Q&A about the Consumer Price Index (Answers)

The CPI is meant to indicate average fluctuations in prices of commodities (goods and services) purchased by households nationwide. In other words, using the consumption by households at a given time as the base period, the index shows changes in the total amount of expenditure required to purchase the equivalent goods and services purchased by households in the base year, setting the consumption structure.
Thus, the CPI is intended to measure changes in prices themselves. We should note, therefore, that the index does not take into consideration fluctuations in living expenses caused by changes in the kind, quality, and/or quantity of commodities purchased by a household as its lifestyle, tastes, etc. change.

Prices tend to inflate as the economy becomes more active and the demand expands, while we usually come down as the economy grows sluggish and the demand diminishes. The CPI, which shows price fluctuations, is often called “a thermometer of the economy”, and a crucial index as the government tries to make effective and appropriate economic policies. In addition, the index is used as a “deflator,” for some other important economic indices, such as those calculated from Family Income and Expenditure Survey(FIES), household consumption expenditure from GDP statistics, and others. Furthermore, the relevant laws require that the actual standards of the National (“Kokumin Nenkin”) and Employee (“Kosei Nenkin”) Pension payments should be adjusted in accordance with price fluctuations. Such fluctuations are measured in terms of CPI. Elsewhere, Bank of Japan considers this index as it decides financial policies. People look to the index when we determine wages, house rents, utility charges, etc. As you see, the index is broadly used both in the public and private sectors.

To calculate the CPI, we first choose those commodities (goods and services) that occupy larger shares in household consumption expenditure. These commodities are used in the calculation. Next, based on the shares in the consumption expenditure of these respective chosen items, we determine the weight of each of these items. These shares are based on the results from the FIES and other surveys.
The prices of the chosen items are mainly taken from the monthly Retail Price Survey (RPS).
Now, to calculate the CPI, we calculate the price index of each of the chosen items, using its average price across the municipalities surveyed, with the price in the base year set at 100. Next, we calculate the weighted average of each item’s price, using the weight (its share in the whole household consumption expenditure) determined above. Furthermore, out of the weighted averages calculated this way, we calculate several different price indices, such as those of Subgroups, of Ten Major Groups, the General (All items) price index, and so on.
As of 2010, the base year in CPI calculation is 2005. This base year is updated every 5 years (“base revision”).
At every base revision, the items to use in CPI calculation as well as their respective weights are reconsidered. During the time period whose base year is 2005, 584 items are employed in CPI calculation. Also, there is a regulation about CPI calculation that we should reconsider these items even outside the regular update, whenever there are new commodities that rapidly find their ways into the majority of households. The midpoint-year revision was conducted in January 2008, when we abolished two of the existing items and added three new ones. Thus, since January 2008 onwards, 585 items in all have been in use in the CPI calculation.

For CPI, the International Labour Organization (ILO) has set up an international standard. The 17th International Conference of Labour Statistics convened in Geneva in December 2003, reconsidered the older international standard for CPI and adopted a resolution that set up a new standard. Concurrently, a revised international manual for CPI, entitled “Consumer Price Index Manual: Theory and Practice,” was completed and published on the ILO’s website as well.
Still, the world has yet to establish the best, single method of calculating CPI, with much room for further arguments left. Japan’s relevant authorities basically follow the existing international standard as we calculate the CPI.

Basically, the CPI is published at 8:30am, on the Friday of the week that includes the 26th day of each and every month. The preliminary CPI of the current month for the Ku-areas of Tokyo and the CPI for the preceding month for Japan (the whole country)are published. In addition, in December and March of each year, the average CPIs for the preceding calendar and fiscal years(*) are also published, respectively.
(*) In Japan, the fiscal year starts in April and ends in March of the next year.

Recently, what are commonly called “core” and “core-core” indices have been growing popular, as indices to some factors in price fluctuations. These are, however, only “popular” names of indices either published or processed and calculated by some users, and not official names.
Now, sometimes “Fresh foods,” whose prices often fluctuate considerably depending on the weather, are deducted from the “General (All items)” index, in order to see the basic trend in the prices of goods and services. The result of this deduction is often called “Core index”.
A similar price index treated as important in the USA and some other nations is known as “USA-type core index” or “Core-core index”. This is calculated by deducting the “Food (excluding alcoholic beverages) and energy” from the “General (All items)” index. Some users in Japan refer to the “General (All items), excluding the elements of fresh food, petroleum products and other peculiar factors” index published by the Cabinet Office in its monthly economy reports and other publications as “Core-core index.” While the Cabinet Office does calculate this index based on the price index of each relevant item, this index is certainly not the “General (All items), excluding food (excluding alcoholic beverages) and energy” index. Please note the difference.

(For your reference)

With the CPI and other monthly statistics, some commodity prices rise and fall after a similar pattern every year. For instance, many seasonal clothes’ prices hike early in their season and fall later, around the end of the season, at sales. Price fluctuations of this kind are known as “seasonal variations.”
We publish seasonally adjusted indices excluding these seasonal changes for the following eight series: “General (All items),” “General (All items) excluding fresh foods,” “General (All items) excluding imputed rent,” “General (All items) excluding imputed rent and fresh foods,” “General (All items) excluding food (excluding alcoholic beverages) and energy,” “Goods” , “Goods , excluding fresh food,” and “Semi-durable goods.”
Since the seasonally adjusted index is adjusted based on average seasonal changes calculated by past indices, price changes in particular months may affect other months (e.g., school fees, medical treatment). Thus, the seasonally adjusted index may change in the months when no price changes actually occurs.
At the same time, change over the year, comparison of the current index with that of the same month in the previous year is also useful to eliminate the seasonality. The rate of Change over the year does not change for one year as long as prices do not change. Thus, in the CPI, change over the year is often used to see the price trend.
As for fresh food for which prices fluctuate greatly, and electricity and gas for which prices change only in the price revision month, change from the previous month, and comparison of the current index with that of the previous month, seems to be close to the real feeling of consumers. Thus, change from the previous month without seasonal adjustment is also published.

[For your reference]
For information on seasonal adjustments to CPI, consult “Chapter 7: Seasonal adjustment” (PDF:1,762KB) in p.38 of “Outline of the 2005-Base Consumer Price Index” as well.
Also, you can download a seasonal adjustment program, “X-12-ARIMA (including Part X-11),” from the website of the Census Bureau, the US Department of Commerce.
In addition, the Bank of Japan’s website features a “X-12-ARIMA Operation Manual.”(in Japanese only)

Needless to say, buying a house or a piece of land is a form of property acquisition and not consumption expenditure. Such a purchase, therefore, is not counted in the CPI. Still, it is an undeniable fact that a household living in a house it owns receives some service from the house. This service originally comes from the household’s purchase of the house and/or the premises. Also, many households are paying a mortgage. Here arises a problem that, one way or another, the housing expense of a owner-occupied house should be counted in the CPI calculation.
So, let us suppose a case where a house-owning household is renting the same house from someone else. Then, the household has to pay some rent, needless to say. Then, we can estimate the value of the service from an owned house which is equivalent to the worth of the service from a rent house and treat the value as housing expense. An “imputed rent of an owner-occupied house” refers to the rent paid to owner-occupied houses assuming that owned house were rented. Such imputed rents are taken into the CPI calculation.
To include imputed rents, we determine the calculation weight of each and every type (size and structure) of house, depending on the imputed rent estimated by the Ministry of Internal Affairs and Communications in its “National Survey of Family Income and Expenditure.” Then, we use the fluctuations in rents of ‘house rent, private’ by RPS, to estimate changes in the imputed rents of owner-occupied houses of all the various types.
Note that this method of calculating imputed rents is employed by many advanced economies in their consumer price indices as well as in the System of National Accounts (93SNA).

With these commodities that do not appear at all in the market during some period of time in the year or appear only for a strictly limited period of time, the RPS surveys their prices only during those months when each of them is distributed and sold. This naturally means that such commodities are not covered by the survey in some months (uncovered months). Now, in such months, the weights of those uncovered seasonal commodities are proportionately distributed to the other commodities covered by the survey. Yet this results in the average weight of the same item in some months differing from its annual average weight. To avoid this problem, the average index of a seasonal item from the last survey is applied to (carried over through) the months immediately before the next survey.

Accompanying the regular base revision of 2010, the RPS personnel added new items to the survey, and reconsidered the City groups covered by the survey and the numbers of samples to collect, in order to reflect changes in the structure of consumption by households more faithfully, in commodity distribution and services in the market, etc. In addition, we reconsidered the survey months for clothes and other seasonal commodities, in January 2010.
Consumer price indices whose base is 2005, however, which are to be calculated until December 2011, need calculating based on the weights determined in accordance with the consumption structure of 2005. For this reason, the Survey handles those items whose survey months were changed in January 2010 following the rules (1) through (5) below. You need to keep this change in mind as you read the changes over the year of those items calculated by comparison with January through December 2010.

(1) Items whose survey months end earlier than before
<EG> Men’s coats
Before the change: November through February of the following year → After the change: November through January of the following year

Since no survey takes place in February starting in 2010, the indices from January are carried over to the following February. (The change ratios for February from January, therefore, are 0.0%.)

In addition to Men’s coats, those items whose survey periods now end earlier than before include Men's suits (for summer), Men's suits (for winter), Men’s jackets, Men’s slacks (for winter), Women's coats, Women's jackets, Blouses (long sleeves), Blouses (short sleeves), Desks, and School knapsacks.

(2) Item whose survey months begin later than before (men’s jackets)
Before the change: September through March of the following year → After the change: October through February of the following year

Since no survey takes place in September starting in 2010, the index from August is “carried over” to the following September. (The change ratio for September from August, therefore, is 0.0%).

(3) Items whose survey months begin earlier than before
<EG> Men’s suits (for summer)
Before the change: April through September → After the change: March through August

Although from 2010, this item is covered by the March survey, the surveyed prices are not employed in the index calculation. As before the change, the March index is the average of the index from the most recent preceding survey (April through September of the year before).
In addition to men’s summer suits, those items whose survey periods now begin earlier than before include men’s pants (for summer), blouses (short-sleeved), reading desks, and students’ bags.

(4) Item whose survey months have been reset (Pencil cases)
Before the change: Throughout the year → After the change: December through February of the following year
As a result of this change, from 2010, no survey of Pencil cases takes place from March through November. The index for February is, therefore, carried over. (The change from the previous month is, therefore, 0.0% for March through November.) Note that the index value to carry over is not the average from the survey period but that of February, when the survey ends, since this particular item shows almost no seasonal change.

(5) Item whose survey months have now been abolished (Girls’ skirts (for summer))
Before the change: March through September → After the change: Throughout the year

As a result of this change, from 2010, the survey is conducted in January, February, and October through December as well. The results from these survey periods, however, are not employed in the index calculation for October through February of the following year. This index remains, as it has been, the average from the preceding period (March through September).

As we calculate the CPI, we classify commodities (including both goods and services) households purchase into 585 items, and then determine the weight (share in the whole consumption expenditure) of each item, depending on the amount spent on that item. Each item’s price are collected monthly. Now, in defining these items, we attempt to group commodities of similar functions into a single item, so that the various commodities classified under the same item will show similar price fluctuations. For instance, Fresh milk (sold in stores) and Fresh milk (delivered) are both “Fresh milk.” Yet their prices behave differently in reality, since Fresh milk (delivered) carries the added value of the delivery, which is a form of service provided. For this reason, we place Fresh milk (delivered) and Fresh milk (sold in stores) in two differentitem.
Ideally, as we try to learn trends in the price of each item, it would be best to survey the price and amount sold of every single product of an item at every single store, every single month, and then use a Fisher (ideal) index formula to calculate the index. Practically, however, it is impossible to survey all such product sales and prices at every store, every month. In view of this reality, we carefully set up the items to survey as described above, and survey the price changes of every item in terms of a single specification that represents the item.
Another feasible method of price surveys, for instance, covers multiple brands of a single item and determines the weight of each brand surveyed in accordance with the sales of each specification as of the base month. Then, the weighted average of the prices of these specifications is calculated. Yet any method that employs fixed weights can very often produce more deviant averages than the current method does, since changes in relative prices of the multiple specifications of the same item can actually lead to considerable changes in the shares occupied by the different specifications.
(“Items” and “specifications” employed in the CPI calculation)
There are numerous kinds of commodities (goods and services) households can buy. In calculating the CPI, these commodities are classified into 585 groups, according to similarities in their functions and price trends. Also, each of these 585 groups must have at least a 1/10,000 share in the whole household expenditure as found by the FIES. These 585 groups are called “items.”
Every single “item” contains several specifications of different qualities, standards, volumes, etc. To calculate the CPI, we choose out of those specifications one that represents the whole item. This specification is called the “survey specification”, and its monthly prices are collected. Thus, the “specification” in the CPI is the one chosen from the multiple specifications of a single item, after careful consideration of their qualities, standards, volumes, etc.. Sometimes, “survey specification” does not determine brands, but shows only the characteristics of the item. In other cases, only one brand is chosen and employed for single “survey specification”. There are yet other cases in which multiple brands are chosen and employed for single “survey specification”.
Considering differences in distribution among different regions, a price collector picks up one of the nominated products which sells best at each store surveyed, and its price is surveyed continuously. (For example, in case the item is “yogurt” and the specifications are “plain yogurt, 400 to 450g in a package [the survey specifications of yogurt as of October 2009],” the best-selling specification at each store surveyed is chosen).

Concerning the survey specifications, we investigate the distribution and sales of the major specifications of each and every item and, if we find any existing survey specification declining in distribution, we switch to another specification that is selling more. This switch takes place regularly, twice every year. In addition, we try to cover new products in the survey on a timely basis. For instance, in case a manufacturer abolishes the production of a survey specification and launches a new, follow-up product, we change the survey specification in response, in addition to the regular switches. We watch any major change in distribution. In fact, there are several dozen survey specification changes each year. Thus, we always keenly follow the market in an effort to survey the prices of specifications that truly represent the items.

* To see what items and specifications are covered by the RPS, see “List of Items” in “Retail Price Survey.”
See “Q & A about the Retail Price Survey” for questions about the surveys, such as:
Q Are private brand commodities surveyed?
Q Are bargain sales investigated?
Q Are discount stores included as survey subjects?

The CPI shows how much the average prices of commodities (goods and services) have changed since the base year. Goods and services purchased by households change with time, every time new commodities appear on the market or many consumers’ tastes change. For this reason, if the base year remains long, the index tends to deviate from actual conditions. Therefore, we revise the base year (base revision) regularly, reconsidering the items employed in calculating CPI and their weights, etc. In the case of the Japanese CPI, the base year is changed every 5 years, in years that end in “0” or “5” in the Western calendar. From a global perspective, many nations revise their CPI every 5 years or so. In Japan, the Statistical Committee said in its report in 2010 that revisions every 5 years were appropriate for the CPI as well as many other economic indices.
In addition, the items are subject to reconsideration and changes even outside the regular revisions whenever a new kind of goods and/or services appears on the market and gains considerable share in it.

Actually, two types of the CPI are published. One is the “official” one, whose calculation weights are based the shares of FIES and updated every 5 years. The other is called a “Laspeyres” index, calculated on a chain basis, whose weights are updated every year. This chain index, in order to measure effects of changes in the consumption structure more frequently, has been published each year since 1975. Since 2005, the chain index has been published every month as well, to reflect changes in consumption even more faithfully.
In addition to these, we also publish a “midpoint-year basket index”, calculated with weights that reflect the consumption structure of a year between a base year and an observation year. This is yet another way to reflect actual changes in the household consumption structure, including substitution effects.
Also note that, in Japan, the standards of most economic indices are updated every 5 years, including price indices such as the CPI and the Corporate Goods Price Index (CGPI), as well as quantity indices like the Index of Industrial Production and the Transportation Index.

At the base revision, held every 5 years, to the items selected for use in calculating the CPI, we consult the results from the FIES to choose those items occupying a share larger than the set minimum in the whole expenditure. The revision held in 2005 added some new ones to the survey items, especially information technology devices, such as TV sets (LCD) and DVD recorders.
Furthermore, reconsideration of and revisions to the survey items are to be made in years other than the base years as well, as we expect some new goods and services will emerge and enter the market in the future.

Since we change the items and update the calculation weights at the base revisions, no doubt the CPI before and after the revision refer to two different things, precisely speaking. In order to facilitate timeline analysis of price trends and changes over a long period of time, however, we do some processing to link old and new CPI, at every revision.
To do so, we convert the old-base CPI, using the ratio of the old annual average indices to the new indices (the new-base indices = 100). This adjustment is applied to each and every article of calculation independently and the upper level indices by the linked indices are not recalculated. Also, the rates of change (from the previous month, over the year, from the previous year and from the previous fiscal year) are not recalculated with the index adjustment. Their respective figures, based on the old and new bases, are published and used as they are. In each base year, also the figures calculated with the old-base are used as they are, for the change from the previous month in January, the change over the year from January to December, the change from the previous year, and the change from the previous fiscal year.

Considering the rates of contribution (influence) of to the difference by the revision, -0.5, the difference between the bases of the change over the year of the “General (All items)” of Japan in July 2006, is mainly composed of -0.16 by “Transportation & communication” group and -0.18 by “Reading & recreation” group.

Now, we can consider the following three factors as the causes of this decline between the old and new bases:

(1) Many consumers’ purchases shifted to those commodities whose prices fell remarkably.
In the ”Transportation & communication” group, “Mobile telephone charges” almost tripled its weight from 74/10,000 to 208/10,000. At the same time, many households switched to mobile phone contracts with greater discount.

(2) Influence of new information technology devices
The “Reading & recreation” group was greatly affected by information technology devices such as “TV sets (LCD)”, added in at the base revision, whose prices fell drastically. This had a tremendous impact on the General (all items) index’s decline.

(3) Indices for items showing a remarkable price drop were reset to 100.
The index for “Personal computers (notes),” whose prices fell remarkably, was about 16 at the old base. The base revision reset the index to 100, resulting in some considerable contribution to the General (all items) index’s decline.
(100 / 16 nearly equals 6).

For more specific examples, please consult the file below:
“Factors contributing to the difference between old and new bases” (July 2006, Japan) (PDF:17KB) (in Japanese only)

A base revision can reset the index of an item to 100. Such a reset alters the contribution of the item to the total change, even with its weights remaining almost the same. This effect is mainly ascribable to the “Index of Item A in the previous period” in the numerator of the formula below, which calculates the contribution of an item A to the total change:

Durable goods, such as Personal computers, Cameras, TV sets (LCD), etc., currently show extremely low indices after adjustment was made to reflect quality improvement (See D-1.), which was necessitated by accelerated technological innovations in these products. With these items whose indices have declined after quality adjustment, an index reset can have greater impact than with other types of items.
It is possible to estimate how much influence a reset has, in advance (see example below).

The CPI, since it is meant to purely measure changes in prices, should basically continue to track down the prices of the same items (goods and services). Some of the items surveyed, however, can become obsolete and no longer reflect price changes. Production of some items can be discontinued, or replaced by new items. In these cases, we need to replace the items to survey. In such a replacement, the differences in functions, volume, characteristics, etc., between the old and the new can alter the price index. Now, the “quality adjustment” is applied to eliminate the influence of such differences.
With the CPI, whenever items to survey are updated, we have quality adjustment, choosing the best method from several such as the overlap method, adjustment by the quantity ratio method, the regression method, the option cost method, the class mean imputation method, the direct comparison method, and others. For PCs and digital cameras, whose quality improvement is very rapid and product cycles are quite short, the hedonic method with some information from POS systems is employed to directly calculate price changes after quality adjustment. This method can also be used for quality adjustment when the survey specification is changed for PC printers.

(POS information)
This information contains each product’s sale price(s) and volume, as well as its characteristics, and is collected from retailers’ POS (Point Of Sales) systems. The CPI calculation employs POS information on each and every product’s sale prices, volumes and characteristics for all the relevant products sold in major mass retailers, PC specialist stores and other outles in our country (about 3,400 outlets).

The hedonic method, one of the methods used in quality adjustment, stands on the supposition that the quality of a product can be reduced to several characteristics (performance) that make up the quality, and that the product’s price is determined by the performance. The method uses multiple regression analysis to identify the relationship between each product’s price and its characteristics (for instance, in the case of a PC, its screen size, the CPU clock frequency, hard disk drive capacity, etc.) This way, the method quantitatively determines to what degree the product quality accounts for price differences among different products.
The CPI calculation employs this hedonic method to directly calculate price changes after quality adjustment of PCs and digital cameras, whose technological improvement is very rapid and product cycles are quite short. Since this method requires a huge amount of data on prices, volumes, and characteristics of numerous products in order to provide a reliable and objective multiple regression analysis, we use POS information with the hedonic method.
In case we change the survey specification of PC printers, we consider the hedonic method as one of the alternative ways to perform quality adjustment. From 2003 through 2007, the survey specification was changed in October each and every year, with a linking coefficient calculated with the hedonic approach to bring together an old index and a new one. In 2008, however, direct comparison was employed in place of the hedonic. In the years to come, whenever we change the survey specification of PC printers, the hedonic method is to be considered as an alternative way of quality adjustment, if there is a remarkable improvement in the quality of printers.
Also note that, in applying the hedonic method, we compare an index calculated with the method to one calculated with matched-model method, in addition to verifying how well the regression model fits the raw data, the significance and consistency of descriptive variables, etc. The matched-model method uses the Fisher formula to calculate the price change ratio of a single product that sold well both in the current month and the month before, using its price and sales volume in the two months. In fact, as far as PCs are concerned, there has not been much difference between the index computed with the matched-model method and that with the hedonic method, which shows that the latter method has not resulted in excessive quality adjustment.

[For your reference]
“About the Hedonic Method(PDF:643KB)” by Shimizu, Makoto & Nagai, Keiko(in Japanese only), Published in the monthly magazine Statistics - November 2006, Japan Statistical Association.

In reality, the hedonic method calculates the relationship between the price and the characteristics (the performance) of a product statistically, using a huge volume of sales data (POS information). Therefore, actual calculations of how each element of a product’s performance affects its price do not result in half the price when the product’s performance doubles.
For example, the “Annual Report on the Consumer Price Index - 2007” contains some hedonic calculation results for November 2007, which suggested that “a PC’s price rises by 2.9%, when the HDD capacity expands by 10GB.” Following this, as illustrated below, when we “compare old and new prices, they deduct 2.9% off the actual price of the PC when a new version of the PC whose HDD capacity has increased by 10GB is released.” Thus, in case a PC’s HDD capacity has increased from 80GB to 160GB, memory from 512MB to 1GB, and CPU upgraded to dual core, the price consumption is reduced by 36.0% compared to the actual new price, as quality adjustment.
Thus, the term “performance” itself requires careful definitions, and there is no simple way of determining quality adjustment for performance improvement.

For specific methods of calculation, consult "2005-Base Explanation of the Consumer Price Index", p.41-42: "Appendix 1 Calculation of the price index for PCs and Cameras by the Hedonic method (PDF:1,762KB)"
Also, for the regression models and other formulas, etc., employed in actual price index calculations for 2007, consult "Annual Report on the Consumer Price Index - 2007", "Appendix 6 The Hedonic regression formula for PCs and cameras (PDF:21KB)" (in Japanese only).

[For your reference]
The Corporate Goods Price Index (CGPI), calculated by Bank of Japan, involves quality adjustment that employs the hedonic method with PCs, video cameras, etc.
“Application of the hedonic method in the Corporate Goods Index with a 2005 base” (in Japanese only)

With CPI calculations, 59 items, namely airline fares, electricity, mobile telephone, etc., the fare/rate systems are complicated and diverse, and it is hard to select a single specification to represent each item. Also, the fares/rates themselves are also diverse. To make the CPI correctly reflect changes in these rates and fares, therefore, we calculate the monthly indices for these items using the special calculation methods which employ prices from the RPS as well as other data obtained from routine statistical work, etc.
For how to calculate CPI with these special calculation methods, consult "2005-Base Explanation of the Consumer Price Index", p.43 -70: "Appendix 2 Special calculation of prices in the observation period and the price indices for some items (PDF:1,762KB)".

To calculate the CPI for mobile telephone charges, we calculate a weighted average of prices of the three major carriers, namely, NTT DoCoMo, KDDI, and Softbank. The weights are determined in accordance with the numbers of subscriptions * (updated yearly) to the respective carriers and the telecommunication systems (2nd generation, 3rd generation).
The CPI calculation follows the following regulations in handling discounts, etc.
(1) We do not adopt “family discounts,” since their numbers of subscribers, actual use among subscribing family members, etc., are not available.
(2) We do not adopt discounts set to the years of use as well, since the numbers of subscribers after respective numbers of years of subscription, the average number of years of subscription, etc., are not available.
(3) We also do not adopt those toll plans conditional on the number of years of subscription, since these are different from “common” plans in that the number of subscribers covered by such conditional plans are not available, and in that these plans have restrictions with respect to the time period of subscription as well as a penalty for breaking the contract.
The actual CPI calculation selects the “most reasonable toll plans,” considered to be the most favorable to consumers, out of three types of “common” toll plans offered by each of the three major carriers, namely, <20-minute talk, 4,100-packet communication>, <200-minute talk, 11,300-packet communication>, <660-minute talk, 23,400-packet communication>. The “common” toll plans here refer to those that meet the criteria mentioned just above. These three types have been determined by totaling the amounts of cellular phone toll expenditures taken from the FIES. More specifically, we calculate the arithmetic product by multiplying the number of households in each “class” of mobile phone toll expenditure by the amount of expenditure (medium value in each class of values), and accumulate the amounts of expenditure in each value class. Next, we divide this cumulative product into three to calculate the average expenditure for each of the three classes. Then, for these averages, the length of talk and amount of packet traffic are estimated.

* The numbers of subscribers to each cellular phone carrier (Telecommunications Carriers Association)

Currently, we choose the most reasonable plans available to consumers, after comparing the toll plans available without any restrictions of subscription time length, etc., regardless of the actual numbers of subscribers to such plans, since no information is available on the numbers of subscribers to each toll plan of each carrier, the numbers of subscribers switching to different plans, etc.

“Daredemo Wari” (introduced on September 1st, 2007), “Hitoridemo Wari 50” (on August 22nd, 2007), “Shin Jibun Waribiki” (September 1st, 2007), and “Jibun Waribiki 50” (September 1st, 2007) all provide the maximum discount possible on the monthly basic use charge regardless of the time length each subscriber has been using the phone service, on condition that the subscriber holds on to his/her plan for two years or longer. “Fami Wari MAX 50” (introduced on August 22nd, 2007) and “Kazoku Waribiki MAX 50” (on September 1st, 2007), meanwhile, cuts down the monthly basic use charge to a half, again regardless of the time length each subscriber has been using the phone service, on condition that the subscriber holds on to his/her plan for two years or longer. In the cases of the last two plans, the subscriber has to have signed up for a discount plan on transactions with members of his/her family.
Based on the regulations described under E-2 (a) above, these discount plans carrying a condition or available to transactions among members of the same family alone are excluded from the CPI calculation. This exclusion will remain for the time being.
Also, as of August 2007, KDDI offers a discount plan named “MY Wari” (“My Discount”), which is to automatically transfer to “Daredemo Wari” in September 2007, and Softbank “Jibun Waribiki” (“My Own Discount”), which should automatically transfer to “Shin Jibun Waribiki.” Both of these two plans carry a condition of a minimum of 2 years’ staying with the same plan. Therefore, none of these is employed in the CPI calculation.

“au Kaikata Select Simple Course” and “Simple Orange” (both introduced on November 12th, 2007) and “Blue Plan Value” (on December 5th, 2007) have the condition that a subscriber must sign up a new subscription with the carrier or buy a new cellular phone handset. “Value Course” (introduced on November 26th, 2007) requires that a subscriber should buy a new handset of the 905i Series or later. An existing subscriber cannot switch to any of these toll plans without meeting one condition or another.
In this respect, none of the plans is considered as we look for the most reasonable toll plan available. Thus, we choose the most reasonable one from the toll plans already offered before the introduction of those ones mentioned here. In more specific terms, these “toll plans already offered before the introductions” include KDDI’s toll plans already offered on or before November 11th, 2007, NTT DoCoMo’s plans other than “Value Course” and “Basic Course,” and Softbank’s plans other than “Simple Orange,” and “Blue Plan Value.”

For the time being, the CPI calculation will use a plan offered before any of the four plans mentioned, “au Kaikata Select Simple Course,” “Value Course,” “Simple Orange,” and “Blue Plan Value,” were launched, for two reasons. Firstly, there is no way to know, at this moment, how many subscribers will switch to the four plans. Secondly, existing subscribers (of the other plans) are required to buy a new handset to sign up for any of the four.
In the future, as the four plans grow and they become available to more subscribers, and if it is confirmed that “au Kaikata Select Simple Course,” “Value Course,” “Simple Orange,” and “Blue Plan Value” have become mainstream toll plans, the CPI calculation might take the four into consideration in the selection of the “most reasonable” toll plan.

As far as new subscribers to the carrier who join in the two plans and those existing subscribers switching to a new handset and either of the two plans are concerned, there is no difference. With the two plans, however, an existing subscriber to Softbank can switch to either of the two with no requirement to meet, while “au Kaikata Select Simple Course,” “Value Course,” “Simple Orange,” and “Blue Plan Value” require subscribers to buy a new handset. This is where the difference lies.

Yes, they are, since the “CDMA1x WIN” toll plans and “Orange Plan (WX)” are “common” toll plans with no requirement imposed on those wanting to sign up for them. Therefore, they are considered in the selection of the most reasonable plan.

In most cases with the RPS, the survey specifications for the next month are published at the end of each month (the publishing day). This holds true for the cellular phone toll plans and services surveyed. Together with how they are taken into the calculation, they will be published on the website.(Japanese only). In some exceptional cases, however, if for instance a new service is launched all of a sudden or a new plan / service takes some time before it is determined whether or not to include it in the calculation, the above regular rule cannot apply.
Also, for some of the items whose index is calculated with the special calculation method, some classified corporate information is employed. Such basic data, therefore, cannot be published.

The monthly survey of mobile phone tolls begins on the Friday of the week including the 12th of each month.
Meanwhile, the day when a new toll plan becomes effective can change, depending on the last day of each subscriber’s phone toll month. While for all subscribers to NTT DoCoMo and KDDI the toll month ends on the last day of each calendar month, Softbank has three different toll months, which end on the 10th, 20th, and end, respectively, of each calendar month. For this reason, the first day a new toll plan applies to a subscriber differs depending on his/her toll month. In case the earliest date when a new toll plan applies falls on any date before the calendar month’s price survey, the toll plan is considered in that month’s CPI calculation.

The new plan is applicable before the month’s price survey
In this case, the new plan is applicable before the month’s price survey. The new plan, therefore, is employed in the CPI calculation for the month.

The new plan has yet to become applicable as of the current month’s price survey
Since the new plan has yet to become applicable as of the current month’s price survey, the plan is employed in the CPI calculation for the following month.

Since we calculate the CPI weighted arithmetic mean with a fixed basket in the base period preceding the observation period (Laspeyres method), basically there is no plan to change the weights or the special calculation methods as long as the 2005 base holds. Still, in case such plans that separate the handset’s price and the phone toll become the mainstream, we might reconsider the relevant special calculation methods, etc.
Meanwhile, in order to reflect changes in the consumption structure of households in the indices, we will annually update the weights used in calculating the consumer price indices by the Laspeyres' Chain Index Method (reference indices).

The tolls of “National expressway tolls” and “City expressway tolls” are surveyed on the Friday of the week including the 12th of each month. If the day when a new toll system becomes applicable falls on a day before that month’s price survey, the new system is employed in that month’s CPI, on the assumption that the new toll system is applicable to all the days of the week.

Both the “social experiment ahead of schedule” and the “Comprehensive Immediate Policy Package” have been reflected in the indices starting with their results in October 2008.

The new reduction, applicable to national expressways and those vehicles that pass through ETC gates wirelessly, applies specified discounts during specified hours of the day and days of the week. (For more details of the expressway toll reduction, see the website of the Ministry of Land, Infrastructure, Transport and Tourism) (in Japanese only).

In the CPI calculation, “expressway tolls” cover tolls for standard-sized vehicles in both the Distance toll system (of both the ordinary [rural districts], and special [or major city districts, i.e., Tokyo and Osaka regions]) and Flat toll system. The tolls as of the day of the monthly RPS are used in the CPI calculation. (See E-3 (a).) In considering discounts for cars equipped with an onboard ETC unit, the toll rates are obtained by averaging the ordinary and ETC-discounted toll rates by weight of the ETC utilization factor (fixed at the mean rate of 2005). As for the said ETC-discounted rates, reduced normal toll rates are obtained by averaging the rate of discounts by hour of the day by weight of traffic volume (common to weekdays and holidays, and to all road sections; fixed at the 2005 figures) using available information, and further averaging the result by weight of the numbers of weekdays and holidays (deemed 5 to 2). For this reason, the ETC discounts, whose rates differ depending on the hour of the day and the day of the week, are to be reflected in the CPI, as they have been so far.

For the method of calculating the index, see "2005-Base Explanation of the Consumer Price Index" (p.56), "Appendix 2 Special calculation of prices in the observation period and the price indices for some items (PDF:1,762KB)").” Please read the complementary explanation below as well.

Following what is described in E-3 (a), the “rural maximum toll of 1,000 yen on weekends and holidays,” to take effect on March 28th, is to be reflected in the CPI for April 2009. Since, however, with the 2005 base it is impossible to learn the traffic volumes of all the different driving routes (rural-to-urban-to-rural routes, other routes, etc.), the other toll reduction, “the same maximum toll applied to a rural-to-urban-to-rural drive for rural districts on weekends and holidays” beginning on April 29th cannot be accurately counted in. In short, the CPI calculation can only cover the maximum toll of 1,000 yen in rural districts, on weekends and holidays. (For more details on the expressway toll reduction to begin in March 2009 as part of the “Economic Policy Package:Measures to Support People’s Daily Lives,” etc., consult the website of the Ministry of Land, Infrastructure, Transport and Tourism) (in Japanese only).

For the driving distances to which the maximum of 1,000 yen applies, we adjust the ETC discount ratios to provide the toll of 1,000 yen and use these adjusted ratios in the CPI calculation.

The “social experiment for free expressways 2010,” to begin on June 28th, 2010, is to be reflected in the CPI for July 2010 and onwards. (For the relationship between the first date of a new toll system and the timing of its reflection in the CPI, see E-3 (a)).
When including the effects of the social experiment for free expressways in the CPI calculation, the indices for different expressway segments (ordinary and special [major city regions], etc.) * are multiplied by the discount ratios of those segments not made free, as shown below.

* The indices of “National expressway tolls” are calculated based on the average of the indices for the different toll classes weighted by the corresponding business revenues.

(For more details of the “social experiment for free expressways 2010,” see the website of the Ministry of Land, Infrastructure, Transport and Tourism) (in Japanese only).

The RPS selects a specified number of schools from each of private junior high schools, high schools (both public and private), universities (both national and private), junior colleges, and professional schools, in the order of enrollment size, *1 and surveys their tuitions and admission fees.
The CPIs for School fees are calculated monthly, by
(1) Summing up the “average tuition *2” and “average admission *3” of each grade of each chosen school, and
(2) Calculating the average of the above sums weighted by each school’s enrollment size.

*1 The schools with the largest enrollments are chosen, from private junior high schools and high schools (both private and public) of the municipalities covered by the surveys, and from universities (both national and private), junior colleges, and professional schools of each prefecture.
*2 As tuition fees of private junior high schools and high schools (both private and public), the monthly amount of each grade is employed each month, which is calculated by dividing the annual tuition fee of each grade by 12.
*3 As admission fees, the monthly amount of the fee paid by a student of each grade when he/she entered the school is employed each month. For instance, a second-grade high school student’s admission fee is the amount he/she paid in the year school year before. The monthly amount is calculated by dividing the admission fee paid by 36, in the case of junior high and high schools.

These have already been reflected since the CPI for April 2010. *1 Note, however, that the index of “High school fees, public” is not 0.0, since the index reflects admission fees as well, which new students still have to pay.
Meanwhile, the RPS records “0 yen” for public high schools that are tuition-free. With private high schools, the tuition fees paid by students, after including the “High school enrollment support fund system,” are considered in the CPI calculation. *2

*1 For more information on the influence (degree of contribution) in the General (All items) index of “Free tuition fee at public high schools and high school enrollment support fund system”, etc., consult:
“The influence of new subsidies for tuition and enrollment fees for high schools - Japan Results for April 2010” (PDF:18KB)(in Japanese only)
“The influence of new subsidies for tuition and enrollment fees for high schools - the Ku-area of Tokyo, Results for the middle of April 2010 (PDF:19KB)”(in Japanese only)
*2 Although there is additional aid for those households whose annual income is less than 3.5 million yen or so, this is not considered in the index calculation, since it is applicable only to some households.
For the treatment of “High school fees” in the RPS, see “Q & A about the Retail Price Survey”

The CPI includes almost 50% more (in terms of the calculation weights) items than does the CGPI, which does not cover such items as School fees, Rent, Eating out, and so on. Also, prices of services tend to fluctuate less than those of goods, since labor costs account for more service prices than prices of goods.
Also, the CPI covers goods purchased by households and does not include crude oil and other raw materials, intermediate materials like electric components, construction machinery and other types of equipment, etc. Thus, a price hike in these goods affects the CPI only indirectly.
For these reasons, the CPI and the CGPI do not necessarily show similar changes.
Note, however, that, if the scopes of the two indices are narrowed down to cover the same items as far as possible, i.e., if we compare the CPI for “Goods excluding fresh foods” with the domestic CGPI for “Final consumption goods” alone, the two indices show similar fluctuations.

The discrepancy between the CPI and the GDP deflator is mainly ascribable to the different things they cover. Other causes of the discrepancy include, among others, the different calculation formulae employed.

(1) The Target
While the CPI focuses only on household consumption, the GDP deflator covers business investments in equipment, etc., in addition to household consumption. Since much of the investment in equipment today is made in information technology goods, whose quality is rapidly improving, price falls in such goods considerably affect the deflator. For this reason, the change ratios of the deflator tend to be lower than those of the CPI.
Also, while the prices of petroleum products and other imported goods are rising, the CPI is usually pulled up. On the other hand, the deflator tends to drop until such price hikes are all reflected in the relevant product prices. Thus, the discrepancy between the two grows wider.
If the scopes of the two indices are narrowed down to cover the same items as far as possible, i.e., if we compare the General (All items) CPI with the GDP deflator for “Final household consumption expenditure” alone, the two indices show quite similar fluctuations.

(2)The Formula
While the CPI calculation employs the Laspeyres formula, the GDP deflator employs the Paasche.
Generally, the Paasche formula, which calculates a weighted average using the quantitative weights at the time of comparison, tends to provide a lower index, while the Laspeyres, which employs quantitative weights at base period, usually produces higher values. In addition, since quality improvement is reflected in the form of an increase in volume, Paasche gives a larger weight to an item whose price has fallen due to quality improvement. For this reason, the GDP deflator, which employs Paasche, tends to show larger decline ratios.
Also note that the GDP deflator employs a “chain method” with the base years – it updates bases, to minimize the bias accompanying calculation of the index. Such a chain method is also used with the CPI as well, to provide and publish an additional, referential value to the index.

Although we often simply talk about “TV sets,” the plain fact is that there is a huge diversity of TV sets on the market --- from a small one with limited functions and a friendly price tag, designed for personal viewing, to a large, expensive set with many functions, made for a whole family to watch together in the living room. Consumers, out of such a huge variety available, select the right commodities (goods and services) that suit their budgets and needs. The FIES reveals the average purchase prices of and other data about major consumer electronic products, foods, etc., that consumers buy. These average purchase prices provide the average purchase prices of those items purchased, disregarding the quality (functions and characteristics) of individual products. The CPI, for its part, tracks price changes for commodities of the same quality and, when any of the subjects of the survey are changed, excludes the portions of prices changes that are ascribable to quality differences between the obsolete products and their new counterparts. This is because the CPI is intended to measure changes in prices themselves.
Here, comparison is made between movements of the average purchase price index and the CPI, in a specific example.
Before the “bubble” economy burst, the average purchase price of a TV set almost leveled off, which suggested that many manufacturers sold new TV sets with enhanced functionality in the same price ranges, with many consumers buying them accordingly. After the “bubble” burst, however, with the phenomenon called “price collapse” pervading in society, the average purchase price tended to decline. Then, in 2004 and 2005, the average suddenly picked up, partly thanks to the expanding sales of flat-screen TV sets, whose unit prices were higher. Turning our eyes to the CPI of TV sets, the index has always been on a steady decline since even before the bubble burst, since the CPI calculation excludes price rises ascribable to functional improvement. (Figure 1)

With foods as well, transitions in an average purchase price and the CPI can differ from each other, as many consumers’ tastes, etc., change. For instance, the average purchase price of 1 liter of milk has been on a greater decline than that of the CPI, as more consumers have shifted to low-fat milks, whose unit prices are lower. Meanwhile the average purchase price of 100 ml of soy sauce has been rising more rapidly than has its CPI, with more consumers favoring soy sauces from “raw” (not defatted) soybeans. Such soy sauces are more expensive (Figure 2).

As shown so far, the differences differ between movements of average purchase price indices and of CPIs, depending on the item. Generally, with TV sets and other consumer electronic devices, whose functions are always improved by one technological innovation after another, the CPIs tend to drop more sharply than do average purchase prices. Elsewhere, like Soy sauce, with an item where more consumers go for luxury specifications when the economy is picking up and switch to cheaper ones in economic slumps, the average purchase price usually fluctuates more drastically than does the CPI.

This opinion seems to be from a paper written by a certain official of Bank of Japan, which was published in 1998 to describe the author’s personal views. This paper has many questionable points. For example, the author makes some estimate based on the US method of indexing, which is completely different from that of Japan, with all its biases (See also “ Accuracy of CPI”) (in Japanese only).
To inquiries concerning this paper, Bank of Japan responded that its claims were nothing but the author’s own opinions and not the official ones of the Bank at all.
There is a great variety of arguments about the accuracy of the CPI and we, the Statistics Bureau, intend to learn more and undertake more research on it. (See “Documents To Study about CPI”) (in Japanese only).
The CPI tends to become higher than the actual price level as the time after the base year grows longer. This is because the current CPI calculation employs the Laspeyres method, which employs items and weights fixed to the base year. Seen from the 2000 base, the “General (All items)” and “General (All items) excluding fresh foods” indices of 2006, 6 years after the base year, carried Change over the year higher by 0.4 to 0.6 points than those compared to the latest base year, 2005.