Abstract
This study aims to explore second-level agenda-setting at the national level. In particular, it examines the relationships among the citation bias of the New York Times, national level public opinion, and Congressional policies from 1956 to 2004 in order to better understand mass media's role in national policymaking. In addition, it also tests one important intervening variable of the relationships among the three attribute agendas (the media agenda, the public agenda, and the policy agenda)—the president's policy liberalism.
Notes
Note. The media bias score is calculated by the equation: (brookings *53.3 + civil *49.8 + naacp *75.4 + strategic *46.3 + amnesty *57.4 + foreign *60.2 + sierra *68.7 + enterprise *36.6 + rand *60.4 + rifle *45.9 + retired *66 + carneige *51.9 + heritage *20.0 + urban *73.8 + responsive *66.9 + consumer *81.7 + christian *22.6 + cato *36.3 + women *78.9 + intereco *48.8)/total number = media bias score. NYT = New York Times.
1Groseclose and Milyo (2005) used several multinomial logistic models to estimate parameters and seed values to compute the ADA scores. Please see details of their models in pages 1208 to 1211 of their article. The values calculated in this study are systematically lower than the ADA scores calculated in Groseclose and Milyo's study. In particular, their average ADA score of the New York Times from 7/1/01 to 5/1/02 is 73.7, whereas our scores are 56.21 for 2001 and 56.72 for 2002.
2The forecast values of the time series are called impulse response functions, revealing what happens to the time series in the future, if each shock goes up by one unit at t. By using the estimates, this exercise can be simulated a number of times and 95% confidence intervals can be constructed.
Note. The numbers in the table are Z statistics; the numbers in parentheses are p value. VAR contains four time lags. There were 37 observations in the series, running from 1960 to 1996. IV = independent variable; DV = dependent variable.
3The impulse response functions involve introducing a shock to the impulse variable in the system and tracking movements in the response variable using the VAR estimates for computing a forecast. Shocking a variable means increasing the independent series by 1 standard deviation and estimating the impact the increase has on the other series in the system (Peake, Citation2001). In Figure , each chart represents the direction and magnitude of the response variable over 8 years to a 1 standard deviation shock in the impulse variable. Calculated with the adjusted ADA scores, 1 standard deviation is 4.3 for media bias, 4.3 for public mood, and 7.8 for policy activity liberalism. Those in the first column illustrate the effects of each agenda of a 1 standard deviation impulse in media bias; graphs in the next two columns do the same for public mood and policy activity liberalism, respectively.
4The value of Y axis for the 0 on X axis is the response in the 1st year after the shock; the value for 1 is the response in the 2nd year after the shock, and so on.