University of Gothenburg

Swedish politics in the post-war period has presented a picture of high stability. The most conspicuous indicator of this stability is, perhaps, the longstanding Social Democratic administration, despite the fact that its parliamentary base has been at times quite meager or has been broadened by coalition formation. In a broader sense, stability has been also a characteristic feature of Swedish political life in general during the last decades. Although conflicts have by no means been absent, they have hardly reached the stage of acute crises or even prevented a far-reaching understanding among the parties. With the exception of a temporary increase in the Communist vote at the end of World War 11, political extremism has gained only insignificant support.

The party balance in the electorate gives a similar impression of stability, if one considers only the combined voting strength of the "bourgeois" - or nonsocialist — parties, on the one hand, and the Social Democratic vote, on the other. As Table 1 indicates, only minor shifts have occured from one election to another. However, stability in voting behavior has not been so complete as to exclude significant tendencies toward political change. Thus it can be seen from the table that the combined non-socialist share of the vote showed a continuous increase in the Lower House elections from 1948 to 1958. The same trend appeared — although on a lower level — in the communal elections from 1946 to 1958. If we consider the entire sequence of elections during this period, the trend becomes more irregular because in all communal elections the bourgeois share of the vote has been less than in the previous parliamentary election. In contrast, beginning with the 1958 parliamentary election, the Social Democratic Party has shown a series of uninterrupted gains up until the communal elections in 1962.¹

The non-socialist part of the electorate is divided among three "bourgeois"

---

* This article is a revised version of "Politisk rörlighet och stabilitet i valmanskåren", published in Statsvetenskaplig Tidskrift, 1964:4. The Swedish manuscript was completed in July, 1964. The study does not cover the 1964 parliamentary election, although a few references to this election appear in the footnotes.

parties, i.e. the Conservative Party, the People's Party (Liberals), and the Agrarian Party, which has changed its name during this period to Center Party but still draws the bulk of its support from the farming population. (Cf Table 2.) The percentage of the votes obtained by the bourgeois parties combined, and the percentage for the Social Democratic Party as well, varied between slightly more than 44% and slightly more than 50% in the elections from 1946 to 1962. Although the shifts occuring from one election to another can be considered moderate, it should be noted that they nevertheless took place at the very limit of majority formation.

The fact that no non-socialist government was formed during the period examined here cannot be explained as being entirely due to the stability of the division of party preferences in the electorate. It was also due to the way the two-chamber system of the Parliament functions and to conditions created by parliamentary party politics. In terms of parliamentary seats there was, in fact, a bourgeois majority in the Lower House after the 1956 elections, and a sort of tie-vote situation arose in 1958-1960. During the whole period, however, the Social Democrats could muster a majority in the Upper House which is elected by county and city councils under a system of successive renewal.² Contributing to the same effect was the split between the non-socialist parties, as reflected in the fact that the Agrarian Party joined the Social Democrats in a coalition government during 1951—1957. In addition, finally, the Social Democrats have usually been able to rely on the support — although uncertain — of the Communist parliamentary group.

We may now turn to another aspect of political change within the electorate, i.e. the variations in the proportional strength of each of the non-socialist parties. These are somewhat greater than the shifts discussed above. In the elections between 1946—1962 the difference between the highest and lowest percentages was 8.1 per cent for the Conservative Party and 8.8 per cent for the People's Party, while the total bourgeois percentage for the same period varied within 6.3 percentage points. The People's Party reached its highest percentage of the vote — 24.4 per cent — in the 1992 parliamentary election, and the lowest in the 1946 and 1958 communal elections, in both cases 15.6 per cent. The Conservative Party had its greatest success in the 1958 communal elections (20.4 per cent of the votes) and received the weakest support in the 1948 and 1950 elections (12.3 per cent).

The relation between the trends in the development of the electoral support of the bourgeois parties can briefly be described as follows. The People's Party started with a marked election success in 1948, when it received as much as 22.8 per cent of the votes, and remained on about this level until 1956, while the Conservative Party, in contrast, was considerably weakened at the beginning of the period but then began a series of gains with the 1952 elections.³ After that the strength of these two parties developed again in opposite directions in both of the 1958 elections and in the parliamentary elections of 1960. Serious losses suffered by the People's Party in the 1958 elections were thus followed by simultaneous but somewhat weaker increases for the Conservative Party. In 1960

the tendencies were again reversed since the election resulted in some recovery for the People's Party, parallelling a somewhat more marked weakening of the Conservative Party. Finally, both parties suffered some setbacks in the 1962 elections.

This development partly corresponds to a series of complementary increases and decreases in the percentage of votes cast for the Agrarian/Center Party. An examination of the changes from 1946 to 1956, on the one hand, and from 1956 to 1960, on the other, will serve to bring out this relationship. In the 1956 election, with 40.9 per cent, the Conservative and the People's Party received their largest combined share of the vote (the Conservatives got 17.1 % and the People's Party 23.8 %). Thus, the two parties had attained an increase in their electoral strength of 10.4 percentage points since 1946. But, correspondingly, there was a simultaneous weakening of the Agrarian Party from 13.6 to 9.4 per cent of the vote, i.e. a decrease of 4.2 percentage points. As these figures indicate, the Agrarian Party had to face the fact that a substantial part of its adherents abandoned the party and switched to the bourgeois opposition during the Social Democrat-Agrarian coalition.

As mentioned above, the main trend changed in 1958 when the Social Democratic Party entered a new period of growth. In the years from 1957 to 1960, the predominating issue in Swedish politics was a tense and prolonged controversy over a great reform of the pension system.⁴ Ultimately, the reform was carried through in accordance with a Social Democratic proposal but it was not until after a referendum in 1957, a governmental crisis, and a dissolution of the Lower House of the Parliament that the new pension act could be enacted and brought into force in 1960. The non-socialist parties were all opposed to the Social Democratic pension scheme but, at the same time, they were badly split. During the course of the struggle, the People's Party suffered a severe electoral defeat in 1958, when it tried to advocate a compromise plan which no other party was willing to join. In the 1960 election campaign, the Conservative Party tried to revive the battle by demanding that the newly introduced pension system should be immediately abolished, and this attack was combined with other proposals involving a considerable cut in the costs of the social welfare system. At this stage, both the People's Party and the Agrarian (Center) Party chose to disassociate themselves from the Conservative's "extremism" with regard to social welfare policies. As a consequence, the People's Party decided to change their standpoint and sided with the Social Democrats on the pension question. Thus isolated in a "rightist" position, the Conservative Party met with a severe reversal in the 1960 election. Finally, the Agrarian (Center) Party left the coalition government in 1957 and took from the beginning a strongly negative position on the pension question. In the 1960 campaign, however, the Party chose to approach the standpoint of the People's Party. In this way, the Agrarian/Centre Party managed not only to win back what it had lost of farmers' votes but also to gain new support within the urban middle class at the cost of the two other opposition parties.

As a result of this development, the share of the votes received by the

Conservatives and the People's Party was again brought down to 34 per cent in the election of 1960 (with 16.5% and 17.5% respectively). Once more, however, a compensating change took place among the bourgeois parties, since a substantial portion of the decrease was balanced by a gain of 4.2 percentage points for the Center Party from 1956 to 1960. The 1958 and 1960 parliamentary elections provide striking examples of this internal complementarity of the changes in the support of the bourgeois parties. In the Lower House election in 1958 the People's Party lost as much as 5.6 percentage points, but despite of that the combined strength of the bourgeois parties did not decrease. In the 1960 election, it was the Conservatives' turn to loose 3.9 percentage points as compared with the preceding communal elections in 1958, but the gains made by the Center Party and the People's Party compensated for more than one half of the loss.

Actually, despite the fact that there have been marked fluctuations in the strength of the bourgeois parties in the electorate, there has been no case of simultaneous gains for all of these parties in the post-war elections. The Conservative Party and the People's Party, both in opposition throughout the period, have achieved a simultaneous increase only in the 1952 and 1956 elections. To sum up, we find that the development of the non-socialist parties forms a sequence of largely counterbalancing changes.

Consequently, stability in voting behavior has been considerably less than may be suggested by the aggregate figures in Table 1. Especially in the case of the People's Party, fluctuations in the number of votes received are quite impressive. This is most clearly brought out if the gain or loss in each election is calculated as a percentage of the number of votes cast for the party in the immediately preceding election. By using this method we find that the People's Party showed a 70 per cent increase in the number of votes received in the 1948 election, a 24 per cent decrease in the 1958 Lower House election, and 23 per cent increase from the communal elections in 1958 to the Lower House election in 1960.

It is obvious that to a large extent changes in the party division of the electorate took place, within the electoral base of the bourgeois parties. This becomes particularly noteworthy if we take into account the fact that public political debate in Sweden has centered on the position of the Social Democrats as the party in power. Nevertheless, the non-socialist parties have fought a series of battles between themselves and as a consequence they have been engaged in a continuous internal competition for voting support from the urbarr and rural middle class. This forms the political backgound of the peculiar wavering in the partisan division of the bourgeois vote that becomes visible in the election returns. No doubt, these conditions also provide a partial explanation of the fact that the political fluctuations occuring in the electorate did not affect the Social Democrats' position as a Government party.

The tendencies we have dealt with thus far are those apparent in the distribution of votes among the parties. This kind of aggregate statistics can indicate only the net effect of all the changes that take place in the enfranchised population during a sequence of elections. In reality, each such net effect must be

due to a multitude of shifts in individual behavior which constitute flows in different directions and with different characteristics in the electorate. When dealing with "political mobility" — in terms of electoral behavior — we may distinguish between two main components of change: first, there is the shifting of party preferences which leads to an exchange of supporters between the parties, and secondly, there is the shifting between voting and non-voting. Both of these components contribute to the number of votes a given party will receive, and the total amount of change within its electoral support must naturally be greater than the net result. In addition, there is also the slow but continuous renewal of the enfranchised population that is due to mortality and the inclusion at each election of young citizens reaching the voting age. In the latter case, we also have to take into account the significance of short-term fluctuations, since the division of party preferences as well as the interest in exercising the right to vote in the newly enfranchised age-group may vary from one election to another.

In the following, we shall try to illuminate the extent and nature of political mobility in Sweden through an inquiry into these components of electoral change. For this purpose, we shall utilize data concerning individual voting behavior which have been gathered through a series of nationwide interview sample surveys, carried out by the Institute of Political Science, University of Gothenburg. The surveys have been undertaken in cooperation with the Survey Research Center of the Swedish Bureau of Statistics, which was in charge of the field work and the sampling procedure. The main body of the data utilized in the present study are drawn from an interview survey conducted at the time of the 1960 election, but in some parts of the analysis the material is supplemented with data provided by interview studies of the 1956 election and the referendum in 1957.⁵

The samples used in these surveys may be described as two-stage probability samples, designed to be representative of the enfranchised population up to 80 years of age. The selection of the interviewees was made by sampling from the population registers in 70 areas. The sample of the 1960 survey comprised 1,589 individuals of whom 1,466 (or 92 %) were actually interviewed. The interview material has been supplemented in several ways with the aid of official registers. In the present study, we can thus take advantage of the fact that the voters' registers have been checked in order to determine whether the interviewees have actually voted or not, and the data obtained in this way, rather than interview responses, will be utilized in the analysis. It may also be mentioned that the population registers have been used as the source of information concerning residential mobility in the 1960 sample of respondents. In case of the 1960 study, the procedure of checking the participation in the voting registers was furthermore extended to cover a three-election sequence, comprising the communal elections in 1958, the parliamentary election in 1960, and the communal elections in 1962. The material thus provides particular possibilities for assessing the regularity of the voters' electoral participation. A major part of the following analysis will be devoted to this aspect of political mobility in the Swedish electorate.

Political Mobility in the Electorate 1956-1960

In this part of the study an account will be given of the Swedish electorate in order to illuminate the extent of variability in party choice and voting turnout during a period when three parliamentary elections and one election to the communal councils were held. In particular, we shall give attention to the extent of political mobility within the electoral bases of the four main parties.

The data utilized in the following analysis are drawn from a series of questions in the 1960 election survey which was designed to give as complete a picture as possible of the "voting history" of the respondents.⁶ For the 1956—1960 period the questionnaire contained detailed queries concerning participation and party choice in each of the elections. Subsequently, respondents were asked in more general terms whether they had shifted their party allegiances in past elections and how regularly they used to vote. Finally, there was a specific question about the respondent's party preference in the first election in which he had voted.

Admittedly, there are inherent deficiencies in the data obtained through these retrospective questions. There must have been, as a matter of course, some errors of recollection in the responses and, in addition, the answers to our questions may have been influenced to some extent by a lack of willingness to give information about non-voting.⁷ No doubt, some respondents may also have been reluctant to reveal changes in their partisan persuasions. It may be presumed that such errors work in the direction that stability in party preferences, and regularity in participation become somewhat overemphasized in the interview material. It should be kept in mind, however, that at least one source of error has been counteracted by the checking of the respondents' participation in two of the elections in the voters' registers. Furthermore, we may assume that errors in recalling party choice in earlier elections have occured mostly in indicating one's voting in each specific election, while it must have been easier for the respondents to remember whether they had shifted their party preferences at all

over a longer period of time. In other words, changes from one party to another
that took place in any of the four elections 1956-1960 may be reflected in the
responses even if they happened to be located at wrong points of time.

In Table 2 the distribution of voters among the parties in the interview sample is compared with the corresponding distributions in the entire electorate for the 1958 communal election and the 1960 parliamentary election. The table indicates that the proportion of Social Democrats is somewhat too high while that of the Communist Party is somewhat too low in the sample; on the whole, these deviations balance each other. It can also be seen from the table that the set-back experienced by the Conservatives is reflected in the interview data only as a slight trend. The deviations as such are not great enough to cast any particular doubts on the representativeness of the sample.⁸ They are pointed out here only to emphasize that the survey sample cannot be expected to reflect in an altogether precise way the changes that have occured in the partisan division of the electorate from election to election.

In order to obtain a general picture of the scope of variability in participation and partisanship we have undertaken a comprehensive analysis of the whole interview material concerning voting behavior in all of the elections between 1956 and 1960. The results are shown in Table 3. The material presented in this table includes only those respondents who were enfranchised during the entire period and participated in at least one of the elections.⁹

In Table 3, in the columns referring to different parties are listed all voters who in at least one election cast their vote for the party in question. The column headed "Bourgeois party" includes all those who have voted for a non-socialist party in at least one of the elections, i.e. all respondents who are included in the separate columns for the three bourgeois parties. This method of presenting the material, of course, implies that party changers must appear in at least two of the party columns in the table. Obviously, it must also mean that the number of interviewees in each party column considerably exceeds the number of votes which the party could receive in any specific election. In the "Total" column of the table the percentage distribution is in turn based upon the net number of respondents in each category.

As is shown in the "Total" column of Table 3, those who participated in all of the elections during the period and, in addition, consistently voted for the same party make up about two thirds of all voters included in the analysis.¹⁰ Somewhat less than one fourth failed to vote on at least one occasion but never switched from one party to another. Party changers represent 13 per cent. Although it is not indicated in this table, some of these party changers also abstained from voting in one or more elections. About one half of the party changers category was oscillating only among the non-socialist parties, while the rest was shifting between bourgeois and Social Democratic or Communist voting; there was also a small group who changed between the Social Democrats and the Communists. Since the number of Communists in the sample is very limited, they are not treated separately but are included in the "Total" column (Communists who changed parties also appear in the "other party changers" category in other columns).

In addition to the data given in the table, it may be mentioned that more than half of those voters who shifted only among bourgeois parties changed their sympathies between the Conservative Party and the People's Party. In the exchange of voters between non-socialist parties and the Social Democratic Party, Conservatives played a quite insignificant role; only 16 per cent of this category voted on some occasion for the Conservative Party.¹¹

A comparison of the party columns in Table 3, finally, brings out a very significant feature in the picture of the Swedish electorate provided by our data. It is apparent that the electoral support of the Social Democratic Party must have been considerably more stable than that of the bourgeois parties. The data suggest quite clearly, too, that this was due to a relatively great amount of

internal shifting of voters among the Conservative Party, the People's Party and
the Center Party.

It should be noted that this disparity between the Social Democratic Party and the bourgeois parties is — at least to some extent — conditioned by the existing distribution of electoral strength in the party system, i.e. the fact that the party system comprises one very large party along with several smaller ones. Even if there would have been a perfect equilibrium - for example created by the exchange of numerically equal flows of party changers among the parties — the "turn-over" in the voting support would have been proportionally greater for the bourgeois parties than for the large Social Democratic Party.

If voting behavior in the 1960 election is taken as the starting point for the analysis, we may calculate for each of the parties the proportion of its adherents who had ever voted for some other party in past elections. Then a still more marked difference becomes evident. As can be seen from Table 4, nearly nine tenths of those who voted for the Social Democratic Party in 1960 said that they had never voted for another party, while the corresponding proportions for the non-socialist parties are consistently lower than 60 per cent. Virtually the same results were obtained through a parallel analysis of the data provided by the 1956 interview survey, which is also presented in the table.¹²

We have already observed that this difference in the stability of party support may be due to the structure of the party system. However, the prerequisites of political opinion formation which are created by the party system have further implications which deserve some attention. Apparently, the Swedish multi-party system functions in such a manner that shifts in party affiliation involve a new choice between the opposition parties about as often as a reconsideration of the voter's decision whether to support the government policy or the opposition. This signifies a differentiation with regard to the voters' partisan attachment which seems, in turn, to be closely connected with the working of the party

system. The nature of this differentiation may be illuminated if we turn our attention to interview data reflecting the voter's attitude toward his party. This is illustrated in Table 5. The table is based upon an interview question which provides a simple measure of the strength of the voters' partisan convictions. In this question, the respondents were asked whether they felt themselves to be "strongly convinced" adherents of the party they supported.¹³

As indicated in Table 5, Social Democrats feel themselves to a particularly great extent to be convinced supporters of their party. It seems to be a wellfounded inference that the internal competition among the non-socialist parties plays a major role among the factors that contribute to make bourgeois voters less likely to form a strong sense of attachment to some specific party. That is, a considerable portion of these voters may be determined in their aversion to the Social Democratic government but they are less convinced in regard to the best way of expressing it. The other side of the coin, then, appears in the peculiar mobility of party choice in this part of the electorate which has already been documented. To explain why the People's Party shows an even lower proportion of convinced adherents than the two other bourgeois parties we have to make an additional assumption about the linkage between the party system and the strength of voters' party allegiances. It seems plausible that the People's Party's middlemost position in the party system makes them particularly able to compete with the Social Democrats for the support of those voters who are vacillating between approval and criticism of Social Democratic policies and — for that very reason — are unlikely to become convinced adherents of the party they choose to vote for. As a consequence of its location in the party system the People's Party thus attracts votes from two different sources of halfhearted support.¹⁴

The interpretation offered here requires that those differences between the
parties form a persistent pattern. The good convergence between the results
obtained in the 1956 and 1960 election surveys lends support to this inference.

Stability and Variations in Electoral Participation 1958-1962

For more than fifty years the Swedish Central Bureau of Statistics has conducted
large-scale investigations of voting behavior in parliamentary elections with

the aid of data drawn from voters' registers. As a result of these investigations, official election statistics bulletins contain both extensive and penetrating information regarding voting turnout in different population groups as weH as in the electorate as a whole.

Although much more limited in scope, the interview survey of the 1960 election also provides data that offers particular advantages. In the first place, we are able to undertake the analysis with the aid of a more exhaustive set of variables, including the voter's party choice and political attitudes as well as his electoral participation. Our collection of voters' registers data concerning the respondents in the sample is not restricted to one single election. Hence we have at hand a very reliable measure of the regularity in the individuals' participation and we can also introduce this factor as a behavioral variable in the analysis.

A general picture of the rate of variability in voters participating in three consecutive elections is presented in Table 6. The table includes all respondents in the 1960 survey sample who were enfranchised throughout the whole period; first time voters in the 1960 election and respondents who died between 1960 and 1962 have been excluded.¹⁵ In the table, the sample is divided into two broad socio-economic strata — middle class and working class — both of which are subdivided into more specific categories. It should be noted, that the term "middle class" is used here to denote all occupational groups which have not been defined as belonging to the "working class". We have refrained from defining an "upper class" category for the simple reason that we would have gotten numerically too small categories in our sample if the employer and white collar groups hade been further divided by social status criteria. The "working class" has been split into two groups of approximately equai size. One of the groups has been defined so as to comprise workers engaged in manufacturing, construction, and mining. Other wage earners in the working class — engaged in commerce, transportation, agriculture, lumbering, and various domestic and service occupations — have been put in an "other workers" category which is, of course, occupationally very heterogenuous. This division by no means implies a sub-classification in respect of social status, i.e. it does not define an "upper" and a "lower" working class. It is justified, instead, by the fact that "workers engaged in manufacturing, construction, and mining" can be distinguished as a politically very homogeneous segment of the working class that consistently shows an especially heavy predominance of socialist voting. In the 1960 election, for example, the bourgeois parties received only 12 per cent of the votes in this group as compared to 23 per cent in the remainder of the working class (cf Table 17).

The socio-economic classification described here includes all respondents in the sample. Family members who were not employed have been classified according to the occupation of the head of the household; married women were classed according to the husband's occupation whether or not they were themselves employed. Retired persons have been classified according to their earlier occupations.¹⁶

As is shown by the marginal distribution in Table 6, about two thirds of the enfranchised persons participated in all three elections. We also find that there was a very thin layer of citizens — consisting of 5 per cent — who remained altogether politically passive throughout the period. On the other hand, one fourth failed to vote in one or two elections but exercised their voting right on at least one occasion. In fact, the scope of this kind of change in voting behavior is considerably greater than that of variability in party affiliation (cf. the marginal distribution in Table 3).

Scholarly studies and official election statistics have long since recognized that there exist relationships between electoral participation and determining factors which are bound up with the individual's social environment and conditions of life. Age, sex, marital status, and social status may be listed to exemplify such determining factors which have been found to be valid in several countries.¹⁷ Among these, socio-economic status has a particular importance for the present study since the partisan division of the Swedish electorate follows socio-economic division lines to a great extent.

In Sweden the investigations undertaken by the Central Bureau of Statistics have shown that the size of the turnout in parliamentary elections is usually somewhat lower among workers than among employers and white collar employees. The general observation that the voting rate tends to increase with increasing status is thus sustained, but the differences prove to be quite moderate when broad occupational groups are compared (see Table 7).

In his book Political Behavior, Herbert Tingsten has presented a theory about the existence of a regular relationship between the level of participation in the electorate as a whole, on the one hand, and the magnitude of group differences in turnout, on the other hand.¹⁸ According to this theory - "the law of dispersion" — an increase in electoral participation is always followed by a decrease in group differences, while decreasing participation leads to an increase of such differences. In both cases, "the law of dispersion" expects the change to be especially great in those categories of voters who usualy show a relatively low voting rate,

i.e. young people, unmarried, low-income groups, etc. The theory thus implies a particularly high degree of variability in the level of turnout of such population groups. As Tingsten points out, the existence of such a relationship would lead us to expect political effects of fluctuations in turnout because of the differences in the social anchoring of the Swedish parties. The same idea has appeared rather often in public political debate; it is then supposed that the Social Democratic party is favored if turnout becomes high but suffers if it becomes low. As will be shown in the following, the results of the present inquiry provide good reasons to doubt the validity of this generalization.

To a certain extent, Table 7 lends support to Tingsten's "law of dispersion". Thus it shows that differences in participation frequency among occupational groups were particularly small in the 1960 election, when the turnout was higher than in any previous election. However, the table shows also an example of a contrary tendency, that is the 1958 parliamentary election, when the voting rate decreased somewhat more among employers and salaried employees than among workers.¹⁹ /'

If we now return to the socio-economic classification employed in Table 6, we find, again, that on the whole differences in electoral participation are quite small. The "other workers" category, however, forms an exception. Obviously there is a larger proportion of irregular voters and consistent non-voters iomong "other workers" than in the rest of the enfranchised population. Of special interest is a comparison with "workers engaged in manufacturing, construction, and mining". Admittedly, the division of the working class employed here does not square particularly well with conventional definitions of social stratification and, in addition, it may well be argued that it is not based on wholly unambiguous criteria. Nevertheless, it does have the merit of allowing us to define an important Social Democratic stronghold in the working class which shows about the same amount of regularity in participation as do the middle class groups. In fact, we have found the same differences between these two segments of the working class throughout our series of voting surveys.²⁰

A further illustration of the fluctuations in voting turnout in different socioeconomic
strata is presented in Table 8, where the percentage of participants
is given for each of the elections: 1958, 1960, and 1962. The comparisons made

possible by this table are of particular interest, since the Central Bureau of Statistics does not conduct investigations on the basis of the voters' registers for communal elections. However, it must be emphasized that the results derived from the interview sample do not have as high precision as to permit reliable conclusions from small differences among the population groups.

Nevertheless, the interview sample seems to reflect changes in voting turnout which are worthy of note. Thus we find that the increment of participation that occured from the communal elections in 1958 to the parliamentary elections in 1960 is more noticeable in the working class than in the categories of employers and white collar employees. Although the differences in Table 8 are quite small, the tendency may be confirmed by a corresponding comparison between the data from the parliamentary elections in 1958 and 1960 appearing in Table 7, which is based on much larger samples. The decline in turnout in the 1962 communal elections, on the other hand, seems to have been of about equal size in the middle class and the working class; in both cases about 6 percentage points. In each of the elections we find that "workers engaged in manufacturing, construction, and mining" show a higher voting rate than the category of "other workers". Moreover, the table suggests that the decrease in turnout was relatively small among voters in the first-mentioned group.

By using a more elaborate social status classification than the one employed here, we would have been able to distinguish certain smaller categories with extremely high voting rates, e.g. owners of large industry and business firms, salaried employees in managerial or higher civil servant positions, etc. Similarly, working class categories showing especially low participation frequencies could be obtained, for example, by controlling the income factor.²¹ However, Tables 6—8 serve to bring out the still more important fact that within a very wide sector of the enfranchised population electoral participation has reached a high and even level without any considerable differences among occupational groups.

Actually, other social environment and social role factors have as great, or greater importance for the individual's propensity to exercise his voting right although none of these factors show more than a weak relationship to the voter's party choice. Among such factors, marital status is perhaps the most outstanding. Married voters — in all occupations and age groups — show a consistently higher voting rate than the unmarried. In the 1960 election 92 per cent of the enfranchised married population voted, while participation among unmarried and formerly

married in the electorate was 74 per cent.²² Another example is provided by the relatively low turnout in the youngest and the oldest age groups, although abstention from voting among the latter is, of course, to a great extent caused by sickness and other disabilities. In contrast, the difference between the voting rates of men and women has been almost completely eliminated in Sweden. (See Table 7)

Geographic mobility is also one of the environmental factors associated with a low propensity to vote, although this is perhaps less well known. For the purpose of exploring the effects of geographic mobility we have collected data from the population registers which indicate how long respondents in the 1960 survey sample had been living in the parishes where they were last registered. With the aid of these data we have divided the interview sample into the three categories appearing in Table 9. One of these categories includes persons who have been residing in the same parish since 1946, while those who have moved since 1946 have been classified in two groups according to the year they have moved to their present places of residence.²³

From the point of view of voting, a migration often implies that in the election following his move a voter will have to cast his ballot at a new polling station. If he still remains in the voters' register of his former election district, he must either travel to that place on the election day or use the possibility of voting at a post office. In either case it may become somewhat more difficult — although certainly not much — for a person to participate in the election. In the table, time intervals have been chosen so that those who have moved during the last of the periods must have been subjected to this kind of inconvenience in at least one of the elections between 1958 to 1962.

Table 9 indicates very clearly that migration to a new place of residence decreases the propensity to vote in the following election. However, this effect is by no means permanent, as can be seen from the fact that those who had moved between 1947—1956 show as high and regular participation as the category of people who had not changed their place of residence.

The explanation that first comes to mind is, of course, that the discrepancy appearing in the table is due to the above-mentioned temporary inconveniences in the voting procedure which usually accompany a change of residence.²⁴ It seems plausible, however, that the pattern appearing in the table also reflects a more complex relationship. We know that residential mobility is often an indication of social mobility. This implies that to a comparatively great extent movers must be exposed to new or changed social norms of political behavior. As a consequence, residential mobility may be assumed to increase the probability that the individual will experience a conflict between mutually counteracting expectations concerning both his group identification and his party affiliation. Several studies have provided evidence that such a "social cross pressure" situation has a dampening influence on political interest and participation ²⁵ However, a full documentation of such an interpretation of the data given in Table 9 would have required an inquiry into individual-group relations, which cannot be undertaken within the scope of the present article.

Party Preference and Regularity in Electoral Participation

We may now turn to the discussion of the problem of how the balance of partisan strength is affected by fluctuations in turnout. To provide a starting-point for the following inquiry, we shall first present an analysis of the relation between party preference and regularity in electoral participation from 1958 to 1962.

The relevant data are presented in Table 10. In the table, supporters of the four main parties in the 1960 election are classified according to their party choice on that occasion. Those who abstained in 1960 but voted in the 1958 communal elections are classified according to their party preference in the latter election. Here — as in several following tables — the analysis is restricted to include only those respondents who were enfranchised throughout the period. Of course, persons who did not vote in either 1958 or 1960 or did not reveal any party preference must also be excluded.

A comparison between the columns in Table 10 makes it quite clear that there are only slight differences among the parties. The proportion of voters who consistently participated in the whole sequence of elections from 1958 to 1962 is about the same for all of the parties, and proportions of irregular voters who cast their votes in only one or two of the elections are about equal, too. Our data thus lead to the conclusion that there was virtually no relationship between the voter's partisan orientation and the degree of regularity in his participation. It is obvious that in order for some party — or parties — to be generally favored by a high turnout or disfavored by a low turnout the results must have been different. The period dealt with here includes one occasion of extremely high turnout — i.e. the 1960 election — together with one previous and one subsequent election

with voting rates on a level more normal for the post-war era. Consequently, had
some party been especially sensitive to changes in the level of turnout it would
have appeared in Table 10 through a higher percentage of irregular voters.

Inspection of the aggregate statistics for the whole series of post-war elections appearing in Table 1 lends support to this inference. We find in this array of data no regularity to the effect that a decline in turnout from one election to the following is necessarily accompanied by a concomitant increase in the bourgeois share of the vote, nor have decreases in that proportion always been coupled with rises in the level of turnout. It might be added that a rank-order correlation based on the bourgeois percentages and the participation figures in Table 1 confirms that one can find only a very slight relationship between these two factors (r = - 0.185).

It must be emphasized that these findings do not prove that in the elections under study the aggregate result of individual shifts between voting and nonvoting must have been politically neutral. But we may infer that // fluctuations in turnout have contributed to changes in the bourgeois-socialist division of the vote, it cannot have been due to the existence of a peculiar variability in electoral participation among adherents of the Social Democratic Party. Variability in participation must have had a much less one-sided effect on the balance of electoral strength. The implications of this inference will be illuminated in the following.

Political Attitudes and Variability in Turnout

In Public Opinion and Congressional Elections, William A. Glaser has presented a study of the fluctuations in participation in American elections. In this searching analysis, Glaser has propounded a valuable theoretical framework for this type of research. Another recent treatment of the same subject is to be found in Angus Campbell's penetrating study of the political shiftings that occurred in the American electorate from 1948 to 1952 and from 1956 to 1958 ("Surge and Decline: A Study of Electoral Change"). Campbell demonstrates that on these occasions the alterations in the party division of the vote were determined, at least to a considerable extent, by changes in the rate of turnout. Although Glaser's study is more directed toward the elaboration of a formalized theory than that of Campbell, both authors reach very similar conclusions.²⁶

We may briefly summarize here some of the main points that characterize both Campbell's and Glaser's discussion. Both authors distinguish between a long-range disposition to participate in elections and the specific motivation to vote that is derived from stimuli elicited by the campaign of each individual election. The latter factor can, of course, vary in strength. Some elections will generally be considered as especially crucial or dramatic, while others evoke less excitement. Long-range motivation on the contrary, is determined by persistent social norms and role expectations, or by relatively stable political attitudes, such as the degree of interest in politics in general and the strength of the individual's partisan conviction. We may use here the term "propensity to vote" to denote

the resulting long-range disposition. The two types of motivation work together: The weaker the voter's propensity to vote, the stronger stimuli must be elicited by the campaign for the voter to find reasons enough to exercise his right to vote in a given election. The stronger the propensity to vote, the less the importance of variations in the intensity of campaign stimuli from one election to another; if the propensity to vote is strong enough even a subdued campaign will offer a sufficient stimulus for the voter to participate. With the aid of his theoretical model and empirical survey data, Glaser demonstrates that this relationship really exists and, furthermore, he shows that it must entail a greater variability in participation among persons who have a weak propensity to vote than among those with a stronger propensity.

In his analysis of electoral behavior in the "Eisenhower era", Campbell points out the triangular relationship between the nature of the election campaign, the size of the turnout, and the election outcome. Since campaigns differ in regard to their capacity to engender public concern, Campbell distinguishes between "high stimulus" and "low stimulus" elections. "High stimulus" elections — like those in 1952 and 1956 — have the effect of mobilizing voters having low propensity to vote and weak sense of party attachment to a particularly great extent. Campbell shows that the characteristic campaign themes of the "high stimulus" elections had an especially great influence on party choice precisely among the irregular voters who were thus mobilized. Consequently, the high level of turnout in the 1952 and 1956 elections strengthened the success of the Republican Party, which was favored in these campaigns because of the unusual attractiveness of its presidential candidate. The decline in interest and participation in the "lowstimulus" mid-term elections had the reverse effect, i.e. Republicans suffered the greatest loss of inconsistent supporters.²⁷ In The American Voter, Campbell, et al. complete the picture by analysing the relationships between a variety of political attitudes and electoral participation. In short, these analyses demonstrate that there is a strong relationship between the strength of the individual's political involvement and the intensity of his partisan preference, on the one hand, and the probability that he will cast his vote in a given election, on the other.

In Sweden participation in elections has become stabilized on a higher level than in the United States; fluctuations are weaker and differences among population groups are smaller. Yet it is obvious that Campbell's and Glaser's research directs attention to factors that are relevant also for the study of Swedish politics. It is true that our comparison of the various parties indicated that there were virtually no party differences in regard to variability of participation in the course of a sequence of elections. However, we should by no means overlook the fact that there exists a considerable variation in the intensity of partisan feelings and political interest among individuals which cuts across the party system. All parties, indeed, rely on the votes of both convinced partisans and lukewarm supporters. The differences in outcome among the elections we have dealt with may imply, furthermore, that the political characteristics of the campaign situations have alternatively favored and disfavored different parties in precisely the way that the American investigations suggest. As a matter of fact, the

concept of "election wind" has long been accepted into Swedish political vocabulary. Consequently, we may assume that in each election the party (or parties) which — because of the general political situation - becomes especially attractive and successful will be more able to mobilize supporters with low propensity to vote than the party (or parties) which in the campaign come to be put in an unfavorable position.²⁸ It should be noted that such a relationship would not be incompatible with the results presented above, since short-term fluctuations in the party divison of the votes cast by irregular participants may well be evened out as soon as more than two elections are taken into account.

From the theoretical framework presented above, we may deduce the following two hypotheses which we shall test with the aid of the Swedish election surveys: (1) The stronger the voter's political motivation - in terms of political involvement and intensity of partisan preference - the greater will be his propensity to vote. (2) In the enfranchised population, individual shifts between voting and non-voting from one election to the next will lead to an aggregate result which is to the advantage of the party that is favored by the general trend in the outcome of the later election.

Propensity to Vote, Political Involvement, and Strength of Party Attachment

It follows from the foregoing reasoning that we may take the degree of variability in the voter's electoral participation as a measure of his propensity to vote. In the following analyses we shall employ two measurements of regularity in voting for the purpose of discriminating between voters with "higher" and "lower" propensity to vote. Thus we have obtained one simple dichotomy by dividing the sample into those who participated in all of the three elections we have checked in the voters' registers, on the one hand, and those who abstained on one or more occasions, on the other. As an additional measure, we have utilized an interview question concerning regularity in participation. Respondents were asked whether in previous elections they had voted "always, nearly always, only sometimes, or never". By combining these measurements we have obtained a more restrictively defined category of voters with "high propensity to vote", i.e. those who voted in all the three elections and, in addition, said that they had "always" exercised their voting right.

In Table 11 an index of "campaign exposure" is utilized as a measure of political interest. The index is designed to provide a combined measurement of active information seeking, and for this purpose it summarizes exposure to campaign propaganda on television and radio, reading of political contents in the daily press, reading of election pamphlets, and participation in political discussions in the primary group environment. The category "Low" includes respondents with a low degree of exposure to all of these channels of informations, while respondents in the "High" category showed a high degree of exposure at least to three of them.²⁹

The results in Table 11 confirm our first hypothesis (1), inasmuch as the

proportions of consistent participants in the 1958—1962 elections among individuals with definitely high and definitely low campaign exposure differ markedly from the average. The relationship becomes still more clear if we take into account the respondents' self-evaluation of their participation in earlier elections. As the second row of the table indicates, propensity to vote increases distinctively all along the continuum represented by our index of campaign exposure.

Perhaps not the least interesting information is to be found in the bottom row of the table, which illuminates how insignificant a size the wholly apolitical stratum must have in the Swedish electorate. Even among the one fifth of the sample who almost completely refrained from taking an active interest in sources of political information, only 8 per cent withdraw entirely from electoral participation throughout the 1958—1962 period.

For measuring the intensity of the motivation to vote, we may also choose as the starting point the voters' attitudes toward the controversial issues of party politics. Thus we may presume that the more decidedly and completely the bourgeois oriented voters disapprove of the current Social Democratic policy, the more eager will they be to participate in elections in order to bring about a change of government. Likewise, the willingness to vote among citizens with a Social Democratic party affiliation should be expected in increase with increasing extent of agreement with Social Democratic policy. This presumption — which is, of course, only a reformulation of hypothesis (1) — is substantiated by the data presented in Table 12. In this table, we have grouped bourgeois and Social Democratic voters along a "scale" of intensity in partisan orientation, which has been obtained by summarizing responses to ten political attitude questions.³⁰

Inevitably, our measurements of "campaign exposure" and "intensity of party preference" must have been influenced to some extent by the voters' short-term reactions to the specific political contents of the 1960 campaign propaganda. Yet, the patterns appearing in Table 11—12 show very convincingly that they reflect also persistent attitudes which contribute to determine the individual's long-range propensity to vote. The same seems to be true of the attitudes measures that are included in the analysis in Table 13. In this table, we utilize the aforementioned interview question concerning "conviction" in party preference to obtain a measure of the strength of the voters' party attachment. It is combined in the table with an interview question which asked respondents to judge themselves on their degree of political interest. The question was phrased so as to offer the following four response alternatives: "very", "rather", "not especially", and "not at all interested". The category of "interested" appearing in the table includes persons who answered that they were "very" or "rather" interested in politics.

Table 13 provides a further confirmation of the hypothesis (1) that we are attempting to test here. "Convinced" partisans prove to be regular participants much more often than those who have only a weak sense of affiliation to the party they support. As can be seen in the table, the inclusion of "political interest" in the analysis leads to a very clear gradation of the propensity to vote among the four sub-categories. If we exclude from the analysis the voters' selfevaluation of their regularity in participation the relationship persists but it becomes somewhat less apparent. Among "convinced" partisans 86 per cent were consistent in their participation in the 1958—1962 elections, while the corresponding proportion was 76 per cent for other voters. In Table 11 we found a similar difference between the two measures of the voters' propensity to vote, i.e. differences in the variability of turnout became more moderate when the

analysis was limited to the short period, 1958—1962. If we take into account the generally high participation level this is, of course, precisely what we would expect. (It should also be noted that Tables 11—12 comprise only those who voted in the 1960 election, and, as a consequence, we have excluded one group of irregular voters.) Even among citizens who are little concerned about politics, failure to vote is a rather unusual way of behaving (or misbehaving). In short, the present analysis is not especially designed to single out those small categories of persons with special ways of life or personality characteristics who are particularly unlikely to participate in politics.

We have, so far, discussed political attitudes and social environment factors without considering their interdependence. The existence of an interplay between the individual's political attitudes and the experiences he makes in the group environment can, of course, be taken for granted although it will not be fully explicated in the present study. However, we shall give some further attention to the aforementioned theory of cross pressure. Because of the social structure of the Swedish party system it can be presumed that both middle class voters with Social Democratic preferences and working class people who are supporters of the bourgeois parties must be comparatively often exposed to social cross pressure, which will appear in the form of a conflict between the individual's political party affiliation and the predominant political orientation within the socio-economic stratum to which he belongs. Consequently, we can formulate the complementary hypothesis that variability in electoral participation should be comparatively high within these categories. The analysis presented in Table 14 was undertaken in order to test this hypothesis.³¹

As can be seen in Table 14, the hypothesis is confirmed insofar as working class adherents of the bourgeois parties show a definitely lower degree of stability in turnout than voters in the other categories. However, we do not find a corresponding dampening of participation among Social Democrats in the middle class. The absence of such a tendency might, of course, lead to the inference that the combination of middle class status and a socialist party preference does not actually increase the likelihood of cross pressure. Another — and perhaps more plausible - explanation may be that middle class Social Democrats more often experience a strong pressure of a social norm which prescribes electoral participation as a civic duty than do supporters of the bourgeois parties in the working class.

The results presented in Table 14 also complete the foregoing analysis in another way. As may have been noted, our interpretation of Tables 6 and 10 led to a seeming inconsistancy between the inferences drawn from the data. Thus it was found, on the one hand, that one segmet of the working class — i.e. the "other workers" category — showed less regularity in participation than other groups, and on the other, that Social Democrats were as regular in their turnout as those who voted for the bourgeois parties — in spite of the fact that the Social Democratic Party draws a major part of its support from the working class. Now, Table 14 reveals that there is, in fact, no contradiction between these observations, since it shows that a peculiar variability in turnout was a characteristic only of the non-socialist voters in the working class. This result of the analysis is, of course, in good agreement with our earlier findings concerning differences in political behavior between the category of "workers engaged in manufacturing, construction, and mining" and the "other workers".

It should be kept in mind, that the concept of "social cross pressure" is used to denote the presence of certain conditions in the group environment which have the effect of increasing the likelihood that the individual will experience an attitude conflict. In several studies it has been demonstrated that a political attitude conflict — whether or not it has its roots in contradictory group pressures — not only evokes an inclination to withdraw from active participation, but also leads to uncertainty in party choice, i.e. the individual becomes more likely to postpone his voting decision or to vacillate between different preferences." As a consequence, one may advance the hypothesis that changeability in party

choice is connected with variability in participation, i.e. a comparatively lowrange propensity to vote.³³ The interview data provide some support to this hypothesis. Among those who were consistent in party choice, 62 per cent were also consistent participants in the 1958—1962 elections as well as in earlier elections, while the corresponding proportion among those who had changed parties on some occasion was 54 per cent.³⁴ The difference between the two proportions is certainly rather moderate. However, we could hardly expect to find a strong relationship, because of the fact that change in party choice and withdrawal from participation can be concieved — at least to some extent — as two alternative responses to the same conflict situation.

If this interpretation is correct, it may, however, lead us to conceive of a more complicated relationship between voting behavior and different aspects of political motivation than the one presupposed above. Thus we may expect propensity to vote to be determined by an interplay between political involvement and firmness in partisan conviction, that is, among voters who do not feel themselves to be strongly attached to any specific party a high degree of political involvement would be expected to counteract the inclination to withdraw from participation. In contrast, a low degree of political involvement would be expected to decrease the propensity to vote even among those voters who never find any reason for changing their party affiliations.

This is precisely the pattern that appears in the data presented in Table 15. Here, the sample is divided according to consistency in party choice as well as the degree of political involvement, as measured by the index of campaign exposure. The categories thus obtained represent — as shown in the table — four different levels of regularity in participation. Those who combine a high degree of political involvement with a stable partisan allegiance are most likely to vote in all elections, while the least willingness to vote is found among those who are low-involved and, in addition, have changed their party preferences.³⁵ In this case too, the analysis thus leads to a result which is in good agreement with our hypothesis (1) about the relationship between electoral participation and political attitudes.³⁶

Turnout and Election Outcome

What impact do shifts between voting and non-voting among irregular voters have on the party division of the vote? In order to answer this question we shall attempt an inquiry into voting behavior in 1956, 1960, and 1962 by a comparative analysis of the data provided by our election surveys. It should be made clear in advance, however, that the survey material sets rather narrow limits to the analysis because of the fact that the various types of behavioral change are represented by numerically small groups in the samples. In order that the samples should not be partitioned into too many sub-categories we have thus had to put all supporters of the bourgeois parties in one common category. As a consequence, the following analysis will be limited to illuminating the influence of variability in participation on the relative strength of the bourgeois parties, on the one hand, and the Social Democrats, on the other. Yet, the basis

of the analysis is unavoidably precarious because of the smallness of some
categories.

In the 1960 election 12 per cent of the enfranchised population changed from non-voting in 1958 to voting, while 4 per cent changed their behavior in the opposite direction. Table 16. A shows the distribution of party preferences within both of these categories, as well as among voters who participated on both occasions. Consistent voters and those who abstained in 1958 are classified according to their party choice in 1960, while non-voters in the 1960 election are classified according to the way they voted in the 1958 election. (Consistent nonvoters and respondents who did not reveal their party preferences are excluded from the analysis in this case.) In 1960 the rise in turnout produced a net increase of both bourgeois and Social Democratic voters. The fourth row of the table accounts for the party balance in this net increase. This has been computed by subtracting the number of "lost" voters from the number of "gained" voters in each of the columns in the table. The percentage calculation is based on the remainders thus obtained. Newcoming voters, of course, weigh especially heavily in this percentage distribution, since they are three times as many as those earlier participants who abstained from voting in 1960.

The parties also gained additional strength from young voters, coming of voting age in 1960. In the fifth row of the table this addition to the electorate has been included in a calculation of the party division of the aggregate accession of voters. It should be noted that this row does not strictly represent a "net effect", since the corresponding "loss" that was caused by mortality during the 1958-1960 period is not taken into account here.

Although the relevant categories are numerically small, Table 16.A indicates a tendency which is in accordance with the hypothesis (2) that has been advanced here: As would be expected, the Social Democrats — whose position was somewhat strengthened in 1960 — are favored by the aggregate effect of changes in participation among irregular voters. It is of interest to note that this effect was reinforced by the addition of first-time voters, who showed a substantial Social Democratic majority in 1960.

In the 1962 communal elections turnout decreased again, though it remained at a somewhat higher level than in 1958. The Social Democratic proportion of the vote increased rather considerably. In the interview survey sample, 8 per cent shifted from earlier participation to non-voting in 1962, while 3 per cent who had abstained in 1960 came to the polls in 1962. Since no interview data are available on party preference in the 1962 election the analysis cannot be carried out in the same way as for the 1958-1960 development. We know, however, the 1960 party prefences of those who were consistent participants in 1960—1962, and we have the same kind of information about those non-voters in 1962 who participated in the 1960 election. The first and second rows in Table 16.8 present these percentage distributions. In this case, we may consider the hypothesis (2) to be supported if the proportion who voted for the bourgeois parties in 1960 is greater among those who abstained in 1962 than among those who participated in both of the elections, since such a difference must imply that the bourgeois parties

suffered a greater loss due to the decline in turnout. A comparison between the first and second rows in Table 16.8 indicates, in fact, that there is such a tendency in the data. If we venture to make the assumption that the proportional distribution of the votes cast by earlier non-voters was the same in 1962 as in the corresponding group in the 1960 election the party division of the "net loss" can be calculated. The result is shown in the third row of Table 16.8. Finally, the fourth row accounts for the result of a calculation which — in addition — includes both the accession of first-time voters in 1962 and the loss of voters through mortality during the period, 1960—1962. Deceased persons were classified according to party choice in 1960. As regards first-time voters, we have quite simply assumed that this group had the same magnitude and partisan division in 1962 as in 1960. Even this calculation leads to a "net loss" in all the categories appearing in the table. Both calculations point to a decisively unfavorable trend for the bourgeois parties, i.e. the bourgeois share of the "net loss" is clearly greater than the percentage of bourgeois voters among those who participated in both of the elections in 1960 and 1962. Consequently, we may infer that the aggregate effect of all changes in participation contributed to the increase in the Social Democratic percentage share of votes from 1960 to 1962.

We shall now return to the 1956 election in order to include in the analysis also an occasion when the Social Democrats were the losing party. In comparison with the 1956 communal elections, as well as with the parliamentary election in 1952, the Social Democratic Party suffered a setback in 1956, while at the same time the level of turnout showed a slight rise. (See Table 1) To analyze this development we shall utilize the data provided by the 1956 interview sample survey.

For the 1956 study participation in the parliamentary election of that year has been checked in the voters' register, while the data concerning voting behavior in 1954 have been obtained through the interviews. As a result, our data give a considerable overestimation of the participation frequency in the 1954 election and hence the material is not appropriate for calculation of the "net effect" of the variability in participation during 1954—1956. Nevertheless, it is possible to ascertain the partisan division of the three relevant categories, namely, consistent participants 1954—1956, voters in 1954 who abstained in 1956, and finally nonvoters in the 1954 election who participated iri 1956. These percentage distributions are given in Table 16.C. In accordance with the procedure followed in the foregoing tables, those who participated in 1956 are classified according to their party preferences in that election. Those who voted in 1954 but failed to vote in the 1956 election are, of course, classified according to their party choice in 1954.

The percentage distributions in Table 16.C bring out a very clear tendency. Thus we find that the percentage of Social! Democrats is greatest among voters who withdrew from participation in 1956, while Social Democratic support was clearly weakest among earlier non-voters who went to the polls in 1956. Among regular participants the proportion of Social Democrats is found to be at a medium level. Because in the 1956 election setback affected both parties in the

government coalition, it is of interest to carry out the same type of calculatio for the Social Democratic Party and the Agrarian/Center Party combined. Sue a calculation shows that the coalition parties received 58 per cent of the vote cast by regular participants in the 1954—1956 elections. Among irregular votei who abstained in 1956 as much as 74 per cent were former Social Democrat and Agrarian Party adherents, while the two parties received only 48 per cer of the votes cast by those irregular voters who participated in the 1956 electioi As these figures indicate, the tendency appearing in Table 16.C remains eve if we consider the combined trend of the government parties. Our data thu suggest in a very convincing way that the electoral success of the bourgeoi opposition parties was reinforced by the type of change in voting behavior w are investigating hen

Of course, election returns have been affected by other factors, too. A further analysis of the various components of change thus shows that the Social Democratic Party received 54 per cent of the votes cast by voters coming of voting age in 1956, while the corresponding proportion went up to 67 per cent in the 1960 election. We may also look into the effect of shifts in party preferences from election to election (although it is necessary to stress again that the numerical base is quite meagre). A comparison between the interview surveys in 1956 and 1960 shows that the "election wind" may blow in different directions, indeed. However, the consequences are limited because party changers always constitute a tiny minority in the electorate. In both of the election surveys dealt with here, it turned out that only 6 per cent of the respondents reported a change in party choice since the previous election. Among those who changed between a bourgeois and a Social Democratic party preference from 1954 to 1956, 85 per cent were former Social Democrats who went over to bourgeois voting. In contrast, in the 1960 election the Social Democrats obtained 57 per cent of the votes cast by those who changed between the bourgeois parties and the Social Democratic Party from 1958 to 1960.³⁷

The point of departure for these analyses was our second hypothesis which in substance implied that ordinarily a winning party not only gains additional strength through changes in party preferences, but also derives advantage from the political net effect of participation changes among irregular voters, while a party suffering a decline in its percentage share of the vote will usually lose in both ways. As we have seen from the 1956 and 1960 data, the distribution of party preferences among first-time voters also seems to change from one election to another in accordance with the general trend in the election returns.

Perhaps it should be stated explicitly that the hypothesis — in the form presented in the present study — does not necessarily presuppose a change in the general level of turnout. If the two flows of irregular voters are different in respect of partisan division this will lead to a political net effect even in cases when the total voting rate remains unchanged. We need only to assume that a given campaign situation — because of the nature of the predominant issues — stimulates political interest within some groups of voters but has the opposite effect within other groups in order to realize that the hypothesis may well be generalized in this way. Likewise, among irregular voters the "potential" supporters of different parties may respond differently to the same campaign situation.

With regard to the division of the vote between the Social Democrats and the bourgeois parties, we have now tested the validity of the hypothesis (2) in three elections representing different combinations of changes in the size of turnout and changes in the balance of electoral strength. In all of these cases, we have found that the hypothesis has been supported by the survey data. Admittedly, our inferences are based on tendencies appearing in the form of numerically small differences in the material, and actually both of the interview samples are inappropriately small for the purpose of this part of the analysis.³⁸ However, it must be kept in mind that the alterations occuring within the electorate, too, are

characterized by small marginal shifts. And these shifts were, of course, partly
determined by other factors than variability in electoral participation.

Structural Stability in the Swedish Party System

Lastly, in this exposition of stability and change in the Swedish electorate we shall focus on the stability that characterizes the parties' anchoring in different socio-economic strata. The main features of the social structure of the party system are well established.³⁹ Election after election the Social Democratic Party has received about three fourths of its support from the working class, just as the Agrarian/Center Party has obtained the bulk of its vote from the farming population. Likewise, the Conservative Party and the (Liberal) People's Party — who have to a great extent a common social base — have consistently drawn the greater portion of their electoral support from the urban middle class. Within the "upper middle class", the Conservative Party clearly predominates.

Yet, the correspondence between social stratification and lines of political cleavage is by no means complete in the electorate. In the 1956 and 1960 elections, every fourth Social Democratic voter was recruited from the ranks of the middle class, and about the same proportion of the bourgeois vote came from the working class. The picture can, of course, be further elucidated with the aid of more elaborate stratification criteria. Thus, the division of the middle class into employers and salaried employees shows that the Social Democratic middle class support is concentrated in the latter category. Neither is the working class altogether politically homogenous. As has been indicated above, one finds a particularly strong predominance of Social Democratic voting in the category of "workers engaged in manufacturing, construction, and mining". A further specification of the relationship between social status and party preference is possible, although it will not be attempted in the present study.

Our purpose here is only to point out the stability of the partisan division of the vote within broad socio-economic strata. This particular aspect of the structure of the party system is illustrated in Table 17 which shows the distribution of votes in different social strata. The table comprises interview survey data from the election in 1956, the referendum on the pension question in 1957, and the election in 1960. Because of the relatively high degree of similarity in their social base, the Conservative Party and the People's Party are combined in the table, and on the same ground Social Democrats and Communists are put in a common category. Admittedly, the 1957 referendum was not strictly a party contest since voters were confronted with a choice between three pension schemes instead of the usual party ballots. Parties and referendum proposals were, however, so closely connected with each other that it may be deemed justifiable to include the referendum in the time-series presented by the table.⁴⁰ Consequently, we have entered the percentage of the vote for each of the referendum proposals in the appropriate party column. The period of time covered by the table is rather short but, nevertheless, our material has a considerable scope since it reflects voting behavior in three distinctively different political situations. In the first part of

the period, the Social Democratic Party was obviously weakened while the
verdict of the 1960 election undisputably was in its favor.

Stability in party support within different population groups is a prominent feature of the data presented in Table 17. In particular, this is true of the socialist share of the vote. The pattern characterizing the election data also persisted in the 1957 referendum. Almost consistently, however, the support of the Social Democratic pension proposal was somewhat weaker than the proportion of the vote that Social Democrats and Communists jointly achieved in the 1956 parliamentary election.

Changes in the balance of party strength are not altogether absent. One imoortant after-effect of the pension conflict becomes visible in the 1960 data, i.e. the increased capacity of the Center Party to attract voters from outside the agricultural population. In addition, it turns out that the farmers' support of the Center Party recovered from its bottom level at the end of the coalition period. Moreover, the losses suffered by the Conservative Party and the People's Party from 1956 to 1960 were due to a rather even decline across the whole array of population groups. Thus, there is nothing in the material to indicate that the

moderate Social Democratic increase in the 1960 election constituted a "breakthrough" into the middle class. In fact, an examination of all the changes in party support during the 1956-1960 period seems to indicate that those fluctuations that did occur in the electorate were reflected in the distributions of votes within all or at least several of the social strata. On the whole, the pattern remained unchanged in spite of such fluctuations.

---

NOTES

1 This development was broken in the 1964 elections, when the Social Democratic Party suffered a setback. The Social Democratic percentage of the vote declined to 47.3 %, while the Communist percentage increased to 5.2 %.

2 Both houses of the Parliament are elected by a method of proportional representation. In the case of the Upper House, however ,the method of election has a slight tendency to favor the largest of the parties, i.e. the Social Democrats.

3 In the 1952 and 1956 parliamentary elections, the proportion of persons voting for the People's Party reached a somewhat higher level than in 1948 — 24.4 % and 23.8 %, respectively — while the proportion was slightly lower in the 1950 and 1954 communal elections, in both cases 21.7 %.

4 Särlvik, 8., Opinionsbildningen vid folkomröstningen 1957 (Opinion Formation in the 1957 Referendum), Stockholm 1959. — Molin, 8., Tjänstepensionsfrågan. En studie i svensk partipolitik (The Supplementary Pensions Question. A Study in Swedish Party Politics. — Summary in English), Göteborg, 1965.

5 At the Institute of Political Science, University of Gothenburg, these interview sample surveys have been directed by the author of the present article and Professor Jörgen Westerståhl. As a part of this research program an interview sample survey was also conducted in connection with the 1964 parliamentary election. The 1957, 1960 and 1964 surveys were mainly financed with the aid of grants from the Government but the research program has also received generous and continuous support from the Swedish Social Science Research Council. — Reports on 1960 and 1964 surveys have been published in the official election statistics publications: 5.0.5., Allmänna val. Riksdagsmannavalen åren 1959—1960, 11, Stockholm 1961 and Riksdagsmannavalen åren 1961—1964, Stockholm 1965, II (Official Statistics of Sweden. Elections). A study of the 1957 referendum is published in Särlvik, Opinionsbildningen vid folkomröstningen 1957. Further analysis of the election surveys has been presented in Särlvik, Skiljelinjer i valmanskåren (Political Cleavages in the Electorate), Statsvetenskaplig Tidskrift 1965:2—3. See also: Westerståhl and Särlvik, Svensk valrörelse 1956 (Swedish Election Campaign 1956), Gothenburg, 1957 (mimeo).

6 Through the kind cooperation of the Swedish Institute of Opinion Research and its director, Sten Hultgren, the author had the opportunity to test a series of similar questions in an interview sample survey conducted by the Institute in the autumn of 1959. A summary of the findings is published in a press release from the Institute, dated August 9, 1960.

7 A checking of the interview answers in the voters' registers for the 1956, 1957, and 1960 samples has definitely shown that inaccurate answers about participation are heavily concentrated in the category which, according to voters registers, abstained from voting.

8 Tables showing confidence intervals for certain percentage proportions and certain group sizes in the samples can be found in the reports published in the official election statistics publications (se footnote 5). In the present study, we have not indicated confidence limits or results of significance tests. — Perhaps it should be mentioned that the sample is not representative in the same sense for the 1958 electorate as it is for the 1960 electorate. Voters who had died between 1958 and 1960 cannot be included in the 1960 sample and, furthermore, the age-limits becomes 78 instead of 80 in the case of the 1958 election.

9 A small category consisting of respondents who failed to reveal their party preferences in any of the elections was excluded from the analysis presented in Table 3. — In the table, both categories designated as consistent partisans include study of the relationship between party preference and social status can be found in: Janson, C. G., Mandattilldelning och regional röstfördelning (Method of Election and Regional Distribution of the Votes), Stockholm 1961. See also Carlsson, G., "Partiförskjutningar som tillväxtprocesser" (Party Changes as Processes of Growth), Statsvetenskaplig Tidskrift 1963:2—3.

a small group of persons who did not give complete information about their voting behavior. Thus, they have answered "cannot remember", or the equivalent, for some elections. Such incompleteness in the responses is true of 7 per cent of the total number of respondents; the proportion is about the same for all parties. — In coding interview responses concerning participation in the 1956 and 1958 parliamentary elections we have considerered the following two categories as "probable non-voters": (a) respondents who explicitly said that they did not vote, (b) respondents who "did not know" whether they voted. By using this definition, 12 per cent of the sample were classified as non-voters in 1956, and 17 per cent as non-voters in 1958. This means, in fact, an overestimation of the level of turnout.

10 A detailed account of changes in voting behavior from 1958 to 1960 can be found
in the election statistics publications (see footnote 5). Riksdagmannavalen åren 1959—
1960, 11, pp. 58—59. See also: Riksdagsmannavalen 1961—1964, 11, p. 99.

11 These percentages are basedon numerically small categories in the sample. 80 respondents changed between bourgeois parties, and 57 % of this category shifted between the Conservative Party and the People's Party. 74 respondents changed between a bourgeois party and the Social Democratic Party.

12 Table 4 is basedon the whole series of questions concerning voting behavior in
past elections. A corresponding — but somewhat less detailed — series of questions
was included in the 1956 questionnaire.

13 In a series of studies, Campbell et al. have utilized the concept of "party identification" to designate the voter's affective orientation toward his party. Campbell et al. have shown that both the direction and the strength of the voter's partisan orientation exercise a profound influence on his perception of political issues and his voting behavior. See: Campbell, A. et al. The Voter Decides. Evanston 1954, pp. 88—111, and Campbell et al., The American Voter. 1960, pp. 120 ff. — For an application of this concept on Swedish data, see Särlvik, 8., The Role of Party Identification in Voters' Perception of Political Issues. A Study of Opinion Formation in Swedish Politics 1956—1960. (Paper prepared for the Fifth World Congress of the International Political Science Association, Paris 1961, mimeo.) — Särlvik, "Skiljelinjer i valmanskåren", Statsvetenskaplig Tidskrift 1965:2—3.

14 In the 1960 election survey, the interview question concerning "conviction" in party affiliation was administered to one half of the sample before and to the other half after the election. In the 1956 study the question was asked in an interview before the election. In the present analysis, all those respondents who changed their voting intentions during the campaign have been classified as "not convinced", irrespective of their answers to the interview question. — In 1960 the proportion of "convinced supporters" rose from the first interview to the second among supporters of all parties except the Conservative Party; among Conservative voters the trend was in the opposite direction. The author plans to discuss in another study the relationships between party identification and other attitudes toward different asoects of the political system.

15 Both of the categories which have been excluded have a comparatively low voting rate. There is an additional "loss" in the interview sample due to persons who could not be interviewed. This "sample-loss" also includes a great proportion of non-voters. Cf. Riksdagsmannavalen 1959—1960, 11, p 51 f. Partly because of the exclusion of these categories, the voting rate has become somewhat higher in the interview material than in the entire electorate.

16 In the socio-economic classification an office memorandum prepared by the Swedish Central Bureau of Statistics was utilized. This memorandum contains a classification code which was used in the official election statistics up to the 1960 elections. In the present study we have made a number of modifications in this classification scheme, inter alia the division of the working class which appears in the table. All the groups in the table have been obtained by combining several subcategories in the original classification. — Concerning the simple division into a middle class and a working class, cf. Bonham, J., The Middle Class Vote. London, 1954, pp. 52-53.

17 A summary of research results from various countries is given in: Lipset, S. M., Political Man. The Social Bases of Politics. 1960. See also: Duverger, M., La Participation des femmes a la vie politique. Paris, 1955. Among earlier studies in this field we may particularly mention Tingsten, H., Political Behavior. Studies in Election Statistics. Stockholm, 1937.

18 Tingsten, op.cit., p. 230. According to Tingsten, "the law of dispersion" is valid not only for socio-economic group differences but also for other groupings which include categories with different voting rates, e.g. sex and age groups. Cf. Hastad, E.: "Några sifferreflexioner kring valet" (Some quantitative observations concerning the election), Statsvetenskaplig Tidskrift 1936, pp. 340—363.

19 The 1948 election provides another example. In this election, there was a strong increase in turnout but the increase was somewhat greater — for men as well as women — in the middle class than in the working class (i.e. social class II and 111, respectively, in the election statistics). Cf. Riksdagsmannavalen 1945—1948, p. 67.

20 Cf. Särlvik, Opinions bildningen vid folkomröstningen 1957, p. 27. In the 1956 election study, the voting rate was found to be 87 % among "workers engaged in manufacturing, construction, and mining", and 80 % among "other workers". In the "lower middle class" (small businessmen and various white collar groups), the voting rate was 87 %. Westerståhl and Särlvik, Svensk valrörelse 1956.

As an example of underlying social factors, it may be mentioned that the "other workers" category contains a relatively great proportion of persons with low family income. Participation increases with increased income level among workers as well as among other occupational groups. Moreover, the bourgeois parties have a somewhat greater support among workers with low income than among those with high income, cf. Riksdagsmannavalen 1959—1960, 11, pp. 57 and 84; Riksdagsmannavalen 1961—1964, 11, p. 93.

21 Cf. Riksdagsmannavalen 1959—1960, 11, pp. 80 and 84.
22 Cf. Riksdagsmannavalen 1959—1960, 11, pp. 80—84.

23 Data concerning residential mobility during the period 1946—1960 were obtained from the voters' registers which are basedon the population registers. Because such information was not readily available in the voters' registers for the city of Stockholm, this part of the sample has been excluded from the analysis presented in the table. Only migrations involving a change of parishare considered in this case. The year of migration indicated in the table refer to the first year the individual was registered in the parish he lived in according to the 1960 voters' registers. Information on later migrations was obtained from population registers in connection with a checking of the voters' register for the 1962 communal elections. For 1960—1962, it was necessary to treat each change of address as a migration, although it may not always have involved a change from one parish to another.

24 An example of how even rather insignificant obstacles to electoral participation may reduce the willingness to vote is given in: Faul, E. (ed.), Wahlen under Wähler in Westdeutschland, Villingen Schwarzwald 1960. The author shows that there is a correlation between voting rate and rather small differences in distances to the polling places.

25 The concept of "cross pressure" is introduced in Lazarsfeld, P. F. et al., The People's Choice, 2nd ed. 1952. A more complete theoretical treatment appears in: Lane, R. E., Political Life. Why People Get Involved in Politics, 1959, pp. 197 ff. Himmelstrand, has undertaken a detailed analysis of the relationships between different types of cross pressure situations, personality factors and participation; cf. Himmelstrand, U., Social Pressures, Attitudes and Democratic Processes (see pp. 303 ff). Uppsala, 1960. See also Särlvik, Opinionsbildningen vid folkomröstningen 1957, p. 56. For a study of electoral participation in Finland, see: Allardt, E., Social Struktur och politisk aktivitet. En studie av väljaraktiviteten vid riksdagsvalen i Finland. Helsinki 1956. Tingsten's "law of the social centre of gravity" may be considered as a specific formulation of the cross pressure theory.

26 Glaser, W. A.: "Fluctuations in Turnout", in McPhee, W. N. and Glaser, (ed.), Public Opinion and Congressional Elections. 1962. — Campbell, "Surge and Decline. A Study of Electoral Change", Public Opinion Quarterly XXIV (1960), pp. 397-418. — It should be noted that Glaser's theory implies a further development of Tingsten's "law of dispersion". —

"Concerning relationships between political attitudes and electoral participation, see Campbell et al., The American Voter, pp. 96 ff. Glaser arrives — partly in the form of hypotheses — at the same conclusions as Campbell, see: Public Opinion and Congressional Elections, pp. 44 ff.

28 A theory of this nature has earlier been presented in: Westerståhl and Särlvik,
Svensk valrörelse 1954. Icke-röstning. Mediastudier (Swedish Election Campaign, 1954.
Non-voting. Mass Media). Mimeo, Gothenburg, 1956. (See pp. 2—10).

29 For each of the four variables, criteria of "high" and "low" exposure were defined. The "index of campaign exposure" is based an a classification according to the number of such criteria of "high exposure" which respondents fulfilled. The methodological procedure will be discussed by the author in detail in another connection.

30 This attitude measurement has been obtained by combining two basic attitude indexes. One of these indexes measures attitude toward social welfare policy (along a left-to-right continuum), while the other is a measure of approval or distrust of government policy in four vital areas (housing, employment, etc.) See: Särlvik, "Skiljelinjer i valmanskåren", Statsvetenskaplig Tidskrift, 1965:2-3.

31 In the strict sense of the concept, the criteria utilized in the table — i.e. social class and party preference — do not define a cross pressure situation, because they are not sufficient for determining whether the individual has been really subjected to conflicting group pressures. However, our hypothesis presupposes only that the probability of cross pressure becomes greater under the circumstances stated above. — In the table, voters in the 1960 election have been divided according to their party preference in that election, while non-voters have been classified according to their party choice in the 1958 communal election. Communist voters, respondents who failed to reveal their party preferences, first time voters of 1960, and those persons in the sample who died before the 1962 election have been excluded from the analysis.

32 Campbell et al. define cross pressure exclusively in terms of attitude conflict.
See: The American Voter, pp. 81—88, and The Voter Decides, pp. 157 ff and 183.

33 This hypothesis can be deduced from Glaser's theory. See: Public Opinion and
Congressional Elections, p. 50.

34 In this case, the analysis includes only those respondents who had reached the voting age in or before 1954. This means that we have included only voters who have had a real chance to show variability in both partisanship and participation over a longer period of time. — Respondents who gave no indication of their party affiliation and those who had never voted have been excluded. The number of respondents in the "consistent partisans" category is 818, while the number of "party changers" is 335.

35 This analysis may be compared with the results of a local interview study conducted in Gothenburg during the 1954 election campaign. A report on the study is found in: Westerståhl and Särlvik, Svensk valrörelse 1954. Två lokala studier (Swedish Election Campaign 195. Two Local Studies), Gothenburg 1955 (mimeo.). — In this study, it was found that the following three criteria correlated with both variability in partisanship and non-voting in the 1954 election: 1) migration since 1952, 2) deviating party preference or voting intention as compared to the political opinions in the voter's family or among his closest friends, 3) the belief that a change of government would not bring forth a great political change. — The study defined a category of "politically mobile voters" which included party changers since the previous election and respondents who intended to change their party affiliation but returned to their earlier party preferences during the campaign or ended up as nonvoters. (Respondents were interviewedon three occasions.) If "politically mobile" respondents and non-voters are combined, we find that 58 % of this category met at least two of the criteria stated above. Among "politically mobile" only, the proportion who met at least two criteria was also 58 %. The corresponding figure for those who voted in the election and had a stable party prefenrence was 35 per cent.

36 The same pattern appears if the respondents' self-evaluation of their political interest is utilized as a measure of political involvement (cf. Table 13). Among voters who have always voted for the same party and indicated themselves to be "very" or "rather" interested in politics, 69 % had a "high" propensity to vote according to the criteria stated above, while this proportion was 5,0 % among non-interested party changers.

37 All of these percentage are basedon very small groups in the samples. First time voters in 1956: n = 35; first time voters in 1960: n = 42; voters changing between the bourgeois parties and the Social Democratic Party 1954—1956: n = 20; the corresponding party changers 1958—1960: n = 28.

38 The probability that a hypothesis which has been deduced from a strict theory would be confirmed by "chance" must, of course, decrease decisively if we demand that is must be supported by three different tests and by two different probability samples. (Cf. Almond, G. A. and Verba, S. The Civic Culture. Political Attitudes and Democracy in Five Nations. 1963, pp. 253 ff.) — However it ought to be mentioned here that the 1964 election in Sweden seems to represent a deviant development. In this case, the Social Democratic Party showed a decline as compared to the 1962 parliamentary election in spite of the fact that it was favored by the net increase in turnout from 1962 to 1964. Instead, there seems to have been a heavy decrease in the Social Democratic support among young voters. Turnout among young voters declined from 1960 to 1964 and — in accordance with the findings of the present study — the Social Democratic percentage of the votes cast by first time voters decreased from 67 °'„ ih 1960 to 50 °/₀ in 1964. The author intends to undertake a more detailed analysis of this development in a special study.

39 A detailed account of party division among various population groups can be found in the reports for the official election statistics publications which have been prepared by the author of this article. See Riksdagsmannavalen åren 1959—1960, pp. 48 ff and Riksdagsmannanvalen åren 1961—1964, pp. 93 ff. It should be noted, however, that the occupational classification utilized in these reports is not entirely indentical with the classification appearing in the present article. — An extensive ecological

---

9 A small category consisting of respondents who failed to reveal their party preferences in any of the elections was excluded from the analysis presented in Table 3. — In the table, both categories designated as consistent partisans include study of the relationship between party preference and social status can be found in: Janson, C. G., Mandattilldelning och regional röstfördelning (Method of Election and Regional Distribution of the Votes), Stockholm 1961. See also Carlsson, G., "Partiförskjutningar som tillväxtprocesser" (Party Changes as Processes of Growth), Statsvetenskaplig Tidskrift 1963:2—3.

40 Among those who voted in the referendum, 79 % voted in accordance with their party preferences in tjhe previous election. See: Särlvik, Opinionsbildningen vid folkomröstningen 1957, pp. 44 ff.