Introduction 19 stituent establishments comprise the statistical units of the statistical structure. In delineating the statistical structure, functional or other groups in the organizational structure may be ignored and the constituent units regrouped to form the units of the statistical structure. For multi-establishment enterprises, the statistical structure may not coincide with the legal structure in which ownership of assets is registered. 97. The source of information for production statistics and income statistics is often the management and cost accounts of businesses. These accounts record operat- ing revenues earned from the sale of goods and services produced and the associated costs, wages and salaries, depreciation and operating profits. Countries that are more interested in the higher level of autonomy in the structure of enterprises than in the geographical location of the activity may prefer to delineate and use the kind-of-activ- ity unit. On the other hand, if they are more interested in compiling comprehensive production statistics at a subnational level of geographic detail, then it will be neces- sary to delineate the smallest unit (a) that is as homogeneous as possible in terms of activity and geography, and (b) for which statistics on revenues from the sale of goods and services, associated costs, value added and gross fixed capital formation under- taken can be compiled or estimated. This is the level at which the establishment (local kind-of-activity unit) is delineated as the statistical unit. 98. Whenever the legal structure and the statistical construct based on pro- duction units do not coincide, statistical agencies will need to articulate the statistical structure and compile data with the help of surveys. The legal structure may consist of units created purely for tax purposes that are completely irrelevant to the produc- ing units of the enterprise. However, if it is necessary to draw on tax records for the required data or if survey data need to be supplemented with tax data, statistical agen- cies will have to decide whether (a) they can find a way to map the legal and statistical structure of the enterprise, or (b) they prefer to use the legal structure selectively as a proxy for the statistical structure. 99. The statistical structure delineates and identifies the units about which data are to be compiled. However, the data may have to be collected from higher- or lower- level units, which are then described as collection entities. As a result of globalization, some multinational global enterprises are keeping integrated accounting records at the global or the regional levels only; it may thus be increasingly difficult to separate and extract complete accounts for all the activities taking place within each domestic economy if such data are not obtained from the main or regional head office of the global enterprise. 5. National differences in selecting statistical units 100. An establishment undertakes an activity at or from a particular location. Thus, the concept of the establishment combines two dimensions—an activity dimen- sion and a locality dimension. The concept is based on the assumption that the aim of the statistical programme is to compile data that is classified both by activity and by location, e.g., by geographical region. However, ISIC can be used to classify many other variables needed to analyse production and industrial performance. In examin- ing the accounting and operating structures of enterprises, it is possible that producing units with differing levels of homogeneity with respect to activities and geographical precision will be found and that they may be suitable for the compilation of data on selected variables, such as number of persons employed; they may even be suitable for the compilation of production statistics if all the information needed with respect to meaningful operating profit is available.
International Standard Industrial Classification of All Economic Activities (ISIC), Revision 4 20 101. Many different factors play a role in defining the best statistical unit for a given form of data collection, such as the structure of the legal system in a coun- try, including regulations for the organization of businesses; the particular structure of the industries involved; the type of data collection involved; and the purpose and targeted level of data collection. The statistical units discussed in this section can be taken to serve as models, but the actual choice of the type of statistical unit to be used by countries would depend upon national specifics and the type and the purpose of data collection involved. 102. Finally, it should be noted that in order to ensure reliability and compara- bility of data in delineating statistical units, such units should be defined in a consist- ent manner. The consistent definition of these units is equally as important as their accurate classification. C. Classification of statistical units 1. General guidelines 103. In the following paragraphs, a number of general rules of interpretation are given that should be followed when classifying more complex statistical units. It should be noted that the explanatory notes to some sections and divisions of ISIC (see part three below) also indicate how to treat such cases. 104. A unit may perform one or more economic activities falling into one or more categories of ISIC, Rev.4. Units are classified according to their principal activity. In practice, the majority of production units perform activities of a mixed character. The identification of a principal activity is necessary to allocate a unit to a particular ISIC, Rev.4 category. 105. The activity classification of each unit is determined by the ISIC class in which the principal activity—or range of activities—of the unit is included. All activi- ties are considered when determining the principal activity, but only the principal activity is used to classify a unit. The principal activity of the unit in general can usu- ally be determined from the goods that it sells or ships or the services that it renders to other units or consumers. However, the descriptions and explanatory notes of the individual classes in ISIC (see part three below) should be used to determine the activ- ities carried out in terms of ISIC categories, using not only the output structure but also the input structure and most importantly the production process. 106. Ideally, the principal activity of the unit should be determined with refer- ence to the value added to the goods and services produced. The relevant valuation concept is gross value added at basic prices, defined as the difference between output at basic prices and intermediate consumption at purchaser’s prices. (a) Value-added substitutes 107. In order to determine the principal activity of a unit, the shares of value added by that unit to the activities falling into different categories of ISIC, Rev.4 have to be known. In practice, however, it is often impossible to obtain the information on value added of the different activities performed and the activity classification has to be determined by using substitute criteria, such as: (a) Substitutes based on output, such as:
(i) The gross output of the unit that is attributable to the goods or serv- ices associated with each activity;
Introduction 21
(ii) The value of sales or shipments of those groups of products falling into each category of activity; (b) Substitutes based on input, such as:
(i) Wages and salaries attributable to the different activities;
(ii) Hours worked attributable to the different activities;
(iii) Employment according to the proportion of people engaged in the different activities of the unit. 108. Such substitute criteria can be used as proxies for the unknown value- added data to obtain the best approximation possible compared to the result which would have been obtained on the basis of the value-added data. The use of substitute criteria does not change either the methods used to determine the principal activity or the rules of the top-down method. Substitutes are only operational approximations of value added. 109. The simple use of the above listed substitute criteria may sometimes be misleading, however; this will always be the case when the structure of the substitute criteria is not directly proportional to the (unknown) value added. (b) Problems with output substitutes 110. When using sales (turnover) or shipments as a proxy for value added, it immediately becomes evident that in certain cases turnover and value added are not proportional. For example, trade turnover usually has a much lower share of value added than a manufacturing activity. Other examples are turnover of for- warding agents or of general contractors. Even within manufacturing, the relation between sales and the resulting value added may vary between and within activi- ties. If a significant portion of production goes into inventory and is therefore not sold within the same reporting period, the value of sales may grossly underestimate the value added. In addition, in some cases a turnover figure makes no sense or does not exist, e.g., for financial intermediation activities or insurance activities. Similar considerations should be borne in mind when using gross output data as substitute criteria. 111. Many units perform trade and other activities. In such cases, trade turno- ver figures are the most unsuitable indicators for the unknown value-added share of the trade activity. A much better indicator is the gross margin (difference between the trade turnover and purchases of goods for resale adjusted by changes in stocks). The trade margins may vary within a single wholesale or retail trade activity, however, and may also vary between trade activities. In addition, the specific classification rules for retail trade should be considered as set out below. (c) Problems with input substitutes 112. Similar precautions have to be considered when input-based substitute criteria are applied. The proportionality between wages and salaries or employment and value added is not reliable if the capital intensity of the various activities is differ- ent. Higher capital intensity normally implies higher depreciation and a lower share of wages and salaries in the gross value added. Capital intensity varies substantially between different economic activities and also between activities of the same ISIC, Rev.4 class. For example, the activity of producing items by hand will have a lower capital intensity than the mass production of the same item in an industrial plant; both activities, however, fall in the same ISIC, Rev.4 class.
International Standard Industrial Classification of All Economic Activities (ISIC), Revision 4 22 2. Treatment of mixed activities 113. Instances may arise in which considerable proportions of the activities of a unit are included in more than one class of ISIC. These cases may result from the vertical integration of activities (for example, tree felling combined with sawmilling or the manufacture of textiles with subsequent production of wearing apparel); the horizontal integration of activities (for example, the manufacture of hides and skins in slaughterhouses); or any combination of activities that cannot be separated at the level of the statistical unit. In such situations, the unit should be classified according to the rules set out below. 114. Although typically applied at the class level, the rules set out below are valid for applying the classification at any level of the coding structure. (a) Treatment of independent multiple activities 115. If a unit is engaged in several types of independent activities but cannot itself be segregated into separate statistical units (when, for example, manufacture of bakery products is combined with manufacture of chocolate confectionery), the unit should be classified according to the activity that contributes most to the value added of the unit (i.e., the principal activity), as determined by the “top-down” method described in paras. 123–131 below. 116. The “top-down” method also applies wherever a unit performs only one activity or one activity accounts for more than 50 per cent of the value added. How- ever, in such cases the classification of the unit is rather straightforward and no step- by-step application of the “top-down” method is necessary. In such cases, the unit will be classified to the single ISIC category representing the activity carried out, or in the latter case, to the ISIC category that represents the activity that accounts for more than 50 per cent of the value added. (b) Treatment of vertical integration 117. The vertical integration of activities occurs wherever the different stages of production are carried out in succession by the same unit and the output of one process serves as input to the next. Examples of common vertical integration include tree felling and subsequent on-site sawmilling, a clay pit combined with a brickworks or production of synthetic fibres in a textile mill. 118. In ISIC, Rev.4, vertical integration should be treated like any other form of multiple activities, so that a unit with a vertically integrated chain of activities should be classified to the class corresponding to the principal activity within this chain, i.e., the activity accounting for the largest share of value added, as determined by the top- down method. This treatment has changed from previous versions of ISIC. It should be noted that the term “activity” in this context is used for each step in the production process that is defined in a separate ISIC class, even though the output of each step may not be intended for sale. 119. If value added or substitutes for the individual steps in a vertically inte- grated process cannot be determined directly from accounts maintained by the unit itself, comparisons with other units (e.g., based on market prices for intermediate and final products) could be used. The same precautions for using substitutes as listed above apply here. If it is still impossible to determine the share of value added (or its substitutes) for the different stages in the chain of production activities, default assign- ments for typical forms of vertical integration can be applied. The Companion Guide to ISIC and CPC provides a set of examples for such cases.15 15 Companion Guide to ISIC and CPC, Statistical Papers, Series F, No. 101 (United Nations publication, forthcoming).