1. The Sentiment Era

“Sentiment” expresses the unseen mentality of people and atmosphere of the real world. I believe that that sentiment, or the human mentality called the animal spirit, is important in monetary economy and corporate management, the focus of my research. While we work, study, live with our family, shop, and take breaks, I think we feel hope for, or conversely, anxiety for, or do not know something about facing our future dreams and lives. Sentiments such as that human mentality and atmosphere, as human behavior or a result, become the driving force behind the economy. Put another way, sentiment influences economic trends and shapes expectations for the future economy.

It is in the market that those sentiments are truly reflected. It is thought that stock markets etc. lead the actual economy. And it can be thought that sentiment guides the current stock market, or the near future, or, at the least, exerts an influence.

In line with that idea, research interpreting human sentiment and trying to predict markets and the economy is currently extremely popular the world over.

2. Nowcast～forecast the present

There are various approaches to analyzing sentiment. For a long time, there have been experiments where business sentiment is investigated by way of survey and converted into numerals. Examples of collected surveys of consumers and business managers include the Cabinet Office Business Watcher Survey, the Bank of Japan’s Tankan (Short-Term Economic Survey of Enterprises in Japan), and the University of Michigan Consumer Sentiment Index. These indexes are surveys conducted by public institutions and have the advantage of having accumulated data over many years. However, one disadvantage is that time is required to make these results public.

For example, when taking stock prices in the market into consideration, all the latest information is reflected instantly, resulting in fluctuating stock prices. It includes economic, business, political and social news that comes every moment from around the world and sentiment of how people receive such information. Accordingly, if we make past economic phenomenon and business administrative action, which has been well covered in economic science and finance, the subject of analysis, up-to-date information is not really required. However, in order to analyze the real economy of “now,” up-to-date information is extremely important, and above all, is the key to understanding human sentiment “right now.” That is to say, how we nowcast=how we understand the present has become a theme of current financial research.

3. Monetary economy research following real time

Let me give a few examples of research analyzing monetary economy in real time. The largest exchange dealing stocks in Japan is the Tokyo Stock Exchange. It introduced the system called “arrowhead”, an ultra-high speed system which can process transactions in milliseconds (1/1000 of a second), on January 4, 2010. Not limited to stocks, the basic principle of business is “buy cheap, sell high.” If that is the case, for investors, being able to buy and sell stock even 1 millisecond faster than your rivals becomes a condition for winning. In recent years, as most transactions are “algorithmic trading”, where buy-sell orders are placed by computer programs, ultrafast and high frequency transactions, the so-called “HFT (high frequency trading)” is becoming a keyword, not only in business, but also in academic circles. But it also has its problems. HFT can be said to have played a part in the “flash crash” on the U.S. stock markets on May 6, 2010, where the Dow Jones Industrial Average plunged by about 1000 dollars in several minutes.

As another example, let me introduce research focusing on real time. In Abenomics, which began at the end of 2012, as one of its standards for success or failure, a price growth rate, or 2% inflation has been set as a target. This price growth rate is measured in the CPI. CPI stands for Consumer Price Index, and expresses, in numerical form, the sum of prices for goods and real assets (housing and cars etc.) consumed in family budgets. That price growth rate is inflation. Publicly released by the Statistic Bureau, Ministry of Internal Affairs and Communications since August 1946, it is useful as a main economic indicator. However, those announcements are delayed on a monthly basis and lack real time nature, and the existence of bias has been pointed out since they are sampled by particular people. In order to eliminate this problem, and grasp raw economic activity in real time, academic circles have taken the initiative in MIT’s Billion Price Project, followed by the UTokyo Daily Price Project.

4. Text mining～sentiments closely analyzed from Japanese

The sentiment analysis I am involved in is an approach that extracts mentality from Japanese language texts. We live in an era where social media, cloud computing and big data are keywords. While we reap many benefits from these, we react immediately to people’s words and produce our own words. Those kinds of words are directly reflected in people’s mentality. The most straightforward and direct emotions are positive emotions and negative emotions. For example, the Japanese phrase “Keiki ha yoi to omawareru.” reflects positive emotions. In this way, using modern technology, I have investigated what kind of emotions reflected in the vast Japanese language delivered through the media. Then, I have converted them into numerals from one hour to another.

The idea is very simple, but actually very hard work. For example, as the Japanese phrase given above translates into English, “The economic condition seems good.” For a computer, which is easier to handle, English or Japanese? English, of course. In English, each word is separated by a space, and if you remove articles and punctuation marks, this example is in an adjective-noun-verb-adjective sequence, making it easy to distinguish. On the other hand, for Japanese, each single word is not separated by a space, and processing such as inserting a space between words is necessary. I think that, in an earlier era, I would have never thought of attempting to convert Japanese emotions into numerals. However, recently, research related to natural language processing and text mining, which extracts useful information from Japanese, has developed rapidly, and computer software which implements those results has become available. Also, even in the academic world of monetary economy, the establishment of behavioral finance and behavioral economics focusing on the mental side lies in the background. In addition to that, many funds are already created by applying text mining to trading such as stocks in business practices centered in Europe and North America.

On the other hand, finance research using sentiment analysis in Japan has only just begun. First of all, I gathered about 30 years’ worth of the Nikkei, which is read by many businesspeople. On top of that, we made every article and headline from the past 10,920 days the scope of my analysis, converting sentiment into numerals on a daily basis. In the US, research papers using the Wall Street Journal and Twitter are well known, but the Nikkei has a circulation of 1.5 times that of WSJ (2011). That is to say, if you take the size of the populations into account, there is a possibility that sentiment figures from the Nikkei will have a stronger influence on the market compared to those in the US. Now, can sentiment predict stock prices? What do the results of empirical research show?

As expected, we found that in Japanese markets sentiment possesses sustained and meaningful stock price predictivity (for more information, please read the essay I wrote while at Columbia Business School.)

5. Silver lining～New research areas created by sentiment analysis

Last year I was given the opportunity to spend a year on sabbatical at Columbia Business School’s Center of Japanese Economy and Business, and at the Ivy League campus with brick paths and grass surrounded by short cut taxus cuspidata, I learnt many things which affected me greatly. One phrase which remained me was “every cloud has a silver lining.” My interpretation of this is that a dismal, cloudy sky will always be shining in a silver colour from the sun’s rays on the other side. This is also a proverb that became the title of a movie (Silver Linings Playbook). I think that “every cloud has a silver lining” is a perfect expression for sentiment analysis. Using ICT in the “cloud computing” era, we interpret unseen human sentiment hidden in “clouds” from big data, including Japanese text. Practical application is not limited to financial research. Because we can expect the creation of interdisciplinary research fields from areas close to finance such as accounting, marketing and management strategy, to election predictions, policy and administrative evaluation, and changes in the Japanese language itself, I want to take a positive approach to collaboration with people from various fields.