Case Study in Public Administration: a Critical Review of Brazilian Scientific Production Introduction

This paper presents a critical review of 47 articles published between 2006 and 2011 to identify how case studies have been applied in Brazilian research on public administration. In addition to their theoretical and methodological characteristics, four further specific topics of interest were addressed: (a) what is meant by case study; (b) the relationship between the phenomenon of interest and the case under investigation; (c) the possibility of replication; and (d) how the supposed method contributes towards the development of the field of public administration. The main inconsistencies found were: the methodological descriptions are confusing; the results are inconsistent compared with data gathering procedures and data analysis techniques; a lack of information about the number of interviewed individuals; and no descriptions of research variables. The results suggest the reviewed case studies present methodological inconsistencies and limitations, which undermine their scientific value and relevance to academic work in Brazil.


Introduction
There are several reports of failures, difficulties, problems and/or dichotomies in public administration studies as to how research questions are defined or what supports their originality and usefulness, or in the theorization process itself.Bertero, Caldas and Wood (1999) analyzed the quality of scientific research of public administration in Brazil and found results that indicated "peripheral, epistemologically flawed, methodologically deficient, unoriginal, consisting to a large extent of illinformed mimicry" (p.148).
In social sciences, the debate about ontological and epistemological aspects of research is not new.However, in organizational studies, this matter has been gaining space since the 1980s, with arguments that contribute to the proliferation of themes and research methodologies (Campos & Costa, 2007).Bertero et al. (1999) highlight that this type of debate is not exclusive to Brazil.In the United States, the Academy of Management Review (Whetten, 1989) published a discussion about what constitutes a theoretical contribution?and the Administrative Science Quarterly (Sutton & Staw, 1995) dedicated an entire issue to the question what is not theory?Tonelli, Caldas, Lacombe and Tinoco (2003) analyzed the main human resources scientific journals published in Brazil during the 1990s.They analyzed the subject matter, the epistemological basis, the methodological guidance and the demographic authorship.In their conclusions, the authors showed that academic production has risen significantly in volume, but the scientific profile of research generates concerns.According to the authors, in the discussion of methodological basis, the reviewed publications had many weaknesses and case studies were used just to illustrate the theory.In another words, case studies were not used to induce or to create new theory.This occurs because many researchers use case study but provide few details about methodological choice, limiting themselves and their research to short descriptions of the phenomenon being studied.These shortcomings contribute to build fragile theoretical models and limit the easy understanding of phenomena of interest, which is one of main objectives of case studies (Voss, Tsikriktsis, & Frohlich, 2002).
Scientific research conducted with consistent theoretical models and based on clear and concise methods generates important benefits to the public administration field.Thus, researchers can help to explain how the results of their investigations were produced, how robust their discoveries are and how the results can be extrapolated (Prowse, 2010).In this context, this present research attempts to identify how case study has been applied in Brazilian research of public administration by identifying and reviewing empirical and qualitative studies.The term method was used between quotations to inquire as to whether case study is a valid scientific procedure, as discussed later in the text.This present study is important because it questions methodological guidelines presented in Brazilian research on public administration, in particular about the use of case studies as a research method.This paper highlights the predominant use of the case study method, but does not discuss methodological procedures that were developed to obtain the research findings in specific studies.

Academic Production in Public Administration
"Is it still worthwhile studying the public sector and public administration through the lens of theory and modern organizational analysis?"this is the question put forth by Thoenig (2007, p. 10), for whom the production of knowledge in relation to organizations appears to be in a state of paralysis, when considering the public administration as a specific field of study.In addition, analyses made of studies published in this area suggest that, in most cases, public organizations are used merely as an empirical illustration during an investigation of more general aspects of organizational theory.
Various reviews of articles written on the subject of public administration, published both in Brazil and elsewhere (e.g., Gill & Meier, 2000;Kettl, 2000;Ongaro, 2008), show the many ambiguities that distinguish this area of knowledge, which result, according to the author, from the interaction between the subject matter, approach and theoretical references used in those articles.The main reviews published on this subject in Brazil are: Fischer (1984Fischer ( , 1993)); Mezzomo and Laporta (1994); Mezzomo and Vaz (1994) and Souza (1998).
According to Kettl (2000), among the main weaknesses of public administration research, it is necessary to underline the predominant use of methodologies that have not been properly elaborated or that are inadequate for the subject matter being studied.Boyne (2004), in turn, also questions the lack of rigor in the description of the way issues related to the services provided by public administration have been evaluated, both in theory and in practice.In other published articles, he also discusses the importance of quantitative methods for publications in this area (Boyne, 2002), and the relevance of the issues and the goals of academic research in public administration (Boyne, 2003).Roesch (2003), writing about the scientific output on administration in Brazil, and suggesting better methodological guidelines, states that research methods and techniques developed abroad are available to be used, tested or changed to suit the situation in Brazil.However, according to her, few researchers use the most up-to-date methodology to analyze empirical research "because it is seen as being very complicated or because it requires a very large data base for which it is claimed that there is no available funding" (p.166).On the other hand, the potential of qualitative research is never adequately explored, which is notable in research developed by case studieswhere an analysis of empirical material is, normally, "weak, masked by prescriptive tones, and where there is little abstraction effort on the part of the researcher either for the purpose of practical generalization, or for creating a local theory" (Roesch, 2003, p. 166).
In a similar vein, those critical of publications in the area of Brazilian public administration refer principally to the methodological dimensions of research.In addition to the aforementioned works, when Fleury, Bloch, Bravo, and Bicudo (2003) analyzed articles published in the Public Administration Magazine (Revista de Administração Pública [RAP]), between 1992 and 2002, he observed that case study was a research strategy used in 32% of the articles reviewed.Findings discovered by Hocayenda-Silva, Rossoni and Ferreira (2008) confirm this observation.Gondim et al. (2005) analyzed 158 articles published between 2001 and 2004 in Brazilian journals in the area of administration and in Annual Meetings of the Brazilian Association of Postgraduate Studies and Research in Administration (EnANPAD).Findings show that case studies are presented in a superficial way and that theoretical construction is inadequate.

Case Study
In qualitative methodology there is a range of techniques available to collect data, including: open or closed interviews, focal groups, ethnography and participative and non-participative observation (Flick, 2009).Some of the available data analysis techniques are content analysis (Bardin, 1977), hermeneutic analysis (Honer, 2004), discourse analysis, amongst others.Creswell (1998) states that there are four basic levels of qualitative research: case studies, comparative studies, retrospective studies and longitudinal studies.
The main objective of a case study consists in providing the most exact description possible of a case or its reconstruction (Creswell, 1998).The first reference to a case study as a form of investigative approach is attributed to Malinowski (1884Malinowski ( -1942)), in the United States, and Le Play (1806-1882), in France.Both stressed the importance of studying small communities to understand the common standards of behavior in the wider society.Another historical reference is the Chicago School: considered to be the first in a tradition of qualitative research in sociology towards the end of the 19th century.The Chicago School went to carry out investigative research based not only on documental and statistical analysis, but also on personal and interactive contact with people within the community.
A case study involves carrying out a qualitative study using different sources of evidence (Leonard- Barton, 1990), analyzing phenomena properties by means of a selected case (Campbell & Stanley, 1963) and gaining knowledge of a wider universe of similar units (Gerring, 2004).Another concept of a case study is the way in which empirical social research is carried out when a phenomenon is investigated within its real-life context, where boundaries between a phenomenon and a context are not clearly defined and in a situation where multiple sources of evidence are used (Yin, 2009).Moreover, in a case study, emphasis is given to provide a complete and detailed description as well as an understanding of the relationship between the factors involved in each situation, irrespective of the numbers involved (Boyd, Westfall, & Stasch, 1989).
In the literature, there are many different references to case study used as a research method.However, in the present study, case study is understood to be a qualitative study, in the sense of an investigation where sampling is made by selecting certain cases, bearing in mind that these make it possible for a researcher to collect information related to the phenomenon, in order to understand it in its entirety and in the context being studied.In other words, that which is usually referred to, in the literature, as a case study method is nothing more than a restricted definition of a qualitative research method.
Based on Yin (2009), a case study can be appropriate to research actual events and when the phenomenon is of an exceptional nature, as well as presenting numerous variables where the occurrence of instability is very frequent.It is also adequate when the research involves a contemporary phenomenon, where the boundaries between the phenomenon and its context are not clearly established.According to Voss, Tsikriktsis and Frohlich (2002), a case study is a unit of research analysis.Furthermore, it is possible to use different cases for the same study of the same phenomenon so as to analyze different factors.Thus, a case study uses cases as a basis for analysis.
The positivist approach, advocated, for example, by Yin, predicts that the research process is linear, ordered, carefully conceived, executed without flaws or errors and independent of the reader.The positivist approach is based on realist ontology and objectivist epistemology, which presupposes the existence of an objective reality independent of human perception and that reality can be understood through rational research methods.In a different paradigm, the interpretive approach of case studies advocates the need for a full description of the case and the phenomenon being investigated.
To Godoy (2006), case studies based on the interpretive approach stand out by offering opportunities to study phenomena involving mankind and its intricate social relations in different contexts.For this approach, the reality occurs within historical contexts and is socially constructed and, therefore, social and human research should not seek only the path of measurement, but also understanding (Guba & Lincoln, 1994).Interpretive ontology is a subject-object interaction, i.e., it does not consider the existence of a totally objective reality; and its epistemology is constructivist, assuming that all knowledge about reality depends on human practices, constructed through the interaction between individuals and transmitted by a social context (Hartley, 1999;Stake, 2000).
As it might be expected, there are controversies in the understanding of the generalizability of the findings in the positivist and interpretive approaches.In positive research, generalization occurs usually by statistical proceedings, i.e., extrapolating the results for one sample onto an entire population.On the other hand, in the interpretive approach, this ideal of generalization is not sought; rather, the intent is to understand the deep structure of a phenomenon which, it is believed, can be used to inform another environment (Orlikowski & Baroudi, 1991).
In general, studies carried out in Brazil show that the main definitions refer to Yin's work, in their different published editions.It is important to highlight the fact that, in the different versions of this work, Yin has been credited with the operational definition which states that a case study should necessarily be developed through an interview or questionnaire, observation and documental analysis.
However, this does not appear to be the definition as presented in Yin's work.Whether discussing aspects of sample selection, or how and when cases should be generalized, Yin (2009) considers that, to validate research in general, such as research undertaken by means of case studies, it is necessary to have data triangulation, obtained through different techniques.
The proposal outlined by Yin ( 2209) is that a case study chosen for its phenomenon representativeness and the case selected would be the locus of the analysis of the phenomenon, since the relative sampling size does not give validity to a study.Special attention should be given to establishing criteria which are validated as being sufficient to address the phenomenon being studied.However, as Tonelli et al. (2003) argues, research which opts for case studies should present in detail the methodological decisions to support such a choice.In other words, a case study should be selected as a means to select the phenomenon of interest, and not just the locus where such a phenomenon occurs.
The case study is widely used in Brazilian research work in various areas of public administration, and is criticized for its lack of methodological rigor.In spite of these criticisms, some authors (e.g., Benbasat, Goldstein, & Mead, 1987;Bonoma, 1985;Burns, 1990;Eisenhardt, 1989;Gondim et al., 2005;Leonard-Barton, 1990) defend the development of case studies, believing it to be suitable for describing the research subject in depth; carrying out a broad range of observations using various data gathering sources; reconstructing the object of research by relating the parts to the whole; and constructing theories and improving abstract terms and concepts, prepared from a description of the case study based on concrete evidence.
Analyzing evidence is one of the most difficult steps when conducting a case study (Yin, 2009).In most cases, the researcher begins study without having a very clear view of the evidence to be analyzed and may experience difficulties when carrying out this stage of the research.Scandura and Williams (2000), in turn, assert that there is evidence that researchers are being trained to focus far more on applied case studies, rather than on laboratory studiesa fact that concerns the above-mentioned authors, especially in terms of validity of the knowledge being produced.Those authors indicate triangulation, which refers to the multiplicity of theoretical-methodological approaches (external triangulation) and a combination of different data gathering procedures and data analysis techniques (internal triangulation).The objective is to balance the strengths and weaknesses of different patterns of research and produce a more consistent and profound knowledge about social and complex organizational phenomena.

Method
To achieve the proposed objective, which is to identify how case studies have been applied in Brazilian academic research on public administration, an analysis was made of the research work that has been published on the subject, from the date when a common understanding was established about what is meant by a case study: the operational definition ascribed to Yin (2009), or in other publications of his work.
The search for the study of interest was restricted to publications in the period 2006-2011, in major journals related to the subject of administration in Brazil, according to CAPES, the agency responsible for evaluation and qualification of Brazilian journals, namely: Brazilian Administration Review (BAR), Brazilian Business Review (BBR), Gestão & Produção, Organização & Sociedade, Revista de Administração Contemporânea (RAC), Revista de Administração de Empresas (RAE), Revista de Administração Mackenzie (RAM), Revista de Administração Pública (RAP), Revista de Administração da USP (RAUSP), and Cadernos EBAPE.BR.
In the process of sifting through this material, initially all articles that included the word case in their title or abstract were selected, with the corresponding term in Portuguese (caso).Among the articles first selected, only those that had predominantly qualitative empirical research as their objective were considered useful.In total, 47 articles were selected from five journals: Revista de Administração Pública (RAP), Cadernos EBAPE.BR, Revista de Administração de Empresas (RAE), Organização & Sociedade (O&S) and Revista de Administração Contemporânea (RAC).
All the articles selected were initially analyzed according to their contexts, making it possible to identify the period of publication, the journals that published the greater part of the material, as well as the institutions that had the highest output on the subject.Then, based on the outline of previous research on administration published in Brazil (Bertero, Caldas, & Wood, 1999;Bertero, Vasconcelos, & Binder, 2003;Calixto, 2009;Gondim et al., 2005;Hocayen-da-Silva, Rossoni, & Ferreira, 2008;Pacheco, 2003;Tonelli, Caldas, Lacombe, & Tinoco, 2003;Vergara & Peci, 2003), the following general analysis criteria were used: Despite the importance of the context and the general criteria, the central point of the analysis consists of the following four specific questions: (a) What was meant as case study?(b) What was the relationship between the phenomenon of interest and the case being investigated?(c) How successfully could the studies be replicated?(d) How did the so-called method contribute towards the development of public administration research?
Based on the first specific question, the publications were identified according to which definition(s) had been used.Thus, the aim was to seek an overall understanding of the meaning of case study to authors who have published works in the area of public administration.By means of a second specific question, an analysis was made of the reasonsthe theoretical and practical argumentsgiven by the authors as to why it was important to study that particular case.
In general, it has been observed that researchers consistently confirm that using the case study method makes the results of the findings impossible to generalize.This does not seem to be Yin's understanding of the matter, since he proposes that case selection should be made in accordance with the representativeness of the phenomenon, and not the number of cases.Thus, if the phenomenon can in fact be addressed by means of a selected case study, then it is possible to generalize the results of the findings with respect to the phenomenon, and not with respect to the case study itself.The question of its replicability was therefore addressed by means of this third issue.Finally, we have the specific question of how the authors justified using a case study as a means to develop research capabilities in the field of public administration.

Context of the publications
Most of the articles analyzed (70%) were published in two journals, both specialized in public administration: Revista de Administração Pública (RAP) and Cadernos EBAPE.BR.It is worth noting that almost 83% of the works that were published, including the six articles published in Revista de Administração de Empresas (RAE), were printed by three journals from the same institution, the Getulio Vargas Foundation (FGV).This does not mean that the criticisms raised in this study for the revised publications should be directed to the editorial boards of the aforementioned journals.While there are certainly weaknesses in the publishing processes, which may reflect low maturity on the part of reviewers and editors, paper quality depends, before anything else, on the effort expended by authors.The fact that some journals have few or even no articles analyzed in the research may be due to an evaluation process in which articles with certain characteristics have a high probability of being rejected and thus were left out of the sample.For example, single case studies, representing 94% of the research sample, may eventually be accepted for evaluation in a more prestigious journal; however, the study should address an important situation or organization, with interesting contributions to theoretical discussion or practice.
The findings showed that of all the research institutions that had ties with the authors of the reviewed studies, the University of São Paulo (USP) came in first place, with eleven authors; the Getulio Vargas Foundation (FGV) came second with ten; the Federal University of Minas Gerais (UFMG) came next with eight authors; both the Federal University of Bahia (UFBA) and the Pontifical Catholic University of Rio de Janeiro (PUC/RJ) produced seven authors each.In addition, 32 other research institutions were also represented in the article samplings, and two international institutions: the Technische Universität of Berlin and the London School of Economics and Political Science were each represented by one author.
The institutional origin of the authors who had written the articles reviewed in this sampling show that there was a moderate concentration in only a few of the institutions.Of the total of 106 authors involvedwith a total of 27 articles -36 of them (34%) are active in the five institutions mentioned previously -USP, FGV, UFMG, UFBA and PUC/RJ.Finally, it should be mentioned that only three articles were published in English: Albuquerque, Prado and Machado (2011), Saccol, Manica and Elaluf-Calderwood (2011), published in RAE;and Cohen and Silva (2010), published in RAP.

General characteristics of the studies
The general criteria used to classify the studies were: number of case studies, selection strategy used in each case, temporal cut-off, level of governmental research, data gathering procedure, data analysis and theoretical base technique.The findings in each of these criteria will be shown in the following paragraphs.Table 2 shows the quantitative and the percentage of articles classified based on those analysis criteria.Nearly all of the case studies analyzed (94%) had only one case study as their analysis focus.Of the 47 articles, only three were characterized as a multiple case study.The use of multiple cases, amongst other features, is important when one wishes to carry out a comparative study.It is also useful for the validity of the research, since it indicates the possibility of replication.In this sense, the predominance of single case studies in Brazilian literature on public administration may be an indication that the research carried out was not sufficiently reliable to suggest replication in different contexts.
Most of the studies reviewed (59%) can be classified as a convenience sampling case study.It is interesting to note that none of these studies specifically stated that the cases under investigation should be chosen by convenience.On the other hand, of the six studies (14%) where the selection strategy was by typicality, three of them indicated this as their choice.There were practically no other case selection strategies found among the reviewed studies.This seems to indicate a lack of maturity on the part of the Brazilian researchers, thus corroborating the opinions of Gondim et al. (2005), who claimed "we still have not overcome the challenge of making the methodological outline of a case study less dependent on the individual experience of the researcher" (p.66).Another explanation could be related to the availability of funding to carry out research.When choosing a case by conveniencemore immediate, more accessible, etc.a researcher is far more concerned with the practical and economic side of the research work rather than questions related to the relevance of the case from the point of view of its phenomenon of interest, which are of secondary concern.
The temporal cut-off in most of the cases reviewed (80%) was of a transversal nature, that is to say, the data collected refers to a picture of the phenomenon.In the other 20%, the temporal cut-off presents a longitudinal nature, in that seven articles are classified as partially longitudinal, that is to say, transversal, though of a longitudinal nature, while only two articles are classified as longitudinal.The fact that documental research that uses a historical perspective is a transversal study, albeit with a longitudinal nature, was taken into account.
The findings show that the phenomena of public administration that have been addressed are mainly static; this is contrary to the nature of these phenomena, the majority of which are long-term, such as, for example, findings resulting from changes to the management model and impacts on the organizational structure resulting from alterations to the Fiscal Responsibility Law.It is important to mention that some scholars (e.g., Flick, 2009) argue that if the temporal component in qualitative studies is taken into account, the validity and the reliability of the findings should improve.With respect to the studies hereby reviewed, it seems that these recommendations are not being carried out.
Results show an equitable distribution in the studies carried out concerning the level of governmental research.The analysis focus of 18 studies (38%) is on federal public administration, in another 15 studies (32%) on state governments and, lastly, 14 studies (30%) had municipal governments as their analysis focus.The health sector has been the most investigated, with seven studies, followed by agriculture and tourism, with four studies each, and education and science & technology, with two studies.Other sectors also investigated were justice, cultural, employment, finance and housing.
Another criterion refers to data gathering procedures.Taken individually, the technique which reoccurs more often in the studies that were reviewed was documental research, found in 40 (85%) of the 47 articles.This was followed by interviews, structured, semi-structured and open, which were used in 34 studies (72%); observation, participative and non-participative, in 15 (32%) of the cases; and questionnaires, found in four studies (8%).Other data gathering procedures appear only in isolated studies (8%), such as, for example, focal groups, used in conjunction with documental research and the Cherchiglia and Dallari (2006) interview.It is worth noting that of the 34 studies that used this interview, six did not provide information on the individuals who were interviewed.
It should be emphasized that some articles stated that they used bibliographic research as data gathering procedures.In the case of this research, it was decided to consider bibliographic research not as a data gathering procedure, but rather as a necessary and preliminary stage of empirical research.Thus, the most frequently used technique, found in 16 studies (see Table 2), combined documental research with interview.Another technique, in particular, that should be mentioned, found in eleven studies, was a combination of documental research, observation and interview.Among the 35 studies which explicitly indicated the use of a case study as a research method, the most frequently used data gathering procedure was documental research combined with an interview, found in eleven studies (31%).The combination which characterizes a case studydocumental research, observation and interviewwas used in only nine studies (19%).The majority of the studies, 33 (70%), used a combination of at least two data gathering procedures.The use of a combination of more than three different techniques was observed in 15 (32%) of the 47 articles reviewed.
It is possible to infer through these findings that, among those reviewed, only a small number of publications actually conducted what is commonly meant by case study.The reason for this, given the general understanding of what is meant by the case study method, that is to say, qualitative research which involves an interview, observation and a documental analysis, is that few researchers were committed to carrying out this combination of techniques.In addition, it is possible to observe an absence of methodological rigor in these studies, especially the methodological justification and the theoretical and practical basis to carry out the procedures developed.
One of the most troubling findings of the review is the complete lack of information on data analysis techniques used in 30 studies (64%).That is to say, in two-thirds of the articles reviewed, the authors did not consider it relevant to inform the reader how the collected data had been analyzed.It may be inferred that the authors considered that all data is analyzed by using one method only and that all their readers know what method this is.It can also be assumed that the data used in the 30 aforementioned studies was not properly examined, since the authors did not attempt to describe what procedures had been carried out.
Among the 17 (36%) studies that provided information about how the data had been analyzed, the most widely-used technique was content analysis, described in ten studies.Bardin (1977) was the most cited author for the theoretical basis of this technique.Three other studies used the discourse analysis technique: Ferreira and Araújo (2006), Mello and Amâncio (2010) and Saraiva and Nunes (2011).There were three other authors who reported that they had used more unusual techniques, such as observation protocol (Raupp & Pinho, 2011) and semantic meaning analysis (Siqueira & Mattos, 2008).Only two studies developed a statistical analysis in conjunction with qualitative analyses: Cabral, Barbosa and Lazzarini (2008) and Carvilhe, Pereira and Kato (2007).
The most used theoretical bases in these studies consist of classical issues related to public administration: state reform and new public administration.Each of these subjects was addressed in six separate studies.Another recurring theoretical basis was alliances and strategic networks, addressed in five of the studies.The theoretical bases of local development and participative democracy were addressed in four studies each.Together, these five theoretical bases were addressed in 53% of the articles analyzed.At first glance, this convergence of studies in only a few theoretical bases could mean that the theoretical evidences of previous studies are being tested with the replication of research in certain sectors.However, when one compares the theoretical bases of the reviewed studies and the respective sectors in which the objects of analysis are found, it is seen that alignment, with the exception of the tourism sector, was not addressed in two of the studies, where their theoretical basis was an overview of local development.
It is necessary to mention that subjective judgment was indispensable in the case of many of the articles, due to the lack of information.For example, most of the articles did not provide information about the strategy involved in the case study, even though there was relatively clear evidence that made it possible to infer how the article should be classified.Furthermore, in some situations, their judgment does not appear to have been appropriate.

Specific questions
The four specific questions were: what was meant as case study?What was the relationship between the phenomenon of interest and the case being investigated?How successfully could the studies be replicated?How did the so-called method contribute towards the development of public administration research?
Of the 47 articles that were analyzed, 12 (25%) did not mention the method that was used to define the case study, neither did they present any of the concepts or make any reference to any of the authors involved.Of the 36 publications that mentioned some type of definition or cited an author so as to conceptualize the case, 16 (34% of the total) made reference to Yin (2009) or other published editions of his work.The most frequent definition given was: "a case study is an adequate method when it involves a contemporary phenomenon, where the boundaries of the phenomenon and its context are not clearly established" (Yin, 2009, p. 18).Only Tinoco and Macedo-Soares (2008) mentioned Eisenhardt (1989) when categorizing a case study, even though they did not explicitly mention the definition given by the author.
For the second question, it was noted that the relationship between a phenomenon of interest and the case that was investigated seems to occur by convenience, which was the most used sampling strategy.Few of the studies actually present the relationship between phenomenon and case, something that would justify carrying out the case study in question.Amongst these, Sano and Abrúcio (2008), for instance, analyzed one of the experiences of a Brazilian social organization in São Paulo, and stated that the case selected was described as an intensity case, since the first social organizations were created by the Federal Government in 1997, and the State of São Paulo was the federal organ where this type of management was most prolific.Two issues were observed with respect to the possibility of replicating the studies: limitation to generalize results and lack of descriptive details of the case.In general, the authors limited themselves to stating that the use of a case study made the results obtained impossible to generalize.The authors do not seem aware of the need for research replication so as to consolidate study findings, and thereby ensure theoretical-scientific validation.Another possible inference is that, by stating that there are widespread limitations in a case, its findings appear more reliable since the results are not questioned due to the lack of conflicting or divergent observations.The difficulty of replicating the studies could, indirectly, be minimized if the case in question was thoroughly explained in detail, especially the characteristics that prompted the investigation, as confirmed by the representativeness of the phenomenon studied, and its peculiarity in relation to the case under study.However, these characteristics were not presented in the majority of the publications reviewed, as shown above.
The last specific question deals with the description and classification of the methodology of the reviewed studies, as well as their consistency in relation to the research work findings.An analysis was carried out about whether the authors proposed to use a case study or if they did so in accordance with the operational definition, that is to say, by means of an interview or questionnaire, observation and documental analysis.The main inconsistencies found were: the absence of a methodological description, as well as a specific methodology section; methodological procedures described in a confusing and/or erroneous manner; description of results inconsistent with data gathering procedures and data analysis techniques; incorrect application of the data triangulation technique; failure to mention the data analysis technique that was used; no mention of the number and characteristics of those interviewed or addressed in the questionnaires; absence of variables; and generalizations made based on insufficient data.
The absence of a methodological description disqualified the scientific validity of the published articles.In order for the reader to understand how the results were produced and to ensure that it may be possible to replicate the study, a description of the methodological procedures must be developed.An example of this type of methodological inconsistency can be seen in the work of Melo, Monteiro and Fadul (2007), classified as a case study.The aim of the study was to portray the changes that had been put into effect in the municipality's financial administration, during the period between 1989 and 2003.The research is limited to an analysis of the financial statements and even fails to provide a section describing the methodological choices, such as case strategy selection and the techniques used to collect data.
The article by Dossa and Segatto (2010), in turn, is an example of a publication that presents confusing definitions of the methodological procedures that were carried out.According to the authors, "semi-structured interviews were held, semi-structured questionnaires were used and documental research was carried out with all the respondents" (p.1340).Documental research is not a data gathering procedure that is applied to people.
Other serious problems identified are: 23 studies (48% of the total) did not include a description of the technique used to analyze the data that had been collected; 17 of the studies (35%) did not explain the variables analyzed in the study; and 8 studies (17%) did not state the number of interviews or questionnaires that had been carried out.Special mention must be given to the work published by Chagas and Ichikawa (2009), whose aim was to show projects carried out in the C&T network at the Agronomic Institute of the State of Paraná (IAPAR).These authors state that "an understanding of the phenomenon is based fundamentally on an interpretation of what is said by those participating in the survey" (p.9), even though only three people were interviewed by them.Similarly, Bonacim and Araújo (2010), for a study that attempted to analyze the costing system at the Hospital das Clínicas at the Faculty of Medicine of the University of Ribeirão Preto in São Paulo (HCFMRP-USP), as well as to evaluate a proposal to change the Hospital's running costs methodology, interviewed only one person.
Finally, but no less serious, some studies make generalizations that are based on insufficient information.A typical example of this type of inconsistency appears in the work of Saraiva and Nunes (2011) in relation to the effectiveness of ProUni, a public program created to increase access to higher education in Brazil.Based on only 11 interviews with students from a private higher education institution, the authors stated that "ProUni has fulfilled its objectives to make higher education available to all the most disadvantaged segments of society from a social-economic point of view, even though inclusion through quotas based on a person's ethnic background is condemned" (Saraiva & Nunes, 2011, p. 961).

Final Considerations
By means of a systematic revision, the purpose of this study was to identify the way case studies have usually been used in public administration research in Brazil.In general, the main inconsistencies found concern the fact that no descriptions are given about the method being used, the description of the findings is incompatible with the data gathering procedures and analysis techniques, the technique used to analyze the data is rarely divulged, the number of people interviewed is not mentioned and no information is given about the variables.
The findings showed that the publications that were analyzed provided few details about the methodological issues which supported the selection and development of the case studies that were carried out.The findings did not confirm the operational definition normally described in the literature and commonly attributed to Yin, that is to say, the research work did not combine the three suggested data gathering procedures and analysis techniques: interview or questionnaire, participant observation and documental analysis.It was noted that little effort was made on the part of the authors to establish a practical generalization or a local creation theory.In summary, serious methodological weaknesses were observed, confirming the opinion of Roesch (2003), for whom the potential for qualitative research was inadequately investigated because of the very lax methodological approach shown by the researchers.
In the majority of the studies, few of the authors sought to describe the characteristics of the case.Such description would justify its investigation, ensure that it was a representative study of the phenomenon in question, or attribute special characteristics to the case under study.The absence of information shown in the case studies helped confirm the statement made by Bertero et al. (1999) about the quality of Brazilian scientific research on public administration.The results evidenced the difficulty researchers face to establish boundaries for the phenomenon they were studying and to not confuse this with the locus of the analysis.In other words, the case (locus) selected for research, for example, a municipality, a public company, or a government office, cannot be confused with the phenomenon of interest, for example, precarious financial management in municipalities, corruption in public companies, or the relationship between secretariats and civil society organizations.Another difficulty noted was the validity of the studies, which, in general, made little theoretical contribution towards the field of public administration.
However, criticism of the manner in which the national researcher conducts a case study is not restricted to the field of public administration.The use of case study has been questioned for years due to its fragility in social sciences, in both national and international literature.There is nothing new regarding the issues discussed here.This study's contribution is to show that, regarding the use of case study, public administration faces the same problems encountered in other fields.
As summarized in Table 3, future qualitative studies in public administration are recommended to involve stricter methodological approaches and greater attention to some essential points: the case selection strategy, the methodological classification of the study, the methodological procedures for data collection and analyses, and the presentation and description of the results found in the research.The first point concerns the strategy used to select a researched case.As shown in the review, in Brazil, the vast majority of authors do not bother to justify why the case was selected.This suggests that personal convenience is the main criterion in the selection, not the relationship between the case and the phenomenon of interest.In other words, the selection of the case study seems to be made solely based on the locus where phenomenon occurs, and not the phenomenon itself.When authors make it clear why the case was selected, and what the relationship between the case and the theory of reference is, research credibility tends to increase.As a suggestion, one of the sampling strategies pointed out by Patton (2002) (extreme case, typical case, fragile case convenience, maximum intensity variation, and critical case), could be used as a guideline for selection of cases that are most relevant to the study.
What can be observed in the studies reviewed in this paper is that the term case study has been used in different ways, with numerous definitions.The attempts at methodological classification usually result in an unnecessary task, which creates even more confusion.In this sense, efforts could be directed to a clear definition of case study used in research and detailed description of the methods and techniques used.Some authors (e.g., Benbasat et al., 1987;Bonoma, 1985;Burns, 1990;Eisenhardt, 1989;Gondim et al., 2005;Leonard-Barton, 1990) indicate the situations in which the case study brings more benefits to research: describing the research subject in depth; reconstructing the object of research by relating the parts to the whole; and constructing theories and improving abstract terms and concepts, prepared from a description of the case study based on concrete evidence.
The combination of data gathering procedures is a common practice in the studies reviewed.Furthermore, one of the points omitted concerns data analysis.Among the studies, most authors do not detail the relationship between the techniques and the research objectives.Authors should indicate what the expected results for each technique are, i.e., what data will be collected, what information the data can generate and how the information generated (data analysis) can help in achieving the objectives of the research.The proper response to these questions enables a deeper analysis of the case and increases the validity of the research.
The definition of the methodological procedures, a priori, would bring clarity to the techniques used for both collection and analysis, delimiting the scope of the results and the limitations imposed by the techniques used.The proper response to these questions enables a deeper analysis of the case and increases the validity of the research.
With the gaps in the methodological process, the exploitation of results and the presentation of explanations and/or deductions for gaps presented in the literature become impaired.In the description of the results in the reviewed studies, two aspects deserve consideration: inappropriate generalization of the findings and the lack of connection between results and theory.Future studies should make clear how the findings relate to previous studies, i.e., how they help to reinforce or mitigate the theory of reference.Future studies should also explain how findings can be extrapolated to other cases.Such concerns provide conditions for the study replication and contribute to theoretical development.
Finally, it is important to remember that the studies reviewed in this research have been published in prestigious national journals in the field of public administration, and, therefore, went through a theoretically more rigid peer review process compared to other less prestigious journals.Thus, there are biases in the analysis that may represent potential limitations to the results of this research.Choosing a sample of studies for review considering only a select group of journals rated by CAPES means giving up the analysis of several other case studies, including unpublished research, e.g., papers presented at meetings and congresses, and studies published in less prestigious journals.Underestimating the deficiencies of publications in a field is one of the main problems that can occur in a review that considers only the publications of the best journals in the field.
As a final suggestion for future studies, considering the limitations presented in the previous paragraph, it would be interesting to compare the quality of national journals in the field of public administration and the quality of case studies published in these journals.In other words, to check whether articles published in journals with lower qualifications according to CAPES classification have the same problems encountered in the studies reviewed in this research, and the frequency and intensity of these problems.Thus, it would be possible to observe whether the classification of these journals reflects the quality of their published articles denominated as case studies.
(a) number of case studies investigated; (b) selection strategy used in each case; (c) temporal cut-off; (d) level of governmental research on the subject; (e) data gathering technique; (f) data analysis technique; and (g) theoretical basis.

Table 2 General Characteristics of the Studies Reviewed
Note.Source: Prepared by the authors.*Percentage in relation to the total number of articles (47) reviewed.