Time of Vote Choice in India

From Lokniti s Data Archive Time of Vote Choice in India Shreyas Sardesai 1 Jyoti Mishra 2 Studies in Indian Politics 5(1) 82 91 2017 Lokniti, Centre for the Study of Developing Societies SAGE Publications sagepub.in/home.nav DOI: 10.1177/2321023017698269 http://inp.sagepub.com Abstract This piece analyzes the subject Time of Vote Choice in India, a hitherto under-researched aspect of voting behaviour in the country. Using longitudinal data from National Election Studies conducted by Lokniti between 1996 and 2014, it shows that voting intentions of India s voters are increasingly crystallizing before the start of the campaign period of an election. On the other hand, the proportion of voters postponing their vote choice to after the start of a campaign has declined over the last two decades. This declining trend is unique to India as most Western democracies have registered an overtime increase in campaign-period decision-making. While Indian voters who are less political, less partisan and less exposed to media are more likely to be late deciders than early ones, nevertheless, we also find that late deciders both campaign-period and last-minute have gradually become more politically sophisticated and attentive over the years. In this respect, our findings validate both the traditional floating voter theory of the Columbia School and some of the newer studies that have countered it. Keywords Vote choice, TOVC, NES, last minute, before campaign, during campaign, late deciders, early deciders, pre-campaign, campaign period, voting day, national, state, elections Introduction In the study of voting behaviour in India, the subject of people s vote choice has attracted the greatest amount of attention from scholars. The bulk of research in this field thus far pertains to the political Over the last two decades, Lokniti has conducted large-sized sample surveys in India during every national election as well as nearly all state elections. These surveys have attempted to study voters political choices, their voting behaviour and their opinions on a whole host of political issues. The consistency of the exercise has resulted in the creation of a vast archive of data on Indian democracy. While several scholars of Indian politics have referred to these from time to time for their own research needs, an adequate exploration of this rich archive with a view to inform and ideate about trends in Indian politics is lacking. From this issue onwards, this section of Studies in Indian Politics attempts to fill this gap by bringing to its readers short analytical pieces that make use/sense of Lokniti s longitudinal survey data sets. The idea is to present empirical insights and research ideas emanating from the Lokniti archives to students and scholars of politics. 1 Research Associate, Lokniti, Centre for the Study of Developing Societies, Delhi, India. 2 Research Officer, Lokniti, Centre for the Study of Developing Societies, Delhi, India. Corresponding author: Shreyas Sardesai, 29, Rajpur Road, Civil Lines, Delhi 110054, India. E-mail: shreyas.sardesai@gmail.com

Sardesai and Mishra 83 preferences of voters (who are people voting for) and the socio-demographic determinants of these preferences (what decides their vote or why do they vote for the parties or candidates they do). However, relatively little attention has been given to the issue of Time of Vote Choice (TOVC from hereon); that is, when do people make up their mind about who they are going to vote for early or late and why do they make their vote choices when they do? Also, is there any overtime trend with respect to TOVC? For instance, is the proportion of late deciders in Indian elections increasing or decreasing with each passing election or is there no clear pattern? Furthermore, who are the people who decide their vote early and who are the voters who do so at the eleventh hour? Is there also a significant chunk that neither decide early nor late but instead during the campaign period? These are questions that no scholar of Indian elections has attempted to answer despite the fact that data that will make these answers possible has been available for a long time. The standard TOVC question ( When did you make up your mind about whom to vote for ) after all has been asked consistently in post-poll National Election Studies (NES) conducted by Lokniti-CSDS during every Lok Sabha election since 1996 and in post-poll studies conducted by it during most state assembly elections in the same period. 3 Yet, no serious exploration of the question has been undertaken. The aim of this article is to address this significant lacuna in voting behaviour research in India by making use of this data. We also believe that apart from its importance to the academic realm, such knowledge on this aspect may also have practical implications for political actors and campaign strategists, who may benefit from knowing which type of voter is likely to make voting decisions at specific times and for what reasons. Trends of TOVC in India An analysis of TOVC data from NES conducted during Lok Sabha or national elections since 1996 indicates a clear trend towards early decision-making by voters in India. Whereas in the three national elections held in the late 1990s (1996, 1998 and 1999), last-minute (LM) decision-making (those who decided either on voting day or just a few days before it) outnumbered before-campaign (BC) decision-making (those who decided their vote choice before the commencement of the campaign), the last three elections (2004, 2009 and 2014) have seen the pattern getting reversed (Figure 1). A far greater proportion of voters in these elections decided their vote choice before campaigning started than they did last minute. Precampaign TOVC, in fact, has been rising steadily in India over the last two decades. In the 1996 election, only a quarter of all voters had decided their vote choice before the campaign. By 1998 and 1999, early deciders comprised about a third of the electorate. In the 2004 and 2014 elections, nearly half the voters were pre-campaign deciders. In between, in 2009, there was a slight dip in the proportion of pre-campaign deciders, with a sizeable chunk of voters deciding their vote choice during campaign (DC). However, even in 2009, the share of early deciders was higher than the elections of the late 1990s. Thus, broadly speaking, early choice making in India during national elections has been a rising trend. On the other hand, the trend with respect to last-minute crystallization of vote choice has been opposite. The proportion of voters deciding their vote choice very late has consistently declined with each successive national election held since mid-1990s. In 1996, nearly three in five voters fell in this category. By 1999, their proportion declined to two in five, and in 2009 and 2014, such voters comprised only a quarter of all voters. Even if we were to disaggregate this category of LM deciders into two separate categories, namely, voting day deciders and eve of voting day deciders, a declining trend is still seen (Figure 2). The trend also stays 3 For more information on surveys conducted by Lokniti and the methodologies adopted, please refer to its website www.lokniti. org.

84 Studies in Indian Politics 5(1) Figure 1. TOVC in India s National Elections Note: Answer categories for the TOVC question were not uniform across all years and differed slightly. In 1996, 1998 and 1999, LM includes those who decided on election day and a few days before voting. In 2004, 2009 and 2014, LM includes those who decided on election day and a day or two before voting. In 1996, 1998 and 1999, DC includes those who decided after candidates were declared. In 2004, 2009 and 2014, DC includes those who decided during campaign. In all years, BC includes those who decided before the start of the campaign. Figure 2. Last-minute TOVC the same even when we treat late deciders as an aggregation of last-minute deciders and campaignperiod deciders (Figure 3). These findings also mean that the Indian voter seems to be behaving differently from voters in Western democracies with respect to TOVC. In the West, overtime data indicates that more and more citizens are deciding their vote choice after the start of the campaign. In the USA, for instance, the percentage of post-convention deciders (late deciders) was 33 per cent in 2008 as opposed to 28 per cent in 1948 (Box- Steffensmeier, Dillard, Kimball & Massengill, 2015). In the most recent Presidential election of 2016,

Sardesai and Mishra 85 Figure 3. Late versus Early Deciders this figure was about 38 per cent according to the exit polls (CNN.com, 2016). 4 In Germany, it rose from 5 per cent in 1965 to 40 per cent in 2009 (Schmitt-Beck & Partheymuller, 2012) and in the Netherlands, from 10 per cent in 1971 to 45 per cent in 2006 (Irwin & Holsteyn, 2008). Similar increases have been witnessed in Australia and Britain (McAllister, 2000). Meanwhile, in their study of 12 countries, Dalton, McAllister and Wattenberg (2000) also found the phenomenon of late decision-making having gone up in 11 countries. One of the reasons for Indian voters to be behaving differently from their Western counterparts could be a further strengthening of partisan commitment among them. While we do not have concrete evidences to support this supposition, there is however some data that at least indicate no weakening of the partisan attachment to have taken place in India. In 1996, 28 per cent of the voters reported feeling close to a party. In subsequent years, the figure either went up or remained the same. Never has it shown a decline below 28 per cent. While people may have shifted their loyalties from one party to another, nevertheless, their degree of attachment to whichever party they support has remained steady, on the whole. Significantly, not only is the overtime shift away from an LM TOVC true for India as a whole but also the sub-national level shows a similar trend to a large extent. When we disaggregated the NES data statewise for three national elections 2004, 2009 and 2014 (for other elections, the sample size was not large enough to be broken down) we found LM TOVC to be consistently on the decline in 16 out of 26 states (61.5 per cent). An increase in it was seen in only three states with the remaining seven states showing no clear pattern. These TOVC trends in states during Lok Sabha elections, however, do not get replicated in the Vidhan Sabha or state assembly elections. Our analysis of TOVC data from three or more successive assembly elections of 17 states found that out of 10 states that showed a declining LM TOVC during national elections, only three showed the same pattern during state elections. These are 4 However, if we were to focus only on proportion and not on the overtime trend, we find that India reports lower figures of precampaign deciders (47 per cent in 2014) when compared to the USA, where, in the 2016 presidential election, 62 per cent of voters decided their vote at least two months before the voting day (CNN.com, 2016). Similarly, while only 12 per cent of American voters decided in the last week or last few days of the 2016 election, in India, the proportion of the eleventh hour deciders was 27 per cent in the 2014 election.

86 Studies in Indian Politics 5(1) Assam, Tamil Nadu and West Bengal. The remaining states either showed a rise (Uttar Pradesh and Haryana) or a fluctuation (Punjab, Uttarakhand, Himachal Pradesh, Maharashtra and Madhya Pradesh) in LM TOVC. Moreover, none of the three states that showed an increasing LM TOVC in national elections Delhi, Bihar and Rajasthan also showed a similar upward trend in their assembly elections. This finding is in consonance with a recent study by Simon Willocq (2016) of national, municipal and regional elections held in Canada, France, Germany and Spain, in which he found that in each of the countries, vote intentions tended to crystallize earlier in national elections than in local and regional elections. To digress a bit, while most states show divergent overtime trends with respect to the LM TOVC during national and state elections, many do show a pattern when we only take into account the consistency of an LM TOVC. For instance, if we keep 20 per cent as our parameter, seven states Bihar, Chhattisgarh, Delhi, Jharkhand, Madhya Pradesh, Maharashtra and Rajasthan can be classified as strong late deciding states because in three successive national elections and three consecutive state elections held in them, the proportion of LM vote deciders has consistently been over 20 per cent. Three states Gujarat, Tamil Nadu and Uttar Pradesh can be categorized as moderately strong late deciding states because over one out five voters in these states decided their vote at the last minute during three successive national and two consecutive state elections or vice versa. Who Decides Late and Early in India Returning to the broader national picture, even as the trends of TOVC in India do not match with the Western trend, there are however a great deal of similarities between Western studies and our findings for India when it comes to the characteristics of late and early deciders. Table 1 shows that in election after election, LM deciders in India have been less likely to be found among partisans or those who feel close to a particular party. They are also more likely to be found among those voters who give importance to the candidate rather than the party. Conversely, those who vote on the basis of party rather than candidate are more likely to decide before the campaign. Eleventh-hour deciders are also more prone to be found among those who are less interested in politics and election campaigns (Table 2). Table 1. TOVC by Partisanship (%) Feel close to/like a party Don t feel close to/like a party Voted on basis of party Voted on basis of candidate Voted on basis of caste 1996 1999 2004 2009 2014 LM DC BC LM DC BC LM DC BC LM DC BC LM DC BC 48 14 32 33 23 42 27 14 56 21 25 46 21 18 55 61 10 20 45 16 31 37 16 43 30 25 35 30 18 43 36 15 45 28 14 56 25 25 42 23 18 52 42 31 25 33 17 48 31 29 33 33 18 43 47 19 23 41 13 41 Note: Figures for 1998 are not available.

Sardesai and Mishra 87 Table 2. TOVC by Political Sophistication (%) 1996 1999 2004 2009 2014 LM DC BC LM DC BC LM DC BC LM DC BC LM DC BC Interest in politics High 42 18 35 24 12 63 24 22 49 Some 55 16 25 29 16 53 26 27 39 None 60 9 22 35 14 47 30 25 35 Awareness about politics High 48 18 29 32 25 39 28 15 55 25 24 43 25 19 48 Moderate 60 9 24 44 17 32 33 13 51 28 27 37 35 16 38 None 63 7 19 49 10 32 36 14 44 29 25 25 31 12 43 Interest in campaign High 46 19 31 30 23 46 22 12 64 22 23 49 Some 57 15 23 34 25 38 30 18 51 25 27 40 None 59 8 23 46 16 32 35 13 48 30 25 35 Consultation while voting Consulted no one 55 12 26 29 14 55 23 26 43 24 18 52 Consulted someone 62 11 18 37 16 43 33 25 33 33 19 40 Level of Media Exposure High 47 16 34 30 28 40 24 14 61 21 24 47 26 17 49 Moderate 57 14 23 35 19 43 30 14 54 25 26 41 25 20 49 Low 57 12 24 40 18 37 33 16 49 27 26 38 27 18 47 None 62 8 19 50 12 27 38 13 43 33 25 29 32 16 39 Notes: *Two awareness questions in each year were used to construct an awareness index. Awareness questions dealt with knowledge about major political events and leadership. Figures for 1998 are not available for any of the independent variables. Pre-campaign deciders on the other hand are more likely to be found among those more interested in politics and election campaigns. They are also more likely to be drawn from the ranks of those who are more aware about political issues, more exposed to the media and are more independent in deciding who they are going to vote for. Meanwhile, the in-between category of campaign-period deciders presents a mixed picture. While they exhibited traits similar to those of pre-campaign deciders in the late 1990s, for the last decade or so they have been showing traits of both pre-campaign and LM deciders, hence showing no clear pattern. These findings are more or less in alignment with the traditional floating voter theory of the Columbia School (Berelson, Lazarsfeld & McPhee, 1954; Lazarsfeld, Berelson & Gaudet, 1948) and the party identification theory of the Michigan School (Campbell, Converse, Miller & Stokes, 1960). At the same time, we also came across evidence that validates some of the more recent Western studies (Chaffee & Choe, 1980; Dalton, 2013; McAllister, 2000; Whitney & Goldman, 1985) that have found many of the late deciders to be getting more politically sophisticated and paying greater attention to the campaign. For instance, while in the 1996 Lok Sabha elections, 39 per cent of LM deciders and 55 per cent of campaign-period deciders were exposed to the news media; by 2014, the figures had jumped to 65 and 71 per cent respectively. Knowledge levels about politics among the two categories have also gone up. Whereas 57 per cent of LM deciders in 1996 were politically aware; in 2014, 84 per cent of them were

88 Studies in Indian Politics 5(1) found to be aware. Similarly, knowledge levels among campaign-period deciders went up from 77 to 90 per cent between 1996 and 2014. With respect to interest in campaign, the trend is not as neat, however. While LM deciders showed a jump from 34 to 37 per cent between 1996 and 2009 5 in their interest levels, campaign-period deciders showed a decline from 54 to 43 per cent in the same period. What is however significant is that be it 1996 or 2009 or 2014, political sophistication and attentiveness to the campaign among campaign-period deciders have always been higher than LM deciders. Our analysis of the NES data also reveals relationships between several socio-demographic variables and TOVC (Table 3). In India, women are more likely to postpone their vote choice than men. During all national elections since 1996, a greater proportion of women than men have decided their vote choice just on the eve of voting. Men on the other hand have always shown a greater propensity than women to decide earlier be it during the campaign or before it. Although the gap between the two genders across all three categories LM, DC and BC has narrowed over the years, the pattern has remained unchanged. In terms of locality, voters living in rural areas have consistently tended to decide last minute, much more than voters living in urban areas who are mostly early deciders. Age does not show much of a pattern. Younger voters were slightly more likely to decide at the eleventh hour than older voters in 1996 and 1998; however, this pattern reversed in 1999, and in the elections held thereafter, there has not been much of a difference between the young and the old with respect to TOVC. Education however impacts TOVC in a major way. The less educated have always tended to delay their voting decision whereas the more educated have been more likely to decide in advance. This too validates the political sophistication theory. An equally neat pattern is witnessed when we examine TOVC by the economic class. The poorer and lower classes have for the last two decades shown a higher proclivity towards deciding last minute whereas the rich and the middle classes have been more prone to deciding during campaign or earlier than that. A reflection of the class findings can also be seen to some extent in the findings by Hindu castes and communities. Hindu Adivasis or STs, a sizeable proportion of who are poor, have always been more likely to delay their TOVC than other Hindu communities. Hindu upper castes, the most prosperous of all, meanwhile showed the opposite tendency. Class and caste do not seem to neatly overlap, however, with respect to Dalit TOVC. Dalits, also among the poorest, are found to be less likely than the OBCs to decide last minute. Instead Dalits, like the upper castes, are more prone to arriving at a decision pre-campaign. In terms of religious backgrounds, we find that except for the 1996 election, Christian voters have almost always been more likely to decide their vote much in advance as compared to voters belonging to other religious communities. Since 2004, Sikh voters have also shown a similar tendency. Meanwhile, during the same period (2004 2014), Hindus and Muslims have been more likely to decide last minute than others. Moreover, the two communities have usually mirrored each other with respect to TOVC. Finally, our analysis of TOVC by party choice finds that voters of the Left parties have always locked their vote much earlier compared to voters of the BJP alliance or the Congress alliance. Barring 2004 and 2014, those voting for the BSP have mostly tended to decide at the eleventh hour rather than much in advance. Conclusion and Discussion Our article has attempted to analyze the time of people s vote choice in India, a hitherto under-studied aspect of voting behaviour in the country. We have shown that voting intentions of India s voters are 5 Figures for 2014 are not available.

Sardesai and Mishra 89 Table 3. TOVC by Socio-demographics 1996 1998 1999 2004 2009 2014 LM DC BC LM DC BC LM DC BC LM DC BC LM DC BC LM DC BC Gender Men 53 14 27 52 14 31 37 22 38 29 15 53 26 27 39 26 19 48 Women 61 9 20 57 10 27 45 14 32 35 14 46 29 23 36 28 17 45 Age Over 50 years 55 11 25 53 12 30 43 15 34 31 13 51 27 23 40 28 17 46 41 50 56 12 24 55 11 28 40 18 36 31 14 51 26 25 39 26 18 47 31 40 57 11 24 54 13 28 40 20 34 32 16 50 28 26 37 26 19 48 23 30 58 12 22 55 12 30 40 20 36 33 15 50 27 26 37 28 18 47 18 22 60 11 22 56 10 30 39 18 39 34 15 49 29 25 35 28 19 46 Locality Urban 51 14 29 46 12 39 35 21 39 29 15 53 25 26 39 24 17 51 Rural 59 11 22 57 12 26 42 18 34 33 14 49 28 25 37 29 18 45 Education Graduate 49 20 28 43 19 36 32 27 39 24 16 59 21 26 42 24 18 51 Up to matric 52 14 28 51 15 32 34 23 40 30 16 52 26 27 39 28 19 46 Up to primary 57 13 24 55 10 31 42 18 36 33 15 50 29 24 38 26 19 47 Non-literate 62 7 19 60 9 24 46 14 31 37 13 45 31 23 34 30 16 43 Class Rich 44 19 30 36 24 35 26 15 56 21 26 44 20 19 53 Middle 55 13 26 40 22 34 31 16 51 24 27 40 28 18 47 Lower 56 11 24 42 18 34 34 15 48 27 25 38 27 18 46 Poor 61 9 19 42 15 37 36 13 47 30 24 35 27 17 46 Caste Upper caste 53 13 27 47 13 35 38 23 35 30 16 52 25 25 40 26 17 50 OBC 60 10 21 57 12 28 42 19 34 35 13 48 29 25 37 28 17 45 SC 59 10 22 55 11 28 38 14 39 34 13 50 28 25 37 28 16 48 ST 73 7 13 61 10 21 50 12 29 35 15 46 32 25 31 31 21 35 Religion Hindu 58 11 23 55 12 29 40 18 36 33 14 50 28 25 37 28 18 46 Muslim 52 14 30 56 11 31 43 18 33 31 16 49 28 26 35 27 18 48 Christian 54 11 27 50 15 34 41 19 38 24 16 57 19 20 53 16 15 59 Sikh 44 24 11 55 17 24 42 15 19 20 18 57 10 24 57 14 29 51 Party vote Cong+ voters 55 10 28 56 11 30 40 15 39 34 14 50 26 23 42 30 17 45 BJP+ voters 58 11 25 58 11 27 40 19 37 33 14 52 31 27 33 30 18 45 Left 48 11 33 39 14 40 30 17 47 20 13 65 16 30 47 14 24 54 BSP voters 56 10 24 54 12 29 49 21 20 32 11 50 37 22 30 27 10 57 Others 61 14 17 53 12 30 47 27 21 34 19 44 27 26 38 26 17 50 increasingly crystallizing before the start of the campaign period of an election. Conversely, the proportion of voters postponing their vote choice to after the start of a campaign has declined over the last two decades. This declining trend is unique to India as most Western democracies have registered an overtime increase in the campaign-period decision-making. Our findings validate both the traditional floating

90 Studies in Indian Politics 5(1) voter theory of the Columbia School and some of the newer studies that have countered it. While Indian voters who are less political, less partisan and less exposed to media are more likely to be late deciders than early, nevertheless, we also find that late deciders both campaign-period and LM have gradually become more politically sophisticated and attentive over the years. We believe that our results have two paradoxical implications for political actors and their campaigning strategies. First, the lessening share of late deciders in the electorate means that there are fewer voters for them to target during the peak campaigning period and the fact that they are more likely to be drawn from those with low exposure to news media, low levels of education and low interest in politics makes them more difficult to reach to and influence. At the same time, their proneness to be apartisan, coupled with the fact that they have over the years become more media-exposed and attentive, make them promising subjects for campaign effects, which in a close election can prove to be crucial. Second, since voters in India are increasingly firming up their vote before the start of the campaign, parties may be tempted to start getting their message across to the voters several months before the voting day. However, they may not find this to be feasible since early deciders are more likely to be found among partisans, hence making them least susceptible to campaign influence and messaging. We believe that we have only scratched the surface on the subject of TOVC and that there is scope for more rigorous research on it. The theories of cross-pressure and ambivalence, for instance, are definitely worth testing. It has been argued that voters who delay their timing of vote choice do so because of identity and opinion-related cross-pressures experienced by them that pull them in different political directions (Berelson et al., 1954; Lazarsfeld et al., 1948). It has also been asserted that those torn between their party identification and their candidate evaluation decide late (Campbell et al., 1960). Related to this early cross-pressure theory are the recent concepts of internal and external ambivalence (Lavine, 2001; Mutz, 2002; Nir, 2005). A voter is said to be internally ambivalent when he or she has inconsistent political attitudes or when he or she perceives several parties or candidates or competing considerations as being similarly attractive. External ambivalence refers to a situation in which a voter s multiple crosscutting social identities make him or her encounter contradictory political messages within his or her social environment. Both aspects of ambivalence have been shown to be positively associated with latedecision making in the West. Young researchers may want to dig deeper into the NES data and explore if this holds true for India as well. References Berelson, Bernard R., Lazarsfeld, Paul F., & McPhee, William N. (1954). Voting. Chicago: University of Chicago Press. Box-Steffensmeier, Janet, Dillard, Micah, Kimball, David, & Massengill, William. (2015). The long and short of it: The unpredictability of late deciding voters. Electoral Studies, 39(September), 181 194. Campbell, Angus, Converse, Philip, Miller, Warren, & Stokes, Donald. (1960). The American voter. New York: Wiley. Chaffee, Steven, & Choe, Sun. (1980). Time of decision and media use during the Ford-Carter campaign. Public Opinion Quarterly 44(1), 53 69. CNN.com. (2016). Exit polls: 62% decided vote before September. Retrieved 15 December 2016, from http:// edition.cnn.com/videos/politics/2016/11/08/first-exit-polls-hillary-clinton-donald-trump-tsr.cnn Dalton, Russell. (2013). The apartisan American: Dealignment and changing electoral politics. Washington, DC: CQ Press. Dalton, Russell, McAllister, Ian, & Wattenberg, Martin. (2000). The consequences of partisan dealignment. In Russell Dalton & Martin Wattenberg (Eds), Parties without partisans (pp. 37 63). Oxford: Oxford University Press. Irwin, Galen, & Van Holsteyn, Joop. (2008). What are they waiting for? Strategic information for late deciding voters. International Journal of Public Opinion Research, 20(4), 483 493.

Sardesai and Mishra 91 Lavine, Howard. (2001). The electoral consequences of ambivalence toward presidential candidates. The American Journal of Political Science, 45(4), 915 929. Lazarsfeld, Paul F., Berelson, Bernard R., & Gaudet, Hazel. (1948). The people s choice: How the voter makes up his mind in a presidential campaign (2nd ed.). New York: Columbia University Press. McAllister, Ian. (2000). Rational or capricious? Late deciding voters in Australia, Britain and the United States. A paper prepared for the workshop Do Campaigns Matter? 28th Joint Sessions, European Consortium for Political Research, Copenhagen, 14 19 April 2000. Mutz, Diana. (2002). The consequences of cross-cutting networks for political participation. The American Journal of Political Science, 46(4), 838 855. Nir, Lilach. (2005). Ambivalent social networks and their consequences for participation. International Journal of Public Opinion Research, 17(4), 422 442. Schmitt-Beck, Rüdiger, & Partheymüller, Julia. (2012). Why voters decide late: A simultaneous test of old and new hypotheses at the 2005 and 2009 German federal elections. German Politics, 21(3), 299 316. Whitney, Charles, & Goldman, Steven. (1985). Media use and time of vote decision: A study of the 1980 presidential election. Communication Research, 12(4), 511 529. Willocq, Simon. (2016). Why do they take longer to make up their mind? A comparative study on late deciding voters in western democracies. Paper presented at EPOP Conference University of Kent, 9 11 September 2016.