Path: janda.org/c24 > Syllabus > Topics and Readings >
Pedersen

Electoral Volatility

Pedersen Reading

ELECTORAL VOLATILITY IN WESTERN EUROPE, 1948-1977

MOGENS N. PEDERSEN

Mogens N. Pedersen, excerpted from 'The Dynamics of European Party Systems: Changing Patterns of Electoral Volatility', European Journal of Political Research, 7/1 (1979), 1-26. Copyright 1979. Reprinted with permission of Kluwer Academic Publishers.

 
The Problem

During the 1960s it was a widely held view among political scientists that European party systems were inherently stable structures which--with a few exceptions--reflected the societal cleavage structures of the past.'[1] This view was even bolstered with strong empirical evidence pertaining to the party system at the level of the electorate. Thus Rose and Urwin were able to conclude that 'the electoral strength of most parties in Western nations since the war had changed very little from election to election, from decade to decade, or within the lifespan of a generation.'[2]

Recent European political history has produced some rather unexpected events which make it somewhat difficult to reconcile this theoretical view with the political realities. Thus the number of competing parties has increased considerably in some of the Northern European countries over the last few years, and the distribution of electoral strength in several countries has changed in unpredictable ways. The relationship between voters and parties has apparently undergone an alteration, in some countries detectable as a tendency towards decreasing party identification; in some as a tendency towards an increased frequency of unconventional political behaviour; and in a few cases as an outright defection of large portions of the electorate from older parties to new parties which are either classifiable as traditional mass parties, nor as 'catch-all' parties in the sense of the late 0. Kirchheimer.[3]

These are developments which are by now clearly visible. They seem to indicate that even if party systems may still reflect he traditional cleavage structure in the society, the significant exceptions that Rokkan and Lipset were talking about are no longer few, but constitute a larger and growing part of all European party systems.

We want to know if these phenomena of change which are visible primarily, but not solely, in the Northern European countries represent fundamental transformations of the party systems, or if they are better interpreted as mere fluctuations, temporary deviations from an otherwise stable pattern? Do they fit into a pattern, or are they just some atypical events with no connection with the past history of these societies or with the development going on in other countries in Europe? Are party systems in Europe converging, becoming more uniform, or are they diverging through the 1970s? Is it possible to speak about distinct periods of stability and change, and, if so, are these periods country-specific, or do they apply to larger parts of the European polities? Are fluctuations randomly distributed across time and countries, or do they make up trends of increasing and decreasing rates of change?

Phrased this way we are dealing with a very broad problem with a great many facets. When we want to map and understand the extent and the direction of party system change in Europe, the first problem we encounter is one of delimitation of the problem, the narrowing down of a broad question to a manageable problem. This delimitation goes as follows.

The concept of party system change is not a simple and straightforward one. We may learn that a party system is a 'system of interaction resulting from inter-party competition.[4] But such interaction and competition can be observed at different levels, or--to put it in other terms--this system can itself be interpreted as consisting of various subsystems. A concise mapping of party system change would have to cover the levels of parliament and government, the level of the party as an organization, and the level of the electorate. Party system change can be defined as the total set of changes in patterns of interaction and competition at these three levels as well as between them.

It is not possible to encompass all these dimensions at this stage. I have deliberately chosen to concentrate attention on the level of the electorate. This analysis is thus based upon the assumption that 'election results are important to politicians and to political scientists'. One may also say that even if elections are far from always being decisive events, they are still the best available vantage point for a study of change, because change will either be a result of elections, or elections will register any change which may occur in the party system.

At the electoral level, as well as on other levels, the party system may be described in terms of various theoretical concepts. Party systems differ with regard to polarization, fragmentation, institutionalization etc.

As long as a party system can be considered a system of 'parts', we may, however, from time to time find it useful to return to the simplest definition of party system format. The format of the electoral party system can be described in terms of the number of parties contesting the elections, and the distribution of electoral strength among these parties.

Given the central role of elections in the process of party system change, it is relevant--and we hope not over-ambitious--to examine the evolving patterns of format change in Europe, i.e. changes in the number of parties, and in the relative distribution of electoral strength among the competing parties. The phenomenon which we are singling out for analysis is electoral volatility, by which will be meant the net change within the electoral part y system resulting from individual vote transfers.[ 6 ]

Even a concept as simple as this is not without its problems. Traditionally political scientists have been very much preoccupied with the number of parties, because this number was considered important for the mechanics of the party system. Since the late ig6os it has become customary to describe the format in terms of number of parties as well as distribution of party strength. Many attempts have been made to provide single number quantitative indices of format which would make it possible to compare party systems and to describe intra-system changes. These measures like Rae's index of fractionalization[7] have mostly been locational or static measures for which reason alone they are not optimal for a study of the dynamic properties of party systems.[8]

If we want to concentrate attention on ongoing format change, we therefore have to devise measures of change that will discriminate among systems; which will reflect similarities and differences between diachronic patterns; and which are fairly easy to interpret in a theoretically meaningful way. These requirements can be met by using one or another kind of summary measure of rate of change in the party system. Several such measures have been proposed and applied recently.[9] In this paper we will use a measure of electoral volatility which is derived in the following manner.

Let pi,tstand for the percentage of the vote, which was obtained by party i at election t. Then the change in the strength of i since the previous election will be:

Ępi,t = pi,t - pi,t- 1

and if we do not consider sign differences we have the following relation for the party system:


where n stands for the total number of parties competing in the two elections.

Remembering that the net gains for winning parties numerically are equal to the net losses of the parties that were defeated in the election, one may also wish to use another indicator which is slightly easier to calculate and to interpret, namely:

Volatility (Vt) = 1/2 x TNCt

Vt is simply the cumulated gains for all winning parties in the party system, or--if the symmetrical interpretation is preferred--the numerical value of the cumulated losses for all losing parties. Its range of variation has a straightforward explanation, and it can be expressed in terms of percentage.

A description of the European systems in terms of volatility will at least give a partial answer to the broader questions about change and stability in European party systems. It may thus lead to an identification of what are the typical and what are the deviant patterns of development.

2. National Patterns of Volatility

The measure of volatility tells to what extent party strength is being reallocated from one election to the next between losing and winning parties. An examination of national patterns will thus indicate, if the relative positions of parties are fairly constant, or if they fluctuate in ways which may eventually reflect basic electoral realignments. The extent of volatility may differ across countries as well as over time; differences may be of a random character or they may be subject to to certain political regularities. A first task for any analysis is to look for such regularities.

If the measure of volatility is applied to data from the thirteen countries, a picture is obtained which in the light of earlier analyses contains expected as well as not so familiar features (see Fig. 13-1),

That France comes out as the country in which average electoral volatility is the highest will probably not surprise many; nor will the fact that the post-war politics of Austria and Switzerland have been characterized by negligible volatility. All those who are familiar with the long-term trend in the German electorate since 1949, will not be surprised to find that the average net gains/losses in the German system amount to approximately to per cent per election. On the other hand it was probably not expected by many that the Nordic countries differ widely in this respect, nor that Denmark scores second-highest among the thirteen nations, nor that the European party systems on the whole differ considerably with regard to volatility.

Such observations may conceal a lot. An average may not be typical for the diachronic pattern, but may for instance reflect the occurrence of one or a few highly atypical elections. In order to pass judgements on volatility, we will therefore as a natural next step examine the dispersion of observations around their respective national means. In Fig. 13.2 this has been done by means of a plot of the relationship between means and their corresponding standard deviations. This plot affirms that Austria, Switzerland, and also Sweden are nations in which the relative strength of parties has hardly changed from one election to the next. It further identifies a group of six other nations, United Kingdom, Finland, Belgium, Ireland, Italy, and the Netherlands, with a somewhat higher average volatility and a uniform pattern of dispersion.

One intriguing finding appears in the plot next. The two big nations, Germany and France, and the two smaller nations, Denmark and Norway, apparently come out as fairly alike in their statistical patterns: a relatively high average volatility is combined with a high degree of fluctuation around that mean.

The interpretation of these findings is far from simple. A great many diachronic patterns of volatility might produce such means and standard deviations. But Fig. 13.2 provides a clue for further examination, because it shows in clear statistical terms that at least two 'families' of party systems exist, viz. a group of party systems with a relatively stable distribution of party strength, and a group of systems in which volatility has been changing considerably over time.

In passing it is tempting to mention that these two clusters of party systems do not have any affinity with the way in which students of party systems and democratic political systems traditionally classify European countries. The two clusters transcend the classifications of Almond (immobilist versus working multiparty systems); of Duverger and Neumann (classification according to number of parties); of Lijphart (centrifugal, centripetal, consociational and depoliticized democracy); and of Sartori (moderate versus polarized plural-isru etc.). Such a discrepancy may be of no theoretical import-ance; it may, however, also imply that the usefulness of these classifications is restricted and even decreasing over time.

A step towards further clarification can be taken by introducing the time variable, i.e. by changing the perspective from an examination of the range of fluctuations to an examination of the diachronic patterns of volatility. In Table 13.1 the two groups of party systems have been sorted by means of a crude periodization.

TABLE 13-1 The Volatility of European Party Systems 1948-77:
Average Net Gains Compared Across Time and Countries
Country
Period
National average
No. of election periods

1948-59
1960-69
1970-77

Switzerland

1.9
3.7
6.4
4
6

Austria

4.1
3.9
3.1
3.7
8

United Kingdom

4.4
5.2
7.9
5.9
8

Finland

4.4
6.9
9.1
6.8
8

Sweden

4.8
4.3
6.6
5.2
9

Netherlands

6.3
7.9
12.7
9.1
8

Belgium

7.9
10.3
5.5
7.9
9

Italy

10.3
8
6.8
8.4
6

Ireland

10.9
6.8
5
7.9
8

Norway

3.4
5.2
17.1
8.1
7

Denmark

5.5
8.9
18.7
11
12

Germany

15.2
9.5
4.9
9.8
7

France

21.8
11.9
10.6
16.8
7

Period average

7.8
7.3
9.2
8.1

No. of election periods

36
34
33
103

At once a clear pattern of change and stability emerges from the data. The group of party systems with a high level of volatility and high standard deviations clearly falls into two distinct subgroups: the German and the French party system, in which volatility has decreased over time considerably, especially at the beginning of the period, and the Norwegian and the Danish party system, in which exactly the opposite trend is visible, i.e. where volatility has tended to increase, especially in the last part of the period. But it is also important to note that the party systems in the other group also tend to fall into two subgroups: in Austria, Belgium, Ireland, and Italy volatility has tended to diminish over time, while it has increased in Switzerland, the United Kingdom, Finland, Sweden, and the Netherlands.

This finding suggests a new way of classifying European party systems according to their trends of volatility: the data apparently falls into three, or maybe even four, fairly distinct groups or 'families' of party systems.

This idea can be pursued further by adding a new dimension to the search for order. If it holds true that the thirteen party systems differ with regard to their secular trends, then an obvious next step in the mapping operation consists in examining the data for trends and for statistical linearity. The question becomes then if volatility tends to increase or decrease in a linear way, or if the apparently orderly picture in Table 13.1 reflects more complicated developmental patterns, e.g. patterns in which decades or other temporal sequences play an ordering role. Table 13.2 gives the answer.

This set of regression equations and correlation coefficients tells a clear story. Party systems in Europe do indeed differ with regard to the level of volatility, and with regard to the direction of change in volatility. The overall tendency towards a relatively stable pattern which comes out if data from all nations are artificially aggregated, disappears and is replaced by an array of widely differing trend lines. The data suggests a classification with three classes, differentiated from each other by means of the rate of change, i.e. the slope of the regression line.

The two small Scandinavian countries, Norway and Denmark, evidently stand distinctively apart from most other European countries in terms of volatility patterns, and so do the two major nations, France and Germany. In both cases the regression analysis also supports the impression which could be had from a visual inspection of period averages (Table 13.1), namely that the trends for these four countries fit very well with a linear model.

3. A Classification of European Elections

From what was said in the previous section it follows that high-volatility elections as well as low volatility elections are not randomly scattered across time and nations. It may be useful to identify various types of elections across Europe. By means of such a mapping we get not only a picture of the occurrence of these events which are often dramatic, but also a more simple, and yet informative summary of the national patterns.

The 103 elections which were held between 1948 and 1977, are not normally distributed around the European mean of 8.1 per cent volatility. Fig. 13.3 shows that the distribution is highly skewed, with 64 elections below the mean and 39 elections above. A few elections stand out because of their extraordinary degree of volatility. This is true of the1973 election in Denmark, when the number of parties in the system doubled; of some of the French elections in the1950s; and of the German election in 1953 when the CDU won its biggest victory in the post-war period.

The very shape of the distribution leads us to draw some distinctions, based upon pure statistical reasoning. Thus the elections in what is approximately the lowest quartile stand somewhat apart from those in the two middle quartiles, and these again form a cluster distinct from that formed by the elections in the highest quartile.

The map of European elections that can be drawn by means of this tripartite classification is presented in Fig. 13.4. This figure recapitulates some of the findings which were reported in the previous section, but it also adds some new features. Thus we see, first, that some countries, viz. Switzerland, Sweden, and Austria, have never in the post-war period experienced high-volatility elections. Second, one notes that some countries on the contrary never experienced low-volatility elections. This holds true not only for France, but for Italy, Belgium, and the Netherlands as well. Third, we note that the European countries differ considerably in the sense that in some of them high-volatility elections are to a considerable extent a thing of the past--of the 1950s while such elections are very much a 1970s phenomenon in some other countries. Finally, a comparison of decades indicates a general European trend towards relatively more high-volatility elections and relatively fewer low volatility elections, with the ig6os being the decade in which these two types ofelections were relatively limited.

As late as 1970 it was possible and also valid to argue that 'the first priority of social scientists concerned with the development of parties and party systems since 1945 is to explain the absence of change in a far from static period in political history'.[10] In the light of the preceding pages we are entitled to, and even forced to, qualify this statement. At the end of the1970s the first priority is to understand why some party systems still appear to be stable while other systems either have been undergoing a transformation or have gone through a period of considerable instability. Even if electoral volatility is only one among several possible indicators of persistence/change, it is a sufficiently important indicator to warrant this statement.

 


1. S.M. Lipset and Stein Rokkan, 'Cleavage Structures, Party Systems and Voter Alignments: An Introduction', in Lipset and Rokkan (eds.), Party Systems and Voter Alignments (New York: Free Press, 196 7), 50-

2. Richard Rose and Derek Urwin, 'Persistence and Change in Western Party Systems since 1945', Political Studies, 18/3 (1970), 295-

3. Otto Kirchheimer, 'The Transformation of the Western European Party Systems', in Joseph LaPalombara and Myron Weiner (eds.), Political Parties and Political Development (Princeton: Princeton University Press, 1966), 17 7-200.ELECTORAL VOLATILITY IN WESTERN EUROPE

4. Giovanni Sartori , or Ana,5sisPart,es and Party Systems: A Framework f (Cambridge: Cambridge University Press, 1976), 1- 44.

5. Rose and Urwin, 'Persistence and Change in Western Party Systems since 1945', P. 288.

6 W. Ascher and S. Tarrow, 'The Stability of Communist Electorates: Evidence from a Longitudinal Analysis of French and Italian Aggregate Data', American Journal of Political Science, 19/3 (1975), 48o-i.

7. Douglas Rae, The Political Consequences ofelectoral Laws (New Haven: Yale University Press, 1967).

8. Mogcns N. Pedersen, 'La Misurazione del Mutamento nei Sistemi Partitici: Una critica', Rivista Italiana di Scienza Politica, 8/2 (1978), 243-61 [English version in Comparative Political Studies, 12/4 (1980), 387-4031

9. Rose and Urwin, 'Persistence and Change in Western Party Systems since 1945'; Adam Przeworski, 'Institutionalization of Voting Patterns, or Is Mobilization the Source of Decay?', American Political Science Review, 69/1 (1975), 49-67; Ascher and Tarrow, 'Stability of Communist Electorates'.

10. Rose and Urwin, 'Persistence and Change in Western Party Systems since 1945', P. 295; italics in the original.