HANDBOOK OF THE INTERNATIONAL COMPARISON PROGRAMME.

4. PRICE ESTIMATION AND ADJUSTMENT PROCEDURES


1. Indirect price comparisons
1. Education
2. Collective consumption of government
3. Medical services
2. Hedonic estimation and other price-slope adjustments
1. Hedonic price estimation
2. Price-slope adjustments
1. Size of purchase and package
2. Consumer durable and machinery and equipment
3. Construction
4. Some miscellaneous pricing issues
1. Equivalence in use
2. Identity of product

4. PRICE ESTIMATION AND ADJUSTMENT PROCEDURES

146. This chapter takes up a number of practical problems encountered by countries in providing prices and related salary and direct quantity information for the ICP. Indirect price estimation is discussed in section A, and price adjustments, including hedonic regression estimates, as well as a number of other price-slope adjustments, are taken up in section B. Section C discusses the construction sector, and section D discusses some miscellaneous questions, including equivalence in use.
147. Throughout this chapter a recurring theme is that the prices used in the ICP must be consistent with the way quantities have been valued in the national accounts. Food items, such as rice, mealy meal, cooking oils or sugar, are often distributed through a ration system at subsidized prices. Usually, there is a second market with higher prices. Whatever the system, the guiding principle for the ICP is that, however the item is valued in the GDP, that is the price that should be used for comparison. Some countries, for example, value a price-controlled cereal at the legal price, even though there are many "illegal" transactions at higher prices. In such a case, the price for ICP purposes should be the legal price because it is the price that when divided into expenditures will yield the correct quantity. In countries where two markets exist and the item is valued at the quantities in each market, the ICP price should be the quantity-weighted average of the controlled and free price; the two markets should also be reflected in the national accounts.

148. Indirect price comparisons are used when it is difficult to obtain direct price comparisons, most usually where output is not sold, as in general government, education and for some medical services, depending very much on the country. Before taking up specific sectors, it should first be noted that the problems posed for the ICP in health care, education and government are analogous to the problem of obtaining quantity comparisons over time for the national accounts. Suppose all health services were provided free of charge by government. To obtain a deflator for national accounts, it would still be necessary to know what part of the increase in government expenditures on health from the previous year to the current year are due to increased quantities of services provided and how much are due to increased costs of inputs per unit. Health expenditures may rise due to increased cost of inputs per unit, but if there are productivity gains by providers of health services, then part of the input price increase is absorbed by productivity gains. If there is some measure of changes in health service prices over time, then the deflation problem becomes much simpler.
149. This line of reasoning leads to the principle that comparisons should be as close to the final consumption of the service as possible. Consider the case of hospital services, which in most ICP work are compared at least partially on an input basis. It is possible to obtain direct quantity comparisons for hospital services, which are number of beds, or number of bed days. Clearly, bed days comes closer to final use than does simply the number of beds. But bed days does not hold constant the service that goes with a stay in a hospital bed. One could come closer to service by comparing also the number of nurses and physicians, as well as equipment, that are part of the package provided by a hospital in a bed day. In either case, a direct quantity measure is being used to approximate the volume of hospital services. By dividing, say, bed days into hospital expenditures, one would obtain an indirect price measure, namely, expenditures per bed day, that can be used to estimate price ratios across countries.
150. If there are hospitals in all countries that bill patients for the full cost per bed day, then it is better to simply compare charges per bed day. However, since cost per bed day depends on the medical service involved, it would be better still to know cost per bed day for an appendectomy. A still better comparison would be to simply compare the total charges in two countries for an appendectomy of "identical quality", separating the operation costs and surgeon's fee from the hospital charges. Whichever bed day price were used, the quantity would be indirectly derived by dividing price into expenditure. While moving closer to the final service usually improves comparisons, the extent to which health services can be matched across countries will remain less than for most commodity and many service comparisons.
151. In the hospital example, both indirect and direct price and quantity comparisons are considered. There is another intermediate approach that bases comparisons on prices of inputs rather than prices of outputs. This is most common in education and government, where output prices are not known but where quantities and prices of the main input, labour, are known. The advantages of using wages and salaries versus numbers of employees will be discussed below. This type of comparison is also termed indirect because the prices of outputs are being inferred from the prices of inputs.
152. The principle of basing indirect price comparisons on indicators as close to final output as possible is also appropriate for deflators over time in a country, as well as for comparisons across countries where it may be more difficult to hold quality constant. The organization of this section is to move from quantity-based comparisons to those more directly related to the price of the final service. We begin with education, move to general government, and then to health services. 13/

153. Schools are run by governments, by private groups for profit and by non-profit organizations, often religious. Some educational services provided by government may also involve fees to users that cover part of the costs, with the remainder financed by taxation. The services provided by private non-profit institutions and government are also included in the final consumption of households at their cost.
154. There is very little basis for making price comparisons for education on the basis of tuition or fees because they usually do not cover full cost. However, in some regions there are enough private schools with similar financing patterns that it is possible to use school fees as a price of some portion of educational expenditures.
155. The two main quantity indicators available for education are number of pupils and number of teachers, usually broken down into primary, secondary and higher education. These inputs do not, of course, measure the output of education. Attempts have been made to compare student performance on standardized tests across countries, but as yet that work is too fragmentary to offer any alternative quantity indicator. When numbers of teachers have been used, attempts have been made to standardize their quality based on their education. Because it is thought that number of pupils is also a measure of educational output, both teachers and pupils have at times been combined to obtain one direct quantity indicator for education. When this approach is used, the teacher-pupil measure is assumed to be the quantity of educational output, and the price is indirectly obtained by dividing the quantity into expenditures.
156. This direct quantity approach has been thought less satisfactory than use of the price of inputs. Salaries of teachers are a major cost in education and it is usually possible to compare average salaries for similarly trained teachers across countries. It is simpler to control for the quality of teachers by salary comparisons than to adjust quantities. If price parities for education are approximated by salary comparisons, it is assumed, given the same non-salary expenditures, that the educational output of a teacher of similar training is the same across countries.
157. Price comparisons for non-salary expenditures in education are imputed from parities for related detailed headings such as fuel, transport, publications and the like. Where possible, depreciation charges are separately treated and price comparisons are imputed from parities for maintenance expenditures. This is one of several ICP expenditure headings where parities are imputed from parities from other categories.

158. Conventionally, the value of the output of these non-market and therefore non-price collective services is measured by the costs of providing them. The most important cost incurred in all cases is the compensation of employees, but total costs also include intermediate consumption and capital consumption outlays. Variations in labour costs, however, do not properly reflect differences in productivity, although, in the techniques applied in ICP, some attempt is made to take account of the different skill inputs. This is done through the explicit categorization of employment by occupation type, qualifications, length of experience, and skill levels.
159. In calculating the compensation of general public administrative services, including defence, details of the occupational wage structure are required. Estimates are also needed of employers' and employees' total pension, superannuation and social security contributions, as well as of other benefits normally provided as part of total effective personal emoluments. Total compensation for each occupation compared should be an average of all levels of government. Some typical job descriptions for general government are included as core items in the core commodity list.
160. The parities for non-compensation parts of government are usually imputed from parities for related components for which price comparisons have been carried out, such as clothing or publications. A set of weights based on typical expenditure distributions are used to average the imputed parity over all of government intermediate consumption.

161. The reader may recall that according to the ICP practice, health consumption covers not only expenditures on health by households but also expenditures by government (e.g., services provided free of charge by government hospitals). This is one of the few differences between the 1968 SNA expenditure breakdown and the ICP breakdown of the main expenditure categories.
162. One set of problems in obtaining parities for conversion from national currency to a common currency unit for health services stems from the fact that in a number of countries there are at least three ways in which the pricing of these services can be distinguished from one another:


1. Health services provided by private practitioners or by private health units (e.g., in private hospitals);
2. Health services provided by government or non-profit health units (e.g., government or non-profit hospitals), which, however, still have a market character. These health services may be cheaper than similar private health services; nevertheless, the price at which they are sold covers the whole, or a least a substantial part of the cost of these services;
3. Health services provided by government health units (e.g., government hospitals), which have less of a market character. These government health services are either entirely free of charge or provided for only a nominal fee.


163. The health services provided in each of these three types of units may or may not differ substantially from one another in respect of their qualities. However, there is no price, only costs, associated with free health care, so the only basis of a price comparison would be to assume it is the same or some percentage of the private charge, or to base the comparison on costs. This is an area where ICP methods continue to evolve.
164. What countries actually do is some combination of input and output comparisons, trying always to get as close to output as possible. In practice, this has usually meant carrying out price comparisons for all of Medical Care and Pharmaceutical Products (1.5.1) and Therapeutic Appliances and Equipment (1.5.2), and comparisons of medical and dental fees for well specified procedures for Medical Services Outside Hospitals (1.5.3). While, in some countries, insurance or subsidies operate to reduce the price to consumers, in most cases, the practitioner charges full fee, is reimbursed for part or all of the fee, and the residual is paid by the consumer. For all three of these summary categories, it is possible to specify some drugs or medical procedures as part of the core commodity list.
165. Most free services or medical services charged at less than cost occur in Hospital Care (1.5.4), which also include government clinics. Here, the basis of comparison has varied during the ICP, but a common method is to break up these expenditures into salaries of professionals and of other workers, maintenance expenditures, such as food, fuel, laundry and the like, and depreciation of hospitals. Wage and salary comparisons are carried out for specified occupations, for example, physicians, lab technicians, nurses, orderlies, or maintenance workers, where it may be assumed that there is no or some objectively estimated productivity differential in these workers across countries. The core items for such comparisons become the occupation specifications for which wages and salaries are compared. Usually, the comparisons for other hospital costs are imputed from parities for analogous detailed headings, for example, food purchases from foods or maintenance from fuels and repair charges.
166. If it is possible to obtain hospital costs per bed day for specified types of illness, and for specified rooms, such as double or four or more, this is an alternative to the input comparison. Sometimes, this information can be used to supplement data on input prices. In any event, it is important enough as an approximation that it is also included in the core commodity list.

167. Hedonic regressions are appropriate when there is a strong relationship between objective quality characteristics and the market price. Hedonic regressions were originally used within a country for holding constant the quantity of consumer durables over time by isolating the price augmenting effect of additional features of the item in temporal price index construction. These techniques have a natural application across countries where it is hard to match identical items but where price-determining characteristics of the good are identical. Once the key price-determining characteristics of an appliance or a vehicle have been identified and their empirical relation to price determined, estimates of price in different countries for given characteristics of, say, a rental dwelling can be easily made. These techniques have been used for a number of producer durables, automobiles and some appliances, as well as house rents, for which some detail is given below. In general, hedonic techniques allow price comparisons across countries for specifications for which regression estimates can be made of the price, even though the price for the exact specification is not observed in the country.
168. Household rents are a significant percentage of consumer expenditures in most countries, and appropriate quantity comparisons across countries, or even between regions of a country, are difficult. There are direct quantity comparisons between countries that look at persons per room or per unit area of housing. But variation in housing quality within and between countries is so large that such a comparison on the basis of housing census data would not be comparing like with like. The way the ICP has dealt with this problem is partly through hedonic regression equation methods. In the case of housing, hedonic techniques require that there be a rental survey, and typically only a portion of participating countries have such a survey. These hedonic methods based on rental surveys are outlined here and are described more fully in Kravis, Heston and Summers (1982, pp. 54-59).
169. Price-determining characteristics for rental housing include size of dwelling, availability of amenities such as water, electricity, central heating and bathroom facilities, and usually age of dwelling. The ICP has created rent cells for about 60 dwelling types that represent combinations of the above characteristics. Many participating countries have rental surveys running from hundreds to thousands of observations. The monthly rent is regressed against the above amenities, size and age of dwelling, and usually additional survey variables that may be particular to each country. When more information is available in the survey, such as location, whether the dwelling is furnished, whether it is rent-controlled or subsidized, availability of a garage, and air-conditioning, these additional variables are often included in the regression equation of an individual country in order to sharpen the estimates of the coefficients on the rent-determining variables defining the cells across countries.
170. It is quite valuable for countries to estimate hedonic regressions from rent surveys for national purposes. First, such a regression approach will almost certainly improve the constant price estimates of rental services in household consumption. Secondly, hedonic rent regressions provide a basis for estimates of the value of housing stock to be included in capital stock. Thirdly, rental surveys typically permit estimation of dwelling rents for the same specifications in different regions of a country, which may be useful in determining housing allowances and/or construction priorities.
171. For ICP purposes, the task of the country is to provide rental estimates for a variety of dwelling sizes and associated amenities that adequately represent the housing stock of the country. Countries, whether or not they use the hedonic approach, will provide estimates for a number of ICP rent cells that are important for the country. Important cells are determined by examining housing census information to ascertain which of the 60 dwelling types will be important enough, say, 3 per cent of the dwelling stock, in a particular country to warrant providing a rent estimate for that cell.
172. The actual rent estimate for a cell may be constructed separately for rental and owner-occupied dwellings or all together. If it is all dwellings in a country, then it is important to give appropriate weight to both groups of dwelling. For countries using hedonic regressions, the coefficients in the regression equation will be weighted by housing census proportions, for example, if there are rent coefficients by rural versus urban, then the weight to each coefficient will be the proportion of dwellings in rural and urban areas of both rental and owner-occupied dwellings. For those characteristics defining the cell, say, area of 30 square metres, age of building 20 years, indoor plumbing, electricity and water, these values would be plugged into the regression equation. Values for any additional variables in the regression equation would be set to match the specification so as to produce estimates of national average rents. 14/
173. Comparisons of rents, whether estimated from hedonic regressions or otherwise, must also take into account different methods of subsidizing housing space. The housing "price" may depend upon whether it is the renter or the property that is subsidized. If the renter is subsidized, then house rents paid will reflect market rents, and the national accounts should embody such rents in their estimate of house rents and the imputed rental value of owner-occupied dwellings. When dwellings are subsidized, this may show up as a government expenditure for maintenance and capital cost of the subsidized units, or, for older buildings owned by government but fully depreciated, it may not show up in government at all. In the latter case, the national accounts do not reflect the market value of these dwellings and some decision has to be made as to whether to take this into account. As discussed in chapter II, rental services consumed, not rental expenditures of households, is the relevant concept for the ICP; any expenditures of government for housing should be transferred to household consumption.
174. If there is partial rent control, then the national accounts should embody part of the rental dwellings at market prices and part at controlled prices. Further, in order to get the appropriate rent for owner-occupied dwellings, it is necessary to know what is done in the national accounts. The SNA recommendation is that market rents should be used for owner-occupied dwellings, though, in some countries, controlled rental prices have been used. In any case, the principle should be that the rent for any cell should be based on the proportion of dwellings in that cell at market and controlled rents.
175. The whole question of the treatment of imputed rents in the national accounts can be usefully discussed in regional meetings. The experience of neighbouring countries is likely to better inform national accounts statisticians in all of the countries. Such discussions should improve and make more comparable the national accounts treatment of this difficult area, quite independent of the ICP.
176. To summarize, countries provide rents of dwellings for a number of different housing types based upon rental or housing surveys that in a number of countries have been analysed using hedonic regression techniques. What is crucial is that rent estimates be provided for dwelling types that are important for the country and that the estimate of rental price reflect the same valuation used in the national accounts for both rental and owner-occupied dwellings. Finally, since quantities are known from housing censuses for most countries, these weights should also be provided. As will be discussed in chapter V, such weights are used for several other basic headings to obtain the parities at the basic heading level.
177. Hedonic price estimation has been carried out for automobiles and appliances such as refrigerators. The methods used for automobiles are very similar to those for house rents. Generally, a grid of horsepower and weight cells defines a number of automobile models, and within each cell from one to four actual models are described. Countries then provide prices for those models based on their importance, using sales in the benchmark year. The prices may be based on hedonic regression equations or on other survey data. As in the case of house rents, there are weights associated with each automobile price that are taken into account in building up the basic heading parity. Passenger automobiles are purchased as final product and as intermediate product by firms, so it is important that the auto comparison be done carefully. Because there are discounts from list price in some countries, these must be taken into account when estimating the national average price.
178. It has been found for most appliances that usually one or two factors are enough to explain price, such as total litres and freezer size in the case of refrigerators. So, in practice, the ICP has used a number of other procedures, which are termed price-slope adjustments, to modify actual prices to meet a specification.

179. The character of the remaining problems usually concern situations where it is not possible to match items exactly, but there is information available that makes it possible to relate the differences in the match to differences in price. These procedures have been grouped into two types, those arising from size of package or purchase, and those arising from the technical capacity of the item.

180. For most items, the price per unit of weight declines with the size of purchase because packaging and retailing charges do not rise proportionally with size of purchase. Basic items such as rice can often be purchased in boxes of under a kilogram, as well as in bags of over 10 kg, within the same country, and even in the same outlet. The premium in price per kg paid on small-sized purchase quantities can be considered as the price paid by the consumer for more distributive services. Most ICP specifications include weight or volume as a part of the specification, so that, in effect, a 0.5-kg box of rice is a different item from a 5-kg bag of rice.
181. Often, a specification will give a range of weight, for example, a 0.3 to 0.5-kg tin of peas. If in one country the commonly purchased tomato sauce is in a 0.33-kg tin and in another country a 0.45-kg tin, then each country would convert it to a price per kg on a proportional basis. In other words, the relationship of weight to price is assumed linear within small ranges, as in this example.
182. Often, choosing the correct unit of comparison for an item is in effect a price-slope adjustment that puts a wide variety of items into a standard frame. Household textiles and floor coverings are an example. It turns out that one can hold constant many of the price-determining characteristics by specifying price per square metre for a given material of a given weight. For example, cotton towels come in a bewildering variety of sizes and weights, but the main element in price is the amount of material. This can be standardized by suggesting an approximate range of towel sizes of a given weight per square metre and converting the collected price to a price per square metre. This applies to many types of floor coverings, curtain materials, suiting material and the like.
183. Differences in packaging for the same unit of purchase of the same commodity often do not affect relative prices across countries. Frequently, wage rates, recycling legislation or relative costs of paper will lead one country to adopt one form of packaging as less expensive, while another form of packaging may be cheaper in another country. If volume sellers use different packaging in the two countries, it is a difference that can usually be ignored.
184. However, ICP specifications commonly do distinguish between an item sold in bulk and an item that is pre-packaged and treat them as different items. This is because purchases in bulk often involve less service to the consumer and higher unit prices.

185. For many home appliances and for a variety of capital equipment items, one dimension of the product may be price determining. For example, an electric motor specification may be for a three-quarter horsepower unit but in a particular country the unit available is one horsepower. So long as there is information on the relationship between price and horsepower in a range bounding these motor sizes, it is reasonable to adjust the one-horsepower price to the three-quarter price. Use of such price-slope adjustment factors is frequent because it permits the use of already collected consumer, wholesale and import prices, with the consequent savings in resources needed for additional collection.
186. In consumer goods, these types of adjustment are frequently used for refrigerators, air-conditioners, heating appliances, fans and the like. Where countries can easily find the same brand and model widely sold, simple identity of item is to be preferred. But where regions are heterogeneous, or import restrictions limit brand representation on local markets, pricing identical items may not be possible. To illustrate, most information about room air-conditioners is contained in the number of BTUs of capacity. This is a case where it may not be necessary to compare exactly the same size of air-conditioner across countries, because the relationship of price to BTUs provides a gradient that allows for the adjustment of actual prices to a normalized price for a slightly different size. Refrigerators present a slightly more complicated case because both litre of freezer space and total litres affect the price. However, where exact matches of freezer and total size are not possible, it is usually straightforward to convert a price for a slightly different size to that of the specification. A summary of the experience in pricing capital equipment in Africa treats a number of these questions. 15/
187. Price-slope adjustments are frequently necessary and are fairly easy to apply in the case of machinery and equipment. Representative profiles of a wide range of equipment goods have been drawn up (and are continually being revised to take account of changing and improving technologies) for ICP machinery pricing. Identification and matching of these items depends on a precise and well-defined description of the technical characteristics of each product, for which there is typically a specification sheet, with technical descriptions and usually illustrations. The technical characteristics listed for each item include all features relating to size, power, performance (output, speed, capacity etc.) and weight, and these are usually arranged in some order of relative importance. The ICP list provides these details and also contains an illustrative guide, together with some suggested models and brand names in certain cases, to aid the process of product identification and pricing. One advantage for most developing countries is that a single price quotation from the main franchise or importing agency is usually all that is necessary. The problem for many countries is that they will not be able to supply prices for the exact specification.
188. One reason for this is that when Governments directly encourage and protect their country's capital goods industry, then it is difficult to find exactly the same item in every country. Matching also proves to be especially difficult where, for reasons of statistical confidentiality, it is impossible to identify any specific models and brand names. The process of comparing equivalent equipment goods in these situations is carried out on the basis of matching, as closely as possible, pre-defined physical specifications and output characteristics. When this method of matching is followed, it is important that any comparison is made primarily on the basis of the most essential characteristics. Price-slope adjustments will often be needed because of small differences in capacity of, say, a pump, in different countries. These adjustments are typically done by the regional organization with consultation with the country when necessary. The main responsibility of the country is to provide along with the price as much technical information as is available about the particular capital good.
189. Some of the important factors that affect actual purchase prices of machinery are installation costs, delivery charges, after-sales services, contractual maintenance, accessories, discounts for quantities purchased, financing arrangements, tax charges and other terms of sale. For consistent pricing in all countries, only normal sales conditions should be taken into account and specially negotiated terms of payment and variable quantity purchase should be excluded.
190. The problem of identifying and matching investment purchases of machinery and equipment in many developing countries is, in fact, often less complex than in advanced industrialized countries. For a significant number of items, the goods will have been originally produced in Japan, Europe or the United States of America and will have been imported into the country. They are likely to be items already in the ICP specifications. There are likely to be only a few distributors, usually in the major cities, so outlets are easy to identify. Further, since many of the items are imported, price information may already (or should) be collected as part of import price index construction. While experience of statistical offices in pricing machinery and equipment is often modest, this is an ICP pricing area that in practice is fairly straightforward for countries to meet the needs of the ICP and also expand their own pricing base. Most of the problems of comparability can be handled at the regional level by careful matching, use of some price-slope adjustments, and consultations with countries.
191. In general, the price of a piece of capital equipment should include all taxes but not optional service charges. Manufacturers often provide some type of guarantee on part or all of their equipment for a limited period of time, with the cost included in the selling price. Dealers also offer service contracts along with the equipment. Whenever there is an additional charge for the service contract, that charge should not be included in the price. Any finance charges associated with the purchase of the equipment should also not be included in the price.

192. In the ICP, direct price or cost information are usually not used for comparisons for construction items. There are four main reasons for this. First, many kinds of construction projects are not carried out every year in all countries, so actual prices or construction costs of equivalent building and construction projects are simply not available. Secondly, actual construction projects are unique, often differing in shape, size, technical characteristics or materials used, and so they are hard to match across countries.
193. Thirdly, construction costs for the same type of project in the same country with the same input prices tend to have a high variance because of random factors, including weather. Therefore, the actual cost of a particular construction project may be substantially above or below what would be typical for that type of construction in the country. Fourthly, and operationally most important, is that there is an alternative approach that appears to produce very reasonable results.
194. This approach employs a series of standard building models, which are costed in all countries using a common bills-of-quantities approach. These model construction products, which may never actually be built in a country, can be broken down into a variety of common components and activities. These components include initial ground clearance and excavation, foundations, walls and roofing, which in turn can be disaggregated into uniform bills-of-quantities, such as a running metre of masonry of given height and width or a cubic metre of excavation. These well specified bills-of-quantities permit matching of items based on a range of identical materials and quantities used as basic inputs that can be priced in many countries.
195. In this approach, a given construction project is modelled as some weighted sum of the bills-of-quantities involved. These models have been developed by architects or quantity surveyors in a number of countries and represent a variety of building types. The bills-of-quantities approach was pioneered by EUROSTAT and was developed for use in many Latin American comparisons. However, there is much geographical diversity in construction practice, so it has been important to develop and adapt models that are particular to regions, such as has been done in Africa and Asia.
196. While the detailed specifications closely resemble actual buildings or construction activities in terms of their shape, dimensions, methods of construction etc. they are, in fact, standard models or prototypes. Once the details have been fully specified, the process of costing can be undertaken by both government and professional quantity surveyors, architects or engineers. Often, public works departments and other government agencies will regularly maintain files with costs of bills-of-quantities that are the building blocks in the many infrastructural development activities and contractual building projects in which Governments are involved. In the core commodity list, a number of the basic building models have been included for which countries would build up cost estimates from the detailed bills-of-quantities contained in the construction specification manual.
197. As construction methods respond to changes in types and relative costs of materials available, to real wage changes, and to new techniques and improvements in construction machinery, the prototype construction projects used by the ICP are also updated. These changes may show up in the relative weights or the number and technical descriptions of the bills-of-quantities.
198. The advantage to participating countries of the bills-of-quantities approach is that it lends itself to methods of national deflation of construction costs over time that take account of productivity changes. Most national construction indexes are based on movements in the prices of inputs, with or without some fairly arbitrary adjustment for productivity improvements. A time-to-time price index based on the bills-of-quantities approach by its nature takes into account productivity changes.

199. There are many problems that come up in price comparisons for particular basic headings, such as fish, which present problems because of the use of the same names for different fish, and different names for the same fish, not just between countries, but often in different parts of the same country. In this particular case, the ICP has used several common names in addition to Latin names, which any marine expert in a country can identify. Most of these pricing issues become clear in the specifications or in consultations and will not be treated here. However, two general questions are taken up in this section, equivalence in use and identity of products.

200. Different countries consume cereal products in different proportions, some consuming mainly rice and others mainly corn. For this reason, there are those that have advocated converting a kg of each type of cereal into, say, wheat equivalents, and comparing that price across countries. In general, the ICP has not adopted equivalence in use as a basis for comparison, so that comparison of like with like has been the guiding rule.
201. Some examples may illustrate instances where this rule has sometimes been violated. Consider the case of wheat flour, where both whole wheat and white flour are the specifications. Usually, there is a small premium for whole wheat flour over white flour in countries like the United States, because of the much larger volume of white flour that is consumed. But, in some countries, whole wheat flour is more common, especially where grinding is done locally and many breads are prepared at home, and it sells at a lower price than enriched white flour. Where countries have both items, both can be priced. However, this is a case where it is useful if countries can give an indication of relative importance of the items, as in the asterisked (*) item method, as discussed in the following chapter. This is also a case that could be treated as equivalence in use and where simply the average price of white and whole wheat flour were compared across countries, as was done in phase I of the ICP in developing binary comparisons.
202. Another example occurs for countries where no distinction is made in the market between goat and lamb. In fact, the ICP has specifications for both, but in practice, if a country does not distinguish between the two, one price may be all that can be supplied. What should be done? To some extent, this problem solves itself because usually countries that would distinguish the two also have systems of meat grading that include a general carcass weight specification. Usually, carcass weight comparisons produce matching between countries and little is lost by equating goat with lamb for countries that make no distinction.
203. There are also several other areas where equivalence in use seems more sensible than pursuing exact like with like. For example, most countries have several forms of squash or pumpkin that may look very different but for all practical purposes serve a common function and sell for the same unit price. The same is true for leafy greens. The ICP specification in such cases is generic and countries must judge whether their local item is a variety of a common item. A similar case is ketchup, which in some countries has a tomato base, in some countries bananas, and in still others, pumpkin, or some mixture. In some regions, these are treated as identical.
204. Another case that is fairly important is cooking oil. In addition to comparisons for specific oils, such as sunflower or sesame oil, there is also an ICP specification for a general purpose cooking oil. Typically, the common vegetable oil is a blend of available oils, the blend varying both within and between countries, depending on the relative price of the usual oils used in the blend, for example, corn, groundnut and the like. In this case, all blends are considered equivalent in use, even if not identical.
205. A final illustration is provided by clothing. Usually the ICP offers a number of specifications of clothing materials, with technical specifications as to whether it is wool, a blend, dacron or other synthetic, or a cotton, the thread count where relevant, its weight per square metre, if price determining, type of weave, such as corduroy, and finish. For many countries, purchase of materials is a first step in clothing purchase, the second step being tailoring, which may be in the home or by a tailor. As part of the price collection in these categories, tailoring charges for a variety of tasks are specified. For some clothing items, such as a dress or a pair of pants, it may be less expensive for a consumer to buy material and pay tailoring charges than to buy them ready-made in a store. In such cases, it is the final consumption of the dress by the consumer, not the manner of its production, that is being treated as equivalent.

206. In this section, we return to an issue discussed in chapter III, namely, how close should countries try to stick to identity of brand in pricing items. One general procedure used in the OECD countries is to try to provide prices for identical brands but to mark with an asterisk (*) items that are important. As will be discussed in chapter V, this means that if a particular brand is priced in a country that is not important in consumption and is not given an asterisk (*), it will get less weight in determining the parity for the basic heading. The system of using asterisks for certain items appears to have promise, but like all comparisons, it is much easier where countries are homogeneous.
207. Consider beverages and tobacco comparisons across all of Asia. The list of items will include a number of international brands, often selling at premium prices because of tariffs or bans on imports, plus a large number of local brands. Usually, the local brands are the volume sellers, so that there will be a specification for a bottle or can of beer that may say "local volume selling beer", where identity of brand is not required. Countries may also provide prices for internationally known brands if they are commonly consumed, which is likely to be the case with cigarettes, where both national and international brands share the market. Whenever countries feel they can match a brand but do not regard the item as characteristic of their consumption, they should indicate this when supplying the price, or choose to simply not price that brand.
208. Sometimes, countries use identical names for goods and services that are not identical. In purchased transport, for example, there are many problems of international comparability, mostly associated with holding quality constant. It is easy enough to stratify by types of urban transport, cycle rickshaws, three- and four-wheel taxis, jitney buses, regular buses, deluxe buses, trolleys and subways. Suppose that in one country, on bus trips a person can expect only standing room, a wait for the bus of 30 minutes or more, and that the bus itself will have few amenities, while the opposite is true in another country; should the bus trips in the two countries be treated as like to like? Is the comparison only for purchase of transport from A to B by road, or do we want to hold constant the quality of the ride and the time input of consumers? The answer is that some compromise must be reached because we enter a very uncertain world if we try to take account of time in queues. Probably, the best that can be done is to not simply equate first class train in one country to first class train in another, but to inquire exactly what amenities the consumer is buying. In fact, second class in Europe has been equated with first class in some South Asian countries. However, for many urban commuting trips, crowding and delays are common in rich and poor countries, and it is not easy to stratify adequately to take account of many quality differences.
209. The first choice for countries is to price identical goods and services. However, countries must remain prepared to question the obvious identity of brand with brand or a first class ticket in one country with a first class ticket in another. These questions should be raised with regional coordinators or in meetings with pricing counterparts in neighbouring countries. It is important to know exactly what service is being provided and just how important particular brands are in local markets, and here the knowledge of the staff of national statistical offices combined with the experience of ICP coordinators can often come up with an appropriate solution.