Arend Lijphart

The Westminster model and Westminster adapted

The Westminster model has been extremely influential in the shaping of modern democracies, particularly, of course, in democracies that were formerly ruled by Britain. In my recently published book Patterns of Democracy, I analyse the universe of well-established modern democracies, defined as all countries with a population of at least 250,000 that were democratic in the late 1990s and that had been continuously democratic since 1977 or earlier. There are 36 democracies that fit these criteria. Of these 36, no less than 15–42 per cent of the total, not counting the United Kingdom itself—are democracies with a history of being under British rule. [1]

However, in most of these democracies formerly ruled by Britain, the Westminster model was not adopted without various modifications. Anthony Payne uses the term `Westminster adapted' in characterising the governmental systems of the former British colonies in the Caribbean, [2] and this term can be appropriately applied to former British dependencies elsewhere, too. The main institutions of democracy tend to vary among countries along two dimensions, which I have called the executives-parties dimension and the federal-unitary dimension. Where the Westminster model has been influential, it is along the first of these dimensions that it has often been adopted, but along the second dimension that it has usually been adapted, or even changed radically.

The first (executives-parties) dimension groups together five characteristics of executive power, executive-legislative relations, and the party, electoral, and interest group systems. In majoritarian, Westminster-style, democracy, power is concentrated in the hands of the majority, and majoritarian democracy has the following institutional characteristics:

(1) one-party majority cabinets,

(2) executive dominance over the legislature,

(3) two-party systems,

(4) majoritarian and disproportional electoral systems, and

(5) pluralist interest group systems with free-for-all competition among groups.

Consensus democracy, in contrast, is characterised by sharing, dispersing, and limiting power instead of concentrating power, and it has the following typical features:

(1) executive power-sharing in broad multi-party coalitions,

(2) executive-legislative balance of power,

(3) multi-party systems,

(4) proportional representation (PR), and

(5) a coordinated, `corporatist' interest group system aimed at compromise and concertation.

The second federal-unitary dimension also consists of five elements: the federal-unitary contrast, bicameralism vs. unicameralism, the degree of difficulty in amending constitutions, judicial review, and the degree of independence of central banks. Here, too, majoritarian systems are systems of concentrated power:

(1) unitary and centralised government,

(2) concentration of legislative power in a unicameral legislature,

(3) flexible constitutions that can be amended by simple majorities,

(4) legislatures that have the final word on the constitutionality of their own legislation, and

(5) central banks that are dependent on the executive.

Consensus systems typically have

(1) federal and decentralised government,

(2) division of legislative power between two equally strong but differently constituted houses,

(3) rigid constitutions that can be changed only by extraordinary majorities,

(4) laws that are subject to a judicial review of their constitutionality by supreme or constitutional courts, and

(5) strong and independent central banks.

Because not only the first of these differences, but also the other four are commonly associated with the contrast between federalism and unitary government, this second dimension can be called the federal-unitary dimension.

The clearest examples of Westminster-inspired systems that are majoritarian on the executives-parties dimension, but consensual-federalist on the federal-unitary dimension are Australia, Canada, and the United States. But it is true, more generally, that the Westminster model has been much more influential with regard to the first than to the second dimension. For instance, the Caribbean democracies are very strongly majoritarian on the first dimension; with regard to the second dimension, they are unitary instead of federal—they are obviously too small for federalism to make sense—but they do have the other federalist characteristics: bicameral legislatures, constitutions requiring super-majorities for amendment, judicial review, and fairly independent central banks. In fact, the only example of a former British colony that faithfully followed both dimensions of the Westminster model was New Zealand; since the adoption of proportional representation (PR) in 1996, however, New Zealand has moved in the direction of consensus democracy on the first dimension and is therefore no longer a good example of the Westminster model. There are other exceptions to my general proposition that could be noted, but that, for brevity's sake, I shall not mention; the one exception worth pointing out, because of its relevance in the context of this paper, is the use of the single transferable vote (STV) system for parliamentary elections in Ireland and Malta—which makes these countries cases of `Westminster adapted' instead of pure Westminster systems on the first dimension.

STV and the Australian majoritarian/federal system

Then, of course, there is the unusual case of `Westminster adapted' represented by the use of STV elections by the Australian Senate. From the perspective of the two-dimensional majoritarian-consensus contrast, this adaptation makes Australian democracy slightly, but by no means insignificantly, more consensual on both dimensions.

Let me discuss the second (federal-unitary) dimension first. The relevant variable here is the contrast between unicameral legislatures at one extreme (the most majoritarian-unitary characteristic) and strong bicameralism (the most consensual-federal characteristic) at the other extreme. The strength of bicameralism depends on two criteria: the symmetry and incongruence of the two houses of the legislature. Two houses are symmetrical if they are equally powerful and if they are both directly elected by the voters and therefore both enjoy full democratic legitimacy. Two houses of a bicameral legislature are incongruent if they clearly differ in composition. The House of Representatives and the Senate in Australia do not have equal power, but by comparative standards the Senate is a very powerful body, and the relationship between the two houses can therefore be classified as only moderately asymmetrical; moreover, both houses are popularly elected. The two houses are also clearly incongruent in their composition. They already qualify for the label of strong bicameralism in this regard as a result of the equal representation of the states in the Senate in spite of the states' highly unequal populations—a feature of many federal systems. The difference in the methods of election—the majoritarian alternative-vote system for the House of Representatives and PR for the Senate—makes the two houses even more different in composition and reinforces their incongruence. STV therefore has the effect of strengthening bicameralism and also the federalist character of Australian democracy on the second dimension.

As far as the first (executives-parties) dimension is concerned, the election by STV of one chamber of the legislature—the less important but still very strong chamber—adds an element of proportionality to the political system and therefore also at least a small measure of consensus democracy. But how proportional is the STV election system in Australia really, given the divergent effects of PR on the one hand and the equal representation of the states, and hence the highly unequal and disproportional representation of the populations of the different states (and territories), on the other hand? This question can be answered easily because we can measure the degree of disproportionality of the 20 Senate elections conducted by STV so far and compare it with both PR and majoritarian elections held in other countries. Of the several alternative measures of disproportionality that are available, I prefer the index proposed by Michael Gallagher. [3] It measures the total percentage by which the over-represented parties are over-represented (which is, in principle, the same as the total percentage of under-representation), but it does so in such a way that a few large deviations between parties' vote and seat shares are weighted more heavily than a lot of relatively small vote-seat share deviations. The highest percentage of disproportionality in the Australian Senate STV elections was 9.68 per cent (in 1961) and the lowest was 1.44 per cent (in 1980). The average for the 20 elections was 4.15 per cent. (These percentages were calculated on the assumption that the Liberal and National parties can be regarded as one party.)

Table 4.1 shows average disproportionalities of elections in 36 democracies in the post-World War II period (from the first democratic election in each country until 1996). For most countries, the figures shown in the table are the indices of disproportionality in the elections of the lower or only house of the legislature. I made two adjustments, however, that require explanation. First, I counted closely allied parties (the Liberal and National parties in Australia, the Christian Democratic Union and the Christian Social Union in Germany, and three pairs of Belgian parties) as one-and-a-half parties, that is, halfway between one and two parties. And I made a similar adjustment, but in the opposite direction, for factionalised and uncohesive parties in Colombia, India, Italy, Japan, and the United States—counting each of these as one-and-a-half parties instead of one party. (The main purpose of these adjustments was to measure the degree of multi-partism more accurately, but they have consequences for measuring the degree of disproportionality, too.) Because these operationalisations may be controversial, I also provide alternative measures of disproportionality in the table—based on the one-party assumption for tightly twinned parties (in accordance with the prevailing Australian academic custom of thinking in terms of a single `Liberal-National party' instead of two parties) and the same one-party assumption for factionalised parties. As the table shows, the alternative indices for these countries tend to be higher, but only very slightly.

Table 4.1 Average Electoral Disproportionality and Type of Electoral System
in 36 Democracies, 1945–1996

Disproportionality	Alternative Disproportionality	Electoral System
Netherlands	1.30	-	PR
Denmark	1.83	-	PR
Sweden	2.09	-	PR
Israel	2.27	1.75	PR/(PM)
Malta	2.36	-	PR-STV
Austria	2.47	-	PR
Germany	2.52	2.58	PR
Switzerland	2.53	-	PR
Finland	2.93	-	PR
Belgium	3.24	3.23	PR
Italy	3.25	3.49	PR
Luxembourg	3.26	-	PR
Ireland	3.45	-	PR-STV
Portugal	4.04	-	PR
Iceland	4.25	-	PR
Norway	4.93	-	PR
Japan	5.03	5.30	SNTV
Greece	8.08	-	PR
Spain	8.15	-	PR
Australia	9.26	9.57	AV
Papua New Guinea	10.06	-	Plur.
United Kingdom	10.33	-	Plur.
Colombia	10.62	3.38	PR/Pres
New Zealand	11.11	-	Plur
India	11.38	12.37	Plur
Canada	11.72	-	Plur
Botswana	11.74	-	Plur
Costa Rica	13.65	4.13	PR/Pres
Trinidad and Tobago	13.66	-	Plur
Venezuela	14.41	4.28	PR/Pres
United States	14.91	5.33	Plur/Pres
Bahamas	15.47	-	Plur
Barbados	15.75	-	Plur
Mauritius	16.43	-	Plur
Jamaica	17.75	-	Plur
France	21.08	11.84	Maj/Pres

Source: based on data in Arend Lijphart, Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries, New Haven, Conn., Yale University Press, 1999.

Key : PR Proportional Representation Plur Plurality

STV Single Transferable Vote Pres Presidential

SNTV Single Non-Transferable Vote Maj Majority

AV Alternative Vote PM Parliamentary Majority

My second adjustment, for presidential democracies, has more far-reaching consequences. Presidential elections in these countries are at least as important as legislative elections, and I therefore averaged the disproportionalities in the two types of elections (using the geometric mean). Because presidential elections entail the selection of a single winning candidate, they are inherently majoritarian and disproportional; the combined indices of disproportionality are therefore much higher than the indices for the legislative elections. The alternative disproportionality column in Table 4.1 ignores the presidential elections and gives the values for legislative disproportionality only (in Colombia, Costa Rica, Venezuela, the United States, and France). [4]

As Table 4.1 shows, the disproportionality in PR systems ranges from a low of 1.30 per cent in the Netherlands to a high of 8.15 per cent in Spain. The legislative disproportionalities in presidential Colombia, Costa Rica, and Venezuela (which use PR in their congressional elections) fall within this range, too. The mid-point of this range is 4.72 per cent. The average 4.15 per cent disproportionality in the STV elections of the Australian Senate is actually a bit lower than this mid-point, indicating a slightly more proportional performance than what PR systems generally achieve. A less favourable picture emerges when we insert the Australian Senate's 4.15 per cent figure in the table between Portugal and Iceland. The 4.15 per cent is higher than the percentage of disproportionality in 14 of the PR systems (16, if we include Colombia and Costa Rica) and lower than that of only four other PR systems (five, if we include Venezuela). The table also shows a strikingly clear line dividing the PR (and parliamentary) systems from the plurality (first-past-the-post) and majority systems. The Australian Senate's 4.15 per cent disproportionality is well above this line. Therefore, even though Australian Senate elections are only partly proportional (because of the STV system) and partly disproportional (because of the population disparities between the states), in practice the system works mainly like a PR system.

There are two other respects in which the Australian Senate elections since 1949 resemble PR elections in other countries. One is that PR is generally associated with a change from two-party to multi-party systems or an increase in multi-partism. Australian Senate elections show a very clear trend toward multi-partism. A good measure is the effective number of parties represented in the Senate; it counts the number of parties, but weights them by their sizes, which means that larger parties are counted much more than small parties. For instance, in a pure two-party system with two equally strong parties, the effective number of parties is exactly 2.0, and in a three-party system with three equally strong parties, the effective number is 3.0. If one of the three parties is considerably weaker than the other two, the effective number will be somewhere between 2.0 and 3.0, depending on the relative strength of the third party. [5] Counting the Liberals and Nationals as one party, the first three PR elections yielded virtually pure two-party systems (effective numbers of 1.98 and 1.99); since 1980, the system has been more like a two-and-a-half party system (with between 2.40 and 2.68 effective parties). When the effective number of parties is regressed on the election year, the correlation coefficient is a very strong 0.80 and the adjusted r-squared 0.63 (statistically significant at the 0.1 per cent level).

Second, the degree of proportionality in PR elections depends a great deal on the average magnitude of the election districts (that is, the average number of representatives elected in each district): proportionality tends to increase with increasing district magnitude. In the Australian Senate elections, the magnitudes have varied considerably as a result of the increases in the number of senators for each state, the addition of the ACT and the Northern Territory as new districts, and the fact that while most elections have been half-senate elections, there have been five elections of the entire Senate following double dissolutions. The usual negative relationship between magnitude and disproportionality turns up in the 20 Senate elections, too, although it is not exceptionally strong; the correlation coefficient is 0.32, which is statistically significant at the 10 per cent level. Average magnitudes have ranged from 4.25 to 10 senators elected per district, and for each additional senator elected, disproportionality decreases by 0.35 per cent.

As I mentioned earlier, in the discussion of Table 4.1, most of the disproportionality percentages are based exclusively on elections of the lower or only houses of parliaments; the one exception is that in presidential democracies, presidential elections were also taken into consideration. Elections of upper houses were never counted. When upper houses are relatively powerful, however, a very good case can be made for adding the disproportionality of the upper house election to the overall disproportionality figure for a political system. In the Australian case, this would mean a lower percentage than the 9.26 per cent (or 9.57 per cent) indicated in Table 4.1—which is already relatively low compared with the other plurality and majority election systems. This adjustment would not change Australia's classification from a majoritarian to a consensus democracy on the first dimension, but it would give Australia a somewhat lower overall majoritarian score on this dimension. More importantly, in substantive terms, the PR election of the Senate in Australia clearly brings an element of consensus into what remains, on the first dimension, a basically majoritarian democracy.

What would be needed to turn Australia into a consensus democracy? My answer to this hypothetical question is: the adoption of PR for the election of the House of Representatives. This is probably not just a necessary, but also a sufficient condition. The case of Malta shows that it is possible to have PR and still maintain a strict two-party system, but the Australian experience with PR in Senate elections suggests very strongly that Malta does not provide a good analogy. My prediction would be that PR for House elections in Australia would almost certainly lead to considerably lower electoral disproportionality (lower than that in Senate elections because of the larger size of the House and the election of the entire House at one time instead of having staggered elections), a multi-party system (even if the Liberal and National parties are counted as one party), multi-party coalition cabinets or minority cabinets, and a less dominant executive and more assertive legislature.

What would be the consequences of a switch to PR in terms of government performance and policy outputs? In the next section, I turn to the comparative evidence that we have on this question.

PR versus plurality: does it make a difference?

Does the difference between PR and majoritarian elections make a difference for the operation of democracy, especially for how well democracy works? The conventional wisdom is that there is a trade-off between the quality and the effectiveness of democratic government. On the one hand, the conventional wisdom concedes that PR may provide more accurate representation and, in particular, better minority representation and protection of minority interests, as well as broader participation in decision-making. On the other hand, the conventional wisdom maintains that the one-party majority governments typically produced by plurality elections are more decisive and hence more effective policy-makers. This view is reflected in the adage, recently restated by Samuel Beer, that `representative government must not only represent, it must also govern.' [6] This has a clear implication that representativeness comes at the expense of effective government.

This conventional wisdom has long been widely accepted without adequate empirical examination, perhaps because its logic appears to be so strong that no test was thought to be needed. For instance, A. Lawrence Lowell wrote in 1896 that it was a self-evident `axiom' that one-party majority cabinets, resulting from plurality elections, were needed for effective policy-making. [7] I shall first examine the second part of the conventional wisdom, which posits a link between PR and ineffective decision-making; in the next section, I shall discuss the first part, which concerns democratic quality.

The theoretical basis for Lowell's axiom is certainly not implausible: concentrating political power in the hands of a narrow majority can promote unified, decisive leadership, and hence coherent policies and fast decision-making. But there are several counter-arguments. Majoritarian governments may be able to make decisions faster than the coalitions produced by PR, but fast decisions are not necessarily wise decisions. In fact, the opposite may be more valid, as many political theorists, notably the venerable authors of the Federalist Papers, [8] have argued. The introduction of the so-called `poll tax', a new local government tax in Britain in the 1980s, is a clear example of a policy, now universally acknowledged to have been a disastrous policy, that was the product of fast decision-making; in all probability, the poll tax would never have been introduced had it been more carefully, and more slowly, debated.

Moreover, the supposedly coherent policies produced by majoritarian governments may be negated by the alternation of these governments; this alternation from Left to Right and vice versa may entail sharp changes in economic policy that are too frequent and too abrupt. S.E. Finer, in particular, has forcefully argued that successful macro-economic management requires not so much a strong hand as a steady one, and that PR and coalition governments are better able to provide steady, centrist policy-making. [9] Finally, policies supported by a broad consensus are more likely to be successfully carried out and to remain on course than policies imposed by a `decisive' government against the wishes of important sectors of society. These counter-arguments appear to be at least slightly stronger than the argument in favour of majoritarian government that is based narrowly on the speed and coherence of decision-making.

The empirical evidence is mixed, but gives a slight edge to PR and consensus government, too. For instance, Richard Rose and Francis G. Castles find no significant differences in economic growth, inflation, and unemployment between PR and non-PR systems among the industrialised democracies. [10] On the other hand, Peter Katzenstein and Ronald Rogowski have shown that small countries adopted PR and corporatist practices in order to compensate for the disadvantages of their small size in international trade; [11] that is, these consensus elements served as sources of strength instead of weakness.

The analyses by the above four scholars all had to do with aspects of macro-economic management; these are indeed excellent performance indicators because they involve crucial functions of government and because precise quantitative data are available. Because the theoretical arguments and the empirical evidence are mixed, but are slightly more favourable to PR, my working hypothesis will be that PR produces better results—but without the expectation that the differences will be very strong and significant. Another reason not to expect major differences is that economic success is not solely determined by government policy. As far as British macro-economic policy is concerned, for instance, Rose points out that

many influences upon the economy are outside the control of the government … Decisions taken independently of government by British investors, industrialists, consumers and workers can frustrate the intentions of the government of the day. In an open international economy, Britain is increasingly influenced too by decisions taken in Japan, Washington, New York, Brussels, or Frankfurt. [12]

Rose's point obviously should not be exaggerated: the fact that governments are not in full control does not mean that they have no control at all. When the economy performs well—when economic growth is high, and inflation, unemployment, and budget deficits are low—governments routinely claim credit for this happy state of affairs. And voters are known to reward government parties in good economic times and to punish them when the economy is in poor shape.

However, Rose's argument does highlight the need to take these other influences into account. To the extent that they are identifiable and measurable variables, they should be controlled for in the statistical analyses. One such potentially important explanatory variable is the level of economic development. Another is population size, if only because our democracies differ widely in this respect. It is not clear, however, whether size is a favourable or an unfavourable factor: large countries have greater power in international relations, which they can use to gain economic benefits, but greater international influence also means more responsibility and hence higher expenses, especially for military purposes.

There may also be fortuitous events that affect economic success, such as the good luck experienced by Britain and Norway when they discovered oil in the North Sea. The effects of such fortuitous events, as well as external influences that cannot clearly be identified and controlled for, can be minimised when economic performance is examined over a long period of time and for a large number of countries. These two desiderata frequently are in conflict with each other: extending the period of analysis often means that some countries have to be excluded. And they may both conflict with a third desideratum—that the most accurate and reliable data be used. Therefore, in the analysis below, I shall usually report the results for different periods, different sets of countries, and different types of data, in order to provide as complete and robust a test of the hypotheses as possible. Finally, I limit the potential disturbing impact of external forces on economic performance by excluding the five smallest democracies with populations of less than half a million—the Bahamas, Barbados, Iceland, Luxembourg and Malta—from the analysis because these small countries are obviously extremely vulnerable to international influences. [13]

Table 4.2 shows the results of the bivariate regression analyses of the effect of PR on five sets of macro-economic variables: economic growth (average annual growth rates), inflation (both in terms of the conventional consumer price index and the more comprehensive GDP deflator), unemployment (both the standardised annual averages and the unstandardised percentages—less reliable but available for more countries), strike activity (working days lost per thousand workers per year), and budget deficits (average annual deficits as a per cent of GDP). My independent variable is the degree of electoral proportionality (as in the first column of Table 4.1); because all of the economic variables are for the 1970s or later years, the proportionality variable that I used is also for the 1971–96 period (instead of the longer 1945–96 period shown in Table 4.1). [14]

The estimated regression coefficient is the increase or decrease in the dependent variable for each unit increase in the independent variable—in our case, each increase by one percentage point of proportionality. Because the range in the degrees of proportionality is about 20 percentage points, the distance between the `average' PR system and the `average' non-PR system is about 10 points. Therefore, in answer to the question `how much difference does PR make?', the reply can be—roughly—ten times the value of the estimated regression coefficient. For instance, based on the fourth row of Table 4.2, the effect of PR on inflation is approximately ten times the estimated regression coefficient (which is -0.412 per cent): about 4.1 per cent less annual inflation in PR than non-PR systems.

Because the table reports bivariate regression results, the standardised regression coefficient in the second column equals the correlation coefficient. The statistical significance of the correlations depends on the absolute t-value, shown in the third column, and the number of cases, shown in the fourth column. Whether or not the correlations are significant is indicated by asterisks; three levels of significance are reported, including the least demanding 10 per cent level. If the number of countries is 21 or lower, the countries are usually the OECD countries and the data are usually the most reliable OECD data; when the number is above 21, the developing countries are also included to the extent that the necessary data on them are available.

Table 4.2 Bivariate Regression Analyses of the Effect of Electoral Proportionality on 16 Macro-Economic Performance Variables

Estimated RegressionCoefficient	Standardised RegressionCoefficient	Absolute-value	Number of countries
Economic growth (1980–93)	-0.032	-0.104	0.562	31
Economic growth (1970–95)	-0.002	-0.019	0.076	18
Economic growth (1980–95)	0.003	0.025	0.110	21
GDP deflator (1980–93)	-0.412**	-0.350	2.010	31
GDP deflator (1970–95)	-0.093	-0.254	1.049	18
GDP deflator (1980–95)	-0.074	-0.108	0.473	21
Cons. price index (1970–95)	-0.104	-0.303	1.274	18
Cons. price index (1980–95)	-0.094	-0.139	0.610	21
Unempl., standard. (1971–95)	-0.142*	-0.396	1.432	13
Unempl., unstand. (1971–95)	-0.092	-0.199	0.812	18
Unempl., standard. (1980–95)	-0.172	-0.262	0.939	14
Unempl., unstand. (1980–95)	-0.120	0.170	0.750	21
Strike activity (1970–94)	-7.743	-0.253	1.045	18
Strike activity (1980–94)	-5.786	-0.119	0.575	25
Budget deficits (1970–95	0.035	0.066	0.247	16
Budget deficits (1980–95)	0.031	0.049	0.204	19

* Statistically significant at the 10 percent level (one-tailed test).

** Statistically significant at the 5 percent level (one-tailed test).

Source: based on data in Arend Lijphart, Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries, New Haven, Conn., Yale University Press, 1999.

The bivariate relationships in Table 4.2 indicate mixed results. Electoral proportionality is associated with less economic growth (according to two of the three measures) and with higher budget deficits, but, on the positive side, with less inflation (as measured by the GDP deflator and the consumer price index), less unemployment (according to both standardised and unstandardised statistics), and fewer strikes (working days lost per thousand workers); all of these measures are annual averages. The picture turns uniformly favourable for PR, however, when the level of economic development and population size (logged) are controlled for. First, all three of the estimated regression coefficients for economic growth are now positive. Second, the negative bivariate relationship between proportionality and the GDP deflator in 1980–93 is strong and statistically significant at the 5 per cent level, and it remains significant when level of development and population are controlled for; moreover, the other four relationships remain negative and, in addition, become extremely strong and highly significant—at the 1 per cent level—when these controls are introduced. Third, with the same controls, all four relationships between proportionality and unemployment remain negative, and the first achieves statistical significance, albeit merely at the 10 per cent level. Fourth, the relationships with strike activity also remain negative with the same controls in place, and the relationship in the 1970–94 period becomes significant at the 10 per cent level. Fifth, the two positive—that is, unfavourable—relationships between proportionality and budget deficits become negative when the controls are introduced.

These findings warrant three conclusions. First, PR has a uniformly better macro-economic performance record than majoritarian systems, especially with regard to the control of inflation, but also, albeit more weakly, with regard to all of the other economic performance variables. Second, however, only a few of the correlations are statistically significant, and they clearly do not permit the definitive conclusion that PR systems are better policy-makers than majoritarian systems. Therefore, third, the most important conclusion is a negative one: majoritarian democracies are clearly not superior to PR systems as policy-makers—and the conventional wisdom is clearly wrong in claiming that this is the case.

PR and democratic quality

The conventional wisdom argues—erroneously, as I have just shown—that majoritarian systems are better at governing, but admits that PR is better at representing—in particular, representing minority groups and minority interests, representing everyone more accurately, and representing people and their interests more inclusively. This section will examine the relationship between PR and five frequently used sets of indicators of the quality of democracy and democratic representation: women's representation (in parliaments and in cabinets), income equality, voter turnout, satisfaction with democracy, and proximity between governments and citizens. Table 4.3 shows the bivariate relationships—all of which are statistically significant at the 5 or 1 per cent levels, and all of which show that PR works better than non-PR.

Women's political representation is an important measure of the quality of democratic representation in its own right, and it can also serve as an indirect proxy of how well minorities are represented generally. The fact that there are so many different kinds of ethnic and religious minorities in different countries makes comparisons extremely difficult, and it therefore makes sense to focus on the `minority' of women—a political rather than a numerical minority—that is found everywhere and that can be compared systematically across countries. The average percentage of women's parliamentary representation—which ranges widely, from a high of 30.4 per cent in Sweden over the period 1971–1995 to 0.9 per cent in Papua New Guinea—is about 5.8 percentage points higher (ten times the estimated regression coefficient, as explained earlier) in PR than in non-PR systems. Women tend to be better represented in developed than in developing countries, but when the level of development is controlled for, the relationship between proportionality and women's legislative representation weakens only slightly and is still significant at the 1 per cent level. The pattern is similar for the representation of women in cabinets—ranging from 42.1 per cent in Norway to 0 per cent in Papua New Guinea—although the correlation is significant only at the 5 per cent level.

Political equality is one of the basic goals of democracy, and the degree of political equality is therefore an important indicator of democratic quality. Political equality is difficult to measure directly, but economic equality can serve as a valid proxy, since political equality is more likely to prevail in the absence of great economic inequalities.

Table 4.3 Bivariate Regression Analyses of the Effect of Electoral Proportionality on 10 Indicators of the Quality of Democracy

EstimatedRegressionCoefficient	StandardisedRegressionCoefficient	Absolutet-value	Number ofcountries
Women's parl. repr. (1971–95)	0.577**	0.455	2.983	36
Women's cab. repr. (1993–95)	0.620*	0.348	2.161	36
Rich-poor ratio (1981–93)	-0.255*	-0.465	2.462	24
Decile ratio (c. 1986)	-0.080**	-0.578	2.747	17
Voter turnout (1971–96)	0.890**	0.400	2.543	36
Voter turnout (1960–78)	1.014**	0.534	2.960	24
Satisf. with dem. (1995–96)	1.936*	0.487	2.228	18
Differential satisf. (1990)	-1.090*	-0.634	2.46	11
Government distance (1978–85)	-0.055*	-0.580	2.251	12
Voter distance (1978–85)	-0.864*	-0.611	2.439	12

* Statistically significant at the 5 percent level (one-tailed test).

** Statistically significant at the 1 percent level (one-tailed test).

Source: based on data in Arend Lijphart, Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries, New Haven, Yale University Press, 1999.

The rich-poor ratio is the ratio of the income share of the highest 20 per cent to that of the lowest 20 per cent of households. The ratio varies between 16.4 in highly inegalitarian Botswana and 4.3 in egalitarian Japan. Electoral proportionality and inequality as measured by the rich-poor ratio are negatively and very strongly related (statistically significant at the 5 per cent level and almost at the 1 per cent level). The more developed countries have less inequality than the developing countries; when the level of development is controlled for, the correlation between PR and equality weakens only slightly and is still significant at the 5 per cent level.

The decile ratio is a similar ratio of income differences: the income ratio of the top to the bottom decile. It is available for most of the OECD countries, based on the most painstaking comparative study of income differences that has been done so far. [15] Finland has the lowest decile ratio, 2.59, and the United States the highest, 5.94. PR systems are again the more egalitarian; the correlation is significant at the 1 per cent level and is not affected when level of development is controlled for.

Voter turnout is an excellent indicator of democratic quality for two reasons. First, it shows the extent to which citizens are actually interested in being represented. Second, turnout is strongly correlated with socio-economic status and can therefore also serve as an indirect indicator of political equality: high turnout means more equal participation and hence greater political equality; low turnout spells unequal participation and hence more inequality. The basic measure is the number of voters as a percentage of voting-age population; Italy had the highest average turnout, 92.4 per cent, and Switzerland the lowest, 40.9 per cent, in the 1971–96 period. PR and voter turnout are positively and very strongly correlated (significant at the 1 per cent level). Turnout in PR systems is about 8.9 per cent higher than in non-PR systems. However, several controls need to be introduced. First of all, compulsory voting, which is somewhat more common in PR than in majoritarian countries, strongly stimulates turnout. Second, turnout is severely depressed by the high frequency and the multitude of electoral choices to be made both in the Swiss PR system and in the majoritarian United States. Third, turnout tends to be higher in more developed countries. When compulsory voting and the frequency of elections (both in the form of dummy variables) as well as the level of development are controlled for, the effect of PR on voter turnout weakens only slightly (and is still significant at the 5 per cent level). The regression analysis was repeated with the average turnout data collected by G. Bingham Powell for an earlier period (1960–78)—with virtually identical results. [16]

Does PR affect citizens' satisfaction with democracy? Hans-Dieter Klingemann reports the responses to the following survey question asked in a large number of countries, including eighteen of our democracies, in 1995 and 1996: `On the whole, are you very satisfied, fairly satisfied, not very satisfied, or not at all satisfied with the way democracy works in (your country)?' [17] The Danes and Norwegians expressed the highest percentage of satisfaction with democracy: 83 and 82 per cent, respectively, said that they were `very' or `fairly' satisfied. The Italians and Colombians were the least satisfied: only 19 and 16 per cent, respectively, expressed satisfaction. Citizens in PR systems are significantly more satisfied with democratic performance in their countries than citizens of majoritarian democracies; the difference is approximately 19.4 percentage points.

In an earlier study of eleven European democracies, Christopher J. Anderson and Christine A. Guillory found that, in each of these countries, respondents who had voted for the winning party or parties were more likely to be satisfied with how well democracy worked in their country than respondents who had voted for the losing party or parties. [18] Because it is easy to be satisfied when one is on the winning side, the degree to which winners and losers have similar responses can be regarded as a more sensitive measure of the breadth of satisfaction than simply the number of people who say they are very or fairly satisfied. The largest difference was in Greece (37.5 percentage points) and the lowest in Belgium (4.7 percentage points). As Table 4.3 shows, the difference in satisfaction is almost 11 percentage points smaller in PR than in non-PR systems. The correlation is significant at the five per cent level.

The final two variables can be used to test the following key claim that is often made on behalf of majoritarian systems: because in the typical two-party system the two major parties are both likely to be moderate, the government's policy position is likely to be close to that of the bulk of the voters. John D. Huber and G. Bingham Powell compared the government's position on a ten-point Left-Right scale with the voters' positions on the same scale in 12 Western democracies in the 1978–85 period. [19] One measure of the distance between government and voters is simply the distance between the government's position on the Left-Right scale and the position of the median voter; this measure is called `government distance' in Table 4.3. The other measure is the percentage of voters between the government and the median citizen, called `voter distance' in the table. The smaller these two distances are, the more representative the government is of the citizens' policy preferences. Government distance ranges from a high of 2.39 points on the ten-point scale in the United Kingdom to a low of 0.47 in Ireland; voter distance was found to be the greatest in Australia, 37 per cent, and the smallest in Ireland, eleven per cent. Contrary to the majoritarian claim, both distances are actually smaller in PR than in majoritarian systems. Both correlations are significant at the five per cent level.

The general conclusion is that PR has a much better record than majoritarian democracy on all of the measures of democratic quality, and that, as the previous section showed, majoritarian systems do not have a better record of governing. This means that there is no trade-off and no difficult choice to make in electoral engineering: PR systems clearly outperform non-PR systems.

---

Footnotes

* I am very grateful for the comments that I received from participants in the Representation and Institutional Change Conference and, in particular, for the many helpful suggestions by Murray Goot. [1] Arend Lijphart, Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries, New Haven, Yale University Press, 1999. These 36 countries are also the countries used in the statistical analyses in the second part of this chapter. They are: Australia, Austria, Bahamas, Barbados, Belgium, Botswana, Canada, Colombia, Costa Rica, Denmark, Finland, France, Germany, Greece, Iceland, India, Ireland, Israel, Italy, Jamaica, Japan, Luxembourg, Malta, Mauritius, Netherlands, New Zealand, Norway, Papua New Guinea, Portugal, Spain, Sweden, Switzerland, Trinidad, United Kingdom, United States, and Venezuela.

[2] Anthony Payne, `Westminster adapted: the political order of the Commonwealth Caribbean', in Jorge I. Domnguez, Robert A. Pastor and R. DeLisle Worrell, eds., Democracy in the Caribbean: Political, Economic, and Social Perspectives, Baltimore, Ma., Johns Hopkins University Press, 1993, pp. 57–73.

[3] Michael Gallagher, `Proportionality, disproportionality and electoral systems', Electoral Studies, vol. 10, no. 1, March 1991, pp. 33–51.

[4] In Israel, I also included the direct popular election of the prime minister in 1996—a highly disproportional election, too. Because it entailed only one election, however, the alternative Israeli percentage, based solely on legislative elections, is barely lower than the original percentage.

[5] For the exact formula, see Patterns of Democracy, op. cit., p. 68. Independents are counted as tiny one-member parties.

[6] Samuel Beer, `The roots of New Labour: liberalism rediscovered', The Economist, vol. 346, no. 8054, 7 February 1998, p. 25.

[7] A. Lawrence Lowell, Governments and Parties in Continental Europe, Boston, Houghton Mifflin, 1896, pp. 73–74.

[8] Alexander Hamilton, John Jay and James Madison, The Federalist, New York, McLean, 1788.

[9] S.E. Finer, Adversary Politics and Electoral Reform, London, Anthony Wigram, 1975.

[10] Richard Rose, What Are the Economic Consequences of PR?, London, Electoral Reform Society, 1992; Francis G. Castles, `The policy consequences of proportional representation: a sceptical commentary', Political Science, vol. 46, no. 2, December 1994, pp. 161–71.

[11] Peter J. Katzenstein, Small States in World Markets: Industrial Policy in Europe, Ithaca, N.Y., Cornell University Press, 1985; Ronald Rogowski, `Trade and the variety of democratic institutions', International Organization, vol. 41, no. 2, Spring 1987, pp. 203–23.

[12] Rose, op. cit., p. 11.

[13] The remaining 31 countries are listed in note 1 above.

[14] I also reversed the sign of the percentages in order to measure the degree of proportionality instead of the degree of disproportionality.

[15] Anthony B. Atkinson, Lee Rainwater and Timothy M. Smeeding, Income Distribution in OECD Countries: Evidence from the Luxembourg Income Study, Paris, Organisation for Economic Co-operation and Development, 1995.

[16] G. Bingham Powell, Jr., `Voting turnout in thirty democracies: partisan, legal, and socio-economic influences', in Richard Rose, ed., Electoral Participation: a Comparative Analysis, Beverly Hills, Calif., Sage, 1980, pp. 5–34.

[17] Hans-Dieter Klingemann, `Mapping political support in the 1990s: a global analysis', in Pippa Norris, ed., Critical Citizens: Global Support for Democratic Government, Oxford, Oxford University Press, 1999.

[18] Christopher J. Anderson and Christine A. Guillory, `Political institutions and satisfaction with democracy: a cross-national analysis of consensus and majoritarian systems', American Political Science Review, vol. 91, no. 1, March 1997, pp. 66–81.

[19] John D. Huber and G. Bingham Powell, Jr., `Congruence between citizens and policymakers in two visions of liberal democracy', World Politics, vol. 46, no. 3, April 1994, pp. 291–326.