This preview shows page 1. Sign up to view the full content.
Unformatted text preview: however, as to whether this relationship is meaningful
in light of the potential impact of other variables. In Block 2, the ﬁrst of two sets
of covariates is entered into a regression equation along with religious concentration. With the inclusion of these 11 covariates in the analysis, the impact of
religious concentration on divorce, that is, β = –.136, is only slightly reduced from
the zero-order effect, that is., –.144, and remains signiﬁcant at the < .05 level of
probability. Thus, there is only a 5.6 percent reduction in the overall effect of
religious concentration on divorce when controlling for 11 of the factors that are
known to be associated with divorce. In short, the degree of religious concentration in the United States still has a statistically signiﬁcant and independent
effect on the rate of currently divorced vis-à-vis this set of variables.
This regression analysis also shows that the divorced rate (signiﬁcant at the
p < .05 level) is directly related to percentage population change, percentage
Asian/Paciﬁc Islander, percentage unemployed, percentage employed in manufacturing jobs, and percentage urban. The same analysis shows that the divorced
rate (again signiﬁcant at the < .05 level probability) is inversely related to
percentage of persons aged between 15 and 34, percentage Hispanic, percentage
of females in the population, and median family income. Thus, the ﬁndings
from previous studies regarding inﬂuences on divorce are generally supported
in this research, including all four of the “economic instability” measures. On the
other hand, there is an absence of statistically signiﬁcant relationships between
persons currently divorced and percent white and percent Native American.
By examining the Beta coefﬁcients, that is, the standardized partial regression coefﬁcients, in Block 2 of Table 2, one can assess the relative impact of
each of the 12 independent variables in the regression equation on the dependent
variable. Percent urban (β = .397) and percent population change ( β = .372) clearly
have the greatest impact on the divorce rate. When these 12 variables are rankordered, religious concentration ranks seventh relative to impact on the rate of
Block 3 shows the regression analysis with “region” (i.e., West, Northeast,
and South, with Midwest as the criterion variable) entered as additional covariates
beyond those examined in Block 2. Because region has been shown to be an
important contextual variable in the examination of divorce and of religious
orientation, the analysis examines region as an added emphasis.
To test the impact of region, a separate analysis of variance procedure was
conducted in order to determine if there are signiﬁcant differences among
regions in the currently divorced rate. The results conﬁrm the importance of this
contextual variable. The mean divorce rate in the West (91.51) is signiﬁcantly
greater ( p ≤ .05) than the rates in each of the other three regions (Midwest 350 LARRY C. MULLINS ET AL. rate = 68.91, Northeast rate = 71.25, and South rate = 76.85). The mean divorce
rate in the Midwest is signiﬁcantly less than rates in the West and South, but not
signiﬁcantly different from the Northeast. The rate in the Northeast is signiﬁcantly
lower than the West’s rate, but not that different from the other two. The South’s
rate is signiﬁcantly less than the West’s rate, but signiﬁcantly greater than the
rate for the Midwest. Thus, as expected the divorce rate varies among regions.
Again, as in Block 2, the overall effect of religious concentration on the
divorced rate, after controlling for all covariates including region, is somewhat
reduced from what it was in the analysis presented in Block 2, that is, from β =
–.136 to β = –.119—a 12.5 percent reduction. The independent effect of religious
concentration on the divorced rate, however, remains statistically signiﬁcant.
Examination of the regression results in Block 3 indicates that when
divorce is regressed on these social, economic, and geographic factors, three
variables do not show a statistically signiﬁcant effect, that is, percent aged
between 15 and 34 years, percent Native American, and residence in the Northeast.
Percent urban shows the largest effect on the divorced rate ( β = .382), followed
by residence in the West (β = .327), and percent change in population ( β = .294).
Rates of currently divorced are signiﬁcantly higher in the West and the South,
but not in the Northeast. The low zero-order correlation between divorce and the
Northeast is indicative of the low divorce rate in that region.
The addition of region to the equation in Block 3 increases the explained
variance from 33.8 to 43.3 percent. This indicates a real increase in explained
variance of 9.5 percent...
View Full Document
- Fall '13