Goal

The Society for Human Resource Management (SHRM) defines diversity as the 'collective mixture of differences and similarities that includes for example, individual and organizational characteristics, values, beliefs, experiences, backgrounds, preferences, and behaviors.' More broadly, the trifecta of diversity, equity and inclusion (DEI) has been at the forefront of socio-political and economic discussion, with widespread recognition of systemic problems of inequity and exclusion in both the public and private spheres across industries, educational access and quality of life measures.Although the benefits of DEI are well established, especially in the workplace, there is surprisingly little public information about the impacts of DEI, both causally and correlatively, at different scales of space (e.g., country vs. state) and time, on human wellbeing and productivity measures. For example, education and exposure to more diverse populations have often been mentioned as important for the development of a well-rounded workforce ( and is embodied in universities' admittance of cognitively and racially diverse student bodies). Comprehensive studies at global scale have been lacking, however. The goal of this project is to conduct such an ambitious study in a data-driven and non-partisan manner by integrating, and using robust statistical methodologies on, multiple government and privately generated datasets.

Opportunity

Diversity, equity and inclusion are at the forefront of socio-political discussion, because of evidence that has arisen in specific contexts (such as the workplace, including company profitability and employee motivation) about their benefits. Yet, the full and global impacts of DEI on measures of human flourishing and wellbeing have still not been quantified and explored across space and time. Due to recent advances in data integration, statistics, as well as availability of data sources from government, non-profits and polling organizations, we have an opportunity to map these impacts and put the question to a rest.

Roadblocks

The main challenge is to infer causality by using a robust set of statistical methodologies, including methods that correctly distinguish between correlation and causation, that have been developed in multiple contexts and fields. Unlike the medical arena, we will not have the ability to run controlled experiments to measure impacts of DEI variables. However, the breakthroughs subsequently described would help us infer impact even without randomized trials.A secondary challenge is to refine the accuracy and usability of our AI and data integration tools to collect data across many sources over the Web and combine them for our studies.

Breakthroughs Needed

In order to truly measure global impacts of diversity, equity and inclusion (DEI) on human wellbeing, established datasets need to be collected, homogenized and integrated. These require advanced techniques in artificial intelligence, data integration and web sciences, along with access to datasets such as the Gallup World Poll, as well as scalable techniques to crawl public data from the United Nations and sources such as the Federal Reserve. Recent research in our group has made promising strides in all these directions.On the statistical front, we must use, refine and combine advanced methods (examples including difference-of-differences, Bayesian models and changepoint detection) refined across diverse fields like machine learning and econometrics on the integrated suite of datasets to measure impact. In some cases, only correlative relationships may be feasibly detected, while in other cases (such as factors for which more complete data and observations have been observed either cross-sectionally or longitudinally, or both) strong evidence of causality can be quantified. Recently, there has been enormous progress across the research communities noted above that lends credence to the belief that, with careful and dogged research, we can make the needed breakthroughs to quantify causal impacts of independent DEI variables.

Key Indicators of Success

By the 3rd year, studies using integrated data from the United States, Canada and Western Europe, on which DEI measures and policies have provable causal impact on human wellbeing measures, should have been completed and published.By the 5th year, a greater range of causal factors would be discovered (or debunked) and some correlative factors would be elevated to definitively causal (or non-causal). The range of countries should be expanded to Asia and South America.In the first 10 years, the findings above would be further validated and strengthened through field experiments and pilot policy implementations.

Additional Information

Technical Summary:A study published in the American Sociological Review found that companies with the highest percent in racial or gender diversity have higher sales revenue, more customers, higher than average market share and profitability. Other studies have concurred; yet, there is surprisingly little information, especially quantitatively, on (i) what variables can be used to measure different aspects of diversity, equity and inclusion (DEI), (ii) the possible context-dependence of any such variables (e.g., workplace vs. political representation vs. judiciary), (iii) cross-sectional variance and causal impacts of DEI on productivity and wellbeing, (iv) longitudinal change, and (v), impacts of policy measures on DEI, among other (secondary) research questions. Answering these questions, even tentatively, requires access to a range of polling and public datasets (including census data). Since different countries release data at different granularities, and even levels of accuracy or trustworthiness, a robust data integration scheme is necessary. At the same time, even given such an integrated and homogenized corpus, the lack of controlled experiments begs the question of whether we can truly move beyond correlation-analyses. We argue that methods like difference-in-difference and changepoint detection could be conceivably used, if the questions are scoped and the methodology is rigorous. Some of the more important findings could be validated through select field trials and pilot policy implementations that achieve the balance between natural and controlled experiments.

1. SHRM defines inclusion as the achievement of a 'work environment in which all individuals are treated fairly and respectfully, have equal access to opportunities and resources, and can contribute fully to the organization's success.' Source: https://www.shrm.org/resourcesandtools/hr-topics/pages/diversity-equity-and-inclusion.aspx
2. A good source for understanding difference-in-difference estimation: https://www.publichealth.columbia.edu/research/population-health-methods/difference-difference-estimation
3. A preprint describing an evaluation of changepoint detection algorithms: https://arxiv.org/abs/2003.06222
4. Multiple research projects conducted in the proposer's group on data and knowledge integration: https://usc-isi-i2.github.io/home/
5. Evidence that DEI impacts workplaces and company culture is described comprehensively in several sources.

We cite as an example, a recent book: 'Success through Diversity' by Carol Fulp. ISBN 9780807039854Collaborators who could serve as advisors/run projects to execute the idea:
1. Brittany Morey, University of California at Irvine
2. Nancy D Spector, Drexel University
3. Kristina Lerman, University of Southern California
4. Ryan Adams, Princeton University
5. Patrice Buzzanell, Purdue University