Deliberative Democracy and Citizenship: In Search of the Efficacy Effect

Enthusiasm for deliberative democracy has grown in recent years, as many believe that it can create better citizens generally, and particularly increase their perceptions of political efficacy. Although the “efficacy effect” is frequently touted as a rationale for engaging in deliberative processes, there is little empirical research on the subject. This study examines the impact on perceptions of internal and external political efficacy of participation in one manifestation of a deliberative democracy process, the AmericaSpeaks 21 Century Town Meeting. Using quasiexperimental survey data collected from subjects at several different points in time, the study finds partial support for the claim that deliberative democracy produces the efficacy effect. Implications of the findings and directions for future research are discussed. 2 Journal of Public Deliberation, Vol. 6 [2010], Iss. 2, Art. 8 https://www.publicdeliberation.net/jpd/vol6/iss2/art8


INTRODUCTION
Deliberative democracy seeks to infuse government decision making with the reasoned discussion and collective judgment of citizens (Chambers 2003;Cunningham 2002;Elster 1998;Gutmann and Thompson 2004).In recent decades, deliberative democracy has resurged in popularity in the United States and in other Western nations.The increased interest is, in part, a function of the claim that deliberative democracy can create "better citizens" by fostering and improving civic attitudes and behaviors, such as political sophistication, interest, participation, trust, respect, empathy, and sociotropism or public spiritedness (Luskin and Fishkin 2003;Mansbridge 1995).One of the most important citizenship characteristics thought to be developed by democratic participation is political efficacy (Pateman 1970), defined as the "feeling that individual political action does have, or can have, an impact upon the political process" (Campbell, Gurin, and Miller 1954: 187).Although the "efficacy effect" is frequently touted as a rationale for engaging in deliberative processes, there is little empirical research on the subject.
This study seeks to fill part of this research gap by examining the impact on political efficacy of participation in one manifestation of the AmericaSpeaks 21 st Century Town Meeting, a deliberative democracy process that enables large groups of citizens to deliberate about important community issues and provide recommendations to government officials.The study uses quasi-experimental, longitudinal survey data to address three questions: Do participants and non-participants differ in terms of political efficacy before participation in the Town Meeting?Do participants have a higher sense of political efficacy after participation in the Town Meeting?
If participants have a higher sense of political efficacy after participation in the Town Meeting, do those effects last over time?
The article first explores the concept of political efficacy and examines its supposed decline in the United States.The second section defines deliberative democracy and explains how deliberative participation is theorized to affect the skills and dispositions of citizenship generally and political efficacy specifically.It also briefly examines previous research on the relationship between deliberation and political efficacy.The third section turns to methodology, discussing the AmericaSpeaks 21 st Century Town Meeting, the research design, variable construction, and analytic methods.The results of the study are presented in the fourth section, and the article concludes with a discussion of the implications and directions for future research.

THE DECLINE OF POLITCAL EFFICACY
Many scholars claim that the United States and other western nations are experiencing a "citizenship deficit" or a broad erosion of the skills and dispositions of citizenship among the general public (Frantzich 2005;Macedo et. al 2005;Nabatchi 2010;Putnam 2000;Skocpol and Fiorina 1999).Indicators of this decline are varied and numerous, including voter turnout, political participation, activity, and knowledge, engagement in campaign activities, trust in government, and associational memberships among others.A review of these indicators has led to a "growing sentiment among contemporary political scientists and political analysts that the foundations of citizenship and democracy in America are crumbling" (Dalton 2006: 1).For example, Macedo and colleagues (2005: 1) assert that: American democracy is at risk.The risk comes not from some external threat but from disturbing internal trends: an erosion of the activities and capacities of citizenship.Americans have turned away from politics and the public sphere in larger numbers, leaving our civic life impoverished.Citizens participate in public affairs less frequently, with less knowledge and enthusiasm, in fewer venues, and less equally than is healthy for a vibrant democratic polity….Americans should be concerned about the current state of affairs.The risk is not to our national survival but to the health and legitimacy of our shared political order.
The most important indicators of citizenship for the purposes of this study are internal and external political efficacy.Internal political efficacy refers to one's feelings of personal competence "to understand and participate effectively in politics" (Craig, Niemi, and Silver 1990: 290).It represents beliefs about the impact a person can have on politics and the political process as a result of her or his own skills and confidence.External political efficacy refers to one's perceptions about the responsiveness of the political system, both governmental authorities and institutions, to citizen demands (Craig, Niemi, and Silver 1990: 290).It is the belief that the political system is both receptive and responsive to the interests and actions of citizens.Since their development, internal and external political efficacy have become among of the most frequently used measures of general political attitudes (Craig, Niemi, and Silver 1990).
The American National Elections Survey (ANES) has measured political efficacy since the 1950s.As shown in Table 1, most Americans felt relatively efficacious about government through the early 1960s (Nie, Verba, and Petrocik 1976).This result was interpreted as evidence of American governmental stability and effectiveness in responding to popular concerns (Craig, Niemi, and Silver 1990).However, during the late 1960s and throughout the 1970s, feelings of cynicism and powerlessness grew among citizens, and perceptions of political efficacy declined.
For example, from 1966 to 1998, between 63 and 73 percent of ANES respondents agreed with the statement, "government is so complicated that a person like me can't really understand what's going on."The high level of agreement with this statement reflects a low level of internal political efficacy among Americans.In 2008, 68 percent of respondents agreed with the statement.
The results for external political efficacy, which is measured with two questions, are similar.Again, a high level of agreement reflects a low level of efficacy.In 1964, 29 percent of The percentage in agreement with that statement rose to an all time high of 66 percent in 1994, and fell slightly to 60 percent in 2008.Although ANES data show fluctuation in political efficacy among the general population over the decades, the declines among the disadvantaged and the young are more apparent.
Since political efficacy is thought to be a key indicator of the overall health of democratic systems (Craig, Niemi, and Silver 1990), scholars have suggested that negative political efficacy trends could threaten the established democratic order (Easton 1965(Easton , 1975;;Miller 1974).It is important to note that although perceptions of political efficacy have bounced up and down over the decades, a growing number of scholars assert that at the very least, these long-term negative developments in political efficacy "suggest, perhaps simplistically, that most Americans no longer feel that their views are represented adequately in traditional political venues" (Dennis and Owen 2001: 401-402).

DELIBERATIVE DEMOCRACY
Deliberative democracy connects public decision making to participation, reasoned discussion, and the collective judgment of citizens (Cohen and Fung 2004).Broadly defined, deliberative democracy is A form of government in which free and equal citizens (and their representatives), justify decisions in a process in which they give one another reasons that are mutually acceptable and generally accessible, with the aim of reaching conclusions that are binding in the present on all citizens but open to challenge in the future (Gutmann and Thompson 2004: 3-7).
In contemporary political theory, deliberative democracy stands in contrast to aggregative democracy.While both models share assumptions about the structuring of democratic institutions, they focus on different decision making processes (Young 2000).The aggregative model, which forms the basis of representative government, is "vote-centric" (Chambers 2003); it relies on the aggregation of individual preferences to arrive at public policy decisions, and uses voting and bargaining to determine how those individual preferences are cumulated (Mansbridge 1980;Young 2000).Because voting and bargaining encourage strategic behavior based on individualist and economic incentives (Barber 1984;Buchanan and Tullock 1962;Riker 1962Riker , 1982)), the aggregative model is also an adversarial model of democracy (Mansbridge 1980;Miller 2000).In this adversarial model, public policy (and other governmental) decisions are seen as a zero-sum game where majority rules (for a discussion, see Radcliff and Wingenbach 2000).Moreover, this model leaves the main work of governance to professional political elite; "the concept of [citizen] participation has only the most minimal role.Indeed, not only has it a minimal role but … the emphasis [is] placed on the dangers inherent in widespread popular participation in politics" (Pateman 1970: 1).
In contrast, deliberative democracy is "talk-centric" (Chambers 2003); it emphasizes well-informed citizen participation such that "citizens can pool information and ideas, bring local knowledge to the table, establish greater levels of equality and political opportunity, and the like" (Leib 2004: 3).Deliberative democracy fosters cooperation and mutual understanding rather than winning and losing (as adversarial democracy seems to); it purports to give all citizens a "voice" rather than just the most powerful or the most numerous (as tends to occur in majoritarian democracy); and it encourages citizens to make decisions based on "public reasons" that can be supported through deliberation rather than on individual prejudices that thrive in the privacy of the voting booth (Levinson 2002).
Before examining the potential benefits of deliberative democracy, three additional points must be made.First, "[a]lthough theorists of deliberative democracy vary as to how critical they are of existing representative institutions, deliberative democracy is not usually thought of as an alternative to representative democracy.It is rather an expansion of representative democracy" (Chambers 2003: 308).This relates to the second point, namely that most deliberative processes include some form of voting or preference aggregation mechanism; however, unlike in traditional aggregative systems, the focus is on the processes of preference formation that precede voting.Finally, deliberative democracy does not translate into a single method or process; rather, it is an umbrella term for a wide variety of processes, such as the Kettering National Issues Forum, deliberative polling, consensus conferences, planning cells, citizen juries, study circles, the AmericaSpeaks 21 st Century Town Meeting, and various e-democracy initiatives, among others (Gastil and Levine 2005; for a chart describing dialogue and deliberation processes, see: http://www.thataway.org/exchange/files/docs/ddStreams1-08.pdf; for process descriptions, see: http://www.thataway.org/exchange/files/docs/Framework_Links.doc).
A discussion about the specific design of these deliberative democracy processes is beyond the scope of this article; however, key features shared among the processes include: 1) a focus on action, 2) an appeal to values, 3) the absence of pre-existing commitments, 4) mutuality of focus, 5) the free exchange of knowledge and information, and 6) activities occurring within small groups, although the overall process may involve thousands of people (Lukensmeyer and Brigham 2005;Torres 2003).Differences among the processes include, but are not limited to 1) who participates in deliberation, 2) how participants exchange information and make decisions, and 3) the link between the deliberations and policy or public action (Fung 2005;Bingham, Nabatchi, and O'Leary 2005).

THE EDUCATIVE EFFECTS OF DELIBERATIVE DEMOCRACY
According to theorists, deliberative democracy may have educative effects for citizens.Carole Pateman's (1970) seminal work, Participation and Democratic Theory, is widely noted for its exploration of this issue, but other authors also articulate the theory well (e.g., Mansbridge 1995;Morrell 1998Morrell , 2003Morrell , 2005)); thus, only a brief overview is provided here.Pateman (1970: 24-25) asserts that "the central feature of participation is an educative one, using the term education in the widest sense.[The system] is designed to develop responsible individual social and political action through the effect of the participatory process."Deliberative participation is especially expected to have significant impacts on political efficacy (Pateman 1970; see also Finkel 1985;Mansbridge 1995;Morrell 1998Morrell , 2003Morrell , 2005)).Pateman (1970) explains that as one participates in politics, s/he acquires political skills and perceptions of self-competence, which are qualities necessary for popular self-government and effective control over one's environment.
In addition, participation creates a circular causal process (Finkel 1985) whereby "[p]articipation develops and fosters the very qualities necessary for it; the individuals participate, the better able they become to do so" (Pateman 1970: 42-43).
Noting that Pateman's (1970) articulation better represents the concept of internal political efficacy, Finkel (1985: 893-894) asserts that the effects of participation on external political efficacy can be understood with the "mobilization of support" theory, articulated by scholars such as Ginsberg (1982), Weissberg (1975), andWright (1976).In this view, participation should increase the belief that regime authorities are responsive to attempted influence from citizens (Craig 1979).The idea is that participation will promote citizens' feelings of legitimacy toward the political system and increase acquiescence to government (Finkel 1985).
In turn, regime stability is enhanced by "inducing citizens to believe that the government is responsive to their own needs and wishes" (Ginsberg 1982: 182) and by "encouraging [citizens] to believe that they are ultimately controlling the government …and keeping them committed to the existing system" (Olsen 1982: 6).
Despite the claims about and interest in deliberative democracy, as well as calls for more studies, empirical research lags behind both theory and practice (Delli Carpini, Cook, and Jacobs 2004).The few empirical studies on deliberative processes that do exist show mixed results.On the one hand, several studies suggest that under certain conditions, deliberation can produce more sophisticated, tolerant, and participative citizens (Fung 2001;Fung and Wright 2003;Gastil and Dillard 1999b;Gastil, Deess, and Weiser 2002;Luskin and Fishkin 1998;Sulkin and Simon 2001).Deliberation has also been found to increase participants' levels of political knowledge (Cook and Jacobs 1999;Fishkin and Luskin 1999;Hansen and Anderson 2004;Luskin, Fishkin, and Jowell 2002) and help participants form more reflective, coherent, and comprehensive judgments about the issue at hand (Carpella, Price, and Nir 2002;Fishkin and Luskin 1999;Luskin, Fishkin, and Jowell 2002;Mathews 1994;Sturgis, Roberts, and Allum 2005).
On the other hand, some studies show that the short-term gains from deliberative participation result in little long-term civic engagement activity (Kimmelman and Hall 1997;Delli Carpini and Keeter 1996).Deliberation can also reduce the consistency between attitudes and behavior among subjects, produce decisions that conflict with both expert decisions and subjects' own personal opinions, and lead to decisions participants later regret (Holt 1993(Holt , 1999;;Wilson, Dunn, Kraft, and Lisle 1989;Wilson and Schooler 1991).Similarly, studies show that deliberation can cause participants to doubt that a "correct" decision exists (Armor and Taylor 2003;Iyengar and Lepper 2000), which can leave them feeling more anxious and frustrated about the issue under discussion after deliberation than before (Button and Mattson 1999;Hendricks 2003;Kimmelman and Hall 1997).
Little research has directly examined the relationship between deliberation and political efficacy (Morrell 2005), and the few studies that do exist provide mixed results.Some studies suggest increases in perceptions of political efficacy (Gastil andDillard 1999a, 1999b;Doble, Higgins, Begasse, and Fisher 1996;Smith 1999), but other studies show no effects of participation on political efficacy (Gastil 2004;Walsh 2003).Clearly, more research is needed.
This study is a first step in that direction.

STUDY OVERVIEW
This study examines the effects on political efficacy of participation in the AmericaSpeaks 21 st Century Town Meeting as convened by the United Agenda for Children.
AmericaSpeaks is a non-partisan, nonprofit organization based in Washington, D.C. that uses the 21 st Century Town Meeting (hereafter the "Town Meeting"), to engage large, demographically representative groups of citizens (from 100 to 5,000+) in simultaneous deliberation around a specific policy issue in a particular political community (Lukensmeyer and Brigham 2005).The goals of the Town Meeting are to ensure that: 1) all voices, including the general public and key stakeholders are at the table, 2) the priorities of the public get the attention of decision makers and the media, and 3) a substantial segment of the public supports the results of the forum and has a stake it its implementation (Lukensmeyer, Goldman, and Brigham 2005: 157).
To meet these goals, the Town Meeting has several critical components (for a detailed discussion, see Lukensmeyer, Goldman, and Brigham 2005;Lukensmeyer and Brigham 2005; see also www.americaspeaks.org).First, AmericaSpeaks works closely with sponsors to conduct widespread outreach and targeted recruitment such that the assembled group of citizens resembles the demographic makeup of the community.Second, all participants receive materials that neutrally detail and balance the perspectives on the issue(s) under consideration.Third, participants are randomly assigned to tables that seat between ten and twelve people including a trained facilitator who helps keep the dialogue on target.The fourth and fifth components are participation technologies, immediate reporting, and theming.Networked laptop computers are used at each table to instantly record the ideas of participants.This information is transmitted to a central database, where the "Theme Team" distills it into major topics, issues, ideas, and messages.The ideas from this facilitated theming are then presented back to all participants in the room to "build collective ownership of the group's work" (Lukensmeyer, Goldman, and Brigham 2005: 159).In addition, all participants are given a wireless polling keypad for voting on issues and comparing their personal positions to those of others.The final component links the process and outcomes to decision makers, who are not only participants in the process, but who also pledge (and are expected) to seriously review and consider the input of participants.
The design of the Town Meeting makes it an excellent deliberative process to study, as many of its design choices create a context within which one would expect to find evidence of the efficacy effect (Fung 2003(Fung , 2005;;Williamson 2004).
The United Agenda for Children (UAC) is a coalition of public, private, and non-profit organizations in Mecklenburg County, North Carolina.The UAC goal is to create and implement a "united action plan that would ensure that all children in Mecklenburg County are healthy, safe and well-educated" (UAC 2006).The first step toward that goal was the UAC Town Meeting, held on December 11, 2004.At that meeting, more than 1,000 people who live in Mecklenburg participants, and about 240 were walk-ins that were not accounted for at the registration table.A total of 525 participants completed surveys, yielding a response rate of about 47.6%.Nonparticipants were not surveyed at Time 2.  16.2%.Non-participants returned 40 surveys, producing a response rate of 32%.It is important to note that the original research design provided for individual level links between subjects at Times 1 and 3; however, due to low responses rates, only 17 participants and 36 non-participants can be linked between these times.Given these low numbers, comparisons using date from Time 3 are given minimal attention in this article.

Variable Measurement
Internal and external political efficacy are measured with standardized summative indices consisting of several different survey items.All items are 5 point Likert scale questions with the anchors of 1 = strongly disagree and 5 = strongly agree.Several steps were used to test the survey items and the indices for validity and reliability.1 First, items were grouped into a potential index based on face validity, and principal components factor analysis was conducted to test the fit of each item within the preliminary index.Second, items were selected based on both the unrotated and quartimax rotated factor loadings, then the psychometric adequacy of each index was evaluated using Cronbach's alpha, an estimate of the reliability and internal consistency of responses to the items comprising a given scale (Cronbach 1951).Finally, items were combined into a summative index, which was then standardized.Cases where a respondent did not answer a question that was included in a final index were omitted from the analyses.

Internal political efficacy (IPE)
is measured with the following three items, which ask how strongly the subject agreed that: 1) Sometimes politics and government seem so complicated that a person like me can't really understand what's going on [COMPLICATED]; 2) I consider myself well-qualified to participate in politics [MEQUAL]; and 3) I often don't feel sure of myself when talking about politics or government [NOTSURE].

External political efficacy (EPE)
is measured with the following four items, which ask how strongly the subject agreed that: It is important to note that these four questions relate to perceptions of government's responsiveness, not to the actual responsiveness of the convening organization (United Agenda for Children) to the Town Meeting input.
Table 2 displays the rotated factor loadings for each item in each index, as well Cronbach's reliability score. 2 The factor loadings in both indices are all reasonable, perhaps with the exception of LOCRESP in the EPE index.The reliability scores, .610and .676for internal and external political efficacy respectively, are less than .70, the standard cut-off for high reliability, which suggest that there is some measurement error in theses variables.3However, scholars have long associated measure of political efficacy with a "lack of validity and reliability" (Craig, Niemi, and Silver 1990: 289; for a discussion particularly about internal political efficacy, see Morrell 1999Morrell , 2003Morrell , 2005)).

Analyses
Because this study uses non-equivalent comparison groups without random assignment, it suffers from selection bias threats.Of particular concern is the threat of self-selection, because the outcome "effect may be due to the difference between the kinds of people in one experimental group as opposed to another" (Cook and Campbell 1979: 53).The potential for selection bias is manifest in at least three ways in this study: 1) potential differences between participants and non-participants; 2) potential differences between the pre-registrant and walk-in participants; and 3) potential differences among those who completed surveys and those who did not (these are complicated by the fact that the survey was administered at multiple points in time).
There are two types of self-selection bias: selection on the observables and selection on the unobservables.Selection on the observables, concerns differences between groups that can be estimated from the data.In this study, it is highly probable that variables besides participation status affect political efficacy among the subjects.For example, a significant body of research indicates that political efficacy is correlated with demographic characteristics and that demographic characteristics are related to participation (Dowse and Hughes 1972;Kleiman 1976;Lyons 1970;Martinussen 1972;Orum 1989;Verba and Nie 1972).To control for this problem, estimates of the difference in political efficacy between two groups (i.e., nonparticipants and participants, or participants at different points in time), are presented with OLS regression, where IPE and EPE are separately examined as a function of participation status and demographic variables, including race, age, gender, employment status, parental status, education level, and SES.
When comparing participants and non-participants, the OLS models are specified as: These OLS models control for the potential selection on the observables problem in that they examine the IPE and EPE differences between the groups while holding demographic variables constant.However, the OLS estimates of the difference between the two groups may be biased due to selection on the unobservables, or information not captured in the survey that affects the decision to participate in the program and the outcome, in this case, political efficacy.
To control for this problem, I also analyzed the data using Heckman treatment effect models for comparisons of participants and non-participants, and propensity score matching models for comparisons of participants at different points in time.
Heckman treatment effect models (TEM) use a two equation approach that addresses several classes of selection bias, including bias due to self-selection (Briggs 2004;Heckman 1978Heckman , 1979;;Heckman andRobb 1985, 1986).The first equation is a selection equation, which predicts some binary outcome.In this case, the binary outcome is participation (i.e., whether one is a participant or non-participant), as predicted by the demographic variables and the variables that indicate previous political participatory activities.The second equation is a structural equation, which examines the ultimate dependent variable(s) of interest.In this case, the dependent variables are internal and external political efficacy, which were regressed as a function of participation and all demographic variables.Using these two equations, the marginal effects of participation on political efficacy are assessed.The Wald chi-square statistic (a measure of goodness of fit) and the likelihood ratio test of independent equations (LR test, a measure of the relationship between the two equations), are used to determine whether the OLS regression models or the Heckman models provide more consistent estimators of the efficacy differences between participants and non-participants.The TEM models are specified as: Propensity score matching (PSM) is a non-experimental method of sampling that deals with the problem of selection bias, and thus allows for more robust comparisons of two groups.
PSM produces a control group whose distribution of covariates is similar to that of the treated group (Rosenbaum and Rubin 1983).Conditioning many covariates produces a problem of dimensionality that calls for a method of summarizing multi-dimensional covariates.PSM adopts one-dimensional propensity scores to achieve the "dimension reduction" (Hahn 1998: 317;D'Agostino and Rubin 2000).The PSM method consists of four steps: 1) estimating the propensity score; 2) checking the balance of covariates; 3) matching subjects either through pair matching or sub-classification; and 4) calculating the average treatment effects (Becker and Ichino 2002;Dehejia andWahba 1999, 2002;Rosenbaum andRubin 1983, 1984).
In this study, a binary probit model is employed to estimate the propensity score.Time is the dependent variable, and covariates include all demographic variables and variables that indicate previous political participatory activities.The second equation examines IPE and EPE as a function of time and demographic characteristics.Participants at each time of comparative interest are then "matched" using one-to-one pair matching without replacement.Paired t-tests are conducted to get consistent estimates of program effects, that is, to determine whether participation in the program had an impact on subjects' perceptions of political efficacy as measured by the changes in political efficacy among participants at various points in time.The specification of the PSM models is: where T is a binary variable representing time, with 0 = participants at earlier time, and 1 = participants at later time, ∑X k = demographic variables, and ∑V k = previous participatory activities.

RESULTS
The results are presented in the following order.First, participants and non-participants at Time 1 are compared to determine whether they had different perceptions of political efficacy before the UAC Town Meeting.Second, participants at Time 2 are compared to participants at Time 1 to determine whether political efficacy increased after participation in the UAC Town Meeting.To determine whether the effects of participation on political efficacy persisted over time, participants at Time 3 are compared separately to participants at Time 2 and Time 1.
Participants and non-participants at Time 3 are then compared, as are non-participants at Time 3 and Time 1.

Participants and Non-Participants at Time 1
Both the OLS and TEM models were employed to find statistically consistent estimators of the differences in IPE and EPE between participants and non-participants at Time 1.For IPE, the TEM results show a large and significant chi-square statistic (52.40, p < .000)and a significant LR test (14.45,p < .0001),indicating that we must reject the null hypothesis that rho is zero.This means that the two equations are related, there is self-selection in the samples, and that the OLS estimate is biased.Therefore, the Heckman results provide a better estimate of the effect of participation on IPE.The Heckman structural model indicates that, holding all else constant, participation was related to IPE in a statistically significant way (p < .000),with participants feeling less internally efficacious than non-participants.These results are displayed in Table 3.

Sample Size 315
In terms of EPE, the chi-square statistic is fairly small and significant (21.49, p < .006),but the LR test is not significant (1.26, p < .2611),indicating that we should not reject the null hypothesis that rho is zero.This means that the two equations are not related and that there is no self-selection in the samples.Thus, the OLS estimate for EPE is not biased and is better than the Heckman result.The OLS results indicate that when controlling for demographic characteristics, participants had significantly lower levels of EPE (p < .037)than non-participants.These results are displayed in Table 4. Together, these results suggest that prior to the UAC Town Meeting, participants had significantly lower perceptions of internal and external political efficacy than non-participants..055

Participants at Time 1 and Time 2
The analyses of participants at Time 1 and Time 2 get to the heart of the study as they search for evidence of the efficacy effect among participants.OLS and PSM were used to estimate the effects of participation on IPE and EPE.Since PSM does not provide statistics with which to compare the efficiency of estimators, both sets of results are presented and discussed.
The OLS results, displayed in Table 5, indicate that when controlling for demographic characteristics, time had no significant effect on IPE (p < .506);however, it did have a statistically significant effect on EPE (p < .024).In other words, the OLS results suggest that immediately following the UAC Town Meeting, participants experienced a statistically significant increase in external political efficacy, but not internal political efficacy.The results from the PSM models are displayed in Table 6.The one-to-one pair matching of 93 participants suggests that after the UAC Town Meeting, perceptions of IPE increased by about 7.4%, although this increase is not statistically significant (p < .55).In contrast, EPE increased by over 31%; this result is statistically significant (p < .0059).Together the OLS and PSM results show that participation had a positive and statistically significant impact on external political efficacy.Internal political efficacy also increased, though not in a statistically significant way.

The Persistence of the Efficacy Effect
Three sets of comparisons were made to examine the persistence of the efficacy effect.
First, IPE and EPE among participants at Time 3 were compared to participants at Time 2 and Time 1 using OLS and PSM methods.Second, participants at Time 3 were compared to nonparticipants at Time 3 using OLS and TEM.Finally, non-participants at Time 1 were compared to non-participants at Time 3 using paired t-tests.The results of these analyses are below.
First, the results of both the OLS and PSM models indicate that internal and external political efficacy did not change in a significant way for participants between Time 2 and Time 3.
The OLS results, displayed in The OLS results comparing participants at Time 1 and Time 3 are displayed in Table 9.
They indicate that when controlling for demographic characteristics, time had a statistically significant effect on IPE (p < .008)but not on EPE (p < .121).In other words, participants experienced a statistically significant increase in internal political efficacy from immediately before the UAC Town Meeting until 24 months after the UAC Town Meeting.External political efficacy also increased, but not in a statistically significant way.As noted previously, it is probable that there is significant endogeneity in these models; therefore, the OLS estimates are likely to be biased upward.
Moreover, the TEM comparisons of participants and non-participants suggested that the OLS estimates for IPE suffered from selection bias.Given this, it is reasonable to assume that the PSM results are more consistent than those from the OLS model.Following this logic, it appears that only EPE increased in a statistically significant way among participants between Time 1 and Time 3. Second, OLS and TEM models were used to compare participants and non-participants at Time 3. The TEM results show a small but significant chi-square statistic for IPE (19.24, p <.0136), and small but insignificant chi-square statistic for EPE (11.50, p < .1751).The LR tests indicate that we should not reject the null hypothesis that rho is zero for either IPE (1.18, p < .2775)or EPE (0.62, p < .4328).Therefore, the two equations are not related, there is no selfselection in the samples, and the OLS estimates of the IPE and EPE differences between participants and non-participants are better than those generated by the TEM.Accordingly, only the OLS results, displayed in Table 11, are presented here.These results indicate that when controlling for demographic characteristics, there were no significant differences in IPE between participants and non-participants at Time 3 (p < .446).This result suggests that the gains in internal political efficacy that participants experienced, though not statistically significant, were large enough to catch them up with non-participants.In contrast, however, EPE was significantly lower among participants at Time 3 than non-participants at Time 3 (p < .085).A brief summary of the results is in order.First, before the UAC Town Meeting, participants had significantly lower perceptions of both internal and external political efficacy than non-participants.This suggests that in this case, AmericaSpeaks was effective in engaging the less efficacious among the citizenry in the UAC Town Meeting process.Second, after the UAC Town Meeting, participants experienced an increase in both internal and external political efficacy; however, only the increase in external political efficacy was statistically significant.
These results suggest that the UAC 21 st Century Town Meeting produced the efficacy effect, particularly with respect to external political efficacy.Third, there were no significant changes in internal or external political efficacy among participants at Time 2 and Time 3, suggesting that the efficacy effect persisted over time.Moreover, there were no statistically significant changes in internal political efficacy among participants at Time 1 and Time 3; however, the difference in external political efficacy was statistically significant, again providing partial evidence of the efficacy effect in the case of the UAC Town Meeting.Finally, the strength of the efficacy effect is evidenced by other results.There were no significant changes in non-participants' perceptions of IPE or EPE between Time 1 and Time 3, and unlike before the UAC Town Meeting, there were no statistically significant differences in internal political efficacy between participants and non-participants at Time 3. However, there were significant differences in external political efficacy between participants and non-participants 24 months after the UAC Town Meeting.
participation, such as voting, campaigning, and protesting.The deliberative nature of the 21 st Century Town Meeting, which is fundamentally different from that of traditional political activities, may have attracted less efficacious citizens.Perhaps these individuals viewed the UAC Town Meeting as an opportunity to express their interests and exact responsibility and accountability from government officials.A second, and even more plausible explanation is that the context and subject matter of the UAC Town Meeting (youth policy pertaining to health, education, and safety) was important and controversial enough to motivate the less efficacious to participate.Thus, maybe political participation in such events is contingent on the subject matter of deliberation and on the context in which political participation is to take place.Whatever the explanation, it appears that in the case of the UAC Town Meeting, the AmericaSpeaks process was successful in engaging less efficacious citizens.
The engagement of the less efficacious, while good, may not be a sufficient reason to use deliberative processes; however, if such processes generate the efficacy effect, then the strength of advocates' arguments grows.The results of this study show that there were statistically significant increases in subjects' perceptions of external political efficacy after participation in the UAC Town Meeting.Moreover, internal political efficacy also increased after the UAC Town Meeting, but not in a statistically significant way.Thus, in this case, the UAC Town Meeting was at least partially successful in generating the efficacy effect among participants in that it significantly impacted external political efficacy, or feelings about the responsiveness of government and its authorities.In general, these findings buttress the mobilization of support of theory more so than Pateman's (1970) theory.
The potential value of deliberative democracy increases to the extent that deliberative processes result in long-term and sustainable benefits to citizenship indicators.The results of this study show that among participants, internal and external political efficacy neither increased nor decreased in a significant way 24 months after participation (i.e., between Time 2 and Time 3).
Moreover, the results indicate that among participants, external political efficacy was significantly higher at Time 3 than at Time 1. Internal political efficacy also increased among participants between these time periods, although not in a statistically significant manner.These results suggest that the increases in perceptions of external political efficacy persisted over time after participation.It is, however, important to note that participants at the UAC Town Meeting had access to data showing that the meeting helped produce an increase in funding for school nurses and launched a unified oversight committee to advocate for children's issues.This may have influenced their perceptions of external political efficacy.
Together, these results provide interesting insights about the nature of the relationship between deliberative participation and political efficacy.Research shows that external efficacy is profoundly affected by political events (Clarke and Acock 1989;Gurin and Brim 1984); it evolves and transforms in response to changes in the political landscape.Moreover, deliberative theory rests on the observation that effective participation requires the belief that participation matters, that government is listening and will be responsive (Finkel 1985).It is not surprising then, that external efficacy in this study is reactive to interventions that are explicitly supported by (and have the explicit support of) key political decision makers, as is found in the 21 st Century Town Meeting.
Whereas external efficacy is malleable, internal efficacy may be more stable.Internal efficacy is a product of political socialization and all of its accompanying social, familial, educational, and other forces (Almond and Verba 1963).As such, it develops over time in response to personal experiences, vicarious experiences, social persuasions, and physiological factors (Bandura 1994).Thus, while perceptions of internal political efficacy can change, it is not surprising that a one-time, eight-hour intervention did not change significantly the way people feel about their abilities to participate in politics.This then, may be the lynchpin on which the educative effects of participatory theory turns.It may not be enough to have a singular deliberative experience; deliberative events may need to happen several times before internal efficacy will change in a statistically meaningful way.This issue merits future testing.
Two additional results from this section are worth discussing.First, statistically significant differences in internal political efficacy between participants and non-participants did exist at Time 1, but not at Time 3. Therefore, the results at Time 3 are evidence that participation did impact internal political efficacy, albeit not in a statistically significant manner.Significant differences in external political efficacy between participants and non-participants at Time 3 remained.Second, the results show that there were no significant differences in non-participants' perceptions of internal or external political efficacy between Time 1 and Time 3. This is important because it demonstrates that the comparison group experienced no changes in political efficacy over time.When viewed holistically, these results add strength to the findings regarding the effects of participation in the UAC 21 st Century Town Meeting on perceptions of political efficacy, and particularly external political efficacy.
Several limitations to this study are worth noting, each of which gives rise to directions for future research.First, although the 21 st Century Town Meeting has many of the design features one would expect from an "ideal" deliberative process (Williamson 2004), it is still very specialized and highly unique.The Town Meeting process involves lengthy and elaborate preparation, significant use of new technologies, and varying modes of deliberation.It is a highly managed process, from the development of a neutral statement of the issue (including issue definition), to intensive facilitation and deliberation, to the writing of outcomes and recommendations for decision makers.The process is intensive for participants, especially when compared to many other deliberative processes, involving at least 8 full hours of deliberation and discussion.When coupled with the design elements of the Town Meeting, these features raise important questions about the generalizability of findings from this study.
I believe, however, that the findings are generalizable to other manifestations of the AmericaSpeaks 21 st Century Town Meeting.AmericaSpeaks conducts several Town Meetings a year.Although the Town Meetings occur in different communities and focus on different issues, the format of the Town Meeting process stays the same, increasing the likelihood of generalizability in these settings.Nevertheless, it is plausible that the effects of participation on political efficacy may change depending on the subject matter of deliberation.Moreover, the findings may also hold in studies of deliberative democracy where the observed process shares similar procedural elements or structural characteristics with the 21 st Century Town Meeting.
These are matters for future research.
Second, individual respondents in this study cannot be linked over time, and we must be wary in asserting the overall impact of the process on perceptions of political efficacy.Given the low response rates, this is particularly true for the results on the persistence of the efficacy effect.
The linking of respondents over time would be a feature in the ultimate deliberative democracy research design, as it would readily allow scholars to assess the individual level changes that accrue as a result of participation.Moreover, it would be useful to examine multiple examples of a singular deliberative democracy process and to engage in comparative studies of two or more deliberative democracy processes.
Finally, it is important to recognize that the dependent variables for internal and external political efficacy have measurement error.This may raise questions about the strength of the results reported here.Despite the measurement error, the effects of participation on external political efficacy were significant above the 90% confidence level.If political efficacy is a key characteristic to be developed by participation, then it is necessary to have reliable and consistent measures of the concept.While some work in this area is progressing, particularly with respect to internal political efficacy (Morrell 2003(Morrell , 2005)), a considerable amount of additional work is needed.On a related note, scholars also need to develop reliable and consistent measures of other indicators of citizenship skills and dispositions.For example, researchers need to develop measures for political empathy, trust, sophistication, respect, and sociotropism.Until these measures are developed and tested, the impacts of deliberative democracy on the skills and dispositions of citizenship cannot be fully understood and appraised.

CONCLUSION
In addition to showing that the UAC 21 st Century Town Meeting was successful in encouraging less efficacious citizens to participate, this study provides partial support for the idea that deliberative democracy can produce the efficacy effect.Most notably, the results show that external efficacy, which regards perceptions about the responsiveness of government to citizen demands, increased in a statistically significant way following participation in the UAC Town Meeting.Internal political efficacy, which regards perceptions of one's competence to engage in politics, also increased after participation, although not in a statistically significant way.These results provide some support for the claims that deliberative democracy produces the efficacy effect; however, the results also suggest a need for additional theorizing and testing.
Despite its limitations, this research still informs the study of deliberative democracy and has theoretical and practical importance.The results give much needed indications about the impacts and outcomes of deliberative democracy processes, in this case, the AmericaSpeaks 21 st Century Town Meeting.Scholars can use this information to refine theory, enhance practice, and design better research studies.Research needs to move beyond normative speculation about the intrinsic benefits of deliberative democracy to theoretical development and empirical assessment.
Scholars need to better articulate the theory of deliberative democracy, and develop specific ideas about how various deliberative features and practices contribute to outcomes.Until such theorizing and testing is underway, we cannot determine whether institutionalizing deliberative democracy in the regular practices of government is warranted.

Figure
Figure 1: Subjects and Response Rates for Data Collection Time 1 Time 2 Time 3 1) Elected officials don't care what people like me think [NOCARE]; 2) People like me don't have any say about what the government does [NOSAY]; 3) Elected officials are only interested in people's votes [OFFVOTE]; and 4) Local government is responsive to citizen concerns [LOCRESP].
a binary variable representing participation, with 0 = non-participant and 1 = participant, and ∑X k includes all demographic variables.When comparing participants at different points in time, the OLS models are specified as: a binary variable representing time, with 0 = participant at the earlier time, and 1 = participant at the later time, and ∑X k includes all demographic variables.
a binary variable representing participation, with 0 = non-participant and 1 = participant, ∑X k = demographic variables, and ∑V k = previous participatory activities.

Table 1 : Internal and External Political Efficacy, 1964-2008 a
Numbers represent percentage of respondents who agreed with the statements measuring internal and external political efficacy.High agreement signals low levels of political efficacy.b Internal political efficacy is measured a statement that asks subjects how strongly they agree that "Sometimes politics and government seem so complicated that a person like me can't really understand what's going on."c,d External political efficacy is measured with two statements that ask subjects how strongly they agree that "People like me don't have any say about what the government does" (NO SAY), and "I don't think public officials care much what people like me think" (NO CARE).Source: ANES Guide to Public Opinion and Electoral Behavior.Available at: http://www.electionstudies.org/nesguide/gd-index.htm#5 respondents agreed that "people like me don't have any say about what the government does."This rose to an all time high of 56 percent in 1994, and was at 49 percent in 2008.Likewise, in 1964 only 36 percent of respondents agreed that "I don't think public officials care much what people like me think."

Table 2 : Rotated Factor Loadings and Alpha Scores for IPE and EPE* Factor Loadings
In addition to the IPE and EPE indices, several demographic characteristics, including race, age, gender, employment status, parental status, educational level, and socioeconomic status (SES), are used as control variables.Previous political participatory activities, including whether in the last year the subject had attended a neighborhood or town-hall meeting, a meeting involving school affairs, or contacted a local public official, and whether the subject voted in the November 2004 presidential election, are also used as control variables in some models.All variables for demographic characteristics and previous political participatory activities are dichotomous.4

Table 7 : OLS Results -IPE and EPE among Participants at Time 2 and Time 3
Table 7, indicate that when controlling for demographic characteristics, time had no significant effect on IPE (p < .675)or EPE (p < .291).The PSM results, displayed in Table 8, agree; participants did not experience statistically significant changes in IPE (p < .8108)or EPE (p < .6753) between Time 1 and Time 2. These results suggest that the increases in perceptions of IPE and EPE as a result of participation in the UAC Town Meeting persisted, but did not change significantly over time.

Table 9 : OLS Results -IPE and EPE among Participants at Time 1 and Time 3
Interestingly, the PSM results, displayed in Table10, suggest the opposite, that participants experienced a statistically significant change in EPE (p < .09),but not IPE (p < .4792)during this time period.