Content-Type: text/html Intermedia Agenda Setting in the 1996 Presidential Election Thomas P. Boyle Susquehanna University 514 University Ave. Selinsgrove, PA 17870 717.372.4079 [log in to unmask] Submitted to: Mass Communication & Society Division Association for Education in Journalism and Mass Communication Baltimore, MD August 5-8, 1998 1996 Intermedia Agenda Setting Intermedia Agenda Setting in the 1996 Presidential Election Submitted to: Mass Communication & Society Division Association for Education in Journalism and Mass Communication Baltimore, MD August 5-8, 1998 Abstract This study focuses on the network news coverage across two time periods during the 1996 presidential election. Results from a content analysis of 116 political advertisements, 818 newspaper stories, and 101 network news stories found support for the belief that advertising influences the prominence of an issue within network newscasts, but not in the overall time devoted to it. Also, findings suggest prestige newspapers play a role in determining what issues lead television newscasts. List of Tables Table 1: Combined Mentions of the Political Advertisement Agenda Table 2: Combined Time of the Political Advertisement Agenda Table 3: Pearson Coefficients of the 1996 Presidential Political Advertisement Mentions Table 4: Pearson Coefficients of the 1996 Presidential Political Advertisement Seconds Table 5: Combined Mentions of the Network News Agenda Table 6: Combined Time of the Network News Agenda Table 7: Prominence of Issue Stories in First Segment of Network Newscasts Table 8: Pearson Coefficients of 1996 Network Presidential Campaign Coverage Table 9: Combined Mentions of the Newspaper Agenda Table 10: Combined Column Inches of the Newspaper Agenda Table 11: Combined Prominence of the Issues of the Newspaper Agenda Table 12: Pearson Coefficients of 1996 Presidential Coverage by Newspaper Column Inches Table 13: News Mentions and Seconds at Time 2 Regressed on Advertising, Prestige Press, and Real-World Indicators Table 14: News Prominence at Time 2 Regressed on Advertising, Prestige Press, and Real-World Indicators Introduction The goal of this study is to analyze on a national level the influence of the television advertising agenda on the network news agenda during the 1996 presidential political campaign season. This study is driven by two areas of inquiry. First, considering the large number of agenda-setting studies considering the media agenda's relationship the public or policy agendas, it is crucial to understand the construction of the media agenda. This is an area outlined by other researchers (Rogers, Dearing, and Bregman, 1993). Second, if campaigns do play a major role in determining the media agenda, what does this say about the importance of campaigns in choosing our president. Recent Agenda-Setting Research In 1993, Rogers, Dearing, and Bregman noted the original McCombs and Shaw study (1972) has been cited in nearly 200 agenda-setting studies. They also outline four areas of present and future research: 'setting the media agenda, real-world conditions, individual-level cognitive processes, and increasing the accuracy of the public agenda' (1993, p. 80). This study falls into the first of these areas: intermedia agenda setting. In this area, one study of special note is by Marilyn Roberts and Maxwell McCombs and supports a strong link between television advertising and the television news agenda (1994). Considering the 1990 Texas gubernatorial race, Roberts and McCombs found support for a strong relationship between newspaper and television news content (Roberts & McCombs, 1994, p. 257). Also, a strong correlation was found between television political advertising, and television news and newspaper coverage (Roberts & McCombs, 1994, p. 257). In a related study, Roberts considered how the agenda-setting function of the news and political advertising helped predict who voters supported during the 1990 Texas gubernatorial battle (1992). Roberts used discriminant analysis to see if the three different agendas -- voters, news, and advertising -- had a strong relationship with the vote reported in the post-election survey. Initially, this classification technique had a 70 percent success rate (Roberts, 1992, p. 892). An important note is Roberts' explanation of the limitations of the agenda-setting process to explain voting behavior. Only a "subset" separates the voters of each candidate and even when an issue is high on the "advertising agenda" it does not always transfer into a predictor of the vote (Roberts, 1992, p. 889). Yet when "partisanship, gender, degree of media reliance" and other controls are added to the analysis, the predictive power of the issue agenda increases -- in some cases up to 90 percent or more (1992, pp. 890-892). In addition to considering the influences of the television advertising agenda on the news agenda, the consideration of the "prestige press" and its impact on network news programming has been examined by media scholars. Danielian and Reese (1989) conducted an extensive 40-week study of intermedia agenda-setting focusing on the cocaine issue and reported the results in two book chapters. They found network news programs followed the lead of the New York Times in their coverage of the issue (Reese & Danielian, 1989, p. 41). As in the original agenda-setting study (McCombs & Shaw, 1972), Danielian and Reese found great similarity within media categories (1989, pp. 52-53). Danielian and Reese conclude while the New York Times lead coverage initially, other media also lead at various points and in some cases "converged" on the issue (1989, p. 56). Along with considering the influence of one agenda on another, the placement of news stories within their larger program vehicle has been considered by the researchers. In an extensive agenda-setting study, Iyengar and Kinder (1987) found the placement of news story within simulated broadcasts had a great influence on what stories viewers believed important. With another variable -- real-world indicators, Behr and Iyengar (1985) studied survey data and a selected network newscast from 1974 through 1980. The researchers found changing national indicators influence the amount and prominence of an issue's treatment by network news (Behr & Iyengar, 1985, p. 45). The researchers suggest real-world indicators lead to an increase in an issue's saliency on the news and public agendas (Behr & Iyengar, 1985, pp. 53-54). Overall, the literature supports the importance of political advertising, prestige press coverage, and real-world indicators to network news coverage. The study here focuses on their influence during the 1996 presidential content. Summary of the Design Primary data were collected through a content analysis of four sources: the universe of the advertisements produced by the three major candidates in the 1996 presidential race, a text-based analysis of randomly constructed weeks of the network news verbatim transcripts for eight weeks prior to election day, a text-based analysis of news stories in three major daily newspapers, and an analysis of changes in real-world indicators for a subset of the forty issue categories. Issue mentions and the time or column issues devoted to it were coded by the researcher. A second coder analyzed ten percent of the advertisements and texts. The intercoder reliability rate was similar to other content analyses using the appropriate formula for two coders: 84 percent for the advertisements and newspapers, 83 percent for the network newscasts (for formula see North, Holsti, Zaninovich, & Zinnes, 1963). Along with coding for mentions and amount of space or time, the study analyzed the placement of the issues within the network newscasts and major dailies. Campaign coverage on the network news agenda was the dependent variable. The unit of analysis was the forty issue categories. The units of observation were 116 political advertisements, 101 campaign-focused network news stories, and 817 campaign newspaper stories. In summary, the study sought to find support for or against two hypotheses. H1 The greater an issue is mentioned by televised political advertisements in Time 1, the greater an issue is mentioned by network news programs in Time 2. H2 The greater an issue is mentioned by televised political advertisements in Time 1, the greater prominence of an issue within network newscasts in Time 2. Analysis of the Data The advertising, network news, and newspaper agendas are presented below followed by the findings for correlation and regression analyses. After a discussion of the statistical significance of each agenda, a discussion of the substantive significance of individual variables occurs in the discussion section that follows. Table 1 Combined Mentions of the Political Advertising Agenda: Top 15 Issues Time 1 Mentions Time 2 Mentions 1. Taxes 68 1. Leadership 140 2. Leadership 60 2. Political Process 74 3. Political Process 54 3. Taxes 45 4. Drugs 44 4. Children 34 5. Family 17 5. Family 32 6. Campaign 15 6. Elderly 28 Employment 15 7. Drugs 27 7. Economy 13 8. Morality 25 8. Budget 12 9. Education 21 9. Children 9 10. Immigration 20 Morality 9 11. Crime/Violence 19 10. Government Sp. 8 12. Campaign 16 11. Elderly 6 Employment 16 12. Education 5 13. Economy 15 13. Crime 3 14. Environment 11 Experience/Ability 3 Gun Control 11 Trade Issues 3 15. Budget 9 14. Compassion 2 Government Sp. 9 Environment 2 Welfare 2 15. Gun Control 1 Knowledgeable 1 N=40. Unit of observations = 116 advertisements. Total mentions, Time 1 = 352 and Time 2 = 574. Time 1: Sept. 2 - Sept. 29, 1996. Time 2 : Oct. 9 - Nov. 5, 1996. Table 2 Combined Time of the Political Advertising Agenda: Top 15 Issues Time 1 Seconds Time 2 Seconds 1. Political Process 211.2 1. Leadership 546.7 2. Leadership 203.0 2. Political Process 253.7 3. Taxes 195.6 3. Taxes 154.1 4. Drugs 166.0 4. Children 116.5 5. Family 63.9 5. Elderly 102.5 6. Morality 58.1 6. Family 101.6 7. Campaign 50.2 7. Morality 101.2 8. Children 39.3 8. Drugs 95.9 9. Employment 36.3 9. Education 64.3 10. Budget 33.1 10. Campaign 61.3 11. Economy 30.4 11. Crime/Violence 60.4 12. Government Sp. 29.3 12. Immigration 60.0 13. Experience/Ability 20.0 13. Economy 44.4 14. Compassion 15.0 14. Employment 41.1 15. Elderly 14.6 15. Environment 38.5 N=40. Unit of observations = 116 advertisements. Total time, Time 1 = 1215.0 seconds and Time 2 = 2010.9 seconds. Time 1= Sept. 2 - Sept. 29, 1996. Time 2 = Oct. 9 - Nov. 5, 1996. Advertising: Tables 1 and 2 show the major issues on the combined advertising agendas for the Clinton, Dole and Perot teams. In Table 1, taxes, leadership, political process, drugs, family, children, and morality were issues in the top ten ranks while campaign, employment, economy, budget and government spending all fell between time 1 and time 2. In the later time period, elderly, education and immigration all moved up to become top ten issues. The leadership category garnered a large portion of the issue mentions -- almost a quarter of the total mentions Table 3: Pearson Correlation of 1996 Presidential Political Advertisements Mentions Clinton Mentions T1 Clinton Mentions T2 Dole Mentions T1 Dole Mentions T2 Perot Mentions T1 Clinton Mentions T2 .61 Dole Mentions T1 .84 .35 Dole Mentions T2 .61 .56 .69 Perot Mentions T1 .08 .20 .04 .27 Perot Mentions T2 .35 .40 .29 .47 .92 Table 4: Pearson Correlation of 1996 Presidential Political Advertisements Seconds Clinton Seconds T1 Clinton Seconds T2 Dole Seconds T1 Dole Seconds T2 Perot Seconds T1 Clinton Seconds T2 .56 Dole Seconds T1 .80 .40 Dole Seconds T2 .66 .60 .80 Perot Seconds T1 .07 .18 .04 .29 Perot Seconds T2 .38 .50 .30 .54 .86 Table 2 focuses on the actual time each advertisement devoted to an issue. Political process, leadership, taxes, drugs, family, morality, campaign, and children all remained as top ten issues throughout the two time periods. Employment and budget dropped significantly while elderly and education emerged on the top ten issue agenda for the network news agendas at time 2. Pearson correlation coefficients in table 3 suggest a strong relationship between the two major party candidate advertising at time 1 with a r = .84. At time 2, the statistical relationship drops to a moderate one with a r = .56. There is a weak relationship between the Dole advertisements at time 1 and Clinton ads time 2 (r = .35). The reverse relationship is moderate: Clinton ads at time 1 and Dole at time 2 (r = .61). Moderate coefficients are also found between each candidate's advertising in time 1 and time 2. The issues in the Perot advertisements in time 1 have a weak relationship with major party candidate ads. Pearson coefficients range from .04 to .29 in tables 3 and 4. Perot mentions at time 2 were stronger in relationship to Dole and Clinton than in time 1 with coefficients range from .29 to .54. Perot mentions and seconds between time 1 and time 2 provide coefficients of .92 and .86 respectively. Advertising seconds in table 4 reflect the same positive relationships described above for the Democratic and Republican nominees. A fairly strong relationship (r = .80) at time 1 and a moderate one (r = .60) at time 2 between the major party candidates. A stronger relationship was seen between Clinton issues at time 1 and Dole time 2 (r = .66) than the reverse relationship (r = .40). The relationship between the two time periods for Dole advertising was higher (r = .80) than with the Clinton campaign (r = .56). Discussion: The advertising data reflects a number of substantively significant findings when put in the context of the 1996 campaigns. Changes from time 1 to time 2 in the sum of mentions and seconds are supported by the issues focused on by Clinton, Dole, and Perot. The leadership category is the number one category by far in time 2 as the candidates focused on their vision for America. For example, one of the themes of the Clinton campaign was his vision for "building a bridge to the 21st century" -- Dole often asked if was a toll bridge. Also seen in tables 1 and 2, immigration was an issue that rose from zero mentions to 20 mentions or 10th in the rank-order on the advertising agenda at time 2. A similar rise is seen in table 2. The immigration issue was substantively used by the Dole campaign, especially in California where Dole spent most of his resources in the closing month of the campaign. It was speculated by pundits the immigration issue would be a wedge issue in the golden state with its 54 electoral votes. Along with leadership, political process was a major focus of the Perot ads. Taxes, morality, and drugs were also major issues in Dole advertising. Clinton's camp focused on a variety of family and working class issues: children, family, elderly, and education. The Pearson correlation coefficients in tables 3 and 4 reflect a number of elements in the issue context of the 1996 campaign. Perot's weak correlation coefficients with the major candidates reflects his limited resources -- constrained by spending limits since accepting federal campaign money. The Perot team's issues had low coefficients at time 1 with the major party candidates while rising to moderate ones during time 2. In the context of the campaign, Perot was forced to focus on only a few issues because he could only buy so much air time. His advisors eventually bought 15-second spots instead of the normal 30-second spots in an effort to disseminate his message further with limited funds. Perot's coefficients between his own advertising at time 1 and time 2 was higher than the relationship between the major party candidates with their own advertising. This consistency of messages is substantially supported by two elements of the presidential race. First, Perot's messages was that of the "outsider" and he continued that theme throughout the two time periods. Second, the Perot team didn't have the resources or willingness to re-evaluate their messages and make daily strategic changes in advertising copy. Tables 3 and 4 reflect the dominance of the Clinton campaign in setting the issue agenda between the campaigns. The Clinton team had gained a reputation in 1992 to quickly respond to charges within one day's news cycle and this dominance was even further perfected in 1996. The strong coefficients between Clinton time 1 and Dole time 2 reflects the possibility the Dole campaign attempted to distinguish itself from the earlier views communicated by Clinton who used the bully pulpit very effectively to get his message out. This belief has further substantive support in light of the late leadership changes the Dole campaign made in their advertising team. The Clinton ads in time 2 showed a weaker relationship with Dole in time 1. Network News Agenda Table 5 shows the issues important on the network news agenda. Campaign-related coverage was dominated by the campaign's race aspects: who was leading, team strategy, polling results, and media buys. In time 1, campaign elements made up 60 percent of news mentions while in time 2 those elements made up 77 percent of coverage. Other issues that were top ten topics in both time periods: economy, education, taxes, political process, crime/violence, and leadership. Three issues fell considerably: drugs, foreign policy, and environment. In time 2, morality, elderly, budget, and civil rights/race relations emerged as important issues. Table 5 Combined Mentions of the Network News Agenda: Top 15 Issues Time 1 Mentions Time 2 Mentions 1. Campaign 204 1. Campaign 444 2. Drugs 35 2. Morality 135 3. Economy 27 3. Education 75 4. Education 26 4. Economy 52 5. Taxes 21 5. Elderly 30 6. Political Process 19 6. Taxes 15 7. Crime/Violence 18 7. Crime/Violence 13 8. Foreign Policy 12 8. Economy 11 9. Leadership 11 Leadership 11 10. Environment 10 9. Political Process 10 11. Budget 9 10. Budget 9 12. Immigration 8 Civil Rights/Race 9 Government Sp. 8 11. Gun Control 7 13. Abortion 7 12. Environment 6 Employment 7 13. Children 5 Morality 7 Drugs 5 14. Welfare 4 14. Family 4 U.S. Military Inv. 4 Government Sp. 4 15. Children 3 Immigration 4 Elderly 3 Poverty 4 Family 3 15. Compassion 4 Gun Control 3 Health 2 Poverty 3 Welfare 2 Trade Issue 3 Women's Issues 2 N=40. Unit of observations = 101 network news stories. Total number of mentions, Time 1 = 468 and Time 2 = 864. Time 1= Sept. 2 - Sept. 29, 1996. Time 2 = Oct. 9 - Nov. 5, 1996. Table 6 Combined Time of the Network News Agenda: Top 15 Issues Time 1 Seconds Time 2 Seconds 1. Campaign 1906.51 1. Campaign 4416.93 2. Economy 262.30 2. Morality 1015.62 3. Drugs 247.00 3. Education 464.45 4. Education 207.00 4. Economy 430.65 5. Political Process 173.61 5. Elderly 310.75 6. Employment 170.82 6. Taxes 152.00 7. Crime/Violence 113.78 7. Leadership 103.75 8. Taxes 109.58 8. Political Process 94.21 9. Foreign Policy 104.58 9. Civil Rights/Race 86.69 10. Environment 82.58 10. Employment 77.28 11. Leadership 67.73 11. Crime/Violence 75.40 12. Abortion 56.44 12. Budget 60.75 13. Budget 54.83 13. Gun Control 51.61 14. Immigration 53.73 14. Environment 43.28 15. Morality 48.76 15. Children 43.17 N=40. Unit of observations = 101 network news stories. Total seconds, Time 1 = 3971.6 and Time 2 = 7682.5. Time 1= Sept. 2 - Sept. 29, 1996. Time 2 = Oct. 9 - Nov. 5, 1996. Table 6 focusing on the actual time devoted to the issues reflects much the same issue landscape as table 5. Drugs, foreign policy, and environment drop out of the top ten in addition to another issue not seen in table 5: crime/violence dropping to 11th. Morality, elderly, and civil rights/race relations emerged as important issues in addition to a new one: leadership. Again, coverage focusing on the race aspects of the campaign reigned. Campaign race coverage made up 48 percent of network stories at time 1. The percentage increased in time 2 to more than 57 percent. Table 7 reflects the overall dominance of the campaign " horse race" aspects of network news coverage in relationship to other issues that make-up the predominate content of campaign stories on the network news agenda. In time 1, the campaign race category included 8 stories that were in the first segment (before the opening and first commercial) of the newscast. Overall, this reflects 8 out of the 42 prominent campaign stories or 19 percent Table 7 Prominence of Issue Stories in First Segment of Network News Program Time 1 # Stories Time 2 # Stories 1. Campaign 8 1. Campaign 36 2. Economy 3 2. Morality 21 Environment 3 3. Economy 6 Political Process 3 4. Leadership 5 Morality 3 5. Civil Rights/Race 3 3. Crime/Violence 2 Crime/Violence 3 Drugs 2 Drugs 3 Education 2 6. Children 2 Elderly 2 Family 2 Government Sp. 2 Government Sp. 2 Taxes 2 7. Education 1 Morality 2 Employment 1 4. Budget 1 Environment 1 Family 1 Gun Control 1 Health 1 Religion 1 Housing 1 Women's Issues 1 Immigration 1 Foreign Aid 1 Infrastructure 1 Trade Issues 1 Experience/Ability 1 N=40. Unit of observations = 101 network news stories. Total number of mentions, Time 1 = 42 and Time 2 = 98. Time 1= Sept. 2 - Sept. 29, 1996. Time 2 = Oct. 9 - Nov. 5, 1996. in the first segment were about the horse race aspects of the presidential campaigns. In time 2, the number of campaign horse race stories jump significantly as election day approaches and the network agendas turn their attention to the presidential contest. Nearly double the time 1 results, 36 of 90 stories in the top segment or 40 percent of the campaign-related stories now focused on the horse race aspects of the contest. Another significant number of stories was those focusing on morality -- 21 stories total. From there the number of stories focusing on an issue drops considerably with only 6 stories focusing on the economy, third in the rankings. Table 8: Pearson Correlation of 1996 Network Presidential Campaign Coverage Network Mentions T1 Network Seconds T1 Network Prominence T1 Network Mentions T2 .94 .94 .79 Network Seconds T2 .96 .96 .79 Network Prominence T2 .85 .85 .78 Note: The network news coverage is represented by randomly constructed weeks and it is appropriate to consider the statistical significance of the correlation coefficients. All of the coefficients in the table are statistically significant at the p _ .05 level. The other data sets, political advertisements and prestige press, are the universe of those stories and advertisements and are not samples. Table 8 shows strong correlation coefficients between the network news mentions, seconds, and prominence. With coefficients of .78 up to .96 when considering the 40 issue categories. In light of the rank-order tables of the network news agendas above, it is not surprising to find high correlation coefficients between these three levels of measurement between the two time periods. Discussion: A number of findings appear substantively significant in the context of the 1996 presidential race and the advertising agenda. Tables 5 and 6 indicate issues at time 2 -- the month before election day -- include many of the major issues of the advertising agenda. The morality issue focused on by the Dole camp late in the campaign was ranked number two on the network news agenda. A focus on the elderly -- a Clinton campaign issue -- was also high on the list. Medicare funding became a controversial issue with Clinton noting Dole's congressional votes in favor of program cuts and Republicans accusing Clinton of scaring older Americans. Leadership was another issue each candidate tried to claim with their own vision of America. The civil rights/race relations category emerged as a top issue in the time devoted to it as is seen in table 6. This issue is not a major issue at time 1 or at either time period in the network mentions categories (table 5). Yet, affirmative action became a very controversial issue in California where it was a referendum question focusing on the role of affirmative action in school admissions. This could explain the issue's rise on the rankings, especially as the presidential race focused on California and the candidates positioned themselves accordingly. One thing is clear -- campaign horse race coverage dominated all other issue categories. The rest of the findings of the network news agenda and the advertising requires further analysis of the issues to determine any influences on the network news agenda. Prestige Press Coverage Considering the campaign coverage by the Los Angeles Times, New York Times, and Washington Post, the rankings in table 9 reflect -- like their electronic counterparts -- campaign horse race issues dominated overall coverage. Campaign race mentions constituted more than 55 percent of those in time 1 and more than 60 percent of those of time 2. Fifteen issues make up the top ten at time 1 and 19 issues at time 2. Most remained consistently in the top group, but government spending dropped off considerably from between the two time periods. Defense spending, budget, women's issues, environment, and defense spending all emerged as major issues. The column inches for the combined prestige press newspapers are show in table 10. Not surprising, campaign horse race coverage dominated both time periods. Similar to the newspaper mentions table, government spending was a highly ranked issue at time 1, yet fell dramatically -- here out of the top 15 list. Yet, different than the mentions table, drugs and family fell out of the top ten between time 1 and time 2. Column inches of immigration, leadership, and education emerged as top issues in time 2. The majority of campaign-related coverage on the front pages of the newspapers consist of two issue categories (table 11). Again, stories about the campaign race, polling, strategy, and media buys dominated the number of position on both rankings. Campaign race coverage far outpaced all other issue stories making up 50 of 77 stories or 65 percent of time 1. At time 2, campaign race coverage made up 65 percent of campaign coverage. The only other category in double-digits was morality in time 2. Table 9 Combined Mentions of the Newspaper Agenda: Top 15 Issues Time 1 Mentions Time 2 Mentions 1. Campaign 201 1. Campaign 275 2. Drugs 26 2. Morality 57 3. Taxes 17 3. Leadership 13 4. Government Sp. 16 4. Civil Rights/Race 10 5. Crime/Violence 14 Foreign Policy 10 6. Economy 9 Defense Spending 10 7. Education 8 5. Crime/Violence 9 Foreign Policy 8 Immigration 9 8. Family 7 Taxes 9 Morality 7 6. Economy 7 9. Elderly 5 7. Education 6 Immigration 5 8. Budget 5 10. Abortion 4 Elderly 5 Civil Rights/Race 4 Women's Issues 5 Leadership 4 9. Drugs 4 11. Environment 3 Family 4 Traditional Values 3 10. Abortion 3 12. Employment 2 Environment 3 Infrastructure 2 Defense Spending 3 Poverty 2 11. Gay Rights 2 Religion 2 Government Sp. 2 Nuclear Arms Race 2 Health 2 U.S. Military Inv. 2 Infrastructure 2 13. Children 1 12. Poverty 1 Gay Rights 1 Welfare 1 Health 1 U.S. Military Inv. 1 Veteran's Issues 1 Compassion 1 Welfare 1 Compassion 1 Experience/Ability 1 N=40. Unit of observations = 817 newspaper stories. Total number of mentions, Time 1 = 364 and Time 2 = 453. Time 1= Sept. 2 - Sept. 29, 1996. Time 2 = Oct. 9 - Nov. 5, 1996. Table 10 Combined Inches of the Newspaper Agenda: Top 15 Issues Time 1 Inches Time 2 Inches 1. Campaign 4353.51 1. Campaign 8063.86 2. Drugs 567.56 2. Morality 1338.06 3. Taxes 349.22 3. Civil Rights/Race 418.69 4. Government Sp. 333.77 4. Immigration 281.73 5. Crime/Violence 279.22 5. Economy 245.05 6. Foreign Policy 207.52 6. Foreign Policy 236.37 7. Family 134.14 7. Leadership 224.39 8. Morality 130.61 8. Crime/Violence 164.51 9. Civil Rights/Race 113.47 9. Education 158.74 10. Economy 107.29 10. Taxes 158.32 11. Immigration 97.15 11. Women's Issues 125.19 12. Environment 95.98 12. Drugs 105.25 13. Traditional Values 90.63 13. Elderly 95.93 14. Elderly 70.66 14. Family 85.80 15. U.S. Military Inv. 53.61 Defense Spending 85.50 15. Budget 74.11 N=40. Unit of observations = 817 newspaper stories. Total number of inches, Time 1 = 7529.9 and Time 2 = 12196.0. Time 1= Sept. 2 - Sept. 29, 1996. Time 2 = Oct. 9 - Nov. 5, 1996. Pearson correlation coefficients in table 12 show very strong correlation coefficients between the three prestige presses coverage of the presidential candidates in 1996. The correlation coefficients suggest strong and positive relationships between the three newspapers in their coverage of the 1996 election. Table 11 Combined Prominence of Issues on the Newspaper Agenda: Time 1 # Stories Time 2 # Stories 1. Campaign 50 1. Campaign 91 2. Drugs 3 2. Morality 16 Economy 3 3. Civil Rights/Race 6 Immigration 3 4. Economy 5 Taxes 3 5. Education 3 3. Crime/Violence 2 Immigration 3 Government Sp. 2 Taxes 3 Foreign Policy 2 Leadership 3 Morality 2 6. Women's Issues 2 Foreign Policy 2 7. Budget 1 Crime/Violence 1 Drugs 1 Government Sp. 1 Traditional Values 1 Defense Spending 1 N=40. Unit of observations = 817 newspaper stories. Total number of page one stories, Time 1 = 77 and Time 2 = 140. Time 1= Sept. 2 - Sept. 29, 1996. Time 2 = Oct. 9 - Nov. 5, 1996. Discussion: The results from the newspaper agendas reflect a consistency seen in the network news coverage: predominant coverage of the race aspects of the presidential race. Several other issue categories at time 2 are especially noteworthy. First, the top non-horse race issue on the network news agenda was also seen in the newspaper mentions and seconds. Substantially, this supports the belief by the Dole team stated publicly numerous times during October that they would focus on making the "character" issue stick on Clinton. In addition, this was an issue the Perot team attempted to bring into focus. Table 12: Pearson Correlation of 1996 Presidential Campaign Coverage by Newspaper Column Inches Los Angeles Times T1 New York Times T1 Washington Post T1 Los Angeles Times T2 .98 .98 .97 New York Times T2 .96 .96 .96 Washington Post T2 .98 .98 .97 After campaign horse race and morality, newspaper inches includes civil rights/race relations, immigration, economy, foreign policy, and leadership. Again, as the race focused more on California and both major party candidates aimed their advertising on those media markets, civil rights/race relations (affirmative action) and immigration become important issues as the candidates lined up their positions on those two controversial issues for the golden state. Like many presidential races before, the economy was an important issue with Clinton claiming his leadership was in part responsible for the strong economy while Dole criticized his policies. In addition, foreign policy in the Middle East and other parts of the globe were debate topics. Consequently, leadership was high on the newspaper agenda with all three candidates try to tie-in issues with their leadership in the past and their vision for the future. Television has been called a headline service and no doubt newspapers are often better vehicles for more complex stories. This could have played a role in the differences between the newspaper agenda and the network news agenda. Specifically, the network news agenda had foreign policy in their top ten listings of mentions and seconds at time 1. Yet, at time 2 the issue didn't appear on the top 15 rankings. On the newspaper agendas, the same issue appeared in the top ten in both time periods in mentions and inches. Furthermore, newspaper inches of foreign policy rose from the seventh place in time 1 to the 4th place in time 2. This discussion also makes another point -- the need to go beyond mere rankings to more sophisticated measures of analysis. The analyses above point to two major findings: newspaper campaign coverage consistently reflects a positive strong relationship between the three newspapers included in the study, and the relationship between the newspaper, advertising, and network news agendas have some issues in common. Below a more statistically sophisticated level of analysis is used to better understand the relationships between the variables. Regression Analysis Network News Mentions and Seconds: The advertising agenda, network news agenda, prestige press agenda, and real-world indicators at time 1 were all considered for their predictive ability in determining the network news agendas at time 2. In table 13, the first two ordinary least squares models and their weighted least squares transformations show the coefficients for network news mentions at time 2 regressed on advertising mentions, prestige press stories, and network news mentions at time 1. Two variables are statistically significant in at least two of the models. Advertising mentions in the first model is statistically significant and produces a negative coefficient of -00.42. In the other three models, with network news mentions at time 2 as the dependent variable, this variable is not statistically significant. In all of the first four models, network news at time 1 is statistically significant, positive, and has a substantial coefficient ranging from 1.41 to 2.99. In the OLS models, the variable is significant at the .05 level. The addition of the real-world indicator variable in the model produced a coefficient not statistically significant. The real-world coefficients reduced the adjusted R2 of the models. The standard errors were reduced for the OLS models and increased for the WLS models. Clearly, their inclusion in the models is of questionable analytic value. In the remaining four models with network news seconds at time 2 regressed on the predictor variables, two variables are statistically significant. Prestige press inches is significant at the .20 level in the OLS model. Yet, the model could be problematic due to heteroscedasticity. The weighted least squares model attempting to address this problem transforms the variable into one that is not statistically significant. Again, the results show the network news coefficients -- measured as seconds in the final four models -- are statistically significant in three of the models. The network news seconds at time 1 produce coefficients ranging from 1.48 to 1.76. The introduction of the real-world indicators using a subset of the data produces coefficients not statistically significant. The adjusted R2 and the standard error of the model were reduced in the OLS models. Yet, the standard error increased in the WLS models when real-world indicators were added. Table 13 News Mentions and Seconds at Time 2 Regressed on Advertising, Prestige Press, and Real World Indicators Dependent Variable Network News Network News Mentions T2 Seconds T2 Independent Variables T1 OLS WLS OLS WLS OLSb WLS OLS WLS Constant term 01.22 00.85 02.50 02.78 -02.13 00.24 52.59 08.44 (4.58) (1.37) (9.91) (3.54) (35.52) (2.92) (94.72) (14.80) Ad mentions -00.42* -01.48 -00.08 -09.64 (0.25) (3.52) (1.28) (13.01) Ad seconds -0.69 -03.12 -03.52 -34.80 (0.53) (23.15) (7.05) (110.01) Prestige press stories 00.60 00.57 -01.42 -02.27 (0.81) (0.89) (2.07) (2.13) Prestige press inches 0.36* 00.18 -0.11 -0.23 (0.22) (0.27) (0.71) (0.69) Network news 01.59*** 01.41** 02.63*** 2.99* 01.48*** 01.65***1.76* 01.51 (0.80) (0.81) (1.03) (1.55) (0.51) (0.57) (0.97) (1.36) Real world indicators -0.56 -01.19 -03.94 -12.53 (0.66) (1.94) (6.41) (22.45) R2 89.70 69.60 75.70 75.20 93.30 79.00 56.50 55.60 Adjusted R2 88.80 64.50 56.30 25.60 92.70 75.50 21.60 00.00 Standard Error of Model 24.34 04.30 16.93 04.42 191.60 09.86 162.70 15.28 Durbin-Watson 00.93c 01.36d 01.34d 02.27d 01.04c 01.65d 01.96d 02.46 N 40 40 10 10 40 40 10 10 Note: The Weighted Least Squares are transformed variables constructed by taking the OLS coefficients and dividing through by the Ad Mentions or Seconds. This is a remedial measure designed to reduce heteroscedasticity. In addition, autocorrelation is especially problematic in data across time. These ad variables were chosen for the transformations because they are theoretically linked to network news mentions and seconds. Also, regression diagnostics indicated they are a good choice for the transformation. Without this transformation, OLS coefficients with autocorrelation and heteroscedasticity present are biased and unreliable (see Gujarati, 1995, pp. 366-367, 411). *p _ .20 aBreusch-Pagan-Godfrey test indicates heteroscedasticity is likely in OLS model. **p _ .10 b Heteroscedasticity is possible in OLS model. ***p _ .05 cIndicates autocorrelation is likely. d Indicates autocorrelation is possible. Standard errors in parentheses. Time 1= Sept. 2 - Sept. 29, 1996. Time 2 = Oct. 9 - Nov. 5, 1996. Discussion: The results from table 13 would appear to not support the substantive significance of advertising mentions and seconds on the network news agenda. The one statistically significant variable would seem to support the reverse hypothesis. In addition, the one prestige press variable that is statistically significance is problematic because of heteroscedasticity. Its statistical significance disappears in the WLS model. Most noteworthy is the size and statistical significance of the network news variables. These variables support the routinization of news beats and other influences of the gatekeeping process that would encourage coverage of the same issues over time. Substantively, the importance of this variable is reflective of the consistent agenda with which networks cover presidential campaigns and their candidates. Network News Prominence: The models in table 14 focus on the explanatory power of variables to determine which issues were covered in the first segment of the network newscast. Advertising mentions and prestige press stories both had statistically significant variables. Ad mentions were significant at the .20 level in the OLS model with the real-world indicators included. The prestige press variable was significant in three of the four models. When the first model was transformed due to an indication of heteroscedasticity, the prestige press variable increased slightly and remained significant at the .05 level. The only other variable that appears statistically significant is network news prominence, but in the transformed model it loses its statistical significance. The last four models consider advertising seconds and prestige press inches. Advertising seconds is statistically significant at the .05 level in the WLS model when the OLS model indicates the presence of heteroscedasticity. In the second OLS model -- with real-life indicators included -- it remains statistically significant, but at the .20 level. Prestige press inches is a statistically significant variable in three of the four models. Since the variables are measured on different units of measurements (i.e. newspaper inches versus advertising seconds) it is difficult to interpret the size of the variables. Yet, the prestige press inches appear to be small -- only .01 across all four models. Network news prominence at time 1 is statistically significant in one of the models. When the model is transformed, this significance disappears. Unlike the network news mentions and seconds at time 1 discussed in table 13, the dependent variable presented at the earlier time period for news prominence didn't produce consistently statistically significant coefficients. Discussion: Two independent variables appear significant in the models presented in table 14. Advertising mentions and seconds, and prestige press stories and inches are statistically significant. Greater focus should be placed on the first and third WLS models where transformation supported by the presence of heteroscedasticity in the original OLS models, and the second and fourth OLS models where heteroscedasticity is not evident. Focusing on those models, prestige press stories are statistically significant at the .05 level in two models and advertising seconds in two. Also, ad mentions is statistically significant in one of those models. Of special note is the third WLS model in the table. Advertising seconds is significant at the .05 level and has a coefficient of .62. This weighted model is appropriate because the original OLS model indicates heteroscedasticity. The transformed model has an R-squared of 66.70 and a standard error of the model is low at .11. Moving from the total time and mentions in table 13 to the prominence of which issues are considered important enough to be included in the lead block of stories (table 14), the greatest agenda-setting influences are support by the regression analyses. The statistical significance of the dependent variable at time 1 is gone except in two biased OLS models. Of the other independent variables, only prestige press and advertising seconds are statistically significant at the .05 level. More importantly, the substantive significance of these models is supported by the theoretical basis of the study. Major newspapers setting the agenda of network newscasts is certainly not an astonishing finding. Considering the variables in the models in table 14, for every six stories in the prestige press newspapers at time 1 there would be one network news story either leading the news or soon following it in the lead block. Prestige press inches follows that for every 100 inches of newspaper coverage there would be one network news story in the lead block. Of even more substantive significance is the importance of the advertising mentions and seconds. For every six times an issue like the economy was mentioned in advertising at time 1, there was a network news story whose predominate focus was the economy in the first block of the newscast at time 2. On the level of advertising seconds, every two seconds devoted to an issue by one of the candidates in their advertising spots at time 1 would produce 1.24 stories about the issue at time 2 in the lead segment of the network newscast. It is logical that campaign advertising is representative of larger campaign efforts. Since most presidential advertising is only 30 seconds, every second is important real estate and its occupation is planned extensively by media consultants. Table 14 supports the Table 14 News Prominence at Time 2 Regressed on Advertising, Prestige Press, and Real World Indicators Dependent Variable Network News Prominence T2 Independent Variables T1 OLSa WLS OLS WLS OLSa WLS OLS WLS Constant term 00.08 00.13 -01.19* 00.22 00.06 00.01 -01.21 00.08 (0.66) (0.18) (0.81) (0.21) (0.67) (0.03) (0.84) (0.11) Ad mentions -00.01 00.18 00.16* -00.88 (0.04) (0.42) (0.09) (0.69) Ad seconds 00.01 00.62*** 00.07* -00.54 (0.01) (0.24) (0.04) (0.78) Prestige press stories 00.13*** 00.17*** 0.18 00.31* (0.03) (0.05) (0.14) (0.11) Prestige press inches 00.01*** 00.01***00.01 00.01* (0.00) (0.00) (0.01) (0.01) Network news prominence 01.23** 00.04 00.50 -00.72 01.15** -00.14 00.70 -00.11 (0.65) (0.38) (0.64) (0.83) (0.65) (0.20) (0.55) (0.81) Real world indicators 0.07 00.12 00.06 -00.00 (0.05) (0.12) (0.05) (0.14) R2 76.40 47.70 73.90 89.60 76.30 71.40 72.6 79.2 Adjusted R2 74.50 39.00 53.00 68.90 74.30 66.70 50.7 37.5 Standard Error of Model 03.26 00.55 01.34 00.21 03.27 00.11 01.37 00.10 Durbin-Watson 01.05c 01.20c 01.99d 01.87d 01.14c 01.32c 02.13d 01.59d N 40 40 10 10 40 40 10 10 Note: The Weighted Least Squares are transformed variables constructed by taking the OLS coefficients and dividing through by the Ad Mentions or Seconds. This is a remedial measure designed to reduce heteroscedasticity. In addition, autocorrelation is especially problematic in data across time. These ad variables were chosen for the transformations because they are theoretically linked to network news mentions and seconds. Also, regression diagnostics indicated they are a good choice for the transformation. Without this transformation, OLS coefficients with autocorrelation and heteroscedasticity present are biased and unreliable (see Gujarati, 1995, pp. 366-367, 411). *p _ .20 aBreusch-Pagan-Godfrey test indicates heteroscedasticity is likely in OLS model. **p _ .10 b Heteroscedasticity is possible in OLS model. ***p _ .05 cIndicates autocorrelation is likely. d Indicates autocorrelation is possible. Standard errors in parentheses. Time 1= Sept. 2 - Sept. 29, 1996. Time 2 = Oct. 9 - Nov. 5, 1996. belief that their decision to devote two seconds of a 30-second spot to an issue could influence the network news agenda. Conclusions and Implications In review, the study sought to test the hypotheses below. H1 The greater an issue is mentioned by televised political advertisements in Time 1, the greater an issue is mentioned by network news programs in Time 2. H2 The greater an issue is mentioned by televised political advertisements in Time 1, the greater prominence of an issue within network newscasts in Time 2. Hypothesis 1 The first hypothesis sought to investigate whether support could be found for the belief that political advertisements had an influence on the amount of time or mentions devoted to an issue during network newscasts of the 1996 presidential campaigns. The findings would appear to be inconclusive of supporting the belief in the ability of advertising to influence overall network news content. Based on the consistency across the models, the findings are unable to determine support for the hypothesis. In addition, only one of the prestige press coefficients are statistically significant. The strongest support is found for the influence of the network news agenda at the previous time period influencing time 2. In summary, the analyses does not support the influence of television mentions and seconds influencing the network news agenda. Hypothesis 2 The second theoretical statement sought to investigate whether support could be found for the belief that political advertisements had an influence on the decision of media gatekeepers in the placement of stories in the most important part of the newscast. The study finds support -- although not overwhelming -- in favor of this belief. With news prominence as the dependent variable, three of the statistically significant advertising coefficients in the regression analysis were positive. Also, six of the prestige press variables were statistically significant. Two of models found support for the network news prominence at time 1 as an indicator of time 2 prominence. In summary, the analyses here supports the belief that the content of the political advertisements at time 1 influence the placement of news stories devoted to those issues in the lead-off segment of the network newscasts. Implications for Intermedia Agenda-Setting The research hypotheses focused on the influence of televised issue advertising on the network news coverage of a presidential political campaign. Three major implications are suggested by the findings. First, influences of other media are suggested when network news prominence is the dependent variable. It is a logical conclusion that where the influences of agenda-setting would be found the most would be in the placement of issue stories within a newscast. The overall number of mentions and time in a newscast would be expected to be less impacted by advertising influence. This study does not focus on the public agenda, but research considering the public agenda suggests the importance of stories that lead the newscast. Iyengar and Kinder (1987) found issues focused on in the lead stories were issues important on the public agenda. Also, the researchers found stories contained elsewhere in the network news programs did not have any influence on the public agenda. In the context of the 1996 presidential campaign coverage, the Iyengar and Kinder (1987) study supports the significance of issues contained in the lead block of the network newscasts. Second, support found in this study -- while not overwhelming -- would indicate the influence of presidential political advertisements in the 1996 presidential campaign coverage. This finding supports one of the only similar studies focusing on the intermedia agenda-setting influence of advertising (Roberts & McCombs, 1994). Furthermore, it is important to note Roberts and McCombs observation that advertising acted as a "parsimonious surrogate for the campaign" (1994, p. 258). In the 1996 presidential campaign, it is logical to note that the advertising variables are the variables representing the influence of the overall campaign messages. Political campaigns develop messages used in paid media (advertising) and earned media (news coverage). Hopefully, between the two channels their message will resonant with voters. When political advertising works best is when it receives news attention and thus the message is disseminated well beyond the reach of the original advertising audience. Standard procedure in large campaigns -- especially presidential ones -- is to release advertising spots directly to journalists, often before they air on television. This is sometimes done for strategic reasons to get the other campaign to expend resources to develop, produce and air response advertising for an ad critical of their candidate. Campaigns have been know to release spots with much fanfare to get the other campaign to expend their resources for an advertisement that -- unlike originally stated -- only runs briefly. Important in considering intermedia agenda setting is that campaigns directly release advertisements to the media. Consequently, journalists are exposed to advertisements and their issue content regardless of how many times the spots actually ran in media markets. Third, the prestige press variable was significant in an overwhelming number of the models. This finding is supported by other intermedia agenda-setting studies. Newspapers, those in this study and smaller dailies, have been found to lead network news coverage of issues and campaigns (Merriam, 1989; Reese and Danielian, 1989; Danielian and Reese, 1989). Even the seminal agenda-setting analysis by McCombs and Shaw (1972) found a moderate to strong relationship between the newspaper and network news agendas. Overall, the network news agenda from the two time periods support the belief that television is a headline service taking cues from other sources -- advertising and the prestige press. In considering the total time and mentions, there is some support for the consistency of the network news agenda from time one to time two. The campaign issues on the network news agenda at time one have predictive value for all of the issues in the newscast at time two, but this relationship disappears when greater focus is put on the lead story or block (issue prominence). Issues may remain on the network news agenda for two months time because of the routinization of news beats and other factors influencing decisions by network gatekeepers. Yet, the networks survey the news media environment to help decide what issues will get their greatest attention. In the future, greater attention needs to be focused on how issues rise on the network news agenda to become lead stories. Bibliography Behr, R. L., & Iyengar, S. (1985). Television news, real-world cues, and changes in the public agenda. Public Opinion Quarterly, 49, 38-57. Brians, C. L., & Wattenberg, M. P. (1996). Campaign issue knowledge and salience: Comparing reception from TV commercials, TV news, and newspapers. American Journal of Political Science, 40(1), 172-193. Danielian, L. H., & Reese, S. D. (1989). Chapter 6: A closer look at intermedia influences on agenda setting: The cocaine issue of 1986. In P. J. Shoemaker (Ed.), Communication campaigns about drugs, Government, media and the public (pp. 47-66). Hillsdale, NJ: Lawrence Erlbaum. Gujarati, D. N. (1995). Basic econometrics. (3rd ed). New York: McGraw-Hill. Hall, M. (1997). One-to-One Politics in Cyberspace. Media Studies Journal, 11(1), 97-103. Iyengar, S., & Kinder, D. R. (1987). News that matters: Agenda-setting and priming in a television age. Chicago: University of Chicago. McCombs, M. E., & Estrada, G. (1997). The news media and the pictures in our heads. In S. Iyengar & R. Reeves (Eds.), Do the media govern? Politicians, voters, and reporters in America (pp. 237-247). Thousand Oaks, CA: Sage. McCombs, M. E., & Shaw, D. L. (1972). The agenda-setting function of mass media. Public Opinion Quarterly, 36, 176-187. North, R. C., Holsti, O. M., Zaninovich, M. G., & Zinnes, D. A. (1963). Content analysis, A handbook with applications for the study of international crisis. Evanston, IL: Northwestern University. Patterson, T. E., & McClure, R. D. (1976). The unseeing eye: The myth of television power in national politics. New York: G.P. Putnam's Sons. Reese, S. D., & Danielian, L. H.(1989). Chapter 5: Intermedia influence and the drug issue: Converging on cocaine. In P. J. Shoemaker (Ed.), Communication campaigns about drugs, Government, media and the public (pp. 29-45). Hillsdale, NJ: Lawrence Erlbaum. Roberts, M. (1992). Predicting voting behavior via the agenda-setting tradition. Journalism Quarterly, 69(4), 878-892. Roberts, M., & McCombs, M. (1994). Agenda setting and political advertising: Origins of the news agenda. Political Communication, 11, 249-262. Rogers, E. M., Dearing, J. W., & Bregman, D. (1993). The anatomy of agenda-setting research. Journal of Communication, 43(2), 68-84.