On the inevitability of divided government and improbability of a complete separation of powers
This paper provides a tightly written overview and modest extension of the constitutional exchange and evolution model developed in Perfecting Parliament and uses that approach to analyze the division of authority that one would expect to see in contemporary constitutional governments. The analysis suggests that constitutions tend to be written, based on the king and council template, and buttressed by a more or less independent court system. Moreover, it suggests that constitutions change at the margin through time as constitutional bargaining takes place. This suggests that a complete separation of power is unlikely to be observed in the long run. Empirical evidence developed from the IAEP data base is consistent with these predictions.
KeywordsConstitutional reform Separation of powers Divided government Constitutional exchange
JEL ClassificationH11 D72 D86
There are positive and normative approaches to analyzing divided government. Positive research explores the properties and origins of governments with several more or less independent centers of policy-making authority. Normative analysis attempts to assess the relative merits of alternative divisions of authority. With respect to the latter, it has often been suggested that legislative, executive, and judicial authority should be exercised by separate, independent centers of policy making: a parliament, prime minister, and court system. In some cases, the two approaches are joined in order to simultaneously analyze and justify particular architectures and procedures of governance. As a result, government institutions may be designed or modified with their anticipated normative properties as main objectives. It is arguable, for example, that the contemporary division of legislative, executive, and judicial authority in the United States owes much to the analyses and normative theories of the separation of powers worked out in the seventeenth and eighteenth centuries.1
In this paper, I suggest that divided government is inevitable, although a complete separation of powers is unlikely. This is partly because it is not functionally possible for a single individual to enact and enforce a complex body of law over a large territory in which a large number of individuals live. It is also because constitutional gains to trade exist from time to time, and realizing those gains tends to favor power-sharing arrangements over unitary ones and mixed over completely separate assignments of authority. Indeed, the advantages of shifting authority away from a unified government are so great that no sensible person or small group of such persons would voluntarily hold on to all authority. The essential question about governance in the long run is not whether it will or should be divided, but how authority will be reassigned through constitutional bargaining, given an original assignment and amendment procedure.
The first parts of the paper discusses why constitutional gains to trade exist, why this implies that authority over public policy tends to be divided, and why it is unlikely to be distributed in the manner proscribed by the separation of powers doctrine. In the course that analysis, how and why “political property rights” are initially assigned, why they tend to be written down, and why a process of constitutional review tends to be common within representative governments are also briefly discussed. The last part of the paper uses data on contemporary constitutions to determine if the predictions of that analysis are consistent with the formal divisions of authority observed in contemporary governments. It turns out that most governments are divided, have written constitutions, include procedures for constitutional review, and evidently have a distribution of authority that reflects ongoing constitutional bargaining. The conclusion summarizes the analysis and results, and their relevance for the doctrine of the separation of powers.
2 Divided government, the market for power, and constitutional reform
Several definitions of divided government are possible. Perhaps the most general is that a government may be said to be divided if more than one person has influence over policy decisions.
Divided government is more likely in large than small organizations, but policy-making authority is often divided even in relatively small organizations, such as partnerships. Partnerships often makes policy decisions through consensus (unanimous agreement), in which case each partner has veto power over their organization’s policies. The partners may also share agenda authority—the ability to propose new policies to fellow partners—insofar as suggestions for new policies or changes in policy are developed through formal discussion at meetings or in informal discussions before such meetings take place. In relatively large organizations, several more or less independent groups of persons may jointly determine policy through standing procedures that determine how a series of more or less independent decisions produce policies. For example, adopting major policies in a large organization often requires unanimous recommendations among its various agencies of governance, which may themselves rely on majority rule. In such cases, the multi-person “branches” of government have veto power, although no single member of those branches may. Such organizations often have a formal “architectural division of authority.” There may, for example, be a chief executive officer, a board of directors, and a body of shareholders.
In large organizations, a precise statement of the manner in which policies are to be selected and reviewed is often valuable, because this reduces conflict over the procedures themselves. The simplest way to reduce conflict over policy-making procedures is to write down the rules for making such decisions in as unambiguous a manner as possible. By doing so, the formal (de jure) distribution of policy-making authority within an organization is characterized. Such written statements are an organization’s constitution and normally describe the architecture of its government, how senior officeholders are selected, and the procedures through which policies are adopted by those officeholders. In large organizations, “the government” is largely a process through which more than one person chooses and implements policy. It is neither a unique decision maker nor, in large organizations, a single decision-making body.
For the purposes of this paper, it is important to note that the division of authority characterized by a constitution is rarely entirely fixed by law or custom. Most constitutions include formal amendment procedures and also informal ones through which constitutional gains to trade can be realized.
2.1 A digression on the exchange of authority
For the present analysis, it is important to note that what is traded is not actually goods, but authority over goods. Exchange requires a voluntary transfer of authority over goods and/or services from one party to another. If Alice simply took goods from Bob and gave him some of her own, this would not be an exchange unless Bob agreed to shift authority over his goods to Alice. If he did not, he could veto such transfers even if such a transfer made him better off. Moreover, both civil law and the courts would support his veto even in such cases. Trade occurs only when all parties agree to particular shifts of authority over goods from themselves to other persons. Without such a transfer of authority, trade does not take place, although theft and redistribution may.
This implies that the usual way of labeling the diagram is a bit misleading. The horizontal and vertical axes are normally said to represent quantities of goods (and/or services). However, what is actually characterized is the distribution of authority over the existing goods and services in the economy of interest (e.g., of ownership rights). What points inside an Edgeworth box actually characterize is the distribution of authority over a predetermined quantity of two goods.
For example, if X in the diagram above is tomatoes, Z is peppers, and the two private garden outputs can be represented by the endowment combination labeled 1, with each gardener having complete authority of his or her own production, gains to trade exist. These gains can be realized by shifting some of the authority that each initially has over their own homegrown produce to the other. This shift occurs only for the subset of the tomatoes and peppers traded. Technically, what is actually exchanged is not the produce per se, but rather (legitimate) authority over part of their garden’s output.
The exchange may be expressed in terms of the number of peppers that Bob gives to Alice in exchange for a number of tomatoes. However, only the initial distribution of authority is changed, the total number of peppers and tomatoes is unaffected.
Note that voluntary exchange requires Alice and Bob to have two levels of authority, the first over the goods themselves—they may use them more or less as they see fit—and the second over the assignment of those rights. With respect to the latter, they must have the authority to shift their authority over their goods and/or services from themselves to other persons. And normally, as owners, they will also have the authority to propose particular terms of trade and to reject shifts of their own initial authority. Because voluntary exchange requires participants with both veto and agenda-setting power, the act of trade may be said to require the existence of a “divided government,” although this is not true for acts of taking or theft.
2.2 Trading authority to make policy
One might ask what the above legalistic analysis of exchange has to do with the topic of this paper, namely the concepts of divided governance and separation of powers as applied to politics?
There are several parts to the answer, but the four main ones should be clear from the above. First, organizations—even temporary organizations such as those involving two traders—very often exhibit divided authority. Second, authority is inherently transferable. Third, transfers of authority may be voluntarily undertaken. Whenever particular transfers exist that can make all of the parties with veto authority over the exchange better off, the parties will attempt to reassign authority to realize those gains. Together these three imply the fourth: therecan be, and are likely to be, markets for power (authority). Indeed, essentially all retail, wholesale, stock, insurance, and futures markets are settings in which authority over goods and services is traded among participants.
Using Edgeworth boxes to analyze markets for political as opposed to economic authority “simply” requires reinterpreting the axes. The horizontal axis might, for example, be reinterpreted as the degree of budget authority over education and the vertical axis as the degree of budget authority over healthcare services. The two persons represented may be pivotal members of a central and local government. The indifference curves may represent the preferences of pivotal decision makers over those areas of authority (which are partly institutionally induced by various rules and regulations). Insofar as the indifference curves capture essential aspects of the goals of the relevant decision makers, the logic of the Edgeworth box is unaffected. There may be gains to trade in political authority between Alice and Bob. Power may grow out of the barrel of a gun, as claimed by Mao Zedong, but once possessed, gains to trade occasionally exist and can be realized through entirely voluntary shifts of authority from one locus of authority to another.
Note that such gains to trade do not require an initial “mixed government” in which all authority is initially shared among several officeholders or policy-making bodies. Gains to trade can exist when there is initially a complete separation of power as well as with mixed assignments. In the illustration, gains exist at assignment 2 where a complete separation of authority exists. Note that both pivotal officeholders prefer distribution 3 to distribution 2. Thus, even with an initially complete separation of authority, gains to trade may exist, and authority may be shifted among the persons or agencies of interest in a manner that “undermines” the initial separation of authority.2
Such constitutional exchanges are common in all organizations with governments whose constitutions include methods through which policy-making authority can be reassigned. Such procedures can be regarded as amendment procedures, although they may not always be called that within the organization itself.
Markets for power, for example, clearly exist within modern corporations. Senior management’s influence over “their” corporation’s future policy decisions (products, production process, distribution network, and employee compensation schemes) is clearly divided. Senior managers share policymaking authority with each other and with their board of directors, their shareholders, and other financiers. Authority might initially be held entirely by an organization’s formeteurs, who may subsequently decide to “go public.” This act involves trading authority from themselves to others in exchange for capital or other goods and services. Such trades of authority resemble those that begin at the endowment labeled 2 in Fig. 1. Individual stockholders may increase their authority by purchasing it (shares) from other stockholders or paying a higher price for it in the initial offering. When there are large numbers of shareholders, even a modest increase in percentage of all shares held can significantly increase one’s influence over policies.3
To the extent that the usual assumptions about indifference curves are empirically relevant, the Edgeworth box representation of constitutional gains from trade implies that authority tends to be mixed. The contract curve lies more or less in the middle of the box, and in equilibrium authority tends to be divided and mixed along most of the contract curve, rather than undivided or separate as it would be at or very near the four corners of the box.
A complete separation of powers is highly unlikely if indifference curves over policy-making authority have their usual shapes—whether in terms of legislative, executive, and judiciary powers, or not.4
2.3 Relevance of constitutional exchange for normative constitutional theory
The possibility of constitutional exchange is a serious problem for constitutional designers whenever they prefer a complete separation of authority or any other distribution of authority that is not on the contract curve. A complete separation of powers could be specified in constitutional documents, but this would be unlikely to survive in the long run, because gains to constitutional exchange would eventually be realized.
Preventing such gains from trade from being realized requires other supporting institutions that block, or raise the transaction costs, of constitutional bargains among those holding authority, whether separate or not. Formal amendment procedures and constitutional courts may, for example, be included in constitutional designs that attempt to reduce the scope for constitutional bargaining (Congleton and Rasch 2006). However, it must be acknowledged that formal procedures do not necessarily prevent or significantly increase the transaction costs of informal constitutional bargains.
To make sure that formal procedures of constitutional bargaining and exchange (reform) are used, rather than informal procedures, constitutional designers may specify an initial division of authority in written constitutional documents and provide a method of constitutional review that can block informal reforms by ruling them “unconstitutional.” Together a formal amendment procedure and constitutional review process would not eliminate constitutional exchange, but would constrain the set of bargains that can be lawfully made and thereby restrict the trajectory of constitutional development.
For positive, as opposed to normative, theory, the existence of constitutional gains to trade away from the contract curve is less problematic. It simply implies that both formal and informal constitutional reforms tend to be adopted whenever they benefit the parties with sufficient authority to modify standing procedures. As this process takes place, the realization of constitutional gains to trade tends to produce stable outcomes as the participants reach the relevant contract curve. Once reached, the result is a temporarily stable distribution of authority, until a shock alters the indifference curves of pivotal decision makers.5
2.4 On the “breakdown” of unitary assignments of authority
Two points on the constitutional contract curve of our illustrating Edgeworth box may be thought to be especially stable, namely, those that assign all authority to one or the other person (or one or the other rule-making body). Complete undivided political authority may conceptually be grounded on coercive ability or agreement, according to whether we follow Mao’s (1938/1990) or Hobbes’ (1651/1982) reasoning.
An undivided assignment of authority would place us at either the lower left-hand or the upper right-hand corner of the Edgeworth Box. To demonstrate that potential gains from sharing authority exist even in this case requires moving beyond the standard Edgeworth box analysis. To do so, it is useful to examine more closely the nature and implications of an initial unitary distribution of policy-making authority in a regional government.
All organizational governments make rules that are to be followed by their team members. Regional and national governments, however, differ from other organizational governments in that their rules bind most persons living within a given geographical area, including persons who are not members of their organizations. An undivided allocation of political authority, thus, would allow the person(s) possessing that authority to direct all the actions of all the other persons in their dominion. Suppose that all the residents of the territory of interest are completely cowed by their ruler or have agreed to a unitary assignment and completely defer to his authority, and so do whatever the “slave-master authoritarian ruler” commands. This is what completely undivided authority requires.
To use this undivided authority effectively, Queen Alice must know what it is that Slave Bob must do to advance her interests in every possible circumstance. And, Bob must be able to follow all the rules (conditional strategies) created for him by Alice. If the rules are complete, Bob will have no autonomous control over policy and little or none over his “own” life.
In practice, the effectiveness of such rules is limited by the abilities of the human mind and the complexity of the decision and action environments. In complex settings, the perfect use of undivided authority is physically and mentally impossible. Imagine, for example, the complexity of the conditional rules required to solve the relatively straightforward problem of having Bob move the pieces on a chessboard exactly as Alice would have in every possible circumstance.
In all but the simplest of settings, delegating some decision-making authority to a subset of the slave-residents can make the authoritarian materially better off by increasing the effectiveness of the organization. These slaves, now agents to whom authority is delegated, become less slave and more freeman as their scope for independent decision making and action increase. And, insofar as increased authority (freedom) on the part of the slaves makes them better off, there will be gains from trade. However, in this case the gains arise because shifts of authority expand the Edgeworth box, rather than through reassigning authority within predetermined and bounded policy domains. As a result of advantages of delegation, unitary authority is replaced with divided authority, however limited it may be. Once divided, the logic of the Edgeworth box can again be applied to analyze subsequent bargaining, given the new division of authority.
Only if Queen Alice cannot tolerate any shift of authority, because “power” and power alone is the measure of her welfare (her utility), would such delegations not occur. Unfortunately for such rulers, the consequent reduction in organizational effectiveness often matters. In a world in which other organizations control adjacent territories, such sovereigns would tend to lose ground and eventually their authority to others who are not quite as “power hungry,” who are willing to give up some decision-making authority to obtain greater organizational output, and thus rule over more effective organizations and larger territories.6 By sharing some of their authority, such rulers may be said to use their territory’s human and other resources more effectively, but “use” does not imply “fully control” in such cases. Their overall authority may be said to increase insofar as giving up authority in some policy areas extends or produces new authority in others.
The claim here is not that historical governments were ever all-powerful slave masters, but that even in such cases, advantages of divided governance exist. Authority can remain concentrated in a single decision maker or decision-making body only in very small organizations operating in very simple environments. Large organizations operating in complex environments cannot afford to completely concentrate policy-making authority.
3 On the initial assignment of authority and architecture of governance
We now turn to the initial assignment and division of authority, which has to this point been left outside the discussion. Again the analysis is positive, although the initial assignment may reflect the constitutional norms of those creating new organizations and the subsequent division of authority may reflect the norms of those participating in constitutional bargaining and reform.
In the usual analysis of an Edgeworth box, the initial endowment is left outside the scope of analysis. For the present analysis, however, the source of the initial endowment is also of interest. All formal organizations have a beginning, at which point the procedures for policy-making authority are created by the organization’s formeteur(s). When a new organization is created, policy-making authority is created out of thin air. This can be done because authority is relational and social rather than physical.7
This is not to say that each organization’s architecture and assignment of authority is also created out of thin air. Most organizations draw their governments from well-known, well-functioning templates familiar to their formeteur(s). These include more or less conventional architectures and divisions of authority, as between a ruler and his advisory council, a king and his council of state, a prime minister and parliament, president and congress, and so forth. The “conventional” templates at a moment in time reflect the past successes (and implicitly the failures) of many generations of previous formeteurs and their governments. Insofar as some templates for governance advance formeteur interests better than others in the long run, it is these that form the menu of alternative governments that new formeteurs choose from. Only organizations that survive and flourish are likely to attract the attention of new formeteurs.
Although the original design is not likely to be permanent for reasons discussed above, the original constitutional design has significant effects on the distribution of authority that emerges from intra-organizational bargaining through its specification of the architecture of government, the initial allocation of authority, and the formal amendment process. Differences in starting points have significant effects on the sequence of subsequent bargains struck for the same reasons that initial endowments in an Edgeworth box imply different Pareto sets. As a consequence, constitutions tend to exhibit what others have called path dependency, although path dependency does not prevent constitutional convergence from occurring in the long run.
3.1 The king-and-council or prime minister and parliament
Among the most common architectures for sharing authority in large formal organizations are ones drawn from the king-and-council template, a general governance structure in which a “king” (chief, baron, king, president, prime minister, etc.) shares policy-making authority with a “council” (council of wise men, grand council, cabinet, parliament, congress, etc.). Authority can be divided in a large number of ways in that template, and both kings and councils can be selected in a number of ways, creating a continuum of both authority and government types. Other templates are also possible, but this is an amazingly common structure (Congleton 2001).
In large organizations and composite ones, authority can also be delegated to “lower” levels of government, which themselves may have divided king-and-council structures. Conversely, in cases in which several organizations band together to form a larger one, new “higher levels” of government may be created, and these are also likely to be divided and based on the king-and-council template, as in the present-day European Union.
The template solves a variety of practical informational, incentive, and successional problems (Congleton 2011a). It is also easily adjusted at a number of margins, which allows it to be fine-tuned to cope with new circumstances and with variation in the talents of the men and women that hold positions of authority. The flexibility of this template allows it to be used by a variety of formeteurs to advance a variety of objectives. For example, the same template can be used to advance both dictatorial and democratic goals, by adjusting the manner of selecting officeholders and the division of authority among them.
The procedures chosen and initial assignment of authority tends to align the interests of officeholders with those of the formeteurs. For example, a kingdom founded by a successful general will normally begin with an executive-dominated (general-dominated) version of the king and council template. Given an executive with relatively broad powers of appointment, promotion, and demotion, pleasing the general will be one of the surest ways to rise to wealth and authority. In contrast, a government that emerges out of a constitutional convention is likely to include decision rules that align the interests of future governments with those of the citizenry. Elections tend to “force” rivals for office to pay attention to citizen-voters in a manner that executive appointment and heredity do not. Thus, constitutions drafted to advance general interests (or reformed to advance such interests) tend to use election-based procedures to select senior office holders (as with members of parliament), rather than executive appointment or hereditary.
In both cases, subsequent bargaining is likely to lead to shifts favoring either parliament or the executive according to the external shocks experienced, the amendment process, and the initial assignment of authority among officeholders.8 The many dimensions of authority and flexibility of the template provide many possible directions for constitutional bargaining and exchange. Nonetheless, the initial assignment of authority matters. The bargains realized at each moment in time reflect the preexisting division of authority, the particular people holding positions of authority in the branches of government, as well as the nature of the environment in which the organization operates.
The reforms may or may not exhibit trends according to the various shifts in the contract curve between the king and council. When liberal norms gain ground within government or among those who select government officials, the reforms adopted tend to better align the interests of policy makers and executives with the interests of the citizenry. Contrariwise, shocks favoring authoritarian rule tend to produce laws and institutional reforms that better align a “subject’s” interest with those of its ruler. In the first case, governments may be constrained to avoid both malfeasance and large mistakes. In the latter case, it is the subjects that would be so constrained.
3.2 Controlling the scope for constitutional exchange: written constitutions and judicial review
In ordinary exchange, it is usually known who has authority over what and the extent to which authority can or cannot be transferred. This is often a matter of custom and routine rather than formal deeds and contracts. Nonetheless, written deeds and contracts supported by a civil law system can increase that understanding and reduce transaction costs of some kinds of trades. They do so by increasing the certainty of both formal and informal claims of authority and by characterizing the legitimate procedures through which authority may be transferred. Most civil law systems also adjust transaction costs in various ways to promote a subset of trades and discourage others.
For example, under most civil law systems, contracts based on misrepresentations (fraud) are not enforced, nor are contracts between politicians and their supporters (bribes, etc.). Under contemporary Western law, cartel agreements are not enforced and one cannot sell oneself into slavery no matter how much one wishes. Thus, it can be said that contemporary civil law systems facilitate some shifts of authority and rule others out, thereby promoting economic development, civil liberties, and safety. Only when a dictator or parliament has a very short planning horizon would he or it prefer arbitrary rule to a predictable, law-bound civil law system.
A similar role clearly exists for written constitutions and review systems that enforce political property rights and amendment procedures. A written constitution establishes “political property rights” (distributes authority), much as a deed establishes the boundaries of a particular’s owner’s land claims. Its formal amendment procedures establish procedures through which that authority can be lawfully traded among members of the existing government. These, as in the Edgeworth Box illustration, imply that authority may shift among branches of government as circumstances change. An independent constitutional court can increase the value of what might be called political property rights by enforcing constitutional assignments of authority and restricting non-constitutional (both voluntary and involuntary) transfers of authority. A well-constructed court assures that policymakers abide by constitutional procedures and constraints.
A variety of institutional devices can produce such courts, including, for example, careful nonpartisan screening of potential judges, lifetime appointments, and commitments to not reduce the salaries of judges. A conservative legalistic perspective and non-decreasable salaries increase the likelihood that constitutional excesses of the legislative and executive branches will actually be overturned by increasing the average judicial affinity for existing constitutional law and reducing prospects for retribution. The overall result in a well-functioning system tends to reduce intragovernmental conflict and other risks by clarifying the distribution of authority among officeholders, the procedures for selecting officeholders, and the subsequent policy-making process. When clear violations are all overturned by constitutional review, fewer will be attempted.
Well-constructed constitutional courts thus increase the value of constitutional documents by defending the pre-existing distribution of authority and blocking illegitimate shifts of that authority. This, together with supermajority or other demanding requirements for formal amendments, tends to increase the transaction costs of reform, increasing constitutional stability and somewhat constraining the trajectory of constitutional evolution.9
However, the advantages of constitutional review differ for autocrats and governments based on popular sovereignty. Neither parliaments nor autocrats wish to be (unnecessarily) constrained by constitutional courts. However, in cases in which constitutions are created by communities to advance shared goals—whether piecewise or whole cloth—voters with a long-term perspective are likely to want their governments to follow the proscribed procedures and stay within the constraints characterized by their constitution. This suggests that independent supreme courts or similar review procedures are likely to be more common in polities whose constitutions are developed through constitutional conventions or reformed in a liberal direction than in ones developed by monarchs or dictators.10
Note that this analysis, in combination with a king and council architecture for policy making, implies that liberal governing systems tend to include three branches of government, a king, council, and a more or less independent court system with at least limited powers of constitutional review. It also suggests that policy-making authority will be divided among the three branches of government. It does not imply, however, a complete separation of policy-making responsibilities. The logic of the Edgeworth Box suggested that shared authority tends to be more common in the long run.
3.3 Other methods of sharing authority: subsidiarity and advantages of local control
Another common institutional device for sharing authority is decentralization. A complete analysis is beyond the scope of this paper, but several points are relevant. First, the goals of the formeteurs and the circumstances of a state’s founding will influence the initial distribution of authority. It is, for example, physically impossible for a single ruler or parliamentary system to police all laws or implement all policies throughout a large territory. Thus, some degree of decentralization is necessary. Even a rent-extracting authoritarian central government will delegate authority to regional agencies in areas in which doing so is likely to increase the rents that can be extracted. A profit-maximizing government would not necessarily neglect “its” citizens, but would be concerned with them only insofar as their responses to tax and other rules affect net revenues (Congleton and Lee 2009). In contrast, the delegation of authority to a central government by pre-existing local or state governments will be limited to policy areas in which clear local advantages can only be realized through centralization.11
Again it is possible that normative theories may influence the initial distribution of authority between central and regional governments. Normative theories of decentralization imply that if economies of scale in services are limited and externalities modest, the bulk of government authority should reside at local levels, because local governments can better tailor their services and rules to local demands and resources (Tiebout 1956; Oates 1972). When the economics of local governance varies among policy areas, some services and rule-making authority will be retained locally, and others transferred to the central government. The optimal division varies in such cases, but not the conclusion that divided government is better than unitary government.12
Regardless of the origin of decentralization (delegation or amalgamation), every division of authority between central and local governments establishes “property rights” that can be traded between levels of government. In the long run, the actual division of authority is ultimately determined by a long series of constitutional exchanges, rather than by the constitution per se (Congleton et al. 2003). Bargaining in most federal systems is nearly continual and authority shifts among levels of government at the margin essentially every year.
Again, constitutional and quasi-constitutional exchange implies that there is unlikely to be a complete separation of authority in the long run, even if initially there is a sharp division of responsibilities—as marble cakes replace layer cakes, to use Oates’ characterization.
4 Some empirical evidence on the nature of divided government
The above analysis implies that the distribution of authority to make rules reflects (a) the goals of those initially forming the government, (b) the available templates when a government is first formed, and (c) subsequent bargaining that takes place after a government is created. As a consequence, essentially all governments (1) are divided, (2) use a structure based on the king-and-council template, (3) are written to reduce political uncertainties and reduce intra-organizational conflict. Policy-making authority is normally shared between (4) the executive and parliament and (5) among levels of government. In some cases, (6) constitutional courts will be created to bind government to constitutional procedures and restrictions. Constitutional courts are more likely to be adopted (7) when constitutions have liberal foundations or reflect liberal reforms.
Binghamton University’s Institutions and Elections Project (IAEP) has recently produced a data set that allows the above predictions to be checked. The project digitized constitutional documents and other data for about 150 countries (all those with populations >500,000) for a 34-year period (1972–2005). The data are not complete, and digitizing institutional data is a nontrivial task, so the data are not perfect, but are the best available at this time for such an analysis. Its various constitutional measures tend to be binary rather than continuous, which implies that the fine-grained bargaining that generates a nation’s institutions cannot be directly examined with the IAEP data set. Nonetheless, the main implications of the above theory of constitutional design, reform, and evolution can be assessed.
The IAEP data base includes “yes/no” tabulations of whether a country has a formal written constitution, includes various architectural features, and about many aspects of its policy-making process. These provide a good deal of information about the structure of governance and distribution of authority within the countries and time period examined. The purpose of the data and estimates below is to test the main implications of the theory outlined in the first part of the paper, and may be regarded as a complement to the extensive inductive literature on constitutions that has emerged in the past two decades [see Voigt (2011) for a survey of that literature].
4.1 Frequency distributions of government architecture and authority
The discussion of the king-and-council template suggests that governments are very often organized into two parts, an executive and a council or parliament. It also suggests that the manner in which the executive and parliament are chosen and the distribution of authority between them varies according to the conditions of their founding and the course of constitutional bargaining.
4.1.2 Written constitutions
4.1.3 Constitutional courts
Federalism in its constitutional sense normally requires explicit mention of subnational governments (provinces, states, lander, etc.) and often includes a federal chamber in the parliament. This constitutional definition is far more restrictive than the one that economists use when analyzing decentralization (fiscal federalism), which focuses on the autonomy of state and local governments at taxation, expenditures, and regulation, rather than constitutional architecture (Ahmad and Brosio 2006). Political federalism often reflects a nation’s history more than the merits of decentralization, in that federal countries are often amalgamations of formerly more or less independent states, as in the United States, Germany, Spain, Canada, and Switzerland.
4.1.5 Division of authority
Consistent with the above analysis, most governments have written constitutions, use the king-and-council template, and have constitutional courts. However, more central to this paper is the division of authority. A written constitution might, for example, concentrate most authority in the legislature or the executive. The IAEP data set does not provide a single measure of executive or legislative authority but does provide sufficient detail about the procedures of governance to construct such measures. If the main hypothesis of this paper is correct, divided government should be commonplace, and vary somewhat through time. Most national governments will include legislatures and executives that share considerable authority.
Together with the legislative indices, it is clear that most governments are divided in most places and at most times, with a fairly even distribution of authority between the executive and legislatures, although significant variation in the distribution of authority exists. Overall, the frequency distributions suggest that a complete separation of authority between the parliament and executive with respect to legislation is rarely observed.
4.2 Some evidence of constitutional negotiations, OLS estimates
If the distribution of authority reflects constitutional exchange, there should be a negative relationship between executive and legislative authority, but not a one-to-one relationship. And, if opportunities for constitutional exchange are produced by exogenous shocks, the path of constitutional bargains should be substantially unpredictable. In such cases, it would resemble a random walk, as modest but unpredictable reforms are adopted given previous assignments of authority. These predictions can be explored by regressing executive authority against legislative authority.
Executive authority index estimates as functions of legislative authority and other constitutional variables (1972–2005)
Amendments, proposed by executive
Amendments proposed by legislature
Age of const. × amends proposed by executive
Age of const. × amends proposed by legislature
Number of observations
Consistent with the constitutional exchange hypothesis, the amendment process plays a significant role in determining executive authority. When the legislature has veto power over amendments, the executive tends to be weaker, other things being equal. When the executive has veto power over amendments, the executive tends to be more powerful, other things being equal.
The last column includes several interaction terms that seemed likely to be relevant given the discussion in the first half of this paper. Legislative and executive bargaining roughly offset each other through time in countries where both the legislature and executive can propose amendments.13 The formal institutions of constitutional bargaining and reform evidently matter.
Overall, the coefficient estimates are of stable magnitudes across estimates and the regressions all have a statistically significant fit (at the 0.001 level). All the coefficients have the predicted signs and most are statistically different from zero at very high levels of significance. The R2 are relatively low, however, which suggests that idiosyncratic aspects of the environment in which constitutional bargaining takes place are often decisive.
5 Summary and conclusions
This paper has provided a tightly written discussion of the origin and dynamics of constitutional design, and provided evidence that the theory has relevance for contemporary constitutional design. Much of the general discussion is a condensed version of the theory worked out in Part I of Perfecting Parliament, but the particular models used to motivate the discussion, its application to separation of powers, and the empirical tests using contemporary constitutions are new. The empirical evidence developed in the last part of the paper is consistent with the constitutional exchange and survivorship perspective. Essentially all of the governments in the IAEP sample have written constitutions and are based on the king-and-council template. Most have constitutional courts and exhibit intermediate divisions of authority among the various centers of governance.
In the long run, the divisions of authority that we observe tend to be less formal and less separate than many normative constitutional theorists might wish, because constitutional and quasi-constitutional exchange tends to shift authority within and among all the centers of governmental authority. New circumstances, new theories, and new norms may require modifications of existing distributions of authority if a nation is to survive and prosper.
There is, of course, no assurance that every reform is an improvement. A fairly well-designed constitution can be undermined by a series of reforms (as arguably happened to the Weimar constitution in the late 1920s and early 1930s). However, constitutional exchange can also enhance the effectiveness of a constitutional design by shifting authority about an organization in a manner that increases its effectiveness or enhances its normative appeal. Indeed, Congleton (2011a) suggests that the emergence of Western democracy and its associated prosperity were largely consequences of fortuitous constitutional bargaining and reforms adopted during the nineteenth century. The present paper suggests that the theory of constitutional origin and bargaining worked out to explain the emergence of Western democracy can also shed light on the nature and division of authority present in contemporary constitutions.
With respect to the separation of powers doctrine, the analysis provides a positive rationale for the emergence of three-branch governments in polities with liberal constitutional designs, but it also suggests that a complete separation of authority among the branches is very unlikely to persist in the long run. Authorities will tend to be shared more than the separation of powers doctrine supports.
See Vile (1968) for an excellent intellectual history of normative theories of the separation of powers doctrine from the sixteenth century through the post-World War II period. See Buchanan and Tullock (1962) for the classic rational choice-based analysis of the constitutional architecture of the United States. Congleton (2011a, b) suggests that the American division reflects both liberal constitutional theory and a century of constitutional experimentation in the British colonies.
Congleton (2011a, chaps. 5–6) provides a more technical analysis of bargaining over veto and agenda control in a setting of exogenous shocks—an analysis that persons who are skeptical of the Edgeworth box representation of constitutional exchange should consult. It shows how changes in circumstances (initial conditions and/or probability distributions) may change policies and also interests in bargaining over policy-making authority.
In the two-person case, this analogy is less perfect, because whichever person has more than half the shares would control a company if it used majority rule. Thus, only four assignments of authority are possible in two firms in which shares are divided between two persons, and only two of that involve sharing overall authority. The analytics of divided authority can be very complex. See Congleton (2011a, b) for an effort to characterize how assignments of veto power and agenda control might be exchanged.
Bargaining itself might also be analyzed, as with a Tullock success function or Nash bargaining game. Such an analysis would, however, only partly characterize the stochastic path of divided authority that tends emerge through time as exogenous shocks are experienced. Alternative theories of bargaining and amendment could then be tested using fine-grained constitutional time-series data, although it would be the shocks as much as the bargaining that determine the course of constitutional development.
Indeed, the necessity of sleep would make holding on to power by oneself, even with very loyal slaves, exceedingly difficult.
The actual exercise of authority, however, does require the use of scarce resources, including physical ones as well as what might be called psychological and sociological ones.
Congleton (2011a) provides a good deal of historical evidence that Western democracy emerged from such bargaining.
It bears noting that a constitutional court can also violate its constitution. For example, it can (indirectly) transfer authority to the judicial domain from the legislative and executive domains by both blocking legislation and expanding its jurisdiction (as arguably occurred with Marbury v. Madison  in the United States, and many times afterward).
An interesting constitutional device for escaping from constitutional constraints was developed by Saddam Hussein, who modified the pre-existing constitution of Iraq to exempt the highest office (himself)) from constitutional constraints. Article 38c of the 1990 constitution grants the Revolutionary Command Council the authority for “Accusing and prosecuting members of the Revolutionary Command Council, Vice-Presidents, and Ministers.” Note that the president is not so constrained by the council under the 1990 constitution. (http://www.servat.unibe.ch/icl/iz01000_html#A038_). Not surprisingly, there was no supreme court in Iraq under the 1990 constitution. It was otherwise a fairly liberal constitution.
The piecemeal creation of federations and confederations, as opposed to “whole cloth” constitutional design, is historically more common than the reverse. Historically, most towns had their own governments and these had been “glued” together through confederal contracts, treaties, conquest and the threat of conquest, and marriages (in Europe). As a result many decentralized systems have different starting points and different initial divisions of authorities between the central and local governments.
Nonetheless, even in autocracies, most towns have some independent authority to write and enforce their own laws, although those laws cannot ordinarily overturn those of “higher” levels of government. (This is essentially what the term “higher level” means in most cases.) Historically, the extent of local authority varies by country and place, partly because of differences in starting points but also because different quasi-constitutional bargains have been struck through time.
- 12.A useful rule of thumb for welfare-enhancing decentralization is the principle of subsidiarity embodied in Article 5(3) of the Treaty on European Union. Paraphrasing slightly, it can be written as follows:
The [central government] shall act only if and insofar as the objectives of the proposed action cannot be sufficiently achieved by the Member States, either at the central level or at the regional and local level, but can rather, by reason of the scale or effects of the proposed action, be better achieved at [the central government] level.
In many cases, however, the central government is constitutionally advantaged in these exchanges (through various superiority clauses and a broader tax base), which tends to promote the gradual centralization of policymaking authority beyond that recommended by economic analysis.
The estimated coefficient for the interaction between executive ability to propose amendments and age of constitution is less accurately estimated than the interaction between legislature ability to propose amendments and constitutional age. Introducing these two interaction terms caused the value for authority of the legislature to cease being significantly different from zero. Dropping that variable and the also insignificant dictator variable caused the interaction term between executive authority to propose amendments and constitutional age to become statistically significant.
- Ahmad, E., & Brosio, G. (Eds.). (2006). Handbook of fiscal federalism. Cheltenham UK: Edgar Elgar Publishers.Google Scholar
- Buchanan, J. M., & Tullock, G. (1962). The calculus of consent. Ann Arbor, MI: University of Michigan Press.Google Scholar
- Congleton, R. D. (2011a). Perfecting parliament: Liberalism, constitutional reform and the rise of western democracy. Cambridge: Cambridge University Press.Google Scholar
- Congleton, R. D., & Rasch, B. E. (2006). Amendment procedures and constitutional stability. In R. D. Congleton & B. Swedenborg (Eds.), Democratic constitutional design and public policy. Cambridge: MIT Press.Google Scholar
- Hobbes, T. (1651/1982). Leviathan. Baltimore: Penguin Books.Google Scholar
- Mao, Z. (1938/1990). Quotations from Chairman Mao Tse Tung. San Francisco: China Books and Periodicals.Google Scholar
- Oates, W. E. (1972). Fiscal federalism. New York, NY: Harcourt, Brace, Jovanovich.Google Scholar
- Vile, M. J. C. (1968). Constitutionalism and the separation of powers. Oxford, UK: Oxford University Press [reprinted by Liberty Fund in e-book form in 2012].Google Scholar
- Voigt, S. (1999). Explaining constitutional change: A positive economic approach. Cheltenham: Edward Elgar.Google Scholar