NGO Activism: Exposure vs. Influence
Extracted source text · arXiv 2411.06875
\documentclass[12pt,english]{article}
\usepackage[hyphens]{url} % Allow breaks after "-" in url
\usepackage{afterpage}
\usepackage{float}
\usepackage{etoolbox}
\AtBeginEnvironment{tabular}{\normalsize}
\usepackage{textcomp}
%\usepackage[labelformat=parens,labelsep=quad, skip=3pt]{cax ption}
% \usepackage[margin=1in]{geometry}
\usepackage[skip=0.1\baselineskip]{caption}
\usepackage[a4paper,
left=1in,
right=1in,
top=1in,
bottom=1in]{geometry}
\usepackage[english]{babel}
\usepackage[utf8x]{inputenc}
\usepackage{amsmath}
\usepackage{amssymb}
\usepackage[retainorgcmds]{IEEEtrantools}
\usepackage[flushleft]{threeparttable}
\usepackage{rotating,booktabs}
\usepackage{graphicx}
\usepackage{tabularx}
\usepackage{kpfonts} % for nice fonts
\usepackage{microtype}
\usepackage{booktabs} % for nice tables
\usepackage{bm} % for bold math
\usepackage{listings} % for inserting code
\usepackage{verbatim} % useful for program listings
\usepackage{color}
%\usepackage[colorlinks=true]{hyperref}
% colors and hyperlink to bibliography and objects
\usepackage{hyperref}
\hypersetup{
colorlinks = true,
linkcolor = blue,
citecolor = blue,
runcolor = blue,
urlcolor = blue
}
\usepackage{url}
% Start TOC for appendix https://tex.stackexchange.com/questions/419249/table-of-contents-only-for-the-appendix
\usepackage[toc,page,header]{appendix}
\usepackage[]{minitoc}
\noptcrule %<<<<<<<<<<<<<<<<<<< HERE
% Make the "Part I" text invisible
\renewcommand \thepart{}
\renewcommand \partname{}
% end
\usepackage{lscape}
\usepackage{subcaption}
\usepackage{mwe}
% use for hypertext
%\usepackage[colorinlistoftodos]{todonotes}
\usepackage{natbib}
\usepackage{multibib}
\newcites{sec}{References in the Online Appendix}
\usepackage{setspace,caption}
\usepackage{lipsum}
\captionsetup{font=doublespacing}% Double-spaced float captions
\usepackage{amsthm}
\theoremstyle{plain}
\newtheorem{theorem}{Theorem}
\newtheorem{prop}{Proposition}
\newtheorem{result}[theorem]{Result}
\newtheorem{acknowledgement}{Acknowledgement}
\newtheorem{algorithm}[theorem]{Algorithm}
\newtheorem{axiom}[theorem]{Axiom}
\newtheorem{case}[theorem]{Case}
\newtheorem{claim}[theorem]{Claim}
\newtheorem{conclusion}[theorem]{Conclusion}
\newtheorem{condition}[theorem]{Condition}
\newtheorem{conjecture}[theorem]{Conjecture}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{criterion}[theorem]{Criterion}
\newtheorem{definition}[theorem]{Definition}
\newtheorem{example}[theorem]{Example}
\newtheorem{exercise}[theorem]{Exercise}
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{notation}[theorem]{Notation}
\newtheorem{problem}[theorem]{Problem}
%\newtheorem*{proposition*}[theorem]{Proposition}
\newtheorem{remark}[theorem]{Remark}
\newtheorem{solution}[theorem]{Solution}
\newtheorem{summary}[theorem]{Summary}
\newtheorem*{proposition*}{Proposition}
\newtheorem*{proof*}{Proof}
\setcitestyle{round,citesep={,}}
\captionsetup[figure]{% use subfigure to confine changes to subcaptions
justification = RaggedRight, % Or justified
format=hang}
\usepackage{subcaption}
\raggedbottom
%% table size
\usepackage{adjustbox}
%%
\usepackage{pdflscape}
\usepackage{dsfont}
\begin{document}
\title{\Huge{NGO Activism:\\ Exposure vs. Influence}\thanks{This study received funding from a Sciences Po–McCourt Institute grant. We are grateful to Martino Banchio, Jacob Conway, Filippo Lancieri, Carlo Schwarz, Catherine Thomas, and participants at the ESSEC Research Day for their constructive comments. The views expressed herein are those of the authors and do not necessarily reflect the views or policies of the Board of Governors of the Federal Reserve System. All errors and omissions remain our own.}}
\author{\textbf{Michele Fioretti}\thanks{Bocconi, Department of Economics. Corresponding author. e-mail: \href{fioretti.m@unibocconi.it}{fioretti.m@unibocconi.it}} \and \textbf{Victor Saint-Jean}\thanks{ESSEC Business School, Department of Finance. e-mail: \href{saintjean@essec.edu}{saintjean@essec.edu}} \and \textbf{Simon C. Smith}\thanks{Federal Reserve Board. e-mail: \href{simon.c.smith@frb.gov}{simon.c.smith@frb.gov}}}
\date{September 30, 2025}
\maketitle
%\thispagestyle{empty} % Remove page number from title page
\begin{abstract}\singlespacing
We examine how NGOs time campaigns to influence corporate decisions at Annual General Meetings (AGMs). Using data on 2,500 U.S. campaigns, we find that NGOs often campaign on the AGM date, which increases media exposure and future shareholder proposals but does not affect current AGM votes or consumers. As NGOs gain reputational capital, they shift to earlier pre-AGM campaigns to support related shareholder proposals and directly improve corporate behavior. The evidence underscores a strategic trade-off between visibility and influence, showing how NGO tactics evolve with NGOs’ credibility and the changing expectations of socially conscious stakeholders about NGO activities.\\
%We analyze the timing of NGO campaigns to shed light on NGOs' strategies and how they evolve over time. Data from 2,500 campaigns show that NGOs are six times more likely to launch campaigns on their target's Annual General Meeting (AGM) date. Although this strategy increases media exposure and stakeholder scrutiny, resulting in consumer backlash and related shareholder proposals at the following AGM of the targeted firm, it has no impact on current AGM votes. As NGOs build reputational capital, they adjust their timing to influence AGM votes, revealing the trade-offs they face to drive corporate change.\\
\noindent \textit{JEL classifications:} L21, L31 \\
\textit{Keywords:} Non-Governmental Organizations, NGO campaigns, social and environmental responsibility, ESG, firm dynamics, non-profit objectives
\end{abstract}
\newpage\onehalfspacing
\section{Introduction}
Non-governmental organizations (NGOs) are increasingly prominent actors in shaping firms’ environmental and social policies \citep{baron2001private, egorov2017private, yaziji2009ngos}. Yet, their strategic choices remain poorly understood. Much of the literature treats NGOs as either principled advocates motivated by social good \citep{besley2005competition, benabou2006incentives, daubanes2019rise} or as dysfunctional entities prone to corporate capture and strategic manipulation \citep{grossman1994protection, bertrand2020tax, akey2021hacking}. We argue and demonstrate empirically that NGOs adapt their strategies in response to incentives, shaping which stakeholders they engage and ultimately affecting the social and environmental performance of targeted firms.
To study how NGOs respond to incentives and their implications for firms, we focus on a setting where firms are especially exposed to outside pressure: the Annual General Meeting (AGM). AGMs are predictable focal points at which executives present results, shareholders vote on proposals, and media attention peaks \citep{fioretti2024shared}. Figure~\ref{fig:intro_fig} shows that NGO campaigns cluster heavily around AGM weeks: incidence is four to six times higher than in non-AGM weeks (Panel a), and media coverage spikes simultaneously (Panel b). These striking patterns raise a central question: do NGOs target AGMs mainly to maximize \textit{exposure}, or do they also use campaigns to exert \textit{influence} over corporate outcomes?
\begin{figure}[!htbp]
\centering
\caption{Timing of NGO campaigns and media attention around AGMs \label{fig:intro_fig}}
\begin{subfigure}{0.49 \linewidth}
\centering
\includegraphics[width=\linewidth]{Figures/sum_stat_ngo.pdf}
\caption{Campaigns peak during AGMs \label{fig:NGO_time}}
\end{subfigure}
\begin{subfigure}{0.49 \linewidth}
\centering
\includegraphics[width=\linewidth]{Figures/sum_stat_news.pdf}
\caption{Media coverage rises during AGMs \label{fig:AGM_coverage}}
\end{subfigure}
\begin{minipage}{1 \textwidth} % choose width suitably
{\footnotesize Note: Panel (a) displays the number of NGO campaigns launched relative to the date of the Annual General Meeting of the targeted firm (week 0), based on 1,544 NGO campaigns. Panel (b) compares the weekly average count of news during the AGM week (gray bars) and outside of the AGM week (white bars), based on weekly observations for 5,211 US firms. The NGO data come from Sigwatch; the news data are from Ravenpack. \par}
\end{minipage}
\end{figure}
Examining campaign timing reveals that younger NGOs concentrate campaigns on the AGM date, where visibility is maximized but most shareholder votes are already casted \citep{gent2015reputation, bebchuk2020dancing}. These campaigns cannot change immediate voting outcomes but are found to mobilize other stakeholders, such as future shareholder proponents. For instance, as the NGO \textit{Public Eye} put it in its 2025 campaign timed on the AGM date of \textit{Inditex}: ``As shareholders, you have the power to put Inditex on a sustainable course'' \citep{cleanclothes2025inditexagm}. Indeed, we find little evidence that other stakeholders are moved by these campaigns. For example, we find no evidence that sales are differently affected based on campaign timing relative to the AGM date, indicating that consumers may not be the main audience of NGOs.
While they do not sway votes or sales, AGM-timed campaigns significantly increase the visibility of NGOs themselves. Using Google search data, we find that such campaigns generate sharp increases in search intensity for the campaigning NGO, particularly when it is relatively young. These visibility gains persist for several months and are substantially larger than those observed for campaigns launched at other times. This suggests that AGM-day actions, while limited in shifting immediate corporate outcomes, are instrumental in boosting public recognition of NGOs.
As NGOs accumulate visibility through AGM-date campaigns, their strategy changes. Because most shareholder votes are cast well before the meeting, campaigns on the day itself rarely sway immediate outcomes. By contrast, campaigns in the months preceding the AGM raise the likelihood of proposal success. Campaigns one to three months before the meeting increase support by 8\%, while earlier ones often induce settlements that improve firms’ environmental and social performance.
Consistent with this shift from exposure to influence, we find that NGOs progressively move their campaigns earlier in time as they gain AGM exposure. First-time campaigns are concentrated on the AGM date, where visibility is highest, while subsequent efforts are launched progressively earlier, when they can still shape shareholder decisions. This pattern is especially pronounced among NGOs with high search intensity, which may not drive substantial visibility benefits from AGM-day campaigns.
Repeated targeting of AGMs also plays a distinct role. By returning to these focal moments, NGOs build a “reputation for toughness”—demonstrating their ability to amplify negative media coverage when firms are most exposed—which can enhance the effectiveness of subsequent campaigns. Yet, once reputation is established, further AGM-day campaigns yield diminishing benefits, generating smaller increases in NGO search intensity, lower success rates for related shareholder proposals, and less negative media coverage on the target firm. This pattern reflects NGOs’ changing incentives because of evolving stakeholder expectations. As NGOs become recognized and credible actors \citep{bar2008seller, baur2011moral, mitchell2017reputations}, their actions are interpreted differently: what initially signals resolve later signals routine. Established NGOs are therefore expected to engage in more constructive tactics, such as supporting shareholder proposals, rather than relying on visible confrontation.
Taken together, this lifecycle of tactics—initially focused on visibility, later on direct influence—illustrates how NGOs adaptively balance exposure, reputation, and pressure on firms, with implications for their social and environmental impact. In this sense, AGM-day campaigns act as costly signals of commitment that help NGOs build credibility \citep{kreps1982reputation, milgrom1982limit}, but the meaning and effectiveness of those signals evolve as stakeholders update their expectations about each NGO’s toughness and legitimacy.
This dynamic lifecycle highlights NGOs’ role as coordinators that align stakeholder actions and channel them into constructive corporate change, as their strategies shift over time to balance visibility and influence \citep{briscoe2016social}, with campaign tactics evolving dynamically across the life cycle to shape both corporate responses and stakeholder engagement \citep{cabral2010dynamics}.
By documenting this strategic adaptation, we provide new conditions under which firms internalize their externalities \citep{fioretti2022caring,allcott2022economic}. More broadly, our work offers empirical support for theories asserting that stakeholders contribute resources only when they expect to capture part of the value created, underscoring the interplay between stakeholder involvement, reputation, and corporate change \citep{Porter2011, Kaplan2020, McGahanPongeluppe2023}.
Prior studies highlight the role of shareholders influencing firms via voice through voting or exit via divestments \citep[e.g.,][]{cunat2016say,broccardo2020exit, bonnefon2022moral, hart2022private, saint2024exit}, and of consumers through willingness to pay \citep{gneezy2010shared}. We instead show that NGOs act as coordination mechanisms, enabling diverse stakeholders to collectively discipline firms \citep{koenig2019social, chuah2024shareholder, mazet2023some, mcgahan2023new}. In doing so, NGOs expand the set of channels through which civil society shapes corporate conduct.
Finally, our methodology relates to the literature on the strategic timing of actions to exploit fluctuations in media attention. Prior work shows that politicians and governments time executive orders, military operations, and even disaster relief to coincide with or avoid competing news \citep{eisensee2007news, durante2018attack, djourelova2022media}, and that social movements strategically leverage focal dates to amplify visibility \citep{madestam2013do}. We show that NGOs adopt similar tactics in the corporate domain.
The paper is organized as follows. Section \ref{s:data} outlines the data, and Section \ref{s:empirical} provides background on AGM dates. Sections \ref{s:media} and \ref{s:influence} examine NGO campaigns as tools to gain media exposure or to influence firms through key stakeholders, respectively. Section \ref{s:reputation} brings these channels together by analyzing how NGO reputation enhances the success of shareholder proposals. Section \ref{s:discussion} discusses the results and concludes.
\begin{comment}
\newpage
\section{Old Introduction}
Over recent decades, non-governmental organizations (NGOs) have emerged as influential actors shaping corporate strategies and regulations outside traditional political channels \citep{baron2001private, egorov2017private}. Yet, NGOs' strategic incentives and actions are difficult to observe \citep{yaziji2009ngos}, making them appear either as principled advocates for social good \citep{daubanes2019rise} or dysfunctional entities prone to corporate capture \citep{bertrand2020tax}. This paper provides new insights into NGO activism by demonstrating how NGO's strategic incentives shape stakeholder mobilization and corporate responses.
Leveraging a unique dataset of 2,500 NGO campaigns targeting 523 publicly listed U.S. firms, we exploit the predictable scheduling of firms' Annual General Meetings (AGMs) to study how campaign \textit{timing} affects outcomes. NGOs often target AGM dates to capitalize on heightened media scrutiny, as executives present firm performance and shareholders vote on proposals affecting corporate policies and governance \citep{dyck2008corporate}. We find that less experienced NGOs prioritize AGM-day campaigns to maximize media \emph{exposure} and build their reputations \citep{bar2008seller}. Although these campaigns rarely influence shareholder votes, since most voting occurs beforehand \citep{bebchuk2020dancing,gent2015reputation}, \textcolor{red}{they successfully mobilize other stakeholders, prompting new shareholder proposals at future meetings.}
As NGOs mature and build reputational capital, their strategy evolves. Experienced NGOs increasingly shift their campaigns earlier, aiming to directly \emph{influence} shareholder votes and corporate behavior. Such pre-AGM campaigns significantly boost shareholder support for environmental and social proposals at the current AGM, resulting in measurable improvements in corporate policies. These findings reveal a dynamic lifecycle in NGO tactics: initial high-visibility campaigns build credibility, enabling subsequent constructive engagement with firms and stakeholders for direct impact \citep{kreps1982reputation,milgrom1982limit,baur2011moral, mitchell2017reputations}. Our results highlight NGOs' critical coordinating role in aligning stakeholder actions and driving corporate social responsibility.
% %%%%%%%%%%%%%
% Over recent decades, non-governmental organizations (NGOs) have emerged as influential actors shaping regulations and corporate strategies outside traditional political channels \citep{baron2001private,egorov2017private}. However, their actions are difficult to observe and evidence on their strategic incentives remains limited \citep{yaziji2009ngos}. As a result, NGOs are often viewed either as principled actors promoting social good \citep{daubanes2019rise} or as dysfunctional entities vulnerable to corporate capture \citep{bertrand2020tax}. This paper provides new insights into NGO activism by analyzing how NGOs coordinate stakeholders and shape corporate social value. %through the timing of their campaigns. %In particular, we focus on campaign timing—one of the few observable aspects of NGO activity—and show that it reflects a key strategic trade-off NGOs face between gaining exposure versus exerting influence.
% Employing over ten years of campaign data against large U.S. corporations, we find that NGOs are significantly more likely to launch campaigns against a firm on its Annual General Meeting (AGM) date. AGMs attract heightened media coverage as managers and key shareholders convene to discuss the firm’s performance and outlook \citep{dyck2008corporate}. NGOs exploit this peak in attention by timing campaigns during the AGM period, increasing their visibility and signaling their commitment to the issue, thereby accumulating reputational capital. While timing might seem like a minor feature of activism, our analysis reveals that it is pivotal: when NGOs launch a campaign can determine which stakeholders are mobilized and how firms respond.
% In this paper, we find that NGOs’ campaign strategies evolve over time in response to this exposure-vs-influence trade-off. Early-stage NGOs tend to prioritize maximum exposure by launching campaigns on the target firm’s AGM date, when executives face shareholder scrutiny and media attention. This timing yields widespread publicity – triggering consumer backlash and prompting subsequent shareholder proposals on the campaign’s issues – but it rarely produces immediate changes in the targeted firm’s environmental or social performance.
% % We focus on the timing of NGO campaigns—one of the few observable aspects of NGO actions—to reveal a key trade-off NGOs face: gaining media exposure versus exerting direct influence on firm behavior. Employing over ten years of campaign data against large U.S. corporations, we find that NGOs are significantly more likely to launch campaigns against a firm on its Annual General Meeting (AGM) date. AGMs attract heightened media coverage as managers and key shareholders convene to discuss the firm’s performance and outlook \citep{dyck2008corporate}. NGOs exploit this peak in attention by timing campaigns during the AGM period, increasing their visibility and signaling their commitment to the issue, thereby accumulating reputational capital \citep{bar2008seller,gent2015reputation,mitchell2017reputations}. However, because most shareholders have already voted by proxy ahead of the AGM, a campaign launched on the AGM date is unlikely to sway the current vote \citep{bebchuk2020dancing}. Consistent with this, we find that while high-visibility AGM-date campaigns do not materially affect contemporaneous voting outcomes, they do trigger other stakeholder responses: consumer backlash and related shareholder proposals at the target’s next AGM are both more likely after an NGO campaigns on the AGM day, compared to campaigns launched at other times.
% Over time, as NGOs build their reputation, their strategic focus shifts from exposure to direct influence \citep{bar2008seller}. Early-stage NGOs prioritize media exposure by targeting AGM dates, effectively using these high-profile campaigns to signal toughness and build a reputation for advocacy \citep{kreps1982reputation,milgrom1982limit}. We find that this strategy yields substantial publicity and future credibility but comes at the cost of limited immediate impact on firm behavior. As an NGO matures and accumulates reputational capital, it increasingly launches campaigns in the months before the AGM, aiming to influence shareholder voting outcomes directly. Campaigns timed one to two months prior to the AGM prove more effective at changing firms’ policies: they significantly increase shareholder support for environmental and social proposals at the current AGM and lead to measurable improvements in corporate behavior. Consistent with reputation-building models, we show that the success of a shareholder proposal at an AGM is positively associated with the number of past AGM-date campaigns by the supporting NGO \citep{bar2008seller}. However, this effect exhibits diminishing returns as the NGO conducts more such campaigns, suggesting that after a certain point additional AGM-day campaigns contribute less to perceived credibility. More experienced NGOs thus shift toward earlier campaigns, engaging constructively with stakeholders ahead of the AGM to maximize direct impact \citep{baur2011moral,mitchell2017reputations}.
Our identification strategy takes advantage of the timing of firms’ AGMs, which is exogenously determined by regulation and remains stable year to year \citep{fioretti2024shared}. This predictable corporate calendar allows us to separate an NGO’s timing decision from confounding events: while current news or events may influence \emph{whether} an NGO decides to launch a campaign, they cannot change \emph{when} the firm’s AGM occurs. Importantly for the interpretation of our results, campaign timing is not systematically related to target firm characteristics such as size or ESG ratings. This alleviates selection concerns for instance, it rules out that only less experienced NGOs target certain types of firms on AGM dates, supporting a causal interpretation of the trade-off between campaigning on AGM dates for exposure versus earlier for influence.
Our findings reveal a novel trade-off between gaining exposure and exerting direct influence as a critical driver of the societal consequences of NGO activism \citep{briscoe2016social}. While much existing literature characterizes activists primarily as prosocial or morally driven actors \citep{besley2005competition, benabou2006incentives, baron2010morally}, or conversely, as entities vulnerable to corporate capture \citep[e.g.,][]{grossman1994protection, bertrand2014whom, bertrand2020tax, akey2021hacking}, our paper emphasizes NGOs’ strategic incentives. NGOs are not passive actors: their campaign strategies dynamically evolve over their life cycle, shaping both corporate responses and stakeholder engagement \citep{cabral2010dynamics}. By highlighting this strategic adaptation, we offer new conditions for firms to internalize their externalities.
Further, we contribute to the literature examining firms’ societal impacts \citep{flammer2018competing, fioretti2022caring, fangwa2024governance}. While prior studies emphasize shareholders influencing firms via voice through voting \citep{cunat2016say} or exit via divestments \citep{broccardo2020exit, bonnefon2022moral, hart2022private, saint2024exit}, and consumers through willingness to pay \citep{gneezy2010shared}, we highlight NGOs' role in coordinating these stakeholders \citep{koenig2019social, chuah2024shareholder, mazet2023some, mcgahan2023new}. NGOs serve as coordination mechanisms, enabling stakeholders with varied preferences to collectively demand corporate change.
%Lastly, by focusing empirically on NGO strategies over their lifecycle, we integrate stakeholder and signaling theories \citep{Barney2018, mcgahan2023new} to demonstrate how NGOs leverage reputation-building to shift from attention-seeking campaigns to more influential engagements.
Our work thus provides empirical support for theories asserting stakeholders contribute resources primarily when they anticipate capturing some of the generated value, underscoring the interplay between stakeholders' involvement, their reputation, and positive corporate change \citep{Porter2011, Kaplan2020, McGahanPongeluppe2023}.
The paper is organized as follows. Section \ref{s:data} outlines the data and provides summary statistics. Section \ref{s:empirical} describes the empirical strategy. Sections \ref{s:media} and \ref{s:influence} media exposure and direct influence at the AGM, respectively, as potential drivers of campaign impact. Section \ref{s:reputation} explores the role of NGO reputation in enhancing shareholder proposal success. Section \ref{s:discussion} discusses the results and concludes.
% %%%\newpage
% This paper provides new insights into NGOs as coordinators of various stakeholders (e.g., shareholders and consumers), empirically highlighting trade-offs that arise from stakeholder involvement \citep{Porter2011,Kaplan2020}. We shed light on the incentives NGOs face and how they can help align a firm’s actions with the interests of its stakeholders \citep{McGahanPongeluppe2023}. In this direction, our results also suggest that stakeholders can gain privately (through reputation) by engaging with a firm, lending empirical support to theories positing that stakeholders would not contribute resources to an organization without the potential to claim at least some of the value created (cf. \citealp{Barney2018,mcgahan2023new}).
% %Finally, our findings contribute to a growing literature on the impact of activist campaigns on corporate outcomes.
% Finally, we also contribute to the growing body of literature on firms’ objectives and societal impact \citep{fioretti2022caring}. Firms can help address global challenges like pollution and inequality by internalizing their externalities, but the mechanisms driving such changes outside of formal regulation remain unclear. Prior studies have typically exploited the exogenous occurrence of NGO or consumer campaigns to assess changes in firm performance and prosocial behavior (e.g., \citealp{koenig2019social,mazet2023some,conway2024consuming}). We turn the focus to the NGOs themselves, examining how their strategies evolve over their life cycle. Our results underscore the role of NGOs as stakeholder coordinators and highlight how reputation dynamics drive a strategic shift from gaining attention to exerting influence. In doing so, we bridge insights from stakeholder theory and signaling theory to better understand the evolving tactics of activist organizations striving to drive corporate social change.
% The paper is organized as follows. Section \ref{s:data} outlines the data and provides summary statistics. Section \ref{s:empirical} describes the empirical strategy. Sections \ref{s:media} and \ref{s:influence} media exposure and direct influence at the AGM, respectively, as potential drivers of campaign impact. Section \ref{s:reputation} explores the role of NGO reputation in enhancing shareholder proposal success. Section \ref{s:discussion} discusses the results and concludes.
% \section{Introduction (longer)}
% Over recent decades, non-governmental organizations (NGOs) have emerged as influential actors shaping regulations and corporate strategies outside traditional political channels \citep{baron2001private,egorov2017private}. However, their actions are difficult to observe, and evidence on their strategic incentives remains limited \citep{yaziji2009ngos}. As a result, NGOs are often viewed either as principled actors promoting social good or as dysfunctional entities vulnerable to corporate capture \citep{daubanes2019rise,bertrand2020tax}. This paper provides new insights into NGO activism by analyzing how NGOs coordinate stakeholders and shape corporate social value through the timing of their campaigns. In particular, we focus on campaign timing—one of the few observable aspects of NGO activity—and show that it reflects a key strategic trade-off NGOs face between gaining exposure versus exerting influence.
% Using a unique dataset of 2,500 campaigns by NGOs targeting 523 U.S. publicly listed firms, we exploit the predictable scheduling of firms’ Annual General Meetings to study how campaign timing affects outcomes. While timing might seem like a minor feature of activism, our analysis reveals that it is pivotal: when NGOs launch a campaign can determine which stakeholders are mobilized and how firms respond.
% In this paper, we find that NGOs’ campaign strategies evolve over time in response to this exposure-vs-influence trade-off. Early-stage NGOs tend to prioritize maximum exposure by launching campaigns on the target firm’s AGM date, when executives face shareholder scrutiny and media attention. This timing yields widespread publicity – triggering consumer backlash and prompting subsequent shareholder proposals on the campaign’s issues – but it rarely produces immediate changes in the targeted firm’s environmental or social performance.
% As NGOs gain experience and build reputational capital, they shift toward launching campaigns before the AGM. Such early campaigns sacrifice some media spotlight but directly influence corporate decision-making – significantly increasing shareholder support for related proposals at the current AGM and leading to measurable improvements in corporate behavior. In short, NGOs follow a dynamic life-cycle in their tactics: newer NGOs initially pursue exposure at the expense of immediate impact, whereas mature NGOs time their campaigns to maximize direct influence on firm policies.
% Consistent with reputation-building models, we show that an NGO’s history of high-visibility campaigns enhances its credibility in the eyes of stakeholders. Shareholder proposals are more likely to succeed when accompanied by an NGO campaign on the same issue – especially if the NGO has a track record of targeting AGM days. However, this reputational boost exhibits diminishing returns: the positive effect on proposal success fades after an NGO has staged multiple AGM-day campaigns. In contrast, an NGO’s past campaigns that were not on AGM days have no significant influence on proposal outcomes.
% These patterns suggest that NGOs strategically use attention-grabbing AGM-day campaigns early on to signal “toughness” and build name recognition, even though such campaigns have limited direct effect on that year’s votes. As a result, more mature NGOs can leverage their established reputation to engage in earlier, pre-AGM campaigns focused on constructive influence, aligning with theories of cooperative engagement between stakeholders and firms \citep{McGahanPongeluppe2023}. Through this shift, NGOs increasingly work with stakeholders (e.g. long-term investors) to press for change, rather than just shining a spotlight on corporate misbehavior.
% To identify these effects, we leverage the exogenous timing of corporate AGMs. In the U.S., AGM dates are fixed by regulation (SEC rules) and remain stable for each firm year-to-year, providing a natural experiment for NGO visibility opportunities \citep{fioretti2024shared}. This predictable calendar allows us to isolate how NGOs respond to scheduled moments of public scrutiny, holding firm-specific factors constant. Our main variable of interest is the relative timing of an NGO’s campaign (months before or on the AGM). We show that more experienced NGOs systematically campaign earlier – trading off media visibility for greater impact – while less experienced NGOs cluster their campaigns on the AGM date.
% Importantly, we find that key characteristics of the targeted firms (such as size and ESG ratings) do not differ systematically by NGO experience or campaign timing. This alleviates concerns about selection bias: for example, it rules out the possibility that only inexperienced NGOs target certain “easier” or smaller firms. We further control for unobserved heterogeneity by comparing outcomes across campaigns within the same firm-year and issue, accounting for time-varying factors like organizational learning or issue salience. Finally, while NGOs may also engage firms through partnerships or behind-the-scenes lobbying, such interactions are often voluntarily disclosed and subject to selection biases. In contrast, public campaign timing is an observable strategic choice, making it a suitable focus for empirical analysis of NGO behavior.
% Overall, our results highlight a previously underexplored trade-off between maximizing exposure and achieving direct impact as a driver of NGO campaign outcomes. This finding contributes to the literature on activism and private governance by demonstrating that NGOs are not passive or monolithic actors; rather, their tactics evolve over time to balance visibility and influence. Whereas much of the existing literature assumes activists primarily seek immediate policy changes from their targets, our evidence shows that building reputational capital through media exposure can be a strategic end in itself, enabling greater influence down the line.
% At the same time, some studies have noted that NGOs can be co-opted or “captured” by firms’ influence (e.g., through partnerships or donations). Our analysis offers a more nuanced perspective: by understanding NGOs’ incentives and constraints, we see that even when early campaigns appear ineffectual, they may serve a long-run purpose in the NGO’s strategy. In doing so, this work bridges the gap between viewing NGOs as purely principled altruists versus dismissing them as ineffective or compromised.
% We also contribute to the growing body of literature on firms’ objectives and societal impact \citep{fioretti2022caring}. Firms can help address global challenges like pollution and inequality by internalizing their externalities, but the mechanisms driving such changes outside of formal regulation remain unclear. Prior research emphasizes that shareholders influence firms through voice (voting at AGMs) or exit (divestment), and that consumers exert pressure through their purchase decisions. Our findings highlight the pivotal role of NGOs as coordinators of these stakeholder pressures. By orchestrating campaigns that galvanize consumers and shareholders in tandem, NGOs empower dispersed stakeholders with unaligned preferences to demand change from corporations. In essence, NGO activism serves as a catalyst in the corporate social value creation process, helping to channel public and investor scrutiny into tangible action by firms. Understanding this dynamic role of NGOs is therefore essential to explain when and why firms voluntarily adopt socially responsible changes in the absence of explicit public policy mandates.
% %%%
% \newpage
% \section{Old Intro}
% Over recent decades, non-governmental organizations (NGOs) have emerged as influential actors shaping regulations and corporate strategies outside traditional political channels \citep{baron2001private,egorov2017private}. However, their actions are difficult to observe, and evidence on their strategic incentives remains limited \citep[][]{yaziji2009ngos}. As a result, NGOs are often viewed either as principled actors promoting social good \citep{daubanes2019rise} or as dysfunctional entities vulnerable to corporate capture \citep{bertrand2020tax}.
% \sloppy This paper provides new insights into how NGO campaigns coordinate stakeholder actions and ultimately shape the creation of corporate social value. We focus on campaign timing—one of the few observable aspects of NGO activity, which is far more opaque than that of for-profit firms.\footnote{A parallel literature uses administrative data to study charitable donations to NGOs \citep[e.g.,][]{cage2023far,cage2025charitable}, offering valuable insights into their funding base. Our focus is instead on NGOs’ strategic choices—an important aspect of their behavior that remains relatively underexplored, especially outside developing-countries contexts \citep[e.g.,][]{deserranno2025impact}.} While timing may seem like a minor feature, our approach reveals a key trade-off NGOs face: gaining media exposure, which can be leveraged in future campaigns, versus exerting direct influence on firm behavior. We identify the main determinants of how NGOs resolve this trade-off and show that their choices shape which stakeholders are mobilized and how firms ultimately respond.
% % Over recent decades, non-governmental organizations (NGOs) have emerged as influential actors shaping regulations and corporate strategies outside traditional political channels \citep{baron2001private,egorov2017private}. However, their actions remain difficult to observe, and evidence on their strategic incentives is limited \citep[][]{yaziji2009ngos}. Consequently, NGOs are often viewed either as principled actors promoting social good \citep{daubanes2019rise} or dysfunctional entities susceptible to corporate capture \citep{bertrand2020tax}. %This paper introduces a novel framework for understanding NGOs’ strategic choices and their implications for corporations and stakeholders.
% % Consider an NGO that discovers corporate wrongdoing. It must decide whether to campaign immediately or wait for a more strategic moment when the firm is particularly vulnerable. Immediate action may foster dialogue, but delayed campaigns can amplify impact through increased stakeholder awareness and stronger signals of opposition.
% Our findings reveal a “life cycle” in NGO strategies \citep{wernerfelt1988umbrella,cabral2000stretching}. Early-stage NGOs prioritize \textit{exposure} by campaigning on a target firm's Annual General Meeting (AGM) date, where executives face shareholder scrutiny and heightened media attention, triggering consumer backlash and prompting future shareholder proposals. However, these high-visibility campaigns rarely improve firms’ environmental and social outcomes directly. Over time, NGOs shift to campaigning before a target’s AGM date to exert direct \textit{influence}, increasing the success rate of current shareholder proposals and leading to measurable improvements in corporate behavior. %Understanding NGO's strategies matters for policy, as it illustrates how an external actor influences the creation of corporate social values by coordinating actions by other stakeholders.
% Consistent with reputation-building models, shareholder proposals are more likely to succeed when accompanied by NGO campaigns on the same issue—especially when the NGO has a history of targeting AGM dates \citep{bar2008seller}. However, this effect diminishes with the number of past AGM-date campaigns, suggesting decreasing returns. In contrast, past non-AGM campaigns have no influence on proposal success. NGOs thus strategically use AGM-date campaigns to signal toughness and build their reputation, initially benefiting from substantial media attention but paying the cost of limited direct influence on AGM voting \citep{kreps1982reputation,milgrom1982limit}. As a result, more mature NGOs shift toward earlier, pre-AGM campaigns, aiming to build credibility through constructive stakeholder engagement \citep{baur2011moral,mitchell2017reputations}.
% % Therefore, AGM exposure enhances NGO reputation. Supporting this view, we find shareholder proposals on a given issue are more likely to succeed when accompanied by NGO campaigns in the same year, particularly if the NGO previously campaigned on AGM dates. However, such returns diminish with the number of past AGM-date campaigns, while past non-AGM campaigns have no impact. Thus, NGOs likely use AGM campaigns as a strategic signal to build their reputation and credibility, consistent with models of reputation-building \citep{bar2008seller}. Specifically, initial AGM-date campaigns generate substantial negative attention, bolstering an NGO's "reputation-for-toughness," but later campaigns offer minimal additional impact \citep{kreps1982reputation,milgrom1982limit}. Consequently, mature NGOs shift their strategy toward earlier campaigns, not just due to diminishing returns from visibility but also to establish multidimensional credibility through constructive stakeholder engagement \citep{baur2011moral,mitchell2017reputations}.
% % Our interpretation is that AGM exposure can enhance NGO reputation. Supporting this view, we find that the success of a shareholder proposal on a given cause increases when an NGO campaigns on the same issue and year, with past AGM-date campaigns by the NGO amplifying this effect. However, the returns to AGM-date campaigns diminish as their frequency rises. In contrast, past NGO campaigns on non-AGM dates have no impact.
% % We view AGM-date campaigns as a way for NGOs to signal their type to stakeholders and build their ``reputation,'' as evidenced by the diminishing returns of AGM-targeting over time, which is consistent with reputation models \citep{bar2008seller}.\footnote{This practice also generates additional negative news to a target firm only for an NGO's first few campaigns. Later campaigns inflict no additional damage. Thus, the NGO may be willing to incur a cost to build its ``reputation-for-toughness,'' which it can then leverage in future campaigns \citep{kreps1982reputation,milgrom1982limit}.} Hence, more mature NGOs shift to campaign before AGMs. Furthermore, NGO reputation is likely multidimensional, shaped both by a sustained track record of activism and by a demonstrated commitment to a cause. As such, NGOs may transition to early campaigning not only due to diminishing returns on visibility but also to establish credibility through constructive engagement with key stakeholders \citep{baur2011moral,mitchell2017reputations}.
% Our identification strategy leverages the timing of firms' AGMs, which is exogenously fixed by SEC regulations and stable within firms over time. This enables us to analyze how NGOs respond to predictable opportunities for visibility, holding firm strategies constant. Our main endogenous variable, therefore, is NGO campaign timing, which we explain by reference to each NGO’s prior campaign experience: more experienced NGOs campaign earlier, trading visibility for impact. While unobservable factors might also influence timing decisions, we verify that key characteristics of the target firms—such as size and ESG scores—do not systematically correlate with NGO experience or campaign timing. This reduces concerns about selection bias, for instance, ruling out that less experienced NGOs strategically target smaller or less prosocial firms for easier gains.
% % Our identification strategy leverages the timing of firms' AGMs, which are exogenously fixed by SEC regulations, enabling us to study how NGOs respond to predictable opportunities for visibility keeping firms' strategies fixed. Still, the timing of campaigns remains endogenous, as NGOs self-select when to act. We address this by showing that experience shapes timing decisions: more experienced NGOs campaign earlier, trading visibility for impact. While unobservable factors could influence timing, we verify that target characteristics such as size and ESG scores do not systematically vary with NGO experience, mitigating concerns about target selection. Thus, experience emerges as a key dimension explaining the timing of NGO campaigns, which we show to have important consequences for stakeholders and target firms.
% By comparing NGO campaign outcomes across firms and issues within years, our analysis accounts for factors that vary across years within NGOs, firms, and issues, such as organizational competence, learning, moral authority, resources, and the salience of specific issues. NGOs also engage firms through partnerships and other channels, but since such activities are voluntarily disclosed---and thus subject to selection biases \citep{burger2010promoting}---we rather focus on campaign timing, which must be public to be effective and, thus, observable to researchers.%Because our empirical approach compares the outcomes of NGO campaigns across multiple firms within years, our analysis accounts for factors all factors varying across years within NGOs, firms, and causes, such as an NGO’s organizational competence, learning, moral authority, resource endowment, and the salience of the issue. NGOs also engage with firms through various other channels, such as forming partnerships. However, data on partnerships and other NGO activities are typically disclosed voluntarily, making them susceptible to selection bias \citep{burger2010promoting}. Thus, we restrict our analysis to campaign timing, which, by definition, must be public—and thus observable to researchers—to be effective.
% Our findings open new avenues for research, as future work may empirically explore how NGOs choose between campaigning and other activist tactics or whether these different approaches act as complements or substitutes. The exogenous timing of AGMs provides a useful identification strategy for such analyses, as it holds the target firm's strategic choices constant, allowing researchers to isolate NGOs' choices and their incentives.
% Our results reveal a novel tradeoff between exposure and direct impact as a key driver of the societal consequences of NGO activism \citep{briscoe2016social}. Unlike our framework, much of the existing literature assumes that activists primarily pursue social objectives, either stemming from prosocial behavior \citep{besley2005competition,benabou2006incentives} or moral preferences \citep{baron2010morally}, without discussing their incentives. %Their level of commitment can also depend on how closely they align with the cause, as noted by \cite{besley2005competition}. As regulatory capture by firms has become an enduring feature of modern economies, NGO activism has emerged as a counterbalance to realign public and private interests \citep{glaeser2003rise,egorov2017private,daubanes2019rise}.
% At the same time, several studies highlight instances in which NGOs themselves are subject to firm capture \citep[e.g.,][]{grossman1994protection,grossman2001special,bertrand2014whom,drutman2015business,bertrand2021hall}. Firms may also strategically leverage their relationships with NGOs—through donations or partnerships—to advance political agendas \citep{bertrand2020tax}, or to restore consumer and investor trust following reputational damage \citep{akey2021hacking}.
% Our main contribution in this literature is to show that NGOs are not passive actors. Rather, their tactics evolve over time, shaping firms and their stakeholders' responses in different ways depending on the strategies they employ \citep[e.g.,][]{hendricks2009information,cabral2010dynamics}. Thus, understanding NGOs' incentives is essential to learn why firms internalize their externalities.
% We also contribute to the growing body of literature on firms' objectives and their societal impact \citep{flammer2018competing,fioretti2022caring,fangwa2024governance}. Firms can help address significant global challenges such as pollution and inequality by internalizing their externalities. However, the mechanisms by which firms do so without explicit public policies, like Pigouvian taxes, remain unclear \citep[e.g.,][]{hawn2016mind}. The literature emphasizes that shareholders can influence firms through either voicing concerns---especially through voting at AGMs \citep{cunat2016say}---or selling shares \citep{broccardo2020exit,bonnefon2022moral,hart2022private,saint2024exit}, whereas consumers exert influence through their willingness to pay \citep[e.g.,][]{gneezy2010shared}. However, other agents may play a pivotal role. Among them, we highlight that NGOs can serve as a coordination mechanism among stakeholders with potentially unaligned preferences and empower them to demand change \citep{koenig2019social,fioretti2024shared,chuah2024shareholder,mazet2023some,mcgahan2023new}.
% %A theoretical literature regards socially responsible acts as a hedge against different risks \citep{albuquerque2019corporate}, including reputation risk \citep{vanhamme2009too,elfenbein2012charity,kitzmueller2012economic,bakos2011cooperation}. Reputational concerns thus shape corporations' interactions with stakeholders, including activist shareholders \citep{dimson2015active} and corporate donations \citep{akey2021hacking}. Building on this body of work, we show how NGOs leverage reputational threats to drive socially responsible corporate behavior.
% The paper is organized as follows. Section \ref{s:data} outlines the data construction and provides summary statistics. The empirical approach is described in Section \ref{s:empirical}. Sections \ref{s:media} and \ref{s:influence} examine media exposure and influence at the current AGM as potential drivers of AGM actions. Section \ref{s:reputation} explores the role of NGO reputation in supporting related shareholder proposals. Section \ref{s:discussion} discusses the results and concludes.
% % play a crucial role in shaping corporate behavior through private politics, using tactics such as public shaming and boycotts. Their influence hinges on their reputation, which is shaped by past successes \citep{yaziji2009ngos}. This paper explores how NGOs manage the trade-off between maximizing influence on firms and enhancing their reputation.
% %%%%
% % \newpage
% % This shift is not driven by learning effects but by an NGO's accumulated exposure, which enhances its ability to shape shareholder votes. We further show that an NGO’s history of AGM-timed campaigns predicts its future influence on shareholder proposals, reinforcing the importance of reputation as a signaling mechanism \citep{bar2008seller}.
% % A major empirical challenge in studying NGOs is the scarcity of data. Unlike publicly listed firms, NGOs face fewer transparency requirements, making it difficult to track funding and strategic decisions \citep{maxwell2012economic,fioretti2022caring}. Existing studies often emphasize extreme NGO behaviors, from advancing social good \citep[e.g.,][]{daubanes2019rise} to corporate capture \citep[e.g.,][]{bertrand2021hall,bertrand2023investing}. Our study moves beyond these extremes by providing systematic evidence on how NGOs time their campaigns to balance exposure and influence.
% % Taken together, our results describe a novel trade-off between exposure and (immediate) impact as a key driver of the societal consequences of NGO activism. Unlike our paper, these activists are often assumed to pursue social objectives, which may stem from prosocial behavior as described by \cite{benabou2006incentives} or from moral preferences, according to \cite{baron2010morally}. Their level of commitment can also depend on how closely they align with the cause, as noted by \cite{besley2005competition}. \cite{baron2010morally} further illustrates that NGOs can emerge when socially motivated individuals coordinate efforts, drawing on frameworks outlined by \cite{dixit2004lawlessness} and \cite{tabellini2008scope}. When firms become targets of activist campaigns, firms may engage in ``self-regulation'' to preempt opposition and reduce potential harm. As firms' capture of regulators has become an unavoidable feature of our economies, NGO activism surges to rebalance public and private interests \citep{glaeser2003rise,egorov2017private,daubanes2019rise}.%
% % %Our results also relate to a literature studying how and why interest groups seek to influence government policy \citep{grossman1994protection,grossman2001special,bertrand2014whom,drutman2015business,bertrand2021hall}. a more recent discussion Baumgartner et al., 2009; Bertrand et al., 2014; Drutman, 2015). We differ from much of this prior work in our focus on influence via expert commentary, rather than through financial contributions and, much more importantly, in documenting one mechanism by which private interests may cloak biased advice by inducing its provision by a non-obviously aligned party. This finding has implications for how we model the process of governmental information acquisition (Austen-Smith, 1993; Laffont and Tirole, 1993), and is also of direct policy relevance for corporate disclosure requirements (Bebchuk and Jackson, 2013; Peng, 2016).
% % \footnote{A related literature highlights several instances in which NGOs can become subject to firm capture \citep[e.g.,][]{grossman1994protection,grossman2001special,bertrand2014whom,drutman2015business,bertrand2021hall}. For example, firms may strategically exploit their relationships with NGOs—through donations or partnerships—to further political agendas \citep{bertrand2020tax} or to regain consumer and investor trust after reputational damages \citep{akey2021hacking}.}
% % This paper also contributes to the growing body of literature on firms' objectives and their societal impact \citep{flammer2018competing,fioretti2022caring,george2023purpose,fangwa2024governance}. Firms can help address significant global challenges such as pollution and inequality by internalizing their externalities. However, the mechanisms by which firms do so without explicit public policies, like Pigouvian taxes, remain unclear, even when CSR investments increase market value \citep[e.g.,][]{hawn2016mind,hartzmark2019investors,hasan2022brand}.\footnote{A recent literature also suggests that government initiatives might even hinder private ones \citep{akey2024carbon}.} The literature emphasizes that shareholders can influence firms through either voicing concerns---especially through voting at AGMs \citep{cunat2016say,gine2017governance}---or selling shares \citep{broccardo2020exit,bonnefon2022moral,hart2022private,saint2024exit}, whereas consumers exert influence through their willingness to pay \citep[e.g.,][]{gneezy2010shared,elfenbein2010greater}. However, other agents may play a pivotal role. Among them, NGOs can serve as a coordination mechanism among stakeholders with potentially unaligned preferences and empower them to demand change \citep{fioretti2024shared,chuah2024shareholder,koenig2019social}.
% % Therefore, this short paper sheds new light on NGOs as coordinating various stakeholders such as shareholders and consumers, empirically highlighting a set of tradeoffs that arise from stakeholder involvement \citep{porter2011creating,kaplan2020social}. We provide new insights into the incentives NGOs face and how they can help align a firm's action with the interests of its stakeholders \citep{mcgahan2023there}. In this direction, the results also suggest that other stakeholders can gain privately (i.e., reputation) by engaging with a firm, bringing empirical evidence to theories suggesting that stakeholders would not contribute resources to an organization without the potential to claim at
% % least some of the value created \citep[cf.,][]{barney2018resource,mcgahan2023new}.
% % Finally, our findings contribute to a growing body of literature that exploits the exogenous occurrence of NGO campaigns to asses changes in the performances and prosocial outcomes of targeted firms \citep[e.g.,][]{koenig2022effects, mazet2023some, conway2024consuming}. Our study turns the table by focusing on the incentives that non-profit firms face when campaigning against for-profit firms.
% % %A theoretical literature regards socially responsible acts as a hedge against different risks \citep{albuquerque2019corporate}, including reputation risk \citep{vanhamme2009too,elfenbein2012charity,kitzmueller2012economic,bakos2011cooperation}. Reputational concerns thus shape corporations' interactions with stakeholders, including activist shareholders \citep{dimson2015active} and corporate donations \citep{akey2021hacking}. Building on this body of work, we show how NGOs leverage reputational threats to drive socially responsible corporate behavior.
% % The paper is organized as follows. Section \ref{s:data} outlines the data construction and provides summary statistics. The empirical approach is described in Section \ref{s:empirical}. Sections \ref{s:media} and \ref{s:influence} examine media exposure and influence at the current AGM as potential drivers of AGM actions. Section \ref{s:reputation} explores the role of NGO reputation in supporting related shareholder proposals. Section \ref{s:discussion} discusses the results and Section \ref{s:conclusions} concludes.
% % % play a crucial role in shaping corporate behavior through private politics, using tactics such as public shaming and boycotts. Their influence hinges on their reputation, which is shaped by past successes \citep{yaziji2009ngos}. This paper explores how NGOs manage the trade-off between maximizing influence on firms and enhancing their reputation.
% % %%%%
% % \newpage
% % NGOs have become influential actors outside traditional political channels, influencing corporate strategies through high-profile campaigns \citep{baron2001private,egorov2017private}. While research has documented the impact of such campaigns \citep[e.g.,][]{spar2003power,yaziji2009ngos}, little is known about NGOs’ strategic objectives and trade-offs over their life cycle. For example, just before McDonald’s 2006 AGM, Greenpeace’s McAmazon campaign led to a rapid shift in McDonald’s sourcing policies \citep{goldberg2007}. Similar campaigns have targeted Nike, Shell, Monsanto, and BNP Paribas. Understanding the strategic timing of such campaigns is key to unpacking NGO decision-making.
% % A major empirical challenge in studying NGOs is the scarcity of data. Unlike publicly listed firms, NGOs face fewer transparency requirements, making it difficult to track funding and strategic decisions \citep{maxwell2012economic,fioretti2022caring}. Existing studies often emphasize extreme NGO behaviors, from advancing social good \citep[e.g.,][]{daubanes2019rise} to corporate capture \citep[e.g.,][]{bertrand2021hall,bertrand2023investing}. Our study moves beyond these extremes by providing systematic evidence on how NGOs time their campaigns to balance exposure and influence.
% % Using campaign data spanning over a decade, we find that NGOs improve their reputation by launching campaigns around AGMs, leveraging heightened media attention to increase their visibility and credibility \citep{dyck2008corporate}. However, these late campaigns are unlikely to influence shareholder votes directly, as most votes are cast in advance \citep{bebchuk2020dancing}. Instead, we show that such campaigns increase the likelihood of related shareholder proposals in the following year and subsequent boycott campaigns.
% % Methodologically, we leverage the relative stability of AGM timing, regulated by the SEC, to isolate exogenous variation in media exposure. This approach accounts for time-varying confounders at the firm, industry, and campaign levels, allowing us to identify the trade-off NGOs face between exposure and immediate impact.
% % Our results align with firm life-cycle models, where reputation investment is key in early stages \citep[e.g.,][]{wernerfelt1988umbrella,cabral2000stretching}. For NGOs, delaying campaigns until AGMs is a costly investment in reputation, as it foregoes immediate impact in favor of long-term credibility. Over time, as NGOs build reputation, they shift toward earlier, more targeted influence strategies.
% % Taken together, our findings reveal a fundamental trade-off in NGO activism: reputation-building through exposure versus immediate corporate influence. Unlike previous studies that assume NGOs pursue purely social objectives \citep{benabou2006incentives,baron2010morally}, we demonstrate that NGOs make strategic decisions based on their life-cycle stage. These insights contribute to a broader understanding of how NGOs shape corporate behavior and societal change \citep{glaeser2003rise,egorov2017private,daubanes2019rise}.
% % \newpage
% % Little is known about the incentives that drive activists when they engage with firms on social or environmental issues. This paper provides empirical evidence on the incentive structure of NGOs, a central player in today’s economy. We show that these incentives shape how NGOs engage with firms, influencing both their strategies and the resulting impact on stakeholders and society.
% % Activists engage with firms in multiple ways. They may form partnerships, demonstrating that corporate objectives can align with broader social goals—for example, Conservation International helped Starbucks secure high-quality coffee from Mexican farmers whose livelihoods were threatened by multinational competition. When partnerships fail, NGOs often turn to public campaigns \citep{baur2011moral,chatain2019ngos}. Because visibility and publicity are central to their mission, these campaigns are among the few observable NGO activities that do not rely on voluntary disclosures, which are prone to selection bias \citep{burger2010promoting}. Our analysis starts from the premise that NGO campaigns are strategically timed. By leveraging campaign timing, we uncover NGOs’ underlying tactics and immediate motives, and we show how these actions affect firms and their stakeholders.
% % As many organizations, we view NGOs' needs and goals as changing over time. Although it widely thought that NGOs can achieve both (e.g., mediatic) exposure and influence on firms--as a matter of fact, they may achieve both at the same time when they can show a successful track record--we show that, in the immediate, they can either achieve one or the other.
% % \vspace{5cm}
% % NGOs are often regarded as guardians of the common good. They engage in negotiations with firms that violate social or environmental standards and may resort to public shaming if these efforts fail. Their influence depends on their reputation, which, in turn, is shaped by their past influence \citep{yaziji2009ngos}. However, naming and shaming a firm can generate significant public exposure while exerting little actual influence, as firms may retaliate against the NGO. Conversely, tactics that effectively influence management may be less visible to the broader public, limiting their impact on the NGO’s ``reputation.'' How do NGOs navigate this trade-off?
% % % We exploit an institutional set up that enables us to seperate the timing of influence from that of exposure: the Annual General Meeting of companies. It is exogenous, attracts media attention, and matters for corporate policies (cite classical papers on voting etc). However, much of decisions are taken months before the AGM, at times when media attention is lower. Thus, we exploit the timing of NGO campaigns related to AGM to analyze their exposure vs influence trade-offs. \\
% % % Over recent decades, non-governmental organizations (NGOs) have emerged as influential social actors, shaping regulations and corporate strategies outside traditional political channels. Through tactics of ``private politics,'' such as boycotts and shaming campaigns, NGOs have effectively pressured companies to reform their practices \citep{baron2001private,egorov2017private}. Many studies examine how campaigns by both large and small NGOs impact %and delegitimate // it is unclear what deligitimate means here
% % % targeted firms, often showing enduring effects \citep[e.g.,][]{spar2003power,yaziji2009ngos}. For instance, just weeks before McDonald's 2006 Annual General Meeting (AGM), Greenpeace launched a campaign branding the company `McAmazon` for alleged links to Amazon deforestation, prompting McDonald's to adopt a new sustainable sourcing policy within days \citep{goldberg2007}. Similar campaigns have targeted companies across various sectors, including Nike, Shell, Monsanto, and BNP Paribas. Despite the significance of these campaigns in shaping NGO strategies and influencing corporate behavior, we know relatively little about NGOs’ objectives, trade-offs, and how these evolve over the course of an NGO’s life.
% % Researchers addressing these questions face several challenges, particularly due to the scarcity of data on NGOs. For instance, donation data for non-political nonprofits is difficult to access because of donor privacy concerns.\footnote{Data on donations to political parties and charities are more readily available due to disclosure rules and surveys \citep[e.g.,][]{cage2015improving,perez2017partisan,bouton2022public,guillot2023charitable}. This paper differs from this literature in that it does not focus on the connection between NGO and politics, but rather on the incentives that NGOs face when deciding to campaign against a firm.} NGOs are also subject to much less stringent transparency requirements than other firms such as publicly listed corporations, which complicates the tracking of their fund usage. Moreover, the absence of a quantitative framework to systematically analyze NGO objectives presents another significant challenge, as NGOs' diverse missions cannot easily be reduced to profit maximization \citep{maxwell2012economic, fioretti2022caring}. To obviate to this problem, the literature emphasizes extreme behaviors of NGOs, with some contributing to the greater good \citep[e.g.,][]{daubanes2019rise}, while others are captured by corporate interests \citep[e.g.,][]{bertrand2021hall,bertrand2023investing}.
% % % Researchers addressing this questions face several challenges, particularly due to the scarcity of data on NGOs. Donation data for non-political nonprofits is difficult to access because of donor privacy concerns,\footnote{Data exist for donations to political parties and charities thanks to disclosure rules and surveys \citep[e.g.,][]{cage2015improving,perez2017partisan,bouton2022public,guillot2023charitable}. This paper differs from this literature in that it does not focus on political donations or the resulting campaigns, but rather on the incentives that NGOs face when deciding to campaign against a firm.} and NGOs have much more lenient transparency requirements than publicly listed corporations, complicating the tracking of their fund usage. Also, the absence of a quantitative framework to systematically analyze NGO objectives presents another significant challenge, as NGOs' diverse missions cannot be easily reduced to profit maximization \citep{maxwell2012economic, fioretti2022caring}. The literature emphasizing extreme behaviors of non-profits, with some contributing to the greater good \citep[e.g.,][]{daubanes2019rise} whereas others are captured by corporate interests \citep[e.g.,][]{bertrand2021hall}. In this paper, we conduct a forensic investigation that infers NGOs' objectives from their reactions to external incentives, focusing on the primary—and often only—observable data on NGOs: their campaigns, which must necessarily be public and observable to influence firms and their stakeholders, drawing from data o To illustrate how objectives vary over time and in the life of an NGO, our analysis spans a wide range of organizations, from smaller entities like the `Physicians Committee for Responsible Medicine` and the `European Center for Digital Rights` to major players like `PETA` and the `Sierra Club.`
% % %As a result, the primary—and often only—observable data on NGOs come from their campaigns, which, by necessity, must be public to influence a firm and its stakeholders.%By analyzing their objectives in connection with various private politics tactics, this paper sheds light on the inner workings of NGOs, revealing the trade-offs they face and how they navigate them over time, with significant implications for both firms and society.
% % The starting point of our analysis is that NGO campaigns are strategically timed. Employing over ten years of campaign data against the largest US corporations, we identify two key factors that influence an NGO's behavior. First, we find that NGOs can improve their reputation by launching campaigns near the target firm’s Annual General Meeting (AGM) date, as it attracts the highest media coverage with media outlets engaging with managers and key shareholders to discuss the firm's achievements and outlook. However, such late campaigns are unlikely to influence shareholder votes at the AGM, as most shareholders have already voted by proxy \citep{bebchuk2020dancing}. Therefore, we NGOs must navigate a tradeoff between exposing themselves, their cause, and the target firms, but having small chances to directly affect a firm's actions or constructively engaging with the firm,
% % %Drawing on campaign data from a wide range of NGOs---spanning smaller entities like the `Physicians Committee for Responsible Medicine` and the `European Center for Digital Rights` to major players like `PETA` and the `Sierra Club`---we identify two key factors that influence an NGO's ability to capture stakeholders' attention. First, media coverage amplifies NGO campaigns, making efforts like shaming and boycotting more effective.\footnote{Consistent with this, \cite{hawn2021media} finds that media coverage of a firm's irresponsible behavior can significantly amplify negative public perceptions \citep{shea2019microfoundations}, potentially resulting in the failure of planned merger.} Second, NGOs with stronger reputations are better positioned to influence target firms, as they can leverage broader support from stakeholders such as workers, consumers, shareholders, suppliers, and the public. We demonstrate that these two factors create a trade-off regarding the strategic decision of campaign timing. NGOs must choose between maximizing influence on firms and enhancing their reputation.
% % % Employing over ten years of campaign data against the largest US corporations, we find that NGOs can improve their reputation by launching campaigns near the target firm’s AGM. The AGM attracts the highest media coverage, as media outlets engage with managers and key shareholders to discuss the firm's current achievements and future outlook \citep{dyck2008corporate}. This heightened media attention motivates NGOs to campaign during this period, increasing their visibility and positively influencing the perception held by the public and the target firm regarding the NGO’s stance on the issue, as well as the NGO’s commitment to driving change, thereby accumulating reputation \citep{bar2008seller,gent2015reputation,mitchell2017reputations}. Since our estimation routine compares the timing of the campaigns of NGOs campaigning against multiple firms in a year, we can account for any confounding factors varying within an NGO (and target firms) across years through fixed effects, such as their size.
% % A late campaign is unlikely to influence shareholder votes at the AGM, as most shareholders have already voted by proxy \citep{bebchuk2020dancing}. Nevertheless, we find that the heightened visibility of campaigns timed with the AGM increases the likelihood of (i) a shareholder proposal on a related cause the following year and (ii) a boycott campaign in the subsequent months, compared to campaigns launched at other times. Campaigns initiated one to two months before the AGM are more likely to lead to successful votes on related proposals at the current AGM, revealing a new trade-off between influence and exposure.
% % We argue that NGOs navigate this tradeoff differently depending on their reputation. We first show that NGOs are more likely to campaign on AGM dates early in their history but gradually shift to launching campaigns two or three months prior the AGM to impact votes directly. Our analysis rules out learning this aspect of proxy voting as an explanation for our finding, as we observe the same trend for NGOs whose first campaign occurred just after an AGM, meaning they already knew they could not influence the vote by campaigning that late. We also find no evidence that NGOs select different target firms as they gain more campaigning experience.
% % Second, we demonstrate that the success of a shareholder proposal at a current AGM is linked to how many times the NGO backing the proposal has campaigned on AGM dates in the past. We interpret this as a reduced-form proxy for the reputation built by the NGO for fighting over issues. As we document diminishing returns to targeting AGM dates, once an NGO has accumulated sufficient reputation, it is more effective at influencing shareholder votes through early campaigns. Conversely, we find that accumulating experience at campaigning outside AGM dates does not help NGOs at swaying shareholder proposal votes.
% % Empirically, our approach leverages exogenous variation in the timing of AGM meetings, which in the U.S. we show to be relatively stable within firms over time due to regulations by the Securities and Exchange Commission (SEC) designed to give shareholders adequate time to review proposals before voting them. Across analyses, our approach also allows us to exclude time-varying confounding factors at the firm or industry, and campaign cause levels. In particular, while current events and news may influence an NGO's decision to launch a campaign, they cannot influence the AGM calendar. As a result, objectives are identified by the tradeoff between targeting a firm's AGM date to increase their exposure and impacting the firm's decisions by campaigning early in support of the news or event.
% % Although we do not structurally estimate stakeholders' beliefs about an NGO's type (reputation)---an approach unfeasible with campaign data alone---our analysis indicates that the success of an NGO campaign in supporting a shareholder proposal at an upcoming AGM relies on the NGO's past activism, highlighting the importance of reputation as a signaling mechanism \citep{bar2008seller}.
% % We interpret our results through the lens of firm life cycle models, where consumers initially lack information about a firm's product quality, spurring firms to invest in reputation, especially when firms are young \citep[e.g.,][]{wernerfelt1988umbrella,cabral2000stretching,cabral2009umbrella}. In this context, waiting for the target firm's AGM is a costly investment for NGOs, as they must weigh the immediate recognition gained from protesting as soon as discovering a project they oppose against the benefits of waiting to protest when a firm is under substantial media scrutiny at a time that is salient for shareholders and decision makers.
% % Over time, NGOs may recognize that their campaigns garner significant support from key stakeholders like consumers and shareholders, reflecting a better reputation. Once established, they can leverage this reputation to directly impact firms' decisions, engage in negotiations with firms, or spend their image to directly support shareholder proposals.
% % % \newpage
% % % In April 2006, Greenpeace launched an aggressive public campaign against McDonald's labeled ``McAmazon'', as the fast-food giant was accused of participating in the Amazon deforestation. Within days, McDonald's executives engaged in discussion with the activist group, which ultimately led to a new sustainable sourcing policy \citep{goldberg2007}. As highlighted in this well-known case, non-governmental organizations (NGOs) have been prominent social actors since the 90s \citep{edwards1996too}, and often praised as an essential counterweight to state power \citep{world1994governance}. It is important to understand what incentives NGOs respond to, which ultimately shape the way they can achieve their objectives and help firms internalize their externalities.
% % % Yet, very little is known about how NGOs operate. NGO campaigns are the only observable action to researchers and constitute the backbone of their actions. A critical aspect of those campaigns is their timing and underlying objectives. While much research has focused on the outcomes of these campaigns \citep[for instance in the development literature, e.g.][]{vivalt2020much}, relatively little attention has been paid to the strategic considerations that precede them. Specifically, the question of when NGOs choose to target firms and why they do so at particular moments remains underexplored. These factors are vital as they can significantly impact the effectiveness of the campaigns and the responses from firms.
% % % Using data on NGO campaigns from Sigwatch, this paper seeks to fill this gap by examining the timing and objectives of NGO campaigns against US firms from 2010 to 2021. Our starting point is that NGO campaigns are strategically timed: they are eight times more likely to be launched on the day of the targeted firm's Annual General Meeting of Shareholders (AGM) than on any other day of the year. This observation suggests a deliberate effort by NGOs to capitalize on the heightened visibility that AGMs offer \citep{fioretti2024shared}.
% % % It is crucial to note that shareholder proposals and votes are typically decided well before the AGM date. Therefore, NGOs are not trying to influence the voting outcomes through their campaigns. Instead, we propose that NGOs deliberately time their campaigns to coincide with AGMs to leverage the increased media exposure that these events generate, as firms have on average X\% more news written on them during the AGM week. Indeed, a campaign launched during the month of the firm's AGM can increase the firm's media presence by 20\% more than one initiated three months earlier. This timing strategy enables NGOs to maximize public awareness of their campaign's objectives, using the AGM as a platform to amplify their message.
% % % These campaigns have tangible effects on firms by provoking reactions from different stakeholders. Specifically, campaigns launched during an AGM lead to a higher likelihood of shareholder proposals on related causes at the subsequent year's AGM and trigger a noticeable decline in sales, indicating a rise in consumer awareness and potential backlash. In this sense, NGOs are not merely drawing attention; they are actively shaping the agenda and influencing stakeholder behavior.
% % % Moreover, this exposure plays a critical role in helping NGOs build a reputation for effective activism. Our findings suggest that as NGOs gain experience in targeting AGMs, they tend to initiate campaigns further away from AGM dates, indicating a shift toward a more strategic approach to influencing firm behavior over time. This shift reflects an increasing ability to exert pressure on firms and drive meaningful change, beyond merely "stealing the spotlight" at AGMs.
% % % We further demonstrate that NGOs must balance the trade-off between reputation building and campaign efficacy. Campaigns initiated by NGOs with prior AGM-targeting experience are X\% more effective than those launched by inexperienced NGOs. However, this experience yields diminishing returns, as evidenced by a concave relationship between experience and campaign effectiveness. This suggests that while initial experience enhances an NGO's ability to exert pressure, additional experience beyond a certain point does not necessarily translate into greater effectiveness.
% % % In summary, our study sheds light on the strategic timing of NGO campaigns and their role in shaping corporate behavior. By highlighting the importance of media exposure, reputation building, and the trade-offs that NGOs face, we provide a comprehensive framework for understanding the determinants of their real social impact and their capacity to drive corporate change. \\
% % % This paper contributes to the growing literature on NGO campaigns as a form of stakeholder pressure influencing firms' decisions. Direct evidence of this influence includes findings by \cite{koenig2019social}, who demonstrate that NGO campaigns act as catalysts in punishing corporate social irresponsibility, and \cite{chatain2019ngos}, who explore how collaborations between firms and NGOs can lead to improvements in supply chains. Additionally, NGOs are more likely to target firms that exhibit higher levels of disclosure and transparency \citep{schafer2023interplay}, suggesting that firms with greater openness are more vulnerable to public scrutiny. Building on this insight, \cite{hombach2019firm} argue that regulators could leverage additional disclosure requirements as a tool to expose firms to public pressures, such as NGO campaigns, thereby nudging them toward more socially responsible behavior.
% % % Moreover, firms themselves may strategically leverage relationships with NGOs, either through donations or partnerships, to advance their political agendas \citep{bertrand2021hall} or to restore consumer and investor trust following a reputation-damaging event \citep{akey2021hacking}.
% % % As other publicly available signals of socially irresponsible behavior \citep{meier2023consumers, saint2024exit}, trigger shareholder and consumer reactions, . These papers collectively demonstrate that media exposure plays a critical role in disciplining corporate behavior, serving as a foundation for interpreting our results. In a related context, \cite{dimson2015active} find that ESG-motivated shareholder activism tends to be more effective when a firm's reputational risk is elevated, implying that media exposure acts as a catalyst for observable corporate social responsibility actions \citep{fioretti2024shared}.
% % % The effectiveness of NGO actions is closely tied to their reputation, a cause explored in other settings as well. For example, \cite{usmani2022ngos} find that prior engagement with target communities can increase the effectiveness of NGO actions by 30\%, emphasizing the importance of established credibility. However, it's worth noting that even well-intentioned campaigns can potentially damage an NGO's reputation if not executed carefully \citep{will2017discourse}. The concept of NGOs building their reputation, sometimes at the expense of immediate efficiency, can be rationalized through the broader literature on firm reputation and profitability. In this context, \cite{shapiro1983premiums} provides a relevant model, demonstrating how firms make "investments in reputation" by initially selling products at a loss to reveal their quality to a wide audience, with the expectation of reaping long-term profits. %Analogously, NGOs may engage in seemingly ineffective but highly publicized campaigns initially to establish their presence and credibility, which later enables them to engage more effectively with firms, often well before AGM dates.
% % % \vspace{1ex}
\end{comment}
\section{Data}\label{s:data}
Our dataset integrates four sources: NGO campaigns, online attention, shareholder proposals, and firm-level news.
First, we use data from \textit{Sigwatch}, a consulting firm tracking global NGO campaigns against corporations \citep{hatte2020geography}. The data record the campaign’s public release date, cause (e.g., ``GMOs in food''), the targeted company and its prominence, and sentiment (ranging from --2 to +2, typically negative). We focus on single-firm, negative-sentiment campaigns where the firm appears in the headline.\footnote{Since our analysis links campaign timing with a firm’s AGM date, we exclude multi-firm campaigns as their AGM dates may differ. Single-firm campaigns represent 93\% of the original sample.} The final dataset spans 2,512 campaigns launched by 731 NGOs against 523 firms.
Second, we collect monthly Google Trends scores for all NGOs in our sample from 2010 to 2022. Scores are normalized on a 0--100 scale, either (i) relative to the NGO’s own peak searches (\textit{within-NGO}) or (ii) relative to a benchmark NGO in the same month (\textit{between-NGOs}). We use Greenpeace---the most searched NGO---as the benchmark for comparability across organizations.
Third, we use data from \textit{Institutional Shareholder Services} (ISS) on proposals at the Annual General Meetings (AGMs) of the 3,000 largest US-listed firms. We select 4,483 environmental and social (E\&S) proposals filed between 2010 and 2021 (837 firms in total). About one-third of these proposals are successful, either withdrawn before the AGM (after firm commitments) or passed in a vote. Both the number of E\&S proposals and their success rate increased steadily over the period.
Lastly, we obtain firm-level news from \textit{RavenPack}, which transforms unstructured news into structured data with sentiment and event tagging. We count news items related to US-listed firms from 2010 to 2020, retaining only items with a \textit{Relevance Score} of at least 75\%. The final sample includes 5,212 firms.
We link Sigwatch, ISS, and RavenPack data using standard identifiers (CUSIP, ISIN, year). Appendix Table \ref{tab:summary} reports descriptive statistics. Panel (a) shows that, on average, NGOs launch 3.2 campaigns during the sample, 24\% of which are on the target AGM date, though over half of NGOs appear only once. Most campaigns focus on social rather than environmental issues, but the distribution is highly skewed: a small set of NGOs accounts for a disproportionate share of activity, while the median NGO runs only a single campaign. Panel (b) documents substantial variation across targeted firms in size and news coverage. Targets range from very large firms with extensive media presence to smaller firms with limited coverage, suggesting that NGOs are not exclusively focused on high-visibility corporations. 53\% of the campaigns are on social causes, with the remaining being environmental. We return to the remaining panels of Table \ref{tab:summary} in Section \ref{s:media}, after first providing background on AGM dates.
\begin{comment}
% \subsection{Data Sources}
Our dataset integrates four sources. First, we use data from \textit{Sigwatch}, a consulting firm tracking global NGO campaigns against corporations \citep{hatte2020geography}. The data includes the campaign’s public release date and cause (e.g., ``GMOs in food''), the targeted company and its prominence (e.g., if mentioned in the headline), and the sentiment (varying from negative, -2, to positive, +2), which is generally negative. We focus on campaigns against US-listed firms with negative sentiment, where only one firm is mentioned, and it appears in the headline.\footnote{Since our analysis links campaign timing with a firm's AGM date, we exclude campaigns against multiple firms as their AGM dates may differ. Those single-firm campaigns represent 93\% of the original sample.} The dataset spans 2,512 NGO campaigns launched by 731 unique NGOs and targeting 523 unique firms. The most active NGOs in our sample include well-known organizations such as the Sierra Club, PETA, and Greenpeace. The most targeted companies are Monsanto, McDonald's, Facebook, Coca-Cola, Apple, and Exxon Mobil. The most common causes are ``environment/pollution,'' ``human \& labor rights,'' ``animal rights,'' and ``health.''
Second, we collect monthly Google Trends scores for all NGOs in our sample from 2010 to 2022. Scores are normalised on a 0–100 scale, either (i) relative to the NGO’s own peak month of searches (\textit{within-NGO}), or (ii) relative to searches for another keyword in the same month \textit{(between-NGOs}). We construct both series, using Greenpeace--the most frequently searched NGO in our sample--as the benchmark for the \textit{between-NGOs} measure to ensure comparability across organisations.
Third, we use data from \textit{Institutional Shareholders Service} (ISS), a key provider of corporate voting data, on shareholder proposals at the Annual General Meetings (AGMs) of the 3,000 largest US-listed firms by market capitalization. For those firms, ISS data report detailed information on shareholder proposals for these firms, including proposals withdrawn before the AGM.
We select environmental and social (E\&S) proposals (4,483 in total) from the complete set of proposals filed at US firms between 2010 and 2021 (837 firms in total). Around a third of these E\&S proposals are successful: either withdrawn from the voting agenda before the AGM or passed after going through a shareholder vote.\footnote{Evidence shows that shareholders generally agree to withdraw the proposal only if the company commits to its implementation \citep{harvard2023}.} We document two facts: the number of E\&S shareholder proposals has increased by 40\% between 2010 and 2021, and their success rate has increased from 35\% in 2010 to almost 50\% in 2021. Appendix Figure \ref{fig:sum_stat_proposals} illustrates these trends.
Lastly, we source firm-level news data from Ravenpack. RavenPack is a data analytics platform that transforms unstructured news content into structured data, offering sentiment analysis and event detection. The data allows us to count the news related to every firm headquartered in the US.\footnote{We only retain news with a \textit{Relevance Score} of at least 75\%, which ensures the relevance of the news for the firm.} We include only US-listed firms from 2010 to 2020. The sample includes 5,212 US firms.
We link Sigwatch, ISS, and RavenPack using standard identifiers, including year, firms' CUSIP, and ISIN numbers. Appendix Table \ref{tab:summary} presents summary statistics for NGO-related variables (Panel a). On average, NGOs launch 3.40 campaigns during the sample period, but the distribution is skewed, with over 50\% of NGOs initiating only one campaign. Most campaigns focus on social issues rather than environmental ones. Panel (b) displays the variation in key variables for the targeted firms, including market capitalization, assets, and news coverage.
The following three sections examine an NGO's payoffs in connection with the timing of its campaigns, and the reactions of central stakeholders. We present the key takeaways from each analysis at the end of each section. Before that, we detail our empirical approach.
%\noindent \textbf{NGO reports data.} We employ data from Sigwatch, a European consulting firm tracking and analyzing NGO campaigns globally \citep{hatte2020geography}. A campaign is defined as an event launched by an NGO against one or several companies on a given cause. The dataset indicates the date when a campaign has been made public on an NGO's website, the targeted company, its cause (e.g., ``GMOs in food"), a measure of the sentiment ranging from --2 to +2, and the prominence of the firm in the campaign (e.g., whether the firm is mentioned in the headline of the report or elsewhere in the communication).
% We keep only campaigns targeting US-headquartered listed firms with a negative tone (sentiment $< 0$), where only one firm is mentioned, and the mention appears in the headline (prominence = 4). We exclude cases where multiple firms are targeted because, having different AGM dates, the time from the start of the NGO campaign and the AGM dates will differ across firms, polluting our estimates. Overall, these filters ensure that these firms are at the center of a negative controversy.
% \vspace{1ex}
% \noindent \textbf{News data.} Firm-level news data come from Ravenpack. We keep only news with a relevance score of at least 75\%.\footnote{Entity relevance is a metric between 0 and 100 computed by RavenPack, indicating how strongly related a firm is to the underlying news. A high score ensures that the event is highly relevant for a firm.}
% Due to our interest in AGMs, we consider only US-listed firms from 2010 to 2020 (4,676 firms).
% \vspace{1ex}
% \noindent \textbf{Shareholder Proposals data.} Data on shareholder proposals at firms' AGMs come from \textit{Institutional Shareholders Service} (ISS), the main provider of corporate voting data. Starting from the entire universe of shareholder proposals, we select the environmental or social shareholder proposals filed at US firms from 2010 to 2021, as identified by the data provider. The final sample has a total of 2,078 shareholder proposals submitted at 459 different firms.
% \vspace{1ex}
% \subsection{Summary Statistics on NGO Campaigns' Timing}
% \vspace{1ex}
% \noindent \textbf{Timing of campaigns.} We recenter each firm's year around its AGM, defining every week as $k$ weeks from the AGM, with $k$ ranging from -24 to 24. Panel (a) of Figure \ref{fig:sum_stat} shows the average weekly number of news (blue bars) and NGO campaigns (gray bars). News coverage follows a cyclical pattern, peaking every three months in line with quarterly earnings announcements, with an average of 12 to 20 news items per week. NGO campaigns, however, are concentrated around the AGM: while a typical week sees 13 campaigns, AGM weeks receive around 100. Appendix Figure \ref{fig:cattaneo} tests \citep{cattaneo2020simple} and confirms that the density of the NGO campaigns exhibits a discontinuous drop right after the AGM date (p-value $\leq 1\%$), suggesting that NGO target AGMs.
% We recenter every firm's year around their AGM, such that every week is defined as being $k$ weeks away from the AGM, with $k$ ranging from -24 to 24. In Figure \ref{fig:sum_stat}, we plot the average weekly number of news (blue bars) and the number of NGO campaigns in the sample (gray bars). We first notice the cyclicality of news: a firm's media coverage peaks every three months, in line with its corporate calendar (earnings announcements are quarterly). The average weekly number of news varies from 12 to 20. On the other hand, the NGO campaign calendar is centered only around the AGM: while a typical week receives on average 13 NGO campaigns, the AGM weeks have 100 campaigns.
\end{comment}
\section{Institutional Background on AGM Dates}\label{s:empirical}
Our empirical approach examines the timing of NGO campaigns relative to AGMs, leveraging the exogeneity of AGM dates, which remain stable within firms across years due to SEC regulations. Rule 14a-8 mandates shareholder proposals be filed 120 days before the proxy statement, anchoring AGM dates across years \citep{glac2014influence}. Appendix Figure \ref{fig:agm_persistence} confirms this stability, with exceptions largely due to pandemic-related adjustments allowed by the \cite{securities2020guidance} to facilitate shareholder attendance at AGMs.\footnote{Results remain robust excluding the covid period (2020–2021).}
While external events may shape NGOs’ incentives to campaign, they do not alter the timing of AGMs. This stability makes AGMs a useful benchmark: NGOs can choose to react immediately to a negative event concerning a firm, or alternatively wait and align their campaign with its AGM, when the firm is most exposed to scrutiny, as shown in Panel (b) of Figure \ref{fig:intro_fig}, which shows that news coverage spikes in AGM weeks. Consistent with this view, we show in the Appendix that AGM-timed campaigns generate more negative news coverage than those launched at other times (Appendix Figure \ref{fig:ngo_agm_bad_news}). Taken together, these patterns highlight why AGMs provide a natural focal point to study the strategic trade-off between maximizing visibility and exerting direct influence.
% Our empirical approach investigates the time gap between a firm's AGM and an NGO campaign, while controlling for time-varying confounding factors at the firm, industry, and campaign cause levels across years. Importantly, our identification strategy relies on the independence of the corporate calendar from events that may motivate NGOs to launch campaigns. The timing of AGMs is largely exogenous, as SEC regulations mandate a stable schedule across years. Specifically, Rule 14a-8 ensures shareholders have ample time to review proposals before voting, requiring that shareholder proposals be filed at least 120 days before the proxy statement is released, based on the previous year's AGM timing. This stabilizes AGM dates across consecutive years \citep{glac2014influence}, as shown in Appendix Figure \ref{fig:agm_persistence}, which plots AGM months across adjacent years. Most variation outside the diagonal line—indicating significant fluctuations in AGM months—is attributed to the pandemic, as the SEC allowed firms to adjust AGM dates in 2020 to prevent COVID-19 spread and facilitate shareholder attendance \citep{securities2020guidance}. Our results remain robust even when excluding data from 2020 and 2021.
% While current events may influence an NGO's decision to launch a campaign against a firm, they are unlikely to affect the timing of AGMs. These external factors may shift an NGO's objectives by altering incentives to target a firm. Our analysis leverages these shifts to identify a key trade-off for NGOs: whether to influence a firm's strategies by launching campaigns early or to gain greater visibility by timing campaigns around the firm's AGM.
% \vspace{1ex}
% \noindent\textbf{Target selection.} While our analysis centers on campaign timing—leaving NGOs’ target selection to future research—Appendix Table \ref{tab:agm_selection} shows that NGOs campaigning on AGM dates do not select firms that differ in observable characteristics (market capitalization, assets, sales, media coverage, or environmental and social scores) from those targeted at other times, and campaign sentiment is likewise comparable across both groups. Moreover, Table \ref{tab:summary} demonstrates that target choice is independent of an NGO's campaign experience: Panel (c) reveals no significant differences in targets' market cap, assets, news mentions, ESG scores, or campaign history between targets of small and large NGOs, and Panel (d) finds no correlation between targets' characteristics and NGOs’ past campaigns. However, we see that NGOs increasingly focus on environmental over social issues in later years, reflecting the rising importance of environmental concerns. %Our results highlight NGO experience as a crucial factor in this trade-off. Before interpreting our findings,
% \textcolor{red}{For identification reasons, our analysis focuses on campaign timing while abstracting from the first stage---NGOs' investment in gathering information on specific firms---which we leave for future research.} Importantly for our analysis, Appendix Table \ref{tab:summary} suggests that NGOs' targets do not differ based on the number of campaigns an NGO has previously launched. For instance, Panel (c) shows that both small and large NGOs (in terms of number of campaigns) target firms with similar market capitalization and total assets. We also find no differences in news mentions, environmental scores, social scores, or the number of environmental and social campaigns received.
% Because of identification reasons, our analysis focuses on the timing of a campaign without considering the first stage that is the NGO's investing into finding information regarding a specific firm, which we leave for further research. However, results in Appendix Table \ref{tab:summary} suggest that NGOs do not select targets based on their own experience, namely the number of campaigns they launched in the past. Panel (c) reveals that both small and large NGOs---in terms of their number of campaigns in the sample---target firms with similar market capitalization and total assets. We also detect no differences in news mentions, environmental scores, social scores, and the number of environmental and social campaigns received.
% Panel (d) examines target selection by regressing each firm-level variable on the number of past campaigns launched by the NGO targeting the firm, excluding (Column 1) or including (Column 2) firm-by-year, NGO-by-year, and cause-by-year fixed effects. The results show that NGOs' campaign targets do not correlate with the past experience of the campaigning NGO, though they increasingly focus on environmental issues over social ones in later years, reflecting the growing prominence of environmental challenges.
% Our empirical approach examines the time gap between a firm's AGM and actions taken by NGOs or other stakeholders while controlling for time-varying confounding factors at the firm, industry, or campaign cause levels. Crucially, our identification strategy relies on the independence of the corporate calendar from events that might motivate NGOs to launch a campaign. AGM timing is largely exogenous to concurrent events, as Security and Exchange Commission (SEC) regulations require AGMs to follow a stable schedule across years. Specifically, Rule 14a-8 ensures shareholders have sufficient time to review proposals before voting on them. As a result, shareholder proposals must be filed at least 120 days before the proxy statement is released, based on the previous year's AGM timing. This effectively stabilizes AGM dates across consecutive years \citep{glac2014influence} as observed in Appendix Figure \ref{fig:agm_persistence}, which plots AGM months across adjacent years. Most of the variation outside of the diagonal line---referring to wild variation in AGM months across consecutive years---is due to the pandemic, as the SEC permitted firms to adjust AGM dates in 2020 to avoid AGM meetings becoming covid-spreader events and allow shareholders to attend them \citep{securities2020guidance}. Our results remain robust even when excluding data from 2020 and 2021.
% Therefore, while current events or news breaks may influence an NGO's decision to launch a campaign against a firm, they are unlikely to affect the timing of AGMs. These external factors ``shift'' an NGO's objectives by modulating its incentives to target a firm. Our analysis leverages these shifters to identify a novel trade-off NGOs face: influencing a firm's strategies by launching campaigns early or gaining greater visibility by timing campaigns around the firm's AGM.
% Our results indicate NGO experience as a central variable influencing this tradeoff. Therefore, before interpreting our results, we show in Appendix Table \ref{tab:summary} that NGO do not seem to select different firms as target based on their experience. Separating firms based on whether they have above or below the long-run number of campaigns, Panel (c) finds that small and large NGOs target firms with similar market capitalization, total assets--we use deciles of the underlying yearly distributions for these two variables because these variables have increased for some deciles substantially overtime, mechanically influencing the estimates otherwise--number of yearly news, environmental and social scores, and also display similar number of envirnomental and social campaigns. Panel (d) examines selection over these variables within firms, by regressing each of this variables on the number of past campaigns that the relevant NGO had launched, excluding (column 1) or including (2) fixed effects at the level of firm-by-year, ngo-by-year, and campaign-by-year. The correlation analysis shows that ngos campaign target do not change over time. However, NGOs are more likely to campaign for environmental rather than social issues in later years.
% \vspace{1ex}
% \noindent \textbf{AGM targeting.} Is the NGO campaign peak on the week of the AGM significant? For more granularity, we look at the density of NGO campaigns happening on and around the day of the AGM. More specifically, we conduct a \cite{cattaneo2020simple} tets of density with a cutoff on the day after the AGM. The estimated coefficient is negative and statistically significant at the 1\% level, indicating that there is indeed a drop in the probability of observing an NGO campaign right after the AGM.
% \vspace{1ex}
% \section{Campaign Timing and Reputation Building}
\section{Timing Campaigns for Visibility} \label{s:media}
\begin{comment}
NGOs are substantially more likely to launch their campaigns around AGM dates, on average. According to Figure \ref{fig:NGO_time}, NGO campaigns are about 4 to 6 times more common on AGM weeks compared to other weeks; while a typical week sees 28 campaigns, AGM weeks have more than 100.\footnote{Appendix Figure \ref{fig:cattaneo} confirms this finding showing a significant drop in campaign density immediately after AGM dates (p-value $\leq 1\%$) \citep{cattaneo2020simple}.}
\begin{figure}[t]
\centering
\caption{Timing of NGO campaigns and media attention around AGMs \label{fig:sum_stat}}
\begin{subfigure}{0.49 \linewidth}
\centering
\includegraphics[width=\linewidth]{Figures/sum_stat_ngo.pdf}
\caption{Campaigns peak during AGMs \label{fig:NGO_time}}
\end{subfigure}
\begin{subfigure}{0.49 \linewidth}
\centering
\includegraphics[width=\linewidth]{Figures/sum_stat_news.pdf}
\caption{Media coverage rises during AGMs \label{fig:AGM_coverage}}
\end{subfigure}
\begin{minipage}{1 \textwidth} % choose width suitably
{\footnotesize Note: Panel (a) displays the number of NGO campaigns launched relative to the date of the Annual General Meeting of the targeted firm (week 0), based on 1,489 NGO campaigns. Panel (b) compares the weekly average count of news during the AGM week (gray bars) and outside of the AGM week (white bars), based on weekly observations for 5,211 US firms. The NGO data come from Sigwatch; the news data are from Ravenpack. Calendar weeks are re-centered around the firm's AGM. \par}
\end{minipage}
\end{figure}
This finding could be related to the heightened stakeholder attention during AGMs \citep[][]{fioretti2024shared}, as illustrated in Figure \ref{fig:AGM_coverage}, where news coverage peaks in the 10 days around an AGM (gray bars) compared to other periods (white bars). %To test if media exposure of NGO campaigns is highest during the AGM week, we aggregate negative E\&S news by firm ($n$) and week-by-year ($w,y$) and perform the event study:
To test if NGO's visibility is highest when they campaign on AGM day, we perform the following event study:
\end{comment}
Having shown that NGO campaigns cluster around AGM dates and that they increase a target's negative media coverage, we next ask whether such timing enhances NGOs’ visibility. We use Google search activity as a proxy for NGO visibility and estimate event-study regressions comparing campaigns launched on the AGM date with those launched at other times:
\begin{equation}
\label{eq:gt_model}
\begin{aligned}
\text{Google Trends Score}_{g,m,y} & = \sum^{\tau=5}_{\substack{\tau=-4 \\ \tau\neq -1}} \left(\beta_{\tau}^{AGM} \cdot \text{AGM Campaign}_{c}\cdot \mathbf{1}_{\{M_{c,m,y}=\tau\}}\right) \\
&\quad + \sum^{\tau=5}_{\substack{\tau=-4 \\ \tau\neq -1}} \left(\beta_{\tau}^{Non-AGM} \cdot \mathbf{1}_{\{M_{c,m,y}=\tau\}}\right)
\\
&\quad + \alpha_{c,y} + \alpha_{m,y} + u_{g,m,y}.
\end{aligned}
\end{equation}
Because NGOs typically campaign multiple times, we follow \citet{baker2022much} and stack all campaign events. Each event is denoted $c(i,n,g,y,m)$, where campaign $c(\cdot)$ is identified by cause $i$ (e.g., pollution), NGO $g$ (e.g., Greenpeace), target firm $n$ (e.g., Shell), calendar year $y$, and month $m$. We then analyze variation in the Google Trends score for NGO $g$ within a five‐month window around the start of $g$’s campaign; this window may spill into years $y-1$ or $y+1$ if the campaign occurs early or late in year $y$.
The main explanatory variables are event-time dummies, $\mathbf{1}_{\{M_{c,m,y} = \tau\}}$, equal to one if month $m$ is exactly $\tau$ months from campaign $c$, with $\tau \in [-5,5]$. The omitted category is the month immediately preceding the campaign ($\tau = -1$). These indicators are interacted with $\text{AGM Campaign}_{c}$, a dummy equal to one if campaign $c$ was launched on the target firm’s AGM date. Thus $\beta^{AGM}_\tau$ estimates the increase in visibility for NGO campaigning on an AGM date, over the direct effect of campaigning at any date, $\beta^{Non-AGM}$.
The remaining terms include fixed effects at the NGO–firm–year level ($\alpha_{c,y}$), which absorb all variation across NGO-firm pairs over years, as well as month-by-year fixed effects ($\alpha_{m,y}$). Our main coefficients of interest ($\beta^{AGM}_\tau$) are therefore identified from within-pair variation across months. Standard errors are clustered at the NGO level.
Under the hypothesis that AGM campaigns affect NGO visibility differently depending on their reputation, we estimate \eqref{eq:gt_model} separately for an NGO’s first few campaigns, when its recognition is still limited, and for later ones. Panel (a) of Figure~\ref{fig:gtrends_campaigns} reports the estimated coefficients of $\beta^{AGM}_\tau$, splitting the sample into an NGO’s first three campaigns (blue dots) and subsequent campaigns (red triangles). The results indicate that AGM campaigns generate a sharp and statistically significant increase in NGO visibility, raising Google Trends scores by almost 50\% in the following three months. For later campaigns, the effect is weaker—about 40\% in the first month and around 10\% in the next two months—and not statistically significant in any period, consistent with diminishing visibility returns as NGOs become more established. These results are estimated within NGOs: early in their lifecycle, NGOs benefit more from AGM timing to build visibility, but the returns to this tactic decline as their public recognition grows.
\begin{figure}[!htbp]
\centering
\caption{Increased NGO visibility around an AGM campaign}
\label{fig:gtrends_campaigns}
\begin{subfigure}{0.56 \linewidth}
\centering\includegraphics[width=\textwidth]{Figures/gtrends_campaigns.pdf}
\caption{The first AGM campaigns bring visibility}
\end{subfigure} \hfill
\begin{subfigure}{0.43 \linewidth}
\centering\includegraphics[width=\textwidth]{Figures/selection.pdf}
\caption{No evidence of Selection}
\end{subfigure}\hfill
\begin{minipage}{1 \textwidth} % choose width suitably
{\footnotesize Note: Panel (a) plots the estimated $\beta^{AGM}$ coefficients from \eqref{eq:gt_model}. The dependent variable is the monthly Google Trends score for the NGO. Blue dots are estimated by restricting the sample to an NGO’s first three campaigns, while red triangles consider later campaigns. Standard errors are clustered at the campaign level. Panel (b) reports the estimated coefficients of the regression of an indicator that an NGO campaign was launched on the target firm's AGM date on the displayed (standardized) variables, and year fixed effects. Error bars shows 90\% (shaded) and 95\% confidence intervals using standard errors that are clustered by firm. }
%The sample period goes from January 2010 to December 2019 and covers 4,676 firms. NGO campaign data from Sigwatch. AGM data from ISS. News data from Ravenpack. Industry data from Compustat.
\end{minipage}
\end{figure}
We show that these results are robust to addressing three main concerns. First, the stacked event-study framework implicitly treats campaign timing as simultaneous across NGOs, whereas in reality campaigns are staggered across organizations and calendar months. This misalignment can distort estimates when treatment effects are heterogeneous across groups or over time. Second, some NGOs launch multiple campaigns within the same year, which may confound the analysis if two campaigns fall within five months of each other. To address both issues, Appendix \ref{apndx:specs} replicates the analysis using the staggered difference-in-differences estimator of \citet{sun2021estimating}, which explicitly accounts for variation in treatment timing, as well as a distributed-lag specification that flexibly controls for past campaigns within a year. Both approaches yield similar results, with substantially larger gains for early campaigns.
Third, although we find no evidence of pre-trends, NGOs might still select targets strategically to raise visibility—for example, by focusing on firms that are more vulnerable to pressure. Panel (b) of Figure \ref{fig:gtrends_campaigns} addresses this concern by regressing an indicator for AGM-timed campaigns on firms’ financial and ESG characteristics and the cause of the campaign, with year fixed effects and firm-clustered standard errors.\footnote{The independent variables include the target firm's market capitalization (\textit{Market Cap}), yearly sales (\textit{Sales}), total assets (\textit{Total Assets}), number of news over the past 12 months (\textit{News count}), the firm's Environmental and Social scores (\textit{Env. score}, \textit{Soc. score}), how negative the campaign is towards the firm (\textit{Campaign sentiment}), and an indicator for the campaign cause being related to an environmental topic (\textit{Env. campaign}).} Note that no coefficient is significantly different from zero. We find that a one standard deviation increase in any of these variables moves the likelihood of an AGM campaign by less than 5\% of a standard deviation, indicating little selection on observables.
Further evidence on potential selection based on the number of past NGO campaigns is presented in Table \ref{tab:summary}. Panel (c) compares firms targeted by NGOs that had previously campaigned on a target’s AGM date with those that had not, finding no significant differences across the same variables analyzed in Panel (b) of Figure \ref{fig:gtrends_campaigns}. This suggests that NGOs do not appear to select targets based on their past campaigning strategies. Panel (d) corroborates this result by regressing firm characteristics on the number of past NGO campaigns and showing no systematic relationship once fixed effects are included. The only consistent pattern is that NGOs increasingly focus on environmental rather than social issues over time, reflecting their growing salience.%\footnote{Results are robust to including NGO-by-year, firm-by-year, NGO-by-firm, and month-by-year fixed effects.}
Thus, the evidence presented so far points to:
\begin{result}\label{res:1}
Visibility gains from AGM‐day campaigns are greatest during an NGO's initial campaigns.
\end{result}
\noindent For clarity, each of the next sections highlights its main findings in a similar manner.
% \newpage
% Having shown in Figure \ref{fig:intro_fig} that NGO campaigns cluster around AGM dates, we begin by asking whether such timing enhances NGOs’ visibility.\footnote{Appendix Figure XXX shows that AGM-timed campaign increases negative news coverage more than campaigns timed at other time.} We estimate event-study regressions of Google search activity around campaigns, as a measure of an NGO's visibility, distinguishing between those launched on the AGM date and those launched at other times, using the following specification:
% \begin{equation}
% \label{eq:gt_model}
% \begin{aligned}
% \text{Google Trends Score}_{g,m,y} & = \sum^{\tau=5}_{\substack{\tau=-4 \\ \tau\neq -1}} \left(\beta_{\tau}^{AGM} \cdot \text{AGM Campaign}_{c}\cdot \mathbf{1}{\{M_{c,m,y}=\tau\}}\right) \\
% &\quad + \sum^{\tau=5}_{\substack{\tau=-4 \\ \tau\neq -1}} \left(\beta_{\tau}^{Non-AGM} \cdot \mathbf{1}{\{M_{c,m,y}=\tau\}}\right)
% \\
% &\quad + \alpha_{m,y} + \alpha_{c,y}+ u_{g,m,y},
% \end{aligned}
% \end{equation}
% \noindent
% Here, we follow \citep{baker2022much} and stack all NGO campaign events, denoted by $c(g,n,y,m)$ where we identify a campaign by an NGO $g$, a target firm $n$, in year $y$ and month $m$, and consider how the the NGO's Google Trends score varies in the five months around its campaigns.
% The main explanatory variables are a set of event-time dummies, $\mathbf{1}\{M_{g,m,y,c} = \tau\}$, which takes the value of one if the distance between campaign $c(\cdot)$ and month $m$ of year $y$ is exactly $\tau$ months, with $\tau \in [-5,5]$. These indicators are interacted with $\text{AGM Campaign}_{c}$, a dummy equal to one if campaign $c$ was initiated on the target firm's AGM day. The reference period is $\tau = -1$, corresponding to the month immediately preceding the campaign launch.
% We also include campaign-by-year fixed effects ($\alpha_{g,y}$), topic-by-year fixed effects ($\alpha_{t(g),y}$), and month-by-year fixed effects ($\alpha_{m,y}$) to control for yearly changes within NGOs, targeted firms, and NGO-firm relationships, as well as time-varying topic-relevant changes. Standard errors are clustered at the NGO level.
% Figure \ref{fig:gtrends_campaigns} plots the estimated coefficients for the campaign event study. The left panel shows results for the first three campaigns launched by an NGO, while the right panel focuses on later campaigns. In both cases, we find no evidence of pre-trends. For inexperienced NGOs, launching a campaign on the target firm’s AGM day leads to a sharp rise in Google searches in the month immediately following the campaign--almost 0.8 of a standard deviation--followed by a gradual decline that remains statistically significant for up to three months. By contrast, NGOs that have already launched more than three campaigns do not experience any visibility gains. Moreover, campaigns launched on days other than the target firm’s AGM generate no discernible increase in attention, regardless of NGO experience.
% \begin{result}\label{res:1}
% Gains in visibility are greater for unexperienced NGOs campaigning on the target's AGM day.
% \end{result}
\subsection{Stakeholders' Reactions}
We next examine how campaign timing affects stakeholders and firms’ ESG outcomes.
\subsubsection{Shareholders}
Shareholders can be seen as primary proponent of firm's social values \citep[e.g.,][]{naaraayanan2021real,baker2022investors}. Shareholders may respond more to causes campaigned during AGMs, when media coverage is highest. We define as ``related'' a shareholder proposal on the same cause of the campaign in the previous year. Appendix \ref{apndx:match} describes the matching procedure.
For each campaign in our sample, we examine the probability that the next AGM includes a related shareholder proposal, depending on the campaign’s timing. To this end, we recalendarize time so that month 0 corresponds to the AGM month for each firm. As discussed in Section \ref{s:empirical}, AGM months are stable over time; thus, this recentering does not alter the distribution of months across years or risk double counting.\footnote{All subsequent analyses use the same recalendarization. Months –5 and +5 are defined as 45-day intervals to ensure the recalendarized year covers the entire calendar.} We estimate the following regression:
\begin{equation}\label{eq:shareholders}
\begin{aligned}
%\text{Proposal}_{i,n,m,y} & = \sum^{\tau=24}_{\substack{\tau=1 \\ \tau\neq 1}} \left(\beta_\tau \cdot \mathbf{1}{\{m=\tau\}} \cdot \text{NGO Campaign}_{i,n,m,y}\right) + \gamma_n + \alpha_{i,s(n),m,y}+ u_{i,n,m,y}.%\\ & +(\text{Industry} \times \text{Year} \times \text{Topic})_{i,s(n),m,y} + u_{i,n,m,y}.
\text{Proposal on Same Cause}_{i,n,y} &= \sum^{\tau=-4}_{\substack{\tau= -15 \\ \tau\neq -4}} \left(\beta_{\tau} \cdot \mathbf{1}_{\{\text{Distance to AGM$_{c(i,n,g,y,m)}$} = -\tau\}}\right) \\
&+ \alpha_{i,s(n),g,m,y} +\gamma_n + u_{i,n,y}
\end{aligned}
\end{equation}
For each campaign $c(i,n,g,y,m)$ launched by NGO $g$ against firm $n$ on cause $i$ in year $y$ and month $m$, we analyze the likelihood that the next AGM of firm $n$ includes a shareholder proposal on the same cause. Formally, $\text{Proposal on Same Cause}_{i,n,y}$ equals one if such a proposal is observed within 4 to 15 months after campaign $c(\cdot)$, and zero otherwise, with timing measured by the indicator $\mathbf{1}_{\{\text{Distance to AGM}_{c(i,n,g,y,m)} = -\tau\}}$. Since SEC 14a8(e)(2) rule prevents shareholders from including proposals in the proxy statement within 120 days of its release, we cap the analysis at four months before the previous year’s AGM and set the reference month to $\tau =-4$ because of SEC rules.
Although firm-by-time fixed effects are not possible in this setting as they would absorb the variation used to estimate $\beta_\tau$, this analysis purges all the variation from cause-by-industry-by-NGO-by-month-by-year through $\alpha_{i,s(n),g,m,y}$ and firm invariant characteristics through $\gamma_n$, which we view as the main confounding factor since issue salience varies systematically across firms, industries, causes, and NGOs over time. We cluster standard errors by firm.
Figure \ref{fig:campaign_proposal} shows that only campaigns launched twelve months before the next AGM increase the likelihood of a shareholder proposal on the same issue. The effect is substantial—an increase of 12 percentage points, or 44\% relative to the baseline. This timing suggests that campaigns around an AGM may spur proposals at the next one, possibly due to heightened media attention.
Importantly, we can rule out the possibility that NGOs partnering with shareholders drive this outcome. If that were the case, submitting a shareholder proposal followed by an early campaign would allow the partners to immediately gain broad support on the proposal, without having to wait a year to advocate for change at the next AGM.
\subsubsection{Consumers}
Drawing from quarterly data from Compustat, we investigate the responses of consumers to the timing of NGO campaigns by studying the relation between the target's ``Sales to Asset ratio'' and the distance between the campaign and the target's AGM through
\begin{equation}\label{eq:sales}
\begin{aligned}
\frac{\text{Sales}_{n,q}}{\text{Assets}_{n,q-1}} & = \sum^{\tau=5}_{\substack{\tau=-4 \\ \tau\neq -1}} \left(\beta_\tau \cdot \text{NGO Campaign}_{n} \cdot \mathbf{1}_{\{\text{distance}=\tau\}}\right) + \alpha_{n,y} +\alpha_{s(n),q} + u_{n,q},
\end{aligned}
\end{equation}
\noindent where $q$ indicates a quarter and $\alpha_{n,y}$ and $\alpha_{s(n),q}$ are fixed effects at either the firm-by-year level or the industry-by-quarter-by-year level, respectively. The dependent value normalizes $n$'s sales by its asset to account for time variation in relative firm sizes, which is large in our ten-year sample. Standard errors are clustered at the firm level.
Appendix Figure \ref{fig:campaign_sales} shows the estimated ${\hat{\beta}_\tau}$ coefficients separately for NGO campaigns (i) planned on an AGM date (Panel a) and (ii) not planned on an AGM date (Panel b). We find no consumer response in either case, indicating that shareholders may be the primary audience for NGO campaigns.\footnote{Unreported results show also no difference between early and late campaigns.}
% do so to campaigns targeted on the AGM date, implying a loss of sales to total assets of 0.7pp, or about 5\%---a magnitude consistent with other papers \citep{conway2024consuming,saint2024exit}.%We find that consumers do not respond to NGO campaigns not timed on an AGM day, but they do so to campaigns targeted on the AGM date, implying a loss of sales to total assets of 0.7pp, or about 5\%---a magnitude consistent with other papers \citep{conway2024consuming,saint2024exit}.
\subsection{Does AGM-Targeting Create Positive Change?}
\label{s:ES}
If public shaming effectively influenced firm behavior, the AGM month should deliver the largest improvements in E\&S scores, given that it triggers heightened negative news coverage of the target firm (Appendix Figure \ref{fig:ngo_agm_bad_news}). However, Appendix Figure \ref{fig:campaign_esg_score} shows that only campaigns launched 3 to 4 months before the AGM lead to subsequent increases in firms' E\&S scores—rising by around 3\% over the following year.\footnote{Acknowledging the limits of ESG scores as measures of social and environmental impact \citep[e.g.,][]{kotsantonis2019four,allcott2022economic}, we use MSCI to mitigate signal risk: its ratings are more time-consistent and widely used by sustainable investors \citep{berg2021history,berg2024economic}.} By contrast, campaigns launched at other times, including during the AGM month, have no economically or statistically significant effect.
This finding suggests that NGOs' preference for targeting AGM dates may be driven by factors beyond their \textit{direct} influence on firms. Therefore, the results in this section suggest that:
% timing campaigns around AGMs is not driven by an overlooked mechanism.
\begin{result}\label{res:2}
NGO campaigns on AGM dates spur shareholders to place related proposals at future AGMs of the targeted firm, but they produce no contemporaneous improvements in the firm’s E\&S scores. %(i) Shareholders are more likely to make related proposals the following year. (ii) The target's sales drop in the months after a campaign.
\end{result}
\section{Timing Campaigns to Influence AGM Outcomes}\label{s:influence}
Early campaigns could benefit an NGO aiming to influence current AGM outcomes. Since most votes are cast by proxy before the AGM, targeting the AGM date may limit an NGO’s ability to sway a significant share of the votes. For instance, over 85\% of vote-eligible shares for Microsoft’s 2023 AGM were cast by proxy before the meeting, indicating that campaigning around AGM dates may have little impact on shareholders' votes \citep{microsoft2023}.
We define a proposal as successful if either it is withdrawn by the shareholder before the AGM or it secures at least $50\% + 1$ vote at the AGM. We first look at the influence of NGO campaigns on the success of related shareholder proposals using the following specification on the whole sample of shareholder proposals:
\begin{equation}\label{eq:success}
\text{Success}_{i,n,m,y} = \sum^{0}_{\substack{\tau=-6}} \beta_\tau \cdot \text{Campaign}_{i,n,m+\tau,y} + \alpha_{n,y} + \alpha_{i,y} +\alpha_{m}+u_{i,n,m,y},
\end{equation}
where ``$\text{Success}_{i,n,m,y}$'' is an indicator for the success of proposal $i$ at firm $n$ in month $m$ in year $t$, and ``$\text{Campaign}_{i,n,m+\tau,t}$'' is an indicator for the occurrence of an NGO campaign on the same cause as proposal $i$, $\tau$ months before the AGM. The reference period covers the months that are more than six months before the AGM of year $y$---i.e., $m\in[-11,-7]$, as campaigns launched during that period are more likely to be in reaction to the previous year's AGM results. We account for firm-by-year, and cause-by-year fixed effects, as well as seasonality with month fixed effects, thereby exploiting only variation across campaigns and causes within firm-year pairs. The standard errors are clustered by firm.
\begin{figure}[t]
\centering
\caption{NGOs' campaigns before the AGM help E\&S shareholder proposals succeed \label{fig:ngo_support}}
\begin{subfigure}{0.49 \linewidth}
\centering\includegraphics[width=\textwidth]{Figures/timing_months_firm_level_withdrawn_es.pdf}
\caption{Proposal accepted or withdrawn }
\end{subfigure} \hfill
\begin{subfigure}{0.49 \linewidth}
\centering\includegraphics[width=\textwidth]{Figures/timing_months_pct_es.pdf}
\caption{Vote share in favor of proposal %\label{fig:ngo_support}
}
\end{subfigure}\hfill
\begin{minipage}{1 \textwidth} % choose width suitably
{\footnotesize Note: The figure plots the estimated coefficients from (\ref{eq:success}). The dependent variables are an indicator variable that is one if a shareholder proposal is withdrawn or successful if voted (911 proposals) in Panel (a), and the percentage of votes in favor of a proposal (2,078 proposals) in Panel (b). Standard errors are clustered at the firm level. \par}
\end{minipage}
\end{figure}
Panel (a) Figure \ref{fig:ngo_support} plots the estimated ${\beta}_\tau$. We find that NGO campaigns happening 6 months and 1 month before the AGM positively influence the outcome of related shareholder proposals. They increase, respectively, the probability of success by 2 percentage points and 3 percentage points more than a campaign launched more than six months earlier, corresponding to a relative increase between 6 and 8\%.
To understand why campaigns launched six months before the AGM are also somewhat successful, we re-estimate \eqref{eq:success} on the subset of proposals that went to a vote, using the percentage of votes in favor as the dependent variable.\footnote{Since this analysis focuses on voted proposals, we also control for the voting recommendations of management and proxy advisors such as ISS (``For,'' ``Against,'' or ``Abstain'').} Panel (a) of Figure \ref{fig:ngo_support} shows that only NGO campaigns launched one or three months before the AGM increase support for related shareholder proposals, with an increase of nearly 2 percentage points. Given the generally low baseline support for these proposals (24\%), this corresponds to an 8\% increase in shareholder support when an NGO campaign addresses the same cause, on average.
%\textcolor{red}{XXX I don't understand this paragraph. Do we add more variables now all of the sudden? aren't these variables captured by the fixed effects? The part from "Note that..." feels detached from the rest of the paragraph. XXX} To understand why campaigns six months before the AGM are also somewhat successful, we estimate (\ref{eq:success}) on the subset of voted proposals, with the percentage of votes in favor as the dependent variable. Note that every proposal to be voted on at the AGM is accompanied by a vote recommendation by the management (``For'', ``Against'', or ``Abstain''). Institutional investors often follow proxy advisor recommendations, particularly from ISS \citep{shu2024proxy}. We include management and ISS recommendations as covariates in (\ref{eq:success}).
% Figure \ref{fig:ngo_support} shows that only NGO campaigns launched one or three months before the AGM increase support for related shareholder proposals, with an increase of nearly 2 percentage points. Given the generally low baseline support for these proposals (24\%), this corresponds to an 8\% increase in shareholder support when an NGO campaign addresses the same cause, on average.
In contrast to Panel (a), Panel (b) detects no significant increase in support for campaigns initiated as early as six (or five) months before the AGM. This suggests that the difference between the two figures arises from shareholder proposals being withdrawn after reaching an agreement with the firm. As a result of these early campaigns, the target firm may take actions to resolve the issue highlighted in both the campaign and the shareholder proposal, leading to an improvement in its E\&S score, which supports the finding in Section \ref{s:ES} that only early campaigns result in better E\&S scores.
Note that the subset of shareholder proposals going to a vote is not random; an argument raised, for instance, by \cite{flammer2015does}. Given the generally low support of shareholder proposals going to vote, managers may let proposals that are the least likely to succeed go to a vote. However, those votes can still be helpful in disciplining firms as large shareholder support to a cause may be hard to conceal for managers \citep{levit2011nonbinding}. This reinforces the importance of analyzing all E\&S proposals, including those withdrawn as in Panel (a).
The evidence in this section points to:
\begin{result}\label{res:3}
Pre-AGM campaigns raise shareholder support and proposal success, while AGM-day campaigns have little effect.
\end{result}
\section{Exposure vs. Influence}\label{s:reputation}
Sections \ref{s:media} and \ref{s:influence} outlined a tradeoff: campaigns launched close to an AGM maximize \textit{exposure}, while those launched earlier maximize \textit{influence}. In this section, we examine how NGOs navigate this tradeoff as their visibility increases. We first study whether greater past exposure leads NGOs to shift campaign timing from the AGM date to earlier months. We then ask whether such reputation translates into greater effectiveness at influencing shareholder proposals.
\subsection{The Timing of Past and Current Campaigns}
We first examine whether NGOs shift from targeting AGMs to launching earlier campaigns as they campaign more. Specifically, we estimate
\begin{equation}\label{eq:distance}
\begin{aligned}
\text{Distance to AGM}_{c(i,n,g,y,m)} &= \sum_{\tau=0.1}^{0.9} \beta_\tau \cdot \mathbf{1}_{\left\{ \text{Campaign \#}_{g,y-1} = \tau \right\}} \\
&\quad + \alpha_{t(\cdot),y} + \alpha_{g,y} + \alpha_{n,y} + u_{c(i,n,g,y,m)}.
\end{aligned}
\end{equation}
where the dependent variable measures the number of months between campaign $c(\cdot)$ and the targeted firm’s AGM, and the key explanatory variable is the sequence number of campaigns by NGO $g$ by year $y$. $\alpha_{t(\cdot),y}$, $\alpha_{g,y}$, and $\alpha_{n,y}$ are topic-by-year, NGO-by-year, and firm-by-year fixed effects, respectively.
%Fixed effects absorb NGO size, firm-specific shocks, and cause-specific shocks.
Identification thus comes from NGOs launching multiple campaigns in the same year.\footnote{Appendix Figure \ref{ngo_campaign_timing_robustness_fe} shows similar results when controlling only for firm and NGO fixed effects, so that the estimation of $\hat{\beta}_\tau$ come from variation across both campaigns and years.}
Panel (a) of Figure \ref{fig:timing_exp_gt} shows that first campaigns tend to occur right at the AGM, while subsequent ones are launched progressively earlier. This pattern is consistent with NGOs initially using AGMs to gain visibility and later exploiting their visibility to influence voting outcomes.%\footnote{Large negative coefficients for NGOs with many campaigns are driven by a few outliers with unusually high activity.}
These patterns are not driven by alternative explanations.
First, they are unlikely to reflect NGOs learning that campaigning on the AGM date does not affect voting, since most shareholders have already voted by proxy. Indeed, the same downward trend in campaign timing holds when focusing on NGOs whose first campaign occurred after the AGM (Appendix Figure \ref{ngo_campaign_timing_before_after}).
Second, they are not driven by selection of more ``influenceable'' targets: neither target size (market capitalization) nor ESG scores covary with NGO past AGM exposure (Appendix Figures \ref{ngo_campaign_experience_mc} and \ref{ngo_campaign_experience_esg}).
\subsection{Visibility and Campaign Timing}
We next use Google Trends scores to proxy for an NGO’s reputation. We compute the average search score for NGO $g$ in year $y$ relative to Greenpeace in year $y-1$, discretized into deciles and regress:
\begin{equation}\label{eq:distance_google}
\begin{aligned}
\text{Distance to AGM}_{c(i,n,g,y,m)} &= \sum_{\tau=0.1}^{0.9} \left( \beta_\tau \cdot \mathbf{1}_{\left\{ \text{Relative Google Score}_{g,y-1} = \tau \!\right\}} \right)\\
&\quad + \alpha_{t(\cdot)} + \alpha_{y} + u_{c(i,n,g,y,m)},
\end{aligned}
\end{equation}
which is similar to model \eqref{eq:distance_google} with $\text{Relative Google Score}_{g,y-1}$ as a measure of NGO visibility. $\alpha_{t(\cdot)}$ and $\alpha_{y}$ are topic and year fixed effects. Panel (b) of Figure \ref{fig:timing_exp_gt} shows that only the most visible NGOs---those in the top deciles of relative searches---campaign several months ahead of AGMs, while less visible NGOs continue to concentrate on AGM dates.
\medskip
Taken together, these results indicate a systematic shift: NGOs start by targeting AGMs to maximize exposure, and as their reputation grows, they move campaigns forward in time to increase their chances of influencing firm decisions.
\begin{result}\label{res:4}
NGOs initially target AGM dates to gain visibility. As their exposure increases, they advance campaigns by several months to influence the outcome of shareholder proposals.
\end{result}
\begin{figure}[t]
\centering
\caption{NGO reputation and campaign timing \label{fig:timing_exp_gt}}
\begin{subfigure}{0.49 \linewidth}
\centering\includegraphics[width=\textwidth]{Figures/campaign_timing_experience_exclude.pdf}
\caption{Later campaigns pre-date the target's AGM date \label{fig:timing_experience}}
\end{subfigure} \hfill
\begin{subfigure}{0.49 \linewidth}
\centering\includegraphics[width=\textwidth]{Figures/gtrends_campaign_timing.pdf}
\caption{More visible NGOs campaign earlier \label{fig:timing_gtrends}}
\end{subfigure}\hfill
\begin{minipage}{1 \textwidth} % choose width suitably
{\footnotesize Note: This figure plots the estimated coefficients from (\ref{eq:distance}). The dependent variable is the months between an NGO campaign and the targeted firm's AGM date. Standard errors are clustered at the NGO level. \par}
\end{minipage}
\end{figure}
\begin{comment}
\section{Exposure vs. Influence}\label{s:reputation}
Sections \ref{s:media} and \ref{s:influence} outline a tradeoff: late campaigns \textit{expose} a firm's wrongdoing to its stakeholders and boost the NGO's visibility, while early ones are more effective at \textit{directly influencing} a firm's decisions. How do NGOs navigate this tradeoff?
In this section, we first explore whether NGOs, after gaining visibility by targeting AGMs in the past, switch to campaigning before the AGM.
\subsection{NGO Experience and Campaign Timing}
Targeting AGM-dates could enahnce an NGO's visibility and its reputation for commitment to a cause. In this section, we explore whether NGOs, after accumulating some visibility, begin launching campaigns earlier---before the target firm’s AGM---to improve their chances of influencing the outcome as suggested in Section \ref{s:influence}.
We investigate this mechanism by:
\begin{equation}\label{eq:distance}
\begin{aligned}
\text{Distance to AGM}_{c(i,n,g),y} &= \sum_{\substack{\tau>1}} \beta_\tau \cdot \mathbf{1}{\{\text{Campaign \#}_{c(\cdot),g,y} = \tau\}} \\
&+ \alpha_{g,y} + \alpha_{n,y} + \alpha_{i,y} + u_{c(g,n,i),y},
\end{aligned}
\end{equation}
which examines ``$\text{Distance to AGM}_{c(i,n,g),y}$,'' the distance (in months) of campaign $c(\cdot)$ to the AGM of targeted firm $n$, on topic $i$, as a function of the number of past campaigns by NGO $g$, using an indicator, ``$\text{Campaign \#}_{c,g,y}$,'' that is one if campaign $c(\cdot)$ is the $\tau^{\text{th}}$ campaign of NGO $g$ in year $y$ and zero otherwise.
Since our campaign data start in 2010, we do not know how many campaigns an NGO launched before 2010. To control for this truncation as well as time varying confounders such as NGO size, we include NGO-by-year fixed effects. We also include firm-by-year and cause-by-year fixed effects to account for firm and cause-specific occurrences that, varying over time, may incentivize an NGO to launch a campaign. Hence, the identification of $\{\beta_\tau\}_\tau$ comes from multiple campaigns $c(\cdot)$ by the same NGO $g$ within year $y$.\footnote{In Appendix Figure \ref{ngo_campaign_timing_robustness_fe}, we present similar results when controlling only for firm and NGO fixed effects, so that the estimation of $\hat{\beta}_\tau$ come from variation across both campaigns and years.} Because of the scope of this analysis, we only consider NGOs with at least two campaigns, but the results still hold in the full sample. The standard errors are clustered at the NGO level.
To interpret the estimated $\{\hat{\beta}_\tau\}_\tau$ coefficients, consider that, on average, the first NGO campaign is exactly on the month of the AGM. Thus, Figure \ref{fig:timing_experience} indicates that as the NGO's experience increases, the new campaigns are launched before the AGM date.\footnote{The large negative coefficients estimated for NGOs with many campaigns---e.g., more than 9---are driven by outlier NGOs with several campaigns per year.} We postpone the interpretation of this result to Section \ref{s:discussion}, which sums up and discusses the results.
\vspace{1ex}
\noindent\textbf{Alternative mechanism---learning.} This result is unlikely due to NGOs learning that campaigning on the AGM date does not influence voting decisions, as most shareholders have already voted by proxy. Appendix Figure \ref{ngo_campaign_timing_before_after} shows the same downward trend observed in Figure \ref{fig:timing_experience} when subsetting to NGOs whose first campaign took place within 2 weeks \textit{after} the AGM, when the NGO surely knew that it could not affect votes.
\vspace{1ex}
\noindent\textbf{Alternative mechanism---selection.} NGO do not increasingly launch campaigns against firms that may be more easily influenced over time. Appendix Figures \ref{ngo_campaign_experience_mc} and \ref{ngo_campaign_experience_esg} exclude selection on observables. For instance, target sizes do not covary (market capitalization) with NGO experience. In addition, environmental (social) scores do not covary with an NGO's experience in environmental (social) campaigns.
\subsection{Reputation and Campaign Timing}
We further validate the link between past exposure and AGM timing by estimating a similar specification to \eqref{eq:distance} using \textit{between-NGO} relative Google search scores as a proxy for reputation. Specifically, we measure NGO $g$’s reputation in year $y$ as its average Google Trends score relative to Greenpeace in year $y-1$ with the variable $\text{Relative Google Score}_{g,y-1}$ which is discretized across deciles. For example, an NGO with an average relative score of 45 in 2015 is assigned a reputation of 0.45 in 2016 and $\text{Relative Google Score}_{g,y-1}=1$ for the fifth $\tau$ decile. Formally, we estimate
\begin{equation}\label{eq:distance_google}
\begin{aligned}
\text{Distance to AGM}_{c(i,n,g),y} &= \sum_{\tau=0.1}^{0.9} \beta_\tau \cdot \mathbf{1}\!\left\{ \text{Relative Google Score}_{g,y-1} = \tau \right\} \\
&\quad + \alpha_{t(\cdot)} + \alpha_{y} + u_{c(g,n,i),y}.
\end{aligned}
\end{equation}
Figure~\ref{fig:timing_gtrends} shows that only the most visible NGOs campaign ahead of the target's AGM date. In particular, NGOs with average Google search scores between 80\% and 90\% of Greenpeace’s score tend to launch campaigns four to six months before the meeting.
Therefore, we conclude the following:
\begin{result}\label{res:4}
NGOs with little reputation or experience target AGM dates to gain visibility. They schedule subsequent campaigns a few months before AGM dates to influence the outcome of shareholder proposals.
\end{result}
\end{comment}
\subsection{NGO Reputation and AGM Influence}\label{sec:success}
We now ask whether NGO reputation, as measured in the previous sections, translates into greater influence over shareholder proposals. To test this, we estimate
\begin{equation}
\begin{aligned}\label{eq:main_reg}
\text{Success}_{i,n,y} &= \beta_1 \, \text{Campaign on Same Cause}_{c(i,n,g,y,m)} + \beta_2 \,\text{\# Past Campaigns}_{g,y-1} \\
&+ \beta_3\, \text{Campaign on Same Cause}_{c(i,n,g,y,m)} \cdot \text{\# Past Campaigns}_{g,y-1} \\
&+ \alpha_{n,y} + \alpha_{g,y} +\alpha_{i,y} + \alpha_{m(c(i,n,g))} + u_{i,n,y},
\end{aligned}
\end{equation}
\sloppy where ``$\text{Success}_{i,n,y}$'' indicates whether a proposal is withdrawn or passes a vote, as in equation \eqref{eq:success}. The key variable ``$\text{Campaign on Same Cause}_{i,n,y}$'' equals one if NGO $g$ campaigned against firm $n$ on cause $i$ in the ten months preceding the AGM. ``$\text{\# Past Campaigns}_{g,y-1}$'' counts NGO $g$’s past campaigns. We absorb firm-by-year, NGO-by-year, cause-by-year, and month variation through fixed effects, and cluster standard errors at the firm level.
\input{Tables/ngo_experience_success_fe}
Column (1) of Table \ref{ngo_experience_success} shows that campaigning on the same cause as a shareholder proposal has little effect on its own ($\beta_1 \approx 0$). What matters is the interaction: proposals are more likely to succeed when the campaigning NGO has already campaigned ($\beta_3 > 0$). A one-standard-deviation increase in past campaigns on the same cause raises proposal success by about 8\% on average. The regression also includes the squared term of ``$\text{\# Past Campaigns}_{g,y-1}$,'' showing decreasing marginal returns of campaigning.
We unpack the mechanism behind these coefficient estimates by distinguishing between prior AGM campaigns and non-AGM campaigns in Column (2). We find that only the former contribute to proposal success. In fact, one-standard-deviation more AGM campaigns increases the probability of success by roughly 4.5 percentage points (about 15\%), while non-AGM campaigns have no measurable effect.
Column (3) tests for diminishing returns by including squared terms. The results confirm a concave pattern: additional AGM campaigns continue to improve proposal success, but at a decreasing rate. Importantly, the combined effect remains non-negative even for the most active NGOs.\footnote{The marginal effect of AGM campaigning never turns negative in our sample since no firm launches more than five AGM-date campaigns, so our proxy for reputation does not imply adverse returns.}
We corroborate these findings with an alternative proxy: NGO visibility measured by Google Trends. Specifically, Column (4) replaces ``$\text{\# Past Campaigns}_{g,y-1}$'' by ``\text{Google Search Popularity$_{g,y-1}$},'' which we define as the average search score of NGO $g$ relative to Greenpeace in year $y-1$.\footnote{The table drops the direct effect of this variable on the dependent variable, as it is absorbed by NGO-by-Year fixed effects.} The interaction between this measure and campaigning on the same cause is positive, with a concave shape. Thus, both proxies---past AGM campaigns and Google visibility---suggest that reputation enhances the effectiveness of NGO campaigns, albeit with diminishing marginal returns.
These results support the following result:
\begin{result}\label{res:5}
Targeting AGMs builds NGO reputation and increases their ability to raise the support of related shareholder proposals, but with diminishing returns.
\end{result}
\section{Discussion and Conclusion}\label{s:discussion}
Examining NGO campaign timing, we show that not just the decision to intervene but also its timing can materially affect firm behavior and shape the course of future activism. Over time, NGOs shift from visibility-driven \textit{exposure} tactics to those designed to \textit{influence} corporate strategies, highlighting how the temporal dimension of their actions channels pressure that leads firms to internalize social and environmental externalities.
We first show that NGOs’ early campaigns often coincide with the target’s AGM date, when media attention is highest, and that such timing significantly increases their visibility (Result~\ref{res:1}). These visibility gains are concentrated in NGOs’ first few campaigns, indicating that AGM-day actions serve primarily to build name recognition rather than to influence firms directly. By contrast, later AGM-day campaigns have no immediate effect on firms’ E\&S scores, which improve only when campaigns are launched ahead of the AGM (Result~\ref{res:2}).
Hence, campaigning before the AGM offers direct advantages. When an NGO targets a firm on an issue for which a shareholder proposal has already been filed, it boosts the proposal’s chance of success—often leading to its withdrawal from the ballot—and raises average shareholder support by 8\% (Result~\ref{res:3}). Given the typically low baseline support for such proposals, early NGO interventions can effectively discipline managers who must contend with heightened shareholder pressure \citep{levit2011nonbinding}. Importantly, all analyses include NGO-by-year and cause-by-year fixed effects, thereby accounting for time-varying confounders at both the NGO level—such as experience or reputation—and the cause level—such as specific news events. This ensures that globally renowned NGOs (e.g., PETA) do not drive the results. Furthermore, the analyses in Sections~\ref{s:media} and \ref{s:reputation} show little evidence of selection on observables, alleviating concerns that NGOs disproportionately target firms that are easier to influence or more likely to generate visibility.
Our results suggest that NGOs face a trade-off that they navigate by adapting their strategies over time. NGOs with limited activism histories (or low visibility) tend to campaign on the AGM date, whereas more established NGOs prefer targeting firms just before the AGM to directly influence shareholder votes (Result~\ref{res:4}).
As a result, having targeted AGMs in the past increases the likelihood that an NGO can help a shareholder proposal pass related to its campaign. However, repeatedly targeting AGMs leads to weaker influence over firm behavior (Result~\ref{res:5}).
One interpretation is that NGOs, having already demonstrated commitment to their cause, face weaker incentives to keep signaling through costly AGM-day campaigns. In this view, declining returns simply reflect a reduced need for visibility once reputational capital is established.
Yet the evidence points to a different mechanism. Within the same NGO, AGM-day campaigns progressively lose their marginal effect on media attention (Figure~\ref{fig:gtrends_campaigns}) and their ability to damage a firm’s reputation (Appendix Figure~\ref{fig:media_coverage_experience}), suggesting that the audience’s response—rather than the NGO’s effort—changes over time. As NGOs become recognized and credible actors, stakeholders interpret their campaigns differently: what initially signals toughness later signals redundancy or routine. Established NGOs are expected to engage in more constructive dialogue, such as supporting shareholder proposals, rather than resorting to public confrontation. Consequently, the same action produces less reputational harm for firms and fewer visibility gains for NGOs.
The diminishing returns to AGM targeting thus stem from evolving stakeholder expectations and a changing perception of NGO legitimacy, rather than from reduced internal pressure to seek exposure.\footnote{This interpretation does not rule out signaling motives altogether. Early AGM-day campaigns are consistent with “reputation-for-toughness’’ models, in which actors bear short-term costs to establish credibility that can be leveraged in future interactions \citep{kreps1982reputation,milgrom1982limit}.}
Other interpretations are unlikely to be relevant unless they covary with the number of AGM-date campaigns within a given year, NGO, firm, and cause. For instance, organizational aging (e.g., older NGOs becoming larger and more complex) or experience in managing AGM-related information is accounted for through NGO-by-year fixed effects. These interpretations open avenues for future research, contributing to the broader discussion on civil society lobbying and activism \citep{bertrand2021hall,bertrand2023investing}.
In conclusion, NGOs without an established reputation coordinate stakeholders over time by campaigning on AGM dates, drawing attention to issues that spur future engagement. Once reputable, they coordinate stakeholders within the same AGM cycle by supporting existing shareholder proposals and directly shaping firm decisions. Thus, NGOs act as a coordination device throughout their lifecycle, with the form of coordination evolving with their perceived toughness and credibility. In this way, NGOs’ incentives shape their strategies and, ultimately, the social and environmental impact of firms.
We also acknowledge that NGOs increasingly engage not only in campaigns but also in collaborations with corporations \citep{dahan2010corporate}. Future research could examine when confrontation shifts to cooperation and how such collaboration reshapes the balance between visibility, credibility, and influence.
% \singlespacing
\bibliographystyle{ecca-mod}
\bibliography{covid_bibliography}
\appendix
\section*{\Huge{Online Appendix}}
\section{Omitted Tables}
\setcounter{table}{0}
\renewcommand{\thetable}{A\arabic{table}}
\setcounter{equation}{0}
\renewcommand{\theequation}{A\arabic{equation}}
\setcounter{figure}{0}
\renewcommand{\thefigure}{A\arabic{figure}}
\input{Tables/chatgpt}
%\begin{landscape}
% \input{Tables/agm_selection}
%\end{landscape}
%\input{Tables/regs_experience}
% \input{Tables/ngo_experience_coverage}
\clearpage
\section{Omitted Figures}
\setcounter{table}{0}
\renewcommand{\thetable}{B\arabic{table}}
\setcounter{equation}{0}
\renewcommand{\theequation}{B\arabic{equation}}
\setcounter{figure}{0}
\renewcommand{\thefigure}{B\arabic{figure}}
% \begin{figure}
% \caption{Greater media coverage around AGM periods (red) than in other periods (blue) \label{fig:news_count_agm}}
% \centering
% \begin{minipage}{0.48\textwidth}
% \centering
% \includegraphics[width=\linewidth]{Figures/sum_stat_news_month.pdf}
% \subcaption{AGM Period defined as AGM Month}
% \end{minipage}\hfill
% \begin{minipage}{0.48\textwidth}
% \centering
% \includegraphics[width=\linewidth]{Figures/sum_stat_news_3m.pdf}
% \subcaption{AGM Period defined as Within 3 Months}
% \end{minipage}
% \begin{minipage}{1 \textwidth} % choose width suitably
% {\footnotesize Note: News coverage is much higher during AGM periods (red bars) than during other periods (blue bars). Extending the AGM period from a window of $\pm$ ten days (Panel a) to $\pm$ three months (Panel b), the distance between the two columns decreases, as the gap between news coverage during AGM periods and other periods is washed out by a greater interval. %However, this gap has stayed positive for all the years. In each panel, the columns compute the average weekly number of news for 4,676 firms listed in the US from January 2010 to December 2019, depending on whether the nearest AGM is less than ten days (Panel a) or three months (Panel b) away.
% News data come from Ravenpack. AGM dates come from ISS. \par}
% \end{minipage}
% \end{figure}
% \begin{figure}[h]
% \centering
% \caption{Count of E\&S shareholder proposals and their success \label{fig:sum_stat_proposals}}
% \centering
% \includegraphics[width=0.75\linewidth]{Figures/sum_stat_proposals.pdf}
% \begin{minipage}{\textwidth} % choose width suitably
% {\footnotesize Note: This figure plots the number of E\&S proposals at Russell 3,000 firms from 2006 to 2021 (solid line) and the number of successful E\&S proposals (blue area) in the same period. A successful proposal is defined as a proposal that is either withdrawn before the AGM, or that goes to vote and passes. Data from ISS. Based on 5,937 shareholder proposals filed at 979 firms. \par}
% \end{minipage}
% \end{figure}
\begin{figure}[h!]
\centering
\caption{Persistence of the AGM month over time}
\label{fig:agm_persistence}
\includegraphics[width=0.75\textwidth]{Figures/agm_month.pdf}
\begin{minipage}{1 \textwidth} % choose width suitably
{\footnotesize Note: This figure plots the percentage of AGMs happening in each month of the year (y-axis), given the month of the previous year's AGM (x-axis). A larger and darker circle indicates a greater proportion of meetings. Plot based on 38,471 meetings of 5,745 US companies, from 2010 to 2020. \par}
\end{minipage}
\end{figure}
% \begin{figure}[h!]
% \centering
% \caption{Testing NGO campaign timing around the AGM date \label{fig:cattaneo}}
% \vspace{-0.15cm}
% \includegraphics[width = 0.75\textwidth]{Figures/density_test.pdf}
% \begin{minipage}{1 \textwidth} % choose width suitably
% {\footnotesize Note: This figure plots the density test for the occurrence of NGO campaigns around firms' AGM, following \cite{cattaneo2020simple}. The test is based on 521 NGO campaigns launched within 50 days of the firm's closest AGM. The NGO data come from Sigwatch; the AGM dates are from ISS. \par}
% \end{minipage}
% \end{figure}
\begin{figure}
\centering
\caption{Firms’ negative media coverage increases if NGOs campaign during AGMs}
\label{fig:ngo_agm_bad_news}
\includegraphics[width=0.75\linewidth]{Figures/ngo_agm_bad_news_weekly.pdf}
\begin{minipage}{1 \textwidth} % choose width suitably
{\footnotesize Note: The figure plots the coefficients $\beta_\tau$ from the following event study regression where we re-calendarize time around an AGM date so that the AGM month is month 0 in each year for each firm:
\begin{equation*}
\begin{aligned}
News_{n,w,y} &=
\sum_{\substack{\tau=-5 \\ \tau \neq -4}}^{5}
\!\left( \beta_{\tau} \cdot 1\!\left\{ M_{n,w,y} = \tau \!\right\} \cdot NGO\,Campaign_{n,w,y} \!\right) \\
&\quad + X_{n,w,y}\gamma' + \alpha_{n,y} + \alpha_{w,y,s(n)} + u_{n,w,y},
\end{aligned}
\end{equation*}
where $News_{n,w,y}$ is an indicator for whether firm $n$ in week $w$ of year $y$ experiences at least one negative E\&S news event. The set of dummies $1\{ M_{n,w,y} = \tau \}$ indicate the distance in months from the closest AGM. The function $M_{n,w,y}$ assigns week $w$ to each month; the indicator is 1 if the week falls in
month $\tau$, where a month is made up of 4 weeks, and zero otherwise. We interact these event-study indicators, with the variable $NGO\,Campaign_{n,w,y}$, an indicator variable equal to 1 if firm n experienced an NGO campaign in week $w$ of year $y$, and 0 otherwise. The vector $X_{n,w,y}$ includes the direct effect of these variables. The reference month is $\tau =-4$ following SEC rules on shareholder proposals. Note that, because of the stability of AGM dates across years within firms (Appendix Figure \ref{fig:agm_persistence}), the distribution of months fixed effects is unaffected by the re-calendarization, so that months are not double counted across years. Standard errors are clustered by firm.
\par}
\end{minipage}
\end{figure}
%\begin{figure}
% \centering
% \caption{Examples of positive and negative E\&S news}
% \label{fig:news_examples}
% \begin{subfigure}[t]{0.9\textwidth} % Use a fraction instead of \textwidth directly
% \centering
% \caption{Example of a positive E or S news}
% \includegraphics[width=0.9\linewidth]{Figures/company_example_positive.png}
% \label{fig:sub1}
% \end{subfigure}
% \hskip 3cm % Or remove this if not needed
% \begin{subfigure}[t]{0.9\textwidth}
% \centering
% \caption{Example of a negative E or S news}
% \includegraphics[width=0.9\linewidth]{Figures/company_example_negative.png}
% \label{fig:sub2} % Fix duplicate label
% \end{subfigure}
% \begin{minipage}{0.9\textwidth} % Use a proper fraction
% {\footnotesize Notes: This figure displays examples of a positive (top) and negative (bottom) E\&S news. News data from RavenPack. \par}
% \end{minipage}
%\end{figure}
%\begin{figure}[t]
% \centering
% \caption{Media sentiment of NGO campaigns around an AGM}
% \includegraphics[width=0.8\textwidth]{Figures/ngo_agm_sentiment.pdf}
% \label{fig:media_sentiment}
% \begin{minipage}{1 \textwidth} % choose width suitably
%{\footnotesize Note:The figure presents estimates from (\ref{eq:proximity}). The dependent variable is the average sentiment of E\&S news received by firm in a week. %The sample period goes from January 2010 to December 2019 and covers 4,676 firms. NGO campaign data from Sigwatch. AGM data from ISS. News data from Ravenpack. Industry data from Compustat.
%Standard errors are clustered at the firm level. %We regress the number of negative E\&S news about firm $f$ in month $t$ (capped at 150 to avoid outliers) on a dummy indicating the publication of a negative NGO report about firm $f$ in month $t$, a dummy equal to one when month $t$ is $k$ months away from the nearest AGM, where we take $k$ to vary between -5 and +5 months, the interaction between these two dummy variables, and firm and month by industry fixed effects. Month -5 and +5 are ``one and a half'' months, so the year still consists of 12 full months. The figure also reports the 95\% confidence interval when the standard errors are clustered at the level of the firm. The sample period goes from January 2010 to December 2019 and covers 4,676 firms. NGO campaign data from Sigwatch. AGM data from ISS. News data from Ravenpack. Industry data from Compustat. The regression also includes firm and year-by-industry fixed effects.
%\par}
%\end{minipage}
%\end{figure}
\begin{figure}[h!]
\centering
\caption{NGO campaigns and shareholder proposals}
\includegraphics[width=0.8\textwidth]{Figures/campaign_proposal.pdf}
\label{fig:campaign_proposal}
\begin{minipage}{1 \textwidth} % choose width suitably
{\footnotesize Note: This figure plots the estimated coefficients of the regression of a dummy indicating the occurrence of a shareholder proposal on month $t$ at firm $f$, relative to an NGO campaign targeting firm $f$ on the same cause, $k$ months ahead. The regression also includes campaign and month-by-year-by-industry-by-topic fixed effects. The regression is based on 1,227 NGO campaigns. Standard errors are clustered at the firm level. \par}
\end{minipage}
\end{figure}
\begin{figure}
\centering
\caption{NGO campaign timing and consumer backlash}
\label{fig:campaign_sales}
\begin{subfigure}[t]{0.49\textwidth} % Use a fraction instead of \textwidth directly
\centering
\includegraphics[width=\textwidth]{Figures/campaign_agm_sales_inter.pdf}
\caption{NGO Campaigns on AGM dates}
\label{fig:sub1}
\end{subfigure}
\begin{subfigure}[t]{0.49\textwidth}
\centering
\includegraphics[width=\textwidth]{Figures/campaign_agm_sales_no_agm.pdf}
\caption{NGO Campaigns on Other Days}
\label{fig:sub2} % Fix duplicate label
\end{subfigure}
\begin{minipage}{\textwidth} % Use a proper fraction
{\footnotesize Notes: The figure plots the estimated ${\beta}_\tau$ from \eqref{eq:sales}. Panel (a) focuses on NGO campaigns on AGM dates, while Panel (b) focuses on campaigns on dates other than AGM dates.
% \begin{equation*}
% \begin{aligned}
% \frac{\text{Sales}_{n,q}}{\text{Assets}_{n,q-1}} & = \sum^{\tau=5}_{\substack{\tau=-4 \\ \tau\neq -1}} \left(\beta_\tau \cdot \text{NGO Campaign on AGM}_{n} \cdot \mathbf{1}{\{\text{distance}=\tau\}}\right) \\
% & + \sum^{\tau=5}_{\substack{\tau=-4 \\ \tau\neq -1}} \left(\tilde{\beta}_\tau \cdot \mathbf{1}{\{\text{distance}=\tau\}}\right) + \alpha_n +\alpha_{s(n),q} + u_{n,q},
% \end{aligned}
% \end{equation*}
% which interacts a time dummy, the occurrence of a campaign, and a dummy indicating that the campaign happened on AGM day. Panel (b) plots the estimated coefficients of the regression of sales over (lagged) total assets, on a set of time dummies interacted with a dummy indicating the occurrence of an NGO campaign targeting firm $n$ from (\ref{eq:sales}). The regressions also include firm-by-year and quarter-by-industry fixed effects.
The reference period is -1 quarter, consistent with the other analyses. The number of observations is 239,818 in Panel (a) and 239,925 in Panel (b). Standard errors are clustered at the firm level. \par}
\end{minipage}
\end{figure}
\begin{figure}
\centering
\caption{Campaign Timing and E\&S Score Improvement}
\label{fig:campaign_esg_score}
\includegraphics[width=0.75\linewidth]{Figures/campaign_esg_score.pdf}
\begin{minipage}{1 \textwidth} % choose width suitably
{\footnotesize Note: This figure plots the estimated coefficients of the following regression:
\begin{equation}\label{eq:ES}
\Delta_5 \text{E\&S Score}_{n,q}=\sum_{l=-6}^5 \text{Campaign}_{n,q,l}+X_{n,q}\gamma'+\alpha_n+\alpha_{s(n),q}+u_{n,q},
\end{equation}
where $\Delta_5 \text{E\&S Score}_{i,q}$ is the change in firm $n$'s average of its Environmental and Social scores between quarter $q-1$ and quarter $q+4$, and $\sum_{l=-6}^6 \text{Campaign}_{n,q,l}$ is a set of dummy variables indicating the occurrence of an NGO campaign against firm $n$ in quarter $q$, being $l$ months away from the closest AGM. The regression also controls ($X_{n,q}$) for firm size (logged total assets), lagged average of E and S scores, firm fixed effects ($\alpha_n$), and industry-by-quarter fixed effects ($\alpha_{s(n),q}$). ESG scores are from MSCI, NGO campaigns data from Sigwatch, firm characteristics from Compustat, and AGM dates from ISS. The regression has 47,082 observations. Standard errors are clustered at the firm level.
\vspace{1ex} Unlike an event study, equation (\ref{eq:ES}) is a distributed leads-and-lags regression. We omit a reference period, so the coefficients are interpreted relative to firms that did not experience a campaign. This choice allows us to directly compare the effects of early campaigns with those launched on the AGM date relative to firms without campaigns. Results remain similar if we instead set month –4 as the reference period, as in the other analyses.
\par}
\end{minipage}
\end{figure}
\begin{figure}
\caption{NGO past campaigns and the timing of current NGO campaigns -- Alternative set of Fixed Effects} \label{ngo_campaign_timing_robustness_fe}
\centering
\begin{subfigure}[t]{0.8\textwidth}
\centering
% \caption{NGOs whose first campaign was after an AGM}
\includegraphics[width= \linewidth]{Figures/campaign_timing_experience_robust_fe.pdf}
\end{subfigure}
\begin{minipage}{1\textwidth}% choose width suitably
{\footnotesize Note: This figure plots the estimated coefficients from the regressions of the distance between NGO campaigns and the targeted firm's closest AGM (in months) on dummies indicating the number of campaigns the focal NGO had already launched. The regression includes replaces NGO-by-year fixed effects with separate fixed effects for each NGO and each year. Hence the coefficients $\{\hat{\beta}_\tau\}_\tau$ in (\ref{eq:distance}) is also identified from variation in the numbers of NGO campaigns over years. Standard errors are clustered at the NGO level. \par}
\end{minipage}
\end{figure}
\begin{figure}
\caption{Campaign timing for NGOs whose first campaign is after the AGM} \label{ngo_campaign_timing_before_after}
\centering
\begin{subfigure}[t]{0.8\textwidth}
\centering
% \caption{NGOs whose first campaign was after an AGM}
\includegraphics[width= \linewidth]{Figures/campaign_timing_experience_after_agm.pdf}
\end{subfigure}
\begin{minipage}{1\textwidth}% choose width suitably
{\footnotesize Note: This figure plots the estimated coefficients from the regressions (\ref{eq:distance}) of the distance between NGO campaigns and the targeted firm's closest AGM (in months) on dummies indicating the number of past campaigns of the focal NGO. The panel focuses on NGOs whose first campaign was within the two weeks following the AGM. Standard errors are clustered at the NGO level. \par}
\end{minipage}
\end{figure}
\begin{figure}
\caption{Past campaigns and target firm's size}
\centering
\includegraphics[width=0.8\linewidth]{Figures/campaign_experience_mc.pdf}
\label{ngo_campaign_experience_mc}
\begin{minipage}{1\textwidth}% choose width suitably
{\footnotesize Note: This figure plots the estimated coefficients from regressing the target firm's market cap (in deciles to control for the fact that the tails of the market cap distribution have thickened in recent years) on dummies indicating the number of campaigns the focal NGO had launched in previous years as in (\ref{eq:distance}). Standard errors are clustered at the firm level. The number of observations is 1,338. \par}
\end{minipage}
\end{figure}
\begin{figure}
\caption{Past campaigns and target firm's E\&S characteristics} \label{ngo_campaign_experience_esg}
\centering
\begin{subfigure}[t]{0.8\textwidth}
\centering
\caption{Target firm's E score for Environmental campaigns}
\includegraphics[width= \linewidth]{Figures/campaign_experience_e_score.pdf}
\end{subfigure}
\begin{subfigure}[t]{0.8\textwidth}
\centering
\caption{Target firm's S score for Social campaigns}
\includegraphics[width= \linewidth]{Figures/campaign_experience_s_score.pdf}
\end{subfigure}
\begin{minipage}{1\textwidth}% choose width suitably
{\footnotesize Note: This figure plots the estimated coefficients from regressing either a target firm's environmental score (top panel) or its social score (bottom panel) on dummies indicating either the number of environmental (top panel) or social (bottom panel) campaigns the focal NGO had launched in previous years as in (\ref{eq:distance}). Standard errors are clustered at the firm level. The number of observations is 636 and 698, respectively. \par}
\end{minipage}
\end{figure}
\begin{figure}
\caption{Media coverage of NGO-target firms around their AGM}
\label{fig:media_coverage_experience}
\centering
\includegraphics[width=0.9\linewidth]{Figures/ngo_agm_bad_news_experience.pdf}
\begin{minipage}{1\textwidth}% choose width suitably
{\footnotesize Note: To test if media exposure of NGO campaigns is highest during the AGM week, we aggregate negative E\&S news by firm ($n$) and week-by-year ($w,y$) and perform the event study:
\begin{align*}
\text{News}_{n,w,y} &=
\sum_{\substack{\tau=-5 \\ \tau \neq -4}}^{5}
\left( \beta_{\tau} \cdot \mathbf{1}_{\{M_{n,w,y} = \tau\}} \cdot \text{NGO Campaign}_{n,w,y} \right) \\
&\quad + X_{n,w,y}\gamma' + \alpha_{n,y} + \alpha_{w,y,s(n)} + u_{n,w,y},
\end{align*}
where ``$\text{News}_{n,w,y}$'' is an indicator for whether firm $n$ in week $w$ of year $y$ experiences at least one negative E\&S news event. The independent variables include a set of dummies ``$\mathbf{1}_{\{M_{n,w,y} = \tau\}}$'' indicating the distance in months from the closest AGM. The function ``$M_{n,w,y}$'' assigns week $w$ to each month; the indicator is 1 if the week falls in month $\tau$, where a month is made up of 4 weeks, and zero otherwise. We interact these event-study indicators, with the variable ``$\text{NGO Campaign}_{n,w,y}$'' an indicator variable equal to 1 if firm n experienced an NGO campaign in week $w$ of year $y$, and 0 otherwise. The vector ``$X_{n,w,y}$'' includes ``$\mathbf{1}_{\{M_{n,w,y} = \tau\}}$'' and ``$\text{NGO Campaign}_{n,w,y}$'' separately. We also include firm-by-year ($\alpha_{n,y}$) and industry-by-week-by-year fixed effects($\alpha_{w,y,s(n)}$). The reference month is $\tau =-4$ following SEC rules on shareholder proposals. Blue dots are estimated using campaigns that are among the first three for each NGO, while red dots are estimated using later campaigns. Standard errors are clustered at the firm level.
\par}
\end{minipage}
\end{figure}
\clearpage
\section{Robustness of Google Trends Results}\label{apndx:specs}
\setcounter{table}{0}
\renewcommand{\thetable}{C\arabic{table}}
\setcounter{equation}{0}
\renewcommand{\theequation}{C\arabic{equation}}
\setcounter{figure}{0}
\renewcommand{\thefigure}{C\arabic{figure}}
This section discusses the robustness of the results presented in Figure \ref{fig:gtrends_campaigns} to alternative specifications. In particular, we replicate the analysis using (i) a staggered difference-in-differences estimator \citep{sun2021estimating}, which accounts for heterogeneous campaign timing across NGOs, and (ii) a distributed lags specification, which controls for the possibility of overlapping campaigns within the same year.
\subsection{Staggered Difference-in-Differences Specification}
We first replicate Figure \ref{fig:gtrends_campaigns}, applying the staggered difference-in-differences approach to account for heterogeneous campaign timing across NGOs \citep{sun2021estimating}.
Estimated coefficients $\beta_{\tau}^{AGM}$ are reported in Figure \ref{fig:gtrends_campaigns_sunab_first_later}. The results are consistent with our baseline findings: AGM-timed campaigns (blue dots) generate sizeable visibility gains for early campaigns, raising Google search intensity by about 58\%. By contrast, NGOs see a more modest increase (33\%) in search intensity for later campaigns, and only in the month of the campaign.
\begin{figure}[h]
\centering
\caption{Staggered difference-in-differences specification}
\label{fig:gtrends_campaigns_sunab_first_later}
% \begin{subfigure}{0.49 \linewidth}
% \centering\includegraphics[width=\textwidth]{Figures/gtrends_campaigns_staggered.pdf}
% \caption{Main specification}
% \end{subfigure} \hfill
% \begin{subfigure}{0.49 \linewidth}
\centering\includegraphics[width= 0.75\textwidth]{Figures/gtrends_campaigns_sunab_first_later.pdf}
% \caption{\cite{sun2021estimating} specification}
% \end{subfigure}\hfill
\begin{minipage}{1 \textwidth} % choose width suitably
{\footnotesize Note: The Figure plots the estimated $\beta^{AGM}$ coefficients from \eqref{eq:gt_model} using a staggered difference-in-differences estimator \citep{sun2021estimating}. The dependent variable is the monthly Google Trends score for the NGO. Blue dots are estimated by restricting the sample to an NGO’s first three campaigns, while red triangles consider later campaigns. Standard errors are clustered at the campaign level.}
\end{minipage}
\end{figure}
\subsection{Distributed Lags Specification}
In the baseline event-study, if an NGO launches several campaigns in close succession, their effects may overlap in calendar time. For example, if one campaign takes place at $t=0$ and another at $t=2$, the months around $t=2$ capture the lingering impact of both events. This overlap can bias the estimated effects in the previous analyses. To address this, we estimate a distributed lags specification that explicitly includes \textit{leads} and \textit{lags} of both AGM and non-AGM campaigns:
\begin{equation}
\begin{aligned}\label{eq:dist_lag}
\text{Google Trends Score}_{g,m,y} & = \sum^{\tau=5}_{\substack{\tau=-5}} \left(\beta_{\tau}^{AGM} \cdot \text{AGM Campaign}_{g,m+\tau} \right) \\
&\quad + \sum^{\tau=5}_{\substack{\tau=-5}} \left(\beta_{\tau}^{Non-AGM} \cdot \text{Non-AGM Campaign}_{g,m+\tau}\right)
\\ & \quad + \alpha_{m,y} + \alpha_{g,y}+ u_{g,m,y},
\end{aligned}
\end{equation}
By doing so, the regression attributes observed variation in search activity in each month to the full sequence of campaigns occurring in that window, rather than assigning the entire effect to whichever campaign happens to be closest in event time. Put differently, the distributed lags approach disentangles the impact of overlapping campaigns by allowing each campaign to contribute separately to visibility over several months. This ensures that the estimated $\beta^{AGM}_\tau$ captures the marginal visibility effect of a single campaign, net of any other campaigns that may have occurred nearby.
\begin{figure}[!htb]
\centering
\caption{Distributed lags specification}
\label{fig:gtrends_campaigns_distr_lags}
\begin{subfigure}{0.49 \linewidth}
\centering\includegraphics[width=\textwidth]{Figures/gtrends_first_campaigns_distr_lags.pdf}
\caption{First three campaigns}
\end{subfigure} \hfill
\begin{subfigure}{0.49 \linewidth}
\centering\includegraphics[width=\textwidth]{Figures/gtrends_later_campaigns_distr_lags.pdf}
\caption{Later campaigns}
\end{subfigure}\hfill
\begin{minipage}{1 \textwidth} % choose width suitably
{\footnotesize Note: The two plots present the estimated $\beta^{AGM}$ with blue circles and $\beta^{Non-AGM}$ with red triangles coefficients from \eqref{eq:dist_lag} on two different samples. The first panel subsets the data to each NGO's first three campaigns, while the second panel examines later campaings. The dependent variable is the monthly Google Trends score for the NGO. Standard errors are clustered at the campaign level.}
%The sample period goes from January 2010 to December 2019 and covers 4,676 firms. NGO campaign data from Sigwatch. AGM data from ISS. News data from Ravenpack. Industry data from Compustat.
\end{minipage}
\end{figure}
Figure \ref{fig:gtrends_campaigns_distr_lags} reports the estimated coefficients. When restricting attention to an NGO’s first three campaigns, the results confirm that AGM-timed campaigns lead to significant short-run increases in visibility (blue dots), while later campaigns generate little or no effect. Whether early or late, non-AGM campaings have a limited effect on an NGO's visibility.
Overall, these robustness checks confirm that our main results are not driven by campaign timing heterogeneity across NGOs or by overlapping campaigns within a year.
\clearpage
\section{Matching Proposals And NGO Campaigns\label{apndx:match}}
\setcounter{table}{0}
\renewcommand{\thetable}{D\arabic{table}}
\setcounter{equation}{0}
\renewcommand{\theequation}{D\arabic{equation}}
\setcounter{figure}{0}
\renewcommand{\thefigure}{D\arabic{figure}}
This section outlines the process of matching each shareholder proposal in the ISS dataset with a relevant NGO campaign cause.
Sigwatch assigns a``Generic Topic" to each campaign in their dataset. To streamline the analysis, we combine certain causes when they are difficult to distinguish or when one category has few campaigns and can reasonably fit within a broader category. Specifically, we merge ``Labor Rights" with ``Human \& Labor Rights," ``Children" with "Human \& Labor Rights," ``CSR / ESG" with ``CSR," and ``Biodiversity" and ``Wildlife'' with "Environment." For campaigns under the ``Safety" cause, which includes only 12 entries, we examine each campaign’s specific purpose and classify it as either ``Product Safety" or ``Food Safety."
Using these consolidated ``Generic Cause Names'', we then categorize each shareholder proposal based on its short description. When a shareholder proposal is submitted, it includes a brief, one-sentence description, which is provided in the ISS dataset and serves as the foundation for cause classification. We start by assigning proposals to causes based on keywords in their descriptions and then manually classify any remaining proposals.
\vspace*{\fill}
\input{Tables/proposals}
\vspace*{\fill}
\end{document}
\begin{table}[!htbp]
\caption{NGO's strategy and support for related shareholder proposals} \label{ngo_experience_success}
\begin{center}
\resizebox{ 0.95\textwidth}{!}{
\begin{tabular}{lcccc}
\tabularnewline \toprule
Dependent Variable: & \multicolumn{4}{c}{Proposal Passes (0/1)}\\
\cmidrule(lr){2-5}
Model: & (1) & (2) & (3) & (4) \\
\midrule
% {Main Variables}\\
{Campaign on Same Topic} & -0.0003 & -0.013 & -0.005 & -0.001\\
& (0.026) & (0.028) & (0.025) & (0.030)\\
\# Past Campaigns & -0.034 & & & \\
& (0.024) & & & \\
{Campaign on Same Topic} $\times$ \# Past Campaigns & 0.078$^{***}$ & & & \\
& (0.028) & & & \\
\# Past Campaigns squared & 0.008$^{**}$ & & & \\
& (0.003) & & & \\
{Campaign on Same Topic} $\times$ \# Past Campaigns squared & -0.015$^{**}$ & & & \\
& (0.007) & & & \\
\# Past Non-AGM Campaigns & & 0.0008 & & \\
& & (0.017) & & \\
{Campaign on Same Topic} $\times$ \# Past Non-AGM Campaigns & & -0.011 & & \\
& & (0.024) & & \\
\# Past AGM Campaigns & & 0.005 & -0.034 & \\
& & (0.020) & (0.023) & \\
{Campaign on Same Topic} $\times$ \# Past AGM Campaigns & & 0.048$^{*}$ & 0.091$^{***}$ & \\
& & (0.026) & (0.032) & \\
\# Past AGM Campaigns squared & & & 0.010$^{**}$ & \\
& & & (0.005) & \\
{Campaign on Same Topic} $\times$ \# Past AGM Campaigns squared & & & -0.013$^{**}$ & \\
& & & (0.005) & \\
{Campaign on Same Topic} $\times$ Google Search Popularity & & & & 0.071$^{*}$\\
& & & & (0.040)\\
{Campaign on Same Topic} $\times$ Google Search Popularity square & & & & -0.021$^{*}$\\
& & & & (0.012)\\
\midrule
\emph{Fixed-effects}\\
CUSIP x Year & \checkmark & \checkmark & \checkmark & \checkmark \\
NGO x Year & \checkmark & \checkmark & \checkmark & \checkmark \\
Topic x Month-Year & \checkmark & \checkmark & \checkmark & \checkmark \\
\midrule
%\emph{Fit statistics}\\
Observations & 36,543 & 35,778 & 36,543 & 33,142\\
R$^2$ & 0.61894 & 0.61698 & 0.61995 & 0.62078\\
\bottomrule
% \multicolumn{5}{l}{\emph{Clustered (CUSIP) standard-errors in parentheses}}\\
\multicolumn{5}{l}{ *--p$< 0.1$; **--p$< 0.05$; ***--p$< 0.01$.}\\
\end{tabular}}
\end{center}
\begin{tablenotes}
\footnotesize \vspace{-1.5em}
\item Note: This table reports the estimated coefficients from (\ref{eq:main_reg}) in Column 1. In Column 2, we also consider the number of campaigns the NGO has launched in the past which were not on the firm's AGM day (\textit{\# Past Non-AGM Campaigns}) and the number of campaigns launched by the NGO on the targeted firm's AGM date in the past (\textit{\# Past AGM Campaigns}), and their interaction with \textit{Campaign on Same Topic}. In Column 3, we focus on the number of campaigns launched by the NGO on the targeted firm's AGM date in the past (\textit{\# Past AGM Campaigns}), its square, and their interaction with \textit{Campaign on Same Topic}. In Column 4, we replace \textit{\# Past Campaigns} from model \ref{eq:main_reg} by NGO $g$'s average Google Trends Score relative to GreenPeace the year before the campaign. All regressions also include firm-by-year, NGO-by-year, topic-by-month-by-year fixed effects. Clustered (firm) standard errors are reported in parentheses.
\end{tablenotes}
\end{table}
\begin{table}[H]
\caption{Summary Statistics \label{tab:summary}}
\begin{center}
\vspace{-0.5cm}
%\captionsetup{justification=centering}
\resizebox{0.65\textheight}{!}{
\begin{tabular}{lccccc}
\toprule
\multicolumn{6}{l}{\textbf{Panel a:} \textit{Variables varying by NGO}} \\
\midrule\midrule
& Mean & 25\% & 50\% & 75\% & 95\% \\
\cmidrule(lr){1-6}
Number of campaigns by NGO & 3.20 & 1 & 1 & 3 & 13 \\
\hspace{3mm} -- Campaigns on Env. topic & 1.51 & 0 & 1 & 1 & 5 \\
\hspace{3mm} -- Campaigns on Soc. topic & 1.70 & 0 & 1 & 1 & 6 \\
\hspace{3mm} -- On AGM date & 0.24 & 0 & 0 & 0 & 1 \\
Campaign sentiment & -1.72 & -2 & -2 & -1 & -1 \\
\cmidrule(lr){1-6} \\
% \\
% \midrule
\multicolumn{6}{l}{\textbf{Panel b:} \textit{Variables varying by target firm}} \\
\midrule\midrule
& Mean & 25\% & 50\% & 75\% & SD \\
\cmidrule(lr){1-6}
Market Cap (USD bn) & 175.66 & 21.71 & 75.05 & 209.47 & 263.66 \\
% Market Cap (decile) & 9.20 & 9 & 10 & 10 & 1.64 \\
Total Assets (USD bn) & 98.81 & 21.80 & 59.27 & 138.68 & 124.36 \\
% Total Assets (decile) & 9.26 & 9 & 10 & 10 & 1.52 \\
Yearly Sales (USD bn) & 37.26 & 6.85 & 17.35 & 42.41 & 150.58 \\
Yearly Number of News & 1,183 & 328 & 616 & 1,046 & 1,695.25 \\
Environmental score & 5.66 & 4 & 5.58 & 7.3 & 2.22 \\
Social score & 4.44 & 3.30 & 4.70 & 5.60 & 1.58 \\
Campaign on Env. topic & 0.47 & 0 & 0 & 1 & 0.50 \\
Campaign on Soc. topic & 0.53 & 0 & 1 & 1 & 0.50 \\
\cmidrule(lr){1-6} \\
% \\
% \midrule
\multicolumn{6}{l}{\textbf{Panel c:} \textit{Selection---Avg. target characteristics' based on whether the campaigning NGO campaigned on AGM dates in the past}} \\
\midrule\midrule
& \multicolumn{2}{c}{NGO w/ Past AGM Campaigns} & \multicolumn{2}{c}{NGO w/o Past AGM Campaigns} & Test Difference \\
\cmidrule(lr){2-3} \cmidrule(lr){4-5} \cmidrule(lr){6-6}
& Mean & S.E. & Mean & S.E. & p-value \\
\cmidrule(lr){1-6}
Market Cap (decile) & 9.23 & 0.10 & 9.07 & 0.04 & 0.13 \\
Total Assets (decile) & 9.26 & 0.08 & 9.23 & 0.04 & 0.72 \\
Sales (decile) & 9.31 & 0.09 & 9.14 & 0.04 & 0.07 \\
Yearly Number of News (lag) & 1,132.37 & 114.74 & 1,368.67 & 47.67 & 0.06 \\
Environmental score & 5.64 & 0.12 & 5.72 & 0.06 & 0.58 \\
Social score & 4.44 & 0.09 & 4.45 & 0.04 & 0.89 \\
Campaign on Env. topic & 0.46 & 0.03 & 0.5 & 0.01 & 0.15 \\
Campaign on Soc. topic & 0.54 & 0.03 & 0.49 & 0.01 & 0.12 \\
\cmidrule(lr){1-6}
\\
% \midrule
\multicolumn{6}{l}{\textbf{Panel d:} \textit{Selection---Target's characteristics based on NGO's past campaigns regressing} $z_{ny} =\beta_0 +\beta_1 \cdot \textit{\# Past Campaign}_{g(n)y} +\varepsilon_{ny}$ } \\
\midrule\midrule
& \multicolumn{2}{c}{ $\hat{\beta}_1$ (SE) w/o FEs} & \multicolumn{2}{c}{$\hat{\beta}_1$ (SE) w/ FEs} & N \\
Dependent Variable ($z_{ny}$): & (1) & (2) & (3) & (4) & \\
\cmidrule(lr){1-1}\cmidrule(lr){2-3}\cmidrule(lr){4-5}\cmidrule(lr){6-6}
Market Cap Decile & -0.002 & (0.003) &0.003 & (0.016) & 1,338 \\
Total Assets Decile & 0.003 & (0.004) &-0.002 & (0.012)& 1,338 \\
Sales Decile & 0.001 & (0.004) & -0.002 & (0.03) & 1,337 \\
Yearly Number of News & 0.996 & (12.4) & -4.12 & (16.5) & 1,323 \\
Environmental score & 0.004 & (0.009) & 0.041 & (0.041) & 1,338 \\
Social score & -0.0006 & (0.006) & 0.019 & (0.032) & 1,338 \\
Campaign on Env. topic & 0.006$^{***}$ & (0.002) & 0.021$^{**}$ & (0.010) & 1,338 \\
Campaign on Soc. topic & -0.006$^{***}$ & (0.002) &-0.018 & (0.012) & 1,338 \\
\cmidrule(lr){1-6}
\bottomrule
\end{tabular}}
\end{center}
\begin{tablenotes}[flushleft,para]
\footnotesize
\item Note: This table reports summary statistics and robustness checks across four panels.\\
\item \textbf{Panel (a).} Summary statistics on NGO campaigns (Sigwatch); AGM dates (ISS).\\
\item \textbf{Panel (b).} Target-firm statistics: accounting (Compustat), ESG (MSCI), news counts (RavenPack).\\
\item \textbf{Panel (c).} No-selection-on-observables test comparing targets where the NGO had previously campaigned on at least one of its targets’ AGM dates. \textit{Market Cap}, \textit{Total Assets}, and \textit{Sales} are deciles using all U.S. Compustat firms in the campaign quarter. Deciles avoid mechanical correlation with \textit{\# Past NGO Campaigns}, which grows over time.\\
\item \textbf{Panel (d).} Change in targets’ characteristics as an NGO campaigns multiple times. Each row reports $\beta_1$ from the panel-header regression of a dependent variable (e.g., \textit{Market Cap Decile}) on the NGO’s number of past campaigns up to the previous year. Columns (3)–(4) include \textit{NGO$\times$Year}, \textit{Firm$\times$Year}, and \textit{Topic$\times$Year} fixed effects; \textit{Topic$\times$Year} is omitted in the last two rows due to collinearity. Standard errors clustered by firm. Similar results by clustering by NGO.
\end{tablenotes}
\end{table}
% latex table generated in R 4.0.4 by xtable 1.8-4 package
% Mon Jun 06 11:28:20 2022
\begin{table}[h!]
\caption{Proposal Causes} \label{tab:proposals}
\begin{center}
\resizebox{\textwidth}{!}{
\begin{tabular}{lcr}
\toprule
NGO Campaign & Category & Example of Shareholder Proposals \\
Topic Classification & (E or S) & \\
\midrule
Climate Change & E & Encourage and Support Carbon Tax to Address Climate Change \\
Environment & E & Stop financing mountain top removal coal mining \\
Pollution & E & Report on Efforts to Reduce Pollution From Products and Operations \\
Rainforest & E & Report on impact on rainforest \\
Sustainability & E & Establish board committee on sustainability \\
Waste Generation & E & Do not increase radioactive waste production \\
Water & E & Report on supply chain water impacts \\
\midrule
Animal Rights \& Welfare & S & Review suppliers' animal slaughter methods \\
CSR & S & Disclose Charitable Contributions \\
Food Safety & S & Label Genetically Modified Ingredients\\
Health & S & Reduce Tobacco Harm to Health \\
Human \& Labor Rights & S & End Sri Lanka Trade Partnerships Until Human Rights Improve \\
Indigenous People & S & Adopt Global Policy Regarding the Rights of Indigenous People \\
Intellectual Property Rights & S & Implement a Process to Fully Vet Whether a \\
& & Supervisor is a Bona Fide Inventor Before a Patent is Filed \\
Politics & S & Affirm Political Nonpartisanship \\
Product Safety & S & Report on Nanomaterial Product Safety \\
Worker Safety & S & Report on Reducing Risks of Accidents \\
\bottomrule
\end{tabular}}
\end{center}
\begin{tablenotes}
\footnotesize \vspace{-1.5em}
\item Note: This Table reports the 17 topics assigned to each shareholder proposal in the sample. Shareholder proposals data from ISS. The topics are from Sigwatch's categorization of campaigns.
\end{tablenotes}
\end{table}
\begin{thebibliography}{55}
\providecommand{\natexlab}[1]{#1}
\bibitem[{Akey \textit{et~al.}(2021)Akey, Lewellen, Liskovich and Schiller}]{akey2021hacking}
\textsc{Akey, P.}, \textsc{Lewellen, S.}, \textsc{Liskovich, I.} and \textsc{Schiller, C.} (2021). Hacking corporate reputations. \textit{Rotman School of Management Working Paper}, ~(3143740).
\bibitem[{Allcott \textit{et~al.}(2022)Allcott, Montanari, Ozaltun and Tan}]{allcott2022economic}
\textsc{Allcott, H.}, \textsc{Montanari, G.}, \textsc{Ozaltun, B.} and \textsc{Tan, B.} (2022). An economic view of corporate social impact. \textit{Available at SSRN 4047276}.
\bibitem[{Baker \textit{et~al.}(2022{\natexlab{a}})Baker, Larcker and Wang}]{baker2022much}
\textsc{Baker, A.~C.}, \textsc{Larcker, D.~F.} and \textsc{Wang, C.~C.} (2022{\natexlab{a}}). How much should we trust staggered difference-in-differences estimates? \textit{Journal of Financial Economics}, \textbf{144}~(2), 370--395.
\bibitem[{Baker \textit{et~al.}(2022{\natexlab{b}})Baker, Egan and Sarkar}]{baker2022investors}
\textsc{Baker, M.}, \textsc{Egan, M.~L.} and \textsc{Sarkar, S.~K.} (2022{\natexlab{b}}). \textit{How do investors value ESG?} Tech. rep., National Bureau of Economic Research.
\bibitem[{Bar-Isaac \textit{et~al.}(2008)Bar-Isaac, Tadelis \textit{et~al.}}]{bar2008seller}
\textsc{Bar-Isaac, H.}, \textsc{Tadelis, S.} \textit{et~al.} (2008). Seller reputation. \textit{Foundations and Trends{\textregistered} in Microeconomics}, \textbf{4}~(4), 273--351.
\bibitem[{Baron(2001)}]{baron2001private}
\textsc{Baron, D.~P.} (2001). Private politics, corporate social responsibility, and integrated strategy. \textit{Journal of economics \& management strategy}, \textbf{10}~(1), 7--45.
\bibitem[{Baur and Palazzo(2011)}]{baur2011moral}
\textsc{Baur, D.} and \textsc{Palazzo, G.} (2011). The moral legitimacy of ngos as partners of corporations. \textit{Business Ethics Quarterly}, \textbf{21}~(4), 579--604.
\bibitem[{Bebchuk \textit{et~al.}(2020)Bebchuk, Brav, Jiang and Keusch}]{bebchuk2020dancing}
\textsc{Bebchuk, L.~A.}, \textsc{Brav, A.}, \textsc{Jiang, W.} and \textsc{Keusch, T.} (2020). Dancing with activists. \textit{Journal of Financial Economics}, \textbf{137}~(1), 1--41.
\bibitem[{B{\'e}nabou and Tirole(2006)}]{benabou2006incentives}
\textsc{B{\'e}nabou, R.} and \textsc{Tirole, J.} (2006). Incentives and prosocial behavior. \textit{American economic review}, \textbf{96}~(5), 1652--1678.
\bibitem[{Berg \textit{et~al.}(2021)Berg, Fabisik and Sautner}]{berg2021history}
\textsc{Berg, F.}, \textsc{Fabisik, K.} and \textsc{Sautner, Z.} (2021). Is history repeating itself? the (un) predictable past of esg ratings. \textit{SSRN Electronic Journal}, \textbf{3722087}.
\bibitem[{Berg \textit{et~al.}(2024)Berg, Heeb and K{\"o}lbel}]{berg2024economic}
\textsc{Berg, F.}, \textsc{Heeb, F.} and \textsc{K{\"o}lbel, J.} (2024). \textit{The economic impact of ESG ratings}. Tech. rep., SAFE Working Paper.
\bibitem[{Bertrand \textit{et~al.}(2021)Bertrand, Bombardini, Fisman, Hackinen and Trebbi}]{bertrand2021hall}
\textsc{Bertrand, M.}, \textsc{Bombardini, M.}, \textsc{Fisman, R.}, \textsc{Hackinen, B.} and \textsc{Trebbi, F.} (2021). Hall of mirrors: Corporate philanthropy and strategic advocacy. \textit{The Quarterly Journal of Economics}, \textbf{136}~(4), 2413--2465.
\bibitem[{Bertrand \textit{et~al.}(2020)Bertrand, Bombardini, Fisman and Trebbi}]{bertrand2020tax}
\textsc{Bertrand, M.}, \textsc{Bombardini, M.}, \textsc{Fisman, R.} and \textsc{Trebbi, F.} (2020). Tax-exempt lobbying: Corporate philanthropy as a tool for political influence. \textit{American Economic Review}, \textbf{110}~(7), 2065--2102.
\bibitem[{Bertrand \textit{et~al.}(2023)Bertrand, Bombardini, Fisman, Trebbi and Yegen}]{bertrand2023investing}
\textsc{Bertrand, M.}, \textsc{Bombardini, M.}, \textsc{Fisman, R.}, \textsc{Trebbi, F.} and \textsc{Yegen, E.} (2023). \textit{Investing in influence: Investors, portfolio firms, and political giving}. Tech. rep., National Bureau of Economic Research.
\bibitem[{Besley and Ghatak(2005)}]{besley2005competition}
\textsc{Besley, T.} and \textsc{Ghatak, M.} (2005). Competition and incentives with motivated agents. \textit{American economic review}, \textbf{95}~(3), 616--636.
\bibitem[{Bonnefon \textit{et~al.}(2022)Bonnefon, Landier, Sastry and Thesmar}]{bonnefon2022moral}
\textsc{Bonnefon, J.-F.}, \textsc{Landier, A.}, \textsc{Sastry, P.~R.} and \textsc{Thesmar, D.} (2022). \textit{The moral preferences of investors: Experimental evidence}. Tech. rep., National Bureau of Economic Research.
\bibitem[{Briscoe and Gupta(2016)}]{briscoe2016social}
\textsc{Briscoe, F.} and \textsc{Gupta, A.} (2016). Social activism in and around organizations. \textit{Academy of Management Annals}, \textbf{10}~(1), 671--727.
\bibitem[{Broccardo \textit{et~al.}(2022)Broccardo, Hart, Zingales \textit{et~al.}}]{broccardo2020exit}
\textsc{Broccardo, E.}, \textsc{Hart, O.}, \textsc{Zingales, L.} \textit{et~al.} (2022). Exit versus voice. \textit{Journal of Political Economy}, \textbf{130}~(12), 3101--3145.
\bibitem[{Cabral and Hortacsu(2010)}]{cabral2010dynamics}
\textsc{Cabral, L.} and \textsc{Hortacsu, A.} (2010). The dynamics of seller reputation: Evidence from ebay. \textit{The journal of industrial economics}, \textbf{58}~(1), 54--78.
\bibitem[{Chuah \textit{et~al.}(2024)Chuah, DesJardine, Goranova and Henisz}]{chuah2024shareholder}
\textsc{Chuah, K.}, \textsc{DesJardine, M.~R.}, \textsc{Goranova, M.} and \textsc{Henisz, W.~J.} (2024). Shareholder activism research: A system-level view. \textit{Academy of Management Annals}, \textbf{18}~(1), 82--120.
\bibitem[{{Clean Clothes Campaign}(2025)}]{cleanclothes2025inditexagm}
\textsc{{Clean Clothes Campaign}} (2025). Ngos demand end to airborne fashion and justice for bangladeshi workers at inditex agm. \url{https://cleanclothes.org/news/2025/ngos-demand-end-to-airborne-fashion-and-justice-for-bangladeshi-workers-at-inditex-agm}, accessed: 2025-09-29.
\bibitem[{Cu{\~n}at \textit{et~al.}(2016)Cu{\~n}at, Gin{\'e} and Guadalupe}]{cunat2016say}
\textsc{Cu{\~n}at, V.}, \textsc{Gin{\'e}, M.} and \textsc{Guadalupe, M.} (2016). Say pays! shareholder voice and firm performance. \textit{Review of Finance}, \textbf{20}~(5), 1799--1834.
\bibitem[{Dahan \textit{et~al.}(2010)Dahan, Doh, Oetzel and Yaziji}]{dahan2010corporate}
\textsc{Dahan, N.~M.}, \textsc{Doh, J.~P.}, \textsc{Oetzel, J.} and \textsc{Yaziji, M.} (2010). Corporate-ngo collaboration: Co-creating new business models for developing markets. \textit{Long range planning}, \textbf{43}~(2-3), 326--342.
\bibitem[{Daubanes and Rochet(2019)}]{daubanes2019rise}
\textsc{Daubanes, J.} and \textsc{Rochet, J.-C.} (2019). The rise of ngo activism. \textit{American Economic Journal: Economic Policy}, \textbf{11}~(4), 183--212.
\bibitem[{Djourelova and Durante(2022)}]{djourelova2022media}
\textsc{Djourelova, M.} and \textsc{Durante, R.} (2022). Media attention and strategic timing in politics: Evidence from u.s. presidential executive orders. \textit{American Journal of Political Science}, \textbf{66}~(4), 813--834.
\bibitem[{Durante and Zhuravskaya(2018)}]{durante2018attack}
\textsc{Durante, R.} and \textsc{Zhuravskaya, E.} (2018). Attack when the world is not watching? us news and the israeli-palestinian conflict. \textit{Journal of Political Economy}, \textbf{126}~(3), 1085--1133.
\bibitem[{Egorov and Harstad(2017)}]{egorov2017private}
\textsc{Egorov, G.} and \textsc{Harstad, B.} (2017). Private politics and public regulation. \textit{The Review of Economic Studies}, \textbf{84}~(4), 1652--1682.
\bibitem[{Eisensee and Str{\"o}mberg(2007)}]{eisensee2007news}
\textsc{Eisensee, T.} and \textsc{Str{\"o}mberg, D.} (2007). News droughts, news floods, and u.s. disaster relief. \textit{The Quarterly Journal of Economics}, \textbf{122}~(2), 693--728.
\bibitem[{Fioretti(2022)}]{fioretti2022caring}
\textsc{Fioretti, M.} (2022). Caring or pretending to care? {S}ocial impact, firms’ objectives, and welfare. \textit{Journal of Political Economy}, \textbf{130}~(11), 2898--2942.
\bibitem[{Fioretti \textit{et~al.}(2025)Fioretti, Saint-Jean and Smith}]{fioretti2024shared}
\textsc{Fioretti, M.}, \textsc{Saint-Jean, V.} and \textsc{Smith, S.~C.} (2025). The shared costs of pursuing shareholder values.
\bibitem[{Flammer(2015)}]{flammer2015does}
\textsc{Flammer, C.} (2015). Does corporate social responsibility lead to superior financial performance? a regression discontinuity approach. \textit{Management science}, \textbf{61}~(11), 2549--2568.
\bibitem[{Gent \textit{et~al.}(2015)Gent, Crescenzi, Menninga and Reid}]{gent2015reputation}
\textsc{Gent, S.~E.}, \textsc{Crescenzi, M.~J.}, \textsc{Menninga, E.~J.} and \textsc{Reid, L.} (2015). The reputation trap of ngo accountability. \textit{International Theory}, \textbf{7}~(3), 426--463.
\bibitem[{Glac(2014)}]{glac2014influence}
\textsc{Glac, K.} (2014). The influence of shareholders on corporate social responsibility. \textit{Economics, Management, and Financial Markets}, \textbf{9}~(3), 34--79.
\bibitem[{Gneezy \textit{et~al.}(2010)Gneezy, Gneezy, Nelson and Brown}]{gneezy2010shared}
\textsc{Gneezy, A.}, \textsc{Gneezy, U.}, \textsc{Nelson, L.~D.} and \textsc{Brown, A.} (2010). Shared social responsibility: A field experiment in pay-what-you-want pricing and charitable giving. \textit{Science}, \textbf{329}~(5989), 325--327.
\bibitem[{Grossman and Helpman(1994)}]{grossman1994protection}
\textsc{Grossman, G.~M.} and \textsc{Helpman, E.} (1994). Protection for sale. \textit{The American Economic Review}, \textbf{84}~(4), 833.
\bibitem[{Hart \textit{et~al.}(2022)Hart, Thesmar and Zingales}]{hart2022private}
\textsc{Hart, O.~D.}, \textsc{Thesmar, D.} and \textsc{Zingales, L.} (2022). \textit{Private sanctions}. Tech. rep., National Bureau of Economic Research.
\bibitem[{Hatte and Koenig(2020)}]{hatte2020geography}
\textsc{Hatte, S.} and \textsc{Koenig, P.} (2020). The geography of {NGO} activism against multinational corporations. \textit{The World Bank Economic Review}, \textbf{34}~(1), 143--163.
\bibitem[{Kaplan(2019)}]{Kaplan2020}
\textsc{Kaplan, S.} (2019). \textit{The 360\textdegree Corporation: From Stakeholder Trade-offs to Transformation}. Stanford, CA: Stanford University Press.
\bibitem[{Koenig and Poncet(2019)}]{koenig2019social}
\textsc{Koenig, P.} and \textsc{Poncet, S.} (2019). Social responsibility scandals and trade. \textit{World Development}, \textbf{124}, 104640.
\bibitem[{Kotsantonis and Serafeim(2019)}]{kotsantonis2019four}
\textsc{Kotsantonis, S.} and \textsc{Serafeim, G.} (2019). Four things no one will tell you about esg data. \textit{Journal of Applied Corporate Finance}, \textbf{31}~(2), 50--58.
\bibitem[{Kreps and Wilson(1982)}]{kreps1982reputation}
\textsc{Kreps, D.~M.} and \textsc{Wilson, R.} (1982). Reputation and imperfect information. \textit{Journal of economic theory}, \textbf{27}~(2), 253--279.
\bibitem[{Levit and Malenko(2011)}]{levit2011nonbinding}
\textsc{Levit, D.} and \textsc{Malenko, N.} (2011). Nonbinding voting for shareholder proposals. \textit{The journal of finance}, \textbf{66}~(5), 1579--1614.
\bibitem[{Madestam \textit{et~al.}(2013)Madestam, Shoag, Veuger and Yanagizawa-Drott}]{madestam2013do}
\textsc{Madestam, A.}, \textsc{Shoag, D.}, \textsc{Veuger, S.} and \textsc{Yanagizawa-Drott, D.} (2013). Do political protests matter? evidence from the tea party movement. \textit{The Quarterly Journal of Economics}, \textbf{128}~(4), 1633--1685.
\bibitem[{Mazet-Sonilhac and M{\'e}sonnier(2023)}]{mazet2023some}
\textsc{Mazet-Sonilhac, C.} and \textsc{M{\'e}sonnier, J.-S.} (2023). Some don't like it hot: Bank depositors and ngo campaigns against fossil fuels.
\bibitem[{McGahan(2023)}]{mcgahan2023new}
\textsc{McGahan, A.~M.} (2023). The new stakeholder theory on organizational purpose. \textit{Strategy Science}, \textbf{8}~(2), 245--255.
\bibitem[{McGahan and Pongeluppe(2023)}]{McGahanPongeluppe2023}
\textsc{McGahan, A.~M.} and \textsc{Pongeluppe, L.~S.} (2023). There is no planet b: Aligning stakeholder interests to preserve the amazon rainforest. \textit{Management Science}, \textbf{69}~(12), 7860--7881.
\bibitem[{Microsoft(2023)}]{microsoft2023}
\textsc{Microsoft} (2023). Microsoft corporation 2023 final proxy voting results. \url{https://www.microsoft.com/en-us/investor/corporate-governance/votingresults.aspx}.
\bibitem[{Milgrom and Roberts(1982)}]{milgrom1982limit}
\textsc{Milgrom, P.} and \textsc{Roberts, J.} (1982). Limit pricing and entry under incomplete information: An equilibrium analysis. \textit{Econometrica: Journal of the Econometric Society}, pp. 443--459.
\bibitem[{Mitchell and Stroup(2017)}]{mitchell2017reputations}
\textsc{Mitchell, G.~E.} and \textsc{Stroup, S.~S.} (2017). The reputations of ngos: Peer evaluations of effectiveness. \textit{The Review of International Organizations}, \textbf{12}, 397--419.
\bibitem[{Naaraayanan \textit{et~al.}(2021)Naaraayanan, Sachdeva and Sharma}]{naaraayanan2021real}
\textsc{Naaraayanan, S.~L.}, \textsc{Sachdeva, K.} and \textsc{Sharma, V.} (2021). The real effects of environmental activist investing. \textit{European Corporate Governance Institute--Finance Working Paper}, ~(743).
\bibitem[{Porter and Kramer(2011)}]{Porter2011}
\textsc{Porter, M.~E.} and \textsc{Kramer, M.~R.} (2011). Creating shared value: How to reinvent capitalism and unleash a wave of innovation and growth. \textit{Harvard Business Review}, \textbf{89}~(1--2), 62--77.
\bibitem[{Saint-Jean(2024)}]{saint2024exit}
\textsc{Saint-Jean, V.} (2024). Exit or voice? divestment, activism, and corporate social responsibility. \textit{Divestment, Activism, and Corporate Social Responsibility (April 16, 2024)}.
\bibitem[{{S}{E}{C}(2020)}]{securities2020guidance}
\textsc{{S}{E}{C}} (2020). Staff guidance for conducting shareholder meetings in light of covid-19 concerns. \url{https://www.sec.gov/ocr/staff-guidance-conducting-annual-meetings-light-covid-19-concerns}, accessed: 2022-03-08.
\bibitem[{Sun and Abraham(2021)}]{sun2021estimating}
\textsc{Sun, L.} and \textsc{Abraham, S.} (2021). Estimating dynamic treatment effects in event studies with heterogeneous treatment effects. \textit{Journal of Econometrics}, \textbf{225}~(2), 175--199.
\bibitem[{Yaziji and Doh(2009)}]{yaziji2009ngos}
\textsc{Yaziji, M.} and \textsc{Doh, J.} (2009). \textit{NGOs and corporations: Conflict and collaboration}. Cambridge University Press.
\end{thebibliography}