Survey
design, methodology, and other housekeeping details
The Social
Capital Community Benchmark Survey was designed by the Saguaro
Seminar: Civic Engagement in America, a project at the John
F. Kennedy School of Government at Harvard University. The
principal investigator on this project was Prof. Robert D. Putnam,
and the survey drew upon the lessons learned from a Social Capital
Measurement Workshop held at Harvard University in October 1999.
In addition, there was a Scientific Advisory Committee convened
to advise on survey construction, consisting of some of the
leading scholars on measuring social capital and cross-racial
social trends. [7] All efforts were made, where
possible to use questions extensively tested in previous surveys.
The survey,
averaging 26 minutes, was conducted by telephone using random-digit-dialing
during July to November, 2000, although interviewing in the
national survey and in most of the community surveys was concluded
by October. TNS Intersearch, an international survey firm,
was commissioned to conduct the interviewing, and prepare the
data for analysis. Roughly 29,200 people were surveyed. The
national sample (N = 3,003) of the continental U.S. contains
an over-sampling of black and Hispanic respondents; 501 non-Hispanic
blacks were surveyed and 502 Hispanics participated.
In addition,
each sponsoring organization (largely community foundations)
decided on the size and sampling geography for each community
sample. Most of the samples range in size from 500-1,500 interviews.
(A complete list of communities surveyed, their sample size
and geographic definition are shown in Table 1 below.)
[8]
The
confidence interval for the various communities is given in
Table 2 below and ranges from +/- 2.1% for the national sample
to +/- 5.8% for some communities like Seattle.
Response rates averaged 28.9%
for the community samples and 28.7% for the national sample, using
the AAPOR RR2 formula. The adjusted cooperation rates, examining
the percent of those contacted who agreed to participate and completed
the interview, were 41.6% for the community samples and 42.3%
for the national
sample.
[9]
Up to 10
additional callbacks after the initial call were made to try
to get respondents to participate and calls were made at various
times of the day and various days of the week, spread out over
the call period. All but very hard refusals were attempted
to be converted into participants.
*The
survey data, survey instrument, and codebook are now available
from
the Roper
Center for Public Opinion Research at Storrs Connecticut.
