- Email: [email protected]
Materials information systems for public policy making
Materials information systems for public policy making
Leonard L. Fischman
The
objective
system
of
should
decision
maker,
decision,
with
accurate
information US
at the
time acute,
of the his
at
in
relationship
to such
importantly,
it attempts
ments
that
the such
the
of his and
varying
options.
the
systems
government
systematically
provide
a full, of
looks
information
to
awareness
consequences article
an
be
This
materials used its
by
the
general
a goal.
More
to evaluate
specific
require-
systems
should
meet to satisfy such a goal. The author Future,
is with
1755
Resources
Massachusetts
NW, Washington,
DC 20036,
for the Avenue USA.
This paper is based upon a report prepared by the author for the US National Commission on Supplies and Information Shortages, published in Studies, US Systems Government Printing Office, Washington, 1976.
’ ‘Information’ is employed here in the broadest possible sense - including both quantitative and non-quantitative data and all manner of manipulation of such both repetitive and ad hoc. data, ‘Information systems’, however, refers only to on-going procedures, and covers the complete information process, from initial collection of data by the federal government, directly or from other sources, through tabulation and analysis, to ultimate communication. Infirmities in continued
RESOURCES
POLICY
on p 780
September
Policy support is not the only objective of government information systems, and more often than not is not even its principal purpose. Information collected by the US government on materials supply, use, etc is organized and published mainly as a ‘common good’, regarded as having a utility not readily allocable (except for reproduction costs) to particular users. Even much of the analysis is regarded as primarily a general public service. Many evaluations, however, and most of the detailed records are designed essentially for internal consumption by government policy makers and are never published, or are published only partially, belatedly, or through unsystematic revelation. As the policy analysis utility of federal materials information systems is a joint product, it is difficult to allocate costs to it. What these costs reap in the way of benefits is measured by resulting changes in public policy, and the values of such changes are largely subjective. Given such problems on both sides of the balance, meaningful estimates of the net benefits of materials information systems remain elusive.’ For our purpose, materials may be defined as those products which represent the least differentiated (most homogeneous) kinds of input into industrial processes or final consumption. In this sense, copper metal is more nearly a material than is copper ore or brass mill products, since the one may well yield several metals and the other is sufficiently differentiated that its potential uses are fewer than the uses of unwrought copper. Materials may also be viewed more broadly as a common denominator for purposes of measurement or analysis. Materials information extends to the metal content both of ore and of advanced products. The term also comprehends non-metallic mineral commodities, such as sulphur, asbestos, and calcium carbonate, and vegetable materials, such as cotton, rubber, and timber. It is worth distinguishing materials from the resource commodities from which they are derived (eg ores, crops, logs) and from the natural resources themselves (eg mineral deposits, cropland, and forests). In our usage, foodstuffs are not materials, except for their non-food uses, nor are energy commodities, except for their non-energy uses.
1977
179
Materials information systems forpublic continued
from p 179
any one of these three broad stages will detract significantly from the capacity of the system as a whole to achieve its policy-supportive purpose. * For a fuller account, see Leonard L. Fischman, ‘Materials information systems for federal policy making’, in Information Systems Studies. U S Government Printing Off ice, Washington, 1976, p 31ff. 3Almost from the start, data were collected on the industry affiliation of ‘gainful workers’; at present, through data now collected on a sample basis, it is possible to distinguish, by sex, race, and geographical area, those employed (and unemployed) in forestry, metal mining, non-metallic mineral mining, logging, and some dozen material processing industries. Distribution of incomes earned in each of these industries, by sex, race, and geographical area, are also tabulated, and there are various industryoccupation cross-tabulations. Almost as longstanding a tradition are the so-called ‘economic’ censuses, the core of which is an accounting of industry revenues in terms of materials costs, payrolls, and other ‘value added’. Physical quantities of materials shipped and consumed are also shown. The first Census of Manufactures was in 1809. Its subsequent periodicity varied from ten to two years, and it is now taken at five year intervals. Annual Surveys of Manufactures, conducted on a cut-off sampling basis, now provide interim data. The Census of Minerals Industries dates back to 1860 for metal mining and to 1880 for other non-fuel minerals. Its general form and content are similar to that of the Census of Manufactures, and it has been taken in the same years. The Census of Agriculture dates back to 1840, was taken every ten years until 1920, and then every five years since 1925. “A third census system, dating to the production control effort of World War II, includes current information on materials shipments in physical units, and comprises about 100 monthly, quarterly, in the Current or annual releases industrial Reports series. About 5 000 individual products or commodities are covered in total, but only a fraction of these may be considered as materials, and coverage within the materials area is haphazard. A fourth census system, a by-product of social security (old age pension) administration, provides detailed establishment and employment counts, by industry (including the materials industries) and by small geographical area. A fifth provides data on external commodity trade. Although relatively new to the Bureau of the Census (since 1941 ), these kinds of statistics are among the governments oldest collected by anywhere. Compilation of annual data for continoedonp 181
180
policy
making
History and background* Materials information systems in the US government date back to the early days of the republic. More often than not, such information has been generated as either a component of, or an incidental by-product of, information systems designed for other or broader purposes.
Blmauofthe census One of the earliest systems is that established pursuant to Article 3 of the US Constitution, providing for a census of the USA ‘within three years after the first meeting of the Congress of the United States, and within every subsequent term of ten years . ..’ This decennial population census, initially carried out by the ‘Census Office’, has developed from a mere count to a complex survey of demographic, social, and economic characteristics.3 A distinguishing characteristic of the economic censuses - one that limits their usefulness for materials analysis - is their adherence to an ‘industrial’ classification. An industry, by definition, is the aggregate designation for a group of ‘establishments’, each of which is allocable to the particular industry on the basis of its primary revenue source. Though the output of virtually the whole range of raw materials is covered in various censuses (forest products are an exception), the reported inputs of materials (together with labour, capital, energy and water) are those which support the total output of each industry, whatever the particular product mix might be. Technical input-output relationships are consequently obscured.4
Geolospiarl &.m~ey and BWWU of
Mines
Information systems concerned principally with minerals have existed for about a century, the US Geological Survey having been established in 1879. Its first director saw his responsibility for natural resource inventories as confined to public lands. By 1882, however, the US Congress had made specific provision ‘for geologic surveys in various portions of the United States’, and the Survey then assumed a nationwide responsibility for information on mineral reserves and resources, which has since been extended worldwide. A systematic series on mining statistics (published as Mined Resoumes) was started at about the same time, but it was not until after spin-off of the Bureau of Mines in 1910 that the latter became seriously involved with mineral commodity statistics. (Its initial preoccupation was with mining technology and safety). The Bureau has since become the leading purveyor of US mineral commodity statistics, and a leading purveyor of world mineral data too. For non-fuel minerals alone, it regularly canvasses tens of thousands of respondents - many of them monthly and quarterly - with some 200 different report forms. In addition to publishing the annual Mirxeds Yecmfwok, it distributes both statistical and narrative information to the public through the quinquennial Mined Facts andPrvblems, and a changing variety of other releases and publication series of both short and long standing. Most of the Bureau of Mines output relates to physical quantities of commodities.5 other agjwlch The foregoing only hints at the multiplicity of federal agencies which, by reason of history and/or related functions, contribute to total
RESOURCES
POLICY
September
1977
Materials
continued
from p 180
the USA by the Treasury Department began in 1790 and monthly compilation started in 1866. In 1903, responsibility for trade statistics was transferred to the newly created Department of Commerce and Labor, and was subsequently moved around within that department. (The Census Bureau is part of the Department of Commerce; Labour has been split into separate department.) Commodity a detail in US trade statistics differs for imports and exports, although in recent years a certain degree of comparability has been introduced by converting both export and import data to a standard industrial classification (SIC) basis, and for international comparison to a standard international trade classification (SITC) basis. 5 During much of the period since the Geological Survey spawned the Bureau of Mines, parent and progeny have been at odds over the locus of responsibility for mineral reserve and resource estimates. More than one formal agreement has been reached, basically charging the Survey with the origination of information on ‘resources’ (total mineral occurrence and its characteristics) and the Bureau information on with origination of (identified, economically ‘reserves’ recoverable quantities of the valuable material). but there has been some mutual trespassing. Both agencies have in recent years instituted computerized record-keeping and analytical systems, based essentially (though the Survey system is somewhat more flexible) on records for each individual mineral deposit. There is an on-going effort to link the two systems. Neither system, nor any other work of either organization, so far provides anything close to a meaningful estimate of the mineral materials which might be economically recoverable from deposits outside known mining areas (‘prospective reserves’). 6 For a fuller account, see Fischman, op cit. Ref 2, pp 19ff and 175ff. 7There have been a number of fairly comprehensive other than attempts, Fischman (op cit. including Ref 21, particularly the study by the US Congress Office of Technology Assessment, An Assessment of Information Systems Capabilities Required to Support US Materials Policy Decisions, U S Government Printing Office, Washington, 1976. The report of the National Commission on Supplies and Shortages, Government and the Nation’s Resources, US Government Printing Office, Washington, 1976, devotes a chapter to ‘Improving data collection and analysis’, as does a Report to the US Congress by the Comptroller General (‘Programs for gathering commodity information and making forecasts’, chapter 5, US Actions Needed to Cope with Commodity Shortages, General Accounting Office, 29 April 1974).
RESOURCES
POLICY
September
information
systemsforpoblicpolicy
making
materials data. The International Trade Commission, for example, compiles detailed output data on organic chemicals, because its predecessor, the Tariff Commission, was charged with overseeing the protection of an infant coal-tar derivatives industry in the wake of World War I. The connection between prices and real wages is the reason the Bureau of Labor Statistics is the principal producer of wholesale as well as retail price series, including the series on material commodities. Across-the-board data on materials like cotton, wool, and hides are in the domain of the Agriculture Department, because they are agricultural commodities, but not so rubber, since there is no domestic production (even though coffee, a ‘food’, remains an agricultural item). The Bureau. of Domestic Commerce, in the Department of Commerce, maintains and publishes information on a variety of materials industries, but industrial production indexes fall to the Federal Reserve Board, and textile commodities, since the onset of international ‘arrangements’, have been confined to a separate Commerce entity. Although, as a service to the public, some government agencies and private organizations attempt to assemble in their publications the different data from various official and non-official sources, it is the exception, rather than the rule, to find in one place a reasonably complete, current, continuing compilation of all the data on any given material. This is true not only for information released in print, but for that which the federal government maintains for its own policy analysis purposes. Such assembly, and its location within the government, for the most part responds to the emergence of intragovernmental politics, and the availability of problems, appropriate personnel.
Substantive
requirements6
Since the various materials information systems of the US government have been established at different times, for different reasons, both to meet policy analysis purposes and to provide a general public good, they do not conform to an overall systematic plan, nor can their policy utility be appraised according to some clearcut set of substantive standards, for the following reasons:’ 0 0 0
there are several crosscutting dimensions to a policy-supporting materials information system; there are pros and cons to almost any system design; in the final analysis, the worth of any system rests much less m the excellence of its design than in the excellence of the personnel that operate it.
What is appropriate for any policy-supportive system also depends heavily on the style, comprehension, and background knowledge of the policy makers and analysts who use it; this is hardly a fixed parameter. Some mode of appraisal of the information systems is desirable whether or not it is clearcut. We consider four different viewpoints from which an appraisal may be made and some desiderata identified.
Relationship
to policy decisions
A principal barrier to either designing organizational structure for materials
1977
or appraising a particular information is overlap of
181
Materials in@matbn
systems
for public policy making
a One reason for this is that the adequacy of other available information is rarely clear-cut. When, for example, the US Environmental Protection Agency needed to consider the potential impact of specific effluent limitations on the paper and other industries, there was no apparent thought that the work could be within accomplished elsewhere the federal establishment, using existing information banks and analytical models. EPA was probably correct in assuming that adequate impact models did not exist, especially since, for its purposes, ‘adequate’ meant being good enough not only to satisfy the agency’s decision makers, but to meet the challenge of every affected interest that could contest its decisions. The agency’s reaction was first to contract out for an emergency type of analysis and, once the initial pressure was reduced, to contract further for a definitive series of more larger, ‘management by objective’ (MB01 studies. The econometric and other models that are part of these studies are intended to be the definitive way of evaluating regulatory impact on various industries. This course of action might be acceptable, save for the fact that the locus of industry expertise in the US if anywhere, is in the government, Department of Commerce, that no single approach or evaluation is about to be accepted throughout the government as definitive, and that EPA’s MB0 analyses have yet to be satisfactorily linked with some approach to generalized economic impact.
182
materials information with other kinds of information. We may define ‘materials information’ as including impacts of the materials economy on the rest of the economy. One would not want to decide any policy question affecting materials without assessing such indirect impacts, but by the same token, the kinds of information affecting materials must also be considered as being environmental information, foreign affairs information, labour information, consumer information, etc. If this breadth of expertise could conceivably be embodied in individuals or small units, should it then be duplicated over and over within the government to serve as many kinds of policy makers? There is certainly virtue in the specialized marketing of a particular product in this case an information product - but there is also virtue in developing specialized skills in its production. Since, even for more tangible goods, it has never been quite settled whether there is superior merit in ‘knowing the territory’ or ‘knowing the product’, there is no obvious answer for the organization of government information systems. Whether through established coordinating procedure or through the alertness of personnel, the desiderata which really have to be met are that neither the direct nor the indirect impact on or through materials of any policy bearing on materials, nor any intragovernmental source for such information, is inadvertently overlooked. In other words, whether the relevant data and analysis are adduced by materials specialists or by any other kind of specialists is inconsequential, so long as, if significant, they are adduced. Any duplication of effort should be deliberate - either to ensure the consideration of possibly conflicting data, analyses, or biases, or to avoid place or time disutility. Timing is critical. At the point that a particular policy decision is under consideration, there is the choice of using various combinations of already gathered facts, already completed analysis, newly gathered facts, and new analysis. The choice will rationally depend upon the adequacy of what has gone before and the convenience of assimilating it. A critical information-system challenge is to have anticipated the extent to which timeliness of response and economy of effort, as well as the conservation of fugitive information, dictate advance information-system investment. More often than not, in the federal government, the answer to such questions has been improvization, duplication, and over-investment. It is physically and psychologically easier to meet each new information or analysis requirement afresh than to retrieve - let alone ascertain the present worth of - any previous or concurrent research or information activity. Once having been faced with an information ‘gap’, an agency tends to be restrained only by budget from setting up a system with which it will permanently be prepared for repetition of the need.s Given the current state of the evaluative art, it is a reasonable conclusion that no one agency should be investing any more than is immediately necessary in methodological innovations until it or some other (possibly more neutral) agency has more thoroughly explored on a prototype basis the methodological alternatives and their costs and benefits.
The time dimension One of the problems of deciding on a particular form of analytical modelling is that different kinds of model are appropriate to different
RESOURCES
POLICY
September
1977
Materials information systemsforpublicpolicy
making
forward-looking time spans. The most complex impact modelling is suitable for short-term questions, measured in months rather than years. Though dynamic models have been designed which proceed by fractions of a year to as far ahead as the computer budget will allow, it is doubtful that their output is useful for more than a few years. As the time horizon lengthens, structurally simpler models are usually more appropriate, and the type of basic data required also changes.
requirements. One reason for the establishment of the National Commission on Supplies and Shortages was congressional frustration with the commodity shortages of 1973-74, and the feeling that if there had been an adequate early warning system ameliorative action could have been taken. The usual mental concept of such a system involves models and automated data banks. The Department of Commerce Materials Division, which has been trying to establish an early warning system, is bringing into use a number of computerized systems and models, though it still relies fundamentally on the knowledge and alertness of its individual industry specialists9 The inconclusiveness of present analytical systems with regard to indirect consequences has already been noted. Determining the proximate consequences of a government decision is not much less of a challenge. Were this not so, there would not be a buyer for every seller on the commodity exchanges or a variety of public and private forecasting systems and prophets. Whatever the methodology, no model or analyst can do better than project probable outcomes or specify a high-probability range of outcomes. Unfortunately for both models and modellers, ‘validation’ by past successes or retrospective fittings is no guarantee of current accuracy. Potential consequences do have to be forecast somehow, however, and the most dependable information will presumably come from the combination of an individual analyst familiar with his subject and inspired impact modelling. Whatever their design, and whether econometrics, judgment, or engineering data are the source of their parameters, the most useful models will be those which reflect real-life behaviour rather than economic theory or rational optimization. They will be economical as well as useful if they have been stripped of the variables and equations that add little either to understanding of the behavioural mechanism or precision of the results. To predict short-term shortages, materials models need not be complicated and may have only limited utility.‘* Short-range
’ It is doubtful that any elaborate system is actually either needed or useful to provide early warning of imminent materials developments requiring policy policy decisions. The history of 1973 deliberations preceding the imposition of export controls on ferrous scrap shows that there is no lack of early warning of emerging short-range problems by affected interests. (See Fischman, op cit. Ref 2, pp 107ff and 273ff. Another illustration is the export control episode with regard to cattle hides, described on pp 131ff and 345ff.) Executive branch and congressional show decided makers also policy reluctance to take unpalatable action without the prior build-up of enough of a visible crisis to generate a resounding positive demand. A hesitance to act on early warning signals may well have substantive justification. More critical than early warning per se is the ability to distinguish false real alarms. and Furthermore, ~najor information the requirement is related not with alertness, but with understanding the direct and indirect of alternative consequences actions or inaction. lo The reason for this is that the controlling factor in shortages (or in shortage-induced high prices) is not the nature of the particular commodity but the state of US and world business cycles. If the macro-economic predictions which them are incorrect, feed commodity models, no matter how exquisite, will do a poor job of forecasting coming demand and supply. A reasonable understanding of what happened at the last peak, plus some rough knowledge of what has since happened to industry capacity and product flows, will permit reasonable predictions of what will happen at the next peak, without a model.
RESOURCES
POLICY
September
Medium-range requirements. For medium-range purposes (eg 2-10 years), systems and models become more useful to the analyst. Even here, however, industry and commodity specialists have an advantage over straight mechanical projections of being able to use a changing variety of updated information on capacity planned, under construction, and existing, as well as on shifts in supply patterns and consumption habits. An optimum arrangement is the kind of model which can receive and make use of such ‘forerunner’ information inputs. Modelling for this medium range is particularly difficult, as many variables will still be significantly affected by the state of the business cycle, which at this range can only be guessed. Much of the difficulty will disappear, however, if the model or information system is not
1977
183
Mahiak
i$onnation systems for public policy making
asked the wrong questions. Most policy questions looking forward into this period are best served by infdrmation on peaks, troughs, and trend values, and rarely require projections of actual quantities or system for this prices. The alertness function of an information medium range is important, because there will be less forward-looking reportage in the trade press, and it may not be possible, without detailed analysis, to foresee the constraints and problems which may develop under various sets of circumstances.
I1 For the very long term (over 50 years), there is no need for an information system as such. Persons are required who are sufficiently familiar with the analytical in literature are so placed and government that they can keep policy makers within reasonable limits. I2 A particularly difficult question is the handling of microdata. Certain types of policy analysis depend heavily upon detailed data for individual firms and establishments, yet it is not easy to preserve the current norm of maximum confidentiality and still make these data available to the extent required. A further problem is that many of the kinds of data which particular policy questions seem to call for do not exist as such in all plants and enterprises. To require such data therefore either imposes a burden, which it should be specifically estimated is worth its economic cost, or generates pm fom replies whose inaccuracy may put their value in doubt. It is incumbent upon the initiators of government surveys not only to review the p&m f& results of survey pretests, but to look with care into the procedures respondents resort to to provide their replies.
184
Long-range w@wwzts. Most of the shorter range modelling detail and other shorter range considerations clearly become irrelevant when looking 1650 years ahead. Only trend values, or at most the trend values for peaks and troughs, are of real use. At this range, one has to accept that predictions can be reasonably made only within widening bounds of error. Many of the grosser demand-determining relationships are persistent, however, and this is what makes longterm projection plausible. Supply is a much more difficult question. Even up to 50 years, there is little reason to expect materials consumption to be limited by fundamental resource constraints, but effective resource adequacy still depends upon timely inputs of exploration, R&D, capacity investment, and planning. The policy role of materials information systems for this long-range horizon is essentially one of uncovering the areas for vigilance (eg international political forecasting) or which require current governmental action because short-range commercial interest may differ from what is of long-range public advantage. Long-range models and projections may also contribute significantly by persuading against ill-advised action. Long-range supply schedules can be devised, as in the Bureau of Mines data bank (the ‘minerals availability system’), which provide some guidance on the possible extent of prospective supply stringency. However, given the broad uncertainties posed by prospective technological change, by as yet undiscovered reserves, and by poorly predictable rearrangements in world trade, such supply schedules cannot very credibly be applied as feedback into demand-supply equilibrium models. It would be misleading in most cases to pretend that long-range forecasts are based on anything but broad judgments.” Information
system stages
An information system was defined earlier as comprehending the acquisition of quantitative and qualitative data, analysis of data, and communication of findings. Acquisition and storage. It is easy to specify the kinds of data consumption, imports, exports, inventory changes, (production, prices, production capacity, natural resource limitations, production costs, etc) which are useful as decision-making inputs, but this does not help in concluding what kind of detail, at what order of accuracy, ought to be repetitively collected and tabulated. The many inconsistencies of data series over time, among issuing agencies, and among different materials, clearly illustrate the welter of technical choices facing data gatherers and manipulators.” The question of confidentiality extends to storage decisions. If returns from surveys are permanently stored (either in original form or on magnetic tape) complete with individual identification, there are
RESOURCES
POLICY
September
1977
Materials I3 Restraint in survey activity, uniformity of standards, and quality of statistical practice are in theory enforced by the Division of Statistical Standards in the Office of Management and Budget (part of the President’s Executive Office), but owing partly to resource limitations, its effectiveness has been limited. ‘*For materials, the most fundamental analytical requirement is to have reasonably reconciled data on supply and consumption. This has long been recognized by the US Department of Agriculture in its ‘supply and distribution’ (sometimes ‘supply and disappearance’) tables and has become the organizing principle in the Bureau of Mines’ recently inaugurated annual series, Miner& in the US Economy. The amount of estimation that goes into the latter is an indication of the shortfall in basic data that calls for survey consideration of possible improvements. On the other hand, there is a tendency to try to fit data and analysis into unnecessarily constricting patterns. For example, it is at least questionable whether the SIC industry allocation which the Bureau is now attempting for its consumption distribution, and which for the typical mineral commodity involves considerable estimation, is sufficient of an advantage over more traditional consumption categories to warrant the unavoidable inaccuracies and extra effort. ” Much gathering analysis of and materials (and other) data is for routine periodic summaries, ranging in complication from ‘value of mineral production’ tabulations to the National Income Accounts large-scale and input-output models. These comprehensive compilations are almost as demanding in processing unimportant important data, or in providing as sometimes spurious precision. There is their little practical choice as to continuation, since once however, initiated, each has tended to generate its own market. The time for choice and discrimination in such tabulations is in the early stages, before the market has had a chance to develop. I6 Some agencies in the federal government have gained the reputation of being beyond influence by parties whose fortunes may be affected by the data they issue: other materials information agencies (eg the Bureau of Mines and the Bureau of Domestic Commerce) have not. In general, it is the closeness of association between agency personnel and the industries they ‘serve’ that leads to suspicion. The agencies, of course, consider this closeness the very basis of the accuracy and instructiveness of their analyses and and there is data, considerable merit in that view. Such agencies are likely to achieve more trusted status only if their expositions and evaluations carefully distinguish industry views from their own and do not give the impression of uncritical acceptance of
continued on p 186
RESOURCES
information
systems
forpublicpolicy
making
permanent risks, regardless of safeguards, of compromise of confidentiality. If they are stored without identification, checkbacks for error and cross-tabulations with other data sets are impossible. US federal agencies vary considerably in their handling of surveys and microdata, and some (partly at their own insistence) are even governed by separate statutes. Recent rulings and court decisions, however, have led to a greater concern with confidentiality, as an offset to ‘freedom of information’ procedures and magnetic-tape data distributions which have combined to open up the federal government’s internal statistical workings to more general public inspection.13 Selection of methods of aggregating and cross-tabulating data involves many of the same criteria that apply to data collection. The decision on what to collect and how to specify it should depend largely on a prior design of the tables, charts, and other modes in which the data are to be used.14 The role of price is of particular interest in many materials problems. Price series are known to have defects, varying from material to material, as representations of actual transaction prices, yet average-value-of-output series are rarely substituted. Apart from seasonal adjustment and annual averaging, there is little routine manipulation of price data in the federal government, the series being considered to have analytical significance in and of themselves. Many of the resource and materials price series relied upon most for policy analysis are those collected and published privately, rather than the corresponding series included in the Bureau of Labor Statistics wholesale price indexes.15
Analysis.
Communication. There is no optimum formula for communicating materials information. Federal agencies seem to have become increasingly adept at communicating principal interrelationships, and have effectively used graphics to display some of the subtleties. A considerable challenge remains, however, which cannot be met except by constant interaction between information and policy personnel. Clarity in communication is often achieved only at the expense of making complicated things seem simple, which can badly prejudice the resulting decisions. While a reasonable mastery of communications techniques is essential (frequently to the point of interposing a communicator between the information specialist and his decision-making clients), an equally important requisite is adaptability to the varying styles, capabilities, backgrounds, and time commitments of the clientele to be served. A good communication system will provide the generalist with the essence of a situation, without either leading or misleading, and the specialist with all of the details required for complete analytical understanding. Credibility is said to be one of the most serious problems with some sectors of the current materials information system. Communication failures are part of the reason for this lack of confidence. Technicians are usually only frustrated by such problems as discrepancies in figures issued at different times and places; policy makers, or their aides, are likely to lose faith. A more copious accompaniment of . changed or alternative figures by reconciliations or other explanations would improve the image of issuing agencies with both types of users.16
POLICY September
1977
185
Materials
information
systems forpublicpolicy
making
Purposes of systems Materials information systems can be said roughly to serve one of three kinds of purposes - alertness to new circumstances that require policy or decision making attention (‘monitoring’ or ‘early warning’); formulation and periodic revision of standby-resource strategies and contingency plans for dealing with possible adverse developments (such as supply interruptions or cartel actions); and capacity, in terms of information base and continuing ‘overhead’ analysis, to handle important ad hoc evaluations in a timely and economical fashion.
continued from p
185
information from industry sources. While is communication alone good not sufficient to establish a reputation for independence, it is an essential minimum. It must be borne in mind that can flow two ways. communication Ultimate responsibility for the quality of information systems lies not with the but with producers, information information consumers. Policy makers who view information systems merely as devices to provide briefs and backup for views they reach independently will get what they are looking for. Congressmen and executive branch officials who do not respond to information flows with some critical interest will get even less.
186
Early wartzing. A variety of potentially emerging situations is involved, ranging from the near to long term. However, it is doubtful that any elaborate information system is either very much needed or very useful in providing early warning of such problems as impending materials shortages. This does not mean that the trade press and price trends should not be monitored, or that the information should not be communicated to policy makers without waiting for it to come to their attention through complaints or from their more general reading. The principal purpose, however, would be to encourage advance thought about the situation and avoid hasty action. A risk in responding positively to an early warning is that the action may exacerbate the situation it is designed to remedy. This is even more true of precursors of positive action, such as attempts to get more information on the seriousness of a supply or price perturbation. Early warning systems for supply or price problems thus involve a dilemma. If they provide for adding monitoring functions in the face of an impending crisis, they may well precipitate such a crisis; if they introduce monitoring functions on a continuing basis, in a non-crisis atmosphere, they may burden both the government and respondents with unnecessary cost and effort. Another dilemma facing early warning systems is the trade-off between persuasiveness and objectivity. When warnings are confident and clear they are more likely to be acted upon, whereas dispassionate, ‘two-handed’ projections generally receive little attention. Yet the latter type of projection usually reflects more accurately the forecaster’s ability to predict. In judging forecasts, little can be made of a forecaster’s ‘track record’. Retrospective evaluations of early warnings can never be unequivocal, especially since such warnings tend to be either self-fulfilling or self-defeating. Ideally, a warning should be given early enough and persuasively enough to trigger the responses that will prove it wrong. The longer range the pending problem, the easier it is to put off action. The dilemma of choice between dramatic and well balanced particularly to long-range forecasting. applies predictions Fortunately, at least some long-range problems can be attacked with relatively modest, readily acceptable beginnings which, if sustained over long periods, can have significant cumulative effects. An information system that keeps its balance, therefore, can still hope to accomplish something without running the risk of stimulating major national commitments - to action or to inaction - that prove to be major blunders. Contingency pt!cmning. This category of information systems, required to deal with such contingencies as supply interruptions and sharp price rises, has a predictive aspect of seemingly ‘early warning’
RESOURCES
POLICY
September
1977
Materials in@mation
I7 More precisely, what is needed is a schedule of probabilities for different magnitudes. For example, the probability of a partial interruption of US imports of metallurgical manganese is greater than the probability of a complete interruption, and the probability of a short interruption is greater than the probability of a long information system, by The one. assembling all available intelligence and processing it through personal judgment or through gaming or other models, needs to put probability numbers on a matrix of size-duration combinations. are A Technology them ” Among of Economic Stockpile Assessment Policy, Office of Technology Assessment, Critical 1976 and us Congress, Materials: Commodity Action Analyses: Aluminum, Chromium, Platinum, and Palladium, Office of Minerals Policy Development, US Department of the March 1975. A mostly Interior, conference on methodological Contingency planning is being planned for December 1977 under the sponsorship of and MITRE Corporation the the Engineering Foundation. 19The importance of evaluating such been consequences has ancillary recognized by the US Congress in its frequent specification of economic impact analysis as part of a regulatory procedure, and by the Executive in a broadside order calling for ‘Inflation Impact Statements’ (subsequently renamed ‘Economic Impact whenever rules and Statements’) regulations may affect costs or prices. Such analysis calls for an information base which includes not only basic data on materials supply and distribution and on price and production arrangements, models of the macroeconomic but relationship of the materials economy to economy, and rest of the the microeconomic models of the relationship of individual firms and producing units to the materials economy.
RESOURCES
POLICY
September
systems
for public
policy
m&zing
nature, which it may well have to incorporate. What is ultimately needed, however, as a foundation for contingency policy planning is prediction, not that an event is likely to occur, but of the degree of risk of its occurrence, and with what magnitude both in volume and in duration.” Probability of occurrence is by itself inconclusive for action purposes. It is equally important to be able to predict the consequences of each potential size-duration combination. A potentially cataclysmic event merits attention even if it has relatively minute possibility of occurrence; an event of relatively minor consequence may deserve attention if its occurrence is rather certain. In this sense, early warning is a special, prejudged case of contingency analysis; the forecaster has predicted that there is sufficient probability of occurrence of an event of sufficient gravity that action should be considered to avert it. Measures to avert and to compensate are sometimes identical - in kind, if not in magnitude. Stockpiles, for example, will tend to discourage some types of price increase or supply interruption, at the same time as they provide a means of compensating for the price or supply contingency if it occurs. A common element in aversive and compensatory measures is that they both aim at avoiding the damage of a price rise or supply interruption. Most of the calculations connected with contingency analysis are well beyond any current state of the art. The tendency in recent studiesI has been to rely on neoclassical demand-supply analysis, and to calculate net balance between loss of consumer, and increase in producer, surplus. Such analysis is sensitive to rather shaky elasticity parameters, in addition to being insensitive to the practical possibilities for reallocation of resources. Contingency analysis requirements differ little from what is required for the wider objective of being able to predict the consequences of alternative actions or inaction in a variety of circumstances. The areas of decision can vary widely, to include export control policy, tax policy, environmental and health and safety regulation, price and wage controls, transportation rates and arrangements, public land policy, forest management, etc. Decisions in all these fields have an impact on materials, but since the impact is not always direct, data and models should be available somewhere in government to establish the links between the proximate effect of each such policy and the consequent impact on materials and through materials on the rest of the economy.‘g On-line capacity. In deciding how far to go in setting up a continuing information base, the considerations are in many respects similar to such private yardsticks as life-cycle costs and net present value of return on investment. The difficulty in applying the latter concept in government is that the value of better decision making is largely a matter for political determination. The life-cycle cost criterion is a little easier to apply. Federal agencies can estimate the maintenance cost of any information system they introduce and can consider periodically whether the continued patching of an existing system might not exceed the cost of installing and maintaining a new one. They can also take into account that a dollar of expenditure deferred is at least five cents saved. A good rule is that materials information systems should attempt
1977
187
Materials information systems for public policy making to anticipate decision-making requirements, but not too thoroughly and not too far in advance. Time should be allowed for ‘market’ evaluation of systems, despite the problems such evaluation presents. There are economies of scale in information systems (both crosssectionally and over time) and system managers need to have due regard for the differences between overhead and incremental cost. Once the basics of a system have been worked out (investment), the marginal costs of maintaining it or of extending its scope may be small enough to balance even uncertain expectations as to its future use. Again, in conducting field surveys, it is cheaper to obtain additional information at the same time than to come back for more later (though it is also necessary to consider decreasing marginal quality of response). Conservation of information is another criterion for determining on line capacity. Information that is not obtained today may not be available tomorrow, or may not be available with comparable accuracy. Timing may be critical to its utility. Data and analysis which are not ready when most needed provide less than optimal return on expenditure. It is generally better to gather too much too soon than too little too late. Information systems need constant planning, care, and attention. There is little sense in investing heavily in grand designs without providing architectural supervision. Competent, experienced, imaginative personnel are an essential. The best, most respected information systems in the US government, such as those of Census and the Bureau of Labor Statistics, have enough resources to engage in planning and methodological analysis, whereas the most able personnel in the poorer systems tend to be employed on miscellaneous duties or are largely occupied with crises. Excess resources can easily be squandered, however, especially if an information system is allowed to become too much a creative plaything, without constant exposure to a cost-utility standard. It is the policy analysts who have to undertake the principal responsibility for specifying which kinds of information are useful and which are relatively superfluous for policy analysis purposes. They must also remain clear-headed as to what kind of analysis is both illuminating and likely to have ultimate decision-making impact.
188
RESOURCES
POLICY
September
1977
Leonard L. Fischman
The
objective
system
of
should
decision
maker,
decision,
with
accurate
information US
at the
time acute,
of the his
at
in
relationship
to such
importantly,
it attempts
ments
that
the such
the
of his and
varying
options.
the
systems
government
systematically
provide
a full, of
looks
information
to
awareness
consequences article
an
be
This
materials used its
by
the
general
a goal.
More
to evaluate
specific
require-
systems
should
meet to satisfy such a goal. The author Future,
is with
1755
Resources
Massachusetts
NW, Washington,
DC 20036,
for the Avenue USA.
This paper is based upon a report prepared by the author for the US National Commission on Supplies and Information Shortages, published in Studies, US Systems Government Printing Office, Washington, 1976.
’ ‘Information’ is employed here in the broadest possible sense - including both quantitative and non-quantitative data and all manner of manipulation of such both repetitive and ad hoc. data, ‘Information systems’, however, refers only to on-going procedures, and covers the complete information process, from initial collection of data by the federal government, directly or from other sources, through tabulation and analysis, to ultimate communication. Infirmities in continued
RESOURCES
POLICY
on p 780
September
Policy support is not the only objective of government information systems, and more often than not is not even its principal purpose. Information collected by the US government on materials supply, use, etc is organized and published mainly as a ‘common good’, regarded as having a utility not readily allocable (except for reproduction costs) to particular users. Even much of the analysis is regarded as primarily a general public service. Many evaluations, however, and most of the detailed records are designed essentially for internal consumption by government policy makers and are never published, or are published only partially, belatedly, or through unsystematic revelation. As the policy analysis utility of federal materials information systems is a joint product, it is difficult to allocate costs to it. What these costs reap in the way of benefits is measured by resulting changes in public policy, and the values of such changes are largely subjective. Given such problems on both sides of the balance, meaningful estimates of the net benefits of materials information systems remain elusive.’ For our purpose, materials may be defined as those products which represent the least differentiated (most homogeneous) kinds of input into industrial processes or final consumption. In this sense, copper metal is more nearly a material than is copper ore or brass mill products, since the one may well yield several metals and the other is sufficiently differentiated that its potential uses are fewer than the uses of unwrought copper. Materials may also be viewed more broadly as a common denominator for purposes of measurement or analysis. Materials information extends to the metal content both of ore and of advanced products. The term also comprehends non-metallic mineral commodities, such as sulphur, asbestos, and calcium carbonate, and vegetable materials, such as cotton, rubber, and timber. It is worth distinguishing materials from the resource commodities from which they are derived (eg ores, crops, logs) and from the natural resources themselves (eg mineral deposits, cropland, and forests). In our usage, foodstuffs are not materials, except for their non-food uses, nor are energy commodities, except for their non-energy uses.
1977
179
Materials information systems forpublic continued
from p 179
any one of these three broad stages will detract significantly from the capacity of the system as a whole to achieve its policy-supportive purpose. * For a fuller account, see Leonard L. Fischman, ‘Materials information systems for federal policy making’, in Information Systems Studies. U S Government Printing Off ice, Washington, 1976, p 31ff. 3Almost from the start, data were collected on the industry affiliation of ‘gainful workers’; at present, through data now collected on a sample basis, it is possible to distinguish, by sex, race, and geographical area, those employed (and unemployed) in forestry, metal mining, non-metallic mineral mining, logging, and some dozen material processing industries. Distribution of incomes earned in each of these industries, by sex, race, and geographical area, are also tabulated, and there are various industryoccupation cross-tabulations. Almost as longstanding a tradition are the so-called ‘economic’ censuses, the core of which is an accounting of industry revenues in terms of materials costs, payrolls, and other ‘value added’. Physical quantities of materials shipped and consumed are also shown. The first Census of Manufactures was in 1809. Its subsequent periodicity varied from ten to two years, and it is now taken at five year intervals. Annual Surveys of Manufactures, conducted on a cut-off sampling basis, now provide interim data. The Census of Minerals Industries dates back to 1860 for metal mining and to 1880 for other non-fuel minerals. Its general form and content are similar to that of the Census of Manufactures, and it has been taken in the same years. The Census of Agriculture dates back to 1840, was taken every ten years until 1920, and then every five years since 1925. “A third census system, dating to the production control effort of World War II, includes current information on materials shipments in physical units, and comprises about 100 monthly, quarterly, in the Current or annual releases industrial Reports series. About 5 000 individual products or commodities are covered in total, but only a fraction of these may be considered as materials, and coverage within the materials area is haphazard. A fourth census system, a by-product of social security (old age pension) administration, provides detailed establishment and employment counts, by industry (including the materials industries) and by small geographical area. A fifth provides data on external commodity trade. Although relatively new to the Bureau of the Census (since 1941 ), these kinds of statistics are among the governments oldest collected by anywhere. Compilation of annual data for continoedonp 181
180
policy
making
History and background* Materials information systems in the US government date back to the early days of the republic. More often than not, such information has been generated as either a component of, or an incidental by-product of, information systems designed for other or broader purposes.
Blmauofthe census One of the earliest systems is that established pursuant to Article 3 of the US Constitution, providing for a census of the USA ‘within three years after the first meeting of the Congress of the United States, and within every subsequent term of ten years . ..’ This decennial population census, initially carried out by the ‘Census Office’, has developed from a mere count to a complex survey of demographic, social, and economic characteristics.3 A distinguishing characteristic of the economic censuses - one that limits their usefulness for materials analysis - is their adherence to an ‘industrial’ classification. An industry, by definition, is the aggregate designation for a group of ‘establishments’, each of which is allocable to the particular industry on the basis of its primary revenue source. Though the output of virtually the whole range of raw materials is covered in various censuses (forest products are an exception), the reported inputs of materials (together with labour, capital, energy and water) are those which support the total output of each industry, whatever the particular product mix might be. Technical input-output relationships are consequently obscured.4
Geolospiarl &.m~ey and BWWU of
Mines
Information systems concerned principally with minerals have existed for about a century, the US Geological Survey having been established in 1879. Its first director saw his responsibility for natural resource inventories as confined to public lands. By 1882, however, the US Congress had made specific provision ‘for geologic surveys in various portions of the United States’, and the Survey then assumed a nationwide responsibility for information on mineral reserves and resources, which has since been extended worldwide. A systematic series on mining statistics (published as Mined Resoumes) was started at about the same time, but it was not until after spin-off of the Bureau of Mines in 1910 that the latter became seriously involved with mineral commodity statistics. (Its initial preoccupation was with mining technology and safety). The Bureau has since become the leading purveyor of US mineral commodity statistics, and a leading purveyor of world mineral data too. For non-fuel minerals alone, it regularly canvasses tens of thousands of respondents - many of them monthly and quarterly - with some 200 different report forms. In addition to publishing the annual Mirxeds Yecmfwok, it distributes both statistical and narrative information to the public through the quinquennial Mined Facts andPrvblems, and a changing variety of other releases and publication series of both short and long standing. Most of the Bureau of Mines output relates to physical quantities of commodities.5 other agjwlch The foregoing only hints at the multiplicity of federal agencies which, by reason of history and/or related functions, contribute to total
RESOURCES
POLICY
September
1977
Materials
continued
from p 180
the USA by the Treasury Department began in 1790 and monthly compilation started in 1866. In 1903, responsibility for trade statistics was transferred to the newly created Department of Commerce and Labor, and was subsequently moved around within that department. (The Census Bureau is part of the Department of Commerce; Labour has been split into separate department.) Commodity a detail in US trade statistics differs for imports and exports, although in recent years a certain degree of comparability has been introduced by converting both export and import data to a standard industrial classification (SIC) basis, and for international comparison to a standard international trade classification (SITC) basis. 5 During much of the period since the Geological Survey spawned the Bureau of Mines, parent and progeny have been at odds over the locus of responsibility for mineral reserve and resource estimates. More than one formal agreement has been reached, basically charging the Survey with the origination of information on ‘resources’ (total mineral occurrence and its characteristics) and the Bureau information on with origination of (identified, economically ‘reserves’ recoverable quantities of the valuable material). but there has been some mutual trespassing. Both agencies have in recent years instituted computerized record-keeping and analytical systems, based essentially (though the Survey system is somewhat more flexible) on records for each individual mineral deposit. There is an on-going effort to link the two systems. Neither system, nor any other work of either organization, so far provides anything close to a meaningful estimate of the mineral materials which might be economically recoverable from deposits outside known mining areas (‘prospective reserves’). 6 For a fuller account, see Fischman, op cit. Ref 2, pp 19ff and 175ff. 7There have been a number of fairly comprehensive other than attempts, Fischman (op cit. including Ref 21, particularly the study by the US Congress Office of Technology Assessment, An Assessment of Information Systems Capabilities Required to Support US Materials Policy Decisions, U S Government Printing Office, Washington, 1976. The report of the National Commission on Supplies and Shortages, Government and the Nation’s Resources, US Government Printing Office, Washington, 1976, devotes a chapter to ‘Improving data collection and analysis’, as does a Report to the US Congress by the Comptroller General (‘Programs for gathering commodity information and making forecasts’, chapter 5, US Actions Needed to Cope with Commodity Shortages, General Accounting Office, 29 April 1974).
RESOURCES
POLICY
September
information
systemsforpoblicpolicy
making
materials data. The International Trade Commission, for example, compiles detailed output data on organic chemicals, because its predecessor, the Tariff Commission, was charged with overseeing the protection of an infant coal-tar derivatives industry in the wake of World War I. The connection between prices and real wages is the reason the Bureau of Labor Statistics is the principal producer of wholesale as well as retail price series, including the series on material commodities. Across-the-board data on materials like cotton, wool, and hides are in the domain of the Agriculture Department, because they are agricultural commodities, but not so rubber, since there is no domestic production (even though coffee, a ‘food’, remains an agricultural item). The Bureau. of Domestic Commerce, in the Department of Commerce, maintains and publishes information on a variety of materials industries, but industrial production indexes fall to the Federal Reserve Board, and textile commodities, since the onset of international ‘arrangements’, have been confined to a separate Commerce entity. Although, as a service to the public, some government agencies and private organizations attempt to assemble in their publications the different data from various official and non-official sources, it is the exception, rather than the rule, to find in one place a reasonably complete, current, continuing compilation of all the data on any given material. This is true not only for information released in print, but for that which the federal government maintains for its own policy analysis purposes. Such assembly, and its location within the government, for the most part responds to the emergence of intragovernmental politics, and the availability of problems, appropriate personnel.
Substantive
requirements6
Since the various materials information systems of the US government have been established at different times, for different reasons, both to meet policy analysis purposes and to provide a general public good, they do not conform to an overall systematic plan, nor can their policy utility be appraised according to some clearcut set of substantive standards, for the following reasons:’ 0 0 0
there are several crosscutting dimensions to a policy-supporting materials information system; there are pros and cons to almost any system design; in the final analysis, the worth of any system rests much less m the excellence of its design than in the excellence of the personnel that operate it.
What is appropriate for any policy-supportive system also depends heavily on the style, comprehension, and background knowledge of the policy makers and analysts who use it; this is hardly a fixed parameter. Some mode of appraisal of the information systems is desirable whether or not it is clearcut. We consider four different viewpoints from which an appraisal may be made and some desiderata identified.
Relationship
to policy decisions
A principal barrier to either designing organizational structure for materials
1977
or appraising a particular information is overlap of
181
Materials in@matbn
systems
for public policy making
a One reason for this is that the adequacy of other available information is rarely clear-cut. When, for example, the US Environmental Protection Agency needed to consider the potential impact of specific effluent limitations on the paper and other industries, there was no apparent thought that the work could be within accomplished elsewhere the federal establishment, using existing information banks and analytical models. EPA was probably correct in assuming that adequate impact models did not exist, especially since, for its purposes, ‘adequate’ meant being good enough not only to satisfy the agency’s decision makers, but to meet the challenge of every affected interest that could contest its decisions. The agency’s reaction was first to contract out for an emergency type of analysis and, once the initial pressure was reduced, to contract further for a definitive series of more larger, ‘management by objective’ (MB01 studies. The econometric and other models that are part of these studies are intended to be the definitive way of evaluating regulatory impact on various industries. This course of action might be acceptable, save for the fact that the locus of industry expertise in the US if anywhere, is in the government, Department of Commerce, that no single approach or evaluation is about to be accepted throughout the government as definitive, and that EPA’s MB0 analyses have yet to be satisfactorily linked with some approach to generalized economic impact.
182
materials information with other kinds of information. We may define ‘materials information’ as including impacts of the materials economy on the rest of the economy. One would not want to decide any policy question affecting materials without assessing such indirect impacts, but by the same token, the kinds of information affecting materials must also be considered as being environmental information, foreign affairs information, labour information, consumer information, etc. If this breadth of expertise could conceivably be embodied in individuals or small units, should it then be duplicated over and over within the government to serve as many kinds of policy makers? There is certainly virtue in the specialized marketing of a particular product in this case an information product - but there is also virtue in developing specialized skills in its production. Since, even for more tangible goods, it has never been quite settled whether there is superior merit in ‘knowing the territory’ or ‘knowing the product’, there is no obvious answer for the organization of government information systems. Whether through established coordinating procedure or through the alertness of personnel, the desiderata which really have to be met are that neither the direct nor the indirect impact on or through materials of any policy bearing on materials, nor any intragovernmental source for such information, is inadvertently overlooked. In other words, whether the relevant data and analysis are adduced by materials specialists or by any other kind of specialists is inconsequential, so long as, if significant, they are adduced. Any duplication of effort should be deliberate - either to ensure the consideration of possibly conflicting data, analyses, or biases, or to avoid place or time disutility. Timing is critical. At the point that a particular policy decision is under consideration, there is the choice of using various combinations of already gathered facts, already completed analysis, newly gathered facts, and new analysis. The choice will rationally depend upon the adequacy of what has gone before and the convenience of assimilating it. A critical information-system challenge is to have anticipated the extent to which timeliness of response and economy of effort, as well as the conservation of fugitive information, dictate advance information-system investment. More often than not, in the federal government, the answer to such questions has been improvization, duplication, and over-investment. It is physically and psychologically easier to meet each new information or analysis requirement afresh than to retrieve - let alone ascertain the present worth of - any previous or concurrent research or information activity. Once having been faced with an information ‘gap’, an agency tends to be restrained only by budget from setting up a system with which it will permanently be prepared for repetition of the need.s Given the current state of the evaluative art, it is a reasonable conclusion that no one agency should be investing any more than is immediately necessary in methodological innovations until it or some other (possibly more neutral) agency has more thoroughly explored on a prototype basis the methodological alternatives and their costs and benefits.
The time dimension One of the problems of deciding on a particular form of analytical modelling is that different kinds of model are appropriate to different
RESOURCES
POLICY
September
1977
Materials information systemsforpublicpolicy
making
forward-looking time spans. The most complex impact modelling is suitable for short-term questions, measured in months rather than years. Though dynamic models have been designed which proceed by fractions of a year to as far ahead as the computer budget will allow, it is doubtful that their output is useful for more than a few years. As the time horizon lengthens, structurally simpler models are usually more appropriate, and the type of basic data required also changes.
requirements. One reason for the establishment of the National Commission on Supplies and Shortages was congressional frustration with the commodity shortages of 1973-74, and the feeling that if there had been an adequate early warning system ameliorative action could have been taken. The usual mental concept of such a system involves models and automated data banks. The Department of Commerce Materials Division, which has been trying to establish an early warning system, is bringing into use a number of computerized systems and models, though it still relies fundamentally on the knowledge and alertness of its individual industry specialists9 The inconclusiveness of present analytical systems with regard to indirect consequences has already been noted. Determining the proximate consequences of a government decision is not much less of a challenge. Were this not so, there would not be a buyer for every seller on the commodity exchanges or a variety of public and private forecasting systems and prophets. Whatever the methodology, no model or analyst can do better than project probable outcomes or specify a high-probability range of outcomes. Unfortunately for both models and modellers, ‘validation’ by past successes or retrospective fittings is no guarantee of current accuracy. Potential consequences do have to be forecast somehow, however, and the most dependable information will presumably come from the combination of an individual analyst familiar with his subject and inspired impact modelling. Whatever their design, and whether econometrics, judgment, or engineering data are the source of their parameters, the most useful models will be those which reflect real-life behaviour rather than economic theory or rational optimization. They will be economical as well as useful if they have been stripped of the variables and equations that add little either to understanding of the behavioural mechanism or precision of the results. To predict short-term shortages, materials models need not be complicated and may have only limited utility.‘* Short-range
’ It is doubtful that any elaborate system is actually either needed or useful to provide early warning of imminent materials developments requiring policy policy decisions. The history of 1973 deliberations preceding the imposition of export controls on ferrous scrap shows that there is no lack of early warning of emerging short-range problems by affected interests. (See Fischman, op cit. Ref 2, pp 107ff and 273ff. Another illustration is the export control episode with regard to cattle hides, described on pp 131ff and 345ff.) Executive branch and congressional show decided makers also policy reluctance to take unpalatable action without the prior build-up of enough of a visible crisis to generate a resounding positive demand. A hesitance to act on early warning signals may well have substantive justification. More critical than early warning per se is the ability to distinguish false real alarms. and Furthermore, ~najor information the requirement is related not with alertness, but with understanding the direct and indirect of alternative consequences actions or inaction. lo The reason for this is that the controlling factor in shortages (or in shortage-induced high prices) is not the nature of the particular commodity but the state of US and world business cycles. If the macro-economic predictions which them are incorrect, feed commodity models, no matter how exquisite, will do a poor job of forecasting coming demand and supply. A reasonable understanding of what happened at the last peak, plus some rough knowledge of what has since happened to industry capacity and product flows, will permit reasonable predictions of what will happen at the next peak, without a model.
RESOURCES
POLICY
September
Medium-range requirements. For medium-range purposes (eg 2-10 years), systems and models become more useful to the analyst. Even here, however, industry and commodity specialists have an advantage over straight mechanical projections of being able to use a changing variety of updated information on capacity planned, under construction, and existing, as well as on shifts in supply patterns and consumption habits. An optimum arrangement is the kind of model which can receive and make use of such ‘forerunner’ information inputs. Modelling for this medium range is particularly difficult, as many variables will still be significantly affected by the state of the business cycle, which at this range can only be guessed. Much of the difficulty will disappear, however, if the model or information system is not
1977
183
Mahiak
i$onnation systems for public policy making
asked the wrong questions. Most policy questions looking forward into this period are best served by infdrmation on peaks, troughs, and trend values, and rarely require projections of actual quantities or system for this prices. The alertness function of an information medium range is important, because there will be less forward-looking reportage in the trade press, and it may not be possible, without detailed analysis, to foresee the constraints and problems which may develop under various sets of circumstances.
I1 For the very long term (over 50 years), there is no need for an information system as such. Persons are required who are sufficiently familiar with the analytical in literature are so placed and government that they can keep policy makers within reasonable limits. I2 A particularly difficult question is the handling of microdata. Certain types of policy analysis depend heavily upon detailed data for individual firms and establishments, yet it is not easy to preserve the current norm of maximum confidentiality and still make these data available to the extent required. A further problem is that many of the kinds of data which particular policy questions seem to call for do not exist as such in all plants and enterprises. To require such data therefore either imposes a burden, which it should be specifically estimated is worth its economic cost, or generates pm fom replies whose inaccuracy may put their value in doubt. It is incumbent upon the initiators of government surveys not only to review the p&m f& results of survey pretests, but to look with care into the procedures respondents resort to to provide their replies.
184
Long-range w@wwzts. Most of the shorter range modelling detail and other shorter range considerations clearly become irrelevant when looking 1650 years ahead. Only trend values, or at most the trend values for peaks and troughs, are of real use. At this range, one has to accept that predictions can be reasonably made only within widening bounds of error. Many of the grosser demand-determining relationships are persistent, however, and this is what makes longterm projection plausible. Supply is a much more difficult question. Even up to 50 years, there is little reason to expect materials consumption to be limited by fundamental resource constraints, but effective resource adequacy still depends upon timely inputs of exploration, R&D, capacity investment, and planning. The policy role of materials information systems for this long-range horizon is essentially one of uncovering the areas for vigilance (eg international political forecasting) or which require current governmental action because short-range commercial interest may differ from what is of long-range public advantage. Long-range models and projections may also contribute significantly by persuading against ill-advised action. Long-range supply schedules can be devised, as in the Bureau of Mines data bank (the ‘minerals availability system’), which provide some guidance on the possible extent of prospective supply stringency. However, given the broad uncertainties posed by prospective technological change, by as yet undiscovered reserves, and by poorly predictable rearrangements in world trade, such supply schedules cannot very credibly be applied as feedback into demand-supply equilibrium models. It would be misleading in most cases to pretend that long-range forecasts are based on anything but broad judgments.” Information
system stages
An information system was defined earlier as comprehending the acquisition of quantitative and qualitative data, analysis of data, and communication of findings. Acquisition and storage. It is easy to specify the kinds of data consumption, imports, exports, inventory changes, (production, prices, production capacity, natural resource limitations, production costs, etc) which are useful as decision-making inputs, but this does not help in concluding what kind of detail, at what order of accuracy, ought to be repetitively collected and tabulated. The many inconsistencies of data series over time, among issuing agencies, and among different materials, clearly illustrate the welter of technical choices facing data gatherers and manipulators.” The question of confidentiality extends to storage decisions. If returns from surveys are permanently stored (either in original form or on magnetic tape) complete with individual identification, there are
RESOURCES
POLICY
September
1977
Materials I3 Restraint in survey activity, uniformity of standards, and quality of statistical practice are in theory enforced by the Division of Statistical Standards in the Office of Management and Budget (part of the President’s Executive Office), but owing partly to resource limitations, its effectiveness has been limited. ‘*For materials, the most fundamental analytical requirement is to have reasonably reconciled data on supply and consumption. This has long been recognized by the US Department of Agriculture in its ‘supply and distribution’ (sometimes ‘supply and disappearance’) tables and has become the organizing principle in the Bureau of Mines’ recently inaugurated annual series, Miner& in the US Economy. The amount of estimation that goes into the latter is an indication of the shortfall in basic data that calls for survey consideration of possible improvements. On the other hand, there is a tendency to try to fit data and analysis into unnecessarily constricting patterns. For example, it is at least questionable whether the SIC industry allocation which the Bureau is now attempting for its consumption distribution, and which for the typical mineral commodity involves considerable estimation, is sufficient of an advantage over more traditional consumption categories to warrant the unavoidable inaccuracies and extra effort. ” Much gathering analysis of and materials (and other) data is for routine periodic summaries, ranging in complication from ‘value of mineral production’ tabulations to the National Income Accounts large-scale and input-output models. These comprehensive compilations are almost as demanding in processing unimportant important data, or in providing as sometimes spurious precision. There is their little practical choice as to continuation, since once however, initiated, each has tended to generate its own market. The time for choice and discrimination in such tabulations is in the early stages, before the market has had a chance to develop. I6 Some agencies in the federal government have gained the reputation of being beyond influence by parties whose fortunes may be affected by the data they issue: other materials information agencies (eg the Bureau of Mines and the Bureau of Domestic Commerce) have not. In general, it is the closeness of association between agency personnel and the industries they ‘serve’ that leads to suspicion. The agencies, of course, consider this closeness the very basis of the accuracy and instructiveness of their analyses and and there is data, considerable merit in that view. Such agencies are likely to achieve more trusted status only if their expositions and evaluations carefully distinguish industry views from their own and do not give the impression of uncritical acceptance of
continued on p 186
RESOURCES
information
systems
forpublicpolicy
making
permanent risks, regardless of safeguards, of compromise of confidentiality. If they are stored without identification, checkbacks for error and cross-tabulations with other data sets are impossible. US federal agencies vary considerably in their handling of surveys and microdata, and some (partly at their own insistence) are even governed by separate statutes. Recent rulings and court decisions, however, have led to a greater concern with confidentiality, as an offset to ‘freedom of information’ procedures and magnetic-tape data distributions which have combined to open up the federal government’s internal statistical workings to more general public inspection.13 Selection of methods of aggregating and cross-tabulating data involves many of the same criteria that apply to data collection. The decision on what to collect and how to specify it should depend largely on a prior design of the tables, charts, and other modes in which the data are to be used.14 The role of price is of particular interest in many materials problems. Price series are known to have defects, varying from material to material, as representations of actual transaction prices, yet average-value-of-output series are rarely substituted. Apart from seasonal adjustment and annual averaging, there is little routine manipulation of price data in the federal government, the series being considered to have analytical significance in and of themselves. Many of the resource and materials price series relied upon most for policy analysis are those collected and published privately, rather than the corresponding series included in the Bureau of Labor Statistics wholesale price indexes.15
Analysis.
Communication. There is no optimum formula for communicating materials information. Federal agencies seem to have become increasingly adept at communicating principal interrelationships, and have effectively used graphics to display some of the subtleties. A considerable challenge remains, however, which cannot be met except by constant interaction between information and policy personnel. Clarity in communication is often achieved only at the expense of making complicated things seem simple, which can badly prejudice the resulting decisions. While a reasonable mastery of communications techniques is essential (frequently to the point of interposing a communicator between the information specialist and his decision-making clients), an equally important requisite is adaptability to the varying styles, capabilities, backgrounds, and time commitments of the clientele to be served. A good communication system will provide the generalist with the essence of a situation, without either leading or misleading, and the specialist with all of the details required for complete analytical understanding. Credibility is said to be one of the most serious problems with some sectors of the current materials information system. Communication failures are part of the reason for this lack of confidence. Technicians are usually only frustrated by such problems as discrepancies in figures issued at different times and places; policy makers, or their aides, are likely to lose faith. A more copious accompaniment of . changed or alternative figures by reconciliations or other explanations would improve the image of issuing agencies with both types of users.16
POLICY September
1977
185
Materials
information
systems forpublicpolicy
making
Purposes of systems Materials information systems can be said roughly to serve one of three kinds of purposes - alertness to new circumstances that require policy or decision making attention (‘monitoring’ or ‘early warning’); formulation and periodic revision of standby-resource strategies and contingency plans for dealing with possible adverse developments (such as supply interruptions or cartel actions); and capacity, in terms of information base and continuing ‘overhead’ analysis, to handle important ad hoc evaluations in a timely and economical fashion.
continued from p
185
information from industry sources. While is communication alone good not sufficient to establish a reputation for independence, it is an essential minimum. It must be borne in mind that can flow two ways. communication Ultimate responsibility for the quality of information systems lies not with the but with producers, information information consumers. Policy makers who view information systems merely as devices to provide briefs and backup for views they reach independently will get what they are looking for. Congressmen and executive branch officials who do not respond to information flows with some critical interest will get even less.
186
Early wartzing. A variety of potentially emerging situations is involved, ranging from the near to long term. However, it is doubtful that any elaborate information system is either very much needed or very useful in providing early warning of such problems as impending materials shortages. This does not mean that the trade press and price trends should not be monitored, or that the information should not be communicated to policy makers without waiting for it to come to their attention through complaints or from their more general reading. The principal purpose, however, would be to encourage advance thought about the situation and avoid hasty action. A risk in responding positively to an early warning is that the action may exacerbate the situation it is designed to remedy. This is even more true of precursors of positive action, such as attempts to get more information on the seriousness of a supply or price perturbation. Early warning systems for supply or price problems thus involve a dilemma. If they provide for adding monitoring functions in the face of an impending crisis, they may well precipitate such a crisis; if they introduce monitoring functions on a continuing basis, in a non-crisis atmosphere, they may burden both the government and respondents with unnecessary cost and effort. Another dilemma facing early warning systems is the trade-off between persuasiveness and objectivity. When warnings are confident and clear they are more likely to be acted upon, whereas dispassionate, ‘two-handed’ projections generally receive little attention. Yet the latter type of projection usually reflects more accurately the forecaster’s ability to predict. In judging forecasts, little can be made of a forecaster’s ‘track record’. Retrospective evaluations of early warnings can never be unequivocal, especially since such warnings tend to be either self-fulfilling or self-defeating. Ideally, a warning should be given early enough and persuasively enough to trigger the responses that will prove it wrong. The longer range the pending problem, the easier it is to put off action. The dilemma of choice between dramatic and well balanced particularly to long-range forecasting. applies predictions Fortunately, at least some long-range problems can be attacked with relatively modest, readily acceptable beginnings which, if sustained over long periods, can have significant cumulative effects. An information system that keeps its balance, therefore, can still hope to accomplish something without running the risk of stimulating major national commitments - to action or to inaction - that prove to be major blunders. Contingency pt!cmning. This category of information systems, required to deal with such contingencies as supply interruptions and sharp price rises, has a predictive aspect of seemingly ‘early warning’
RESOURCES
POLICY
September
1977
Materials in@mation
I7 More precisely, what is needed is a schedule of probabilities for different magnitudes. For example, the probability of a partial interruption of US imports of metallurgical manganese is greater than the probability of a complete interruption, and the probability of a short interruption is greater than the probability of a long information system, by The one. assembling all available intelligence and processing it through personal judgment or through gaming or other models, needs to put probability numbers on a matrix of size-duration combinations. are A Technology them ” Among of Economic Stockpile Assessment Policy, Office of Technology Assessment, Critical 1976 and us Congress, Materials: Commodity Action Analyses: Aluminum, Chromium, Platinum, and Palladium, Office of Minerals Policy Development, US Department of the March 1975. A mostly Interior, conference on methodological Contingency planning is being planned for December 1977 under the sponsorship of and MITRE Corporation the the Engineering Foundation. 19The importance of evaluating such been consequences has ancillary recognized by the US Congress in its frequent specification of economic impact analysis as part of a regulatory procedure, and by the Executive in a broadside order calling for ‘Inflation Impact Statements’ (subsequently renamed ‘Economic Impact whenever rules and Statements’) regulations may affect costs or prices. Such analysis calls for an information base which includes not only basic data on materials supply and distribution and on price and production arrangements, models of the macroeconomic but relationship of the materials economy to economy, and rest of the the microeconomic models of the relationship of individual firms and producing units to the materials economy.
RESOURCES
POLICY
September
systems
for public
policy
m&zing
nature, which it may well have to incorporate. What is ultimately needed, however, as a foundation for contingency policy planning is prediction, not that an event is likely to occur, but of the degree of risk of its occurrence, and with what magnitude both in volume and in duration.” Probability of occurrence is by itself inconclusive for action purposes. It is equally important to be able to predict the consequences of each potential size-duration combination. A potentially cataclysmic event merits attention even if it has relatively minute possibility of occurrence; an event of relatively minor consequence may deserve attention if its occurrence is rather certain. In this sense, early warning is a special, prejudged case of contingency analysis; the forecaster has predicted that there is sufficient probability of occurrence of an event of sufficient gravity that action should be considered to avert it. Measures to avert and to compensate are sometimes identical - in kind, if not in magnitude. Stockpiles, for example, will tend to discourage some types of price increase or supply interruption, at the same time as they provide a means of compensating for the price or supply contingency if it occurs. A common element in aversive and compensatory measures is that they both aim at avoiding the damage of a price rise or supply interruption. Most of the calculations connected with contingency analysis are well beyond any current state of the art. The tendency in recent studiesI has been to rely on neoclassical demand-supply analysis, and to calculate net balance between loss of consumer, and increase in producer, surplus. Such analysis is sensitive to rather shaky elasticity parameters, in addition to being insensitive to the practical possibilities for reallocation of resources. Contingency analysis requirements differ little from what is required for the wider objective of being able to predict the consequences of alternative actions or inaction in a variety of circumstances. The areas of decision can vary widely, to include export control policy, tax policy, environmental and health and safety regulation, price and wage controls, transportation rates and arrangements, public land policy, forest management, etc. Decisions in all these fields have an impact on materials, but since the impact is not always direct, data and models should be available somewhere in government to establish the links between the proximate effect of each such policy and the consequent impact on materials and through materials on the rest of the economy.‘g On-line capacity. In deciding how far to go in setting up a continuing information base, the considerations are in many respects similar to such private yardsticks as life-cycle costs and net present value of return on investment. The difficulty in applying the latter concept in government is that the value of better decision making is largely a matter for political determination. The life-cycle cost criterion is a little easier to apply. Federal agencies can estimate the maintenance cost of any information system they introduce and can consider periodically whether the continued patching of an existing system might not exceed the cost of installing and maintaining a new one. They can also take into account that a dollar of expenditure deferred is at least five cents saved. A good rule is that materials information systems should attempt
1977
187
Materials information systems for public policy making to anticipate decision-making requirements, but not too thoroughly and not too far in advance. Time should be allowed for ‘market’ evaluation of systems, despite the problems such evaluation presents. There are economies of scale in information systems (both crosssectionally and over time) and system managers need to have due regard for the differences between overhead and incremental cost. Once the basics of a system have been worked out (investment), the marginal costs of maintaining it or of extending its scope may be small enough to balance even uncertain expectations as to its future use. Again, in conducting field surveys, it is cheaper to obtain additional information at the same time than to come back for more later (though it is also necessary to consider decreasing marginal quality of response). Conservation of information is another criterion for determining on line capacity. Information that is not obtained today may not be available tomorrow, or may not be available with comparable accuracy. Timing may be critical to its utility. Data and analysis which are not ready when most needed provide less than optimal return on expenditure. It is generally better to gather too much too soon than too little too late. Information systems need constant planning, care, and attention. There is little sense in investing heavily in grand designs without providing architectural supervision. Competent, experienced, imaginative personnel are an essential. The best, most respected information systems in the US government, such as those of Census and the Bureau of Labor Statistics, have enough resources to engage in planning and methodological analysis, whereas the most able personnel in the poorer systems tend to be employed on miscellaneous duties or are largely occupied with crises. Excess resources can easily be squandered, however, especially if an information system is allowed to become too much a creative plaything, without constant exposure to a cost-utility standard. It is the policy analysts who have to undertake the principal responsibility for specifying which kinds of information are useful and which are relatively superfluous for policy analysis purposes. They must also remain clear-headed as to what kind of analysis is both illuminating and likely to have ultimate decision-making impact.
188
RESOURCES
POLICY
September
1977