Explaining Political Efficacy in Deliberative Procedures-A Novel Methodological Approach

So far, not much research has been done explaining the change of political efficacy in deliberative procedures, and case studies or experiments prevail in the field. Quantitative, systematic studies of real-life cases are missing. This article contributes to filling this gap. It identifies factors which lead to increased group-related political efficacy in deliberative procedures applying an almost novel method, i.e. a quantitative meta-synthesis combining and aggregating data from case studies. The study focuses exemplarily on Germany. The findings indicate that an improvement of political efficacy is more likely when deliberative procedures take place in a municipality, which has institutionalized citizens’ involvement in a local ‘participatory plan’ (‘local constitution’) and provides respective staff.


Introduction
Effects on political efficacy, i.e. citizens' perceptions that they have an influence on public issues, are of great interest in research on deliberative procedures (e.g.Delli Carpini, Cook & Jacobs, 2004;Fishkin & Luskin, 1999;Rosenberg, 2007).As a crucial predictor of political participation, efficacy is pivotal for striving democracies and thus a significant concept in respective theories (Conway, 2000;Pateman, 1970).
Theorists have claimed for a long time that participation in deliberative procedures would improve citizens' political efficacy.Empirically, the results are mixed.Recent works have shown that deliberative procedures can affect political efficacy positively, negatively or not at all.Some studies detected an increase of political efficacy in deliberative procedures (Fishkin, 1995;Grönlund, Setälä, & Herne, 2010;Nabatchi, 2007), but Morrell (2005) and others noticed little or no impacts (e.g.Gastil, 1999;Morrell, 1998;Stromer-Galley & Muhlberger, 2009;Walsh, 2003).A few scholars even noticed decreased efficacy when people are confronted with disagreement (e.g.Mutz, 2008).Obviously, impacts of deliberative procedures on efficacy depend on specific factors.
Accordingly, scholars of deliberation have stressed the need to examine which factors influence the improvement of political efficacy (e.g.Geissel, 2009;Mutz, 2008;Thompson, 2008).Up to now research mainly focused on deliberative experiments or single events (Fishkin, 1995;Fung, 2004;Fung & Wright, 2003;Gastil et al., 2010;Grönlund, Setälä, & Herne, 2010;Gutmann & Thompson, 1996;Knobloch & Gastil, 2015;Nabatchi, 2007;Smith, 2009).Hardly any large-n studies have been conducted on variables influencing efficacy in real-life deliberative procedures, and generalizable results are missing altogether.This article will fill this gap and answer the following question: Under which conditions do real-life deliberative procedures enhance political efficacy?
Methodologically, we address the lack of (generalizable) findings by applying a rather novel methodthe accumulation of "the intellectual gold of case study research" (Jensen & Rodgers, 2001, p. 235; see also Smith et al., 2015).Recently scholars have tried to summarize case study findings in narrative synopses (e.g.Delli Carpini, Cook, & Jacobs, 2004, p. 200;Goodin & Dryzek, 2006).In contrast, we intend to generate a quantitative large-n synopsis allowing for statistical calculation.By aggregating and integrating a large number of case study findings, we aim to test hypotheses and identify generalizable results (Borroso et al., 2003, p. 154).
The aggregation of case study information is methodologically demanding.Conventional meta-analyses take into account so-called effect sizes such as mean values, correlation coefficients, or standard deviation values.Such effect sizesand other quantitative data-are seldom available in case studies on deliberative procedures (see "Data and methods" section), and therefore a conventional metaanalytical approach cannot be applied.A different approach is necessary and attempted in this article, i.e. an approach called quantitative meta-synthesis1 .
Within meta-synthesis, researchers do not gather the primary data themselves, but rely on mostly narrative information provided by case studies.Since abundant case studies on deliberative procedures are available, it is worth testing whether the findings of these case studies can be aggregated and integrated in a large-n data set.Accordingly, this article also addresses a methodological question: Is a metasynthesis an appropriate approach for the accumulation of narrative case studies findings on political efficacy in deliberative procedures?
Via a quantitative meta-synthesis, our study identifies variables that foster the improvement of efficacy in real-life deliberative procedures at the local level in Germany.The Federal Republic of Germany is an especially interesting case for scrutinizing local deliberative procedures.Since unification, the local level is becoming a particularly dynamic field for participatory approaches.Germany's local representative democracy is increasingly complemented with deliberative procedures, and an abundance of case studies is available.
We focus on Participatory Budgeting (PB) and Local Agenda 21 (LA 21) because of their comparability.LA 21 and PB procedures share several similar features. 2They are both consultative procedures with the purpose of influencing policies.Both apply similar formats with an information phase, a discussion phase, and a phase in which the proposals are finalized and handed over to local decision-making bodies, and both miss decision-making competence. 3Until now in Germany, more than 20 percent of all municipalities have decided to conduct an LA 21 procedure (see website: http://www.agenda21-treffpunkt.de), and in more than 130 municipalities PB procedures have been implemented (see website: http://www.buergerhaushalt.org). 4he article is structured in the following way: In the next chapter, we introduce the state of the art on political efficacy in deliberative procedures before developing the hypotheses.Next, the method and the data used to test the hypotheses are described.The findings of the meta-synthesis are then presented.The article concludes with a summary of the study's research contributions and directions for future research.
The term 'deliberative' is applied in this article according to generally accepted definitions.Nabatchi et al. (2012, p. 6f) define 'deliberative' as "a process characterized by deliberation, or the thoughtful and reasoned consideration of views, experiences, and ideas among a group of individuals."It is a particular type of communication, in which participants discuss political topics from different perspectives and agree on well-thought-out solutions (similar to Smith, 2009).Accordingly, we use the terms deliberative, deliberation, or deliberative procedure to describe participatory procedures based on dialogue aimed at achieving problemsolving suggestions (for the theoretical debate, see e.g.Kahane, Loptson, Herriman & Hardy, 2013;Parkinson & Mansbridge, 2013;Talpin, 2013).

Political Efficacy in Deliberative Procedures:
State of the Art and Hypotheses Campbell, Gurin, and Miller (1954) conceptualized political efficacy as the "feeling that individual political action does have, or can have, an impact upon the political process, namely, that it is worthwhile to perform one's civic duties" (Campbell,  Gurin, & Miller, 1954, p. 187). 5In this article we examine group-related efficacy, which is defined in the literature in two different ways.The first definition, which we apply in our study, specifies group-related efficacy as the aggregated political efficacy of participants (Bandura, 2000).The second concept of group-related efficacy as collective political efficacy refers to the perception of the empowerment of a group and is mainly applied in research on mobilization and collective action. 6e are interested in aggregated efficacy, not in collective efficacy.
The concept of aggregated efficacy is not novel in research on deliberation.Aggregated efficacy and individual efficacy are often not clearly differentiated.
Efficacy is considered as a political attitude, which exists, deteriorates, or improves within an individual, a group, or society more or less simultaneously and synchronously.For example, Nabatchi (2010) showed the increase of political efficacy within the AmericaSpeaks 21 st Century Town Meeting by comparing the sample of participants with a sample of non-participants, i.e. looking at grouprelated efficacy as aggregation of individuals' efficacy.
Although the research community on deliberation has attached crucial importance to efficacy, systematizing studies on fostering factors are rare.The few quantitative studies on deliberative procedures examined neither the potential improvement of political efficacy nor the factors influencing this improvement (Beierle & Cayford, 2002;Newig et al., 2013;Ryan, 2014;Ryan & Smith, 2012;Spada, 2010).This gap will be addressed with the quantitative meta-synthesis in this article.
The hypotheses were developed in an iterative way, oscillating between theoretical assumptions, case study findings and data availability (e.g., Font, Pasadas del Amo & Smith, 2016).An abundance of possible hypotheses can be found within publications on deliberation.However, in the process of data gathering, we realized that due to the lack of data availability, not all hypotheses could be tested.For example, information about the content of participatory proposals was rare.Accordingly, respective hypotheses and variables had to be eliminated from our study.
In the literature on factors explaining the impacts of deliberative procedures on political efficacy, three branches of explanations can be identified (Font & Navarro, 2013;Font, Pasadas del Amo & Smith, 2016), and our study is structured along these lines: 1) Actors, e.g. the support of local politicians for a deliberative procedure; 2) Participatory commitment, e.g. the implementation of special staff supporting citizens' involvement; and 3) Social-political and economic context, e.g. the size of municipality.
Another branch of explanation examines the effects of different types of deliberative procedure (e.g.Font & Navarro, 2013).However, in our metasynthesis, similar types are examined, and, accordingly, this variable is a constant.

Actors
The support of local politicians for a deliberative procedure ('event-limited support') seems to enhance the probability of its success.For example, examining a participatory process centered on health-policy issues in Canada, Barrett et al. (2012, p. 199) emphasized the attendance of decision-makers as the crucial factor for an effective deliberative procedure.Similarly, Ryan (2014, p. 71) had shown that the success of PB procedures in Great Britain depended mainly on the support of the mayor (for Germany : Oels, 2003 28, p. 243f;for Brazil: Wampler, 2007, p. 258).Accordingly, we can develop the following hypothesis for the relationship between the dependent variable (efficacy) and the independent variable (politicians' support and participation): H1: When local politicians support or participate in a deliberative procedure, improvement of political efficacy is more likely.

Participatory Commitment
Some studies on deliberative procedures highlight the relevance of the institutionalized, long-term financial and structural commitments for participatory decision-making of a municipality.Such a commitment seems to have positive impacts on the success of deliberative procedures (Font, Pasadas del Amo, & Smith, 2016;Röcke, 2014).Structural commitments include the establishment of special staff in charge of implementing citizens' involvement and a participatory plan (local constitution).A participatory plan covers criteria and objectives of municipal participation as well as rights and obligations of all actors involved (politicians, citizens, administration).Such institutionalized participatory commitments differ from ad-hoc support by local politicians, because they function independently from particular politicians.Hence, the following hypotheses are stated: H2: When a municipality has implemented a participatory plan, improvement of political efficacy is more likely.
H3: When a municipality has established special staff for citizens' involvement, improvement of political efficacy is more likely.
Additionally, it can be assumed intuitively that the former implementation of participatory procedures affects the impact of subsequent participatory processes in a municipality (Barrett et al., 2012, p. 189;Beaumont, 2011;Schugurensky & Myers, 2008, p. 92).The culture of a municipality might change when the citizenry has experiences with participatory procedures.
H4: When a municipality has experiences with participatory procedures, improvement of political efficacy is more likely.

Socio-Political and Economic Context
Several additional explanatory variables are discussed and have to be taken into account, mainly the size of population, economic strength and the ideology of the party in government (socio-political and economic background of the municipality) (Bryan, 2004;Sintomer, Röcke, & Herzberg, 2016, p. 176ff).7 Participation most likely leads to larger impacts in smaller municipalities (Bryan, 2004;Dahl & Tufte, 1973, p. 43;Denters et al., 2014, p. 152) and in municipalities with economic strength (Boulding & Wampler, 2009).In municipalities with few inhabitants, deliberative procedures seem to be easier to organize and it is "easier for participants to hold the public authority to account for failure to implement proposals" (Font, Smith, Galais & Alarcon, 2016, p. 9).In municipalities with larger populations, it might be more difficult to influence local politics because more citizens and interest groups compete for influence.Economically strong municipalities can invest in more well-conducted procedures and implement expensive suggestions, which might lead to an improvement of efficacy.Contrarily, economically weak municipalities can only implement cost-cutting suggestions, which might have negative impacts on political efficacy.
Traditionally, left-wing parties have emphasized more strongly the virtues of participation.Some of them have given participation a central role in their political program (Cohen & Fung, 2004).There is a huge research literature highlighting links between left-wing local government and the rise of participatory procedures (Sintomer, Herzberg & Röcke, 2008).But, as Galais and Font (2011, p. 10) pointed out, "It is not clear whether all the left [party] families share this same priority."

Data and Methods
Our research is based on the conviction that systemizing research is needed, which allows for generalization and hypothesis-testing.8Therefore, we systemized case study findings through a quantitative meta-synthesis.Up to now, statistical analyses of case study findings on deliberative procedures have been very rare (exception: Beierle & Cayford, 2002).Recently, one study adopted a meta-synthetical approach (also Newig et al., 2013).Smith et al. (2015) examined case data available on the participaedia.netplatform, applying statistical methods.However, they followed a different research question, looking at the impacts of procedural institutional designs.In our study, we examine effects of other explaining variables, i.e. actors as well as commitment and context, and keep the institutional design of the procedure as constant.Furthermore, Smith et al.'s (2015) research relies on documentation reported online by practitioners and scholars, while our study is mainly based on data provided by academic case studies.

Meta-synthesis: From Case Study to Quantitative Dataset
The transformation of qualitative case study findings into a quantitative, large-n meta-synthesis data set requires three complex steps, each of which is confronted with several challenges and potential distortions: (1) Search for case studies and representativeness (2) Quantification and coding of case study data (3) Strategies to cope with missing data

Search for case studies and representativeness
The first potential challenge refers to the availability and the selection of unbiased case studies.To collect all case studies on PB and LA 21 in Germany, we adopted two search strategies: First, several databases were checked using various keywords.The aim was to find all studies on PB and LA 21 published between 1992 and 2015. 9Second, relevant networks, institutions, and stakeholders in the field of participatory procedures were contacted. 10We were able to gather more than 300 case studies, and this entire 'universe of case studies' was included in the meta-synthesis.
Various potential biases within the 'universe of case studies' could be encountered: Often it is assumed that only studies on successful cases are published, which would imply distorted findings.A rigorous screening of the 'universe of case studies' revealed that this critique is unjustified.Case studies mostly start at the beginning of the procedure and at that point nobody knows, whether this procedure will be successful or not.Furthermore, case study authors pursue different definitions of success, which makes a respective bias even more unlikely (similar: Beierle & Cayford, 2002).
Other potential distortions might occur due to the commission by a municipality or due to the Hawthorne effect.About 30 percent of the case studies explored in our meta-synthesis were commissioned, but biased information was not detected in these publications.The Hawthorne effect implies that participants in an observed participatory procedure are aware of being studied and act differently than normal.However, within our 'universe of case studies,' field work conducted via participatory observation and ex post studies provided similar results.Finally, there might be a distortion, because case studies might often be conducted in university cities.We also did not notice any respective distortion in our data.
Since there are no official records on LA 21 and PB procedures, we can neither say with certainty that the cases covered by the case studies are representative nor can it be verified that they are not (similarly, see Font, Pasadas del Amo, & Smith, 2016, p. 3).Nevertheless, some information is available, which allows a tentative check on the representativeness: The distribution of LA 21 and PB procedures among German states within our data set reflects the general distribution: Nordrhein-Westfalen, Hesse, and Saarland are the most active states, and the new East German states are less enthusiastic (Ruschkowski, 2002).It is also assumed that most of the LA 21 procedures take place in mid-size (between 200,000 and 500,000 inhabitants) municipalities (Brunold, 2004) and our data set reflects this pattern, as well.

Quantification and coding of case study data
The second challenge is the standardization and quantification of mostly narrative information in order to accumulate it.Case studies apply different operationalization, and the respective information has to be transformed and converted to a common metric system through standardization.We developed two possible ways of standardization: The first way refers to a breaking down of mostly narrative information into numerical indicators.For example, in one case study, efficacy was measured through surveying participants at the beginning and the end of the procedure and numerical information about the improvement was provided.
Yet, an approach is necessary which transforms narrative information into standardized numerical data.This transformation is mastered by coders.Coders code the information by assigning a numerical evaluation.Based on their knowledge of dozens of cases, they decide on the code.For example, a case study reports that only a few of the participants expressed satisfaction with the procedure.The coder would then assign a 1 for this variable.
Most variables were coded on an ordinal five-point scale.Of course, a five-point scale cannot reflect the diverse and complex reality, but since we aim to create a manageable data structure in order to run quantitative analysis tools, the complexity had to be reduced (similarly, see Font, Pasadas del Amo, & Smith, 2016).
To enhance reliability, each case study was coded by at least two coders, and intercoder reliability as an important indicator of coding quality was tested continuously (similarly, see Newig et al., 2013).Intercoder reliability is measured by Krippendorff's Alpha.In the meta-synthesis presented in this article, intercoderreliability is .86.11 Additionally, information was judged by the coders to ensure sufficient validity.They checked, for example, whether convincing evidence verifies the information (data, description of examples).For the calculation run in this article, only information judged as valid was included.

Strategies to cope with missing data
Often case studies do not contain all information required to test all hypotheses.
From 300 case studies in our data set, 180 case studies provided sufficient information on political efficacy.Strategies to cope with missing data in a quantitative meta-synthesis of case study findings differ from strategies in traditional meta-analyses.Whereas in traditional meta-analyses an imputation of incomplete data is possible, predominantly by statistical estimation techniques, imputing data is not a feasible strategy for meta-syntheses.
Only two strategies are possible, if insufficient data is available.Either the variable is excluded and the corresponding hypothesis cannot be tested, or additional data is generated.Additional information can be collected by relying on experts' assessments, and we conducted interviews with case studies' authors, administrative staff, politicians, citizens, and moderators.These experts evaluated the procedure based on their in-depth knowledge of the specific case.
Assessments of experts are widely used and accepted within social sciences as a useful, reliable source for gaining information, but are also criticized as being subjective.To guarantee intersubjectivity (see, for instance, Coppedge et al., 2011, on measuring democracy), two experts were interviewed for each procedure.
Finally, data was collected by exploring additional documents, e.g.administrative minutes and proceedings.

Operationalization: Dependent Variable
In this meta-synthesis, we search for factors which increase the probability of successful impacts on aggregated political efficacy (dependent variable).12Since none of the case studies explored in our meta-synthesis provides individual-level data on political efficacy, we refer to aggregated efficacy.In most case studies, efficacy was measured by observing the deliberative procedure and attaining a general impression.The unit of observation was the group of participants taking part in the deliberative procedure.
These observations are coded as an ordinal five-category variable.Whereas value 0 indicates no improvement, value 1 identifies slight improvement and value 2 stands for moderate improvement.Value 3 indicates strong improvement.Finally, value 4 stands for a very strong improvement of participants' political efficacy.
Since most data was qualitative, it was not possible to formulate strict thresholds for assigning the codes.

Operationalization: Independent Variables
Referring to the above-mentioned discussion of explaining factors (independent variables), we focused on three basic types of independent variables: those related to actors (support by local politicians) and to institutionalized participatory commitment as well as additional variables on socio-political and economic contexts.
Information on support or participation of mayor or city council members were all dichotomous (Did mayor / city council members participate / support the procedure, with scale points 0=not existing; 1=existing).The institutionalized participatory commitment is measured by the existence of a participatory plan and special staff for citizens' involvement.Both variables are dichotomous (Does municipality have a participatory plan?Is there special staff for citizens' involvement?, with scale points 0=not existing; 1=existing).Former participatory procedures is a dichotomous variable (Were there deliberative procedures before?, with scale points 0=no former participatory procedures; 1=former participatory procedures).
The socio-economic background is recorded by two factors: The size of municipality is measured by the number of inhabitants; the level of municipal economic strength is a numeric variable calculated from municipal revenues per capita minus municipal government debts per capita.
Aspects of the socio-political municipal background are captured by the ideology of the party in government, which is a dichotomous variable (Which parties have the relative majority in council?, with scale points 0=right wing; 1=left wing).There are no intercorrelations between the independent variables (see appendix, Table 3, Descriptive overview of dependent and independent variables).

Steps of Analysis
In the first step of the analysis, bivariate cross tables and calculated effect sizes were produced to investigate which decisive factors led to impacts of deliberative procedures on political efficacy.In the second step, the ordinal regression (PLUM) function in SPSS was applied.Constructing ordinal regression models requires several decisions.We identified the ordinal dependent variable and then we decided which predictors to use.Finally, we chose the type of link function that provides a good fit for the data.

Results: Under which conditions do real-life deliberative procedures enhance political efficacy?
The results of effect size calculation (see Table 1) indicate a statistically significant, medium positive correlation between participatory plan as well as special staff with political efficacy and a statistically significant, weak positive correlation between mayor participation and support, former participatory procedures, ideology, municipal economic strength with political efficacy.Yet, city council participation and support show a statistically significant, weak negative correlation.Municipal size correlates very weakly with political efficacy, and this effect is not significant.We then calculated different regression models using the cauchit link function, which allows for non-standardized normal distributions.This approach was considered to be the most appropriate given the non-normal distribution of outcome variable data (Cohen et al., 2006).Within all models, this link function was the better choice because of its satisfying 'parallel lines' assumption and larger model fitting statistics.Table 2 presents the results of ordinal regression models with political efficacy as dependent variable.We are aware that strong causal inferences are not warranted because our metasynthesis is not based on randomized experiments.Since the representativeness of the case examined in the case studies is unknown, we applied the statistical method of bootstrapping, allowing for testing the robustness of statistical results (see appendix, table 4: Replication of statistical analysis in table 2 using the bootstrapping method).The bootstrapping calculation confirms our findings.Accordingly, modest assertions about estimated associations between variables provided in table 2 can be adopted (see for information on causal relationships, e.g.Rosenthal & Rosnow, 2008).
As expected from the effect-size calculation, it seems to be more likely that deliberative procedures have impacts on political efficacy in municipalities with participatory plan, special staff, a participatory history, and economic strength.
For event-limited mayoral and city council participation and support, the results are inconsistent.The effect-size calculation suggests that it does matter whether mayors or city council members participate in and respectively support deliberative procedures. 13However, the ordinal regression results considering the support and participation of local politicians are not statistically significant.
The size of a municipality has a negative co-efficiency, but is not significant: As municipal size decreases, so does the probability of impacts on political efficacy.Right-wing party ideology has an insignificant negative effect as well.

Discussion and Conclusion
Under which conditions do real-life deliberative procedures enhance political efficacy?Our research implies that deliberative procedures can lead to a greater sense of political efficacyunder certain circumstances.The strongest explanatory power is resting on the institutionalized, comprehensive financial and structural commitments of a government for participatory procedures and participatory decision-making.In short, political efficacy increases if governments make real efforts.This result is connected to the second finding: Political efficacy increases in municipalities with wide-ranging participatory experiences more often than in municipalities without participatory history.Participatory procedures are no fasttrack aid, but their potential requires time to unfold.
Socio-political and economic contexts, such as municipal economic strength, municipal size, and city council ideology, have little or no explanatory power.Also event-limited support by local politicians is less relevant.
Our findings are consistent with recent research demonstrating the importance of a government's participatory commitments for successful participatory procedures.In contrast, former findings about support by mayors or city council members are disconfirmed by our results: Whereas some studies had shown the importance of support by local politicians (for UK: Ryan, 2014;for Brazil: Wampler, 2007), our large-n meta-synthesis demonstrates that the probability of impacts on efficacy is not enhanced by politicians' support and participation.Institutionalized commitment is more vital.This discrepancy between our findings and results in former studies might be due to the fact that up to now institutionalized commitment has hardly been taken into account in empirical research on efficacy in deliberative procedures.
Although our study has certain limits, since it focused on LA 21 and PB in Germany, the findings provide clear implications for the real world of politics.The results indicate that municipal governments should think twice why and whether they want to establish a deliberative procedure such as PB or LA 21 and invest adequate resources.If a government aims at conducting a successful deliberative procedure, it should allocate adequate funds and infrastructure.Special staff for citizens' participation and a local participatory plan are crucial for the success.If the local government plans a 'low-budget' procedure without financial and institutional commitment, the endeavor will most likely not produce any effects.
The following suggestions for further research concern the methodology used in this study.Is a meta-synthesis an appropriate approach for the accumulation of case study findings on political efficacy in real-life deliberative procedures?Yes, but only under certain circumstances and only to a certain degree.An abundance of high-quality, scientifically valid studies must be available on similar cases, and these case studies must provide a large amount of valid and reliable information.Besides these already demanding requirements, scholars must deal with several challenges.
Considering data collection, the main challenge refers to the representativeness of cases and case studies.Examining deliberative procedures, it is hardly possible to prove that the cases studies are representative, because there is no comprehensive data set on the 'universe of cases.'If a study aims at providing an exhaustive overview of all procedures, additional sources and data-collecting tools are necessary.
Once case studies are collected, the crucial challenge is the quantification of narrative information.Quantification requires the breaking down of information into numerical indicators, e.g. the percentage of participants whose efficacy has been enhanced during the deliberative procedure.In most case studies, however, numerical data is rare, and, accordingly, the available narrative information must be transformed.This transformation is to be conducted by experienced coders, who assess the information based on their knowledge and assign a code.To increase reliability, at least two coders assessing the case studies information independently are required (intercoder-reliability).
The information provided by available case studies will often not suffice.Only two strategies are possible if this problem occurs.Either the respective variable is excluded and the corresponding hypothesis cannot be tested or additional information must be collected, e.g.via document analysis or expert interviews.An additional pitfall needs to be mentioned in this context, i.e. the consequences of not including all potentially explanatory variables due to the lack of data ('omitted variable bias').However, this potential bias is widespread in new fields of research such as studies on deliberative procedures and not specific for a meta-synthesis.
We conclude that the meta-synthesis of case study findings meets the requirements of a useful, feasible, reliable and appropriate method, if scholars consider all potential flaws.14A methodological alternative is conducting a large-n study by generating primary data via observation of a large number of deliberative procedures or exploring existing material, e.g.minutes and proceedings, without referring to case study information (Font, Pasadas del Amo, & Smith, 2016).This approach is feasible for ongoing cases or cases providing sufficient text material.
We suggest a mix of different methods.If reliable case studies are available, it saves time and resources to start with the exploitation of case studies and to create a numerical data set.Data collection based on other sources can then be inserted into this data set.In contrast to the conventional approach of meta-analysis, the advantage of our approach is the aggregation of narrative information from qualitative case studies.This approach is challenging, but provides a variety of benefits.

Table 4 . Replication of statistical analysis in table 2 using bootstrapping method
Source: provided by the authors