12 Interim Update on NSF’s Under-representation and Barrier Analysis Tool

August 11, 2023

SUBJECT: CDIO Interim Update on NSF’s Under-representation and Barrier Analysis Tool

Over the past several months, I’ve highlighted an effort led by the CDIO team to develop the “NSF Under-representation and Barrier Analysis Tool.” One of the things I’m most excited about is our development of an Underrepresentation Framework. This tool is complementary to our MD-715 report. It serves as an analytical framework intended to ensure that NSF remains competitive with industry and other sectors by leveraging data from authoritative data sources such as the Census Bureau and Bureau of Labor Statistics (BLS) data to assess whether the full spectrum of diverse talent is fairly represented in our workforce. It will ultimately give us a data-informed foundation for ensuring that diversity and merit work in harmony without compromising one for the other regarding employment decisions.

In recent weeks, we began to review and refine the mapping and tables developed for the NSF STEM occupational specialties with the Standard Occupation Classification (SOC) system as a proof of concept with the model. The concept leverages data from the American Community Survey, the Census Bureau, and the Department of Labor (DOL) to develop analogs to conduct a comparative analysis of the NSF workforce’s demographic information with the society’s representative labor force. The concept also leverages historical NSF workforce data to begin developing predictive analytic capability that will be ameliorated over time. This concept demonstrates the level of diversity that can be achieved if respective segments of society were available (at the National =, State = and = occupational specialty levels). Initial data visuals of initial progress are provided later in this document.

In summary, this is a three Phase effort involving NSF data management, BLS data management, and model development.

Phase 1: NSF Data Management [Complete]

Ten years of NSF data has been reshaped to be suitable for import into Tableau. These data have been visualized as several worksheets and dashboards in Tableau (examples attached). They show current state snapshots of the workforce, broken down by Occupational Series, Sex/Gender, and Race/Ethnicity categories. We have leveraged the native forecasting capabilities of Tableau to generate seven-year forecasts for the trends of each category within each focal demographic metric.

Phase 2: BLS Data Management [In Progress]

We have downloaded nine years of BLS data and are currently reshaping it to match the structure of the NSF data for import into Tableau. Once ready, we will, incorporate it directly into the existing worksheets/dashboards for comparison, again leveraging the native forecasting capabilities of Tableau.

Phase 3: Model Development [Not Started]

Tableau offers built-in forecasting capacity, like Excel. But the models are not modifiable and provide limited predictions based heavily on weighted averages of furnished data.

We have started to build a more rigorous and robust predictive model in an external software platform (R Studio) and utilizing Tableau for visualization. We will work closely with the National Center for Science and Engineering Statistics (NCSES) and our Human Resources Management (HRM) team to begin defining parameters, hypotheses and levers based on known manageable parameters (that is, features for which actions such as position description updates, policy recommendations, use of hiring flexibilities etc.) can be implemented to bring about changes in the makeup of the workforce and the trajectory in which we reach population parity in occupational specialties where we see under-representation. Modifying these parameters will reveal expected outcomes for the workforce and would be observable in model output. We anticipate seeing expected trends and where they occur in the range of possible outcomes to provide insight into the likelihood of the action being positive, negative, or neutral.

Pages 3 and 4 of this document provides some of the initial data visuals that were presented as a progress review. The snapshots below depict under-representation trends benchmarked against the represented labor force with data from the authoritative data sources highlighted previously in this document. The below visuals highlight trends from the Asian American and Black demographics represented in the NSF workforce as well as the initial predictive analytics and level of confidence regarding the projections. The NSF workforce information includes current workforce data and roughly ten years of historical data to depict the initial predictive analytic.

As stated earlier in this document, one of the next phases of the model build out is a more rigorous and robust predictive model in an external software platform like R or Python and utilizing Tableau for visualization. These models would be collaboratively developed with NCSES and HRM staff input and based on known manageable parameters (that is, features for which policies and actions can be implemented to bring about changes in the makeup of the workforce). At end state, the model will reveal how modifying these parameters will reveal expected outcomes for the workforce and would be observable in model output. We anticipate seeing expected trends and where they occur in the range of possible outcomes to provide insight into the likelihood of the action being positive, negative, or neutral.

Trends in NSF STEM Workforce – Asian

Trends in the NSF STEM Workforce – Black

Trends in the NSF STEM Workforce – Hispanic or Latino

Concept for integration into HR Policies, Practices and Processes

The top talent management processes most susceptible to bias are promotions and/or succession, recruiting, performance management, manager interactions, organizational changes, team interactions, total rewards, and onboarding. At times, systemic bias underpins many of the organizational barriers to advancement. Organizations create their cultures and processes and develop stakeholder expectations for majority groups, making it more difficult to hire, promote and advance underrepresented talent. Through this lens and utilizing the under-representation framework explained in this paper, the following outlines an initial approach to integrate the framework into general HR concepts, policies and processes:

• Strategic Recruitment Plan: Develop an NSF Strategic Recruitment Plan lead by HRM and coordinated with the CDIO and other key stakeholders that utilizes insights from NSF’s barrier analysis and underrepresentation tool to focus on untapped talent and underrepresented talent pools. The need to reimagine and broaden NSF’s talent sourcing and recruitment apertures is a significant step to reaching untapped talent and underrepresented talent pools. Additional, specific opportunities to better come alongside, empower, and integrate with our HRM partners are as follows:
• Use the predictive analytic portion of the UR Tool to focus on strategic recruitment priorities and targets as part of the Strategic Recruitment Plan.
• Partner with HRM by ensuring awareness of any potential barriers to entry as HRM advises hiring managers on the most appropriate selection tools to assess required competencies at which phase(s) of the hiring process (e.g., interviews, writing assessments, situational judgement tests, skills assessments, etc.).
• Support HRM in their efforts in ensuring the talent team is involved in the development of any new assessments and advises hiring managers to develop appropriate interview questions and rating guides to reduce bias in the hiring process.
• Utilize applicant flow data (via the Electronic Data Warehouse) to assess how the demographics of the applicant pool shift at each of the major stages in the federal hiring process: application, qualification, referral and selection.
• Customizing, branding, and publishing the employee value proposition (EVP) or “what’s in it for them” to ensure we do not unintentionally divert underrepresented talent from our NSF talent funnel. EVPs appeal differently to different groups, and we need to account for this in our JOAs, branding, engagements, and media footprint.
• Further mitigate potential barriers by ensuring inclusive job descriptions that reframe telling the story of the duties of the position so applicants can better see/envision themselves in that particular role. Oftentimes job descriptions are reused/recycled and become outdated in their relatability to prospective talent.
• Mandating diverse interview panels being included as part of the auditable “recruitment file” whereby the OECR/CDIO ensures/certifies in writing that each hiring panel is sufficiently diverse.

• Measures of Effectiveness and data to determine organizational change and progress of the framework:
• Representation data (UR/Barrier Analysis Tool and MD 715 Report)
• Inflow (e.g., Applicant flow data)
• Outflow (e.g., Attrition & retention data)
• Upflow (e.g., Promotion data)

• HR Policy Development and Governance: Include CDIO as coordinating/contributing partner in HRM Policy development, governance, and review to ensure DEIA principles are woven into the fabric of our HR policies, processes, and procedures. Additionally, identify which policies warrant early and up-front involvement by CDIO in the policy development phase/lifecycle. Specifically, regarding talent management policies and procedures, we can utilize talent analytics as an input to enable effective talent planning and optimization. We can leverage the HR consulting component of the HR Business Partner (HRBP) function in HRM to utilize the data provided by the UR Tool for HR to advise managers before and during the hiring process to enable effective talent planning. This includes leveraging the applicant flow data (once built out) to provide a greater understanding of how demographics change through the various steps in the recruitment, rating, assessment, and selection processes.

• HRStat & Human Capital Operating Plan (HCOP): Partner with HRM to incorporate DEIA measures as part of HRM’s HRStat and Human Capital Operating Plan (HCOP) as core elements of the Human Capital Accountability program. Approaching DEIA as an outcome, integrating CDIO’s data and analytics into these planning and reporting functions to provide additional insight on human capital focus areas, broad initiatives, and outcomes that NSF leadership cares about most. Integrate with HRM to help elevate the HCOP and HRStat reporting to provide stakeholders with the information and evidence needed for effective decision making, as well as re-focusing on strategic aiming points and results.

• Operationalizing the framework as a component of hiring processes: The diagram below depicts the concept for using the data and framework as a part of hiring processes and feedback loop for policy recommendations:

The CDIO Team is on track to demonstrate additional progress of the tool and framework in the coming weeks. As progress continues, the CDIO team is working closely with internal stakeholders to continue refining and integrating of this framework into Human Capital processes and strategies. We will also continue discussions with various groups, advisory committees, and stakeholders supporting NSF to find opportunities to replicate the framework in other areas as we look to increase internal and external capacity to not only create opportunities but genuinely leverage the full spectrum of talent to protect National Security, advance science, and well-being for the good of society.

Dr. Charles (Chuck) Barber
Chief Diversity and Inclusion Officer
U.S. National Science Foundations