Insider stories: analyzing internal sustainability efforts of major US companies from online reviews

Establishing whether a company genuinely supports internal sustainability efforts (ISEs) such as gender equality, diversity, and employee well-being is challenging. The lack of clear methodologies to operationalize these practices and the scarcity of reliable data documenting these efforts further complicate the assessment. This research addresses this gap by developing a novel framework for understanding and measuring ISEs. The study's primary objective is to analyze the relationship between a company's commitment to ISEs and its subsequent success, measured through stock growth and employee ratings. Understanding this relationship is crucial for both companies and policymakers. Companies can use this knowledge to optimize their internal sustainability strategies, while policymakers can leverage it to design more effective policies that encourage and incentivize sustainable business practices. The importance of this research lies in its potential to bridge the gap between theoretical understanding of sustainability and its practical implementation in the corporate world. By providing a robust methodology for measuring ISEs, this study lays the groundwork for future research and informs evidence-based decision-making in the realm of corporate sustainability.

Existing literature often treats corporate sustainability as a monolithic construct, lacking a nuanced understanding of internal practices. While external sustainability initiatives have received considerable attention, internal efforts impacting employees have been relatively understudied. Previous frameworks for internal sustainability often focused solely on social aspects, lacking a comprehensive approach. This study leverages the well-established UN Sustainable Development Goals (SDGs) to create a more holistic and robust framework for understanding and measuring ISEs. This approach addresses the limitations of existing frameworks by incorporating a broader range of internal sustainability aspects. The integration of the UN SDGs provides a strong theoretical foundation, ensuring the framework’s relevance and generalizability across various contexts.

The study employed a three-step mixed-methods approach to define ISEs. **Step 1 (Pre-selection):** The 17 UN SDGs were reviewed, and 13 were deemed relevant to the corporate context by three independent annotators. **Step 2 (Unsupervised discovery):** A deep-learning framework based on the Sentence-BERT algorithm was developed to analyze over 350,000 employee reviews from Glassdoor for 104 major US companies (2008-2020). This framework scored each review against the 13 pre-selected SDGs. Independent annotators validated the framework's relevance, leading to the selection of 8 SDGs. **Step 3 (Consolidation):** The three annotators merged semantically similar SDGs from the previous step, resulting in a final framework with six ISE dimensions: Monetary, Health, Education, Diversity, Infrastructure, and Atmosphere. Each ISE was operationalized using keywords identified through TF-IDF analysis of relevant reviews (see Figure 4). Company-level ISE scores were calculated by averaging the similarity scores between employee reviews and the corresponding ISE vectors (using SBERT similarity scores). A threshold of 0.31 was determined for the SBERT similarity score, paired with ISE-specific thresholds based on the 95th percentile of each ISE's distribution. The rationale for using this thresholding is to select only strong positive signals regarding ISEs. To analyze the relationship between ISEs and company success, two measures were used: online employee ratings (balance, career, culture, management, overall) and stock growth (2009-2019). Principal Component Analysis (PCA) reduced the six ISE dimensions to two main components: Staff Welfare (PC1) and Financial Benefits (PC2). OLS regression was employed to predict employee ratings based on PC1 and PC2, while the relationship between ISE scores and stock growth was visualized in a scatterplot (Figure 5). Industry sector differences in ISE scores were examined using boxplots and MANOVA.

The study's findings demonstrate a strong positive correlation between ISE scores and company success. **RQ1 (Validation of ISE scoring):** The deep learning method for detecting ISEs demonstrated high face validity. Keyword analysis showed that the method effectively captured the nuances of each ISE, with many keywords exhibiting high discriminative power for their respective ISEs. Notably, keywords related to work-life balance and flexible working conditions were linked to both 'Health' and 'Diversity' ISEs, highlighting the interrelation between these dimensions. **PCA analysis:** This revealed that two principal components, 'Staff Welfare' and 'Financial Benefits', explained 88% of the variance in company-level ISE scores. 'Monetary' ISE was relatively independent of other dimensions, while the remaining five showed strong interconnections. **RQ2 (ISEs and company success):** OLS regression analysis indicated that 'Staff Welfare' was strongly positively correlated with all aspects of company ratings (balance, career, culture, management, overall), explaining up to 64% of the variance. Figure 5 shows that companies with high scores on both 'Staff Welfare' and 'Financial Benefits' demonstrated greater stock growth. 'Staff Welfare' exhibited a stronger association with stock growth than 'Financial Benefits'. **RQ3 (ISEs across sectors):** MANOVA analysis revealed significant differences in ISE scores across various industry sectors. The 'Industrials' and 'Information Technology' sectors led in 'Staff Welfare' ISEs, followed by 'Financials'. The 'Health Care' sector showed high variability, possibly due to the demanding nature of the work. 'Consumer Staples' and 'Consumer Discretionary' sectors lagged significantly behind in both 'Staff Welfare' and 'Financial Benefits'. Figure 7 provides a more granular view, illustrating variations within sectors, highlighting differences between companies like Costco, which scored high on financial benefits but low on staff welfare, and others like Dollar General, which performed poorly on both aspects. Even within the high-performing IT sector, there were noticeable variations between companies such as Microsoft and more traditional firms such as Infosys.

This study provides strong evidence for the microfoundations of internal sustainability and its strategic importance for companies. The findings support a two-factor conceptualization of ISEs – 'Staff Welfare' and 'Financial Benefits' – highlighting the need for a holistic approach that balances both aspects. The strong positive correlation between high ISE scores and improved company performance (both in terms of employee ratings and stock growth) underscores the strategic value of investing in ISEs. The sector-level analysis reveals variations in the adoption of ISEs, with industries such as IT and Industrials leading the way. This suggests that policy interventions could be targeted towards specific sectors to maximize their impact. The study’s findings have implications for scholars, policymakers, and company managers, providing a robust framework for analyzing ISEs and informing evidence-based decision-making.

This research contributes significantly to the understanding of internal corporate sustainability by developing a validated six-dimension framework for ISEs, grounded in the UN SDGs. The findings highlight the strong positive relationship between ISEs, especially 'Staff Welfare', and company success, both financially and in terms of employee satisfaction. The sector-specific analysis identifies leading and lagging sectors, offering valuable insights for targeted policy interventions. Future research should investigate the causal relationship between ISEs and company success, explore additional ISE dimensions, and extend the analysis to a broader range of companies and geographical areas. Further work can also use the proposed framework to investigate other relevant aspects of employee experiences.

The study has several limitations. First, the list of ISEs might not be exhaustive, although a rigorous mixed-methods approach was used to establish its validity. Second, data limitations restrict the analysis to companies with sufficient review data, and the analysis does not account for key financial events before 2008. Third, the study lacks causal claims, only demonstrating correlations. Fourth, the data's representativeness may be affected by the unequal distribution of reviews across different sectors. While efforts were made to mitigate this issue, potential self-selection bias remains a concern. Fifth, the focus on employee reviews may not capture the full spectrum of internal sustainability initiatives.