Product Classification System

Discussion Paper on Developing a Classification System

for Products Produced By Service Industries:   Issues and Insights

by Michael F. Mohr [December, 1998]

Michael Mohr is a senior economist at the Bureau of the Census in the U.S. Department of Commerce. The opinions expressed in this paper are the author's and do not necessarily represent those of the Census Bureau or the Department of Commerce.

I. Introduction

In a forthcoming Federal Register notice, the Economic Classification and Policy Committee (ECPC) of the Office of Management and the Budget will announce a new initiative (hereafter, the Initiative) to develop a comprehensive classification system for the products produced by the spectrum of U.S. industries that have recently been defined and classified under the new North American Industry Classification System (NAICS).(1) The Initiative will be implemented in two phases. First, an interim, or exploratory, phase will be launched in early 1999 and it is expected to produce preliminary results for a subset of NAICS service industries; these results will be incorporated into the questionnaires for the 2002 Economic Census.(2) Second, exploiting the lessons and insights gained from the deliberations and the data collection activities of Phase 1, a second, or final, phase of the Initiative will be launched after the 2002 Economic Census to develop a complete and fully integrated product classification system that extends to all NAICS industries. The results of Phase 2 are to be incorporated into the questionnaires for the 2007 Economic Census.

The ultimate objective of the Initiative is to develop a market-oriented/demand-based, classification system for products that (a) can be linked to the NAICS industry classification system but is not industry-of-origin based, (b) is consistent across the three NAICS countries, and (c) promotes improvements in the identification and classification of service products across international classification systems, such as the Central Product Classification System of the United Nations. Towards this objective, the goal of Phase 2 of the Initiative will be an agreed-upon, integrated, and comprehensive list of products, product definitions, and product codes that (1) encompasses both services and goods alike and (2) accommodates a demand-side/market-oriented classification framework for grouping and aggregating these products. In preparation for Phase 2, the goal of Phase 1 is to systematically explore the seminal development of a formal classification system for service products that can be used throughout the business and economics communities of users to coordinate the collection, tabulation, and analysis of data on the value of the detailed products produced by service industries and on the prices charged for those products.

The purpose of this paper is to provide the universe of interested parties with a fuller understanding of and appreciation for the complexity involved in the effort to classify service products. To that end, this paper seeks to highlight and substantively discuss the important issues and problems that will confront and have to be addressed by the Federal agency classifiers in order to develop a seminal and operational classification system for the products of service industries. In particular, the paper argues that the process of classifying service products must be a comprehensive effort that addresses identification, measurement, data collection, and classification; and it presents and explains a detailed methodology for implementing such an effort. Perhaps most importantly, the paper articulates and illustrates the need to originate the classification process from an industry-specific perspective that is grounded in a good understanding of the industry's production process.

II. The Process of Classifying Service Products

The product classification system presently used for the U.S. manufacturing sector has evolved over nearly a century of time. Its present form, therefore, represents a long period of learning by doing with respect to bringing the several dimensions of data collection into alignment. For example, Census has learned much over the years about the products that each manufacturing industry produces, how to measure the output of such products, what is the appropriate reporting unit for collecting the output data, and how to create a formal coding system for classifying the multitude of manufactured products.(3) The situation in service industries, however, is quite different; here no formal classification system exists and even the identification of the products produced by service industries is, at best, very limited. As such, the process of classifying service products must necessarily be a comprehensive effort that addresses both the conceptual issues and the data collection issues necessary to insure that the classification system created for services is conceptually sound, feasible to implement, and satisfies the operational objectives set for it.

A. Guiding Principles

In his insightful and instructive papers on industrial classification issues, Jack Triplett enunciated two distinct conceptual paradigms to guide the overall process of economic classification under the new North American Industry Classification System (NAICS):

  1. the classification of industries should be erected on a production-oriented, or supply-based, conceptual framework, and
  2. the classification of products produced by industries should be based on a market-oriented, or demand-based, conceptual framework.(4)

Both of Triplett's classification frameworks were adopted by the three countries that are party to NAICS, as the bedrock concepts for classifying industries and products.(5) For example, the preface to the just published NAICS classification manual for industries states:

Economic units that have similar production processes are classified in the same industry, and the lines drawn between industries demarcate, to the extent practicable, differences in production processes.(6)

And with respect to product classification, the three-country agreement states:

The statistical agencies of the three countries also agree that market-orientated, or demand-based, groupings of economic data are required for many purposes, including studies of market share, demand for goods and services, import competition in domestic markets, and similar studies.(7)

Due to differences in the analytical uses served by the end products of each classification system, however, Triplett [1994a, p.7] also advises:

...there is no reason to integrate a product grouping system with an industry classification system, and there is every reason to avoid linking the two when they are in fact different.

Recently, however, the ECPC has adopted the following proposition as the "first" principle for developing a classification system for the products produced by service industries:

An understanding of the production process of the establishments included in the respective service industries is a required first principle for the process of classifying the products produced by those industries.

This principle implies that knowledge of the production process is an essential part of the conceptual foundation for constructing not only the NAICS industry classification system but also a product classification system for services that will satisfy the ultimate objective espoused by Triplett and the ECPC -- a system that accommodates a demand-side/market-orientated grouping and aggregation of the products produced by service industries. Because it may be misinterpreted as contradicting Triplett's aforementioned admonition, it is important to note that Triplett and the ECPC explicitly recognized the importance of the first principle in service industry classification early on:

In order to classify establishments producing services, one must have an understanding of what these establishments are producing. This is as true for a production-orientated classification system as for a market-orientated one.(8)

Put simply, one cannot begin to create a demand-side product classification system without first determining what are in fact the final products produced by, versus the inputs consumed by, a given industry's production process. For the classification of service products, a particularly useful articulation of the first principle is found in Sherwood [1997, p.3]:

To correctly define the output of a service industry it is essential to specify exactly what the industry agrees to sell and what the customer agrees to buy. That is, a determination must be made of what is implicitly "contracted for" when a transaction takes place. Further, it is important to distinguish between the output the industry produces and the activities carried out by the industry to produce the output.

B. What Is a Service Product

As its first objective, the classification process for service products must address and resolve the conceptual issues related to (1) identifying and defining the products produced by each service industry and (2) determining the most appropriate unit for measuring the nominal output of those service products. A sense of the difficulty inherent in this objective is suggested immediately by the intangible nature of the boundary conditions that are widely accepted and used to distinguish a service-type product from a goods-type product; namely, unlike goods, services are consumed on the spot; they cannot be stored, transported, or transferred and may often be invisible. As is evident, these boundary conditions tell us what a service isn't, but they do not tell us what it is. T. P. Hill [1977, p. 318] partially fills this conceptual void with his much cited general definition for a service:

A service is a change in the condition of a person, or of a good belonging to some economic entity, brought about as the result of the activity of some other economic entity, with the approval of the first person or economic entity.

Although still quite abstract, it is clear that Hill's definition can satisfy both dimensions of the objective in simple situations. For example, under Hill's definition, traditional hair cutting at a conventional barber shop fully satisfies the first dimension because it identifies and defines a distinct final product -- a hair cut -- that is produced by the shop and sold to the final consumer. And, it satisfies the second dimension because it implicitly tells us that the revenues derived from producing this service product is an appropriate measure of the shop's nominal gross output for this service product. However, when we move to more complex situations, the number of activities included in the production process proliferates, and Hill's general definition becomes a progressively less useful guide for the classification process. Consider, for instance, hair cutting at a salon where the activity of hair cutting is automatically preceded by hair washing and succeeded by blow drying. The fundamental conceptual question for the classifiers in this example is the following:

Within the production process of the salon, what is the objective of these additional activities -- is it to produce intermediate outputs/services that are subsequently bundled into the salon's final product and make it distinct from that of the barber shop (a salon-type hair cut as opposed to a barber shop-type hair cut) or is it to produce more types of final products (hair washing, blow drying) in addition to hair cutting?

To accommodate these more complex situations, the Bureau of Labor Statistics (BLS) developed a definition that defines a service product in a manner that explicitly recognizes the importance of using the production process framework to guide it conceptually in identifying the products produced by service industries and in collecting revenue and producer price data for these products:

We consider a service to be a bundle of goods and labor activities provided to a customer to accomplish a given function and the service must be consumed at the time it is provided. Both the provider of the service and the consumer of the service must agree on the basic goal of the activity.(9)

Although the goods included in the "bundle" in the BLS definition are physical products that are provided by the service supplier to the customer, it is important to note that these goods are generally intermediate inputs to rather than products of the service supplier's production process. And, unlike the service product per se, these goods may not be consumed when purchased and do not necessarily involve activity by the service supplier when ultimately consumed.(10) Thus, a ticket vendor contracts with a printing company to print and deliver tickets for a future "free" concert; the vendor then bundles the cost of both the tickets and its "vending service" into a ticket price that allows the buyer to attend the conference (consume the ticket) at the stated future date. The BLS definition and this example, however, bring to the fore the second, or unit of measure, dimension of the first objective in classifying service products. This is because the BLS definition seems to suggest that the best nominal output and prices measures for the service product always correspond to the bundled or gross output concept that includes the cost of any goods included in the delivered bundle. If, however, the good and the service(s) in the bundle are functionally independent (say, e.g., if the ticket is separable from the vending service or if loanable funds are separable from a bundle of loan services) then the cost of the goods sold (the ticket or loanable funds) can be netted out in theory, if not always in practice, to obtain nominal output and price measures that correspond better to the pure service or gross margin concept of output(11).

C. Methodology

The foregoing discussion indicates that the process of developing an operational classification system for service products must not only proceed from an industry-specific perspective but also embody a comprehensive methodology that addresses identification, measurement, and data collection, and classification issues. In turn these requirements suggest that the classifiers implement their mandate in the context of a methodology that incorporate the following sequence of six activities or steps:

  1. Developing a heuristic but informed description of the production process for each industry
  2. Identifying/defining the final products produced by each industry
  3. Determining the appropriate unit for measuring the nominal output of a given product and assessing the feasibility of collecting both nominal output and price measures for these units
  4. Researching the reporting unit that is appropriate for collecting product data for each industry
  5. Developing formal definitions for the identified products
  6. Creating a formal classification system that allows users to:
  7. identify the amount of each product produced by each industry,
  8. aggregate common products across all industries, and
  9. group and aggregate products in a manner that satisfies the demand-side classification framework adopted by the three NAICS countries.(12)

Step 1: It is necessary to approach the process of product classification in services from the perspective of the production process because it provides the necessary conceptual framework for: (a) identifying the activities performed by a given industry, (b) facilitating an ordered consideration of information and competing hypotheses about the role/objective of those activities in the production process, (c) developing informed judgements about the final products produced by the industry, and (d) providing insights into the transaction unit that is appropriate for measuring the respective products and the reporting unit that is appropriate for collecting the data. This approach is especially important in the case of service industries where, in contrast to, say, manufacturing, there exists much confusion, indeed ignorance, about what many service industries do and how they do it. Accordingly, as their first order of business, the classifiers should develop an informed heuristic description of each industry's production process.

Step 2: Given this model, the classifiers can use it to formulate a cogent hypothesis about the ultimate objective or output(s) of those activities, separate the output of those activities into final products and intermediate products, and identify for inclusion in the product classification system only the final products -- those produced for sale outside the reporting units in each industry. Thus, the classification system should not include intermediate products -- those produced by the unit but then consumed by it as an input to its final output -- because the value of such products is already counted in the value of the unit's final products.(13) Integral to this step of the classification process is the need to define and consider what the industry intends its output to be. In that regard, a study by Sherwood [1994, p. 12] suggests that the role of the consumer in the industry's production process must also be considered before one can identify and define the service product produced by that process:

To specify a service output correctly, we need to separate the change in the consumer brought about by the supplier [of the service] from other changes that might occur when the service is provided.

For example, T.P. Hill [1977, pp. 322-23] defines the service product of, perhaps, most visits to a doctor's office as the advice given, or course of treatment recommended, by the doctor to the patient. In other words, the change in the patient, which is required under Hill's previously noted general definition for a service product, is the knowledge/advice transferred to him by the doctor, not a cure:

The services provided by doctors, dentists, etc. are the changes in the condition of their patients which are directly attributable to their own actions....It is important not to over-step the boundary of production by seeking to attribute to doctors what is beyond their powers to provide. Doctors do not provide cures, still less good health.(14)

Thus, even if the patient's condition is worsened by following this advice, the office has still produced a service product because it is the advice, not the outcome, which identifies and defines the service product under Hill's definition. This example can be extrapolated to any number of service industries where advice is given or subjects are taught in order to identify and define the service product produced by those industries.(15)

Step 3: Although inextricably related to the two previous steps, the third step is unique in that it moves the classification process from the conceptual to the operational level, through explicit consideration of how the industry identifies and measures its outputs. It does this by addressing the need to define the unit of output in a manner that is (a) consistent with industry record- keeping practices and (b) appropriate for measuring both the nominal output and the price dimensions of the respective service products generated by a given industry. In this regard, the experience of BLS in collecting data for its producer price indexes for services is instructive. For example, BLS found that it could not collect price data for the service product produced by the reporting units in the Property and Casualty Insurance industry when it used the U.S. national income accounts measure of output (premiums less claims) but that it could collect data for a gross output measure defined as premiums plus the return on the invested portion of the premiums.(16)

In general, this step of the process also requires that the classifiers answer two fundamental questions: (1) is the output of a given service product best estimated by a gross margin or gross output measure and (2) what data must be collected to create that measure? As with the task of identifying the products produced, having a model of the industry's production process also provides valuable guidance on the unit of measure decision. To illustrate, consider that it is impossible in many service industries to identify and measure the nominal output of each specific service that the reporting unit performs for its varied customers. Rather, what is measurable is the revenue that the reporting unit receives from the bundle of services and goods combined within either a given class of activity in the unit's production process or a collection of such activities . Thus, we can observe the revenues (interest income) generated by the overall loan-making activity of a bank, but we cannot observe those associated with the individual services and activities (e.g., purchasing and transporting funds, matching term structures of borrowers with fund providers, credit investigation, loan administration, etc.) embodied in that extended activity. For most industries, these activity-based revenues, which represent the gross output from that activity, will be best for measuring the nominal output of a given service product. But, in some cases (the service product produced by trade and transportation industries or loan service products produced by financial intermediaries, for instance), one may conclude that the production process indicates that the gross margin earned by the activity is best. And this decision has data collection implications; namely, data must be collected not only for the gross revenue earned by the activity-bundle but also for the cost of goods sold that are included in the bundle. Finally, since the aforementioned choices address only the nominal output side of data collection; it is important to strive for choices that are also operational on the price collection side.

Step 4: The fourth step relates to the fact that the production process (and/or the record keeping practices) of the businesses in a given service industry may require collecting product data from a reporting unit that differs substantively from the time-honored establishment concept employed in manufacturing. While this step is often successively disposed of during the industry classification process, it is also likely that the correct reporting unit (indeed, the correct industry classification) may not become apparent until a description of the industry's production process has been prepared and understood. Marcus [1994], for example, reports that commercial banks frequently distribute the many activities embodied in their overall production process -- making loans, collecting deposits, etc. -- across several locations which either do not generate and/or record revenue for the activity performed at that location. Similarly, in a study of the medical services industry in Australia, White [1994] finds that many doctor's create legal entities that, at best, embody modest facilities (perhaps a phone) and whose sole activity, tax avoidance, bears no functional relationship to the production process by which the doctors supply medical services to their patients.

Step 5: The fifth step addresses the need to prepare formal definitions or descriptions for each service product identified for inclusion in the classification system.(17) For example, the classifiers for the finance sector of NAICS might identify "loan services" as a generic product class for banks and then define several specific types of loan services as distinct products within that product class.

Step 6: The sixth and final step produces the ultimate objective of the classification process. Here the classifiers will translate their considered deliberations from the previous steps into a formal coding system. And, this system must not only accommodate the demand-side grouping and aggregation of service products but also allow users to identify the value of each product produced on both an economy-wide/wherever-made basis and an industry-of-origin basis.

III. Example Applications of The Production Process

This section examines the production process of several industries to illustrate the fundamentally important role that knowledge of the production function plays in the product classification process. One of these industries--integrated steel--illustrates the importance of using this framework to guide the process of product classification in the manufacturing sector, while the remaining four industries--retail trade, banking, insurance, and doctor's offices and hospitals -- illustrate its importance to product classification for services industries. In all cases, however, the treatment and description of the industry's production process is intended to be primarily heuristic and suggestive rather than technical and definitive.

A. Integrated Steel Plant

The iron and steel mill industry (NAICS 331111) illustrates an important and rather ubiquitous problem that is encountered in the first or identification step of the product classification process for many goods- and servicing-producing industries. Suppose one examines the production accounts of an integrated iron and steel mill and observes that (a) the plant produces not only a wide array of steel products but also pig iron and coke and (b) the plant consumes both primary inputs -- capital and labor -- as well as a variety of intermediate inputs, including iron ore, coal, pig iron, and coke. Given this information, a natural first inclination would be to include the plant's steel products as well as its pig iron and coke products in the product classification system for NAICS 331111.(18) However, suppose that a flow chart description of the mill's overall production process reveals that it embodies four distinct and essentially separable activities: (1) a coke-making activity that creates coke from coal; (2) a blast furnaces that consumes all the coke produced and transforms iron ore into pig iron; (3) a steel-making furnace that consumes all the pig iron to generate molten raw steel; and (4) a continuous caster that transforms all the raw steel into a wide variety of steel products that can be further refined. Given this information about the production process, it becomes apparent that the coke and pig iron produced by these mills should be excluded from the product classification system because they are intermediate products whose value is already reflected in the value of the final

(i. e., steel) products produced by these mills.

B. Doctor's Offices and Hospitals

Analogous to the production process of the integrated steel plant, the process of the reporting units for doctor's offices that are included in NAICS subsectors 621 (Ambulatory Health Care Services) and for Hospitals (NAICS subsector 622) is also composed of several activities that produce intermediate outputs that are completely consumed in creating the final product(s) of the reporting units. Consider, for instance, an itemized health insurance bill that is received pursuant to a visit to a doctor's office or a stay in the hospital. In either case, the document is likely to display line-item charges for a wide menu of activities, including one or more of the following: x-rays, blood tests, other diagnostic tests/procedures, room charges (for hospital stay), administering drugs, remedial treatment, etc. Without benefit of understanding the production process, some observers might conclude that each of the charges on the itemized bill is prima facie evidence that the establishment in question is producing several distinct service products that should be included in the product classification system for these industries. However, the classifiers may discover that a systematic description of the production process reveals that none of these activities constitutes a final output of either a visit to a doctor's office or a stay in the hospital. Rather, this description might suggest that all of the listed activities are best interpreted as intermediate inputs that are either produced internally or purchased externally and then consumed by the health provider in the course of obtaining the final output/ultimate objective of its production process -- a diagnosis or a defined course of treatment.(19) If so, then the nominal value of the medical services product produced would be computed as the sum of the listed charges.(20)

C. Retail Trade

The two previous examples related to industries where the ultimate objective/output of the production process is best measured by gross output. For the industries contained in NAICS sector 44-45 (Retail Trade), however, it is widely believed that the output of their services product can be and should be measured by the gross margin earned by these industries.(21) The reason for this difference lies in the fact that the overall production process of the establishments classified into a particular retail trade industry is generally thought to be separable such that the process used to produce the service product of the establishment is functionally independent of the product(s) it vends. True, the sales price paid by the consumer for a retail trade transaction incorporates not only the costs of the goods vended by the establishment but also the price of a complex bundle of services provided by the establishment. However, the production process of most retailers is usually directed at producing services which make the goods vended accessible, desirable, and convenient to its customers, and it generally leaves the goods vended in their original state. Oi [1992, p.168] identifies several of the individual services often found in the service bundle produced by retail establishments:

consummating transactions; assembling, displaying, and providing information on goods; making the goods available at times and places convenient to customers; supplying additional services, such as delivery and credit; and packaging and processing goods into more sustainable forms (for example, cold cuts at the delicatessen counter of a supermarket.(22)

Thus, if the technology allows separation of the services bundle from the goods in the overall bundle purchased by the customer, it follows that the best nominal measure of the retailer's service product is computed as its sales receipts less its cost of goods sold.

D. Banking

Perhaps more than any other industry in the services sector, it is banking and related industries contained in NAICS subsector 522 (Credit Intermediation and Related Activities) that pose some of the most complex conceptual challenges that the classifiers will encounter in implementing the classification process for the products produced by service industries. In view of this fact, this section offers an extensive summary of the issues involved as well as suggestions for grounding the classification process on a solid understanding of an industry's production process.

D.1 The Controversy

Triplett succinctly states the banking problem:

....the measurement of banking [output] has been inhibited by two major unresolved questions: (1) What are the outputs? And (2) What are the inputs?(23)

The failure to obtain agreement on these two fundamental questions has contributed to more than a half century of much discussion, confusion, and disagreement about what banks produce.(24) The oldest and most important contributor to this controversy is the traditional treatment of interest flows by national income accountants in the U.S. and elsewhere. This treatment has two overriding characteristics:

  1. It treats interest paid by borrowers as an expense and a charge against bank value added rather than as the largest source of bank revenue -- a payment for the loan services provided by the bank, and
  2. It treats interest paid by banks as a source of bank value added rather than as a cost of acquiring loanable funds.

Although national income accountants apply this net interest approach to all industries, its application to banking has two serious and undesirable consequences:

  1. It implies that banks make a negative contribution to the value added of the overall economy,(25) and

  2. It implies that banks have a convoluted production process, wherein outputs become inputs and vice versa.

Bailey [1971, p. 260] provides a rationale for the income accounts treatment of interest flows and acknowledges the problem this treatment causes for measuring the value-added form of bank output:

...[w]hen an enterprise receives capital from the public and transfers it to other is reasonable to follow the general rule that the income and product originating in a firm shall depend only on its net interest payments, that is, interest paid minus interest received. If this rule were applied to banks, however, their contribution to income and product would be negative....

To overcome the first, or negative value-added, consequence of applying the net interest convention to banking, Ynmenta [1947] introduced the imputed-interest convention. Under this convention, banks are credited with making "imputed" sales of deposit services (in an amount about equal to the excess of interest earned over interest received) to account for the "free" services rendered to depositors (particularly, demand depositors), and with making an equal amount of "imputed" interest payments to those depositors in lieu of direct interest payments. Although it returned the banking industry to good economic standing, the imputed-interest approach to defining and measuring bank output has long been very unpopular in the literature because it doesn't address the second consequence noted above.(26) Representative of the opposition's fundamental argument, Speagle and Silverman (S-E) [1953, p. 130] suggest that the imputed-interest approach is premised on the proposition that the banking industry, "makes no explicit charges for the bulk of its output, which [it contends] consists chiefly of services such as bookkeeping, monetary transfers, and investments for depositors." S-E [1953, p. 129] reject this view and advocate instead an alternative -- services-approach -- that treats sales revenue to banks like other industries in the national accounts:

It should not matter what name, popularly, is given to this flow of business revenue -- receipts for goods sold, rent received, interest received -- as long as the recipient furnishes this service for economic gains.

More specifically, the services-approach advocates argue that the definition and measurement of both gross output and value-added in banking should be derived from the perspective of its apparent production function -- what banks do primarily is buy loanable funds and sell loan services. However, Berger and Humphrey [1992, p. 247] point out that this simple version of the services-approach model applies to only large money-center banks because these banks do in fact operate by obtaining the bulk of their funds by buying them (paying interest) from other banks and large depositors and by converting these purchased funds into loans. Nevertheless, there is now widespread agreement on an expanded version of the services-approach model, wherein the set of service products produced by most banks is recognized to include both loan and deposit services. Indeed, in recent years, even the national income accountants of most industrialized nations have abandoned the historical treatment of interest flows discussed in D.1 of this section. Instead, they have adopted the procedure recommended in the 1993 version of the System of National Accounts (SNA93), which allocates net interest received by banks between loan services and deposit services provided by banks.(27) Thus, SNA93 does not share the characteristics and consequences of the traditional SNA method that were discussed earlier in this section. But, there remains considerable disagreement as to whether deposit services provided by banks (in lieu of direct interest payments and service charges to, principally, demand depositors) constitutes an output or an input in the bank production function. Hancock summarizes the current state in banking as follows:

Although it is agreed that banking firms produce heterogeneous services, there has been little consensus on the measurement of their outputs and inputs. The outputs used by various researchers include total assets, earning assets, loans, total deposits, produced deposits, demand deposits in dollar terms, the number of deposits and loan accounts, gross operating income, and combinations of these measures....The central questions in what can be termed "the calculation problem" are (1) Which balance sheet items produce services that are net outputs, and which ones are net inputs? And (2) How does one measure the outputs and inputs, or put prices on them.... Another way of posing the problem is whether stock or flow variables measure the relevant concept of bank output and input.(28)

D.2 Heart of the Problem

At the heart of most all of the aforementioned difficulties in identifying, defining, and measuring the output of service products in banking is the fact that banks historically have not, either by virtue of regulation or by custom, fully transacted their deposit purchase and deposit services activities in an "arms-length" and explicit-price market. In such a market, banks would be required to purchase deposits -- an indispensable intermediate inputs for their loan making activities -- via the payment of explicit and market-determined interest rates and, in turn, they would require depositors to pay the full-price of the services consumed by them. In actual fact, however, banks use a combination of market and non-market transactions to purchase their inputs of loanable funds and to obtain payment for their output of deposit service and loan service products. Prior to deregulation and the abolition of Regulation Q in 1980, for example, commercial banks were subject to restrictions on the interest rate they could pay for deposits generally and they were prohibited from paying any interest for a major category of deposits -- demand deposits. As a result, commercial banks competed for demand deposits (and essentially circumvented Reg Q) by bartering; i.e., they paid for deposits by providing a variety of "free" or unpriced deposit services -- check cashing, ATM's, etc.

Assuming that were the whole story, one could argue that the free deposit service activities performed by banks are akin to the coke-making and iron-making activities in the integrated steel plant example above. That is, the primary purpose of these deposit activities is to create intermediate outputs (loanable funds), which are then used as inputs to the banks primary activity and product (loan making) and whose cost is fully reflected in the prices banks charge for and the revenues they receive from loan making. Mohr [1990, pp.24-9] and Berger and Humphrey [1992, p. 252], however, suggest that there is more to the story. Mohr, for example, articulates a widely held position by bank researchers:

There is, however, a fundamental difference between [unpriced] deposits and coke. Namely, the act of depositing funds in a bank simultaneously enables the bank to make loans which earn interest, some portion of which the depositor is entitled to as a reward for keeping the funds in the bank.

Further, Berger and Humphrey's research indicates that depositors, not borrowers, pay for most, if not all, of the services provided to them through a combination of explicit and implicit service prices. The latter represents the imputed interest rate (known as the earnings credit rate or ECR) that banks routinely apply to demand deposits that have a "minimum" or "compensating balance" restriction. Finally, as a result of the initial deregulation of banks in 1980 and the subsequent deregulation in 1982, which produced a surge of substitutes for conventional deposit accounts (such as NOW and money market accounts with check writing privileges) offered by nonbank deposit institutions and brokerage firms, the percentage of deposits acquired through direct interest payments and the percentage of explicitly and implicitly priced deposit services has risen sharply.(29) Nevertheless, a large portion of bank demand deposits are still acquired by barter.

D.3 The SNA93 Solution

In the wake of the release of the aforementioned SNA93 by the United Nations, a much improved consensus now appears to be emerging on what banks produce, how they produce it, and how to measure it. There are three hallmarks of the SNA93 approach to identifying and measuring the products produced by banking. First, there is the fact that SNA93 abandons the contentious practice of applying the traditional income accountant's treatment of interest flows to banking and related industries. Thus, in his discussion of banking under SNA93, Obst [1998, p. 5] states:(30)

Perhaps a fundamental point to make is that, in the calculation of the output of banks, the normal [i.e., traditional] role of interest in the calculation of bank output is not applied. That is, while the value of interest received and paid is not considered in the definition of outputs or inputs for any other industries, interest flows are an integral part of the calculations for banks. Indeed, it is the fact that interest is such an integral part of banks operations compared to other industries that a different treatment is required. Certainly, in measuring productivity it is clear that the SNA treatment of interest is a contentious point, but for banking no such contention should exist.

Second, to avoid not only misallocating gross interest receipts from loans entirely to loan services but also double counting for unpriced deposit services that are already included in interest receipts, SNA93 treats the interest receive by banks from its loan-making activities as a composite flow that can be decomposed into three components: (1) a pure interest charge, (2) a loan service charge, and (3) a return to cover the charge for unpriced deposit services. Arithmetically, this decomposition evolves from the following accounting identity:

GI = cost of funds + return to loan services + return to unpriced deposit services

= (r1 + sL)L + sDD

= rLL + sDD

where GI = gross interest flows from loan activities,

rL = (r1 + sL) = loan rate in the absence of unpriced deposit services,

r1 = imputed pure interest rate cost of funds,(31)

sL = loan service charge rate,

sD = rate of return to unpriced deposit services,

L = dollar value of loan balances, and

D = dollar value of deposit balances.

Third, SNA93 defines the output of unpriced deposit services, sDD, as the margin

(r1 - rD)D by which imputed interest paid to depositors exceeds the interest actually paid, where rD is the average rate of explicit deposit interest payments by the bank that is obtained by dividing explicit deposit service revenues by the value of deposits. And, it computes the nominal output of loan services as the imputed gross margin, (rL - r1)L, that is earned by banks in excess of the imputed pure interest cost of funds r1L. This process of allocating interest receipts and indirectly measuring the aggregate values of both loan services and unpriced deposit services, known as Financial Intermediation Services Indirectly Measured (FISIM) in SNA93, can be expressed in terms of the GI equation above as:

GI - rDD - r1L = net interest received - r1L

= (rL - r1)L + (r1 - rD)D

= sLL + sDD

In light of the foregoing, the classifiers might consider approaching the identification, definition, measurement and data collection of the service products produced by the reporting units in banking within a production process framework that has the following characteristics:(32)

  1. It combines primary and intermediate inputs to generates a variety of products final to the industry, including primarily loan services but also including priced deposit services as well as fee-for-service products related to the sale of investment, insurance, fiduciary, accounting, computer, and other services.
  2. It defines and measures the nominal output of priced deposit services by the reporting unit as the sum of explicit revenues earned by these services.
  3. It defines and measures the nominal output of unpriced deposit services by the reporting unit as the sum of the implicit revenues earned by applying the unit's ECR to its deposit balances.
  4. It computes the output of loan services as a gross margin measure.(33)
  5. It defines loan service revenues broadly to include not only the interest received from traditional loan accounts but also the income (dividends and capital gains) generated from investments in marketable securities such as stocks and bonds.(34)
  6. It measures the value of other service products generated by the reporting unit's production process by the explicit revenues earned by those products.

Finally, in adopting the foregoing framework, the classifiers will need to carefully consider what data the statistical agencies will be required to collect in order to produce both nominal and price-adjusted output measures for the range of banking products they identify for inclusion in the classification system.

E. Insurance

NAICS subsector 524 (Insurance Carriers and Related Activities) provides another interesting demonstration of the importance of the first principle of product classification in identifying, defining, and measuring the output of service industries.(35) This subsector contains a wide variety of industries whose component businesses operate in one or more segments of the overall insurance market -- life insurance, health insurance, property and casualty insurance, automobile insurance, etc. Regardless of the segment(s) they operate in, however, all insurance carriers share a common attribute: the ultimate objective of their production process is to insure policyholders against the financial risk that emanates from a defined risk factor or factors-- death, illness, loss of job, fire, flood, driving, etc. Entities (businesses, households, etc) contract with insurance companies to perform this risk-absorption function by taking out policies which set forth the exact terms of the contract and by paying premiums. In part insurance companies achieve their risk-absorption commitment not only by collecting premiums from individual policyholders but also by performing a variety of activities that are designed to create and manage a large pool of policyholders in order to spread the risk associated with any individual or group of claims against it. Given this fact, several analysts, including national income accountants in the U.S. and elsewhere, have adopted the convention that the nominal output of the insurance industry can be measured as the net premiums received by industry (gross premiums received less claims paid). This convention is justified by Hirshhorn and Green [1980, p.1522], who argue that

A life insurance company is able to offer protection because it has created the facilities for pooling risks; and it is the range of activities a life insurance company undertakes as part of its efforts to maintain its capacity for pooling risks that constitutes the services provided by the company.

Sherwood [1997] characterizes the net premiums approach as a value-added measure of output that fails not only to properly identify and define the ultimate objective/output of the industry's production process (the contractually assumed quantity of risk) but also to distinguish between the output of the total process and that of the administrative activity component of the process:

Thus, the net premiums notion is consistent with a concept of output that is [derived from] the activities carried out by the industry to maintain the capacity for pooling risk.(36)

From this perspective, then, gross premiums is a better measure of the service product produced by insurers because it includes claims, which by definition represent a major portion of the quantity-of-risk assumed.

Moreover, it is arguable that the annual output measure for the insurance industry should include more than the gross premiums collected in a given year. Consider, for example, what is required of an insurance carrier if it is to fully and responsibly discharge the risk protection commitment made to its policyholders. As a general rule, it is clear that this commitment requires the carrier not only to collect premiums but also to invest them in order to provide a stock of fungible assets that can cover both the level of anticipated claims submitted to it each year as well as some level of catastrophic claims. Indeed, it is also likely that the targets for both types of claims are adjusted annually to reflect inflation in both the general price level and in more specific prices such as, say, the size of awards awarded to plaintiffs in civil suits. Should the carrier fail to operate in such a prudent manner, it could be forced through bankruptcy to default on its fundamental obligation to policyholders. Sherwood and others adopt this view and posit that the production process of the typical insurance provider (particularly, life insurers and property and casualty insurers) can be expected to show a structure that devotes much, if not most, of its inputs to finding and developing avenues of investment for the premiums and other funds raised by its on-going activities. It follows, therefore, that the sum of the annual proceeds generated by the carrier's premium investment activities -- interest, rents, dividends, capital gains, etc. -- plus the gross premiums generated from its insurance vending activities is a legitimate measure of both the carrier's current success in carrying out its risk absorption obligations and of its annual gross output.(37)

IV. Conclusions

The effort to create an operational classification system for service products will require the classifiers to undertake a considerable learning process in many industries in order to address and resolve the difficult conceptual and measurement issues discussed in this paper. Accordingly, it would be most helpful if industry experts from the business and academic communities would volunteer their expertise and knowledge to the classification groups assembled from the Federal agencies. We believe that such cooperation would not only expedite the learning process but also improve the quality and usefulness of the final product obtained from the overall classification process.


Bowman R. T. and R. A. Easterlin [1958], "The Income Side: Some Theoretical Aspects," in Victor Fuchs, ed., Production and Productivity in the Services Industries, NBER Studies in Income and Wealth, vol. 34, Irvington-on-Hudson, New York: Columbia University Press, pp.149-86.

Bailey, M. J. [1971], National Income and the Price Level, second edition, New York: McGraw-Hill Book Company.

Benston G. J., G. A. Hanweck, and D. B. Humphrey [1982], "Scale Economies in Banking: A restructuring and Reassessment," Journal of Money, Credit, and Banking, vol. 14, pt.1, November, pp. 435-50.

Berger A. N. , A. K. Kashyap, and J. M. Scalise [1995], "The Transformation of the U. S. Banking Industry: What a Long, Strange Trip It's Been," Brookings Papers on Economic Activity, No.2, pp. 55-218.

_______ and D. B. Humphrey [1992], in Z. Griliches, ed., Output Measurement in the Services Sector, NBER Studies in Income and Wealth, vol. 56, Chicago and London: The University of Chicago Press, pp. 245-79.

Economic Classification Policy Committee [1994], "Standard Industrial Classification Replacement," Federal Register, July 26, p. 38092-96.

_______ [1993a], "Issue Paper No. 6, Services Classifications," Bureau of Economic Analysis (BE-42), U.S. Department of Commerce, Washington: D.C., September.

_______ [1993b], "Issue Paper No. 1, Conceptual Issues," Federal Register, March 31, pp.16991-17000.

Gerduk, I. B. [1998], "PPI Experience in Formulating Product Lines in the Service Sector," paper presented at Thirteenth Meeting of Voorburg Group on Service Statistics, Rome: Italy, September 22, 1998.

Gorman, J. A. [1969], "Alternative Measures of the Real Output and Productivity of Commercial Banks," in Victor Fuchs, ed., Production and Productivity in the Services Industries, NBER Studies in Income and Wealth, vol. 34, Irvington-on-Hudson, New York: Columbia University Press, pp.155-89.

Griliches Z. (ed) [1992], Output Measurement in the Services Sector, NBER Studies in Income and Wealth, vol. 56, Chicago and London: The University of Chicago Press.

Gullickson, W. [1995], "Measurement of Productivity Growth in U.S. Manufacturing," Monthly Labor Review, (July), pp. 13-28.

Hill, T. P. [1977], "On Goods and services," Review of Income and Wealth, vol. 123 (4), pp. 315-38.

Hill, P. [1998], "Interest in National Accounts," paper prepared for Workshop on Measuring Bank Output, The Brookings Institution, Washington, DC, November 20, 1998.

Mamalakis, M. J. [1987], "The treatment of Inteest and Financial Intermediaries in the National Accounts: The Old "Bundle Versus the New "Unbundle Approach," Review of Income and Wealth, June, pp.169-92..

Marcus, S. J. [1994], "Expanding Service Coverage in the U.S. Economic Censuses," in Papers and Final Report, Ninth Meeting of Voorburg Group on Service Statistics, Sydney:Australia, October, pp.6-18.

Mohr, M. F. [1990], "Economic Accounting for the Banking Industry and Interest Flows in the NIPA's: The Case for the Services Approach," mimeo, January.

Obst, C. [1998], "A Note on Banking Output: Recent National Accounts Developments," paper prepared for Workshop on Measuring Bank Output, The Brookings Institution, Washington, DC, November 20, 1998.

Office of Management and Budget [1998], North American Industry Classification System, United States, 1997, Bernan Press: Landham, Maryland.

_______ [1997], "1997 North American Industry Classification System--1987 Standard Industrial Classification Replacement," Federal Register, April 9.

Oi, W. Y. [1992], "Productivity in Distributive Trades," in Z. Griliches, ed., Output Measurement in the Services Sector, NBER Studies in Income and Wealth, vol. 56, Chicago and London: The University of Chicago Press, pp. 161-93.

Salem, M. [1998], "Banking Services in Canadian SNA," paper prepared for Workshop on Measuring Bank Output, The Brookings Institution, Washington, DC, November 20, 1998.

Sealey, C. W. and Lindley J. [1977], "Inputs, Outputs, and a Theory of Production and Cost at Depository Financial Institutions, Journal of Finance, Vol.32 (4), pp. 1251-66.

Sherwood, M. J. [1994], "Difficulties in the Measurement of Service Outputs," Monthly Labor Review, March, pp. 11-19

[1997], "Output of the Property and Casualty Insurance Industry," paper (revised February 1998) presented at the Centre for the Study of Living Standards Conference on Service Sector Productivity and the Productivity Paradox, Ottawa, April 11-12, 1997.

Triplett, J. E. [1994b], "Economic Concepts for Economic Classifications, Survey of Current Business, November, pp. 45-49.

_______ [1994a], "Economic Classification in the New North American Industry Classification System (NAICS)," paper presented at Seminar on New Directions In Statistical Methodology, Washington, DC, May 25, 1994.

_______ [1991], "Measuring the Output of Banks: What Do Banks Do?" BEA discussion paper no.53, Washington, DC: Department of Commerce

_______ [1990], "The Theory of Industrial and Occupational Classifications and Related Phenomena," in 1990 Annual Research Conference, Proceedings in Arlington, Virginia, March 18-21, 1990, by the Bureau of the Census (Washington, DC: U.S. Government Printing Office), pp. 9-25.

United Nations, Organization of Economic Cooperation and Development, World Bank, International Monetary Fund, Commission of European Communities [1993], System of National Accounts, 1993, Brussels/Luxembourg, New York, Paris, Washington, D.C.

White, P. [1994], "Medical Services Industries: Report on Research in Progress," in Papers and Final Report, Ninth Meeting of Voorburg Group on Service Statistics, Sydney:Auatralia, October,1994, pp. 454-475.

Industry Appendix

U.S. Industries Arranged by NAICS Sectors, Subsectors, and Industry Groups


511 Publishing Industries

5111Newspaper/Periodical/Book/Database Publishers

51111 Newspaper Publishers

51112 Periodical Publishers

51113 Book Publishers

51114 Database and Directory Publishers

51119 Other Publishers

511191Greeting Card Publishers

511199 All Other Publishers

5112 Software Publishers

51121 Software Publishers

512 Motion Picture & Sound Recording Industries

5121 Motion Picture & Video Industries

51211 Motion Picture & Video Production

51212 Motion Picture & Video Distribution

51213 Motion Picture & Video Exhibition

513 Broadcasting & Telecommunications

5131 Radio & Television Broadcasting

51311 Radio Broadcasting

513111Radio Networks

513112 Radio Stations

51312 Television Broadcasting

5132 Cable Networks & Program Distribution

51321 Cable Networks

51322 Cable & Other Program Distribution

5133 Telecommunications

51331 Wired Telecommunications Carriers

51332 Wireless Telecom Carriers (exc Satellite)

513321 Paging

513322 Cellular & Other Wireless Telecommunications

51333 Telecommunications Resellers

51334 Satellite Telecommunications

51339 Other Telecommunications

514 Information & Data Processing Services

5141 Information Services

51411 News Syndicates

51412 Libraries & Archives

51419 Other Information Services

514191 On-Line Information Services

514199 All Other Information Services

5142 Data Processing Services

51421 Data Processing Services


521 Monetary Authorities - Central Bank

5211 Monetary Authorities - Central Bank

52111 Monetary Authorities - Central Bank

522 Credit Intermediation & Related Activities

5221 Depository Credit Intermediation

52211 Commercial Banking

52212 Savings Institutions

52213 Credit Unions

52219 Other Depository Credit Intermediation

5222 Nondepository Credit Intermediation

52221 Credit Card Issuing

52222 Sales Financing

52229 Other Nondepository Credit Intermediation

522291 Consumer Lending

522292 Real Estate Credit

522293 International Trade Financing

522294 Secondary Market Financing

522298 All Oth Nondepository Credit Intermediation

5223 Activities Related to Credit Intermediation

52231 Mortgage & Nonmortgage Brokers

52232 Financial Clearinghouse & Reserve Activities

52239 Other Credit Intermediation Activities

523 Security, Commodity Contracts & Like Activity

5231 Scrty & Comdty Contracts Intermed & Brokerage

52311 Investment Banking & Securities Dealing

52312 Securities Brokerage

52313 Commodity Contracts Dealing

52314 Commodity Contracts Brokerage

5232 Securities &Commodity Exchanges

52321 Securities &Commodity Exchanges

5239 Other Financial Investment Activities

52391 Miscellaneous Intermediation

52392 Portfolio Management

52393 Investment Advice

52399 All Other Financial Investment Activities

523991 Trust, Fiduciary & Custody Activities

523999 Miscellaneous Financial Investment Activities

525 Funds, Trusts & Other Financial Vehicles

5251 Insurance &Employee Benefit Funds

52511 Pension Funds

52512 Health & Welfare Funds

52519 Other Insurance Funds

5259 Other Investment Pools & Funds

52591 Open-End Investment Funds

52592 Trusts, Estates &Agency Accounts

52593 Real Estate Investment Trusts

52599 Other Financial Vehicles


541 Professional, Scientific & Technical Services

5411 Legal Services

54111 Offices of Lawyers

54112 Offices of Notaries

54119 Other Legal Services

541191 Title Abstract & Settlement Offices

541199 All Other Legal Services

5412 Accounting/Tax Prep/Bookkeep/ Payroll Services

54121 Accounting/Tax Prep/Bookkeep/ Payroll Services

541211 Offices of Certified Public Accountants

541213 Tax Preparation Services

541214 Payroll Services

541219 Other Accounting Services

5413 Architectural, Engineering &Related Services

54131 Architectural Services

54132 Landscape Architectural Services

54133 Engineering Services

54134 Drafting Services

54135 Building Inspection Services

54136 Geophysical Surveying & Mapping Services

54137 Surveying/Mapping (exc Geophysical) Services

54138 Testing Laboratories

5414 Specialized Design Services

54141 Interior Design Services

54142 Industrial Design Services

54143 Graphic Design Services

54149 Other Specialized Design Services

5415 Computer Systems Design &Related Services

54151 Computer Systems Design &Related Services

541511 Custom Computer Programming Services

541512 Computer Systems Design Services

541513 Computer Facilities Management Services

541519 Other Computer Related Services

5416 Management, Sci & Tech Consulting Services

54161 Management Consulting Services

541611 Admin & Gen Management Consulting Services

541612 Human Res & Exec Search Consulting Services

541613 Marketing Consulting Services

541614 Process, Phys Dist & Log Consulting Services

541618 Other Management Consulting Services

54162 Environmental Consulting Services

54169 Oth Scientific & Technical Consulting Services

5417 Scientific R&D Services

54171 R&D in Physical, Engineering & Life Sciences

54172 R&D in Social Sciences & Humanities

5418 Advertising & Related Services

54181 Advertising Agencies

54182 Public Relations Agencies

54183 Media Buying Agencies

54184 Media Representatives

54185 Display Advertising

54186 Direct Mail Advertising

54187 Advertising Material Distribution Services

54189 Other Services Related to Advertising

5419 Oth Professional/Scientific/ Technical Service

54191 Marketing Research & Public Opinion Polling

54192 Photographic Services

541921 Photography Studios, Portrait

541922 Commercial Photography

54193 Translation & Interpretation Services

54194 Veterinary Services

54199 All Oth Prof, Scientific & Technical Services


561 Administrative & Support Services

5611 Office Administrative Services

56111 Office Administrative Services

5612 Facilities Support Services

56121 Facilities Support Services

5613 Employment Services

56131 Employment Placement Agencies

56132 Temporary Help Services

56133 Employee Leasing Services

5614 Business Support Services

56141 Document Preparation Services

56142 Telephone Call Centers

561421 Telephone Answering Services

561422 Telemarketing Bureaus

56143 Business Service Centers

561431 Private Mail Centers

561439 Oth Business Service Centers (incl Copy Shops)

56144 Collection Agencies

56145 Credit Bureaus

56149 Other Business Support Services

561491 Repossession Services

561492 Court Reporting & Stenotype Services

561499 All Other Business Support Services

5615 Travel Arrangement & Reservation Services

56151 Travel Agencies

56152 Tour Operators

56159 Oth Travel Arrangement & Reservation Services

561591 Convention and Visitors Bureaus

561599 All Oth Travel Arrange & Reservation Services

5616 Investigation & Security Services

56161 Investigation, Guard & Armored Car Services

561611 Investigation Services

561612 Security Guards & Patrol Services

561613 Armored Car Services

56162 Security Systems Services

561621 Security Systems Services, (except Locksmiths)

561622 Locksmiths

5617 Services to Buildings & Dwellings

56171 Exterminating & Pest Control Services

56172 Janitorial Services

56173 Landscaping Services

56174 Carpet & Upholstery Cleaning Services

56179 Other Services to Buildings & Dwellings

5619 Other Support Services

56191 Packaging & Labeling Services

56192 Convention & Trade Show Organizers

56199 All Other Support Services

562 Waste Management & Remediation Services

5621 Waste Collection

56211 Waste Collection

562111 Solid Waste Collection

562112 Hazardous Waste Collection

562119 Other Waste Collection

5622 Waste Treatment & Disposal

56221 Waste Treatment & Disposal

562211 Hazardous Waste Treatment & Disposal

562212 Solid Waste Landfill

562213 Solid Waste Combustors & Incinerators

562219 Other Nonhazardous Waste Treatment & Disposal

5629 Remediation & Oth Waste Management Services

56291 Remediation Services

56292 Materials Recovery Facilities

56299 All Other Waste Management Services

562991 Septic Tank & Related Services

562998 All Other Miscellaneous Waste Management

1. See Office of Management and Budget [1997, 1998].

2. Phase 1 will be confined to classifying the products produced by the industries in four selected NAICS service sectors -- Information (sector 51), Finance and Insurance (sector 52) except Insurance (subsector 524), Professional, Scientific, and Technical Services (sector 54), and Administrative and Support, Waste Management and Remediation Services (sector 56). The detailed U.S. industries included in these sectors are listed in the Industry Appendix at the end of this paper.

3. The coding scheme and description of the products included in it are published periodically by the Census Bureau in the Numerical List of Manufactured and Mineral Products.

4. Among the papers included in this body of work are Triplett [1990, 1994a, and 1994b]; see also Economic Classification and Policy Committee [1993b].

5. See Economic Classification and Policy Committee [1994] and Office of Management and Budget [1997].

6. See Office of Management and the Budget [1998], p. 3.

7. See Economic Classification and policy Committee [1994], p. 38094.

8. See Economic Classification Policy Committee [1993a], section 6.5.

9. See Gerduk [1998], p. 3.

10. Outside of trade and transportation industries where it is common, it is perhaps rare that the goods included in the bundle are not consumed when purchased.

11. The BLS methodology also recognizes and incorporates these unit of measure decisions; see Gerduk [1998], p. 4.

12. In this regard, Triplett [1994a, p. 6] says of a product grouping system, "It should incorporate, and facilitate the analysis of, the relationships among products--demand relations, substitution relations, marketing relationships, uses by consumers or by other ultimate purchasers.

13. An ideal classification system would allow for the elimination of products produced and consumed at each level of aggregation; for example, the product output measures for each NAICS national industry and sector would include only the output of products final to the industry and sector. This requirement is more demanding than the procedure presently used by the Census Bureau to collect product shipments data from the establishments in the manufacturing and mining sectors. That procedure collects data on the unduplicated shipments of products at the establishments level within each NAICS national industry of each sector. However, it does not necessarily follow that the aggregate of the establishment-level output measures for each product is equal to the unduplicated output of the product for the national industry because of intra-industry production and consumption of products.

14. See Hill [1977], p. 322.

15. Interestingly, however, Hill [1977, p. 324] maintains that no output is produced when a teacher fails to improve/change the condition of poor students because they are unable to absorb the knowledge the teacher is trying to transmit to them. In contrast to Hill's position, Sherwood [1994, p. 15] states that, " it is reasonable to argue that if someone is paying the teacher to be in the classroom, the teacher must be providing some service."

16. See Gerduk [1998], p. 6.

17. An example of task (1) is the "Numerical List of Manufactured and Mineral Products" prepared by the Census Bureau for the 1997 Economic Census of these industries.

18. This approach, often found in the productivity literature, can be described as the duplicated output and input approach. Under this method, the same values of coke and pig iron that are defined as outputs would also be included in the input vector along with the coal, iron ore, and other inputs used to make steel, coke, and pig iron. However, this method of accounting produces a downward bias in total-factor productivity growth estimates because it adds equal amounts to output and input. For this reason, BLS employs the unduplicated output and input approach discussed in the text; see Gullickson [1995].

19. Following Hill [1977, pp. 322-23], this definition of nominal output excludes "cures" which may or may not follow from the treatment received from the doctor's office or hospital. Sherwood [1994, p. 13] suggests treating the cure factor in medical treatment as part of the quality component in the price paid for such treatment.

20. Research conducted by BLS suggests that the above scenario fits better the record keeping practices of hospitals because, unlike doctor's offices, hospitals clearly earmark the itemized expenses to a particular treatment code. In addition, BLS finds that the charges attached to activities on a bill for a visit to doctor's office generally include a markup over costs. As a result, BLS has chosen to treat the listed activity charges for a visit to a doctor's office as products rather than inputs whose sum equals the value of production. However, if doctors are simply spreading the total markup for their final product over the listed activities, simple summation would still produce a correct measure of the nominal value of final product. And, a price index for deflating this measure could be estimated by creating a composite price index from the unit "prices" and "output" values collected for each activity.

21. This belief also generally extends to other distributive industries, including most of those classified by NAICS into Wholesale Trade (sector 42) and Transportation and Warehousing (sector 48-49).

22. This passage, which rephrases Oi's original document, was taken from Sherwood [1994], p.12.

23. See comments by Triplett in Griliches [1992], section III.A: Measuring the Output of Banking and Services, pp. 287-88.

24. Triplett [1991] provides an excellent review of the literature on output measurement in banking.

25. Net interest paid (interest paid - interest received) by banks is not only negative but also larger than the sum of the other factor incomes generated by the industry -- wages and salaries, depreciation, profits, etc.

26. See, for example, Speagle and Silverman [1953], Bowman and Easterlin [1958], Warburton [1958], Sealey and Lindley [1977], Mamalakis [1987], Mohr [1990], and Triplett [1991]. Gorman [1967], however, defends BEA's use of the imputed-interest approach vis-a-vis the services approach.

27. See United Nations et al [1993].

28. See comments by Hancock in Griliches [1992], section III.A: Measuring the Output of Banking and Services, p. 296.

29. The history and details of recent bank deregulation acts are well documented in Berger et al [1995].

30. Also see papers by Hill [1998] and Salem [1998].

31. Also called a reference rate in the SNA93 literature, r1 appears to be conceptually equivalent to the ECR discussed in section D.2.

32. This approach is in strong contrast to what Triplett [1992, p. 290] calls the "activity" approach to discriminating between bank outputs and inputs. Using this approach, Benston, Hanweck, and Humphrey [1982, p. 440] concluded that, "Output should be measured in terms of what banks do that cause operating costs to be incurred." The difficulty with this standard for classification is demonstrated by the fact that it would double count the output of integrated steel mills because it would treat the output of their coke and pig iron activities as final rather than intermediate outputs.

33. Although not apparent, the output measure for banking found in BEA's input-output (I-O) tables implicitly incorporates a gross margin measure for loan services. This fact is easily demonstrated. First, let Y = I-O output measure and note that Y is defined by I-O as:

Y = imputed interest paid + monetary service charges received from all sources.

= {monetary interest received + imputed interest received + dividends and other property income} - {monetary interest paid + nondeposit monetary interest paid}+ monetary service charges received from all sources

= gross margin income from loan services + gross monetary service charges received from all sources.

34. Hill [1998], for example, argues that SNA93 should be generalized to recognize and allow for the fact that banks can provide loan services not only when they intermediate deposits and purchased funds into loans but also when they lend their own (i.e. equity) funds.

35. Sherwood [1997] provides an excellent review and treatment of the conceptual issues related to defining and measuring output in the insurance industry.

36. See Sherwood [1997], p.5.

37. A variation of this prescription is found in the output definition used by the United Nations system of national accounts (SNA). Unlike the U.S. national accounts, the SNA defines the output of the insurance industry to include not only net premiums but also the return obtained from investing the gross premiums. For further discussion, see Sherwood [1997, p. 10] and the references cited there.