Portal de Periódicos da CAPES

Late-Night Comedy as a Gateway to Traditional News: An Analysis of Time Trends in News Attention Among Late-Night Comedy Viewers During the 2004 Presidential Primaries

2008; Taylor & Francis; Volume: 25; Issue: 4 Linguagem: Inglês

10.1080/10584600802427013

1091-7675

Lauren Feldman, Dannagal G. Young,

Media Influence and Health

Abstract Recent reports published by the CitationPew Research Center for the People and the Press (2000, Citation2004) propose that young audiences are abandoning traditional news as a source of election information in favor of late-night comedy programs. However, additional evidence (CitationYoung & Tisinger, 2006) suggests that exposure to late-night comedy programming is positively correlated with traditional news exposure. This study extends this body of research by offering evidence that exposure to late-night comedy is associated with increases in attention paid to the presidential campaign in national network and cable news. The analysis uses data collected via the CitationNational Annenberg Election Survey during the 2004 presidential primary season, between October 30, 2003, and June 4, 2004. Cross-sectional results demonstrate that viewers of late-night comedy programs—specifically viewers of The Tonight Show with Jay Leno and The Late Show with David Letterman, as well as Comedy Central's The Daily Show with Jon Stewart—pay more attention to the campaign in national network and cable news than nonviewers, controlling for a variety of factors. An analysis of time trends also reveals that the rate of increase in news attention over the course of the primary season is greater for viewers of Leno or Letterman than for those who do not watch any late-night comedy. Keywords: late-night comedysatirenews attractiongateway hypothesis Notes 1. This and subsequent joke content was compiled as part of a content analysis conducted by the second author. 2. Although our measurement and analyses treat Daily Show viewers, Leno/Letterman viewers, and those who do not watch any late-night comedy as three distinct subgroups, it is possible that people who report most often watching The Daily Show also watch Leno or Letterman on occasion, and vice versa (see CitationPrior, 2007, for a similar discussion involving cable news audiences). To acknowledge the possibility of cross-viewership, we label the two late-night comedy viewing subgroups as "Daily Show dominant" and "Leno/Letterman dominant." 3. We recognize that the large sample size of the NAES could render even small effects statistically significant. For this reason, we ran a second regression analysis that was confined to respondents between the ages of 18 and 29 (N = 5,879), which is the age group considered both to be the primary audience for late-night comedy and the least attentive to traditional news (Pew Center, 2000, 2004). In this analysis, the relationship between late-night viewing and news attention remains significant among Daily Show dominant viewers (β = .06, p < .001) and Leno/Letterman dominant viewers (β = .05, p < .001). Of note, with this younger subset, the association between late-night viewing and news attention is even stronger than it was with the full sample. In fact, interactions between age and both the Leno/Letterman and Daily Show variables in the full sample model were significant and negative (both βs = −.04, p < .01), indicating that the relationship between late-night comedy viewing and news attention is stronger for younger than older adults. 4. The average daily sample sizes for each subgroup were as follows: non-late-night (M = 126.6, SD = 40.5), Leno/Letterman dominant (M = 46.8, SD = 16.3), and Daily Show dominant (M = 6.3, SD = 3.2). 5. Moving averages, which pool data across days, reveal important patterns in the data that would otherwise be obscured by sampling error (CitationJohnston, Blais, Brady, & Crete, 1992). The centered moving average for a particular day is that day's value averaged with a specified range of values around it. For example, the 21-day centered moving average is the average value obtained from 10 days before through 10 days after a given day. This is why, in Figure 1, the trend lines begin and end 10 days after/before the true endpoints of the campaign period being analyzed (i.e., October 30, 2003, and June 4, 2004). 6. Although the trend in news attention among Daily Show dominant viewers appears to fluctuate toward the end of the campaign, a closer examination of the time series using autocorrelation techniques suggests that this movement is random error, likely owing to the variation in daily sample sizes for the Daily Show subpopulation during this time. After removing the linear and quadratic trends, the series is comprised only of white noise (i.e., it has a mean of 0, constant variance, and no serial correlation). This indicates that there is no systematic relation between time periods beyond the general linear and quadratic trend. 7. The discontinuity in the time trend appears in the graph on January 13 rather than on January 19. Although this slight discrepancy could be attributed to anticipation of the Iowa caucuses, it is more likely due to the fact that a centered moving average incorporates the values of data following a given day, and therefore changes tend to appear in the graph before they actually occurred (CitationKenski, 2006b). Thus, in this case, a change in attention levels on January 19 was plausibly picked up in the 21-day centered moving average on January 13. 8. A cubic component was also tested but was not significant. 9. Note that when the Iowa predictor is added in Model 2, the linear trend is reduced to nonsignificance. Although this might suggest that most of the linear trend in news attention for Leno/Letterman dominant viewers and non-late-night viewers is captured by the effect of the Iowa caucuses, this should be interpreted cautiously because of the collinearity between the day and Iowa predictors (estimates for day2 should be unaffected since centering was used to reduce its correlation with day). Multicollinearity also likely explains why the estimated effects of the linear and quadratic predictors drop to nonsignificance when the effect of the Iowa caucuses is added to the Daily Show model. 10. The test statistic is obtained by calculating the difference between the estimates of two slopes (or intercepts) and dividing this by the standard error of the difference in slopes (or intercepts), as captured in the following formula: The computation of the denominator, S B 1 − B 2 , involves pooling and summing the residual mean-square errors for each regression equation. The complete formulas used for comparing slopes and intercepts can be found in CitationKleinbaum et al. (1998, pp. 322–326). Necessary computational statistics are provided in Appendix B. The t statistic obtained is evaluated using (N 1 + N 2 – 2k) degrees of freedom, where N is the number of days in each series and k is the number of parameters in each model. 11. Note that nearly identical results were obtained when comparing the constants across the Model 1 equations. In both cases, the significant differences detected between the intercepts for Daily Show dominant and non-late-night viewers, and between Leno/Letterman dominant and non-late-night viewers, withstand adjustments for multiple comparisons. Using the Bonferroni correction, which divides the Type I error rate (0.05) by the number of comparisons (here, three), .017 becomes the criterion for significance. The p values obtained fall well below this cutoff. 12. Prior to conducting the regression, we examined the time series and determined that it did not contain any AR or MA components. 13. A quadratic time trend was also tested but was not significant. 14. Interactions between time and differences in both age and education were also tested; neither was significant. This further suggests that the audience differences in age and education did not, over time, become more or less important in predicting differences in news attention. 15. The nonsignificant effect of the Iowa caucuses among Daily Show dominant viewers does not necessarily indicate that this effect is significantly less than that for the Leno/Letterman dominant group. Thus, we used a t test to compare the coefficients for the Iowa effect found for Leno/Letterman dominant viewers and Daily Show dominant viewers. This difference was not significant; however, given the relatively smaller daily sample sizes and wider sampling variability of the Daily Show audience, we likely did not have ample statistical power to detect a difference here. Similar conclusions can be drawn with regard to differences in the size of the quadratic effects across the Daily Show and Leno/Letterman equations, which were also not significant.