The role of users’ engagement in shaping financial reporting: should activists target accounting more?

We define accounting engagement as stakeholders’ actions taken with the intention of influencing corporate reporting. Using this definition, we review the literature on such activism and discuss avenues for research. The evidence reviewed suggests accounting engagement is rare. We reflect on the reasons of this, given evidence on increasing overt engagement on other corporate issues, such as managerial compensation and governance, social, and environmental responsibility. Both information production and information acquisition costs have decreased over time, raising further questions about why engagement has not increased. We consider potential reasons linked to concerns over whether financial reporting meets users’ information needs, particularly, given the emergence of new users and the role of new technologies in the diffusion and processing of information. These concerns have accompanied claims of increasing complexity of financial accounting and the threat of information overload.


Introduction
Conceptual frameworks for financial reporting start by defining financial statement users as including existing and potential investors, lenders, and other creditors (IASB 2018, FASB 2018). Yet our understanding of the use of financial accounting information by these actors and others remains limited (e.g. Cascino et al. 2014. This is problematic, as price efficiency and capital formation require a broad base of financial statement users who can promptly process corporate information. Recent developments in information technologies and the growing connectedness of users, via social networks, permit disseminating information at low cost, almost in real time (Chen et al. 2014, Lee et al. 2015, Miller and Skinner, 2015, blurring the line between financial information (prepared under Generally Accepted cost (e.g. Bao et al. 2020, Cao et al. 2020, Cohen et al. 2020. Nor can it be explained by a lack of coordination, given the ubiquity of social media (e.g. Lee et al. 2015, Acemoglu et al. 2018. Financial reporting may be at the onset of what we could call a 'reverse expectation gap.' The audit profession for years has been aware of the existence of an expectation gap, where users hold higher expectations than auditors in internal control assurance, fraud, and illegal-operations detection (Porter 1993, McEnroe and Martens 2001, Asare and Wright 2012. Financial reporting is at risk of devolving in the opposite direction, whereby regulators and practitioners develop ever more complex disclosures (e.g. Lev and Rajgopal 2016), but nonprofessional users are ill-prepared to process these, while professionals increasingly disregard them and over-weight less reliable but more timely sources of information.
The remainder of the paper is organised as follows. Next, we define engagement and review the literature on actions taken by users to influence corporate reporting. In doing so, we focus on traditional financial statement users, because they are the ones that substantially engage with firms, managers, and directors. We then explore the role of other users, with particular emphasis on media and public attention, in both initiating and upholding engagement. Finally, we reflect on the reasons for the lack of accounting engagement.

Accounting engagement by large investors and other traditional shareholders
We define user engagement as 'actions taken by financial statement users with the ultimate intention of influencing corporate reporting.' 3 Research has studied such actions, where direct, publicly overt engagement is often denoted as 'voice' or 'direct intervention' (Hirschman 1970, McCahery et al. 2016 and is viewed as more effective and socially desirable than the alternative 'exit' or 'voting with the feet' (Dasgupta andPiacentino 2015, Broccardo et al. 2020). Although we do not review inaction (i.e. exit or threat of exit) strategies, they may also influence reporting practices (e.g. Dou et al. 2016).
We review overt engagement with the firm, which we define as encompassing public 'voice' actions targeted at the firm, its managers, or directors, such as announced opposition to management (through Schedule 13D or similar filings), submission of shareholder proposals, negotiation with management, or initiation of proxy contests. Out of these, proxy contests, while effective, have become uncommon since corporate takeover activity declined in the 1980s, and other forms of engagement have emerged (Jensen 1993, Gillan andStarks 2000), with cumulative engagement being common (David et al. 2001). Toward the end of this section, we briefly review overt engagement with accounting standard setters, which bypasses the firm and thus meets our definition only partly. 4 Other less overt (often private) engagement actions exist, from sending e-mails, to participating in corporate calls, attending industry, broker-hosted, and bank conferences or presentations in investors' offices (road shows), and firms' headquarters (in-house) (Solomon and Soltes 2015). These events give access to management and key personnel and are important for growing, hard-to-value, small firms that are rich in intangibles Miller 2012, Green et al. 2014). Opaque firms benefit from such engagements, and, for example, Green et al. (2014) report that conference firms gain 0.34 analysts and experience a 1.24 percent increase in institutional ownership following a broker-hosted conference, while Kirk and Vincent (2014) report similar benefits for firms that create professional investor relation offices. Hence the literature suggests this type of engagement is beneficial, although the aim is usually limited to gaining access, and it only shapes financial reporting indirectly. These engagements are usually private or restricted (by invitation) and nonconfrontational, in contrast to the overt engagements. Because they are not readily observable, research on these engagements is limited.
Overall, the evidence suggests non-accounting engagement is on the rise and recently accelerating. 5 For example, shareholder-initiated proposals, which often accompany substantial press coverage and institutional investor involvement, have been surging since the mid-1980s. Despite this trend, accounting engagement is rare. People do not take to the streets because of controversial accounting treatments. The anecdotal survey evidence of Goldstein (2014) suggests that only 15 percent of the different forms of engagement studied link somewhat to accounting, as captured by engagement surrounding quarterly earnings. 6 Goldstein (2014, p. 13) notes an asymmetry in perceptions: preparers think of quarterly earnings calls and associated conversations with investors as engagement, while stakeholders think of engagement as we have defined it in this review, as 'voting a concrete action. ' No doubt the most overt and measurable engagement is voting. Proposals can be presented for approval, both by managers and shareholders, and, depending on the relevant securities stock exchange, refer to routine or nonroutine issues. 7 Research on proposals and general assembly meeting (GAM) voting has been conducted in the United States, where proposals focus on governance issues, such as rescinding poison pills, or requiring confidential voting, which may have spillover effects on accounting. The lack of accounting proposals can be explained, at least partly, by a legal limitation: accounting issues are considered 'ordinary business,' which is a reason to exclude a proposal from voting in the United States. The Securities and Exchange Commission (SEC) has not had a clear stance on how to interpret what constitutes ordinary business, allowing, for example, proposals on executive pay (Ferri and Sandino 2009) but not on auditing, with SEC staff issuing 'no action letters,' and excluding pension funds proposals 5 In 2017-2020, Lazard (2020) identifies over 2,000 campaigns, noting a growth in 'amicable agreements.' Activism against non-US targets accounts for 40 percent of activity. The academic literature suggests similar patterns. Ertimur et al. (2011) report that compensation-related proposals to S&P 1500 firms went from 320 in the period -2002, to 720 in the period 2003. Ertimur et al. (2010 also document an increase over time in proposal receiving majority support, which goes from 10.5 percent in 1997, to more than 29 percent in 2004. 6 User engagement may start with a letter/e-mail or telephone call (Goldstein 2014, p. 14), making it opaque. Regulation Fair Disclosure (Reg. FD), enacted by the SEC in October 2000, mandates that, when a firm discloses non-public material information to an individual or entity, that information must be publicly disclosed. Reg. FD reduced private communications temporarily. However, Solomon and Soltes (2015) since 1942, under SEC Rule 141-8). However, their usefulness was questioned early on, given low voting support Pound 1993, Gillan andStarks 2000) 9 and that they are nonbinding. Thus they were largely ignored by boards and managers and had a weak impact on share prices and managerial turnover, whether successful or not (Karpoff et al. 1996, Levit andMalenko 2011). This trend started to shift in the 1990s, with shareholder proposals gaining traction Cotter 2007, Bauer et al. 2015) as (i) influential institutional investors got behind them, (ii) they started to gain majority votes (Cuñat et al. 2012), and (iii) media coverage of the issues underlying proposals grew, increasing scrutiny. This shift means that managers and directors who ignore proposals may suffer reputational consequences, such that even a small percentage of votes against may reduce abnormal CEO pay (Cai et al. 2009), increase performance-sensitivity of compensation (Ertimur et al. 2013), induce CEO replacement (Buchanan et al. 2015), increase director turnover (Brochet and Srinivasan 2014), or lower the likelihood of directors being re-elected (Ertimur et al. 2010). In fact, engagement may target director elections with 'just vote no' campaigns, which publicly embarrass directors (Del Guercio et al. 2008) or lead to proxy contests and loss of further directorships (Fos and Tsoutsoura 2014).

Which users engage with the firm?
Due to free riding problems, theory predicts that only investors with sufficiently large stakes have incentives to undertake monitoring or costly control actions, as only large investors are likely to cover their monitoring costs with the increased return from monitoring Hart 1980, Shleifer andVishny 1986). Still, there is evidence of engagement by (1) individuals, (2) private institutions, (3) public institutions, (4) religious groups, (5) social activists, and (6) labour unions (e.g. Cotter 2007, Baloria et al. 2019). Individual investors played a role in early engagements, but increasingly it is institutional investors that get involved (Smith 1996, Gillan and Starks 2000, Ferri and Sandino 2009. Institutional investors have such large stakes that selling shares is not always an option (without driving prices down and further losses). They also have information advantage, given their resources and expertise (Piotroski andRoulstone 2004, Solomon andSoltes 2015). Out of all institutional investors, public and union pension funds abandoned their passive role first (in the late 1980s), led by the California Public Employees' Retirement System (CalPERS), which has voted, for example, against auditors that provide significant non-audit services (Krishnan and Ye 2005). More recently, hedge funds have dominated engagement (Klein and Zur 2009, Renneboog and Szilagyi 2011, Baloria et al. 2019). 8 Proposals can be submitted by any shareholder continuously holding shares worth $2,000 (or 1 percent of market value of equity) for one year, at least 120 days before the proxy statement is mailed. One proposal per shareholder is allowed. If the shareholder has complied with Rule 14a-8, does not fall within the 13 bases for exclusion and not withdrawn it, the company should include it in the proxy statement. The board may obtain permission from the SEC to exclude it, and then the SEC will issue a 'no action letter.' The board may also negotiate with the proponent. 9 Karpoff et al. (1996) report that 22 of 521 (4.2 percent) corporate governance proposals in the period 1986-1990 receive majority support, while Gordon and Pound (1993) report 12 out of 266 (4.5 percent) in the 1990 proxy season. In Thomas and Cotter (2007), 341 proposals (23.5 percent) receive majority support, dominated by corporate governance proposals. One hundred and three were fully implemented, as measured by corporate announcements within one year.
Shareholders perceive these different users as differently aligned with their interests. While all may counter agency problems (Harris and Raviv 2010), concerns exist that certain users pursue narrowly self-interested objectives (Cai and Walkling 2011, Agrawal 2012, Del Guercio and Woidtke 2018. Some institutional investors, such as banks, mutual funds, or insurance companies, may have incentives to 'curry favour with managers' to secure future business (Baloria et al. 2019, p. 911), while others may not have the skills to improve on managerial decisions. Thus their actions may just distract managers from more important tasks. Labour unions, in particular, have been singled out on this (Jensen and Murphy 1990, Schwab and Thomas 1998, Ferri and Sandino 2009. The literature is mixed on whether engagement, by any user, is beneficial (Brav et al. 2008, Renneboog andSzilagyi 2011). Early work by Karpoff et al. (1996) or Gillan and Starks (2000) looking at shareholder proposals, fails to find benefits in terms of improving firm operations, 10 while more recent evidence contends that it generates value, in particular, with respect to hedge funds. Hedge funds have fewer conflicts of interest, since they do not sell products or services (Cheng et al. 2015), and can hold highly concentrated positions in a small pool of firmsunlike regulated mutual funds and pension funds (Brav et al. 2008). Recent work finds that hedge funds orchestrate 'wolf packs' of investors willing to buy shares in target firms ahead of public disclosure of campaigns (Wong 2020) and create value by (i) firing management; (ii) forcing operational, financial or governance reforms (e.g. spinning-off noncore assets); or (iii) getting target firms taken over (Greenwood and Schor 2009). The literature documents positive abnormal returns to the announcements of hedge fund engagement via filling of Schedule 13D (Brav et al. 2008, Klein andZur 2009). 11 However, these benefits are also questioned. Aslan and Kumar (2016) provide mixed evidence of spillovers to market rivals. Coffee and Palia (2016) argue that hedge fund engagement produces cutbacks in long-term investment (particularly in R&D). Finally, Klein and Zur (2011) report that it reduces bondholders' wealth.
In line with this discussion, even though collaboration improves the probability of success (Dimson et al. 2015), it is not common for users to join forces. This suggests that free-riding and lack of alignment dominate. Gillan and Starks (2000) report that only 10 percent of proposals are submitted by coordinated groups, a number that rises to 25 percent in recent samples like Dimson et al. (2015).
Voting also reflects these misalignments. Proposals from individual investors (so-called 'gadflies') rarely receive majority support, and, when they pass (often due to directors' career concerns), they tend to destroy value (Gantchev and Giannetti 2020). Proposals from religious groups and social activists often receive the lowest support (Thomas and Cotter 2007). These groups are less studied, but the evidence suggests they often push for corporate social responsibility actions (Grewal et al. 2016). Flammer (2015) finds religious groups sponsor the largest number of CSR proposals (30.56 percent of 2,729 proposals between 1997-2012) and achieve the lowest percentage of approved proposals (0.60 percent).
In voting, proxy advisors emerge as relevant users (SEC 2013). They advise institutional investors on a subscription basis and may vote on their behalf (Ertimur et al. 2013(Ertimur et al. , 2017. For 10 Karpoff et al. (1996) looking at the period 1986-1990 fail to find evidence of benefits of shareholder activism as measured by market-to-book ratio, operating return on sales, and sales growth rate. They also do not find evidence of increased managerial turnover. Gillan and Starks (2000) find no general evidence of market reaction, albeit some of their results suggest a negative reaction. 11 When Schedule 13D is filed with the SEC, additional details are provided, usually in Item 4 or in a public letter attached. Schedule 13D is only required (within 10 days) for investors that acquire 5 percent or more of any class of securities Zur 2009, Cheng et al. 2015), but the form is also filed by smaller investors. example, mutual funds rely almost exclusively on their recommendations for proxy votes (Belifanti 2010). The dominant proxy advisor is Institutional Shareholder Services (ISS), and its key competitor is Glass Lewis (Choi et al. 2010), with a joint market share of 97 percent in United States (Copland and Larcker 2018) and 66 percent in Europe (Hitz and Lehmann 2018). Proxy advisors process large amounts of publicly available information, including adjusted accounting data. Their engagement with reporting is indirect, either by disclosing the data they use, 12 thereby potentially motivating managers to manage those numbers, or by participating in 'engagement meetings' organised by issuing firms prior to the proxy season (in person or by phone) that provide additional context or clarification of public disclosures (Hayne and Vance 2019). Proxy advisors influence shareholder voting in corporate governance outcomes, such as director elections or executive compensation (Bethel and Gillan 2002, Ertimur et al. 2013, 2017, Larcker et al. 2015, McCahery et al. 2016). Studies place their influence at 13-25 percentage points (Bethel and Gillan 2002, Cai et al. 2009, Larcker et al. 2015, though it is not entirely clear that their recommendations are value increasing (Larcker et al. 2013). In July 2020, the SEC amended the rules governing proxy solicitation, to improve the transparency, completeness, and accuracy of recommendations, requiring sharing rebuttals to their advice from managers. This area no doubt provides opportunities for research.

Which topics receive greater engagement?
The evidence, drawn mostly from studies that examine shareholder proposals, suggests limited accounting engagement, and, when it exists, it is often to request disclosures that may allow acting on other issues. Engagement focuses on external and internal corporate control and governance issues Cotter 2007, Grewal et al. 2016). Prior work classifies these two broad topics into further categories: (1) auditor, (2) compensation, (3) environmental/social (including reporting and reducing greenhouse gas emissions, waste storage, or the use of clean and renewable sources), (4) external corporate control and governance, (5) internal corporate control and governance, and (6) other social responsibility (Karpoff et al. 1996, Thomas andCotter 2007). 13 This last category may include topics such as political donations, codes of conduct, fair employment, or sexual orientation anti-bias policies.
Corporate governance engagement is viewed as raising substantive issues and targeting underperforming firms where improvements are necessary. In contrast, engagement on CSR issues was initially seen as having a more tenuous link with value, and typically attained low shareholder approval (Campbell et al. 1999, Romano 2001). 14 Recent work by Flammer (2015) still reports little voting support (about 75 percent of CSR proposals get less than 20 percent votes in favour), while Grewal et al. (2016) find that, from 1999 to 2013, the number of CSR proposals has doubled. Support, while low, has grown from 9 percent of votes in 12 For example, ISS, in its 'Pay-for-Performance Mechanics' documentation, noted that, in the 2020 proxy season, it replaced GAAP-based metrics, such as ROE, ROA, ROIC, and EBITDA growth, with EVA-based metrics, 'after engaging with investors and issuers on the limitations of GAAP metrics' for financial performance assessment, as EVA improves 'comparisons between companies,' because it 'cuts through accounting distortions.' 13 Early work uses less extensive categories. For example, Gillan and Starks (2000) find the most common types of proposals link to 1) repealing a classified board, 2) eliminating a poison pill, and 3) executive compensation. 14 Part of this mixed evidence reflects the mixed views on the impact of CSR on firm value (Bénabou and Tirole 2010). CSR may allow managers to take a long-term perspective, maximizing intertemporal profits, which benefits universal owners. Alternatively, responsible businesses may act as a channel for delegated philanthropy, revealing a managerial agency problem and destroying value (Dimson et al. 2015).
favour from 1997-2004, to 17 percent from 2005-2012. 15 Despite low support, this research argues that these proposals have a symbolic value, bringing social issues to the attention of management and the public and lending them legitimacy and support.
Perhaps the most comprehensive study on engagement linked with social and environmental issues is by Dimson et al. (2015), who had access to the data of an asset manager and its engagement actions, which exemplifies the above argument that accounting engagement serves as a channel for other actions. Dimson et al. (2015, p. 3236) split accounting topics along three categories and consider 'audit and control' and 'transparency and performance' as corporate governance themes, 'emission management and reporting' as an environmental issue (within climate change), and 'sustainability disclosure and reporting' as a social issue. This illustrates the integration of accounting in social, governance and environmental topics but also its lack of consideration as a separate engagement topic in prior work.

Auditor ratification and audit quality
Engagement may directly target the auditor-client relationship, via GAM ratification of the auditor (e.g. Dao et al. 2012) or in voting for directors or audit committee members. Such engagement clearly links to financial statements quality. For example, ISS US guidelines (ISS 2019a), recommend case-by-case vote on members of the audit committee and the board if poor accounting practices are identified. ISS advises voting against (or withholding votes for) members of the audit committee when non-audit (NAS) fees are excessive or the company has an adverse audit opinion. 16 ISS (2019b) UK guidelines also note a concern for excessive NAS fees, unexplained auditor switches, or audit partners 'linked to some controversy.' In the United States, where much of the research has been conducted, auditor ratification is voluntary and thus a management initiative, instead of direct user engagement. Sainty et al. (2002) find that 24.2 percent of EDGAR firms have auditor ratification votes in 1997, while subsequent work reports 65 percent to 70 percent for large firms (Raghunandan 2003, Krishnan and Ye 2005, Liu et al. 2009). Overall, larger, better performing firms with higher quality governance (Krishnan and Ye 2005) and higher audit quality (Dao et al. 2012) are more likely to seek auditor ratification. The exception is foreign firms listed in the United States. According to Tanyi and Cathey (2020), they rarely seek auditor ratification approval (only 13 percent in their sample do), and their requests tend to reflect shareholders' concerns with audit quality.
Auditor ratifications typically receive 98 percent to 99 percent of votes in favour (Sainty et al. 2002, Barua et al. 2017. Even in times of great scrutiny of the audit profession (such as surrounding the demise of Andersen), Mishra et al. (2005) report a minor increase in votes against ratification, up to 4.95 percent in 2003. 17 According to Cunningham (2017), proxy advisors' recommendation may somewhat increase voting against auditors (by 6.7 percent points). 15 In Flammer's (2015) sample (2,729 CSR proposals over 1997(2,729 CSR proposals over -2012, the most common types of proposals are split into environmental/sustainability issues (where environmental issues (504 proposals) dominate sustainability ones (144)), social issues (where the major ones relate to labour (455)), and political issues (444 proposals), followed by health issues, human rights, and animal rights. The topic of adding minorities and women to the board is the least common category disaggregated (79 proposals), though Flammer (2015) has a further category 'other social issues' where no disaggregation is provided. 16 Excessive NAS in the United States is (ISS 2019a) 'non-audit ('other') fees > audit fees + audit-related fees + tax compliance/ preparation fees.' In the United Kingdom (ISS 2019b), it is non-audit services that 'routinely exceed standard audit-related fees.' 17 As an example, Mishra et al. (2005)  Thus approval rates approaching 90 percent are seen as negative outcomes, where the evidence suggests voting against management's auditor choice acts a monitoring mechanism, as votes against are associated with poor performance (Liu et al. 2009), low corporate governance quality (Raghunandan andRama 2003, Tanyi andCathey 2020), and proxies for lack of auditor independence and low accounting quality, such as greater audit tenure (Dao et al. 2008), earnings restatements (Hermanson et al. 2009, Liu et al. 2009), internal control weaknesses (Hermanson et al. 2009, Cunningham 2017, and high NAS fee ratios (Raghunandan 2003, Dao et al. 2008, Liu et al. 2009, Mishra et al. 2005. Auditor disapproval is associated with increased likelihood of subsequent dismissals (Sainty et al. 2002, Barua et al. 2017) and negative market reactions to 8-K announcing the results of the vote (Tanyi and Roland 2017).
This monitoring role is, however, questioned by Cassell et al. (2020), who study restatements and suggest that in firms with greater percentage of retail investors, auditor dissatisfaction, and dismissals may reflect only poor performance. They find voting against the auditor, even when the current auditor differs from the one that audited the original statements. Greater institutional/ block-holder ownership appears to lower dissatisfaction with the auditor (Dao et al. 2008, Liu et al. 2009, Cunningham 2017 and to attenuate the effect of uninformed voting (Cassell et al. 2020). This speaks of interrelations between different types of monitoring and potentially of less need for overt opposition in settings where private engagement may be more effective.

Changes in accounting and disclosure policies
Research provides ample evidence consistent with users' demand for high quality reporting and disclosure (Bushee et al. 2003, Bird and Karolyi 2016, Lawrence 2013). This accounting engagement is only rarely channelled through shareholder proposals. 18 An exception is studied in Ferri and Sandino (2009), who study shareholder proposals on the accounting treatment of employee stock options (ESO). They study 153 proposals (90 percent presented by union pension funds) to expense ESO, out of which 107 were voted upon, receiving an average of 47 percent votes in favour, with 51 proposals gaining a majority vote. These proposals followed criticism to the accounting rule that allowed reporting no compensation expense for ESOs, potentially incentivising firms to use this compensation vehicle in excess.
More recent research focuses almost exclusively on engagement on environmental, social and governance topics, which link to disclosure. Baloria et al. (2019) study shareholder proposals related to political spending. 19 Around 20 percent are implemented, indicating substantial success, but perhaps of more interesting is that half of the proposals are withdrawn (not voted on), indicating private negotiation. Baloria et al. (2019) find the market reaction to 18 In contrast to the United States, in many countries in Europe, shareholders vote on the acceptance of financial statements and statutory audit reports. Proxy advisors, like ISS, recommend, as a rule, approving them. For the United Kingdom and Ireland, ISS (2019b) recommends voting against in case of fraud or material misstatements during the year. For continental Europe, it recommends voting against when the company is not responsive to shareholder questions about items that should be publicly disclosed (ISS 2019c). These guidelines indicate that overall quality of disclosure is considered. announcements of proposal implementation is negative, particularly, for those that did not go to vote, questioning whether such proposals increase value. Despite this, overall the research on CSR-related engagement suggests it leads to better climate-risk disclosures (Flammer et al. 2020), greater CSR as measured by KLD scores, better operating performance (plausibly through greater customer/employee satisfaction), and positive market reactions to successful proposal announcements and implementations (Flammer 2015, Dimson et al. 2015. Work on consequences of accounting engagement provides mixed findings. For example, Gal-Or et al. (2016) and Kachelmeier et al. (2016) examine directors' votes and find that voting against audit committee members leads to lower tax NAS services purchased and lower absolute discretionary accruals, even at low levels of vote withholding (on average 5.1 percent). Cheng et al. (2015) find evidence of greater conservatism after first-time schedule 13D and 13D/A filings by hedge funds, and Fan et al. (2020), using shareholder voting on corporate governance proposals, find evidence of lower earnings management in firms where shareholder proposals on corporate governance narrowly pass, as compared to firms where they narrowly fail. However, the findings of Gal-Or et al. (2016) are restricted to firms with nonstaggered boards, and Sun et al. (2013) find evidence of an association between pay-for-performance proposals and greater subsequent earnings management and target beating. This comports with the finding of Bertomeu et al. (2017) that corporate governance mechanisms that lead to steeper performance pay may increase earnings management in equilibrium.
While, as noted, engagement rarely targets accounting, two further results stand out with respect to the role of accounting information. First, Bourveau and Schoenfeld (2017) argue that a way to avoid being targeted is to improve financial disclosure. This is because mispriced firms are more likely to be targeted, and disclosure can make prices converge to fundamental value. Also, disclosure can enhance communication with stakeholders, improving the firm reputation and credibility and lowering litigation risk. Second, Adams and Neururer (2020) suggest that low accounting quality is associated with future activism, in that restatements and internal control weaknesses, small auditors, high abnormal audit fees, and high abnormal accruals may drive investor attention and lead to corporate governance proposals.

Which firms get targeted and how do they and other stakeholders react?
Firms may be targeted for years (Karpoff et al. 1996, Gillan and Starks 2000, Dimson et al. 2015. Engagement targets are firms that reflect the proponent's motivations but also where their actions are likely to achieve visibility (firms that attract greater press coverage are more susceptible) and, in the case of proposals, where high voting support is likely (Smith 1996). Of course, where the firm targeted is also representative of the issue at hand. Targeted firms usually perform poorly, justifying stakeholder dissatisfaction, and engagement is less costly than a takeover bid or proxy contest. Firms with entrenched managers (Ertimur et al. 2010) and more takeover defences are more likely to be targeted as are those that do not pay dividends (Klein and Zur 2009).
When management becomes aware of an engagement, it is common for the parties to meet, beginning negotiations and discussions, which may be optimal, particularly if users' demands are value reducing (Renneboog and Szilagyi 2011). Managers may attempt to convince the user to withdraw. This is clearly the case for proposals. Prevost and Rao (2000) argue that when a proposal is submitted it signals that management has been unwilling or unable to prevent the submission. Thus, as noted before, proposals are impactful even when withdrawn. Bauer et al. (2015) study over 12,000 proposals filed with S&P 1,500 companies from 1997-2009 and find that more than 20 percent are withdrawn. Withdrawn proposals are often presented by institutional investors and coordinated groups with great negotiating ability (Bauer et al. 2015).
Because of their influence, they can engage in private negotiations or draw contracts with the firm (Schoenfeld 2020).
Research on such private negotiations, because of its very nature, is limited. Early work by Carleton et al. (1998) studying private negotiations by TIAA-CREF pension fund and 45 firms contacted about governance issues in 1992-1996, finds agreements with targeted firms in over 95 percent of cases. 20 More recently, Bebchuk et al. (2020) study settlement agreements, in which investors agree to a standstill provision, during which they refrain from trying to influence policies of the firm, and the incumbent directors agree to take steps in the favoured direction. Such settlements are growing, going from 3 percent for campaigns launched in 2000-21 percent in 2013.
Thus, while withdrawn proposals may be implemented by the board (fully or in a modified manner), often symbolic changes are agreed upon to appease and continue to resist substantive changes. Grewal et al. (2016) find greater firm reaction to immaterial ESG issues, 21 while David et al. (2007) find that agreements are more likely with salient stakeholders but also that symbolic settlements reduce CSP. At the core of responsiveness is the idea of Mitchell et al. (1997) that managers cannot respond to all demands from all stakeholders and must allocate their attention, usually based on salience. Salient stakeholders are those with power, legitimacy, and urgency to their claims. In line with this view, Bebchuk et al. (2020) find settlements are more likely when there is a credible threat to lose board seats in a proxy fight or when reputation concerns are stronger.
In addition to implementing proposals, either fully or symbolically, managers may respond pre-emptively to pressure. When managers expect a contentious GAM, they react by issuing good news beforehand (Dimitrov and Jain 2011). Managers also strategically react to the increased participation of shareholders by attracting retail investors to voting (Lee and Souther 2020). This is done by sending by post full sets of printed proxy materials, increasing retail turnover. Greater turnover is beneficial because retail investors usually vote with management.
Finally, managers may react with corporate lobbying. Gleason and Glendening (2019)  Overt engagement makes other agents react, too. Evidence suggests auditors react increasing audit fees and issuing more adverse internal control opinions (Guo et al. 2021), consistent with increased auditor concerns about litigation and reputational risk. The limited evidence that exists suggests that auditors are aware of overt as well as more subtle engagements, like negative media coverage of client-specific CSR issues (Burke et al. 2019), and incorporate them into their pricing and decisions to retain clients. Sunder et al. (2014) examine debtholder reactions, as measured by loan spreads before and after activism. They find that the spreads decrease when activism targets managerial entrenchment but increase when it relies on the market for corporate 20 TIAA-CREF is the Teachers Insurance and Annuity Association of America-College Retirement Equities Fund. It is one of the largest pension funds in the United States and has been very active in private negotiations and in sponsoring shareholder proposals (Carleton et al. 1998). 21 Material and immaterial are, as defined by SASB, which develops sustainability accounting standards for the disclosure of sustainability issues in SEC filings, such as the Form 10-K and 20-F (SASB 2017). control or significant financial restructuring. This is in line with the expropriation from debtholders concerns of Klein and Zur (2011).

2.4.
Overt engagement with the standard setter: lobbying and academic engagement An alternative to engaging with the firm to influence accounting is to engage with the standard setter and participate in the political process, lobbying to influence or obstruct accounting standards or their implementation (e.g. McLeay et al. 2000, Durocher et al. 2007, Hochberg et al. 2009, Stenka and Taylor 2010. Zimmerman (1978, 1979) argue that there are incentives to lobby on accounting procedures because financial statements are used in the regulatory process (e.g. to set rates or to provide import relief), thereby playing a role in wealth transfers (Stigler 1971, Peltzman 1976).
Lobbying is difficult to observe because it may occur before issues are added to regulators agendas (Stenka and Taylor 2010), and users prefer informal channels, 22 such as pressure through media reports and campaigns, appeals to trade associations or government officials and technical staff in regulatory bodies, rather than writing public comment letters on exposure drafts (Weetman et al. 1996, Georgiou 2010). This preference is perhaps suboptimal, given that, when informed users such as portfolio managers, brokerage firms, or venture capitalists participate in lobbying, they are influential (Hillman and Hitt 1999), and pursue standards 'that provide useful information' (Durocher et al. 2007, p.55).
Stakeholders lobby when their benefits, adjusted by the probability of influencing the regulatory outcome of interest, exceeds the costs (Sutton 1984). Friedman and Heinle (2016) show that, as the uniformity of standards increases, incentives to lobby diminish and rules become less vulnerable to political influence. In addition, regulators are usually aligned with users' interests (Saemann 1999, Weetman 2001). This 'public interest' view of politics, coupled with the lengthy horizons involved, is a further explanation for the limited incidence of formal lobbying. 23 Therefore, while it is not uncommon for users as pension funds, unit trusts, or mutual funds to lobby (Hochberg et al. 2009, Giner and Arce 2012), free-riding problems, alignment with the regulator, long-term horizons, and the fact that firms are usually unlikely to be affected by any one accounting standard change, add up to explain limited formal lobbying (Weetman et al. 1996, Durocher et al. 2007, Georgiou 2010, Jorissen et al. 2012. 24 Anecdotally, these sorts of users appear more likely to lobby for enforcement. This was the case of the implementation of the Sarbanes-Oxley Act of 2002 (Hochberg et al. 2009) 25 and the Dodd-Frank Reform Act whistle-blowing provisions (Baloria et al. 2017).
As an example of the long-term horizons involved in lobbying, currently, under the framework of 'Better Communication in Financial Reporting,' the IASB has had three interlinked 22 According to Georgiou (2010, p. 113), on a scale of effectiveness perceptions from 1 to 5, commenting in the media and appealing to trade association are rated 3.09, while writing comment letters is rated at 2.57. 23 Durocher et al. (2007) also mention lack of knowledge of the process and issues as potential explanations. 24 Giner and Arce (2012) find that, out of 539 letters sent to the IASB with respect to IFRS 2 -Share-based Payments, Consultation Process, only 41 (7.6 percent) are classified as having been sent by users. Jorissen et al. (2012) analyse all comment letters sent to IASB during the period 2002-2006. Out of 3,234 letters analysed, 47 originate from users (investors, financial analysts, consumer organizations, and others, all of them users of financial statements for decision-making) and 110 from individuals. In their sample, the major participants are preparers (1,425), accounting professionals (835) and national standard setters (483). 25 Hochberg et al. (2009) find that, out of 1,948 comment letters received, 125 are from investor groups and 240 from individuals. Out of these letters, 104 (189) lobby for stronger measures than those proposed by the SEC, respectively, while seven (nine) take a neutral position, and 10 (13) lobby for exemptions or delays in implementation.
consultations on IAS 1-Presentation of Financial Statements improvements. (i) General Presentation and Disclosure, focusing on the statement of profit or loss (IASB 2019a), was open for comment until September 30, 2020, and received 213 comment letters. (ii) Disclosure Initiative-Accounting Polices, developing guidance to apply materiality judgments (IASB 2019b), was open for comments until November 29, 2019, and received 89 letters. (iii) Disclosure Initiative-Principles of Disclosure was open until October 2, 2017, and received 108 letters. Appendix 1 (see Supplementary data) lists participants in this process, which exemplifies that lobbying is dilated in time and requires a long-term commitment. We identify 44 parties that commented in all three consultations, and only one group representing users: The Corporate Reporting Users Forum (CRUF). The other 43 are 15 accounting standard setters, 4 securities regulators, 7 audit firms, 6 professional associations of accountants, 2 corporate preparers, and 9 preparer associations.
A final group that engages with the financial reporting process but typically falls out of the user definition, is academics. Academics engage indirectly in several ways: they produce research that may form the basis for new regulation. They may help firms introduce changes to their management accounting and control systems, which can, in turn, improve financial reporting (e.g. García Osma et al. 2021). More directly, academics have taken up high-profile positions at regulatory and standard setting bodies (such as, FASB, IASB, or EFRAG). Also, academics provide literature reviews requested by accountancy bodies, join committees that revise national GAAP standards, or participate in conferences and events organised by both regulators and the profession. Despite these engagement efforts, little is known about the role of academics in standard setting and practice. In lobbying research, academics represent a small percentage of all comment letters sent (around 5% in Giner and Arce (2012), and 2% in Jorissen et al. (2012)), and concerns exist that pressures to publish and trends in the development of academia in recent years may lower engagement with practice (e.g. Barth 2018, Schrand 2019). Thus such research into the real impacts of accounting academics is needed.

The role of media and public scrutiny in engagement
A fundamental piece of the engagement puzzle is the media. The media helps disseminate both corporate information and the concerns of users, raising awareness. 26 Theoretical arguments on the impact of media are mixed. Greater coverage reduces information acquisition costs, leading to more complete information incorporated into prices Stiglitz 1980, Bloomfield 2002). However, relevant corporate news releases may be highly monitored by traders so that the media results in no marginal increase in dissemination (Bushee et al. 2010). The empirical evidence suggests that media facilitates information gathering Zingales 2002, Miller 2006), improving the firm information environment Peress 2009, Soltes 2010), particularly when uncertainty is high (Bonsall IV et al. 2020), 26 A further, growing channel is social media, which firms use to communicate all types of information, and appears dominated by discussion from outside stakeholders on topics linked to social responsibility (Gomez-Carrasco et al. 2021). Corporations increasingly use social media to disseminate financial information. In social media, the online community can converse among themselves as well as with the firm 'in a multi-way dialogue viewable by all' (Lee et al. 2015, p. 370). Social media thus reduces firms control over the information environment (Miller and Skinner 2015), although the evidence is supportive of social media use lowering information asymmetry (Blankespoor et al. 2014) and attenuating negative stock price reaction to recall announcements (Lee et al. 2015) and positive spillover attention from consumers to financial markets (Madsen and Niessner 2019). and reaching and influencing more market participants than other information sources (Bushee et al. 2010(Bushee et al. , 2020. 27 The media includes, in the classification of Drake et al. (2017), professionals (e.g. Dow Jones, Newswires), semi-professionals (e.g. newspapers, business news, investment research) and nonprofessionals (e.g. blogs, forums), which jointly cover all types of corporate information, from tax evasions (Chen et al. 2019) to earnings announcements (Bonsall et al. 2020). Financial corporate events, such as earnings releases, dividends, M&A announcements, or changes in credit ratings, are covered more extensively by professional and semi-professional media (Drake et al. 2017).
Dissemination via professional business media has positive market consequences (Li et al. 2011, Twedt 2016, Fang and Peress 2009, Drake et al. 2014, Kothari et al. 2009). Nonprofessional media outlets have been studied less, and the evidence is more mixed, with critics indicating the noise-to-signal ratio is high. Drake et al. (2014Drake et al. ( , p. 1675 argue that reporters may be susceptible to the same behavioural biases (limited attention, bounded rationality, and bottomline fixation) that contribute to mispricing and lack the expertise to understand accounting information. In contrast, recent evidence suggests nonprofessional media may provide value-relevant information. For example, Huang et al. (2020) analyse Glassdoor.com (a platform where employees voluntarily and anonymously share opinions) and find that opinions posted there predict firm operating performance.
Media also has an information creation role as it produces new corporate information (Drake et al. 2014), which is often used by market participants (Zingales 2000). This literature analyses the effects of media information related to insider trading filings Suk 1998, Dai et al. 2015), SEC filings (Li et al. 2011), corporate governance violations (Dyck et al. 2008), tax evasion (Chen et al. 2019), or accounting fraud (Miller, 2006) and finds that, when media creates content that exposes governance problems, it disciplines managers (Baloria and Heese, 2018), acting as a monitor.
There is limited research on the role of public pressure in shaping accounting. The roots of this research can be found in the work of Watts and Zimmerman (1978, p.115), who argue that high profits draw attention of the public, who may associate these profits with monopoly rents, creating incentives to ask for the 'nationalization, expropriation, break-up, or regulation' of industries or firms. Holthausen and Leftwich (1983) similarly argue that accounting numbers indirectly affect the extent to which firms are criticised by such parties as consumers or politicians. The general prediction in this literature is that managers manage earnings downward, to reduce visibility and engagement, obtain political favours, or at least limit political intervention in rate or price negotiations (Navissi 1999, Gill-de-Albornoz andIllueca 2005), import relief investigations (Jones 1991), or anti-trust investigations (Cahan 1992). The evidence, particularly in early studies, using size as a proxy for scrutiny, was mixed (e.g. Dhaliwal et al. 1982, Lilien and Pastena 1982, Daley and Vigeland 1983, Christie 1990, Han and Wang 1998. More recently, evidence indicates that activists effectively use public scrutiny to pressure firms to comply with disclosure regulations. For example, Dyreng et al. (2016) cite the case of ActionAid International, a nonprofit group in the United Kingdom that discovered in 2010 that around half of the firms in the FTSE 100 did not comply with the Companies Act of 2006 to disclose the 27 Aware of this importance, managers seek to influence media coverage (Westphal and Deephouse 2011). A wide literature exists on the role of media in financial markets; see Thompson et al. (1987), Klibanoff et al. (1998), Tetlock (2007, Tetlock et al. (2008), Solomon and Soltes (2012) name and location of all subsidiaries. ActionAid launched a campaign that, among other actions, consisted in 'broadly publicizing all of FTSE 100 firms' lists of subsidiaries to highlight tax haven use' (Dyreng et al. 2016, p. 149). Pressure was also articulated as a threat of negative publicity. 28 Public scrutiny often relies on the financial press and increasingly on social media, to induce responses, not only from firms but also via accounting regulation or political intervention, although there is limited research on these topics. Accounting rules are, in general, determined by specialists (Ramanna, 2015), but the economic and social consequences of those rules influence voters (Linn et al. 2010), who may, in turn, penalise politicians who do not react to their concerns. 29 Thus public pressure and scrutiny may generate incentives for politicians to intervene in accounting. Evidence is largely consistent with the principal-agent theory of regulation (Zeff, 2005, Farber et al. 2007, which assumes that politicians are self-interested (Giner and Mora 2020) and intervene in accounting to influence their likelihood of re-election. Firms react to this threat, working to gain the favour of politicians (Yu and Yu 2011). 30 Public pressure may also lead to the type of actions that may be classified as 'exit' strategies, such as consumer boycotts. 31 Perhaps more importantly, the recent work of Samuels et al. (2021) analytically shows that the relation between public scrutiny and misreporting is not unambiguously negative. This is because, at low levels of scrutiny, pressures for earnings are low (i.e. earnings response coefficients are low), limiting managerial incentives for biasing accounting numbers. When public scrutiny increases, so do ERCs and thus the incentive to manage earnings. The probability of detection also increases with scrutiny. This monitoring channel dominates the valuation channel at high levels of scrutiny, leading to a unimodal relation between public scrutiny and misreporting.

Why is there limited accounting engagement?
Extensive research documents real effects of accounting: it influences investment efficiency, risk management, liquidity, cost of capital, and procyclicality (e.g. Kanodia and Sapra 2016). Thus one might wonder why users do not engage more? Do they care less about financial reporting than other corporate issues? A possible answer is that financial reporting is not relevant. This is a longstanding concern, given evidence of declining trends in relevance and quality Zarowin 1999, Rajgopal andVenkatachalam 2011). However, our analyses in Appendix 2 (see Supplementary data), replicating and extending seminal work, suggests these trends have stabilised and even reverted over time, invalidating that explanation. This is in line with Beaver et al. (2020), who find that financial information is increasingly important for investors. A second potential explanation is that the market shapes financial reporting on a continuous 28 ActionAid submitted to the Companies House a list of noncompliant FTSE 100 firms (Dyreng et al. 2016, p. 154). 29 For example, SEC-relevant politicians have a 31 percent higher probability of not being re-elected after a local firm faces SEC enforcement for financial misconduct (Mehta and Zhao 2020). 30 Research documents that politically connected firms are less likely to face SEC enforcement (Correia 2014) and have lower earnings quality (Chaney et al. 2011, Gross et al. 2016. See also Faccio et al. (2006), Gul (2006), Leuz and Oberholzer-Gee (2006), Wu et al. (2012), Goldman et al. (2013), Batta et al. (2014), or Guedhami et al. (2014). For studies on political influence on accounting, see Stigler (1971), Peltzman (1976), Snyder (1990), Grossman and Helpman (1994), Gordon and Hafer (2005), Farber et al. (2007), or Ramanna (2008). 31 Consumer associations view boycotts as a fundamental way to exert economic pressure on large companies. See for a list of boycotts: https://www.ethicalconsumer.org/ethicalcampaigns/boycotts, accessed October 2020. basis, serving as a more effective channel than user engagement. To the extent, for example, that research finds greater conditional conservatism in firms that depend on debt (e.g. Ball et al., 2008) or that have greater information asymmetry (e.g. LaFond and Watts, 2008), one can infer that the market already shapes financial reporting, at least within regulatory thresholds, because supply and demand of capital is involved. Thus any more direct, targeted engagement would only exist in cases where the market may be less likely to act efficiently. This likely reason does not preclude the existence of other alternative explanations, perhaps of second order.
Ultimately, user engagement depends on the usefulness of accounting information, which hinges both on the information needs of users and on what financial reporting can and cannot achieve. Information production and dissemination is increasing McDonald 2014, Gao and, as firms and information intermediaries provide disclosures through multiple channels. This trend suggests a reduced cost of information production, which runs parallel to a decrease in information acquisition costs. However, this may also link to potential increases in the cost of information awareness and processing, and growing concerns about information overload (e.g. Monga and Chasan 2015), although evidence on overload is limited (Chapman et al. 2019.

Who are the relevant users?
Standard setters have, for years, kept the concept of 'user' on the backburner, seemingly unwilling to extend the user base or to design standards targeted at increasing the usefulness of information for less sophisticated users. Thus, while accounting regulation aims to serve and protect the interests of financial statements' users (Young, 1994, Cooper and Morgan 2013, paradoxically, there is limited user involvement in standard setting (Durocher et al. 2007, Durocher andGendron 2011). Users are 'represented rhetorically' (Hopwood, 1994, p. 248), where standard setters treat 'users' as a single, standardised yet abstract category of idealised economic actors who make economically rational decisions and are self-interested in a consistent way (Williams and Ravenscroft 2015, Stenka and Jaworska 2019). The literature denotes them 'made-up' users (e.g. Pelger 2016, Georgiou 2018). Oberwallner et al. (2020) suggest preparers also construct these abstract users and their information demands, with limited realism, when deciding what information to include in annual reports.
Therefore critics say that research has failed to fully uncover actual users' needs (Power 2010, Williams andRavenscroft, 2015). And in fact, there is limited research on users beyond traditional capital providers (shareholders and debtholders). 32 Even for these users, there is limited understanding of how they use and combine different sources of information for decision-making , Cascino et al. 2020. Increasingly, research acknowledges the potential for other users' engagement. For example, Henrichs et al. (2019) study access to corporate conference calls and report participation by individuals who would not be typically classified as traditional information intermediaries or users, such as suppliers, strategic partners, bank advisers, consultants, and the media.
There is however little research about some of these other users. We discuss some briefly, where we see opportunities for further work. With respect to internal users, the role of employees and middle managers, and how management accounting and control systems influence financial reporting is a potentially fruitful area. Prior work has focused on how labour unions, in 32 Stenka and Jaworska (2019, p. 3) list other potential real users: 'employees, regulatory agencies, governments, consumers, and special interest groups-with significantly different (and potentially conflicting) needs.' negotiating, use financial information to infer financial stability and set their demands (Faleye et al. 2006). Seminal papers model labour unions as rent seekers that extract quasi-rents from firms under the threat of a strike, giving rise to incentives to manage earnings downward to minimise this rent extraction (Liberty andZimmerman 1986, DeAngelo andDeAngelo 1991). To date, however, the literature has failed to find definitive evidence for this hypothesis (Garcia Osma et al. 2015), while it does find that managers, to minimise labour-related costs (strike costs, inefficient production, or damaged reputation in labour markets), withhold good news (Chung et al. 2016), miss earnings targets (Bova 2013), or smooth income (Hamm et al. 2018).
A unique form of employee engagement is whistleblowing. Dyck et al. (2010) note that employees assist in fraud detection, because they can gather fraud-relevant information as a product of their work (see also Stubben and Welch 2020). 33 Recent US regulation encourages and rewards whistleblowing (SEC 2015), while other countries trail in regulating it, and the EU has only recently provided protection to whistle-blowers. 34 With respect to external users, new technologies are giving rise to new users, where opportunities for research exist, such as robo-advisers, which are information intermediaries that analyse financial statements based on computer algorithms with little or no human intervention (Hodge et al. 2021). 35 The SEC has made robo-advisers an examination priority to understand how they present information to retail investors (SEC 2017a(SEC , 2017b. When contracting with a roboadviser, customers complete a questionnaire with information about age, monthly income, savings objective, level of financial knowledge, and risk tolerance. The robo-adviser then uses mathematical algorithms, based on portfolio management strategies, combining stock market information, investors' resources, and their replies to the questionnaire to provide advice. While financial statements, and particularly, information on accruals, changes in earnings, and asset growth are relevant for portfolio management strategies (Hand and Green 2011), the use that robo-advisers make of financial information is unclear. Study of them is relevant. Investors increasingly use them, because they have lower investment minimums and charge lower fees than human advisers and returns appear comparable (Uhl and Rohner, 2018).

Information acquisition and production: the problems of overload and complexity
To improve user engagement and decision-making, the literature suggests (i) the information set may be altered, including its presentation features (Hodge et al. 2004, Libby andEmett 2014), 36 33 While accounting concerns may exist, they are rarely reported by whistle-blowers, Stubben and Welch (2020) find that most cases relate to discrimination, sexual or other forms of harassment, and violations of human resources policies (54.9 percent of cases). They also find that 15.7 percent relate to business integrity concerns (i.e., conflicts of interest, falsification of company records, or bribery), misuse of corporate assets (11.8 percent), and workplace safety concerns (8.1 percent). Accounting, and financial concerns account for only 0.7 percent of the reports. 34 For further information about higher EU protection of whistle-blowers, see European Commission (2018). Whistleblowing protection does not include classified information. 35 Other advisers that have not been studied in detail are compensation consultants, who advise on compensation (level and structure) for managers and directors (Cadman et al. 2010, Murphy and Sandino, 2010, Armstrong et al. 2012, Cho et al. 2020. In reaction to concerns over the role of compensation consultants, the SEC had required, since 2006, that firms disclose in proxy statements the identity and role of all consultants. Critics accuse these consultants of encouraging excessive compensation, saying they bias their advice to receive greater fees from clients Fried, 2006, Waxman, 2007). On compensation consultants and CEO payment, see also Conyon et al. (2009) or Chu et al. (2018. 36 Libby and Lewis (1977, figure 1) model the inputs of the information set along the following dimensions: 1) scaling (measurement, discrete/continuous, deterministic/probabilistic); 2) statistical properties (number of cues, distributional characteristics, interrelationships of cues, and dimensionality); 3) information content but also, of course, (ii) it may be necessary to educate decision-makers to improve their information processing and awareness. Alternatively, Libby (1976) notes decision-makers (iii) may be substituted with a model. Increasingly, all three elements are shifting, particularly, with new technologies emerging that allow for faster, consistent, and unbiased decision-making. One potential risk is that a machine may prepare disclosures that are read by another machine. The work of Cao et al. (2020) points in this direction and finds that disclosure is affected by machine processors, used by algorithmic traders, robo-advisers, and quantitative analysts. This shifting background matters in understanding why users engage. Conceptually, to benefit from a specific disclosure, users must know of its existence, access it, and must be able to process the information it contains. These various steps can be denoted as 'information acquisition.' Users differ in their (i) awareness of disclosures, (ii) understanding of the relevance of specific disclosures for their decisions, and (iii) the ability and time to process and appropriately incorporate accounting information into their decision-making (Blankespoor et al. 2019(Blankespoor et al. , 2020. While rational behaviour models predict that investors are efficient in gathering and processing all available information (Sharpe 1964, Lintner 1965, Black 1972, Merton 1973, information acquisition is costly, in terms of both time and effort (Hirshleifer and Teoh 2003). Thus market participants are selective in what information they acquire and how they use it (Hodge and Pronk 2006). Investors with limited processing and capital capacity appear to be easily distracted (Madsen and Niessner 2019). They appear to prefer easier-to-analyse information and invest more in firms with more readable, shorter annual reports (Lawrence 2013) while underreacting to earnings announcements (Bernard and Thomas 1989), not fully identifying changes in disclosures (Cohen et al. 2020), failing to process information in footnotes and the different persistence of the cash flow and accrual components of earnings (Sloan 1996, Xie 2001). In turn, information processing problems could lead to delays in the impounding of information into asset prices (Cohen and Lou 2012).
With increasing geographical dispersion of investors, disclosure through the internet is cost effective, flexible, and easily accessible (Debreceny et al. 2002), yet little is known about internet intermediaries (see, e.g. Chen et al. 2014) or where users obtain information. In addition to buying research from analysts, users may monitor repositories such as EDGAR for filings, subscribe to blogs, do web searches, track social media, etc. However, information gathering is costly (Merton 1987, Hirshleifer andTeoh 2003), and, for nonprofessional investors, this may be a major hurdle. In addition, while greater dissemination reduces acquisition costs, concerns exist that it may crowd out information production (Gao and Huang 2020). This may happen because the advantage for professional investors decreases as dissemination increases (Irani andKaramnou 2003, Koch et al. 2013). A further concern is that information production by nonprofessional internet intermediaries with diverse backgrounds, expertise, and incentives may hinder price formation, if they provide information that is noisy, untimely, or inaccurate (Drake et al. 2017). 37 Overall, this brief discussion raises numerous research questions regarding the incremental information added by new information intermediaries, their role in shaping disclosure and reporting, the extent to which they should be regulated, or whether increasing complexity generates (bias, error); 4) presentation (format: numerical, graphical; verbal: sequence and aggregation); 5) context (viewing conditions, objective, net costs, task characteristics, and feedback). 37 Professionals earn their living by covering firms and have technical skills as well as financial/accounting backgrounds. This ensures they have incentives to expend time, effort, and resources to produce useful information. This is clearly not the case for nonprofessional intermediaries, who may be at the opposite end of the spectrum along one or several of these dimensions. overload. Indeed, concerns about information or disclosure overload have existed for years (e.g. ICAS 2011, ICAEW 2013. The AICPA, in the Jenkins report, almost three decades ago (AICPA 1994, p. 3) complained that the 'undisciplined expansion of mandated reporting could result in large and needless costs.' Many years later, the FASB launched, in 2009, the 'Disclosure Framework' project, with the aim of producing more 'useful, organized and consistent disclosures.' Hans Hoogervost, then chairman of the IASB, shared overload concerns in 2012 while also acknowledging that what one user may regard as clutter could be a 'golden nugget of information' for another (IASB 2012).
These concerns map back into a reality: annual reports have become alarmingly long, containing about five times more text in 2017 than in 1995, and 12 times the number of annual textual changes, i.e. text removed, added, or modified (Cohen et al. 2020). They also contain more boilerplate and redundant disclosures, less specific language, and are less readable (Dyer et al. 2017). Critics contend that this evolution is partly due to accounting standards becoming more complex (Lev and Rajgopal 2016). 38 This is particularly the case with respect to changes in the disclosure requirements for fair value accounting, derivatives, and hedging (Monga andChasan 2015, Guay et al. 2016). 39 Unsurprisingly, some evidence suggests preparers, particularly in small, private firms, may be increasingly unaware of regulations (Gassen and Muhn 2018) While it seems unlikely that regulation can encourage user engagement, it can incentivize the production of information, its processing, and dissemination. For example, the recent regulation on short-selling in the United States incentivized short-sellers' discovery of hidden bad news (Fang et al. 2016).
Overall, despite calls for simplified disclosure, financial reporting appears to move in the opposite direction. 40 This has coincided with a seeming loss of readership, as Loughran and McDonald (2017) note that, immediately after filling, the annual report of an average publicly traded firm is downloaded from the Securities and Exchange Commission (SEC) website 28.4 times. In contrast, investors' Google searches increase two weeks before corporations' earnings announcements, suggesting heightened interest (Drake et al. 2012). Cohen et al. (2020) find that a trading strategy that goes long on firms that have textual changes in their annual reports and short on firms that do not is profitable (up to 7 percent in value-weighted annual abnormal returns), which suggests that investors do not carefully read the filings. 38 Khan et al. (2018) study 138 accounting standards issued by the FASB during from 1973-2009 and find no market reactions to events that changed the probability of issuing the standards for 104 FASB standards, while 25 were associated with increases in estimation risk. 39 Guay et al. (2016) argue that regulation increases information complexity, as measured by different proxies of lower readability and greater length of annual reports (10-K filings). They find greater complexity following the adoption of SFAS 157, 'Fair value measurement,' which provides a framework for the measurement of financial assets and liabilities at fair value, increasing information processing costs, and SFAS 133 'Accounting for derivative instruments and hedging activities,' which created uncertainty about the hedging of derivates positions. 40 The audit report recently underwent an expansion as well, with the new expanded reports (IAASB 2015, PCAOB 2017) that go beyond the traditional binary fail/pass model and include information on accounting and risk issues (key/critical audit matters) that enable users to better understand the audit opinion. This new report has not been in place for long, and evidence on its consequences is still developing (Minutti-Meza, 2020). But early evidence suggests these additional disclosures add limited new information, do not improve audit quality, and affect audit fees (Gutierrez et al. 2018, Bédard et al. 2019). This may be the case because information on client's risk is likely available elsewhere (Gray et al. 2011, p. 662), which suggests that users rarely read the report, as they focus on whether the report is unqualified and the name of the accounting firm signing it.
Research on the actual use of fillings is needed as well as on the use of new technologies, to explore how to better communicate with users. Recent lab experiments examine these issues. Investors increasingly consult financial information on mobile phones (with small screens), and this appears to harm decision-making (Grant 2020). Users over-focus on headlines (Brown et al. 2020), which is consistent with distraction. This research suggests the medium of disclosure influences decisions. For example, CEO communication style is differently perceived over social media than in conference calls , and it appears that CEOs can mitigate the effects of negative information by communicating news via their personal Twitter account (Elliott et al. 2018) or using online video (Elliott et al. 2012). 41 Moving from this experimental research to naturally occurring environments is, however, not straightforward. In the field, evidence on whether information overload is a real concern is limited. It is a longstanding contention that disclosure overload may make decision-making less efficient (e.g. Casey 1980). 42 Overload problems extend to management accounting and auditing (Alles et al. 2006, p. 158), where data analytics implementation may identify numerous problems ('alarms'), 43 leading the organisation to ignore the alarm system altogether ('a particular case of information overload.') Despite this, the literature on 'Big Data' generally views the growing volume of information in a positive light.

Conclusion
We define accounting engagement as stakeholders' actions taken with the aim of influencing corporate reporting. Building on this definition, this review reflects on the role of user engagement in accounting and discusses avenues for future research. Our definition is narrow, to keep the review tractable, and we acknowledge that we do not review in detail important actors (as they do not meet our definition) or actions that, while not targeting financial reporting, influence it indirectly.
Our review concludes that user engagement, as we have defined it, is growing in many areas of corporate actions-but not in accounting. This is puzzling against a backdrop characterised by the emergence of new users (potentially new information objectives). But so far researchers have a limited understanding of the role of new technologies in the diffusion and processing of financial information.
Despite an ample literature on accounting quality and its links with capital market consequences, research into how well accounting quality maps into users' needs and decisionmaking is also limited, and there is a dearth of research on the consequences of overt and particularly private engagement in accounting quality. We identify areas where further research maybe fruitful, such as the processes by which information is acquired and processed (and the costs) as well as potential issues linked to information overload. Moving forward, it is imperative to improve our empirical approaches and gather direct evidence by engaging users directly, to 41 Fidelity Investments, with 15 million retail brokerage customer accounts, reported in 2013 that mobile trading grew 63 percent over the previous year. Their financial apps had been downloaded 3.6 million times (Patel 2014). 42 Early work referred to overload as the number of inputs or cues provided to a decision-maker (Iselin 1988), where, as data is processed, the cognitive structure becomes overtaxed, and resources committed to perception are deployed to integrating cues, such that new cues are less well perceived. As dimensionality increases, so does information diversity, which Iselin (1988) concludes is the real threat to accurate and timely decision-making. 43 If high-volume, high-velocity, high-variety information can be collected, managed, and analysed in a timely manner, it may help auditor decision-making (risk assessments). However, Brown-Liburd et al. (2015) and  note that processing and obtaining value from this data is challenging (partially because of outlier generation), and thus these tools are still rarely used in audits. create a better understanding of how information is used and for which decisions. Examples of innovative empirical approaches, such as the work of Fishbane et al. (2020), which suggests that changing the presentation of information affects important day-to-day decision-making, could be embraced.
Research provides concerning evidence on limited use of annual reports, begging the question of what exactly is the role of annual financial statements and its competitive advantage? It might also be interesting to study whether users use all disclosures and whether disclosures are increasingly prepared to be interpreted by machines. Finally, future research on overt engagement could analyse whether there is an 'activism loop,' in which activist groups target other activists and firms targeted by activists lobby the government.