Measuring the Transaction Sector in the Polish Economy, 1996 – 2002 Aleksander Sulejewicz * Warsaw School of Economics Phone (48-22) 337 93 76 Phone/Fax: (48-22) 64 66 128 [email protected] Patrycja Graca ** Warsaw School of Economics Phone (48-22) 337 93 76 Phone/Fax: (48-22) 64 66 128 [email protected] Paper presented at The 9th Annual Conference of International Society for New Institutional Economics Barcelona, 22-25 September, 2005 * Pronounced [Sooleyevitch] * * Pronounced [Gra:tsah] 1 Abstract Empirical validation of the transaction cost concept proved to be a major intellectual endeavour that has yielded only partial success. Particular difficulties have been encountered in the measurement of such costs at the micro or macro level. The article of Wallis and North (1988) is one attempt to provide a measure of transaction costs in the national economy. Their attempt is to define “transaction sectors” and relate the levels of output (i.e. costs incurred) in such sectors to the level of gross national / domestic product. Among these costs one finds: a) costs of management, sales, administration and control, b) costs of financing, insurance, distribution, c) (some of the costs) of the public sector / the State. Apart from the original research concerning the US, there have been very few studies describing other economies (e.g. Australia, Argentina). The paper joins the discussion on macroeconomic interpretation of transaction costs started by Wallis and North. While we had hoped to trace the evolution of the transaction sectors as well as the pattern of transaction activities in non- transaction sectors as defined above, the availability of data prevented us from accomplishing ambitious research tasks. This paper is basically a replication of the study Wallis and North did for the US and B. Dollery and Leong did for Australia albeit for a much shorter time span (seven years). It contains a short description of the methodology used by these authors, the application of the method to the data on the Polish economy in the 1990s and 2000s. We compare the findings with Wallis and North, Dollery and Leong and provide some preliminary interpretation of the results. Basically, our findings are remarkably close to the estimates of both these teams. The paper ends with remarks on the merits of this approach within neo-institutional economics with some implications for the understanding of economic growth. 2 1. Introduction The concept of transaction costs As we know, the dominant paradigm in economic thinking, i.e. neoclassical economic theory1 privileges the market and the use of pricing mechanisms as a form of economic activity. The theory contends that, under most circumstances, free market exchanges lead to optimum results in allocation of scarce resources. It was the fundamental insight of Ronald Coase to note that having adopted such a view, one has difficulty in explaining other forms of economic organization, notably the firm. Why should firms exist (where market forces are suspended), if it is the market that solves the economic problem?2 His answer was simple but rather unusual: Market is not a free good, or more precisely, the use of the price system is costly. He called the costs of using the price system ‘transaction costs’ (TCs).3 ‘In order to carry out a market transaction it is necessary to discover who it is that one wishes to deal with, to inform people that one wishes to deal with and to what terms, to conduct negotiations leading up to a bargain, to draw up the contract, to undertake the inspection needed to make sure that the terms of the contract are being observed, and so on’.4 The costs of using a market are not merely the costs of information gathering, assessment and use (a lot of that is contained in prices) but also safeguarding contractual rights (and hence writing, protecting, executing contracts). Thus critical economists have identified a) search and information costs (finding buyers and sellers, communicating with them through a stock exchange, fair, or advertising, etc.), b) bargaining and decision costs (when writing a contract it is crucial to decide and accept – and pay for - the degree of detail of the agreement), c) supervision and enforcement costs (we need to monitor the behaviour of our contractors – delivery timing, amount and quality of the good/service etc.) If buyers or sellers have an incentive to behave opportunistically, they might cheat at every stage of the process, they may hide information, pretend to be somebody else (for instance a trustworthy company), renege on the contract if they can go away unpunished or simply steal the money (e.g. by not replacing a faulty product). Discovering what they are like in reality is no easy matter – and is obviously costly. Thus ‘the costs of running the economic system’ as described by K. Arrow are a separate and nontrivial category of economic life. One should not think that only ’outside’ market transactions are costly in this sense. Internal costs of management / organization may also be considerable: a) the costs of setting up, maintaining and changing an organization (design), b) the costs of running an organization (information, costs of decision making, monitoring the execution of tasks / orders, measuring performance of employees. 1 Just for the uninitiated we provide a short list of names. Most of (quasi) Nobel prize winners in economics are in this current of thought: Paul Samuelson, Milton Friedman, Kenneth Arrow, Gary Becker, Gerard Debreu, George Stigler, Robert Mundell, etc. The neoclassical theory is above a theory of price. 2 The economic problem is precisely “the allocation of scarce resources that have competing uses” as introduced into textbook economics by Lionel Robbins in the 1930s. 3 R. Coase, The Nature of the Firm, Economica nr 4, 1937, pp. 386-405, 4 R. Coase, The problem of social cost, Journal of Law and Economics, nr 3, 1960, p.15 [1-44] 3 It is argued that transactions are responsible for most overhead costs, so that management is a key function contributing to their level but also allowing for their reduction.5 Let us be clear: cost is not something an economist is happy about, economizing on (the sum of) production costs and transaction costs is the objective function of any organization, especially, of course, of the business firm.6 Understanding that execution of something that is assumed totally unproblematic in neoclassical economics provides the cornerstone to the institutional analysis of the economic activity. If transaction costs are zero (as assumed by neoclassicals) then there is no rational basis for institutional choice. It does not matter. On the other hand, non-zero transaction cost world (as much closer to reality) makes the economic understanding of institutional setups and their change possible. Consulting any standard textbook shows that in ‘pure economics’ markets and managerial transactions are implicitly assumed to take place against a well defined and well behaved political background. ‘Institutional arrangements consistent with a capitalist market order hold, and this means that a particular local, national, or international organization of the political economy exists. Of course, the provision of such an organization and the public goods associated with it also involve costs. These are political transaction costs. These are, in a general sense the costs of supplying public goods by collective action…’7 These can be the costs of a) setting up, maintaining and changing a system’s formal and informal political organization (establishment of the legal framework, the administrative structure, the judiciary, educational system, etc.); b) running a polity (to build a monopoly of organized violence [M. Levi], defense, legislation, administration of justice, etc.) Just as before, information, decision making, monitoring, enforcement of compliance are all costly. All the previously mentioned types of costs can most probably be divided into ‘variable’ transaction costs, in the sense that they depend on the volume or number of certain transactions and into ‘fixed’ ones. The latter would be the set-up (or change) costs of a new basic organization. From the point of view of received theory, these are the costs that have to be added to production and transportation costs, the costs that are normally recognized in accounting systems of, say, a firm. Unfortunately, such an approach which amounts to complementing standard private and public accounting framework (and the theoretical model of neoclassical economics) with positive transaction costs is deceptively simple (see below). Measuring transaction costs ‘Empirical studies of transaction costs almost never measure these costs directly’.8 Nevertheless, if the transaction cost as a heuristic device is to settle fully into mainstream economics, we need to provide it with precise empirical content and a definition allowing measurement. The predominant application of the concept and the need for measurement arises in the context of comparative governance of business dealings. For instance, is a joint venture better an arrangement than sales agency, or franchise, or fully owned subsidiary? At the sectoral level, there have been several, not very formal, approaches to the practical measurement of transaction costs in the – so far only national – economy. 5 J. G. Miller, T.E. Vollman, The Hidden Factory, Harvard Business Review, 55, no. 5. 1985, pp.142-150. 6 O.E. Williamson, Economic Institutions of Capitalism, The Free Press, 1985. 7 E. G. Furubotn, R. Richter, Institutions and Economic Theory, The University of Michigan Press, 1998, p.47. 8 O.E. Williamson, op. cit. p.35 [Polish edition, PWN, Warszawa 1998] 4 Let us take market TCs. For instance, prices of similar or identical goods are not usually the same over adjacent or even in the same market. Although consumers know that price differences are significant for certain products, many avoid expending time, effort, and money to find the lowest-price seller. One can reasonably hypothesize that the relative price differences (as reported to the average price) may be interpreted as measures of search cost avoided by consumers, and therefore, as measures of their own transaction costs. What they do not pay as costs they pay as price. Thus costs of marketing would qualify as transaction costs. There are standard microeconomic estimates of the difference between production costs and the price paid for the commodity by the final consumer.9 When average transport costs and average indirect tax are excluded, the average marketing cost may be around 40-50% of the final price. Tentative measures of internal managerial transaction costs yield similar magnitudes. If we accept, as suggested above, that overhead costs are an approximation of TCs, their share in total costs (or value added) has, in the US, reached levels between 35% and 60%.10 Since these general administrative costs include production costs as well (e.g. utilities, depreciation, repair) a very crude (split) estimate of ½ would yield a share of circa 20% transaction costs in total costs of a firm. If profit margins are included and ‘outside’ market transaction costs added the sum of market and organizational TCs may reach 40-60% of the final price paid by consumers. We have indicated above that some of these costs are recurrent, i.e. variable. What is clear is that, especially, in periods of radical social and economic change, that is in periods when new institutions are being built, there must be substantial investment outlays, the costs of setting up the new system must be borne. Market transformation in Eastern Europe is a paramount example of such systemic change. Is there a way to measure these transaction costs? Other challenges are in store as well. In practice, it is not easy to distinguish between production costs and transaction costs. An example (provided by Shepard): if production is lost due to delays in planning, is it the result of slow planning [TCs] or of a technology that cannot adapt quickly to late changes in the plan [PCs]? Double sourcing (maintaining supplies from two plants) rather than one may be seen as increasing production costs (because a firm is renouncing economies of scale) or increasing transaction costs (securing supplies given uncertainty and/or opportunistic behaviour on the part of a would-be monopolistic supplier). On the other hand, choice of that particular structure may be an ex-post indication that other forms of organizing supply are even more expensive (in terms of transaction costs). Thus a neat separation of production costs dependent on technology and transaction costs dependent on behaviour is not (always) tenable.11 Measurement of macroeconomic transaction costs Given the importance of both theoretical concept of transaction costs and of the magnitude of empirically defined costs crudely guesstimated above, the need to reliably measure the transaction costs has been recognized. Douglass North and John Joseph Wallis did the pioneer attempt at macroeconomic measurement of TCs for the US economy12. Their precise 9 F.M. Scherer, The selling costs, in: J. Eatwell, M. Milgate, P. Newman, eds. The New Palgrave: A Dictionary of Economics, vol. 4, p. 300-301, London, Macmillan 1987. 10 E. G. Furubotn, R. Richter, op. cit. p.51. 11 ‘The transaction costs and transformation costs of buying (and selling) a house are, at the appropriate margins, substitutes for one another and therefore can be treated the same theoretically’ J.J. Wallis, D. C. North, 1988, Measuring the transaction sector in the American Economy, in: S.L. Engerman, R.E. Gallman, Long-term Factors in American Economic Growth, Studies in Income and Wealth no. 51, Chicago and London, University of Chicago Press. p. 99 12 J.J. Wallis, D. C. North, 1988, Measuring the transaction sector in the American Economy, in: S.L. Engerman, R.E. Gallman, Long-term Factors in American Economic Growth, Studies in Income and Wealth no. 51, Chicago and London, University of Chicago Press. pp. 95-161 5 methodology of measurement is briefly described in the section below. Here we only indicate the main aspect of their thinking. Their ‘basic approach is to segregate economic activities and actors into those that are primarily associated with making market exchanges and those that are not. The sum of resources used by those associated with transacting make up our estimate of the transaction sector’. (p.97) ‘Transaction function’ is an equivalent of the ‘transformation function’ as costs are only incurred in either area when the expected benefits from doing do exceed the costs of doing so. Inputs into a transaction function are no different from ordinary inputs (land, labour, capital, entrepreneurship). ‘When we speak of transaction costs we mean the economic value of the inputs used in performing the transaction function… Transaction costs include the value of labor, land, capital and entrepreneurial skill used in making exchanges. We measure the size of the transaction sector by determining which labor, land and capital costs should be included in the transaction sector’. (p.97) Since part of the transaction costs are not observable (e.g. time spent looking for appropriate houses to buy or waiting for buyers to come by) Wallis and North introduce the measure of the ‘transaction services’ not full transaction costs: transaction services are the observable part of transaction costs (e.g. services of lawyers and realtors; p.99). Using (their) analogy with national income accounts, they only try to capture that part of transaction costs that flows through the market. Transaction costs inside firms are also identified by the function that particular employees (i.e. these are labour costs) perform. They ‘regard the firm as a bundle of contracts’ whereby a hierarchy of contracts involves owners, managers, foremen, workers. At the top of the sequence (hierarchy)13 transaction costs involve processing and conveying information, at lower levels conveying information is complemented with monitoring the labour contract. In the simple stylized example they provide, Bill Gates (they use Henry Ford) first, purchases the firm’s output and the producers (sellers) are the people actually making the [products]; and secondly, he purchases the transaction services of the intermediate occupations in order to coordinate, enact, and monitor the exchange he makes with those who provide transformation services. (p.100) They are aware that ‘making detailed decisions on who does and who does not perform transaction functions in a given firm or industry is impossible short of an intimate and exhausting study of the process of transforming inputs into outputs in each industry. We have chosen a compromise method to get at transaction services within firms.’ (p.100) They single out occupations that provide primarily transaction services to the firm and the residual (‘by elimination’) are transformation services. ‘The wages of employees in these “transaction occupations” constitute our measure of the transaction sector within firms’. (p.100) Wallis and North consider also a third category: intermediaries. Here, all of the resources, that is total value of the inputs used by the intermediaries, are included in the transaction sector and provide the measure of transaction costs. The problem in this case is to decide which firms/industries are properly classified as “transaction industries” (intermediaries). The authors include real estate and finance, banking and insurance, the legal profession, wholesale and retail trade. With some hesitation they add ‘protective services’ (police, guards, etc.) (unfortunately) necessary for enforcing one’s property rights. All in all, according to Wallis and North, increases in transaction costs reduce net social welfare. (p.103) And yet, “none of our transaction services are unproductive. They all represent the resource costs of making exchanges which, on net, make the parties to those exchanges better off (even when transaction costs are included). (p. 103-104) 13 Wallis and North mix temporal and spatial (vertical) metaphors in describing the ‘bundle’. 6 2. Methodology of measurement of transaction costs in Poland The present paper tries to apply the theoretical framework described above to the Polish economy in transition. Our estimate of the transaction costs in the Poland covers the period of 1996-2002. The length of the time span covered is clearly unsatisfactory for an enterprise of empirical analysis of transaction costs in the long term. However, we have been severely constrained by the availability of reliable and detailed data as compiled by the Polish Statistical Office. Nevertheless, we tried to follow the approach of Wallis and North (1986) and Dollery and Leong (1998)14 as closely as possible precisely for the reason of comparability of results, rather than for a more substantive investigation of the Polish economy as such.15 We are envisaging a calculation of transaction costs for a longer period with some modifications of the methodology but this is not a subject of the present paper. To replicate the study, our estimate of the magnitude of transaction costs is based on the one used in the study of Dollery and Leong. Following Wallis and North, they drew a basic distinction between the private and public sectors of the economy. These were in turn subdivided into “non-transaction” industries and “transaction” industries (see: Table 1). Table 1: Transaction and non transaction industries and services Private sector Public sector Production subsector Transaction subsector Production subsector Transaction subsector Agriculture Financial intermediaries Education Public administration Construction (w/o insurance) Health Public order Mining Insurance Rail/Air transport Defense Manufacturing Real estate Public utilities Postal services Transport / Storage Wholesale trade Social welfare Services Retail trade Communications Source: Dollery, Leong (1998), table 1, p. 209 The Polish Central Statistical Office (Główny Urząd Statystyczny), however, defines the respective categories differently. It uses the Polish Classification of Activity PKD16 (Polska Klasyfikacja Działalności) which is compatible with NACE (the statistical classification of economic activities in the European Union). According to PKD economic activity is divided into sections, sub-sections17, divisions, groups, classes and sub-classes. For instance, we have section ‘manufacturing’, sub-section ‘manufacturing of food products, beverages and tobacco’, group ‘processing and preserving fruit and vegetables’, class ‘processing and preserving fruit and vegetables not elsewhere classified’, sub-class ‘services related to processing and preserving fruit and vegetables’. Given the fact that not all data collected and published by GUS are of equal degree of detail, the most detailed parts of economic activities that are the basis of our analysis are at the level of the division.18 The PKD classification as detailed into sections and some divisions is as follows19: 14 B. E. Dollery, Wai Ho Leong, 1998.Measuring the transaction sector in the Australian Economy, 1911-1991 in: Australian Economic History Review, vol. 38, no. 3, pp.207-231. 15 Wallis and North cover the period of one century (1870 -1970), Dollery and Leong cover eighty years (1911- 1991). 16 Polska Klasyfikacja Działalności PKD, Dz. U. Nr 33, appendix to the ordering [rozporządzenie] of the Council of Ministers of 20.01.2004. 17 This concerns some sections only. 18 Even here lack of data is manifest. One can even say that rarely if ever detailed data is available at a level lower than section and division. You will find comments below when specific information is discussed. 19 This is the classification which has been the basis for the calculation of transaction costs in this paper. In reality, the classification is more encompassing. 7 A. Agriculture, hunting and forestry B. Fishing C. Mining D. Manufacturing E. Electricity, gas and water supply F. Construction G. Wholesale and retail trade, repair of motor vehicles, motorcycles, personal and households goods H. Hotels and restaurants I. Transport, storage and communication 60. Land transport; transport via pipelines 61. Water transport 62. Air transport 63. Supporting and auxiliary transport activities; activities of travel agencies 64. Post and telecommunications J. Financial intermediaries K. Real estate, renting and business services, real estate activities L. Public administration and defense; compulsory social security M. Education N. Health and social work O. Other community, social and personal service activities 90. Sewage and refuse disposal, sanitation and similar activities 91. Activities of membership organizations n.e.c. 92. Recreational, cultural and sporting activities 93. Other service activities20 Each of the PKD sections is further subdivided into private and public sectors. According to Dollery and Leong, groups of activities are subdivided into private and public. This is not compatible with the definition employed by GUS. A GUS division of sections into market and non-market would be a better approximation of Dollery and Leong’s approach but even in this case the distinction would not be entirely adequate.21 So for the purpose of this study we have attempted to find equivalents of the sectors as defined by Dollery and Leong (Table 1) in the form of sections and divisions of PKD classification. Transaction costs have been estimated for each of the subsectors of the private and public sectors. Time series of comparable data are only available after 1992. For the private sector transaction activities we have used data pertaining to the global output in respective sections of the PKD classification. (see: Table 2) Global output in this case is an estimate of resources used in the transaction activities in the private sector, that is, of its transaction costs. Next, we have calculated global output in each of the sections as the share (calculated a ratio – see the discussion below) in GDP in current prices and have added these values together to obtain the share of transaction costs in the private sector in transaction activities in the GDP for each year. 20 For technical reasons, sections P and Q have not been included: households employing outside persons, organizations, ex-territorial teams. 21 For instance, among market sections one would include “transport, storage and communication” section which contains “post and telecommunications” division. In Dollery and Leong, this is classified as a part of transaction activities of the public sector. 8 Table 2: Global output in millions of zlotys in the private sector – transactional activities. Dollery- Polish 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Leong GUS equi- classi- valents fication Finance Financial Insurance interme- diaries 4009,4 3327,5 4523,9 5773,9 10012,8 16965,1 20690,7 31990,8 40392,5 37353,4 37119,6 Real estate 14542,9 19694,8 25481,2 26126,2 39859,9 51904,7 72705,8 99182,2 115163,4 137817 156913,8 14542,9 Wholesale Trade and trade repairs Retail trade 31707,5 49803,6 64158,5 63236 94891,1 125576,9 154890,4 179099,7 207165,9 238289,9 254582,8 31707,5 * In 1992-1994 (the first column) at producer prices, in the second column 1994 at base prices (w cenach bazowych) as in the following years. The differences due to the change in data presentation are not significant for the transaction cost analysis. The difference between values expressed in producer prices and those expressed in base prices for any particular year does not exceed 2.5%. Source: Rocznik Statystyczny [Statistical Yearbook], GUS, Warszawa, vols. 1995-2003. The data on global output published by GUS are subdivided into sections only and not into divisions. For this reason, we cannot make the distinction between one part of “insurance” and the remaining part of “financial intermediaries” nor can we analyze separately wholesale and retail trade. This makes a very detailed analysis awkward, but does not influence the transaction costs as such. Table 3: Gross Domestic Product at current prices in millions of zlotys Year 1995 1996 1997 1998 1999 2000 2001 2002 2003 GDP 329567,1 414424,7 504133 589361,3 652517,1 723886,3 760595,3 781112,4 814698,1 Source: Rachunki kwartalne PKB w latach 1995-2003 [Quarterly Accounts, GDP in 1995-2003], GUS, Warszawa, October 2004; Rocznik Statystyczny [Statistical Yearbook], GUS, Warszawa, vols. 1993-1995. GDP data for 1995-2003 is calculated according to ESA’95. Rachunki kwartalne PKB w latach 1995-2003 [Quarterly Accounts of GDP in 1995-2003] is the source for the modifications that had been carried out in the national accounts statistics, in annual and quarterly data for the period 1995-2003. As a result the GDP data contain “time series that are comparable to the ESA principles of 1995 which complement the recommendations of the European Commission”.22 Transaction costs in the transaction subsector of the public sector have been estimated on the basis of data describing State expenditure. According to Dollery and Leong, the value of state expenditure on transaction activities defines the volume of transaction costs in this subsector. Individual categories of state expenditure of transactional nature have been expressed as a percentage of GDP in current prices. Having added all the values of any particular year we have obtained the level of annual transaction costs in the public sector, in its transaction subsector. Table 4: Public expenditure in millions of zlotys in the public sector - transaction activities Dollery’s Polish GUS categories equivalents 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 State administration, public 704,4 1022,2 1601 2310,7 3115,8 3950 4927,1 5851,5 5163,6 5484 6150 6601,1 Public Justice and administration attorneys 339,9 627,6 930,4 1333,8 1713,1 2345,5 2790,2 3085,4 3188,5 3481,8 4405 5095,3 Public security (and fire Public order protection) 849,3 1434,5 2043,8 2710,6 3380,4 4318,3 4901,8 5575,7 2635,5 2729,3 7757,7 7943,2 National Defense defense 1807,1 2536,5 3309 4127,5 5249,4 6003,3 7275 8358,7 6773,4 7251,6 8875,1 9400,6 Postal services Source: Rocznik Statystyczny [Statistical Yearbook], GUS, Warszawa, 1992-2003. 22 Rachunki kwartalne PKB w latach 1995-2003, [GDP Quarterly Accounts in 1995-2003], GUS, Warszawa, October 2004, p. 14. 9 State expenditure published In ‘Rocznik Statystyczny’ do not include post office. Such an item (“poczta i telekomunikacja”) is classified in PKD at the level of section „transport, gospodarka magazynowa i łączność”. One could establish the value of global output for ‘post and telecommunications’ but, firstly, GUS does not publish data on global output at the level of divisions, and secondly, the division “poczta i telekomunikacja” contains a group “telekomunikacja” which, according to Dollery and Leong, is a part of non-transaction activities of the public sector. We have solved the problem as follows. The way postal services are classified – either as non-transaction or transaction subsector of the public sector does not affect the volume of total transaction costs. The only impact is upon the degree of detail of the analysis. As we do not have state expenditure on postal services we classify this item as non-transaction activity. We should remember that transaction costs in the public sector – transaction subsector – will thus be underestimated. The degree of this underestimation seems small, however. 23 The complexity is greater when it comes to the estimation of transaction costs in production activities in both public and private sector. The method used is the same for both sectors. First, we identify persons whose jobs are of the transactional nature in the production subsector. Data pertaining to the number of persons working in each profession (subdivided into sections)24 compiled by GUS are not available in publications. We enumerate below the professions which are included into the transaction category. These professions are called ‘Type I Professions’ by Wallis and North: a) Managers and owners (Head-masters, managers, directors, brokers, traders, bankers), b) Foremen (inspectors, controllers, post masters, head-masters), c) Traders (agents, low level dealers, sales agents – sea, insurance, real estate – sales persons and other persons related to the sales process, newspaper agents) d) Office employees (accountants, cashiers, secretaries, shorthand typists, business machine operators, telephone operators, shorthand typists, ship transport officers, secretarial employees, other office workers), e) Liberal professions (accountants, lawyers, judges, notaries, recruitment officers), f) Security and defense employees (policemen, security men, watches, public order service, detectives, constables). Then we have expressed individual type I professional groups in production sections / divisions as a share of average employment in these sections/ divisions. The shares have been multiplied by gross wages and salaries in these sections / divisions to give us the value of transaction costs in production activities in public and private sectors. Finally, all values have been expressed as the shares in annual GDP at current prices. 23 Average employment in the division ‘post and telecommunications’ does not, in the period studied, exceed 28% of the section ‘transport, storage, and communications’. In addition, the division ‘post and telecommunications’ includes not only postal services, but also communication. If we assume that post and telecommunications constitute about ½ of the division ‘post and telecommunications’, then average employment in postal services does not exceed 14% in the above period. 24 GUS does not collect data on the number of persons employed in each profession subdivided into divisions. Sections provide the greatest degree of detail available. 10