Articles THE THEORY OF AGENCY: THE POLICING "'PARADOX'" A N D REGULATORY BEHAVIOR Barry M. Mitnick Relations of agency, or "acting for," are pervasive in complex societies. Examples include the worker-boss, physician-patient, adviser-administrator, and parent-child relations. Functional dependencies, among other reasons, determine that agency relationships will be extremely common. A key problem that principals in such relations face is that of insuring that the agent does in fact act for the principal. This paper will present a model of agency with policing and apply it to discuss regulatory behavior. The policing "paradox" to be described is only an apparent, ascribed paradox, evident mostly in the short-run, and due generally to information effects. It occurs to observers of policing because of lack of knowledge about the goal state and information level of policing participants, and about the dynamic characteristics of the policing process. By identifying it we wish to emphasize policing as a process, with definite stages, rather than compress the argument in the way the usual holistic economic approach would proceed. The policing arguments are developed in the context of the theory of agency, presently under development (see Mitnick 1973, 1974, 1975; for the economic theory of agency, see Ross 1973, 1974; see also Goldberg 1973, 1974; for a related model of policing in the context of the firm, see Alchian and Demsetz 1972), and managerial discretion models (see, e.g., williamson 1964; Migue and Belanger 1974). We have chosen a fiduciary function representation (see below), derived from Williamson (1964), rather than the budget-output line employed by Migue and *Assistant Professor of Public Administration and Political Science, Ohio State University.
28
PUBLIC CHOICE
Belanger (1974) to depict discretionary resources. This is partly because we seek a general agency model of policing, wishing to examine returns to agent and principal directly rather than use the surrogate "output" for principal returns, and partly in order to be able to compare agent and principal returns directly. Both Williamson and Migue and Belanger have noted, as we do, that income and substitution effects may occur in managerial discretion models. The sources and circumstances of these effects in each case, however, are different. Williamson (1964)describes the effects of changes in tax rates in each of his models. Migue and Belanger note the effects of changes in demand, cost, and demand and cost elasticities on the budget-output function. The present paper discusses policing effects using the fiduciary function representation. (See also Mitnick 1973, 1974.) We begin by describing fiduciary relations in the context of agency.
L Agency and Fiduciary Relations The agent holding the fiduciary norm must act diligently, with the skills at his disposal, for the principal's goal, without regard for any other goals that may bear . . . . . on his relation with the principal, including any self-go a ls. 1 The norm may be expected in contractual discretionary agency, typically under conditions of trust, under principal dependency, or under agent domination of the principal's interests. The fiduciary function, relation (in the mathematical sense), or set will be said to be the set of collections of specifications of the returns to each of the agent's goals, self- and other-, that are subject to the possible choice of the agent, given his contractual arrangement with the principal. We may also term this set the institution function, relation, or set since the possibly policed formal norms or rules of an institution, 2 perhaps an organization, may constrain the set of selections available to the agent. The pure fiduciary will choose that collection of specifications which contains the highest returns to his other-goals, regardless of the level of return to his self-goals. The lexicographic, or "lexical" fiduciary will after choosing the collections with the highest return to the principal, choose that collection which also has the highest return to his self-goals. These types may be contrasted with types of agents who choose for their self-goal first: The pure 1See Mitnick (1975). The arguments below regarding the policing "paradox" and regulatory behavior are from Mitnick (1974). Note that the usage of "fiduciary" is different from that employed by Curry and Wade (1968), who use it for entrepreneurs conceived as agents for groups, and do not require tt/e restriction on consideration of competing, including self-, interests. General usage (see e.g., Riker 1962, pp. 24-28; Pitkin 1967; Seavey 1964) agrees with the sense here employed. A self-goal is an objective of self-regarding preferences; an other-goal is an objective of other-regarding preferences. Here "self-regarding preferences" relate to concerns that are private, personal, egoistic, selfish, self-bettering, and so on; "other-regarding preferences" relate to such concerns of the "other" party. 2Blake and Davis (1964, p. 464) note that "norms clustered around a given functional requirement are often coUectively designated as 'institution'."
AGENCY
29
self-interest agent chooses that collection of specifications which contains the highest returns to his self-goals, regardless of the level of return to his other-goals. The lexical self-interest agent chooses, first, those collections of specifications which contain the highest returns to his self-goals and, second, from that group of collections, that collection which has the highest return to his other-goals. 3 We posit that policing may involve a general developmental pattern (allowing skips in stages) from the pure self-interest agent to the lexical self-interest agent to the lexical fiduciary and finally to the pure fiduciary.
IL Diagrammatic Exposition of Agency with Policing Assume that the agent and principal are rational, have preferences that reflect "greed," i.e., "more is always better," and that the agent possesses a genuinely preferred self-goal and a genuinely preferred other-goal (the goal of the principal). Assume that the agent's indifference curves between "a" (return to self-goal as measured in "resources" devoted to it) and " p " (return to other-goal as measured in "resources" devoted to it) may be constructed, are continuous and that they have negative slope. Assume also that the slopes of the indifference curves at constant "a" are increasing (absolute value decreasing) as we reduce "p," Le., that " p " is more highly valued with respect to "a," the less of " p " there is. Another way of lookin~ at this is to note that this implies that " a ' " s self-interests are not inferior goods. We could, of course, assume that they are inferior goods (e.g., the manager who insists on a certain level of discretionary return of his perrequisites or emoluments, let us say, as the store of discretionary resources shrinks, no matter what the resulting possible return to the owner's goal of profit), since such cases are not uncommon. For illustrative purposes, however, we assume the normal goods case.
Assume that the agent administers some quantity of these resources, subject to his discretion, which in continuously varying amounts may be either devoted to return to the principal's goal, diverted to the agent's self-goal, or "lost." The agency relation involves the supply or conversion of this resource to increase the return to the principal's goat. Assume that only the agent can convert this resource or otherwise supply it so that it increases return to the principal's goai. The principal will expect some minimal level of performance from his agent, that is, minimal ievel of resources devoted to return to his goal, below which the principal will seek to discharge his agent. Above this level, Pi' the agent has discretion. Here "discretion" means that the agent may choose to perform acts that affect return to the principal's goal. The agent has a similar minimum level, ai, below which he would 3For a r g u m e n t s on the comparative behavior o f these four types o f agents, see M~tnick
(1974, 1975). 4See Williamson (1964, p, 47). The model of the policing process discussed here was suggested by Williamson's treatment of managerial discretion and Alchian's comments on Witliamson. See Alchian (1971), The approach in no way depends on the theory of the firm, however, or on its origin in economics except as the assumptions are similar.
30
PUBLIC CHOICE
quit as agent. The agent's discretionary choice of returns to his self- and to his other-goal are constrained by a feasibility set of possible combinations of agent-principal receipts, the fiduciary function, p(a). Although we represent it as a function and as a function only of "a," the true fiduciary set may consist of the region below p(a) and depend on other factors, here controlled, besides "a." The assumption of greed, however, makes only the outer boundary of this region relevant, and this is what is indicated by p(a). The shape of the fiduciary function may depend in an organization, for example, on organizational rules, technology, and so on. We assume that through the range of study the agent's preferences will be independent of the fiduciary function. This may of course not be true in general; what is preferred is frequently determined by what one can get. Where the slope of the fiduciary function is positive, an increase in return to the agent's self-goal also implies an increase in return to the principal; where this slope is negative, increase in return to the agent's self-goal means a decrease in return to the principal.
P
I II
ill
iii
I
I
I
II
iiiii
Ua
iiiiiiiiii
I
I
I
I
Pl p(a) Pi
Figure I
a
Consider the fiduciary function depicted in Figure 1. If the agent is a pure fiduciary, he will select the greatest feasible return to the principal's goal, and will be indifferent regarding the resultant return to any self-goals that may be affected by the agency relation. His indifference curves will be horizontal lines rather than negatively sloped, and he will choose the level of return Pl to the principal's goal. He Mll be indifferent between the returns to his self-goal at points A and B and between. If, however, the agent is a lexicographical fiduciary, he will select from among those points that equivalently maximize return to the principal that point that yields the highest return to his self-goals, i.e., point B. For the general case where the agent has genuine preferences for both a self- and an other-goal that are not constrained in choice by a fiduciary or other norm, the agent's indifference curves are negatively sloped and determine an "operating point" at the "highest"
AGENCY
31
point of tangency between the indifference curves and the fiduciary function, as in Figure 2 (point A). Note that in Figure 2 the tangency occurs below the point of maximum return to the principal (point B). (See Williamse-1 1964; Migue and Belanger 1974.) This operating point is allowable only if we assume that the cost to the principal of policing his agent so that he selects the point of maximum principal return exceeds the value of the principal of the additional return that the policing could obtain. (See Alchian 1971; Atchian and Demsetz 1972.)
P
U0 p(o)
Figure 2 For the purpose of developing a simple model here, assume that p(a) is linear and downward-sloping~ i.e., that constant additions to return to the self-goal mean constant reductions in return to the other-goal of the agent. This is similar to Williamson's emoluments model (1974, pp. 49-52), where we will consider lump sum withdrawal o f resources rather than a proportional tax as the first policing stage. Assume that the principal for some reason wishes to police his agent. Note that one form of policing would be encouraging the agent to hold the fiduciary norm; we will not, however, consider here the actual mode of policing. Assume that the principal has no outside source of resources to devote to a policing mechanism. He must then divert some of the total discretionary resources potentially available to be distributed towards his own and the agent's ends into a policing apparatus. Note that we have assumed that only the agent may supply or convert the resource so that it increases return to the principal. Alternatively, at this first stage, we may simply consider the effect of withdrawing some part of the discretionary resources without assuming the creation of a policing apparatus. Any withdrawal of resources, or conversion of discretionary to nondiscretionary resources, whether or not policing is intended, is relevant here. Thus, after resource withdrawal but before inauguration of the policing u n i t - or before that unit begins to have any
32
PUBLIC CHOICE
effect - we have the situation depicted in Figure 3. For convenience in notation, let P-Pi be represented by "p," and a-a i by "a."
P
Pt
Pmo
Oi
Ot°
Ore° Ot
Or" 0
Figure :5 The policing cost in resources is Pm-Pt" The operating point before is (amo, Pmo) and after, (ato , Pto)' ato < a m o because the assumption that "a" is not an inferior good. (In the economic formalism, something similar is called the "income aP m a + Pm" Substituting values before m and after resource withdrawal, and then subtracting:
effect.") The fiduciary function is p -
Pm
Pmo
=
-
a~
a
mo
+
Pm
m
Pm am
Pto = -
a
to + P t
(Pmo - Pto ) = - p--m(aa mo - ato) + (Pm - Pt ) m
Pm
(Pm . Pt.) . (Pmo . . Pto )
a (amo - ato) m
(i)
AGENCY
or
(Pm - Pt ) Pm
(Pmo - Pto ) Pm
33
(amo - ato) am
(2)
Eq. (1) suggests the following general law for fiduciary functions of this type: The net gain in agent fidelity is directly proportional to the difference between the discretionary resources withdrawn and the resultant difference between the principal's operating points. Eq. (2) states this in terms of proportionate costs and losses relative to total potential discretionary return to each party. Furthermore, Pm given the assumed preference structure of the agent, if 5---,the absolute value of the m slope of the fiduciary function, is >~ 1 (and it may be some undetermined amount less), that is, if the potential resources that can be realized by the principal exceeds that which could be realized by the agent, the resources withdrawn to be devoted to policing costs (if that is where they go) will always exceed the immediate gain in agent fidelity as measured in resource loss to the agent. For aP2 sufficiently small (that is, linear functions sufficiently more horizontal), however, the resource toss to the agent may exceed the potential resource cost to the principal. Of course, we have not indicated (nor do we know) how the agent and principal will value their respective loss and cost. It is clear that in the limit of some sufficiently large resource withdrawal the principal may obtain a perfectly fiduciary agent, given "room" above the minimum to operate. We note that the theorem above may be subject to empirical study. According to the above assumptions, a principal may under some specified conditions exchange discretionary resources for an agent acting with greater fidelity to the principal's goal. That is, (amo , Pmo) > (ato ~ Pto)" This is a paradoxical situation in the short run, since the principal has voluntarily yielded valued resources, and we may ask under what conditions would a rational principal do this? Clearly, if the principal has another goal that is satisfied by having a truer agent, he may prefer to do this, i.e., if he has a goal of agent fidelity. The principal may also do this if he expects a net gain as the result of expected future receipts. Only an observer's lack of information about the fact he is observing a stage in a process, and about the principal's full goal set, makes the situation appear paradoxical, of course. Thus we consider now stage two of the policing process: the policing apparatus is in operation; the fiduciary function is altered as the policing unit changes the permissible combinations of resources to the agent and resources to the principal, that is, changes the operating distributive rules. The situation is indicated for the above restrictive model in Figure 4. The policing device alters the slope of the fiduciary function; the greater the success of the device, the more steeply negative that slope becomes. That is, the policing device makes "'a" relatively more expensive to the agent with respect to
34
PUBLIC CHOICE
Pm
Uo Pto
p(a)
aso
ato as
amoat
Figure 4
am o
Thus, given the same overall level of "satisfaction," i.e., the same indifference curve, " p " is substituted for "a". (In the economic formalism, something similar is called thd "substitution effect.") 5 In Figure 4, we see that the fidelity of the agent is further increased (loss of ato - aso resources to the agent), and the principal has a gain of Pso - Pro over stage one. The net effect over stages one and two (policing device funded out of total discretionary resources and in operation) is that the principal exchanges operating point (amo , Pmo) for point (aso , Pso)" Now although the agent's receipts will always be tess than at the start under the present model, the receipts by the principal may be either less than or greater than (or, of course, equal to) receipts at the start. Which of these results obtains depends on the efficiency of the policing device-as indicated by the slope of the fiduciary function - and the shape of the indifference curves of the agent. If Pso ~ P m o ' we have the policing "paradox" noted above. Again, the principal may be coming out ahead because he also has a goal of agent fidelity, or, again, the principal may have a future reward in mind, which we will consider under stage three. If Pso ~ Pmo' the principal has 5Note that the source of this substitution effect is different from those of the effects discussed in WiIliamson (1964) and Migue and Belanger (19 74), involving as it does changes in the fiduciary function due to organizational changes resulting from policing.
AGENCY
35
succeeded in taxing the agent not only for paying the total cost of policing, but also into contributing to the principal's receipts. Now, aso - ato represents the agent's response to policing. Given that the agent is rational, he will clearly yield resources only until the point at which it is no longer rational to do so. Thus the expected value of the negative sanctions of the policing device (including net loss in return to the agent's self-goal) must be greater than or equal to the value of the agent of aso - ato. The value of aso - ato may at that point be the expected value of the loss to the agent if the sanctions are applied times the probability that the sanctions will in fact be applied. For example, although the value to the corrupt policeman of the loss of career plus criminal penalties may be very high, he may perceive that the probability of catching him is very low. It may thus be rational for him to respond minimally to the policing attempt. In stage three, the principal has either succeeded in training his agent to value the principal's goal more highly, perhaps through the policing device, or, if there are multiple agents, replaced those who value "p" very low with agents who value it highly. Thus, for example, the few "rotten apples" in the police department - men with indifference curves with slopes everywhere much more negative (more steeply downward sloping than the average) - have been "weeded out." We then have for our simple model the situation in Figure 5.
°1
Pm ~
Ps
ili Pro
Pt
Oro (]so OtoO=
xt
u" )(o)
(ltamo
Figure 5
a
36
PUBLIC CHOICE
The tangency of the new indifference curves U 2a is further to the left " The new operating point is (aro, Pro)' where aro < aso < ato < amo, and Pro > P s o > Pto' but Pro either < Pmo o r / > Pmo" After stage three the principal may in fact have finally realized his delayed net gain; or he may be left in the situation of the policing "paradox." As above, he may, of course, have a net gain because of the value he places on the increased fidelity of his agent. Note again that the rational principal would not have elected to police his agent i£ he did not expect a net gain from the attempt. Factors such as error due to uncertainty regarding the effectiveness of the policing mechanism in obtaining the results intended and regarding the choices of the agent, and other factors, may intervene. In addition, the principal may have imperfect information regarding the fiduciary function (which may involve the terms of contract between agent and principal) and the preferences of his agent in various situations. Or the bounded rationality (see March and Simon 1958, Chapter 6; Simon 1957) of the principal and agent may limit their ability to calculate the probable outcomes from electing any action. Or factors relating to strategic interaction between the parties may affect the result at any stage. If the agent at stage three has a hostile reaction to policing, 6 his preferences may shift the opposite way, so that the tangency would be further to the right. The principal would then have a net loss between stages two and three. In order to rationally elect policing, the principal must therefore be able to predict any agent reactions of this type. We note an additional factor that may be relevant if the policing process occurs over some time. The principal may have to discount his expected return at each stage due to the delay in receiving it. If discounts of this type are required, the principal will demand higher net returns than under the short term or static case. This may necessitate a better or more efficient policing mechanism. Note also the importance of the fiduciary norm under policing: the fiduciary polices himself. This economizes on policing costs, including costs attached to metering the effects of the policing mechanism and reporting this feedback to the principal. (el. those cited by Kaufman 1973.) Thus savings (-to whoever pays such costs - most likely, the principal) occurs on both policing costs and specification costs - the costs of choosing agent acts. In conclusion, we have formulated the policing process in three stages: 1) diversion of resources to policing or other uses; 2) implementation of policing mechanism; and 3) agent's reaction to policing. These stages are, of course, very broad and would require substantial elaboration in a reasonably complete theory of agency. A sense of "paradox" may be experienced by an observer of the policing process because of lack of knowledge regarding the fact that 1) the principal may have a hidden or nonobvious goal, and/or 2) the principal may err, having poor information due to the factors noted above (the principal and agent may also engage in strategic interaction with a paradoxical result occurring because each party may possess poor information on the ultimate effects of their joint 6On reaction, see, e.g,, Brehm (1966), Day and Hamblin (1964), WiUiamson(1973).
AGENCY
37
behaviors), and/or 3) the principal may be in the midst of a policing process and expect future returns, the mechanism and dynamics of which are not obvious to the observer.
IIL The Policing "Paradox" and Regulatory Behavior Under conditions of "organizational slack" (see Cyert and March 1963), the managers of an organization may be viewed as possessing some Fuced quantity of discretionary resources to dispose of. (See williamson 1964). In the firm, for example, revenue that is discretionary may be devoted to increments of profit to the "guardian" stockholders or to increased emoluments or staff that benefit the managers. Similar arguments can be developed for some administrative agencies (cf. Migue and Belanger 1974; Niskanen 1971; Weatherby 1971). In particular, the regulatory agency can be modelled in this fashion. We assume that the discretionary resources o f the agency may go either toward extra satisfaction of some public interest criteria 7 through their indicators, e.g., careful examination of rate applications, or toward increasing the commissioner's status, easing his workload, and insuring the likelihood for him of lucrative future employment, where these rewards may be offered largely by the regulated industry, and may be attained by commission activities that favor the industry, g The commissioner's preferences reflect a trade-off between choice of return to self-goals, as represented by activities favoring the industry, and return to public interest criteria. We assume, however, that the commissioner will act more like a self-interest agent than a fiduciary. We expect that, through some range, return to public interest criteria and to self-goals will be instrumental to each other; past this range, self-goals and public interest criteria will be in antinomy. The commissioner enjoys some discretionary return for discretionary return to the public interest, and vice versa; he does gain some added status, for example, for added efforts in behalf of the public interest. But his return from activities generally favoring industry (and against the public interest) soon dominates his sources of return to self-goals. Note that, in the early range, some extra activities favoring the industry may also be in the public interest. The fiduciary function is therefore shaped something like that in Figure 2. 7These may be viewed as a set of criteria, aggregated by some calculus into a single dimension. This is obviously difficult to do, if only because of problems of measurement and of operationalization. The welfare tradeoffs involved could require better definition of social values than is now available. In principal, however, such tradeoffs could be established and such a dimension constructed; decisions, after all, are made on such matters all the time. But because of such analytical problems, the subsequent discussion should be thought of as a "gedanken experiment." Employment of the arguments we offer here as explanation does not require that we actually construct operational indices of the variables considered, but only acceptanee of the assumption that this could in principle be done. 8For arguments regarding the goals of regulators, see e.g., De Alessi (1974), Eckert (1973), Mitnick and Weiss (1974), Noll (1971).
38
PUBLIC CHOICE
The fixed supply of discretionary resources may include such factors as the commissioner's supply o f incentives to distribute within his agency to produce behavior toward one of these ends, i.e., reflect the garnard (1938) "economy of incentives" model, and the commissioner's disposition of that portion of his own time not already appropriated by certain commission activities. Public interest activist groups may successfully divert part of the discretionary overhead into extensive litigation. 9 I f the model developed above is applicable, a policing " p a r a d o x " may occur. Because of diversion of resources alone, the first stage of the policing model, we may observe the short run "paradox": The commissioners, in re-distributing their reduced supply of discretionary resources, may act with greater fidelity to the given public interest criteria, i.e., devote less return to industry goals that are instrumental to commissioner self-goals. But the return to the public interest, involving, for example, extensive delays in litigation, may be less than previously. The agency may even be effectively paralyzed, unable to dispose of the cases before it. Once the diverted resources have been converted into a functioning policing mechanism, perhaps through the adoption o f agency regulations requiring use of agency resources to consider public interest questions, the return to that interest may increase. Since the discretionary resources are limited in amount, this involves re-distribution from industry ends instrumental to commissioner goals to public interest ends. It is possible, as we noted in the section above on the policing "paradox," that the increased return to the public interest criteria may stilI be below its original level. The added delay in litigation may still offset, for example, the gain in respect for public interest criteria reflected in agency decisions. In time, the commissioners may be "educated" to respect and prefer the public interest group position more, and simuhaneously to prefer the goals whose return is mediated by the regulated industry less. The policing mechanism constitutes a shift in the incentive system that may thus, over time, produce 9Migue and Belanger (1974, p. 46) note that "politicians and through them other parties in the decision-making process can reduce the bureaucrats' margin of discretion." This may occur through constraining their budget. But we note that constraints on true discretionary budget may have sources other than constraints on total budget, Migue and Belanger assmne an instantaneous response to an increase in demand: the budget line rises. They assume that demand leads to a rise in discretionary budget. But it is perhaps more likely that in the short run discretionary budget cannot be increased in response to demand; there is a lag due to the political process of approving a new budget. In the short run it is possible, then, that increased demand will simply lead to reduction in the discretionary budget as more of the total budget is consigned to satisfy sharply increased demand. (This could conceivably be represented in the Migue and Belanger model by an increase in minimum "cost.") The fall of the budget line under these conditions may he similar in effect to the policing effect we des6ribe. Note that extensive litigation may involve both increased cost for the bureau (e.g., more rules to be made and followed) and increased demand (e.g., more rules and dec!sior~s are demanded as caseload increases), which would reduce discretionary budget according to our argument above. In either case, according to this modification of the Migue and Belanger model, the budget line would fall, giving rise to policing effects.
AGENCY
39
genuine changes in preferences. The commissioners may follow a developmental pattern, changing from self-interest agents to lexical self-interest agents to lexical fiduciaries and finally, if unlikely, into pure fiduciaries. The commissioners may also discover that the new incentive system permits them to satisfy their old preferences for self-goals through other means, e.g., through the public interest groups. This would allow them to remain essentially pure self-interest agents (through lexical self-interest could occur'if choosing in this way were consistently instrumental to self-goals). They may find, for example, that status accrues to the position of noble defender of the public interest, and that this reputation could be converted into remunerative future employment. Because of turnover in commissioners, the new incentive system may tend to "recruit" commissioners whose preferences are satisfied by the new system better than those of the old commissioners. At any rate, in the long run the "paradox" may resolve itself. Because the "paradox" may involve severe degradation in the actual, net return to the public interest, a question is in order about the rationality of public interest groups that challenge the commissions. The rational principal attempts policing only if he expects a net return, discounted for the intervening adjustment time when his return is below the original level. We suggest that because of bounded rationality and information costs, not to speak of fundamental problems in valuation, the return to the given public interest criteria may be more difficult to measure and predict than certain obvious indicators wrongly assumed to be correlated with return to that interest. (Cf. Niskanen's comments, Migue and Belanger 1974, p. 43.) And these indicators are then relied on in actions that public interest groups consider very rational indeed. 10 A major indicator is the degree of apparent commissioner fidelity, i.e., how much of discretionary agency resources he diverts to his self-goals, perhaps through the instrumental means of appearing in his conduct to favor the regulated industry. Does the commissioner-agent appear to be a self-interest agent or fiduciary, with gradations between? Note that the real measure of agent fidelity is the difference between what he appropriates for his self-goals and the level of discretionary return to agent self-goals at which discretionary return to the principal is maximized (see points A, B in Figure 2). Since the latter may be unknown, the focus inevitably shifts to simply reducing discretionary return to the agent. If by some chance the organization is already operated to give maximum discretionary return to the principal, then reduction in agent return will result in a reduction in principal return irrespective of any policing "paradox." Public interest groups that object to this "necessary" level of discretionary agent return, e.g., comfortable offices, will thus be acting rather directly against their interest. If, given the information that this is so, such groups persist in such behavior, we would assume either that they are irrational, which is not likely, though possible, or that what they really prefer is 10A1chian and Demsetz (1972) note in their model of the firm that the difficulty of monitoring marginal productivity leads to the metering of inputs. This result is clearly related to our own and Niskanen's (in " C o m m e n t " on Migue and Belanger 1974) arguments.
40
PUBLIC CHOICE
the appearance of agent fidelity and not the maximization of return to the public interest. In addition, the legal and other weapons that such groups may use can frequently be directed only against evidence of agency manager fidelity, rather than to insure some given return to the public interest. For example, a group may use litigation to force an agency to prepare an environmental impact statement, or to force an agency that prepared only a pro forma statement to put some content into it. But the group generaiIy cannot litigate to determine that content, ke., to insure the level of return to the interest they represent that they prefer. Thus even if a group focuses on return to the public interest rather than on fidelity it may only be able to act on commissioner infidelity, hoping the outcome will not be perverse. At the heart of the policing "paradox" is the observation that agent fidelity in the sense used here does not necessarily correlate with level of principal return. By focusing on apparently improper behavior by the commissioners, such as reliance on the industry for helpful advice, "ex parte" contacts with he industry that may involve exchange of inside information, field visits and spending time with industry representatives rather than with commission staff or "impartial" experts, allocating discretionary resources within the agency to engineering sections with close ties to the industry rather than, say, environmental evaluation sections, and so on, the public interest groups may succeed only in securing an honest agency that doesn't regulate, and, ultimately, a righteous government that cannot govern. IV. Conclusion
In conclusion, we have developed a model of policing in the context of agency relations and managerial discretion. The model had three stages: 1) diversion of resources to policing or other uses; 2) implementation of policing mechanism; and 3) agent's reaction to policing. We then applied the model to the case of regulatory behavior. We argued in part that public interest groups are constrained (and perhaps in some cases may elect) to police the manifestations of agent fidelity in the regulatory agencies rather than adherence to public interest criteria. This has possibly paradoxical consequences in that return to public interest criteria may thereby be reduced.
AGENCY
41
REFERENCES Alchian, Armen A. "The Basis of Some Recent Advances in the Theory of Management of the Firm." Reprinted on pp. 131-139 in W. Breit and H. M. Hochman (eds.), Readings in Microeconomics, Second Edition. New York: Holt, Rinehart and Winston, 1971. Alchian, Armen A. and Harold Demsetz. "Production, Information Costs, and Economic Organization." Am. Econ. Rev., 62 (December 1972), 777-795. Barnard, Chester I. The Functions of the Executive. Cambridge, Massachusetts: Harvard University Press, 1938, 1968. Blake, Judith and Kingsley Davis, "Norms, Values, and Sanctions." Pages 456-484 in Robert E. L. Faris (ed.), Handbook of Modern Sociology. Chicago: Rand McNally, 1964. Brehm, Jack. A ~l~eory of Psychological Reactance. New York: Academic, 1966. Curry, R. L., Jr. and L. L. Wade. A Theory of Political Exchange. Englewood Cliffs, New Jersey: Prentice-Hall, 1968. Cyert, Richard M. and James G. March. A Behavioral Theory of the Firm. Englewood Cliffs, New Jersey: Prentice-Hall; 1963. Day, tk C. and R. L. Hamblin. "Some Effects of Close and Punitive Styles of Supervision." Am. Jour. of SocioL, 69 (March 1964), 499-510. De Alessi, Louis. "An Economic Analysis of Government Ownership and Regulation: Theory and the Evidence from the Electric Power Industry." Public Choice, 19 (Fall 1974), 1-42. Eckert, Ross D. "On the Incentives of Regulators: The Case of Taxicabs." Public Choice, 14 (Spring 1973), 83-99. Goldberg, Victor P. "Consumer Choice, Imperfect Information, and Public Policy." IGA Research Report No. 26 (Davis, California: University of California, Institute of Governmental Affairs, August 1973). • "Regulation and Administered Contracts." Unpublished paper (October, December 1974). Kaufman, Herbert. Administrative Feedback. Washington, D.C.: The Brooldngs Institution, 1973. March, James G. and Herbert A. Simon. Organizations. New York: Wiley, 1958. Migue, Jean-Luc and Gerard Belanger. "Toward a General Theory of Managerial Discretion." Public Choice, 17 (Spring 1974), 27-47, including "Comment" by William A. Niskanen and "Reply" by Migue and Belanger. Mitnick, Barry M. "Fiduciary Rationality and Public Policy: The Theory of Agency and Some Consequences." Paper delivered at the 1973 Annual Meeting of the American Political Science Association, New Orleans, Louisiana. • The Theory of Agency: The Concept of Fiduciary Rationality and Some Consequences. Unpublished doctoral dissertation, Departm6nt of Political Science, University of Pennsylvania (1974). . "The Theory of Agency: The Fiduciary Norm." Paper prepared
42
PUBLICCHOICE
for the 1975 Annual Meeting of the American Sociological Association, San Francisco, California. Mitnick, Barry M. and Charles Weiss, Jr. "The Siting Impasse and a Rational Choice Model of Regulatory Behavior: An Agency for Power Plant Siting." Jour. Environ. Econ. Mgt., 1 (1974), 150-171. Niskanen, William A., Jr. Bureaucracy and Representative Government. Chicago: Aldine-Atherton, 1971. Noll, Roger G. Reforming Regulation. Washington, D.C.: The Brookings Institution, 1971. Pitkin, Hanna Fenichel. The Concept of Representation. Berkeley: University of California Press, 1967. Riker, William H. The Theory of Political Coalitions. New Haven: Yale University Press, 1962. Ross, Stephen A. "The Economic Theory of Agency: The Principal's Problem." Am. Econ. Rev., 62 (May 1973), 134-139. "On the Economic Theory of Agency: The Principle of Similarity." In Proceedings of the NBER-NSF Conference on Decision Making and Uncertainty, forthcoming. Seavey, Warren A. Handbook of the Law of Agency. St. Paul, Minnesota: West, 1964. Simon, Herbert A. "A Behavioral Model of Rational Choice." In Simon, Models of • Man: Social and Rational. New York: Wiley, 1957. Weatherby, James L., Jr. "A Note on Administrative Behavior and Public Policy." Public Choice, 11 (Fall 1971), 107-110. Williamson, Oliver E. The Economics of Discretionary Behavior: Managerial Objectives in a Theory of the Firm. Englewood Cliffs, New Jersey: Prentice-Hall, 1964. "Some Notes on the Economics of Atmosphere." Fels Discussion Paper No. 29 (Philadelphia: Fels Center of Government, University of Pennsylvania, March 1973).