Purpose. The Early Career Doctorates Survey (ECDS) is designed to provide nationally representative statistics on recent doctorate (or equivalent) recipients working at U.S. master's degree- or doctorate-granting academic institutions (excluding medical schools and centers) and federally funded research and development centers (FFRDCs). Established by the National Center for Science and Engineering Statistics (NCSES) within the National Science Foundation to address the need for greater information on postdoctoral researchers (postdocs) and individuals working in the United States who earned their first doctoral degree abroad, the ECDS provides the most comprehensive data collected to date on the demographics, labor market experiences, and jobs held by doctorates in the first decade after earning their degree. These data include information on job quality and training, professional activities and achievements, work-life balance, mentoring, research opportunities, and career plans.

Data collection authority. The information is solicited under the authority of the National Science Foundation Act of 1950, as amended, the America COMPETES Reauthorization Act of 2010, and the Confidential Information Protection and Statistical Efficiency Act of 2002. The Office of Management and Budget control number for this survey is 3145-0235, which expired on 30 September 2019.

Survey contractor. RTI International.

Survey sponsors. NCSES and National Institutes of Health (NIH).

Frequency. The frequency of this data collection has not been established.

Initial survey year. 2017. This was the first full-scale ECDS. The pilot study was conducted in 2015.

Reference period. The week of 1 October 2017.

Response unit. Individuals working at U.S. academic institutions (excluding medical schools and centers) and FFRDCs.

Sample or census. Sample.

Population size. Approximately 186,700 individuals. Note: Estimating the size of the U.S. early career doctorate population is a goal of the ECDS.

Sample size. 15,465 individuals.

Target population. The 2017 ECDS target population was all individuals who

• Earned their first doctoral degree (PhD, MD, or equivalent) between 1 July 2007 and 30 June 2017, and
• Were working in a master’s degree– or doctorate–granting U.S. academic institution (excluding medical schools and centers) or FFRDC during the week of 1 October 2017.

Sampling frame. Lists of potential early career doctorates—persons receiving their first doctorate within the past 10 years—working at institutions sampled from the set of academic institutions included in the 2016 Survey of Graduate Students and Postdoctorates in Science and Engineering (GSS), and all FFRDCs listed on the FFRDC Master Government List maintained by NSF (https://www.nsf.gov/statistics/ffrdclist/).

Sample design. The ECDS employed a two-stage sample design.

The first stage was a stratified sample of 344 institutions with academic institutions, FFRDCs, and the NIH Intramural Research Program (NIH IRP) placed in separate strata. Within academic institutions, large universities with medical schools and centers were split into medical and non-medical sampling units. The non-medical academic sampling units were then stratified into three subgroups based on Carnegie classification.

In the second stage, a stratified sample of potential early career doctorates was selected from each of the responding institutions. The sample was stratified by postdoc status, citizenship, race, and sex to enhance representation across these key domains.

Due to low response rates and the resulting potential for nonresponse bias in subpopulation estimates, data for the medical schools and centers and the NIH IRP strata are excluded from published tables and figures. However, for research purposes, restricted-use data are available that include responses from early career doctorates working at medical schools and centers and the NIH IRP. (Learn how to apply for access to restricted use microdata at https://www.nsf.gov/statistics/license/.)

Data collection. As noted above, the ECDS was a two-stage data collection. The first stage included obtaining lists of all potential early career doctorates working at the sampled institutions the week of 1 October 2017 and who had earned their first doctorate or equivalent between 1 July 2007 and 30 June 2017. When highest degree or doctoral award date were missing for some early career doctorates within the sampled institution’s administrative data, those institutions were asked to include all individuals with job titles that were likely to be filled by doctorate holders.

For the second stage of data collection, two modes of data collection were available to early career doctorates: a Web-based survey and computer-assisted telephone interviewing (CATI). The recruitment protocol included a notification contact followed by an e-mail containing a unique link to the Web survey. A series of multimodal reminders (e-mail, mail, and phone) were used to improve response rates. These reminders ceased when the sample member completed the survey, was determined to be ineligible, or refused to participate. No incentives were offered to participants.

Mode. Stage 1 respondents uploaded their lists of early career doctorates through a secure Web portal (SSL FTP site). Almost all stage 2 respondents completed the survey on the Web (91% used a computer to access the instrument, and 6% used a tablet or cellphone); 3% responded using CATI or a mix of CATI and Web-survey modes.

Response rates. The response rate calculations below adhere to American Association for Public Opinion Research standards for computing response rates.^{​See response rate calculations in American Association for Political Opinion Research (AAPOR). 2016. Standard Definitions: Final Dispositions of Case Codes and Outcome Rates for Surveys. 9th ed., p. 61. Deerfield, IL: AAPOR.See response rate calculations in American Association for Political Opinion Research (AAPOR). 2016. Standard Definitions: Final Dispositions of Case Codes and Outcome Rates for Surveys. 9th ed., p. 61. Deerfield, IL: AAPOR.See response rate calculations in American Association for Political Opinion Research (AAPOR). 2016. Standard Definitions: Final Dispositions of Case Codes and Outcome Rates for Surveys. 9th ed., p. 61. Deerfield, IL: AAPOR.} For the academic institutions (excluding medical schools and centers) and FFRDCs, the institution response rate was 77.0%. The individual-level response rate was 65.1%. (See technical tables A-1 and A-2 for the stages 1 and 2 response rates.)

As noted above, due to low response rates, data for the medical schools and centers and the NIH IRP strata are excluded from published tables and figures. For research purposes, the restricted-use data include responses from early career doctorates working at medical schools and centers and the NIH IRP.

Data editing. The data collected in the ECDS were subject to both manual and automated data processing and correction procedures. After data collection, several types of data edits were applied.

These included the application of “reserve codes” for logical skips and different types of missing data, as well as rule-based and case-level edits. If an item was skipped due to the routing logic of the instrument, the missing value was replaced with a value of -1, which is the code reserved for logical skip. All remaining missing responses were set to a missing reserve code (-8) if the respondent stopped responding prior to reaching the item or -9 for all other types of nonresponse.

Imputation. Questionnaire items with missing data within the ECDS were imputed through an approach that included the use of three imputation techniques: (1) cold-deck imputation, (2) logical imputation, and (3) hot-deck imputation. Some items experienced all three techniques, whereas other items experienced only one or two of the techniques. As an example, for the questionnaire items in the education history section, only cold-deck imputation was used, and as a result, these items could include some missing values after the imputation process. Cold-deck imputation used information from the 2015 ECDS pilot study frame data or the Doctorate Records File to fill missing data within common items (e.g., gender and year born) based on an exact, unique match for certain respondents. Logical imputation was used when a relationship among missing items in the questionnaire could be deduced (e.g., if a respondent is a U.S. citizen, logically the respondent should skip the Green Card question as it does not apply). Data still missing after the use of the first two imputation techniques were then imputed through a cyclical weighted sequential hot-deck (WSHD) method which replaces missing data by imputing plausible values from statistically selected donor cases (stochastic imputation) (Cox 1980; Iannacchione 1982). The screener and basic demographics sections of the questionnaire and certain additional questionnaire items were selected and imputed first. These core questionnaire items were then available to be used in the construction of the imputation cells for all other sections. Items were imputed by questionnaire section and the order they were asked within the section.

Weighting. The ECDS used a complex sample design. To produce population estimates, person-level analysis weights were calculated that accounted for the differential sampling rates used in the first and second stages and were then adjusted to account for nonparticipation and unreleased institutions in the first stage and nonresponse and unknown eligibility in the second stage. Estimates were calculated as the sums of the final person-level analysis weights. Replicate weights used in the jackknife variance estimation technique were calculated in a similar manner.

Detailed information on the weighting procedures is contained in the ECDS Methodology Report, available upon request from the ECDS Survey Manager.

Variance estimation. For a clustered sample like ECDS (where early career doctorates are clustered within institutions), the estimate of the variance is calculated based on the variability across cluster-level estimates. With a stratified sample, this variability is calculated within each sampling stratum and then aggregated across strata. To account for complex sample design and weighting adjustment steps we developed replicate weights for variance estimation (Wolter 2007), specifically using the delete-1 jackknife method.

Users of the data can calculate estimates and their variance estimates (or standard errors) using the full-sample weights and the 183 replicate weights.

Disclosure protection. To protect against the disclosure of confidential information provided by ECDS respondents, the estimates presented in data tables and the InfoBrief are rounded and small cells are suppressed. All weighted counts are rounded to the nearest 100, standard errors are rounded up to the nearest 50, although calculations of percentages are based on unrounded estimates. Median salaries are rounded to the nearest $1,000, and the associated standard errors are rounded up to the nearest $500. Cells are suppressed when the raw count is less than 5, as indicated by the letter D in tables.

Sampling error. ECDS estimates are subject to sampling errors. Estimates of sampling errors associated with this survey were calculated using the jackknife replicate weights. Data table estimates with coefficients of variation (that is, the estimate divided by its standard error) that exceed 50% are deemed unreliable and are suppressed. The letter S indicates this type of suppression in a table cell.

Coverage error. At the institution level, coverage extends to all institutions in the GSS universe (a census of U.S. academic institutions granting master’s or doctorate degrees in science, engineering, or health) and all FFRDCs (based on a Master Government List maintained by NSF) and thus coverage error in minimal. A few small universities were dropped from the stage 1 sampling frame as they would not have had enough early career doctorates to obtain the required stage 2 sample (even if combined with another institution). The potential undercoverage associated with the dropping of these institutions is minimal (estimated at 0.2%) due to the small number of early career doctorates at these institutions.

At the individual level, undercoverage would result if an institution omitted some early career doctorates from the lists they provided (e.g., by leaving out individuals working at a particular center). Similarly, overcoverage could result if an institution’s list includes ineligible individuals (i.e., those who do not have a doctoral degree or whose degree was awarded more than 10 years ago). Throughout the list development process, list coordinators received guidance to help avoid these types of errors; and lists were checked against expected counts to minimize coverage error.

Nonresponse error. For the academic institutions (excluding medical schools and centers) and FFRDCs, the institution response rate was 77.0% and the individual-level response rate was 65.1%.

To examine the potential nonresponse bias in the 2017 ECDS data, a nonresponse analysis study was conducted. Results of the study showed that all detectable differences were properly addressed by the nonresponse weighting adjustments of the survey data.

As noted above, due to low response rates and the resulting potential for nonresponse bias in subpopulation estimates, data for the medical schools and centers and the NIH IRP strata are excluded from published tables and figures. For research purposes, the restricted-use data include responses from early career doctorates working at medical schools and centers and the NIH IRP.

Measurement error. Cognitive testing of the survey instrument informed decisions to help minimize measurement errors from ambiguous questions and from a multimode survey approach.

This is the first full-scale implementation of the ECDS. Limited data elements are comparable with those from the ECDS pilot study and the Survey of Doctorate Recipients. As a result, use caution when comparing ECDS data with data from other sources.

Changes in survey coverage and population. As noted above, because of low response rates and the resulting potential for bias in subpopulation estimates, data for the medical schools and centers and the NIH IRP are excluded from the 2017 ECDS published tables and figures whereas the InfoBrief for the 2015 ECDS pilot study included data for medical schools and centers and the NIH IRP.

Changes in questionnaire. The 2017 ECDS instrument retained the core data elements of the 2015 ECDS pilot study instrument. However, a few major changes were made to the pilot study instrument for inclusion in the full-scale data collection. These changes were implemented to improve the quality and analytic utility of the data.

• The 2015 ECDS pilot study collected information on the first full-time paid position, whereas the 2017 ECDS collected information on the first position, whether it was paid or unpaid, full-time or part-time.
• The 2015 ECDS pilot study collected data on the current position; that is, the position the early career doctorate held when completing the survey. In contrast, the 2017 ECDS asked about the position that the early career doctorate held at the sampled institution during the week of 1 October 2017.
• The order in which data on these positions were collected differed as well. In the 2015 ECDS pilot study respondents were asked about their positions chronologically: first full-time paid position followed by current position. In the 2017 ECDS instrument, respondents were asked to report on the position they held during the week of 1 October 2017 followed by first position.
• To minimize item nonresponse on questionnaire items used to produce key estimates, the demographic information section was split into two parts for the 2017 ECDS; one of these parts was moved closer to the beginning of the survey, so complete information on sex, race and ethnicity, birth year, U.S. citizenship, and English as a first language was collected for most respondents.

Changes in reporting procedures or classification. The 2017 ECDS added the Institutional Portal, a website created for the full-scale survey to facilitate the completion of the stage 1 data collection by institutional staff. The portal allowed institutions to grant survey approval, appoint coordinators for the ECDS, upload the list of early career doctorates, and send the prenotification e-mails out to begin stage 2 data collection.

Citizenship and race. Institutions in stage 1 provided combined citizenship and race data on the lists of early career doctorates to be used in stage 2 sample selection. Non-U.S. citizens were collected as a single group, and for U.S. citizens, a simplified 3-category race variable was collected, corresponding to the stage 2 sampling strata: Asian, White, and underrepresented minority.

Doctoral or doctorate degree. For this survey, the terms “doctoral degree” and “doctorate degree” were used interchangeably to refer to any kind of doctoral degree or international equivalent, aside from a juris doctor (JD) degree. Examples included PhD, EdD, DSc, MD, DO, DDS, DVM, and MBBS, among many others.

Early career doctorate. An individual who received their first doctoral degree between 1 July 2007 and 30 June 2017 and was working at one of the institutions selected for the survey during the week of 1 October 2017. The individual could have held any type of position (e.g., faculty, staff, postdoctorate, nonfaculty research) and received their doctoral degree from an institution in any country and in any discipline. All doctoral-level degrees were eligible (e.g., PhD, EdD, DSc, MD, DO, DDS, DVM, MBBS, and international equivalents), with the exception of JDs.

ECDS methodological study. From July 2012 through February 2013, NCSES conducted a methodological study with 81 institutions to test the feasibility of obtaining lists of early career doctorates and then sampling and surveying the early career doctorates based on these institutional lists.

ECDS pilot study. Conducted between July 2014 and March 2015, the pilot version of the survey involved 176 unique institutions and was designed to improve upon the methodology and initial study design for full scale data collection.

Federally funded research and development center (FFRDC). One of the three types of institutions included in the survey, an FFRDC is a unique type of nonprofit organization that is sponsored and funded by the U.S. government to meet a special long-term research or development need (https://www.nsf.gov/statistics/ffrdclist/).

Field of doctorate. The doctoral field is as specified by the respondent in the ECDS at the time of the survey, coded using classification of instructional program (CIP) codes. The CIP codes were subsequently recoded to the NCSES Taxonomy of Disciplines (ToD) used in the ECDS tables. See technical table A-3 for the crosswalk between the CIP codes and ToD fields used in the ECDS tables. The CIP-ToD crosswalk used to create these fields and to enable comparison among NCSES surveys is available upon request from the ECDS Survey Manager.

Foreign doctorate degree. A respondent’s doctorate or equivalent was earned outside of the United States.

Postdoctoral researchers (postdocs). The definition of a postdoc varies by institution and individuals. Respondents were provided the following guidance: “For the purposes of this survey, a postdoctoral appointment, or “postdoc,” is a temporary position awarded in academe, industry, government or a non-profit organization primarily for gaining additional education and training in research.”

NCSES defines a postdoc as meeting both of the following qualifications:

• Holds a recent doctoral degree, generally awarded within the past 5 years, such as PhD or equivalent (e.g., ScD or DEng); a first professional degree in a medical or related field (MD, DDS, DO, DVM); or a foreign degree equivalent to a U.S. doctoral degree.
• Has a limited-term appointment, generally no more than 5–7 years, primarily for training in research or scholarship and working under the supervision of a senior scholar in a unit affiliated with the institution.

Race and ethnicity. Ethnicity is defined as Hispanic or Latino or not Hispanic or Latino. Values for those selecting a single race include American Indian or Alaska Native, Asian, Black or African American, Native Hawaiian or Other Pacific Islander, and White. Those persons who report more than one race and who are not of Hispanic or Latino ethnicity also have a separate value. Race and ethnicity were collected for all respondents, regardless of citizenship status.

Survey of Graduate Students and Postdoctorates in Science and Engineering (GSS). The GSS is an annual census of all U.S. academic institutions granting research-based master's degrees or doctorates in science, engineering, and selected health fields as of fall of the survey year. Eligible institutions offer at least one graduate program in science, engineering, or health. The GSS is sponsored by NCSES and by NIH.

Stage 1. Stage 1 of data collection involved working with a list coordinator at each institution to develop a list of early career doctorates to be used as a sampling frame.

Stage 2. Stage 2 of data collection involved inviting a sample of early career doctorates at each institution to participate in the Web-based ECDS to collect information about their academic and professional experiences.

Temporary visa holders. Individuals in the United States on temporary U.S. resident visas.

Underrepresented minorities (URMs). This category comprises three racial or ethnic minority groups (Blacks or African Americans, Hispanics or Latinos, and American Indians or Alaska Natives) whose representation in science and engineering education or employment is smaller than their representation in the U.S. population.

University-affiliated medical schools and centers. Medical schools, hospitals, or medical centers that were eligible for the GSS are affiliated with a university that was eligible for the GSS.

U.S. citizens and permanent residents. U.S. citizens, including those from Puerto Rico and the U.S. territories, and permanent residents holding permanent U.S. resident visas (Green Cards).