Individual consumption is a highly important component in monitoring the domestic economy. We look at both mined data and hard data.
We think that in the long term, consumption is what creates the ups-and-downs in an economy.
For instance, many American financial policies are made based on retail sales. Of course, consumption details are imperative in forecasting GDP and stock prices, but are also significant in following financial policies.
This goes for Japan as well. In Japan's case, what is most critical is whether consumption eventually leads up to inflation or not. Therefore we combine a variety of CPIs and consumption statistics in order to forecast the Bank of Japan's policy making.
For example, if a company raises prices and experiences falls in sales, it would restore its previous prices. On the contrary, if consumption does not fall in spite of the raises, we may say that increases in consumption and prices is evident, and hence may assume that the BOJ will not conduct further monetary easing.
Macro data is very useful in seeing the full picture. If for example we are talking about securities for department stores, we would make sure to examine the monthly sales report published by the Japan Department Stores Association. We also use macro statistics to find out what the long-term trends and situations of different industries are, because at our firm, proposing mid-to-long-term investment strategies is a great priority.
This may be said for other countries as well, but we feel that the sampling of these data are often uneven. For example, the people surveyed in a household budget survey could only be representing a portion of the entire population.
To correct these disparities, it's best if surveyors expand their breadth of samples.
Not only expanding the size of data, but the speed of research could be improved as well to increase the responsiveness of these surveys.
It's a different story if we could perceive everybody's household expenditures, but if C2C (customer-to-customer) models of business are becoming more common and consumption patterns themselves are changing dramatically, we should take them into the picture in order to understand the market.
Exactly.
We are different from economists, so our main job is to propose investment strategies, not pure market research. Hence we routinely will examine economic indicators, but we also always check market consensus. And this is why we have high interest in perceiving trends and movements that are normally undetectable with other statistics.
Let us pretend that the numbers were low during the Christmas season in existing statistics. Ordinarily, this brings us to the conclusion in the market that consumption is weak. However, we would go a step further and doubt whether this is true. If it is proven that online consumption is rising and that people simply stopped going to the stores themselves, we could analyze that consumption is perhaps high. There are not many data that we could use to make analyses like this, but if we can, our operations would greatly expand as strategy planners.
We are incredibly excited about it.
For example, the consumption of services has risen steadily compared to goods. We think that this service will be useful in perceiving these trends. If we could look at consumption activities by gender using credit card attributes, we may be able to say something like that the economy is improving because more men are buying clothes. If we look by age, we may find out that elderly people, who have lots of savings but have not had enough time to use them, might be consuming very actively. The data will be imperative for testing such hypotheses as these, because existing data, which we have no other choice but to rely on, is mostly qualitative. We find much potential in this service for improving these issues.
Diversity of the indexes is key in this service. We can no longer identify trends by measuring each slice of the pie in overall consumption, because today we are not in an era of simple mass production and consumption. If we can analyze consumption from different perspectives, like age, gender, or region, we may find a wide spectrum of trends. Whether for policy or investment, it is necessary to examine the pieces, not simply the whole.
More focus should be placed on the peripheries or in smaller details.
Focusing on the whole may have sufficed in the previous era of mass production and consumption, because everyone bought similar products. But because consumption is so diversified today and sales forms also differ greatly in each industry, without understanding consumption by sectors, it is hard to identify the ups-and-downs in demands.
If we take the recent "bakugai", or "explosive buying" activities of Chinese tourists in Japan, this is an incredibly interesting topic in investment. However, you cannot understand it simply by looking at the averages. Ideally, we want to narrow down to foreign visitors in Japan and examine consumption trends, but this was not possible with conventional data; hence we could only discuss these topics qualitatively. If big data could resolve some of these issues, that would be an incredible achievement. We would be very thankful if each segment had its own sub-indices for analysis in particular investment themes.
It differs by the objective, but if it is to forecast GDP, we mainly use household budget surveys and industry statistics. However, these statistics are difficult to use in tracking monthly and quarterly movements. The truth is that we are also constantly in doubt of their reliabilities and accuracies. The BOJ publishes consumption indexes, but they only come out 40 days after the research, which is a little slow. Department stores, supermarkets and convenience stores are slightly faster since they publish in 20 days, and food services in 25 days. Despite the admirable speed, the statistics lack completeness because we cannot see data on other service consumptions.
We are very interested in them too. A statistic that uses interviews at airports for quarterly data is probably replete with bias, but is still speedy and useful. We think it's pretty quick for statistics from October to December be published in January.
Sometimes we discuss quantitative, macro-level topics like the current survey of commerce or household budget surveys, but also qualitative topics like consumption of foreigners in Japan, which we get information from our analysts.
The primary issue with all of these information is bias. Qualitative data unquestionably contain bias, but the people who answer quantitative surveys are also slightly unique, and so the samples become biased too.
For instance, how common is it actually to keep a household account book? Perhaps the problem could be alleviated by multiplying the sample numbers by 100, or like current household consumption research, people could instead fill in their account books once a month. Although, this would still leave us with the issue that single-person households are not taken into account, even though they represent one-third of all households in Japan. In these ways, the current survey of commerce and industry-provided statistics are very reliable. However, because service consumption is not included, many problems are yet to be solved.
We are interested in a wide variety of services, such as theme parks and hotels, fitness clubs and nursing care, but our biggest interest lies in online services. Sadly, there is not much information and there are many things we don't understand about it.
We agree. For example, you can look at reservation numbers for new games in the gaming industry, or visitor numbers for theme parks. The transportation industry also has some statistics, albeit a little slow.
But online companies often lack statistics entirely, making information inaccessible.
We think that categorizing different online services and understanding the scale of their businesses is sufficient in following online trends. Like for example, how much money 10,000 people would spend annually. According to household budget surveys, 8,000 households spent 3 billion yen in a year. It would be interesting to see how much spending online services accounted for.
Credit card payment details are very intriguing. If there are large disparities between government and industry statistics, the bigger it is, we could use that to figure out more interesting things.
Although biases in samples are problematic in existing statistics, JCB Consumption NOW should also be capable of minimizing them. The effects of price fluctuations can be greatly reduced too, such as in vegetables. Aside from supermarkets, few people would probably buy vegetables from a vender using a credit card. For these reasons, the project and the idea are very suitable for investment.
Nonetheless, there probably will be different sample biases compared to household surveys. Take age for example. The bias will likely decrease if there are greater amounts of past data to deal with, but since credit card payments also lean towards a few specific industries, the data easily could become biased too.
Also, as time passes by the portion of credit card payment is increasing, and so we need to take care this bias too.
We think that by mixing gender, age, and industries in use, the service could create specific investment-themed categories. Once data includes the price information, the data would be much more attractive.
The main actors in a given economy are households, corporations and governments. Society's primary objective is to make households happier.
For instance, corporations are making lots of profit and increasing their retained earnings alone cannot make a society happier. Likewise, the government amassing more money (through increased taxes and such) cannot make a society happy. What's most crucial then, is the household.
One way we measure "happiness" in households is through assessing incomes. Increase in incomes is a definite positive. Yet, there is something off about simply examining incomes. What's important is not whether someone's incomes and savings increased, but whether at the end of the day, if one's happiness is increased by spending their money. Of course, money cannot buy all kinds of happiness. Many factors other than purchasing things with money deeply contribute to our well-being.
However, the prospect that consumption makes us happy is paramount in our capitalist society. Hence, consumption statistics may be considered one kind of measure for happiness.
The government and the central bank acknowledge similarly, which is why tracking consumption holds special meaning for any country. Of course GDP is another big measure of a country's economy, but consumption accounts for 60%, making it the largest piece of the pie.
This is why consumption is crucial in determining whether an economy is performing well or not.
The most widely used statistics are the household budget survey published by the Statistics Bureau.
This asks over 10,000 households to keep an account book to research how their expenditure. The survey is also used as raw data to determine the size of consumption within Japan's GDP. I hear however, that it's quite difficult to find people that are willing to participate in this survey, and I presume this is because most people are too busy in their everyday lives to cooperate with the government's research. The large burden on survey participants is one issue that should be solved.
Besides household budgets, there's also the current survey of commerce published by the Ministry of Economy, Trade and Industry. While the household budget survey researches consumption using household budgets,the current survey of commerce takes its numbers from stores that sell to households. Although purchasing and selling correspond to each other closely, there are many instances in which statistics show discrepancies.
"Bakugai", or explosive-buying as they say has supposedly waned, but the trend only appears in the seller side's current survey of commerce - not in the buyer side's household budget surveys.
Regardless, both statistics have proven to be too slow in publication, meaning that statistical data is lagging behind in our lives and businesses today.
A new committee on revising statistics was recently set up in January, 2017. Led by Chief Cabinet Secretary Suga, the committee is currently working on extensively renewing government statistics. I am also a member of the group, but I would say it is a very large-scale project. There are two main reasons for such moves.
The first is stagnance of Japan's economic growth. Compared to its heyday, Japan's growth has been sluggish in the last two decades. Nowadays its numbers only hover closely around zero. Growth is negative when the economy is unwell, and becomes positive if it recovers, but either way statistics never are far from zero. Although the differences may seem insignificant, growth being positive or negative mean very different things. If the growth rate is negative, the government or the central bank must churn out policies to revive the economy. Consequently, it becomes necessary to measure the growth rate with precision, so as to distinguish between positive and negative. If this was a while back when Japan's growth rate was at 10% or so, a few decimals wouldn't have mattered so much. In that case you do not need to care about precision so much. On the other hand, in the economy with low growth rate, we need so precise measurement, i.e. economic statistics. But this is a widespread issue among most developed nations; Japan is no exception.
Secondly, demands for real-time statistics have gone up significantly.
Massive developments in information technology have enabled us to speed up everything in our lives and businesses, but in turn we become pressed to make decisions more quickly too. Ironically, Japan's statistics system seems as if it were stopped in time. Rules that were decided over seven decades ago still prevail.
It is inevitable to maintain a certain level or conservativeness to guarantee consistency, but current government statistics are far too obsolete to be used today.
Corporate decisions cannot be made with statistics that are two-months old; they are of the past and are hardly useful. The government recognizes this issue, and is making moves to speed up statistical procedures.
As I said earlier, I believe that the project is something that can responds to today's needs for faster and more accurate statistics.
My colleague, Professor Noriyuki Yanagawa and I have been proposing what we call "privatization of statistics". Up until now, the state had a monopoly over statistics, whereby data compilement, analysis, and publication were all done by the government. I presume this was because generating statistics was too costly for the private sector, leaving the government the only capable organization.
Today however, the private sector holds both data and the technology to analyze statistics. This is why we think that the they should be opened to the public. I personally see this project as the first step in privatizing statistics.
That being said, there are two things that I would ask JCB and Nowcast to keep in mind upon the service's release.
First of all, immense care must be placed on extracting information from credit card data. Collecting and analyzing the payment details of individuals would obviously help illustrate a better picture of consumption trends in the entire Japanese economy, but that is still inadequate.
For example, let's pretend that overall payments increased by 10% one month. Maybe every user had increased their credit card usage by 10%, but it also could be that big spenders increased their use by 20% and that raised the average instead.
If the former is true we may say that everyone is "happier", but if the latter is true, perhaps it is only a certain portion of people that are "happier" than before. Moreover, the latter would indicate that the consumption gap is increasing - an undesirable phenomenon for society.
This is merely one example, but what I want to emphasize is that some important detail may be lost if the data is compiled haphazardly. Because the service deals with an invaluable sort of big data, I hope that good technology is developed to better extract information.
Secondly, I would like for the companies to be careful of the some of the data's partialities.
Although government survey samples lack in size, it ensures diversity by collecting data from different kinds of people. As opposed to that, big data is derived from businesses carrying out their operations, so there is the possibility that the data is somewhat partial. If we leave this problem, the output data doesn't express true circumstance of Japanese economy. Credit cards are no exception, because you can't tell the what the consumption activities of people who do not have and use credit cards are.
There are a couple of ways in which these "sample biases" may be resolved, but unfortunately no panacea exists for now.
Both points that I have just raised consist of difficult technical issues. Nonetheless, I am confident in Nowcast's abilities and experiences in big data analytics to contribute in solving these challenging issues. The road is tough, but I personally am very excited for the fruits of this service and would like to actively offer my own experiences and knowledge in helping them do so.
The new committee on revising statistics is also attempting to revamp statistics through big data. I hope that this service will become an example to follow.
See Data