CONVERGENCE OF
INDUSTRIAL CLASSIFICATIONS: NAICS-NACE
The agreement signed by the heads of the statistical agencies of Canada, the European Union and the United States concerning a project on the convergence of NAICS and NACE outlines a series of phases towards the achievement of the objectives of the project. The first phase calls for the preparation of a report on the technical and cost implications of convergence, to be completed by October 2000. It is to undertake this analysis that a meeting of representatives from the three statistical agencies met on August 30 to September 1, 2000.
This report summarizes the discussions of the group and represents the first deliverable outlined in the project agreement. It will not, however, provide a full accounting of the technical and cost implications of convergence. It was realized at this first meeting that to do so would require an in-depth review of all areas of divergence between NAICS and NACE, and a negotiation over which of the two, or both, would require change. Such an exercise could only be carried out over an extended period of time and require consultation with the areas concerned.
To launch such a process, however, certain basic agreements must be reached. First, there should be some assurance that there are no ineluctable conceptual differences standing in the way of convergence. If agreement cannot be reached on these basic principles, convergence could not be achieved and there would be little point in pursuing more detailed examination of differences. Secondly, assuming that there are no insurmountable conceptual differences, agreement must be reached on what is meant by convergence. While all can agree that convergence means making both classifications more similar to each other than they currently are, there can be differences in degree in the expectations of what this particular convergence project can be expected to achieve. Finally, assuming that agreement can be reached on the meaning of convergence, there remains an issue as to what level of the two classifications convergence should be sought. Only then can the full technical and cost implications be fully ascertained.
This report presents the working group’s consensus on each of these three fundamental questions. It will describe the conceptual differences between NACE and NAICS and justify the conclusion that these are not insurmountable. It will provide a working definition of convergence for the purposes of this project and it will propose the level of detail at which this convergence should be sought. Based on these conclusions, it will then propose a program of work to be carried out over the next year to identify the areas of change that would be required in each classification, with associated technical and cost implications.
Conceptual differences:
A discussion paper prepared by the U.S. representatives outlined a number of conceptual differences between NACE and NAICS and these formed the basis of the discussions of the working group. Through discussion, a few more were identified.
The first of these is the supply-side conceptual basis of NAICS, which establishes the production process as the main determinant for grouping economic activities. This principle is one of the hallmarks of NAICS and there would be no room for compromise on its application in the convergence project. The Eurostat representative recognized that there was no single principle underlying NACE and that this was a difference with NAICS. He pointed out, however, that in practical terms, the classification structures of NAICS and NACE are quite similar and that where they are different, a pragmatic approach could be adopted to make them converge. He also agreed that the adoption of the production process principle provides a firm basis for making classification rulings and more rigorous guidance for the ongoing development and maintenance of the classification. As such, having such a principle is a desirable feature of a classification and he signified Eurostat’s willingness to consider its application in the convergence project. In any event, all agreed that pragmatic solutions to deal with this issue were at hand.
A second difference between NAICS and NACE is the frequency of revisions. The NAICS partners are committed to a 5-year revision cycle, in recognition of the rapid evolution of economic activities in the North American economies. Tying the two classifications together would imply harmonizing the periodicity of their revisions. The Eurostat representative recognized that more frequent revisions would be desirable for NACE as well, but that these revisions must be of an evolutionary rather than a revolutionary nature. The NAICS partners indicated that they foresaw little change in the traditional primary and manufacturing sectors and that revisions were likely to concern emerging industries in the information and services sectors. They also foresaw changes in the distribution sectors (wholesale, retail and transportation and storage) for 2007. A consensus was reached on the requirement for more frequent evolutionary revisions for both classifications. Particular emphasis was placed on the need for more regular and detailed communication on the interpretations of the classifications over time, another implication of closer alignment of the classifications.
A general rule established during the development of NAICS was the treatment of activities that did not exist in all partner countries. In such cases, classes were defined and carried even in the country where the activity did not exist, so long as it existed in more than one of the partner countries. For example, there is a class for orange groves in NAICS Canada, even though there is no such activity in Canada, as it exists in both the United States and Canada. It was agreed that a similar rule would be adopted for the convergence project.
In addition to these general differences, there are more specific differences regarding the treatment of various economic activities. A major difference is the delineation of the Information sector in NAICS. This is a major difference but Eurostat is already committed to introducing a similar concept in NACE, as an alternate aggregation structure in the short term, but as a distinct class in the future.
There are also differences in the treatment of repair and maintenance, which has been generally regrouped under a single sub-sector in NAICS. In discussion, it was ascertained that there is a similar treatment in NACE and NAICS of rebuilding, considered as manufacturing, and repair. The repair activity in NACE, however, can be found throughout the classification, as it was in the previous North American classifications. While Eurostat is willing to consider the regrouping of all repair and maintenance, it would be informative to learn more about the relative size and structure of the sub-sector based on data collected under NAICS. The U.S. has collected data on this sector in the 1997 Economic Census and will make it available to the working group.
Another area of difference is the treatment of wholesale and retail in NAICS, which is based in the first instance on the production process, rather than class of customer. As this area is slated for review for NAICS 2007, there is an opportunity to bring about greater convergence through a pragmatic approach if agreement cannot be reached on the production process principle. The fact that retail sales is such an important economic indicator for all the partners in convergence provides an additional constraint on what can be accomplished in this area.
The next item in the U.S. discussion paper was the treatment of ancillary or auxiliary units. There was agreement on the need to record both the activity of these units as well as the activity of the units they served. The U.S. indicated that they were refining their treatment of these units for the 2002 Census, and were preparing a paper on the subject, which will be made available to the working group.
The treatment of units engaged in more than one economic activity was identified as another area of potential difference between NACE and NAICS. These include combined activities, vertically integrated activities and horizontal integration. While a number of specific cases were brought up, there appears to be no deep ideological differences between the two classifications in the treatment of these activities. There does seem to be a lack of consistency in NAICS as regards some combined activities. In a number of cases, there are exceptions to the principal activity rule (for example the 50% rule in the agriculture sector) that are not documented or justified. These should be inventoried and considered case by case. There are also inconsistencies in the treatment of support activities to specific sub-sectors and industry groups. In certain cases, these are placed in specific industries of their own, in others they are include in the industry supported. There are no clear criteria or documentation explaining this different treatment and these should also be inventoried and considered on a case by case basis.
Overall, these issues do not present insurmountable obstacles to convergence. There would be value therefore in defining more precisely what convergence means for the purpose of this project and at what level of detail it would be sought.
Convergence options
The working group developed a number of possible scenarios to bring about convergence between NACE and NAICS. These are summarized below and recommendations are made regarding which of these should be subjected to more in-depth analysis of costs and implications.
The Group identified two categories of needs that would be addressed by convergence: those of analysts interested in transnational studies and comparisons, and those of organizations, enterprises or institutions, which code their constituent units, clients and/or suppliers according to industry. While the former could be satisfied by better concordances between the two classifications, the latter would be better served by the adoption of a common structure and coding for both classifications.
The group agreed to examine these two fundamental forms of convergence: the adoption of the same structure or an improvement over existing concordances between NACE and NAICS. The objective of the better concordance form of convergence would be the establishment of 1 to 1 links between NAICS and NACE, with minimal “noise”, at an agreed target level of the classifications. Each of these approaches could be applied at various levels of detail of the classifications: at the top level only; at the most detailed level (defined for this purpose as the level at which the 3-country NAICS is harmonized, representing 478 classes) or at some middle level. Together, these dimensions form a 2x3 matrix, yielding 6 possible options.
Same structure |
Better concordance |
|
Top level |
Top level |
|
Mid-level |
Mid-level |
|
Detailed level |
Detailed level |
The group agreed to define two variants of the middle level. Both would be constrained at the bottom by the 3-country NAICS level, as foreseen for 2007. For example, if it were foreseen that the lowest level of NAICS harmonization in Public Administration will be at the 2-digit level, then NACE-NAICS convergence would also go to this level only. The first middle level variant would be at the NAICS sub-sector – NACE division level, representing approximately 100 classes. NACE rev.1 only has 72 divisions currently but Eurostat has expressed a desire to increase this number. The other mid-level would be at the NAICS industry group – NACE group level, for approximately 300 classes. This also represents the level at which a relationship would be established with ISIC rev. 4 and proposals will be made to the United Nations in this respect. Finally, a hybrid level combining the mid-level and the detailed level for, at minimum, the Information sector, and other sectors as identified, was defined. Each middle level option includes the higher levels of the subject classifications. For example, the same structure-middle level 100 class option implies that the top level is harmonized.
Each of these options was discussed in terms of their costs and benefits. Development (research, consultation, documentation) and implementation (conversion, data collection and recoding) costs were considered in a general way based on an appreciation of the extent of change implied by each option. Relevant observations are shown in the following table.
Same structure
|
Better concordance
|
|
Top level Advantages: Comparability at
top Flexibility below Future
harmonization possible Likely to remain
constant over time Visible result Common coding Disadvantages: Only 17 to 25
classes in common Requires a lot of
change for little payoff Can only be seen as
a start, not enough on its own to warrant cost |
Top level Advantages: Convertability at
top Full flexibility Likely to remain
constant over time Less costly than
adopting same structure as costs associated with renumbering are avoided. Disadvantages: No classes in
common Result not visible No common coding Requires a lot of
change for little payoff |
|
Mid-level – 100
class Advantages: Comparability
pushed down a level Flexibility below Future
harmonization possible Visible result Common coding for
approximately 100 high level classes Disadvantages: Significant change
required Higher development
and implementation costs that top level options or concordance at same level |
Mid-level – 100
class Advantages Convertability
pushed down a level Full flexibility Less costly than
adopting same structure at same or higher level as costs associated with
renumbering are avoided. Disadvantages: No classes in
common Result not visible No common coding More difficult to
maintain than high level concordance Requires a lot of
change for little payoff |
|
Mid-level – 300
class Advantages: Comparability
pushed down a level Greater
harmonization achieved Very visible result Common coding for
approximately 300 classes Disadvantages: Less flexibility Significant change
required Higher development
and implementation costs than higher level options or concordance at any
level |
Mid-level – 300
class Advantages Convertability
pushed down a level Full flexibility Less costly than
adopting same structure at same or higher level as costs associated with
renumbering are avoided. Disadvantages: No classes in
common Result not visible No common coding More difficult to
maintain than high level concordance Significant change
required |
|
Hybrid level Advantages: Comparability
pushed down to the detailed level for Information sector Greater
harmonization achieved Very visible result Common coding for
approximately 400 classes Eurostat already
committed to detailed Information sector in NACE Disadvantages: Less flexibility Significant change
required Higher development
and implementation costs |
Hybrid level Advantages Convertability
pushed down a level Full flexibility Less costly than
adopting same structure at same or higher level as costs associated with
renumbering are avoided. Disadvantages: No classes in
common Result not visible No common coding More difficult to
maintain than high level concordance Significant change
required so that marginal cost of going to full detailed level is justified |
|
Detailed level Advantages: Full comparability Harmonization
achieved Very visible result Common coding for
approximately 500 classes Disadvantages: Less flexibility Significant change
required Highest development
and implementation costs of all options |
Detailed level Full convertability
(subject to 3-country NAICS constraint) Full flexibility Less costly than
adopting same structure at same or higher level as costs associated with
renumbering are avoided. Disadvantages: No classes in
common Result not visible No common coding More difficult to
maintain than high level concordance Significant change
required |
The working group retained neither option that stopped at convergence at the top level as it considered that costs significantly outweighed benefits. It also considered that the same structure at the detailed level option was too ambitious an undertaking, involving very high costs. The group also believes that the mid-level and full detail concordance options are roughly equivalent in costs and that therefore, only the better concordance at the detailed level would be retained as a viable option.
This leaves four viable options, in increasing order of costs and benefits:
a. Better concordance at the detailed level
b. Same structure at the 100 mid-level
c. Same structure at the 300 mid-level
d. Same structure at the hybrid level
To fully assess the costs and implications of any of these options requires an in-depth analysis of differences and a series of negotiations to ascertain which of either party or both would agree to change to achieve the desired outcome. This is a very time-consuming and labour intensive activity, which the Working Group believes should be undertaken for, at most, two of these options, over the next year. The recommended options for this purpose are a) and d). It is recognized that it is necessary to achieve option a) in order to implement option d).
Workplan
The group discussed a workplan to reach agreement on the areas and extent of changes required to achieve options a) and d) and to develop detailed estimates of the remaining development and of the implementation costs for these options.
To assist in this process, the group reviewed a preliminary list of differences between NAICS and NACE prepared by the American representatives. This list is not necessarily exhaustive but it does highlight many, if not most, of the main differences between the two classifications. The results of this review are attached as Appendix 1.
The group considered, by NAICS sector, each difference and classified each according to the degree of difficulty expected to achieve agreement. The 76 inventoried differences broke down as follows:
1. easily resolved or tolerated (acceptable “noise”) 28
2. true
problems 15
3. require
discussion 14
4. require
discussion, NACE likely to change 10
5. required
discussion, NAICS likely to change 4
6. not
clear, requires more research 5
The group was encouraged by the outcome of this review, which established that the majority of the cases could readily be resolved and that serious problems were few. Based on this analysis, it was determined that a program of work covering a 1-year period would be required to fully identify and negotiate agreements on these differences.
The budget required for this activity over the next year is salary to support 1 full-time equivalent for each agency and travel money to cover 4 one-week meetings to be held in November in Washington, in February in Mexico, in May in Luxembourg and in August in Ottawa. Funding to cover the participation of the UN at these meetings must also be secured.
The first meeting according to this plan will be held in Washington on November 6-8, 2000. At this time the group will determine the process by which all differences will be identified, establish criteria for evaluating the significance of each difference and agree on the method and timing of negotiations on required changes. Agreement will also be reached on the specific sectors to be reviewed in the following quarter and reported upon in Mexico in February. The results of these discussions will be presented in the working group’s second report, by the end of November 2000.