The Influence of Presumed media Influence on Strategic Voting

Communication Research http://crx.sagepub.com The Influence of Presumed Media Influence on Strategic Voting Jonathan Cohen and Yariv Tsfati Communication Research 2009; 36; 359 originally published online Apr 1, 2009; DOI: 10.1177/0093650209333026 The online version of this article can be found at: http://crx.sagepub.com/cgi/content/abstract/36/3/359 Published by: http://www.sagepublications.com Additional services and information for Communication Research can be found at: Email Alerts: http://crx.sagepub.com/cgi/alerts Subscriptions: http://crx.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav Citations http://crx.sagepub.com/cgi/content/refs/36/3/359 Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009  Communication Research Volume 36 Number 3 June 2009 359-378 © 2009 Sage Publications 10.1177/0093650209333026 The Influence of Presumed http://crx.sagepub.com hosted at Media Influence http://online.sagepub.com on Strategic Voting Jonathan Cohen Yariv Tsfati University of Haifa, Israel An increasingly influential line of research on media effects suggests that some of the effects of news media on society take place because people perceive media as influen- tial. In this article, the authors test this notion, in the context of voting decisions. The authors propose that voters’ perceptions regarding the influence of media will be related to their intention to vote strategically—that is, to vote for a party they favor less than their most preferred option. If news media are perceived to persuade other voters to switch their votes, it will more likely be necessary to switch one’s vote to either conform to or counterbalance the effects of media on others. Two studies, utilizing three data sets, collected in the context of the Israeli Knesset elections of 2003 and 2006 using different measures of strategic voting were used to test the hypothesis that perceptions of media influence on others will relate to strategic voting. The hypothesis was confirmed in both studies. Keywords:  presumed media influence; third-person effect; strategic voting; elections E lectoral choices, like other choices in life, are made within a social context and are affected by social norms and perceptions and by expectations of how others will behave. Such expectations are, in turn, based on information about public opin- ion, candidates, and past political behavior. Much of this social information is avail- able to voters through mass media that is simultaneously transmitted to them and to other voters. By assessing the nature of the political information provided by news media, and weighing it against their estimate of its impact on the behavior of others, voters can estimate expected shifts in public opinion and consequently calculate the impact of their own voting choices. Following this argument, we propose that among the factors shaping voting choices are beliefs about media influence. Specifically, using data from two Israeli elections, we test the proposition that voters’ beliefs about news media influence are related to their propensity to vote strategically. Strategic Voting Social information is important in many decision-making processes. In some cases, knowing how others are likely to behave is important for the purpose of adapting one’s 359 Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009 360   Communication Research behavior to conform to that of others. But information about others’ behaviors or inten- tions is also used to calculate how one’s actions will interact with what others will do and affects the chances of achieving one’s goals. This latter—coordination—function of social information is likely to be of primary importance especially when the behav- ior is instrumental rather than primarily expressive and when outcomes are dependent on how others behave (Cox, 1997). Sometimes, it is more valuable to avoid doing what others are likely to do (e.g., choosing a less crowded road during heavy traffic). In other cases, it may be more profitable to anticipate what others will do and preempt them (e.g., being an early adopter of interactive innovations; Rogers, 1995). Whether to conform to or to deviate from the norm, knowing what others are likely to do helps in calculating how one’s own decisions will impact a social outcome. In a political context, each voter’s single vote is counted among millions of others to determine the outcome of elections, and depending on the electoral system, votes for losing parties may become insignificant in forming the government. Research has found that while most voters vote sincerely (i.e., a vote for the party or candidate that is closest to their worldview and about whom they feel most favorably; Lanoue & Bowler, 1992), others do not. Sophisticated voters (Abramson, Aldrich, Paolino, & Rohde, 1992) or “strategic” voters (Blais & Nadeau, 1996; Downs, 1957) are those who base their choice not only on their opinion of parties and candidates but also on calculations of possible outcomes. The most likely calculation that motivates strategic voting is the probability of candidate viability; voters wish to avoid “wasting” their votes on a candidate who will not be able to represent them (Lanoue & Bowler, 1992). In parliamentary sys- tems, where people vote for one of many parties (with no one party having a major- ity), the largest of which must then agree to form a majority coalition with some smaller ones, voters may also wish to maximize their vote’s impact on the nature of the future coalition. A voter may vote for a large party rather than a preferred smaller one in order to influence which large party will form a coalition (Cox, 1997). Or, if a voter believes he or she knows who will win the election and consequently assem- ble the coalition, he or she may vote for a second-favored party in order to make it is a more attractive coalition partner (Benoit, Giannetti, & Laver, 2002). Both these strategies require strategic voters to have a sense of how other voters are likely to vote and of election results. For example, in closely contested elections, strategically voting for a smaller party is less likely (Burden, 2005), but when a large margin is expected and one’s favorite candidate is perceived to be a “sure winner” or “sure loser,” strategic voting is more likely. As Cox (1997) points out, politics in complex and large societies require exten- sive social coordination. Information about the behavior or intentions of others and their expected outcomes are necessary for strategic calculations, and media serve as primary sources of information allowing voters to assess the political climate. Voters can receive predictions from media, but they also follow public opinion polls, Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009  Cohen, Tsfati / Influence of Presumed Media Influence   361 political advertising, and coverage, and they estimate the effectiveness of these on how others will vote. Thus, beliefs about media impact should be related to beliefs about the behavior of other voters and consequently to voting decisions. The Influence of Presumed Media Influence People’s perceptions of media and their presumption of media influence have been shown to be biased, but they have also been found to be important. People perceive media coverage of issues they care about to be overly negative (Gunther & Liebhart, 2006), and people patently believe that they are less influenced by harmful media content (and more influenced by prosocial content) than others (Davison, 1983). But despite the inaccuracy of perceptions about media and their influence, recent research has found that these biased perceptions have very real consequences. This process, titled “the influence of presumed media influence” (Gunther & Storey, 2003), has been tested in a host of social and political contexts: People who believe media have stronger influence on their children tend to report stronger monitoring of harmful content (Hoffner & Buchanan, 2002), and Gaza settlers who estimated stronger media impact declared they view democracy as less legitimate and admitted they were more willing to engage in violent protest (Tsfati & Cohen, 2005). Taken together, this literature points out that people act upon their perceptions of media influence regardless of whether these perceptions are accurate. But why should the belief in media power affect voting patterns? Because media cover elections and if one wishes to coordinate his or her vote with the votes of oth- ers, then the degree to which media coverage is seen as potent should be a major determinant of one’s expectations of how others will vote. This is especially true in an age when media increasingly emphasize strategic information and personal cam- paign coverage (i.e., horse race coverage; Cappella & Jamieson, 1997). For example, if media are seen to favor a leading party either through a positive tone or through providing it with extensive and prominent coverage (especially as the campaign comes to an end) and this media support is seen as potentially effective, voters may be confident enough in the election results and hence shift their vote from this large party they believe is sure to win to their second-choice party: a smaller, preferred coalition partner or a smaller party they wish to see pass the election thresh- old. Conversely, if media are seen as favoring a smaller, feared party and as effec- tively persuading voters to increase support for this party, this may motivate a vote for a second-choice smaller party—rather than one of the large parties—in order to influence the makeup of the coalition. But if media are seen as inconsequential, then it will be hard to anticipate others’ votes and less logical to vote strategically. Though there are many potential scenarios, it would seem that in many cases, the belief in media influence is likely to shift perceptions of probable outcomes and make the tactics of voting more salient. Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009 362   Communication Research Hypothesis 1 (H1): The stronger a voter’s belief in media influence on the voting deci- sions of others, the more likely he or she will be to vote strategically. The Israeli Context Established in 1948 as a parliamentary democracy, Israel has a 120-member parliament, the Knesset, which is elected by a closed-list system of proportional representation with the entire country serving as one constituency. It is a multiparty system in a multicleavaged society (Lijphart, 1993), with an average number of parties in each Knesset that is usually no lower than a dozen (Rahat & Hazan, 2005). In the 2003 elections, 13 of 27 participating parties were successful in get- ting representatives elected (i.e., passed the legal threshold of 1.5% of the vote), and in the 2006 elections, 12 of 31 parties passed the newly raised threshold of 2%. From a comparative perspective, these election thresholds are very low (e.g., as compared to 5% in New Zealand and Germany [Weaver, 2002] or 10% in Turkey [Yavuz & Ozcan, 2006]). In combination with a system that has no regional repre- sentation, low thresholds provide incentives for very small parties to compete and thus create an extreme multiparty system that offers Israeli voters ample choice between parties with relatively similar platforms. On the other hand, since many parties do not pass the threshold, voters are also faced with the very real possibility of wasting their votes. The Elections of 2003 and 2006 Israelis went to the polls five times in national elections from 1996 to 2006. Three of these elections resulted in major power shifts between the traditional Right and Left block leaders: Likud and Labor. During this time, the electoral system changed twice, and many new parties entered parliament (immigrant parties, pensioners’ party, etc.). In 2003, the contenders for forming a coalition were Likud’s Ariel Sharon and Labor’s Amram Mitzna, and Likud enjoyed a significant lead in preelection polls. Following a rift within the Likud over the Gaza withdrawal, Sharon and several other prominent Likud leaders, joined by major figures from Labor and other par- ties (e.g., Shimon Peres), formed a new centrist party—Kadima. This party enjoyed a dramatic advantage over both Likud and Labor. Following Sharon’s sudden hospitalization before the 2006 elections, and his replacement by Ehud Olmert, Kadima’s gap in the polls decreased. In both campaigns, then, voters were faced with a choice of whether to vote for the leading party or whether to support a smaller party that is closer to their worldview but may not be a relevant political force after the election. The following studies, then, test the possible role of presumed media influence in citizens’ sincere or strategic voting within the Israeli context. Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009  Cohen, Tsfati / Influence of Presumed Media Influence   363 Measuring Strategic Voting Alvarez and Nagler (2000) distinguish three approaches for measuring strategic voting in the literature. The first, the aggregated inference method, involves using aggregate election results in different ways to discern the extent of strategic voting. For example, Burden (2005) compares election returns with polling data to tap stra- tegic shifts in votes. In the second, the self-reporting intentions approach, research- ers rely upon the reports of survey respondents about the motivations for their own voting behavior. For example, Arian and Shamir (cited in Abramson et al., 2004, p. 715) asked respondents directly if they would change their vote if their favorite candidate had no chance of winning. The third, the direct measurement approach, “tries to model strategic voting directly as the objective differences between the stated vote and the preference rankings of individuals” (Alvarez and Nagler, 2000, p. 63). This methodology, which enjoys several advantages over the others, accord- ing to Alvarez and Nagler, is the approach we adopt in Study 2. However, we first conducted an exploratory test of our hypothesis utilizing a self-report measure of strategic voting. Study 1: Testing H1 Using Self-Reported Measures of Strategic Voting In this study, we tested H1 using secondary analysis of data gathered by the Hertzog Institute for Media, Society and Politics, before the Israeli 2003 (n = 1,219) and 2006 (n = 509) elections. In both cases, respondents were a random sample of adult Israelis interviewed by phone, 4 to 5 weeks prior to election day. Response rates calculated according to the guidelines of the American Association for Public Opinion Research were RR1 = .20 in 2003 and RR1 = .19 in 2006. In 2003, 87.0% of respondents were Jews, and the rest were Arab Israelis; 49.3% of respondents were female. On average, respondents had 13 (SD = 2.90) years of schooling, and their age was 41.20 (SD = 16.78). According to self-reports, 64.0% were secular, 22.0% traditional, 8.3% religious, and 5.6% Ultra-Orthodox. A comparison of these distributions to population parameters was reported elsewhere (Tsfati & Peri, 2006). In the 2006 sample, 85.7% were Jews, and 49.7% female, with an average of 13.57 (SD = 5.65) years of schooling, and the average age was 44.39 (SD = 17.51). According to self-reports, 56.0% were secular, 25.9% traditional, 10.6% religious, and 6.9% Ultra-Orthodox. Measures Strategic voting. In both cases, the self-report measure of strategic voting was worded, Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009 364   Communication Research When deciding about their vote, some people take into account additional consider- ations, in addition to the extent of their agreement with parties and candidates, such as how the coalition will look, whether or not a party will pass the threshold etc. Other voters do not take into account such considerations. To what extent will such consider- ations take part in your voting decision in the coming elections?1 Response categories varied between 1 = not at all and 4 = to a great extent (2003: M = 1.87, SD = 1.01; 2006: M = 1.92, SD = 1.01). Presumed media influence. As in previous presumed influence and third-person perception research, presumed media influence on others was measured using a single survey question worded, “To what extent will other voters be influenced by news media coverage of the elections when they make up their mind about their vote?”2 Response categories for both items varied between 1 = not at all and 4 = to a great extent (2003: M = 2.35, SD = 0.89; 2006: M = 2.26, SD = 0.90). A similar item asked about presumed media influence on self (2003: M = 1.51, SD = 0.78; 2006: M = 1.50, SD = 0.79). Third-person perception scholars debate how to best model presumed media influence when examining its outcomes. Some use a differ- ence score of the effects on self and others, whereas others these media effects on self and others items separately, controlling for each other.3 Hoffner and Buchanan (2002) used both strategies and found that the difference scores approach concealed the reverse effects of the self and others items (p. 241; a similar pattern was identi- fied by Tsfati, Ribak, & Cohen, 2005). Given that presumed media influence on self emerged as a significant predictor of the outcome variables in these past “presumed influence” studies, we controlled for this variable in the model reported below. In the context of strategic voting, as in others, what is important is not the general assess- ment of how powerful media are but rather the assessments of how media affect others, net of its general impact (as measured by its perceived effects on the self).4 Covariates. Given that strategic voting is predicted not only by people’s percep- tions of media influence, several variables known as predictors of strategic voting (e.g., Cain, 1978; Lanoue & Bowler, 1992) were included in the model are described below. Political interest. Respondents were asked to what extent they “follow election campaign news coverage” and “tend to talk to friends about politics.” Response cat- egories varied between 1 = not at all and 4 = to a great extent. In both 2003 and 2006, both items loaded on a single factor in an exploratory factor analysis (EFA; principal components, varimax) explaining 62.79% of the variance in 2003 and 78.89% in 2006. The items were significantly correlated in both years, (2003: r = .57, p < .001; 2006: r = .38, p < .001) and hence were averaged to measure political interest (2003: M = 2.38, SD = 0.76; 2006: M = 2.54, SD = 0.79). Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009  Cohen, Tsfati / Influence of Presumed Media Influence   365 Political participation. Respondents were asked whether, in the past year, they “participated in demonstrations or political rallies,” “presented political stickers or had hung political ribbons,” and “signed petitions.” Response categories were 1 = no; 2 = yes, once; and 3 = yes, more than once. In both cases, the three items loaded on a single factor in an EFA (Principal Components, Varimax) explaining 57.0% of the variance in 2003, and 68.6% in 2006. The items were averaged to create a participa- tion scale (2003: Cronbach’s α = .62, M = 1.15, SD = .37; 2006: Cronbach’s α = .77, M = 1.30, SD = .53). Political cynicism. A battery of five questions requested respondents to asses their attitudes toward the political system and their ability as citizens to influence it: (1) “To what extent you and your friends can influence government policy?” (1 = not at all, 4 = to a great extent; reverse coded). (2) “Politicians care only about their per- sonal future; they do not care about the future of the country.” (3) “I do not care who wins the coming elections.” (4) “There are almost no differences in the positions of the major parties on the different issues.” (5) “It does not matter whom you vote for, it doesn’t really change anything.” Response categories varied between 1 = strongly disagree and 4 = strongly agree. In an EFA conducted separately for each sample (principal components, varimax, explaining 57.45% of the variance in 2003 and 61.3% in 2006), Item 1 loaded separately than the other items, and hence, it was used as a single-measure item of efficacy (2003: M = 1.68, SD = 0.80; 2006: M = 1.74, SD = 0.85), and the other items were averaged to create a measure of political cynicism (2003: Cronbach’s α = .59, M = 2.19, SD = 0.73; 2006: Cronbach’s α = .68, M = 2.41, SD = 0.79). Construct validity for the interest, participation, efficacy, and cynicism measures was tested by comparing respondents who intended to vote to those who did not intend to vote, on these measures. As expected, voters who reported that they intended to vote scored significantly higher on the political interest and efficacy scales and, in 2003, also on the political participation scale, compared with those who did not intend to vote. In contrast, those intending to vote reported, in both years, lower levels of political cynicism than those not intending to vote. These comparisons provide evidence of construct validity for these measures. News media exposure. Respondents in both surveys were asked how many times a week they read newspapers, listen to the news on the radio, watch news or current affairs programs on television, and browse news sites on the Internet.5 In both samples, an EFA (principal components, varimax, explaining 57.45% of the variance in 2003 and 40.5% of the variance in 2006) revealed a two-factor structure with the Internet item loading separately. Thus, two variables were created, one representing each factor: The first was the average of the television, newspaper, and radio item (2003: M = 2.19, SD = 0.73; 2006: M = 3.89, SD = 2.01), and the second was the Internet item (2003: M = 2.19, SD = 0.73; 2006: M = 2.77, SD = 3.18).6 Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009 366   Communication Research Results Our hypothesis predicted that perceptions of media influence will be associated with strategic voting. Indeed, results for both election years demonstrate signifi- cant and positive correlations between presumed influence of campaign coverage on others and self-reported strategic voting (2003: r = .13, p < .001; 2006: r = .17, p < .001), such that a stronger belief in media influence was associated with an increased tendency for strategic voting. To examine the contribution of presumed media influence to predicting strategic voting, controlling for the effects of potential political and demographic predictors, cumulative logistic analysis was conducted with strategic voting as the dependent variable and presumed media influence and the control variables as predictors. Cumulative logit is the most recommended method for ordered categorical depen- dent variables (Allison, 1999; McCullagh, 1980). Unlike ordinary least squares regression, it does not require that the dependent variable be continuous or nor- mally distributed. Results for the cumulative logit models are presented in Table 1. In both election years, Arab voters had higher odds of scoring higher on the strategic voting scale than their Jewish counterparts (2003: b = 0.70, SE = 0.22, p < .01; 2006: b = 1.35, SE = 0.31, p < .001). Presumed influence on self was also consistently associated with strategic voting (2003: b = 0.38, SE = 0.08, p < .001; 2006: b = 0.48, SE = 0.14, p < .05), such that the more respondents felt news media had an influence on them, the more likely they were to report that they take into account strategic consider- ations in their voting decisions. Higher political interest was also associated with higher strategic voting scores in 2003 (b = 0.22, SE = 0.10, p < .05), and political participation was borderline significantly associated with strategic voting in that sample (b = 0.29, SE = 0.17, p = .069). In 2006, more efficacious respondents were more likely to report strategic voting (b = 0.27, SE = 0.13, p < .05), and politically cynical respondents were less likely to report strategic voting (b = –0.33, SE = 0.14, p < .05). Over and above the effects of these control variables, presumed media influence on others was positively and consistently associated with strategic voting. In 2006, the p value for this association was exactly on the border of accepted significance (b = 0.24, SE = 0.12, p = .05), but in 2003, the association was statistically signifi- cant (b = 0.16, SE = 0.08, p < .05): The more respondents perceived that news media have an influence on other voters, the more likely they were to report that they take into account strategic considerations when voting. Study 2: Testing H1 Using Direct Measurement of Strategic Voting The main methodological limitation of Study 1 is its use of self-report measures for strategic voting, which could reflect self-perceived sophistication rather than Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009  Cohen, Tsfati / Influence of Presumed Media Influence   367 Table 1 Cumulative Logistic Regression Models Predicting Self-Report of Strategic Voting (Study 1) 2003 2006 Threshold: Strategic = 1 2.20 (.62)*** 2.85 (.87)*** Threshold: Strategic = 2 3.16 (.63)*** 3.84 (.89)*** Threshold: Strategic = 3 4.75 (.64)*** 5.84 (.92) Political efficacy 0.06 (.09) 0.27* (.13) Political interest 0.22* (.10) 0.21 (.16) Political cynicism –0.04 (.09) –0.33* (.14) Political participation 0.29# (.17) 0.30 (.19) Sex (1 = female) –0.08 (.13) 0.13 (.22) Age 0.00 (.01) 0.01 (.01) Religiosity –0.10 (.08) –0.09 (.12) Education 0.02 (.03) –0.01 (.01) Arab voters (= 1) 0.70*** (.22) 1.35*** (.31) News media exposure Factor 1 (TV, newspapers, radio) 0.04 (.04) 0.10 (.06) Factor 2 (Internet) 0.02 (.03) –0.02 (.04) Presumed media influence on self 0.38*** (.08) 0.48** (.14) Presumed media influence on others 0.16* (.08) 0.24^ (.12) Nagelkerke R2 0.08 0.20 N 844 363 Note: Table entries are cumulative logistic regression coefficients. Standard errors are reported in parentheses. # p < .10. ^p = .05. *p < .05. **p < .01. ***p < .001. actual strategic voting. Indeed, Alvarez and Nagler (2000) demonstrate that “relying on the self reports of respondents to measure the incidence of strategic voting is problematic and is likely to produce overstated estimates of the proportions of the electorate who are voting strategically” (p. 63). Study 2 overcame this limitation by using a more direct measure of strategic voting. Another purpose of Study 2 was to try to decipher some specific ways in which perceptions of media influence influ- enced strategic voting in the context of the 2006 Israeli elections. Remembering that strategic voting entails voting for the party or candidate that is different from the one that is closest to voters’ worldview and about whom they feel most favorably, the direct measure we applied in this study was based on sepa- rate items asking about voting intention and attitudes toward parties. This approach is identical to that used by Blais, Aldrich, Indridason, and Levine (2006) in their study of the Israeli elections of 2003. Given that in the Israeli parliamentary system, voters cast ballots for party lists only (and not for individual candidates), we asked about attitudes toward parties (in a similar manner to Blais et al., 2006). Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009 368   Communication Research Method Respondents were a random phone sample of likely voters who said they were going to vote or those who had not yet decided (n = 517). Unlike in Study 1, which surveyed respondents weeks prior to the election, in this study, respondents were contacted within the final 10 days before the elections. Response rate calculated according to the guide- lines of the American Association for Public Opinion Research was RR1 = .35. Jews made up 83.6% of the sample, and the rest were Arab Israelis; 56% of respondents were female. Respondents had on average 14.11 (SD = 3.20) years of schooling, and their average age was 46 (SD = 16.68). The distribution of self-reported religiosity was 52.2% secular, 26.8% traditional, 14.2% religious, and 6.9% Ultra-Orthodox. Measures Strategic voting. Respondents were asked about their attitudes toward the party for which they planned to vote and toward two additional parties they felt were clos- est to them. Question wordings were, “To which of the parties running for the 17th Knesset election do you plan to vote?”7 This question was followed by another item, worded, “Is there any other party you feel is close to you in its positions, that you have considered voting for, or that you would vote for under other circumstances?” Respondents could name two such parties.8 Respondents replying they still have not decided about their voting intention were probed to name all parties they were con- sidering and to name the one for whom it is most likely they will end up voting. Only 409 respondents were willing to name the party for which they were most likely to vote, of which 173 (42.30%) named a second party and 25 (6.11%) named three parties. A single-item measure was used to gauge respondents’ attitudes toward the parties they mentioned. Question wording was, “How close, on a 1 (= ‘very far’) to 10 (= ‘very close’) scale, do you feel toward the positions of (each of the parties they mentioned)?” These two measures (voting intentions and attitudes toward party positions) were used to identify strategic voters. When respondents had less favorable attitudes toward the party’s positions for which they planned to vote, in comparison to one of the two additional parties the mentioned, they were labeled strategic voters.9 There were 58 such voters, compris- ing 11.2% of likely voters. This figure is similar to Abramson et al. (1992) who, utilizing a similar measurement strategy, reported 14% sophisticated voting in U.S. primaries. The present estimate of strategic voting also falls in the range of findings, reported by Alvarez and Nagler (2000), who reviewed 13 studies and found esti- mates of strategic voting between 5.1% and 17.0 %. Presumed media influence. As in the previous study, respondents were asked, “To what extent will other voters be influenced by news media coverage of the elections when they make up their mind about their vote?” Response categories for both items Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009  Cohen, Tsfati / Influence of Presumed Media Influence   369 varied between 1 = not at all and 4 = to a great extent (M = 2.32, SD = 0.80). A similar item again asked about presumed media influence on self (M = 1.56, SD = 0.78).10 A follow-up item to the question on presumed influence on others asked respondents whether they think news media coverage will increase support of a certain party or parties. Political interest. Respondents were asked to what extent they “follow election campaign news coverage,” “are interested in political issues,” “tend to talk to friends about politics,” and “follow election polls.” Response categories for all items varied between 1 = not at all and 4 = to a great extent. All four items loaded on a single factor in an EFA (principal components, varimax) explaining 56.63% of the vari- ance. The items were averaged to create an index of political interest (Cronbach’s α = .74; M = 2.35, SD = 0.66). Political participation. As in Study 1, respondents were asked whether, in the past year, they “participated in demonstrations or political rallies” or “presented political stickers or hung political ribbons.” Response categories were 1= no; 2 = yes, once; and 3 = yes, more than once. The correlation between the items was r = .39 (p < .001). The items were summed to create a political participation scale (M = 2.73, SD = 1.15). Political cynicism. Respondents were presented with the same battery of five questions used in Study 1. In an EFA (principal components, varimax, explaining 36.76% of the variance), Item 1 again loaded separately than the other items, and hence, it was again treated as a single-item measure of efficacy (M = 1.93, SD = 0.77), and the other items were averaged to create a measure of political cynicism (M = 2.17, SD = 0.65). While the reliability of this construct was rather low (Cronbach’s α = .55), it was retained in order to remain consistent and allow com- parability to the models reported in Study 1 and given the fact that its inclusion or exclusion in the models reported below did not change the pattern of results. News media exposure. As in Study 1, respondents were asked about the frequency of their exposure to newspapers, news on the radio, news or current affairs programs on television, and news pages on the Internet. An EFA (principal components, vari- max, explaining 58.73% of the variance) revealed a two-factor structure but this time with the Internet item loading together with newspapers and television loading together with radio news. Thus, two separate variables were created, one represent- ing each factor: The first was the average of the television and radio items (M = 3.92, SD = 2.24), and the second was the average of the Internet and newspapers items (M = 2.82, SD = 2.13).11 Like in Study 1, construct validity for the political interest, participation, efficacy, and cynicism measures was established by comparing respondents who intended to Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009 370   Communication Research Table 2 Logistic Regression Model Predicting Strategic Voting 2006 (Study 2) b SE Political efficacy –0.23 0.28 Political interest –0.03 0.40 Political cynicism 0.11 0.34 Political participation –0.05 0.19 Sex (1= female) 1.39*** 0.43 Age 0.01 0.01 Religiosity 0.18 0.25 Education –0.04 0.07 Arab (= 1) 0.38 0.60 News exposure Factor 1 (radio, TV) –0.03 0.11 Factor 2 (newspapers, Internet) 0.05 0.10 Presumed media influence on self –0.38 0.28 Presumed media influence on others 0.54* 0.27 Nagelkerke R2 0.17 n 136 Note: Only respondents deliberating between two or more parties were included in the analysis. # p < .10. ^p = .05. *p < .05. **p < .01. ***p < .001. vote to those who did not know whether they would vote or were not yet certain on these measures. As expected, voters intending to vote scored higher on the political interest, participation, and efficacy scales, compared to those uncertain about whether they would vote. In contrast, those intending to vote reported lower levels of cynicism, compared to those not certain. Results Our hypothesis predicted that perceptions of media impact will be associated with strategic voting. Logistic regression analysis was conducted with strategic voting as the dependent variable in order to examine the effect of presumed media influence on others on strategic voting controlling for the effects of potential intervening vari- ables.12 As Table 2 demonstrates, the only control variable significantly affecting strategic voting was sex, with females having higher odds of being strategic voters, compared to males (b = 1.39, SE = 0.43, p < .001, eb = 4.02). Ceteris paribus, pre- sumed media influence had a significant effect on strategic voting (b = 0.54, SE = 0.27, p = .05). Each one-unit increase on the presumed influence scale was associ- ated with an increase of 71% in the odds of being a strategic voter. Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009  Cohen, Tsfati / Influence of Presumed Media Influence   371 Post Hoc Analysis The data demonstrate that perceptions of media influence on other voters are asso- ciated with strategic voting. But what kind of strategic shifts in voting are accounted for by voters’ perceptions of news media influence? To tackle this question, one has to examine the direction of shifts sparked by perceptions of media influence. While there are numerous possible scenarios for strategic voting, the most commonly dis- cussed is a shift from a small party to a larger one due to viability considerations. As explained above, the possibilities for strategic voting are diverse, and thus, modeling the exact type of strategic voting in a statistical model may be complex. A likely assumption that helps to slightly simplify such an analysis is that Israelis per- ceive that media influence people to support the large parties and their candidates rather than small ones: This is because most coverage of Israeli elections focuses on the heads of the large parties that are candidates for premiership and because large parties run larger campaigns, enjoy better ties with journalists, and are more news- worthy because of their candidates’ current status (as ministers and opposition lead- ers). This assumption was examined in our survey by asking respondents about the direction of media influence.13 Respondents reporting that they thought coverage would help a party were asked which parties would likely benefit from media cover- age. Expectedly, 84.3% of respondents answering the question mentioned at least one of the three leading parties. If one assumes that most voters perceive that media help the large parties, in what way does this shape strategic voting? Supporters of small parties may feel that if media coverage causes people to support larger parties, the small parties they sup- port may not pass the threshold, and thus, they risk losing their vote (Lanoue & Bowler, 1992). This may cause them to vote strategically for the large party closest to their political attitudes (though not as close as a small party). Thus, one may hypothesize that presumed media influence will be related to strategically voting for a large party instead of a favorite small one. To examine this notion, a logistic regression was run, comparing sincere voters (= 0) to the 18 strategic voters who deserted small parties in favor of larger parties (those who reported feeling closest to a small party but that they were likely voting for one of the three larger parties; = 1). Results are presented in Table 3. As the table shows, sex was again the only significant control variable predicting strategi- cally deserting a small party in favor of a larger party (b = 1.59, SE = 0.58, p < .01, eb = 4.90) . As predicted, and all other things being equal, presumed media influence on others was also positively and borderline significantly associated with strategic voting while deserting a small party in favor of a larger party (b = 0.64, SE = 0.36, p = .07, eb = 1.89). Each one-unit increase on the presumed influence item was asso- ciated with an 89% increase in the odds of strategically voting this way. Though several other scenarios of strategic voting are possible, given the relative rarity of this type of voting, our data set was not large enough to model the less common ones. Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009 372   Communication Research Table 3 Logistic Regression Model Predicting Strategic Voting While Deserting Small Parties in Favor of Larger Parties (1 = strategic voters, 0 = sincere voters) b SE Political efficacy –0.61 0.40 Political interest –0.65 0.55 Political cynicism –0.07 0.45 Political participation 0.26 0.22 Sex (1 = female) 1.59** 0.58 Age 0.01 0.02 Religiosity 0.44 0.30 Education –0.08 0.09 Arab (= 1) –0.23 0.92 News exposure –0.07 0.16 Factor 1 (radio, TV) Factor 2 (newspapers, Internet) 0.01 0.14 Presumed media influence on self –0.36 0.41 Presumed media influence on others 0.64# 0.36 Nagelkerke R2 0.32 n 117 Note: Only respondents deliberating between two or more parties were included in the analysis. Strategic voters not deserting small parties in favor of larger parties were also excluded from the analysis. # p < .10. ^p = .05. *p < .05. **p < .01. ***p < .001. Discussion Three data sets, from two election campaigns, utilizing different measures of stra- tegic voting and conducted at various stages of the campaign, demonstrate that per- ceptions of media influence on others are associated with strategic voting. One can interpret this evidence as an indirect media effect on voting decisions. What we think media do to other voters sometimes influences our own voting decisions. This was the case even after controlling for political and demographic variables and for percep- tions of media influence on the self. Though in Study 2 the effects of perceptions of media influence were not common (i.e., they did not move many voters to vote stra- tegically), such perceptions played a role for a significant portion of voters who were deliberating how to vote (i.e., the 173 who mentioned at least two parties). We believe the results of this study open a new avenue for media effects research in the context of political campaigns, one that is subtler and relatively unrelated to the amount of media exposure, compared to other theories of media influence on elections. It is important to note that in both self-report data sets and the data where strategic voters were directly identified, the majority of voters seemed to vote sincerely. But even though strategic voters are a minority, strategic voting can play an important role Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009  Cohen, Tsfati / Influence of Presumed Media Influence   373 in elections. For example, in our 2006 Study 2 data, four respondents (0.8%) reported they would vote for the Chail party, an extreme rightwing party that did not ultimately pass the threshold. However, eight additional likely voters were identified as strategic voters who felt closest to this party but reported intentions to vote for a larger party. This suggests that if all voting was sincere, this party would be represented in the Knesset. On an anecdotal level, it is interesting to note that six of these eight strategic voters who deserted Chail reported high or very high presumed media influence. Thus, it may well be that this Knesset is free of the extremist representatives of this party because its supporters believe that media coverage works against them and believe media are powerful. Though many scenarios of strategic voting were possible in the 2006 elections, only one was found to be significantly related to presumed media influence: the move from smaller parties to larger ones apparently based on viability consider- ations. This trend of strategic voting is also the one most commonly mentioned in the literature (e.g., Abramson et al., 1992). However, given the small number of strategic voters, it is quite likely that perceived media influence has other, smaller influences that we were not able to detect. A closer examination of the data, however, did reveal a few cases of people who felt closest to large parties but said a few days before the election that they were probably going to vote for a smaller one. For example, of the 143 respondents who mentioned Kadima as one of the three parties they were considering, 11 claimed they would most likely vote for another, smaller party though they felt closer to Kadima. However, this trend was not found to be related to perceived media influ- ence. This is noteworthy simply because it demonstrates that in some political systems, strategic voting may take on more than one route and may involve moti- vations other than voting for viable parties or candidates. The exact direction of the influence of presumed influence in such contexts, then, still remains largely unknown. What are the implications of our results to the theory of presumed influence? In most studies of the effects of presumed media influence, findings tend to be orga- nized into two major theoretical categories: perceived influence as promoting con- formity to social norms or as causing defiance of such norms (Katz, 2005). The findings of this study, however, do not fit either category but rather suggest that presumed media influence may simply be one of many social beliefs that form the basis upon which people consider their actions and their potential outcomes—what may be termed a coordination effect. In this sense, this study provides a new direction for presumed influence research. No less important is that for those interested in strategic voting, our results indicate a new variable—presumed media influence—to help explain why people vote strategically. Traditional strategic voting research has underscored the importance of information in voters’ strategic considerations. The current investigation elaborates on the media and the perceptual sources underlying the process. Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009 374   Communication Research Are our results generalizable to other electoral systems? The answer to this ques- tion is likely to be positive, so long as those systems have more than two options on the ballot and voters are able to form perceptions about the influence of news media in the campaign. Arguably, if presumed media influence had an impact on strategic voting in the complex Israeli system (with approximately 30 parties on the ballots and endless possible coalitions), it should have even stronger and clearer effects in simpler systems (where perceptions of media influence may be easier to form). For example, in the 2008 U.S. Democratic primaries, a voter favoring a candidate that he or she feels is mistreated or ignored by media (say, Dennis Kucinich) may vote for another candidate (say, Barack Obama) because he or she feels that this candi- date enjoys a more positive and possibly influential news coverage. Another interesting result is that in Study 2 (which utilized a direct measure of strategic voting), women were significantly more likely to vote strategically. This is even more interesting because at the bivariate level women reported less politi- cal interest (M = 2.29, SD = 0.66) than men (M = 2.43, SD = 0.67), t(515) = 2.38, p = .02, though there were no such differences in participation, efficacy, or cynicism. Though we cannot explain this finding, we did find several studies that have exam- ined the connection between gender and strategic voting: Like in Study 1, Lanoue and Bowler (1992) and Skabalan (1988) used self-report measures to test for a gender effect on strategic voting but did not find any. In contrast to our finding, in a study of the 2003 California recall election, Shaw, McKenzie, and Underwood (2005) found men to be more likely to vote strategically for Schwarzenegger. This may suggest that in some cases, the gender effect may be due to the identity of the candidates. Interestingly, in both the self-report studies, Arabs reported being more likely to vote strategically, though this same pattern was not found in Study 2. Though it makes sense that Arabs who are more likely to feel closest to small Arab parties would be especially prone to strategic consideration, and though they perceive them- selves as more strategic in their voting, this did not seem to be the case in the 2006 elections. However, given the findings regarding both women and Arabs, future research should explore whether minorities who may be especially concerned about having their political voices being heard are more strategic voters. It is also noteworthy that in both self report data sets (2003 and 2006), presumed media impact on self is a stronger and positive predictor of strategic voting than per- ceived impact on others. A possible interpretation is that those seriously considering their voting decisions tend to use media more reflexively when deliberating about their choices and thus may be more aware of this use. This interpretation is consistent with the fact that presumed influence on self was associated with political interest and efficacy and with less certainty in voting in all three data sets. This suggests that perhaps more involved respondents do not necessarily see media influence on voting as inherently negative and are willing to publicly admit such influence, compared to less involved respondents. Thus, presumed influence on self may relate to self-reports of strategic considerations through higher involvement and more deliberation. Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009  Cohen, Tsfati / Influence of Presumed Media Influence   375 A related noteworthy pattern was that the self-reported measures of strategic vot- ing in Study 1 were generally related to the political covariates, while the direct measure of strategic voting in Study 2 was not. Two possible explanations seem plausible. First, the strategic voting question in Study 1 asked about taking into account strategic considerations (and the threshold and coalition structure were pro- vided as examples). Some people may reasonably have indicated that they consider such strategic considerations but still end up voting sincerely. In contrast, in Study 2, strategic voters were identified on a much more stringent basis. Perhaps interested and involved voters are likelier to weigh strategic considerations but not necessarily to vote strategically. Second, it may very well be that response bias, a probable limitation of self-report accounts of strategic voting (Alvarez & Nagler, 2000), is partly responsible for the different effects of the political variables in the two types of studies: Obviously, people who view political interest and participation as positive may also view the use of strategic considerations as more legitimate or even desirable and report higher levels of both. This is exactly the reason we employed a direct measure in Study 2, and despite the differences in methodologies and results, the main finding that per- ceived influence on others is positively related to strategic voting was consistently confirmed. Additional limitations of the current investigation include the very early measurement of strategic voting in Study 1, and in Study 2, given the number of parties running for the Knesset and the distribution of voting intentions for these parties in our data, small ns prevented us from testing for the influence of presumed media influence on more specific shifts in voting. Though much work remains before we can fully outline how perceptions of media influence shape voting in general and strategic voting specifically, our study goes a long way in establishing that such a connection exists. Notes   1. This wording is different from the self-report measures used by Arian and Shamir, reported above, and from the measures used by Niemi, Whitten, and Franklin (1992) and by Heath and Evans (1994), in that it taps the use of strategic considerations and not actually changing one’s vote due to strategic con- siderations. Given that the poll was conducted early on in the campaign, before many voters have made up their mind, it seemed to us more sensible, and less hypothetical, to ask about the such considerations and not about actual shifts in voting.   2. In order to examine possible effects of different possible question wordings, a random half of the 2003 respondents were asked a somewhat more positively worded set of items: “To what extent will news media coverage of the elections help other voters/you when they/you make up their/your mind about their vote?” Since no significant effects of question wordings emerged, the differences in question wordings were ignored, and the items were collapsed.   3. Other scholars (Lo & Wei, 2002; Neuwirth & Frederick, 2002) have recently suggested using additive scales, consisting of the sum of the items measuring presumed influence on self and others. Given this recommendation, we also ran models with such additive scales. Results in both 2003 and 2006 showed a significant effect (2003: b = 0.27, SE = 0.05, p < .001; 2006: b = 0.32, SE = 0.08, p < .001), Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009 376   Communication Research but the overall explained variance in the model was in both years somewhat lower, so we opted for report- ing the models with both items separately as independent variables.   4. As expected, given previous research on the third-person perception, in both 2003 and 2006, the dif- ference between the perceived influence on self and others items was statistically significant—in 2003, paired t(994) = 26.30, p < .001; and in 2006, paired t(421) = 14.95, p < .001. As in previous third-person perception research, the items were significantly correlated (2003: r = .35, p < .001; 2006: r = .32, p < .001).   5. For various reasons, answer categories to the closed-ended question were different in the differ- ent exposure items and in the different election years. In order to standardize them, they were translated to the estimated number of days per week (e.g., if an answer category was 4-6 times a week, it was recoded as 5).   6. Reliability analysis for the media exposure variables resulted with rather low reliabilities (in 2006, Cronbach’s α = .50). This is not unusual in the communication literature, given that items measuring exposure should be conceptualized as “causal indicators” of exposure (indicators that determine a latent construct on a theoretical basis rather than the more common “effects indicators” that are affected by the latent construct). A high intercorrelation between causal indicators is not a necessary condition for reli- ability. This is because a respondent could have a high score on one indicator (say, the TV indicator) without necessarily having a high score on the others.   7. The distribution of answers was 20.7% Kadima, 11.8% Labor, 9.5% Likud, 6.4% for the National Religious Block, 5.0% Shas, 5.4% Meretz, and 4.4% Israel Beytenu. The remaining respondents reported they planned voting for smaller parties.   8. For the second party, the distribution (of the total sample) was 6.2% Kadima, 4.3% Labor, 4.3% Likud, and 3.3% for the National Religious Block; the remaining respondents reported feeling close to smaller parties. Each of the parties mentioned as third parties were mentioned by no more than 0.80% of the total sample.   9. Similar direct measures of strategic voting were used by Blais and Nadeau (1996) and Abramson et al. (1992). 10. As in Study 1, the difference between the two presumed influence items was statistically significant— paired t(425) = 15.53, p < .001—and they were significantly correlated (r = .24, p < .001). 11. As in Study 1, and for the same reasons, the interitem correlations were rather low: for the first factor, r = .23, p < .001, and for the second factor, r = .10, p < .05. 12. As in the previous study, we also used an additive scale, composed of the sum of the presumed media influence on self and others (as recommended by Lo & Wei, 2002; Neuwirth & Frederick, 2002). In this case, the effect of this additive scale was insignificant (b = 0.08, SE = 0.17; ns). This is probably because the effects of presumed influence on self and others worked in different directions and thus canceled each other out (similar to what happened in the Hoffner & Buchanan, 2002, study). 13. Question wording was, “In general, do you think news media coverage will increase support of a certain party or parties?” References Abramson, P. R., Aldrich, J. H., Diamond, M., Diskin, A., Levine, R., & Scotto, T. J. (2004). Strategic abandonment or sincerely second best? The 1999 Israeli prime ministerial election. Journal of Politics, 66, 706-728. Abramson, P. R., Aldrich, J. H., Paolino, P., & Rohde, D. W. (1992). “Sophisticated” voting in the 1988 presidential primaries. American Political Science review, 86(1), 55-69. Allison, P. (1999). Logistic regression using the SAS system: Theory and application. Cary, NC: Wiley-SAS. Alvarez, R. M., & Nagler, J. (2000). A new approach for modeling strategic voting in multiparty elections. British Journal of Political Science, 30, 57-75. Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009  Cohen, Tsfati / Influence of Presumed Media Influence   377 Benoit, K., Giannetti, D., & Laver, M. (2002, April). Strategic voting in mixed-member electoral systems. The Italian case paper prepared for delivery at the annual meeting of the European Public Choice Society, Belgirate, Italy. Blais, A., Aldrich, J., H., Indridason, I. H., & Levine, R. (2006). Do voters vote for government coali- tions? Testing Downs' pessimistic conclusion. Party Politics, 12(6), 691-705. Blais, A., & Nadeau, R. (1996). Measuring strategic voting: A two-step procedure. Electoral Studies, 15, 39-52. Burden, B. C. (2005). Minor parties and strategic voting in recent U.S. presidential elections. Electoral Studies, 24, 603-618. Cain, B. E. (1978). Strategic voting in Britain. American Journal of Political Science, 22, 639-655. Cappella, J. N., & Jamieson, K. H. (1997). Spiral of cynicism: The press and the public good. Oxford, UK: Oxford University Press. Cox, G. W. (1997). Making votes count: Strategic coordination in the world’s electoral systems. Cambridge, UK: Cambridge University Press. Davison, W. P. (1983). The third person effect in communication. Public Opinion Quarterly, 47, 1-15. Downs, A. (1957). An economic theory of democracy. New York: Harper & Row. Gunther, A. C., & Liebhart, J. L. (2006). Broad reach or biased source? Decomposing the Hostile Media Effect Journal of Communication, 56(3), 449-466. Gunther, A. C., & Storey, J. D. (2003). The influence of presumed influence. Journal of Communication, 53, 199-215. Heath, A., & Evans, G. (1994) Tectical voting: Concepts, measurement and findings. British Journal of Political Science, 24, 557-561. Hoffner, C., & Buchanan, M. (2002). Parents’ responses to television violence: The third-person percep- tion, parental mediation, and support for censorship. Media Psychology, 4(3), 231-252. Katz, E. (2005, June). The myth of media impact is influential, too. Keynote lecture delivered to the Influence of Presumed Media Influence Research Workshop, Haifa, Israel. Lanoue, D. J., & Bowler, S. (1992). The sources of tactical voting in British parliamentary elections, 1983-1987. Political Behavior, 14(2), 141-157. Lijphart, A. (1993). Israeli democracy and democratic reform in comparative perspectives. In E. Sprinzak & L. Diamond (Eds.), Israeli democracy under stress (pp. 107-123). Boulder, CO: Lynne Rienner. Lo, V., & Wei, R. (2002). Third-person effect, gender, and pornography on the Internet. Journal of Broadcasting & Electronic Media, 46, 13-33. McCullagh, P. (1980). Regression models for ordinal data. Journal of the Royal Statistical Society, Series B, 42, 109-142. Neuwirth, K., & Frederick, E. (2002). Extending the framework of third-, first-, and second-person effects. Mass Communication & Society, 5, 113-140. Niemi, R. G., Whitten, G., & Franklin, M. N. (1992). Constituency characteristics, individual character- istics and tactical voting in the 1987 British general election. British Journal of Political Science, 22, 229-240. Rahat, G., & Hazan, R. Y. (2005). Israel: The politics of an extreme electoral system. In M. Gallagher & P. Mitchell (Eds.), The politics of electoral systems (pp. 333-351). New York: Oxford University Press. Rogers, E. M. (1995). The diffusion of Innovations (5th ed.). New York: Free Press. Shaw, D., McKenzie, M. J., & Underwood, J. (2005). Strategic voting in the California recall election. American Politics Research, 33, 216-245. Skabalan, A. (1988). The mostly-sovereign people: Sophisticated voting and public opinion about term limits in California, Political Behavior, 20(1), 35-51. Tsfati, Y., & Cohen, J. (2005). The influence of presumed media influence on democratic legitimacy: The case of Gaza settlers. Communication Research, 32, 794-821. Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009 378   Communication Research Tsfati, Y., & Peri, Y. (2006). Mainstream media skepticism and exposure to sectorial and extranational news media: The case of Israel. Mass Communication & Society, 9, 165-187. Tsfati, Y., Ribak, R., & Cohen, J. (2005). Parents’ third person perceptions regarding the influence of television: Rebelde Way in Israel. Mass Communication & Society, 8(1), 3-22. Weaver, K. R. (2002). New Zealand: The supreme political football. Boston: Center for Retirement Research. Yavuz, M. H., & Ozcan, N. A. (2006). The Kurdish question and Turkey’s Justice and Development Party. Middle Easy Policy, 13(1), 102-119. Jonathan Cohen (PhD, University of Southern California, 1995) is a senior lecturer in the Department of Communication at the University of Haifa, Israel. His research interests include the study of mediated relationships with TV characters, presumed media influence, media perceptions and effects, and Israeli television audiences. Yariv Tsfati (PhD, University of Pennsylvania, 2002) is a senior lecturer in the Department of Communication, University of Haifa, Israel. His research focuses on audience perceptions of news media. Downloaded from http://crx.sagepub.com at University of Haifa Library on May 2, 2009