News Bundling in Emissions Targets Announcement

Authors:

Lilly Xu, Ling Li, Shirley Lu

Date Created:

2025-01-01

Course Title:

Professor:

Not specified

About Paper:

Across Global Industry Classification Standard (GICS) industry firms across various industry groups and applied both traditional groups and subgroups, corporations frequently form alliances and staggered difference-in-differences models, controlling for focused on setting emissions reduction targets. These targets can firm- and year-fixed effects, to evaluate how corporate emissions range from broad commitments (e.g. net zero carbon emissions by reporting changes in response to alliance announcements. These 2050) to highly specific goals (e.g. increase of the modal share of findings offer important insight into the potential spillover effects rail freight from currently 18% to 30% by 2030). We hypothesize of industry-wide climate commitments on firm-level target-setting that the announcement of such targets by industry-representative behavior. alliances leads to an increase in the average number of emissions targets reported by firms within the corresponding industry group. In addition, we hypothesize that country-level environmental policies may increase climate announcements in companies after To test this hypothesis, we compiled a dataset of alliances across being released. Similarly, we will test this with a difference-in- GICS industry groups, identifying those that publicly committed differences model. A GPT model will further separate the policies to emissions reductions and recording the earliest year of such announcements. Wethenmergedthisdatawithdataofover37,000 by ambiguous, enforceable, temporary, and reversible, to explore how different policy characteristics shape corporate responses. 80 Program for Research in Markets and Organizations StrategicNarrativeasSignal: UsingLanguageModelstoAnalyzeCEORhetoric Across High- and Low-Performing Firms Lynn Collardin, Luciana Silvestri, Ranjay Gulati Stanford University | Psychology | 2026 This project explores how corporate leaders use narrative to dimensions. shape organizational identity and signal strategic direction. We Wearenowbuildingandtrainingalargelanguagemodel(LLM)to analyze how executives frame their companies’ purpose, approach scale this coding process across thousands of firm documents. The to uncertainty, relationship with customers, and stance toward LLM is trained on the hand-coded excerpts and is currently being technology. The central question is whether high- and low- performing firms construct fundamentally different narratives and evaluated for its accuracy. One area of focus involves analyzing how those stories diverge thematically. how firms frame artificial intelligence, specifically comparing early vs. late adopters and examining whether AI is portrayed as We began by manually coding a range of company documents, opportunity, risk, or both. including earnings calls, annual reports, and shareholder letters, Preliminary patterns suggest that stronger performers use more from over 20 successful firms. This qualitative analysis revealed emotionally resonant, morally aspirational, and future-focused six recurring rhetorical dimensions: customer framing (rational vs.narratives. Meanwhile, struggling firms tend to default to rational, emotional), purpose (moral vs. utilitarian), response to uncertainty (consistencyvs. paradox), technologyorientation(future-vs. past- defensive messaging. If these narrative patterns consistently anchored), goal setting (ambitious vs. modest), and overall tone emerge ahead of major shifts, they could offer a unique method for identifying early signals of company success and how firms (optimistic vs. defensive). Among the sources reviewed, we respond to change. identified shareholder letters as the richest for narrative analysis. Future work will refine the model, expand the sample, and Using the 2024 S&P 500 total return ranking, we then created a explore how narrative framing relates to long-term resilience, and balanced dataset with 20 top-performers and 20 underperformers organizational identity. and manually coded their shareholder letters across the six

Abstract:

Across Global Industry Classification Standard (GICS) industry firms across various industry groups and applied both traditional groups and subgroups, corporations frequently form alliances and staggered difference-in-differences models, controlling for focused on setting emissions reduction targets. These targets can firm- and year-fixed effects, to evaluate how corporate emissions range from broad commitments (e.g. net zero carbon emissions by reporting changes in response to alliance announcements. These 2050) to highly specific goals (e.g. increase of the modal share of findings offer important insight into the potential spillover effects rail freight from currently 18% to 30% by 2030). We hypothesize of industry-wide climate commitments on firm-level target-setting that the announcement of such targets by industry-representative behavior. alliances leads to an increase in the average number of emissions targets reported by firms within the corresponding industry group. In addition, we hypothesize that country-level environmental policies may increase climate announcements in companies after To test this hypothesis, we compiled a dataset of alliances across being released. Similarly, we will test this with a difference-in- GICS industry groups, identifying those that publicly committed differences model. A GPT model will further separate the policies to emissions reductions and recording the earliest year of such announcements. Wethenmergedthisdatawithdataofover37,000 by ambiguous, enforceable, temporary, and reversible, to explore how different policy characteristics shape corporate responses. 80 Program for Research in Markets and Organizations StrategicNarrativeasSignal: UsingLanguageModelstoAnalyzeCEORhetoric Across High- and Low-Performing Firms Lynn Collardin, Luciana Silvestri, Ranjay Gulati Stanford University | Psychology | 2026 This project explores how corporate leaders use narrative to dimensions. shape organizational identity and signal strategic direction. We Wearenowbuildingandtrainingalargelanguagemodel(LLM)to analyze how executives frame their companies’ purpose, approach scale this coding process across thousands of firm documents. The to uncertainty, relationship with customers, and stance toward LLM is trained on the hand-coded excerpts and is currently being technology. The central question is whether high- and low- performing firms construct fundamentally different narratives and evaluated for its accuracy. One area of focus involves analyzing how those stories diverge thematically. how firms frame artificial intelligence, specifically comparing early vs. late adopters and examining whether AI is portrayed as We began by manually coding a range of company documents, opportunity, risk, or both. including earnings calls, annual reports, and shareholder letters, Preliminary patterns suggest that stronger performers use more from over 20 successful firms. This qualitative analysis revealed emotionally resonant, morally aspirational, and future-focused six recurring rhetorical dimensions: customer framing (rational vs.narratives. Meanwhile, struggling firms tend to default to rational, emotional), purpose (moral vs. utilitarian), response to uncertainty (consistencyvs. paradox), technologyorientation(future-vs. past- defensive messaging. If these narrative patterns consistently anchored), goal setting (ambitious vs. modest), and overall tone emerge ahead of major shifts, they could offer a unique method for identifying early signals of company success and how firms (optimistic vs. defensive). Among the sources reviewed, we respond to change. identified shareholder letters as the richest for narrative analysis. Future work will refine the model, expand the sample, and Using the 2024 S&P 500 total return ranking, we then created a explore how narrative framing relates to long-term resilience, and balanced dataset with 20 top-performers and 20 underperformers organizational identity. and manually coded their shareholder letters across the six

Source:

Harvard / Harvard College | Kirkland House | Economics | 2027 / 2025

Topics:

firm, industry, emission, target, acros, narrative, announcement, group, alliance, model, policy, company

Co-authors:

@lillyxu230 , @lingli231 , @shirleylu232