14 NSF Workforce Demographics Forecasting
NSF Workforce Demographics Forecasting
The National Science Foundation (NSF) has undertaken an effort to develop tools that enable us to leverage the full spectrum of diverse talent that society has to offer in the way we “recruit, retain and develop a high-performing workforce that values fairness, diversity, and inclusion to promote the progress of science.” This allows us to rethink how we harmonize diversity of talent, skills, and abilities with merit principles without compromising one for the other. To this end, the NSF has set forth three Strategic Goals (SG):
- Promote workforce diversity by recruiting from a diverse, qualified group of potential applicants to secure a high-performing workforce drawn from all segments of society.
- Foster workplace inclusion by cultivating an environment that encourages collaboration, flexibility, and fairness to enable individuals to contribute to their full potential and promote retention.
- Ensure the sustainability of these efforts by developing structures and strategies to equip leaders with the ability to manage diversity, be accountable, measure results, refine approaches based on data, and institutionalize a culture of inclusion.
The Office of the Chief Diversity and Inclusion Officer (OCDIO) has launched a project aimed at assessing the current level of diversity among the NSF workforce (to inform SG1), measuring the results of on-going and pending actions aimed at promoting and sustaining workforce diversity, and evaluating the efficacy of these actions to refine them empirically (to inform SG3). Consistent with SG2, the OCDIO has been communicating and collaborating with relevant departments and offices (Office of General Counsel, Human Resources Management, and the National Center for Science and Engineering Statistics) to ensure that pertinent data are being used, stakeholder utility is maximized, and appropriate methodologies are employed.
The OCDIO’s project is a broad effort to capture the historical and current dynamics of the NSF workforce as compared to the civilian workforce in general, as well as to project the dynamics of the NSF workforce to develop baseline expectations against which to compare moving forward. This structure is useful and informative on its own, but it also serves as the scaffolding to build hypothetical scenarios of impending and potential actions that are likely to affect recruitment and retention processes, such as position description modernization, authorization to expand hiring flexibility, targeted hiring, and the implementation of new policies (actions which are referred to as “events” in the forecasting tool). These actions are either in the process of being rolled out or are in discussion and/or development within various NSF departments. Initial results with a simple model (using generalized linear regression of a Poisson family with a logit link function) indicate good statistical fit of the model to the data.
While each of these events is expected to have some influence on the workforce, the exact nature of the influence (i.e., its direction and strength, or magnitude) is difficult to predict in advance and there is a paucity of data from which to pull to inform expectations. This leaves practitioners having to rely on intuition to estimate a range of potential magnitudes with great uncertainty (i.e., wide margins of error). The forecasting tool itself can be leveraged to circumvent this complication by introducing simulated effects of magnitudes that span the range of possible (or even plausible) values. That is, for example, allowing the magnitude of the influence of an event to vary from 1% to 50%, including all values in between. This produces a spectrum of potential outcomes from different events compared side-by-side for decision makers to evaluate at no cost and with no risk.
The remainder of this document is dedicated to the use and interpretation of the Underrepresentation Tool dashboard. The data demonstrated in the dashboard and the model used to develop forecasts are described below. There are also instructions for navigating and interacting with the dashboard to explore the subsets of data considered.
