Resumé

Opinion research has for long discussed the robustness of individuals' political opinions. Panel data have been a much used method to study opinion change and stability. This method can tell how many change position from one point of time to another, but it can hardly be used to tell why individuals change or maintain their attitudes. To be able to study the significance of information upon people's political views, opinion research has in recent years adopted an experimental approach. This article examines the results from a survey experiment, based on the counter-argument technique developed by Sniderman, Piazza and Kendrick. The technique makes it possible to study how readily people can be persuaded to leave a political position when confronted with a counter-argument. The subject of the experiment was taxation of old age pensioners. The analyses showed that it was easier to persuade respondents to abandon a position in old age pensioners' disfavour than a position in old age pensioners' favour. Respondents with a low educational level were easier to persuade than respondents with a high level of education.

Introduction

The purpose of this article is to investigate how readily people can be persuaded to leave a position on a political issue. Using data from a surveybased we try to answer two research questions. First, how many on each side change their minds under the pressure of a counter-argument? Second, which segments of the population are easiest to persuade?

The questions posed in this article arise out of a classic and ongoing discussion between political scientists about the robustness of people's politicalopinions. contemporary agenda for this debate was set by Philip Converse in his famous article 'The Nature of Belief Systems in Mass Publics' (1964). He found that when ordinary citizens were repeatedly asked the same questions about their policy preferences, they tended to give differentanswers different times. And the changes in responses did not seem to fit any reasonable pattern. Converse argued that this lack of consistency was an expression of 'non-attitudes' or 'doorstep opinions', constructed to

---

* Ann-Helen Bay, NOVA Norwegian Social Research. Munthes gt 29. N-0260 Oslo. Norway.

conceal the fact that the respondents actually had no opinion at all. Subsequently,several have argued that the instability of individual citizens' responses to surveys may partly result from measurement errors (Achen 1975; Judd & Milburn 1980), from ambiguous questions and from errors in preparing the data. According to these opponents, making statisticalcorrection measurement errors gives the researcher access to people's true attitudes.

In recent years a third school of thought has been established within public opinion research (Zaller & Feldman 1992). This school of thought draws heavily on cognitive schema theory, first elaborated within cognitive psychology and later adopted by social psychologists (Lau & Sears 1986; McGraw & Lodge 1995). According to schema theory, individuals relate to problems in general by means of a set of cognitive structures or heuristics. A precoded set of associations enables individuals to form opinions and reach decisions on a broad range of problems (Fiske & Taylor 1991). Applied opinion research, opinions are not regarded as fixed entities, but rather as distributions of considerations - 'the attitudes, beliefs, values, and information that can be retrieved in response to a question' (Feldman 1995, 266). As these considerations or heuristics may not only vary in character, but also have contradictory implications for the individual's opinion about a problem, the frame of reference for the problem faced will often be of decisive importance (Sniderman et al. 1991a; Zaller 1992). Frames influence the person's understanding of which considerations are relevant in order to think about an issue, and consequently influence the person's opinion. According to this school of thought, opinions have to be understood as an interplay between situational factors (frames) and individual (heuristics) (Feldman 1995).

The conventional opinion survey will seldom be a suitable tool for studying this interplay, because the researcher has little control over information respondents have been exposed to (McGraw & Lodge 1995). Methodologically the focus on heuristics and frames has led opinion research experimental methods. Experiments allow the researcher to decompose the complex phenomenon information and in a controlled manner to study the effects of it upon people's opinions (McConahay 1973; Kinder & Palfrey 1993).' Experiments within opinion research take place in laboratories (Nelson et al. 1997) and in non-representative (Green et al. 1994) and representative (Sniderman et al. 1991c) public opinion surveys. The experiment analysed in this article was part of an extensive representative survey. Its aim was to measure the robustness of people's opinions towards a counter-argument. The counter-argument technique was developed by Sniderman et al. (1991c), with the aim of paving the way for the study of political persuasion. We will describe the technique more closely as we explain the subject of the experiment.

The Experiment: Subject and Design

Advocates of experimental methods within opinion research regard them as a supplement to conventional opinion survey, not a replacement. Experiments deepen our understanding of how individuals reason about political Can an experiment elaborate findings that remain puzzles in conventional opinion surveys? The experiment we report here aimed to broaden understanding of Norwegian public opinion towards policy programmes the elderly.

Political issues involving the elderly play a predominant role in Norwegian Policy measures concerning the aged attract a lot of attention the population (Hellevik 1997), and such policy programmes weigh heavily in public budgets. From political as well as social science quarters it has been contended that heavy expenditure on the elderly sector will lead to conflicts of interest between workers and pensioners. The elderly will use their numerical strength to force through increased transfers, whereas the economically active part of the population will be opposed to financing them (Light 1988; Johnson et al. 1989; Hippe 1990). However, so far it has proven difficult to find evidence of a war against welfare policy programmes directed towards the elderly. On the contrary, numerous studies show broad consensus in giving political priority to the elderly. Large portions of the electorate want government to spend more money on the elderly, and very few support cuts in public spending in this field (Colbjørnsen et al. 1985; Svallfors 1989; Pettersen 1995). If a political conflict interests is to take place between workers and pensioners it most probably has to be initiated from above - through arguments from political leaders or other opinion leaders. Our experiment aimed to investigate the persuasive power of arguments highlighting the conflicting interests between and pensioners upon people's opinions.

The counter-argument technique is designed to grasp the dynamic between argumentation and people's attitudes. According to its inventors, people who form an opinion on a political issue are exposed to influence from political leaders, the mass media or through discussions with friends, colleagues or family members. Typically, the public comes to a controversial political issue with an established sympathy for one or the other side. Political struggles usually boil down to each side attempting to persuade those who are initially opponents. As the conventional opinion survey interview is designed to minimise pressures upon respondents, it taps people's opinions in a rather artificial situation. The counter-argument technique is designed to mimic real-life conversation, by creating a situation where the respondent's opinion evokes a reaction from another."

Some general reservations have to be made concerning the counterargumenttechnique.
technique does not enable us to rank objectively

the strength of alternative counter-arguments (Sniderman et al. 1991b). The information given in the counter-argument is also very selective. The respondent is allowed to consider only one aspect of the problem and contrary to a real-life discussion there is no genuine exchange of views. These reservations are important to keep in mind when considering the possibility of generalising rindings from such an experiment.

Our experiment was carried out in the autumn of 1996 on a representative sample of the Norwegian population. The experiment was done by telephone dealt with the question of how much tax should be paid by old age pensioners. The Norwegian tax system treats old age pensioners favourably with economically active persons. The political discussion concerning old age pensioners revolves around whether the advantages enjoyed by old age pensioners under the tax system should be abolished. One political argument for raising taxes paid by old age pensioners is that it would ease the economic burden upon workers in relation to pensioners. If old age pensioners paid more tax they would appear to help finance the increase in public expenditure resulting from the growing share of elderly in the population.

All the respondents in the sample were asked to take a view on the alternatives of (1) lower tax for old age pensioners than for economically active persons and (2) equal tax for old age pensioners and the economically active.³ The first alternative is in favour of old age pensioners, the second alternative in their disfavour. Those who had agreed with the first alternative, alternative in favour of old age pensioners, received a counterargument emphasised the interests of the economically active: 'Do you stick to this view even if it may mean that the economically active will have to pay higher tax than at present?' Those who agreed with the second alternative, the alternative in disfavour of old age pensioners, received a counter-argument that emphasised the interests of pensioners: 'Do you stick to this view even if it may mean that old age pensioners will attain a lower net income than at present?'⁴

First we report how many on each side changed position when confronted with the counter-argument. In the second part of the analysis we study which segments of the population are easiest to persuade, based on a model developed by Sniderman et al. (1991b).

Which Standpoint Was It Easiest to Persuade Respondents to Abandon?

The aggregated results of the experiment are summed up in Figures 1,2,3
and 4.5 Figure 1 shows the distribution of replies to the initial question.

Figure 4 shows the distribution after both groups have heard a counterargument.
Figure 1 and Figure 4 gives the net result of the
persuasion.

Sixty percent supported at the outset the alternative of equal tax for old age pensioners and workers, 40 percent the alternative that old age pensioners pay less tax. In other words a fairly large majority supported at the outset the alternative entailing a reform of old age pensioner taxation, which is in the old age pensioners' disfavour. Figure 2 and Figure 3 illustrate which standpoint it was easiest to persuade respondents to abandon. In Figure 2 we have added the respondents who were persuaded to leave the standpoint of lower tax for pensioners (35 percent of those who at the outset took this view) to the respondents who at the outset were of the view that old age pensioners and the economically active should pay equal tax. Figure 3, on the other hånd, shows the sum of those initially holding the standpoint of lower tax for old age pensioners and those who were persuaded to leave the standpoint of equal tax for old age pensioners and the economically active (44 percent of those who at the outset took this standpoint). As can be seen from the figures, it was easiest to persuade the respondents to abandon the standpoint of equal tax for old age pensioners and the economically active; the standpoint in disfavour of the old age pensioners.⁶

After both groups have heard a counter-argument, 52 percent support the lower-tax alternative, while 48 percent advocate equal tax. Although the change from initial to final distribution is not a dramatic one, it does show that the majority standpoint has shifted as a result of the persuasion attempt. At the outset the majority was in favour of old age pensioners paying the same tax as workers. In the final distribution, on the other hand, there is a majority, albeit a narrow one, in favour of old age pensioners being taxed more favourably.

The findings are interesting in relation to earlier studies. First, the experiment that even in such a key issue as elderly policy a relatively large portion of the population can be persuaded to leave their standpoint when confronted with a counter-argument. With the reservation that the respondents this experiment have been exposed to very selective information, it indicates that public opinion is a fragile entity. This is in line with the picture Converse (1964) gave of mass opinion based on panel data from conventional opinion surveys. On the other hand, the findings support the number of opinion studies showing that elderly and elderly policy programmes sympathy in the population. There turned out to be greater potential for persuasion in emphasising the interests of the pensioners than of the economically active. A greater number were moved by the argument that the elderly may become worse off financially than by the argument that the economically active will pay higher tax.

Which Respondents Are Persuaded to Change Standpoint: A Model of Persuasibility

The counter-argument technique is quasi-experimental (Cook & Campbell 1979). The counter-argument that respondents encounter depends on the side of the issue they select. Contrary to an ideal experiment, the participants differ systematically on other variables than the treatment given in the experiment. We are not fully able to isolate the effect of the persuasive argument from other variables that may influence change of opinion. The treatment itself is also ambiguous. The effect of the persuasive argument can be traced back either to the content of the argument in particular or to the general situation of being counter-argued. The inventors of the technique, et al. (1991b), have formulated a model of persuasibility which to a certain extent allows us to sort out the relative importance of the content of the argument.

They maintain that there are generally two main reasons why individuals change their opinion when confronted with a counter-argument. This applies not only in political issues, but in all situations where an attempt is made by argument to persuade someone to abandon their point of view. Individuals change their standpoint either because they are convinced by the counterargument question, or because they are persuaded to do so by counterarguments general. Sniderman et al. (1991b) characterise the first explanation 'topic bound' and the second as 'topic free'.

The distinction is crucial for whom we expect to be persuaded to abandon a position. The issue-specific explanation is linked to the issue the respondents are asked to take a position on. This does not apply to the general explanation. We would expect the same factors to have a bearing on whether a person can be persuaded to abandon their standpoint regardless of which issue the person in question takes a position on. The issue-specific explanation presupposes, in line with schema theory as dealt with in the introduction of this article, that the subject matter of the counter-argument puts people on the track of what considerations they should pursue when taking a position on a political issue. The counter-argument may appeal to the individual's interests, his or her political values or another consideration. explanations whose significance we are examining must consequently justified in terms of their connection with the issue. Figure 5 illustrates this reasoning.

We have used the counter-argument technique to investigate whether highlighting interests of workers and pensioners manages to mobilise a conflict interests between the two groups. The issue-specific hypothesis is that pensioners and workers change position in accordance with their interests.

To pursue the general explanation, we investigate whether education
and political awareness have a bearing on change of position. There are

several reasons for expecting that persons with high education are more difficult to persuade than persons with low education. First, education is a measure of a person's cognitive competence (Sniderman et al. 1991a). Cognitively competent persons must be assumed to be better equipped to perceive dilemmas inherent in an issue than are cognitively less competent persons (Converse 1964; Zaller 1992). Counter-arguments probably have less effect on them, since they have anticipated them before they adopted their standpoint. Another reason why it may be difficult to persuade people with a high level of education is that they are generally more antiauthoritarianthan with a low level of education (Jenssen 1993). This may entail that they are less likely to yield when someone attempts to pressure them to change their view. Thirdly, education can also be a measureof awareness. Individuals with a high level of education will be more likely to be active in and well informed about politics (Martinussen 1993). This increases the possibility that the political standpoints they expresshave thought through and are in harmony with each other, and therefore are more difficult to budge.

However, education is a controversial measure of political awareness (Luskin 1987; Sniderman et al. 1991a; Zaller 1992). We therefore also make a more direct examination of the significance of political awareness for change of position. We investigate whether there is a connection between change of position and the individual's knowledge of politics. We expect persons with a wide knowledge of politics to be more difficult to persuade than persons with little knowledge of politics.

We will emphasise that the dependent variable in this study is change of opinion. We do not intend to explain the attitudes respondents hold before or after they have encountered a counter-argument. Neither do we set forth to examine all possible variables that may have a bearing on change of opinion. Our aim when it comes to studying why the respondents have changed their opinion is more limited: to see if old age pensioners and the economically active change opinion in accordance with their interests.

Did the Counter-Arguments Appeal to the Interests of Old Age Pensioners and the Economically Active?

Table 1 shows the share of the economically active and old age pensioners who changed their view after being confronted with a counter-argument.⁷ Contrary to our expectation, the old age pensioners show the strongest propensity to abandon the standpoint in favour of lower tax for old age pensioners. On the other hand, the pattern is in keeping with expectations in regard to the standpoint of equal tax for economically active and old age pensioners. Fifty-seven percent of the old age pensioners abandon this standpoint, 42 percent of the economically active. Hence where the standpoint equal tax is concerned, the relationship between workers and old age pensioners is in keeping with the hypothesis. The fact that the old age pensioners show a stronger propensity than the economically active to abandon a standpoint implying a financial loss for them may indicate that they take their own interests into account to a greater degree than do economically active persons. However, as the old age pensioners tend to abandon both standpoints, another and in our judgement equally credible explanation, is that old age pensioners in general are easier to influence than workers. The reason may be that they are less familiar with and interested in political issues. Studies have shown that the elderly are less interested in and participate less in political life than do younger age groups (Martinussen 1975). We will return to this reasoning later in the analysis.

Income has a bearing on how the economically active and the old age pensioners are affected by these alternatives. High-income groups among old age pensioners and the economically active have most to lose from raised taxes. For the old age pensioners an increase in pensioners' tax will not apply to old age pensioners on low incomes. As regards the economicallyactive, progressive tax system ensures that tax increases have the biggest impact on those on high incomes. Income may therefore have a bearing on how old age pensioners and workers react to the attempt at persuasion. We have tested this hypothesis, but found no support in the

data.8 Persons on low incomes were more likely to change their position
among both the old age pensioners and the economically active, and
independent of their initial standpoint.

We have linked the issue-specific explanation in the Sniderman et al. (1991b) model of persuasibility to schema theory. We expected that highlighting interests of workers and pensioners would persuade the two groups to follow their interests. This expectation received no empirical support. We must emphasise that our intention has not been to make an empirical test of schema theory as such. We used schema theory as a point of departure for investigating the potential for creating a conflict of interests between workers and old age pensioners. The reason why we did not find support for our expectation may be that narrow self-interest lacks persuasive in this area of public policy. This conclusion is contrary to the anticipated clash of interests between workers and pensioners, but it is in line with other studies of public opinion towards policy programmes for the elderly. Applying conventional survey data, Huddy (1989) found no evidence of generational conflict on old age issues in American society. Goul-Andersen (1993) found that in Denmark the economically active had a stronger propensity to favour increase in old age pensions than old age pensioners. Bay (1998) found no differences between the economically active and old age pensioners in their generosity towards welfare programmes the elderly in Norway.

The Issue-Specific Explanation versus the General Explanation: Regression Analysis⁹

The analysis has given little support to the hypothesis derived from the issue-specific explanation. Change of standpoint was not in keeping with the interests of the old age pensioners and the economically active. We conclude the empirical analysis with a regression analysis, where we include both the issue-specific and the general explanation variables.

Gender is included in the analysis as a control variable. Men have a value of 1, women 0. Age is included partly as a control variable and partly as a variable capturing the individual's interest in relation to old age pensioners' taxation. Education and political knowledge are derived from the general explanation; economic activity and income are derived from the issuespecific A high value on the education variable stands for a high level of education. A high value on the political knowledge variable stands for a wide knowledge of politics.¹⁰

To retain the entire sample the economic activity variable is split into
old age pensioners, non-economically active and disabled, and economically
The economically active are the reference group.

The bivariate analyses in Table 1 showed that old age pensioners had a greater propensity to change position than the economically active. The same applied to persons on low incomes. These differences do not appear in the regression analysis in Table 2. This indicates that the bivariate effects of these variables were primarily an expression of educational level or of political awareness. The reason why old age pensioners and persons on lowincome easier to persuade may be that they are not used to relating to political problems. A central idea of, among others, Converse (1964) is that the ability to sort and structure political issues is determined by cognitive and political competence.

Education, being a general explanation variable, has a significant effect in both groups. Regardless of standpoint, it proved easier to persuade people with lower rather than higher education. Among those who initially favoured equal tax for the economically active and old age pensioners, political knowledge also has a significant effect. If one pushes the argument a little to the extreme, greater political sophistication was required to maintain a standpoint in old age pensioners' disfavour than to maintain a standpoint in old age pensioners' favour. This difference in effect can be interpreted in two ways. Inasmuch as the question of old age pensioners' tax is so complex, it is conceivable that those with little political knowledge have not realised that a system of equal tax for old age pensioners and the economically active will entail lower net incomes for old age pensioners. Upon being told this, they revise their position. Another, more subtle explanation is that it takes a certain sureness and confidence in one's own views to stick to a position that someone tells you is in the elderly's disfavour.The

favour.Thefaet that there is a broad consensus both in political circles and in the population at large that the government has a responsibility for the welfare of the elderly may make it uncomfortable to be associated with such a standpoint (even though the person in question may not at the outset have seen any contradiction between being positive towards the elderly and being in favour of equal tax for the economically active and old age pensioners).

Summary and Discussion

In this article we have studied how readily people can be persuaded to leave a political position when confronted with a counter-argument. The counter-argument technique is developed to enable opinion research to grasp the dynamic between political argumentation and public opinion. It is based on the theoretical assumption that a person's opinions have to be understood as an interplay between individual predispositions and an informational The objective of the technique is to have the survey interview selectively mimic conversation, by creating a situation where the respondent's opinion evokes a reaction from another. The analysis revealed that a substantial share, altogether 41 percent of the entire sample, changed their view after being confronted with a counter-argument. However, when considering this number it is important to keep in mind that the information given in the experiment is rather selective compared with real-life discussions.

However, we have not primarily been interested in the total share of persuaded respondents. We wanted to know which standpoint showed itself easiest to persuade people to leave, and which respondents changed opinion. The fact that it was easier to persuade persons who had taken a standpoint that was in favour of the elderly than persons who had taken a standpoint in the elderly's disfavour supports earlier studies showing widespread sympathy for the elderly and their needs. The much mentioned of interests between workers and pensioners is not revealed in the data. The economically active did not show a greater propensity to abandon a standpoint in favour of old age pensioners or vice versa.

As the counter-argument technique is in an initial stage within opinion research, there are not many comparable studies. We have adopted the technique from Sniderman et al. (1991b). They used it to study the stability of people's attitudes to public programmes targeted at the black section of the population in the US. First they asked the entire sample if the governmentshould spending on programmes to help blacks. Respondents that favoured raised transfers were given a follow-up question asking whether they would stick to this view even if it meant that people would get

special treatment just because they were black. Respondents that went against raised transfers were given a follow-up question asking whether they would stick to this view even if it meant that black people would be poorer and more frequently jobless than white people. As in our experiment, although the attempt at persuasion caused many on both sides to abandon their position, the aggregated differences between the initial and the final distribution were not wide. However, also in the US experiment the minoritystandpoint the majority standpoint (Sniderman et al. 1991b. 229). Initially the majority favoured increased spending. After the counterargumentsthe went against increased spending.

The effect of our persuasion attempt shares another feature with the result from the study by Sniderman et al. (1991b). In both experiments it was easier to persuade respondents with lower rather than higher education. The findings indicate that the less educated a person is, the more susceptible he or she is to political persuasion. This is in line with studies using other measures of opinion stability (Converse 1964), and is an explanation relevant test out in future studies applying the counter-argument technique.

We also examined the effect of political awareness, here measured as knowledge about the political system. It turned out that this variable had a bearing only on changes in standpoint in favour of equal tax for old age pensioners and the economically active, i.e. the standpoint that was in old age pensioners' disfavour. The lower the political knowledge, the easier it was to move the respondent away from this position. Since we have only addressed one delimited issue, we are cautious not to generalise on the basis of this experiment. However, we have indicated that this result may suggest that it takes greater insight and confidence in one's own views to maintain a position that is counter to perceived popular political attitudes than a position in keeping with that of a perceived popular standpoint.

In line with schema theory, we expected that accentuating the interests of old age pensioners and the economically active would activate the interests these groups as a heuristic. This hypothesis received no empirical support. Old age pensioners and the economically active did not change standpoint in accordance with their interests. Sniderman et al. (1991b), on the other hand, found strong support for their issue-specific explanation. They studied the effect of ideological orientation - between conservatives and liberals (Sniderman et al. 1991b, 232). To be able to draw general conclusions about the relevance of the heuristic perspective, more studies of persuasibility are needed.

---

NOTES

1. This type of experiment should be kept separate from methodological experiments for which there are long traditions in opinion research (e.g. Bishop et al. 1978; Schuman & Presser 1981). Methodological experiments are concerned with how the method used to measure public opinion may colour the replies received. The purpose is to arrive at sound measuring instruments. The experimental tradition we refer to here is targeted at the substantive content of the questions; whether a particular problem formulation induces 'different attitudes' from another problem formulation.

2. Sniderman & Grob (1996) make a distinction between postdecisional and predecisional interventions in survey-based experiments. The counter-argument technique is postdecisional. It occurs after a respondent has chosen a position. The split-half method is an example of predecisional intervention. This involves splitting a sample into two or more groups. Respondents in the groups are asked the same question, but accompanied by different information (Piazza et al. 1990; Sniderman et a1.1991c). The purpose is to create a different frame of reference for the same problem to examine if this has any significance for people's opinion and for their thinking on the problem.

3. 'Views vary on how much tax should be paid by old age pensioners. Which of the following assertions do you agree with most? (1) Old age pensioners should pay less tax than the economically active, even if they are on the same income. (2) Old age pensioners and the economically active with the same income should pay equal tax.'

4. Sniderman et al. (1991b) characterise this experimental technique as the counterargument It is intended to capture the number of respondents who are persuaded counter-arguments to change standpoint. Sniderman et al. make an important clarification as regards the term 'persuade'. Persuasion may give the impression that someone has been convinced that a particular standpoint, and not another, is correct. We have no grounds for saying that this is the case basedon the present experiment. The term is only intended to express that a person verbally changes his or her mind. We have spoken of people having changed their minds and meant by this that they say they have changed them. Saying that you have changed your position on an issue need not be the same as changing it. But it is precisely getting people to say that they have changed their mind that is the chief aim of political argumentation. (Sniderman et al. 1991a, 236)

5. This way of presenting the results is taken from Sniderman et al. 1991b.

6. The difference between the two groups when it comes to change of standpoint is significant at 99 percent level.

7. The economically active are defined as persons who work more than ten hours per week, and who do not have an old age pension as their main source of income.

8. The share of the economically active and the old age pensioners who changed standpoint, by income (percent); 1996 (N in parentheses): Initially for lowcr tax Initially for equal tax Low High Low High income income Pvalue income income Pvalue I coiiomically active 36 23 0.005 45 35 0.005 (116) (36) (232) (101) Pensioners 39 43 0.720 59 48 0.324 (25) (9) (38) (13)

9. The literature points to problems associated with using linear regression analysis for dichotomous dependent variables. One of the assumptions underlying linear regression analysis is that errors for the observations are unsystematic and that residuals therefore have a normal distribution. This assumption cannot be fulfilled in the case of dichotomous variables. The residuals can assume only two values and cannot therefore have a normal distribution. If the dependent variable has a skewed distribution, can also be obtained producing probabilities that exceed 1 or that are less than 0, in other words impossible results (Aldrich & Nelson 1986; Sørensenl9B9). Logistic regression is therefore recommended as an alternative to linear regression analysis. The dependent variables here are dichotomous. but do not have a markedly skewed distribution. To check the analysis results we have performed the same analysis using logistic regression. This analysis did not change the picture produced by linear regression analysis. The direction of the significant relations is identical and a significant is shown by the same variables.

---

9. The literature points to problems associated with using linear regression analysis for dichotomous dependent variables. One of the assumptions underlying linear regression analysis is that errors for the observations are unsystematic and that residuals therefore have a normal distribution. This assumption cannot be fulfilled in the case of dichotomous variables. The residuals can assume only two values and cannot therefore have a normal distribution. If the dependent variable has a skewed distribution, can also be obtained producing probabilities that exceed 1 or that are less than 0, in other words impossible results (Aldrich & Nelson 1986; Sørensenl9B9). Logistic regression is therefore recommended as an alternative to linear regression analysis. The dependent variables here are dichotomous. but do not have a markedly skewed distribution. To check the analysis results we have performed the same analysis using logistic regression. This analysis did not change the picture produced by linear regression analysis. The direction of the significant relations is identical and a significant is shown by the same variables.

10. The education variable has four values - elementary school, college I. college 11, postcollege / university - and is treated as continuous. In this study political awareness is measured by the individual's knowledge of the political system. Knowledge can be said to be an objective measure of political awareness. It gauges a person's involvement in politics by how much he or she has absorbed political information (Zaller 1992). The question measuring political knowledge is as follows: Governmental responsibility in this country is split between state. county and municipality. It may be difficult to keep track o[ how various tasks are shared between the three levels of government. I am now going to mention some governmental and for each task I want you to sav where you think the main responsibility lies. Is it the state, county or municipality which has the main responsibility for a. Running primary and lower-secondary schools. b. Running upper-secondary schools. c. Home-helps. d. Disbursement of unemployment benefit. e. Disbursement of social assistance. f. Running old people's and nursing homes. g. Disbursement of disability benefit. h. Granting liquor licences. An index was constructed by means of the count command in SPSS (Statistical Package for the Social Sciences). The index ran from 0 to 8. The mean is 5.77 and the standard deviation is 1.60.

ACKNOWLEDGEMENTS

The author would like to thank Ola Listhaug. Margaret Ford and three anonymous referees
{or valuable comments on earlier drafts.

REFERENCES

Achen, C. H. 1975. 'Mass Political Attitudes and the Survey Response". American Political
Science Review 69, 1218-31.

Aldrich, J. H. & Nelson, F. 1986. Linear Probability. Logit ami Probit Models. Quantitative
Applications in the Social Sciences 45. Newbury Park, CA: Sage.

Bay, A.-H. 1998. Opinion og eldrepolitikken (Mass opinion and public policy towards elderly).
Rapport 24/98. Dissertation. Oslo: NOVA.

Bishop, G. F., Tuchfarber, A. J. & Oldendick, R. W. 1978. 'Change in the Structure of
American Political Attitudes: The Nagging Question of Question Wording", American
Journal of Political Science 22, 250-69.

Colbjørnsen, T., Fennefoss, A. & Hernes, G. 1985. Så samles vi på valen . . . (Then comrades
come rally . . .). Oslo: Forskningsstiftelsen FAFO.

Converse, P. E. 1964. 'The Nature of Belief Systems in Mass Publics", in D. E. Apter. ed..
Ideology and Discontent. New York: Free Press.

Cook, T. D. & Campbell, D. T. 1979. Quasi-Experimentation: Design and Analysis. Issues
for Field Settings. Chicago: Rand McNally.

Feldman, S. 1995. Political Judgment. Structure and Process. Ann Arbor: University of
Michigan Press.

Fiske. S. T. & Taylor, S. E. 1991. Social Cognition. New York: McGraw-Hill.

Goul Andersen, J. 1993. "Sources of Welfare-State Support in Denmark: Self-Interest or
Way of Life', in Hansen, E. J. et al, eds, Welfare Trends in Scandinavian Countries. New
York: M. E. Sharpe.

Green, D. P., Kahneman, D. & Kunreuther, H. 1994. 'How the Scope and Method of Public
Funding Affect Willingness to Pay for Public Goods', Public Opinion Quarterly 58, 49-67.
Hellevik, O. 1997. 'Valgkampens vinnere og tapere', Tidens Tegn 8, 29-34.

Hippe, J. 1990. Mye vil ha mer. En analyse av formuesfordeling og familieoverføringer (Have
much, want more. An analysis of wealth distribution and family transfers). Rapport 101.
Oslo: FAFO.

Huddy, L. 1989. Generational Agreement on Old-Age Policies: Explanations Based on Realistic
Interests, Symbolic Political Attitudes, and Age Identities. Dissertation. Los Angeles:
University of California.

Iyengar, S. 1990. 'Framing Responsibility for Political Isuues: The Case of Poverty', Political
Behavior, 12(1), 19-39.

Jenssen, A. T. 1993. Verdivalg. Ny massepolitikk i Norge (Policy crossroads. New mass
politics in Norway). Oslo: Ad Notam. Gyldendal.

Johnson, P., Conrad, C. & Thomson, D., eds. 1989. Workers versus Pensioners. IntergenerationalJustice
an Ageing World. Manchester: Manchester University Press.

Judd, C. M. & Milburn, M. A. 1980. 'The Structure of Attitude Systems in the General
Public: Comparisons of a Structural Equation Model', American Sociological Review 45,
627-43.

Kinder, D. R. & Palfrey, T. R. 1993. 'On Behalf of an Experimental Political Science', in
Kinder, D. R. & Palfrey, T. R., eds, Experimental Foundations of Political Science. Ann
Arbor: University of Michigan Press.

Lau, R. R. & Sears, D. 0., eds. 1986. Political Cognition. Hillsdale, NJ: Lawrence Erlbaum
Associates.

Light, P. C. 1988. Baby Boomers. New York: W. W. Norton.

Luskin, R. C. 1987. 'Measuring Political Sophistication', American Journal of Political Science
31,856-99.

Martinussen, W. 1975. 'Politiske skillelinjer og politisk deltakelse' (Political dividing lines
and political participation), in Ramsøy, N. R. & Vaa, M., eds, Det norske samfunn
(Norwegian society). Oslo: Gyldendal Norsk Forlag.

Martinussen, W. 1993. Velferdsfellesskap. Medvirkning og solidaritet i nærmiljø, yrkesliv og
politikk (Welfare community. Participation and solidarity in local environment, at the work
place and in politics). Trondheim: Allforsk AVH, Senter for samfunnsforskning.

McConahay, J. B. 1973. 'Experimental Research', in Knutson, J. N., ed., Handbook of Political
Psychology. San Fransisco: Jossey-Bass.

McGraw, K. M. & Lodge, M. 1995. 'Introduction', in Lodge, M. & McGraw, K. M., eds,
Political jJudgment. Structure and Process. Ann Arbor: University of Michigan Press.

Nelson, T. E., Clawson, R. A. & Oxley, Z. M. 1997. 'Media Framing of a Civil Liberties
Conflict and Its Effect on Tolerance', American Political Science Review 91(3), 567-83.

Pettersen, P. A. 1995. 'The Welfare State: The Security Dimension', in Borre, O. & Scarbrough,
E., eds, The Scope of Government. Oxford: Oxford University Press.

Piazza, T., Sniderman, P. M. & Tetlock, P. E. 1990. 'Analysis of the Dynamics of Political
Reasoning: A General Purpose Computer-Assisted Methodology', in Stimson, J., ed.,
Political Analysis, Vol. 1. Ann Arbor: University of Michigan Press.

Schumann, H. & Presser, S. 1981. Questions and Answers. New York: Academic Press.

Sniderman, P. M., Brody, R. A. & Tetlock, P. E. 1991a. 'Introduction: Major Themes', in
Sniderman, P. M., Brody, R. A. & Tetlock, P. E., eds, Reasoning and Choice. Explorations
in Political Psychology. Cambridge: Cambridge University Press.

Sniderman, P. M. & Grob, D. B. 1996. 'Innovations in Experimental Design in Attitude
Surveys', Annual Review of Sociology 22, 377-99.

Sniderman, P. M., Piazza, T. & Kendrick, A. 1991b. 'Ideology and Issue Persuability: Dynamics of Racial Policy Attitudes', in Sniderman, P. M., Brody, R. A. & Tetlock, P. E., eds, Reasoning and Choice. Explorations in Political Psychology. Cambridge: Cambridge University Press.

Sniderman, P. M., Piazza, T., Tetlock, P. E. & Fjeld, P. J. 1991c. 'The American Dilemma: The Role of Law as a Persuasive Symbol,' in Sniderman, P. M., Brody, R. A. & Tetlock, P. E., eds, Reasoning and Choice. Explorations in Political Psychology. Cambridge: Cambridge University Press.

Svallfors, S. 1989. Vein dlskar vdlfdrdsstaten? Attityder, organiserade intressen och svensk
vålfårdspolitik 1969-1993 (Who loves the welfare state? Mass opinion, organised interests
and the Swedish welfare policy). Lund: Arkiv avhandlingsserie, 30.

Sørensen, R. 1992. 'Logitmodellens Analyse av diskret avhengig variabil', Tidsskrift for
samfunnsforskning 30, 61-86.

Thomson, D. 1989. 'The Welfare State and Generational Conflict: Winners and Losers', in
Johnson, P., Conrad, C. & Thomson, D., eds, Workers versus Pensioners. Intergenerational
Justice in an Ageing World. Manchester: Manchester University Press.

Tocqueville, A., de, 1835, 1840, 1945. Democracy in America, ed. P. Bradley. New York:
Knopf.

Zaller, J. R. 1992. The Nature and Origins of Mass Opinion. Cambridge: Cambridge University
Press.

Zaller, J. R. & Feldman, S. 1992. 'A Simple Theory of the Survey Response: Answering
Questions versus Revealing Preferences', American Journal of Political Science 36, 579-616.