The relationship between transparent and ... - BigIdeasProject

The relationship between transparent and participative government: A study of local governments in the United States Abbreviated Running Head: Transparency and Participation of US Local Governments Eric W. Welch Associate Professor Department of Public Administration 412 South Peoria Street University of Illinois at Chicago Chicago, Illinois 60304 [email protected] Abstract The relationship between transparency and participation of government is not well articulated in the literature. Transparency provides stakeholders with knowledge about the processes, structures and products of government. Participation refers to the quantity, quality and diversity of input of stakeholders into government decision making. Greater transparency and participation are often considered to operate side-by-side. However, in the Internet age the change in the magnitude of information disclosure may outweigh the change in the level of participative government. This paper uses data from a 2010 national survey of five US local government agencies to test hypotheses about the relationship between transparency and participation and the factors that affect them. Findings show that participation is positively associated with transparency, but transparency does not lead to participation. Additionally, organizations that are under stronger influence from external stakeholders report higher levels of participation but in some cases higher levels of external influence dampen transparency. Points for Practitioners For practitioners, the paper provides a framework for understanding the relationship between transparency and participation as two dimensions of open government. While in some departments these two dimensions are present at similar levels, in others they are not. For example, among five departments, police report the lowest overall participation and transparency levels, while Parks & Recreation and the Mayor’s Office report the highest transparency and participation levels, respectively. Additionally, Police departments report significantly higher transparency than participation, while the reverse is true for community development departments. The paper provides public administrators with greater insight into causes of and relationships between transparency and participation. Keywords Local Government, Openness, Participation, Transparency

1

The relationship between transparent and participative government: A study of local government administrations in the United States

Introduction It is often expected that transparency of and participation in government are linked. Typically, it is expected that these two constructs are positively correlated such that governments high in transparency also exhibit high participation, as it is thought that transparency is necessary for participation. Yet little research has actually investigated whether this true, and in fact, at the outset of this paper the opposite should be considered. It is entirely possible that greater levels of transparency are associated with lower levels of participation, or that the relationship between transparency and participation depends upon different organizational and environmental factors. Moreover, transparency and participation are not monolithic – there are multiple definitions and perspectives about their boundaries, or rather their distinctness. In the transparency literature, several authors recognize that transparency concerns the extent to which an organization provides information about its own internal processes, decision and performance (Welch & Wong 2001; Gerring and Thacker 2004; Curtin and Meijer 2006). Heald identifies three different dimensions of transparency: decision making, policy content and policy outcomes (2006). The participation literature considers the ways in which stakeholders ideas, concerns, and interests are integrated into organization decisions and actions (. Some research examines how participation processes are affected by such factors problem salience and power distribution among actors (Downes & McMillan, 2000; McMillan, 2002; Heeter, 1989). Other research

2

examines how stakeholder input is integrated into decision outcomes (Masters 2002; Macintosh 2004; Alter et al. 2008). This paper distinguishes between the two constructs based on the level of interaction the decision maker has or intends to have with a stakeholder. Transparency is defined as the active disclosure of information by an organization that enables external actors to monitor and assess its internal workings, decisions and performance (Grimmelikhuijsen and Welch, 2010). This is generally a one-way communication process in which the organization provides information to other stakeholders. Participation is defined as the involvement of one or more stakeholders in decision-making or policy in such a way that the stakeholder input is considered during the decision making process and influences the decision outcome (Bickerstaff, 2001; Rowe and Frewer, 2000; Roberts 1997; 2004). This is generally a two-way communication process in which stakeholder is considered a participant in the decision making process (Fulla and Welch 2005).i Transparency and participation have strong normative foundations: in general people often believe that more is better. More information is important for citizens and other stakeholders to understand how their taxes are spent and help them make decisions about who to support in the next elections (Etzioni, 2010). In public administration, transparency is often linked to more trustworthy, accountable and open government (Hood 2006; Meijer 2009). Similarly, government decisions are expected to reflect the interests and needs of stakeholders and participation is an important component of that process (O’Connor et al 2000). In newer governance models, citizens and other stakeholders are participants in the governing process (Henderson 2007).

3

Nevertheless, public organizations are likely to be strategic about transparency and participation. In a turbulent environment in which stakeholders are always presented with the results of media inquiries, budget cuts, and whistleblowers, public organizations manage their exposure. Hence, this paper seeks to understand whether organizations vary in terms of transparency and participation, how the two dimensions are related, and what factors predict them. It asks three questions: How does transparency affect participation and vice versa? How is the use of technology associated with transparency and participation? How does external stakeholder influence affect the levels of transparency and participation in public organizations?

This paper takes advantage of a recent study of local governments in the United States in which survey data was collected from five different government departments: the mayor’s office, community development, parks and recreation, finance, and police. The survey sought to understand both basic levels of transparency and participation, but also how technology is used by the organization for both information dissemination and participation with stakeholders. This paper first develops a set of hypotheses and then tests the hypotheses using two-stage least squares regression. Findings show that participation in local governments is positively associated with transparency, but that the reverse does not hold; transparency does not always lead to participation. Additionally, organizations that are under stronger influence by external stakeholders report higher levels of participation, but in some cases higher levels of external influence dampen transparency. The use of information and communication technologies in local government agencies is positively associated with transparency but not associated with participation. Conclusions reflect on these findings.

4

Separating Transparency and Participation It is possible to consider transparency and participation as separate constructs, located on two different axes that range from low to high. The resulting diagram, presented in Figure 1, shows the two dimensions give rise to four quadrants. Quadrants two and four are the easiest to explain and conform to general expectations regarding the relationship between transparency and participation. In quadrant two both constructs are on the higher end, indicating that organizations found in this quadrant are both highly transparent and highly participative. To some extent, an example of such a situation may be found at multilateral international negotiations where decisions are based on consensus, participation in deliberations is open to all member nations, and desired insertions and deletions in negotiated text are clearly visible. Organizations in quadrant four are less transparent and less participative. There are probably many examples of these agencies, but it is safe to identify national security agencies as most likely to fit into this quadrant. For such organizations, transparency and participation represent activities that may jeopardize the mission of the agency. Recent work has shown that despite the rhetoric of openness, these agencies are increasingly secretive and offer very limited access to highly scrutinized outsiders (Lewis 2010). In quadrants one and three, the situation is mixed. In quadrant one, where there is higher participation and lower transparency, organizations are more likely to interact with stakeholders and provide opportunities for input into their decision processes; however they disseminate less information to the stakeholders. Such a situation may occur in highly turbulent or politicized environments, where there is a perceived need to maintain control over the interpretation of the disclosed information. Transparency, which is the one-way disclosure of information, may be perceived to be more risky than a more participative approach where information can be 5

provided in a more controlled, contextualized, give and take environment. Additionally, greater interaction with stakeholders is also a potential source of power for more political (or politically vulnerable) agencies, whereas the distribution of information may not result in relationship building. It is likely that the mayor’s office or the city’s executive council office would be more politicized than, for example, parks and recreation agencies. As such, we would expect the mayor’s office to respond to external stakeholders through participation more than transparency. In quadrant three there is lower participation by external stakeholders in internal decision making and higher transparency. In such situations, accountability may be less related to political responsiveness and more related to professionalism (Romzek and Dubnick, 1987; Bovens, 2005). Police agencies may be more likely to provide information about recent crimes and the extent to which crimes are solved, than they are to allow external stakeholders to provide input into decision making; participation may reduce control while selective information disclosure helps maintain control and demonstrate professionalism. Figure 1: Four potential relationships between participation and transparency High Participation

I

II

Low Transparency IV

High Transparency III

Low Participation

This framework helps visualize that the relationship between transparency and participation may be less straight forward than is commonly understood. Some agencies may find participation a

6

less risky choice, while others feel that dissemination of information is more appropriate. Two caveats. First, this discussion recognizes only the relative variation in transparency and participation, not their overall levels, which could be continually increasing or decreasing. Second, this framework is not normative. Rather, it considers that agencies may be using transparency and participation strategically, as two complementary responses to external stakeholders.

A Strategic Response The political context of public organizations requires managers to simultaneously address external political demands and maintain sufficient control to carry out the tasks required of the agency (Bozeman and Straussman 1990). As a result, managers must “negotiate, bargain, and compromise with the external forces in the environment in order to set and carry out an organization's goals (Knott 2003).” Additionally, managers must be strategic. They make decisions and plan in ways that optimize the ability of employees to accomplish work and the responsiveness to political exigencies (Bryson 1995). Given that managers are essentially information brokers between the internal organization and external environment, they develop strategic means of providing information in ways that simultaneously satisfy external groups and maintain organizational control. Because transparency and participation are two potentially different means of communicating and conveying information, it is likely that managers will use different mixes of the two depending upon the political context of the organization.

7

Principal-agent theory provides a mechanism for understanding how the political context of public organizations will affect the information provision behavior of managers. A principal sets a contract with an agent to conduct an activity that benefits the principal, but is costly for the agent to carry out (Laffont & Mordimort, 2002; Mitnick, 1975; Moe, 2006; Pratt & Zeckhauser, 1991). Because explicit contracts are rare, observation and monitoring of agents is difficult, and the goals of the principal and agent differ. For example, when the agent’s activity is difficult to observe and monitor, information asymmetry – an imbalance in the amount or quality of information held by the two parties – may result (Laffont & Mordimort, 2002; Mitnick, 1975; Moe, 2006). When an agent’s behavior is difficult or costly to monitor, principals cannot determine if the agent has conducted the contracted activity. Agents resist the provision of information, preferring to minimize uncertain response by external stakeholders. Thus, even though an agent may commit to a contract that serves the goals of the principal (ie. open democratic processes), information asymmetry provides discretionary power to the agent and enables the agent to pursuit its own interests. In this case, external stakeholders (the public) are principles who have an implicit contract with government to develop programs, establish policy, and deliver services that are in the best interests of the public. Without including incentives or disincentives that increase the willingness of the agent to provide information, the agent may not abide by the implicit contract and, consequently, not fulfill the goals of the principal. This problem is called moral hazard. To guard against this, citizens and other stakeholders demand more information and more access to and inclusion in decision processes as a means of holding government accountable to the implicit contract.ii As noted above, it is likely that transparency and participation represent two different mechanisms through which public organizations can respond to stakeholders, demonstrate 8

openness, and satisfy accountability demands. Transparency provides explicit information on the organization, its processes, events, activities, and outcomes. Often this is in the form of written formal reports, data, memos, schedules, and meeting agendas (Pitrowski and Borry 2009; Meijer 2009). But it can also be in the form of allowing observation to decision making activity. Participation, as defined above, is the opportunity for external stakeholders to provide input into decision making (Norris 2001; Jennings and Zeitner 2003; Ho 2002). Such activities create a more complex decision environment where government and stakeholders disclose preferences and positions. While this interaction is essentially a bargaining situation in which neither side is completely open, participation does provide stakeholders access to information about the public agency (Roberts 1997; 2004). Moreover, because information is disclosed in a richer contextualized setting, the government may be able to control the way in which decision preferences are presented. Such control may reduce the risk that one-way transmission of information (transparency) is misinterpreted or misapplied. When an organization provides explicit information in a one-way process its opportunity to control the interpretation and use of the information occurs prior to delivery, during the content development stage. An organization that engages in participation provides information as well – both explicit and tacit – but it is delivered within a more highly controlled context. Given this, it is likely that the two forms of openness will be used at different times or for different circumstances. For example, if the political environment is turbulent (or potentially turbulent) the agency may be less likely to provide explicit information and more likely to seek mechanisms to present contextualized information within a controlled context. When the political environment is less uncertain, agencies may be more able to anticipate interpretation of explicit information, and be more able to control the message before it is released. By controlling 9

the message, agencies may be able to maintain discretion and power, and at the same time, comply with accountability requirements. H1: Higher levels of external stakeholder influence (lower relative levels of agency control) will be associated with higher levels of participation. H2: Higher levels of external stakeholder influence (lower relative levels of agency control) will be associated with lower levels of transparency. When an organization increases participation, such that it increases the ability of stakeholder to provide input into decision making, it also gives up some level of control over information disclosure. In some cases, the agency may prefer contextualized disclosure over explicit disclosure, as noted in the framework above. However, greater participation may also require additional disclosure of explicit information; external stakeholders, especially non-government stakeholders (citizens, businesses and advocacy groups), will require some level of additional information in order to effectively participate in decision making or policy making. As a result, increased participation may lead to greater transparency. While, the agency’s calculus is strategic and continues to result in information asymmetry, it considers that further loss of control through participation is likely to be lower than the loss that would have occurred through transparency. H3: Higher levels of participation will be associated with higher levels of transparency. By contrast, higher levels of one-way disclosure of information may not lead to two-way engagement with stakeholders in ways that garner input to decision making. Transparency enables visibility into the organization; participation increases visibility into the organization as well, but it also enables external actors to actively contribute to the decision processes of the 10

organization. To change visibility into active engagement, some kind of additional incentive or requirement may be needed. For example, a third party (political body or court of law) may demand greater citizen involvement in agency decision making. Hence, release of more information (quantity or quality) is not expected to lead directly to greater participation. The role that technology plays in transparency and participation is potentially large. New information and communication technologies enable government agencies to more easily provide information and citizens and other stakeholders to access it (La Porte et al. 2001; Tolbert and Mossberger 2006; West 2004, Musso and Weare; Jun and Weare 2010; Bimber 1999; Welch and Fulla 2005). More than a convenience, new online applications provide the means for stakeholders and government to communicate, and to record the interactions. A perennial question in the field of electronic government concerns the relationship between technology and behavior. The perspective taken here is that the effect of technology on organizations depends upon socio-technical context in which it is embedded. This helps explain why “thirty years of systematic empirically grounded research about IT and organizational change suggests that many organizations have trouble in readily changing their practices and structures to take effective advantage of IT (Kling and Lamb, 2000).” In line with prior work, it is expected that organization that use internet technologies more will report higher levels of transparency and participation. H4: Greater use of information technologies and the internet will be positively associated with transparency and participation.

11

Method, Data and Measures Method The models comprise a set of equations in which the first equations specifies the Decision Transparency model and the second equation specifies the Decision Participation model. The two equations are estimated simultaneously due to the potential for endogeneity, or mutual causality, that may exists among Decision Transparency and Decision Participation. The estimation technique is appropriate here because it is possible that 1) a more active participation process may lead to an increase in the need for information dissemination, and 2) a more transparent process may lead to greater levels of participation. Although, the cross-sectional nature of the data used in this study does not allow strict testing of causal relationships, the simultaneous estimation of the two equations does recognize the potential two-way associations that are of interest in this paper. To carry out the two staged least squares estimation technique, endogenous variables are first regressed on all exogenous variables. In the second stage, predicted values obtained through the estimation are then substituted on the right hand side of the equation. This reduces the correlation between the variable and the error term for the equation and results in more consistent estimates than would have been obtained using ordinary least squares regression (Pyndyck and Rubinfeld 1991; Greene 2000). The two equations are: Information Dissemination = (Decision Participation, External Stakeholder Influence, Organization Characteristics, Information Technology, City Size) Decision Participation = (Information Dissemination, External Stakeholder Influence, Organization Characteristics, Participation Technology, City Size)

12

In these equations, the endogenous variables are found on both sides of the equation, while the exogenous (those determined outside of the equation) are only on the right side of the equation. It is important to point out that the analysis will conduct two types of estimation.

Data The analysis uses data from a web survey on e-government technology and civic engagement conducted for the Institute of Policy and Civic Engagement at the University of Illinois at Chicago. The survey was administered to government managers in 500 local governments with citizen populations ranging from 25,000 to 250,000. Because larger cities often have greater financial and technical capacity for e-government, all 184 cities with a population over 100,000 were selected while a proportionate random sample of 316 out of 1,002 communities was drawn from cities with populations under 100,000. For each city, lead managers were identified in each of the following five departments: general city management, community development, finance, the police, and parks and recreation. A total of 2,500 city managers were invited to take part in the survey. The survey began on August 2, 2010 and closed on October 11, 2010. A total of 902 responses were received for a final response rate of 37.9%.iii Endogenous Variables The two endogenous variables used in the paper are transparency and participation. Transparency is measured using an index of 12 survey items asking respondents about the frequency with which the agency uses different mechanisms to disseminate information. Participation is measured as an index of the perceived frequency with which 13 different stakeholders participated in agency decisions and policy making. The correlation between these 13

two indexes is 0.20 indicating that while they are significantly related in the positive direction, the level of correlation is low. Correlations between these two variables are similar across city sizes. However, they vary substantially by department type. For example, the correlations between participation and transparency are 0.54, 0.42, and 0.31 for finance, police and parks and recreation departments, respectively. ANOVA tests were conducted to understand more about how transparency and participation vary across department. Results presented in Table 1 show that in two departments – Mayor’s office and community development – participation is statistically higher than transparency. This compares to parks and recreation and police where transparency is perceived to be higher than participation. There is no difference between the two in finance departments. These findings not only demonstrate the potential importance that organizational and environmental factors have on the two variables, but they also provide some evidence of construct validity for the analysis. As discussed above, agencies that are more political or politically vulnerable are relatively more focused on participation than transparency. The opposite is true for those agencies that are more professionalized and less political. Table 1. Relative Differences in Transparency and Participation by Department Transparency Greater than Participation Mayor’s Office Community Development Finance Parks and Recreation Police

No Difference Between Transparency & Participation

Participation Greater than Transparency *** ***

NS *** ***

14

Means and standard deviations for these variables are presented in Table 2 below. However, departments also vary in terms of the absolute levels of transparency and participation. For example, respondents from Mayor’s offices reported the highest levels of participation, followed in order by community development, finance, parks and recreation and police. The department reporting the highest transparency was parks and recreation, followed by the Mayor’s office, finance, community development, and police.

Exogenous Variables The first equation for Transparency has multiple exogenous variables on the right hand side. Three measures used to operationalize external influence were constructed based on a two stage statistical process. First, factor analysis with varimax rotation was used to factor respondent perceptions of the influence of ten different external actors. The results, shown in Table 2, identified three different factors with loadings over 1.0 – Civil Society Influence, City Influence, and State Influence. The civil society influence factor includes advocacy groups, businesses, media and public opinion. The city influence factor includes the mayor, city departments, and the city council, while the state influence factor includes the governor, state courts and state legislature. As a second step, the different measures were then combined into additive scales and tested for reliability (Chronbach’s alpha, reported in Appendix 1). The variables used in the equations are the additive scales.

15

Table 2. How much influence do the following institutions or individuals exert over your department?

Advocacy Groups Businesses Media Public Opinion Mayor City Departments City Council Governor State Courts State Legislature

Civil Society Influence 0.84 0.72 0.76 0.82 0.11 0.26 0.13 0.12 0.13 0.12

City Influence

State Influence

0.12 0.24 0.13 0.16 0.86 0.58 0.81 0.20 0.03 0.12

0.09 0.24 0.14 0.03 0.09 0.09 0.15 0.80 0.81 0.88

Other organization related exogenous variables in the transparency equation include one variable that captures the existence of an information security policy in the organization (Security Policy); one capturing perceived level of employee discretion over the disclosure of information in the agency (Employee Disclosure Discretion); one that measures the percentage of employees that use the Internet (Percent Internet Use); and one, called Organization Size, is log of the number of full time equivalent employees in the organization. In addition, the equation contains dummy variables for department and city size where the reference groups are police and large city size, respectively. The five city size variables are coded based on the population of the 500 cities in the sample: small (under 50,000); small-medium (50,000 to 99,999); medium (100,000 to 149,999); medium-large (150,000 to 199,999); and large (200,000 to 250,000). Several of the same exogenous variables appear in the Participation equation including the three stakeholder influence, organization size, department and city size variables. However, this second equation removes the three variables related to information provision and substitutes them with three others related to participation. The Participation equation includes one 16

participation technology variable that captures the extent to which the agency uses different technologies to enable citizen and stakeholder participation (Participation Technology), and two variables capturing the perceived effect of technologies on participation (Bad Democratic Outcome and Good Decision Outcome). The last two outcome variables were created using the same two stage method described for the influence variables in the previous paragraph. Results are presented in Table 3. The Good Decision Outcome factor includes perceptions that electronic and communication technologies can lead to revitalization of public debate, improved government decision making and better policies. The Bad Democratic Outcome factor also includes three responses: increased conflict with citizens, undermining of democratic practices, and distortion of political information and facts. Factor loadings exceeded 1.0 and all reliability measures are included in Appendix 1. Table 3. In your opinion, to what extent do electronic information and communication technologies lead to the following outcomes?

Increase conflict with citizens Undermine democratic practices Distort political information and facts Revitalize public debate Improve governmental decision-making Lead to better policies

Good Decision Outcome 0.07 -0.06 0.01 0.76 0.92 0.92

Bad Democratic Outcome 0.77 0.85 0.84 0.12 0.07 0.04

Descriptive statistics for the variables used in this paper are presented in Table 4. A chart containing all variables, measures and reliability coefficients are presented in Appendix 1. In most cases, multiple measures were combined to enhance construct reliability. All reliability coefficients range between 0.68 and 0.90.

17

Table 2. Descriptive Statistics

Variable Name

N

Endogenous Variables Transparency Participation Stakeholder Influence Civil Society Influence State Influence City Influence Organization Factors Employee Disclosure Discretion Percent Internet Use Participation Technology Security Policy Mayor’s Office Community Development Finance Parks and Recreation Police Organization Size (logged) Perceptions on Participation Bad Democratic Outcome Good Decision Outcome City Size Small Cities Small to Medium Cities Medium Cities Medium to Large Cities Large Cities

Mean

Standard Deviation

780 798

3.01 3.04

0.62 0.93

831 833 823

2.41 1.95 3.13

0.75 0.82 0.86

889 902 830 887 902 902 902 902 902 902

3.28 70.98 4.21 2.07 0.16 0.24 0.16 0.23 0.22 3.74

1.16 30.50 4.50 0.91 0.36 0.43 0.37 0.42 0.41 1.61

858 875

6.78 9.43

2.60 2.47

902 902 902 902 902

0.39 0.28 0.20 0.08 0.05

0.49 0.45 0.40 0.27 0.21

Findings Table 4 presents results from the two-stage least squares regression model. The sample size has dropped due to listwise deletion of missing values. In general, the findings show mixed support for expectations that external stakeholder influence will affect transparency and participation. Participation model results show that as organizations experience higher levels of influence from civil society, they are more engage in more frequent participation activities that involve

18

stakeholders in decision making. Additionally, organizations that perceive higher amounts of influence from city government report more frequent participation activity. These findings are in line with prior literature and support the first hypothesis (H1). State influence does not seem to affect participation frequency. This is probably because many of the city level participation activities would be more related to local issues; decisions or policies that are less likely to result from state pressure. As predicted in hypothesis two (H2), city influence is negatively associated with information dissemination. Managers who perceive higher levels of influence from the city (mayor, other departments, and city council) are more likely to report lower levels of transparency. By contrast, neither state influence nor civil society influence is associated with information dissemination. It is possible to interpret the findings to indicate that as city government seeks greater control over the organization, managers and other employees are less likely to provide explicit information that could further loss of control and more likely to provide contextualized information through participative activity. That is, as the city principal seeks to increase its ability to direct the organization, the agency alters its information provision strategy from explicit (transparent) to contextualized (participative). In this way public agencies respond to external pressure in ways that enable them to simultaneously respond to informational demands and maintain control. In situations where bureaucratic politics are particularly strong, this strategy may be particularly prominent. Additionally, results show that while coefficient for participation in the transparency model is positive and significant, the coefficient for transparency in the participation model is not statistically significant. This provides evidence for the third hypothesis (H3) that greater participation will lead to more information disclosure. There is no evidence that information 19

disclosure does has a direct effect on participation. Further, it is useful to compare the direct and indirect effects of city influence on transparency. The direct effect is negative – a one unit increase in city influence is associated with a 0.09 point decrease in transparency – while the indirect effect city influence on transparency through participation is positive – a one unit increase in city influence results in a 0.07 point increase in transparency (0.07 x 1.03 = 0.07). The compensation is almost equivalent, further supporting the general expectation that transparency and participation are complementary strategies. Table 4. Simultaneous Model of Transparency and Participation Transparency: Information Dissemination Transparency Participation Civil Society Influence State Influence City Influence Security Policy Employee Disclosure Discretion Percent Internet Use Participation Technologies Good Decision Outcome Bad Democratic Outcome Organization Size Mayor’s Office a Parks and Recreation a Finance a Community Development a Small City b Small to Medium City b Medium City b Medium to Large City b Intercept N Adjusted R square Model Significance

Participation in Decision Making and Policy -0.06 (0.38)

1.03 (0.18) -0.12 (0.06) 0.01 (0.04) -0.09 (0.04) -0.07 (0.03) 0.01 (0.03) 0.01 (0.00)

0.03 (0.02) 0.33 (0.14) 0.21 (0.11) 0.25 (0.15) 0.39 (0.14) 0.12 (0.20) 0.09 (0.20) 0.05 (0.21) 0.01 (0.25) 0.43 (0.45) 580 0.14 ***

*p