Economic Analysis Series No.208THE ECONOMIC ANALYSIS

October, 2023

(Articles) Indexes of Business Conditions: Hiroshi YOSHIKAWA
Characteristics of Economic Forecasts by Private Research Institutes: Taro SAITO
The Impact of Recent Economic Structural Changes on Business Fluctuations: Minoru MASUJIMA
A New Index to Capture Business Conditions (Coincident Index): Yasuhisa INO and Susumu KUWAHARA
Verification of Provisional Outlier for Seasonal Adjustment in Quarterly Estimates of GDP: Tadashi GONDA and Yohei MATSUMURA
Opportunities and Challenges in the Use of Alternative Data: Tsutomu WATANABE, Yuki OMORI and Sho YOKOYAMA
Business Cycle Analysis Using Textual Information and Machine Learning: Mototsugu SHINTANI
GDP Nowcasting and Economic Assessment: Applying Real-time GDP Forecasting to Early Assessment of Ongoing Economic Activities: Satoshi URASAWA
Development of a New Multivariate Time Series Analysis Method and Its Application to Business Cycles: Hiroshi IYETOMI
Possibility to Utilize Alternative Data for Analyzing Macro-economic Trend: Decomposing Supply and Demand Driven Inflation by Using POS Data: Yuko UENO and Ryuga KITAGUCHI
Use Cases of New Economic Analysis Methods in Europe: Shuto KOJIMA
Various Types of Alternative Data and Its Use Cases in the U.S.: Keita MIYANO
Efforts on Economic Forecasting: Growing Uncertainty and the Role of Consensus Forecasting: Yukiko ITO and Eriko TAKAHASHI
(Document) ESRI International Conference 2023 “Demographic change and Economic growth”: Economic and Social Research Institute

The full text is written in Japanese.

(Abstract)

Characteristics of Economic Forecasts by Private Research Institutes

By Taro SAITO

One of the important roles of private research institutes in analyzing economic trends and forecasting the economy is to produce highly accurate forecasts of economic growth rate and price inflation rate. The average error in forecasting Real GDP growth rate by private research institutes for 43 years from FY 1980 to FY 2022 is 1.33% (mean absolute error). The actual values are often out of the forecast range (maximum to minimum value of the forecasts), but compared to the government economic outlook, the forecast errors of private research institutes are smaller. Although judgments regarding economic turning points tend to be delayed, there is a tendency for forecasts to move up during periods of economic expansion and move down during periods of economic recession. Therefore, the direction of revision of forecasts can be useful information in determining the turning point of the economy.

With changes in socioeconomic conditions, it has become necessary to develop forecasts in a short period of time after accurately grasping the current economic trends. In addition, the forecasting period has been extended earlier than in the past, and the forecast period has become longer.

In recent years, the need for a more rapid grasp of current economic trends has increased due to an increase in the number of cases in which the economy has fluctuated significantly in a short period of time, and analysis using conventional macroeconomic statistics alone is no longer sufficient to deal with this situation. Against this backdrop, alternative data has become increasingly useful, but there are also data constraints that pose many challenges for its use in continuous economic analysis. If the Cabinet Office were to publish monthly GDP, it could contribute to a quick and accurate assessment of economic trends.

JEL Classification Codes: A11, E32, E37
Keywords: economic forecast, forecast accuracy, monthly GDP

The Impact of Recent Economic Structural Changes on Business Fluctuations

By Minoru MASUJIMA

Over the long term, the trend of the economy has been declining and inventory investment and capital investment have become less volatile. This is due to changes in the structure of the economy, such as the shift of the economy toward services, globalization, digital transformation, and increased investment in research and development. While there was a divergence between the Business Cycle Index (coincident index) and real GDP developments near the peak of the 16th cycle (October 2018), this was largely due to the fact that the pace of decline in external demand was not rapid enough to cause the economy to cool down all at once. In addition, in the labor market, more people, especially women and the elderly, entered the labor market amid the population decline, supporting the employment income environment. After the Corona shock, the momentum of wage increases has strengthened. It is necessary to pay close attention to how structural changes in the labor market will proceed and how they will affect the business cycle.

Given these changes in the characteristics of business cycles and the underlying changes in the structure of the economy, in order to properly assess the business cycle, it is necessary, first to continue to be sensitive to turning points in the business cycle, but, second, to also monitor a broader range of indicators without overemphasizing industrial production, and, third, to also pay some attention to the development of real GDP as well from the perspective of assessing overall trends in the economy.

JEL Classification Codes: C82, E01, E32
Keywords: Business Cycle Index, inventory cycle, capital investment cycle

A New Index to Capture Business Conditions (Coincident Index)

By Yasuhisa INO and Susumu KUWAHARA

The basic structure of the Indexes of Business Conditions, which have been published monthly since August 1960, has been maintained to this day. However, the environment surrounding the Japanese economy has changed, and the need for a fundamental review has been pointed out. In July 2022, the Economic and Social Research Institute, Cabinet Office, Government of Japan released a “new index to capture business conditions (coincident index)” (the “new coincident index”) as the result of a back-to-basics study on the ideal form of business condition indexes. The “new coincident index” was developed by rethinking the conventional way of understanding the economy, which assumes common fluctuations across the economy, and, instead, focusing on fluctuations in the aggregate volume of economic activity, reflecting the decoupling between the movement of goods and services as the economy has become more service-oriented and software-oriented. The basic approach is to combine a wide range of indicators, to capture the three aspects of production, distribution, and spending, and to capture the autonomous economic activities of the private sector. The “new coincident index” fluctuates in the same direction as the current Indexes of Business Conditions (coincident index), but its amplitude is slightly smaller and its movements are more similar to those of real GDP than those of the current Indexes of Business Conditions (coincident index). Whether or not the "new coincident index" can be used as an appropriate indicator of the economy should be thoroughly examined as more data are accumulated in the future.

JEL Classification Codes: C82, E01, E32
Keywords: indexes of business conditions, coincidence index, equivalence of three approaches, aggregate volume, common fluctuations, the Committee for Business Cycle Indicators

Verification of Provisional Outlier for Seasonal Adjustment in Quarterly Estimates of GDP

By Tadashi GONDA and Yohei MATSUMURA

Seasonally adjusted series of Gross Domestic Product (GDP) and its breakdown are revised retrospectively for the entire period every time Quarterly Estimates of GDP (QE) is released, since the seasonal adjustment is conducted with the latest estimates of original series updated. With respect to such revisions, when some enormous economic shock occurs, it has been indicated that the failure to apply a provisional outlier at the preliminary stage can lead to continuous revisions of past growth rates each time subsequent QE is released. Therefore, in this paper, we will focus on two periods: (A) the financial crisis triggered by the collapse of Lehman Brothers and the subsequent recovery period (January-March 2008 to July-September 2009), and (B) the period in which the impact of the enormous economic shock seems to be less significant (January-March 2017 to July-September 2018), we examined whether setting a provisional outlier at the preliminary stage, without waiting for the annual estimates, would suppress revisions to past figures. It is expected that this verification will help to prevent excessive revision of past figures caused by seasonal adjustment and improve the accuracy of the preliminary estimates.

JEL Classification Codes: C32, C52, C82, E01
Keywords: Quarterly Estimates of GDP, Seasonal Adjustment, Outlier Detection

Opportunities and Challenges in the Use of Alternative Data

By Tsutomu WATANABE, Yuki OMORI and Sho YOKOYAMA

The term “alternative data” has become a term frequently seen in newspapers and magazines. Until now, the data used to understand the current state of the economy consisted of government statistics such as GDP data and corporate data such as corporate financial statements. This data was, for example, the source of information investors used to buy and sell securities in financial market. These types of data are called “traditional” data. Now, alternatives to such traditional data called “alternative data” have emerged. Examples include scanner data from supermarket checkout counters, credit card transaction data, and location data from smartphones. While alternative data already existed prior to the pandemic, it was only used by a few financial institutions and investors, and awareness of it was not very high. However, it came into widespread use with the onset of the pandemic, and its use has continued to grow after the pandemic. This chapter first looks at the current state of and future prospects for alternative data and discusses some of the challenges that may arise as its use expands. Then, two examples of the use of alternative data are presented. The first is an analysis of changes in consumer behavior during the pandemic using smartphone location data. The second is an analysis of changes in the network structure between consumers and stores using credit card transaction data.

JEL Classification Codes: C80, C81, C82
Keywords: Alternative data, non-traditional data, smartphone location data, credit card transaction data

Business Cycle Analysis Using Textual Information and Machine Learning

By Mototsugu SHINTANI

This paper provides an overview of techniques for utilizing textual information to compute business cycle indices available prior to official government statistics, along with their applications to the Japanese economy. In the review, we classify the methods into two categories: the lexicon approach, which focuses on term frequencies of predefined keywords, and the machine learning approach, which utilizes the text data to learn language models. In the lexicon approach, classical sentiment analysis remains highly valuable due to its low computational cost and economic interpretation. However, when computing the index, utilizing a domain-specific polarity dictionary for macroeconomic analysis, and conducting careful preprocessing of text data are essential. On the other hand, in terms of forecast accuracy, machine learning approach that can effectively capture textual information, including context, is preferable. We can anticipate a rising trend in the utilization of new language models for conducting business cycle analysis. At the same time, given the rapid evolution of large language models in recent years, ensuring the continuity of reporting the same index, as well as retrospective estimation when models are replaced, is important.

JEL Classification Codes: C53, C55, E17
Keywords: Natural Language Processing, Nowcast, Sentiment Analysis

GDP Nowcasting and Economic Assessment: Applying Real-time GDP Forecasting to Early Assessment of Ongoing Economic Activities

By Satoshi URASAWA

This study discusses the use of real-time GDP forecasting (“nowcasting”) in assessing the state of the economy by examining the relationship between the basic assessment of economic developments in the government’s Monthly Economic Report and the assessment of economic developments obtained in nowcasting.

The results, while based on a limited number of observations, suggest that the government's basic assessment is likely to be revised in a similar direction as the GDP nowcasting estimates in situations where the nowcasting GDP estimates are successively revised in one particular direction by a certain degree. The finding implies that GDP nowcasting can provide useful information for the government to take into account in its economic assessment.

JEL Classification Codes: E37
Keywords: Japan, Monthly Economic Report, Real-time data

Development of a New Multivariate Time Series Analysis Method and Its Application to Business Cycles

By Hiroshi IYETOMI

We regard macroeconomic phenomena such as business cycles and price fluctuations as collective movements of individual agents and approach them empirically using concepts and methods from statistical physics. Following this approach, we have recently developed a Complex Hilbert Principal Component Analysis (CHPCA) method for detecting collective movements in real data sets, which has the same computational complexity as the PCA method based on real data and can be used to easily analyze the dynamic correlation structure among multivariate variables. This analysis method can be used to mechanically extract the lead-lag relationship among basic business trend indicators. That is, there is a possibility that the selection of basic indicators needed to construct leading, coincident, and lagging business trend indices can be made in a more objective way. We also report the results of our examination of the leading nature of the Business Watchers Survey data with respect to the main body of the economy using the CHPCA method. We find that the data are very promising as leading indicators.

JEL Classification Codes: C10, C38, E32
Keywords: econophysics, business cycle, complex Hilbert principal component analysis, business watchers survey

Possibility to Utilize Alternative Data for Analyzing Macro-economic Trend: Decomposing Supply and Demand Driven Inflation by Using POS Data

By Yuko UENO and Ryuga KITAGUCHI

High-frequency real-time data (henceforth “alternative data”) has recently been gathering much attention at home and abroad as an effective tool for the analysis of the most recent economic trend. This paper intents to quantify the degree to which either demand or supply-driven contributions to the inflation derived from the POS data, which is a leading example of alternative data. Our results show that supply is mainly driving inflation after 2022, and that these supply-driven contributions have been widely spreading among various grocery items to affect household consumption behaviors. The use of alternative data is still in its development stage, leaving much to be solved to replace the official statistics for analyzing macro-economic trend. Economists, including those at government, need to explore further possibility to utilize alternative data, given its advantages.

JEL Classification Codes: D12, D40, E21, E31
Keywords: Alternative Data, Assessment of Macro-economic Trend, Inflation, Demand or Supply Factor

Use Cases of New Economic Analysis Methods in Europe

By Shuto KOJIMA

This paper introduces new economic analysis methods using big data and non-traditional data from Europe, taking up examples of their use in the European Commission, the European Central Bank, and the German Federal Government. It includes the utilization of data that has become available due to the progress of digitization, and various efforts to quickly grasp changes in the rapidly evolving economic situation.

JEL Classification Codes: E32, E60
Keywords: Business Cycles, Statistics, Big Data, Alternative Data

Various Types of Alternative Data and Its Use Cases in the U.S.

By Keita MIYANO

This paper will provide an overview of alternative data in the U.S., including nowcasting and various analytical efforts using such data, for both public and private organizations. While public institutions tend to focus on nowcasting of overall economic trends such as GDP, many private institutions are attempting to conduct more micro-level analyses such as people's consumption behavior.

Since the use of alternative data for economic analysis is still a relatively new method, there are concerns about the proper handling of data from the perspective of privacy protection, which may lead to further regulations and restrictions on data use in the future. In addition, a decline in the accuracy of analyses using alternative data has also been pointed out since COVID-19 pandemic happened.

JEL Classification Codes: E29, E59
Keywords: alternative data, nowcast

Efforts on Economic Forecasting: Growing Uncertainty and the Role of Consensus Forecasting

By Yukiko ITO and Eriko TAKAHASHI

The purpose of this paper is to examine the significance of the “JCER ESP Forecast” survey as a private sector consensus forecast, amid heightened uncertainty and growing need for economic forecasting. The development of information and communication technology (ICT) has made it possible to instantly obtain large amounts of information about the economy and construct forecasts. At the same time, along with the development of ICT, uncertainty has elevated, and economic forecasting has become increasingly important. In this context, the role of the “JCER ESP Forecast” is to first provide reference criteria for the economic outlook, second to provide information clarifying the formation of market expectations, and third, to present the diverse views of private sector economists on economic risks and desirable policies. Uncertainty is inherent in forecasting; “JCER ESP Forecast” provides the mean distribution of probabilities as information on uncertainty. Starting in 2022, measurements indicating the direction of the consensus forecast bias, and the 50% probability that the actual value will fall within a specific interval became available to the public. We use ESP Forecast data to calculate economists’ average probability of peaks and troughs of the business cycle; in the case of economic peaks, the index becomes significant when it exceeds 60% and approaches 90%, and for troughs, when it exceeds 70%.

JEL Classification Codes: E37, E3
Keywords: economic forecast, consensus forecast, ESP forecast

(Document)
ESRI International Conference 2023 “Demographic change and Economic growth”

Economic and Social Research Institute

Page top