Retesting the Marketplace Theory of Media Use—9

factor was identified (eigenvalue=1.699, variance explained=42.472). The reliability quotient was

low (alpha=.52). For Newspaper B, one factor was identified (eigenvalue=1.841, variance

explained=46.024). The reliability quotient was also somewhat low (alpha=.60).

*Attitudes toward journalism. *Beaudoin and Thorson (2002) found two dimensions to

attitudes toward journalism: attitudes toward diversity aspects of journalism and attitudes toward

financial aspects of journalism. There were two statements for diversity aspects (see Appendix).

They did not, however, appear to form useful dimensions (Newspaper A, corr.= .115, p < .001;

Newspaper B, corr. = .175, p < .001). Although significant, these correlations are low and, thus,

were deemed to be unreliable measures of attitudes toward diversity aspects of journalism. (In

addition, factor analysis failed to offer different dimensions to these two attitude groupings.)

Thus, we dropped these two statements from further analysis. We implemented two statements to

measure attitudes toward the financial aspects of journalism (see Appendix): Newspaper A

(corr.=.406, p<.001) and Newspaper B (corr.=.463, p<.001). In each case, a higher score indicates

a more positive (or more socially conscious) response.

In addition, demographics were used as control variables. Education was measured on a

4-point scale from “high school grad or less” (1) to “post-graduate training” (4), and income was

measured on a 5-point scale from “less than $25,000” (1) to “$150,000 and more” (5). Also, there

were measures for gender (M=), ethnicity (W=1), and age.

**Analysis **

We conducted structural equation modeling (SEM) with AMOS 4.01. SEM involves

simultaneous regressions and estimates direct and indirect effects. We tested the specified models

for each of the two newspapers. We used two model fit indices—the comparative fit index (CFI)

and the normed fit index (NFI). Values fall between 0 and 1, with 0.95 indicating a good fit

between the observed data and the specified model (Kaplan, 2000). By using critical ratios to

evaluate path significance, we determined if the ratio of an estimate to its standard error exceeded

1.960 and 2.576 (indicating significance at the .05 and .01 levels, respectively) (Hoyle, 1995). In