The 1991 National Organizations Survey
August 1993
Prepared by:
Professor David Knoke
Department of Sociology
909 Social Sciences Bldg.
University of Minnesota
267 19th Avenue South
Minneapolis, MN 55455
(612) 624-4300
Internet: knoke001@maroontc.umn.edu
1. PROJECT OBJECTIVES
The 1991 National Organizations Survey (NOS) is the first representative sample of U.S.
work establishments, physical sites where paid employment occurs. Principal investigators were
Arne L. Kalleberg (University of North Carolina), David Knoke (University of Minnesota),
Peter V. Marsden (Harvard University) and Joe L. Spaeth (University of Illinois). It was
sponsored by a grant from the National Science Foundation (SES-9022192, "Collaborative
Research: U.S. Organizations' Human Resources Policies"), with a subcontract from the U.S.
Department of Labor for the analysis of job training. The main data collection was carried out
by the Survey Research Laboratory at the University of Illinois at Champaign during the director
of Joe L. Spaeth and Diane O'Rourke (SRL Project 666). Additional aggregate-level data were
compiled from secondary sources at the University of North Carolina by Mark Van Buren and
Arne L. Kalleberg. Preparation of the 1991 NOS for distribution was done at the University
of Minnesota by Alisa Potter and David Knoke. The approximate direct costs for data collection
were $100,000. These data were released to the public on September 1, 1993, to be distributed
by the Inter-university Consortium for Political and Social Research.
The NOS consists of data on 727 employers of the respondents and their spouses in the 1991
General Social Survey, an annual cross-sectional sample of the adult population. The NOS
concentrated on the establishments' human resources policies and practices. Items asked about
current staffing procedures, internal job ladders and promotion chains, job training programs,
and employee benefits and incentives. Additional items gathered basic information about each
organization's formal structures, social demography, environmental situation, and productivity
and performance. Four other data sets were merged with the organization survey: (1) all
variables from the corresponding GSS respondents, including more than 50 job-related questions;
(2) industry-level variables at the two-digit SIC code level; (3) county-level and (4) zipcode-level
information about the social and economic setting within which each establishment is located.
These data offer opportunities for researchers to test explanations of organizational structures
and impacts on employees, with a representative sample having high generalizability.
2. SAMPLE DESIGN
The sampling design is basically a simple one that is capable of providing an current sample
of all types, sizes, and ages of work establishments. The basic problem in generating an
establishment or organization sample is the absence of a complete sampling frame. Most list
of organizations are incomplete. They may be limited to organizations of one or a few types,
and they may omit or underrepresent certain organizations, especially small or new ones. With
no sampling frame available and apparently no method for generating a frame, organization
researchers have often used nonprobability sampling methods.
The basic procedure for producing a probability sample of all types, sizes, and ages of
establishments (and of the organizations that contain them) is simple but moderately expensive.
Respondents to a standard survey of a human population are asked to identify the establishment
where they work by name, address, and telephone number. The establishments nominated by
this method are drawn with probability proportionate to size (PPS). All units containing the
survey respondents are drawn PPS, from the smallest work unit to the largest, most
encompassing organization. Such a survey-based sample is as current as the date of each survey
interview, in contrast to the time lag that may be required to enter published or even machine-
readable directories.
This sampling method works because each potential respondent represents the establishment
where he or she is employed. The more employees an establishment has, the more likely it is
to fall into the sample. If the method for sampling persons is equiprobable, the probability that
an establishment will fall into the sample is proportionate to the number of its employees. In
organizational research this method is known as hypernetwork sampling (McPherson 1982). In
statistics, it is a special case of multiplicity sampling (Sudman, Sirken, and Cowan 1988). In
this method, survey respondents are asked to nominate people known to them who have a rare
characteristic, such as a specific disease or being a Vietnam veteran. The nominators must be
in a specific, countable relationship to the nominees. The probability that a nominee will fall
into a sample is proportionate to the number of nominators in the population (known as the
multiplicity). In a survey-based sampling design for establishments, an establishment's
multiplicity is its total number of employees. The design aspect of sampling establishments is
thus rather simple; but realizing the design in practice is another matter.
The 1991 General Social Survey (GSS), conducted by the National Opinion Research Center
(NORC) at the University of Chicago, included a 15-minute module consisting of questions
about respondents' experiences with the organizations for which they worked. In addition, they
were asked to give the names, addresses, and telephone numbers of the establishments where
they worked and where their spouses worked. This information, along with answers to questions
on respondents' and spouses' occupations and industries, were transmitted to the Survey
Research Laboratory (SRL) at the University of Illinois. Under the direction of Joe Spaeth and
Diane O'Rourke, SRL interviewers attempted to conduct telephone interviews with informants
at each of the establishments nominated by the GSS respondents. Thus, there were two stages
in collecting data on establishments: collecting nominations of establishments, done face-to-face
by NORC interviewers, and telephone interviews of establishment informants by SRL
interviewers.
As noted above, respondents were asked three questions regarding their establishments and
those of their spouses. Owing to the possible complexity of persons' relationships with the
establishment for which they worked, three sets of questions were used: a straightforward set
asking for the name, address, and telephone number of the "one place" where they worked, a
set asking for the place where respondents or spouses would "find out where you will be
working," and a set asking if they "have any kind of base operations at all." Photocopies of the
GSS questionnaire pages containing this information, plus data on respondents' and spouses'
occupations and industries, were transmitted from NORC to SRL.
Of the 1,517 GSS respondents, 912 were working, as were 519 of their spouses, for a total
of 1,431. A total of 1,427 nominations were transmitted to SRL by NORC, of whom 909 were
respondents and 518 were spouses. The discrepancy of four cases probably arose because
NORC necessarily transmitted "dirty" data to SRL but released clean data to the social science
community. Of the respondents (who were always the nominators), 86 percent were able to
supply the establishment name, 3 percent did not know the name or did not answer, and 11
percent refused to provide this information. Although the differences between respondents and
spouses were significant at the .05 level, they were not large: respondents were 5 percent more
likely to provide an establishment name for themselves than for their spouses (it should be noted
that several establishments had no names).
Table 1 summarizes the disposition of the 1,427 cases for which GSS nominations were
attempted, that is the ultimate outcome of GSS and SRL attempts, not the nominations
transmitted by the GSS. Of the 1,427 respondents and spouses, 1,056 (74 percent) were initially
adequate; the nominating information was sufficient to begin interviewing immediately. The
other 371 cases (26 percent) provided initially inadequate information. Of these, nearly half
(46.4 percent) were refusals, virtually all of which remained unusable. Of the 53.6 percent that
lacked information on name or address, about two-thirds were usable and one-third remained
unusable. Therefore, ultimately 79 percent of the nominations led to interview attempts, 4
percent were ineligible and 17 percent were unusable. Given the available information and
procedures, incomplete information was not a major problem, but the refusals seemed to be quite
intractable.
3. DATA COLLECTION
Based on the results of an earlier study (Spaeth 1985), it was clear that the SRL data
collection effort would be intensive and time consuming. The study design allowed for many
more contacts than would be attempted in a standard population sample. Especially well-
qualified interviewers were recruited and paid a higher than normal level. The interviewers
received three days' training on the design of the study, its content, and how to persuade
reluctant respondents to cooperate. Data collection began on April 18, 1991 and continued
through November 29. It took a median of two contact attempts to reach an eventual
respondent, with a range of 1 to 29. A median of five contact attempts were required to
complete an interview (range 1-33). In all, the median interview involved contacts with six
persons, the maximum being 58 (i.e., speaking to a few people several times, not 58 different
people). Most interviews were completed in one session, but 18 percent required two sessions
and 8 percent three or more. Since extensive factual data were required, 17 percent of
interviews were done with more than one respondent per establishment. The length of the
interviews range from 10 to more than 100 minutes, with a median of 42 minutes. Clearly, our
expectation that data collection would require considerable time and effort was amply borne out.
Table 2 shows the outcomes of the 1,127 possible interview attempts. There was one major
and one minor contingency in this process. The major contingency occurred when a potential
informant, designated as "the had of the personnel department or the person responsible for
hiring," asked to receive a questionnaire by mail, even though the study design called for a
telephone interview. This request was usually because an informant refused to be interviewed
by telephone, although some simply wanted to be able to examine the questionnaire. Self-
administered questionnaires, changed as little as possible from the interview version, were
mailed to these people. Mailed questionnaires were followed up by phone and by subsequent
mailings.
The minor contingency was the existence of more than one nominator for a given
establishment, labelled "Duplicate Employers" in Table 2. Some of these duplicates resulted
from the GSS multi-stage area probability design, for which the smallest area was a single block.
In one instance, eight GSS respondents worked for the same large employer. Some duplicates
were spouses who worked for (or were) the same employer. In order not to impose unnecessary
respondent burden, informants about establishments were interviewed only once. (One exception
was for several series of questions about occupations held by the GSS respondent or spouse;
informants were asked these series for each separate occupation.) Data from this interview were
simply transferred to the records pertaining to the other nominators.
As shown in Table 2, the SRL attempted to collect data on 1,067 separate establishments.
Of these establishments, 59.4 percent were finalized by telephone (52.6 percent were completed,
1.7 percent pending at the close of the field period, and 5.1 refused to be interviewed).
Questionnaires were mailed to 40.7 percent of the establishments (29.3 percent returned
completed, 44.2 percent refused, and 26.5 percent were pending). The combined telephone and
mail questionnaire completion rate was 64.5% of the establishments, respectable by the standards
of organizational research but still low for most surveys. The cases for duplicate employers
produced an additional 39 completions, 14 refusals, and 7 pending. Of the data collection
attempts, 83 percent of refusals were in response to both mailed questionnaires and pendings.
People in both groups most often said that they did not have time. The 727 cases that provided
complete data on establishments are 50.9 percent of the original 1,427 attempted nominations.
Although even this rate is respectable in organizational research, it leaves much to be desired.
Future NOS projects must make changes in design and data collection to boost the response rate,
especially for the mailed questionnaires.
4. PUBLICATIONS BASED ON DATA
As of the date of this writing, no publications have been reported for the 1991 National
Organizations Survey. However, a special issue of the American Behavioral Scientist is being
prepared for publication in 1994. This issue will consist of methodological and substantive
articles written by the principal investigators. Of especial interest to users of the 1991 NOS will
be an article describing in detail the sampling design and data collection that produced the
survey.
5. TECHNICAL INFORMATION ABOUT THE DATA FILE
The data file was subjected to the following cleaning procedure: All the completed
interviews and mail questionnaires were keyed and verified. A system file was created at the
SRL using the Statistical Package for the Social Sciences (SPSS-X) program for mainframe
computers. Raw frequencies were generated for all variables on the file. The output was
checked for appropriately declared missing values and whether out of range values existed for
any of the variables. All out-of-range values were checked against the questionnaires and
necessary changes were made directly to the raw data. An SPSS export file was created and
transferred electronically to the University of Minnesota. Three aggregate-level SPSS export
files were produced at the University of North Carolina (by Mark Van Buren under the
supervision of Arne Kalleberg): (1) industry-level variables at the two-digit Standard Industrial
Classification for each establishment; (2) county-level and (3) zipcode-level social and economic
indicators for the geographic units in which each establishment was located. These files were
also sent to the University of Minnesota, where they were merged with the corresponding 1991
General Social Survey respondent cases, using SPSS-X on a VAX mainframe computer (by Alisa
Potter under David Knoke's supervision). Where a NOS establishment was the employer of a
GSS respondents' spouse, the individual-level data about the spouse were assigned to the relevant
GSS variables (for example, the number of years of education, EDUC, was given the value
reported in SPEDUC). Because GSS respondents provided very little information about their
spouses, substantial missing data occurred for those cases.
The final 727-case NOS data set (MAINDAT8) became publicly available as an SPSS export
file for mainframe computers. Although missing data codes were created where appropriate,
the archived version of the SPSS-X file did not declare any of these values to be missing. Users
should consult the codebook in this manual to determine which codes for which variables they
wish to treat as missing values. To safeguard the confidentiality of the GSS respondents, the
county- and zipcode-level variables were not released to the public along with the other
variables. An SPSS export file containing these sensitive variables, which can be merged to
recreate the entire NOS data set, requires a written plan of research usage, a signed pledge of
confidentiality, and a security deposit refundable on return of the data. For information on how
to obtain this export data set, write to Professor David Knoke, Department of Sociology, 909
Social Sciences Bldg., University of Minnesota, 267 19th Avenue South, Minneapolis, MN,
55455. A copy of the codebook for all the variables in the the 1991 NOS data is included at the
end of this technical manual.
6. DOCUMENTATION FOR USE OF THIS DATA
The data set provided is the version used by the principal investigators. It is important that
the original collectors of this data as well as the distributor of the data not be held responsible
for secondary use of the data and for interpretations and findings presented in print by secondary
users. The following is a form of a disclaimer that we encourage all users of this data to
employ.
"The data (and tabulations) utilized in this (publication) were made available (in part) by
the archive or agency which distributed the data. The data for the 1991 National
Organizations Survey were originally collected by the Survey Research Laboratory at the
University of Illinois and the National Opinion Research Center at the University of
Chicago. Neither the original sources or collectors of the data nor the distributor of the
data bear any responsibility for the analyses or interpretations presented herein."
An appropriate citation for users of the NOS is:
Kalleberg, Arne L., David Knoke, Peter V. Marsden and Joe L. Spaeth. 1991. The 1991 National
Organization Survey [machine readable data file]. University of Minnesota [producer]
1992. Inter-university Consortium for Political and Social Research (ICPSR) [distributor]
1993.
Individuals receiving and using these data are strongly urged to inform the distributor of the data
of any errors and discrepancies that are discovered during the course of using these data. Users
are particularly urged to contact the archive about problems and difficulties which prevented
effective and convenient utilization of the data. This information is necessary in order to
improve the data and to facilitate more efficient and economic processing of the data. Users are
also asked to provide information as to significant subsets and special aggregations of data that
are developed in using these data. Finally, in order to provide agencies with essential
information about the use of archival sources and to facilitate the exchange of information about
research activities, each user is expected to send two copies of each completed manuscript (or
thesis abstract) to the distributor:
Inter-university Consortium for Political and Social Research
Box 1248
Ann Arbor, MI 48106
(313) 764-2570
REFERENCES
McPherson, J. Miller. 1982. "Hypernetwork Sampling: Duality and Differentiation Among
Voluntary Organizations." Social Networks 3:225-249.
Spaeth, Joe L. 1985. "Job Power and Earnings." American Sociological Review 50:603-617.
Sudman, Seymour, Monroe G. Sirken and Charles D. Cowan. 1988 "Sampling Rare and
Elusive Populations." Science 240:991-996.
Table 1. Results of GSS Nominations Attempts from Occupations Reported for
Respondents or Spouses
______________________________________________________________________________
ACTIVITY N %
______________________________________________________________________________
TOTAL OCCUPATIONAL INFORMATION 1,427 100.0%
INITIALLY INADEQUATE INFORMATION 371 26.0%
GSS Respondent Refusal 172 100.0%
Usable 6 3.5%
Unusable 166 96.5%
Missing Data on Name/Address 199 100.0%
Usable 125 62.8%
Ineligible 5 2.5%
Unusable 69 34.7%
INITIALLY ADEQUATE INFORMATION 1,056 74.0%
Initially Adequate Information 1,056 100.0%
Usable 996 94.8%
Ineligible 54 3.8%
Unusable 6 0.6%
TOTAL OUTCOMES 1,427 100.0%
Usable 1,127 79.0%
Ineligible 59 4.1%
Unusable 241 16.9%
______________________________________________________________________________
Table 2. Results of SRL Interview Attempts
______________________________________________________________________________
ACTIVITY N % N %
______________________________________________________________________________
TOTAL USABLE INFORMATION 1,127 100.0%
DUPLICATE EMPLOYERS 60 5.3%
All Duplicates 60 100.0%
Refused 14 23.3%
Pending 7 11.7%
Completed 39 65.0%
SRL INTERVIEW ATTEMPTS 1,067 94.7%
TELEPHONE INTERVIEWS ONLY 633 100.0%
Refused, No Mail 54 5.1%
Pending, No Mail 18 1.7%
Completed by Telephone 561 52.6%
MAIL QUESTIONNAIRE 434 100.0%
Refused 192 44.2%
Pending 115 26.5%
Completed Mail Questionnaire 127 29.3%
TOTAL OUTCOMES 1,127 100.0%
Refusal 260 23.1%
Pending 140 12.4%
Completed 727 64.5%
______________________________________________________________________________