Michigan State of the State 45
METHODOLOGICAL REPORT
MICHIGAN STATE UNIVERSITY
STATE OF THE STATE SURVEY
[MSU SOSS-45]
Summer 2007 Round
Prepared by:
Larry A. Hembroff
Brian D. Silver
Institute for Public Policy and Social Research
Office for Survey Research
Michigan State University
October, 2007
NOTE TO THE READER
The State of the State Survey [SOSS] is administered by the Institute for Public Policy
and Social Research of Michigan State University.
For the benefit of sponsors, consumers and users of SOSS data, we have prepared this
guide to the purpose, design, methods, and content of the survey.
Please address questions or comments to:
Dr. Larry A. Hembroff, Senior Survey Methodologist, Office for Survey Research,
Institute for Public Policy and Social Research, Berkey Hall, Michigan State
University, East Lansing MI 48824
Phone: (517) 353-1763
Fax: (517) 432-1544
Internet: Hembroff@msu.edu
Dr. Brian D. Silver, SOSS Director, Department of Political Science, Michigan
State University, East Lansing MI 48824
Phone: (517) 355-2237
Fax: (517) 432-1091
Internet: BSilver@msu.edu
1. PURPOSE OF SURVEY
Dr. Jack H. Knott, former Director of the Institute for Public Policy and Social
Research [IPPSR], made the Michigan State University State of the State survey [MSU
SOSS] a reality by promoting the idea throughout the University and convincing the
key sponsors to contribute funds to get the survey off the ground. With funding
assured for the first year, planning began in June 1994. After completing 19 rounds
of SOSS, there was a brief period of inactivity between the Fall of 1999 and the Winter
of 2001 when, for budgetary reasons, no rounds of SOSS were conducted. However,
with the appointment of Dr. Carol Weissert as the Director of IPPSR in the Fall of
2000, there was a resurgence of both interest and funding for the resumption of SOSS
as a longitudinal survey of the state's adult population on policy- relevant issues.
SOSS is a quarterly survey of the citizens of Michigan. It employs Computer
Assisted Telephone Interviewing (CATI) technology to interview a stratified random
sample of Michigan citizens. Conducted by the Office for Survey Research, a division
of the Institute for Public Policy and Social Research, SOSS was inaugurated in
October 1994.
Although dozens of surveys are conducted in Michigan every year, none is
designed to provide a regular systematic monitoring of the public mood in major
regions of the state. SOSS is designed to fill this information gap. SOSS has five
principal objectives.
1. To Provide Information about Citizen Opinion on Critical Issues. In
keeping with MSU's role as the premier Land Grant University in the United States,
MSU seeks to inform the public about the state of the state. Although statistics from
censuses, public records, programs, and services provide important information about
the state of the state, there is no substitute for gathering information directly from the
citizens. By conducting a State of the State survey at regular intervals, IPPSR hopes
to monitor the public's mood about important aspects of Michigan's public life. This
information should be useful not only to citizens at large but also to policy-makers in
the public sector and to other groups and organizations that take an active interest in
the state of the state of Michigan.
By disseminating this information through the mass media and in special
studies, IPPSR hopes to provide baselines for assessing change in the people's sources
of satisfaction and dissatisfaction with the quality of life, the performance of public
institutions, the impact and efficacy of public policy, and the opinions about various
aspects of life in Michigan, such as confidence in the economy and the climate for
business, protection of the environment, freedom from crime, family life, and the
vitality of ethnic groups and communities.
2. To Provide Data for Scientific and Policy Research by MSU faculty. MSU's
faculty will use the data from the State of the State Survey to address a wide variety
of issues in public policy. What are the factors associated with the declining levels of
confidence in governmental institutions? To what extent does social and economic
status affect tolerance and mutual trust between ethnic and racial groups? Are
subjective perceptions of environmental quality related to "objective" measures of
environmental quality in Michigan's counties? These are only a few examples of the
types of questions that the principal researchers will address using the SOSS results.
To serve the interests of a wider scientific community, the SOSS data is deposited in
an international data archive.
3. To Provide Useful Information for Programs and Offices at MSU. IPPSR
has conducted a wide variety of studies for the use of MSU administrators and faculty.
SOSS will also develop data for such internal use as well as provide data for use by the
MSU Extension, the Vice Provost for University Outreach, and other offices.
Generally, the Winter rounds of the survey will assess the public image of higher
educational institutions, which will be useful to many offices at MSU.
4. To Develop Survey Methods. The computer-assisted telephone interviewing
(CATI) technology lends itself to experiments in question wording, question order, and
formatting of response categories. By varying the wording and sequences of questions
and responses, the investigators can study the sensitivity of answers to the format of
questions. Although survey research demands creative skills and remains to some
extent an "art," the scientific study of survey methods is a well established discipline.
Contributing to the scientific literature on survey methods is an important goal of the
OSR; hence, a variety of experiments are built into some of the survey instruments.
5. To Provide Opportunities for Student Training and Research. Data from
SOSS will be made directly available to professors and students for use in instruction
and research in classes at MSU. The availability of up-to-date information on public
opinion and individual perceptions and experiences of the Michigan population will
increase the sense of immediacy and relevancy of educational projects.
2. CALENDAR
People's experiences and the public mood change not only from year to year but
also with the seasons. It is important to establish baselines for understanding what
is a "normal" seasonal fluctuation and what is a more permanent change. For this
reason, SOSS is conducted at regular quarterly intervals. Roughly one-fourth of the
questions are repeated in each quarterly round.
SOSS has seasons itself, however, by focusing the main theme of each round
of the survey on topics that correspond with the annual cycle of major events in
Michigan and at Michigan State University. In general, the intended cycle is as
follows:
Fall. The Fall round in even-numbered years focuses on elections, political
participation, and political attitudes and orientations. In odd-numbered years, the
Fall round tends to focus on health and the environment.
Winter. The Winter round in each year focuses on the state of the state of
Michigan, in particular on the performance of governmental institutions at all levels,
on the subjective quality of life of Michigan's citizens (satisfaction with public
education, work, protection from crime, environmental preservation, and so forth), and
on the desire for reform in Michigan's political economy. This information should help
to inform the public discussion around the time of the Governor's annual budget
message. In addition, questions on the public's perceptions of Michigan's higher
educational institutions should help to inform public discussion around the time of the
annual "State of MSU" address by the President of the University.
Spring. The Spring round has as its main theme the state of Michigan families,
the role and status of women, and the status of children. Assessments of public
opinion concerning issues of women's rights, the status of children, and related issues
will help to inform policy debates.
Summer. The Summer round focuses primarily on the state of ethnic
Michigan, i.e., the vitality and diversity of Michigan's ethnic and racial communities.
SOSS assesses the strength of ethnic ties and identities, perceptions of various ethnic
groups (tolerance, stereotyping), and experiences of intolerance or discrimination. In
addition, the extent of attachment to and vitality of wider communities (towns and
cities) is an important mark of the quality of life in Michigan.
From time to time, SOSS may depart from this thematic plan when particular
sponsorship or pressing issues make it necessary or desirable. Beyond the core set of
interview items, SOSS-45 included questions on charitable giving and volunteer
activities. This included questions on respondents' confidence and trust in charitable
organizations.
3. STRUCTURE OF THE QUESTIONNAIRE
The questionnaires for each round of the survey are designed by a different set
of principal investigators, who are usually faculty and students at MSU, but other staff
or clients also. Each survey instrument consists of three main parts: a demographic
core, a non-demographic core, and the main substantive theme or themes.
The demographic core contains questions on the social background and status
of the respondents (age, sex, education, employment status, type of community, marital
status, number of children, size of household, income, ethnic identity, etc.). This bloc
of questions is repeated in each round, though more detailed questions on some of the
dimensions (e.g., the number and ages of children) might be included in certain
rounds.
The non-demographic core contains additional questions that are repeated in
every round of the survey in order to gauge broad shifts in the economic, social, and
political orientations and status of the population. These include questions about
consumer confidence, self-identification on a liberal-conservative scale, partisan
identification, assessments of presidential performance and gubernatorial
performance, and other issues.
Together the demographic and non-demographic core of the questionnaire take
an average of about 5 minutes of interviewing time to complete.
The remainder of the interview is timed to last an average of 15 minutes, so
that on average the interviews take about 20 minutes of the respondent's time.
The questionnaire consists almost entirely of closed-ended questions. Verbatim
responses are used and open-ended coding are required for these questions.
A word of caution is in order on the use of the data. Because of the inclusion
of question-order and question-wording experiments, the codebook for the survey,
containing the raw or weighted frequency distribution of responses, may be difficult
to interpret and must be used carefully. Often, alternative variants of questions will
be combined into composite measures in the final data that are distributed, but the
original questions also remain in the codebook and data set. Although OSR will do its
best to document such situations, it is the responsibility of the data users and analysts,
not of the OSR, to assure that the appropriate variants of questions are used in
analyses and reports. A copy of the CATI interview program with the skip patterns
indicated by "[goto ...]" commands and "[if ...]" commands accompanies the codebook
to help clarify the paths particular respondents would take through the interview.
4. MANAGEMENT AND ORGANIZATION
IPPSR. The overall SOSS program is directed by Dr. Brian Silver
(Department of Political Science). Overall responsibility for the execution and
management of the SOSS rests with the Office for Survey Research (OSR) of the
Institute for Public Policy and Social Research. The Principal OSR staff for SOSS
consists of Dr. Larry Hembroff, Survey Director and Methodologist, Karen Clark,
Programmer and Project Manager, and the Director of Survey Operations (this position
was vacant during the execution of SOSS-45).
OSR staff is responsible for the technical work of programming the CATI
survey instrument, training and supervising interviewers, selection and
administration of the sample, coding of data, and preparation of the final data set and
documentation. In addition, OSR staff works with and advises the principal
investigators and other researchers in the design of the sample and the survey
instrument. However, final approval of the survey and sample design rests with the
principal investigators, not OSR staff.
For each round of the survey, a small working group of principal investigators
is responsible for the design of the instrument for that round, subject to final approval
by the SOSS Director and OSR staff. The working groups consist primarily of
"principal investigators" for the given round who will conduct the major initial
analyses of the data, provide a public briefing, and have priority in analyzing the data
for publication for the six-month period following the end of the field period for that
round (more on data access below).
The Working Group for the Summer 2007 survey was comprised of:
Mark Wilson, Ph.D., Associate Professor, Department of Geography
Brian Silver, Professor, Dept. of Political Science, MSU, and SOSS Director
5. FUNDING
The following organizations and units on campus have provided funding for
SOSS during the 1995-2007 series of surveys:
Organizations
Area Agencies on Aging Association of Michigan
Aspen Institute
Community Foundation for Southeastern Michigan
C. S. Mott Group for Sustainable Food Systems
Dept. of Political Science, Florida State University
Dept. of Political Science, Tufts University
Nonprofit Michigan Project
University of Michigan
United Way of Michigan
Michigan State University
Applied Policy Grants Initiative
Center for Health Care Studies
Center for Health Promotion and Disease Prevention
College of Communication Arts & Sciences
College of Human Ecology
College of Human Medicine
College of Osteopathic Medicine
College of Social Science
Department of Economics
Department of Political Science
Department of Psychology
Department of Radiology
Department of Sociology
Education Policy Institute
Institute for Public Policy and Social Research
Julian Samora Research Institute
Land Use Policy Institute
Legislative Leadership Program
Managed Care Institute
Michigan Agricultural Experiment Station
MSU Extension
MSU Institute for Children Youth and Families
Office of the Provost
Office of the Vice President for Research and Graduate Studies
Office of the Vice Provost for University Outreach
School of Criminal Justice
School of Labor and Industrial Relations
School of Social Work
6. DISSEMINATION OF RESULTS
To assure timely dissemination of the results and timely and fair access to the
data, early in its deliberations the Advisory Committee approved certain principles.
Each round of the survey has an identified set of Principal Investigators (PI's)
who have priority in access to the data for that round but also certain obligations. The
PI's are responsible for preparing and conducting a press briefing based on results of
the survey within one week of the end of the field date. IPPSR's outreach and design
staff assist in this effort, working with the MSU University Relations.
The PI's have exclusive right to prepare scientific papers for publication from
the data for that survey for a period of six months after the end of the field date.
All data for the survey, however, are made available to offices within MSU for
internal use as soon as the data are available and documentation is prepared.
All data for the survey are made available to instructors in courses at MSU to
use the data for instructional purposes as soon as the data are available and
documentation prepared.
Six months after completion of the field date, the survey data are made
available on an unrestricted basis to all MSU faculty and students.
Originally, it was planned that one year after completion of the field date, the
data and documentation will be deposited at the Inter-University Consortium for
Political and Social Research (ICPSR) in Ann Arbor. However, beginning in the Spring
of 2002, each individual SOSS data set, interview instrument, and methodological
report have been posted in "universally" readable formats to the SOSS section of
IPPSR's webpage for downloading by any interested party. Such a deposition of the
data is intended to facilitate dissemination and use of the data by the wider scientific
and policy community as well put a certain seal of approval on the data quality to
enhance the possibilities for researchers to publish from the data.
7. SAMPLE DESIGN
The referent population is the non-institutionalized, English-speaking adult
population of Michigan age 18 and over. Since the survey was conducted by telephone,
only persons who lived in households that had telephones roughly 97% of the
Michigan population had a chance of being interviewed.
Stratification. To assure representation of major regions within Michigan, the
sample was stratified into six regions, each consisting of a set of contiguous counties,
plus the City of Detroit. The grouping of counties corresponds to that used by MSU
Extension prior to July 2005 with Detroit separated out from the Southeast region.
The six regions are defined as follows (counties listed within regions -- also see
the map in the Appendix):
1. Upper Peninsula (Alger, Baraga, Chippewa, Delta, Dickinson, Gogebic,
Houghton, Iron, Keweenaw, Luce, Ontonagon, Mackinac, Marquette,
Menominee, Schoolcraft)
2. Northern Lower Peninsula (Alcona, Alpena, Antrim, Benzie, Charlevoix,
Cheboygan, Crawford, Emmet, Grand Traverse, Iosco, Kalkaska, Leelanau,
Missaukee, Montmorency, Ogemaw, Oscoda, Otsego, Presque Isle,
Roscommon, Wexford)
3. West Central (Allegan, Barry, Ionia, Kent, Lake, Manistee, Mason,
Mecosta, Montcalm, Muskegon, Newaygo, Oceana, Osceola, Ottawa)
4. East Central (Arenac, Bay, Clare, Clinton, Gladwin, Gratiot, Huron,
Isabella, Midland, Saginaw, Sanilac, Shiawassee, Tuscola)
5. Southwest (Berrien, Branch, Calhoun, Cass, Eaton, Hillsdale, Ingham,
Jackson, Kalamazoo, St. Joseph, Van Buren)
6. Southeast (Genesee, Lapeer, Lenawee, Livingston, Macomb, Monroe,
Oakland, St. Clair, Washtenaw, Wayne [excluding Detroit])
7. Detroit City
To allow reclassification of the place of residence (county) into alternative regional
groupings, each respondent's county of residence is also coded on the data set.
In July 2005, the MSU Extension reconfigured its regions from six to five. The only
region that did not change in terms of the counties comprising it was the Upper
Peninsula. The new regional configuration is as follows:
Region 1 Upper Peninsula: Menominee, Delta, Chippewa, Luce, Mackinac,
Schoolcraft, Alger, Marquette, Dickinson, Iron, Gogebic, Baraga,
Ontonagon, Keweenaw, Houghton.
Region 2 North: Emmet, Cheboygan, Presque Isle, Alpena, Montmorency,
Otsego, Charlevoix, Leelanau, Benzie, Grand Traverse, Kalkaska,
Crawford, Oscoda, Alcona, Iosco, Antrim, Manistee, Missaukee.
Region 3 Central: Kent, Ottawa, Gratiot, Montcalm, Newaygo, Midland,
Isabella, Mecosta, Oceana, Bay, Arenac, Gladwin, Clare, Osceola, Lake,
Mason, Ogemaw, Roscommon, Wexford.
Region 4 Southwest: Lenawee, Hillsdale, Branch, St Joseph, Cass, Berrien,
Jackson, Calhoun, Kalamazoo, Van Buren, Ingham, Eaton, Barry, Allegan,
Shiawassee, Clinton, Ionia, Muskegon,
Region 5 Southeast: Monroe, Wayne, Washtenaw, Livingston, Oakland,
Macomb, St Clair, Lapeer, Genesee, Sanilac, Saginaw, Tuscola, Huron,
Particularly for purposes of maintaining the longitudinal value of the State of the
State Survey data sets, OSR elected to continue using the original regional
configuration as the basis for the stratified sampling design of each survey. OSR will
continue to calculate caseweights that will allow generalizations to these regions that
take full advantage of the disproportionate sampling design. However, to maintain the
utility of the SOSS data sets for MSU Extension purposes, as of SOSS 38, we have
constructed a variable (MSUE2005r5) aggregating counties into the new MSUE
regional groupings and have constructed a separate set of caseweights appropriate for
these regions.
Sampling. Until SOSS-35, all previous respondents were derived only from
random-digit dial samples. Beginning with SOSS-35, a change was made in the
sampling strategy for the State of the State Surveys. The overall intent of the change
was to reduce costs, increase response rates, and shorten the field period needed to
complete each survey. The revised strategy is similar to that used on the University
of Michigan's Survey of Consumer Attitudes. A portion of the sample of interviews is
derived from a new random-digit dial sample of phone numbers in the state. The
details of this are described below. The other portion of the sample of completed
interviews (roughly 40%) is derived from re-interviews of individuals who had been
interviewed in the previous round of SOSS and who had agreed to be re-contacted.
Roughly 90% of all respondents in each round of SOSS agree to be re-contacted. Re-
interviewing individuals who constituted a representative random sample of the state's
adults should still constitute a representative random sample several months later if
adjustments for any non-response are made. Limiting the portion of SOSS-45's sample
of completed interviews derived from re-interviews with SOSS-44 participants to less
than half of the total number of SOSS-45 interviews ensures that there should be
sufficient numbers of respondents who will be willing to be re-contacted and will be
reachable for the next round of SOSS. In addition to the three benefits listed above as
reasons for making the change in sampling strategy, having a portion of each round
of SOSS derived from re-interviews with individuals from a previous round enables a
part of the SOSS sample to constitute a panel so that change can be measured at the
individual level from quarter to quarter a distinct benefit.
Respondents' households newly enlisted to participate for SOSS-45 were
selected using list-assisted random-digit dial sampling procedures. Those being re-
interviewed had been sampled and selected in this same manner when they were first
recruited to participate in the previous round of SOSS. Ordinarily, the initial sample
of randomly generated telephone numbers is purchased from Survey Sampling, Inc
(SSI). SSI begins the process of generating phone numbers with the list of all working
area code and phone number exchange combinations. In the case of this study, the
universe was constrained to include only those telephone numbers that are active in
the state of Michigan. From within this list of possible phone numbers, SSI eliminates
those banks of numbers represented by the 4-digit suffix that are known to be unused
or are known to be used only by institutions. To improve the efficiency of the calling,
we have begun to have SSI stratify this sampling frame into two strata initially, one
comprised of all phone numbers that are listed in phone directories, and the other
comprised of all phone numbers that are not listed in directories but which are
members of banks in which at least one phone number is listed. We then request that
SSI over-sample phone numbers from the listed stratum. Telephone numbers are
selected at random in proportion to the number of households in each county from all
those remaining telephone numbers until the quantity needed within a particular
geographic grouping of counties is obtained.
As a final step, SSI screens the phone numbers generated. The resulting
sample is then checked against SSI's database of business phone numbers and checked
for known disconnected numbers. Ordinarily, these numbers are removed from the
sample and not called.
To determine the total number of telephone numbers to have SSI generate in
order to achieve the desired sample sizes within regions of the state, OSR divided the
number of completed interviews desired by the product of (a) the proportion of numbers
expected to be working household numbers (the Hit Rate), (b) the proportion of
household numbers that would contain an eligible respondent (the Eligibility Rate),
and (c) the proportion of households with eligible respondents who would complete the
interview in the time period available (the Completion Rate). For SOSS-45, a total of
5,610 phone numbers were used, 542 in the re-contact segment and 5,068 in the new
RDD segment. The working phone number rate was 92.8% in the re-contact segment
and 70.5% in the new RDD segment.
The sampling design for the State of the State Survey is a stratified sample
based on regions of the state with the regions sampled somewhat disproportionate to
the actual sizes of the populations within each region. The purpose of the stratification
is to assure a sufficient minimum number of respondents from each of the strata to
permit detailed analysis.
The typical sampling design for SOSS calls for approximately 150 interviews
from the East Central Region, the Southwest Region, and the combined Upper
Peninsula and Northern Lower Peninsula Regions. Approximately two hundred
interviews are to be completed in the West Central Region and the Southeast Region.
And approximately 150 interviews are to be completed from the City of Detroit. The
total sample size is to be approximately 1,000.
Sample Weights. Because of the split sample approach, we have weighted
each segment regarding selection probabilities and then combined them into a single
file. The combined data file is then weighted to be representative of the geographic
regions and the state as a whole. The details for weighting each segment are provided
below.
Because of the stratification (i.e., geographic strata, listed vs. not-listed phone
number strata) and the unequal sampling rates across the strata, it is necessary to use
"weights" to bring the characteristics of the sample into line with those of each region,
or with those of the state as a whole (depending on the purpose of the analysis).
Accordingly, the data files contain weights for the original six MSU Extension regions,
for the new Extension regions, as well as for the state as a whole.
As indicated above, the initial frame was stratified into listed numbers and not-
listed numbers in 1+ banks and then listed numbers were over-sampled. Other
information from SSI indicates that 65% of households with phones have listed
numbers. An initial weight, listwt, was constructed to adjust representation of listed
and unlisted numbers in the data file so that listed numbers comprised only 65% of all
data records.
To construct the remaining weights, characteristics of the population of the
regions were drawn from 2000 census data. To make generalizations about
individuals' views and behaviors, it is necessary to ensure that each respondent in a
survey sample has an equal probability of selection or is represented in the data set
as having had equal probabilities of being selected. However, since households with
multiple phone lines have more chances of being selected into the sample than those
with only one phone line, this source of unequal chances has to be adjusted for in
analyzing the data. Consequently, the SOSS interview included a question asking
respondents how many separate phone numbers the household has. In the event of
item non-response, the number of phone lines was assumed to be one. Each case was
then weighted by the reciprocal of the number of phone numbers and then adjusted so
that the total number of cases matched the actual number of completed interviews. In
the data set this weight is named PHWT.
Similarly, an adult in a two-adult household would have half the chance of
being selected to be interviewed as would the only adult in a single adult household.
This, too, requires adjustment to correct for unequal probabilities of selection. The
interview included a question as to the number of persons 18 years of age or older
living in the household. In the event of item non-response, the household was assumed
to have only one adult. Each case was then weighted by the inverse of its probability
of selection within the household, or by the number of adults in the household. This
was then also adjusted so that the total number of weighted cases matched the actual
number of completed interviews. In the data set, this weight is named ADLTWT.
At this point, the adjustment was intended primarily to facilitate accurate
weighting to adjust for non-response based on age, gender, and race within SOSS
regions. It is common for some groups of individuals to be more difficult to reach or
more likely to refuse in RDD (random-digit dialing) surveys. For making
generalizations about the population from which the sample was drawn, the accuracy
of the results can be distorted by these non-response patterns. Consequently, it is
common to weight cases in the sample to adjust for non-response. This is accomplished
by weighting each case so that cases of each type appear in the sample proportionately
to their representation in the general population.
For the State of the State Survey, cases are weighted so that the proportions
of white males, African American males, other racial group males, white females,
African American females, and other racial group females in the sample for each
region matched the proportions each of these groups represent in the adult population
of each of the original MSU Extension regions and the City of Detroit based on the
2000 Census. In the data set, this weighting factor is named RACGENCT.
Furthermore, within each of the original MSU Extension regions and the city of
Detroit, the cases were additionally weighted so that the proportion of cases falling
into each of the following age groups matched the proportions in the 1990 Census for
each region: 18 - 24 years old, 25 - 29, 30 - 39, 40 - 49, 50 - 59, 60 - 64, and 65 or older.
In the data set, this weighting factor is named AGEWT (since rounding and missing
data sometimes result in the weighted number of cases differing slightly from the
actual number, AGEWT is adjusted slightly with ADJWT to ensure that the number
of cases for each region in the weighted data set is the same as the actual number of
interviews completed). Detroit continues to be a separate stratum to this point, but a
new variable MSUEREGN was constructed to fold Detroit proportionately into the
Southeast region within that variable. A new weighting variable (MSUEWT) was
constructed to represent Detroit proportionately correctly within the southeast
MSUEREGN.
Since the sample was drawn disproportionately across the original six MSUE
regions of the state (with Detroit in the Southeast region), statewide estimates of the
citizenry's opinions require post-stratification weights to adjust for the over-sampling
of some regions and the under-sampling of others. Thus each case was weighted so
that the proportion of cases from each region in the total sample matched the
proportion of adults from the corresponding region in the state's population based on
2000 Census data. The weighting factor for this post-stratification weighting in the
data set is named STATEWT.
It is important to note that these weight factors were constructed sequentially
and build on the earlier steps. Thus, AGEWT weights cases adjusting for the number
of phone lines, the number of adults in the household, the number of respondents from
each county, the gender x race category proportions within the region, and the age
category proportions within regions. STATEWT weights cases by all of those
adjustments implied by AGEWT and adjusts the proportions of cases across regions.
For developing statewide results, the user should use the data weighted by
STATEWT and select only those cases for which the value of the variable SAMPLE is
less than 3. For comparing the results among regions -- if Detroit is to be separate --
the user should use the data weighted by ADJWT, again selecting only cases for which
SAMPLE is less than 3. To compare directly the original MSUE regions, the data
should be weighted by MSUEWT and cases selected for which SAMPLE is less than
3.
As we noted above, beginning with SOSS-38, we have constructed an
alternative set of weights based on the new MSU Extension regions. To identify
regions, we grouped cases based on the respondent's county of residence into one of six
regional groupings (including Detroit as a separate region) in a variable named
MSUE2005. The race x sex x age profile of the sample (weighted by adltwt) was then
compared to the corresponding profile in the 2000 U.S. Census for each region and the
city of Detroit. For this comparison, respondents' ages were collapsed into one of four
categories: 18-29, 30-44, 45-64, and 65 or older. This variable is labeled AGECAT4.
A weight value (NEWADJWT) was calculated for each case that is intended to adjust
the cases within each region to match the race x sex x age profile while keeping Detroit
separate from the new Southeast Extension region. Another region variable
(MSUE2005r5) was constructed representing only the five new Extension regions with
Detroit included in the Southeast region and then an additional weighting adjustment
was made for cases in the Southeast region so that Detroit cases were proportionately
represented within the region and the total number of weighted cases in each region
equaled the actual number of interviews. This weight variable, MSUE2005WT,
should be used when the new Extension regions are to be compared to each other.
NEWADJWT should be used if the new Extension regions are to be compared to each
other with Detroit separated out for comparison to other regions of the state.
Table A in the Appendix presents the characteristics of the unweighted
respondents on several characteristics, in comparison with the population in each
region and in the state of Michigan as a whole.
Sampling Error. The sampling error can be estimated for each region and for
the state as a whole at the 95% confidence level as follows:
Confidence Interval = +/- 1.96SQRT(PxQ/ (n - 1))
where n is the number of cases within the region or the total sample and P is
the proportion of cases giving a particular response and Q is 1-P. While this may vary
from question to question depending on the pattern of answers, the largest margin of
error would occur when P is .5 and Q is .5. Therefore, the margins of error for each
region and the total statewide sample excluding the supplemental Hispanic/Latino
segment of the sample can be estimated as:
REGION Number of Cases Margin of Sampling Error
Upper Peninsula 46 +/- 14.6%
Northern Lower Peninsula 68 +/- 12.0%
West Central 175 +/- 7.4%
East Central 134 +/- 8.5%
Southwest 116 +/- 9.1%
Southeast 193 +/- 7.1%
Detroit 157 +/- 7.8%
Statewide Total 889 +/- 3.3%
8. FIELD PROCEDURES
CATI System. Interviews were conducted using the Computer Assisted
Telephone Interviewing system (CATI) of IPPSR's Office for Survey Research ( OSR).
OSR uses the CASES (version 4.3.7) software for its CATI system. CASES was
developed by the University of California Berkeley, the U.S. Census Bureau, and the
U.S. Department of Agriculture. In a CATI system, the completed interview is scripted
and then programmed so that, when executed from a computer workstation, each
question or instruction is presented on the computer screen in order to the interviewer.
The program then indicates what numeric codes or text the interviewer is allowed to
enter as responses to each of the questions. When entered, the responses are stored
directly into the data set for the study.
The CASES software enables the interview to be fully programmable. The
software integrates both closed-ended questions and open-ended questions. The
software allows interviewers to record notes along with responses to closed questions.
By default, the software moves directly from one item to the next in the sequence
unless specific program commands are inserted to direct the execution path elsewhere.
Different skip commands can be associated with separate responses to the same
questions. For example, the interview can be directed to a separate battery of follow-
up questions if the respondent answers "<1> YES" to a question on smoking cigarettes,
and to an entirely different series of questions if the respondent answers "<5> NO."
Commands can also be inserted between questions to direct the interview to a
particular battery of questions based on the combination of responses to two or more
previously answered questions. The programming features minimize the opportunities
for many errors since inappropriate questions will not be asked and, as a result,
appreciably less editing is necessary after the interview.
Interviewers and Interviewer Training. New interviewers received
approximately 15 hours of training, including a shift of practice interviewing. Each
interviewer trainee received a training manual with instructions on techniques and
procedures, copies of all relevant forms, and descriptions of operations. The OSR
telephone interviewing training package was developed using "General Interviewing
Techniques: A Self-Instructional Workbook for Telephone and Personal Interviewer
Training", authored by P. J. Guenzel, T. R. Berckmans, and C. F. Cannell (1983) of the
Survey Research Center, Institute for Social Research, University of Michigan.
Experienced interviewers received approximately two hours of study specific
training to acquaint them with the study protocols, the interview instrument, and the
objectives of the various questions. New interviewers were also given this information
as a part of their training. Approximately 55 different interviewers were involved in
data collection on the 45nd State of the State Survey.
Field Period and Respondent Selection in Household. Interviewing began on
July 10, 2007, and continued through September 22, 2007.
In the portion of the sample that involved re-interviewing respondents from the
previous SOSS, interviewers asked to speak with that person when they contacted the
household. When interviewers successfully contacted a household in the new RDD
portion of the sample, the study procedures required them to randomly select an adult
from among those residing in the household to be the respondent. The Trohldal-Carter
technique was used as the mechanism for choosing a respondent within each
household.
Telephone numbers were called across times of the day and days of the week.
If after a minimum of nine call attempts, no contact had been made with someone at
the number, the call schedule for that case was reviewed by a supervisor to see that it
had been tried across a variety of time periods. If it had not, the supervisor would re-
release the number for additional calling in time periods that had not been tried. If,
after additional calls were made, still no contact was made, the number was retired as
a non-working number. If the review of the case indicated that it had been tried at
various times and days, the supervisor might finalize the case as non-working or might
release it for up to six additional tries. In the case contact was established, the number
would continue to be tried until a total of 12 attempts were made or the interview was
completed, the interview was refused, or the case was determined to be ineligible or
incapable.
The average interview lasted approximately 12.0 minutes (standard deviation=
3.2) with a median of 12.0 minutes. In the case of an initial refusal, numbers were
called back after five days (although this was shortened as the end of the field period
neared). Efforts were made to persuade initially reluctant respondents to complete the
interview.
Completion Rate. A total of 880 interviews was completed, 296 with
participants re-contacted from the SOSS-44 survey and 593 with new RDD
participants. The overall completion rate among eligible households for the study was
36.6% (29.5% in the new RDD segment and 70.6% in the re-contact segment).
Of those completing the interview, the mean number of calls required was 4.1
(3.7 among the re-contact cases and 4.17 among the new RDD cases. Interviewers
made a total of 33,043 calls to complete the 889 interviews.
The refusal rate was 18.4%.
9. DOCUMENTATION AVAILABLE
The following documentation is available for this survey:
a. Methodological Report
b. Questionnaire (included in Methodological Report)
c. Codebook (included in Methodological Report)
d. Coding instructions (included in Methodological Report)
e. SPSS (windows) commands to read the ASCII data set
f. SPSS commands for weighting cases in the sample
Michigan State of the State 45