The Dog that Didn t Bark: The Dog that Didn t Bark: Would turnout of non-voters make a difference? Would turnout of non-voters make a difference?

Transcription

1 Revised 27 March 2004 The Dog that Didn t Bark: Would turnout of non-voters make a difference? Per Arnt Pettersen and Lawrence E. Rose The Dog that Didn t Bark: Would turnout of non-voters make a difference? Per Arnt Pettersen and Lawrence E. Rose Department of Social Science Bodø Regional University N-8049 Bodø, NORWAY tel: (+47) fax: (+47) per.arnt.pettersen@hibo.no Department of Political Science, University of Oslo P.O Blindern, N-0317 Oslo, NORWAY tel: (+47) fax: (+47) l.e.rose@stv.uio.no Introduction having fun with numbers It is one of those cases where the art of the reasoner should be used rather for the sifting of details than for the acquiring of fresh evidence. The difficulty is to detach the framework of fact of absolute, undeniable fact from the embellishments of theorists and reporters. Then, having established ourselves upon this sound basis, it is our duty to see what inferences may be drawn, and which are the special points upon which the whole mystery turns. Sherlock Holmes to Watson in The Adventure of the Silver Blaze by Sir Arthur Conan Doyle Asking if turnout of non-voters would make a difference is a counterfactual and purely hypothetical question, but this does not prevent political pundits and laymen alike from posing the question. Not only is it intrinsically interesting for psephologists and a source of speculative curiosity for informed observers, it is a matter of concern from a normative perspective, and even more critically so for political parties, for which higher turnout could spell the difference between victory or defeat (see, for example, Bennett & Resnick 1990; Crew et al. 1977; Lijphart 1997; McAllister & Mughan 1986; Petrocik 1987; and Piven & Cloward 1988). Thus, despite the fact that the question cannot be given any totally reliable answer, it continues to attract attention and debate to which we wish to add some additional observations. In doing so, we join the discussion in the spirit of those who acknowledge that it is possible to have fun playing with numbers. Paper prepared for delivery in Workshop 9, Low turnout does it matter? at the ECPR Joint Session of Workshops, Uppsala, Sweden, April 2004 The paper begins with a short review of the principal approaches to the question which have been used previously. After this we sketch out an alternative approach. This approach is then employed for purposes of analyzing six different parliamentary elections held in Norway between 1981 and The paper concludes with a brief summary discussion. 1

2 Traditional approaches to the potential impact of higher voter turnout Answers given to the hypothetical question of what impact increased voter participation would have on electoral outcomes depend at least in part upon the approaches and methods used to make estimates or informed guesses. Among the most common approaches are those using materials from electoral surveys based on large and representative samples. With this as a point of departure, a straightforward way to solve the brainteaser seems to be to ask those who admit not having taken part in the election a follow up question about what party they would have voted for if they had voted. This is a classical approach employed by Campbell and his associates in some of the earliest electoral studies in the U.S. (Campbell et al. 1960: ). Adding these hypothetical votes to the real votes then permits an assessment of whether the election outcome would have changed e.g. resulted in a different president or government, allowed some parties to pass the minimum threshold for representation, or altered the majority of representatives in parliament, congress or local government assemblies. What seems to be a fairly straightforward approach, however, has several serious hitches: - First, a question about how a person would have voted obviously has to be asked after the election as well as after a question about if the respondent had actually taken part in the election. Presuming that respondents know the election result, this opens for obvious bandwagon effects of support for the winner among those who really did not care much about the election and hence did not vote. Typically the conclusion would then be that nothing would have changed. If anything, the victory would only have been greater for the winner(s). - Second, it is difficult to assess the probability of which individuals among the non-voters would with a certain likelihood actually vote, and who would be likely to stay at home under any circumstances. The problem, in other words, is to identify those individuals among the non-voters who may with a reasonable certainty be considered genuine hypothetical voters as opposed to die hard stay-at-homers. - Third, according to the tenets of cross-pressure theory, some individuals are abstainers precisely because they can not make up their mind about what party to choose. But in a survey setting they are invited to solve the uncertainties that were unsolvable at the time of the actual election. Aside from the ethical problems that arise in connection with asking people to make such a choice, the reliability of the answers is likely to be low since the cross-pressure has not vanished, and the pick of party may therefore be as unreliable as a toss of the coin. Another approach to the question of what impact increased voter participation might have is found where panel data are available. In these instances one can arrive at an estimate by assuming that voters who became non-voters at the subsequent election would have voted for the same party as they did last time if they had voted. Adding these hypothetical votes to the real votes again offers an indication of whether the election outcome would have changed. This is an approach employed, for example, by Butler and Stokes in their study of political changes in Britain (Butler & Stokes 1971:339). While this approach avoids the pitfalls of ex post facto bandwagon effects, there is still a problem of making an assessment of which (or how many) non-voters are genuinely potential party supporters as opposed to people who would be likely to stay at home on the sofa in any event. Among other much used approaches to the study of the hypothetical impact of non-voters, one is to investigate which parties have the largest proportion of identifiers among the nonvoters. Here the presumption is that it is these parties that would have the most to gain from increased voter mobilization (cf. Abramson et al. 1982:90, Flanigan & Zingale 1994:44). Another approach is to investigate the socio-demographic composition of non-voters under an assumption that it is those of low socio-economic status who participate least and that such individuals exhibit a tendency to vote more for non-conservative parties. In this case the analytical question is what would have happened if low socio-economic individuals had participated at the same level as those with high socio-economic status (cf. Bjørklund 2001:81; Lijphart 1997:2-5; Piven & Cloward 1988:21-25; Wolfinger & Rosenstone 1980:80-88). The most systematic studies of hypothetical participation of non-voters, however, are those carried out by Paul Abramson and his associates. At every presidential election in the USA since 1980 they have estimated the effect of increased participation among non-voters. Their approach is based on a comparison of the frequency of voting between identifiers of the Republican and Democratic parties. The conclusion drawn is that most of the time increased turnout would not change anything. But in the close election of 2000 they find that strong 2 3

3 identifiers with the Democrats and independents leaning towards the Democrats tended to stay at home more frequently than strong identifiers with the Republicans and independents leaning towards the Republicans (Abramson et al. 2002:92). The conclusion in this instance was that If strong Democrats had been as likely to vote as strong Republicans, and if independents who leaned Democratic had been as likely to vote as independents who leaned Republican, and if Democrats in these to groups had been as likely to vote for Gore as strong Democrats and independents who leaned democratic and did vote, Gore s overall share of the vote overall share of the vote would have increased by 1.8 percentage points. Depending upon in which states these increased votes were cast, even this small increase could have provided Gore with an electoral vote majority (Abramson et al. 2002:93 emphasis added). Notice the assumption, which we have emphasized, that the non-voters would have to behave identically with those having similar sympathies who actually did vote, and that the probability for voting is identical for sympathizers of the two parties. However, no effort is made to validate this assumption. An alternative approach to the potential impact of higher voter turnout An alternative, albeit a bit more indirect, approach to the hypothetical impact of non-voters is one akin to that followed by Flanigan and Zingale, but with an attempt to provide more precise, better grounded estimates of the likelihood of voting among non-voters i.e. estimates based on a more differentiated assessment of the mobilization potential found among the group(s) of non-voters. The basic idea is to investigate the proportion of supporters for different parties among the non-voters in other words to establish an estimation of the number (or volume) of immobilized potential voters for different parties. With this information if one could with some certainty estimate a reasonable probability for actual participation among these supporters, it would then be possible to calculate the extra votes that were lost due to the lack of voter mobilization of the respective parties and in this fashion assess the potential impact of higher turnout. Our approach follows this latter logic, but does so by focusing upon partisan sympathy thermometers commonly used in electoral surveys rather than party identification. The advantage with this approach is that the question of sympathy is asked independent of any questions about voting and the answer is given without any reference to participation in the election. Equally if not more important, whereas an individual is generally only able to choose one party at an election, individuals (both voters and non-voters) can theoretically have positive sympathy (or antipathy) with respect to more than one party. Under these conditions bandwagon effects are minimized if not completely eliminated. Note, however, that under these circumstances the crucial question is at what point such sympathy is likely to be transformed into actual voting for a party. Presumably the more sympathy an individual expresses regarding a party, the higher the probability of voting for that party. Hence, the critical analytical challenge is to find the level of sympathy that transforms sympathy into voting with a certain degree of probability. One way of meeting this challenge it by studying the sympathy level of actual voters for various parties found through various election surveys and then use this information to calculate more realistic estimates of the consequences of a higher level of mobilization among non-voters. This approach has the advantage of providing contextually sensitive, party specific estimates about the mobilization potential for each party and the probable impact on the electoral outcome in the event sympathizers among the nonvoters had displayed the same propensity to vote for a specific party as had sympathizers among those who did vote. With this starting point, the approach we adopt can be described in four steps: 1) The first step involves inspecting the sympathy scores for those who actually voted for different parties to investigate if there is any (common) pattern of sympathy that seems to suggest a high likelihood of transforming sympathy into actual votes. For Norwegian voters and parties an intriguing picture emerges, in that a 70 degree positive sympathy score seems to be a significant threshold with respect to transforming sympathy into actual voting. Over time results indicate that about 90 percent of those who actually vote for each and every party express a positive sympathy of 70 degrees or more. 2) However, lots of respondents (both voters and non-voters) have positive sympathy ratings of 70 degrees or more for more than one party, and the probability for choosing a specific party among those with 70 degrees or more for purposes of casting a ballot is unknown. Yet this can be estimated in several ways for example by probit regression, using sympathy scores to predict actual voting for the same party (or even inspecting percent tables since this is a bivariate analysis). This probability among party sympathizers to vote for a party will vary from party to party and from election to election. 4 5

4 3) The third step is crucial inasmuch as it is assumed that the probability for transforming a given level of sympathy into a vote for a specific party is the same among non-voters as it is among actual voters. So if a certain proportion of individuals among the non-voters has a positive sympathy of say 70 degrees or more for a specific party, given the probability estimates calculated in step two above it is possible to calculate how many of the nonvoters have a reasonable chance of actually voting for the party in question had they in fact cast a ballot. 4) Since no party can get less than the actual votes cast on election day, using these estimates every party will increase its electoral support, even if by only a small fraction for some parties. But just adding to the real percentage of votes could lead to a total support for all parties of more than 100 percent of the electorate. This means that when calculating changes in the support for parties it is most appropriate to use actual votes cast for every party as found in official election statistics and add the number of new votes coming from non-voters for every party had the non-voters voted, and then recalculate the hypothetical outcome for purposes of comparison with the actual outcome. It is appropriate that we elaborate upon and illustrate these steps before turning to our results. nonetheless comparable across all years (see Appendix 1). The response scales used allow sympathy scores ranging from 0 to 100 to be established, where 0 represents a strong negative orientation, and 100 represents a strong positive orientation to the party in question. As can be seen in Figure 1, which is based on responses from the 1981 election survey, the likelihood of voting for a party is highly contingent upon the degree of sympathy an individual holds for a specific party. Rarely if ever are people with a negative orientation to a party likely to vote for that party. 1 Only when one has a more positive orientation does voting for a party become more likely, and this likelihood increases sharply once a certain sympathy threshold is passed. As Figure 1 indicates, this threshold (as well as the likelihood of voting for a given party at a specific level of sympathy) varies a bit from one party to another, but as a general tendency the likelihood of voting for a party only begins to increase significantly once a sympathy score of 70 has been registered. The only exceptions of any note relate to the Labour and Conservative parties, where a 60 degree sympathy score appears to be sufficient Figure 1. Percent of individuals voting for specific parties dependent upon different sympathy scores for the party in question, What level of sympathy is needed to turn sympathy into voting? The first question that needs to be answered is what level of sympathy is found among those who actually voted for each party. This knowledge is necessary if we are to argue that a certain amount of sympathy for a party is needed if a specific voter (and by the same logic any non-voter) should be inclined to vote for a specific party. It is reasonable to believe that this level of sympathy would be rather similar across all parties since it is hard to believe that many voters who feel indifferent towards a specific party or have negative feelings would vote for that party. However, the degree of positive feelings may of course vary from party to party. Obtaining this knowledge is a purely descriptive task, but it helps produce knowledge about sympathy among actual voters which shall be used to identify those non-voters that would be likely to vote for a specific party had they voted. Percent voting for party Soc. Left Labour Liberal Christ. P. Center Conserv. Progress Sympathy score Wording of the sympathy thermometer items used in the national election surveys carried out in Norway from 1981 to 2001 has varied slightly, but the essence of the question is 1 That there are exceptions, however, is not totally unreasonable, since there can be various reasons for such behavior tactical voting being one of them. 6 7

5 to increase the likelihood of actually voting for these parties. That it should be precisely these two parties which deviate from the main tendency is not surprising, however, since these are the two parties which for many decades have been the dominant actors along a left-right dimension found in the Norwegian political battlefield. Despite inter-party differences, the principal message of Figure 1 (and similar figures which could be presented for all other elections from 1985 to 2001) is that a positive sympathy score of 70 represents something of a general turning point or threshold with respect to the likelihood of voting for a party. Looking across all of the major parties and all elections in the 1981 to 2001 period, almost nine out of every ten individuals (89.3 percent) expressed a sympathy score of 70 or above for the party for which they voted. 2 Variation in this percentage, moreover, is limited. As results in Table 1 demonstrate, in only five instances does this percentage drop below 85 (the minimum being 83 percent for the Progress in Table 1. Proportion of voters for different parties expressing sympathy scores of 70 or more, 1981 to Percent. Year Soc. Left Labour Liberal Christ. P. Centre Conserv. Prog (N=) (61) (491) (58) (111) (93) (416) (57) (N=) (104) (679) (68) (174) (125) (545) (65) (N=) (222) (608) (77) (161) (109) (399) (206) (N=) (134) (668) (62) (123) (311) (276) (81) (N=) (118) (608) (84) (264) (133) (267) (192) (N=) (235) (354) (70) (215) (93) (428) (169) Average In our analyses we have limited our attention to the seven parties which have been the principal actors achieving parliamentary representation in all elections during the period. There are, of course, other parties that compete in national elections, but only rarely have they gained parliamentary representation, and the number of their supporters found in the national election surveys are so few as to make it difficult to establish reliable parameter estimates for these parties. 1989) and in only two instances does this exceed 95 percent (the maximum being 96.2 percent) for the Socialist Left in 1985). As furthermore is evident from the bottom row of Table 1, across all six elections the Socialist Left has had the highest average percentage of voters expressing a sympathy score of 70 or above (92.9 percent), whereas the old Agrarian, which is now known as the Centre, has the lowest average percentage (86.9). What proportion of non-voters sympathize with various parties? Using a sympathy score of 70 as a reference point, the table may be turned in order to ask what proportion of non-voters express a sympathy score of 70 degree or more for each of these seven major parties. The answer to this question is displayed in Table 2. First of all notice the sum of sympathies as displayed in the right-hand most column of the table. The fact that sympathy scores sum to more than 100 for all years serves to underline that non-voters, just as is the case for voters, tend to express sympathy for more than one party. This is a consideration that may reflect cross-pressure which in some instances may produce nonvoters. Table 2. Proportion of non-voters with a sympathy score of 70 or more for different parties, 1981 to Percent. Year Soc. Left Labour Liberal Christ. P. Centre Conserv. Prog. Sum (N=) (186) (188) (186) (187) (188) (188) (188) (N=) (238) (237) (238) (241) (238) (237) (237) (N=) (229) (234) (228) (231) (231) (234) (234) (N=) (350) (352) (342) (347) (347) (351) (352) (N=) (264) (278) (259) (277) (270) (272) (281) (N=) (335) (336) (333) (336) (331) (337) (337) Average

6 Look next at the average proportion of sympathizers among the non-voters, which is found in the last row of the table. Results here show that the Labour is by far the winner or perhaps one should more correctly say the loser! For all of the elections covered, an average of about 43 percent of non-voters express a sympathy of 70 degrees of more for the Labour. This percentage reaches a zenith in 1993 when just over 50 percent of the non-voters expressed a sympathy above 70 degrees or more, after which the percentage has declined, reaching a low of roughly 30 percent in Second best is the Conservative with an average of just over 30 percent of strong sympathizers among the non-voters. For all of the remaining parties the average percentage of non-voters expressing a sympathy of 70 degrees or more is substantially lower a little over 20 percent for the Left Socialists and Progress, roughly 17 percent for the Christian People s and Centre, and just under 10 percent for the Liberal. But the respective percentages displayed in Table 2 vary much more from one election to another than was the case for voters displayed in Table 1. What probability of voting for various parties does sympathy produce? These results suggest that partisan sympathies may be much more subject to temporally related factors among non-voters than is the case for voters, and that certain parties stand to gain more than others were they to be capable of exploiting these swings in sympathy among non-voters. But it is necessary to exercise caution in drawing such conclusions. The amount or level of sympathy expressed among non-voters is not enough information to say anything about the proportion of non-voters with a sympathy of 70 degree or more that might reasonably be expected to go to the polls and cast a ballot with minimal urging, thereby being in a position to add their votes to the actual number of voters for each party. The reason is that both voters and non-voters tend to have sympathy for more than one party (as we have already seen demonstrated in Table 2). We therefore also have to calculate the probability that a sympathy of 70 degrees or more really generates a vote for each party. This probability coefficient, moreover, may well vary for each party and between elections. To estimate this probability one may employ probit regression analysis with respect to the likelihood of voting for a specific party as this is contingent upon the degree of sympathy an individual has for the same party. Results from such regression analyses are displayed in Table 3. Table 3. Probit regression of voting for a specific party by sympathy for the same party. Probability increases. (Z-values in parentheses) * Year Soc. Left Labour Liberal Christ. P. Centre Conserv. Prog (11.53) (23.12) (10.84) (13.69) (11.91) (20.08) (10.09) (11.44) (27.69) (12.73) (16.87) (14.14) (24.58) (12.87) (15.40) (2945) (13.27) (16.80) (14.69) (23.57) (19.03) (13.59) (22.11) (11.20) (15.19) (19.74) (21.68) (14.17) (14.32) (23.63) (11.51) (16.33) (16.95) (21.32) (18.87) (17.0) (22.0) (12.8) (17.0) (15.5) (22.4) (16.6) Average * Minimum N = 1527, Maximum N = 2126 While the proportion of voters for each party with at sympathy level of 70 degrees or more varied little between parties, as the coefficients found in Table 3 readily demonstrate, there are major differences between parties regarding the probability that a sympathy of 70 degrees or more will lead to a vote for the different parties. On average the largest parties the Labour and the Conservatives have an advantage in this respect. Thus, over the period from 1981 to 2001 the average probability of voting for the Labour if a person first expresses a sympathy for the party of 70 degree or more was 54 percent (see the last row in Table 3). The probability was highest in 1985 (63 percent) and lowest in 1993 (46 percent). For the Conservative the comparable average was 49 percent, with the highest probability being in 1985 (nearly 56 percent) and the lowest in 1993 (44 percent). The contrast between these two parties and the Liberal is striking. Hence, although people may sympathize with the Liberal, there is on the average a probability of only a little over 20 percent that these good feelings for the party will translate into votes. In 1993 the probability coefficient is as low as.159. The Socialist Left party is in a situation rather similar to that of the Liberals, with an average probability of only roughly 25 percent. For the 10 11

7 other parties The Christian People s, the Centre and the Progress the average probability is around 30 percent. What is noteworthy with these parties is that all have been subject to significant electoral ups and downs in the course of recent years. Would higher electoral turnout have made a difference? Against this background, the question of importance is whether these swings in probabilities and/or the higher, more stable probabilities for the Labour and Conservative parties provide the basis for arguing that a higher level of voter turnout would have had an impact on the electoral outcome. To answer this question it is necessary to estimate what each party would probably gain should turnout indeed have been higher, and then compare the new distribution of the vote with the original distribution. To estimate this electoral add on for each party the following procedure is used. First the number of actual non-voters that it is possible to assume has a sympathy of 70 degrees or more for a given party is calculated. This is done by using the proportion of non-voters in the survey data found to have a sympathy of 70 degrees or more (from Table 2) and multiplying this by the total number of actual non-voters known from official electoral statistics. This pool of sympathetic non-voters is then multiplied by the probability of voting for the party given a sympathy of 70 degrees or more as this was estimated using survey data (from Table 3). Doing so yields a number of non-voters which can be added to those who actually voted for the respective party. Once this is done for all parties it is then possible to calculate a new (hypothetical) distribution of votes for purposes of comparison with the actual electoral outcome. 3 This procedure has been employed for all seven parties for all six election years. Before turning to the results, a few remarks about the Norwegian electoral system are in order, and some expectations about the hypothetical impact of higher levels of turnout may be offered. The Norwegian parliamentary electoral system is based on 19 electoral districts in which a modified Sainte-Laguë system of proportional representation is used in allocating seats. The largest electoral districts are the counties of Hordaland (17), Oslo (16) and Akershus (15), while the counties of Aust-Agder and Finnmark send only four representatives to the parliament. It can also be mentioned that Norwegian electoral law diverges from the one-man, 3 A more detailed example of this procedure based on data for 1981 is found in Appendix 2. one-vote principle such that there are far more voters behind every representative from the larger electoral districts than is the case in the smaller ones. But what is important is that only when the last representative from an electoral district is won by a small margin is it likely that the seat could (hypothetically) be lost if an alternative distribution of the vote showed either a drop in the proportion of votes for the party of the marginal seat or a growth in support for the party competing most closely for the last seat from that district. Within such a system, in which individual may like and have sympathy for more than one party (as well as dislike one or more parties) but are only able to vote for one, it is reasonable to expect that each and every party would experience a positive impact of increased voting by non-voters expressing sympathy of 70 degrees or more for at least one party. The impact on the overall electoral outcome, however, is more difficult to predict. At least three competing hypotheses may be formulated regarding the consequences of such increased support: One possibility is that the largest parties will be those which benefit most from the mobilization of sympathizers among non-voters. This expectation is predicated on the idea that non-voters, much as voters, are inclined to favor parties having the best prospects for being election winners. If so, then it should be the larger parties that have an extra non-mobilized potential among non-voters. An alternative and contrary hypothesis is that smaller parties would benefit most from a higher rate of mobilization among non-voting sympathizers. This expectation is based on the idea that some voters abstain, even if they like a party a lot, because there is little chance for the party especially small parties to achieve a significant position in government. For these people the usual choice is either to vote for another party or to abstain. But were they in fact to vote for the party of their first choice, this would benefit the smaller parties. Finally, a third hypothesis is that the parties farther out on the left-right axis stand to benefit most from increased mobilization of sympathizers. Here the underlying logic is that these people, perhaps even more so than supporters of small parties, do not find it likely that their preferred party will end up in a position to be a major shaper of public policy and therefore again either chose neighboring parties or abstain. But were they to 12 13

8 be mobilized on the basis of their first preferences, the parties farther out on the wings would stand to gain the most. Did the smaller parties stay small merely because they were not able to mobilize latent sympathy? Results In investigating these alternatives it would ideally be preferable to have a random sample of elections in order to eliminate or control for the effects of special circumstances surrounding any given election. But relevant data are only available from six elections in the period from 1981 to Fortunately, however, these elections have properties which nonetheless make them very suitable for our investigation, since the outcomes vary in an interesting way. Thus, 1981 is the election when the Conservative achieved its greatest success in the post war period, followed by 1985 were the Labour gained some of what was lost in Then in the 1989 election both of the largest parties the Labour and the Conservative lost quite substantially. After this the 1993 election is characterized by the conflict over Norwegian membership in the EU and votes for the opposition parties especially the Centre increased dramatically. In 1997 the Progressive not only had become larger than the Conservative, but was actually the largest of the non-socialist parties, and the Christian People s experienced their best election ever. Finally, the 2001 election was characterized by catastrophic results for the Labour, at least in light of their previous post war electoral record. There is, in other words, very nice variation in the electoral outcomes over the 20 year span in time under investigation. Let us therefore turn to these elections and look at the potential impact of increased mobilization of non-voters using the model of estimation outlined above in each case. 1981: Victory for the Conservatives The first election under study is that of 1981, the best election ever for the Conservative, which achieved support of over 30 percent of the voting electorate. However, even if the election was considered a defeat for the Labour, the party still received close to 40 percent of the vote. Note also that no other party achieved support above 10 percent. Three questions therefore seem obvious in connection with this election: (1) Had the Conservative fully mobilized and utilized its potential of sympathy in realizing the upsurge of Conservative support? (2) Did the Labour have an unused reservoir of sympathetic nonvoters which made them the genuine election loser compared to earlier electoral results? (3) Using information regarding sympathy scores, the probability estimates of voting among persons with a specific minimum level of sympathy and knowledge about the number of actual voters as well as non-voters, it is possible to calculate a new electoral result in which a certain number of non-voters are presumed to be mobilized and added to the votes of those who actually did vote in order to answer these questions. The results of doing this are displayed in Table 4. 4 This table suggests that three parties the Left Socialists, the Christian People s and the Conservative would have suffered a relative loss if non-voters had in fact participated at a higher rate. It is also noteworthy that the Conservatives seemed to have taken out what the party could (and indeed even a little more) of the potential support available among its sympathizers. Table 4. Difference between the actual and a hypothetical distribution of the vote based on an estimated electoral add on from non-voters who express a positive sympathy of 70 degrees to the party election. * Actual distribution of vote (percent) Hypothetical distribution of vote with electoral add on based on non-voters (percent) Difference between hypothetical and actual distributions Soc. Left Labour Liberal Christ. P Centre Conserv Progress (Absolute) Sum *Support is calculated only for the parties with lists in all electoral districts. 4 In Table 4 and subsequent tables for the parliamentary elections in the period from 1985 to 2001, the results are at the aggregate national level. In reality voting occurs in sub-national electoral districts according to a system of proportional representation using party lists. It is possible to argue that to be most robust, our analyses should be carried out at the district level and then aggregated, but to do so in a reliable fashion would require much larger sample sizes for the national surveys which we used in some of our estimation procedures

9 By comparison only one party would have enjoyed a gain worth mentioning had their sympathizers turned out in greater numbers in 1981, and that is the Labour which would have gained 0.6 percent according to the estimation procedure used here. But even if that gain is worth noticing, it is unlikely it would have made any difference for the election outcome. Only if the 0.6 percent had been concentrated in election districts which were very close and a Labour nominee was next in line for a seat in parliament might they have gained extra parliamentary mandates. Notice also that the total impact of the new distribution of votes had non-voters of a certain level of partisan sympathy been mobilized to the same degree as all other persons is only an absolute change of 1.4 percent, that is, the sum of differences between actual and estimated vote distribution, either positive or negative. If the entire difference had been related to only one party, it might have made a difference, but it would probably not have changed the election result in any major way. Table 5. Difference between the actual and a hypothetical distribution of the vote based on an estimated electoral add on from non-voters who express a positive sympathy of 70 degrees to the party election. * Actual distribution of vote (percent) Hypothetical distribution of vote with electoral add on based on non-voters (percent) Difference between hypothetical and actual distributions Soc. Left Labour Liberal Christ. P Centre Conserv Progress (Absolute) Sum *Support is calculated only for the parties with lists in all electoral districts. 1985: The Labour strikes back support, while on the left side of the political spectrum increased mobilization of non-voters would have made no difference. The 1985 election is similar to the previous one in several ways. Again only the Conservative and Labour parties have electoral support that exceeds 10 percent of the electorate. But in this case the Conservative has lost a little compared to 1981 (declining to about 30 percent) whereas the Labour has gained a little (to about 40 percent). All the other parties left, right or centrist have below 10 percent electoral support. At this election, therefore, the basic question is: Has there been any changes since the previous election regarding the impact of hypothetical participation from non-voters? The answer, provided by results displayed in Table 5, is that the total impact which was rather small in 1981 is even less. Across all parties, the sum of all (absolute) changes is no more than 0.7 percentage points. This is a figure which indicates that mobilizing non-voters would have had virtually no impact on the election outcome whatsoever. The only thing worth commenting upon is that the small gains that might have been registered are on the right side of the political spectrum, and would especially have implied a gain for the Conservative as well as a 0.1 percent gain for the Progressive. By contrast, all three of the centrist parties would presumably have suffered a relative loss of 0.1 percent in 1989: Both Labour and the Conservatives lose The 1989 election saw some major changes regarding election results. The Conservative surge from the early 1980 s is over. The party is basically back to its normal strength at around 20 percent. This decline, however, does not favour the Labour, which also lost ground and ended up at the level that became the standard for the 1990 s namely at about 35 percent. The winners of the election were rather the parties on the wings of both sides of the left-right political continuum. On the left the Left Socialist broke through the 10 percent level, and on the right the Progressive fared even better, achieving support from over 13 percent of the electorate. All the centrist parties, meanwhile, continued to stay below the 10 percent level. But what would have happened had non-voters to a greater degree been roused from their sofas? At this election increased mobilization would, according to our calculations, only provided a further boost to the election winners on each side of the political spectrum (see 16 17

10 Table 6. Difference between the actual and a hypothetical distribution of the vote based on an estimated electoral add on from non-voters who express a positive sympathy of 70 degrees to the party election. * Actual distribution of vote (percent) Hypothetical distribution of vote with electoral add on based on non-voters (percent) Difference between hypothetical and actual distributions Soc. Left Labour Liberal Christ. P Centre Conserv Progress (Absolute) Sum *Support is calculated only for the parties with lists in all electoral districts. Table 6). Yet the boost would again have been relatively modest (an increase of 0.3 and 0.4 percent respectively). On this occasion the losers would have been the Labour (-0.3 percent) and the Christian People s (-0.3 percent). At this level, it is unlikely that either winners or losers would have gained or lost any extra seats. It can also be observed that just as in 1981, the total impact on the distribution of support to various parties with a high level of mobilization would have been limited to a total of 1.4 percent change. Even if benefiting only one party, this is not a level of change that would have been likely to dramatically affect the electoral outcome 1993: The EU and the Centre The 1993 parliamentary election was, as noted previously, highly focused on the issue of Norwegian membership in the EU, which was to be decided upon by a national referendum the following year. The most strident opponents to membership were the Centre and the Left Socialist, where both leaders and followers were almost unanimously against membership. Among parties favoring membership only the Conservative had supporters that basically held the same view as their leaders. All other parties were divided both at the leadership level as well as among supporters. Most visible was the split inside the Labour, where the opponents to membership formally organized their own faction. By comparison, whereas the leadership of the Christian People s and the Liberal parties took a position against membership, a fair proportion of the supporters of these parties favored joining the EU. The leadership of the Progressive also held a pro-membership position, but knew that among activists, as well as ordinary supporters, there were a large proportion of opponents. At the election a year prior to the referendum the Progressive lost half of the voters it had attracted in The chairman Carl I. Hagen later expressed regret and considered the party s position on the EU-issue in 1993 to be one of the biggest mistakes the party ever made. Indeed, in subsequent years the Progressive has declined to take a stand on the EU-issue, leaving it up to the individual leader and voter if they wanted to support membership or not, since the issue will ultimately be decided by a national referendum. Compared to the 1989 parliamentary election, however, the major change to be observed in 1993 concerned the other non-socialist parties, especially the major electoral victory of the Centre which enjoyed an increase of about 10 percent in electoral support and suddenly became a prominent non-socialist party. On the left side of the political spectrum the Labour enjoyed increased support, primarily at the expense of both the Conservatives and the Progressive. But the Socialist Left was surprisingly unable to benefit from its opposition to EU membership. While the Centre was able to exploit the issue to a major victory, the Socialist Left actually lost a couple of percentage points compared to the election in Would this scenario a one issue election with clear winners and losers change the impact of increased mobilization of non-voters with sufficiently high levels of partisan sympathy on the election result? As the results presented in Table 7 reveal, the aggregate differences are larger than at the previous election (2.0 compared to 1.4), but still remarkably small considering individual parties. The Centre appears to have won about as much as it possibly could based on its anti-eu membership stance, and mobilization of sympathizers for the Conservative and Progressive parties would also have had marginal effects according to our estimates. Only the Labour seems to have been in a position to have benefited significantly from increased mobilization. The Christian People s, on the other hand, would have lost support by about the same amount had all sympathizers actually voted. Again 18 19

11 Table 7. Difference between the actual and a hypothetical distribution of the vote based on an estimated electoral add on from non-voters who express a positive sympathy of 70 degrees to the party election. * Table 8. Difference between the actual and a hypothetical distribution of the vote based on an estimated electoral add on from non-voters who express a positive sympathy of 70 degrees to the party election. * Actual distribution of vote (percent) Hypothetical distribution of vote with electoral add on based on non-voters (percent) Difference between hypothetical and actual distributions Soc. Left Labour Liberal Christ. P Centre Conserv Progress (Absolute) Sum *Support is calculated only for the parties with lists in all electoral districts. Actual distribution of vote (percent) Hypothetical distribution of vote with electoral add on based on non-voters (percent) Difference between hypothetical and actual distributions Soc. Left Labour Liberal Christ. P Centre Conserv Progress (Absolute) Sum *Support is calculated only for the parties with lists in all electoral districts. the result regarding numbers and distribution of parliamentary seats would have changed only under very specific circumstances. Losses and gains would have to have been related to election districts where a party only needed few extra votes to win or loss a seat. could have gained about one-half a percent, while the hypothetical biggest losers the Liberals and Conservatives would have only dropped in support by 0.2 percent each. 2001: Disaster for the Labour 1997: The Progressive becomes the second largest party The 1997 election was basically back to normal with only one exception the exception being that the Progressive had become the largest of the non-socialist parties. As a mirror image of the victory for the Progressives, the Conservative continued to lose among non-socialist voters. Among the centrist parties the Centre was back to normal with support of a little less than 10 percent of the electorate while the Christian People s was the victorious party, reaching nearly 15 percent support. Among the socialist parties, both Labour and Left Socialists, lost support compared to Looking at the potential effects of increased mobilization among sympathetic non-voters (see Table 8), there is absolutely nothing to write home about. The total change is only about one percentage point, and this change is so evenly distributed that it is highly unlikely that anything could have happened regarding seat changes. Together the two socialist parties During the period of post war social-democratic hegemony from 1945 to 1969 the Labour alone would carry well above 40 percent of the electorate. From the election of 1973 onward, however, this support dropped below the 40 percent level, the only exception being in During the 1990 s the party seemed to stabilize its support a little above 35 percent. But this changed dramatically at the election in 2001, where electoral support for the party plunged another 10 percent, to around the 25 percent mark. The Lefts Socialists gained some from voters leaving the Labour, but so also did the Conservative and the Christian People s (Aardal 2003:36). Together four parties were considered winners of the election: the Left Socialists accomplished their best election ever, the Conservatives regained its position as the leading party among the non-socialists, while the Christian People s and Progress parties consolidated their positions. The losers were in addition to the Labour already mentioned the Liberal, for which support declined to within a fraction of depriving the party of its parliamentary seats, and the Centre, which only two elections before had been dominant among the non-socialists

12 Concluding discussion Perhaps equally noteworthy was the fact that aggregate electoral turnout was again down (to 75.5 percent) by comparison with most previous elections in the post war period. What would have happened had sympathetic non-voters been mobilized to a greater degree? The answer is found in Table 9. Table 9. Difference between the actual and a hypothetical distribution of the vote based on an estimated electoral add on from non-voters who express a positive sympathy of 70 degrees to the party election. * Actual distribution of vote (percent) Hypothetical distribution of vote with electoral add on based on non-voters (percent) Difference between hypothetical and actual distributions Soc. Left Labour Liberal Christ. P Centre Conserv Progress (Absolute) Sum *Support is calculated only for the parties with lists in all electoral districts. The objective of this paper has been to estimate an alternative model with respect to the proportion of non-voters each party would receive in addition to their actual voter support under hypothetical circumstances where they were able to mobilize sympathizers among the ranks of the non-voters. In estimating this model we use expressions of partisan sympathy that are independent of actual voting behavior. The main theoretical assumption is that the relationship between expressed sympathy with a party (70 degrees or more on the sympathy thermometer in this instance) and the probability that this sympathy may be transformed into a vote for the party in question is the same for non-voters, were they in fact to vote, as it is for actual voters. Our expectation was that results predicted by such a model would be different from those predicted by most other approaches, where the conclusion is generally that nothing would have changed, except perhaps for the very close presidential election in the United States in Obviously in a two party system, where the winner takes all, every close election could be changed had extra votes been mobilized for the losing party. But systems of proportional representation such as that found in Norway are presumably less sensitive to an impact of increased voter mobilization. Even so, we had expected that our estimated model could potentially have generated three alternative results: - First, that the largest parties would benefit most from higher mobilization of Only in 1993 would the mobilization of sympathetic non-voters have had the same degree of impact as in 2001, yet the combined effect continues to be rather limited a total change of only two percentage points in absolute terms. But at least this time there is a certain pattern related to the hypothetical mobilization of non-voters: The winners would have done better and the big loser would have done even worse. The Left Socialists could have gained another half percent, and so could the Conservatives, while the Labour even at an election with at 10 percent drop in support could have lost even more (0.6 percent) if sympathetic non-voters of all parties had been mobilized. In other words, compared to other parties the Labour did not have any extra sympathy among non-voters that could have come to their rescue. Our calculations suggest a similarly pessimistic situation for the Liberals and the Centre since there is no additional gain to be found by higher levels of mobilization among those who preferred to stay at home on election day in sympathizers since there is reason to believe that non-voters just as voters tend to sympathize more with larger, frequently winning parties. - Alternatively, it is possible that the smaller parties would profit most since small parties normally have little chance of being represented in government. Precisely for this reason sympathizers of these parties may to a greater extent either vote for other parties or abstain a response that would imply noteworthy effects of higher electoral turnout. - Third, similar reasoning relates to the parties located on the wings of the left-right political spectrum. This is because to date these parties have never been invited to participate in government coalitions in Norway, and voting for these parties could therefore be considered a waste of ones vote, again leading to abstention. But should they nonetheless be mobilized, an impact would be registered