Experience tells us that guessing the top of the market is far easier than guessing the bottom of it. The market top usually shows up after a long climb. Of course, there is no clear definition as to how long the “long” should be. We could say, however, that when the market climbs to the top, various indicators begin to signal “overheat,” and at least for some short duration, and sometimes for some considerable duration, the market experiences a correction.

In contrast, even when “oversold” signals flash and we see temporary rebounds, in many cases the market will continue to correct. We may say that here again, human psyche is at work.

As the phrase “profit take” suggests, when we achieve certain returns (out of greed), fear begins to grip our soul. This is the first reason for a market correction. And once the correction begins, the fear amplifies, and unless investors see clear signals, they tend to stay away from the market. This may be the primary reason why it is difficult to guess the bottom of the market.

For humans, it may be easier to eradicate greed than fear. In this chapter, we will consider a few indicators that may tell us the top and bottom of the market, as we continue to struggle with greed and fear.

Volatility as Indicator

When the equity market rises steadily, after a while the market volatility drops to a level not seen or seldom seen in the past. This is a danger signal, and we may wish to check futures open interest levels and cash arbitrage positions, in addition to various technical indicators, to be discussed later. Although there are no “absolute” levels for these indicators, we can at least make a probabilistic judgment from historical examples.

If, after a long ascent of the market, we see significant policy announcements, clear signs of a weakening economy, or large movements in interest rates or FX, there is a good chance of a market crash, and we need to be extra cautious in these cases. As an action plan, it may prove prudent to take a profit and shift funds to more defensive stocks. Defensive stocks are typically those with low historical volatility, and in terms of sectors, generally belong to the food, telecom, utility, and railway sectors.

In recent years, we hear a lot about hedge funds such as commodity trading advisor (CTA) funds and risk parity funds. In contrast with CTA funds, which are basically momentum funds that follow a trend, risk parity funds place volatility in their core strategy (these categorizations are oversimplified for our discussion purposes).

Risk parity funds commonly practice diversification by investing globally in equities, commodities, and bonds, and their strategy is to move their funds from high volatility assets to low volatility assets. They abhor high volatility, and as soon as volatility begins to rise, shift funds, but as touched upon in the “GPIF” section in Chapter 2, there are some equity‐only funds that follow similar investment disciplines.

These funds utilize various technical indicators to forecast future market volatility, and one of the indicators is likely a volatility level far from the historical average. Here, market volatility refers to the volatility of major equity indices such as the Nikkei 225 and TOPIX.

There are a few reasons why the market volatility drops. One is a fall in volatility of each index‐constituent stock, and another reason is a fall in the correlation among constituent stocks, or both. That a fall in correlation causes a drop in market volatility is easy to understand: if the index consists of just two equally‐weighted stocks, A and B, and if A goes up by 5% and B comes down by 5%, the index will not move (strictly, the index will move somewhat due to the change in weight), and the index volatility on that day will be 0%.

Ordinarily, an index consists of many stocks, and they move by different percentage points. If the correlation is +1, all the stocks moved in the same direction by the same percentage. If the correlation is –1, an equal number of stocks moved in the opposite direction by the same percentage. In actuality, the correlation fluctuates between +1 and –1. And with other conditions kept constant, the higher the correlation, the higher the index volatility, and vice versa.

As noted in the “Put/Call Ratio” section in Chapter 3, a volatility spike can be used as a “buy‐on‐the‐dip” opportunity, but the returns are not necessarily spectacular and the win‐ratio is contingent on market sentiment. This is because, as discussed in the “VIX Index” section in Chapter 3, particularly when the economy is in deterioration, volatility spikes tend to occur in succession.

While a difficulty in calculation makes it less usable, a more reliable “buy‐on‐the‐dip” signal is the correlation coefficient. This may be due to the fact that the correlation moves between +1 and –1 and tends to mean‐revert more than the volatility. Table 4.1 lists the 10‐day, 20‐day, and 3‐month Nikkei 225 index returns when the Nikkei 225 was bought with the signal of its 1‐month “realized correlation” rising above the 1.5σ of the 6‐month average.

The above statistics are based on the data sample from 2003 to 2009, but the “win ratio” is largely unchanged whether the market was tending upward or downward. Unfortunately, the “realized correlation” is not something readily available to ordinary investors and thus its practical usage as a signal may be limited.

Also, we note that Table 4.1 only refers to the case where the correlation rose to 1.5σ or above, and we do not vouch for equivalent results when the correlation falls. In other words, using the correlation as a “sell” signal may not be a valid strategy.

This said, stocks do not perennially move with negative correlations, meaning that probability dictates that a negative correlation cannot last forever. If that is the case, what reverses low correlation and makes stocks begin to move more or less in the same direction?

Economy and Policy as Indicators

A terrorist attack such as 9/11 or a major natural disaster such as the Great East Japan Earthquake inevitably sinks the market, but since these events are unpredictable, we will not take them into consideration. Apart from these events, one of the major elements that drive the market in one direction is the state of the economy.

As described earlier, our experience tells us that to know the state of the economy, we only need to follow a few crucial leading indicators. We have mentioned that these indicators include the ISMPMI and the OECD CLI. When these indicators show signs of sudden deterioration, we are almost certain to see the elevation of market volatility.

Whether we can use these indicators to see where the economy is going before their official release of data, we claimed for the OECD CLI in the “More on OECD CLI” section in Chapter 1, that forecasting was possible to some extent by isolating the country‐by‐country constituents and their respective weights. For the ISMPMI, we should remember that the indicator is based on a survey.

If manufacturers believe that the economy is on the mend, they will likely increase capital expenditure and inventory, but if they believe the opposite, they will likely refrain from additional investments and may even lower prices to reduce inventory. All of these corporate behaviors directly impact the ISMPMI. When corporate heads see the worsening economic indicators, they generally further rein in their investments. This is the “downward spiral” mechanism.

Of course, in reality, there are as many corporate heads as there are numbers of corporations, and what they see and think are not necessarily uniform. Undoubtedly, however, they are a group of learned and experienced individuals who share a certain common sense, and when the common sense becomes the consensus, corporate activities may dwindle or resurge.

One such commonsense response may be formed around the policy and action of the central bank. As written in “The Fed” section in Chapter 2, since the central bank acts to stabilize the economy, if the action is deemed to destabilize the economy contrary to the intent, corporations will be more cautious regarding their future outlook and activities.

While the Fed's role is to “stabilize the economy,” history shows that the Fed's job has largely been that of fighting inflation. Indeed, fighting inflation has been the job of the world's central banks.

Deflation is an offspring of recession and thus has existed long into the past, but the word “deflation” did not become part of the household lexicon until about twenty years ago, and as far as I know, the epicenter was Japan. Thus began the anti‐deflation measures by the BoJ, followed by the Fed and ECB post‐2008, only to be expanded into Kuroda Bazooka, as discussed in the “Abenomics” section in Chapter 2.

Returning to the Fed, since the Fed has historically been an inflation fighter, its main duty is that of braking the economy. Inflation may simmer down when a brake is applied to the economy, but the economy may also come to a grinding halt as a result.

The BoJ's interest rate policy contributed to the collapse of the '80s bubble in Japan, for example. Similarly, it is easy to imagine how the Fed's interest rate policy contributed to the collapse of the internet bubble in 2000 and the credit bubble in 2008.

A recent example may be the Fed rate hike in December 2015. Whether we look at the GDP growth rate, durable goods new orders, housing starts, or the ISMPMI, the US economy was not strong enough to absorb the impact of the rate hike. Unsurprisingly, coinciding with the timing of the Fed rate hike, global equity markets began to crumble. The initial reaction was mild, but as the December ISMPMI showed a sudden and substantial decline, the equity market came tumbling down. This was particularly apparent in Japan.

These examples suggest that in passing judgment on market tops, we must ponder the balance between the state of the economy and the level of the market. If the market is believed to be lagging the economy, the market will continue to rise, and vice versa. When the market is not overheated and, at the same time, the economy is robust, even in the wake of rate hikes by the central bank, the market will likely remain strong.

In addition to the human psyche element, a difficulty in guessing the bottom of the market arises because the bottom often does not hinge on the state of the economy. A simple example involving the rate of unemployment, a frequently quoted statistic, may aid in understanding this.

While it somewhat differs from country to country, generally full employment is deemed achieved when the rate of unemployment is somewhere around 3% or 4%. In other words, an economy will not improve much more from these unemployment levels, and the top of the market is likely close at hand.

When we think of the bottom of the market, however, there is no clear unemployment level that signals it. Needless to say, the unemployment level will never reach 100%, but this is an extreme case. The problem is that there is no historical answer as to what rate of unemployment foretells the bottom of the market. In other words, whatever the rate of unemployment may be, the market can rebound, the onset of which forms the bottom of the market.

In most cases, the market does not rebound due to a sudden improvement in economic statistics. Rather, the market rebounds due to policy announcements designed to improve the economy or due to technical factors discussed later. Only insiders know when and what policy will be announced. This is just another source of the difficulty in guessing the bottom of the market.

It generally takes months or even years for a given economic policy to impact the real economy. Accordingly, when a policy is announced, the market is not responding to an actual improvement in the economy but the content of the policy and expectation of the policy's impact. Whether the resulting market rebound becomes sustainable depends on the actual improvement in the economy. That said, although it may take months before the policy reveals its impact, the harbinger of the impact may show much sooner, and even if it does not, often it is wise to invest in equities assuming the effectiveness of the policy.

The reason, again, is that corporations and investors often share a common thought. When the central bank cuts interest rates and engages in quantitative easing, or when the government announces a large‐scale supplementary budget to stimulate the economy, corporations, believing that the economy will improve, are likely to increase spending (i.e., they may increase capital expenditure or hire more workers).

Investors, too, may decide to increase equity holdings or buy real estate even on borrowed money, and ordinary citizens, attracted by a favorable interest rate environment, may decide to buy housing units or luxury items. All these activities pile onto each other to boost the economy, forming something opposite to the “death spiral.” This is not just an economic theory but is a phenomenon that happens in front of our very eyes.

In using the state of the economy as a yardstick for the top and bottom of the market, a measure often quoted is the Buffett Indicator. The name springs from the famed US investor Warren Buffett, and the ratio is a quotient of the total equity market capitalization divided by the nation's GDP. If in the US, the equity market capitalization is that of the S&P 500, and if in Japan, TOPIX, and the GDP should be the inflation‐reflecting nominal GDP.

The usefulness of this ratio arises from history, as the collapse of major “bubbles” in the past was triggered when the Buffett Indicator was either above or close to 1. Also, the market, after collapse, tended to find a bottom when this ratio was close to 0.5. The usefulness of this ratio is limited, however, by the scarcity of such examples (i.e., the Buffett Indicator has seldom gone above 1 or below 0.5).

Additionally, even when the ratio goes beyond 1, the condition can last for quite some time. Good examples of this can be seen in the recent equity market in the US and in Japan. As of November 2017, the US equity market has been shifting above the Buffett Indicator of 1 for a few years, and the Japanese equity market for at least several months.

Valuations and Technical Indicators

So far, we have considered the ways to forecast peaks and troughs of the market from the viewpoints of volatility, the economy, and policies. In this section, we will look into more traditional methods, use of valuations and technical indicators.

Having said this, the realm of valuations and technical measures is broad and deep so that we cannot possibly cover the entirety of these subjects in this book, and perhaps moreover, unless history repeats itself 100%, examining every minute detail and situation where these factors were relevant in the past may not be all that helpful. Accordingly, we will limit our discussion to more or less representative and well‐known valuation and technical measures.

We note that while economic indicators and policy responses may be more suitable in forecasting a mid‐ to long‐term stock market movement, valuations and technical indicators are more often used for short‐term predictions. The success of the OECD CLI may be impressive, but, as discussed earlier, this indicator is more suited for forecasting mid‐ to long‐term trends of the equity market. Valuations and technical indicators, therefore, can possibly compensate for any defects of the longer‐term economic indicators.

One area of technical indicators is the chart. There are those analysts and traders called chartists, who employ shapes of the market movements (charts) in timing the market. They not only observe the shapes and trends of the market but also often utilize various technical measures as well. I am a complete stranger in understanding the shapes and forms of charts, and thus have no credibility in commenting on the validity of the subject.

Warren Buffett apparently pays little attention to charts, stating that they look the same when turned upside down, but he is a quintessential fundamental long‐term investor and has no need to understand short‐term fluctuations of the market.

There are, however, day‐traders who anecdotally have claimed to amass considerable wealth by effective usage of charts. As always, if a large group of people see the same phenomenon and move in the same way, stocks and stock markets will move. In this sense, charts cannot be easily dispensed with as “nonsense,” but whether their validity depends on something more than a beauty pageant can be debated.

If historical investor returns are anywhere near normal distribution, then there should be those that boast outstanding returns almost consistently, and it is possible that those investors might have been simply “lucky.” Whatever the case may be, my own ignorance on the subject means that I cannot discuss the validity of their methods. What can be said is that the validity may be evidenced by the simple existence of those who have claimed to benefit from charts.

Valuations generally refer to those traditional measures, such as P/B (price‐to‐book ratio), P/E (price‐to‐earnings ratio), and EV/EBITDA (enterprise value/earnings before interest, taxes, depreciation, and amortization). These measures are used to judge whether the share price is cheap or expensive relative to the fundamentals.

Apart from the extremes, there is no definite and absolute fair value of these measures. Depending on the market sentiment, and for single stocks, their historical valuations and valuations of their peers, fair values are determined. Once again, the beauty pageant aspect of the stock market asserts itself here.

Can, then, these valuation measures tell us where to find the top and bottom of the equity market? If valuations are the beauty pageant tools, then the answer seems to be yes, but historical examination tells us that this is not necessarily so. Said another way, investors can suffer a great loss, at least in the short‐term, if they pass judgment on market tops or bottoms by using valuation measures alone (mid‐ to long‐term effectiveness will be discussed later).

We often hear an argument that P/B = 1 is the bottom price, whether for the entire market or single stocks. The textbook argument goes this way: since the “B” of P/B is the book value, the share price cannot theoretically fall below its book value. This argument, however, overlooks the fluctuation in the book value. When the market collapses, often the book value itself suffers a great deal of damage.

In the fall of 2008, during the middle of the global financial crisis, some argued that since the TOPIX P/B fell below 1, the Japanese equity market was a “buy.” Needless to say, the market kept on falling afterward, and had we followed the above advice, the trade would have incurred a loss at least for months to come.

We may think that if the share price and book value fall alike, the P/B measure may still prove effective. The problem is that while the share price changes are monitored and observed constantly, the book value can only be known by company disclosure and generally is kept under a veil until the time for company reports. The P/B's falling below 1 is a phenomenon that takes place because investors assume before disclosure that the book value has already suffered impairment.

A similar situation can be said of P/E and EV/EBITDA. In calculating P/E or EV/EBITDA, a common practice is to use estimated future earnings numbers for the denominator “E.” Take a company whose fiscal year‐end falls at the end of March, for example. If “now” is September 2017, the earnings number to be used is the forecast number as of March 2018, and the forecast number comes either from industry analysts of brokerage firms or fund managers, or their consensus.

Analysts typically form their views based upon company forecasts. In general, the views are not renewed frequently but every quarter or even with less periodicity. When the economy shows sudden weakness (which can trigger or be caused by an equity market collapse), company forecasts and analyst views are often unable to keep up with the sudden change, and thus P/E or EV/EBITDA becomes unsuitable in predicting the market bottom.

Here, we looked at 2008, a very special year, as an example, but whenever we have seen major market collapses, similar phenomena existed. As it is difficult to use valuations to detect market bottoms, it is also difficult to use valuations to detect market tops, since the same issues with the book value and the forecast earnings still exist when the market is strong.

To test all existing technical indicators under all conceivable circumstances is not feasible. The argument here, therefore, will be once again limited. Suffice it to say that as far as my experience goes, I have never met a trader who was perennially successful by simply following technical indicators. In this regard, the usefulness of technical indicators is probably also limited.

A difficulty with technical indicators is that even for a single indicator, many versions exist, and they need to be selectively used depending on market conditions and prevailing sentiments. Moreover, even the selective usage does not guarantee a positive outcome, and more effective usage may need to wait for the completion of AI as the trading tool. With these understandings, let us now look at the more popular technical indicators: RSI, Bollinger Bands, and Toraku ratio.

RSI (relative strength index) is a ratio between the rate of upside price moves and downside price moves in a given period of time. It offers a way to quantify “oversold” and “overbought” conditions. When this indicator is below 25–20, it signals a “buy,” and when above 70–80, it signals a “sell.” RSI is expressed by the following equation:

At a glance, RSI appears to be an effective tool. Indeed, equity prices and indices tend to bounce up when RSI is close to or below 20 and fall after RSI's climb close to or above 80.

The problem is that equity prices and indices often bounce up long before RSI gets close to or below 20 and turn down well below RSI = 80. Also, there is no clear answer as to how long we need to hold the equities or indices after purchase at RSI = 20. Even if we conduct an extensive examination of the subject, the likelihood is that the answer varies widely, depending on market conditions.

Other outstanding issues may include the occasional cases where stocks or indices continue to rise after RSI reaches 80 or fall after RSI reaches 20. The obvious implication is that strictly obeying RSI as the buy‐sell indicator may put us on the short end of the stick. Additionally, the definition of RSI leaves “a period” uncertain. Commonly, a period of 14 days or 25 days is used, but we do not know if these periods are ideal. Once again, the “ideal” period likely depends on market conditions and sentiments.

Needless to say, RSI is a statistical measure, and therefore, we expect occasional misses. Accordingly, we may be tempted simply to follow the “buy‐at‐RSI=20 and sell‐at‐RSI=80” strategy. A backtest of this strategy, however, did not generate a positive return either consistently or over the long term.

Bollinger Bands are similar to RSI in principle. The approach takes the standard deviation of the price history to set up the upper and lower limits. The buy or sell signal will be lit when these limits are breached.

This approach assumes that the historical prices are normally distributed. The standard deviation is expressed as σ, as noted earlier, where 1σ indicates that 68% of the prices and 2σ indicates that 95% of the prices are contained within that range in a given period.

For 2σ Bollinger Bands, for example, if the price goes above the upper band, since such an occurrence was observed only 2% of the time in a given period, the event is considered rare and unlikely to last long. In this case, therefore, a “sell” signal will be lit. If the price falls below the lower Bollinger Band, on the other hand, a “buy” signal will be lit.

When we look at the actual price movement, we indeed see the price fluctuating between the upper and lower bands. This is not surprising since the data is based on the past and is expected from the definition of Bollinger Bands.

Just as the Bollinger Bands principle is similar to that of RSI, so are the problems. First and foremost, there is uncertainty as to what “period” is the appropriate period in which the standard deviations are to be taken. At the threshold of a strong bull or bear market, in particular, a given period in the recent past may not function at all.

There is also uncertainty associated with the level of the σ. That is, determining whether 2σ or 1.5σ or some other σ is appropriate depends on the market conditions of the time. Once again, to identify which level of σ is appropriate under given market conditions, we may need the help of AI. At present, the best we can do is to conduct as thorough a backtest as possible and come up with some σ based on the resulting statistics.

Under the assumption that the past price pattern will be repeated in the future, the Bollinger Bands approach is a valid one and may prove to be effective. Of course, if future price patterns deviate widely from the past ones, the Bollinger Bands approach may result in catastrophe.

The Toraku ratio has been in use for many decades and has many followers. The ratio, unlike RSI and Bollinger Bands, cannot be applied to single stocks. Rather, it is used to judge how “overheated” or “overcooled” the market is. The Toraku ratio also depends on the measurement period. The 25‐day Toraku ratio, for example, is expressed by the following formula:

Commonly, a Toraku ratio above 120–130 represents an “overheated” market and below 70–60 represents an “overcooled” market. Table 4.2 is the backtest result of trading the TOPIX following these measures over the last ten years (using the 25‐day Toraku ratio):

The effectiveness of the Toraku ratio is particularly apparent as a “buy‐on‐the‐dip” measure, after the ratio falls below 60 (although the standard deviation is large). Also notable is that the effectiveness largely disappears after 5 days (the difference is small between 5‐ and 10‐day returns), confirming that the Toraku ratio is a short‐term measure.

Nevertheless, though the Toraku ratio may serve as a measure of the overheated and overcooled markets, it may not be all that effective, on the average, as an inflection point measure.

So far, we have seen limited success in capturing the short‐term fluctuation of the market by valuation or technical measures alone. As discussed in the “VIX Index” section in Chapter 3, the situation where the VIX Index spikes up (valuation and technical measures collapse) leads to investor abhorrence of risk and tendency to switch into more defensive stocks. This inclination suggests favoring of more bond‐like stocks or selling of stocks against bonds.

Also, judging from the Toraku ratio and other measures, the “buy‐on‐the‐dip” strategy does appear to work at least partially. It is a question of probability, of course, but extreme cases in the past, in particular, behoove us to consider the “buy‐on‐the‐dip” strategy.

On February 12, 2016, the Nikkei 225 Index dropped over 12% from its 25‐day moving average. The index bounced up 7.2% the next day, and a week after was still up 6.8%; two weeks after, up 8.3%; and after one month, up 13.3% from the February 12 close level.

Similar cases, where the Nikkei 225 Index fell more than 12% from its 25‐day moving average, have taken place 27 times since 1990, excluding the February 12, 2016, example. Some 63% of these cases are concentrated in October 2008, which makes the period indeed extraordinary.

If we exclude this period from our backtest, the average return 1 day after the decline is 2.5%, with a 90% win ratio; the average return 1 week after is 4.6%, win ratio 100%; 2 weeks after, 4.3% and 90%; and 1 month after, 6.9% and 90%, suggesting that the “buy‐on‐the‐dip” strategy was the correct one.

Similar statistics can be collected by ranking the 1‐day percentage decline of the Nikkei 225 Index. The data is somewhat old, starting from May 22, 2013, but if we collect all the cases where the Nikkei 225 fell more than 5% in a day since 1990, the average return after 1 day is 1.3% with a 71% win ratio; after 1 week, 3.4% and 71%; and after 1 month, 3.5% and 70%. These are all decent statistics, although not as good as the earlier cases probably because this backtest does not exclude October 2008.

These examples may suggest that, even when using technical measures and the VIX Index as inflection point indicators rather than judging from commonly accepted levels, we probably should look at more extreme levels. Needless to say, extreme cases do not happen too often, but market tops and bottoms are indeed those rare cases, and our argument here is simply to examine which measures are the most effective.

In this regard, we should note that although, as discussed earlier, the valuation measures such as P/B and P/E, with book values and forecast earnings subject to change, may not be effective in the short term, they can yet prove their validity when technical considerations are added.

Table 4.3 represents the backtest results of trading the TOPIX when the TOPIX P/E and P/B were either below or above their 7‐year average. The backtest assumes that the TOPIX futures had been bought when the P/E or P/B was below the 7‐year average and sold when it was above the 7‐year average, and it measures the 3‐month, 6‐month, 1‐year, 2‐year, and 3‐year returns afterwards. The CAPE notation in the table refers to the cyclically adjusted price‐to‐earnings ratio, which uses the 10‐year average earnings figures in the denominator of the P/E to avoid the annual fluctuation in earnings.

More strictly, the CAPE calculation needs to take inflation into account, but since inflation has been virtually nonexistent in Japan over the last ten years, this factor is ignored in the present case for the sake of simplicity. The “Nonadjusted P/E” in the table, on the other hand, denotes the ordinary P/E, which uses the single‐year earnings in the denominator.

The reason the 7‐year average is being used is simply due to the lack of data. In fact, using the 5‐year average instead of the 7‐year average generates results not too distinct from Table 4.3. What Table 4.3 shows is that the CAPE, P/E, and P/B all have a tendency to return to their long‐term average. It is this tendency that generates the profit when any one of these valuation measures deviates from its average.

As usual, the reality is not as simple as it sounds, however. Looked at closely, the standard deviation of each return is quite high, with the possible exception of the 3‐year return. If we assume that the returns are normally distributed, probability dictates that the returns could have been negative. (For example, the 3‐month return of the CAPE case could have ranged from ‐7.9% to +11.9% with a 68% probability.)

If the 3‐year return is an exception, maybe we should wait for 3 years, but again, the results differ depending upon the period of observation. Table 4.3 is the backtest results obtained since September 2010. The equity market was largely upward‐sloping after the global financial crisis and benefiting from Abenomics.

Table 4.4 represents the backtest results since May 1998 (due to the lack of data, only the P/B was backtested). The 2‐year and 3‐year returns are once again attractive but nevertheless substantially below those obtained earlier.