IT HAS BECOME a common practice to present the problems of economic policy in terms of a dichotomy of targets and instruments. That is, it is assumed to be the first step in policy formation to assign desired values to certain economic variables (target variables) and the second step to contrive methods of influencing the values of other variables (instrument variables) in such a way as to ensure, so far as possible, the attainment of the desired values of the former. This practice, which is usually exemplified in macroeconomic analysis, may perhaps have originated in the discussion of the Keynesian problem of reconciling internal and external equilibrium in an open economy,1 but was first systematically developed by the econometricians for the purpose of providing quantitative advice about economic policy, and is associated in particular with the name Tinbergen.2 More recently, however, as we shall see, the econometricians, including Tinbergen himself, have tended to move away from this approach, or at least to give it a more sophisticated twist, just at the time when it is becoming more fashionable for literary economists, or even official committees, to apply it, or at least to employ its terminology.
A target variable may be a price (e.g., an exchange rate), a price index (e.g., cost of living), a rate of change of prices, a quantity index (e.g., real gross national product (GNP), real wages, real consumption, or the amount of some category of real expenditure), a rate of change of quantities (e.g., a growth target), a net value aggregate (e.g., the balance of payments, or a budget surplus), a value ratio (e.g., the share of wages in the national income, the ratio of savings or domestic investment to GNP), or an amount or ratio of excess demand or supply (e.g., unemployment). What makes a variable a target variable is the fact that some policymaker wants it to have a particular value either as an end in itself or as a means whose influence on ultimate ends is exercised outside the system of economic relationships explicitly assumed. Instruments are variables which the policymaker is in a position to influence, either directly or through channels that lie outside the system of relationships explicitly assumed, for the attainment of his desired objectives. Instrument variables acquire that status from their association with various types of policies: government expenditures and tax rates from fiscal policy; interest rates, the quantity of money, and the amount of credit extended by banks from monetary policy; prices and wages from policies of wage or price control; and amounts of raw materials and consumer goods demanded from policies of allocation and rationing, respectively.
It is a central deliverance of the targets-instruments approach that the number of instruments must be at least equal to the number of targets. In making this count, any variables that necessarily vary in a fixed relation to each other—though together they may be adjusted by policy decision—or any variables whose relative marginal influence with respect to each target variable is the same should be considered as a single instrument. On the other hand, variables that in all circumstances compatible with the prevailing economic relationships remain at desired levels should not be reckoned as target variables, though in other conditions—e.g., if some change occurred in the economic structure or some new policy instrument were introduced—they might become such.
It is not difficult to understand the truth of the proposition. If we start from a position in which there are no targets and in which potential instrument variables are set at a given level, it is clear that there must be at least as many dependent variables as there are independent equations describing structural relationships. Otherwise, the equations will not, in general, be compatible with one another. To set a target value for a dependent variable (one that may differ from the actual value of that variable) is equivalent to turning a dependent into an independent variable; this cannot be done in general unless some other variable is changed from an independent into a dependent one, or unless some structural relationship is altered by the introduction of a new dependent variable. These conditions, however, are fulfilled when some policy instrument, previously exogenously determined, is allowed to be determined by the system, including the desired values of the targets, or some new policy instrument is activated and allowed to be similarly determined.
If all the structural equations are linear and all instruments are independently and indefinitely variable, then the number of instruments need be no more than equal to the number of targets. But if the authorities are unable or unwilling to vary the value of instrument variables beyond certain limits (boundary conditions), or if they cannot vary some instruments without affecting others, or if the structural equations are nonlinear (e.g., if the influence of instruments on targets is subject to diminishing returns), it may be necessary to have more instruments than targets.
The targets-instruments approach has some advantages. Given its basic assumption—the legitimacy of posing the problem of policy formation in terms of the attainment of predetermined values for a limited number of target variables—the propositions of the theory are true and useful. And, in spite of what is said below, some of the most important macro-objectives of economic policy can be legitimately expressed in terms of targets that are approximately fixed—near-zero unemployment, near-zero rates of change of domestic price levels, and near-zero imbalance in international payments. Moreover, the approach in question explicitly favors a number of healthy tendencies, e.g., restraint on the tendency of noneconomists to multiply economic targets by attributing intrinsic importance to variables whose significance is purely instrumental, ingenuity in the search for novel instruments of policy, and boldness in overcoming irrational inhibitions against the use of familiar instruments. For example, the growing recognition of the difficulties of reconciling the three targets mentioned above has induced the economic profession to mount an attack on the inflexibility of the exchange rate instrument while pursuing with increased urgency the search for new instruments such as wage policy, or the more conscious use, as separate instruments, not only of monetary policy and fiscal policy, as such, but also particular categories of operations within these broad fields. Above all, the targets-instruments approach, by facilitating the application of statistical and econometric techniques, has helped to put an end to the gross errors which at times characterized the qualitative or rule-of-thumb approach to economic policy formation.
However, progress in the arts is generally bought at a price, normally in the form of a loss of older skills, and in the present instance progress in the quantification of macroeconomic policy has tended to involve an undue sacrifice of the insights of microeconomic welfare theory.
The inadequacy of the targets-instruments approach is often at its most marked where any weakness is most important, namely, in the choice of the target variables and in the determination of their target levels. The old-fashioned welfare economist derives all his intermediate objectives from two ultimate criteria, of which the first—the maximization of real income—is determined, apart from such allowances as may be made for market imperfections, externalities, and errors of anticipation, by the preferences, expectations, and decisions of individuals, while only the second—the distribution of real income—is inevitably a matter for central evaluation. And even here the welfare economist will tend to guide the policymaker toward a rational evaluation by presenting the problem in terms of the effects of income redistribution on the welfare of classes of individuals. The targets-instruments economist, on the other hand, is apt to regard all his specific targets as data, more or less arbitrarily handed down by the policymaker. It is perhaps an acknowledgment of the arbitrary character of targets arrived at in this way that Tinbergen will casually suggest dropping a target when there are too few instruments to permit all targets to be fully attained.
Often it is the national government or political party that appears to be regarded as the appropriate chooser of targets. Such humility on the part of economists is out of place. Left to themselves, governments are likely to choose target values which, in view of the constraints of the system and the shortage of effective instruments, are incompatible—sometimes rather directly incompatible, as in price stability and a predetermined rate of growth of real income. Moreover, they are very apt to assign end value to variables that have no direct connection with ultimate human needs but are believed, on the strength of some dubious but probably quite complex economic theory, to exercise an influence on economic welfare. For this reason, variables, such as exchange rates, interest rates, or budget balances, that ought to be either treated as primarily instrumental or else left to the determination of the market will be regarded as targets and thereby removed from the category of instruments, or irrational limitations will be put upon their use.
Admittedly, one of the main objects of a quantitative confrontation of targets and instruments is to bring home to the policymaker the fact that his chosen target values are mutually incompatible, given the existing constraints, and to get him to give up some of his target variables, to alter his target values, or to overcome his inhibitions regarding the use of certain instruments. However, such a trial-and-error process is unlikely to be the quickest way of reaching a set of realistic objectives, and the objectives ultimately arrived at by this route, even though mutually compatible, are likely to be less closely related to any rational concept of economic welfare than if the economist had from the start advised his government about its proximate economic goals and policies on the basis of an economic welfare function.
Of course, governments, since they have the power and the votes, must—and should—have the last word in determining objectives. But there is some reason to think that they are sometimes stimulated by the prevailing fashion among econometricians and economists into unnecessary targetry.
Lethargy will often prevent governments from choosing targets unless they are provoked into doing so, and some of the targets they entertain they do not pursue with any seriousness. Unfortunately, however, although the explicit message of the targets-instruments approach is to limit the number of targets, its psychological impact—its demonstration effect—is to make government officials target conscious—to convince them that they are sinning against the light if they do not quantify their national economic goals in as much detail as possible. For example, in the domestic sphere many governments are induced to aim not merely at over-all high employment and price stability but at such in every major industry and region of the country. The politician cannot be expected to realize, if the economist does not explain to him, that these aims, however desirable in themselves, cannot easily be realized without impairing the capacity of the economy to adapt to changing needs.
In the international sphere, similar excesses in targetry are not lacking. For example, the otherwise cautious report on the adjustment process by the Organization for Economic Cooperation and Development (OECD)3 encourages countries to quantify their ambitions with respect not merely to their over-all balances of payments but also to their current and capital accounts taken individually. The national targets would then be reconciled by a process of mutual confrontation. This procedure would be unexceptionable if each country were to fix the target level for its current account balance of payments in relation to estimates of what the medium-term trend of its capital account would be at full employment in the absence of capital restrictions. Even then, it would be necessary (1) to coordinate national targets for the over-all balance of payments so as to fit in to the expected growth in world reserves, and (2) to coordinate national estimates as regards the trends in capital movements. What the OECD Working Party seems to have in mind, however, is for each country to prepare a target for its capital account (and hence its current account) reflecting national preferences. The criteria underlying these national preferences being unspecified, it is difficult to guess on what basis the international coordination or mutual adjustment of national targets could take place. And even if countries were able to agree on adjusted targets, one could be fairly certain that if the targets were achievable at all it could only be by methods contrary to an optimal distribution of capital resources.
While the multiplication of unnecessary targets may be said to represent a popular misapplication of the targets-instruments approach, a number of other weaknesses are inherent in the approach itself. Their root cause lies in the essential characteristics of the approach, namely, (1) the representation of social preference by fixed predetermined values of the target variables rather than by a function of such variables. From this follows (2) the attribution of end value to too few of the variables of the system of relationships under examination, and in particular (3) neglect of the end value inherent in the instrument variables themselves. A further consequence is (4) failure to appreciate that the object of policy is to achieve the best compromise among objectives and not to maximize the number of objectives that are fully attained. Finally, there is a weakness which flows partly from the method as such—in which context it is particularly related to (3) above—and partly from the oversimplified macroeconomic character of the economic models to which the method is applied, namely, (5) an undue neglect of the evils of misallocation, as distinct from unemployment, of resources.
Any rational policymaker who is guided by considerations of social welfare must be able to rank the alternative situations that might conceivably result from the application of different policies in order of preference. Among these conceivable situations some will turn out to be feasible. The object of the exercise is to arrive at the most preferred of these feasible situations.
If the welfare significance of all alternative situations that could result from different values of the instrument variables of a given economic model are regarded as determined by the values, in these situations, of certain of the variables of the model (target variables), then, since there is a vast number of possible combinations of instrument values, there will be a vast number of possible situations to be put in order of preference. Moreover, since no two situations are likely to be characterized by the same values of all of the target variables, it is reasonable to assume that no two situations will be graded precisely alike from a welfare standpoint. Such an ordering can conveniently be represented by a continuous function of which the target variables are the arguments.
Tinbergen’s fixed targets do not fit into this picture. The welfare ordering which they imply is one in which each possible situation is ranked according to the number and identity of target variables having prescribed values (i.e., fixed targets) which it contains. Of these possible situations, there is (given the instruments) a subset of feasible situations, each containing a different set (though possibly the same number) of target variables and hence of fixed targets. Of these, the situation with the preferred fixed targets is to be chosen.
Clearly no one would believe in the validity of such an odd type of preference function in and for itself. If Tinbergen uses it, it may be to some extent because politicians, not being particularly rational, like having fixed targets, but primarily because, given the possible sets of fixed targets and the empirical relationships connecting the variables, it is relatively simple to demonstrate mathematically whether and at what values of the instrument variables each set of targets is feasible.
As is indicated below, the econometricians have found it possible to work with welfare functions of a much more rational type. In any event, literary economists who do not aspire to arrive at the optimal value of instrument variables at one blow through the solution of a set of simultaneous equations do not have the same excuse for adopting a fixed-targets approach.
When the problem of policy is regarded as one of maximizing a welfare function of which certain economic variables (which we may continue to call target variables) are the arguments, rather than one of attaining the maximum number of the most preferred fixed targets, it no longer becomes necessary to limit the number of target variables in relation to the number and nature of the instrument variables that are available. All the variables appearing in the model that either affect welfare directly or influence the value of variables outside the model that in their turn affect welfare directly can be included as arguments in the welfare function. It will not be possible to attain predetermined values for these variables, nor yet to attain the values for each of these variables that would be optimal if they could be varied freely without affecting or being affected by the other variables in the model. But it will be possible to find values for all of them which, given the constraints of the model and the available instruments, are the best that can be attained.
Since most of the variables in any simple macroeconomic model will have repercussions on variables outside the model, most of them should probably be included as arguments in the welfare function, though the weights that should be attached to them will, of course, vary. Among these variables are likely to be found at least some of the instrument variables. In the fixed-targets approach, there has to be a complete dichotomy between targets and instruments. Since the values of the target variables are fixed in advance while the values of the instrument variables must be free to vary, at least to some extent, the latter can never be included among the former. With the welfare-function approach the necessity for such segregation disappears.
One reason for including instrument variables among the arguments of the welfare function is that their manipulation frequently (1) absorbs resources and (2) distorts or (more rarely) improves their allocation, in ways which are rarely taken explicitly into account in simple macroeconomic models. For example, a tax imposed for the purpose of checking price inflation is likely, even if it takes the form of a perfectly general tax on consumption, to create some disincentive to effort and enterprise, and, in a national tax, to encourage outward and discourage inward tourism. If it is an expenditure tax that affects investment, or if it is an income tax, it will, in addition, unduly discourage saving and further penalize enterprise. If, as is more likely, it is an indirect tax of restricted application, it will direct resources unduly to untaxed activities. On the other hand, it may sometimes be possible, given the prior existence of price rigidities and market imperfections, to impose a tax, say for anti-inflationary purposes, that will at the same time improve the allocation of scarce commodities. In either event, unless—as is rare—measures of misallocation are themselves included as variables in the model, it will be desirable to include the instrument variable as one of the arguments of the welfare function.
Almost any imaginable economic model contains more variables, including instrument variables, that should in terms of any reasonable welfare function be classified as target variables than it contains instrument variables. For this and other reasons already discussed, it will be impossible, even where the target variables—apart from the constraints of the model—have finite optimal values, fully to attain the optimal values of all target variables.
In the case of a complete micromodel, this is very obvious. Even if all prices were fully flexible, and fully competitive conditions prevailed throughout the economy, the attainment of an equitable income distribution would necessitate a system of budgetary transfers which would be bound to distort incentives to work, to save, and probably also to carry on particular productive activities. The same results will ensue from efforts to meet through the budget collective needs of various kinds, to promote an adequate scale of activity in industries of increasing return, or to adjust incentives to take account of the various sorts of externalities. Where wage and price rigidities and other distortions are introduced into the working of the competitive system by the action of economic entities which are large in relation to the relevant markets and which the state cannot or will not control directly, it will be still more difficult to bring target variables, which in this case might take the form of price-cost relationships and relative personal incomes, into the vicinity of their optimal values.
Even in macromodels, however, full attainment of targets is likely to be impossible. In this event, the best feasible result, given the constraints of the model, will be one in which all target variables continue to be treated as such (i.e., none are discarded) and in which a compromise is achieved between the various targets—few, if any, being fully attained. This is particularly true of target variables whose optimal value is not at some natural limit of variation and whose marginal utility in the vicinity of the optimum is, therefore, very low. For example, in the sphere of public finance it is better to compromise between the objective of equalizing the utility of the marginal dollar spent on public services and on private consumption and the objective of avoiding the disincentive effects of taxation rather than to give one objective absolute priority over the other. Again, in the sphere of demand policy it is better to compromise between full employment and avoidance of inflation rather than to give one objective complete priority over the other.4
Some target variables, such as the quantity of consumption in general, have no finite optimum value, apart from the constraints of the system, and full attainment of the target in such cases is obviously impossible. Nevertheless, they have constrained optima. These lie at the point at which the marginal utility of increasing the variable in question is equal to the marginal cost in terms of loss of utility from reducing other target variables of a similar character or of removing other target variables having a finite optimum further from their optimal levels.
To a devotee of welfare economics, the most striking inadequacy of the fixed-targets approach lies in its handling of questions relating to the allocation of resources. Admittedly, many of the problems of allocation arise in the microeconomic sphere with which the macroeconomic models, to which the targets-instruments approach is usually applied, do not and need not deal, since the problems in question can in principle be dealt with by instruments other than those that enter into the macroeconomic models. However, a number of allocation problems concerned, for example, with the degree of restriction or encouragement to be applied to international transactions, the disincentive effects of taxes, the appropriate levels of public savings, consumption, and investment, or the diseconomies of inflation do arise in a macroeconomic context—very often, as already indicated, in connection with instrument variables. It is much more difficult to devise plausible fixed targets for desiderata in this sphere than in, say, the degree of employment or the balance of payments where, under any reasonable welfare function, the optimal values of the variables in question would lie within a narrow range. It is where variables affecting the allocation of resources are concerned that the constraints of the system call most clearly for compromise solutions and where the necessity, under the fixed-target approach, either of adopting arbitrary target values or of ignoring the significance of variables is most blatant. Thus, in some policy models import restrictions may be banned entirely while in others their undesirable effects are ignored. In some models, budget imbalances are banned while in others their implications for national savings and for the balance between private and public consumption are ignored. And in many cases arbitrary boundary conditions are put on the use of instruments while diseconomies of use within these limits are ignored.
The flexible-targets approach, principally developed by Professor Theil5 but also discussed at an early date by Tinbergen, side by side with the fixed-targets approach,6 though deriving from the fixed-targets approach and bearing traces of its parentage, represents a considerable improvement on the latter, and is free from a number of the objections to it that have been made above. This serves as an encouraging demonstration that the weaknesses that have been pointed out are not inevitable features of quantitative treatment of economic policy problems.
The flexible-targets approach treats social welfare explicitly as a continuous function that is to be maximized subject to the constraints of the model. Instrument variables are included among the variables that are the arguments of this function, and it is emphasized that there is no need for the number of instruments to equal the number of target variables, and that the best feasible situation may be one in which none of the target variables is dropped and none attains its optimal value. This approach is, therefore, free of objections (1), (2), (3), and (4) leveled at the fixed-targets approach on page 392 above.
It is greatly to be regretted that though the econometricians had introduced this approach, and contrasted it with the fixed-targets approach, almost as early as they introduced the latter,7 literary economists who talk about targets and instruments have usually interpreted targets in the sense of fixed targets, as can be seen in the emphasis they lay on comparing the respective numbers of the two classes of variables.
Even in its flexible form, however, the approach of the Netherlands school of econometricians to the problem of optimal policy formation retains several of the characteristic weaknesses of the fixed-targets method. This is true despite the great sophistication of Theil’s work. Here, once more, the root of the trouble lies in the form of the social welfare function that is generally adopted, namely, a particular type of quadratic function. Welfare is regarded as the negative of a weighted sum of squares of the values of the target variables, when these are measured as deviations from their optimal values. Maximum feasible welfare is attained when the weighted sum of squared deviations is minimized, subject to the constraints of the model. Both the optimal values and the weights are fixed in advance. This implies that each target variable has a predetermined optimal value such that (given the values of other variables) any deviation from that value will reduce welfare, that it is a matter of indifference whether that deviation be positive or negative, and that as the optimal value is approached the marginal gain to welfare from a further approximation to the optimum declines.
Now, as was pointed out by Tinbergen himself in his original treatment of flexible targets,8 a distinction has to be drawn between those target variables such as income ratios among classes of income receivers, production ratios among commodities, payments balances, employment ratios, or, one might add, rates of price inflation, where a finite optimum value is relevant, and those, such as real per capita income, where there is no finite saturation point. Predetermined fixed optimal values cannot possibly apply to the latter. Theil, however, gaily sets optimal values for such variables as the levels of consumption and of investment which are clearly of the insatiable type.9 He is, of course, well aware of the obvious objections, but thinks that if his optimal values are chosen with sufficient art (in relation to what is feasible, given the constraints) his preference function will approximate true preferences well enough, over the relevant range. What this appears to mean is that the relative optimal values of different target variables must be set before the calculation is made sufficiently, but not too much, in excess of what the best feasible values will be after the calculation is made to ensure (1) that the absolute marginal utilities of such variables are all positive, and (2) that the relative marginal utilities are what they ought to be. This procedure seems to me so circular as to be valueless except as a first step in a (hopefully convergent) process of mutual adjustment of optimal and best feasible values.
A less serious difficulty arises where, as in the employment ratio, for example, the target variable has a finite optimum that lies at or near the limit of variation. In such a case, if the best feasible level tends to exceed the optimum, the target variable can be replaced by a fixed target.
Parenthetically, it may be observed that whether a variable has or has not a finite optimum will depend not only on the nature of the variable but on the presence or absence of other variables in the system. For example, if only one form of resource use appears among the variables in the equations describing the system—whether it be the amount of income going to one sector of the community, or the amount of output of some particular commodity—there will be some finite optimal level of that use; but if all the competing uses of resources appear as variables in the system, the optimal level for any one use will be infinity. One might go further and say that wherever target variables with finite optima appear in a system they are proxies for other variables of the insatiable type, which, for good or bad reasons, do not appear in the system. Thus, for example, the employment ratio is a proxy for output and, at a further remove, for real income. Rate of price inflation and the savings ratio derive their importance from their presumed effects on future real income, though the chain of causation, particularly in the first case, may be so complex and uncertain that a complete model into which all these relationships would enter explicitly is impracticable.
Even where the target variable is such as to admit of some finite optimum level that is not at the limit of possible variation, it will seldom be precisely a matter of indifference whether the value of the variable exceeds or falls short of its optimal level by a given amount. This is clearly true, for example, of the unemployment ratio, or of the balance of payments.
The least-squares welfare function is open to the further objection that the predetermined optimal values of its target variables and the fixed coefficients by which deviations from these optima are weighted imply that the marginal welfare curves of each of the target variables are independent of the actual values of the other variables. Now this would clearly not hold, for example, in a microsystem in which the supplies of all final products appeared explicitly as arguments of the welfare function; on the contrary, the marginal welfare yield of each product would be a function of the supply of all the others, particularly of substitutes and complements. In some macroeconomic models the value interrelationships between target variables may be less important, though it has to be remembered that such factual interactions as may exist between target variables other than those arising from the explicit constraints of the model should be allowed for in the welfare function. Take, for example, a model in which the target variables are unemployment and import restrictions (expressed in terms of an equivalent ad valorem tariff), and in which optimal unemployment (taking account of its effect on the trend of prices) is positive, and optimal import restriction is zero. In this instance, the optimal level of either target variable will be virtually independent of the actual level of the other. Moreover, the weight to be assigned to deviations from optimal unemployment depends on the interindustrial distribution of unemployment and is probably not greatly influenced by the degree of import restriction. However, the weight to be assigned to deviations from zero import restriction depends primarily on the elasticity of demand for imports, which probably does depend significantly on the degree of unemployment.
There seem to be two main reasons why the least-squares welfare function, in spite of its many and obvious faults, has been used by Theil and other econometricians in their work on economic policy. The first is connected with the fact that the constraints of the system, representing the factual relationships between economic variables, are of a stochastic character. They can only be established, by econometric methods, with some degree of uncertainty, which is expressed by the inclusion of a random variable. Now the least-squares welfare function is one of a class of quadratic welfare functions the expectations of which will be maximized (thus arriving at the probably best feasible outcome) when the stochastic linear equations of the system are replaced by nonstochastic linear equations having the same coefficients, in which the random variables are replaced by their expectations or most probable values. More rational welfare functions would probably not have this convenient property.
This, however, appears to be a case of the metric tail wagging the economic dog. Whatever errors might be introduced into the maximization of the expectation of any given welfare function by taking the random variables in the constraints at their most probable values would surely be small by comparison with the errors that would result from treating the random variables in this way and, to quiet one’s statistical conscience, altering the welfare function that one is setting out to maximize.
Another probable reason for adopting the least-squares type of welfare function, however, is that it retains the notion of predetermined quantitative targets which the policymaker can be persuaded to choose, or led to believe that he has chosen. In fact, this impression is even more illusory for flexible than for fixed targets since the best feasible values of the target variables—which is what matters in the end—will depend to a considerable extent on the weighting system adopted in the welfare function—a system the implications of which the policymaker may find it difficult to comprehend. Moreover, as we have seen in the case of variables without a true finite optimum, the choosing of optimal values—in the light of preconceptions as to what is feasible—calls for an art which the policymaker is unlikely to possess.
It is, of course, much easier to point out the faults in any given social welfare function—such as the least-squares function—than to suggest a satisfactory alternative. If the target variables, as in a micromodel, were all of the kind that have no finite optima, a welfare function of the Cobb-Douglas type might be appropriate, and would have the advantage of not assuming independent utilities. In most models, however, some of the target variables, at any rate, would have finite optima, and the required conditions for the use of such a function would not be fulfilled.
It may be possible for the mathematicians to devise types of welfare functions that could be used to accommodate target variables, some of which have, while others have not, a finite optimum. The writer is in no position to offer suggestions in this regard; and, doubtless, even the form of the function should vary with the nature of the problem—the particular economic model—under examination. What seems to him desirable is that any welfare function should be derived, by however intuitive a procedure, from estimates of the probable effects of variations in the target variables on real income, appropriately discounted for futurity and corrected, if alternative policy mixes are likely to have substantially different relative effects on particular groups in the community, for differences in the marginal utilities of income of the groups in question.
It is not, however, the ambition of this paper to enter at all deeply into the problems of the quantitative global planner. The concern of the writer is much more with the impact of the targets-instruments conceptual scheme on the work of the economic advisor who is concerned with the relative merits of alternative policies for effecting an improvement in—rather than an optimization of—the economic situation. From this standpoint, the criticisms of the flexible-targets approach offered in the immediately preceding pages are really much less important than those advanced in earlier pages against the fixed-targets system, for the simple reason that it is the concepts of the latter rather than of the former that have seeped down from the econometric stratosphere to the less rarified atmosphere in which most workaday economists draw breath.
To these the message of this paper is as follows: (1) Keep your eye at all times on the probable effects of different policies on real income (corrected where necessary for distributional effects). (2) Do not multiply proximate targets without need, but, equally, do not drop legitimate objectives simply because you do not have sufficiently numerous or sufficiently effective policy instruments to attain them all. (3) In particular, do not be hypnotized by the alleged importance of attaining a few fixed targets, often of a macroeconomic character, into forgetting the real income effects of misallocations of resources and distortions of demand. (4) Realize the inevitability of second-best or compromise solutions to policy problems.
It may be worthwhile adding a postscript on the question of the appropriate number of instruments that the policymaker should seek to use. Both the fixed-targets approach and the welfare-economics approach tend to favor a multiplication of policy instruments, the former so as to increase the number of targets that can be attained and the latter so as to permit all objectives to be more closely approximated. Since, as we have seen, it will in general not be feasible to attain the optimal values of such of the target variables as have finite optima, and since some of the target variables may be insatiable, having no finite optima, it follows that, if no costs were involved in the use of instruments, all potential instrument variables should be used to some extent. This looks like a charter for interventionism, but the qualification is important. In fact, all instruments involve some costs in use. These costs are of two kinds—costs in terms of labor and capital, and costs in terms of policy management. The latter are analogous to those costs that, after a point, give rise to diminishing returns to scale in an individual enterprise. It is necessary that policies be centrally coordinated, and in each country there is a limit to the number of policies that can be successfully coordinated by the political and administrative machine. For this reason, the costs of applying any given policy instrument will depend not only on the degree of its use but on the number and nature of the instruments already in use. The existence of both of these kinds of cost, and particularly the latter, will set a limit on the number of policy instruments that can appropriately be brought into operation.
Objectifs et instruments de politique économique
La méthode qui consiste à envisager les problèmes de politique économique en considérant séparément les objectifs et les instruments employés pour les atteindre, d’abord utilisée par les économétriciens qui l’ont conçue, a été adoptée ensuite par les économistes moins férus de chiffres, et même par les milieux officiels. La conception erronée que I’on a généralement de cette pratique tend à encourager une prolifération inutile d’objectifs définis quantitativement, qui sont souvent sélectionnés de façon arbitraire. Si elle bien conçue, la méthode de “l’objectif fixe”, telle qu’elle a été définie à l’origine par les économétriciens — étant entendu que, pour qu’elle soit pleinement efficace, le nombre des instruments doit être au moins égal au nombre des objectifs — peut être extrêmement utile et constituer un excellent facteur de discipline. Elle présente, néanmoins, un certain nombre d’inconvénients du point de vue de l’économie du bienêtre. En effet, elle ne permet pas de représenter de façon satisfaisante les gradations des préférences sociales, elle encourage à sélectionner un nombre trop restreint d’objectifs quantifiables, et conduit, en particulier, à ne pas tenir compte d’une part de considérations de répartition des ressources, et d’autre part de la nécessité de concilier tant bien que mal plusieurs objectifs.
Le concept des “objectifs flexibles”, énoncé plus récemment par les économétriciens, constitue, à cet égard, un progrès considérable par rapport à la méthode de l’objectif fixe, mais il n’a pas, d’une façon générale, été bien accueilli par les économistes “littéraires”, qui continuent de lui préférer l’ancien concept de l’objectif fixe. Toutefois, on peut aussi reprocher à la méthode “flexible” d’orienter la politique économique de façon à obtenir des valeurs aussi proches que possible de valeurs prédéterminées des objectifs. Ceci est incompatible à plusieurs égards avec toute fonction de bien-être social jugée acceptable sur le plan économique.
Metas e instrumentos
La práctica de considerar a los problemas de política económica en términos de una dicotomía de metas e instrumentos ha partido de los econometristas, que la han extendido a otros economistas menos dados a los números y asimismo a los círculos oficiales. Tal como se la suele interpretar equivocadamente, tiende a fomentar una multiplicación innecesaria de metas definidas cuantitativamente y que a menudo se escogen sobre una base arbitraria. Entendiéndolo correctamente, el enfoque de metas fijas en la forma presentada originalmente por los econometristas puede ser de utilidad y valor disciplinario considerables, al recordarnos que el número de instrumentos tiene que ser por lo menos igual al número de metas para poder alcanzarlas con plenitud. Sin embargo, está sujeto a una serie de inconvenientes desde el punto de vista de la economía del bienestar, al no permitir una representación adecuada de las situaciones de preferencia social, y al favorecer una insistencia demasiado absoluta en un número demasiado restringido de objetivos cuantificables, lo que conduce, en especial, al olvido tanto de las consideraciones sobre asignación de recursos como de la necesidad de hallar un compromiso entre los objetivos.
El concepto de metas flexibles, creado más recientemente por los econometristas, representa un avance considerable al enfoque de metas fijas en los aspectos anteriormente mencionados, pero en general no ha alcanzado popularidad entre los economistas literarios, que continúan adheridos al anticuado concepto de las metas fijas. Pero aún este enfoque flexible queda abierto a la crítica por supeditar la política a la consecución de la aproximación más cercana posible a valores predeterminados de los objetivos. Esto está en desacuerdo, en varios sentidos, con toda función del bienestar social que parezca aceptable desde el punto de vista de la economía del bienestar.
Mr. Fleming, Deputy Director in the Research Department, is a graduate of Edinburgh University. He was formerly a member of the League of Nations Secretariat, Deputy Director of the Economic Section of the U.K. Cabinet Offices, U.K. representative on the Economic and Employment Commission of the United Nations, and visiting Professor of Economics at Columbia University. He is the author of numerous articles in economic journals.
As set forth, for example, in J. E. Meade, The Theory of International Economic Policy, Volume 1: The Balance of Payments (Oxford University Press, London, 1951).
J. Tinbergen, On the Theory of Economic Policy (Amsterdam, 1952), and Economic Policy: Principles and Design (Amsterdam, 1956).
The Balance of Payments Adjustment Process, A Report by Working Party No. 3 of the Economic Policy Committee of the Organization for Economic Cooperation and Development (Paris, August 1966).
For employment, it may be argued that the unconstrained optimum—apart from the links between employment and inflation—is at the natural limit of full employment. However, the fact that full employment brings with it the diseconomies of labor shortage—probably not included as a separate target variable —may mean that the optimum is somewhat short of absolutely full employment, or at least that the marginal utility of employment declines in the vicinity of full employment. Even if this were not so, the marginal diseconomies of inflation at full employment might be so great that a compromise between the two objectives was desirable.
H. Theil, “Econometric Models and Welfare Maximization,” Weltwirtschaftliches Archiv (Hamburg, 1954); Economic Forecasts and Policy (second edition, Amsterdam, 1961); Optimal Decision Rules for Government and Industry (Amsterdam, 1964); “Linear Decision Rules for Macrodynamic Policy Problems,” Quantitative Planning of Economic Policy, ed. by Bert G. Hickman (The Brookings Institution, Washington, 1965), pp. 18-37.
Tinbergen says that “this somewhat more general treatment was suggested to me by Professor Dr. J. B. D. Schouten and Professor Dr. H. Theil,” Centralization and Decentralization in Economic Policy (Amsterdam, 1954), p. 7. See also Tinbergen, Economic Policy: Principles and Design (Amsterdam, 1956), and the chapter on ‘The Theory of the Optimum Regime” in Jan Tinbergen: Selected Papers, ed. by L. H. Klaassen, L. M. Koyck, and H. J. Witte-veen (Amsterdam, 1959), pp. 264-304.
Possibly even earlier. A flexible approach is implicit in linear programing as applied to a whole economy.
Tinbergen, Centralization and Decentralization in Economic Policy (cited in footnote 6), p. 52.
Theil, Optimal Decision Rules for Government and Industry (cited in footnote 5), p. 80.