Statistics

Home > Statistics > Outline of Statistics and Statistical Release Schedule > Explanations of Statistics > Explanation of TANKAN > FAQ (Frequently Asked Questions) on "Tankan" (Short-term Economic Survey of Enterprises in Japan)

FAQ (Frequently Asked Questions) on "Tankan" (Short-term Economic Survey of Enterprises in Japan)

March 2002
Bank of Japan
Research and Statistics Department

Index

1. Q&A regarding the outline of Tankan

2. Q&A regarding the judgement survey

3. Q&A regarding the quantitative survey

4. Q&A regarding the accuracy of Tankan

5. Q&A regarding measures taken to improve the transparency and users' convenience

6. Q&A regarding miscellaneous topics

1. Q&A regarding the outline of Tankan

1-1. What are the purposes and characteristics of Tankan?

The main purpose of Tankan is to "grasp the actual condition of the domestic economy." Tankan is a nationwide business survey conducted on a quarterly basis (in March, June, September, and December) which covers overall corporate activities. It consists of two parts: a judgement survey and a quantitative survey such as quarterly data and annual projections. The name "TANKAN (note)" is also widely known overseas.

(note): "TANKAN" is an abbreviation of "TANKI KEIZAI KANSOKU CHOUSA" (the Short-term Economic Survey of Enterprises in Japan).

1-2. Does Tankan have legal authority? (Is there an obligation to respond?)

Tankan is a statistical survey notified to the Minister of Public Management, Home Affairs, Posts and Telecommunications based on Article 8 of The Statistics Law (Law No. 18 of 1947). A "notified statistical survey" differs from a "designated statistical survey" (e.g. "Population Census," "Establishment and Enterprise Census," etc.) based on The Statistics Law, as the respondents are not legally enforced to cooperate. As for Tankan, however, thanks to the understanding and cooperation of the sampled enterprises, almost all of them respond every time. According to Articles 14 and 15 of The Statistics Law, the Bank of Japan is obliged to keep the confidential information of the respondent enterprises under strict control.

1-3. What information can we get from the survey results of Tankan?

To grasp the overall corporate activities precisely, Tankan includes not only a "judgement survey," which covers the qualitative aspect of business confidence, but also a "quantitative survey," which covers figures of sales, fixed investment, etc. (quarterly, semiannual, and annual figures). As for these data, "forecasts" are also surveyed in addition to "actual results." Thanks to its long history, Tankan provides a wide range of consistent data over a long period. These features may be counted as strengths of Tankan.

By combining the "judgement survey" and the "quantitative survey" and comparing them with past data in many ways, users can analyze the economic conditions and corporate activities according to their own interests.

The survey results of Tankan are aggregated figures of the respondent enterprises and does not reflect the judgements or projections of the economy by the Bank of Japan.

1-4. What is the difference between "All Enterprises Tankan" and "Principal Enterprises Tankan?"

The questionnaires of "All Enterprises Tankan" and "Principal Enterprises Tankan" are basically the same. Respondent enterprises of "Principal Enterprises Tankan" are a subset of those for "All Enterprises Tankan." The major differences between two Tankans are the sampling method of the respondent enterprises and the aggregation method.

In "All Enterprises Tankan," private enterprises (excluding financial and insurance industries) with 50 and over regular employees (20 and over for wholesaling, retailing, services, and leasing industries) covered by "Establishment and Enterprise Census" conducted by the Ministry of Public Management, Home Affairs, Posts and Telecommunications, are assumed to be the "population enterprises" (currently about 160,000). The population enterprises are stratified into 118 strata by industry and scale (large, medium-sized, and small enterprises; see note 1), and sample enterprises (currently about 9,000) are selected from them. On aggregation, the DI (see 2-1.) of the judgement survey is calculated by using simple aggregates, while population estimates of the quantitative survey are calculated by industry and scale (note 2).

Notes: 1. In "All Enterprises Tankan," sample enterprises are classified into large, medium-sized, and small enterprises based on their number of regular employees, as shown in the table below.

Table :
  Wholesaling Retailing, services,
and leasing
Other industries
Large enterprises 1,000 employees or more 1,000 employees or more 1,000 employees or more
Medium-sized enterprises 100-999 employees 50-999 employees 300-999 employees
Small enterprises 20-99 employees 20-49 employees 50-299 employees

Notes: 2. Aggregates are released by industry and scale (total of 81 strata: 27 industries <17 in manufacturing, 10 in non-manufacturing> divided into 3 scales <large, medium-sized, and small enterprises>). The strata are further segmented (total of 118 strata) for sampling and calculating the population estimates in order to enhance statistical accuracy.

On the other hand, the sample for "Principal Enterprises Tankan" is taken from "stock exchange listed companies with capital of at least one billion yen, excluding the financial and insurance industries" under the condition that they are "principal enterprises deemed to generally reflect the trends in their industries" (currently about 700). Sample enterprises are basically fixed (although there may be cases in which the number of enterprises decreases due to mergers and bankruptcies). Simple aggregates are used for both judgement and quantitative surveys.

Since the surveyed enterprises are fixed in "Principal Enterprises Tankan," the results may diverge from the actual condition of the economy due to the changes in the industrial structure. It should also be noted that, contrary to its name "principal," it includes some enterprises with small numbers of employees (there are cases in which "small enterprises" of "All Enterprises Tankan" are selected as "principal enterprises"). Because of these reasons, "All Enterprises Tankan" has been treated as the main part of Tankan since the March 1999 survey and "Principal Enterprises Tankan" has been treated as a reference.

Since the majority of users have become to focus mainly on "All Enterprises Tankan," we have decided to abolish "Principal Enterprises Tankan" in the next revision scheduled to take place in FY 2003 (the sample enterprises of "Principal Enterprises Tankan" will continue to be surveyed in "All Enterprises Tankan").

1-5. What is the history of Tankan?

There are many business surveys conducted in and out of Japan. Among these surveys, Tankan is one of the oldest. The first business survey ever conducted in Japan was the "Short-term survey of the industries" by the Industrial Bank of Japan, which started in 1951 following the method of the "Economic test" by the IFO Institute for Economic Research of the former West Germany. By taking over and revising that survey, the Bank of Japan has started "Principal Enterprises Tankan" in 1957. In 1974, the Bank of Japan launched "All Enterprises Tankan" in addition. Since then, various kinds of revisions have been made regarding the sampling and the items surveyed in the aim of adequately reflecting the structural changes of the economy.

A Brief Chronology of Tankan

August 1957
Start of "Short-term Economic Survey of Principal Enterprises (Principal Enterprises Tankan)" (coverage: 524 enterprises <346 manufacturing and 178 non-manufacturing enterprises>*; quarterly survey).
May 1960
Start of "Business Conditions Survey of Small Enterprises" (2,666 enterprises <small manufacturing enterprises only>; semiannual survey).
May 1962
Revision of "Business Conditions Survey of Small Enterprises" shifting from a semiannual survey to a quarterly survey.
November 1966
The title changed from "Business Conditions Survey of Small Enterprises" to "Short-term Economic Survey of Small Enterprises (Small Enterprises Tankan)."
May 1974
Coverage extended for the "Short-term Economic Survey of Small Enterprises." The extended survey was named "Short-term Economic Survey of All Enterprises (All Enterprises Tankan)" (5,596 enterprises <4,243 manufacturing and 1,353 non-manufacturing enterprises>). Large and medium-sized enterprises for the manufacturing sector and non-manufacturing enterprises (large, medium-sized, and small enterprises) were added to the sample.
May 1983
Sample size largely extended mainly in the non-manufacturing sector for "All Enterprises Tankan" (from 5,181 to 7,035 enterprises <among which non-manufacturing: from 1,287 to 3,141 enterprises>).
November 1989
Start of the "Fixed investment of financial institutions" as a supplementary survey to "Principal Enterprises Tankan" (207 enterprises).
November 1993
Sample size largely extended mainly in the non-manufacturing sector for "All Enterprises Tankan" (from 7,376 to 10,011 enterprises <among which non-manufacturing : from 3,386 to 5,822 enterprises>).
March 1997
Difference in the questionnaires between "All Enterprises Tankan" and "Principal Enterprises Tankan" eliminated. (In this process, the number of items surveyed were reduced to half for "Principal Enterprises Tankan.")
March 1999
Samples added to "All Enterprises Tankan" (from 9,129 to 9,433 enterprises). Simultaneously, expansion was made to the release data and the status of "All Enterprises Tankan" was upgraded to become the main survey replacing "Principal Enterprises Tankan" which was dropped to a reference .
  • The sample was basically selected from "stock exchange listed companies with capital of at least 100 million yen, excluding the financial and insurance industries" and are also "principal enterprises generally reflecting the trends in their industries."

1-6. Tankan is also released by the branches of the Bank of Japan. What is the difference from the one released by the head office (Research and Statistics Department)?

The head office of the Bank (Research and Statistics Department) releases the survey results of "All Enterprises Tankan" and "Principal Enterprises Tankan." The branches also release the survey results of the enterprises within the region it covers (hereafter "Branches Tankan").

There are two major differences between "All Enterprises Tankan" and "Branches Tankan." These differences mainly stem from the fact that the sample framework of "All Enterprises Tankan" is based on a nationwide basis and is not designed to accurately reflect the industrial structure of local regions covered by individual branches. The two differences are as follows:

  1. (1) From the viewpoint of reflecting the industrial structure of each region, the samples of "Branches Tankan" include some respondents that are not included in that of "All Enterprises Tankan" (such as "local establishments of large enterprises").
  2. (2) Regarding the aggregation method, while population estimates are calculated for the quantitative data of "All Enterprises Tankan," simple aggregates are calculated for the quantitative data of "Branches Tankan."

Users are advised to keep in mind of this difference between "Branches Tankan" and "All Enterprises Tankan" and use "Branches Tankan" as a reference material to grasp the economic conditions of the respective regions. Also from this aspect, it should be noted that it is inappropriate to compare the absolute levels of DIs and figures among branches.

The head office also releases the "Classification by District" (aggregates of 9 regions such as "Hokkaido," "Kanto," and others) for "All Enterprises Tankan." Since these figures are just subtotals by region extracted from the responses of enterprises selected through a nationwide basis sampling, they do not necessarily reflect the developments of the respective regions on a statistical basis. Therefore, users are advised to use them as references to grasp the economic conditions of the regions. The "Classifications by District" released by the head office is planned to be abolished in the next revision scheduled in FY 2003, while "Branches Tankan" will continue to be released by individual branches.

1-7. How are the financial and insurance industries treated in Tankan?

Financial and insurance industries are currently not included in the sample of "All Enterprises Tankan." Nevertheless, the weight of these industries in fixed investment has increased in line with the development of the financial online system. To cope with this situation, a survey on fixed investment of these industries has been added since the November 1989 survey (corresponding to the current December survey) as a supplementary survey of "Principal Enterprises Tankan." (Those surveyed are: banks <city banks, long-term credit banks, trust banks, regional banks, and the members of the Second Regional Banks Association>, major securities companies and insurance companies.)

In the next revision of Tankan scheduled in FY 2003, the sample as well as the items surveyed are going to be expanded, and the status of the survey will be changed into a sample survey supplementing "All Enterprises Tankan." The details are as follows:

  1. (1) As for the range of the types of institutions to be covered, "shinkin banks," "keitou financial institutions etc.," and "non-depository credit corporations such as credit and finance institutions, and investment institutions" will be added in addition to "banks," "securities companies," and "insurance companies,"

The actual sampling process will be as follows:

The set of all enterprises belonging to the above industries will be the population. Among which, all "city banks," "long-term credit banks," and "trust banks" are going to be included in the sample as of present. As for the other industries, sample enterprises will be selected based on "total assets."

  1. (2) As for the contents of the survey, judgement survey of "business conditions" and judgement and quantitative surveys related to "fixed investment" and "employment" are going to be added. Since the impact of financial institutions on business conditions has recently become increasingly significant, reinforcing the coverage of Tankan on the activities of financial institutions should be effective in "reflecting the current condition of the domestic economy" which is the primary aim of this survey.

2. Q&A regarding the judgement survey

2-1. How is the Diffusion Index (DI) calculated?

In the judgement survey, answers from the respondent enterprises ("1,2,3") are aggregated into the Diffusion Index(DI) as calculated below:

DI (% points)= percentage share of enterprises which chose answer 1 minus percentage share of enterprises which chose answer 3

For instance, "business conditions DI" indicates the respondent's judgement on the "overall business conditions mainly in light of profits." Enterprises are asked to choose one out of three answers, "1. Favorable," "2. Not so favorable," and "3. Unfavorable." The percentage share of the number of enterprises which gave each answer is calculated, and the percentage share of those which answered "3. Unfavorable" is subtracted from those which answered "1. Favorable."

With respect to the business conditions (actual result) of "large manufacturing enterprises" in the September 2001 survey, 5 percent replied "1. Favorable," 57 percent "2. Not so favorable," and 38 percent "3. Unfavorable." As a result, business conditions DI was "5 % minus 38% = -33%."

Regarding the magnitude of changes in the DIs, there are some cases in which the figures appear to be inconsistent at first glance. For example, in the following case, the differences between the "actual result" and the "forecast" for the manufacturing and non-manufacturing industries were both "+2," while the difference for the total of all industries was only "+1."

This is because, on calculating the DI, the percentage shares of enterprises which replied "1. Favorable" and "3. Unfavorable," respectively, are rounded to integers (the first decimal place is rounded off) and this is done for figures of "manufacturing," "non-manufacturing," and "all industries" respectively. Consequently, a "gap due to the round off" occurred for these figures.

Example: Business conditions DI of large enterprises in the June 2001 survey (% points)

Table : Business conditions DI of large enterprises in the June 2001 survey (% points)
  Actual result Forecast Changes
(forecast minus actual result)
Manufacturing -16
(9% - 25%)
-14
(7% - 21%)
+2
Non-manufacturing -13
(10% - 23%)
-11
(9% - 20%)
+2
All industries -14
(10% - 24%)
-13
(8% - 21%)
+1

On calculating the DI, enterprises are not weighted according to their sizes, which means "1 vote per enterprise." This differs from the case with the "quantitative survey" in which larger-scale enterprises have larger influences on the aggregates, and therefore, users are advised to keep this in mind when comparing the two surveys.

2-2. Why are the responses of the judgement survey aggregated into DIs?

Regarding the aggregated data, the percentage shares of the answers ("1,2, and 3") can be used directly. But, for instance, if you are interested in their long-term developments, it might be rather tedious or confusing to handle this type of data. The DIs are compact and handy indicators which reflect the essence of the developments of these data.

2-3. How can the DIs be used?

For each judgement survey, the "actual result (judgement at the time of the survey)" and the "forecast (that for three months ahead)" are surveyed and "actual DI" and "forecast DI" are calculated thereby. One way to use this is to see how well the enterprises have done in their forecasts by comparing the "forecast DI" of the previous survey with the "actual DI" of the present survey.

Users can also combine the DIs and the DI-related quantitative surveys ("business conditions DI" and "current profits" or "ratio of current profit to sales," "production capacity DI" and "fixed investment," "employment conditions DI" and "number of employees," etc.) or compare DIs during similar economic phases in the past according to their various analytical purposes.

2-4. What should you keep in mind when you use the individual DIs?

For example, there are four kinds of DIs regarding supply and demand conditions and inventories: (1) supply and demand conditions for products and services ("excess demand" minus "excess supply"); (2) overseas supply and demand conditions for products ("excess demand" minus "excess supply"); (3) inventory level of finished goods and merchandise ("excessive or somewhat excessive" minus "insufficient or somewhat insufficient"); (4) wholesalers' inventory level ("excessive or somewhat excessive" minus "insufficient or somewhat insufficient").

Among these, (3) asks the judgement on the inventory level of the "enterprise itself." On the other hand, (1), (2), and (4) ask the judgement on the supply and demand conditions or the inventory level of the "industry" in which the enterprise's main merchandise (or service) belongs.

Users are advised to keep in mind these differences in definitions (judgement on the "enterprise itself" or the "industry" it belongs to) when using these DIs.

As regards (1) change in interest rate on loans ("rise" minus "fall"), (2) change in output prices ("rise" minus "fall"), and (3) change in input prices ("rise" minus "fall"), not the "level" but the "direction" of changes are asked.

Please note that the DIs of Tankan basically indicate the "level," but those of above are the exceptions indicating "direction."

3. Q&A regarding the quantitative survey

3-1. What is a population estimate? How is it calculated?

The main purpose of Tankan is to give a timely overview of the actual condition of the domestic economy by researching corporate activities. However, if we try to survey all enterprises (this is called a "complete survey"), not only does it entail a vast amount of cost, but also, since considerable time is needed to aggregate and release the figures, it is difficult to grasp the actual condition of the economy on a timely basis.

Hence, samples are selected under certain statistical criteria from the population enterprises (as for Tankan, they are private enterprises with 50 or more regular employees among those listed in "Establishment and Enterprise Census" by the Ministry of Public Management, Home Affairs, Posts and Telecommunications). The sum of the replied figures of the samples is inflated according to the extractive ratio to get the estimates for the aggregates of the overall population. The figures obtained using this method are the "population estimates." (A survey which uses the information obtained from samples to get estimates of the population is generally called a "sample survey.")

An example of calculation of the population estimate.
When 50 enterprises are selected as samples from a population of 200, and the simple aggregates in sales of the sample was 150:
Population estimate = (150/50) x 200 = 600
Which is:
(average sales per enterprise) x (number of population enterprises) = (estimate of sales of 200 enterprises)

For the quantitative survey of "All Enterprises Tankan," population estimates are calculated by industry and scale (total of 118 strata), and these figures are aggregated to obtain the total population estimate.

3-2. What should you keep in mind when you look at the year-to-year percent change or the revision rate of "All Enterprises Tankan?"

In "All Enterprises Tankan," figures of each fiscal year are surveyed six times from the March survey of the previous fiscal year to the June survey of the following fiscal year. The figures (population estimates) of each fiscal year are fixed by those of the June survey of the following fiscal year.

For instance, as shown in the table below, figures for FY 2000 are fixed by those of the June 2001 survey. Accordingly, the "year-to-year percent change" for the figures of FY 2000 is fixed along with the "denominator" to calculate the "year-to-year percent change" for the figures of FY 2001. Thus, regarding the "year-to-year percent change" for the figures of FY 2001, users should take into consideration that there is a difference between the "denominators" (figures of FY 2000) of the March 2001 survey and the other surveys (June 2001 survey to June 2002 survey).

Numerators" and "denominators" to calculate the "year-to-year percent change" for figures of FY 2000 and 2001. "

Note: Figures in parentheses indicate the ordinary numbers of survey for the figures of each fiscal year.

The "revision rate" is calculated by comparing the "figures of the previous survey" and the "figures of the present survey" as shown below:

Revision rate (%) = (figures of the present survey - figures of the previous survey) / figures of the previous survey x 100

It should be noted that the following equation does not necessarily hold: "year-to-year percent change of the previous survey" + "revision rate of the present survey" = "year-to-year percent change of the present survey".

This is because, as shown above, there is usually a difference between the "denominators" used in calculating the "year-to-year percent change" of the March survey and those used in the other surveys. In the example shown below, the year-to-year percent growth is higher (lower) in the June survey compared to the March survey, but the revision rate of the June survey is negative (positive).

Actual calculation behind example 1:

Year-to-year percent change of the March survey = 10,595.8 billion yen (figures for FY 2001 of the March survey) / 10,361.5 billion yen (figures for FY 2000 of the March survey) x 100-100 = 2.3%

Year-to-year percent change of the June survey = 10,547.2 billion yen (figures for FY 2001 of the June survey) / 9,790.2 billion yen (figures for FY 2000 of the June survey) x 100-100 = 7.7%

Revision rate of the June survey = [10,547.2 billion yen (figures for FY 2001 of the June survey) - 10,595.8 billion yen (figures for FY 2001 of the March survey)] / 10,595.8 billion yen (figures for FY 2001 of the March survey) x 100 = -0.5%

3-3. What should you keep in mind when you look at the year-to-year percent change or the revision r ate of "Principal Enterprises Tankan?"

In "Principal Enterprises Tankan" for example, the "year-to-year percent change" for FY 2000 is fixed at the June 2001 survey. This is the same as "All Enterprises Tankan." However, since figures for FY 2000 are not fixed at the June 2001 survey, the "denominator" used in calculating the "year-to-year percent change" for FY 2001 figures is also not fixed. Thus, the "numerator" used in calculating the "year-to-year percent change" of FY 2000 and the "denominator" used in calculating the "year-to-year percent change" of FY 2001 do not necessarily coincide.

Example: The numerator and denominator of the "year-to-year percent change" in the September 2001 survey

"year-to-year percent change" for the figures of FY 2000 (already fixed at the June 2001 survey)

Table : "year-to-year percent change" for the figures of FY 2000 (already fixed at the June 2001 survey)
Numerator: figures of FY 2000 (June survey)
Denominator: figures of FY 1999 (June survey)

"year-to-year percent change" for figures of FY 2001

Table : "year-to-year percent change" for figures of FY 2001
Numerator: figures of FY 2001 (September survey)
Denominator: figures of FY 2000 (September survey)

Unlike "All Enterprises Tankan," population estimates are not calculated in "Principal Enterprises Tankan." When the actual respondents differ among surveys, it is inappropriate to compare the simple aggregates of different surveys. Hence, on calculating the "year-to-year percent change" in "Principal Enterprises Tankan," simple aggregates of the same survey are used for the numerator and denominator.

Regarding the "revision rate," the calculation method is the same for "Principal Enterprises Tankan" as "All Enterprises Tankan," which compares the "figures of the previous survey" and the "figures of the present survey." However, there is a difference in the data used, where the former uses only the "enterprises, which responded to both the previous and present survey." (Enterprises responding only to either one of the surveys are excluded when calculating the revision rate.)

3-4. How is the "predicted exchange rate expected by enterprises" (Yen per U.S. dollar rate) calculated?

The predicted exchange rate is calculated based on the responses from enterprises ("exchange rates for exports" on the survey sheet). The individual rates are weighted by the value of exports (on a yen basis) of the enterprises upon aggregation, unlike DIs which are based on simple averages.

The predicted exchange rate by industry and scale (excluding the non-manufacturing sector except wholesaling from the 118 strata in 3-1.,) is calculated using the following formula, and the overall predicted exchange rate is calculated using these components.

Predicted exchange rate by industry and scale (yen/dollar) =
Population estimate of (exchange rate of the individual respondent x value of its exports <yen basis>) / Population estimate of the value of exports (yen basis)

Example:

A hypothetical case of 3 respondent enterprises and 30 population enterprises

Enterprise A : exchange rate 115 yen/dollar, exports 50

Enterprise B : exchange rate 120 yen/dollar, exports 100

Enterprise C : exchange rate 125 yen/dollar, exports 150

Table : A hypothetical case of 3 respondent enterprises and 30 population enterprises
Predicted exchange rate = [(115 x 50 + 120 x 100 + 125 x 150) / 3] x 30
  [(50 + 100 + 150) / 3] x 30
  = 121.67 (yen/dollar)

3-5. What is "software investment?"

"Software investment" is a new item surveyed from the March 2001 survey. Enterprises are asked to answer their "software investment which is newly listed (or planned to be listed) as intangible fixed assets during each term."

The survey was started based on the idea that, in view of the growing IT-related investment in Japan, software such as computer programs should also be included in fixed investment to grasp the actual condition of the economy more accurately.

This is also the case for the SNA (System of National Accounts) -related statistics made by the Cabinet Office. For instance, the GDP (Gross Domestic Products) based on the "93 SNA" released in 2000 has newly recorded "computer software" as a part of fixed investment.

As for the corporate accounting system, software accounting has also been introduced for business years since April 1999, and it has become possible to grasp software investment through corporate account items.

Under these circumstances, we have started to survey software investment in Tankan in addition to the conventional fixed investment (tangible fixed assets newly listed). However, "software investment" in Tankan and "computer software" in GDP are surveyed on a different basis.

For details, see "Addition of 'software investment' and related change of release forms in Short-term Economic Survey of Enterprises in Japan, TANKAN" (March 2001).

4. Q&A regarding the accuracy of Tankan

4-1. What are the revisions made recently?

From the March 1999 survey, the release data have been expanded in line with the revision of the sample enterprises of "All Enterprises Tankan." Simultaneously, "Principal Enterprises Tankan," which used to be the main part of Tankan, has been changed into a reference index handing over its role to "All Enterprises Tankan." In response to this change, a comparison between "principal enterprises" and "small enterprises of the 'All Enterprises Tankan,'" has been replaced by that among "large, medium-sized, and small enterprises of 'All Enterprises Tankan'" in the release material.

In view of the recent increase in IT-related investment, Tankan has started to survey "software investment" from the March 2001 survey.

Another revision is planned in conjunction with the next regular revision of sample enterprises for "All Enterprises Tankan" (scheduled in FY 2003). The items surveyed and the criteria used for aggregation is planned to be revised in the aim of reflecting the recent structural changes of industry as well as further enhancing the statistical accuracy. For details, see "Final plan on the revision of Tankan" (June 2001<available in Japanese only>).

4-2. Since "All Enterprises Tankan" is a sample survey, there must be inherent statistical errors in its aggregated figures (population estimates). What kind of errors are included in the figures of Tankan?

In "sample surveys" such as "All Enterprises Tankan," two kinds of errors mainly occur which are "sampling errors" and "non-sampling errors."

For example, if the sample enterprises consist of only those that are larger than the average of the population, population estimates such as sales or fixed investment will be larger (overestimated) than the actual figures (aggregates obtained from a complete survey). An error of this type, which occurs from biases included in the sample, is called a "sampling error."

When responses cannot be obtained from the sample enterprises or incorrect figures are replied, the population estimate obtained may possibly diverge from the actual figure. Errors of this type are called a "non-sampling error." Thus, "non-sampling errors" occur not only for sample surveys but also for complete surveys.

4-3. What measures are taken in order to reduce the sampling errors?

Sample enterprises of "All Enterprises Tankan" are reviewed almost every five years, the interval of which is synchronized with "Establishment and Enterprise Census" by the Ministry of Public Management, Home Affairs, Posts and Telecommunications which is used to specify its population enterprises.

In each review, the following statistical criteria are used in order to control the sampling errors. For details, see "The Methodology of the Sampling and Aggregation of the All Enterprises Tankan" (September 1999).

  1. (1) The error ratios (ratios which indicate the deviation of the population estimate from the "actual figure" <standard error/average>) of the population estimates in sales calculated based on the sample enterprises generally remain within the target range (3% for the manufacturing sector and 5% for the non-manufacturing sector).
  2. (2) The distribution of the sample enterprises measured by the number of regular employees does not differ much from the distribution of the population enterprises.

In practice, each stratum classified by industry and scale for sampling (total of 118) is examined by both its "error ratio" and its "distribution," and the sample is revised when necessary to attain the above criteria. Upon its revision, not all of the samples are retaken but new ones are added to the existing respondents (continued samples).

Even if the samples are adequately revised, sampling errors may increase before the next revision due to decreases in the number of samples caused by, for example, bankruptcies or mergers. Therefore, the above statistical examination is conducted regularly on an annual basis, and if the accuracy declines, new sample enterprises are added at the March survey.

4-4. What measures are taken in order to reduce the non-sampling errors?

From the viewpoint of decreasing non-sampling errors, we are endeavoring to increase the response rate by asking the respondent enterprises for understanding and cooperation. (The response rate of the December 2001 survey of "All Enterprises Tankan" was 96.8%.)

We are also making efforts to increase the response rate by allowing round figures when authorized figures are unavailable (for example, when authorized forecast figures are not decided yet, we ask the respondent to fill in approximate figures).

We also go through the responded sheets carefully to prevent false responses.

In "All Enterprises Tankan," the treatment of unanswered responses are as follows. First, for the "judgement survey," in which the simple aggregation method is taken, the responses are excluded from the aggregation. Second, for the "quantitative survey," in which the population estimation method is taken for aggregation, the average figures of the strata in which they belong to (among the 118 strata by industry and scale) are substituted for the unanswered parts.

Currently, all quantitative question items basically maintain high response rates. Therefore, we have not seen any problems with the measures described above. However, when we reviewed the alternative methods in order to handle the unanswered data from a statistical point of view, regarding "sales," "current profits," and "fixed investment," it turned out that "more accurate figures are likely to be obtained as a whole, by substituting the most recent response of the corresponding enterprise (figures of the current or previous fiscal years) for the unanswered parts."

We are, thus, planning to switch to the above method for supplementing the unanswered responses for "sales," "current profits," and "fixed investment" at the next revision of Tankan scheduled in FY 2003.

4-5. Isn't the 5-years interval for the revision of sample enterprises too long to adequately incorporate the rapid changes in the economic structure?

The samples of "All Enterprises Tankan" are regularly reviewed in response to "Establishment and Enterprise Census" conducted every five years by the Ministry of Public Management, Home Affairs, Posts and Telecommunications. In light of the large and rapid changes in the industrial structure nowadays, however, it is desirable to shorten the 5-years interval.

Meanwhile, the Ministry of Public Management, Home Affairs, Posts and Telecommunications has shortened the interval of "Establishment and Enterprise Census" by initiating a "simple survey" in the third year after the principal survey (note) apart from the "principal survey" conducted every five years. Under these circumstances, we decided to use both the "principal survey" and the "simple survey" after the next revision scheduled in FY 2003 for the sample revision of "All Enterprises Tankan" and shorten its interval to every two or three years.

(note): The "simple survey" was first conducted in 1999.

4-6. Mergers and spin-offs of enterprises are increasing. How are these incorporated into Tankan?

We basically take the following procedures when there are mergers or spin-offs among the sample enterprises in "All Enterprises Tankan."

Mergers:

When there is a merger among the sample enterprises, or when a sample enterprise acquires a non-sample enterprise, the enterprise after the merger becomes the sample enterprise.

When a sample enterprise is acquired by a non-sample enterprise, the enterprise after the merger does not become the sample enterprise.

Spin-offs:

For both of the two methods below, we estimate the impact (rate of change made by the spin-off) on the population estimate of sales and fixed investment and choose the one with the smaller impact.

  1. (1) The core enterprise (the enterprise with the largest number of employees) after the spin-off are chosen as a sample, and if there are other enterprises within the same strata after the split-up (118 strata by industry and scale), they are also chosen as samples.
  2. (2) Only the core enterprise after the spin-off is chosen as a sample.

Regarding the sample enterprises, we set up the above basic rule from the practical viewpoint of choosing which enterprise to be included in the sample upon mergers or spin-offs. On the other hand, the number of population enterprises used for population estimation is unchanged in principal, in response to mergers and spin-offs. (Hence, the number of population enterprises used in this survey does not reflect the increases and decreases in the number of sample enterprises.)

Whenever there is a merger or a spin-off among the sample enterprises of Tankan, the same movement obviously occurs for its population. However, it is difficult to precisely grasp the overall movement such as mergers, spin-offs, bankruptcies, or establishment of new firms (changes from the latest "Establishment and Enterprise Census") regarding population enterprises including that of non-sample enterprises. Thus, we think that it is inappropriate to arbitrarily incorporate only the increases and decreases of sample enterprises due to mergers and spin-offs into the population information (note). In this sense, we define "All Enterprises Tankan" as a "fixed sample survey based on the latest population information available."

(note): However, there is an exception which we call the "1% rule" (cases when the number of population enterprises is changed in response to the increases or decreases of sample enterprises). For details, see "The Methodology of the Sampling and Aggregation of the All Enterprises Tankan" (September 1999).

As corporate restructuring such as mergers or spin-offs is expected to intensify further, we are going to continue searching for better ways to treat these changes in the aim of maintaining and improving the statistical accuracy.


4-7. In which industry is a holding company categorized?

According to the final revision plan of "Japan Standard Industrial Classification (note)," we categorize a "holding company" in the main industry of its subsidiaries (group enterprises) as a whole. In cases that "the main industry" cannot be specified, we categorize it into "services" (an interpretation for this kind of treatment is that a holding company is a "services enterprise for its subsidiaries").

(note): "Japan Standard Industrial Classification" is a standard used for showing the results of various kinds of statistical surveys by industry and is released by the Ministry of Public Management, Home Affairs, Posts and Telecommunications. Tankan will revise the industry classification under the next revision (scheduled in FY 2003) based on the revised "Japan Standard Industrial Classification," released in March 2002.

5. Q&A regarding measures taken to improve the transparency and users' convenience

5-1. Could you describe the main workflow towards its release?

For each survey, we mail the survey forms to the enterprises selected as samples (currently about 9,000) about 1 month prior to the scheduled release date of the survey (usually in the beginning of April, July, October, and mid-December).

The answered survey forms are sent back either to the head office (Research and Statistics Department) or the branches and the answers are scrutinized by the sections in charge. Phone calls are made to the respondents for inquires or confirmations if necessary (survey forms cleared by the branches are sent to the head office).

At the head office, the cleared answers are keyed into the aggregation system and the results are calculated.

From the viewpoint of security control, the aggregated results are strictly protected and not a single staff member is allowed access to the results before the day of release (see 6-2.).

5-2. How are the survey results of Tankan released?

The lineup of released materials on the survey results and the corresponding media of release are as follows:

Released materials;

  1. (1) Summary (a collection of main figures <available in Japanese and English>)
  2. (2) Outline (a digest of representative figures extracted from the "Summary" <available in Japanese and English>)
  3. (3) Figures by Industry (a collection of main figures by industry <available in Japanese only>)
  4. (4) Comprehensive data set (a complete set of detailed figures <available in Japanese and English>)
  5. (5) Long-term time series data (a collection of long-term time series data for the main figures <available in Japanese and English>)

Media of release;

Internet website (provides materials from 1 to 5)

Information services corner at the main entrance of the head office (office hours: 8:50 to 17:00; provides materials from 1 to 3)

Facsimile services (03-3279-4441, 48#; provides material 2)

Material (4) is also available as a statistical periodical (users will be charged for this). For details, see here.

5-3. Apart from the survey results, what other related information is released?

As for Tankan, in order to enhance the convenience of users and the transparency, not only the survey results but also related information is released as shown below. All of them are available on the Internet website (the English version is available on the website for [1], [2], [3], and some parts of [5]).

  1. (1) General Explanation of Tankan
  2. (2) The Methodology of the Sampling and Aggregation of the All Enterprises Tankan
    (a paper explaining the details of Tankan sampling)
  3. (3) Survey sheet of Tankan(the survey sheet sent to sample enterprises)
  4. (4) Guideline on filling out the survey sheet (a note sent to enterprises together with the survey sheet<available in Japanese only>)
  5. (5) Released material regarding Tankan (announcements of revision and others)

5-4. Is the time and date of release of Tankan decided beforehand?

In order to enhance the transparency, the release date of Tankan is announced beforehand and the time of release are all the same.

Together with other major statistics compiled by the Bank, the release date of Tankan is announced on the Internet website at the end of March, June, September, and December for the following six months (April-September, July-December, October-following March, and following January-June). All surveys are released at 8:50 a.m.

5-5. Why is Tankan released at 8:50 a.m.?

Tankan is released just before the opening of (or the time when transactions become active) the Japanese markets (monetary, foreign exchange, and stock markets) which is 9:00 a.m. This is based on the policy that the results of Tankan should be digested by the Japanese markets first.

5-6. Why does the release form of Tankan include only plain results and there are no explanations or comments on them?

On releasing statistics, we believe that it is essential to take a neutral stance and refrain from giving any interpretations or policy judgements on the aggregated results. We also believe that it is important to release the figures as soon as the aggregation is completed. Adding interpretations or judgements may delay this procedure. Based on this policy, we only accept questions from users (including the media) which are purely statistical (questions such as when is the last time such a growth rate occurred or whether there are any noises in the results caused by irregular factors).

Users may think that this is "unkind," but our policy is that "interpretations of statistics must first be made by the market or the public (including economists)" as various interpretations are possible.

We will continue to provide useful information on Tankan to promote its adequate use, such as this Q&A.

The official view of the Bank regarding the interpretations of the newly released statistics including Tankan, and the judgement on the financial and economic conditions based on these interpretations, are shown in the Monthly Report of Recent Economic and Financial Developments, based on discussions made at the Monetary Policy Meeting of the Policy Board, and in the speeches and statements by the Governor.

5-7. What measures are taken in order to release the survey results earlier?

Tankan started to be released about one week earlier from the August 1996 survey (the August survey corresponds to the present September survey). This was made possible by the following two measures: (1) revision of the release material; and (2) termination of reporting the survey results to the board before their release. (As a result, it now takes about one month from sending the survey sheets to releasing the survey results.)

From the March 1999 survey, "The Comprehensive Data Set" (paper format) has been released about two weeks earlier as a result of the streamlining of the workflow etc. Also, from the December 2000 survey, "The Comprehensive Data Set" (electronic format) and the "long-term time series data" have been released one business day earlier.

5-8. Statistical figures often show typical patterns in its movement which sometimes distort their outcomes. What is the case with Tankan?

As for Tankan, users are advised to keep in mind that some of the aggregated results show typical patterns. The major types of patterns are as follows:

(1) Fixed investment

With respect to the revision pattern of annual projections on fixed investment, for example, those of large enterprises are revised upwards from the first March survey to the December survey according to the average of past surveys. This is because new plans which became concrete and plans uncompleted in the previous fiscal year are added as time proceeds from the first survey (March survey) to the December survey. After the December survey, however, fixed investment tends to be revised downwards towards the actual results (obtained in the June survey of the following year) because of delays in construction and postponement of plans to the following fiscal year.

On the other hand, many small enterprises do not formulate annual projections beforehand and thus fixed investment of them tends to be recorded along with the actual implementation. Consequently, fixed investment of small enterprises is revised upwards from the first survey (March survey) to the actual result survey (June survey of the following year) according to the averages of the past.

Of course, these patterns of revision are all merely averages of the past and there are cases which largely differ from these patterns. Nevertheless, when evaluating the figures of fixed investment plans, it is useful to pay attention to these typical patterns of revision (Chart 1).

Chart1: Revision Pattern of Fixed Investment

Graphs of revision patterns of past surveys for annual projections of fixed investment of manufacturing, nonmanufacturing and all industries of large enterprises. The details are shown in the main text.
Graphs of revision patterns of past surveys for annual projections of fixed investment of manufacturing, nonmanufacturing and all industries of small enterprises. The details are shown in the main text.

(2) Corporate finance related DIs

As for the corporate finance related DIs such as "financial position of your company" and "lending attitude of financial institutions," enterprises tend to make cautious judgements for the "forecast." Hence, when we compare the "forecast DI at a certain survey" and the "actual result DI at the following survey," the "forecast DI" is more "tight" for the "financial position DI," and more "severe" for the "lending attitude DI" than the "actual result DI" with only few exceptions (Chart 2).

Chart2: Comparison between the "Forecasts" and "Actual Results" of the Corporate Finance-related DIs

Financial position DI (All industries). Graph of actual series of Financial position DI and connected lines between actual and forecast in each survey. The details are shown in the main text.
Lending Attitude of Financial Institutions DI (All industries). Graph of actual series of Lending Attitude DI and connected lines between actual and forecast in each survey. The details are shown in the main text.

In the next Tankan revision (scheduled in FY 2003), we are planning to abolish the "forecast" survey of both "financial position of your company" and "lending attitude of financial institutions," in light of the bias explained above and the reporting burden on enterprises due to additional survey items. (For the same reason, we are also going to abolish the "forecast" survey for the "inventory level of finished goods and merchandise of your company" and "wholesalers' inventory level in your industry.")

5-9. Where should you contact to ask questions about Tankan?

The following sections will answer your questions about Tankan during office hours on weekdays:

  1. (1) Public Relations Division, Public Relations Department, Bank of Japan (tel. 03-3279-1111, ext. 4636-4639)
  2. (2) Business survey group, Economic Statistics Division, Research and Statistics Department, Bank of Japan (tel. 03-3279-1111, ext. 4010)

Comments and requests for new topics to be included in this Q&A are most welcome. Please send them via e-mail to: post.rsd5@boj.or.jp

6. Q&A regarding miscellaneous topics

6-1. How is Tankan used in the process of monetary policy decision making at the Bank of Japan?

The monetary policy of the Bank is discussed and decided at the Monetary Policy Meeting of the Policy Board. At the meeting, discussions are made from a wide-point of view based on various financial and economic data and anecdotal information. The outcome of Tankan is one of the important information in making judgements. In other words, the outcome of Tankan is one of the major data sources, but monetary policy is not decided solely by Tankan.

6-2. Under what kind of security system is Tankan compiled?

The survey results prior to release and the figures of individual enterprises of Tankan are under strict security control based on internal rules.

The entire work of each Tankan survey is done in an area physically separated from other sections, and staff members other than those designated for the work are shut out from the area during the survey period. Confidential information is handled by a limited number of staff members and are kept locked for paper based materials and protected by passwords on access for electronic based materials.

Not a single person is allowed access to the survey results before the day of release, and even on that day, the number of people in charge of handling the material prior to release (released at 8:50 a.m.) is limited.

6-3. How can a company become a respondent of Tankan (Does the Bank of Japan accept applications from individual companies?)?

As in 4-3., the sample enterprises of "All Enterprises Tankan" are reviewed almost every five years (reviewed every two or three years after the next revision; scheduled in FY 2003). Furthermore, the statistical accuracy is examined each year on a regular basis and the sample enterprises are added if necessary. However, as sample enterprises are selected by the "random sampling" method from the population enterprises listed in "Establishment and Enterprise Census," individual companies cannot be automatically included in the sample even if they apply for it.