The Formation of Reputation in CSR Disclosure: The Role of Signal Transmission and Sensemaking Processes of Stakeholders

Abstract

A growing number of companies are issuing corporate social responsibility (CSR) reports to publicize their commitment to sustainable development. However, skepticism remains among stakeholders about firms’ CSR information, which could hinder the success of worldwide CSR campaigns. Our paper examines mechanisms of how CSR disclosure resonates with stakeholders and influences their attitudes towards firms. Extending the current knowledge of CSR signaling effects, this paper provides a framework illustrating the interplay between CSR signaling properties and readers’ sensemaking processes, thereby predicting how corporate reputation is shaped through CSR communication. In order to test our theoretical hypotheses, a survey was conducted on 53 firms with 1521 respondents. The results show that the better the readers’ comprehension of a CSR report is, the stronger the signals of authenticity and corporate social performance (CSP) they perceive, the better the report value and value fit are recognized, and eventually, the more trust they hold for the firm. The relationship between comprehension of CSR reports and trust is partially mediated by the signaling-sensemaking process. Our research contributes to the literature on micro-foundations of strategic CSR by applying signaling theory in the context of CSR disclosure. The research findings have practical implications for firms’ CSR disclosure strategies.

1. Introduction

With the goal of sustainable development, companies are increasingly expected to demonstrate the integration of environmental and social concerns into their business operations [1], such as the reduction of carbon emissions, ecological protection, promotion of social equality, and poverty alleviation. Consequently, more and more firms worldwide have been providing corporate social responsibility (CSR) reports, or sustainability reports, over the last decade. According to the Governance & Accountability Institute (2021), more than 90 percent of firms on the S&P 500 index published standalone CSR reports in 2020, up from only 20 percent in 2011 (https://www.globenewswire.com/news-release/2021/11/16/2335435/0/en/92-of-S-P-500-Companies-and-70-of-Russell-1000-Companies-Published-Sustainability-Reports-in-2020-G-A-Institute-Research-Shows.html, accessed on 16 January 2023). CSR disclosure has been serving as an increasingly important tool for firms to respond to regulatory pressures, manage prosocial images and stakeholders relations, and therefore, promote their legitimacy and competitive positioning [2].

The key mechanisms by which CSR disclosure impacts firms center on its ability to promote corporate reputation. The strategic view of CSR reports holds that a firm’s involvement in CSR activities activates the goodwill of stakeholders and the general public, which in turn enhances their reciprocal behaviors with the firm [3,4]. For example, consumers would reward the companies that have better CSR image with their stronger interest, loyalty, and trust [5]. The reputation from regular reporting CSR could help the firms build ethical capital that may serve as protection from some negative events [6]. Despite these potential benefits, skepticism about firms’ CSR disclosure remains, which could damage corporates’ reputations. The suspicion stems largely from concerns about the latent motivation behind firms’ CSR engagement. Rather than reflecting a genuine desire to do social good, CSR disclosures by firms are often perceived as “window dressing” or “greenwashing” in an attempt to conceal firms’ non-sustainable practices or misconduct [7,8]. As concerned, firms may conduct CSR activities, particularly corporate philanthropy, as a strategy to cover up their delay of goodwill impairment [9]. Driven by opportunistic motives, firms’ responses to social pressures are often decoupled from their practices [10]. Thus, CSR disclosures are viewed by some people as symbolic compliance to social pressures without substantial devotion to CSR activities. Furthermore, when the goals between managers and stakeholders are not aligned, CSR might incur additional agency costs [11] or managers’ wrongdoing [12]. CSR can be arranged by managers to serve their personal needs of image reinforcement rather than the firm’s reputation [13]. In line with skepticism, research has shown that CSR disclosure may backfire, resulting in negative evaluations of firms among their stakeholders [14].

While the extant literature studied firms’ motivation and organizational consequences of CSR, the micro-foundations remain largely underexplored, especially with respect to how CSR disclosure resonates with stakeholders and influences their attitude towards firms [15]. Given that stakeholders’ attitudes are pivotal in mediating the relationship between CSR disclosure and organizational outcomes [16], more research work is needed to improve the understanding of the micro-level mechanisms. Some scholars used signaling theory to understand how CSR information is conveyed to firms’ stakeholders and thus shapes firms’ reputations [17,18]. In particular, CSR reporting is viewed as a way to signal firms’ intentions and goals associated with their CSR engagement. Signals from CSR reporting could be used to reduce information asymmetry between firms and stakeholders. Therefore, it would promote favorable images of firms [19]. However, extant literature in this stream has focused on the side of signal senders, i.e., the CSR content and communication strategy, but has ignored the perception and interpretation of signal receivers. In practice, the signaling effect hinges on not only the signal itself but also the signal receivers who make sense of the information [20]. CSR reports vary in format, content, and quality across firms due to their voluntary nature. They contain multi-dimensional information that can challenge the comprehension of readers. A holistic view is much needed on how CSR signals are perceived, interpreted, and evaluated by readers.

This paper fills the research gap by connecting the CSR signaling properties with the sensemaking processes of signal receivers. Sensemaking here denotes the processes by which readers assign meanings to firms’ CSR information and make interpretations on both the report and the firm [21]. Extending the current knowledge of CSR signaling effects, we outline a holistic picture of how CSR signals and stakeholders’ interpretation would work interactively to affect CSR reputation. Specifically, we reveal two types of signaling properties regarding CSR reports, i.e., the signal of firms’ authenticity (signal intent) and the signal of firms’ CSP (signal quality). We also explain how the two signaling properties interplay with readers’ recognition, ultimately shaping their attitudes toward firms.

In order to test our theoretical hypotheses, a survey study on 53 firms with 1521 respondents was conducted. In general, the better the readers’ comprehension of the CSR report is, the stronger the signals of authenticity and CSP they receive, and the better the report value and value fit are recognized: eventually, readers express more trust in the firm. Furthermore, two of the four signaling-sensemaking pathways are proved to be partial mediators, indicating that the relationship between comprehension of CSR reports and trust is partially mediated by the signaling-sensemaking process. We conducted several analyses to confirm the robustness of the research findings.

The rest of this paper is organized as follows: in the second section, we review the relevant literature and introduce our theoretical hypotheses. In the third section, we present our research methodology, including the survey design, the data collection, the development of our measures, and the reliability and validity of our scales. In the fourth section, we report the empirical results. We summarize the implications, limitations, and future directions of our research.

2. Literature Review and Hypotheses

2.1. The Formation of Reputation in CSR Disclosure

Our theoretical framework is shown in Figure 1. The framework aims to outline a holistic view of the intricate interplay between CSR signals and readers’ sensemaking processes, thereby predicting how corporate reputation is variably shaped. Firms voluntarily disclose CSR information as a way to signal a superior position regarding their social and environmental performance [22,23]. In light of the signaling theory, how CSR information resonates with readers depends on two types of signaling properties, i.e., the signal of firms’ authenticity (signal intent) and the signal of firms’ CSP (signal quality). However, the exclusive emphasis on the signaling properties without considering the heterogeneity of interpretations among signal receivers is inadequate to predict signaling outcomes. With received signals, readers would assign meanings to firms’ CSR information and thus form their attitude towards the firms [21]. Therefore, we delve into readers’ sensemaking processes and elucidate how CSR signals are interpreted by readers with different value orientations and thus variably shape their attitudes towards firms. As such, a signaling-sensemaking framework is provided for understanding the interplay between signaling properties and readers’ sensemaking processes and how they work in tandem to influence the formation of corporate reputation. In what follows, we discuss the development of the hypotheses in detail.

2.2. Signal Transmission Based on CSR Comprehension

According to signaling theory, stakeholders who receive CSR signals tend to be more aware of firms’ efforts on social and environmental issues, leading to a better firm reputation [22]. The change in readers’ attitudes is based on receiving and perceiving positive signals from CSR reports. As argued in prior literature, two types of information could be processed from the signals: information about intent and information about quality [20,24]. Information about intent allows the signal receivers to make attributions based on the behavior of the signalers, e.g., camouflage and requirements. In contrast, the information is considered high-quality if it is received with greater accuracy and completeness of the underlying abilities or characteristics of the signalers. In the context of CSR disclosure, the information about intent engages signal receivers in understanding the “CSR motives” of the firms, which is often referred to as CSR authenticity. Audiences might perceive firms’ CSR disclosure either as their genuine communication or as a way of hypocrisy and ingratiation with their stakeholders [25]. The CSR information quality largely hinges on how the corporate social performance (CSP) of the firm is perceived. High-quality signaling arouses beholders’ awareness of the distinctive characteristics of firms’ efforts on social and environmental issues as well as their potential impact on society. As such, beholders’ reception of firms’ CSR signals is constituted of two aspects, i.e., the signal of firms’ authenticity (intent) and the signal of firms’ CSP (quality).

The comprehension of CSR reports serves as the basis of signal transmission, engaging the beholders in perceiving the authenticity and CSP from the signals. Comprehension refers to “the activity of simultaneously extracting and constructing meaning through interaction and involvement with text” [26]. In this sense, the extent of readers’ engagement and involvement with the information influences the variation in their comprehension of CSR reports. The higher the level of the readers’ comprehension is, the more actively they are engaged in building perceptions. Guthrie et al. (2012) show that only with a certain level of comprehension would readers be able to effectively receive the information from reading the content [27]. Highly engaged readers tend to use more comprehension strategies to gain meaning from text and read deeply. As it is implied, when exposed to various signals from CSR reports, readers with a higher level of comprehension are able to receive the two types of signals more effectively and be more active in building perceptions. Therefore, we propose the following hypotheses:

Hypothesis 1a. 

Readers’ comprehension of the CSR report is positively associated with their perceived CSR authenticity of the firm.

Hypothesis 1b. 

Readers’ comprehension of the CSR report is positively associated with their perceived CSP of the firm.

2.3. Sensemaking Processes and CSR Evaluation

Sensemaking denotes a process by which actors interpret and assign meanings to certain occurrences and draw on the meanings to define an appropriate course of action [21]. With the received signals, readers assign meanings to firms’ CSR information and thus form their value recognition of both the report and the firm. Readers would assess the value of the report, especially in promoting people’s understanding of firms’ CSR efforts and impact. Moreover, readers would assess firms’ value largely on the basis of their own value orientations. The value fit, i.e., the congruence between audiences’ personal values and a given firm’s values, would substantially affect their attitude towards the firm [28].

In our research, perceived authenticity is assumed to affect the recognition of both the report value and the value fit. On the one hand, with perceived authenticity, readers believe the reported CSR efforts and have faith in the resultant impact, and thus tend to place greater value on the CSR information [29]. Therefore, perceived authenticity is expected to be positively associated with the recognition of report value, as expressed in Hypothesis 2a. On the other hand, authenticity can enhance the beholders’ perceived value of the firm. In the context of CSR disclosure, stakeholders, like employees, consider the authenticity of the firm’s CSR programs to ascertain the firm’s internal values and the consistency with their own values [30]. This implies that when readers perceive a higher level of CSR authenticity, they are more likely to recognize a good value fit between themselves and the firm, leading to Hypothesis 2b.

Hypothesis 2a. 

Readers’ perceived CSR authenticity of the firm is positively associated with their recognition of report value.

Hypothesis 2b. 

Readers’ perceived CSR authenticity of the firm is positively associated with their recognition of value fit.

Similarly, we propose that readers’ perception of CSP can affect their recognition of both the report value and the value fit. Firms’ CSP signals their contributions to social communities and the environment, which reflects firms’ value orientation. With a comprehensive understanding of the firms’ CSP, stakeholders could better identify how certain CSR initiatives of the firms fit with their own values [28,31]. Accordingly, we assume that readers’ perception of CSP is positively related to their perceived value fit. This argument corresponds to Hypothesis 3a. On the other hand, the high level of perceived CSP leads to stakeholders’ positive inferences about the firm’s underlying characteristics, consequently making them judge the report’s value more positively [32]. This implies the link between perceived CSP and the recognition of report value, leading to Hypothesis 3b.

Hypothesis 3a. 

Readers’ perceived CSP of the firm is positively associated with their recognition of value fit.

Hypothesis 3b. 

Readers’ perceived CSP of the firm is positively associated with their recognition of report value.

2.4. The Effects on Reputation

In turn, value recognition is considered a crucial factor that can improve a firm’s reputation [33]. CSR activities, as a means of ‘‘doing good”, have a strong positive effect on a firm’s reputation [34]. As a means of communicating how well the firm does, issuing CSR reports can also positively influence the attitude of stakeholders and the general public [35]. Therefore, we propose that readers’ recognition of report value and of value fit are related to a high level of trust toward the corresponding firm, corresponding to Hypotheses 4a and 4b, respectively. The recognition of report value allows readers to “make predictions, with some level of confidence, regarding the likelihood that [the firm] will live up to [its social] obligations” [36]. In this sense, firms that make substantial efforts in CSP have the incentive to convey this information through their reports by compiling CSR reports of high quality. In contrast, firms with poor CSP usually have less incentive to issue CSR reports or choose to disclose only limited information in their reports. In other words, readers can make inferences about not only the firm’s ability to conduct CSR but also its intention concerning CSR activities. Consequently, the recognition of report value can help the readers predict the level of a firm’s CSR expertise, which can be utilized to enhance the firm’s source credibility and trustworthiness. Besides, the recognition of value fit, which is more intrinsic and subjective, can strengthen the readers’ emotional connection with the firm, and thus, it is “capable of eliciting a bond of trust” [37].

Furthermore, we propose that the enhancement of trustworthiness is affected not only by the recognition of report value and value fit but also by the entire signaling-sensemaking process (from perception to recognition). Beholders mobilize the relevant cognitive resources to process the signals conveyed by the CSR reports and then form their attitudes [29]. Sparr et al. (2022) test the indirect effects of different sensemaking processes and performance outcomes [38]. Yu et al. (2022) suggest that trust creation can be viewed as a process of sensemaking in which signals are enlarged through the incremental accumulation of evidence [39]. These arguments imply that the signaling-sensemaking process is likely to mediate the relationship between the readers’ comprehension and their trust in the firm, leading to Hypothesis 5.

Hypothesis 4a. 

Readers’ recognition of report value is positively associated with their trust in the firm.

Hypothesis 4b. 

Readers’ recognition of value fit is positively associated with their trust in the firm.

Hypothesis 5. 

The relationship between the readers’ comprehension of the CSR report and their trust in the firm is partially mediated by the signaling-sensemaking process.

3. Methodology

3.1. Sample and Data Collection

To empirically test our hypotheses, we collected data from 5300 undergraduate students at several universities in China through a quantitative online survey. Regarding the respondents, there are four reasons why we chose university students as our respondents. First, it is convenient and easy for us to mobilize and recruit university students to respond to a survey. Second, they are knowledgeable enough to read and understand a CSR report and then form their own point of view. Third, compared with working people, college students are less influenced by factors such as interest-based relationships and prejudice in their responses to CSR reports. Finally, these students will be future employees of such companies, and thus, they have some interest in the CSR activities of companies, as CSR is one measure of the care of the companies toward customers, employees, and society.

In addition, we try to eliminate the potential influence of actual CSR conduct on readers’ trust in the following ways. First, the actual CSR conduct of all the firms adopted in our study is certified by CSR-CHINA.net and Beijing Rongzhi Corporate Social Responsibility Institution, both of which are professional firms specializing in CSR assessment. Second, all of the 53 firms used in our research are relatively large and well-known firms in China. Their reputation will be severely harmed if they issue CSR reports with falsified CSR performance. Thus, they have no incentive to tamper with their CSR reports. In addition, no misrepresentation by these 53 firms has yet been found and reported. Third, it is much costlier for a firm with weaker performance to engage in voluntary disclosure than it is for a firm with stronger performance.

For this sample, 5300 participants were invited to complete the web survey about CSR reports and their feelings about the corresponding firms. Finally, 1521 students completed our study. Forty-three percent of the participants were female, and their average age was 20.51 years (SD = 1.79 years). As an incentive to participate, each participant was issued a certificate for being a volunteer. In addition, participants were offered the possibility of winning a prize and certification if they completed the survey scrupulously. After constructing the survey database and removing observations that were missing key explanatory variables (approximately 7.6 percent), we have a sample of 1521 observations. The demographic characteristics of our survey sample are shown in

Table 1. Demographic characteristics of the survey sample (n = 1521).

Demographics	Percentage	Demographics	Percentage
Gender		Source of firm information
Male	42.9	Official website	32.3
Female	57.1	Social network	29.2
Age		Newspaper	9.1
18–19 years	27.2	Professional magazine	4.7
20–21 years	54.3	TV	18.7
22–23 years	15.1	Self-observation	5.9
>23 years	3.4	Monthly family income
Concern about firm		≤2000 yuan	15.7
No concern	26.4	2001–6000 yuan	56.4
General concern	51.2	6000–8000 yuan	30.1
Close concern	22.4	>10,000 yuan	7.8

.

Concerning the CSR reports read by students through the online survey, we adopted 53 Chinese firms, among which 33 were state-owned, 10 were private, and the other 10 were foreign corporations. We chose these firms to test our research hypotheses for the following reasons. First, they are relatively large and well-known in their respective industries. Second, all of them have paid some level of attention to at least one dimension of social responsibilities (including the economic, social, and environmental dimensions). Third, all the selected firms have issued CSR reports for more than eight years, and their CSR reports are of comparatively high quality.

3.2. Procedures

We illustrate the procedures of our survey in Figure 2. First, a pilot study was conducted before our main survey to assess the content validity of our scales. In the pilot study, more than 100 participants in managerial business were invited to pretest our questionnaire, and they noted the problems they encountered while answering questions. A number of small revisions to the questionnaire were made on the basis of the pilot study.

Then, we conducted our main survey. In our research, the longitudinal survey consisted of two stages, i.e., Stage 1 and Stage 2, with an interval of four weeks. Specifically, the Stage 1 survey aimed to determine how the readers perceived the CSR report and CSP in the sensemaking process, as well as their original trust in the corresponding firm. In contrast, the Stage 2 survey aims to find how their trust changes after the sensemaking process. Then, we test our theoretical model by incorporating all the items of the independent variables measured at Stage 1 and all the items of the dependent variable measured at Stage 2. We use the dependent variable measured in stage 2 to run the models, where the time-lagged measurement alleviates the endogeneity concerns that the dependent variable and independent variables are simultaneously influenced by unobservable factors.

With the help of China Youth Daily, 5300 participants were recruited by mail to finish the survey at Stage 1, and 3459 people of them replied, resulting in a response rate of 65 percent. Each participant was randomly assigned to read one of the 53 firms’ reports to balance the number of participants allocated to the same firm. Of the 3459 people who finished the survey at Stage 1, 1521 completed the Stage 2 survey, resulting in response rates of 46 percent at Stage 2 and 30 percent overall. There is no significant sample bias among the respondents at Stage 1 or Stage 2.

3.3. Measures

Measures for most of the constructs used in our study were available in past literature. We adapted them to suit the language environment of China. Unless otherwise specified, all items in our survey used a seven-point Likert rating scale ranging from 1, “strongly disagree”, to 7, “strongly agree”.

Readers’ comprehension. In this study, our main foci are whether and how the degree of a reader’s comprehension of a CSR report impacts their trust in the corresponding firm. Thus, we use one item that describes the extent to which readers understand the CSR report, which shows to what extent the participants understood the informative intention of the CSR report. This item used a seven-point Likert rating scale ranging from 1, “completely not understanding”, to 7, “complete understanding”. We also use one item describing to what extent the reader understands how the report is written. Intuitively, if the one-item construct “comprehension” is reliable, these two items should be highly correlated with each other. In fact, the results show that the correlation coefficient between them is significantly high (0.92), implying the validity of our measurement of readers’ comprehension.

Perceived authenticity. Authenticity broadly refers to “being true to oneself”, and one must be true to oneself within a social context in which others identify and accept one’s values and beliefs as appropriate [40]. Regarding CSR, authenticity involves a verification process about whether or not the disclosed CSR reflects the firms’ genuine motivations, organizational identities, and the real social impact that firms make [41]. We measured the perception of CSR report authenticity with three of the six items from the sense framing scale reported by Fiss and Zajac (2006) [42], which includes the items “Compared with other CSR reports or materials, this CSR report gives a more authentic portrait of the firm”, “This CSR report reflects the unique factors of the firm” and “This CSR report can fully reflect the personality of the firm”.

Perceived CSP. CSP is defined as firms’ commitment to principles, policies, and practices related to their social responsibilities and relationship with stakeholders [43]. We measure the perceived CSP by using the eight-item scale suggested by Jones et al. (2014) [44]. CSP can be considered in two aspects: community and environment. The four items for the community aspect of CSP were “This firm gives back to its community (locally, nationally, and/or internationally)”, “This firm takes part in voluntary or charitable activities”, “This firm is active in helping its community”, and “This firm tries to have a positive influence on its community”. The four items for the environmental aspect of CSP were “This firm has good environmental policies”, “This firm is concerned about environmental sustainability”, “This firm is trying to reduce its impact on the environment”, and “This firm is an environmentally friendly firm”. We also verify the reliability and validity of the scales used to measure the perceived CSP. First, Cronbach’s alpha coefficient is higher than 0.9, which verifies the reliability of the scales. Second, the composite reliability (CR) is far above the recommended level of 0.7. The average variance extracted (AVE) is higher than the recommended level of 0.5, and all the standardized factor loadings are close to or higher than the recommended level of 0.6. All this evidence confirms the convergent validity of the scales. Third, the square root of the AVE of CSP is greater than all of the correlation coefficients, which confirms the discriminant validity.

Recognition of report value. We measured the readers’ recognition of the CSR report value using a four-item scale, which embodied different aspects of value a CSR report can transmit to readers. The items are “This CSR report can help me understand the main information of the firm better”, “The firm should issue this kind of CSR report regularly”, “This CSR report improves my recognition of the firm’s overall operating conditions”, and “This CSR report increases my affection for the firm”.

Recognition of value fit. To measure the readers’ recognition of the firm value fit, we used the three-item scale from Cable and DeRue (2002) [45]. The items include first, “The things that I value in life are very similar to the things the firm values”; second, “The firm’s values and culture provide a good fit with the things that I value in life”; and third, “My personal values match the firm’s values and culture”.

Trust. Consistent with previous studies which regard trust as a multi-dimensional construct [46], we measure trust by using the six-item scale designed by Mayer and Davis (1999) to assess the ability, benevolence, and integrity of a firm [47]. The six items are “I think this firm has all the resources and conditions needed for performing social responsibility”, “I think the awareness of this firm of social responsibility is thorough and comprehensive”, “I think this firm has accumulated rich experience to fulfill its social responsibility in its specific field”, “I think this firm has done well in the aspect of philanthropic work”, “I think this firm has established the perfect social responsibility management system”, and “I think this firm will be determinate to perform its responsibility”. Similar to the perceived CSP, we have also verified the reliability and validity of the scales that measure trust.

Control variable 1–2: Firm Ownership Type (based on survey). Because firm trustworthiness is strongly related to firm ownership type, whether state-owned, privately owned or foreign-owned, we used two dummy variables as control variables to measure firm ownership type. The first one indicated whether the firm is state-owned (yes = 1, no = 0), and the second indicated whether the firm is privately owned (yes = 1, no = 0).

Control variable 3: Demographic characteristics (self-report). We controlled gender (female = 1, male = 0) as a demographic variable that might affect our estimates of trustworthiness. Controlling for gender is important because it may affect employees’ cognitive style and readers’ corporate evaluation [48].

Control variable 4: Initial impression (self-report). Since different people may hold different initial opinions of the same firm, we asked the participants to report their initial impression of the firm before reading the CSR report they had received. Initial impression reflects a collection of participants’ perceptions of firms’ important characteristics like size, industry, and public image. We used a seven-point Likert rating scale ranging from 1, “extremely bad”, to 7, “extremely good”.

3.4. Reliability and Validity of the Scales

Table 2. Confirmatory factor analyses for survey data.

Construct		Standardized Loadings
	Cronbach’s Alpha	Composite Reliability	Average Variance Extracted
Perceived authenticity	0.82	0.82	0.60
Recognition of report value	0.84	0.74	0.59
Perceived CSP	0.94	0.94	0.67
Recognition of value fit	0.86	0.74	0.68
Trust (Stage 1)	0.90	0.90	0.60
Trust (Stage 2)	0.92	0.92	0.66

shows that Cronbach’s alpha coefficients for all the constructs in our model are higher than 0.8, implying the reliability of the scales. Regarding the validity, we verify both the convergent validity and the discriminant validity. On the one hand, we find that the composite reliability (CR) of all constructs is higher than the recommended level of 0.7, and the average variance extracted (AVE) for all constructs is higher than the recommended level of 0.5 (see Table 2), both of which confirm the convergent validity. Moreover, all the standardized factor loadings are close to or higher than 0.60 (See

Table 3. Factor loadings in the measurement model.

Item	Stage 1 Loading	Stage 2 Loading
F1: Reader’s comprehension
1. Do you understand what informative intention this CSR report is trying to convey (V1)	0.63
F2: Perceived authenticity
1. This CSR report reflects the unique factors of the firm (V2)	0.61
2. Compared with other CSR reports or materials, this CSR report gives a more authentic portrait of the firm (V3)	0.67
3. This CSR report can fully reflect the personality of the firm (V4)	0.76
F3: Recognition of report value
1. This CSR report can help me understand the main information of the firm better (V5)	0.53
2. The firm should issue this kind of CSR report regularly (V6)	0.64
3. This CSR report improves my recognition of the firm’s overall operating conditions (V7)	0.75
4. This CSR report increases my affection for the firm (V8)	0.67
F4: Perceived CSP
1. This firm gives back to its community (locally, nationally, and/or internationally) (V9)	0.76
2. This firm takes part in voluntary or charitable activities (V10)	0.72
3. This firm is active in helping its community (V11)	0.76
4. This firm tries to have a positive influence on its community (V12)	0.80
5. This firm has good environmental policies (V13)	0.77
6. This firm is concerned about environmental sustainability (V14)	0.78
7. This firm is trying to reduce its impact on the environment (V15)	0.77
8. This firm is an environmentally friendly firm (V16)	0.76
F5: Recognition of value fit
1. The things that I value in life are very similar to the things the firm values (V17)	0.68
2. What I value in my life is very identical to what the firm pays attention to (V18)	0.67
3. My personal values match the firm’s values and culture (V19)	0.66
F6: Trust
1. I think this firm has all the resources and conditions needed for performing social responsibility (V20)	0.76	0.68
2. I think the awareness of this firm for social responsibility is thorough and comprehensive (V21)	0.79	0.75
3. I think this firm has accumulated rich experience to fulfill its social responsibility in its specific field (V22)	0.77	0.72
4. I think this firm has done well in the aspect of philanthropic work (V23)	0.74	0.67
5. I think this firm has established the perfect social responsibility management system (V24)	0.78	0.68
6. I think this firm will be determinate to perform its responsibility (V25)	0.77	0.69

), which also confirms the convergent validity. On the other hand, we compare the square root of AVE with the correlations of the focal variable with any other variables to test the discriminant validity.

Table 4. Means, standard deviations, correlations and internal consistency estimates.

Variable	M	SD	1	2	3	4
1. Reader’s comprehension	4.99	0.78
2. Perceived authenticity	4.91	0.90	0.32 **	(0.77)
3. Recognition of report value	5.27	0.87	0.47 **	0.54 **	(0.77)
4. Perceived CSP	5.45	0.89	0.45 **	0.49 **	0.56 **	(0.82)
5. Recognition of value fit	5.14	1.00	0.40 **	0.43 **	0.47 **	0.74 **
6. Trust (Stage 1)	4.01	0.87	0.39 **	0.48 **	0.59 **	0.55 **
7. Trust (Stage 2)	5.03	0.86	0.45 **	0.41 **	0.48 **	0.45 **
8. State-owned	0.71	0.45	−0.08 **	−0.05	−0.00	−0.02
9. Private-owned	0.14	0.34	0.06 *	0.05	0.04	0.01
10. Gender	0.62	0.48	0.02	0.02	−0.01	0.06 *
11. Initial Impression	4.77	1.02	0.28 **	0.32 **	0.32 **	0.24 **
Variable	5	6	7	8	9	10
5. Recognition of value fit	(0.82)
6. Trust (Stage 1)	0.46 **	(0.77)
7. Trust (Stage 2)	0.39 **	0.47 **	(0.81)
8. State-owned	−0.04	−0.01	−0.04
9. Private-owned	0.02	0.03	0.01	−0.64 **
10. Gender	0.05	0.03	0.04	0.03	−0.02
11. Initial Impression	0.24 **	0.30 **	0.28 **	−0.16 **	0.12 **	0.00

Note. n = 1521. The square roots of AVE for all constructs are displayed in parentheses on the diagonal: * p < 0.05, and ** p < 0.01.

reveals the mean, standard deviation, and square root of AVE of all the constructs in our study as well as their correlations. For each row, the square root of AVE is greater than all the correlation coefficients of the focal construct with any other construct, which confirms the discriminant validity. Furthermore, almost all the estimated correlations were below the recommended cutoff of 0.70 [49]. In addition, we computed the variance inflation factors (VIFs) to investigate whether there is a potential multicollinearity problem. If the VIF value of an independent variable is greater than the recommended threshold, the corresponding variable should be removed from the model since it can be explained by other independent variables. In our case, the maximum VIF of all the independent variables is 2.66, and the mean VIF is 1.94, both of which were far below the rule-of-thumb threshold of 10. Therefore, our empirical results are not influenced by the multicollinearity problem.

3.5. Analyses of Measurement Model

Since our data were subjective assessments from respondents, it cannot be denied that the relationships between the variables may be magnified by common method bias. We conducted Harman’s single-factor test to examine the effect of the common method in both exploratory factor analysis (EFA) and confirmatory factor analysis (CFA). First, by using EFA with maximum likelihood (ML) estimation and varimax rotation, we found that the twenty-five measurement variables in our model formed six factors with eigenvalues > 1 rather than a single factor.

Moreover, all of the variables were entered on one single factor to examine the fit of the CFA model. The results show that the single-factor model do not fit the data well: ${\mathsf{\chi }}^2\left[275\right]$ = 8735.59, NFI = 0.66, CFI = 0.67, RMSEA = 0.14, RMSEA confidence interval = 0.14, 0.15. Even though the results of the above analyses cannot guarantee the nonexistence of common method bias, they do indicate that common method bias is not of great concern and is thus unlikely to confound the interpretations of our constructs.

Furthermore, we assessed the validity and adequacy of the measurement model in our study using CFA and specifying a six-factor solution. The results display an excellent fit with the data: ${\mathsf{\chi }}^2\left[261\right]$ is 1035.12, the ${\mathsf{\chi }}^2/\mathrm{df}$ ratio is 3.96, the incremental fit index (IFI) of 0.96 and the comparative fit index (CFI) values of 0.97 both exceed the 0.90 thresholds, and the root-mean-square error of approximation (RMSEA) of 0.04 is lower than the 0.06 threshold. To examine whether the six-factor model is the most suitable solution, we compared it with plausible alternative solutions. The sequential tests demonstrate that the six-factor model fits the data significantly better than the other models do.

4. Analysis and Results

4.1. Manipulation Checks

To examine whether there was a significant difference between the attitudes of participants toward the same firm at Stage 1 and Stage 2, we conducted a paired t-test and found that all the dependent variables measured at Time 2 were significantly higher than those measured at Time 1. Readers’ trust in the corresponding firm is significantly higher at Time 2 (M = 5.03, SD = 0.87) than that at Stage 1 (M = 4.01, SD = 0.86). We also conducted analyses of variance (ANOVA) on the dependent variable to examine the extent to which readers in the manipulated survey time engaged in the evaluation of the firm before or after reading a CSR report.

Table 5. Mean differences among survey times on trustworthiness.

Variable	Stage 1		Stage 2			t-Test
	M	SD	M	SD	F	1 vs. 2
Item 1	4.28	1.14	5.19	1.05	10.77 **	21.34 **
Item 2	4.02	1.09	5.03	1.02	6.94 *	24.37 **
Item 3	4.06	1.16	5.16	1.00	17.50 ***	25.72 **
Item 4	3.87	1.02	4.94	1.01	5.93 *	26.24 **
Item 5	3.77	1.05	4.94	0.99	7.49 **	27.71 **
Item 6	3.93	1.03	4.95	1.04	4.15 *	24.57 **
Trust	4.01	0.87	5.03	0.86	9.07 **	34.01 **

Note. n = 1521; df for F-tests = 1754; df for t-tests 1520. * p < 0.05, ** p < 0.01, *** p < 0.001.

displays the tests of mean differences at different times throughout the survey procedure, which include omnibus ANOVA and contrasts between the two stages on the dependent variable in our model. The results of manipulation checks further confirm the rationality of the longitudinal design and our use of time-lagged measurement for the dependent variable.

4.2. Structural Equation Model

Our analyses of the effects of the readers’ comprehension of the CSR report on their trust in a firm are consistent with the pattern of relationships we predicted in Hypotheses 1–4. Since the sample size is adequate in terms of the respondent-to-item ratio, we tested our prior theoretical assumptions against empirical data with structural equation modeling (SEM), using AMOS software (version 22.0) with the ML estimation procedures. We examined the full structural equation model both with and without control variables to test our proposed hypotheses by using the Stage 1 data as the independent variable and the mediating variables and the Stage 2 data as the dependent variable.

First, we tested the structural equation model without control variables. As depicted in Figure 3, the model demonstrates a good fit with the data: ${\mathsf{\chi }}^2\left[267\right]$ = 1128.92, NFI = 0.96, CFI = 0.97, RMSEA = 0.04, RMSEA confidence interval = 0.04, 0.05. The results show that the readers’ comprehension of the CSR report has a significantly positive relationship with both the perception of the CSR report’s authenticity (β = 0.63, p < 0.001) and the perceived CSP (β = 0.88, p < 0.001). Furthermore, the relationship between the perception of the CSR report’s authenticity and the recognition of the CSR report value is significant and positive (β = 0.45, p < 0.001). The relationship between the perception of the CSR report’s authenticity and the recognition of value fit is also significant and positive (β = 0.10, p < 0.001). Moreover, the perceived CSP is positively associated with the recognition of value fit (β = 0.79, p < 0.001) and the recognition of report value (β = 0.38, p < 0.001). In addition, both the recognition of report value (β = 0.43, p < 0.001) and value fit (β = 0.22, p < 0.001) are positively associated with trust in the corresponding firm. The above results support Hypotheses 1a, 2b, 2a, 2b, 3a, 3b, 4a, and 4b.

Second, the existence of reverse causality between dependent variables and dependent variables would potentially incur endogeneity problems and generate estimation bias. Therefore, we check for the existence of reverse causality in our model by testing an alternative model. This model changes the final dependent variable in our initial model, trust in the corresponding firm, into the only exogenous independent variable. However, this model displays a poorer fit and fails to achieve the threshold values proposed by Hu and Bentler (1999) [50] (${\mathsf{\chi }}^2\left[268\right]$ = 1478.04, NFI = 0.91, CFI = 0.93, RMSEA = 0.06, RMSEA confidence interval = 0.05, 0.06). Consequently, the nonexistence of reverse causality in our initial model can be proved.

Then, we tested the structural equation model with control variables. As shown in Figure 4, the model still displays a good fit after control variables are added (${\mathsf{\chi }}^2\left[369\right]$ = 2312.48, NNFI = 0.92, CFI = 0.93, RMSEA = 0.06, RMSEA confidence interval = 0.05, 0.06). Both the model with and the model without control variables show evidence to support our hypotheses.

To examine whether the readers’ comprehension of the CSR reports substantially explains the variance in predicting trust in the corresponding firm, we conducted a hierarchical regression analysis. When compared to a model containing only control variables, the inclusion of the independent variable, readers’ comprehension of the CSR report, results in a significant increase in the explained variance (from ${\mathrm{R}}^2$ = 0.08, F [4, 1516] = 32.22, p < 0.001 to ${\mathrm{R}}^2$ = 0.15, F [1, 1515] = 54.63, p < 0.001). Furthermore, readers’ comprehension of the CSR report demonstrates a significant positive relationship with trust in the corresponding firm (β = 0.25, p < 0.001).

4.3. Mediation Analysis

As shown in the previous analyses, the mediators were significantly related to not only the independent variable—the readers’ comprehension of the CSR report—but also the dependent variable—their trust in the corresponding firm. We also examined whether the mediated paths improved the model fit by testing a nested model in which we added a direct path from the readers’ comprehension of the CSR report and trust in the corresponding firm and removed all the mediating sensemaking pathways. Despite the decrease in parsimony, the mediated model displays a significantly superior fit than the nested model (${\mathsf{\chi }}^2\left[21\right]$ = 183.11, NNFI = 0.94, CFI = 0.95, RMSEA = 0.12, RMSEA confidence interval = 0.10, 0.14).

In our model, there are four pathways of mediation. Testing multiple mediators together is a good way to reduce parameter bias as a result of omitted variables. To examine whether the intervening variables mediated the relationship between the readers’ comprehension of the CSR report and their trust in the corresponding firm, we followed the procedures recommended by James et al. (2006) [51]. We estimated the indirect effects with the coefficients from the full model and then used bootstrapping methods to test the model with multiple mediators. We constructed the confidence intervals based on 1000 random samples to take the place of the full sample, along with bias-corrected 95 percent confidence intervals. Compared with other methods to test indirect effects in complex models, this approach indicated the most accurate confidence intervals. Our model is a multiple two-stage mediation model, where each indirect effect is the product of three paths, including [F1$\to$F2] $\times$ [F2$\to$F3] $\times$ [F3$\to$F6] and [F1$\to$F2] $\times$ [F2$\to$F5] $\times$ [F5$\to$F6] for sensemaking about the CSR report, [F1$\to$F4] $\times$ [F4$\to$F5] $\times$ [F5$\to$F6] and [F1$\to$F4] $\times$ [F4$\to$F3] $\times$ [F3$\to$F6] for sensemaking about the CSP of the firm.

Table 6. Results of mediation tests predicting trust in the corresponding firm: indirect effects of reader’s comprehension of CSR report through two mediation pathways.

Indirect Effects	BC 95% CI
	Estimate	Lower Bound	Upper Bound
	Predictor: Reader’s Comprehension
Total indirect effect	0.19 **	0.09	0.32
Unique indirect effects through:
1. [F1→F2] × [F2→F3] × [F3→F6]	0.09 **	0.05	0.15
2. [F1→F2] × [F2→F5] × [F5→F6]	0.00	−0.00	0.01
3. [F1→F4] × [F4→F5] × [F5→F6]	0.00	−0.06	0.09
4. [F1→F4] × [F4→F3] × [F3→F6]	0.10 **	0.05	0.16

Note. BC 95% CI refers to the bias-corrected 95% confidence interval; Estimate refers to the effect estimate using 1000 bootstrap samples: ** p < 0.01.

illustrates the results of the mediation tests. The coefficient for the path of sensemaking about the CSR report [F1$\to$F2] $\times$ [F2$\to$F3] $\times$ [F3$\to$F6] is 0.09, and the ninety-five percent confidence interval is (0.05, 0.15). The coefficient for the path of sensemaking about the CSP of the firm [F1$\to$F4] $\times$ [F4$\to$F3] $\times$ [F3$\to$F6] is 0.10, and the ninety-five percent confidence interval is (0.05, 0.16). The coefficients of the other two paths, [F1$\to$F2] $\times$ [F2$\to$F5] $\times$ [F5$\to$F6] and [F1$\to$F4] $\times$ [F4$\to$F5] $\times$ [F5$\to$F6], are not significantly different from zero after adding the direct effect. Since the estimates of the total effect, direct effect and some indirect effects with ninety-five percent confidence intervals that do not include zero were all statistically significant, two of the four sensemaking pathways are proven to be partial mediators. The indirect effect versus the total effect is 0.35. Thus, part of our expectation about the plausible importance of the sensemaking process of the CSR report in improving the trust in the firm can be confirmed. Based on the above results, Hypothesis 5 was partly supported.

4.4. Robustness Analysis

We conducted several additional analyses to examine the robustness of our results. First, since the samples contained participants reading the CSR reports of three types of firms in terms of ownership, we categorized three samples according to the firm ownership type to test whether our results held. As anticipated, we achieved the same results with the different samples (see

Table 7. Results of robustness check.

Hypothetic Pathway	Standardized Coefficient
	Stated-Owned (n = 1085)	Private-Owned (n = 214)	Foreign (n = 222)	Full Sample (n = 1521)
H1a: F1→F2	0.60 **	0.52 **	0.82 **	0.63 **
H1b: F1→F4	0.88 **	0.97 **	0.82 **	0.88 **
H2a: F2→F3	0.42 **	0.45 **	0.61 **	0.45 **
H2b: F4→F3	0.44 **	0.37 **	0.13 **	0.38 **
H3a: F2→F5	0.11 **	0.10 **	0.08 **	0.10 **
H3b: F4→F5	0.78 **	0.81 **	0.82 **	0.79 **
H4a: F3→F6	0.41 **	0.45 **	0.47 **	0.43 **
H4b: F5→F6	0.22 **	0.20 **	0.21 **	0.22 **

** p < 0.01.

). Although the values of the coefficients in the three models differed slightly, both the significance and symbol of the coefficients stayed invariant when compared with our previous analysis results.

Second, we also examined the robustness of our model to other evaluations to provide additional validation. We used the overall impression of the firm as our dependent variable. We found that all our hypotheses held. The results show that the effect of the readers’ comprehension of the CSR report remains significant for both the perception of CSR report authenticity (β = 0.63, p < 0.001) and perceived CSP (β = 0.88, p < 0.001). Furthermore, the relationship between the perception of CSR authenticity and the recognition of report value is significant and positive (β = 0.44, p < 0.001). The relationship between the perception of CSR authenticity and the recognition of value fit is also significant and positive (β = 0.10, p < 0.001). Moreover, the perceived CSP is positively related to the recognition of value fit (β = 0.79, p < 0.001) and the recognition of report value (β = 0.38, p < 0.001). In addition, the recognition of report value (β = 0.28, p < 0.001) and of value fit (β = 0.08, p < 0.001) are positively related to the trust in the corresponding firm.

Third, we also tested the model in which trust is decomposed into three parts, i.e., ability, benevolence and integrity [40]. The empirical results show that both the recognition of report value and the recognition of value fit are positively related to all of the three dimensions of trust, which are consistent with the findings in our original model.

5. Conclusions and Discussion

By combining signaling and sensemaking perspectives, our paper outlines a framework of how reputation is formed after stakeholders receive the signals from CSR reports. The signaling-sensemaking framework highlights two types of CSR signaling properties, i.e., the signal of firms’ authenticity and the signal of firms’ CSP, and elucidates how the sensemaking paths corresponding to each type of signal would influence readers’ trust. We find that the better the readers’ comprehension of the CSR report, the stronger the signals of authenticity and CSP they receive, the better the evaluation of report value and value fit they give, and eventually, the more trust they express in the firm. Furthermore, the relationship between comprehension of a CSR report and trustworthiness is partially mediated by signaling-sensemaking processes, and the mediated paths have been proven to improve the model fit.

5.1. Theoretical Implications

Firstly, this study enriches the literature on micro-foundations of strategic CSR by revealing how CSR disclosure would be perceived by individual stakeholders and thus influences corporate reputation. The strategic view of CSR emphasizes the benefits of CSR for firms’ reputation, goodwill, and stakeholder support [3,4]. While existing researches mainly focus on firms’ motivation and organizational outcomes of CSR disclosure, the underlying mechanisms of how CSR would be perceived and evaluated by stakeholders remain underexplored. This is a critical knowledge gap since stakeholders’ attitude towards firms’ CSR plays a key role in mediating the relationships between CSR disclosure and organizational outcomes [16]. According to instrumental stakeholder theory, the strategic benefit of CSR to firms stems from the value recognition and behavioral reciprocity of stakeholders [52]. Only by winning stakeholders’ recognition and trust could a firm’s CSR promotes its sustainable competitive advantage [4]. The framework proposed and empirically testified in this paper provides a psychological lens for understanding how firms’ CSR disclosure resonates with stakeholders. In light of signaling theory and sensemaking theory, the framework elucidates the processes of how stakeholders receive and evaluate the CSR signals and thereby variably shape their attitudes towards firms’ CSR. As several studies have indicated the potential benefits of CSR disclosure [35,53], our findings complement this strand of literature by unveiling the psychological mechanisms of shaping positive perceptions and evaluations among stakeholders. Our findings also confirm the importance of stakeholders’ perceptions and value orientations in shaping the outcomes of CSR initiatives [54].

Secondly, this research adds new knowledge to signaling theory in the context of CSR disclosure. Signaling theory explains how CSR information activates the stakeholders’ goodwill and thus could positively affect firms’ reputations [22]. The extant research explaining the relationships between CSR signals and reputation has mainly focused on the heterogeneity in CSR content and firms’ communication strategies. However, they largely ignored the value recognition processes on the side of signal receivers. CSR contains multi-dimensional information, which makes it difficult for signal receivers to obtain a consensus comprehension [28]. It is argued that high information asymmetry about CSR activities exists between the firms and the public, which hampers the credibility of CSR reports [54]. Therefore, how CSR signals are perceived, interpreted and evaluated by the readers is crucial for the formation of reputation. The theoretical framework in this study connects signaling effects with the sensemaking processes of signal receivers, thereby providing a holistic view of how the CSR signals would work interactively with stakeholders’ sensemaking processes to affect corporate reputation. Prior researchers proposed two important signal properties, i.e., signal intent and signal quality [20,24]. We annotate the signal properties in the context of CSR disclosure as the signal of firms’ authenticity and the signal of firms’ CSP. In addition, our signaling-sensemaking framework highlights how different signal properties signals would be mediated by readers’ recognition. As such, the findings extend the understanding of CSR signaling by mapping the psychological processes of how reputation is formed after readers receive the signals.

5.2. Practical Implications

Many firms are under suspicion of disseminating false or incomplete CSR information, making it ineffective for firms to gain recognition and support from stakeholders. It is imperative for managers to have a deeper understanding of how CSR is perceived and evaluated by stakeholders. Our results highlight the interplay between the signaling properties of CSR contents and the sensemaking processes of readers for improving the trustworthiness of CSR disclosure. Consequently, managers should not only take into account the contents of the CSR reports but should also try to gain a deeper understanding of the value orientation of their stakeholders who would likely read the CSR reports. By understanding how stakeholders would interpret firms’ CSR information, managers are able to develop CSR reports in ways that enhance stakeholders’ comprehension and engage them more effectively in perceiving authenticity and CSP. Besides, when communicating a firm’s values through CSR disclosure, managers could demonstrate a higher level of value fit with firms’ key stakeholders, thus enabling firms to obtain better evaluation and support.

5.3. Limitations and Suggestions for Future Research

This research has some limitations. First, the participants of our study are all undergraduates. Even though they are less influenced by prejudice and interest-based relationships than are working people, which implies that they are “cleaner” participants, we cannot rule out the possibility that they have their own pre-existing knowledge or evaluation of the firm. In our study, we controlled the readers’ initial impression of firms before they read the CSR reports. However, it might be helpful to use a longitudinal survey design with fictitious firms and include participants with different characteristics to test our theoretical framework. Second, we use the self-report method to measure readers’ comprehension of CSR reports. However, since the self-reported level of comprehension may not reflect the actual extent of comprehension, it would be better to use some objective items to complement this data. We did not do so because, in our research, the CSR reports come from 53 real firms and have different content, and thus, it is difficult to measure readers’ comprehension using the same objective items. Nevertheless, a longitudinal experimental design may make it possible to include objective measurement items because researchers may more easily manipulate the level of clarity and readability of the CSR reports. Third, some firm characteristics, like firm size and industry, are important factors that potentially influence organizational outcomes regarding CSR reputation. Further research could use sub-samples of different firm sizes or industries so as to further test the mechanisms of how stakeholders evaluate CSR reports for different kinds of firms.