21
Appendix
Most of the control variables used in the analyses (see Table A1) are either self-explanatory or
can be found in the documentation at youthandreligion.org. I only note below variables that I
constructed especially for these analyses.
Parent closeness is the maximum value of the closeness variable reported by the respondent for
either parent. The resulting value was standardized (in the full sample) to have mean = 0 and SD
= 1.
Network closure was constructed from three variables that were asked of each respondent’s
social network. For each reported friend, the respondent was asked, which of these friends, 1)
“(do/does) your [PARENT TYPE] not really know that well”; 2) “have parents who know YOU
by name”; 3) “have parents who know your [PARENT TYPE] well enough to call
(him/her/them) on the phone.” These responses were combined (the first was reverse coded) to
give a sense of how much adult networks were closed around the respondent. The resulting sum
(0-15) was divided by 3 to make it comparable to the other network measures.
GPA was constructed from the variable “grades” (y91), which asked, “What kind of grades do
you usually get in school?” The original responses were 10 ordinal categories ranging from “all
As” to “mostly Fs” with an additional category for “mixed” (n=159). The GPA scale used in
these analyses was made into a scale with range 0 to 4 by rescaling the 10 point ordinal scale and
setting the “mixed” responses to the sample mean.
Parent education is the highest level of education for either parent, measured on a 12-point
ordinal scale.
Convention