS Today copyright © 2001 by the Institute for Operations Research and the Management Sciences. All rights reserved.
Lionheart Publishing, Inc.
506 Roswell Street, Suite 220, Marietta, GA 30060, USA
Phone: 770-431-0867 | Fax: 770-432-6969
E-mail: lpi@lionhrtpub.com
URL: http://www.lionhrtpub.com
Web Site © Copyright 2001 by Lionheart Publishing, Inc. All rights reserved.