Resumé

The article analyzes the Danish national election in March 1998. Judged from the aggregate figures, Denmark has stabilized. Net volatility was moderate, 'bloc' volatility was close to zero, and despite forecasts of a non-socialist victory, the Social Democratic-led government managed to continue. Further, in the 19905, the periods between elections have been close to the maximum four years. Therefore, the old picture from the 1970s of Danish politics as highly unstable and as extremely volatile is now outdated. Quite the contrary at the level of individual voters. Close to a third shifted from one party to another, and even among voters who did not shift, a substantial proportion had considered voting for another party. Individual voter mobility seems to be a result of weak voter differentiation between different parties, and not a result of protest and outspoken dissatisfaction. Whatever the explanation, individual volatility is an omen of possible future instability: There is no guarantee that different voter movements will always balance out.

Overview

The Danish election on March 11, 1998 carried the important message that Danish politics has stabilized. First, elections have become more seldom. In the 19705, Denmark had five elections in a decade and in the 1980s, four elections. Elections were often called as soon as the government faced serious difficulties in parliament or could hope to make electoral gains. However, in the 19905, the electoral period has been close to four years with one election in 1990, one in 1994 and one now in 1998.

Second, volatility is now only moderate. Whereas net volatility was 29 percent in 1973 and 18 percent in 1975, it was down to 11 percent in 1994 and 12 percent in 1998. Furthermore, the small changes hardly add up to a general trend. So, at the recent elections the parties behind the present Social Democratic-Social Liberal government gained a poor 0.5 percent of the total vote, and the opposition parties won 0.1 percent while the bill was paid by the group of 'others' with a decline of 0.6 percent of the total vote.

---

*Hans Jørgen Nielsen, Department of Political Science, University of Copenhagen. Rosenborggade 15, DK-1130 Copenhagen K, Denmark. E-mail: hjn(aHfs.ku.dk

For 25 years, three highly unstable elections in the mid-1970s (1973,
1975, and 1977) have shaped the image of Danish politics as highly
unstable. At the system level this image is outdated.

Things differ at the level of individual voters where more indicators point to a substantial, hidden instability. For example, the polls predicted a clear loss for the government, and there may therefore have been a major net swing the last minute before the election, indicating weak bonds between voters and parties. Closer scrutiny of results from the Danish Electoral Survey 1998 also indicates a high proportion of voter shifts. To a large extent, aggregate stability came about because most shifts canceled each other out. Furthermore, even stable voters were often uncertain about their party choice.

Therefore, an important caveat should be added to the main conclusion: Individual level volatility is an omen of possible future gross instability. Different voter swings may not always balance out. In fact, one reason for the 1973 earthquake election was a completely unbalanced voter movement away from the established parties. Such an election may never occur again, but there is no guarantee that stability will last either.

The Long Electoral Period

At the time of the 1998 election, majority patterns in parliament were highly different from the majority patterns at the 1994 election. Four years ago, a strong centrist government faced parties on both sides. In 1998, the election was much more a choice between left and right. The government parties were weakened and would only have a chance of winning in combination with the left-wing parties.

Part of the change goes back to the 1994 election. One of the four parties in the government coalition, the Christian People's Party, did not pass the electoral threshold, and combined the three remaining parties - Social Democrats, Social Liberals and Center Democrats - also lost substantially. The government thereby became strongly dependent on support from other parties which almost inevitably created strains within the government itself about who to cooperate with. Especially the Center Democrats opposed left-wing influence, which led to further erosion. In December 1996, the Center Democrats left the government and increasingly made it clear that they would prefer a non-socialist government after an election. As the Christian People's Party ended up with the same message, it became evident that the prime minister had lost two of his former three non-socialist partners to the opposition on the right. Yet, he still had a chance. The Social Liberals remained faithful and it was given beforehand that the left-wing parties would only support a government led by a Social Democrat.

This change could easily have brought an election in 1997. The economy was blooming, but maybe too much. There were signs of bottlenecks on the labor market, and the surplus on the balance of payment was declining. Some fine-tuning seemed necessary, but it was not easy for the government to get parliamentary backing for more than minor adjustments. According to the pattern from previous decades, it was tempting to call an election if the government had a chance to win. This would fit best with the Social Democratic tradition, whereas the Social Liberals have always been negative early elections.

Twice, it was close. In May 1997, the Conservatives had serious leadership and the government's figures in the polls had improved. Leading Social Democrats pressed for an election, but the Social Liberals said no. However, in October 1997, the Social Liberals had accepted an election, and the government made proposals which the non-socialist opposition was expected to refuse. Surprisingly, the Liberals and the Conservatives conceded, and an election could not be called.

By contrast, the March election came when few expected it. Obviously, the need to fine-tune the economy was growing, and there were only six months left of the electoral period. However, there was also the expectation that an election in the spring of 1998 would mix national politics up with the referendum on the Amsterdam Treaty on May 28. This was deemed dangerous, not least for the Social Democrats, in view of the widespread skepticism towards European integration.

Therefore, the long electoral period resulted from a mix of traditional Social Liberal opposition to early elections (May 1997) and a tactical mishap (October 1997) without any shifts in basic attitudes. Under different circumstances, Denmark could easily return to the system of frequent elections, but whatever the explanations and whatever the forecasts, Denmark got two long periods in the 19905, and frequent elections are therefore not a built-in feature of Danish politics.

The Campaign

The call for election was the first surprise; the outcome was the second. The government victory was minimal, compared with the 1994 election, but substantial compared with the expected non-socialist take-over. These expectations seemed well founded (Table 1). In the polls, the two remaining government parties, Social Democrats and Social Liberals, had lost since the 1994 election, and even when the left-wing parties were included, government support in the polls was hardly higher than 47-48 percent.

This did not compare too well with the support for the traditional opposition, i.e., those parties that had been against the government throughout the electoral period. The core of the opposition was the Liberals and the Conservatives. It was given beforehand that they should be the nucleus of a non-socialist government, and the open question was only whether the prime minister should be a Liberal or a Conservative. Combined, the Liberals and the Conservatives had about the same support as they had at the 1994 election, but the new right had grown. The former leader of the Progress Party had started her own party, the Danish People's Party, which quickly became more popular than the mother party.

So far, there seemed to be an almost even balance between government supporters and the traditional opposition. However, the Center Democrats would bring the combined strength of the opposition up around 50 percent, and in the last week of the electoral period it became increasingly likely that the Christian People's Party would pass the electoral threshold. In that case, odds were 52-48 in favor of the opposition.

All polls during the campaign carried the same message of a clear lead for the opposition and the figures were almost identical from day to day, with no indication of a clear trend. Especially the Social Democrats seemed to have problems with 32-33 percent support in the polls against 34.6 percent at the last election. The main changes occurred when parties passed

the electoral threshold. It was therefore no wonder - but hardly good tactics - that the liberal leader Uffe Ellemann-Jensen proclaimed victory already on election morning. Although he was likely to be right, he was wrong, and so was almost everybody else.

It was hardly a wonder that there were few signs of change during the campaign. The electoral campaign was short, largely without focus, and mostly remarkable for its omissions. First, there was almost no emphasis on economic policies. Indeed, it was difficult for non-socialist parties to make gloomy forecasts when even bankers and leading non-socialist papers pointed to the fine shape of the economy. Therefore, the opposition mainly contended with criticism that the government had failed to use the good times for further improvements in the economy and even this criticism was mixed, as it conceded that the times were good.

Second, the topics shifted all the time. For two or three days, the quality of teaching in public schools was in focus, then for two or three days it was crime, and so forth. In the last days, a non-socialist proposal to lower taxes on house owners received some attention. Ads from trade unions claimed that this would have to be paid by those who rent their housing. After the election, these last-minute ads were given as an explanation that the government had done much better than should have been expected from the polls.

Instead of policy themes, the media focused on 'personalities.' From the very first hours of the campaign, the media saw the election as a presidential contest between the Social Democratic Prime Minister, Poul Nyrup Rasmussen, and the leader of the Liberal Party, Uffe Ellemann-Jensen. This was a daring perspective. Denmark has a lot of parties, only three out of five voters support the two largest parties, and it was even doubtful whether the non-socialist opposition parties were united behind Uffe Ellemann- Jensen, as some preferred the Conservative leader. The presidential perspective have become self-fulfilling through the media's focus on only two politicians and thereby on two parties. Politicians from minor parties complained. Yet, the impact of the media agenda was probably smaller than normally assumed: On election day, the two main parties only progressed from 58 percent to 60 percent of the total vote.

In any case, there was a puzzle after the election: Why did the government maintain and slightly improve its electoral strength when all the polls had predicted that the government would loose? On election day, a GFK poll, conducted the day before, showed a marked gain for the Social Democrats, and later this poll was widely taken as an indication of a last-minute swing. However, even this poll only gave the parties in the government bloc 48 percent support, and the only reason the swing seemed so large was a previous, very low estimate of support to the government bloc. The gap between the polls and the actual outcome remains a puzzle.

What the Voters Found Important in the Campaign

After the election, the Danish Electoral Survey 1998¹ presented a list of campaign issues to the voters and asked how important each issue was for their choice of party.² Table 2 lists the issues according to the percentage who said that the issue had been decisive.³ The ranking is striking. Classic macro-economic issues are low on the list:

unemployment ranks 9th,

taxes rank 1 lth and 12th,

public expenditures and public debts rank 13th.

Amazingly, relations with the European Union rank a modest Bth. After
all, there was going to be a referendum a few months after the election.

These low ranks cast new light on the disappointment of the government that it was not rewarded in the polls for the booming economy. Eventually, the government won the election, but the margin was so small that it almost belittles the widespread theories of a substantial correlation between economic cycles and government success in the electorate. Apparently, the state of the economy was not a burning issue for the voters.

Instead, environment and the 'new right' issues of law and order and
immigration placed high on the list. Even more interesting, three of the top
issues were related to the quality of the welfare state:

There was - and still is - widespread public discontent with waiting lists
in hospitals. Even cancer patients often have to wait several weeks for
an operation and then later for further treatment.

There were - and still are - frequent complaints of inadequate assistance
to the elderly.

Comparative reports place Danish school children below the European
average in reading and other skills.

Danes used to believe that high taxes were compensated by a high quality
welfare state. However, now the quality of the institutions was open to
doubt, which apparently impressed the voters.

This pattern resembles the one Borre & Andersen found after the 1994 election. They saw a pressure from the voters to give priority to issues like health care and the problems of the elderly, while the politicians continued to focus debates on economic issues (Borre & Andersen 1997, 91). The main difference to the 1994 election is that the politicians in the 1998 campaign hardly focused on economic issues, if at all.

The agenda clearly varied between different groups of voters. Those who voted for the government put more emphasis on welfare issues and the environment than voters who supported the opposition, while the opposite holds for issues like relations with the EU, taxes and public expenditures. This difference was probably influenced by the traditional perception - found once again in another series of questions in the Danish Electoral Survey 1998 - that Social Democratic-led governments best handle welfare politics, while non-socialist governments best handle the economy. Budge & Farlie (1983) have explained electoral outcomes by variations in the main agenda from one election to another. However, if the agenda varies from subgroup to subgroup, there is hardly one single main agenda.

Two of the issues on the list relate to two explanations of the gap between the polls and the electoral outcome. First, the trade union campaign against the non-socialist proposal to lower taxes on house owners was assumed to have scared voters back to the government camp. Next, there were guesses that polls showing the chance of getting Uffe Ellemann-Jensen as prime minister in a non-socialist government might have made voters stick to Poul Nyrup Rasmussen. However, both questions were so low on the list that they were not really issues at all.

The Election Result

Compared with expectations from the polls, the election result was very
dramatic: No, the Social Democrats did not lose and no, there was no
change of government. However, compared with the 1994 election, there

were only two significant changes, namely the heavy losses of the
Conservatives and the rise of the Danish People's Party (see Table 3).

The election was obviously a disaster for the Conservatives - going from 15 percent to 9 percent of the total vote. This loss was also the fifth in a row after the 1984 election when the party got 24 percent of the vote. This time, it was almost given beforehand that the party should loose: For one year there had been petty scandals and heavy infighting at the top, and the party was almost an object of public amusement. Still, the magnitude of the loss surprised.

This entailed a further surprise: Ceteris paribus, the Liberals should have been able - at least for the time being - to attract a lot of dissatisfied Conservatives. However, while the Conservatives lost 6 percent of the total vote, the Liberals gained only 0.7 percent. After large advances at the previous elections (1990 and 1994), the Liberal progress clearly had come to a standstill.

Next, the new Danish People's Party got 7.4 percent of the vote. A large part of the votes was won from the Progress Party, but not all, and combined the two parties were back at the level which the Progress Party had reached in the late 1980s.

All remaining changes were very modest and even more negligible when the time perspective is prolonged back to 1990 (not shown in the table): In most cases, the pendulum simply swung back; parties that lost in 1994 regained some of their losses, while winners from 1994 lost some of their voters.⁴ This pattern applied to the two left-wing parties, to the two government and to the two center parties (the Center Democrats and the Christian People's Party).

The modest changes are reflected in a moderate aggregate volatility of 12 percent, and even that figure may be misleading. It is inflated by the split in the Progress Party without which net volatility would only have been 9 percent. Furthermore, bloc volatility - movements between socialists and non-socialists - was a microscopic 1 percent (see the arguments of Bartolini & Mair 1990, ch. 1). In short, Danish politics has stabilized.

A Note On the New Right

The social pattern in party choice did not seem to change much, compared with previous elections. The Conservatives lost, but they lost almost everywhere. However, the other major shift, namely the gain for the new right, deserves a special note. The 'new right' parties, the Danish People's Party and the Progress Party, made a combined gain of 3.4 percent. It was rightly stressed after the 1994 election (Borre 1995, 199; Borre & Andersen 1997, 63) that support for the new right is strongly linked with (poor) education. However, there were also two other patterns.

First, the two new right-wing parties got slightly stronger support from the old generations which contrasts with the normal assumption that it should be easier for parties on the rise to attract young voters without preestablished to other parties.

Next, the two 'new right' parties not only got more than average support from self-employed voters, but also from people who receive their income from disability pensions and from normal old-age pensions. It is surprising that people on welfare support parties which are considered to be on 'the right wing.'

However, since education is closely related both with age and employment situation, a control was made (Table 4). The relation between age and new right vote turned out to be spurious and so did the higher support for the new right among welfare recipients. Two patterns remain:

Support for the new right is markedly stronger among self-employed. This fits with the pattern for the Progress Party in the 1970s (Nielsen 1979, 57-68), and so far there still is an element of "old politics' in the support for the 'new right.'

Support for the 'new right' is strongly related to education, and that is what really counts. However, the relationship is not linear but mostly caused by a very small 'new right' vote among people with a high school exam (a third of the electorate).

Highly educated voters therefore seem almost immunized against the 'new right' tendencies, but not voters with medium education who are just as likely to support the 'new right' as voters with low education. Therefore, the new right may have a broad appeal.

Gross Volatility

This more or less ends the picture of stability. At the level of the individual voters, the Danish Electoral Survey 1998, conducted shortly after the election, shows both a remarkably high amount of individual shifts and widespread uncertainty.

First, according to the weighted data, individual volatility was up to 30 percent among those who voted both in 1994 and in 1998.⁵ The figure is almost identical according to the unweighted data. As the 'normal' level of party shifting is around 20 percent, it seems that voters in 1998 were more volatile than usually.⁶ And while there was almost no net change between socialist and non-socialist parties or between the government bloc and the opposition, around 8 percent of all voters crossed the line between socialist and non-socialist parties, and 9 percent crossed the line between government supporters and opposition. To a very high extent, the individual simply seemed to have canceled each other out at the aggregate level.

Furthermore, a substantial number of those who voted for the same party as in 1994 only did so with some hesitation. The Danish Electoral Survey 1998 asked voters not only which party they voted for, but also whether they had considered voting for another party. This leads to a threefold classification of the voters:

Loyalists (50 percent), who voted twice for the same party without
considering another party.

Doubters (16 percent), who voted twice for the same party but had
considered voting for another party.

Shifters (29 percent), who voted for different parties at the two
elections.

The remaining 5 percent gave no information whether they had considered another party. With so much uncertainty it was no wonder that the campaign was the time for the final decision. This does not necessarily imply that party choice was caused by the campaign. Voters might simply have waited until the last minute. While only 8 percent of the loyalists made their decision during the campaign, 32 percent of all doubters did so, and campaign time shifting topped with 54 percent among the shifters. Overall, a fourth of all voters (26 percent) decided how to vote during the campaign.

Furthermore, many waited quite long: While only a few (2 percent) of the loyalists made their decisions in the last days of the campaign, a fifth of the doubters (22 percent) did so, and among the shifters almost a third (29 percent) waited until the last days before they made up their mind.

According to pollsters, late decisions were the main reason the polls turned out to be poor forecasts of the result and especially why the Social Democrats did much better than expected. Indeed, the Danish Electoral Survey 1998 indicates that the Social Democrats won among those who decided during the campaign, while they had lost slightly among those who decided before the election.⁷ So far, the explanation is sustained.

However, it must be left open whether it happened in the very last days. For those who voted in both 1994 and 1998, the figures in the Danish Electoral Survey 1998 suggest that it happened earlier during the campaign period.

There were also shifts back and forth between voting and non-voting. A tentative analysis of data from the Danish Electoral Survey 1998 concurs with Thomsen's conclusion from an ecological analysis (Thomsen 1998b) that the Social Democratic Party made a net gain from shifts between voting and non-voting. For these voters, the time of decision cannot be fully clarified, but they are likely to have decided late in the campaign.

Therefore, 'late decisions' is a possible factor behind the gap been the
polls and the result.8

Interpreting Volatility

Even when a large number of voters shift from one party to another, they normally move only a short distance. As mentioned, bloc volatility was 8 or 9 percent, depending on how the blocs are defined (socialist parties versus non-socialist parties or government supporters versus opposition), while gross volatility was around 30 percent. It may be argued that the blocs are so heterogeneous that even within-bloc volatility may reflect a profound shift in political orientations. Therefore the parties were categorized and scaled according to their left-right position:

Most shifts took place either within the same category or between neighboring categories. Only 8 percent of all voters made larger shifts, for example from the left wing to the Social Liberals or from the Social Democrats to the Center Democrats and beyond.

This points to the importance of the party format: In a system with a high number of parties it is often easy to be equally - or almost equally - positive towards several parties. The Danish Electoral Survey 1998 asked respondents rate the parties on a scale from 0 ( = highly dislike) to 10 (highly like). It then turned out that

• voters gave their 1998 party a high rating (stable voters with a mean of
8.6 and shifters with a mean of 8.2). Almost nobody rated their 1998
party lower than 6, so there was almost no 'negative voting';

• party shifters were also quite positive towards the party they voted for in 1994. The shifters gave their old party a mean rating of 6.4, and two thirds of all shifters rated their old party at 6 or higher. So far, party shifts can hardly be considered an expression of strong protest. Only the remaining third was either neutral or skeptical towards their former party.

Further, for most voters there was only little difference between the ratings given to the best rated and to the second rated party (Table 5). Thus, a quarter of all voters (25 percent) gave two or more parties tied top ratings, and for another quarter (28 percent) the best and the second-best rated parties were only one point apart on the scale from 0 to 10. Only the remaining half (47 percent) differentiated more sharply between their top priorities among the parties.

The distribution of sympathies is clearly linked with the propensity to
shift (last column in Table 5): 46 percent of those who gave two or more
parties tied top ratings shifted from one party to another, while the figure at

the opposite end was down to 20 percent among voters who had a difference
of three points between the best and the second best rated party.

This points to a situation which resembles model I in Figure 1 more than it resembles model 11. In both models, party A is the favorite party, but in model I, more other parties (B and C) are almost equally well liked, while the other parties in model II are rated much lower than the favorite party. The implications for the interpretation of volatility are clear. If model II

had applied, high volatility would have implied a profound reorientation of voter attitudes towards the political parties. Not so when model I fits best. Temporary leadership problems easily make voters leave one party, while good media performance may attrået them to another, and so forth.

One of the main causes of volatility among Danish voters is probably
the combination of (a) many parties and (b) weak voter differentiation
among these parties, and not negativism towards the parties.

Perspectives

The main lesson from the Danish 1998 election was clearly that Danish politics can be highly stable. The sensational elections are now one generation back in time, and in the long perspective they seem to have been exceptions to the rule rather than the basic truth about Danish politics. Without the three unstable elections, there simply does not seem to have been any long-term trend towards higher volatility since the introduction of proportional election at the beginning of the century.

This is no guarantee that stability also will reign at the next elections. At the individual level, there was a high amount of party shifts and in addition a substantial number of voters who considered shifting to another party. Furthermore, voters often did not differentiate sharply between different parties but liked two or more parties equally. Only as long as these shifts and uncertainties point in different directions is stability likely.

---

NOTES

1. The Danish Electoral Survey 1998 was financed by a grant from the Danish Social Science Research Council to Johannes Andersen and Jørgen Goul Andersen (both University of Aalborg), Ole Borre (University of Aarhus), and the author. The field work was conducted by AIM-Nielsen.

2. The categories were: decisive, highly important, some importance, only minor importance or not important at all.

3. Ten of the fifteen issues were also used in the 1994 election survey. However, the response categories differed and no direct comparison is possible.

4. Thus, the Social Democrats lost 2.8 percent of the total vote in 1994 but now regained 1.3 percent. The Center Democrats had lost 2.3 percent but now regained 1.5 percent and so forth.

5. Obviously, volatility is even higher when shifts between voting and abstaining are included. In that case, only 55 percent of the electorate voted for the same party in 1994 and 1998, 23 percent shifted, and 22 percent abstained at one or both elections.

6. A parallel calculation from the 1990 election survey had 63 percent voting for the same party without hesitation, 17 percent only doing so after considering another party, and 20 percent actually shifting.

7. The loss among those deciding before the campaign corresponds to 0.3 percent of all those voting on March 11 (irrespective of time of decision), while the gain during the campaign was 1.7 percent. However, most was gained earlier in the campaign and not in the last days.

---

8. Only those who voted were asked when they made the decision. The answers indicate that former non-voters made their decisions late. However, there is no information when former voters decided not to vote. It remains a possibility that all the polis had common shortcomings. At the 1994 election, Gallup stressed that the polis had a larger margin of error than earlier because it was difficult for the voters to recall their former vote when the previous election was four - and not two or three - years back in time. No such statements were made in 1998, but the situation was the same.

REFERENCES

Bartolini, S. & Mai, P. 1990. Identity, Competition and Electoral Availabilitiy. The Stabilisation
European Electorates 1885-1985. Cambridge: Cambridge University Press.

Borre, O. 1995. 'Old and New Politics in Denmark,' Scandinavian Political Studies 18, 187-205.

Borre, O. & Andersen, J. G. 1997. Voting and Political Attitudes in Denmark. Aarhus: Aarhus
University Press.

Budge, I. & Farlie, D. 1983. Explaining and Predicting Elections: Issue Effects and Party
Strategies in Twenty-Three Democracies. London: George Allen and Unwin.

Nielsen, H. J. 1979. Politiske Holdninger og Fremskridtsstemme. København: Politiske Studier.