Handbook of Econometrics Vols1-5 _ Chapter 35 pdf

Chupter 35 ECONOMIC POLICY FORMATION: THEORY AND IMPLEMENTATION (APPLIED ECONOMETRICS IN THE PUBLIC SECTOR) LAWRENCE R KLEIN University ofPennsylvaniu Contents Some contemporary policy issues Formal political economy Some policy projections The theory of economic policy Prospects Appendix: An outline of a combined (Keynes-Leontief) input-output/macro model Hundhook of Econometrics, Volume III, Edited by Z Griliches and M.D Intriliguior Elsevier Science Publishers B V, 1986 2058 2063 2070 2072 2087 2088 L R Klan 2058 Some contemporary policy issues Mainstream economic policy, known basically as demand management, and its econometric implementation are jointly under debate now The main criticism comes from monetarists, who focus on versions of the quantity theo_ry of money, from advocates of the theory of rational expectations, and more recently from supply side economists All these criticisms will be considered in this paper, as well as the criticisms of public policy makers, who are always looking for precision in their choice procedures, even when the subject matter is inherently stochastic and relatively “noisy” Demand management is usually identified as Keynesian economic policy, i.e as the type that is inspired by the aggregative Keynesian model of effective demand Also, the mainstream econometric models are called Keynesian type models; so the present state of world wide stagflation is frequently attributed to the use of Keynesian econometric models for the implementation of Keynesian policies These are popular and not scientific views In this presentation the objective will be to put policy measures in a more general perspective, only some of which are purely demand management and aggregative Also, the evolution of econometric models for policy application to many supply-side characteristics will be stressed To a certain extent, the orientation will be towards experience derived from the application of U.S models to U.S economic policy, but the issue and methods to be discussed will be more general For purposes of exposition, two types of policy will be examined, (1) overall macro policies, and (2) specific structural policies Macro policies refer to traditional monetary and fiscal policies, principally of central governments, but the model applications to local government policies are also relevant As the world economy becomes more interdependent, more economies are recognizing their openness; therefore, trade/payments policies are also part of the complement known as macro policy By structural policy I mean policies that are aimed at specific segments of the economy, specific groups of people, specific production sectors, distributions of aggregative magnitudes or markets Economists like to focus on macro policies because they have overall impacts and leave the distributive market process unaffected, able to its seemingly efficient work Most economists look upon the free competitive market as an ideal and not want to make specific policies that interfere with its smooth working They may, however, want to intervene with structural policy in order to preserve or guarantee the working of the idealized market process Macro policies are quite familiar Monetary policy is carried out by the central bank and sometimes with a treasury ministry Also, the legislative branch of Ch 35: Economic Policy Formation 2059 democratic governments influence or shape monetary policy Central executive offices of government also participate in the formation of monetary policy It is a many sided policy activity The principal policy instruments are bank reserves and discount rates Reserves may be controlled through open market operations or the setting of reserve requirements Policies directed at the instrument levels have as objectives specified time paths of monetary aggregates or interest rates At the present time, there is a great deal of interest in controlling monetary aggregates through control of reserves, but some countries continue to emphasize interest rate control through the discount window On the whole, monetary authorities tend to emphasize one approach or the other; i.e they try to control monetary aggregates along monetarist doctrinal lines or they try to control interest rates through discount policy, but in the spirit of a generalized approach to economic policy there is no reason why central monetary authorities cannot have multiple targets through the medium of multiple instruments This approach along the lines of modern control theory will be exemplified below Monetary policy is of particular importance because it can be changed on short notice, with little or no legislative delay It may be favored as a flexible policy but is often constrained, in an open economy, by the balance of payments position and the consequent stability of the exchange value of a country’ s currency Therefore, we might add a third kind of financial target, namely, an exchange value target Flexibility is thus restricted in an open economy Monetary policies that may seem appropriate for a given domestic situation may be constrained by a prevalent international situation There are many monetary aggregates extending all the way from the monetary base, to checking accounts, to savings accounts, to liquid money market instruments, to more general credit instruments The credit instruments may also be distinguished between private and public sectors of issuance The plethora of monetary aggregates has posed problems, both for the implementation of policy and for the structure of econometric models used in that connection The various aggregates all behave differently with respect to reserves and the monetary base The authorities may be able to control these latter concepts quite well, but the targets of interest all react differently Furthermore, monetary aggregates are not necessarily being targeted because of their inherent interest but because they are thought to be related to nominal income aggregates and the general price level The more relevant the monetary aggregate for influencing income and the price level, the more difficult is it to control it through the instruments that the authorities can effect Benjamin Friedman has found, for the United States, that the most relevant aggregate in the sense of having a stable velocity coefficient is total credit, but this is the least controllable.’ The most controllable aggregate, currency plus checking ‘ Benjamin Friedman, “The Relative States: Evidence and Some Speculations”, 645, March, 1981 Stability of Money and Credit ‘ Velocities’in the United National Bureau of Economic Research, working paper No 2060 L R Klein accounts, has the most variable velocity Between these extremes it appears that the further is the aggregate from control, the less variable is its associated velocity This is more a problem for the implementation of monetary policy that for the construction of models But a problem for both policy formation and modeling is the recent introduction of new monetary instruments and technical changes in the operation of credit markets Electronic banking, the use of credit cards, the issuance of more sophisticated securities to the average citizen are all innovations that befuddle the monetary authorities and the econometrician Authorities find that new instruments are practically outside their control for protracted periods of time, especially when they are first introduced They upset traditional patterns of seasonal variation and generally enlarge the bands of uncertainty that are associated with policy measures They are problematic for econometricians because they establish new modes of behavior and have little observational experience on which to base sample estimates Side by side with monetary policy goes the conduct of fiscal policy For many years - during and after the Great Depression-fiscal policy was central as far as macro policy was concerned It was only when interest rates got significantly above depression floor levels that monetary policy was actively used and shown to be fairly powerful Fiscal policy is usually, but not necessarily, less flexible than monetary policy because both the legislative and executive branches of government must approve major changes in public revenues and expenditures In a parliamentary system, a government cannot survive unless its fiscal policy is approved by parliament, but this very process frequently delays effective policy implementation In a legislative system of the American type, a lack of agreement may not bring down a government, but it may seriously delay the implementation of policy On the other hand, central banking authorities can intervene in the functioning of financial markets on a moment’ notice s On the side of fiscal policy, there are two major kinds of instruments, public spending and taxing Although taxing is less flexible than monetary management, it is considerably more flexible than are many kinds of expenditure policy In connection with expenditures, it is useful to distinguish between purchases of goods or services and transfer payments The latter are often as flexible as many kinds of taxation instruments It is generally safer to focus on tax instruments and pay somewhat less attention to expenditure policy Tax changes have the flexibility of being made retroactive when desirable This can be done with some expenditures, but not all Tax changes can be made effective right after enactment Expenditure changes, for goods or services, especially if they are increases, can be long in the complete making Appropriate projects must be designed, approved, and executed Often it is difficult to find or construct appropriate large projects Ch 35: Economic Policy Formation 2061 Tax policy can be spread among several alternatives such as personal direct taxes, (either income or expenditure) business income taxes, or indirect taxes At present, much interest attaches to indirect taxes because of their ease of collection, if increases are being contemplated, or because of their immediate effect on price indexes, if decreases are in order Those taxes that are levied by local, as opposed to national governments, are difficult to include in national economic analysis because of their diversity of form, status, and amount Some tax policies are general, affecting most people or most sectors of the economy all at once But, speci$c, in contrast to general, taxes are important for the implementation of structural policies An expenditure tax focuses on stimulating personal savings Special depreciation allowances or investment tax credits aim at stimulating private fixed capital formation Special allowances for R&D, scientific research, or capital gains are advocated as important for helping the process of entrepreneurial innovation in high technology or venture capital lines These structural policies are frequently cited in present discussions of industrial policy A favorite proposal for strictly anti-inflationary policy is the linkage of tax changes, either as rewards (cuts) or penalties (increases), to compliance by businesses and households with prescribed wage/price guidelines Few have ever been successfully applied on a broad continuing scale, but this approach, known as incomes policies, social contracts, or TIPS (tax based incomes policies), is widely discussed in the scholarly literature These monetary and fiscal policies are the conventional macro instruments of overall policies They are important and powerful; they must be included in any government’ policy spectrum, but are they adequate to deal with the challenge of s contemporary problems? Do they deal effectively with such problems as: -severe unemployment among certain designated demographic groups; -delivery of energy; -conservation of energy; -protection of the environment; -public health and safety; -provision of adequate agricultural supply; -maintenance of healthy trade balance? Structural policies, as distinct from macro policies, seem to be called for in order to deal effectively with these specific issues If these are the kinds of problems that economic policy makers face, it is worthwhile considering the kinds of policy decisions with instruments that have to be used in order to address these issues appropriately, and consider the kind of economic model that would be useful in this connection 2062 L R Klein For dealing with youth unemployment and related structural problems in labor markets, the relevant policies are minimum wage legislation, skill training grants, and provision of vocational education These are typical things that ought to be done to reduce youth unemployment These policy actions require legislative support with either executive or legislative initiative In the case of energy policy, the requisite actions are concerned with pricing of fuels, rules for fuel allocation, controls on imports, protection of the terrain against excessive exploitation These are specific structural issues and will be scarcely touched by macro policies These energy issues also effect the environment, but there are additional considerations that arise from non-energy sources Tax and other punitive measures must be implemented in order to protect the environment, but, at the same time, monitor the economic costs involved The same is true for policies to protect public health and safety These structural policies need to be implemented but not without due regard to costs that have serious inflationary consequences The whole area of public regulation of enterprise is under scrutiny at the present time, not only for the advantages that might be rendered, but also for the fostering of competition, raising incentives, and containing cost elements It is not a standard procedure to consider the associated inflationary content of regulatory policy Ever since the large harvest failures of the first half of the 1970s (1972 and 1975, especially) economists have become aware of the fact that special attention must be paid to agriculture in order to insure a basic flow of supplies and moderation in world price movements Appropriate policies involve acreage limitations (or expansions), crop subsidies, export licenses, import quotas, and similar specific measures They all have bearing on general inflation problems through the medium of food prices, as components of consumer price indexes, and of imports on trade balances Overall trade policy is mainly guided by the high minded principle of fostering of conditions for the achievement of multilateral free trade This is a macro concept, on average, and has had recent manifestation in the implementation of the “Tokyo Round” of tariff reductions, together with pleas for moderation of non-tariff barriers to trade Nevertheless, there are many specific breaches of the principle, and specific protectionist policies are again a matter of concern Trade policy, whether it is liberal or protectionist, will actually be implemented through a set of structural measures It might mean aggressive marketing in search of export sales, provision of credit facilities, improved port/storage facilities, and a whole group of related policy actions that will, in the eyes of each country by itself, help to preserve or improve its net export position We see then that economic policy properly understood in the context of economic problems of the day goes far beyond the macro setting of tax rates, overall expenditure levels, or establishing growth rates for some monetary aggregates It is a complex network of specific measures, decrees, regulations (or their 2063 Ch 35: Economic Policy Formation absence), and recommendations coming from all branches of the public sector In many cases they require government coordination Bureaus, offices, departments, ministries, head of state, and an untold number of public bodies participate in this process It does not look at all like the simple target-instrument approach of macroeconomics, yet macroeconometric modeling, if pursued at the appropriate level of detail, does have much to contribute That will be the subject of sections of this paper that follow Formal political economy The preceding section has just described the issues and actors in a very summary outline Let us now examine some of the underlying doctrine The translation of economic theory into policy is as old as our subject, but the modern formalism is conveniently dated from the Keynesian Reuolution Clear distinction should be made between Keynesian theory and Keynesian policy, but as far as macro policy is concerned, it derives from Keynesian theory The principal thrust of Keynesian theory was that savings-investment balance at full employment would be achieved through adjustment of the aggregative activity level of the economy It was interpreted, at an early stage, in a framework of interest-inelastic investment and interest-elastic demand for cash This particular view and setting gave a secondary role to monetary policy Direct effects on the spending or activity stream were most readily achieved through fiscal policy, either adding or subtracting directly from the flow of activity through public spending or affecting it indirectly through changes in taxation Thinking therefore centered around the achievement of balance in the economy, at full employment, by the appropriate choice of fiscal measures In a formal sense, let us consider the simple model C=f(Y-T) T=tY Z = g(AY) Y=C+Z+G consumption function tax function investment function output definition where G = public expenditures A = time difference operator Y= total output (or income, or activity level) Fiscal policy means the choice of an appropriate value of t (tax rate), or level G (expenditure), or mixture of both in order to achieve a target level of Y This could also be a dynamic policy, by searching for achievement of a target path of Y through time To complement dynamic policy it is important to work with a L R Klein 2064 richer dynamic specification of the economic model Lag distributions of Y-T or AY in the C and I function would be appropriate This kind of thinking inspired the approach to fiscal policy that began in the 1930’ and still prevails today It s an economic system It is inspired thoughts about “fine tuning” or “steering” obviously terribly simplified It surely contains grains of truth, but what are the deficiencies? In the first place, there is no explicit treatment of the price level or inflation rate in this system Arguments against Keynesian policy pointed out the in!ationary dangers from the outset These dangers were minimal during the 1930’ and did s not become apparent on a widespread basis for about 30 years-after much successful application of fiscal policy, based on some monetary policy as time wore on There is no doubt, however, that explicit analysis of price formation and great attention to the inflation problem must be guiding principles for policy formation from this time forward Another argument against literal acceptance of this version of crude Keynesianism is that it deals with unrealistic, simplistic concepts Fiscal action is alnot directed towards “t ” or “G” Fiscal action deals with complicated lowances, exemptions, bracket rates, capital gains taxation, value added taxation, expenditures for military hardware, agricultural subsidies, food stamps, aid to dependent children, and unemployment insurance benefits These specific policy instruments have implications for the broad, general concepts represented by “t ” and “G “, but results can be quite misleading in making a translation from realistic to such macro theoretical concepts The system used here for illustration is so simplified that there is no distinction between direct and indirect taxes or between personal and business taxes The Keynesian model of income determination can be extended to cover the pricing mechanism, labor input, labor supply, unemployment, wages, and monetary phenomena There is a difference, however, between monetary analysis and monetarism Just as the simple Keynesian model serves as the background for doctrinaire Keynesian fiscal policy, there is another polar position, namely, the monetarist model which goes beyond the thought that money matters, to the extreme that says that only money matters The monetarist model has its simplest and crudest exposition in the following equation of exchange Mv=Y For a steady, parametric, value of v (velocity), there is a linear proportional correspondence between M (nominal money supply) and Y (nominal value of aggregate production or income) For every different M-concept, say M,, we would have * MiV, = Y ‘ See the various concepts in the contribution by Benjamin Friedman, op tit Ch 35: Economic Policy Formation 2065 A search for a desired subscript i may attach great importance to the corresponding stability of ui It is my experience, for example, that in the United States, u2 is more stable than ut More sophisticated concepts would be Mu= WiY_,, i=o or or The first says that M is proportional to long run Y or a distributed lag in Y The second says that M is proportional to a power of long run Y or merely that a stable relationship exists between long run Y and M Finally, the third says that M is a function of long run price as well as long run real income (X) In these relationships no attention is paid to subscripts for M, because the theory would be similar (not identical) for any M, and proponents of monetarist policy simply argue that a stable relationship should be found for the authorities for some Mi concept, and that they should stick to it The distributed lag relationships in P_; and X_i are evidently significant generalizations of the crude quantity theory, but in a more general view, the principal thing that monetarists need for policy implementation of their theory is a stable demand function for money If this stable function depends also on interest rates (in lag distributions), the theory can be only partial, and analysis then falls back on the kind of mainstream general macroeconometric model used in applications that are widely criticized by strict monetarists.3 The policy implications of the strict monetarist approach are clear and are, indeed, put forward as arguments for minimal policy intervention The proponents are generally against activist fiscal policy except possibly for purposes of indexing when price movements get out of hand According to the basic monetarist 3The lack of applicability of the monetarist type relationship, even generalized dynamically, to the United Kingdom is forcefully demonstrated by D F Hendry and N R Ericsson, “Assertion without Empirical Basis: An Econometric Appraisal of Friedman and Schwartz’ ‘ Monetary Trends in the United Kingdom,“’ Monetary Trends in the United Kingdom, Bank of England Panel of Academic Consultants, Panel Paper No 22 (October 1983), 45-101 2066 L R Klein relationship, a rule should be established for the growth rate of M according to the growth rate of Y, preferably the long run concept of Y A steady growth of M, according to this rule, obviates the need for frequent intervention and leaves the economy to follow natural economic forces This is a macro rule, in the extreme, and the monetarists would generally look for the competitive market economy to make all the necessary micro adjustments without personal intervention The theory for the steady growth of M and Y also serves as a theory for inllation policy, for if the competitive economy maintains long run real income (ZZw,X_ j) at its full capacity level-not in every period, but on average over the cycle- then steady growth of M implies a steady level for long run price (EqiPpi) The monetarist rule is actually intended as a policy rule for inflation control There are several lines of argument against this seemingly attractive policy for minimal intervention except at the most aggregative level, letting the free play of competitive forces the main work of guiding the economy in detail In the first place there is a real problem in defining M,, as discussed already in the previous section Banking and credit technology is rapidly changing The various M, concepts are presently quite fluid, and there is no clear indication as to which M, to attempt to control To choose the most tractable concept is not necessarily going to lead to the best economic policy of any one of Not only are the Mi concepts under debate, but the measurement them is quite uncertain Coverage of reporting banks, the sudden resort to new sources of funds (Euro-currency markets, e.g.), the attempts to live with inflation, and other disturbing factors have lead to very significant measurement errors, indicated in part at least by wide swings in data revision of various M, series If the monetary authorities not know M, with any great precision, how can they hit target values with the precision that is assumed by monetarists? It was previously remarked that policy makers not actually choose values for “ ” and t “G “ Similarly, they not choose values for “M,“ They engage in open market buying and selling of government securities; they fix reserve requirements for specific deposits or specific classes of banks; they fix the discount rate and they make a variety of micro decisions about banking practices In a fractional reserve system, there is a money multiplier connecting the reserve base that is controlled by monetary authorities to M,, but the multiplier concept is undergoing great structural change at the present time, and authorities not seem to be able to hit M, targets well A fundamental problem with either the Keynesian or the monetarist view of formal political economy is that they are based on simple models -models that are useful for expository analysis but inadequate to meet the tasks of economic policy These simple models not give a faithful representation of the economy; they not explicitly involve the appropriate levels of action; they not take Ch 35: Economic Policy Formation 2079 The problem of optimal policy may, in fact, be one of varying constraints, respecifying F It has been found that the problem of coping with stagflation is intractable in the sense that macro policies cannot bring both unemployment and inflation close to desired targets simultaneously On the other hand, there may exist policies that so if the constraint system is modified By introducing a special TIPS policy that ties both wage rates and profit rates to productivity X/hL = real output per worker-hour it has been found that highly favorable simulations can be constructed that simultaneously come close to full employment and low inflation targets These simulation solutions were found with the same (Wharton) model that resisted full target approach using the methods of optimal control The wage and profits (price) equations of the model had to be re-specified to admit Alnw=Aln(X/hL) Aln(PR/K)=Aln(X/hL) PR = corporate profits K = stock of corporate capital Equations for wages and prices, estimated over the sample period had to be removed, in favor of the insertion of these.t5 A creative policy search with simulation exercises was able to get the economy to performance points that could not be reached with feasible applications of optimal control methods This will not always be the case, but will frequently be so Most contemporary problems cannot be fully solved by simple manipulation of a few macro instruments, and the formalism of optimal control theory has very limited use in practice Simulation search for “good” policies, realistically formulated in terms of parameter values that policy makers actually influence is likely to remain as the dominant way that econometric models are used in the policy process That is not to say that optimal control theory is useless It shows a great deal about model structure and instrument efficiency By varying weights in the loss function and then minimizing, this method can show how sensitive the uses of policy instruments are Also, some general propositions can be developed The more uncertainty is attached to model specification and estimation, the less should be the amplitude of variation of instrument settings Thus, William Brainard has shown, in purely theoretical analysis of the optimum problem, that “L R Klein and V Duggal, “Guidelines Wharton Qwrterly, VI (Summer, 1971), 20-24 in Economic Stabilization: A New Consideration,” 2080 L R Klein Growth assumptions Table and budget deficit fiscal policy planning, Estimated deficit ($ billion) administration Congressional Budget Office Baseline Low alternative 1984a 1987 1988 1989 4.0 4.0 4.0 4.0 4.1 3.6 3.5 -0.9 3.5 2.1 3.5 3.8 3.5 3.1 186 192 211 233 241 248 189 196 197 209 217 261 245 329 212 357 308 390 1984 Real GNP estimates or assumptions administration Congressional Budget OtXce Baseline Low alternative USA February 1985 5.3 4.1 5.4 4.9 1986 (W) Fiscal Years “Source: Baseline Budget Projections for Fiscal Years 1985-1989 Congressional Washington, D C February 1984 Testimony of Rudolph G Penner, Committee tions, U.S Senate, February 22, 1984 Budget Office, on Appropria- policy makers ought to hold instruments cautiously to a narrow range (intervene less) if there is great uncertainty l6 This is a valuable advice developed from the analysis of optimal policy A particular case of uncertainty concerns the business cycle The baseline solution for y, should reflect whatever cyclical variation is present in the actual economy if predictions of y, are at all accurate For example, the existence of a cycle in the United States has been well documented by the National Bureau of Economic Research and has been shown to be evident in the solutions of macro econometric models.” Although the baseline solution of a macro economy extending over to 10 years should reflect a normal cyclical pattern unless some specific inputs are included that wipe out the cycle, that is not the usual practice in public policy planning Policy makers are reluctant to forecast a downturn in their own planning horizon The accompanying table illustrates this point in connection with U.S budget planning in early 1984 The official baseline path assumes steady growth of the economy, contrary to historical evidence about the existence and persistence of a 4-year American cycle An argument in support of this practice has been that the exact timing of the cyclical turning points is in doubt If they 16W Brainard, “Uncertainty and the Effectiveness of Policy,” American Economic Review LVIII May 1967), 411-25 See also L Johansen, “ Targets and Instruments Under Uncertainty,” Institute of Economics, Oslo, 1972 Brainard’ results not, in all theoretical cases, lead to the conclusion that s instrument variability be reduced as uncertainty is increased, but that is the result for the usual case “See I and F Adelman, “The Dynamic Properties of the Klein-Goldberger Model,” Econometrica 27 (October 1959) 596-625 See also, Econometric Models of Cyclical Behavior, ed B G Hickman (New York: Columbia University Press, 1972) Ch 35: Economic Policy Formation 2081 are not known with great precision, it is argued that it is better not to introduce them at all An appropriate standard error of estimate is probably no larger than f 1.0 year; therefore, they ought to be introduced with an estimated degree of certainty The Congressional Budget Office in the United States has a fairly steady expansion path for its baseline case, but introduces a cycle downturn for 1986, in a low growth alternative case, between and years after the last downturn It would seem more appropriate to consider this as a baseline case, with the steady growth projection an upper limit for a more favorable budget projection A series of randomly disturbed simulations of an estimated model F = e;i) t =1,2, H i=1,2 ) R, with R replications of random error disturbances, generates estimated equation system F Each replication produces ’ y{i) \ solutions of the ’Zl y2(i) given ,YP, ‘ :f and initial conditions \ZH, The R stochastic projections will, on average, have cycles with random timing and amplitude They will produce R budget deficit estimates The mean and variance of these estimates can be used to construct an interval that includes a given fraction of cases, which can be used to generate a high, low, and average case for budget deficit values The stochastic replications need not allow only for drawings of e,(‘ they can also be used to estimate distributions ); of parameter estimates for F.18 This is an expensive and time consuming way to generate policy intervals, but it is a sound way to proceed in the face of uncertainty for momentous macro problems It is evident from the table that provision for a business cycle, no matter how uncertain its timing may be, is quite important The higher and steadier growth assumptions of the American administration produces, by far, the lowest fiscal deficits in budgetary planning A slight lowering of the steady path (by only 0.5 percentage points, 1986-89) produces much larger deficits, and if a business cycle correction is built into the calculations, the rise in the deficit is very big In the cyclical case, we have practically a doubling of the deficit in five years, while in lXThe technique employed in G S&ink, Estimation of Forecast Error in a Dynamic and/or NowLinear Econometric Model (Ph.D dissertation, University of Pennsylvania (1971)) can be used for joint variation of parameters and disturbances L R Klein 2082 the cycle-free case the rise is no more than about 50 percent in the same time period Also, optimal control theory can be used to good advantage in the choice of exogenous inputs for long range simulations Suppose that values for T + 30 where T + 30 is 30 years from now areneededfort=T+l,T+2,T+3, (in the 21st century) We have little concrete basis for choice of WT+ 30 ZT-C 30’ By optimizing about a balanced growth path for the endogenous variables, with respect to choice of key exogenous variables, we may be able to indicate sensible choices of these latter variables for a baseline path, about which to examine alternatives These and other analytical uses will draw heavily on optimal control theory, but it is unlikely that such theory will figure importantly in the positive setting of economic policy The role of the baseline (balanced growth) solution for policy making in the medium or long term is to establish a reference point about which policy induced deviations can be estimated The baseline solution is not, strictly speaking, a forecast, but it is a policy reference set of points Many policy problems are long term Energy availability, other natural resource supplies, social insurance reform, and international debt settlement are typical long term problems that use econometric policy analysis at the present time At the present time, the theory of economic policy serves as a background for development of policy but not for its actual implementation There is too much uncertainty about the choice of loss function and about the constraint system to rely on this approach to policy formation in any mechanistic way.” Instead, economic policy is likely to be formulated, in part at least, through comparison of alternative simulations of econometric models In the typical formulation of policy, the following steps are taken: (i) definition of a problem, usually to determine the effects of external events and of policy actions; (ii) carry out model simulations in the form of historical and future projections that take account of the problem through changes in exogenous variables, parameter values, or system specification; (iii) estimation of quantitative effects of policies as differences between simulations with and without the indicated changes; “See, in this respect, the conclusions Policy Optimisation, Report, (London: of the Royal Commission HMSO, 1978) (headed by R J Ball) Committee on Ch 35: Economic Policy Formation 2083 (iv) presentation of results to policy decision makers for consideration in competition with estimates from many different sources Policy is rarely based on econometric information alone, but it is nearly always based on perusal of relevant econometric estimates together with other assessments of quantitative policy effects Among econometric models, several Will often be used as checking devices for confirmation or questioning of policy decisions It is important in policy formulation to have a baseline projection For the short run, this will be a forecast of up to years’ horizon For the longer run, it will be a model projection that is based on plausible assumptions about inputs of exogenous variables and policy related parameters For the longer run projections, the inputs will usually be smooth, but for short run forecasts the inputs will usually move with perceptions of monthly, quarterly, or annual information sources in a more irregular or cyclical pattern The model forecast or baseline projection serves not only as a reference point from which to judge policy effects It also serves as a standard of credibility That is to say, past performance of forecast accuracy is important in establishing the credibility of any model Judgmental information, quantitative reduced form extrapolations (without benefit of a formal model) and estimated models will all be put together for joint information and discussion Models are significant parts of this information source but by no means the whole In many respects, model results will be used for confirmation or substantiation of decisions based on more general sources of information Models are most useful when they present alternative simulations of familiar types of changes that have been considered on repetitive occasions in the past, so that there is an historical data base on which to build simulation analyses A new tax, a new expenditure program, the use of a new monetary instrument, or, in general, the implementation of a new policy that calls on uses of models that have not been examined in the past are the most questionable There may be no historical data base in such situations from which to judge model performance In new situations, external a priori information for parameter values or for respecification with new (numerical) parameter values is needed These new estimates of parameters should be supplied by engineers or scientists for technical relations, by legal experts for new tax relationships, or by whatever expertise can be found for other relationships The resulting simulations with non sample based parameter estimates are simply explorations of alternatives and not forecasts or projections Much attention has been paid, in the United States, recently to changes in laws for taxing capital gains There is no suitable sample that is readily available with many observations at different levels of capital gains taxation Instead, one would 2084 L R Klein be well advised to look at other countries’ experience in order to estimate marginal consequences of changing the tax laws for treatment of capital gains In addition, one could investigate state-to-state cross section estimates to see how capital gains taxes might influence spending behavior Similar analyses across countries may also be of some help Finally, we might try to insert questions into a field survey on people’ attitudes towards the use of capital gains These are all s basic approaches and should be investigated simultaneously There is nothing straightforward to in a new situation, but some usable pieces of econometric information may be obtained and it might help in policy formation Recently, claims were made about the great benefits to be derived from the liberalization of capital gains rates in the United States, but these claims were not backed by econometric research that could be professionally defended For the ingenious econometric researcher, there is much to gain on a tentative basis, but care and patience are necessary In all this analysis, the pure forecast and forecasting ability of models play key roles Forecasts are worthwhile in their own right, but they are especially valuable when examined from the viewpoint of accuracy because users of model results are going to look at forecast accuracy as means of validating models It is extremely important to gain the confidence of model users, and this is most likely to be done through the establishment of credibility This comes about through relative accuracy of the forecast Can forecasts from models be made at least as accurately as by other methods and are the forecasts superior at critical points, such as business cycle turning points? These questions are partly answered by the accuracy researches of Stephen McNees and others.20 The answer is that models no worse than other methods and tend to better at cyclical turning points and over larger stretches of time horizon The acceptability of model results by those who pay for them in the commercial market lends greater support to their usefulness and credibility This supports their use in the policy process through the familiar technique of alternative/comparative simulation International policy uses provide a new dimension for applications of econometric models Comprehensive models of the world economy are relatively new; so it is meaningful to examine their use in the policy process The world model that is implemented through Project LINK has been used in a number of international policy studies, and an interpretation of some leading cases may be helpful “Stephen McNees, “The Forecasting Record for the 197Os,” New England Economic Review, (September/October, 1979), 33-53 Vincent Su, “An Error Analysis of Econometric and Noneconometric Forecasts,” American Economic Review, 68, (May, 1978), 360-72 Ch 35: Economic Policy Formation 2085 Some of the problems for which the LINK model has been used are: exchange rate policy, agricultural policy associated with grain failures, oil pricing policy, coordinated fiscal policies, coordinated monetary policies When the LINK system was first constructed, the Bretton Woods system of fixed exchange rates was still in force It was appropriate to make exchange rates exogenous in such an environment At the present time exchange rate equations have been added in order to estimate currency rates endogenously An interesting application of optimal control theory can be used for exchange rate estimation and especially for developing the concept of equilibrium exchange rates Such equilibrium rates give meaning to the concept of the degree of over- or underevaluation of rates, which may be significant for the determining of fiscal intervention in the foreign exchange market In a system of multiple models, for given exchange rates there is a solution, model by model, for (PX) * Xi - (PM) i * Mi = trade balance for i th country ( PX) i = export price Xi = export volume (goods/services) (PM) i = import price Mi = import volume (goods/services) These are all endogenous variables in a multi-model world system The equilibrium exchange rate problem is to set targets for each trade balance at levels that countries could tolerate at either positive or negative values for protracted periods of time- or zero balance could also be imposed The problem is then transformed according to Tinbergen’ approach, and assumed values are given to s the trade balance, as though they are exogenous, while solutions are obtained for ( EXR) i = exchange rate of the i th country The exchange rates are usually denominated in terms of local currency units per U.S dollar For the United States, the trade balance is determined as a residual by virtue of the accounting restraints i i and the exchange rate in terms of U.S dollars is, by definition, 1.0 As noted earlier, this problem, although straightforward from a conceptual point of view, is difficult to carry out in practice, especially for a system as large L R Klein 2086 and complicated criterion as LINK; therefore, it has to be solved empirically from the ~{[(PX)~*x~-(PM)i*M~]-[(PX)j*Xl-(PM),*M,]*}2 i =min=O, with the entire LINK system functioning as a set of constraints The minimization is done with respect to the values of the exchange rates (instruments) With modern computer technology, hardware, and software, this is a feasible problem Its importance for policy is to give some operational content to the concept of equilibrium exchange rate values Optimal control algorithms built for project LINK to handle the multi-model optimization problem have been successfully implemented to calculate Ronald McKinnon’ s proposals for exchange rate stabilization through monetary policy.21 As a result of attempts by major countries to stop inflation, stringent monetary measures were introduced during October, 1979, and again during March, 1980 American interest rates ascended rapidly reaching a rate of some 20% for short term money One country after another quickly followed suit, primarily to protect foreign capital holdings and to prevent capital from flowing out in search of high yields An internationally coordinated policy to reduce rates was considered in LINK simulations Such international coordination would diminish the possibility of the existence of destabilizing capital flows across borders Policy variables (or near substitutes) were introduced in each of the major country models The resulting simulations were compared with a baseline case Some world results are shown, in the aggregate, in Table The results in Table are purely aggregative There is no implication that all participants in a coordinated policy program benefit The net beneficial results are obtained by summing gains and losses Some countries might not gain, individually, in a coordinated framework, but on balance they would probably gain if coordination were frequently used for a variety of policies and if the whole world economy were stabilized as a result of coordinated implementation of policy Coordinated policy changes of easier credit conditions helps growth in the industrial countries It helps inflation in the short run by lowering interest cost, directly Higher inflation rates caused by enhanced levels of activity are restrained ” Ronald McKinnon, Institute for International Peter Pauly and Christian in International MonetaT Reserve Bank), 1985 An International Standard for Monetary Stabilization, (Washington, D C.: Economics), March, 1984 E Petersen, “ An Empirical Evaluation of the McKinnon Proposal” Issues Policy, Project LINK Conference Proceedings, (San Francisco: Federal 2087 Ch 35: Economic Policy Formation Table Effects of coordinated monetary policy, LINK system world aggregates (Deviation of policy simulation from baseline) 1982 1983 1984 106 125 149 1979 Value of world trade (bill $) Volume of world trade (bill $, 1970) OECD (13 LINK countries) GDP growth rate (W) Consumer price inflation rate (%) 1980 1981 15 53 85 4.7 14.4 20.2 22.8 24.7 26.9 1.9 - 0.2 1.9 -0.5 1.0 -0.4 -0.2 0.1 -0.5 0.3 -0.4 0.3 by the overall improvement in productivity This latter development comes about because easier credit terms stimulate capital formation This, in turn, helps productivity growth measured as changes in output per worker A pro-inflationary influence enters through the attainment of higher levels of capacity utilization, but it is the function of the models to balance out the pro and counter inflationary effects Policy is not determined at the international level, yet national forums consult simulations such as this coordinated lowering of interest rates and the frequent repetition of such econometric calculations can ultimately stimulate policy thinking along these lines in several major countries A number of fiscal and exchange rate simulations along coordinated international lines have been made over the past few years.22y23 Prospects Economic policy guidance through the use of econometric models is clearly practiced on a large scale, over a wide range of countries Fine tuning through the use of overall macro policies having to with fiscal, monetary, and trade matters ‘ *L R Klein, P Beaumont, and V Su, “Coordination of International Fiscal Policies and Exchange Rate Revaluations,” Mode&g the International Transmission Mechanism ed J Sawyer (Amsterdam: North-Holland, 1979), 143-59 Coordination of Economic Policies,” Greek H Georgiadis, L R Klein, and V Su, “International Economic Review I (August, 1979), 27-47 Monetary Policy and the World Economy,” L R Klein, R Simes, and P Voisin, “Coordinated Prt%sion et Analyse iconomique, (October 1981), 75-104 23A new and promising approach is to make international policy coordination a dynamic game See Gilles Oudiz and Jeffrey Sachs, “Macroeconomic Policy Coordination among the Industrial Countries” Brookings Papers on Economic Activity (1, 1984) l-64 L R Klein 2088 has been carried quite far, possibly as far as it can in terms of methodological development There will always be new cases to consider, but the techniques are not likely to be significantly improved upon To some extent, formal methods of optimal control can be further developed towards applicability But significant new directions can be taken through the development of more supply side content in models to deal with the plethora of structural policy issues that now confront economies of the world This situation is likely to develop-further along supply side lines The bringing into play of joint Leontief-Keynes models with fully articulated input-output systems, demographic detail, resource constraints and environmental conditions are likely to be important for the development of more specific policy decisions requiring the use of more micro details from models This is likely to be the next wave of policy applications, focusing on energy policy, environmental policy, food policy, and other specific issues It is clear that econometric methods are going to play a major role in this phase of development Appendix: An outline of a combined (Keynes-Leontief) input-output/macro model The first five sectors listed on p 2067 are the components of final demand as they are laid out in the simple versions of the Keynesian macro model, extending the cases cited earlier by the explicit introduction of inventory investment and foreign trade When the Keynesian system is extended to cover price and wage formation, then the production function, labor requirements, labor supply and income determination must also be included These, together, make up the main components of national income Interest income and monetary relationships to generate interest rates must also be included This outlines, in brief form, the standard macro components of the mainstream econometric model The interindustry relationships making up the input-output system round out the total model The flow of goods, in a numeraire unit, from sector i to sector j is denoted as Correspondingly, the total gross output input-output analysis are defined as of j is X, The technical aij = Xi,/Xj and the basic identity n ” Xi= C j=l of input-output Xi,+&= C a,jX,+