If you are interested in this comment you may perhaps also be interested in the book "Towards a Science of States: their Evolution and Properties".
Erik Moberg ©
The subject matter
The majoritarian-consensus contrast
The main results
The logic of the first result
The properties of some major institutions
Two conceptions of politics
Implications for Lijphart's thesis about logical connections
More about one-party consensus cabinets
More about multiparty majoritarian cabinets
Conclusions about Lijphart's results
THE SUBJECT MATTER
How does the performance of a democracy depend on its institutional characteristics? This is the main problem discussed in Arend Lijphart's Patterns of Democracy. The investigation is theoretical as well as empirical. The empiricism covers all countries in the world which Lijphart classifies as democracies, altogether 36, from 1945 or, for some countries, from the first democratic election if it occurred later than 1945, to 1996.
Lijphart (p 3f) describes the institutions of the countries studied by means of ten variables, belonging to two groups with five variables in each - a particular country is thus characterized by a value for each variable. The first group of variables is presented in table 1, and the second one in table 2. In the empirical parts of the investigation Lijphart relies on operational definitions for all of these variables. Some of these definitions are rather well-known, such as the Laakso-Taagepera index of the effective number of parties for variable 3, while some other ones are developed by Lijphart for the purposes of the investigation.
Going from institutional characteristics to performance
I will first notice the important fact that this issue is a main concern
in Lijphart's analysis. Most political scientists don't discuss the welfare
effects of political systems, or political decision-making, at all, and
Lijphart's work, in this respect, thus stands out in contrast, and commendable
contrast I would say, to main stream political science.
|Institutions, or institutional characteristics, described by the variables||
|1. Cabinets||Concentration of executive power in single-party majority cabinets||Power-sharing in broad multiparty coalitions|
|2. Executive-legislative relations||Executive-legislative relationships in which the executive is dominant||Executive-legislative balance of power|
|3. Party systems||Two-party systems||Multiparty systems|
|4. Electoral systems||Majoritarian and disproportional electoral systems||Proportional representation|
|5. Interest groups||Pluralist interest group systems with free-for-all competition among groups||Coordinated and "corporatist" interest group systems aimed at compromise and concertation|
Even so, however, it is not quite clear for me which parts
of his discussion Lijphart considers to be about performance. In the very
beginning of his work he emphasizes (p xii) the importance of the "so what?"
question. Does "the type of democracy make a difference for public policy
and for the effectiveness of government"? What he has in mind here is the
handling of such matters as macroeconomic problems, violence, gender issues,
welfare policies, environmental problems, criminal justice and foreign
aid. All of this is clearly about performance. Still, the differences which
Lijphart finds with regard to these criteria are to a considerable extent
small, and to a large extent also, in my view, open to very different interpretations.
Furthermore the discussion about these criteria has an ad hoc character
- it comes after the presentation of the main results, of the main discussion
in the book, in chapter 14. For all these reasons I shall not discuss the
criteria mentioned any further. I will rather concentrate on what, at least
for me, is another important aspect of performance, namely the majoritarian-consensus
|Institutions, or institutional characteristics, described by the variables||
|6. Division of power||Unitary and centralized government||Federal and decentralized government|
|7. Parliaments and congresses||Concentration of legislative power in unicameral legislature||Division of legislative power between two equally strong but differently constituted houses|
|8. Constitutions||Flexible constitutions that can be amended by simple majorities||Rigid constitutions that can be changed only by extraordinary majorities|
|9. Constitutions||Systems in which legislatures have the final word on the constitutionality of their own legislation||Systems in which laws are subject to a judicial review of their constitutionality by supreme or constitutional courts|
|10. Central banks||Central banks that are dependent on the executive||Independent central banks|
THE MAJORITARIAN-CONSENSUS CONTRAST
This contrast plays a very important role in Lijphart's analysis - in fact it almost seems to constitute a backbone of the book. Still, I am not sure that the discussion about majoritarianism and consensus, for Lijphart, really is about performance, as it is for me. Even so, Lijphart and I certainly agree that the contrast is of crucial importance, and that, indeed, is enough.
But even if the majoritarian-consensus contrast is important it is also, in Lijphart's version, quite complex. In my description of it I will therefore start with some relatively obvious and simple matters, and then, from there, bring in some of the complexities. At first, it may then be stated that the majoritarian-consensus contrast represents the extremes on a scale, which also contains the continuum in between, and which may be used for characterizing democracies. The most important property of a majoritarian democracy is that a mere majority of the people governs. "The majoritarian model", according to Lijphart (p 2), "concentrates political power in the hands of a bare majority". In a consensus democracy, in contrast, it is required that as many people as possible take part in the governing - the simple majority rule is accepted only as a minimum. The consensus model, says Lijphart (p 2), aims "at broad participation in government and broad agreement on the policies that the government should pursue."
Now, it is interesting to note how close this part of the majoritarian-consensus criterion, which I have presented so far, is to similar welfare criteria used by economists. Obviously the ideal of consensus is very close to the ideal of unanimity, advanced in particular by Knut Wicksell (1896), and later on by James Buchanan and Gordon Tullock (1962). The ideal or criterion of unanimity, in turn, is closely related to the well-known criteria of Pareto and Kaldor-Hicks. It is significant that Lijphart, in support of his consensus ideal mentions still another economist, namely Arthur Lewis (p 31). The similarity of these various criteria means that Lijphart has entered upon a way, which ultimately may lead to an integration of political science and economics. This, by itself, should be noted as an important merit of Lijphart's work.
So far, as I have indicated, I have however only presented one part of Lijphart's majoritarian-consensus criterion, or one of its two dimensions, as he says. According to Lijphart the concentration of power, which is characteristic of the majoritarian model, may be avoided in two ways. "The consensus model", he says (p 185), "is characterized by non-concentration of power, which can take the two basic forms of sharing of power and division of power (my italics). These two forms provide the theoretical underpinnings of the two dimensions of the majoritarian-consensus contrast. The crucial distinction is whether in consensus democracy power is dispersed to political actors operating together within the same political institutions or dispersed to separate political institutions."
Now, using this distinction, what we have discussed so far is dispersal of power to actors operating within the same institutions, and the degrees of dispersal are primarily represented by shares of actors required for taking a decision - the larger the share the greater is the dispersal of power, and the greater the resulting consensus. Lijphart calls this dimension of the majoritarian-consensus contrast the joint-power dimension (p 185), and the five variables in table 1 above are, according to him, associated with this dimension. The other dimension, the second one, is called the divided-power dimension. Dispersal of political power along this dimension means that the power is distributed to separate political institutions - for instance to the institutions belonging to the federal level, and the state level, in a federal system. The variables in table 2 above are associated with this second dimension.
Here I will immediately say that Lijphart's ambition to
link the first dimension, the joint-power one, with the majoritarian-consensus
contrast, in one way or another, makes a lot of sense to me, whereas this
is hardly so for the same ambition concerning the divided-power dimension.
When power is dispersed to a number of separate political institutions
something certainly happens to its concentration, but is it really fitting
to describe this in terms of consensus? The possible results of such a
dispersal may in fact be just opposite to the ones implied by this terminology.
Consider for instance William Riker's assertion (1964) that a federal system
contains mechanisms which make it possible for minorities to impose their
will on majorities, and in that sense make the minorities decisive. What
Riker has in mind in this context, it should be noted, is not the simple
idea that minorities may block changes of the status quo, but the much
more remarkable idea that minorities, in fact, may dictate changes of the
status quo. The examples he presents are all about how an overwhelming
majority of states reluctantly are forced to adopt a policy introduced
by a single, deviating state. If Riker is right in this a movement from
centralized and unitary government to federalism would not mean a movement
to more consensus but rather the opposite. This, I think, is an important
THE MAIN RESULTS
Having introduced the institutional variables and the majoritarian-consensus contrast or scale, I will now summarize the results of Lijphart's work. The presentation differs somewhat from Lijphart's own, both with respect to what I bring together under each point, and with respect to the ordering. Still I hope and believe that Lijphart would agree with the substance. These are the results:
Result 1. All ten institutional variables are associated with the majoritarian-consensus variable. In one place Lijphart thus says (p ix) "that the main institutional rules and practices of modern democracies ... can all be measured on scales from majoritarianism at one end to consensus on the other .. ". In other places (p 2f) he claims more distinctly that the institutional variables are logically connected with the majoritarian-consensus scale. Broadly speaking positions close to the extremes represented by the "from"-columns in the tables 1 and 2 above are connected with majoritarianism, and positions close to the extremes represented by the "to"-columns are connected with consensus. This result is considered by Lijphart, as I understand him, as the most important one of his investigation. But even so its nature is not entirely clear - the problem is the exact meaning of the logical connection. I will come back to this matter in the next section.
Result 2. The variables in the first group are empirically significantly correlated with each other (chapter 14). This correlation is explained by Lijphart by the logical connections stated in result 1. In essence he says (p 2) that since the characteristics "are logically connected, one could also expect them to occur together in the real world".
Result 3. The variables in the second group are also empirically significantly correlated with each other (chapter 14), and again Lijphart explains with the logical association stated in the result 1.
Result 4. The variables in the two groups are not
empirically correlated with each other (chapter 14). Given the logical
associations for all ten variables, as stated in result 1, this
should perhaps, within the framework of Lijphart's ideas, be a bit embarrassing.
Lijphart, however, just registers the fact of the non-correlation without
noticeable concern. Basically he just says (p 2) that "the variables cluster
in two clearly separate dimensions" although "(a)ll ten variables could
... be expected to be closely related".
THE LOGIC OF THE FIRST RESULT
Result 1 thus is about logical connections. A first question about these connections concerns their direction - what is deduced from what? Lijphart's most explicit statement about this matter is, I think, this one (p 2, italics mine): "Ten differences with regard to the most important democratic institutions and rules can be deduced from the majoritarian and consensus principles. ... the majoritarian characteristics are derived from the same principle ... and ... the same applies to the consensus characteristics." This quotation means that the ten institutional characteristics in the "from"-columns of the tables 1 and 2 above could be deduced from the majoritarian principle, and those in the "to"-columns from the consensus principle. Or, in other words, the deductions are from the principles, which form the premises, to the institutional characteristics, which constitute the conclusions.
But even if Lijphart thus explicitly claims that there is a logical direction from the principles to the institutional characteristics he also presents, here and there, arguments in the opposite direction, i.e. arguments leading from institutional traits to principles. A good example is the following chain of arguments (p 10): "Because in the British two-party system the two principal parties are of approximately equal strength, the party that wins the elections usually represents no more than a narrow majority, and the minority is relatively large. Hence the British one-party and bare-majority cabinet is the perfect embodiment of the principle of majority rule: it wields vast amounts of political power to rule as the representative of and in the interest of a majority that is not of overwhelming proportions. A large minority is excluded from power and condemned to the role of opposition." In this deduction it is concluded, basically from the presence of two parties and a one-party bare-majority cabinet, that the majoritarian principle reigns. The deduction is thus from institutional characteristics, forming the premises, to a principle, constituting the conclusion.
Now, before going further, I will make a few comments about these directions. The first direction is in fact, I think, quite remarkable. Let us try to carry out a real deduction, step by step, by starting, for instance, from the majoritarian principle. This principle, we remember, basically says that a bare majority of the people should rule, disregarding the opinions of the minority. Starting with this we should thus, by deduction, arrive at detailed conclusions about the number of parties, about electoral proportionality, about central banks, and so forth. How is this possible? Why, for instance, should there be any parties, elections of representatives, or central banks at all? In which way does the presence of these institutions follow logically from the majoritarian principle? In no way I think. Deductions of the kind mentioned cannot, I submit, be done - and this contention is strongly substantiated by the fact that Lijphart does not present any such deductions. He just presents (p 3f) the alleged results of the deductions, but not, anywhere, the detailed arguments leading from the principles to the institutional traits. The second kind of deduction, where one starts with some institutional characteristics, and tries to find out whether the result is likely to be majoritarian or consensual seems to me, on the contrary, and in principle, perfectly legitimate.
For illuminating the difference between the two kinds of alleged deductions, represented by the two directions, a few words about deductions in general may be in order. Thus, when we are dealing with deductions in an ordinary, conventional verbal context, as is the case here, the conclusions do not usually follow, in a strict logical sense, from the premises explicitly presented. As a rule there are also a number of premises which, although necessary for the conclusions, are not explicitly stated. The reason may be that they are taken for granted for one reason or another, for instance because they are considered as obviously true, and therefore, if presented, just would overload the text without adding anything of value. These things are, however, matters of judgement. Sometimes so much of the premises may be implicit, rather than explicit, that the reader fails to be convinced. Sometimes such a reader may even be right in the sense that there is no reasonable set of implicit premises which, if added, makes the deduction logically valid. For me the first type of deduction above, the one from principles to institutions, lacks validity in exactly this sense. The second one, however, from a particular institution to principles, is much more reasonable in the sense that the missing premises seem to be fewer and more reasonable, or, in other words, much easier for the reader to imagine and add. Still, as I will argue in the following, I hold that the second deduction, even if reasonable in principle, is in fact false.
Anyway, there are two possible directions for the deductions
we are considering. In principle it could be affirmed that it is the first
one which could and should be used, or the second one, or both. Lijphart,
as we have seen, explicitly claims that the first is the one which is important
in his work. This direction is however also a bit absurd, and in fact Lijphart
rather practices the second one. Lijphart's position is thus somewhat unclear.
He may be a proponent of the first direction, or the second, or perhaps
both. Whatever the case, it will, however, as I shall show later on, lead
THE PROPERTIES OF SOME MAJOR INSTITUTIONS
Now and then Lijphart talks about the majoritarian-consensus
properties of specific institutions. His most emphatic statements about
such properties are, in my reading, related to party systems, cabinets
and electoral regimes, and this is the reason why I have chosen this complex
for detailed scrutiny here. He says (p 90), for instance, about the first
variable in table 1 above, dealing with cabinets, that it "can be regarded
as the most typical variable in the majoritarian-consensus contrast: the
difference between one-party majority governments and broad multiparty
coalitions epitomizes the contrast between the majoritarian principle of
concentrating power in the hands of the majority and the consensus principle
of broad power-sharing." In another insistent formulation he asserts (p
143) that "(t)he plurality and majority single-member district methods
are winner-take-all methods - the candidate supported by the largest number
of voters wins, and all other voters remain unrepresented - and hence a
perfect reflection of majoritarian philosophy."
||non-existent by definition|
In the following of this section, with its subsections, I will try to show that these claims of Lijphart are unfounded, and indeed even unlikely to be true. I will start by presenting Lijphart's position in a more exact way, and for that purpose I will use his classification of cabinets as shown in table 3. The classification, which is exhaustive and non-overlapping, is based on two different distinctions: the first one between multiparty and single-party cabinets, the other one between oversized, undersized and minimal winning cabinets. The latter concepts may be defined as follows:
A cabinet is undersized if it represents less than a decisive constellation in the legislature.
A cabinet is minimal winning if it becomes undersized by losing any of its party-members.
A cabinet is oversized if it is neither minimal winning, nor undersized. In other words it is oversized if it has at least one party-member such that, the loss of that member, does not make the cabinet undersized. (An immediate consequence of this definition, as indicated in the table, is that a single-party cabinet cannot be oversized.)
Here, before going further, a comment on the fact that the definitions are phrased in terms of decisive constellations is necessary. The point is important since Lijphart just talks about majorities (p 90). The difference is that the concept of "decisive" constellations indicates that some decision rule is implicitly assumed, whereas this is not necessarily so for the "majority" concept. Let us, for instance, assume that the rules stipulate that a qualified majority of 60 % is required for taking decisions. If so, in order to be minimal winning, a constellation of at least 60 % of the legislators is required - a good 50 %, if that is what is meant by a majority, is not enough. I hope and believe that Lijphart agrees with this. If, however, he holds that support just exceeding 50 % is always enough for being minimal winning, then our definitions are not exactly the same.
Now we are in a position to state what might be called Lijphart's two main claims about institutions more exactly. The first one (p 91, my italics) is that "(t)he most majoritarian type of cabinet is one that is single-party and minimal winning - that is, a one-party majority cabinet." Some pages later on (p 103) he similarly claims that "(i)t is clear that minimal winning and one-party cabinets represent majoritarian characteristics ...". This claim is indicated with an M in the table.
The second one (p 91, my italics) is that "(t)he most consensual type of cabinet is multiparty and oversized." Later on (p 103) he asserts that "(i)t is clear that ... oversized and coalition cabinets express consensus traits." This claim is indicated with a C in the table.
Both of these claims are, I think, disputable, and I shall
now proceed to show that. As a first step in that undertaking I will introduce
a few concepts and ideas in the next two sections.
Two conceptions of politics
Lijphart's classification of cabinets, as he himself informs, is inspired by William Riker (1962). Riker's coalition theory contains, in particular, the prediction that the kind of cabinets which will, in fact, appear are the minimal winning ones. Therefore I will say a few words about this theory, and also about another theory, which is likewise present in many of Lijphart's arguments, namely the one which contains the median voter theorem. These two theories deal, in fact, with different kinds of politics. Still, the important differences are often overlooked, and the two theories are invoked almost simultaneously as if they are applicable in the same contexts, which creates confusion. (The topic discussed here is treated more extensively in the paper Ideological Politics and Interest Politics, in the department Arguments, on this site.)
Riker's coalition theory explicitly belongs to the wider class of zero-sum game theory. The activities studied within this theory can be considered as redistributions of resources among the participants - the sum of the losses, for those hit, equals exactly the sum of the gains, for those favored. In politics we may, for instance, imagine that some groups of citizens, which may be represented by political parties, tax some other groups and share the proceeds. Since some kind of majority rule, often simple majority, is usually used in politics this kind of behavior is, in principle, quite possible. From this it also follows that the exploiting majority, for being as successful as possible, should be as small as possible granted that it remains decisive. When these conditions are fulfilled the number of those exploited is maximized, and the number of those sharing the spoils is minimized - and since people are rational this is the likely, and thus the predicted, outcome. This, in essence, is the theorem of minimal winning coalitions. An obvious, possible application is cabinet formation in democratic systems.
The median voter theorem belongs to an entirely different conception of politics, the spatial one. The fundamental feature here is a scale on which the actors, for instance politicians or voters, take positions according to their views or preferences. The scale is often of an ideological nature, for instance a left-right scale. Decisions are also characterized by their positions on the scale. The closer, on the scale, a decision is to the position of a particular actor, the better it is for that actor - and this is all what politics is about. It is in this world that the median voter theorem holds. The theorem says that, if the simple majority rule is used, then the decision will be the position of the median voter - i.e. the voter having as many voters on her one side of the scale as on her other side. In this world redistributions are no necessary results of the political activity although they may occur as results of the nature of the scale. If, for instance, a leftist position means that the rich should be taxed and the money given to the poor, then a leftist decision will result in that kind of redistributions. This, however, is no necessary attribute of the model. Even with a scale, which does not entail any redistributions at all, the theorem is equally valid.
Summarizing we may thus say that the theorem of minimal
winning coalitions belongs to a conception of politics in which zero-sum
redistributions are fundamental, and in which political ideologies may
be completely absent. This may also be characterized as politics of short-term
interests, or just interest politics. The median voter theorem, on the
contrary, belongs to a conception of politics in which positions on a scale,
and thereby probably ideologies, are fundamental, but where short-term
interests may be completely absent. This, thus, is ideological politics.
For avoiding misunderstanding I shall add that the two kind of politics,
as I see it, certainly can appear together, as different components, in
one and the same country. The politics of a country may thus contain both
interest elements and ideological elements. Still the two kinds of elements
should, as far as possible, be kept apart in the analysis. The two conceptions
of politics are, considered as conceptions, entirely different.
I shall now throw some more light on the majoritarian-consensus contrast by means of two examples. The first example is about a parliamentary country which has majority elections (first past the post) and, partly as a consequence of that, a two-party system. This country, therefore, will usually also have one-party cabinets. I will furthermore assume that the policies considered by the parties may be represented by an ideological scale, say a left-right scale. We are thus dealing with ideological politics. I will also assume that the policies of both parties, according to the median voter theorem, are close to the median voter position. This means that the leftist party stands for the policy represented by its right wing, and that the rightist party, correspondingly, delivers the policy of its own left wing. Granted that these assumptions hold, it is also, I would argue, quite reasonable to talk about consensus politics. The basic reason is that the median voter position is the median position of all voters, and that therefore, in that particular sense, all voters affect the outcome.
The second example is about a parliamentary country with
proportional elections, and a multiparty-system as a consequence of that.
The multiparty-system, in turn, means that the country usually has coalition
cabinets. I will furthermore assume that the agreement, which forms the
base for such a cabinet, essentially is about mutual interest support.
Each party member of the cabinet thus promises the other members to support
their interests in return for support of its own interests. If so we are
dealing with interest politics, and the theorem of minimal winning coalitions
is applicable. We have a typical majoritarian situation: the majority,
represented by the cabinet, satisfies its own interests at the expense
of the outside minority - the outside minority is exploited by the cabinet
Implications for Lijphart's thesis about logical connections
Now, I am not arguing that the characteristics of these two examples are necessary consequences of majority systems with single-party cabinets, and proportional systems with coalition cabinets, respectively. That is hardly the case. I do argue, however, that the two examples are at least quite possible, and furthermore that this possibility has important consequences for Lijphart's thesis about logical connections between institutional characteristics and the majoritarian and consensus principles, whatever the direction of those connections. This can be demonstrated by the following arguments.
If a situation like the one described in the first example
is possible, then majority elections and consensus may obviously coexist.
But if such a coexistence is possible, it is excluded
(i) that majority elections implies majoritarianism, and
(ii) that consensus implies proportional elections.
Similarly, if a situation such as the one in the second
example is possible, then proportional elections and majoritarianism may
coexist. And if so it is excluded
(i) that proportional elections implies consensualism, and
(ii) that majoritariansim implies majority elections.
Here the two (i)-consequences, taken together, obviously disprove the idea that it is possible to derive the principles specified by Lijphart from the institutions he discusses, and in the same way the (ii)-consequences make the idea of deductions from the principles to the institutions impossible. Earlier I said that it was unclear which direction Lijphart actually had in mind when talking about logical connections. Now, however, we see that the direction does not matter. The idea of possible deductions is wrong whichever direction one considers.
But even if the idea about logical connections thus is
wrong, Lijphart may still consider the situations illustrated by the two
examples above quite unlikely. Obviously, as indicated in table 3 above,
he also does so. In the following two sections I will try to show that
this position, at the very least, can be challenged.
More about one-party consensus cabinets
The salient traits in the first example above were the two-party system, the one-party cabinet, the scale of the spatial model, and the resulting median voter politics. Now, it is interesting that some or all of these traits occur together in the writings of several political scientists and economists.
Harold Hotelling thus wrote (1929) that competing political parties, at least when there are just two of them, tend to take positions very close to each other in the middle of the political spectrum. Thus, "(t)he competition for votes between the Republican and Democratic parties does not lead to a clear drawing of issues, an adoption of two strongly contrasted positions between which the voter may choose. Instead, each party strives to make its platform as much like the other's as possible."
Elmer Eric Schattschneider similarly claimed that a two-party system produces moderate parties. "When one stops to consider the amount of thought and energy that has been devoted to the effort to protect people against oppression", he wrote (1942, p 85), "it is difficult to imagine anything more important than the tendency of the parties to avoid extreme policies." Schattschneider also brings A. Lawrence Lowell into his own argument in this way: "A generation ago President Lowell (he was president of Harvard University), writing about English major parties, said that the Liberal and Conservative tended to move toward the political center of gravity, i.e., they tended to be alike. Indeed, the most common criticism made of the American parties is not that they have been tyrannical but that they have been indistinguishable. ... (There is a strong) tendency of the parties to move toward the middle of the road."
By the same token V. O. Key, Jr., stated (1964, p 220) that "Each party leadership must maintain the loyalty of its own standpatters; it must also concern itself with the great blocks of voters uncommitted to either party as well as with those who may be weaned away from the opposition. These influences tend to pull the party leaderships from their contrasting anchorages toward the center. In that process, perhaps most visible in presidential campaigns, the party appeals often sound much alike and thereby contribute to the bewilderment of observers of American politics."
Maurice Duverger also thought along these lines. "Let us take", he wrote (1964, p 387 f), "a precise example, that of contemporary Britain, neglecting the Liberal party which is no longer important. Who decides whether the Conservative or the Labour party shall win the election? Not their fanatical partisans who, being unable to cast their vote for any party further to the Right or to the Left, will naturally vote for them whatever they do, but the two or three million moderate Englishmen, politically situated at the Centre, who vote sometimes Conservative, sometimes Labour. To win their votes the Conservative party is forced to attenuate its Conservatism and Labour its Socialism, both of them adopting a calm tone, a reassuring aspect. Both will have to draw up policies clearly aimed at the Centre and therefore profoundly similar. We arrive at the paradoxical situation that the Centre influences the whole of parliamentary life in the very country in which the electoral system prevents the formation of a Centre party."
It is, in fact, interesting to notice that even Lijphart himself occasionally expresses himself very much like the writers quoted. Thus he says (p 32) about his three models of majoritarianism, Britain, New Zealand and Barbados, "... that their major parties have usually not been very far apart in their policy outlooks because they have tended to stay close to the political center." Furthermore he subscribes to the idea that the politics of these countries to a large extent can be described within the framework of a one-dimensional spatial model. "A corollary trait of two-party systems", he thus writes (p 14), "is that they tend to be one-dimensional party systems; that is, the programs and policies of the main parties usually differ from each other mainly with regard to just one dimension, that of socioeconomic issues. This is clearly the case for the British two-party system. The principal politically significant difference that divides the Conservative and Labour parties is disagreement about socioeconomic policies: on the left-right spectrum ...".
All of these observations together testify, I think, to
the likelihood and realism of the first example above. This in turn means
that a C (for consensus) is more reasonable than the M (for majoritarian)
in the single-party, minimal winning box of table 3.
More about multiparty majoritarian cabinets
I will now take up Lijphart's contention that multiparty, oversized cabinets are the most consensual ones, and I will start by examining in some detail the mechanisms which, according to Lijphart, bring about oversized cabinets. At first it should then be stated that even for Lijphart the reasons for forming minimal winning cabinets are strong, even if not compelling. He also presents a number of reasons why cabinets may depart from the minimal winning size to oversize, or to undersize. Here I will confine myself to the arguments about oversize.
One possible reason for oversizing, according to Lijphart, is the wish to have a reasonable security for the cabinet's winning status, when some participating parties are not completely reliable. Such a security may be achieved by including one or a few additional parties as an insurance (p 99). This notion, however, is just an elaboration of Riker's main idea, and was even put forward by Riker himself, as Lijphart confirms by quoting Riker. Basically it is not a mechanism leading to oversize, but rather a way of securing minimal winning size under real, uncertain conditions.
Another reason given by Lijphart (p 100) for the appearance of oversized cabinets is the existence of external threats, for instance from hostile foreign countries, or internal threats, for instance from anti-democratic movements. This argument is certainly valid in the sense that threats of this kind may give rise to cabinets with a considerably broader support than a minimal, winning constellation. Still the argument is not of much interest here. A first reason is that the situations, in which the enlarged cabinets occur, in fact may be interpreted as corroborating the idea of minimum winning coalitions. If a formally minimal cabinet is threatened in the way assumed this means that there is, in reality, a game which is wider than the normal decision-making one in the legislature and which includes, as participating actors, those who threaten. In this wider game a cabinet, which from a conventional point of view is oversized, may in fact be minimal winning, or, if there are not enough resources for forming even a minimal winning constellation, undersized. Thus, it is perfectly possible to interpret the situations described here in a way which fundamentally is in complete agreement with Riker's theory. Another more important reason for neglecting the argument here is that it is of very limited relevance, since situations with threats of the kinds described are quite exceptional. In this context where we are interested in normal, regular democratic patterns it could, therefore, be disregarded. Our interest here is the rule, not the exceptions.
Still another reason for oversized cabinets given by Lijphart is "particular institutional provisions" (p 102) and he exemplifies with Colombia and Belgium. These provisions basically require that some specified parties, or parts of the population, should be represented in the cabinet. This, however, does not necessarily mean oversize. True, the cabinets are likely to be bigger than in most other countries, but this may just be a reflection of the constitutional rules. The minimal winning concept as defined above, we remember, is related to the constitutional rules, and with the rules mentioned a cabinet usually has to be supported by substantially more than a simple majority in order to be minimal winning. It is only when a cabinet is still broader than that, that it is oversized. The same comment can be made about Lijphart's contention (p 103) that countries, which require qualified majorities for some important decisions, tend to have oversized cabinets. Again, what we see is not necessarily oversize - it may just be an adaptation to the unusually severe rules for being minimal winning. Still Lijphart is certainly right in his assertion that the minimal winning cabinets described here are less majoritarian, and more consensual, than ordinary simple majority, minimal winning cabinets. The purpose of the constitutional rules mentioned probably is to achieve a certain amount of consensus, and that goal is in all likelihood achieved. Anyway, the mechanism discussed is, as the former one, quite exceptional and therefore of limited relevance. Most democratic countries do not have rules of the kind mentioned.
Lijphart also advances another entirely different reason
for oversizing, namely the wish of each party to occupy a position in the
middle of the cabinet, and thereby to strengthen its own position. "In
reality", Lijphart writes (p 100), "the parties' policy preferences may
exert strong pressures to enlarge instead of to minimize the size and range
of coalitions. Each party naturally prefers to form a cabinet that will
follow policies close to its own policy preferences; a cabinet in which
it participates with parties of about equal weight on both its left and
its right is ideal in this respect." This reason, in contrast to the two
last ones above, is not exceptional. If it is valid at all it applies generally
to all multi-party democratic regimes. The validity of the argument can
however be questioned on two grounds. First, since Riker's basic prediction
about minimal winning coalitions is derived within a game-theoretical framework,
it seems logical that arguments about oversize rather than minimal winning
size, are formulated within that framework as well. Lijphart's argument
does however belong to a completely different framework, namely the spatial
one. Second, even within the spatial framework it is far from clear that
the incentives of the parties are such as to bring about cabinets, which
are considerably bigger than minimal winning.
Suppose, for instance, that we have a situation like the one in figure 1, where six parties A-F are characterized by their positions on a left-right scale, and by their number of seats in a legislature with altogether 100 seats. Now, assume that there is a cabinet which consists of the parties B, C and D, and which conducts the policy P. Assume furthermore that D is unsatisfied with this policy and therefore tries to move it further to the right by getting E into the cabinet as well. Certainly this may be attractive for E since a policy to the right of P is better for E than P. For the other parties things are, however, different. Thus for B and C, which are members of the original cabinet, an adjustment of the policy to the right is a deterioration, and these parties are therefore not likely to support D in its efforts to get E into the cabinet. Actually it seems as if any proposal about a changed composition of the cabinet runs into some kind of difficulties like this, and in fact the original assumed composition of the cabinet, having the members B, C and D, is not tenable either. It is refuted by a theorem saying that it is impossible to account for a coalition cabinet within the framework of a one-dimensional spatial model. (This theorem is presented in the section "The impossibility of purely ideological coalition executives" at the end of the paper Ideological politics and interest politics, in the department Arguments, on this site.)
As for Lijphart's arguments for oversize we may thus summarize that they are considerably less far-reaching than one might have presumed. The only regular, non-exceptional, case which passed the scrutiny above was oversize as a security for the cabinet's winning status. This, however, as I argued, is hardly oversize at all, but just an adaptation to real, uncertain conditions - essentially the cabinet is still minimal winning. Anyhow, there is no reason to believe that this kind of oversize, even if we accept to call it so, will be more than marginal. There is no reason to believe that it will bring about the "broad" cabinets which Lijphart talks about, for instance when characterizing variable 1 in table 1 above. It may also be noted that Lijphart, when he reports the empirical finding that oversize cabinets are quite common (p 98), says nothing about the magnitude of oversize. The oversized cabinets behind the figures may thus, to a large extent, be just marginally oversized.
We may thus conclude that although Lijphart's whole idea
about consensual coalition cabinets relies on a fairly common occurrence
of broad, oversized cabinets, he fails to show that the world looks like
that. His survey of incentives for creating oversized cabinets does not
support his idea, neither does his empirical data. In fact, I think that
coalition politics is more likely to enhance majoritarianism than consensus.
The reason is that coalition politics tends to be about interests rather
than ideology, since it is easier to reach agreements about interests than
about ideologies (this theme is further developed in Moberg, 1999). It
is even, as mentioned above, impossible to account for the appearance of
a coalition-cabinet within the framework a one-dimensional spatial model.
Since we are dealing with interest politics, exploitation in Riker's sense
is indeed as likely to be present as consensus seeking. The upshot of all
this is that that an M is at least as appropriate in the multiparty, oversized
box of table 3 above as Lijphart's C.
CONCLUSIONS ABOUT LIJPHART'S RESULTS
The discussion so far has essentially been about various aspects of result 1, and the main conclusion is that Lijphart's association of various institutions with positions on the majoritarian-consensus scale breaks down. In order really to bring this conclusion home it should be underlined that the overwhelming part of Lijphart's empiricism is related to his ten institutional variables, operationally defined. The association of these variables with the majoritarian-consensus contrast, in Lijphart's particular way, relies essentially on his inconclusive logic. After this I will now also say a few words about the results 2, 3 and 4.
Result 2, we remember, is about the empirical correlation of the five institutional variables in table 1 above. On the whole this correlation should not be surprising, and Lijphart himself offers a good explanation for it when he points to the causal connections between four of the five variables. "(E)lectoral systems", he thus affirms (p 181), "shape party systems, which in turn have a strong causal effect on the formation of cabinets, and types of cabinets are further causally related to cabinet duration." Cabinet duration, it should be mentioned, is Lijphart's operational definition of the executive-legislative relations, or the third variable. So far, however, the interest group variable is not brought into the argument, and the problem about the empirical correlation between this variable and the other four thus remains. According to Lijphart (p 181) "(t)he interest group system differs from the other basic variables of the executives-parties dimension in that there is no clear causal connection that links it to the other four variables" and "(t)herefore, the hypothesis that interest group systems are related to these other variables rests entirely on the conceptual correspondence between the corporatism-pluralism distinction and the broad consensus-majoritarian difference." This, I think, is hardly true. It seems indeed quite natural that a coordinated, non-pluralist interest group system, of the kind indicated in table 1 above, goes together with the disciplined, centralized political parties which are likely to exist in a parliamentary system with proportional elections - some arguments for this contention are further developed in Moberg (2000). There is thus no need for invoking the consensus-majoritarian difference in this context.
Going then to result 3, which is about the empirical correlation between the variables in table 2 above, a similar argument can be made. Again there is no need for invoking the consensus-majoritarian difference. There are other good explanations for the correlation as Lijphart himself mentions (p 4ff).
Result 4 finally, which is about the non-correlation of
the variables in two groups, is also quite natural. In fact, without the
consensus-majoritarian contrast, the result is hardly a result at all.
The result would be surprising only if the correlations in the two groups,
that is the results 2 and 3 above, had to be explained by the same factors,
for instance by means of the consensus-majoritarian difference. That, however,
as I have tried to show, is not the case. The correlation in the two groups
are best explained by different sets of factors. The whole idea of "two
dimensions of the majoritarian-consensus contrast" (p 185) is, I think,
In the first sections I complimented Lijphart for bringing in the majoritarian-consensus contrast into his analysis, and I still think that this is appropriate. At the same time however, I think that Lijphart's application of this contrast to his subject matter, and his particular association of various democratic institutions with the contrast, fails to convince. This matter has been the main topic in this comment on Lijphart's Patterns of Democracy and my conclusion is that result 1, as presented above, disappears.
What remains are the results 2 and 3 about two groups
of correlated institutional variables. These results are perhaps not very
surprising, but still they are interesting - and it is good to have them
Buchanan, J M & Tullock, G, 1962, The Calculus of Consent, Logical Foundations of Constitutional Democracy. Ann Arbor: The University of Michigan Press.
Duverger, M, 1964, Political Parties: Their Organization and Activity in the Modern State. New York: Wiley.
Hotelling, H, 1929, "Stability in Competition", in The Economic Journal, vol. XXXIX.
Key, V O, 1964, Politics, Parties and Pressure Groups. New York: Thomas Y. Crowell Company.
Moberg, E, 1999, "The Expanding Public Sector - A Threat to Democracy?", in The Limits of Government - On Policy Competence and Economic Growth, (ed G Eliasson & N Karlson). New Jersey: Transaction Publishers.
Moberg, E, 2000, "The Swedish Model: A Comment on Mancur Olson's Analysis", in A Not-so-Dismal Science - A Broader View of Economies and Societies, (ed M Olson & S Kähkönen). Oxford: Oxford University Press.
Riker, W H, 1962, The Theory of Political Coalitions. New Haven: Yale University Press.
Riker, W H, 1964, Federalism - Origin, Operation, Significance. Boston and Toronto: Little, Brown and Company.
Schattschneider, E E, 1942, Party Government. New York: Rinehart.
Wicksell, K, 1896, Ein neues Prinzip der gerechten Besteuerung, Finanztheoretische Untersuchungen. Jena: Gustav Fischer. An English translation is reprinted in Classics in the Theory of Public Finance (ed R A Musgrave & A T Peacock). London: Macmillan 1967.