University of Lund

I. Introduction

In the study of legislative behavior in the United States, the roll-call analysis has long been regarded as an obvious instrument. The roll calls clearly offer exceptionally good data to determine the degree of party cohesion and party cooperation. Roll-call analysis, of course, does not elucidate dissensions within the party which do not come out into the open. Nor do the roll calls reveal anything of internal agreements, compromises or pressure. Obviously, the roll calls cannot provide complete knowledge of the real cooperation between the parties. They convey no information about the phases of the political process preceding the final parliamentary decision, such as compromises and agreements in committees. It ought to be further observed that the fact that two parties vote alike does not necessarily mean that any cooperation, in a real sense, has occurred.

Open voting was introduced in the Swedish parliament in 1925, and since then roll calls exist indicating how every MP has voted. Actually, it is remarkable that this material, with a minor exception, has not been previously utilized.¹ The following investigation concerns the voting in the lower house in 1964 and 1966.²

The degree of party cohesion and party cooperation is a function of the structure of the party system and of the political system in general. In a politically important question, greater party cohesion is probably demanded than in a question of less importance. And the demands on party cohesion are likely to be intensified, when the distribution of seats among the parties is such that the outcome can hang on one or a few votes. Furthermore, it is probably true that a governing party requires a stricter discipline than opposition parties.

Sweden has long had proportional representation and hence a multiparty system. Five parties have been represented in the Swedish parliament since the 1920s — the Conservatives, the Liberals, the Center party, the Social Democrats and the Communists. The parliament is divided into an upper house and a lower house. From the beginning of 1971, however, the bicameral system will be succeeded by a unicameral legislature. The houses have equal competence and powers

* This article is a summary of a report published in Statsvetenskaplig tidskrift, 1968, pp.
335—405.

in all matters. If they disagree on legislative matters and their decisions cannot be reconciled, the question is dropped. If they differ on matters concerning state expenditures or taxation, each house has to vote on the different proposals. Then the votes of the two houses are added and the opinion that receives a majority of the aggregate number of votes constitutes the decision of parliament. Since the 1930s the Social Democrats have been the dominant party; they have had a majority in the upper house since 1942 and at times also in the lower house. In the elections the distribution of votes between the Social Democrats and Communists on the one hand and the three non-Socialist parties on the other hand has been very stable, oscillating since 1940 between 58/42 per cent (1940) and 49.6/50.3 per cent (1956). Among the non-Socialist parties there have been considerable fluctuations. The Social Democrats have been the governing party, sometimes in cooperation or coalition with the Center party. The non-Socialist parties, or at least the Conservatives and Liberals, have in each election hoped to overthrow the Socialist government but without success. In the 1960s this plea for a change was strong. Joint actions from the non-Socialist parties or from the Liberals and the Center party together sometimes occurred in both the parliamentary and the electoral arenas. Plans for a united non-Socialist party, or at least a merger of the Liberal party and Center party into a middle party, had their advocates.³

II. Major Problems

We are interested in answering the following questions concerning party cohesion
and party cooperation in the lower house of the Swedish parliament in 1964
and 1966:

1. Party Cohesion

Are there differences in cohesion between the parties?

Is party cohesion stronger on politically more important issues than on less
important ones?

On what types of issues is party cohesion weakest?

Are there subgroups within a party exhibiting a different degree of cohesion
than the complements of the subgroups, i.e. the remaining members of the
party?

Party Cohesion and Party Cooperation in the Swedish Parliament in 1964 and 1966

Is the voting behavior of parallel subgroups within two parties more similar
than that of the complements of the subgroups, i.e. the remaining members of
each of the two parties?

2. Party Cooperation

Which parties vote most or least frequently alike?

What is the nature of the cooperation between the non-Socialist parties?
Which parties' voting behavior is most or least similar?

III. Data

The basic data consist of all the main divisions in the lower house during 1964 and 1966, which were officially recorded. Votes are counted by an electrical voting machine, to which buttons are connected from each HP's desk. The number of divisions totalled 366 and 360 respectively; the frequency of divisions was thus practically the same both years. It ought to be stressed that the major portion of matters was decided unanimously and not by a division. Furthermore, it ought to be noted that both years were election years. The divisions have been classified according to whether the vote concerned a government bill and/or a private member's proposal, initiated by a party (party proposal). It has also been noted if the division concerned a private member's proposal, initiated by the Liberal party leader and the Center party leader together. Such proposals were systematically initiated in 1966.

The number of MPs in the lower house was 232 in 1964 and 233 in 1966. With the help of photographs from the electric score board, attached to the voting machine, it has been possible to determine how every MP voted in each division, i.e. if he or she voted yes, no, abstain or was absent. In addition, each member has been classified by party and a number of other attributes. It has been assumed that these data could contribute to explaining the occurrence of deviant voting, and consequently with their assistance it would be possible to establish the existence of groupings within and across party lines. The personal attributes discussed in this article are: age, number of years in parliament (tenure), sex, membership in the "Christian group" and in one of the two temperance groups in parliament (the Social Democrats have such a group, and the non-Socialists have one). The investigation has also dealt with a number of other attributes, e.g. the regional residence of the MPs. These attributes, however, are not discussed in this article due to the fact that they seem to have had only a minor effect on the voting behavior.

The coding scheme for the data presented here is:4

IV. Definitions

In this study, party cohesion is defined as cohesion in voting for a specific line. Party cooperation is said to exist when two or more parties adopt the same line, i.e. vote alike in a division. How is the party line defined? Here the party line denotes the way in which most party members have cast their vote. When the number of "yeas" and "nays" has been equal, the party line has been designated as the "yeas". When the absolute majority of the party members have abstained from voting, the party line however has been designated as "abstain".

The correctness of this definition can, of course, be questioned in individual cases. The majority of the party can deviate from the party's ideological and traditional line for tactical reasons. The party's proposal may have already been rejected during the vote on the counterproposal. But in spite of such considerations, the party line must be the position of the party majority in each specific division.

In previous studies of party cohesion, usually only votes against the party line have been considered as deviant votes. Abstentions and absences have been disregarded. This procedure can be questioned. An abstention may be caused by different circumstances. An MP may abstain because of pairing. Or an abstention may be due to an accident or a defect in the voting machine. In these cases an abstention cannot be viewed as deviant voting. However, these cases are probably relatively rare. In the overwhelming number of cases when an MP has abstained, he has done so for other reasons. He has been unsure of the nature of the issue. Or he has wanted to indicate that his position deviates from the party line.

In this connection we are interested in those who deviate and thus in those
who "abstain against the party line". Cases when "abstain" is the party line, by
definition, are consequently not included.

In order to be able to take various types of "abstentions" into account, the
votes "abstain" have been given the following alternative interpretations:

(1) As a form of absence, i.e. only the yeas and the nays have been considered.
(2) As a wavering vote, i.e. half the abstentions have been regarded as yeas and

the other half as nays.

(3) As a half protest against the party line, i.e. each abstention has been regarded
as a half vote against the party line.

An absence may be due to dissatisfaction with the party line or uncertainty.
But nearly always it is other circumstances which are the cause. Therefore, it has
been considered warranted not to include absences in the analysis.6

Party Cohesion and Party Cooperation in the Swedish Parliament in 1964 and 1966

V. Methods

With these working definitions we can analyze:

(1) The cohesion within a party (a party group). Wc shall investigate the frequency
of deviant voting.

(2) The joint voting of the parties. Two parties are said to have voted jointly
on a question, if the party majorities have voted alike.

Broad measurements of the degree of party cohesion can be obtained by specifying the average number of votes as well as abstentions against the party line per member or per division and by attempting to secure data on the dispersion. Similarly, the degree of party cooperation can be measured by specifying the frequency of joint voting among the parties. These measurements, however, are crude. They do not provide a reliable basis for comparisons between the parties. Difficulties arise when comparing subgroups within one party or subgroups with the same attributes in different parties. What is necessary is that a ratio of affirmative votes to negative votes for every group (subgroup) is calculated for each particular division.

Therefore in our anlysis we shall rely on two well-known methods of measurement, both of which were originally devised by Stuart A. Rice.⁷ Each method is based on a special index — the index of cohesion and the index of likeness.

1. Index of Cohesion

This index provides a measure of the internal cohesion of a group. In this context a group is defined as the aggregate of MPs who have the same attributes for one or more of the variables in the coding scheme. If the votes of a group are divided 50/50 in a division, disunity is total and the index of cohesion = 0. Conversely, if the distribution of votes is 100/0 or 0/100, unity is complete and the index of cohesion = 100. The index of cohesion is usually defined as the absolute difference between the percentage of affirmative votes and the percentage of negative votes. When the distribution of votes is e.g. 70/30 or 30/70, the index of cohesion is 40 in both cases.

Because of the three alternative interpretations given to abstentions, we have
the following three variants of the index of cohesion:

(1) Absolute difference between the percentage of affirmative votes and the
percentage of negative votes. Base: Yeas + nays.

(2) Absolute difference between the percentage of affirmative votes and the percentage of negative votes, when half of the abstentions have been added to the affirmative votes and half to the negative votes. Base: Yeas -f nays + abstentions.

(3) Absolute difference between the percentage of affirmative votes and the
percentage of negative votes, when half of the abstentions have been added

to the smaller sum of the affirmative or negative vones respectively. Base:
Yeas -f nays + 1/2 the abstentions.

The index of cohesion has not been calculated for divisions when two-thirds
or more of the group's members have abstained. Due to limitations of space, only
the first variant is discussed in this article.

2. Index of Likeness

This index provides a measure of the similarity (or dissimilarity) in the voting behavior of two groups. If the votes in one group are divided 50/50 and likewise 50/50 in the other group in a division, the voting behavior is entirely similar and the index of likeness = 100. This is also true if the distribution of votes in the two groups is 100/0 or any other identical distribution. Conversely, if the distribution of votes in one group is 100/0 and in the other group is 0/100, the voting behavior is totally dissimilar and the index of likeness = 0. The index of likeness is calculated by subtracting from 100 the absolute difference between the percentages of affirmative votes cast by the two groups. For instance, if the distribution of votes is 70/30 and 30/70 respectively, the index of likeness is 100 - (70-30) = 60.

Because of the three alternative interpretations of abstentions, we have also
three variants of the index of likeness, viz.:

(1) Subtraction from 100 of the absolute difference between the percentages of
the affirmative votes. Base for each group: Yeas + nays.

(2) Subtraction from 100 of the absolute difference between the percentages of the affirmative votes, when in each group half of the abstentions have been added to the affirmative votes and half to the negative votes. Base for each group: Yeas + nays + abstentions.

(3) Subtraction from 100 of the absolute difference between the percentages of the affirmative votes, when half of the abstentions have been added to the smaller sum of the affirmative or negative votes. Base for each group: Yeas + nays + abstentions.

The index of likeness has not been calculated here for divisions when twothirds
or more of the members in one group or in both groups have abstained.

In comparisons using the index of likeness, one must operate with different
numbers of abstentions in the groups. This complicates the analysis of data.
Therefore the second and third variants have not been employed in this study.

3. Sign Test

In previous research the index scores have usually been analyzed with the aid of
averages and in some cases frequency distributions. Such an anlysis, however,
is not adequate because index scores are comparable only division for division.

Party Cohesion and Party Cooperation in the Swedish Parliament in 1964 and 1966

In addition, the mean as well as frequency distributions are relatively crude and
superficial.

In analyzing differences in voting behavior between various groups, it is essential to know whether or not the differences are randomly dependent. If we want to investigate the difference in the degree of cohesion between Social Democrats who are members of the Christian group and the remaining Social Democrats, the index of cohesion series of the former group has to be tested against the corresponding index series of the latter group. This presupposes that each index score in the one series is compared with the index score for the same division in the other series, which requires that the index series be arranged by pairs. The only possible tests of significance are thus tests for matched pairs. Since it is crucial that divisions with extremely large index differences should not overly influence the results, the most suitable method would seem to be the so-called sign test.⁸ This means that the following procedure is applied to the indices of cohesion: We examine the divisions one by one, to see whether the index score for group A is larger than the index score for group B or whether the opposite is true. No other comparisons are made. If the index scores are the same or if one of the scores has not been calculated, because two-thirds or more of the group have abstained, the division is omitted. After comparing all the index scores, we obtain a sign test value according to the formula

where n_|_ = the number of times group A's index score is larger than group B's and n_ = the number of times group B's index score is larger than group A's. If there is no difference in the voting behavior of the groups, the number of pluses and the number of minuses will be approximately the same. The value of the sign test is low in this case. The larger the difference is between the number of pluses and minuses, the higher the value of the sign test. A high sign test value is an index that the voting behavior of the groups is dissimilar.

The following levels of significance apply to the sign test:

Dissimilarity (or more precisely: observed dissimilarity) between two groups is said to exist if the groups exhibit a difference at the 5 °/o level of significance. If a large number of comparisons are carried out between groups, which in reality are alike, it will be (incorrectly) asserted that the groups are dissimilar in one case out of 20. With a significance level of 1 %, it will be (incorrectly) asserted that the groups are dissimilar in one case out of 100. Thus an observed

difference which is significant at the l % level is a more reliable index that a real difference exists than if the observed difference is significant only at the 5 % level. A similar line of reasoning can be applied to the other levels of significance.

It ought to be borne in mind that with this test, we receive only a measurement
of the reliability of the observation that a difference exists, and not any
measurement of the size of the difference.

4. with the IndexCohesion

Objects Analyzed of
(1) All possible pairs of parties (10 pairs).

Question: Is the cohesion in, e.g., the Conservative party significantly greater
or less than in the Liberal party?

The method of analysis, as already indicated, is constructed so as to take into account that the groups are of varied size, and that the voting behavior is comparable only division for division. The cohesion of various parties is measured by testing the cohesion index scores with the sign test.

(2) A number of subgroups within a party are compared with the complements
of the subgroups, i.e. the remaining party members.

Question: Is the cohesion of a subgroup different from the cohesion of its
complement? Does the cohesion of, e.g., the Social Democratic temperance
group differ from the cohesion of the other Social Democrats?

5. Objects Analyzed with the Index of Likeness

(1) Similarities and dissimilarities in the voting behavior of the parties. To measure party x's voting behavior in comparison with the other parties, the index of likeness scores for each pair of parties, in which x is included, must be compared with the index scores for every other pair of parties, in which x is also included.

Question: Is the similarity in the voting behavior between, e.g., the Conservatives and the Liberals greater or less than the similarity in the voting behavior between the Conservatives and the Center party? A complete answer to the question with reference to all parties and party pairs may be obtained by analyzing the following five units of party combinations:

Within each unit, each pair's index of likeness scores have been tested against
every other pair's index scores with the sign test. For instance, for unit a:

Party Cohesion and Party Cooperation in the Swedish Parliament in 1964 and 1966

Index of likeness for Cons.—Lib. is tested against the index for Cons.—Cent.
Index of likeness for Cons.—Lib. is tested against the index for Cons.—Com.
Index of likeness for Cons.—Cent, is tested against the index for Cons.—Com.

The results of the first sign test indicate thus whether the similarity in the voting behavior of the Conservatives and the Social Democrats is greater or less than the similarity in the voting behavior of the Conservatives and the Liberals or if no difference exists.

(2) Similarities in the voting behavior of parallel subgroups in more than one party. The index of likeness scores of a subgroup within one party and of the corresponding subgroup within another party are compared with the index of likeness scores of the complements of the two subgroups. In this fashion, we eliminate the party factor. Hence the analysis concerns the possible influence of other attributes on voting behavior.

Question: Is the voting behavior of parallel subgroups in different parties more similar than the voting behavior of their complements? Is there greater similarity in the voting behavior of, e.g., Liberals and Social Democrats who belong to the Christian group of the Swedish parliament than in the voting behavior of Liberals and Social Democrats who do not belong to the group? To answer this question, it is necessary to calculate the index of likeness scores for the Liberals and Social Democrats in the Christian group and then to compare them with the scores of Liberals and Social Democrats who do not belong to the Christian group.

6. Paired Comparisons of Individual MP's Voting (Cluster-Bloc Analysis) The methods presented above provide information about the voting behavior of certain groups defined a priori (cf. the variables of the coding scheme). The question can be raised, however, if there are additional groups of MPs, either from one party or from more than one party, who vote alike more often than average in the one party or the parties. This question is particularly relevant for countries where party discipline is relatively weak, but it is also pertinent to Sweden. An analysis to identify additional groups is made possible if all frequencies of agreement in voting for each pair of MPs are computed. This method also gained repute through Stuart A. Rice and has been refined by Herman C. Beyle.⁹ The method is useful for an analysis of groupings within one party and of groupings across party lines. We have investigated the latter but have, in this article, refrained from presenting a detailed account (cf. the summary).

VI. Party Cohesion: Some Broad Measurements

The number of main divisions in the lower house, as previously mentioned, was
366 in 1964 and 360 in 1966. In a substantial number of these, the parties voted
unanimously. The frequency of unanimous votes is shown in Table 1.

The average number of votes against the party- line10 per MP was 12.6 in
1964 and 10.0 in 1966. The corresponding figures for each party are presented
in Table 2.

The average number of "abstentions" against the party line11 per MP was
10.8 in 1964 and 9.3 in 1966. The corresponding figures for each party are presented
in Table 3.

All the MPs voted at least once against the party line, with the exception of
two MPs from the Center party in 1964; however, they were absent for long
periods because of illness.

Party Cohesion and Party Cooperation in the Swedish Parliament in 1964 and 1966

In the Conservative party the highest number of deviant votes cast by a single MP was 43 in 1964 and 27 in 1966, in the Liberal party 31 in 1964 and 33 in 1966, in the Center party 25 in 1964 and 39 in 1966, among the Social Democrats 38 in 1964 and 22 in 1966, and among the Communists 8 in 1964 and 23 n 1966.

The dispersion of deviant votes is partially illustrated in Table 4, which indicates
the number of divisions per party when more than 5, 9, 15 votes and
abstentions respectively have been cast against the party line.

These figures related to party cohesion, as already stressed, must be interpreted with caution. They do not provide any solid knowledge of the parties' cohesion. We do not know if the differences between the parties and the years are significant. The following hypotheses, however, can be made:

(1) Cohesion was strongest among the Communists and subsequently among the
Social Democrats (cf. Tables 2 and 3).

(2) Cohesion, with regard to voting against the party line, was stronger in
1966 than 1964. The Communists, however, represent an exception; in their
case the situation was the opposite.

VII. Party Cohesion: Comparisons between the Different Parties

First, to be able to test the above hypothesis concerning the parties' relative degree of cohesion, the index of cohesion scores have been computed for each party, and each pair of index scores (a total of 10) have been tested with the sign test.

Results of the calculations are presented in Table 5.

The table indicates the following:

(1) The cohesion of the Communist party was significantly stronger than that
of all other parties both in 1964 and in 1966.

(2) In 1964 the Social Democrats, as compared with the non-Socialist parties, had in all cases a greater frequency of higher indices of cohesion. This difference was significant at the 0.05 level vis-a-vis the Conservatives and the Liberals. Vis-a-vis the Center party, however, the difference was not significant. The cohesion of the non-Socialist parties was not significantly different in 1964. Of interest, thus, is the intermediate position of the Center party: neither the difference in relation to the Liberals nor the difference in relation to the Social Democrats was significant, whereas the Conservatives and the Liberals were significantly more disunited than the Social Democrats.

(3) In 1966 the Social Democrats had a significantly stronger cohesion than all the non-Socialist parties (in general at the 0.0001 level). The Conservatives had a stronger cohesion than the Liberals and the Center party. The difference, however, was significant only in relation to the Center party. As in 1964, a significant difference between the Liberals and the Center party did not exist in 1966. It ought to be underlined that the intermediate position of the Center party in 1964 disappeared in 1966.

The hypothesis that cohesion was strongest among the Communists and subsequently among the Social Democrats has thus been verified. Particularly, in 1966 the Social Democrats were significantly more united than the non-Socialist parties. Is it the demand by a governing party for stricter party discipline that basically explains the difference? It is also conceivable that the mounting demands for increased non-Socialist cooperation have created new dissensions in one or more of the non-Socialist parties. What is most striking with regard to the relations between the non-Socialist parties is the stronger cohesion of the Conservatives in 1966. It may be mentioned that of the 8 Conservative MPs who topped the group's list of those voting against the party line, 5 had left parliament, and that of the 8 who topped a similar list of those abstaining against the party line, 7 had left parliament.

It ought to be observed that the method upon which Table 5 is based cannot
be used to test the second hypothesis — viz. the parties' cohesion, excluding the
Communists', was stronger in 1966 than in 1964.

VIII. Party Cohesion on Politically Important Issues

In the foregoing section, all the divisions and hence all issue;; have been regarded as equally important. Such a view, however, is a distortion of reality. Different questions are of varying political significance. A politically important question requires greater party cohesion than does a question of less political importance. How has this requirement been met during 1964 and 1966?

To classify divisions and issues according to their political importance for
the respective parties is virtually an impossible task. The same issue may be of

varying importance to different parties. At any rate, such a solution cannot result
in data which are suitable as the basis for a quantitative analysis.13 Instead
our solution is based on the assumption that the party proposals constitute politically
important questions. In the divisions on party proposals, cohesion in the
parties thus ought to be higher than average with regard to the party's own proposals
and the other parties' proposals. It is relatively difficult to distinguish the
party proposals of the Communists, and therefore they are not dealt with in the
analysis. The Social Democrats, the governing party, did not initiate any party
proposals as such. Instead the government bills have been included. It should be
noted that the non-controversial bills are automatically excluded, because this
analysis deals only with divisions. What is examined is the parties' voting on
government bills and party proposals as they have been reported out of committee
by the committee members of the respective parties. The government bills
have been divided into two categories. The first consists of cases when the
counterproposal is a party proposal; the second consists of cases when the counterproposal
is of a different nature. The first category ought to contain more controversial
issues than the second.
Of special interest are the proposals introduced jointly by the leaders of the
Liberal and Center parties in 1966.
The number of votes as well as abstentions against the party line during the
divisions on party proposals and government bills are presented in Tables 6—9.

The trend is clear. Party cohesion is intensified to the point of virtual unanimity during divisions on the party's own party proposals. Cohesion among the Social Democrats is similarly intensified during divisions on non-Socialist party proposals irrespective of whether or not the party proposals were paired with a government bill. We shall later return to the question of how the other two non-Socialist parties have voted during divisions on the party proposals of the third non-Socialist party. Here it suffices to point out that cohesion in this case is higher than average.

In the same fashion, cohesion is intensified during divisions on government
bills. For understandable reasons this is especially pronounced when the counterproposal
concerns a party proposal.

The proposals, initiated jointly by the leaders of the Liberal and Center parties in 1966 resulted in a total of 25 divisions. The support from the two parties was practically complete. During these divisions no Liberal MP voted against the party line, and only 7 abstentions occurred in the Liberal party. In the Center party, 4 votes were cast against the party line and 10 abstentions. Among the Social Democrats, only 16 votes against the party line and 9 abstentions occurred. Most striking is that the majority of the Conservatives actively supported the motions in only 8 cases. We shall return to this in connection with the discussion of party cooperation.

IX. Party Cohesion on Politically Less Important Issues

The analysis of the parties' internal cohesion during divisions on party proposals and government bills reveals that cohesion is high in politically important questions. Our hypothesis, however, has additional implications; it states that cohesion increases according to the political importance of the issues. It logically follows also that disunity is relatively greatest on issues of little or no importance. To define these issues by means of a method as mechanical as that used to distinguish politically important questions is not possible. However, it is possible to approach the problem by reversing the procedure and analyzing issues on which cohesion has actually been weak. We have examined all divisions where at least 25 %of the votes cast within a party were against the party line.¹⁴ The number of cases when this occurred totalled 66 in 1964 and 51 in 1966. The distribution per party and year is apparent from Table 10. It should be noted that the number of divisions is smaller than the number of cases, since more than one party could display the specified degree of disunity during the same division.

These divisions, with a few exceptions, concerned relatively insignificant
matters. Nor were the exceptions of any real major political importance.

X. Party Cohesion: Different Policy Areas

One question often raised is whether the degree of party cohesion varies from policy area to policy area. As examples of "party splitting" issues, cultural and moral questions are usually cited, although a reduction of differences in these areas may have transpired during the 19605. It has also been common to point to issues which are basically of local or regional interest; according to this view the parties, to a greater degree, can allow or simply cannot prevent splits in these questions. However, these issues are frequently of minor political importance, and they are poor examples to the extent that they fall within several different areas of policy.¹⁵

One feasible approach to answering the question is to attempt to classify the roll calls according to policy areas, e.g. agricultural issues, defense issues, budgetary issues, cultural issues, moral issues, and to attempt to categorize the diviions according to whether they affect national, regional or local interests. This sort of classification, however, presents serious difficulties for the investigator, because the classifications cannot be made mechanically. Frequently an issue encompasses more than one category. Similar difficulties are encountered in classifying issues as national, regional or local. Therefore we have refrained from classifying all the divisions. Instead we have once again only analyzed instances when a party group was so divided that at least 25 °/o of the votes cast within

the party were against the party line. The number of cases was 117, and the
number of divisions 89. What areas of policy did they deal with?

Most conspicuous is that 22 of the divisions (25 cases) concern education and youth, and 7 divisions (11 cases) traffic questions. Among the cultural questions, 3 (7) concerned the theater, while 4 (6) of the welfare questions concerned temperance measures, and of the questions of appointments 4 (7) dealt with creating new professorships in medicine. Only a limited number of the divisions (about 10) concerned questions of purely local interest, Probably the explanation is that questions of purely local interest can only on rare occasions generate a party split of the dimensions we are now discussing.

XI. Party Cohesion of Different Groups within the Parties

In attempting to analyze variations in the degree of party cohesion, we have so far centered attention on the characteristics of the issues voted upon. The problem can, however, be attacked also by focusing on the MPs and investigating if certain of their attributes can contribute to shedding light on party cohesion (cf. code scheme). Our hypothesis has been that persons with certain attributes have formed subgroups, characterized by internal cohesion, within their party and perhaps also between parties.

Thus what we are concerned with here is examining the cohesion of subgroups defined on the basis of the attributes specified, above. For practical reasons, the analysis has been limited to subgroups with at least 5 members. This has resulted inter alia in the Communists being automatically excluded.

As an initial step, the average number of votes against the party line and the average number of "abstentions" against the party line cast within the subgroups have been calculated for each party. Due to limitations of space, these figures have been omitted in this article.

To obtain more reliable measurements, we have tested the cohesion index series of the individual subgroups in a party (e.g. Social Democratic women MPs) against the cohesion index series of the subgroup's complement, i.e. the remaining party MPs (in this instance the Social Democratic male MPs). In evaluating the findings, it should be noted that the sign test is so constructed that cases when the index scores are alike are excluded. This means that all divisions when the party was united have been omitted. The analysis thus concerns divisions when some disunity has occurred in a party.

1. Age

Table 11 shows the findings for age group 1 (49 and under) and age group 3
(60 and over).

In 1964 the youngest age group (49 and under) displayed a significantly stronger cohesion in the non-Socialist parties. In 1966 a significantly stronger cohesion among the younger Liberals and Social Democrats can be observed. For the Conservatives, however, the situation was the reverse. The younger Conservatives now exhibited a weaker cohesion than the rest of the party.

In general, the oldest age group (60 and over) does not form as pronounced
a subgroup as the youngest age group. Significant differences exist only on occasion
and can (as in the case of the Liberals) be indicative of opposite trends.

2. Tenure

Cohesion based on the MP's number of years in parliament is shown in Table 12,
which presents the findings for group 1 (MPs who are serving their first term of
office) and for group 3 (MPs who are serving their fourth term or more).

In 1964, MPs serving their first term had a significantly stronger cohesion
in all parties. In 1966 the difference is significant for the Social Democrats and

the Center party. In the Conservative party, however, as in the case of age,
the situation has changed: MPs serving their first term reveal a tendency toward
weaker cohesion.

MPs serving four terms or more exhibit a weaker cohesion in general. The difference is significant for the Conservatives and Liberals in 1964 and for the Social Democrats in 1966. The Conservatives, however, provide a contrasting picture for 1966. The Conservative MPs with the longest tenure displayed a significantly stronger cohesion.

3. Women MPs

The internal cohesion of women MPs per party is shown in Table 13. Only the
Conservatives and the Social Democrats have sufficiently large subgroups (at
least 5 MPs).

The women MPs consistently exhibited a greater frequency of high cohesion
index scores. The differences are significant, except for the Social Democratic
women MPs in 1964.

4. Members of the Christian Group

The internal cohesion of members of the Christian group is presented by parties
in Table 14.

The Social Democratic members of the Christian group exhibited a significantly stronger cohesion than the remaining Social Democrats in 1964 as well as 1966. The Liberals in the Christian group also exhibited consistently a greater frequency of high cohesion index scores. As for the Conservative and Center party members of the group, the situation is the reverse in 1966.

5. Members of Temperance Groups

The internal cohesion of members of the two temperance groups is shown by
parties in Table 15. It should be noted that the number of Conservative members
was not large enough to constitute a subgroup.

Members of the Social Democratic temperance group have a lower cohesion than the remaining Social Democrats. The difference, however, is not significant in 1964. As for the Liberal party, whose members in the temperance group constituted a majority, there was no uniform trend and no significant differences. The Center party members of the temperance group exhibited a consistent trend of greater unity than the rest of the party group. The difference is significant in 1966.

By combining the attributes, listed in the coding scheme, subgroups within a subgroup can be formed, e.g. Social Democrats who are members of both the Christian group and the temperance group. In many instances, however, there are strong correlations between the various attributes — and primarily between attributes of particular interest — e.g. between age and tenure, or between membership in the Christian group and one of the temperance groups. Of the 21 Liberal members of the Christian group in 1966, no less than 20 also belonged to the non-Socialist temperance group. To examine groups where an extensive overlapping of attributes exists is hardly meaningful, without investigating all combinations, which is out of the question (see below). On the other hand, attributes without such overlapping and which are interesting to investigate result generally in groups with less than 5 members and therefore should not be analyzed. Thus we have not analyzed subgroups with more than one attribute.

XII. Similarities in Voting Behavior of Subgroups within More Than one Party

In order to get additional information about party cohesion, it is of importance to investigate if there are trends of similar voting behavior between parallel subgroups in the various parties, e.g. subgroups formed by members of the Christian group. With the aid of the index of likeness, we can examine for each pair of parties if the subgroups differ from their complements, i.e. the remaining party members.

1. Age

This examination focuses on the youngest group, i.e. 49 and under. The similarity
in voting behavior among members of this group, as compared with the complementary
group, is shown for each pair of parties in Taole 16.

The Social Democratic and non-Socialist MPs who belong to the youngest group exhibit both in 1964 and 1966, in every comparison of party pairs (i.e. Soc.—Lib., Soc.—Cons., and Soc.—Cent.) a tendency toward less similar voting behavior than the Social Democrats and non-Socialists belonging to the older age groups. The differences are significant only in the cases of the Soc.—Lib. and Soc.—Cons, for 1966. As far as the Lib.—Cent, are concerned, the youngest MPs have a significantly more similar voting behavior than the remaining members during both years. The situation is the opposite for the Cons.—Cent.; however, the difference is not significant for 1966.

2. Tenure

This examination focuses on the group of MPs serving their first term in parliament. The similarity in voting behavior among members of this group, as compared with the complementary group, is shown for each pair of parties in Table 17.

The trends apparent in Table 16 recur more pronouncedly here. The less
similar voting behavior between the Social Democrats and non-Socialists serving
their first term of office is consistently significant.

The non-Socialists serving their first term have a significantly more similar voting behavior in 1964 than the remaining members. In 1966, this is true only of the Lib.—Cent. Also of interest is the difference between the age groups and tenure groups for the Cons.—Cent, in 1964. In the case of the youngest age group, the voting behavior is significantly less similar, and in the case of shortest tenure significantly more similar.

3. Women MPs

Only the Conservatives and Social Democrats have sufficiently large groups. A comparison of the series of indices of likeness does not reveal any significantly different voting behavior between the female subgroups of the two parties and their male complementary groups.

4. Members of the Christian Group

The similarity in voting behavior within this group, as compared with the complementary
group, is shown for each pair of parties in Table 18.

Both in 1964 and 1966 the Social Democratic and non-Socialist members of the Christian group (Soc.—Lib., Soc.—Cons., Soc.—Cent.) have in each case a tendency to less similar voting behavior than their fellow party members. The difference is significant for the Soc.—Cent, during both years and for the Soc.— Cons, in 1966. The non-Socialist members of the Christian group in 1966 have a significantly more similar voting behavior than the remaining non-Socialists.

5. Members of the Temperance Groups

Group members in the Soc.—Lib. and Soc.—Cent, do not display any significantly
different voting behavior as compared with the complementary groups

in the Soc.—Lib. and Soc.—Cent, respectively. On the other hand, the similarity in the voting behavior of group members in the Lib.—Cent, is significantly greater than the similarity in the voting behavior of the complementary groups. It should be noted that the Conservatives do not have a sufficiently large (5 members) subgroup.

XIII. Multivariate Analysis?

In the preceding sections we have analyzed various subgroups in the parties and across party lines. We have ascertained differences in the cohesion of the subgroups as compared with their complementary groups,, and we have established similarities and differences in the voting behavior of parallel subgroups in the various parties. The subgroups have been defined a priori on the basis of certain attributes. The result is thus certain correlations (positive or negative) between one particular attribute or several attributes and a certain voting behavior. The question arises, then, if the specified attributes explain the voting behavior brought out in the analysis. In many instances this would seem to be the case. Tenure, e.g., appears to be a significant variable.

Reality, however, is far more complicated. As already pointed out, the attributes are not independent of one another. There is thus a close correlation between age and tenure. Through a more or less skewed distribution in other variables, one variable can determine the findings. To overcome this difficulty, a complete analysis of all combinations of variables contained in the data would be required. The number of combinations is exceedingly large. In addition, our a priori defined variables are not the only ones which can be of importance in this connection.

The enormous amount of work which a complete run of all combinations of variables would require, along with the substantial unreliability which would still remain due to the new variables, has led us to stop at this point. The findings of the analysis can at any rate provide indications, at times relatively clear ones, of the nature of the cause, but a definite warning must be made against too hasty conclusions.

XIV. Joint Voting of the Parties: Some Broad Measurements

It should be recalled that by definition two parties are said to vote jointly if the
party majorities vote alike in a division, i.e. "yes", "no" or "abstain".

We shall begin by presenting the number of times two or more parties have
voted jointly.16 All combinations consisting of two to five parties, which occurred
during the divisions are shown in Table 19.

It should be noted that the two-party combinations include all cases when the respective parties voted alike, and not merely the cases when the parties voted alike in opposition to the remaining parties. This is also true of the three-party and four-party combinations.

Table 20 shows how many times a party has voted alone against the remaining
parties; Table 21 how many times two parties have voted against the

three remaining parties; and Table 22 how many times three parties have voted against the two other parties. A discrepancy in the figures in Tables 21 and 22 arises because the third variable "an absolute majority abstain" has operated in different ways.

The figures provide occasion for comparisons, reflections and appraisals. Undeniably the combinations of the Soc. + Com. and Cons. + Lib. + Cent, respectively, as seen in Tables 21 and 22, are beyond comparison the most frequent. On the other hand, it ought to be underlined that the instances when a strict cleavage between the Socialist parties and the non-Socialist parties has occurred are a minority of all the divisions. It is possible to speak of a non-Socialist bloc and a Socialist bloc, but this depends entirely upon what degree of cohesion is used as the criterion for being able to speak of blocs.

A complementary picture of "blocs" may be obtained if we exclude the Communists. We then obtain the following data for 1964. The Social Democrats voted alone in 22 + 123 divisions, i.e. 40% of the total number. The non stood united in 22 + 118 divisions, i.e. 38 %. In 1966 the corresponding figures were 11 + 122, i.e. 37 % and 11 + 111, i.e. 34 °/o. During the remaining divisions, one or two or, in 18 cases in 1964 and 13 cases in 1966, all three non parties voted in the same fashion as the Social Democrats. The frequencies of "the absolute majority abstains", however, ought to be noted.

The Communists voted in 1964 and 1966 during 78% and 86% of the divisions in the same fashion as the Social Democrats. In 22 and 11 cases respectively, they voted together with the non-Socialist parties against the Social Democrats, in a few cases they voted with only the Liberals or only the Center party, in one case (1966) with only the Conservatives (both parties abstained), and in 17 and 20 cases respectively they formed a common front with the two middle parties.

What was the nature of non-Socialist cooperation during these two years?

The question has already been partially answered by the comments on the cases when all three non-Socialist parties voted alike. What is of particular interest here are the cases — and they are the major portion — when they did not vote alike. To start with the Conservatives, this party to a greater degree than the Liberals and the Center party voted alone during the divisions: 61 times against 20 for the Liberals and 15 for the Center party in 1964, and 70 times against 10 for the Liberals and 13 for the Center party in 1966. Furthermore, the Conservatives, relatively more often than either one of the middle parties, were the only non-Socialist party which voted in the same way as the Social Democrats. The combination Cent. + Soc. + Com. however was more frequent than Cons.

Soc. + Com. in 1964. But this was not the case in 1966.

Of a total of 44 and 71 cases respectively when the Cons. + Soc. or Cons. + Soc. + Com. voted against the Lib. + Cent., 8 and 18 cases respectively concerned proposals for larger appropriations, rejected by the Conservatives and the Social Democrats.

The Liberals and Center party voted together (in toto) substantially more times than with the Conservatives: 265 times against 214 for the Cons, -f Lib. and 193 for Cons. + Cent, in 1964; and 289 times against 172 for the Cons. + Lib. and 156 for Cons. + Cent, in 1966. The changes from 1964 to 1966 are primarily explained by the fact that the Liberals and Center party voted alone during 29 divisions in 1964, while the corresponding figure for 1966 was 64. In connection with this change the figures for the Cons. + Lib. (alone) dropped from 33 to 22, the figures for the Cons. + Cent, rose from 5 to 9, while the figures for the Cons, alone rose from 61 to 70.

Of the non-Socialist parties, the Liberals exhibited the lowest frequency of total joint votes with the Social Democrats, and the Center party the highest. The difference between the Conservatives and the Liberals is, however, negligible.

The Center party was the non-Socialist party which voted most frequently
with the Communists.

Looking at the shifts in the non-Socialist cooperation from 1964 to 1966, it
is conspicuous that the Conservatives became more isolated, while the cooperation
of the middle parties increased.

The same objections as mentioned above (p. 139) can be leveled at these total figures for the joint voting of the parties, viz. that the divisions concerned issues of highly varying political importance. Of crucial interest is joint voting on politically important issus. Therefore in the next section we shall analyze joint voting on government bills and party proposals including proposals initiated by the leaders of the Liberals and the Center party.

In addition, the above measurements of joint voting are crude. Our procedure does not take into account the voting behavior of the entire party. Nor does it allow us to determine if the differences are significant or not. In section XVI we shall obtain a more reliable basis by measuring similarities and differences in the parties' voting behavior with the aid of the index of likeness.

XV. Joint Voting on Politically Important Issues

The Conservative party proposals led to 36 divisions both in 1964 and 1966.
The parties' positions on the proposals and the party combinations voting for
the proposals are presented in Table 23.

In 7 divisions the middle parties voted with the Conservatives, while the
party stood alone in 22 (24) and 21 (23) divisions respectively.

The Liberal party motions caused 45 divisions in 1964 and 15 in 1966. The
parties' positions and the party combinations voting for the proposals are presented
in Table 24.

It is difficult to compare the two years because of the large difference in thi number of divisions. But we can note that the Center party voted with the Lib erals in approximately one-half of the divisions in 1946 and two-thirds in 1966 while the corresponding figures for the Conservatives were one-third and one half.

The Center party proposals resulted in 19 divisions in 1964 and 21 in 1966
The parties' positions and the party combinations voting for the proposals ar<
presented in Table 25.

The Liberals, with minor exception, have voted with the Center party in all the divisions except those when the Center party stood alone. The instances of the Conservatives voting with the Center party have been largely confined to cases when the Liberals also voted for the proposals.

The proposals, initiated by the leaders of the Liberal and Center parties together, which occurred only during the 1966 session, caused 25 divisions. The parties' positions and the party combinations voting for the proposals are presented in Table 26.

The middle parties supported the motions in all the divisions. It is striking that the Conservatives voted with the middle parties in only 8 cases. The party abstained 6 times, and in the remaining cases adopted the same stand as the Social Democrats or the Social Democrats and the Communists together.

The government bills led to 111 divisions in 1964 and 143 in 1966. The
party combinations which occurred during the divisions are presented in Table 27.

Table 28 indicates how many times a party has voted alone against the others; Table 29 how many times two parties have voted against the remaining three; and Table 30 how many times three parties have voted against the remaining two. The figures in Tables 29 and 30 do not agree because the third variable "absolute majority abstains" has operated in different ways. The fourparty combinations are apparent from Table 27.

The tables ought to be compared with the corresponding tables (19—22) for
the total number of votes. To facilitate a comparision, the percentages for the
total number of roll calls are stated in parentheses below.

During the divisions on the government bills, the Social Democrats (here the Communists are excluded) voted alone in 9 + 38 cases, i.e. 42 °/o (40) of the divisions. The non-Socialists were united in 9 + 36 cases, i.e. 41 °/o (38). For 1966 the corresponding figures were 5 + 46, i.e. 36 °/o (37) and 5 + 43, i.e. 34 °/o (34).

The Communists voted in 1964 and 1966 during 82 °/o (78) and 88 °/o (86) of the divisions in the same fashion as the Social Democrats. During 8 °/o (6) and 3 °/o (4) respectively, they voted together with the three non-Socialist parties. During 6 °/o (5) and 5 °/o (6) they formed a common front with the two middle parties.

The Conservatives voted alone in 1964 during 16 °/o (16) of the divisions, against 10°/o (5) for the Liberals and 4 °/o (4) for the Center party; in 1966 during 23 % (19) against 5 °/o (3) for the Liberals and 5 °/o (4) for the Center party. The Conservatives, relatively more frequently than either of the middle parties, was the only non-Socialist party to vote in the same way as the Social Democrats. The combination Cent. + Soc. + Com., however, was somewhat more frequent both years than Cons. + Soc. + Com.

The combination Lib. + Cent, was (in toto) more frequent than the combinations Cons. + Lib. and Cons. + Cent.: in 1964 during 73 °/o (72) of the divisions, against 55 °/o (58) for the Cons. + Lib. and 54 °/o (53) for the Cons. + Cent.; in 1966 during 73 °/o (80) aginst 50 °/o (48) for the Cons, -f Lib. and 45 % (43) for the Cons. + Cent.

The extremely high degree of agreement between the percentages for all the
divisions and the percentages for the divisions on government bills is conspicuous.

XVI. Similarities and Differences in the Voting Behavior of the Parties

In section XII we compared the similarity in the voting behavior of subgroups in more than one party. The index of likeness scores of the subgroups were tested with the sign test against the corresponding scores of the complementary groups; the complementary groups consisted of the remaining members in the particular parties. The same method is now used to measure similarities and differences in the voting behavior of the parties. The index of likeness scores of, e.g., the Cons. —Lib. are computed for each division and are tested agirist the corresponding scores of, e.g., the Cons.—Cent. The sign test value indicates whether the similarity

in the voting behavior of the Conservatives and Liberals is greater or less than the similarity in the voting behavior of the Conservatives and the Center party, or if there is no difference. In this way, we shall test, party by party, all combinations of party pairs in which the one party is included.

1. The Conservatives

It is to be expected that the voting behavior among the non-Socialist parties is
significantly more similar than the voting behavior of a non-Socialist party in
relation to the Social Democrats. What we are primarily interested in is the re

lationship between the non-Socialist parties. Here it can be noted, then, that the Conservatives are significantly more similar to the Liberals than to the Center party in their voting behavior in 1964 but that this difference is not significant in 1966.

2. The Liberals

The similarity in the voting behavior of the Liberals and Conservatives in 1964 is less than, but not significantly different from the similarity in the voting behavior of the Liberals and the Center party. In 1966, however, the latter two parties have a significantly more similar voting behavior than the Liberals and Conservatives.

3. The Center Party

The voting behavior of the Center party for both years is significantly more similar to that of the Liberals than that of the Conservatives. It can be further noted that the Center party in 1964 — as distinct from the Conservatives and Liberals — is significantly more similar to the Social Democrats than the Communists. In 1966 this difference is not significant for any non-Socialist party.

4. The Social Democrats

The Social Democrats and Conservatives exhibit during both years a significantly more similar voting behavior than the Social Democrats and Liberals. At the same time, however, the Social Democrats and the Center party have a significantly more similar voting behavior than the Social Democrats and Conservatives. Thus the Liberals' voting behavior differs most from the Social Democrats

5. The Communists

The Communists and Liberals as well as the Communists and the Center party exhibit during both years a significantly more similar voting behavior than the Communists and Conservatives; the Communists and the Center party a significantly more similar voting behavior than the Communists and Liberals.

XVII. Summary

1. Party Cohesion

The Social Democrats displayed a stronger cohesion than the non-Socialist parties;
the difference in comparison with the Center party, however, was not significant
in 1964. The cohesion of the non-Socialist parties was not significantly

different in 1964. In 1966, however, the Conservatives had a stronger cohesion
than the middle parties; these two parties did not differ from each other. The
Communists had a stronger cohesion than all the other parties.

Party cohesion seems to have been intensified from 1964 to 1966. The Communists
were, however, an exception; the trend here appears to have been the
opposite.

On politically important issues, cohesion was substantially greater. During divisions on party proposals and government bills the parties closed their ranks. During divisions on a party proposal, members of the party sponsoring the proposal have nearly unanimously supported it.

Divisions where sharp splits occurred in a party (21 25 °/o of the votes against the party line), concerned almost invariable issues of minor significance. Of the few exceptions, one or two had some political importance but did not constitute a major issue of political importance.

Issues causing sharp splits (^ 25 °/o of the votes against the party line) fall
within several policy areas. Among such areas with a relatively high frequency
are issues concerning culture, education, youth and traffic.

With regard to the cohesion of various subgroups, it can be pointed out: The youngest age group (49 years and under) in the different parties often exhibited stronger cohesion than the remaining party members, while, on the contrary, the oldest age group (60 years and over) did not form such cohesive subgroups in the respective parties. Similarly, the MPs with the shortest tenure (those serving their first term) as a rule exhibited a stronger cohesion. On the other hand, MPs with the longest tenure (those serving their fourth term or more) as a rule had a weaker cohesion than the rest of the party. The women MP's cohesion was generally stronger than that of the other members. As far as the Christian group is concerned, the Social Democratic members had a stronger cohesion than the remaining Social Democrats. The picture is more varied for the non-Socialist parties. The Liberal members seem to have been more united, the Conservative and Center party members less united. Members of the temperance groups exhibited only in some instances a stronger cohesion.

The analysis of the subgroups' voting behavior across party lines sheds additional light on the findings mentioned above. The youngest age groups and above all the groups with the shortest tenure, in a comparison between the Social Democrats and the non-Socialist parties, generally had a more dissimilar voting behavior than the older groups and groups with longer tenure. The opposite was true for the Liberals-Center party. The voting behavior of women subgroups did not differ from that of the complementary groups. The non-Socialist members of the Christian group had a more similar behavior than the other non As for the Social Democratic members of the Christian group their voting behavior tended to be more dissimilar to that of the non-Socialist members of the group than was the case for the complementary groups. We obtain a similar picture for the temperance groups. The similarity in the voting behavior of the subgroups in the Liberal and Center parties was greater than for the

complementary groups. However there was no significant difference for the
Soc.—Lib. and Soc.—Cent.

2. Party Cooperation

The most frequent two-party combination of joint voting was Soc. + Com.; the most frequent three-party combination was Cons. + Lib. + Cent. On the other hand, it ought to be stressed that the cases when a strict division between the Socialist and non-Socialist parties occurred constituted a minority of the total number of divisions. If we exclude the Communists, we find that the Social Democrats voted alone during 40% (1964) and 37 °/o (1966) respectively of the divisions. The Liberals and the Center party in toto voted together considerably more times than either did with the Conservatives: 265 times against 214 for the Cons. + Lib. and 193 for the Cons. + Cent, in 1964, and 289 times against 172 for the Cons. + Lib. and 156 for the Cons. + Cent, in 1966. Cooperation between the middle parties increased from 1964 to 1966, while the Conservatives increasingly voted alone. The Liberals were the non-Socialist party exhibiting the lowest frequency of total joint voting with the Social Democrats, the Center party the highest frequency. The difference between the Liberals and Conservatives, however, is negligible. The Conservatives, relatively more frequently than the middle parties, were the only non-Socialist party to vote in the same way as the Social Democrats.

With respect to non-Socialist cooperation during the divisions on non-Socialist party proposals, it can be said: The Conservative party proposals received the support of both the Liberals and the Center party in about one-fifth of the cases, and in about three-fifths of the cases the Conservatives voted alone; this is true for both years. The Liberal party proposals received the support of the Conservatives in about one-third (1964) and one-half (1966) respectively of the cases and the support of the Center party in about one-half and two-thirds respectively of the cases. The Center party proposals received the support of the Conservatives in about one-half and one-fifth of the cases respectively and the support of the Liberals in about three-fourths and one-half of the cases respectively. The proposals, initiated by the leaders of the middle parties (1966) received the support of the Conservatives during only 8 of the 25 divisions.

As far as the government bills are concerned, there is a very high degree of
agreement between the percentages of joint voting on the bills and the corresponding
figures for the total number of divisions.

The comparisons of the scores of indices of likeness reveal the following: The Social Democrats and Conservatives exhibited during both years a more similar voting behavior than the Social Democrats and Liberals. At the same time the Social Democrats and the Center party had a more similar voting behavior than the Social Democrats and Conservatives. Thus the voting behavior of the Liberals differed most from that of the Social Democrats. The Conservatives in 1964 but not in 1966 were significantly more similar to the Liberals than to the Center party in their voting behavior.

The analysis of the joint voting of individual MPs discloses that there are some minor groups, as a rule regional groups, in the parties with a high frequency of voting jointly with a corresponding group in another party or with that party in general.

---

NOTES

1 Cf. L. Skold. "Partisammanhållningen i riksdagen", Tiden, vol. XLII, 1950, pp. 278 ff.

2 A more systematic investigation of the divisions in both houses is being prepared by Bjurulf.

3 N. Stjernquist. "Sweden: Stability or Deadlock?", in R. A. Dahl (ed.). Political Oppositions in Western Democracies, New Haven & London, Yale University Press, 1966, pp. 116 ff.; I. Glans. "Sweden: The 1964 Riksdag Election", Scandinavian Political Studies, vol. I, 1966, pp. 225 ff.; B. Molin. "Sweden: Politics Since 1964", Scandinavian Political Studies, vol. 11, 1967, pp. 253 ff.

4 The program is written in ALGOL and can be adapted to data on voting from legislative bodies, irrespective of the number or differentiation of the members and the votes.

5 The mandate period to the lower house was four years.

6 The average number of "absences" per MP was 40.7 in 1964 and 36.8 in 1966.

7 Stuart A. Rice. Quantitative Methods in Politics, New York, Knopf, 1928, pp. 207 ff. Cf. L. F. Anderson, M. W. Watts & A. R. Wilcox. Legislative Roll-Call Analysis, Evanston, Northwestern University Press, 1966.

8 For comments and criticism we wish to acknowledge our debts to Jan Lanke, lecturer at the department of mathematical statistics, University of Lund.

9 H. C. Beyle. Identification and Analysis of Attribute-Cluster-Blocs, Chicago, University of Chicago Press, 1931. L. F. Anderson, M. W. Watts & A. R. Wilcox. Op.cit., pp. 59 ff.

10 It should be pointed out that an MP is said to abstain against the party line when he abstains and "abstain" is not the party line. Therefore we exclude cases when the absolute majority of the party abstained.

11 When the majority have abstained, only the minority of yeas and nays respectively have been counted as votes against the party line.

12 The question is not relevant for the Communists due to their small number.

13 On this problem see, e.g., W. H. Riker. "A Method for Determining the Significance of Roll Calls in Voting Bodies", in J. C. Wahlke &: H. Eulau (eds.). Legislative Behavior, Glencoe, 111., The Free Press, 1959, pp. 377 ff.; Mogens N. Pedersen. "Partiernes holdning ved vedtagelsen af regeringens lovforslag 1945_—66", Historie, vol. VII, no. 3, 1967, pp. 421 ff.

14 Here we have not taken the abstentions into account.

15 P.-E. Back. "Det svenska partivasendet", in Samk'dlle och riksdag, vol. 11, Stockholm, Almquist & Wiksell, 1966, p. 133. Mogens N. Pedersen. »Consensus and Conflict in the Danish Folketing 1964^—65", Scandinavian Political Studies, vol. 11, 1967, p. 148.

16 Mogens N. Pedersen in op.cit. (SPS 2, 1967, p. 152 ff.) has employed an index of distance which measures the distance between the parties by pairs. Cases of joint voting have been assigned the point value of 0, cases of voting in opposition 1 point, and abstentions 1/z point. The total point value has been divided by the number of divisions. The higher the index value is, the greater the distance. Our two-party combinations can serve the name purpose (cases when the absolute majority of the one party has abstained have been regarded as voting in opposition). They provide, however, an inverted picture, i.e. an index of nearness and not an index of distance. The nearness of the parties is however apparent, without undue complication, from the absolute figures in Table 19 because the number of divisions during both years is almost exactly equal.

NOTES

1 Cf. L. Skold. "Partisammanhållningen i riksdagen", Tiden, vol. XLII, 1950, pp. 278 ff.

2 A more systematic investigation of the divisions in both houses is being prepared by Bjurulf.

3 N. Stjernquist. "Sweden: Stability or Deadlock?", in R. A. Dahl (ed.). Political Oppositions in Western Democracies, New Haven & London, Yale University Press, 1966, pp. 116 ff.; I. Glans. "Sweden: The 1964 Riksdag Election", Scandinavian Political Studies, vol. I, 1966, pp. 225 ff.; B. Molin. "Sweden: Politics Since 1964", Scandinavian Political Studies, vol. 11, 1967, pp. 253 ff.

4 The program is written in ALGOL and can be adapted to data on voting from legislative bodies, irrespective of the number or differentiation of the members and the votes.

5 The mandate period to the lower house was four years.

6 The average number of "absences" per MP was 40.7 in 1964 and 36.8 in 1966.

7 Stuart A. Rice. Quantitative Methods in Politics, New York, Knopf, 1928, pp. 207 ff. Cf. L. F. Anderson, M. W. Watts & A. R. Wilcox. Legislative Roll-Call Analysis, Evanston, Northwestern University Press, 1966.

8 For comments and criticism we wish to acknowledge our debts to Jan Lanke, lecturer at the department of mathematical statistics, University of Lund.

9 H. C. Beyle. Identification and Analysis of Attribute-Cluster-Blocs, Chicago, University of Chicago Press, 1931. L. F. Anderson, M. W. Watts & A. R. Wilcox. Op.cit., pp. 59 ff.

10 It should be pointed out that an MP is said to abstain against the party line when he abstains and "abstain" is not the party line. Therefore we exclude cases when the absolute majority of the party abstained.

11 When the majority have abstained, only the minority of yeas and nays respectively have been counted as votes against the party line.

12 The question is not relevant for the Communists due to their small number.

13 On this problem see, e.g., W. H. Riker. "A Method for Determining the Significance of Roll Calls in Voting Bodies", in J. C. Wahlke &: H. Eulau (eds.). Legislative Behavior, Glencoe, 111., The Free Press, 1959, pp. 377 ff.; Mogens N. Pedersen. "Partiernes holdning ved vedtagelsen af regeringens lovforslag 1945_—66", Historie, vol. VII, no. 3, 1967, pp. 421 ff.

14 Here we have not taken the abstentions into account.

15 P.-E. Back. "Det svenska partivasendet", in Samk'dlle och riksdag, vol. 11, Stockholm, Almquist & Wiksell, 1966, p. 133. Mogens N. Pedersen. »Consensus and Conflict in the Danish Folketing 1964^—65", Scandinavian Political Studies, vol. 11, 1967, p. 148.

16 Mogens N. Pedersen in op.cit. (SPS 2, 1967, p. 152 ff.) has employed an index of distance which measures the distance between the parties by pairs. Cases of joint voting have been assigned the point value of 0, cases of voting in opposition 1 point, and abstentions 1/z point. The total point value has been divided by the number of divisions. The higher the index value is, the greater the distance. Our two-party combinations can serve the name purpose (cases when the absolute majority of the one party has abstained have been regarded as voting in opposition). They provide, however, an inverted picture, i.e. an index of nearness and not an index of distance. The nearness of the parties is however apparent, without undue complication, from the absolute figures in Table 19 because the number of divisions during both years is almost exactly equal.