Communal Risk Information Sharing: Motivations Behind Voluntary Information Sharing for Reducing Interdependent Risks in a Community

Abstract

This article extends our understanding of risk communication related to communal risk and risk information sharing. Building on research from risk communication, organizational behavior, and social psychology, it examines individual-, relation-, and community-level motivations to share information about a devastating plant disease. This disease can bring about substantial economic risk to everyone in a farming community. We tested our hypotheses using a national sample of U.S. tomato and potato growers (N = 452). Our findings show that growers were motivated to share information about a communal risk based on (a) individual-oriented concerns for economic costs, (b) relation-oriented concerns for reciprocation and the information recipient’s trustworthiness, and (c) community-oriented concerns comprising a sense of shared responsibility and community cohesiveness.

Keywords

risk information sharing communal risk attitude perceived norm reciprocity shared responsibility cohesion outcome interdependence late blight

In an increasingly global society, we are often reminded of how interdependent we have become concerning the need to rely on others for the surveillance and reporting of shared or communal risks. Communal risks can be defined as risks that pose danger to all members in a collective or a community, and that require some level of social response to manage or reduce. Examples of communal risks range from public health risks such as influenza, tuberculosis, and Ebola; privacy risks such as computer viruses, malware, or hacking; and security risks such as threats of terrorism; to agricultural risks such as avian influenza. A shared characteristic of communal risks is their interdependence, meaning their ability to spread from party to party.

Of particular interest to this study is the role communication may play in the spread or containment of a communal risk. We are specifically interested in what motivates an individual to share with other members of a community that she or he has introduced a risk into the community. Although sharing this information with others can help contain the spread of a communal risk, there are also disincentives to sharing information about one’s own risk status. Some of these disincentives have been well-documented in the literature related to HIV self-disclosure (Parsons, VanOra, Missildine, Purcell, & Gómez, 2004). We question whether motivations to disclose personal health risks in an interpersonal setting (e.g., sharing one’s HIV status with a partner) would be the same for risks potentially affecting a broader community. In other words, understanding what motivates people to share information about a communal risk, which may necessitate explicit cooperation to contain, requires a full consideration of a variety of factors, ranging from self-centered evaluation of immediate costs and gains to concerns for the welfare of the whole community.

This article considers the case of late blight, which poses a significant economic threat to some agricultural communities. Late blight is an infectious plant disease that affects tomatoes and potatoes, killing the plant and rendering the product and seed economically useless to growers. In the 1840s, late blight swept through Europe, depleting potato supplies and causing the Irish potato famine, which resulted in over one million deaths and a similar number of emigrations out of Ireland (Donnelly, 2001). Late blight is still prevalent today. In 2009, one of the largest outbreaks of late blight in recent history spread throughout much of the northeastern United States (Fry et al., 2012). Across the world, the annual loss caused by late blight in potato production has been estimated at more than US$6 billion (Haverkort, Struik, Visser, & Jacobsen, 2009), suggesting the scope of potential damage and societal influences of this disease.

Like valuable information in other contexts such as organizations or interactive media (Fulk, Flanagin, Kalman, Monge, & Ryan, 1996; Kollock, 1999), shared information of a communal risk such as late blight is a collective good. The risk of late blight can be reduced if growers communicate with relevant parties, such as neighbors, buyers, and extension educators when late blight infects their farms. Particularly, risk information sharing enables timely intervention (Yuan, Steinhardt, McComas, Gay, & Smart, 2013). Studies confirm that the responsible pathogen can complete a cycle of reproduction in as quickly as 96 hours and be dispersed in a matter of minutes (see Fry et al., 2012). Although fungicides can curb the contagion, they are expensive and ineffective if applied too early, especially when fungicides are quickly removed by rain—a typical weather condition of late blight outbreaks. Fungicides can also be ineffective if applied too late, which can happen because the development of late blight into visible symptoms often takes about 60 hours, and in a few days, the host plant dies. If growers who have infected crops share the risk information with others, other growers in the community will be able to take timely action to stop the spread of the disease.

Risk information sharing also nurtures cooperation and carries benefits over the long term. Late blight can reoccur (Fry et al., 2012). No growers want to see late blight come back again because others did not or were not able to take timely action due to lack of information (Yuan et al., 2013). Moreover, growers often hope that others will share the risk information with them in the future, so that they themselves can take timely action. Hence, open, timely communication is vital to reducing reoccurring risks of late blight at the collective level.

When exploring factors that can motivate risk information sharing, we found that existing risk communication research could be further developed in the following ways. First, most studies on risk information sharing have focused on individual risks, meaning those that can be managed primarily independently. Although some scholars (e.g., Yang, Kahlor, & Griffin, 2014) have studied risk information sharing about climate change, this type of information sharing does not differ substantially from the propagation of experts’ knowledge to the public and primary groups (e.g., family and friends). Furthermore, participants in this study did not have the degree of interdependence and a heightened sense of urgency that we are investigating. In comparison, we focus on what motivates communication to contain the spread of a risk when each member’s risk depends on the actions or inactions of others.

Second, existing risk communication studies often focus on seeking and processing risk information (e.g., Braun & Niederdeppe, 2012; Griffin, Dunwoody, & Neuwirth, 1999; Kahlor, 2010; Rimal, 2001), but not on information sharing. Although these studies have influenced previous theorization of risk information sharing (e.g., Yang et al., 2014), they have not directly addressed issues beyond individual behaviors. That is, when interdependence among individuals increases, people have to balance conflicting interests between individual members and the collective as a whole (de Dreu, Nijstad, & van Knippenberg, 2008; de Dreu, Weingart, & Kwon, 2000). While others have tested the effect of a general normative belief about risk information sharing (Yang et al., 2014), we believe that more specific motivational factors across different levels of social entity should be examined.

To frame our exploration of the motivational factors, we started from the theory of reasoned action (TRA; see Fishbein & Ajzen, 2011). TRA has been widely used to study how evaluative and normative concerns influence behavior in a variety of settings, including risk information processing (Griffin et al., 1999), health (Montano & Kasprzyk, 2008), organizations (Wang & Noe, 2010), and consumer behavior (Sheppard, Hartwick, & Warshaw, 1988), among others. However, to use TRA as a guide to study communal risk information sharing, we suggest the following extensions regarding individual-, relation-, and community-oriented concerns.

First is explicit examination of the perceived costs of sharing because disclosing that one’s farm has been infected by late blight infection can economically disadvantage the discloser. Altruism should also be included as an alternative motivation because altruistic behaviors can be psychologically rewarding (Piliavin & Charng, 1990), and thereby motivate cooperative behavior regardless of economic cost. Next, moving beyond individual-oriented motivations, we believe that relation-oriented concerns such as a belief in reciprocity can help offset the calculation of immediate payoffs and lead to greater willingness to cooperate for long-term benefits. A discloser’s trust in a recipient’s capacity for benevolent use of shared risk information might also influence the discloser’s motivation to share. Finally, we expand our investigation to community-oriented concerns, including perceptions of shared responsibility and community cohesiveness. A sense of shared responsibility reflects one’s understanding of the interdependence of the outcome, whereas a perception of community cohesiveness indicates solidarity and one’s commitment to a community. It is possible that both concerns motivate cooperative behaviors.

The following sections review the core premises of TRA, along with arguments and findings from studies on organizational knowledge sharing, social exchange, and trust. We borrowed findings on knowledge sharing within organizations because as stated above, existing risk communication research has not paid much attention to the issue of information sharing among stakeholders. We then present a test of our hypotheses using survey data collected from a national sample of tomato and potato growers in the United States. The article ends with a discussion of results, along with theoretical and practical implications, limitations, and future directions.

Attitude and Perceived Norms of Sharing Information

TRA maintains that both attitudinal and normative beliefs can affect behavior (Fishbein & Ajzen, 2011). Attitudinal beliefs refer to one’s evaluations about favorable and unfavorable outcomes led by a behavior, whereas normative beliefs refer to one’s expectation of social rewards and punishments led by this behavior (Fishbein & Ajzen, 2011). These beliefs form an attitude and perceived norm of the associated behavior, respectively. Once formed, the attitude and perceived norm can determine a person’s behavioral intention, which in turn motivates the behavior. As noted above, attitude and perceived norm have been found to predict intentions of various behaviors consistently, including risk information seeking/processing (Griffin et al., 1999) and organizational knowledge sharing (Bock, Zmud, Kim, & Lee, 2005; see Wang & Noe, 2010, for a review). Consistent with conceptual predictions and existing findings per TRA on the influence of attitude and perceived norm on human behavior, we hypothesize the following:

Hypothesis 1 (H1): A favorable attitude toward sharing risk information is positively related to a person’s intention to share information with others.

Hypothesis 2 (H2): A perceived norm of sharing risk information is positively related to a person’s intention to share information with others.

Individual-Oriented Concerns: Payoff and Enjoyment

Concerns for Economic Costs

Although community members might understand that full information sharing across a community is beneficial to both the community and themselves, they might still feel reluctant to do so. A key to the reluctance is concerns for immediate costs and benefits (Hardin, 1982). When blinded by immediate costs, people might lose sight of the long-term benefits of cooperation (Hardin, 1982). According to Cabrera and Cabrera (2002), the most concerned cost for information and knowledge sharing in organizations is post-sharing vulnerability because sharing might reduce the sharer’s chance to outperform others. Communal risk information sharing can bring such vulnerability in the form of direct economic cost. For example, growers are often concerned that after disclosing the information of late blight infection to consumers, consumers would not buy infected produce any more (Yuan et al., 2013). Seed buyers would also avoid buying infected potato tubes because they can spread the disease (Fry et al., 2012). Accordingly, we hypothesize the following:

Hypothesis 3 (H3): Concerns for economic cost are negatively related to having a favorable attitude toward sharing risk information.

Concern for Spending Time and Effort

Information sharing also demands time and effort (Kankanhalli, Tan, & Wei, 2005). Although sharing information about one’s own risk to others in a community might require less formal documentation compared with sharing knowledge in formal organizations, such sharing might still be effortful. On one hand, growers need to compile information carefully so as to reduce potential disadvantages. They want to convey that the infection has been controlled and that the quality of produce is not compromised. On the other hand, the urgency of dealing with a communal risk might limit one’s time and effort spent on information sharing. Because growers might need to spend intensive efforts and time on prevention and/or intervention against late blight given the pathogen’s fast reproduction and spread, risk information sharing may be rendered less a priority. Taken together, a grower’s concern of spending time and effort might compromise a favorable attitude toward risk information sharing, as hypothesized here:

Hypothesis 4 (H4): Concerns for time and effort spent on sharing risk information are negatively related to a favorable attitude toward sharing.

Enjoyableness of Altruism

In the absence of material benefits, psychological rewards such as the enjoyableness of altruistic behaviors can also motivate information sharing (Andreoni, 1990). Here, we adopted a motive-based definition of altruism, which refers to those acts that are motivated “out of a consideration of another’s needs rather than one’s own” (Piliavin & Charng, 1990, p. 30). One source of altruistic motivation is enjoyableness derived from helping behaviors (Smith, Keating, & Stotland, 1989). Studies have shown that a belief in the enjoyableness of altruistic behaviors can motivate information and knowledge sharing in organizations and virtual communities (Constant, Kiesler, & Sproull, 1994; Kankanhalli et al., 2005; Wasko & Faraj, 2000). Such a belief might even be more relevant for sharing risk information, given the lack of immediate material rewards. Therefore, we hypothesize the following:

Hypothesis 5 (H5): The enjoyableness of altruism is positively related to having a favorable attitude toward risk information sharing.

Relation-Oriented Concerns: Reciprocity and Trust

Belief in Reciprocity

Sharing risk information might result in few immediate material rewards when people are responsible for their own risks and cannot receive substantial aid from others immediately after sharing. Despite the lack of immediate benefits, sharing information regarding a communal risk might still be rewarding through future reciprocations. Reciprocity is an important human preference that sustains cooperation with few immediate rewards (Fehr & Gintis, 2007). Cooperation is often not a one-shot deal. As a result, individuals learn to consider not only immediate payoffs but also future ones (Axelrod, 1984). Scholars have found that people in organizations share information and knowledge with expectations of reciprocity in the future (Bock et al., 2005; Kankanhalli et al., 2005). Sharing information about a communal risk can also trigger a consideration of future reciprocity when a risk such as late blight is expected to reoccur (Fry et al., 2012). To reduce the recurring risk, information about late blight infection from other community members is often the most updated and accurate because the infection might not be readily observable outside an infected site. Therefore, we hypothesize the following:

Hypothesis 6 (H6): Belief in the reciprocity in risk information sharing is positively related to having a favorable attitude toward sharing.

Reciprocity is also backed up by norms that internalize reciprocation as morally appropriate (Gouldner, 1960). Reciprocity per se also entails sanctions such as the tit-for-tat strategy that often predominates in iterated social exchanges (Axelrod, 1984). A refusal to reciprocate can meet with punishments by others, even at the cost of the punishers (Fehr & Gintis, 2007). Norms of reciprocity, along with others’ expectations of reciprocation, can be more salient for believers of reciprocity. Particularly, a belief in the reciprocity of communal risk information sharing might raise a normative concern to share the information and/or a pressure about being sanctioned for not sharing. We thus hypothesize the following:

Hypothesis 7 (H7): Belief in the reciprocity of risk information sharing is positively related to a perceived norm of sharing.

Recipient Trustworthiness

Trustworthiness of information recipients is an important factor that might affect information providers’ willingness to cooperate under risky situations (Hardin, 2002; Kollock, 1994; Molm, Takahashi, & Peterson, 2000). A recipient is considered trustworthy if the source believes that the recipient will take into consideration the source’s interests (Hardin, 2002). This interest-based view of trustworthiness is appropriate because as discussed earlier, sharing information of late blight infection at one’s own farm might have significant financial effects. This view of trustworthiness also emphasizes a recipient’s cooperative intention, which was found to be more important than other aspects of trustworthiness (e.g., capability) in risk-management settings (Earle, 2010). Therefore, a recipient’s trustworthiness regarding benevolent intentions can influence a grower’s intention to share, as postulated here:

Hypothesis 8 (H8): Perceived trustworthiness of the recipient is positively related to the source’s intention to share risk information.

Community-Oriented Concerns: Outcome Interdependence and Cohesion

Perceived Shared Responsibility

Communal risk information sharing is affected by both personal interests and community-based concerns. Sharing and reciprocation of sharing are a type of social exchange, which ensues from outcome interdependence among actors who hold resources valued by one another, or who can jointly act to benefit all parties (Emerson, 1972). Studies have shown that outcome interdependence influences information sharing and integration in task teams (de Dreu, 2007). To promote prosocial behaviors in a collective entity, outcome interdependence needs to be salient as a sense of shared responsibility (Lawler, Thye, & Yoon, 2008). Given the interdependent nature of communal risk, perceived shared responsibility can be an important antecedent of risk information sharing, particularly when this perception entails others’ expectations of one’s own cooperative behaviors. We thus hypothesize the following:

Hypothesis 9 (H9): Perceived shared responsibility in addressing a communal risk is positively related to perceived norm of sharing.

Perceived Community Cohesiveness

Social exchange such as risk information sharing entails uncertainties when the other parties refuse to cooperate (Kollock, 1994; Molm et al., 2000). To reduce such uncertainties, having an institution, either formal or informal, often helps (Yamagishi, Cook, & Watabe, 1998). A cohesive community characterized by close relationships among community members and their commitments to the community (Friedkin, 2004) can serve as such an institution. Therefore, perceived community cohesiveness, through promoting collective goods and reassuring one’s concern of others’ uncooperative behaviors (Kanter, 1968), might help to form a favorable attitude toward risk information sharing. A cohesive community can also promote prosocial norms because close relationships among community members often facilitate normative regulations (Coleman, 1988). A commitment to a community can also induce conformism to community norms (Kanter, 1968). When risk information sharing can benefit the entire community, perceived community cohesiveness can render the sharing behavior morally appropriate. Taken together, perceived community cohesiveness might facilitate formation of both a favorable attitude toward and a perceived norm of communal risk information sharing, as postulated here:

Hypothesis 10 (H10): Perceived community cohesiveness is positively related to having a favorable attitude toward sharing communal risk information.

Hypothesis 11 (H11): Perceived community cohesiveness is positively related to a perceived norm of communal risk information sharing.

Method

Sample and Data Collection

We contracted with a survey company to conduct computer-assisted telephone interviews with a sample of U.S. tomato and potato growers. The sample size was targeted at 250 tomato growers and 250 potato growers, which were reached through the following procedure. First, we selected farms that grew more than 1 acre of either tomatoes (8,273 farms, 32.05% of all tomato farms) or potatoes (6,137 farms, 40.88% of all potato farms) through a registration list collected by the U.S. Department of Agriculture. We set the 1-acre threshold to increase the chances of reaching farms that constantly grow the two crops. Second, the company randomly called these farms’ telephone numbers until the target sample size for data collection was reached. In total, 6,984 telephone numbers were dialed, 4,396 were valid numbers, and 3,027 calls were answered. Third, 1,595 growers agreed to proceed to a screening question, which asked about whether their farm grew either tomatoes or potatoes. The telephone interview continued with 466 qualified respondents. These respondents were then informed of the purpose of this study, compensation (a US$35 check) for completing the survey, the estimated interview time (15-20 minutes), and confidentiality policies in accordance with the authors’ institutional review board. Respondents then proceeded to our survey questions. Altogether, 452 respondents (227 tomato growers and 225 potato growers) completed the survey. By standard definitions from the American Association for Public Opinion Research (AAPOR, 2008), the Cooperation Rate 3 was 97%, assuming “those unable to do an interview as also incapable of cooperating” (p. 36). The Response Rate 3 was 22.2%, accounting for eligible respondents in unknown cases with e = .292 (see AAPOR, p. 35).

Measurements

Unless otherwise noted, the survey items were measured using a 6-point scale, ranging from 1 (strongly disagree) to 6 (strongly agree), along with a not applicable coded as missing. Given the length limit of a telephone survey, most multidimensional constructs were measured with one item per dimension. As Cronbach’s alphas underestimate reliability in this case (Little, Lindenberger, & Nesselroade, 1999), they are reported below for references only.

Behavioral intention

As people’s intentions to share with different recipients can be different, when measuring growers’ behavioral intentions to share relevant information, we provided them with a list of recipients who might be interested in staying informed about late blight outbreaks. This list was developed per conversations with experts in this area of research (including university professors and extension educators) and preliminary interviews with the growers. Specifically, the list included (a) “an anonymous service that does not identify my farm such as Blightline or the Late Blight Map, or other similar services,” (b) “adjacent or neighboring growers,” (c) “buyers inside of my county or region of my crops,” (d) “extension educators,” and (e) “commercial third parties such as sales representatives or crop consultants.”

Attitude and perceived norm

We measured attitude as a respondent’s agreement to three items: “If my farm were infected by late blight, sharing the news with others in my county or region would be good for me,” “. . . useful for me,” and “. . . a wise move for me” (α = .84). As perceived norm refers to an overall influence from perceived injunctive and descriptive norms (Fishbein & Ajzen, 2011), perceptions of both types of norm should be measured. For injunctive norm, which refers to perceptions concerning what should be done, we used two items: “Most growers in my county or region think that if late blight has been detected on anyone’s farm, the news should be shared” and “If my farm has late blight, I would be expected by other growers in my county or region to disclose the news;” r (431) = .50. For descriptive norm, which refers to perceptions of others’ behaviors, we used two items: “Growers in my county or region who are important to me would not hide information about late blight on their farms” and “Many growers in my county or region would share the news if late blight were detected on their farms;” r (437) = .56. Cronbach’s α for the four items was .78.

Individual-oriented concerns

We measured concern for economic cost along two dimensions, with one item referring to buyers (“I may lose business if buyers of my seed or crops know my farm has been infected with late blight”) and another referring to other farmers (“I may be disadvantaged to other farmers if buyers know my farm has been infected by late blight”; r (428) = .53). Concern for spending time/effort was measured with two items: “I do not always have time to report late blight” and “I am too busy managing late blight to report it”; r (431) = .43. Enjoyableness of altruism was measured with three items: “I like helping other people,” “It feels good to help others solve their problems,” and “I enjoy helping others in the local community” (α = .84, adapted from Wasko & Faraj, 2005).

Relation-oriented concerns

We measured belief in reciprocity with three dimensions from Perugini, Gallucci, Presaghi, and Ercolani (2003), and adapted three relevant items from previous work on organizational knowledge sharing (Kankanhalli et al., 2005). These items (α = .59) comprise reciprocity belief (“If I share the news with other growers about my late blight, they will share news with me”), positive reciprocity (“I expect other growers to share news with me about their late blight if I share with them”), and negative reciprocity (“I am afraid that if I do not share news with other growers about late blight, they will not share with me”). Given that trust is relational (Hardin, 2002), we measured the perceived trustworthiness of each information recipient, similar to our measurements of behavioral intention. To evaluate whether a recipient was considered benevolent and would take a source’s interests into account, the respondent was asked,

If you were to share the news that your farm were infected with late blight, please indicate how much you trust that each of the following would make good use of the information and not intentionally put you in a bad position.

Options ranged from 1 (not trust at all) to 7 (complete trust), along with a don’t know coded as missing.

Community-oriented concerns

Perception of shared responsibility was measured along two dimensions using three items adapted from Lawler et al. (2008). These items (α = .62) comprise a perception of outcome interdependence (“Late blight is a problem for the whole county or region, not just my farm”) and related normative concerns (“Farmers in my county or region share joint responsibility to deal with late blight” and “Late blight needs to be dealt by all [target crop] farmers in a county or region”). Perception of community cohesiveness was measured along two dimensions from Friedkin (2004), using three relevant items (α = .61) adapted from Bock et al.’s (2005) work on organizational knowledge sharing. These include one item about person-to-group ties (“I feel a strong sense of belonging to the farming community in my county or region”) and two items about person-to-person ties (“The farming community in my county or region is divided and not united” and “Farmers in my county or region treat each other as partners”).

Other related measurements

We asked whether a respondent communicated with other growers through each of the following channels: (a) “face to face interaction,” (b) “telephone,” (c) “text messages,” (d) “email,” (e) “Facebook,” (f) “Twitter,” (g) “citizens band radio,” and (h) “any other unmentioned channels.” We summed the binary answers as the number of communication channels, and used it as a marker for assessing common method variance (CMV; see Podsakoff, MacKenzie, & Podsakoff, 2011, for a review), discussed in the next section.

Analytical Procedures

We tested the hypotheses using structural equation modeling (SEM) because first, SEM can adjust for measurement errors, given that some of our measures, particularly the multidimensional ones, had relatively low reliabilities. Second, SEM can test simultaneously not only the significance level of each hypothesis but also overall model fit. Following conventions (Kline, 2016), we evaluated fits and comparisons of SEM based on four indices. First, the model’s χ² test should be non-significant (e.g., p > .05). Second, root mean square error of approximation (RMSEA) should be less than .05 and have a 90% confidence interval (CI) with the lower bound less than .05 and the upper bound less than .10. Third, comparative fit index (CFI) should be larger than .95. Finally, standardized root mean square residual (SRMR) should be less than .10. The specific models were run using Mplus 7. Below,¹ we describe (a) the model specification, identification, and statistical power; (b) treatments of missing values, non-normality, and collinearity; and (c) assessments of CMV and bias.

Model specification, identification, and statistical powers

Bentler and Chou (1987) suggested that the ratio between a sample size and the number of free parameters should be at least five. Given our sample size, we could not run an SEM with a full measurement model. An alternative is to turn each scale into a single-indicator latent factor, with residual variance specified to adjust for measurement error (Kline, 2016). To identify the model, factor loadings are often fixed at 1, and the percentage of a scale score’s residual variance is fixed at 1 minus reliability score, assuming independent measurement errors. We set reliability scores of unidimensional, multi-item scales (i.e., attitude and altruism) and the bi-item scale (i.e., concern for spending time/effort) using Cronbach’s alpha and Spearman–Brown’s prophecy score, respectively. As these types of reliability scores can overestimate measurement error for multidimensional constructs (Little et al., 1999), we only took 50%² of their calculations for perceived norm, reciprocity, shared responsibility, cohesion, and concern for economic cost. This practice also helps prevent biases induced by overcorrecting measurement errors, especially when the errors are not independent (DeShon, 1998), discussed later. For single-item scales (i.e., behavioral intention and trust), we set reliability scores to .80.

The single-indicator SEM reduced the number of free parameters to 114, including 23 path coefficients, 55 covariances, 18 intercepts, and 18 variances. This model was thereby theoretically identified. However, the ratio between sample size and parameters was still less than five. We hence conducted a Monte Carlo simulation to assess statistical power of the hypothesized model given our sample size (for details of the procedure, see Muthén & Muthén, 2002). To detect relatively small effects, we set the true strengths of all hypothesized relations at .25. We also set correlations among exogenous factors at .25, measurement errors at 20%, and R² of endogenous factors at 50%. The percentage and pattern of missing value were set per the sample. A simulation with 1,000 samples (N = 452) showed that the powers of testing parameters in the model ranged from .94 to 1. Meanwhile, the parameter bias ranged from −1.36% to 3.14%, the standard error bias from −6.36% to 5.91%, and the parameter coverage from 91.8% to 97%. According to Muthén and Muthén (2002), these results suggest sufficient power and acceptable ranges of bias, given our sample size.

Missing value, non-normality, and collinearity

Non-missing values of variables covered 91.81% to 99.78% cases. Non-missing values of correlations covered 88.90% to 100% cases. Missing values were handled using the full information maximum likelihood (ML) approach implemented by Mplus 7. A few scale scores (see Table 1) exhibited problematic non-normality (i.e., absolute skewness > 2 and kurtosis > 7; Curran, West, & Finch, 1996), suggesting possible violations of the multivariate normality assumption of a standard ML estimator. Therefore, we used the ML estimator with robust standard error (MLR) implemented by Mplus 7 because it does not have strict normality assumptions. Concordantly, χ² tests were corrected (Satorra & Bentler, 2001). Finally, collinearity was not a concern because no correlation among scale scores exceeded .85, the suggested threshold for consideration (Kline, 2016).

Table 1.

Summaries of Scale Scores and the Error Covariance Matrix of the Final Model.

	M	SD	SI	KI	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17
1. Intention (A)	4.11	1.59	0.67	0.07	—		.28	.20	.17
2. Intention (N)	5.03	1.01	−2.09	8.34	.14	—	.23	.21	.20
3. Intention (B)	4.67	1.29	−1.42	4.29	.29	.42	—	.18	.16
4. Intention (E)	4.76	1.26	−1.49	4.49	.17	.39	.22	—	.39
5. Intention (C)	4.51	1.38	−1.17	3.32	.15	.31	.17	.39	—
6. Trstwrth. (A)	4.50	2.03	−0.45	1.97	.44	−.03	.08	−.03	.01	—				.43
7. Trstwrth. (N)	5.57	1.55	−1.12	3.69	.11	.22	.20	.13	.10	.17	—
8. Trstwrth. (B)	4.85	1.81	−0.53	2.34	.16	.08	.32	−.08	−.02	.17	.41	—		.60
9. Trstwrth. (E)	5.54	1.59	−1.03	3.41	.13	.12	.10	.31	.23	.15	.43	.12	—
10. Trstwrth. (C)	4.75	1.76	−0.60	2.53	.15	.10	.03	.21	.32	.29	.25	.16	.48	—
11. Norm	4.72	0.84	−1.41	5.66	.19	.52	.35	.32	.24	.04	.29	.16	.18	.12	—	.10
12. Attitude	4.56	1.04	−1.13	3.39	.15	.40	.32	.29	.20	.00	.29	.20	.19	.12	.58	—
13. Economic cost	4.08	1.28	−0.32	2.44	.01	−.01	.16	−.04	−.14	−.11	−.15	.01	−.18	−.13	−.05	−.14	—
14. Time/effort	2.37	0.99	1.06	3.75	−.13	−.25	−.22	−.09	−.16	−.04	−.09	−.13	−.02	−.01	−.23	−.17	.02	—
15. Reciprocity	4.62	0.80	−1.04	4.57	.13	.37	.31	.21	.21	.01	.22	.21	.10	.07	.60	.51	.14	−.20	—
16. Altruism	5.22	0.58	−1.91	15.26	.08	.32	.25	.29	.14	−.06	.06	.08	.06	.09	.36	.35	.11	−.28	.43	—
17. Responsibility	4.99	0.74	−1.36	6.62	.18	.27	.17	.27	.21	.07	.12	.04	.16	.14	.44	.34	.13	−.29	.46	.47	—
18. Cohesiveness	4.51	0.87	−0.76	3.66	.08	.26	.17	.16	.13	−.02	.29	.29	.07	.13	.44	.39	−.03	−.27	.38	.39	.31

Note. N = 452. The lower and upper triangles are pairwise correlation matrix and the measurement residual covariance matrix, respectively. SI = skewness index; KI = kurtosis index; A = anonymous services; N = neighbor growers; B = buyers; E = extension educators; C = commercial third parties; Trstwrth. = trustworthiness.

CMV and bias

Our data might contain CMV (Podsakoff et al., 2011) because first, the data on the dependent and independent variables were collected together. Second, there were semantic similarities when measuring behavioral intention, attitude, and recipient trustworthiness. Finally, social desirability might have also influenced respondents’ answers to questions because they were asked for reasons of their actions or inactions (Fishbein & Ajzen, 2011). CMV might bias correlations among measurements, and therefore needs to be assessed (Podsakoff et al., 2011). As there exists no best way to assess CMV and the biases imposed by it (Podsakoff et al., 2011; Richardson, Simmering, & Sturman, 2009), two techniques, including unmeasured latent method construct (ULMC) and confirmatory factor analysis (CFA) marker, were used for this assessment.

We first conducted an ULMC analysis using CFA models of all scale items. This technique has been used widely, but is also criticized for capturing not only CMV but also all variance unspecified in a model (Podsakoff et al., 2011). Therefore, its result serves as the upper bound of CMV. We followed Richardson et al.’s (2009) procedure and fitted four models, including trait-only, method-only, trait/method, and trait/method-R models (see the authors’ article for detailed specifications). As shown in Table 2, Section A, the method-only model (M2) was significantly worse than the trait-only model (M1, Δχ²(151) = 10,182.48, p < .001), suggesting “observed variance in the independent and dependent constructs is not because of method alone” (Richardson et al., 2009, p. 780). The trait/method model (M3) was significantly improved over the trait-only model (M1, Δχ²(33) = 126.55, p < .001), suggesting the existence of congeneric CMV. However, the fit of the trait/method-R (M4) model was not significantly worse than the trait/method model (M3, Δχ²(153) = 45.92, p = 1), meaning that although CMV existed, it did not significantly bias factor correlations.

Table 2.

Summaries of Model Fits and Comparisons.

Models	CFI	SRMR	RMSEA		χ² test			χ²-difference test
Models	CFI	SRMR	Estimation	90% CI	χ²	df	p	Test against	Δχ²	Δdf	p
Section A: ULMC
M1 trait-only	.91	.05	.05	[.04, .05]	696.25	354	.000	—	—	—	—
M2 method-only	.00	.29	.19	[.19, .20]	9,070.56	505	.000	M1	10,182.48	151	.000
M3 trait/method	.94	.04	.04	[.03, .04]	535.44	321	.000	M1	126.55	33	.000
M4 trait/method-R	.98	.04	.02	[.01, .03]	557.49	474	.000	M3	45.92	153	1.000
Section B: CFA marker
M5 initial CFA	.91	.05	.05	[.04, .05]	720.81	369	.000	—	—	—	—
M6 baseline	.90	.05	.05	[.04, .05]	756.62	388	.000	—	—	—	—
M7 Method-C	.90	.05	.05	[.04, .05]	755.50	387	.000	M6	0.70	1	.403
M8 Method-U	.91	.05	.05	[.04, .05]	701.42	355	.000	M6	53.65	33	.013
M9 Method-R	.96	.05	.03	[.02, .03]	675.93	508	.000	M8	1.26	120	1.000
Section C: SEM modifications
M10 initial	.89	.05	.07	[.06, .08]	256.86	75	.000	—	—	—	—
M11 error-correlated	.95	.04	.06	[.05, .07]	136.54	55	.000	M10	109.06	20	.000
M12 correlated-attitude-norm	.96	.03	.05	[.04, .06]	118.54	54	.000	M11	23.26	1	.000
M13 partial-mediation	.99	.03	.03	[.02, .05]	69.62	46	.014	M12	52.41	8	.000
M14 index-refined	1.00	.02	.01	[.00, .03]	45.99	44	.390	M13	34.68	2	.000
M15 final	1.00	.02	.00	[.00, .02]	51.35	53	.539	M14	5.89	9	.751
Section D: SEM alternatives
M16 reversed-partial	.81	.07	.10	[.09, .11]	365.63	64	.000	—	—	—	—
M17 reversed-full	.63	.10	.10	[.10, .11]	704.12	120	.000	—	—	—	—

Note. χ²-difference tests were corrected (Satorra & Bentler, 2001). See Richardson, Simmering, and Sturman (2009) and Williams, Hartman, and Cavazotte (2010) for specifications of models in Section A and Section B, respectively. CFI = comparative fit index; SRMR = standardized root mean square residual; RMSEA = root mean square error of approximation; CI = confidence interval; ULMC = unmeasured latent method construct; CFA = confirmatory factor analysis; SEM = structural equation modeling.

To corroborate the conclusion above, we conducted a CFA marker analysis, which outperformed other techniques when an ideal marker was used (Richardson et al., 2009; Williams, Hartman, & Cavazotte, 2010). When an inappropriate marker is used, however, the analysis will be a weak test of CMV (Podsakoff et al., 2011), and set the lower bound for CMV assessment. Since a marker should lack in theoretical relations with other variables and tap sources of CMV (Williams et al., 2010), we chose the number of communication channels as the marker. We believe that this marker had no direct relationship with most variables in our hypotheses. Its correlations with other indicators were also small, ranging from −.09 to .15 (see Table 1). This marker might also indicate a respondent’s openness to communication technologies, which might be used to estimate the influence of social desirability when respondents answered our questions. Following Williams et al.’s (2010) procedure and terminology, we fitted five CFA models, including initial, baseline, Method-C, Method-U, and Method-R models (see the authors’ article for detailed specifications). As shown in Table 2, Section B, the significant improvement (Δχ²(33) = 53.65, p = .010) of the Method-U model (M8) and the non-significant improvement of the Method-C model (i.e., M7 with Δχ²(1) = 0.70, p = .400) over the baseline model (M6) suggest the existence of congeneric (rather than non-congeneric) CMV. However, as the fit of the Method-R model (M9) was not significantly worse than the Method-U model (M8, Δχ²(120) = 1.26, p = 1), CMV in the data did not bias correlations among factors. This is consistent with the conclusion drawn from the ULMC technique.

Results

Model Modifications

We estimated the initial model (M10) as hypothesized. This model did not fit the data well, with χ²(75) = 256.86 (p < .001), RMSEA = .07 with a 90% CI = [.06, .08], CFI = .89, and SRMR = .05. To improve model fit, we first correlated measurement errors of behavioral intentions and also those of recipient trustworthiness. Allowing interdependent errors is reasonable given the similarly phrased questions for behavioral intention and recipient trustworthiness, and the existence of congeneric CMV. Summarized in Table 2, Section C, the error-correlated model (M11) improved over the initial model (M10) with Δχ²(20) = 109.06 (p < .001), though its model χ² (136.54, df = 55, p < .001) and RMSEA (.06) were still unsatisfactory.

Next, we correlated the measurement errors of attitude and perceived norm. The two constructs correlate often, especially when predicting social behaviors (Park, 2000). Fishbein and Ajzen (2011) indicated that the construct validity of the two concepts does not preclude their correlation. It is unrealistic to assume that attitude and perceived norm associate only through our measured antecedents. The presence of CMV could also have induced a correlation between the two. The correlated-attitude-norm model (M12) significantly improved over the previous model (M11), with Δχ²(1) = 23.26 (p < .001). Except for the model’s χ²(54) = 118.54 (p < .001), other fit indices met conventional requirements (see Table 2).

We iteratively freed significant (p < .05) direct effects of concern for economic cost, concern for spending time/effort, belief in reciprocity, enjoyableness of altruism, perceived shared responsibility, and perceived community cohesiveness on the behavioral intentions. It might be too restrictive to assume full mediation of attitude and perceived norm because they were measured in a more general sense (i.e., “others in my county or region”) than behavioral intentions regarding specific recipients (Fishbein & Ajzen, 2011, p. 30). The partial-mediation model (M13) significantly improved over the previous model (M12), with Δχ²(8) = 52.41 (p < .001), though its χ² (69.62, df = 46, p = .014) was still significant.

We further consulted the modification indices, which suggested two additional effects. First, concern for economic cost might affect perceived norm. Although this concern is believed to be evaluative, it might suppress perceived norm because the latter is subjective and often correlates with attitudinal concerns, as mentioned previously. Second, the trustworthiness of local buyers might influence one’s intention to share with extension educators. This effect is possible because growers might treat some recipients as interchangeable in terms of sharing risk information. When one type of recipient was believed to be benevolent and trustworthy, growers might not need to share with another type of recipient. Freeing these effects, the index-refined model (M14) significantly improved over the previous model (M13), with Δχ²(2) = 34.68 (p < .001), and its χ² (45.99, df = 44, p = .390) became non-significant. To make this model more parsimonious, we iteratively removed non-significant (p > .1) covariances among measurement errors added before (see the upper triangle of Table 1 for the remaining covariances). This reduction did not worsen the previous model’s fit (M14) as the Δχ²(9) = 5.89 (p = .751), and the final model (M15, illustrated in Figure 1) was also adequately fitted, with χ²(54) = 51.35 (p = .539), RMSEA = .00 with a 90% CI = [.00, .04], CFI = 1.00, and SRMR = .02.

Figure 1.

The final model (M15), including hypothesized (the upper panel) and additional relations (the lower panel).

Alternative models

When conducting SEM, multiple models might fit the data equally well. Therefore, alternative models need to be assessed to disqualify other possibilities (Kline, 2016). As we measured behavioral intentions before attitude and perceived norm, followed by other measurements, it is possible that this ordering might induce a tendency to answer consistently (Budd, 1987), and thus reverse the hypothesized causalities. Although correlational data might not determine these directions, we tested whether alternative models with reversed causal directions fit the data as good as or better than the hypothesized model (i.e., the initial model). Two alternatives were tested. The first (M16) reversed the path from attitudes, perceived norms, and recipient trustworthiness to behavioral intentions. The second (M17) further reversed the remaining hypothesized paths. Summarized in Table 2, Section D, both alternatives fit the data worse than the initial model, suggesting less concern for reversed causality. These results provide empirical support to the robustness of our hypothesized model.

Hypothesis Testing

Figure 1 summarizes the final model, along with results of specific hypothesis testing (see the upper panel), detailed below.

Attitude and perceived norm

Testing of H1 and H2 received partial supports. A favorable attitude was positively associated with intentions to share information about late blight infections with neighboring growers (B = 0.17, SE = 0.07, p = .010) and local buyers (B = 0.29, SE = 0.10, p = .003). However, the attitude was not associated with intentions to share with anonymous services (p = .150), extension educators (p = .093), and commercial third-parties (p = .577). Except for commercial third-parties (p = .845), a perceived norm of sharing was positively associated with intentions to share with other types of recipients, with B ranging from 0.26 to 0.52 (p < .05).

Individual-oriented concerns

Concerns for economic cost was negatively associated with a favorable attitude toward sharing information about late blight infection (B = −0.22, SE = 0.04, p < .001), supporting H3. However, the association between concern for spending time/effort and the attitude was non-significant (p = .386). A belief in the enjoyableness of altruism was also not associated with the attitude (p = .124). Therefore, H4 and H5 were rejected.

Relation-oriented concerns

A belief in the reciprocity of risk information sharing was positively associated with a favorable attitude toward sharing (B = 0.75, SE = 0.09, p < .001), thus supporting H6. Predicted by H7, reciprocity belief was also positively associated with a perceived norm of sharing (B = 0.65, SE = 0.08, p < .001). Predicted by H8, the evaluations of recipient trustworthiness were positively associated with intentions to share with anonymous services (B = 0.43, SE = 0.05, p < .001), local buyers (B = 0.60 SE = 0.05, p < .001), extension educators (B = 0.26, SE = 0.05, p < .001), and commercial third-parties (B = 0.27, SE = 0.05, p < .001). However, the association between trustworthiness of neighboring growers and the intention to share with them was non-significant (p = .239).

Community-oriented concerns

Perceived shared responsibility was positively associated with a perceived norm of sharing (B = 0.15, SE = 0.07, p = .046), thus supporting H9. Predicted by H10 and H11, perceived community cohesiveness was positively associated with a favorable attitude (B = 0.21, SE = 0.08, p = .010) and a perceived norm of sharing (B = 0.19, SE = 0.06, p = .001).

Additional Findings

The final model revealed several unexpected effects regarding local buyers (see the lower panel of Figure 1). First, concern for economic cost was positively associated with the intention to share with local buyers (B = 0.19, SE = 0.05, p < .001) after conditioning on attitude and perceived norm. Second, perceived community cohesiveness suppressed the intention to share with local buyers (B = −0.30, SE = 0.12, p = .011). Finally, the evaluation of local buyers’ trustworthiness suppressed the intention to share with extension educators (B = −0.14, SE = 0.04, p < .001). We elaborate on these findings when discussing practical implications.

Some less unexpected effects were also revealed, including (a) negative direct effects of concern for spending time/effort on intentions to share with neighboring growers (B = −0.20, SE = 0.09, p = .023), local buyers (B = −0.25, SE = 0.11, p = .019), and commercial third-parties (B = −0.21, SE = 0.11, p = .048); (b) a positive direct effect of enjoyableness of altruism on the intention to share with extension educators (B = 0.43, SE = 0.14, p = .002); (c) a negative effect of concern for economic cost (B = −0.20, SE = 0.07, p = .006) and a positive effect of belief in reciprocity (B = 0.44, SE = 0.21, p = .038) on the intention to share with commercial third-parties; and (d) a negative effect of concern for economic cost on the perceived norm of sharing (B = −0.11, SE = 0.03, p < .001). They suggest that the factors entailed in several hypotheses, including the rejected H4 and H5, were still motivating despite lack of relevance to attitude and perceived norm.

Discussion

Although previous work has investigated seeking, processing, and using of risk information (e.g., Braun & Niederdeppe, 2012; Griffin et al., 1999; Kahlor, 2010; Rimal, 2001), fewer studies have examined risk information sharing (Yang et al., 2014). This is particularly true for communal risks that interdependently affect community members and require collective efforts to contain. Our study examined risk information sharing for late blight, which can cause substantial economic risks to a community of growers. Building on the distinction between attitudinal (or evaluative) and normative beliefs (Fishbein & Ajzen, 2011), we tested motivational factors, ranging from individual-oriented concerns and relation-oriented concerns to community-oriented concerns.

Theoretical Implications

This work contributes to the current knowledge of risk communication by identifying a diverse set of factors that motivate information sharing among community members to deal with a communal risk. Some work has been done on this topic—for example, Yang et al.’s (2014) work on information sharing about climate change. However, we know little about how people react when dealing with a risk that has a stronger sense of urgency, particularly when members’ economic outcomes share stronger interdependence and when sharing information to reduce a communal risk might result in heightened immediate risk to the discloser. Previous work on motivations for information sharing has suggested that collective decision making and performance is affected by not only epistemic motivation, that is “the willingness to expend effort to achieve a thorough, rich, and accurate understanding of the world” (de Dreu et al., 2008, p. 23); but also social motivation, which concerns the dialect between individual and collective outcomes (de Dreu et al., 2000). Although epistemic motivation has been found to motivate information sharing about climate change (Yang et al., 2014), we argue that social motivations, especially those regarding specific levels of social coordination, are more relevant to communal risk information sharing.

In the case of sharing information about late blight, we found that growers were motivated by concerns at all three levels, including individual, relation, and community. At the individual level, concerns for cost were salient such that risk information sharing could be forestalled if believed to result in financial losses (H3 supported). Although concern for spending time/effort was not associated with a favorable attitude toward sharing, as hypothesized (H4 rejected), it still suppressed the intention to share. At the relational level, respondents held a favorable attitude and perceived a stronger norm regarding risk information sharing when they believed in reciprocity (H6 and H7 supported). Among other concerns, belief in reciprocity had the strongest effect on attitude and perceived norm (see Figure 1). Another relation-oriented concern—a recipient’s trustworthiness—is also important for risk information sharing (H8 supported). Particularly, recipient trustworthiness was most critical for respondents’ willingness to share risk information with anonymous services, but least for sharing with neighbors (see Figure 1). This pattern is consistent with previous work on social exchange, where trust plays an important role for cooperation without salient institutional assurances (e.g., contracts, organizations, or a sense of community; Kollock, 1994; Molm et al., 2000). At the community level, respondents perceived a stronger norm of sharing risk information when having a sense of shared responsibility (H9 supported), suggesting the importance of salient outcome interdependence for cooperation (de Dreu, 2007). Perceived community cohesiveness was also associated with a more favorable attitude and a stronger perceived norm regarding risk information sharing (H10 and H11 supported). These findings are consistent with Coleman’s (1988) theory of social capital and Kanter’s (1968) work on community commitment.

Our findings suggest that communal risk information sharing can be different from information sharing in other collective contexts. First, normative concerns for communal risk information sharing appear to go beyond attitudinal concerns. As shown in Figure 1, sharing with anonymous services and commercial third parties was motivated by a perceived norm, but not a favorable attitude. The perceived norm also had a stronger effect than the attitude when sharing with neighboring growers. Yang et al. (2014) also found that norms had a stronger effect than other motivators (e.g., epistemic motivation) of sharing climate-change information. Such a strong effect of norm was not often observed in the context of organizational knowledge sharing (see Wang & Noe, 2010). This finding suggests that communal risk information sharing heavily relies on norms and the underlying structural forces (i.e., reciprocal social exchange and social cohesion). A potential reason for the weaker effect of attitude is that the joint benefits of reducing a communal risk (i.e., “benefiting us”) might not be direct and salient enough to be internalized by individuals (i.e., “benefiting me”).

Second, social trust appears to be more important to communal risk information sharing than organizational knowledge sharing. The latter is often backed by dedicated structures (e.g., a formal organization) that regulate cooperation. In an informal community, trust is invaluable to sustain risk information sharing, especially when sharing can induce risks to the discloser, such as with potential financial losses. A discloser might not only consider behavioral outcomes of sharing risk information in a general sense. He or she might also take into account social information about specific recipients’ trustworthiness as part of the reasoning, given a lack of institutional assurance. The effect of such social information appears to be independent of those induced by a general attitude and perceived norm regarding the sharing behavior.

Finally, altruistic motivation did not play a major role in respondents’ views about sharing information about late blight (H5 rejected). This finding is different from previous work on organizational knowledge sharing (Constant et al., 1994; Kankanhalli et al., 2005; Wasko & Faraj, 2000). Wasko and Faraj (2005) suggested that the enjoyableness of helping might become less salient for two reasons: the non-anonymous nature of a collective, and the salience of payoffs. We believe these reasons are also relevant to communal risk information sharing. When community members (e.g., growers) know each other and relevant third-parties (e.g., extension educators), relational and collective-specific mechanisms (e.g., reciprocity and commitment) might override the individual psychological rewards of altruistic behaviors. The potential losses by a communal risk might also direct people’s attentions to immediate payoffs (Kahneman & Tversky, 1979).

Practical Implications

Our findings have practical implications for risk communication to reduce communal risk when economic concerns might be present. First, economic and cooperative concerns should be brought to the table, instead of appealing for non-economic or pure altruistic motivations when encouraging people to share risk information. For example, communication campaigns can be designed to elucidate the interdependence of communal risk and the cooperative nature of risk information sharing. Risk information sharing could be framed as economically beneficial, both in the short and long terms, and at both relational and community levels. Otherwise, the significant influence of economic cost might not only hamper growers’ favorable attitude toward sharing but also undermine their perceptions of norms, as suggested by our additional findings. However, concern for economic cost of sharing risk information should not be downplayed or concealed in communication campaigns. Instead, these concerns could promote sharing with certain parties (i.e., local buyers) because economic costs might motivate disclosure to sustain cooperation in the long term.

Second, risk communication practices can emphasize relationship- and community-building. In doing so, it might be important in the context of late blight to include not only local community members (e.g., neighbors and extension educators) but also any party that might be affected financially (e.g., retailers, produce consumers, and seed buyers). Otherwise, a cohesive community could suppress sharing with the latter, as suggested by the additional finding that perceived community cohesiveness had a negative direct effect on intentions to share with local buyers. To overcome this tendency, interests of parties that are perceived as less aligned with a community might be represented by anonymous services that collect and share risk information for all relevant parties. The goal is not only to increase the appropriateness of risk information sharing with these community members, but also provide institutional assurance for members to trust each other. Each of these strategies also builds on the significant role that perceived norms played in motivations of risk information sharing. Efforts to increase awareness that risk sharing information is a valued norm, and a shared responsibility might encourage sharing.

Limitations and Future Directions

The present study has limitations. First, although most measures were adapted from published studies in other fields, some scales still had relatively lower measurement reliability. While the SEM analysis technique we used can correct for the influence of measurement errors during hypothesis testing, development of more reliable measurement scales is desired in future research. Second, we examined only a specific economic communal risk in agricultural communities. Although we believe information sharing is particularly important for economic communal risk, it remains unclear whether these results would be the same for other types of communal risks. Finally, we focused primarily on the rational aspect of decision making. However, decision making can be affective and less calculative (Kahneman & Tversky, 1979), especially when a communal risk endangers expressive values more than material payoffs (Yang et al., 2014).

Regarding these limitations, we suggest future research should pursue at least two directions. First, the cognitive and affective mechanisms of decision making for sharing risk information can be further elaborated. For example, individuals with different cognitive foci on gains versus losses might be motivated differently when sharing information of one’s own risks. Different affective experiences might also alter the extent of motivation for sharing (see Yang et al., 2014). Second, moving beyond individual psychologies, social influences of community structure, cohesion, and trust can also affect communal risk information sharing, and should be investigated in greater depth.

Conclusion

While communal risks can prevail on the interdependence among us in various aspects of a society, turning interdependence into explicit cooperation can be invaluable for risk reduction. Risk information sharing is one step of such cooperation, which, however, can be forestalled by potential conflicts between individual and collective interests. The present work is one of a few recent efforts to offer better understanding of motivations behind risk information sharing. We believe risk information sharing should be further studied because it is at the core of risk communication to enhance public participation and cooperation—a call that echoes in our increasingly global society.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by a grant awarded to the third author from the National Institute of Food and Agriculture (2011-68004-30154).

Notes

Author Biographies

Wang Liao (MA, Cornell University) is a doctoral candidate in the Department of Communication at Cornell University. He studies structure and micro-mechanisms of group dynamics using quantitative methods combined with network and computational approaches.

Y. Connie Yuan (PhD, University of Southern California) is an associate professor in the Department of Communication at Cornell University. Her research interests include knowledge management, intercultural communication, social network, and computer-supported collaboration in distributed teams.

Katherine A. McComas (PhD, Cornell University) is a professor in the Department of Communication at Cornell University. Her research examines how communication influences risk perceptions, trust in risk managers, satisfaction with decisions, and willingness to engage in future community activities.

References

American Association for Public Opinion Research. (2008). Standard definitions: Final dispositions of case codes and outcome rates for surveys (5th ed.). Lenexa, KS: American Association for Public Opinion Research.

Andreoni

(1990). Impure altruism and donations to public-goods: A theory of warm-glow giving. Economic Journal, 100, 464-477. doi:10.2307/2234133

Axelrod

R. M.

(1984). The evolution of cooperation. New York, NY: Basic Books.

Bentler

P. M.

Chou

C.-P.

(1987). Practical issues in structural modeling. Sociological Methods & Research, 16, 78-117. doi:10.1177/0049124187016001004

Bock

G. W.

Zmud

R. W.

Kim

Y.-G.

Lee

J.-N.

(2005). Behavioral intention formation in knowledge sharing: Examining the roles of extrinsic motivators, social-psychological forces, and organizational climate. MIS Quarterly, 29, 87-111. doi:10.2307/25148669

Braun

Niederdeppe

(2012). Disruption and identity maintenance in risk information seeking and processing. Communication Theory, 22, 138-162. doi:10.1111/j.1468-2885.2012.01402.x

Budd

R. J.

(1987). Response bias and the theory of reasoned action. Social Cognition, 5, 95-107. doi:10.1521/soco.1987.5.2.95

Cabrera

E. F.

(2002). Knowledge-sharing dilemmas. Organization Studies, 23, 687-710. doi:10.1177/0170840602235001

Coleman

J. S.

(1988). Social capital in the creation of human capital. American Journal of Sociology, 94, S95-S120. doi:10.1086/228943

10.

Constant

Kiesler

Sproull

(1994). What’s mine is ours, or is it? A study of attitudes about information sharing. Information Systems Research, 5, 400-421. doi:10.1287/isre.5.4.400

11.

Curran

P. J.

West

S. G.

Finch

J. F.

(1996). The robustness of test statistics to nonnormality and specification error in confirmatory factor analysis. Psychological Methods, 1, 16-29. doi:10.1037/1082-989x.1.1.16

12.

de Dreu

C. K. W

. (2007). Cooperative outcome interdependence, task reflexivity, and team effectiveness: A motivated information processing perspective. Journal of Applied Psychology, 92, 628-638. doi:10.1037/0021-9010.92.3.628

13.

de Dreu

C. K. W.

Nijstad

B. A.

van Knippenberg

. (2008). Motivated information processing in group judgment and decision making. Personality and Social Psychology Review, 12, 22-49. doi:10.1177/1088868307304092

14.

de Dreu

C. K. W.

Weingart

L. R.

Kwon

. (2000). Influence of social motives on integrative negotiation: A meta-analytic review and test of two theories. Journal of Personality and Social Psychology, 78, 889-905. doi:10.1037/0022-3514.78.5.889

15.

DeShon

R. P.

(1998). A cautionary note on measurement error corrections in structural equation models. Psychological Methods, 3, 412-423. doi:10.1037/1082-989x.3.4.412

16.

Donnelly

J. S.

(2001). The great Irish potato famine. Stroud, UK: Sutton.

17.

Earle

T. C.

(2010). Trust in risk management: A model-based review of empirical research. Risk Analysis, 30, 541-574. doi:10.1111/j.1539-6924.2010.01398.x

18.

Emerson

R. M.

(1972). Exchange theory, part II: Exchange relations and network structures. In Berger

Zelditch

Anderson

(Eds.), Sociological theories in progress (pp. 58-90). Boston, MA: Houghton Mifflin.

19.

Fehr

Gintis

(2007). Human motivation and social cooperation: Experimental and analytical foundations. Annual Review of Sociology, 33, 43-64. doi:10.1146/annurev.soc.33.040406.131812

20.

Fishbein

Ajzen

(2011). Predicting and changing behavior: The reasoned action approach. New York, NY: Psychology Press.

21.

Friedkin

N. E.

(2004). Social cohesion. Annual Review of Sociology, 30, 409-425. doi:10.1146/annurev.soc.30.012703.110625

22.

Fry

W. E.

McGrath

M. T.

Seaman

Zitter

T. A.

McLeod

Danies

. . . Smart

C. D.

(2012). The 2009 late blight pandemic in the eastern United States: Causes and results. Plant Disease, 97, 296-306. doi:10.1094/pdis-08-12-0791-fe

23.

Fulk

Flanagin

A. J.

Kalman

M. E.

Monge

P. R.

Ryan

(1996). Connective and communal public goods in interactive communication systems. Communication Theory, 6, 60-87. doi:10.1111/j.1468-2885.1996.tb00120.x

24.

Gouldner

A. W.

(1960). The norm of reciprocity: A preliminary statement. American Sociological Review, 25, 161-178. doi:10.2307/2092623

25.

Griffin

R. J.

Dunwoody

Neuwirth

(1999). Proposed model of the relationship of risk information seeking and processing to the development of preventive behaviors. Environmental Research, 80, S230-S245. doi:10.1006/enrs.1998.3940

26.

Hardin

(1982). Collective action. Baltimore, MD: Johns Hopkins University Press.

27.

Hardin

(2002). Trust and trustworthiness. New York, NY: Russell Sage.

28.

Haverkort

A. J.

Struik

P. C.

Visser

R. G. F.

Jacobsen

(2009). Applied biotechnology to combat late blight in potato caused by phytophthora infestans. Potato Research, 52, 249-264. doi:10.1007/s11540-009-9136-3

29.

Kahlor

(2010). PRISM: A planned risk information seeking model. Health Communication, 25, 345-356. doi:10.1080/10410231003775172

30.

Kahneman

Tversky

(1979). Prospect theory: An analysis of decision under risk. Econometrica, 47, 263-291. doi:10.2307/1914185

31.

Kankanhalli

Tan

B. C. Y.

Wei

K.-K.

(2005). Contributing knowledge to electronic knowledge repositories: An empirical investigation. MIS Quarterly, 29, 113-143. doi:10.2307/25148670

32.

Kanter

R. M.

(1968). Commitment and social organization: A study of commitment mechanisms in utopian communities. American Sociological Review, 33, 499-517. doi:10.2307/2092438

33.

Kline

R. B.

(2016). Principles and practice of structural equation modeling (4th ed.). New York, NY: Guilford.

34.

Kollock

(1994). The emergence of exchange structures: An experimental study of uncertainty, commitment, and trust. American Journal of Sociology, 100, 313-345. doi:10.1086/230539

35.

Kollock

(1999). The economies of online cooperation: Gifts and public goods in cyberspace. In Smith

Kollock

(Eds.), Communities in cyberspace (pp. 219-262). London, England: Routledge.

36.

Lawler

E. J.

Thye

S. R.

Yoon

(2008). Social exchange and micro social order. American Sociological Review, 73, 519-542. doi:10.1177/000312240807300401

37.

Little

T. D.

Lindenberger

Nesselroade

J. R.

(1999). On selecting indicators for multivariate measurement and modeling with latent variables: When “good” indicators are bad and “bad” indicators are good. Psychological Methods, 4, 192-211. doi:10.1037/1082-989x.4.2.192

38.

Molm

L. D.

Takahashi

Peterson

(2000). Risk and trust in social exchange: An experimental test of a classical proposition. American Journal of Sociology, 105, 1396-1427. doi:10.1086/210434

39.

Montano

D. E.

Kasprzyk

(2008). Theory of reasoned action, theory of planned behavior, and the integrated behavioral model. In Glanz

Rimer

B. K.

Viswanath

(Eds.), Health behavior and health education: Theory, research, and practice (4th ed., pp. 67-95). San Francisco, CA: Jossey-Bass.

40.

Muthén

L. K.

Muthén

B. O.

(2002). How to use a Monte Carlo study to decide on sample size and determine power. Structural Equation Modeling: A Multidisciplinary Journal, 9, 599-620. doi:10.1207/s15328007sem0904_8

41.

Park

H. S.

(2000). Relationships among attitudes and subjective norms: Testing the theory of reasoned action across cultures. Communication Studies, 51, 162-175. doi:10.1080/10510970009388516

42.

Parsons

J. T.

VanOra

Missildine

Purcell

D. W.

Gómez

C. A.

(2004). Positive and negative consequences of HIV disclosure among seropositive injection drug users. AIDS Education and Prevention, 16, 459-475. doi:10.1521/aeap.16.5.459.48741

43.

Perugini

Gallucci

Presaghi

Ercolani

A. P.

(2003). The personal norm of reciprocity. European Journal of Personality, 17, 251-283. doi:10.1002/per.474

44.

Piliavin

J. A.

Charng

H.-W.

(1990). Altruism: A review of recent theory and research. Annual Review of Sociology, 16, 27-65. doi:10.1146/annurev.so.16.080190.000331

45.

Podsakoff

P. M.

MacKenzie

S. B.

Podsakoff

N. P.

(2011). Sources of method bias in social science research and recommendations on how to control it. Annual Review of Psychology, 63, 539-569. doi:10.1146/annurev-psych-120710-100452

46.

Richardson

H. A.

Simmering

M. J.

Sturman

M. C.

(2009). A tale of three perspectives examining post hoc statistical techniques for detection and correction of common method variance. Organizational Research Methods, 12, 762-800. doi:10.1177/1094428109332834

47.

Rimal

R. N.

(2001). Perceived risk and self-efficacy as motivators: Understanding individuals’ long-term use of health information. Journal of Communication, 51, 633-654. doi: 10.1093/joc/51.4.633

48.

Satorra

Bentler

(2001). A scaled difference chi-square test statistic for moment structure analysis. Psychometrika, 66, 507-514. doi:10.1007/bf02296192

49.

Sheppard

B. H.

Hartwick

Warshaw

P. R.

(1988). The theory of reasoned action: A meta-analysis of past research with recommendations for modifications and future research. Journal of Consumer Research, 15, 325-343. doi:10.1086/209170

50.

Smith

K. D.

Keating

J. P.

Stotland

(1989). Altruism reconsidered: The effect of denying feedback on a victim’s status to empathic witnesses. Journal of Personality and Social Psychology, 57, 641-650. doi: 10.1037//0022-3514.57.4.641

51.

Wang

Noe

R. A.

(2010). Knowledge sharing: A review and directions for future research. Human Resource Management Review, 20, 115-131. doi:10.1016/j.hrmr.2009.10.001

52.

Wasko

M. M.

Faraj

(2000). “It is what one does”: Why people participate and help others in electronic communities of practice. The Journal of Strategic Information Systems, 9, 155-173. doi:10.1016/s0963-8687(00)00045-7

53.

Wasko

M. M.

Faraj

(2005). Why should I share? Examining social capital and knowledge contribution in electronic networks of practice. MIS Quarterly, 29, 35-57. doi:10.2307/25148667

54.

Williams

L. J.

Hartman

Cavazotte

(2010). Method variance and marker variables: A review and comprehensive CFA marker technique. Organizational Research Methods, 13, 477-514. doi:10.1177/1094428110366036

55.

Yamagishi

Cook

K. S.

Watabe

(1998). Uncertainty, trust, and commitment formation in the United States and Japan. American Journal of Sociology, 104, 165-194. doi:10.1086/210005

56.

Yang

Z. J.

Kahlor

L. A.

Griffin

D. J.

(2014). I share, therefore I am: A U.S.–China comparison of college students’ motivations to share information about climate change. Human Communication Research, 40, 112-135. doi: 10.1111.hcre.12018

57.

Yuan

Y. C.

Steinhardt

J. S.

McComas

K. A.

Gay

Smart

(2013, June). The usage of information and communication technologies to support seeking and sharing information about plant disease. Paper presented at the 63rd annual conference of International Association of Communication. London, England.