Methodological Research Concerning the Survey of Earned Doctorates

Methodological Research Concerning the Survey of Earned Doctorates
Note: Reports available upon request from Lynn Milan (lmilan@nsf.gov).
An Examination of the Institution Eligibility Criteria for the Survey of Earned Doctorates:
This report was the culmination of work conducted during 2013 at the request of NSF that
examined the eligibility criteria for institution inclusion in the SED against a broader national
and international context. The report also made recommendations for an adjudication process for
determining the eligibility of institutions and programs not currently in the SED but appearing to
meet the criteria for inclusion. The report is currently under consideration by NSF. No changes
have yet been implemented to the SED eligibility review process. (Kirby et al., 2014)
Survey of Earned Doctorates Confidentiality Report: This report presents the findings from
the cognitive interviews and focus groups conducted in 2013 and 2014 with doctorate recipients,
graduate deans, institution contacts and institution researchers concerning the confidentiality
procedures employed by the SED. Additional topics covered in the interviews and focus groups
included a revised consent statement and a proposed data linking and data sharing project among
select institutions. The report included findings and recommendations concerning the data
linking and data sharing project under consideration at NSF. No changes have yet been
implemented based on the findings. (Welch et al., 2014)
Timeline Data Quality Improvements for the Survey of Earned Doctorates: This report
presents findings from a study conducted in 2014 that analyzed the current approach the SED
employs to collect, edit and report timeline data. The report provided recommendations for the
increased utility and improved quality of the data through potential questionnaire changes, edit
changes and changes to how the data are contained in the DRF and presented in reports. Findings
were used to inform a number of process revisions, including: expansion of the auto-coding
process for timeline variables; modification of rules used to flag nontraditional timeline
sequences; addition of timeline variables to the Doctorate Records File (DRF) for use in further
research; and revision of select imputation rules. In addition, revisions to timeline questions in
the instrument are included in the planned future cognitive interview activities. (Bautista et al.,
2014)
Enhancements in Auto-Coding in the Survey of Earned Doctorates: This report presents
findings from a study conducted in 2014 on the feasibility of employing an automated coding
application to additional SED variables that are currently coded manually by trained clerks. The
report recommended additional coding rules that would allow for automated coding, reducing the
need for manual coding which in turn would improve quality and lower labor costs. Findings
from this study are being implemented in the SED 2015 round auto-coding activities. (Groenhout
et al., 2014)
After the Breakoff Part 2: Converting Web Break-offs to Completes This presentation was
discussed at the 2014 International Field Directors & Technologies Conference in 2014. The aim
of the presentation was to identify predictors of survey breakoff and identify the most successful
prompts for converting breakoffs to completions. Conclusions from the presentation included (1)

certain characteristics found in survey paradata are significant predictors of survey breakoff (2)
tailored email prompts were much less successful than standard email prompts and (3) standard
email prompts were more successful than mail prompts in converting breakoffs. Drawing on the
third finding listed above, the results of this study informed the 2015 non-respondent prompting
experiment, which is examining if sending email prompts before mail prompts result in a higher
survey completion rate. Respondents eligible for the experiment are those for whom both email
and mailing addresses are available. The control group will receive the standard protocol of mail
prompts first, followed by email prompts, if needed. The treatment group will receive the
prompts in the opposite order: that is, email prompts followed by mail prompts, if necessary.
(Groenhout et al., 2014)
Disclosure Analyses of Tabular Data: In this report, it is recommended to enhance cell
suppression in tabulation by estimating suppressed cells via log-linear modeling. The resulting
complete table is expected to have high utility for users at large because model-based best
prediction is used for suppressed cells from the available unsuppressed information. In particular,
the paper illustrates that estimates can be used to check the underlying trend over different
subgroups cross-sectionally and for a given subgroup longitudinally. The report found that the
proposed method uses only information released under cell suppression in tabulation and,
therefore, does not increase the disclosure risk. Also, it preserves all the unsuppressed cells and
marginal counts. No changes have yet been implemented based on this report. (Cohen et al.,
2013)
Department Coding Feasibility Study: Since 1969, the SED questionnaire has included a
verbatim item asking respondents to name the department of the university that supervised their
doctoral studies. Each year these data are captured in the annual SED data files, but they are not
cleaned, coded, or stored in the DRF. This study was conducted using verbatim responses to this
particular item from the 2011 SED round to examine the feasibility and costs for coding.
According to the report, coding the SED department verbatim responses is feasible for
respondents who reported that their field of study is in science and engineering. If put into
production the entire process could be integrated into standard SED verbatim coding to further
simplify the process and reduce setup and maintenance costs. Hence, the report recommends
adding a coded department verbatim variable to the DRF. No changes have yet been
implemented. (Groenhout and Toit, 2013)
Comparison between Screener/Follow-Up Item Format and Yes/No Item Format: In 2010,
SED changed the format of two demographic items (ethnicity and disability) from
screener/follow-up to yes/no format. This paper explored the impact of this change on the
responses to these items. The report analyzed the four most recent rounds of SED data (20082011); two rounds with screener/follow-up format and two rounds with yes/no format. The
authors did not find evidence of increased data quality in the new format, as they have observed
more “other” responses in the yes/no format. No changes were implemented as a result of this
analysis. (Hernandez, Arakelyan, and Welch, 2013)
Converting Web Breakoffs to Completes and the Effects on Data Quality: This presentation
examines web survey breakoffs in terms of respondent propensity to breakoff, the survey
administrator’s efforts at converting breakoffs to completes, and the data quality from these

surveys. In the first section, we found that the same demographic groups with higher rates of unit
nonresponse are more likely to break off after beginning a web survey. In the second section, we
found that mail prompts were more effective than email prompts at converting web breakoffs to
completed surveys, but recognized some limitations in this comparison and recommended a
prompt mode experiment to evaluate whether email prompts could be more successful. In the
final section, we compared rates of item nonresponse between web surveys completed in a single
session and those that were breakoffs and later converted to completes. We found no significant
difference, concluding that prompting efforts aimed at respondents who have begun but not
completed a web survey are worthwhile as the breakoff has no impact on data quality as
measured by item nonresponse. Findings from this study led to further research in 2014 (see
“After the Breakoff Part 2” above), which had differing results. During the 2015 round, a nonrespondent prompting experiment is being conducted in order to determine if sending email
prompts before mail prompts results in a higher survey completion rate. (Groenhout, Bilgen, and
Latter, 2013)
Mode Effects on Item Response Rates: This report examines item response rate by mode,
controlling for time of completion. According to the findings, web item response rates (RRs)
tend to be higher than paper and pencil item RRs. The presentation concluded that prompts may
play an important role in increasing item RR on the web. Also, the authors found that prompts
may convince early responders to provide very sensitive information (such as responses to Social
Security number and birthday questions), but not late responders. Thus, the web prompts have
remained in the web questionnaire. The web questionnaire redesign in 2015 will further improve
the prompts both for clarity and user experience. The effectiveness of these redesigned prompts
will also be included in the subsequent cognitive interview activities. (Webber, Welch, and
Hernandez, 2013)
Metadata Standards and Technology Development for the NSF Survey of Earned
Doctorates: This presentation was made at the Federal Committee on Statistical Methodology’s
annual conference in the fall of 2013. The presentation discussed the project by NCSES, NORC
and Metadata Technologies North American (MTNA) to develop new technologies to capture
comprehensive metadata, automate the production of essential documentation, and generate an
archival package for the annual NSF Survey of Earned Doctorates (SED). The work on this
project links to a larger project where NCSES is developing a unified data system to store and
disseminate all NCSES survey data. SED materials have been successfully delivered to the
NCSES data system using the automated tools discussed in this presentation. (Noonan et al.,
2013)