University of Goteborg

In the fall of 1966, after the local government elections, it was a widely accepted expectation that the long era of Social Democratic government in Sweden was bound to be brought to an end at the parliamentary election in 1968. Local elections usually reflect trends in national politics, and at that time few would be willing to believe that the Social Democrats could manage to recover from the 1966 defeat and even to win one of its most convincing electoral victories only two years later. In 1966 the Social Democrats found their share of the vote reduced to 42 °/o which was less than the party had received in any election through the past three decades. To the left, the Communist party seemed increasingly succesful in its attempts to widen its base by establishing itself as a democratic left-socialist opposition, freed from its previous linkage to Soviet Communism. At the same time, the non-socialist parties (or "bourgeois" parties as they are more often labeled in Swedish political vocabulary) had attained a voting strength that would have given them a majority in the Second Chamber of the Riksdag in a parliamentary election.

In retrospect it is obvious, of course, that the 1966 election verdict did not foreshadow a durable weakening of the Social Democrats' electoral support. Although we have not yet the benefit of hindsight, there are good reasons to conjecture that the 1968 election outcome, too, should be interpreted as a transient deviation from a persisting "normal" balance of strength in the electorate.

---

* The nationwide interview surveys reportedon in this article were financially supported by Governmental grants and were conducted as a part of the election statistics program of the Swedish Central Bureau of Statistics. The Survey Research Institute of the Central Bureau of Statistics has been in charge of the field work administration and the sampling, while general planning, questionnaire construction and data analysis have been carried out as research work at the Institute of Political Science, University of Goteborg. For this research work, the author has received grants from the Swedish Council for Social Science Research and from The Bank of Sweden Tercentenary Fund. A major portion of the data material presented here also appears in a report that has been prepared for the CBS official election statistics monograph. See: "Valet 1968. En redogorelse for statistiska centralbyråns och statsvetenskapliga institutionens vid Goteborgs universitets intervjuundersokning vid andrakammarvalet 1968". Prepared by Bo Sarlvik in collaboration with research associates in the Swedish Representative Democracy Project. (Mimeo.; the report will appear in print in 1970.)

In 1968, however, the deviation favored the Social Democrats. In the first place, the tremendous enlargement of the turnout yielded a. sizable increment to the Social Democratic voting support. Furthermore, because of the Czechoslovakian crisis the 1968 election campaign was pursued under quite peculiar circumstances. In spite of the fact that the Swedish Communists vehemently repudiated the Warsaw Pact intervention, the party had to face the fact that in this extreme situation its own standing was severely impaired by the widespread agony and distrust that became aroused against international Communism. The party suffered a loss that amounted to about half of its voting strength in 1966. The bulk of this vote went to the Social Democrats. Both of these components of the Social Democratic gain in 1968 are likely to be particularly uncertain assets for subsequent elections.

If the Czechoslovakian crisis at all contributed to the simultaneous weakening of the bourgeois parties, the effect must have been more indirect than in the Communist case. One may speculate about the possibility that anxiety over the international situation led previous bourgeois voters to cast their vote for the established governmental party. However, we have found it difficult to find tangible support in interview survey data for such an inference. Nevertheless, it seems plausible that the Social Democrats were favored in the 1968 election by the mere fact that the events on the international arena detracted attention from the non-socialist parties' campaign against the Government's domestic policy record.

While both of these recent elections could be classified as "deviating", the party division in 1964 was quite close to the "average outcome" of previous elections.¹ We shall therefore treat the 1964 vote as a baseline in much of the following inquiry. Using this baseline, we shall attempt to appraise the significance of the electoral shifts that occurred in the ensuing four year period.

The data base for this study has been obtained through a nationwide interview panel survey that has been conducted by a research project directed by the author at the Institute of Political Science of the University of Goteborg in collaboration with the Swedish Central Bureau of Statistics. The panel study is comprised of two interview sample surveys which were conducted in connection with the parliamentary elections in 1964 and 1968. In addition to the panel sample, the 1968 survey also includes a supplementary sample which consists of respondents that reached the voting age after the 1964 election. For the entire sample, voters' register data pertaining to electoral participation have been collected for the four elections held in 1962, 1964, 1966 and 1968. (It may also be mentioned that a third panel wave to be completed immediately after the 1970 election is now in preparation.)

The first of the following sections will illuminate the exchange of voters among the parties as well as the shifts between voting and non-voting and the changes in the composition of the electorate which brought about an alteration in the party division of the vote between the two parliamentary elections in 1964 and in 1968. Electoral change will then be put into a micro-macropolitical

perspective. Through this part of the analysis we shall attempt to inquire into
the relationships between the partisan change flows and the structural properties
of the party system.

The data material requires some further comments, however, before the analysis results are presented. The actual election returns as well as the distributions of seats in the Second Chamber of the Riksdag have been summarized in Tables 1—2. The corresponding distributions of the party vote in 1964 and in 1968 in the interview sample are shown in Table 3. As can be seen from these tables, the Social Democratic proportion is somewhat oversized in the interview sample while Conservative and Communist voters are somewhat under-represented. It should be noted, however, that the Social Democratic and Communist vote is almost exactly the same as in the electorate as a whole for both elections. As we will be concerned, primarily, with electoral trends it is also reassuring that the changes that occurred in the electorate from 1964 to 1968 are mirrored almost exactly in the sample. As regards the response rate, it may be noted that 92 °/o of the sample was interviewed in 1964. In 1968, interviews were conducted with 88 °/o of the entire survey sample (including the supplementary sample of young voters as well as respondents who belonged to the sample in 1964 but were not interviewed on that occasion). Among those who had been interviewed in 1964 (excluding deceased persons and a few cases of disfranchisement), 91 °/o were also interviewed in 1968. Within the panel sample, 95 °/o of the respondents were interviewed in at least one of the two surveys.

Losses and Gains in Voting Support

As the two interview surveys were conducted at the parliamentary elections, most of the analysis in the present article will bear upon voting behavior in these elections and the changes in the composition of the electorate that occurred between them. By utilizing participation data to fill in the missing link, however, we shall also be able to gain insights about the flows of the vote that led to a sharp deviation from the "average vote" in the 1966 local government election.

In the national electorate, the Social Democrats and the Center party increased their shares of the vote by 2.8 and 2.7 percentage units, respectively, from 1964 to 1968. The People's party and the Communist party suffered corresponding losses of 2.1 and 2.2 percentage units, respectively, while the strength of the Conservative party remained almost exactly the same in 1968 as in 1964.

Table 4 throws light on the changes in individual voting behavior and eligibility to vote which are reflected in these aggregate figures by presenting the survey data in the form of a transition matrix. Then, the direction of the percentage calculation is reversed in Table 5 so as to show how the 1968 voting support of each of the parties as well as the category of non-voters in that election were composed with regard to the individuals' party choice and electoral participation in 1964. A comprehensive overview of the electorate at the two elections is given through the complete cross tabulation in Reference Table I (at the end of this article, p. 277), where all the cell entries are percentages of the grand total for the entire table, i.e. the number of respondents who were eligible to vote in at least one of the elections. It should be noted, however, that no "not ascertained" categories appear in this set of tables, because we have excluded from this part of the analysis all respondents who failed to give information about party choice in an election in which they actually participated according to the voters' registers. On the other hand, these tables include all persons in the entire sample (even non-interviewed) for whom interview data, register data, or a combination of both provide sufficient information about voting behavior and eligibility to vote. (The register data are complete.) Because of the exclusion of missing data cases, the proportion of Social Democratic voters has become somewhat larger in these tables than in Table 3. Nevertheless, the trends in the electorate are reflected correctly in the "change tables" and the data material also shows a quite close congruence to the electorate with regard to the percentage changes in the parties' electoral strength.

The scope of the changes that occurred in the 1964—1968 electorate is illuminated by Table 6. The leftmost column of the table summarizes the entire data material in Reference Table I, while the two remaining percent distributions pertain to those who were eligible to vote in both of the elections and to the 1968 electorate, respectively. As can be seen from the table, 68 °/o of the citizens who were able to vote in both elections actually did so arid supported the same party on both occasions. As usual in Swedish elections, a substantial part of the party change actually involved changes between different opposition parties.

Thus, about half of the 14 °/o that shifted party choice are found to have
switched between bourgeois opposition parties.2

As should be expected, the two winning parties, the Social Democrats and the Center party, maintained particularly large proportions of their 1964 support in 1968 (cf. Table 4). The Conservatives lost about a fifth and the People's party more than a third of their 1964 voters through partisan change. Table 5 completes the picture by showing that the "feeder flow" of party changers makes up about a fourth of the Center party's total voting support in 1968, while the corresponding proportion for Social Democrats is only about 6 %. As a matter of course, this discrepancy is "explainable" as consequence of the fact that the relative size of the Center party gain was much larger than that of the Social Democratic party although both parties got about the same increase in percentage units. Yet, the phenomenon is clearly noteworthy since the degree of stability that obtains in the electoral basis of a political party may well have political implications even if it is "expected" in a statistical sense.

One may distinguish between three components of electoral strength which jointly affect the voting strength of the parties, namely: (1) change in party choice among consistent participants, (2) shifts between voting and non-voting, and (3) changes in the composition of the electorate because of the inflow of first-time voters and the outflow that is due to mortality, etc. The net effects of these components are displayed in Table 7. Just as in Reference Table I all entries are calculated as percentages of the total number of respondents who were eligible to vote in at least one of the two elections (excluding the "missing data" cases defined above), but in Table 7 entries indicate differences between gains and losses. The sum of all these net effects is positive because there was a larger proportion that participated in the 1968 election.

The difference between the Center party and the Social Democratic party that was pointed out above becomes visible also in these summary indexes of net effects. As can be seen from the table, the Center party derived about half of its total net increase from partisan change, while the major part of the Social Democratic increment had its origin in the two other change components. If one thinks of the net effect of shifts between voting and non-voting and changes in the composition of the electorate as constituting a "synthetic" category of voters, he can calculate how the 1968 vote was divided among the parties within that category. It is then found that the Social Democrats obtained 61 °/o of the vote in this synthetic category. This should be compared to the distribution of votes among the consistent participants who did not change between Social Democratic and non-Social Democratic voting; i.e. those who either voted Social Democratic in both elections, or voted for one or two of the other parties in both elections. Within this category, the Social Democrats obtained 52 % of the vote. (It should be kept in mind that the Social Democratic proportion is somewhat too large in our sample, however.)

When making this kind of comparisons, one must of course take note of the fact that the party change component carries more weight than the others upon the percent distribution of the vote among the parties; this is so for the simple reason that every party change has a "double effect" inasmuch as it reduces the strength of one party at the same time as it increases the strength of another. It is also clear that both measurement errors (especially the under-representation of Communist voters in the sample) and sampling errors are present; these calculations must therefore have the nature of somewhat uncertain approximations. On the other hand, we have very reliable measures of some major change factors (e.g. participation, mortality and the size of the first-time vote) and for that reason one would actually have to assume that the sample departs quite wildly from the population in regard to some other components if the population parameters were to differ very much from the picture presented by the sample.

The finding that the increase in participation and changes in the composition of the electorate favored the Social Democratic party in the 1968 election can be traced through all the relevant subcategories in the sample. As is shown in Table 6 where the Social Democratic proportions in 1964 and 1968, respectively, in the "inflow" and "outflow" categories are specified in the rightmost columns, the voters who entered the active electorate in 1968 were more Social Democratic than those who quit. The effect of shifts between voting and non-voting was also amplified by the fact that the "inflow" was more than double as large as the "outflow". At the same time, it can be seen from the top section of these columns that the overall trend in partisan change was favorable to the Social Democrats.

Our inference about the impact of the large turnout in the 1968 election obtains further support through the data in Table 8, where the 1968 voters are classified with regard to eligibility to vote and regularity of participation in the three previous elections. We find, again, that the Social Democratic party received

a particularly large support within categories of voters showing a comparatively low propensity to vote: that is, among older voters who had abstained in two or more of the previous elections (Section A), among young voters entering the electorate as "delayed" first-time voters in 1968 (Section B), and within the comparatively large group of non-_voters in the 1966 local government election (Section C).

Before concluding this section, it may be in order to stress that the differential effect of participation changes that has been displayed in the foregoing tables is not an expression of any strong overall relation between partisanship and regularity in voting behavior. Indeed, the opposite is actually true. There appear to be only slight differences among parties when the 1968 major party voters in the age groups that were entitled to vote in the last previous three elections are compared with regard to their regularity in electoral participation (see Table 9). Actually, the comparisons made in the "change tables" have directed attention to differences that appeared in 1968 among changeable voters. But in the total electorate, these categories do not carry too much weight as compared to the much larger part of the citizenry that participates consistently in election after election. Furthermore it should be noted (because the opposite is often assumed) that in recent elections there does not actually appear any general tendency to the effect that the Social Democratic vote should increase when the turnout is heightened and then go down when the turnout is lowered. As a matter of fact, the Social Democrats received their largest voting support in the two elections during the period which differ most widely with regard to turnout, that is, in the 1962 local government election when the turnout was very low and in the 1968 parliamentary election when it was extremely massive.

Voting within Social Strata 1964-1968

The social bases of Swedish parties form a pattern that has hardly undergone any profound change in regard to its main features since the 1920'5. An overview of this party division of the vote within social strata as it appeared in the 1968 election is provided by the data in Reference Table lI.A-B (pp. 278—79) where the social composition of the major parties is also displayed.³

When comparing the 1968 voting of social groups with previously established alignments, one encounters, however, one exception from this structural stability which has become so marked that it may be taken to signify a real modification of the party system. This has to do with the electoral base of the Center party. Having existed through more than three decades as an almost exclusively agrarian party, the Center party commenced in the late 1950's (during the pension reform controversy) a strategy of appealing to a much wider electorate. Although the transformation of the party has not yet permeated its organizational and parliamentary leadership level, the Center party has now established itself in a position where it enlists substantial voting support from outside the farming population.⁴ In particular, it has been a growth within the middle class strata at the expense of the People's party.

The change in the composition of the Center party vote is evidenced by the data displayed in Table 10 where the 1956 election has been included in order to provide the baseline. In 1968 the farmers' votes make up less than a third of the total Center party vote. Yet, the party has retained its predominance within the farming population as can be seen from Reference Table lI.A. Among occupationally active farmers and their wives the Center Party proportion of the vote even amounts to 70 °/o (this definition of the farmers' category is somewhat more restricted than the one applied in the tables).

At the same time as it has expanded its support, the Center party has thus retained its anchorage in the countryside population. This is also reflected in Table 11 which shows that the party's support is weakest in highly urbanized communities. Actually, the Center party obtains less than a third of its votes from communities where more than 90 °/o of the population lives in a built-up area whereas the corresponding proportion for the People's party and the Social Democrats is about two thirds (cf. Reference Table III). As becomes apparent in the bottom section of Table 10 the explanation is to be found in the fact that the Center party has won its strongest support outside of the farming populaion within the least urbanized areas. That is, the Center party share of the vote increases with decreasing degree of urbanization even if farmers are excluded from the percentage calculation.

The political significance of the displacement of the Center party basis in the electorate is bound up with recent change trends in the Swedish economy. These have led to an intensive economic growth within certain highly urbanized and industrialized centers, whereas there are clear signs of economic stagnation in rural and small-town areas in the country. The resulting discontent within geographically peripheral population groups has obviously become voiced through the new support for the Center party.

In the introduction the 1966 election was classed with regard to the electoral strength of the Social Democratic party as a "temporary deviation" from the more "average" party division in 1964. When looking at the 1968 voting within social groups one finds grounds for that judgement. The change from 1964 to 1968 in the Social Democratic share of the vote comes forth as a fairly even increment due to some increase at the expense of the Communists in the working

class and some gain from the bourgeois parties among enterprisers and white collar employees. (Because of the difficulty of finding differences that could survive significance tests, we shall not attempt to specify any further differentiation in this regard.) The farming population is really the single exception in the entire array of socio-economic groups from the general — but moderate — pro-Social Democratic trend. Wherever the 1966 losses may have occurred, they were apparently recovered in 1968. Indeed there seems to be no segment of the electorate (with the possible exception for very rural areas) where the loss of confidence in 1966 persisted so as to depress the Social Democratic support below its 1964 level.

Strength of Party Affiliation

The tendency of the Swedish multiparty system to function as a two-bloc system provides a framework for partisan competition which appears to bring with it a high degree of instability within the voting support of each of the parties in the non-socialist bloc as an almost unavoidable consequence. In a sense, the difference between the Social Democrats and the non-socialist parties in regard to turnover rates can be seen as a sheer size effect. As long as the Social Democratic party maintains a heavy predominance within one of the blocs, while the smaller parties in the other bloc are involved in intra-bloc competition for votes, each of the latter must show a relatively high turnover rate. Since party strength is a result rather than a cause of voting, "size effects" of this kind can hardly be invoked as an explanation of individual behavior, however. In order to account for the peculiar shifts among the bourgeois parties that have occurred throughout the post-war period, we shall instead inquire into the voters' attitudinal orientations towards the parties.

In a previous study, focusing on electoral behavior during the 1956—1960 period, the author has shown that voters supporting a bourgeois party were much less likely than Social Democrats to think of themselves as "convinced adherents" of the party of their choice.⁵ Furthermore, it was shown that voters having a weak sense of partisan commitment are also most likely to sway from one party to another. These interrelationships between electoral party choice, strength of "convincedness" in party support, and stability in voting behavior come forth in a quite striking way also in our data from the 1964—1968 panel survey. In Tables 12—13, we have taken the voter's party choice and the strength of his party affiliation at the 1964 election as an observation point from which we can obtain a retrospective as well as a prospective view of his voting behavior over time. The measure of strength in party affiliation applied in these tables is provided by an interview question in which respondents were asked whether they thought of themselves as "strongly convinced adherents" of the party they intended to vote for or actually voted for.¹⁵

Our earlier finding that bourgeois voters — in any given election — are
much more likely than Social Democrats to have switched from recent support
of another party obtains a firm corroboration in these data. It is also clear that

there is an extremely high degree of turnover within the Communist voting support. The Communist "hard core" of stable supporters through a sequence of elections is, indeed, so tiny that it becomes barely visible in a sample of the size we are working with.

We furthermore find that inter-party differences in regard to voting stability are fairly moderate among "convinced" party adherents (with exception for Communists), while a quite drastic discrepancy appears when one compares voters with weaker feelings of party affiliation. (See the middle section of Table 12.) It is also noteworthy that the difference between "strong" and "weak" adherents is comparatively insignificant for the Social Democratic party which is, of course, not affected by the changes that occur among bourgeois parties. About half of the party changers had shifted between bourgeois parties, only. The bottom row of Table 12 has finally been included in this exposition of retrospective data only to show that the overall party differences that have been

noted above persist even if one takes into account the individuals' entire voting
records as these are reflected in our data.

In Table 13 the time perspective is reversed so that we are now taking the 1964 election as a baseline for a look ahead to voting in 1968. As is seen from the top row of the table, the proportion of Social Democratic voters who considered themselves as "strongly convinced party adherents" is a great deal larger than for any other party. It might be added that Social Democrats were more confident, too, than the opposition voters about the outcome of the election. In a 1964 pre-election interview, 46 °/o of the Social Democrats declared that they considered a change of government as a "not particularly likely" possibility and an equally large proportion thought it was "not at all likely". To a remarkably large extent, bourgeois voters shared that expectation: 60 °/o considered a governmental change as "not particularly likely" and 26 °/o believed it was "not at all likely".⁷

Four years later, in 1968, the story told by Table 12 was recapitulated. About 93 % of the 1964 nonsocialist voters who participated in both elections reaffirmed their non-socialist outlook in 1968, but about a quarter (24 %) of these consistently bourgeois voters switched to another opposition party. And again, the voters who had been lukewarm in their party support in the previous election were, generally, most likely to change. The relevant data are given in the lower rows of Table 13. As observed in a previous section, the Center

party differs somewhat from the other non-socialist parties in that it lost a smaller proportion of its former supporters than did the other two. This is obviously an expression of the growth process that the Center party is undergoing at present. More remarkable — because it has no counterpart in any of our previous election studies — is the trend in the People's party. As is seen from the table, the landslide away from the party had almost the same magnitude among its once "strongly convinced" adherents as among its less wholehearted supporters. One is led to the inference that the migration of voters from the People's party to the Center party must have cut deeper into the losing party's electoral base than similar moves in the past.

In the 1968 survey, our measure of strength in party affiliation has been redevised so that it can be considered as an adaptation to the Swedish multiparty system of the party identification measure which has been employed in the writings of Angus Campbell and his research associates. Hence, we are now able to distinguish, analytically, between actual voting behavior and feelings of personal affiliation to a political party. (See the technical description in the note, below.⁸) The complete classification gives for each party a three-point gradation ranging from "strong identification" (strongly convinced adherents), over "weak identification" (party adherent but not strongly convinced) and to "party preference" (no party adherent, but feels "closer to" some party than to the others). The last of these categories corresponds to the category of "independents leaning to a party" in the American classification. In addition, there are, of course, respondents lacking (or not confessing) any sense of party affiliation, i.e. the "no party identification" category.

The relationship between party vote and strength of party affiliation as it comes forth in the 1968 election is displayed in Table 14. It should be noted, however, that the party identification measure is applied in a somewhat compressed form in the table, since all voters who were "adherents of" or "preferred" another party than the one they actually voted for have been put in one joint

category, labeled "affiliation to another party". Although this classification is not entirely identical to the one employed in the foregoing tables, it is clear that it measures essentially the same psychological phenomenon. We can also see that the by now well known differentiation among the parties has persisted. In Table 15 we can then compare consistent party supporters with party changers. As expected, the parties' new supporters tend in general to cast their votes with a weaker sense of partisan commitment than the consistent voters. Again, the People's party is an exception, however, in that its stable voters are almost as loosely attached to the party as are the party changers. This is, indeed, both an unexpected and an unprecedented finding. (In the 1964 survey, e.g., 38 °/o of stable People's party voters were "convinced" party adherents as compared to 11 %of those who had voted for another party in a recent election.)

All in all, the data explored here provide abundant evidence to support the conclusion that the vicissitudes in the bourgeois parties' voting support are bound up with the electoral behavior of a sizable mass of voters who are forming only weak and transient feelings of attachment to any particular non-socialist party, although they may be consistent in rejecting the Social Democrats.

We have already implied the presumption that intra-bloc party competition is in itself a major cause of weak partisanship among bourgeois voters.⁹ In the present article, we shall not go much beyond that general presumption. Two contributing features of the party system deserve to be mentioned, however, since they bear directly upon stability in partisanship. In the first place, the tight relationship that exists between the Social Democratic party and the trade unions as well as other "labor movement" organizations brings with it a whole network of supportive group affiliations that may help to explain why Social Democrats more often than others show a strong sense of personal affiliation to their party. If this is true, on the other hand, one should expect that the Center party's equally close links to the farmers' organizations would have similar consequences. Or more precisely, this kind of strong partisanship should appear within the traditional agrarian part of the Center party's electoral base. As is shown by Table 16, this is, indeed, the case. Weak partisanship is a characteristic only of the party's newly acquired non-agrarian support.

The same differentiation within the Center party actually appears in regard to voting stability. We have divided the 1964 Center party voters who also participated in the 1968 election into two major groups: those who were farmers through the entire period and those who had another occupation on both occasions (thus excluding a small category who changed occupations). In the farmers' category only 8 °/o switched to another party in 1968, while the proportion of defectors amounted to 28 % in the other category.

One may furthermore interpret inter-party differentiation in regard to party identification in the light of a theory on partisanship propounded by Angus Campbell.10 From general attitude theory, Campbell derives the proposition that in a party system wherein parties are arrayed along a predominant ideological

axis one should expect, in general, that the individual's feeling of political engagement would be dependent on the degree of extremeness of his party position. The relationship between intensity in political engagement and party choice would then obtain the U-shape that has become well known through the Guttman scaling theory, so that centrist party voters would show a weaker sense of engagement than the supporters of parties located towards the two extreme points of the party dimension. The data in Table 14 lend partial support to this proposition in that it is found that Conservative voters tend to have stronger emotional ties to their party than do the supporters of the two centrist parties. The data for the Social Democrats also fit in with the proposition if one conceives that the Social Democratic party has its center of gravity definitely to the left of the center. The Communists present a clearly deviant case, however.

As the author has shown in a previous study, the functional relationship between political engagement and partisanship actually comes forth with much more clarity (in the Swedish case) if one restates the proposition referred to above, so as to let it pertain to extremeness of partisanship within the framework of a "socialist—non-socialist" two-bloc system.¹¹ One of the analytic tools used for this purpose was based on direction of strength of party identification (toward a party in one of the blocs) in combination with a measure of "negative identification" towards the opposite bloc. A compressed version of this classification — being based on party identification only — is shown in Table 17. In the same table, we have also included a behavioral expression of political involvement, namely, electoral participation. The relationship appearing in the table fits admirably well with our conception of the classification as a partisanship continuum on which political engagement obtains a minimum at the midpoint and increases towards the extremes.¹²

Blocks and Parties

In our reasoning we have relied upon a theoretical model which describes the Swedish party system in the form of a unidimensional party array which is being partitioned into two clusters or party blocs. Like all models, this one depicts the real world in an idealized fashion. Anyone who is familiar with

Swedish politics will be aware, in the first place, of the fact; that the relationship between Social Democrats and Communists at the party leadership level is far less intimate than the usage of the "bloc" concept might imply. The cohesion of the non-socialist bloc is affected by less grave tensions. Nonetheless, the nonsocialist bloc shows incessant signs of falling apart so as to result in an isolated Conservative party and — at best — a two-party Centrist bloc (the People's party and the Center party). In the case of the non-socialist parties, however, threeparty collaboration is consistently recognized as a compelling "minimum winning coalition" requirement for a governmental change.

In a micro-macro political perspective the verisimilitude of our theoretical model must be deemed to be dependent on its capacity to account for the individual voters' responses to party competition within the Swedish polity. From the abundantly rich data material provided by the 1964—1968 panel survey, we have selected some of the relevant measures and analyses for presentation in the article. Although the following exposition is thus not exhaustive, it should be sufficient both to substantiate the tenability of the model and to throw light on its limitations.¹³

In general, the arraying of the parties of a polity along some axis which is generally recognized as a principal evaluative dimension forms a precondition for party politics which would lead us to expect that party competition becomes constrained to "neighborhood competition". Unless some peculiar dislocation occurs in the electorate, one would consequently expect voting change to take the form, primarily, of change flows between parties occupying adjacent positions on the predominant axis. When the distribution of electoral support remains close to equilibrium, it is warranted to add the presumption that there prevails a fair amount of stability and, yet, some vacillation in regard to perceived party positions as well as in regard to the preferred positions held by the individual voters (i.e. their "ideal points" in Coombs' terminology). From these propositions (granting the validity of the added presumption), one can furthermore derive the proposition that change flows between adjacent parties should be composed of voters whose "modal position" should be located near the midpoint between the modal positions of the consistent supporters of the parties they switch between.

In order to test the latter proposition we have inquired into the interdependency between political attitudes and voting behavior. As in the foregoing section, behavior and attitudes in 1964 provide the points of departure for an analysis that is directed backwards as well as forward in the time period spanned by our surveys. The major part of the relevant data material is brought together in Table 18. In the left margin of the table is found a classification pertaining to the individuals' voting records. When applied to the individuals' voting histories up to the 1964 election (in section (a) of the table), the classification takes into account party choice in 1964 and in a sequence of recently preceding elections.¹⁴ When applied to voting behavior at the 1964 and 1968 elections (section (b) of the table), the classification pertains to these two elections only.

Both parts of the table are based, of course, on the 1964—1968 panel of respondents
15

The three attitude measures being deployed in this analysis reflect the voter's outlook at the time of the 1964 election. Our measure of "attitude to -welfare state politics" consists of a set of six questions pertaining to ideologically colored policy issues. It is of course intended to locate the individual's position in relation to the prevailing left-right ordering of the parties. The measure of "Confidence in Government" bears no such explicitly ideological menaing. It has been obtained by means of five interview questions through which respondents were asked to give "good-or-bad" evaluations of the government's achievements in various policy areas. Finally, the measure of "partisan attitudes" is a composite which has been constructed by combining all of the items comprised by the two previous measures; to this set has also been added a question asking respondents whether they felt that current government policies had been mainly favorable or mainly unfavorable to their own social class. In the form these measures are being used here, they are all based on regression analyses in which the appropriate interview responses were entered as independent variables. Voting behavior in 1964 (as defined by a dichotomy comprising only a distinction between Socialist and Non-socialist voting) served as the dependent variable in these regression analyses. Scale scores are actually the predicted values (multiplied by 100) that were obtained by applying the resulting regression equations to each individual's response pattern.¹⁶

By inspecting the data in Table 18 it is found that the modal positions of the various voter categories fit exceedingly well with the expected ordering. Stable Conservative voters are most extreme at the one end of the spectrum in their attitudes to welfare state politics and they are also most consistent in expressing disapproval towards the government. At the opposite end, the Communists occupy an extremely Socialist position on attitudes to welfare state politics. When it comes to confidence in governmental politics they are — as should of course be expected — more ambivalent. As a consequence both for this measure and for the measure of "all partisan attitudes", the mean values increase from a minimum at the stable Conservative position towards a maximum for stable Social Democratic voter, while they decline for the two voter categories at the extreme left.¹⁷

As the two parts of the table are almost identical to each other it may be sufficient to comment in detail upon the data that pertain to voting behavior in the 1964—1968 elections. These are also particularly interesting since they indicate the capacity of attitude measurements obtained four years back in time to predict tendencies in later voting change. (In 1964 the multiple correlation between the items in the "all partisan attitude" set and two-bloc voting was 0.692.) As can be seen from the table, voters who changed between Conservative voting and voting for a centrist party (i.e. the Center party or the People's party or the Christian Democrats) show a mean position in between the stable supporters of these parties. It can be added that those who moved to the Conservative

party did not differ significantly from those who moved in the opposite direction. (The mean score on "All Partisan Attitudes" equals 26 for 1968 Conservatives and 23 for 1968 centrist voters.) What happened was, seemingly, that "leftwing" Conservatives were exchanged for "rightwing" centrist party voters without affecting very much the "attitude average" of any of the parties involved. As is shown by the data given below the "attitude means" for the two major centrist parties do not differ very much, and consequently those who changed between parties in the middle of the party spectrum are not distinguishable from the stable supporters of these parties.

Voters who changed between bourgeois and socialist voting are located very close to the midpoint of the scale (i.e. they had a probability close to 0.5 of voting Socialist in 1964). As can be seen from the following breakdown of the data, they did indeed move from the attitudinal peripheries of their 1964 parties.

In order that our party system model be relevant for the study of micromacro processes it should be a further requirement that the model corresponds to cognitions of the party system which are held by voters in general or, at least, by a quite predominant part of the electorate. This requirement means that within the electorate there should prevail a widely accepted cognitive picture of the party system which could be described as a "map" showing the parties' positioning in relation to each other. The locating of the parties on such a map should then be expected to be concordant with the model.¹⁸ We shall conclude this inquiry by presenting a selection of data which pertain to the tenability of the model in this regard.

From a unidimensional party model one can derive the proposition that the supporters of a given party should be most likely to cognize one of the parties that occupy an immediately adjacent position on the predominant dimension as the party they like "second best" or feel "second closest to". The tenability of that proposition can be judged on the basis of the data in Table 19 which displays

the response distributions obtained when respondents in the 1968 survey were asked about their "second party preferences". Although the corresponding data obtained from the 1964 survey are not shown in this article, they form a pattern which is almost identical to the one appearing in the 1968 data. This lends support to our previous presumption that perceptions of party positions are quite stable over time.¹⁹

Most of the Conservative identifiers rate the People's party as their second preference. However, the fact that a sizable minority mentions the Center party is noteworthy. There prevails, really, no ideologically recognized internal ordering between the two centrist parties on the left-right dimension. On the other hand, there are very clear indications in these and other data that the Center party is becoming cognized as nearer to the Social Democrats than the People's party. At least, that is (as can be seen from the table) the very predominant judgement among Social Democratic identifiers. Also in accordance with the proposition, centrist party supporters are most likely to see the other centrist patry as their second closest choice alternative.

An interesting trend appearing in the table is, furthermore, that the "modal" second preference within each of the parties is most predominantly popular among the party's strong identifiers. Although this trend was not specifically predicted by the proposition it serves to substantiate the general expectation that "weak" party supporters should belong to the attitudinal periphery of their party.

One aspect of the data in Table 19 that deserves particular attention is their striking resemblance to the row-wise rank ordering of the transition probabilites displayed in Table 4. In accordance with our previous reasoning, there appears a very clear tendency for party changers to be most likely to move to the choice alternative that is generally considered as the "second best" within the party they leave. As the author has shown in a previous study, behavioral change rates can indeed serve as a complement to psychological measures of proximity or similarity of parties.20

The interdependency between perceived distances to other parties and party change can also be substantiated with the aid of data pertaining to the voters party choice in previous elections. This is particularly apparent in the case of the centrist party voters, because these voters have several options when choosing the party they could feel "second closest" to. The data given below display the voting records of those 1964 and 1968 centrist party voters who had actually changed parties in comparison to previous elections. (The operational definition of the change categories is the same as in Table 19. Respondents who did not indicate any second preference have been excluded from the classification below.) We can see from these data centrist party voters who consider the Conservatives as the "second best" among the parties and are also most likely to have voted for the Conservative party in a previous election. The same linkage between present second preferences and previous voting appears very clearly in the remaining categories.

Some features of the data in Table 19 are not quite accordant with the clustering of the parties which is implied by the two-bloc conception of the party system. Thus it is found that People's party identifiers are about as likely to choose the Social Democrats as the Conservative party as their secondly preferred party. Among Center party identifiers the proportion favoring the Social Democrats is even somewhat larger than the proportion considering the Con.servatives as their second closest party. Of course, we are not working with complete preference orderings. It may well be true that a very large proportion of the centrist identifiers are considering the Conservatives as "third best" and thereby should move the Conservative party up to more favored average evaluation than the Social Democrats. A much more drastic tendency for an extreme party to repel rather than to attract sympathies becomes apparent in the evaluations of the Social Democratic identifiers. When feelings of friendliness towards other parties are solicited, a large majority of the Social Democrats are obviously not at all inclined to conceive of their party and the Communists as forming a bloc. If we were willing to treat the proportions in the rows of Table 19 as proximity measures, these qualifications would, however, only affect interparty distances along the axis that is defined by the apparent unidimensional ordering of the parties, i.e., Communist, Social Democrats, Center party, People's party and Conservative party.²¹

The Swedish voters' cognitive mapping of their system can also be gauged with the aid of a measure which bears directly upon cognized inter-party distances. This measure was obtained by the use of a simple "cartwheel" technique (cf. Coombs, op. cit.). Thus, for each of the parties, in turn, respondents were asked to indicate which parties they considered as "closest", "second closest" and

"most far away" from that party. From the resulting response patterns, one can recover individual rank orderings of the distances among the parties. These data require a very extensive analysis (making use of non-metric scaling techniques) which cannot be presented in its entirety in the present article. We shall nevertheless draw some partial results from such an analysis, since these are indeed sufficient to substantiate the main features of our party system model.

We shall consider, first, the judgments made by all respondents on the distances
from the extreme parties and to other parties.

The appropriate data are given below. (All percentages are based on the total number of respondents, including those who expressed no opinion.) Within the electorate as a whole there prevails, indeed, an almost complete consensus as to the proper order of the parties in relation to the two extremes of the party system.

The centrist parties were generally seen as closest to eachother. When asked about distances from the People's party, 70 per cent thus mentioned the Center party as closest. Likewise when judging on distances from the Center party, 71 percent thought that the People's party must be its closest neighbor in the party system. When distances to other parties are compared, the People's party is considered as closer to the Conservatives than to the Social Democrats by a large majority. The same judgement is also the most frequent in the case of distances from the Center party but the unanimity is far less pronounced.

Some further penetration of the response frequences will serve to bring the details of these features into light: As regards distances from the People's party, it is found that the Conservatives are most often mentioned as second closest. A total of 76 percent mentioned the Conservatives as either second closest or the closest to the People's party which should be compared to the mere 14 percent who attributed either of these positions of the Social Democrats.

As touched upon above, there was more disagreement about the positioning
of the Center party in relation to the Conservatives and the Social Democrats,
respectively. The Conservatives were considered as second closest or closest to

the Center party by 57 percent when distances from the latter party were judged upon. A sizeable minority disagreed, however: As much as 35 percent thought that the distance to the Social Democrats was the second shortest or even the shortest.

The prevailing view among the centrist parties' voters coincided with the general pattern. The data given below display the modal response pattern for these voters. As is seen, most centrist voters agree that the non-socialist parties are closer to each other than to any other party. Yet, the locating of the Center party at the very midpoint of the party system is sustained by the views of the 22 percent who think of the People's party as closest and the Social Democrats as the second closest to their own party. The clustering of the bourgeois parties is of course also recognized by the Conservative voters.

When gauging distances from the Social Democratic party, 53 percent of all respondents judged the Communists as closest, and a total of 66 percent mentioned the Communists as either closest or second closest to the Social Democrats. The relative smallness of these proportions is due to the presence of a substantial minority considering the gap between Social Democrats and Communists to be wider than that between the Social Democrats and the centrist parties. By way of contrast there was not much disagreement over the location of the Conservatives in relation to the Social Democrats: in comparing distances from the Social Democrats, 76 percent found the Conservatives to be "most far away" from the Social Democrats.

The positioning of the Communists reveals a highly noteworthy differentiation in respect of political outlook among the Social Democratic voters. Of all Social Democratic voters 47 percent considered the Communist party as closest to their own party as compared to 35 percent who chose the Center party. When both "closest" and "second closest" ratings are taken into account, two major categories of Social Democratic voters are formed: One (comprising 45 percent) conceives of the Communists as closest and a centrist party as second closes to the Social Democrats, while another category of about equal size (44 °/o percent) reverses the picture so as to consider a centrist party as closest and the

Communist as second closest. A word might be added about the Communists:
About 80 percent of all respondents thinks of that party as being "most far
away" from each of the non-socialist parties.

In general these data suggest that the voters' cognitions of party distances are more concordant to the two-bloc model than are their second party preferences. Such a comparison also suggests that these two modes of measurement carry somewhat different psychological meanings. Questions asking for "feelings of sympathy" seem to raise the average evaluation of the parties at the middle of the party spectrum. Questions inviting respondents to make judgements on distances among parties, on the other hand, seem more likely to elicit recognition of nearness to an extreme party.²²

We can easily recover from our party distance data one "cognitive map" of the parties' positioning which is consistent with the prevailing views in the electorate. This is given a geometric representation in the figure below. One important qualification is needed, however. Some response patterns suggesting competing pictures of the party system appear so frequently in the data material that they cannot be written off as expressions of idiosyncracies or as sheer noise in the data. None of these raise any real doubt about the prevalence of a basic unidimensionality in the voters' cognitive mapping of the system. The one deviating judgement that is most widely accepted differs from the representation given in the figure in one regard, only: It suggests a larger distance between the Communists and the Social Democrats than between the latter and the Center party. Thus it would place the Communist party far away from all other parties. There is also a sizeable minority who would locate the Center party closer to the Social Democrats than to the Conservatives. Also, there is a minority who think of the People's party as being closest to the Conservatives. Taken together, these versions account for most of the psychologically meaningful patterning that can be found in the data. There are, of course, means to average out all tensions in the data. But we have chosen to let the figure below represent an aggregation of the views on interparty distances that are most prevalent in the electorate.

A word of caution as to the interpretation of these data should be added. They do not prove that all the individuals in the great majority who are aware of the "left-to-right" ordering of the parties must have a distinctively ideological outlook on politics. On the contrary, our survey data evidence that this ordering of the parties may serve as a reference frame also for individuals whose opinions and policy preferences otherwise show very little of partisan consistency.

Why do the bourgeois parties not merge? Or, why could the Center party and the People's party not at least form a joint centrist party? The distinctly ruralagrarian character of the Center party has, of course, posed a major obstacle to any such development in the past. In recent political debate, however, these questions have become highly salient. Party leaderships have not been altogether unresponsive. Thus, the Center party and the People's party have accomplished

a fairly successful policy coordination. Nonetheless, no party merger has occurred,
and none is really expected to occur in the near future.

As can be seen from Table 20 the idea that two or more of the non-socialist parties should merge sounds quite attractive to a large majority of their voters. Conservative voters are close to being unanimously in favor of the establishing of a united bourgeois party. Among supporters of the two centrist parties, opinions are divided over the choice between a three-party merger. Opinions in favor of these two alternatives are about equibalanced.

The opinion profile for the supporters of the two centrist parties is delineated in more detail by the data in Table 21 A-B. It becomes clear that those who consider the Conservatives as a "second best party" are heavily in favor of a three-party merger, while opinions are somewhat balanced in favor of a joint centrist party among the others. Center party voters who have a second preference for the Social Democrats are the least enthusiastic about the three-party merger.

A united bourgeois party was furthermore most attractive to the "rightcenter", that is, to the Center party and People's party voters who hold a strongly negative attitude towards the Social Democrats.²³ These data may serve to bring out a dilemma which the leaderships of the two parties at the middle of the party spectrum must be concerned with. If they choose to strengthen their political position by forming a joint centrist party, they cannot avoid to affront at least a part of their adherents in the "right center". In any case, intra-bloc party competiton would not disappear. On the other hand, if they join a united bourgeois party they take the risk of estranging voting support at their left flank. Under such circumstances, it may indeed seem most prudent to leave things as they are.

Maybe, the majority of the voters agreed upon the prudence of that way of
reasoning when they decided the electoral verdict in 1968. At least, they left
things as they were.

---

NOTES

1 The term "deviating election" has been drawn from Campbell's typological classification of elections. However, when attaching the "deviating" label to the 1968 election, we are obviously extending the meaning of the concept so as to let it denote not only a temporary weakening but also a transient swelling of the predominant party's voting support. Cf. Campbell, Converse, Miller, Stokes. Elections and the Political Order, New York, Wiley, 1966, esp. Chapters 4 and 2. — The background data which provide the ground for considering the 1964 party division as being more "average" or "normal" (especially in regard to the strength of the Social Democratic party) have been analyzed in the author's article on "Political Stability and Change in the Swedish Electorate", Scandinavian Political Studies, vol. I, 1966.

2 These and the following tables showing the size of change categories, require a technical note. There is a small number of persons in the sample who have voted for a non-socialist "coalition ballot" in one of the elections and for a regular party in the other. Persons in this category who voted for the Center party, the People's party, or the Conservative party in the other election are counted as stable voters. If on the other hand, they voted for either the Social Democratic or the Communist party on the other occasion they have been put in the appropriate change category.

3 Comparable data for the 1964 election can be found in: Bo Sarlvik. "Socialeconomic Determinants of Voting Behavior in the Swedish Electorate", Comparative Political Studies, vol. 11, no. 1, April 1969. Data for previous elections are given in: Bo Sarlvik. "Political Stability and Change in the Swedish Electorate", Scandinavian Political Studies, vol. 1, 1966. Additional data material can be found in reports prepared by the author for the official election statistics; these are published in: Riksdagsmannavalen åren 1959_—1960, vol. II and Riksdagsmannavalen åren 1961^—1964, vol. 11, (Sveriges officiella statistik, Stockholm 1961 and 1965, respectively).

4 Of the 39 Center party Members elected in 1968, 24 stated "farmer" as their occupation in the parliamentary calendar. However, of the remaining 15 about half can also be classed as present or former farmers.

5 Bo Sarlvik. "Political Stability and Change in the Swedish Electorate", loc. cit., pp. 196 f. For further discussion of the prerequisites of partisan competition in the Swedish party system, see also the author's article "Party Politics and Political Opinion Formation: A Study of Issues in Swedish Politics 1956—1960", Scandinavian Political Studies, vol. 11, 1967.

6 For a subsample consisting of half of the total sample, the question was asked in a preelection interview (which was later followed by a post-election mail questionnaire) and pertained to the party the respondent intended to vote for. Voters who switched from their preelection voting intentions or had no certain voting intention are classed as "not convinced" in the analysis presented in the present article. (This operation has no significant effect upon the differences and trends discussed here, however.) For the other part of the sample, the question was included in a post-election interview and was asked with reference to the party the respondent had voted for.

7 After a sequence of questions about the current Government, respondents were asked: "How likely do you think it is that we will get another government in this country after this year's election? Is it: Very likely (1%), Rather likely (6 %>), Not particularly likely (50%), Not at all likely (38 %). Marginal proportions are given within parentheses for each alternative; 5 °/o expressed no opinion.

8 Prior to queries about voting, respondents were asked whether they considered themselves as "adherents" of any party. Those responding affirmatively were asked if they regarded themselves as "strongly convinced adherents" of their party. The resulting two groups are denoted as "strong identifiers" and "weak identifiers", respectively. Respondents having no feelings of "adherence" were asked if they felt "closer" to any of the parties than to the others. In this way, the "non-adherents" could be divided into those with a "party preference" and those who had "no party identification". All respondents who indicated any kind of party affiliation were, of course, asked about which particular party they had in mind. For an exhaustive discussion of the concept of party identification and its operationalization, see: Campbell, Converse, Miller, Stokes. The American Voter, New York, Wiley, 1960.

---

8 Prior to queries about voting, respondents were asked whether they considered themselves as "adherents" of any party. Those responding affirmatively were asked if they regarded themselves as "strongly convinced adherents" of their party. The resulting two groups are denoted as "strong identifiers" and "weak identifiers", respectively. Respondents having no feelings of "adherence" were asked if they felt "closer" to any of the parties than to the others. In this way, the "non-adherents" could be divided into those with a "party preference" and those who had "no party identification". All respondents who indicated any kind of party affiliation were, of course, asked about which particular party they had in mind. For an exhaustive discussion of the concept of party identification and its operationalization, see: Campbell, Converse, Miller, Stokes. The American Voter, New York, Wiley, 1960.

9 Frustration due to the seeming inability of the non-socialist parties to defeat the Social Democrats might also be a possible cause of weak partisanship. However, in 1968 the possibility of a governmental change seemed much less remote than in 1964. In pre-election interviews, 7% of the respondents thought a change in government was "very likely", 27% thought it was "rather likely", while 46% considered it as "not particularly likely" and 17% said it was "not at all likely". Among bourgeois voters, 51 % deemed it "very likely or "rather likely" that the election would result in a new government.

10 Angus Campbell. "A la recherche d'un modéle en psychologic électorale comparative", Revue franc. Social., vol. VII, 1966, pp. 579^—597. On the conception of a U-shaped relationship between the intensity component of a one-dimensional attitude and its content component, see: Stouffer et al. Measurement and Prediction, Studies in Social Psychology in World War 11, Princeton, Princeton Univ. Press, 1950, Chapter 7.

11 Bo Sarlvik, "The Relation of Partisan Orientation to Political Engagement in the Swedish Multiparty System", paper prepared for the International Conference on Comparative Political Behavior, Insitute for Social Research, University of Michigan, Ann Arbor, May 1967 (rev. version forthcoming in print). It should be noted that this analysis also led to an alternate model which did not require any two-bloc assumption. As a result, one then obtains a "sine curve" (rather than a U-curve) model with engagement maxima over the "modal" party positions along the "left-right" axis in the party system.

12 As is seen from the bottom row of the table, there is an extremely high degree of coincidence between party identification and voting if the two-bloc division only is taken into account. The comparatively large proportion of defectors in the Socialist "party preference" category is mainly due to defections to the Center party. It should also be noted that a very considerable proportion of the "no party identification" category is actually consistently non-socialist in its electoral choice.

13 For further discussion of the micro-macro direction of analysis, see: Stein Rokkan. Citizens, Elections, Parties. Approaches to the Comparative Study of the Processes of Development, Oslo, Universitetsforlaget, 1970, pp. 18 ff. Theoretical models of party competition are treated, inter alia, in: Anthony Downs. An Economic Theory of Democracy, New York, Harper, 1957; Philip E. Converse, "The Problem of Party Distances in Models of Voting Change", in Jennings — Zeigler (ed.). The Electoral Process, Englewood Cliffs, Prentice Hall, 1966; Gunnar Sjoblom. Party Strategies in a Multiparty System, Lund, Studentlitteratur, 1968; Donald E. Stokes. "Spatial Models of Party Competition", in Campbell, Converse, Miller, Stokes. Elections and the Political Order, New York, Wiley, 1966. See also: Bo Sårlvik. "Partibyten som mått på avstånd och dimensioner i partisystemet" ("Party Change as a Measure of Distances and Dimensions in the [Swedish] Party System"), Sociologisk Forskning, nr. l, 1968.

14 It is obvious that these statements require several qualifications which have been omitted here; see the author's article on "Partibyten som mått på avstånd och dimensioner i partisystemet" ("Party Changes as a Measure of Distances and Dimensions in the [Swedish] Party System"), loc. cit. It should also be acknowledged, that most of the above reasoning has its origin in the work of Downs, op. cit.

15 A technical note is required with regard to the classification that pertains to voting behavior up to the 1964 election. In addition to data concerning voting in 1964 it is based upon interview questions concerning party choice in 1962, 1960 and at the last previous before 1960 in which the respondent had participated. These questions were asked in the 1964 survey. In the classification, respondents who had switched between different bourgeois parties as well as between a socialist and a bourgeois party have been put in the category comprising "change between non-socialist and socialist party". Furthermore, those who had switched between two different centrist parties as well as between a centrist party and the Conservative party have been included in the category of "change between Conservative party and centrist party". — Firsttime voters in 1964 and a small category of voters who gave too incomplete responses to permit classification have been excluded in the classification in section (a) of the table.

---

15 A technical note is required with regard to the classification that pertains to voting behavior up to the 1964 election. In addition to data concerning voting in 1964 it is based upon interview questions concerning party choice in 1962, 1960 and at the last previous before 1960 in which the respondent had participated. These questions were asked in the 1964 survey. In the classification, respondents who had switched between different bourgeois parties as well as between a socialist and a bourgeois party have been put in the category comprising "change between non-socialist and socialist party". Furthermore, those who had switched between two different centrist parties as well as between a centrist party and the Conservative party have been included in the category of "change between Conservative party and centrist party". — Firsttime voters in 1964 and a small category of voters who gave too incomplete responses to permit classification have been excluded in the classification in section (a) of the table.

16 It goes beyond the scope of the present article to discuss the propertics of these measures in any detail or to deal with them in relation to scaling theory. Yet one deficiency of the measurement technique should be mentioned: It does not distinguish between the kind of attitudinal ambivalence that arises when attitudes are somehow equibalanced, and the kind type of ambiguity which is an expression of sheer absence of political cognitions. However, it is significant for the purpose of the present inquiry that both kinds of ambivalent individuals ought to be particularly unlikely to perceive ideological "purity" as an attractive feature in the parties' appearances.

17 One way analysis of variance gives the following F-values and correlation ratios for the section (a) of the table: Attitude to Welfare State Politics: F = 193.72, com ratio = .36; Confidence in Government: F = 177.94, corr. ratio = .34; All Partisan Attitudes: F = 352.36, corr. ratio = .50.

18 This is of course to say that we require the model to be consistent with a "stimulus space" which is being cognized and shared by a very large part of the electorate. For this requirement to be fullfilled, it is not needed that individuals' agree in their evaluations of the parties. It is sufficient that they employ concordant criteria for making judgements on "psychological distances" among the parties and/or distances from their personal positions to the various parties. For an exhaustive treatment of this mode of data analysis, see: Clyde H. Coombs. A Theory of Data, New York, Wiley, 1964.

19 Profound alterations in party strategies may cause such perceptions to change, however. In our 1956 survey (conducted during the Social Democratic — Center party coalition government), 54 °/o of the Center (Agrarian) voters who expressed any second preference chose the Social Democratic party. Only 8 °/o of the People's party voters who indicated a second preference mentioned the Center (Agrarian) Party.

20 See the author's article on "Partibyten som mått på avstånd och dimensioner i partisystemet" (Party Change as a Measure of Distances and Dimensions in the [Swedish] Party System"), loc. cit. In this study transition probabilities as well as a specific kind of "change coefficients" were used as measures of proximity. The "change coefficients" were the ratios between observed and expected frequencies in the change cells. Expected frequencies were calculated with the aid of a technique suggested by Goodman as suitable for the case when one wishes to ignore the main diagonal frequenies in a mobility table. While non-metric scaling of transition probabilities led to a one-dimensional solution, a similar analysis of change coefficients suggested a two-dimensional space in which the Center party and the People's party were located widely apart on a "less important" (conceivably rural-urban) axis. Coombs' technique for multidimensional unfolding as well as the Guttman-Lingoes' computer program for smallest space analysis were applied and yielded very similar results in this case. (Cf. Leo Goodman. "On the Statistical Analysis of Mobility Tables", Amer. Journ. of Social., vol. X, 1965, and Goodman, "A Short Computer Program for the Analysis of Transaction Flows", Behavioral Science, vol. IX, 1964.

21 This would, of course, require that Table 19 can be considered as a conditional proximity matrix wherein the row-wise rank ordering of second preference proportions reflect the rank order of the appropriate inter-party distances. The unfolding of such "stimulus I-scales" would be quite straightforward and lead to the result indicated above. Cf. Coombs, op. cit.

22 Actually, the 1964^—1968 panel survey provides a set of completing proximity and similarity measures which will allow us to present a more searching comparison of different models of data collecting in a subsequent study.

23 There are grounds to believe that the balance of opinion described here is quite stable. In the 1964 survey, respondents were asked whether they considered it a good or a bad idea that all the three bourgeois parties should join in order to form a united party. In 1964, 80 °/o of the Conservative voters were in favor of a united bourgeois party. The corresponding proportion for People's party and Center party voters were 63 °/o and 39 %>, respectively. If one takes into account that the 1964 question did not offer a two-party merge as a response alternative, the 1964 and 1968 response distributions must be considered to show a high degree of concordance.

23 There are grounds to believe that the balance of opinion described here is quite stable. In the 1964 survey, respondents were asked whether they considered it a good or a bad idea that all the three bourgeois parties should join in order to form a united party. In 1964, 80 °/o of the Conservative voters were in favor of a united bourgeois party. The corresponding proportion for People's party and Center party voters were 63 °/o and 39 %>, respectively. If one takes into account that the 1964 question did not offer a two-party merge as a response alternative, the 1964 and 1968 response distributions must be considered to show a high degree of concordance.