An Economic Theory of Political Action in a Democracy Anthony Downs The Journal of Political Economy, Vol. 65, No. 2. (Apr., 1957), pp. 135-150. Stable URL: http://links.jstor.org/sici?sici=0022-3808%28195704%2965%3A2%3C135%3AAETOPA%3E2.0.CO%3B2-F The Journal of Political Economy is currently published by The University of Chicago Press. Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/ucpress.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org. http://www.jstor.org Mon Feb 25 07:40:47 2008 AN ECOSOMIC THEORY OF POLITICAL ACTION IN A DEMOCRACY1 ANTHONY DOWNS Chicago, Illinois I I N SPITE of the tremendous importance of government decisions in every phase of economic life, economic theorists have never successfully integrated government with private decision-makers in a single general theory. Instead they have treated government action as an exogenous variable, determined by litical considerations that lie outside the purview of economics This view is really a carry-over from the classical prenlise that the private sector is a self-regulating mechanism and that any government action beyond maintenance of law and order is "interference" with it rather than an intrinsic part of it." However, in at least two fields of economic theory, the centrality of government action has forced economists to formulate rules that indicate hoLVgovernment "should" make decisions. Thus in the field of public finance, Hugh Dalton states: .ISa result of [the] operations of public finance, changes take place in the amount antl in the nature of the Lvealth which is pl.oduceti, antl in the distribution of that Lvealth among individuals and classes. Are these changes in their aggregate effects socially advantageous? If so theoperations are justified; if not, not. The The argument presented in this article will be tleveloped further in 111y forthcon~ingbook, .4t1 Economic I'lzeory of Democracy, to be published by Harper & Bros. 2 See Gerhard Colm, Essays i n Public Finance and Fiscal Policy (Xew 'T'ork: Oxford University Press, 1955),pp. G8. best system uf public finance is that which secures the maximum social advantage from the operatiolls \vhich it collducts.~ A similar attempt to differentiate the operations "proper" to government from those "proper" to private agents has been made lV. Peck, who writes: "If public operation of an enterP ~ ~ ~ " greater~ ~ ~ socialproduce a net utility, the services rendered by this enterprise in the category Of public goods."' In addition, several welfare economists have posited general principles to guide government acti0n in the 'Or Abba P. Lerner indirectly states such a rule when he says: "If it is desired to maximize the total satisfaction in a society, the rational procedure is to divide income on an basis.''" Admittedly, this list of examples is not long, primarily because Overt 'tatements of a decision rule to guide government action are extremely rare in economic theory. However, it does not unduly distort reality to state that most welfare economists and many public finance theorists implicitly assume that the "proper" function of government is to maximize social welfare. Insofar as they face the problem of government de- 3 The Principles of Public Finance (London: George Koutledge & Sons, Ltd., 1932), pp. 9-10. 'Taxatiott and Welfare (New York: i\Iacmillan Co., 1925), pp. 30-36, as quoted in Harold M. Groves (ed.), Viewpoints i n Public Finance (New York: Henry Holt & Co., 1948), p. 551. The Economics of Control (New York: Macmillan Co., 1914), p. 32. 135 cision-making at all, they nearly all subscribe to some approximation of this normative rule. The use of this rule has led to two major difficulties. First, it is not clear what is meant by "social welfare," nor is there any agreement about how to "maximize" it. In fact, a long controversy about the nature of social welfare in the "new welfare economics" led to Kenneth Arrow's conclusion that no rational method of maximizing social welfare can possibly be found unless strong restrictions are placed on the preference orderings of the individuals in ~ociety.~ The complexities of this problem have diverted attention from the second diflculty raised by the view that government's function is to maximize social welfare. Even if social welfare could be defined, and methods of maximizing it could be agreed upon, what reason is there to believe that the men who run the government would be motivated to maximize it?To state that they "should" do so does not mean that they will. As Schumpeter. one of the few econonlists who have faced this problem, has pointed out: I t does not follow that the social meaning of a type of activity will necessarily provide the motive power, hence the explanation of the latter. If it does not, a theory that co~ltents itself with an analysis of the social enti or need to be served cannot be accepted as an adequate account of the activities that serve it.' Schumpeter here illuminates a crucial objection to most attempts to deal with government in economic theory: they do not really treat the government as part of the division of labor. Every agent in the division of labor has both a private motive and a social function. For ex- 6 Social Choice and Indiuidual Values (hTew York: John Wiley & Sons, 1951). 7 Joseph A. Schumpeter, Capitalisnt, Socialism, and Democracy (New York: Harper & Bros., 1950), p. 282. ample, the social function of a coal-miner is reinoving coal from the ground, since this activity provides utility for others. But he is motivated to carry out this function by his desire to earn income, not by any desire to benefit others. Similarly, every other agent in the division of labor carries out his social function primarily as a means of attaining his own private ends: the enjoyment of income, prestige, or power. hluch of economic theory consists in essence of proving that men thus pursuing their own ends may nevertheless carry out their social functions with great efficiency, at least under certain conditions. In light of this reasoning, any attempt to construct a theory of government action without discussing the motives of those who run the government must be regarded as inconsistent with the main body of economic analysis. Every such attempt fails to face the fact that governments are concrete institutions run by men, because it deals with them on a purely normative level. As a result, these attempts can never lead to an integration of government with other decision-makers in a general equilibrium theory. Such integration demands a positive approach that explains how the governors are led to act by their own selfish motives. In the following sections, I present a model of government decision-making based on this approach. In building this model, 1 shall use the following definitions: 1. Govertzrneuit is that agency in the division of labor which has the power to coerce all other agents in society; it is the locui of "ultimate" power in a given area.8 *This definition is taken from Robert A. Dahl and Charles E. Lindblom, Politics, economic^, dr~d Welfare (New York: Harper & Bros., 1953), p. 42. However, throughout most of my analysis the word "government" refers to the governing party rather than the institution as here defined. 2. :l dcrtto~racyis a political system that exhibits the following characteristics: a) Two or more parties compete in periodic elections for control of the governing apparatus. 11) The party (or coalition of parties) winning a majority of votes gains control of the governing apparatus until the next election. r ) Losing parties never attempt to prevent the \rinners from taking office, nor do ~rinnersuse the powers of office to vitiate the ability of losers to compete in the next election. d) .Ill qane, law-abiding adults who are governed are citizens, and everr citizen has one and only one vote in each election. Though these definitions are both somewhat ambiguous, they will suffice for present purposes. Kext I set forth the following axioms: 1. Each political party is a team of men 1% ho seek office solely in order to enjoy the income, prestige, and power that go with running the governing a p p a r a t u ~ . ~ 2. The winning party (or coalition) has complete control over the government's actions until the next election. There are no votes of confidence between elections either by a legislature or by the electorate, so the governing party cannot be ousted hefore the next election. Iior are any of its orders rcsisted or sal~otaged by an intransigent bureaucracy. 3. (;ovc:rnmcnt's economic powers are unlimitcti. I t can nationalize everything, hand everything over to private interests, or strike any balance between these ex- tremcs. 4. 'I'he only limit on government's powers is that the incumbent party cannot in any \yay restrict the political freedom of opposi".\ ''team" is a coalition whose memhers have identical goals. .\ "coalition" is a group of men ~vho co-operate to achieve some common end. 'l'hese (letinitions are taken from Jacob Narschak, "l'o~vards an Econorr~icTheory of Organization and Information," in Decision Processes, ed. R.11. 'Thrall, C. H. Coonil~s,and R. I,. Davis (New York: John \Vile). & Sons, 1954), pp. 188-89. I use "team" instead of "coalition" in my definition to eliminate intraparty power struggles from consideration, though in hlarschak's terms parties are really coalitions, not teams. t ~ o npartic< or of ~nd~vitlual u n l ~ s \C ~ I I L ~ I I S , they seek to overthrow it by force. 5. Every agent in the model-whether a n individual, a party or a private coalitionbehaves rationally a t all times; that is, it proceeds toward its goals nith a minimal use of scarce resources and undertakes only those actions for ivhich niarginal return cxceeds marginal cost.'O From these definitions and axioms springs my central hypothesis: political parties in a democracy formulate policy strictly as a means of gaining votes. They do not seek to gain otiice in order to carry out certain preconceived policies or to serve any particular interest groups; rather they formulate policies and serve interest groups in order to gain oflice. Thus their social function-which is to formulate and carry out policies when in power as the government-is accomplished as a by-product of their private motive --which is to attain the income, power, and prestige of being in office. This hypothesis implies that, in a democracy, the government always acts so as to maximize the number of votes it will receive. In effect, it is an entrepreneur selling policies for votes instead of products for money. Furthermore, it must compete for votes with other parties, just as two or more oligopolists compete for sales in a market. IIhether or not such a government maximizes social welfare (assuming this process can be defined) depends upon how the competitive struggle for power influences its behavior. 1I-e cannot assume a priori that this behavior is socially optimal any more than we can assume a priori that a given firm produces the socially optimal out- put. *" ?'he tern1 "rational" in this article is synoriytnous with "efficient." This economic definition must not be confused with the logical definition (i.e., pertaining to logical propositions) or the psgchological definition (i.e., calculating or unemo- tional). 138 AX'rHONY DOWNS I shall examine the nature of government decision-making in two contexts: (1) in a world in which there is perfect knowledge and information is costless and (2) in a world in which knowledge is imperfect and information is costly. The analysis of government decisionmaking in a perfectly-informed world is intended only to highlight the basic relationship between a democratic government and its citizens. This relationship can be stated in the following set of propositions : 1. The actions of the government arc a function of the Fvay it expects voters to vote and of the strategies of its opposition. 2. The government expects voters to vote according to (a) changes in their utility incomes from government activity and (h) the strategies of opposition parties. 3. Voters actually vote according to ( 0 ) changes in their utility incomes from government activity and (h) the alternatives offered by the opposition l1 4. Voters' utility incomes from government activity depend on the actions taken by government during the election period. 5. The ~tratcgiesof opposition parties depetitl on their views of the voters' utility income5 from government activity and on the actions taken b>-the government in power. These propositions actually form a set of five equations containing five unknowns: expected votes, actual votes, opposition strategies, government actions, and individual utility incomes from government activity. Thus the political structure of a democracy can be viewed in terms of a set of simultaneous equations similar to those often used to analyze an economic structure. fl In a perfectly informed world, voters al\va)s vote exactly the way government expects them to, so the relationships expressed in Nos. 2 and 3 are identical. But in an imperfectly informed uorld, the government does not alnays know what voters will do; hence Nos. 2 and 3 may differ. Because the citizens of our model democracy are rational, each of them views elections strictly as means of selecting the government most beneficial to him. Each citizen estimates the utility income from government action he expects each party would provide him if it were in power in the forthcoming election period, that is, he first estimates the utility incomeParty A would provide him, then the income Party B would provide, and so on. He votes for whatever party he believes would provide him with the highest utility income from government action. The primary factor influencing his estimate of each party's future performance is not its campaign promises about the future but its performance (luring the period just ending. Thus his voting decision is based on a comparison of the utility income he actually received during this period from the actions of the incumbent party and those he believes he would have received had each of the opposition parties been in power (I assume that each opposition party has taken a verbal stand on every issue dealt with concretely by the incumbents) This procedure allows him to found his decision on facts rather than on conjectures. Of course, since he is helping to choose a future government, he modifies his analysis of each party's past performance according to his estimate of probable changes in its behavior. Nevcrtheless, the current record of the incumbents remains the central item in his evaluation. The government also makes tlecision5 rationally, but its 1)ehavior is not so easy to analyze. l~ecauseit is engaged in political warfare with its opponents. Each party resembles a player in an iY-person game or an oligopolist engaged in cutthroat competition. However, the conjectural variation 1)rot)lern is solnewhat simplified, because the incumbent party E('OS0PIIIC THEORY OF POLTTICrlL AC'I'IOX I S A 1)EhIOCKACY 139 must always commit itself on each issue before the opposition parties do. Since it is in power, it must act whenever the occasion for a decision arises, if failure to respond is counted as a form of action. Hut the opposition, which is not responsible for the government, can wait until the pressure of events has forced the governing party to commit itself. Thus opposition parties have a strategic advantage -which incidentally makes the analysis of interparty warfare simpler than it would be if all parties revealed their strategies simultaneously. However, I shall not explore party strategies in a perfectly informed world, because nearly all the conclusions that could be drawn are inapplicable to the imperfectly informed world in which we are primarily interested. Only one point should 1,e stressed: in a world where perfect knowledge prevails, the government gives the preferences of each citizen exactly the same weight as those of every other citizen. This does not mean that its policies favor all citizens equally, since strategic considerations may lead it to ignore some citizens and to woo others ardently or to favor some with one policy and others with another. But it never deliberately eschews the vote of Citizen .\ to gain that of Citizen R. Since each citizen has one and only one vote, it cannot gain by trading A's vote for R's, ceteris paribus. In short, the equality of franchise is successful as a device for distributing political power equally among citizens. Lack of conlplete inforilzatioil on which to base decisions is a condition so basic to human life that it influences the structure of almost every social institution. In politics especially, its effects are profound. For this reason, I devote the rest of my analysis to the impact of imperfect knowledge upon political action in a democracy. In this model, imperfect knowledge means (1) that parties do not always know exactly what citizens want : (2) that citizens do not always know what the government or its opposition has done, is doing, or should be doing to serve their interests; and (3) that the information needed to overcome both types of ignorance is costly-in other words, that scarce resources must be used to procure and assimilate it. Although these conditions have many effects upon the operation of government in the model, I concentrate on only three: persuasion, ideologies, and rational ignorance. As long as we retain the assuinl~tionof perfect knowledge, no citizen can possibly influence another's vote. Each knows what would benefit him most, what the government is doing, and what other parties would do if they were in pourer. Therefore, the citizen's political taste structure, which I assume to be fixed, leads him directly to an unambiguous decision about how he should vote. If he remains rational, no persuasion can change his mind. But, as soon as ignorance appears, the clear path from taste structure to voting decision becomes obscured by lack of knowledge. Though some voters want a specific party to win because its policies are clearly the most beneficial to them. others are highly uncertain about which party they prefer. They are not sure just what is happening to them or what would happen to them if another party were in power. They need more facts to establish a clear preference. By providing these facts, persuaders can become effective. Persuaders are not interested per se in 140 AX'rHONY DOWNS helping people who are uncertain become less so; they want to produce a decision that aids their cause. Therefore, they provide only those facts which are favorable to whatever group they are supporting. Thus, even if we assume that no erroneous or false data exist, some men are able to influence others by presenting them with a biased selection of facts. This possibility has several extraordinarily important consequences for the operation of government. First, it means that some men are more important than others politically, because they can influence more votes than they themselves cast. Since it takes scarce resources to provide information to hesitant citizens, men who command such resources are able to wield more than proportional political influence, ceteris paribus. The government, being rational, cannot overlook this fact in designing policy. As a result, equality of franchise no longer assures net equality of influence over government action. In fact, it is irrational for a democratic government to treat its citizens with equal deference in a world in which knowledge is imperfect. Second, the government is itself ignorant of what its citizens want it to do. Therefore it must send out representatives (1) to sound out the electorate and discover their desires and (2) to persuade them it should be re-elected. In other words, lack of information converts democratic government into representative government, because it forces the central planning board of the governing party to rely upon agents scattered throughout the electorate. Such reliance amounts to a decentralization of government power from the planning board to the agents.12 The central board continues to decenl2 Decentralization may be geographical or by social groups, depending upon the way society is divided into homogeneous parts. tralize its power until the marginal votegain from greater conformity to popular desires is equal to the marginal vote-loss caused by reduced ability to co-ordinate its actions. This reasoning implies that a democratic government in a rational world will always be run on a quasi-representative, quasi-decentralized basis, no matter what its formal constitutional structure. as long as communication between the voters and the governors is less than perfect. Another powerful force working in the same direction is the division of labor. To be efficient, a nation must devel 013 specialists in discovering, transmitting, and analyzing popular opinion, just as it develops specialists in everything else. These specialists are the representatives. They exercise more power, and the central planning board exercises less, the less efficient are communication facilities in society. The third consequence of imperfect knowledge and the resulting need for persuasion is really a comhination of the first two. Because some voters can be influenced, specialists in influencing them appear. And, because government needs intermediaries between it and the people, some of these influencers pose as "representatives'' of the citizenry. On one hand, they attempt to convince the government that the policies they stand for which are of direct benefit to themselves -are both good for and desired by a large portion of the electorate. On the other hand, they try to convince the electorate that these policies are in fact desirable. Thus one of their methods of getting government lo believe that public opinion supports them is to create favorable opinion through persuasion. Though a rational government will discount their claims, it cannot ignore them altogether. I t must give the influencers more than proportional weight in forming policy, because they may have succeeded in creating favorable opinions in the silent mass of voters and because their vociferousness indicates a high intensity of desire. Clearly, people with an intense interest in some policy are more likely to base their votes upon it alone than are those who count it as just another issue; hence government must pay more attention to the former than the latter. To do otherwise would be ir- rational. Finally, imperfect knowledge makes the governing party susceptible to bribery. In order to persuade voters that its policies are good for them, it needs scarce resources, such as television time, money for propaganda, and pay for precinct captains. One way to get such resources is to sell policy favors to those who can pay for them, either by campaign contributions, favorable editorial policies, or direct influence over others. Such favor buyers need not even pose as representatives of the people. They merely exchange their political help for policy favors-a transaction eminently rational for both themselves and the government. Essentially, inequality of political influence is a necessary result of imperfect information, given an unequal distribution of wealth and income in society. \Then knowledge is imperfect, effective political action requires the use of economic resources to meet the cost of information. Therefore, those who command such resources are able to swing more than their proportional weight politically. This outconle is not the result of irrationality or dishonesty. On the contrary, lobbying in a democracy is a highly rational response to the lack of perfect information, as is government's submission to the demands of lobbyists. To suppose otherwise is to ignore the existence of information costs --that is, to theorize about a mythical world instead of the real one. Imperfect knowledge allows the unequal distributions of income, position, and influence-which are all inevitable in any economy marked by an extensive division of labor-to share sovereignty in a realm where only the equal distribution of votes is supposed to reign. Since the parties in this model have no interest per se in creating any particular type of society, the universal prevalence of ideologies in democratic politics appears to contradict my hypothesis. But this appearance is false. In fact, not only the existence of ideologies, but also many of their particular characteristics, may be deduced from the premise that parties seek office solely for the income, power, and prestige that accompany it.13.4gain, imperfect knowledge is the key factor. In a complex society the cost in time alone of comparing all the ways in which the policies of competing parties differ is staggering. Furthermore, citizens do not always have enough information to appraise the differences of which they are aware. Nor do they know in advance what problems the government is likely to face in the coming election period. Under these conditions many a voter finds party ideologies useful because they remove the necessity for relating every issue to his own conception of "the good society." Ideologies help him focus attention on the differences between parties; therefore, they can be used as samples of all the differentiating stands. Furthermore, if the voter discovers a correlation between each party's ideology l 3 1 define "ideologies" as verbal images of "the good society" and of the chief policies to be used in creating it. 142 ANTHONY DOU'SS and its policies, he can rationally vote by comparing ideologies rather than policies. In both cases he can drastically reduce his outlay on political information by informing himself only about ideologies instead of about a wide range of issues. Thus lack of information creates a demand for ideologies in the electorate. Since political parties are eager to seize any method of gaining votes available to them, they respond by creating a supply. 1;ach party invents an ideology in order to attract the votes of those citizens who wish to cut costs by voting ideologi- cally.'-' This reasoning does not mean that parties can change ideologies as though they were disguises, putting on whatever costume suits the situation. Once a party has placed its ideology "on the market," it cannot suddenly abandon or radically alter that ideology without convincing the voters that it is unreliable. Since voters are rational, they refuse to support unreliable parties; hence no party can afford to acquire a reputation for dishonesty. Furthermore, there must be some persistent correlation between each party's ideology and its subsequent actions; otherwise voters will eventually eschew ideological voting as irrational. Finally, parties cannot adopt identical ideologies, because they must create enough product differentiation to make theii output distinguishable from that of their rivals, so as to entice voters to the 1,011s. However, just as in the product market, any markedly successful ideology is soon imitated, and differentiation takes place on more subtle levels. '"11 reality, party ideologies prol)ably stem originally frorn the interests of those persons who found each party. But, once a political party is created, it takes on an existence of its own and eventually hecomes relatively independent of any particular interest group. When such autonomy prevails, my analysis of itleologies is fully applicable. Analysis of political ideologies can be carried even further by means of a spatial analogy for political action. To construct this analogy, I borrow and elaborate upon an apparatus first used by Harold Hotelling in his famous article "Stability in Competition."15 My version of Hotelling's spatial market consists of a linear scale running from zero to one hundred in the usual left-to-right fashion. To render it politically meaningful, I make the following assumptions: I. The political parties in any society can be ordered from left to right in a manner agreeti upon by all voters. 2. Each voter's preferences are single-peaked a t some point on the scale and slope monotoilicall>. do\\nnard on either side of the peak (unless it lies a t one extreme of the scale). 3. The frequent) distribution of voters along the scale is variable from society to society but fixed iil any one iociety.'" 1. Once placed on the pc~liticalscale, a party can move ideologically either to the left or to the right up to but not beyond the nearest party tonard \r hich it is moving." 5. In a t\\o-party system, if either part) moves anay from the extreme nearest it to\rard the other party, cxtiernist voters a t its end of the scale map abstain because they see no significant difference betn een the choices offeretf thern.I0 Under these conditions IIotelling's conclusion that the parties in a two-party system inevitably converge on the center does not necessarily hold true. If voters lG Actually, this distril)ution may vary in any orle society even in the short run, but I assume it to he fixed in order to avoid discussing the complex of historical, sociological, psychological, and other factors which cause it to change. l 7 I t cannot go beyond the adjacent parties, because such "leaping" would indicate ideological unreliability and would cause its rejection by the electorate. 18 This is equivalent to assuming elastic demand along the scale, as Smithies did in his elaboration of the Hotelling model (see Arthur Smithies, "Optimum Location in Spatial Competition," Joumnl of Political Economy, XLIX [1941], 423-39). 1 I O M I ' H E R 0 P O I 1 1 I ' I I 1 0 . 143 are ciistriljuted along the scale as sho~+,n in Figurc 1, then Hotclling is right. Assuming that Party A starts at position 25 and Party B at 75, both move toward 50, since each can gain more votes in the center than it loses a t the extremes because of abstention. But, if the distribution is like that shown in Figure 2, the two parties diverge toward the extremes rather than converge on the tenter. Each gains more votes by moving toward a radical position than it loses in the center. This reasoning implies that stable government in a two-party democracy requires a distribution of voters roughly approximating a normal curve. IYhen such a distribution exists, the two parties come to resemble each other closely. Thus, when one replaces the other in office, no drastic policy changes occur, and most voters are located relatively close to the incumbent's position no matter which party is in power. But when the electorate is polarized, as in Figure 2, a change in parties causes a radical alteration in policy. And, regardless of which party is in oflice, half the electorate always feels that the other half is imposing policies upon it that are strongly repugnant to it. In this situation, if one party keeps getting re-elected, the disgruntled supportersof the otherpartywillprobably revolt; whereas ~f the two parties alternate in office,social chaos occurs, because government policy keeps changing from one extreme to the other. Thus democracy does not lead to effective, stable government when the electorate is polarized. Either the distribution must change or democracy will be replaced by tyranny in which one extreme imposes its will upon the other. Hotelling's original model was limited to the two-firm (or two-party) case, because, when three firms existed, the two outside ones converged on the iniddle one, which then leaped to the outside to avoid strangulation. Since this process repeated itself endlessly, no stable equilibrium emerged. But, in my model, such leaping is impossible, because each party has to maintain continuity in its ideology. Hence this model can be applied to multiparty systems without resulting in disequilibrium. Multiparty systems are most likely to exist when the distribution of voters is multimodal, as shown in Figure 3. A separate party forms at each mode, and each party is motivated to stay at its mode and to differentiate itself as completely as possible from its neighbors. IE it moves to the left so as to gain votes, it loses just as many votes to the party on its right (or loses them because of a1)stention if it is an extremist party at the right end of the scale), and vice versa. Thus its optimal course is to stay where it is and keep other parties from approaching it. In a multiparty system, therefore, we find conditions exactly opposite to those in a viable two-party system. I17hereasin the former each party links itself to a definite ideological position and stresses its differences from other parties, in the latter both parties move toward the political center so as to resemble each other as closely as pos- sible. This conclusion implies that voters in multiparty systems have a wider range of choice than voters in two-party systems and that each choice in this range is more definitely linked to some ideological position. Thus it appears that the electorate exercises a more significant function in a multiparty system than in a two-party system, because only in the former does it make much difference which party gets elected. However, appearances are deceiving in politics, because in fact the government in a multiparty system is likely to have a less definite, less coherent, and less integrated program than the government in a two-party system. This paradoxical outcome arises from the necessity in most multiparty systems of forming coalition governments. Since voters are scattered among several modes, only rarely does one party obtain the support of a majority of those voting. Yet, in most democracies, the government cannot function ~vithoutat least the indirect support of a majority of voters. Even in systems in which the legislature selects the government, a majority of its members must support the coalition chosen to govern before the coalition can take office. If we assume that representation in the legislature is "fairH--that each member represents the same number of citizens-- then even a coalition government must receive the indirect support of a majority in order to govern. Such support can be maintained only if the government implements at least some policies that appeal to-are ideo logically near-each cluster of voters whose support it needs. If a majority of voters are massed in one relatively narrow band on the left-right scale, then the government can choose all its policies from within this band. Hence its policies will form a fairly cohesive set embodying the ideological viewpoint associated with that area of the scale. This outcome is typical of a two-party system. But in a m~lltipartysystem there are many modes scattered across the whole scale. Therefore, in order to appeal to a majority of voters, the government must be a coalition of parties and must include in its policy-set somepolicies espoused by each party in the coalition. In this manner it "pays off" voters at each cluster in return for their support. However, the result is that its program containspolicies reflecting a wide variety of ideological viewpoints, so that no real cohesion or integration about any one II7eltanschauLing is possible. This outcome necessarily occurs whenever the distribution of voters along the scale is so scattered that only a very wide band can encompass a majority. Consequently, a multiparty system offers voters an ostensible choice between definite, well-integrated policy-sets in each election, but only rarely does one of these sets actually govern. Usually a coalition governs, and its policies are likely to be less definite and less well integrated than those of the government in a two-party system. This is true even though voters in the latter are offered only two relatively unintegrated alter- AN ECONOMIC THEORY OF POLITICAL ACTIOK IN '2 DEMOCRACY 145 natives which closely resemble each other. No wonder politics often seems confusing. IYhether a political system has two or more parties depends on the distribution of voters along the scale and on the electoral rules governing the system. To demonstrate this dual dependence, I use the concept of "political equilibrium." A state of political equilibrium exists when no new parties can successfully be formed and when no existing party is motivated to move away from its present position. The limit to the number of new parties that can be formed successfully springs from my definition of success as ability to gain the income, power, and prestige that go with office; that is, as ability to get elected. If the constitution calls for the election of a legislature by proportional representation and the subsequent formation of a government by the legislature, then many parties can be formed, because any given party can get at least some of its members elected by winning the support of only a small proportion of the citizens. Once elected, these members have a chance to share in the fruits of ofice by joining a coalition government. Hence it follows from my hypothesis about party motivation that many parties are likely to exist in a proportional representation system. Their number is limited only by the number of seats in the legislature and by the necessity of formulating ideologies sufficiently different from those of existing parties to attract votes away from them.'Wew parties continue to form until the distribution of voters is "saturated" -until there is not enough ideological "room" between existing parties to support others significantly different from them. Tn an electoral system in which a lYl'he number of sufticiently different parties as system can support depends upon the shape of the distribution of voters along the scale. plurality is necessary for victory, the limit on successful party formation is much more stringent. Since the only way to insure a plurality against all opponents is to win a majority of votes, small parties tend to combine until two giants are left, each of which has a reasonable chance of capturing a majority in any given election. I17herethese two parties are located on the ideological scale depends upon the distribution of voters, as explained before. Actually, the policy position and stability of the government in a democracy are relatively independent of the number of parties; they follow primarily from the nature of the distribution of voters along the left-right scale.*OIf a majority of voters are massed within a narrow range of that scale, democratic government is likely to be stable and effective, no matter how many parties exist. As noted earlier, the government can formulate a policy-set which appeals to a majority of voters and yet does not contain policies embodying widely disparate points of view. But, if the government can win the support of a majority only by adopting a scattering of policies chosen from a broad range of viewpoints, these policies tend to cancel each other out, and the government's net ability to solve social problems is low. Thus the distribution of voters-which is itself a variable in the long run-determines whether or not democracy leads to effectivegovernment. IYhen information is costly, no decision-maker can afford to know everything that might possibly bear on his decision before he makes it. He must select 20 However, because the preferences of rising generations are influenced by the alternatives offered them, the number of parties is one of the factors that determine the shape of the distribution of voters. only a few data from the vast supply in existence and base his decision solely upon them. This is true even if he can procure data without paying for them, since merely assimilating them requires time and is therefore costly. The anlount of information it is rational for a decision-maker to acquire is determined by the following economic axiom: I t is always rational to perform any act if its marginal return is larger than its marginal cost. The marginal cost of a "hit" of information is the return foregone by devoting scarce resourcesparticularly time-to getting and using it. The marginal return from a "bit" is the increase in utility income received because the information enabled the decision-maker to improve his decision. In an imperfectly informed world, neither the precise cost nor the precise return is usually known in advance; but decisionmakers can nevertheless employ the rule just stated by looking at expected costs and expected returns. This reasoning is as applicable to politics as it is to economics. Insofar as the average citizen is concerned, there are two political decisions that require information. The first is deciding which party to vote for; the second is deciding on what policies to exercise direct influence on government policy formation (that is, how to lobby). Let us examine the voting decision first. BeEore we do so, it is necessary to recognize that in every society a stream of "free" information is continuously disseminated to all citizens. Though such "free" data take time to assimilate, this time is not directly chargeable to any particular type of decision-making, since it is a necessary cost of living in society. For example, conversation with business associates, small talk with friends, reading the newspaper in a barber shop, and listening to the radio while driving to work are all sources of information which the average man encounters without any particular effort to do so. Therefore, we may consider them part of the "free" information stream and exclude them fro111 the problem of how much information a decision-maker should obtain specifically to improve his decisions. The marginal return on information acquired for voting purposes is measured by the expected gain from voting "correctly" instead of "incorrectly." In other words, it is the gain in utility a voter helieves he will receive if he supports the party which would really provide him with the highest utility income instead of supporting some other party. However, unless his vote actually decides the election, it does not cause the "right" party to be elected instead of a "wrong" party; whether or not the "right" party wins does not depend on how he votes. Therefore, voting "correctly" produces no gain in utility whatsoever; he might as well have voted "incorrectly." This situation results from the insignificance of any one voter in a large electorate. Since the cost of voting is very low, hundreds, thousands, or even millions of citizens can afford to vote. Therefore, the probability that any one citizen's vote will be decisive is very small indeed. I t is not zero, and it can even be significant if he thinks the election will be very close; but, under most circumstances, it is so negligible that it renders the return from voting "correctly" infinitesimal. This is true no matter how tremendous a loss in utility income the voter would experience if the "wrong" party were elected. 4nd if that loss is itself small-as it may be when parties resemble each other closely or in local elections-then the incentive to become well informed is practically nonexistent. 7'hercforc, we reach the startling conclusion that it is irrational for most citizens to acquire political information for purposes of voting. ,As long as each person considers the behavior of others as given, it is simply not worthwhile for him to accjuire information so as to vote "correctly" himself. The probability that his vote will determine which party governs is so low that even a trivial cost of procuring inforination outweighs its return. Hence ignorance of politics is not a result of unpatriotic apathy; rather it is a highly rational response to the facts of political life in a large democracy. This conclusion does not mean that every citizen who is well informed about politics is irrational. ,4 rational man can I~ecomewell informed for four reasons: (1)he may enjoy being well informed for its own sake, so that information as such provides him with utility; (2) he may believe the election is going to be so close that the probability of his casting the decisive vote is relatively high; (3) he may need inforination to influence the votes of others so that he can alter the outcome of the election or persuade government to assign his preferences more weight than those of others; or (4) he may need information to influence the formation of government policy as a lobbyist. Neverthe less, since the odds are that no election will be close enough to render decisive the vote of any one person, or the votes of all those he can persuade to agree with him, the rational course of i~ctionfor most citizens is to remain politically uninformed. Insofar as voting is concerned, any attempt to acquire information beyond that furnished by the stream of "free" data is for them a sheer waste of resources. The disparity between this conclusion and the traditional conception of good citizenship in a democracy is indeed striking. IIow can we explain it? The answer is that the benefits which a majority of citizens would derive from living in a society with a well-informed electorate are indivisible in nature. IYhen most members of the electorate know what policies best serve their interests, the government is forced to follow those policies in order to avoid defeat (assuming that there is a consensus among the informed). This explains why the proponents of democracy think citizens should be well informed. But the benefits of these policies accrue to each member of the majority they serve, regardless oi whether he has helped bring them about. In other words, the individual receives these benefits whether or not he is well informed, so long as most people are well informed and his interests are similar to those of the majority. On the other hand, when no one else is well iniormed, he cannot produce these benefits by becoming well informed himself, since a collective effort is necessary to achieve them. Thus, when benefits are indivisible, each individual is always motivated to evade his share of the cost of producing them. If he assumes that the behavior of others is given, whether or not he receives any benefits does not depend on his own efforts. But the cost he pays does depend on his efforts; hence the most rational course for him is to minimize that cost-in this case, to remain politically ignorant. Since every individual reasons in the same way, no one bears any costs, and no benefits are produced. The usual way of escaping this dilemma is f o ~all individuals to agree to be coerced by a central agency. Then each is forced to pay his share of the costs, but he knows all others are likewise forced to pay. Thus everyone is better off than he would be if no costs were borne, because everyone receives benefits which (I here nssume) inore than oflset his sll:tre of the costs This is a hasic rationale for using coercion to collect revenues for national deiense and for many other government operations that yield indivisible l~enefits.~' But this solution is not feasible in the case of political information. The government cannot coerce everyone to be well informed, because "well-iniormedness" is hard to measure, because there is no agreed-upon rule for deciding how much information 01 what kinds each citizen "should" have, and because the resulting interference in personal affairs would cause a loss of ucility that would proba- 1,1y outweigh the gains to be had from a well-informed electorate. 'L'he most any democratic government has done to remedy this situatjon is to co~npelyoung people in schools to take courses in civics, government, and history. Consequently, it is rational for every individual to minimize his investment in political information, in spite of the fact that most citizens might benefit substantially if the whole electorate were well informed. -4s a result, democratic political systems are bound to operate at less than maximum efficiency. Government does not serve the interests of the majority as well as it would if they were well informed, but they never become well informed. I t is collectively rational, but individually irrational, for them to do so; and, in the absence of any mechanism to insure collective action, individual rationality pre-