Explanation of the Tankan (Short-Term Economic Survey of Enterprises in Japan) — Bank of Japan, February 2023
Summary
This February 2023 document from the Research and Statistics Department of the Bank of Japan explains the Tankan, the Short-Term Economic Survey of Enterprises in Japan. It says the Tankan is a statistical survey conducted under the Statistics Law (Law No. 53 of 2007) to give an accurate picture of business trends and support monetary policy. Its target population is private enterprises in Japan with capital of 20 million yen or more, excluding financial institutions and business consultants and pure holding companies, drawn from a Ministry of Internal Affairs and Communications database. The explanation covers survey items, frequency, publication, data collection, the judgment and quantitative surveys, the inflation outlook of enterprises, handling of missing values and sample design.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
- Issuer
- Source documents
- Document type
- Bank Financing For Smes In Times Of Crisis When Whatever It Takes Confronts Black Swans
- Case
- Bank Financing For Smes In Times Of Crisis When Whatever It Takes Confronts Black Swans
Cite this document
“Explanation of the Tankan (Short-Term Economic Survey of Enterprises in Japan) — Bank of Japan, February 2023.”
Original at doi.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/bank-financing-for-smes-in-times-of-crisis-when-whatever-it-takes-confronts-black-swans/
, SHA-256 5ba7fd3901bdfa187270ad1aa806b31d62ce2c4b9f005390d9683fbe73d352db.Full text
TANKAN
(Short-Term Economic Survey of Enterprises in Japan)
Explanation
February 2023
Research and Statistics Department
Bank of Japan
Contents
I. OVERVIEW
1. Conducting Body and Purpose ······················································ 1
2. Coverage ················································································ 1
3. Survey Items ··········································································· 4
4. Survey Frequency ····································································· 5
5. Publication ·············································································· 6
6. Notes for Users ········································································ 12
II. METHOD
1. Data Collection ········································································ 13
2. Survey Content········································································· 13
3. Survey Workflow ······································································ 20
III. CALCULATION AND ESTIMATION
1. Judgment Survey ······································································ 22
2. Quantitative Survey (Annual Projections, Number of New Graduates Hired
and Overseas Business Activities) ·················································· 22
3. Inflation Outlook of Enterprises ····················································· 24
4. Handling of Missing Values ·························································· 24
5. Treatment of Outliers ································································· 25
IV. SAMPLE DESIGN
1. Population ·············································································· 27
2. Basic Concept of Sample Enterprises ·············································· 27
3. Criteria for Statistical Accuracy ····················································· 28
4. Sample Design Procedure ···························································· 30
5. Sample Design of Survey of Financial Institutions ······························· 31
6. Sample Design of Survey of Business Consultants & Pure
Holding Companies ··································································· 33
7. Sample Design of Survey on Overseas Business Activities ····················· 33
8. Sample Maintenance ·································································· 34
Appendix 1 Method of Outlier Detection
Appendix 2 Standard Error Ratios
Appendix 3 Stratification Method of Sample Design
Appendix 4 Chi-square Goodness of Fit Test
i
I. OVERVIEW
1. Conducting Body and Purpose
The Tankan (Short-Term Economic Survey of Enterprises in Japan) is a statistical survey
conducted by the Bank of Japan in accordance with the Statistics Law (Law No. 53 of
2007). It aims to provide an accurate picture of business trends of enterprises in Japan,
thereby contributing to the appropriate implementation of monetary policy.
2. Coverage
(1) Tankan
The Tankan uses a sample survey framework where sample enterprises are selected from
a population in accordance with statistical theory.
The target population of the Tankan are private enterprises (excluding "financial
institutions" and "business consultants & pure holding companies") in Japan with capital
of 20 million yen or more.1
The private enterprises which comprise the survey population of the Tankan are those in
Japan with capital of 20 million yen or more (excluding "financial institutions" and
"business consultants & pure holding companies") and are drawn from the Establishment
Frame Database of the Ministry of Internal Affairs and Communications. It should be
noted that some industries which may have a weak connection to economic conditions
such as "scientific and development research institutes" and "medical and other health
services" are excluded from both the target and survey populations.
1
A target population is the entire group that forms the focus of a particular measurement. However,
there are no exhaustive lists of private enterprises in Japan with capital of 20 million yen or more
corresponding to the target population of the Tankan. This means, in practice, the target population
is a theoretical concept. On the other hand, a survey population is the group that can actually be
measured using information obtained in the actual survey.
1
Sample enterprises are selected from the survey population using stratified sampling to
satisfy established criteria such as that of statistical accuracy; for example, the standard
error ratio of sales is within the target range (3 percent for manufacturing and 5 percent
for nonmanufacturing). The Bank asks newly selected enterprises for their consent to be
added to the Tankan survey sample in advance. Afterwards, those enterprises which have
given consent are added to the sample. The number of sample enterprises and data on
statistical accuracy are available in the accompanying release materials for each survey.
(2) Survey of Financial Institutions
Financial institutions are also surveyed to supplement the Tankan.2
The target population are private financial institutions in Japan with 10 employees or
more.
The private financial institutions which comprise the survey population are those in Japan
with 10 employees or more and are drawn from the Establishment Frame Database of the
Ministry of Internal Affairs and Communications.
Sample institutions are selected from the survey population using stratified sampling to
satisfy established criteria such as that of statistical accuracy; for example, the standard
error ratio of fixed and software investment (excluding land purchasing expenses) for
overall financial institutions is 10 percent or less. The Bank asks newly selected
institutions for their consent to be added to the survey sample in advance. Afterwards,
those institutions which have given consent are added to the sample. The number of
sample institutions and data on statistical accuracy are available in the accompanying
release materials for each survey.
2
The survey on fixed investment of financial institutions was previously conducted to supplement
the Principal Enterprises Tankan from the November 1989 survey until the December 2003 survey.
The Survey of Financial Institutions expanded its coverage and survey items in the March 2004
survey, and has been conducted in order to supplement the Tankan.
2
(3) Survey of Business Consultants & Pure Holding Companies
Business consultants and pure holding companies are also surveyed to supplement the
Tankan.
The target population are private enterprises that are "business consultants & pure holding
companies" in Japan with capital of 20 million yen or more.
The private enterprises which comprise the survey population are "business consultants
& pure holding companies" in Japan with capital of 20 million yen or more and are drawn
from the Establishment Frame Database of the Ministry of Internal Affairs and
Communications.
Sample enterprises are selected from the survey population using stratified sampling to
satisfy established criteria such as that of statistical accuracy; for example, the sample
distributions of sales could be deemed not to deviate from survey population distributions.
The Bank asks newly selected enterprises for their consent to be added to the survey
sample in advance. Afterwards, those enterprises which have given consent are added to
the sample. The number of sample enterprises and data on statistical accuracy are
available in the accompanying release materials for each survey.
(4) Survey on Overseas Business Activities
The Survey on Overseas Business Activities is conducted to supplement the Tankan.
The target population are private parent enterprises (excluding "financial institutions") in
Japan, with capital of 1 billion yen or more, at the top of a consolidated business group
conducting overseas business activities.
The private parent enterprises which comprise the survey population are those in Japan
with capital of 1 billion yen or more, at the top of a consolidated business group with
sales to overseas external customers or overseas subsidiaries/branches (excluding
"financial institutions") and are drawn from the Establishment Frame Database of the
Ministry of Internal Affairs and Communications. It should be noted that some industries
which may have a weak connection to economic conditions such as "scientific and
3
development research institutes" and "medical and other health services" are excluded
from both the target and survey populations.
Sample enterprises are selected from the survey population using stratified sampling to
satisfy established criteria such as that of statistical accuracy; for example, the standard
error ratio of total sales is within the target range (10 percent for both manufacturing and
nonmanufacturing). The Bank asks newly selected enterprises for their consent to be
added to the survey sample in advance. Afterwards, those enterprises which have given
consent are added to the sample. The number of sample enterprises and data on statistical
accuracy are available in the accompanying release materials for each survey.
(5) Revision of Sample Enterprises
In principle, the Bank revises the Tankan sample enterprises once every two or three years.
Surveys which supplement the Tankan (Survey of Financial Institutions, Survey of
Business Consultants & Pure Holding Companies and Survey on Overseas Business
Activities) are also revised at the same time as the Tankan.
Except for these revisions, sample enterprises are basically fixed; however the Bank
examines statistical accuracy regularly (once a year in principle) and if necessary, adds
new sample enterprises, to prevent the lowering of statistical accuracy caused by a
decrease in the number of sample enterprises due to bankruptcies, mergers, and other
factors.
3. Survey Items
The following items in (1) ~ (3) are surveyed on the basis of non-consolidated accounting.
Items in (4) are surveyed on a consolidated basis.
(1) Tankan（4 categories, 26 items）
Business conditions; domestic supply and demand conditions for
products and services; overseas supply and demand conditions for
Judgment survey
products; inventory level of finished goods and merchandise;
wholesalers' inventory level; production capacity; employment
4
conditions; financial position; lending attitude of financial institutions;
conditions for CP issuance; change in interest rate on loans; change in
output prices; change in input prices
Sales; exports; exchange rates (Yen/U.S. dollar); exchange rates
Annual projections (Yen/Euro); current profits; net income; fixed investment; land
purchasing expenses; software investment; R&D investment
Inflation outlook of
Outlook for output prices; outlook for general prices
enterprises
Number of new
Number of new graduates hired (surveyed in June and December only)
graduates hired
(2) Survey of Financial Institutions (3 categories, 8 items)
Judgment survey Business conditions; production capacity; employment conditions
Annual projections Fixed investment; land purchasing expenses; software investment; R&D
investment
Number of new
Number of new graduates hired (surveyed in June and December only)
graduates hired
(3) Survey of Business Consultants & Pure Holding Companies (1
categories, 4 items)
Annual projections Fixed investment; land purchasing expenses; software investment; R&D
investment
(4) Survey on Overseas Business Activities (2 categories, 7 items)
Annual projections Exchange rates (Yen/U.S. dollar); exchange rates (Yen/Euro)
Overseas business Consolidated Sales; overseas sales; consolidated current profits;
activities consolidated fixed investment; overseas fixed investment
4. Survey Frequency
The survey is conducted on a quarterly basis, in March, June, September, and December
each year.
5
5. Publication
(1) Release
Each survey result taken in March, June, September, and December is released at the
beginning of April, July, October, and mid-December, respectively, over two business
days.
The release schedule is available on the Bank's website.
Release dates for the next four weeks are announced. Updated
Release Schedule
every Friday.
Outline of Statistics and Release dates for the next 12 months are announced. Updated
Statistical Release Schedule around the end of June and December.
When released data requires correction, corrected data, in principle, is compiled and
released as quickly as possible. However, in some cases data would not generally be
corrected, for example, where responding enterprises revise their figures after the release.
(2) Publications
The following materials are available on the Bank's website.
Release Timing Name Content
8:50 a.m. on the Summary Main figures of the survey results
release day Outline Selected figures from the Summary
8:50 a.m. on the
following day of The Comprehensive Data Set Detailed figures of the survey results
release
Long-term time series data of the Tankan are available in the "BOJ Time-Series Data
Search."
The sample form, guidelines for filling in the form (Japanese only), and an example
answer sheet (Japanese only) are available on the Bank's website.
6
(3) Release Data Series
A. Tankan
The Bank releases the following data series by industry and enterprise size.
Diffusion index (DI); percentage share of the number of
Judgment survey respondents choosing each alternative; number of reporting
enterprises
Annual projections
Year-to-year percent change; revision rate from the previous
Sales and other items*1 survey; nominal yen-denominated figures; number of reporting
enterprises
Predicted exchange Yen per U.S. dollar; Yen per Euro; number of reporting
rates*2 enterprises
Ratio of current profit
Nominal figures; changes from the previous survey
to sales
Number of enterprises
reporting increase or Nominal figures; number of reporting enterprises
decrease in profits*3
Number of enterprises
Nominal figures
reporting deficits*3
Percentage share of the number of respondents choosing each
Inflation outlook of
alternative; number of reporting enterprises; (reference) average
enterprises
of enterprises' inflation outlook
Number of new graduates Year-to-year percent change; revision rate from the previous
hired survey; nominal figures; number of reporting enterprises
*1 This item corresponds to "domestic sales," "software and fixed investment excluding land
purchasing expenses," and "software, R&D, and fixed investment excluding land purchasing
expenses" in addition to all items described in 3.(1) excluding "exchange rates."
*2 This item corresponds to "exchange rates" described in 3.(1).
*3 Data series are released only for manufacturing, nonmanufacturing, and all industries with regard
to industry classification.
Industry classification
Manufacturing is comprised of 17 industries and nonmanufacturing is comprised of 14
industries based on the Japan Standard Industrial Classification released by the Ministry
of Internal Affairs and Communications.
7
Tankan Japan Standard Industrial Classification (code)
Textiles Manufacture of textile products (11)
Manufacture of lumber and wood products, except furniture (12), Manufacture of furniture and
Lumber & Wood products
Basic materials
fixtures (13)
Pulp & Paper Manufacture of pulp, paper and paper products (14)
Chemicals Manufacture of chemical and allied products (16)
Petroleum & Coal products Manufacture of petroleum and coal products (17)
Ceramics, Stone & Clay Manufacture of ceramic, stone and clay products (21)
Iron & Steel Manufacture of iron and steel (22)
Nonferrous metals Manufacture of non-ferrous metals and products (23)
Food & Beverages Manufacture of food (09), Manufacture of beverages, tobacco and feed (10)
Processed metals Manufacture of fabricated metal products (24)
Manufacturing
General-purpose
General-purpose, Manufacture of general-purpose machinery (25)
machinery
Production & Production
Manufacture of production machinery (26)
Business oriented machinery
Business oriented
machinery Manufacture of business oriented machinery (27)
Processing
machinery
Electronic parts, devices and electronic circuits (28), Manufacture of electrical machinery, equipment
Electrical machinery
and supplies (29), Manufacture of information and communication electronics equipment (30)
Shipbuilding, Heavy
machinery & Other
Transportation Manufacture of transportation equipment (31) (except Motor vehicles, parts and accessories (311))
transportation
machinery machinery
Motor vehicles Motor vehicles, parts and accessories (311)
Printing and allied industries (15), Manufacture of plastic products (18), Manufacture of rubber
Other manufacturing products (19), Manufacture of leather tanning, leather products and fur skins (20), Miscellaneous
manufacturing industries (32)
All industries
Construction work, general including public and private construction work (06), Construction work by
Construction
specialist contractor, except equipment installation work (07), Equipment installation work (08)
Real estate Real estate agencies (68), Real estate lessors and managers (69)
Real estate, Goods
rental & Leasing Goods rental & Leasing Goods rental and leasing (70)
Wholesale trade, general merchandise (50), Wholesale trade (textile and apparel) (51), Wholesale
Wholesaling trade (food and beverages) (52), Wholesale trade (building materials, minerals and metals, etc) (53),
Wholesaling & Wholesale trade (machinery and equipment) (54), Miscellaneous wholesale trade (55)
Retailing Retail trade, general merchandise (56), Retail trade (woven fabrics, apparel, apparel accessories and
Retailing notions) (57), Retail trade (food and beverage) (58), Retail trade (machinery and equipment) (59),
Miscellaneous retail trade (60), Nonstore retailers (61)
Railway transport (42), Road passenger transport (43), Road freight transport (44), Water transport
Transport & Postal activities (45), Air transport (46), Warehousing (47), Services incidental to transport (48), Postal services,
Nonmanufacturing
including mail delivery (49), Postal services (86)
Communications Communications (37)
Information Information services Information services (39)
communication Other information Broadcasting (38), Services incidental to Internet (40), Video picture information, sound information,
communication character information production and distribution (41)
Production, transmission and distribution of electricity (33), Production and distribution of gas (34),
Electric & Gas utilities
Heat supply (35)
Design services (726), Advertising (73), Technical services, n.e.c.(74) (except Veterinary services
(741)), Industrial waste disposal business (882), Automobile maintenance services (89), Machine, etc.
Services for businesses
repair services (90), Employment and worker dispatching services (91), Miscellaneous business
services (92)
Laundry, beauty and bath services (78), Miscellaneous living-related and personal services (79),
Services for amusement and recreation (80), Specialized training colleges and miscellaneous schools
Services for individuals (817), Supplementary tutorial schools (823), Instruction services for arts, culture and technicals (824),
Welfare services for the aged and care services (854), Miscellaneous social insurance, social welfare
and care services (859)
Accommodations, Eating & Drinking services Accommodations (75), Eating and drinking places (76), Food take out and delivery services (77)
Mining & Quarrying of stone and gravel Mining and quarrying of stone and gravel (05)
Note: Each industry is classified according to the small and middle classifications of the Japan Standard Industrial
8
Classification revised in October 2013.
Enterprise size classification
Enterprise size is defined by amount of capital, and divided into three categories.
Enterprises with capital of 1 billion yen or more are large enterprises, those with capital
from 100 million yen to less than 1 billion yen are medium-sized enterprises, and those
with capital from 20 million yen to less than 100 million yen are small enterprises.3
B. Survey of Financial Institutions
The Bank releases the following data series by sector.
DI; percentage share of the number of respondents choosing each
Judgment survey
alternative; number of reporting enterprises
Year-to-year percent change; revision rate from the previous survey;
Annual projections*
nominal yen-denominated figures; number of reporting enterprises
Number of new Year-to-year percent change; revision rate from the previous survey;
graduates hired nominal figures; number of reporting enterprises
* This item corresponds to "software and fixed investment excluding land purchasing expenses," and
"software, R&D, and fixed investment excluding land purchasing expenses" in addition to all items
described in 3.(2).
Sector classification
Financial institutions are comprised of five sectors based on the Japan Standard Industrial
Classification released by the Ministry of Internal Affairs and Communications.
Tankan Japan Standard Industrial Classification (code)
Financial institutions
Banks Banking (62) (except Central bank (621))
Financial institutions for
Financial institutions for cooperative organizations (63)
cooperative organizations
Financial products Financial products transaction dealers and futures commodity
transaction dealers transaction dealers (65)
Insurance institutions, including insurance agents, brokers and
Insurance companies
services (67)
3
Since the March 2004 survey, the amount of capital has been used to define enterprise size instead
of the number of employees, as used previously.
9
Non-deposit money Non-deposit money corporations, including lending and credit
corporations card business (64)
Note: Each industry is classified according to the small and middle classifications of the Japan Standard Industrial
Classification revised in October 2013.
In addition, the Bank releases data on annual projections and number of new graduates
hired for "all industries including financial institutions" and "nonmanufacturing including
financial institutions," which are calculated as the aggregate amount of the Tankan and
the Survey of Financial Institutions.
C. Survey of Business Consultants & Pure Holding Companies
The Bank releases the following data series.
Year-to-year percent change; revision rate from the previous survey;
Annual projections*
nominal yen-denominated figures; number of reporting enterprises
* This item corresponds to "software and fixed investment excluding land purchasing expenses," and
"software, R&D, and fixed investment excluding land purchasing expenses" in addition to all items
described in 3.(3).
"Business consultants & pure holding companies" corresponds to the "728 Business
consultants and pure holding companies" in the Japan Standard Industrial Classification.
In addition, the Bank releases data on annual projections for "all industries including
financial institutions and business consultants & pure holding companies" and
"nonmanufacturing including financial institutions and business consultants & pure
holding companies," which are calculated as the aggregate amount of the Tankan, the
Survey of Financial Institutions and the Survey of Business Consultants & Pure Holding
Companies.
D. Survey on Overseas Business Activities
The Bank releases the following data series for manufacturing (basic materials and
processing, separately) and nonmanufacturing based on the Japan Standard Industrial
Classification.
Annual projections*1 Nominal yen-denominated figures; number of reporting enterprises
Overseas business Year-to-year percent change; nominal figures; number of reporting
activities*2 enterprises
*1 This item corresponds to "exchange rates" described in 3.(1).
10
*2 This item corresponds to "domestic sales" in the consolidated sales, "domestic fixed investment"
in the consolidated fixed investment, "share of overseas sales in total sales," and "share of overseas
fixed investment in total fixed investment" in addition to all items described in 3.(4).
E. Others
The Bank releases the population number and number of sample enterprises as follows.
•The Tankan: The population number and number of sample enterprises by industry4 and
enterprise size listed in A. Tankan.
•The Survey of Financial Institutions: The population number and number of sample
institutions by industry listed in B. Survey of Financial Institutions.
• The Survey of Business Consultants & Pure Holding Companies: The population
number and number of sample enterprises in C. Survey of Business Consultants & Pure
Holding Companies.
•The Survey on Overseas Business Activities: The population number and number of
sample enterprises for manufacturing (basic materials and processing, separately) and
nonmanufacturing listed in D. Survey on Overseas Business Activities.
The Bank also releases standard error ratio as follows.
•The Tankan: Standard error ratio of sales by industry6 and enterprise size listed in A.
Tankan.
• The Survey of Financial Institutions: Standard error ratio of fixed and software
investment (excluding land purchasing expenses) for overall financial institutions in the
Survey of Financial Institutions.
•The Survey of Business Consultants & Pure Holding Companies: Standard error ratio
of fixed and software investment (excluding land purchasing expenses) for business
consultants and pure holding companies in the Survey of Business Consultants & Pure
Holding Companies.
•The Survey on Overseas Business Activities: The standard error ratio of total sales on a
consolidated basis for manufacturing (basic materials and processing, separately) and
nonmanufacturing listed in D. Survey on Overseas Business Activities.
4
Excluding some industries in the upper class such as basic materials and processing.
11
6. Notes for Users
(a) To calculate DIs, the percentage share of the number of respondents who chose each
alternative is rounded to an integer. This rounding-off is conducted for each release
classification DI. Therefore, the degree of change in DIs may sometimes appear
inconsistent among release classifications, for example, between all industries,
manufacturing, and nonmanufacturing, due to rounding error.
(b) There are discontinuities in some data series resulting from the following revisions
and changes.5
March 2004 revision
•Judgment survey: before and after March 2004
•Annual projections: before and after fiscal year 2003
•Quarterly data: before and after December 2003
Changes of the treatment of fixed investment and software investment
•Annual projections: before and after the September 2010 survey
March 2014 revision
•Conditions for CP issuance: before and after March 2014
March 2020 revision6
•Exchange rates: before and after March 2020
(c) For past explanations and other revisions and changes to the Tankan, see the
Explanation and Notices of Changes and Corrections, respectively on the Bank's
website.
5 For further details, see: "Revision of TANKAN in the March 2004 survey –Comparison between the
pre- and post-revision in the December 2003 survey," Research and Statistics Department, Bank of
Japan, March 8, 2004. "Notice for the Tankan," Research and Statistics Department, Bank of Japan,
July 1, 2010. "Planned Revision of the Tankan Survey – Review of survey items, including the
introduction of a survey on Inflation Outlook of Firms –," Research and Statistics Department, Bank
of Japan, March 22, 2013.
6
For further details, see: "Planned Revision of the Tankan," Research and Statistics Department, Bank
of Japan, December 27, 2016 and "Partial Changes to the 'Planned Revision of the Tankan'," Research
and Statistics Department, Bank of Japan, March 4, 2019.
12
II. METHOD
1. Data Collection
The survey is conducted using a paper form or an online form.7,8
The entire period for collecting responses for each survey is about one month from the
day when the paper form or identification code for the online survey is provided to sample
enterprises to the release date. For each survey, the Bank sets a deadline to sample
enterprises of approximately two weeks after the dispatch of the survey form.
2. Survey Content
(1) Judgment Survey
For the following 13 items, sample enterprises are asked to choose the response which,
according to their judgment, best describes the enterprise’s present and future condition
from the three options below (excluding seasonal factors). In the case of a "not applicable
response," they are asked to reply "9."
The 10 items from "business conditions" to "conditions for CP issuance" survey the level
of sample enterprise condition at the time of the survey and in three months’ time. On the
other hand, the three items from "change in interest rate on loans" to "change in input
prices" survey changes in sample enterprise conditions from the three months prior to the
time of the survey and then also from the time of the survey to conditions in three months’
time.
Business conditions:
Judgment of general business conditions of responding enterprises, primarily in light of
their profits.
[1.Favorable 2.Not so favorable 3.Unfavorable]
7
The online survey began from the March 2011 survey. Since the September 2017 survey, the
Government Statistics Sharing System has been used.
8
Sample enterprises are required to choose to complete either the paper survey or online survey in
advance (it can be changed for each survey).
13
Domestic supply and demand conditions for products and services:
Judgment of domestic supply and demand conditions for major products or services in
the industry of responding enterprises, taking into account customer trend, order arrival,
and movements of goods if necessary. Answers including the judgment of overseas
conditions are also acceptable when it is difficult to exclude it.
[1.Excess demand 2.Almost balanced 3.Excess supply]
Overseas supply and demand conditions for products:
Judgment of overseas supply and demand conditions for major products in the industry
of responding enterprises.
[1.Excess demand 2.Almost balanced 3.Excess supply]
Inventory level of finished goods and merchandise:
Judgment of inventory levels of finished goods and merchandise compared with sales of
responding enterprises (only for at the time of the survey).
[1.Excessive or somewhat excessive 2.Adequate 3.Insufficient or somewhat
insufficient]
Wholesalers' inventory level:
Judgment of domestic and foreign wholesalers' inventory levels of major finished goods
and merchandise in the industry of responding enterprises (only for at the time of the
survey).
[1.Excessive or somewhat excessive 2.Adequate 3.Insufficient or somewhat
insufficient]
Production capacity:
Judgment of production capacity or business equipment of responding enterprises,
excluding temporary factors such as regular maintenance of plant equipment.
[1.Excessive capacity 2.Adequate 3.Insufficient capacity]
Employment conditions:
Judgment of the number of employees of responding enterprises.
[1.Excessive employment 2.Adequate 3.Insufficient employment]
Financial position:
14
Judgment of general cash position of responding enterprises, taking into account levels
of cash and cash equivalent, lending attitude of financial institutions, and payment and
repayment terms (only for at the time of the survey).
[1.Easy 2.Not so tight 3.Tight]
Lending attitude of financial institutions:
Judgment of financial institutions' attitude toward lending as perceived by responding
enterprises (only for at the time of the survey).
[1.Accommodative 2.Not so severe 3.Severe]
Conditions for CP issuance:
Judgment of overall issuance conditions for commercial paper (only for at the time of the
survey).
[1.Easy 2.Not so severe 3.Severe]
Responding enterprises for this item are limited to large enterprises with capital of 1
billion yen or more that either issued CP in the past two years or are currently considering
CP issuance.9
Change in interest rate on loans:
Judgment of changes in interest rates on borrowings of responding enterprises
[1. Rise 2.Unchanged 3.Fall]
Change in output prices:
Judgment of changes in yen-based selling prices of major products or services provided
by responding enterprises.
[1. Rise 2.Unchanged 3.Fall]
Change in input prices:
Judgment of changes in yen-based purchase prices of main raw materials (including
processing fees for subcontractors) and/or main merchandise paid by responding
enterprises.
9
The "past two years" means the following:
"from the end of March two years earlier" for the March survey.
"from the end of June two years earlier" for the June survey.
"from the end of September two years earlier" for the September survey.
"from the end of December two years earlier" for the December survey.
15
[1. Rise 2.Unchanged 3.Fall]
(2) Annual Projections
For the following 10 items, annual results and forecasts of "fixed investment," "land
purchasing expenses," "software investment" and "R&D investment" and semiannual
results and forecasts of other items are surveyed in the millions of yen except exchange
rates which are in yen per U.S. dollar and yen per Euro.
The Tankan surveys figures of each fiscal year six times in total, from the March survey
of the previous fiscal year to the June survey of the following fiscal year.
Sales:
Total of items listed as operating revenue on the income statement, such as sales, business
income, and amount of work completed.
Exports:
Including both direct exports and exports via trading companies (except for the
wholesaling industry), but excluding tripartite trade such as exports from a foreign
country to another foreign country.
Exchange rates (yen per U.S. dollar):
Average U.S. dollar-yen exchange rates for actual results and expected rates on which
forecasts of sample enterprises’ overall businesses are based, if projected by those
enterprises.
Exchange rates (yen per Euro):
Average Euro-yen exchange rates for actual results and expected rates on which forecasts
of sample enterprises’ overall businesses are based, if projected by those enterprises.
Current profits:
Current profits listed on the income statement.
Net income:
Net income after tax deduction listed on the income statement.
Fixed investment:
16
Amounts of newly listed tangible fixed assets (including new land purchasing expenses
and lease assets), excluding transfers from other accounts to tangible fixed assets and an
increase in tangible fixed assets due to mergers and transfers of operations.
Land purchasing expenses:
Amounts of new land purchasing expenses (including land developing and preparing)
included in "fixed investment," excluding real estate for sale or sale in lots.
Software investment:
Amounts of newly listed intangible fixed assets (including lease assets) included in
software investment, excluding software recorded as expenses at the time of purchase.
R&D investment:
R&D expenses listed on the income statement. Total of R&D expenses included in general
and administrative expenses as well as in manufacturing expenses for the period.
(3) Inflation Outlook of Enterprises
For the following two items, sample enterprises are asked which of multiple price
developments is closest to their expectations — excluding the effects of institutional
changes such as changes in the consumption tax rate — for 1 year ahead, 3 years ahead,
and 5 years ahead.
Outlook for output prices:
Expectations for the rate of change in the selling price of a responding enterprise’s main
domestic products or services relative to the current level. When responding enterprises
have difficulty in narrowing down to a single core product or service, they are asked to
answer on the basis of the weighted average prices of multiple products and services or
their products and services overall. When they have difficulty in narrowing down the
domestic price of their main product or service, they are asked to answer on the basis of
prices that include overseas prices, excluding the effect of exchange rates to the greatest
extent possible.
[Alternatives]
Rate of changes relative to the current level
17
1. around + 20% or higher (+ 17.5% or higher)
2. around + 15% (+ 12.5% ~ + 17.4%)
3. around + 10% (+ 7.5% ~ + 12.4%)
4. around + 5% (+ 2.5% ~ + 7.4%)
5. around 0% (−2.5% ~ + 2.4%)
6. around −5% (−7.5% ~ −2.6%)
7. around −10% (−12.5% ~ −7.6%)
8. around −15% (−17.5% ~ −12.6%)
9. around −20% or lower (−17.6% or lower)
10. Don't know.
Outlook for general prices:
Expectations of year-on-year rate of change in general prices as measured by the
Consumer Price Index. For those without specific expectations, responding enterprises
are asked to clarify the reason.
[Alternatives]
In annual percent rate changes
11. around + 6% or higher (+ 5.5% or higher)
12. around + 5% (+ 4.5% ~ + 5.4%)
13. around + 4% (+ 3.5% ~ + 4.4%)
14. around + 3% (+ 2.5% ~ + 3.4%)
15. around + 2% (+ 1.5% ~ + 2.4%)
16. around + 1% (+ 0.5% ~ + 1.4%)
17. around 0% (−0.5% ~ + 0.4%)
18. around −1% (−1.5% ~ −0.6%)
19. around −2% (−2.5% ~ −1.6%)
20. around −3% or lower (−2.6% or lower)
* Without specific expectations
21. Uncertainty over the future outlook is high
22. Not really conscious of inflation fluctuations because they should not
influence the strategy of the institution
18
23. Other
(4) Number of New Graduates Hired (surveyed in June and
December only)
For the following item, sample enterprises are asked to provide annual results and
forecasts.
The Tankan surveys figures of each fiscal year five times in total, from the June survey
of the previous fiscal year to the June survey of the following fiscal year.
Number of new graduates hired:
The number of new graduates — employees that responding enterprises consider to be
new graduates — hired and the planned number to be hired during the surveyed fiscal
year.
(5) Overseas Business Activities
For the following 5 items, sample enterprises are asked to provide annual results and
forecasts.
The Tankan surveys figures of each fiscal year ten times in total, from the June survey of
the previous fiscal year to the June survey of the fiscal year after next.
Consolidated sales:
Total of items listed as operating revenue on the income statement, such as sales, business
income, and amount of work completed on a consolidated basis.
Overseas sales:
Sales to overseas external customers (including export) in "consolidated sales," such as
sales to overseas external customers by geographic segment listed in the securities report.
Consolidated current profits:
Consolidated current profits listed on the income statement.
Consolidated fixed investment:
Amounts of newly listed tangible fixed assets (including new land purchasing expenses
and lease assets) on a consolidated basis, excluding transfers from other accounts to
19
tangible fixed assets and an increase in tangible fixed assets due to mergers and transfers
of operations.
Overseas fixed investment:
Amounts of newly listed tangible fixed assets located overseas (including new land
purchasing expenses and lease assets) on a consolidated basis in "consolidated fixed
investment," such as the amounts of newly listed tangible fixed assets of property, plant
and equipment located overseas by geographic segment listed in the securities report.
These amounts exclude transfers from other accounts to tangible fixed assets and an
increase in tangible fixed assets due to mergers and transfers of operations.
3. Survey Workflow
The survey workflow is as follows:
(1) Dispatch of survey forms
The Bank's Head Office (Research and Statistics Department) provides a paper form
or an identification code for the online survey by mail to all sample enterprises, in
principle.
(2) Collection of survey forms
Completed Survey forms are sent back by mail or submitted online to the Bank's
Head Office and responses are entered into the Bank's system. After the deadline
passes, the Bank calls non-respondents and encourages them to complete and send
back forms. Bank staff at the Head Office or branches then go through responses,
calling respondents for confirmation where necessary.
(3) Calculation and estimation
The Bank's Head Office calculates and estimates the survey results of the Tankan,
the Survey of Financial Institutions, the Survey of Business Consultants & Pure
Holding Companies and the Survey on Overseas Business Activities based on the
responded figures.
(4) Publication
20
The Bank's Head Office releases the survey results of the Tankan, the Survey of
Financial Institutions, the Survey of Business Consultants & Pure Holding
Companies and the Survey on Overseas Business Activities.
Flowchart of the Tankan Survey (conceptual diagram)
Dispatch of survey
Encouragement for return of survey
Collection of survey
Response Review
Calculation and estimation
Release
21
III. CALCULATION AND ESTIMATION
1. Judgment Survey
Figures from sample enterprise responses are aggregated into the diffusion index (DI) as
follows:
𝐷𝐼 = 𝑃1 − 𝑃3
𝑃1 : Percentage share of the number of respondents choosing alternative 1
𝑃3 : Percentage share of the number of respondents choosing alternative 3
For calculating DIs, the percentage share of the number of respondents choosing each
alternative is rounded to an integer, then, that of alternative 3 is subtracted from that of
alternative 1.
Example
In the business conditions survey, when 50 sample enterprises choose "alternative 1
favorable," 120 choose "alternative 2 not so favorable," and 30 choose "alternative 3
unfavorable," the percentage shares of the number of respondents choosing each alternative
are 25 percent, 60 percent, and 15 percent respectively.
The "Business Conditions DI" is then calculated as shown below:
Business Conditions DI = 25 (%) — 15 (%) = 10 (% points)
2. Quantitative Survey (Annual Projections, Number of New
Graduates Hired and Overseas Business Activities)
The total values of the survey population (population total) for each survey item such as
"sales" or "number of new graduates hired," is estimated using figures from sample
enterprise responses.10
In the Tankan, the survey population is divided into homogeneous groups (strata) using
three criteria — industry, capital, and sales — in order to select sample enterprises. The
10
If there is a release classification by industry and enterprise size where the number of reporting
enterprises is only one or two, figures are withheld to avoid estimation of the reporting enterprise's
information, in principle.
22
estimate of the population total is obtained by aggregating the population total in each
stratum, which is estimated as follows11:
𝑒𝑖 = average amount of responding enterprises in stratum 𝑖 × 𝑁𝑖
𝑎𝑖
𝑒𝑖 = × 𝑁𝑖
𝑛𝑖
𝑒𝑖 : Estimate of the population total in stratum i
𝑎𝑖 : Simply aggregated amount of responded figures from sample enterprises in stratum i
𝑛𝑖 : Number of responding enterprises in stratum i
𝑁𝑖 : Number of survey population enterprises in stratum i
"Exchange rates" are simple averages of all the enterprises’ responded rates. Only the
figures from enterprises which have provided answers for both "exports" and "exchange
rates" are used to calculate "exchange rates of exporting enterprises."
With respect to release items but not the survey items such as "domestic sales" and
"software, R&D, and fixed investment excluding land purchasing expenses," the actual
amount is obtained as the sum or the difference of the estimate of the population total of
survey items.
Year-to-year percent change and revision rate are calculated using these actual amounts
as estimated and calculated above. Specifically, the revision rate is calculated as shown
below:
𝐸𝑐 − 𝐸𝑝
𝑅𝑒𝑣𝑖𝑠𝑖𝑜𝑛 𝑟𝑎𝑡𝑒 (%) = × 100
𝐸𝑝
𝐸𝑐 : Estimate of population total of the current survey
𝐸𝑝 : Estimate of the population total of the previous survey
"Number of enterprises reporting deficits" is the number of enterprises which reported
deficits in current profits.
11
If there is a stratum where the number of reporting enterprises is zero, the estimate of the population
total is complemented by that of a stratum with smaller sales next to the stratum in the same industry
and enterprise size with respect to the criterion for stratification.
23
3. Inflation Outlook of Enterprises
Percentage shares of the number of respondents choosing each alternative are calculated.
"The average of enterprises' inflation outlook," released as reference data, is the weighted
average of responded rates of price change excluding responses of "Don't know" and
"Don't have clear views on general prices." The responded rates of price change are based
on the following assumptions: for example, "around +15%" and "around +6% or higher"
indicate +15% and +6%, respectively.
4. Handling of Missing Values
Some sample enterprises do not provide answers causing missing sample data. The
missing values are handled as follows.
(1) Judgment Survey and Inflation Outlook of Enterprises
If all or part of the figures from sample enterprise responses in each survey item are
missing, the enterprise is excluded for calculation and estimation. For example, in the
case where a sample enterprise provides only an answer for the present business
conditions without providing the future outlook in the judgment survey, both the actual
result DI and the forecast DI are calculated excluding the enterprise.
(2) Annual Projections, Number of New Graduates Hired and
Overseas Business Activities
If all or part of the figures in each survey item from a sample enterprise response are
missing, the most recently responded figures from the enterprise are individually
substituted for the calculation and estimation in the following imputation methods.12,13
12
The missing value imputation has been introduced since the March 2004 survey.
13
For further details, see: Kiyohito Utsunomiya and Katsurako Sonoda, "Methodology for Handling
Missing Values In TANKAN," Bank of Japan Working Paper Series, No. 01-11, June 2002.
24
With respect to "sales," "exports," "current profits," and "net income," figures obtained
from the non-responding enterprise for each half of the most recent fiscal year, are used
for the imputation: i.e. the figures of the first half and second half of the year are
substituted for the first half and second half of the current survey, respectively.
With respect to "fixed investment," "land purchasing expenses," "software investment,"
"R&D investment," "number of new graduates hired," "consolidated sales," "overseas
sales," "consolidated current profits," "consolidated fixed investment," and "overseas
fixed investment," the figures obtained from the non-responding enterprise for the whole
of the most recent fiscal year, are used for the imputation.
With respect to "exchange rates (yen per U.S. dollar)" and "exchange rates (yen per
Euro)," the figures obtained from the non-responding enterprises for the most recent half
of fiscal year, are used for the imputation.
5. Treatment of Outliers
An outlier is one of the data which significantly deviates from the rest of the data in
general. As stated previously, the estimate of the population total is calculated by stratum
for the items of annual projections such as "sales" and "exports" in the Tankan, using
figures from sample enterprise responses. Under the estimation, if a figure from a sample
enterprise significantly deviates from the rest of the same stratum and it does not represent
the survey population, the estimated values could also significantly deviate from the real
values of the survey population. An example of one of the causes of outliers is the
stratification of the original sampling which is maintained for several years until the next
revision of sample enterprises, by which time the business size or contents of business of
sample enterprises could have greatly changed.
With respect to annual projections for "sales," "current profits," "net income," "fixed
investment," "software investment," and "R&D investment," a sample data that
significantly deviates from the rest of the same stratum beyond a certain criteria is, in
principle, treated as an outlier which does not represent the survey population (see
25
appendix 1 for the method of outlier detection)14. Only when a sample data of "sales" and
"fixed investment" is detected as an outlier is the method of outlier treatment applied to
their sub-items "exports" and "land purchasing expenses" respectively.15,16
Specifically, an outlier is regarded as a missing value and the most recent figures
regarding the survey item of the corresponding enterprise’s response are individually
substituted for the calculation and the estimation (see section 4 for the details). Once an
outlier is detected, this treatment is continued until the upcoming revision of sample
enterprises.
14 Similar treatment is also applied to the Survey of Financial Institutions and the Survey of Business
Consultants & Pure Holding Companies, which have been conducted in order to supplement the
Tankan. With respect to "R&D investment", the Bank does not conduct outlier detection for medium-
sized and small nonmanufacturing in the Tankan nor for the Survey of Financial Institutions (see
appendix 1).
15
For further details, see: Atsushi Ishikawa, Shunsuke Endo, and Tetsuya Shiratori, "Treatment of
Outliers in Business Surveys: The Case of Short-term Economic Survey of Enterprises in Japan
(Tankan)," Bank of Japan Working Paper Series, No. 10-E-8, September 2010.
16
The outlier treatment was introduced from the December 2010 survey.
26
IV. SAMPLE DESIGN
1. Population
The target population of the Tankan are private enterprises (excluding financial
institutions) in Japan with capital of 20 million yen or more. It should be noted that some
industries which may have a weak connection to economic conditions such as "scientific
and development research institutes" and "medical and other health services" are
excluded from the target population. Comprising approximately 210,000 enterprises
(excluding "financial institutions" and "business consultants & pure holding companies")
in Japan, the survey population of the Tankan is drawn from the Establishment Frame
Database of the Ministry of Internal Affairs and Communications, which provides the
most comprehensive information on the target population. As there are non-responding
enterprises in the Establishment Frame Database, there are gaps between the target
population and the survey population, resulting in coverage error. In an effort to limit the
coverage error, enterprises which do not respond to the Establishment Frame Database
are added to the Tankan survey population to the extent possible.
2. Basic Concept of Sample Enterprises
The Tankan uses a sample survey framework, which surveys part of a population (sample)
and estimates the population parameters using the sample data.
In a sample survey, a higher number of sample enterprises improves statistical accuracy
and also increases the total burden on survey respondents. Stratified sampling has been
introduced in the Tankan to satisfy established criteria for statistical accuracy with a
controlled number of sample enterprises. Under this sampling method, the survey
population is divided into homogeneous groups (strata) using criteria such as industry and
capital, then, each stratum is sampled.
It is desirable that sample enterprises are randomly selected in a sample survey so that the
sample closely reflects the features of the population in theory. In this case, however,
many enterprises are newly added or excluded from the sample. This leads to a decrease
27
in the response rate, an increase in respondent errors, and also an increase of the total
burden of producing statistics. From this perspective, in the Tankan, with the current
sample enterprises basically continuing as part of the sample, new sample enterprises are
selected by stratum to satisfy the criteria on standard error ratio and sample enterprise
distribution as described below. With the Tankan’s aim of capturing an accurate picture
of business trends in Japan in mind and the many sample enterprises who are accustomed
to responding to the survey, this method brings about desirable effects such as limiting
decreases in response rates and respondent error.
3. Criteria for Statistical Accuracy
In the Tankan as for any ordinary sample survey, the Bank establishes criteria for
statistical accuracy and selects sample enterprises which satisfy them. The Bank sets the
following two criteria.
(1) Standard Error Ratio
Standard error ratio shows how accurate a survey population mean estimate is. It is
calculated by dividing the standard deviation of the sample mean by the survey population
mean (see appendix 2 for the details). For example, suppose the sample mean of sales is
one million yen and the standard error ratio is three percent, with probability 0.95, the
survey population mean of sales will be between 940 thousand yen and 1,060 thousand
yen. As this example demonstrates, the confidence interval becomes shorter with the
decreased standard error ratio, resulting in improving statistical accuracy.
The population total of quantitative survey items such as "sales" and "fixed investment"
are estimated in the Tankan. A criterion for statistical accuracy is set for the standard error
ratio of certain quantitative survey items. The Bank sets the standard error ratio for sales
in consideration of its importance as a survey item.
The dispersion of sales for each industry is spread out over a wide range and, as is well
known, that of nonmanufacturing tends to be wider than that of manufacturing. With this
in mind, the Bank sets the criteria as three percent or less on the standard error ratio for
28
manufacturing and as five percent or less for nonmanufacturing respectively in the Tankan.
These criteria must be satisfied for overall manufacturing and nonmanufacturing as well
as for each enterprise size — large, medium-sized, and small enterprises.
In addition, the Bank sets a non-binding target of approximately 10 percent or less for
each of the 31 industries by enterprise size.
Criteria for statistical accuracy on standard error ratio of sales
Medium-sized
Large enterprises Small enterprises
enterprises
Manufacturing 3% 3% 3%
Nonmanufacturing 5% 5% 5%
Each industry
approximately 10% approximately 10% approximately 10%
(non-binding)
(2) Sample Enterprise Distribution
In the Tankan, as stated previously, the population data is updated upon revision of sample
enterprises. After the revision, new sample enterprises are selected with the current
sample enterprises basically continuing as part of the sample. This makes it more likely
than in random sampling that the sample distribution may become inconsistent with the
survey population distribution.
On this point, the Bank sets the following criterion: sample distributions of capital size
and sales could be deemed not to deviate from survey population distributions. In
particular, each stratum is further subdivided into several subgroups (minimum strata)
based on capital and sales (see appendix 3 for the details), then the Bank examines each
stratum to check whether there is a statistically significant difference between the
population and sample distributions. This difference is tested by the chi-square goodness
of fit test (see appendix 4 for the details).
29
4. Sample Design Procedure
The Bank has introduced stratified sampling in the Tankan to satisfy established criteria
for statistical accuracy with a controlled number of sample enterprises as described in
Section 2. This section describes the specific procedure of stratified sampling.
In general, a smaller variance of enterprise sales in each stratum can lower a standard
error ratio and satisfy the criteria with a lower number of sample enterprises. This means
that if population is further divided to decrease the variance of each stratum, the criteria
of standard error ratio can be satisfied with a lower number of sample enterprises.
(1) Criteria for Stratification
Survey results of the Tankan are calculated and released by industry and enterprise size
and the criteria of standard error ratio are set for each of these classifications.
Consequently, industry and enterprise size are used as criteria for stratification. Sales is
also a criterion for stratification as the criterion for statistical accuracy has been set for
the standard error ratio of sales.17
(2) Stratification Method
Out of the three criteria for stratification in the sample design of the Tankan, industry and
enterprise size are determined in advance as follows: the number (31 industries and 3
enterprise sizes) and the boundary line or value for strata are defined based on the Japan
Standard Industrial Classification for industry and the amount of capital for enterprise
size. As a result, stratification in the sample design of the Tankan works to determine the
number of strata and the boundary values for 93 strata (31× 3) in order to minimize
variance of sales of the survey population enterprises.
Taking into account the balance between satisfying the criteria for statistical accuracy and
17
Sales was added to the criteria for stratification in the sampling from the March 2018 survey.
Number of employees had been previously used as a criterion for stratification due to lack of
comprehensive data on sales on the survey population. For more details, refer to "Revision of the
Sample Design of the Tankan Using the Economic Census of Japan," Research and Statistics
Department, Bank of Japan, September 23, 2016.
30
lightening the total burden of producing statistics, since the revision of sample enterprises
in 2004, the Bank has divided the survey population into 300 to 400 strata. In fact, there
are 384 strata in the sample design since the March 2022 survey. This means that
stratification in the Tankan works to find boundary values which minimize variance of
sales of the survey population enterprises (see appendix 3).
The Bank places the following restriction on finding the boundary values: each stratum
has five or more population enterprises. This is because the number of sample enterprises
selected from a stratum with less than five population enterprises will inevitably be small,
making it impossible to estimate the population total of the corresponding stratum when
there are non-responding sample enterprises.
(3) Selecting Sample Enterprises
In the Tankan, sample enterprises are selected for each stratum divided as described above
to satisfy the criteria of standard error ratio and sample enterprise distribution as described
in section 3. In particular, with the current sample enterprises basically continuing as part
of the sample, new enterprises are added and current sample enterprises are excluded
from the sample, bearing it in mind to lower the sample size to the extent possible. The
Bank places the following restriction on selecting sample enterprises: the sampling ratio
of each stratum is maintained at 0.5 percent or more.18 This method limits disturbance to
the estimates due to an idiosyncratic factor of a sample enterprise.
5. Sample Design of Survey of Financial Institutions
The Survey of Financial Institutions uses a sample survey framework in the same way as
the Tankan.
18
The Bank eased the restriction on the sampling ratio from one percent or more to 0.5 percent or
more for the March 2018 survey because applying the method of outlier treatment has made it easier
to control an effect on the estimates caused by some irregular movements in values unique to a sample
enterprise.
31
The target population are defined as private financial institutions in Japan with 10 or more
employees. Comprising approximately 2,800 financial institutions in Japan, the survey
population is drawn from the Establishment Frame Database of the Ministry of Internal
Affairs and Communications.
Sample institutions are selected from the survey population using stratified sampling for
each of the 22 strata, which are formed by dividing the survey population based on sector
and the number of employees, in order to satisfy the established criteria for statistical
accuracy with current sample institutions basically continuing as part of the sample.
The Bank sets the following two criteria for statistical accuracy: (1) the standard error
ratio of fixed and software investment (excluding land purchasing expenses) for overall
financial institutions is 10 percent or less; and (2) the sample distribution of number of
employees does not deviate from the survey population distribution.
The survey population comprises seven sectors, used as the criterion for stratification,
based on the Japan Standard Industrial Classification released by the Ministry of Internal
Affairs and Communications as shown below.
Survey of Financial Institutions Japan Standard Industrial Classification (code)
Financial institutions
City banks and trust banks
Member banks of the Regional Banks Banking (62) (except Central bank (621))
Association of Japan and the Second
Association of Regional Banks
Shinkin banks, etc.
Other financial institutions for Financial institutions for cooperative organizations (63)
small businesses
Financial products transaction Financial products transaction dealers and futures
dealers commodity transaction dealers (65)
Insurance institutions, including insurance agents, brokers
Insurance companies
and services (67)
Non-deposit money Non-deposit money corporations, including lending and
corporations credit card business (64)
Note: Each industry is classified according to the small and middle classifications of the Japan Standard Industrial
Classification revised in October 2013.
32
As for the number of employees, the other criterion for stratification, the number of strata
and the boundary value of each stratum are determined in order to minimize variance of
fixed and software investment (excluding land purchasing expenses) of the survey
population institutions on condition that each stratum has five or more population
institutions.
6. Sample Design of Survey of Business Consultants & Pure Holding
Companies
The Survey of Business Consultants & Pure Holding Companies uses a sample survey
framework in the same way as the Tankan.
The target population are private enterprises that are "business consultants & pure holding
companies" in Japan with capital of 20 million yen or more.
Comprising approximately 1,600 enterprises in Japan, the survey population is drawn
from the Establishment Frame Database of the Ministry of Internal Affairs and
Communications.
Sample enterprises are selected from the survey population using stratified sampling for
each of the 4 strata, which are formed by dividing the survey population based on sales
in order to satisfy the established criteria for statistical accuracy with current sample
enterprises basically continuing as part of the sample.
The Bank sets the criterion that the sample distribution of sales does not deviate from the
survey population distribution for its statistical accuracy.
7. Sample Design of Survey on Overseas Business Activities
The Survey on Overseas Business Activities uses a sample survey framework in the same
way as the Tankan.
33
The target population are private parent enterprises (excluding "financial institutions") in
Japan, with capital of 1 billion yen or more, at the top of a consolidated business group
with sales to overseas external customers or overseas subsidiaries/branches.
Comprising approximately 1,600 enterprises in Japan, the survey population is drawn
from the Establishment Frame Database of the Ministry of Internal Affairs and
Communications.
Sample enterprises are selected from the survey population using stratified sampling for
each of the 73 strata, which are formed by dividing the survey population based on sector
and total sales, in order to satisfy the established criteria for statistical accuracy with
current sample enterprises basically continuing as part of the sample.
The Bank sets the following two criteria for statistical accuracy: (1) the standard error
ratio of total sales is within the target range (10 percent for both manufacturing and
nonmanufacturing); and (2) the sample distribution of capital and total sales does not
deviate from the survey population distribution.
8. Sample Maintenance19
(1) Handling of Changes of Sample Enterprises
The Bank examines statistical accuracy — standard error ratios and fitness of the
population and sample distributions as described in section 3 — regularly (once a year in
principle) to prevent the lowering of statistical accuracy due to a decrease in the number
of sample enterprises resulting from bankruptcies, mergers, and other factors. When
accuracy is not sufficient to satisfy criteria, the Bank adds new sample enterprises.
19
The Bank maintains sample institutions of the Survey of Financial Institutions as well as sample
enterprises of the Tankan, Survey of Business Consultants & Pure Holding Companies and Survey
on Overseas Business Activities.
34
(2) Dealing with Mergers and Spin-offs
In general, the Bank takes the following procedures to deal with mergers and spin-offs
involving Tankan sample enterprises.
A. Mergers20
When the core enterprise is a sample enterprise prior to the merger, the merged enterprise
is included in the sample. Specifically, in cases where there is a merger between a sample
enterprise and another sample enterprise or a non-sample enterprise with a smaller sales
volume, the merged enterprise is included in the sample. On the other hand, when there
is a merger between a sample enterprise and a non-sample enterprise with a larger sales
volume, the merged enterprise is excluded from the sample.
B. Spin-offs
Only the core enterprise is included in the sample after the spin-off. However, all
enterprises — the core enterprise as well as other spin-offs — are included in the sample
when; (1) the core enterprise and other spin-offs are grouped in the same stratum for
population estimation and; (2) the impact on the estimate of the population total of sales
or fixed investment turns out to be small with the inclusion of these enterprises into the
sample.
The Bank has introduced the following two exceptional rules applied when the impact of
mergers and spin-offs on the estimate of the population total is above a certain level, such
as large-scale mergers and spin-offs in the Tankan.
C. Distribution method
The Bank distributes the responded figures of merged enterprises and spin-off enterprises
to the core enterprise prior to the merger or spin-off in the Tankan. This means that, in
20
The core enterprise corresponds to the enterprise with the largest sales volume in the Tankan and
the Survey of Business Consultants & Pure Holding Companies, while the core enterprise in the
Survey of Financial Institutions is the institution with the largest number of employees. In the Survey
on Overseas Business Activities, the core enterprise is defined as the enterprise with the largest total
sales.
35
order to decrease technical fluctuations which do not reflect the actual conditions of the
economy, the population total is estimated as if there were no mergers and spin-offs when
the impact of mergers and spin-offs on the estimate of the population total is above a
certain level.
When there is an impact of approximately one percent or more on the estimate of the
population total of sales or fixed investment for the relative category — of the six main
categories: two industries (manufacturing and nonmanufacturing) times three enterprise
sizes (large, medium-sized, and small enterprises) — to which the enterprise belonged
prior to the merger or spin-off, the Bank applies the distribution method.
In the case of mergers:
When there is a merger between enterprises A and B creating merged enterprise C, the
Bank surveys C and then the responded figure of C is distributed to the figures of A and
B. The ratio used in the distribution to A (B) is in principle obtained by dividing the figure
of A (B) by the figure of C at the time of the merger.
In the case of spin-offs:
When enterprise X is spun off into enterprises A, B, and C, the Bank surveys only the
core enterprise (let us assume A) and the responded figure of A is multiplied for
calculation of the figure of X. The multiplying factor is in principle the reciprocal of the
ratio obtained by dividing the figure of A by the figure of X at the time of the spin-off.
D. Incorporating spin-off enterprises in the course of time
A non-core enterprise which was not selected as a sample enterprise at the time of the
spin-off may grow and have an impact above a certain level on the estimate of the
population total in the Tankan. In such a case, the Bank may flexibly incorporate it into
the sample enterprise to capture the whole picture of the pre-existing sample enterprise
even though some time has passed since the spin-off.
Specifically, the Bank may incorporate a non-core enterprise when its figure of sales or
fixed investment has an impact of approximately one percent or more on the estimate of
the population total for the relative category — of the six main categories: two industries
(manufacturing and nonmanufacturing) times three enterprise sizes (large, medium-sized,
36
and small enterprises) — to which the enterprise belonged prior to the spin-off. The Bank
may incorporate such a spin-off enterprise at the time of regular sample addition (once a
year in principle).21
21
The number of population enterprises is changed exceptionally in this case.
37
(Appendix 1)
Method of Outlier Detection
An outlier is one of the data which significantly diverges from the rest of the data.
Detailed procedures of the detection of outliers in the Tankan are as follows.
(1) yit , which denotes the degree of influence of values reported from each enterprise on
year-on-year rates of change or revision rates for the six main categories22 — two industries
(manufacturing and nonmanufacturing) times three enterprise sizes (large, medium-sized, and
small enterprises) —, is calculated as follows:
Degree of change in weighted data
yit = × 100
Estimate of the population total for one of the six main categories
| |
in the previous fiscal year (or previous survey)
wi′ (xit − xit−1 )
= × 100
|Mjt−1 |
xit : Value reported from enterprise i at time t
xit−1 : Value reported from enterprise i at time t-1
wi′ : Adjusted weight of xi *
Mjt−1 : Estimate of the population total for category j (j=1,....,6) at time t-1
* To accurately measure the degree of change in weighted data, despite the difference in the number
of sample enterprises between time t and t-1, wi′ is computed based on the same number of sample
enterprises. For this purpose, when some enterprises provide no response in either of the two surveys
(at different points in time), the missing values are substituted by the sample mean in each stratum
to which each of these enterprises belongs.
(2) Next, zit, the indicator of the degree of divergence based on yit , is calculated as follows:
yit − dj 99
if dj 99 ≤ yit
Dj
zit = dj1 − yit
if yit ≤ dj1
Dj
{ 0 if dj1 < yit < dj 99
If this indicator, zit, exceeds the threshold value23, which is set in advance based on empirical
analyses using past data, the reported value is detected as an outlier.
22
This corresponds to the degree of influence of values reported from each sample on year-on-year
rates of change or revision rates for all the samples in the Survey of Financial Institutions and the
Survey of Business Consultants & Pure Holding Companies.
23 C is set at 50 in the Tankan, and 10 in the Survey of Financial Institutions and the Survey of Business
Consultants & Pure Holding Companies, respectively. With respect to "R&D investment", the Bank
38
dj1 : One percentile24 of yit in category j
99
dj : 99 percentile of yit in category j
Dj : Distance between dj 99 and dj1 (dj 99 − dj1 )
does not conduct outlier detection for medium-sized and small nonmanufacturing in the Tankan nor
for the Survey of Financial Institutions, given that the ratio of samples whose yit is zero is high,
which makes it difficult to set an appropriate threshold value.
24
The one percentile is defined as the value below which one percent of observation can be found.
39
(Appendix 2)
Standard Error Ratios
Standard error ratios represent the variation of sample means relative to the population
mean. Using sales as an example, the formula is shown below.
(Standard deviation of sample mean)
(Standard error ratio)： =
(Population mean)
𝐿
𝑁𝑖 − 𝑛𝑖 𝜎𝑖2
(Standard deviation of sample mean)： = √∑ 𝑊𝑖2
𝑁𝑖 − 1 𝑛𝑖
𝑖=1
𝐿
(Population mean)： = ∑ 𝑊𝑖 𝑌̅𝑖
𝑖=1
𝑁𝑖 ∶ Number of population enterprises in stratum 𝑖
𝑛𝑖 ∶ Number of sample enterprises in stratum 𝑖
̅𝑖 ∶ Population mean of sales in stratum 𝑖
𝑌
𝜎𝑖2 ∶ Population variance of sales in stratum 𝑖
𝑊𝑖 ∶ Share of stratum 𝑖 in population enterprises classified by industry and size
𝐿 ∶ Number of strata in population enterprises classified by industry and size
Note: The formula shown above is derived as follows: when selecting a sample of size n without
replacements from a population of size N (population mean: μ, population variance: 𝜎 2 ),
𝑁𝑖 −𝑛𝑖 𝜎𝑖2
the sample mean and the sample variance are given by μ and , respectively.
𝑁𝑖 −1 𝑛𝑖
40
(Appendix 3)
Stratification Method of Sample Design
In the sample design of the Tankan, the survey population is divided into 93 strata — 31
industries times 3 enterprise sizes (large, medium-sized, and small enterprises). These are
then subdivided into 1 to 6 strata in order to minimize the variance of sales of the survey
population enterprises with a goal of 300 to 400 strata as a whole.
Each stratum is further subdivided into 2 to 4 minimum strata based on capital and sales.
The conceptual diagram below shows the stratification method as outlined above.
Capital
20 million yen to 100 million yen to 1 billion yen or more
100 million yen 1 billion yen
(small enterprises) (medium-sized enterprises) (large enterprises) 31 industries
Find
Find boundary
boundary values
values
w hich
whichminimize
minimize
variance
variance ofof
sales of of
sales thethe
survey
surveypopulation
population
enterprises,
enterprises, forming
forming2 1
toto6 6
strata.
strata.
Sales
Constitute 2 to 4 minimum strata
Minimum stratum
(3)
Sales
(1)
(4) (2)
Capital
41
(Appendix 4)
Chi-square Goodness of Fit Test
The chi-square goodness of fit test is a statistical method to test whether there is a
significant difference between the observed and the expected distributions. Detailed
procedures and the diagram of the test in the Tankan are as follows.
(1) Each stratum is subdivided into several subgroups (minimum strata) i, where 𝑖 =
1,2, ⋯ , 𝑗, based on capital and sales.
(2) In each minimum stratum, the number of survey population enterprises
(𝑁1 , 𝑁2 , ⋯ , 𝑁𝑗 ) and that of sample enterprises (𝑛1 , 𝑛2 , ⋯ , 𝑛𝑗 ) are obtained.
(3) 𝑝𝑖 , which denotes the ratio of survey population enterprises in each minimum stratum,
is calculated as:
𝑗
𝑁𝑖
𝑝𝑖 = (𝑁 ≡ ∑ 𝑁𝑘 )
𝑁
𝑘=1
(4) If there is no significant difference between the sample and survey population
distributions, 𝑒𝑖 , which denotes the number of sample enterprises in each minimum
stratum, is expected to be obtained as:
𝑗
𝑒𝑖 = 𝑛 ∙ 𝑝𝑖 (𝑛 = ∑ 𝑛𝑘 )
𝑘=1
The fitness between the two distributions is tested by the chi-square goodness of fit
test at 5% significance level, where 𝑛𝑖 is the observed value and 𝑒𝑖 is the expected
value.
[Constitution example of minimum strata]
Survey population Sample
(Number of enterprises) (Number of enterprises)
1,000 400 20 8
1,600 500 35 11
(Ratio of minimum stratum (Ratio of minimum stratum
on stratum) on stratum)
0.29 0.11 0.27 0.11
0.46 0.14 0.47 0.15
Testing the difference
between the two
42
File and source
- File
- bank-financing-for-smes-in-times-of-crisis-when-whatever-it-takes-confronts-black-swans.pdf
- Size
- 972,089 bytes
- SHA-256
- 5ba7fd3901bdfa187270ad1aa806b31d62ce2c4b9f005390d9683fbe73d352db
- Original
- doi.org