Scribd Logo
Upload

Welcome to Scribd!

  • KEOHANE, Robert, KING, Gary & VERBA, Sidney-Designing Social Inquiry 1

    Uploaded by

    André Dias
    69 views49 pages
    KEOHANE, Robert, KING, Gary & VERBA, Sidney-Designing Social Inquiry 1

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document
    KEOHANE, Robert, KING, Gary & VERBA, Sidney-Designing Social Inquiry 1

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    69 views49 pages

    KEOHANE, Robert, KING, Gary & VERBA, Sidney-Designing Social Inquiry 1

    Uploaded by

    André Dias
    KEOHANE, Robert, KING, Gary & VERBA, Sidney-Designing Social Inquiry 1

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    of 49
    Selsieesffelemme Designing Social Inquiry ‘Scientific nference in ualtatve Research ‘Cary ing / Robert 0, Kahane / Sidney Verba ‘Ata moment when acute dagreement among colts over he appropra- ‘ess of quiliaive and quantauve serch methods tvetens to undermine ihe validity and coherence af the soll scence, Gary King, Robert Rechans, nd Sidney Verb have writen a tmely and asighied ook tht develop + ‘led approach o valid dseipive and aus inference. The laminate he loge of good quanlatve and good qualitative research designs and ema state that he fo do at fundamental dif Designing Sari nqiy focuses fn improving qualitative research, were numerical measurement i cither imposible or underable. What ae the right questions fo stk? How should You define and make inferences about sual effect? Hove ca you avoid Was? Tow many eases do you need, and how should they be oleted? What ae the onsequences of unavoidable problems in qualtaive research, such 4 he ‘Strement eo, incomplete formation or omiled varies? What sre proper ‘ways to estimate and report the uncertsly of your concusions? How would you know f you were wrong? “nally! For to lng, those of vs in comparative politics and international telatons have been itd fom thereat plies science, Decuse ofthe ‘sues we investigate and the dala we are constrained toemploy we have often [een unable to apply the stata methods of our colleagues. We have isad Innovated a sere of ad hac and idisyncraic methods ith ile systematic ofeach ro the trade arnong them, King. “Hldber Bates, Center or donc Shy i the earl Science "Essen ending for any serious analyt of ne “race Rustl, Yale Uncerily have used Designing Sia nur in det form in. introductory grade ‘seminar, The Conde of Ingulrysor the ast tees years Its 8 marvelous pec of work to which my students responded very peiively. The author: Mle great dea of practical advice about how to ge the mos nilenge’ cut of ‘san peice sia “David W. Rohde, Unters of Flite PRINCETON PAPERBACKS coy 9-69-0947 1-0 screener HINT iia Cn (SR, Salmon Copgnein "Yodk, Hila Reboy Coection ST NTT ci fing / Robert en Designing Social Inquiry SCIENTIFIC INFERENCE IN QUALITATIVE RESEARCH Gary King Robert O. Keohane Sidney Verba PRINCETON, NEW JERSEY Contents ee Preface opp onto Unt Pr 7 nied rn ety Pan 1 The Science tn Social Science 3 In Unies Kangen Poon Uiety Pst 1.1 Intrduction 3 Ghee We se 11.1 Ta Sige of Research, One Lgi of nfezence 3 AR Rexred | 143 Dating Seif Rech i Sal Sees 7 ay Cn Cain i Pantin Do 113 Scence and Complxiy 9 mon 1.2 Major Components of Reseach Design om Dating scm: nine + 12.1 Improving Research Questions 4 seth 7 Gay King abo OKs Sn vo 122 Improving Theory » vs 123 Improving Data Quaity B oP | 124 Improving the Useo Existng Data z 13 Themes of This Volume B So 413.1 Using OlserobeImplations lo Connect Teary shane, fab Owen, and Data 28 finer Sd 13.2 Mainizing Leverage 2 Benisa7 ot sp 133 Reporting Uncertainty 3 re ; 134 Thinking like Social Scientist: Steptcism Pose ha eo compa in debe Pan ua Rol Hiypotieses 2 ‘tan Un pt leper id te ans a iceman drat he Conn 2.1 General Knowledge and Pacticular Facts 3 oc Calis for Bok Leng 211 “Interoreation” and inference 36 Sth Cost en vary Bare i 21.2 “Uniqueness” Complexity and Simplifention 2 ead int Uni Sf Aeron 213 Comparative Case Stas a woeresaazi | 22 Inference: the Seintifc Purpose of Data Collection 6 weer esaszt | 23 Formal Models of Qualitative Research ° om | __ 24 A Formal Model of Data Collection 31 | \ 25 Summarizing Historical Detait 3 | \ 26 Descriptive inference 5 | > 27 ctr for judging Descriptive Infeenes 8 i 271 Unies tnferences 3 | 2272 Eficncy 6 wt» Content {3 Causality and Causal Inference 41 Defining Causality ‘3.414 The Definition and a Quantitative Example 312A Quiitative Example 132 Clarifying Alteative Definitions of Causality ‘321 “Cas Mechs” 3.22 “Multiple Causality” 3.23 “Symmetric” and "Asyometric” Chusality 33 Assumptions Required for Estimating Causal Effects 334 Unit Homogenty 33.2 Conlitiona Independence 34 Ciiteria for Judging Causal inferences 35 Rules for Constructing Causal Theories ‘3.51 Rule 1: Construct Falsfable Thasries 352 Role 2 Build Theories That Are Internally Consistent 3.53 Rule3 Select Deptt Variables Careflly 354 Ruled: Maximize Cnertenes $355 Rule 5 Slate Thoves nas Encompussing Ways 8 Fersibe 4 Determining What to Observe 41 Indeterminate Research Designs 4.11 More nferenees than Olseroations 412 Multcinenity 442 The Limits of Random Selection 43 Ssection Bias 4341 Seleton on the Dependent Variable 432 Selection on an Explanatory Varabie 433 Other Types of Selection Dine 44 Intentional Seletion of Observations 444 Selecting Obserotions on the Explanatory Variable 442 Solectng Range of Values ofthe DeenentVarible 443 Selecting Obserotions on Bal Explanatory and Depenient Variables 444 Selecting Obseoaions So the Key Causal Variable Is Constant 445 Selecting Obseontions So the Depentent Variable 1s Const 45 Concluding Remarks SSereedaaarna S8as 13 15 18 19 m4 2 129 197 138 19 Mo MI i M6 “7 us Contents 5 Understanding What to Avoid 51 Measurement Error 5.11 Systematic Mesurement Eror 5:12 Nowsystennic Messurement Error 52 Excluding Relevant Variables Dias ‘52.1 Gauging the Bis frou Ont Variables 5.2.2 Camples of Ont Varia Bins 53 Including ireevant Variables: Inefficiency 54 Endogencity 541 Corncting Bite Inferences 542 Parsing the Dependent Varale 543 Thasforming Endegeuely into an Onitad Variable Prien | 544 Selecting Obseroatons to Avil Enlageily 54.5 Parsing the Explanatory Variable 55 Assigning Values of the Explanatory Variable 56 Controlling the Research Situation ‘57 Concluding Remarks 6 Increasing the Number of Observations 61 Single Observation Designs for Causal Inference 614 “Crucat” Case Stes 61.2 Reasoning by Analogy 62 How Many Observations Are Enough? 63 Making Many Observations from Few 63.1 Sime Mensures, Ne Units 63. Same Unite, New Mossuns 633 New Measures, New Units 64 Concluding Remarks Beferenees ule 150 151 155 157 168 168 17% 182 185 187 188 19 11 Preface | s 200% we develop a unified approach to valid descriptive and ‘ausal inference in qualitative research, where numerical measure: ‘ment i ether impossible or undesirable. We argue thatthe logic of 00d quantitative and good qualitative research designs donot fn ‘mentally dlifer. Our approach applies equally to these apporenily di ferent forms of scholarship. ur goa in writing this book is to encourage qualitative researchers to take scientific inference seriously and to incorporate it into thle work. We hope that our unified loge of inference, and our atempl to demonstrate that this unified loge can be help to qualitative re. sarchers, will help improve the work in our discipline and perhaps aid research in other social sciences as well. Thus, we hope thet iia book is rend and critically considered by politcal slentists and other ‘social scientists of all persuasions and career stages from qualtalive fold researchers to statistical analysts, from advanced undergraduates snd first-year graduate students to senior scholars. We use some mathe ‘ematical notation because it is especially helpful in clarifying concepts in quolitative methods; however, we assume no prior knowledge of mathematics or statistics, and most ofthe notation can be skipped without loss of continuity University administrators often speak of the complementarity of ‘caching and research. Indeed, teaching and research ae very nevrly ‘coincident, in that they both entail acquiring new knowledge and com. ‘municating itt other, albeit in slightly different forms, This book st. less to the synchronous nature ofthese actives Sinre 189, we have ‘ben working on this book and jointly tenching the graduate seminar “Qualitative Methods in Social Science” in Harvard University’s De- partment of Government. The seminar has beon very lively, and ‘ten has spilled into the halls and onto the pages of lengthy memos. passed among ourselves and our students Our intellectual bates have always been friendly, but our rules of engagement meant that “agrosing to disagree” and compromising were high crimes. I one of us was not truly convinced ofa point, we took it as our obligation to centinue the debate. In the end, we each learned a great deal about ‘qualitative and quantitative esearch from one another and from out sludents and changed many of ou initial positions In addition to is primary purposes, this book is a slatement of our hard-won unani- ‘mous postion on scientific inference in qualitative research K+ Palace \We completed ihe first version of this book in 1991 and have revised it extensively in the years since. Gary King first suggested that we ‘write this book, drafted the fist versions of most chapters, and took the fend througi the long proces of revision. However, the book has ben rewitten so extensively by Robert Keohane and Sney Verb, 08 well as Gary King, that it would be impossible for us lo identify the authorship of many passages and sections reliably, During this long process, we circulated drafts to colleagues around the United States and are indabted to them forthe extraordinary gen- sity of thelr comments. We are also grateful to the graduate si dents who have been expose to this mantuseript bol at Harvard and at other universities and whose reactions have been important 10 us {in making revisions. Trying to ist ll he individuals who were helpful in a proct such as this ts notoriously hazardous (we estimate the probabilly of inadvertenly omitting someone whose comments were Important to us to be 092), We wish to acknowledge the following, individuals: Christopher H. Achen, John Aldrich, Hayward alker Robert H. Dates, james Dats, Nathaniel Beck, Naney Burns, Michael Cobb, David Collier, Gary Cox, Michael C. Desch, David Dessler, Jorge Dominguez, George Downs, Michell Duncier, Matthew Evangelist, John Fereohn, Andrew Gelman, Alexander George, Joshua Goldstein, ‘Andrew Green, David Green, Robin Hanna, Michael Hiscox, James E Jones, Sr, Miles Kalle, Eizabeth King, Alexander Kozhemiskin, Ste [phon D. Krasner, HerbortKritzer, James Kuklinskl, Nathan Lane, Petr Lange, Tony Lavelle, Judy Layzer, Jack S. Levy, Daniel Lite, Sean {yna-ones, Lio L. Martin, Helen Milner, Gerardo L. Munck, Timothy F.Nokken, Joseph S. Nye, Charles Ragin, Swarna Rajagopalan, Sha- ‘mara Shanty Riley, David Rocke, David Rohde, Frances Rosenbluth, David Schwieder, Cllins C. Shackelford, Jr, Kenneth Shepsle, Danie, Walsh, Carolyn Wamer, Steve Aviv Yeti, Mary Zerbings, and Mic ‘chael Zam. Our appreciation goes to Steve Voss for preparing the index, and to the erew at Princeton University Press, Waller Lippin- ‘ott, Malcolm DeBevoise, Peer Dougherty, and Alessandra Bacco. Our thanks also go to the National Science Foundation for research grant ‘SBR-9223697 to Gary King. Robert O. Keoliane is grateful to the John ‘Simon Guggenheim Memorial Foundation fora fellowship during the term of which work on this book was completed, ‘We (in various permutations and combinations) were also extremely fortunate to have had the opportunity to present earlier versions of this book in seminars and panels at the Midwest Political Scenee As ‘sociation meetings (Chicago, 2-6 April 1950), the Politeal Methodol- ‘ogy Group meetings (Duke University, 18-20 July 1950), the American Place + xd oli Sconce Association matings (Wsshingion, DC. 29 August {September 190, the Seminarn the Mathology aed Poa of the Sd Snes (Harvard University, Center rites Ak fairs, 25 September 192), the Collauium Svcs of the Inet plitary Consort for Statistical Appiations Chana Univers 4 Deceer 1990, the tiie for Cabal Conpamion aad Chey seminar series (University of California, erkley, 15 Janay 1038, na the Univers of in, Urbana-Chompaige 8 Mar 3) Gary King Robert ©. Keohanne Sidney Verba Cambridge, Massachusets Designing Social Inquiry CHAPTERT The Science in Social Science 1L1 Inrropucrion ‘Tints noox is about researc inthe socal sciences. Our goal is practical signing research tht wil produce valid inferences abou social and politcal life We focus on political science, but our argument applies 2 ‘other disciplines such as sociology, anthnopology, history, econ ‘and psychology and to nondislplinary arias of study auch aa legal evidence, education research, and clinieal reasoning, > ‘This is neither a work inthe philasophy ofthe socal sciences nor a {guide to specific research tasks Such as the design of surveys, conduct ‘of fie work, or analysis of statistical data, Kathe, this a book about ‘search design: how to pose questions ant fashion scholarly research tormake valid descriptive and causal inferenes. A such it occupies ‘mide ground between abstract philosophical debates and the hands- ‘on techniques ofthe researcher and focuses on the essential logic n= {erlying al social scientific research LAL Taw Styles of Research, One Lagi of Inference ‘Our main goal isto connect the traditions of what are conventionally denoted “quantitative” and “qualtaive” research by applying ui fied logic of inference to both. The two traditions appear quite difer- ‘ent; Indeed they sometimes seem to beat war. Our view le that these dliferences are mainly ones of style and specific technique. The same, underlying logic provides the framework for each research approach. ‘This logic tens to be explicated an formalized clearly in dlcussions ‘of quantitative research methods. But the same logic of inference une derles the best qualitative resoarch, and all qualitative and quantita: live researchers would benefit by mone explicit attention to this logic in the course of designing research. ‘The stycs of quantitative and qualitative research are very diferent. ‘Quantitative research uses numbers and statistical methods, It tends to. be based on numerical measurements of specific aspects of phenom ‘ena; i abstracts from particular instances to seok general descr ‘oF To test causal hypotheses it secks measurements and analyses that are easly epleable by other researchers, 4+ The Scien Soci Scie ‘untae research in contrast, covers wide ange of approaches, but by dttiion, none of thee approaches ries on numa! met surements, Such work has tend fo focis on one or 2 sal une ‘corto se intensive interviews or depth analysis of itr mae tea to bedecursive nth ane tobe concerned witha rounded or comprehensive acount of some event or sit Eventhough they, fave a small numberof ees, qualatve researchers generale ‘rth enormous aunts of infortation rom thei i, Sometings Ih kind of werk inthe socal scence linked with ro ce sic ies wher the focus on n pater event decison instituion, oes ton ue, oF pee of leiation. As isan the cse with quantave ‘cary the instance often portant ins own right a mar "another way of destin cena with rare, ages events to engage in counteract! analysis “the ronal constrton of 8 Ietrattion © 11 course of events which is altered through modifications in one or more ‘conditions'” (Weber [1905] 1949:173. The application of this idea in ‘a systematic, sclenlifc way i lustrated in a particularly extreme ex- ample of a rare event from geology and evationary biology, both historically oriented natural sciences. Stephen J. Gould has suggested that one way to distinguish systematic features of evolution from sto chasticy chance events may be to Imagine what the world would be like if all conditions up to a speci pont were fixed and then the rest, ‘of history were rerun. He contends that it were posible to “replay the tape of if" 1o et evolution occur again fom the beginning, the work's organisms today would be a completly differnt (Could 19853). ‘A unique event on which students of evolution have recently for cused isthe sudden extinction ofthe dinosaurs 65 milion years og Gould (1989316) says, “we must assume that consciousness woud not have evolved on our planet if a cosmic eatatrophe had not ‘claimed the dinosaurs as victims.” I this statement I tue the extine- tion of the dinosaurs was as important as any historical event for human boings; however, dinosaur extinction does na fall neatly into class of events that could be studs! in a systematic comparative fashion through the application of generat laws in a straightforward way. [Neverthles, dinosaur extinction canbe studied scientifically: alter: native hypotheses can be developed and tested with respect to their observable implications. One hypothesis to account for dinosaur ex: tinction, developed by Luis Alvarez and collaborators at Berkeley in the late 1970s (W. Alvarez 1950), pests a cosmic cllsion: a meteor. ‘te crashed into the earth at about 72,000 kilometers an hout, creating a blast greater than that fom a full-sale nuclear wat. If this hypothe: fie ie corvect, I would have tho observable implication that tedium Gan clement common in meteorites but rare on earth) should be found inthe poricular layer of the earths crust tht corresponds to sedi- rent laid down sixty-five million years ago; indeed, the discovery of Tridium at predicted layers in the earth has been taken a8 partial con- firming evidence forthe theory. Although ths is an unambiguously unique event, there are many other observable implication. For one ‘example it should be possible to find the metorite's crater somewhere ‘on Earth (and several candidates have already boen found) > “The issue of the causes) of dinosaur esinction remains unresolved, though the controversy has generated much valuable research. For 2 Htweve an alternative hype hal exam was cb won erp osc th he pra dom and summon cost ha Them hyphae hh ing ae apc ents ed > 12+ The Scr Social cence 5 ‘ur purposes, the point ofthis example is that scintfc generaline Sons areuseflimstoying even highly unsealed ta into lige css of evens, The Alvarez hypothesis cant be tested with efetence toa set of eommon evens tit does have observable implications fr other phenomena Wat cn be evaluated. We should ‘oe however that a hypothesis snot considered a reasonably certain explanation unit has been evaluated emprkally an! passed a mums ber of demanding tests At minim is implications must be con sistent with our knowledge of the external world; at bet it should predietwhat Ince Lakatos (1970) refers to “new fats hat i these formerly unobzerve “The pont stat even apparel unique evens such 35 dinosaur extinction can be studied scentially i we pay altenton fo improv. ing theory, data, and our use of the data Improving our theory rough conceptual clarification and speciction of variables cay {generate more observable implications and even lest éausal theories of unique events such a dinosaur extinction Improving our data a. lows us to observe more ofthese observable mpliations and improv. ing ou we of data permits more ofthese implications to be extracted from existing data That a set of events tobe atid is highly complex cs not render careful research design irlevant Whether we sty ‘many phenomena or few-—or even onthe sadly will be improved i ‘we calle data on as many observable implications of our theory 98 posiblg, ‘12 Mason Components oF Restanc Dasicn Social scence research a its best isa creative process of insight and discovery taking place within a well-ctablished structure of selene nguity. The first-rate socal scientist docs not regard a research design 28 bluopsint for 3 mechaniel process of dat gaticring and eval tion. To the contrary, the scholar must have the flexibility of rind to ‘overturn old ways of looking at the world, to ask new questions, to Fevise research designs appropriately, and then to collect more dat of different type than originally intend. However, ithe escarchers findings are to be valid and accepted by scholars in Us Fed, all these revisions and reconsiderations must take place according to expli Procedures consistent with the rules of inference. A dynamie process Of inquiry occurs within a stable structure of rules Social scientists often begin research with a considered design, cok leet some data, and draw conclusions. But this process is rarcly 3 smooth one andl is not always best done in ths onder: conclusions ‘rarely follow easily from a research design and data collected in accor Major Components of Research Design =” 13, or has collected data as provided by 8 research design, he or she will often find an imperfect ft among the ‘main research questions, the theory and the data at hand. At thisstage, researchers often become discouraged. They mistakenly belive tha ‘ther socal scientists find close, immediate fits between data and re- search. This perception is due to the fact that investigators often take scaffolding after putting up thir intellectual buildings, lav trace of the agony and uncertainty of construction. Ths the rocess of inquiry seems more mechanical and eut-and-dred than it setally i, Some of our advices directed toward reenter who ae trying ig ‘make connections between theory and data, At times, they can design ‘more appropriate data-colletion procedures in order to evaluates Uhcory etter at other times they cin use the data they have nd recast theoretical question (or even pose an entirely different question that ‘was not originally foreseen) to produce a more important research project, The research, if it adheres to rules of inferenee, will sll be Scientific and produce reliable inferences about the worl. Wherever possible, researchers should alzo improve thei research > designs before conducting any field research. However, data has a ‘way of disciplining thought Its extremely common to find thatthe best research design falls apart when the very fist observations are, collected—it isnot that the theory is wrong but that the data are nof suited to answering the questions originally posed. Understanding, from the outset what enn and what eannot Be done at this later stage ‘an help the researcher anticipate atleast some of the problems when first designing the research. For analytical purposes, we divide all esearch designs into four, components: the nscarh quesfon, he theory, he date, andthe we ofthe “ data, These components are not usually developed separately snd ‘scholars do not atfond to them in any preoniained onder In fat, Tor ‘qualitative researchers who begla thelr field work before choosing a precise rescarch question, data comes fis, followed by the others However, this patticular breakdown, which we explain in sections 121-1.24, is particularly usefl for understanding the nature of re scorch designs. In onder to clarity precisely what could be done if re ‘sources were redirected, our advice inthe remainder ofthis section ‘assumes that researchers have unlimited time and resources. Of ‘course, fn any actual research stsation, ane mist always make com Dronises. We believe that understanding the advice inthe four eat {ories that follow will help researchers make these compromises Such a way as to improve thei research designs mos, even when in fact thee esearch is subject to external constants +The Sine in Socal Science %* 121 Improving Rscrch Questions Throughout this book, we consider what to do once we identify the 1 object of research. Given a research question, what are the ways 10 ‘conduct that research so that we can obtain valid explanations of socal and politcal phenomena? Our discussion begins witha research qu tion and then proceeds tothe sages of designing and conducting he ‘esearch, Dut where do research questions originate? Tow docs 9 Scholar choose the topic for analysis? Theres no simple answer to this ‘question. Like others, Karl Popper (196832) has argued that “there is ‘no such thing sa logical method of having new ideas... Discovery contains ‘an irrational element, or a ‘erative intuition: The rules of choice at the earliest stages of the research process ar less formalized than are the rules for other research activities. There are texts on de signing laboratory experiments on social choice, statistical criteria on ‘nal investigator. Since so much of the data infield research involve ‘conversations, impressions, and other unrecorded parlicipatoy infor ‘mation, this reanalysis of results using the same dat snot often done However some important advances might be achieved if more schol. ‘ar tried this type of replication, and it would probably also encourage, ‘others to keep more complete field notes, Occasionally an entire te search projec, including data collection, has boen replicted. Since We ‘annat go back in time, the replication cannot be perfect but can be ‘uite valuable nonetheless. Perhaps the most extensive replication of "An example he ue of mare tha on cole o etc tema norton ‘rm wars ol deh intros I wo eo thse cling ray, en ss hw ton ty pute same hugs thy det pcs ree a ‘re fen we a th ong res mr Sl yop Evert 2 ‘roca ten goat ot the pitino Pen hen ‘oe iyi has oy Major Components of Resch Design °- 27 a qualitative study is the sociological study of Middletown, Indiana, bbogun by Robert and Helen Lynd, Tele fist "Middletown study was Published in 1929 and was replicated in a book. published in 1997, (Over fity years afer the original study, a long series of books andl articles are being. published that replicate these original studies (see Caplow etal, 1983s, 19830 and the citations therein). All qualitative replication need not be this extensive, but this major research projet should serve as an exemplar for what fs possible, All research should attempt to achieve as much repicabilily a8 pos- sible: scholars should always record the exact methods, rules, nd pro ‘calures ust to gather information and draw inferences so tha ane ‘other researcher can do the same thing and draw (one hopes) the same ‘conclusion. Repliability also means that scholars who use unpuly: lished oF private records should endeavor to ensue that future sch ars will have acess tothe material on similar terms taking advantage. ‘of privileged access without seeking access for others precludes repli= cation and calls into question the sienific quality of the work, Usually ‘our work will not be replated, but we have tl 28 if someone may wish todo 5. Even ifthe work fs not replicate, providing the materials for such replication will enable readers tow ‘derstand and evaluate what we have done, 124 Improving the Use of Exiting Data Fixing data problems by collecting new and better data is almost ‘ways an improvement on trying to use existing, lawed data in beter. ‘ways; however, the former approach is not always possible, Soca scientists often find themselves with problematic data ad litle chaned to acquire anything better; thus, they have to make the best of whol, May Bove . Improving the use of previously collected data isthe main topic tought in classes on statistical methods and is, indeed, the chief contri= bution of inferential statistics to the social scleces, The precepts on Is tople that are so clear in the sty of inferential stalisties also apply to qualitative research, The remainder ofthis book deals wi these precepts more fully. Here we provide merely a brief outline of the guldelines for improving the use of previously collected data, Fist, whenever possible, we should use data to generate inferences that are “unbiased,” that is, correct on average. To understand this, very specific idea from staistieal research, imagine applying the same ‘methodology (in quantitative or qualitative research) for analyzing, and drawing conclusions from data across many data sel. Because of small errors in the data or in the application ofthe procedure a single ‘pplication of this methodology would probably never be exactly cor: 2 + The Scion ln Soil Scion eet An “unbiased” procetue will be correct when taken as an aver _2ge aerss many applications even if no single application i correct “The procedure wil not systematically tthe outcome in one direction oranoth ‘Achieving unbiased inferences depends, of course, both on the ofgi- nal collection ofthe data and its Inter use; and, a we pointed out be- fore it is always best to anticipate problems before data collection be- ‘ins, However, we mention these issues briefly here because when Using the data, we need 10 be particularly careful to analyze whether sources of bias were overlooked during data collection. One such source, which ean Tead to biased inferences s that of sclection bins ‘choosing observations in a manner that systematically distorts the ‘population from which they were craven. Although an abviows xan ‘ples deliberately choosing only cases which sipport our theory, sel tion bias can occur in much more subtle ways. Another difficulty can ‘result from omitted variable bas, which refers othe exclusion of some contol variable that might influence a seeming causal connection be {ween our explanatory variables and that which we want to explain We discuss these and numerous other potenti pitfalls in producing unbiased inferences in chapters 2-6 ‘The sacond guideline is based on the statistical concept of “ei ciency": an efficient use of data involves maximizing the information used for descriptive or causal inference. Maximizing efficiency re- ‘quires not only using all our data but also using all he relevant ior tation in the daa to improve inferences. For example, i the data are ‘disaggregated into small geographical units, we should use it that ‘way not jst a anatlonal aggregate, The smaller aggregates will have larger degrees of uncertainty associated with them, but If they a6, at feast in par, observable implications of the theory, they will con: lain some information which can be brought to bear on the inference problem. 13 Tusnes oF Tis Vowume ‘We conclude this overview chapter by highlighting the four important themes in developing research designs that we have discussed here ‘and wil elaborate throughout this book. 1341 Using Oberon Implications to Conncet Theory and Data In this chapter we have emphasized that every theory, to be worth: ‘hile, must have implications about the observations we expect to find if he theory is correct, These obserble inpliations ofthe theory ‘Thomos of This Volume = 29 ‘must guide our data collection, and help distinguish relevant from i relevant facts In chapter 2.6 we discuss how theory affects data clloe- tion, as well as how data disciplines theoretical imagination. Here, we want o stress that theory and empirical research must be tightly con- ‘nected, Any thoory that docs real work for us has implications for em- pirical investigation; o empirical investigation can be successtul with: ‘out theory to guide its choice of questions. Theory and data collection are both essential aspects ofthe process by which we sick to decide, ‘whether a theory should be provisionally viewed true o alse, subject, ‘is in both eases tothe uncertainty that characterizes al infren ‘We should ask of any theory: What are its abervable implisatons? ‘We should ask about any empirical investigations: Are the observ tion relevant to th ise, what de they tenable ws to infor about the correctness of the theory? In any sacl scientific study, the implications of the theory and the observation of focts need to mesh with one another: socal science conclusions cannot be considered ellable if they are not base on theory and dala in strong, conection with one another and forged By formulating and, ‘examining the observable implications of a theory. 132 Masinising Leverage “Te schol who searches or addtional implications of a hypothesis is pursuing one ofthe mt important achievement fal ta ac ‘oe expan ch ys site poste Cood soil Science secks to Inereze the signfennc of wha explained reative {othe information wea inthe explanation Mw con coun Plain what a frat appears to b's complied fect with a age ‘nisl variable ora few variables the ere we have ove problem is very high, Convery if wecan explain many ef onthe basis of nv Oa foe varabes We ao have gh evrgs- Levene so the sul sions nm general and even more 0 fn parialsr sect rss. This may be Betuse salar do nt yet ow how to increase itor boonss nar happen nol tobe creed in coment nr itn or for both of thee reson Arcs oeventonally studied quali ily tn hse nw vg 2 ve Eagan ny thing scons to aque a ost f explanatory variables we sea ato xpi it. In such ches, our goa shuld be to design recarch ‘rh more leverage, “There are various ways in which we can incense our leverage over a remch problem. The prinary woy is fo merase the muror of Sieervabe impictions of eur Bypothesis ar sock cofrmaton of thw inmphistons, As we have described above isan enn vag + The See in Social Science (0 improv the theory 50 that thas more observable implications, improving the data so more ofthese implication are indeed eb sewed and wed fo evaluate the theory and C) improving tne ss ef the datas that more ofthese inplstions are exacted fea exiting da, None ofthese, no the general concept of maximizing leverages fe the same asthe concep of pasnony wich, as we eset Section 122, fan anumption about the nature a he wor ater _ hana rie for esging esa Maximizing leverage 0 ipertant and so general that we strongly ronment resco a ell ose ler inplaing Ft pss ht might be oer it edt o nl dat maybe pablo et some ofthese nv implant the gia dts sts tng onthe ipaton doc nl “eome ut te tts buts hypoth indepen suggested by the ry or dil dataset Buti beter stl ( tur oer daa, Thus we should So camer impeatos that might appear i ther dala as ‘ata aout ther unit, dta tout ote sopcs ofthe unis under Study, data frm diferent lvls of aggregation, hl data fom ster tne periods such as predictions abot the ner fature—a calc the Nypoiesi hae ing, The store evidence we ci Ip tare one he more pera our explanation becomes, a te tore confidence we a ters should have in or conclusions At irs thought, some researchers may cbt tote sof elle Ing servabeimpations rm any source or at ny level gstept ton diferent rom that for hic the theory wos aig. Poser 18 Ltn) ois toga rah esis a “ecloil fallcy”~incoretly sing aggregate data fo make in tenes about Indidualsio wom nas eseiee lo Weert agree that e cn uue aggregate daa lo make fncoet fnfrences abou inividalif we ae rested in individuals then staying divans generally 9 ter sttegy i we ean bain thee daa, However if the inference we seco hake ls more ten 9 ‘ery nao eat hypotss, our Uns may have implication at tay eve of nals and we wil oie be abe tous dal rom ll hse eels to provide sme information about our thor. Thu, even if we are primal intrested nan aggregate level of anti, we ean ea teeny en spree oe corte mest int aned schoo ting win iin cna Carnal ets esate ‘power e ao, mee “Themes of This Volume = 31 ste gin egies eying in eee Fe ay + ty tele on Meee ne ce sree te leone epee neh cme i pane male Cn nay eee picasa coat tem CE ee a cart Se ae not be ae er seman e se ree Pe sre ory ie at ae a li ele Sater eran Cae oe ta ee pm Seether Sree tiene EE eo i ae ite pert tee omer eepe ee se eae a eae mc pene recency read ictal a ene toa ee ee Ne ean Zain ete ee ee te sy a eee tari alte Fence arn fated wate ce pekateentcr ta ree per te 133 Reporting Uncertainty, Al knowledge and alt inference—in quantitative and in qualitative fesearch—is uncertain, Quallaive measurement is efrorprone, 28 {s quantitative, but the sources of error may differ. The qualitative in; terviewer conducting a long, in-depth interview with a respondent whose background he has studied i es key to mismeasure the sub- ject’s real political ideology than is a survey researcher conducting a Nructured interview with a randomly selected respondent about ‘whom he knows nothing, (Although the opposite ts also possible i, Tor instance, he relies too heavily on an informant who Is not trust: [32 «The Seen Socal Sconce worthy) However the survey resarcher is less likely to generalize inappropriately from the particular cases interviewed tothe broader population than is the in-depth researcher. Neither is immune from {he uncertainties of measurement oF the underlying probabilistic na- ture of the social world. ‘All good socal scientists—whether in the quantitative or qualitative trodittone—report estimates ofthe uncetainly of thelr inferences, Per haps the single most serious problem with qualitative research in ps ical sctence isthe pervasive failure to provide reasonable estimates, ‘ofthe uncertainty of the investigator's inferences (se King 199). We ‘an make a valid ference in almost any situation, no matter how Kine ited the evidence, by following the rules inthis book, but we should voit forging sweeping conclusions from weak data. The point i not That reliable inferences are impossible in qualitative research, but alher that we shoul always report a reasonable estimate ofthe deg of cerkainty we have in wach of our inferences, Neustadt and May (1986278, dealing with areas in which precise quantitative estimates are difficult, propose a useful method of encouraging, policymakers (rho are offen faced sith the necesity of reaching conclusions about ‘what policy to follow out of inadequate daa) to judge the uncertainty ‘of thelr conclusions. They ask "How much of your own money would You wager on i” This makes sense as Tong a we also ask, “At what onde 13.4 Thinking tke a Soil Scientist: Sheticism. ‘ul Rita Hype, ‘The uncertainty of causal inferences means that good social scientists do not easily accept them, When told A causes B, someone who “hi ikea social cientis” asks whether that connection is true chusal ones easy fo ask such questions about the research of others, butts more important to ask them about our own research. There are ‘many reasons why we might be skeptical ofa causal account, plausible though it may sound at fst glance. We read in the newspaper thatthe Japanese eat ess red meat and have fewer heart attacks than Amer ‘ans. This observation alone is interesting. In addition, the expl ion—too much steak leads tothe high tale of henet disease in United States plausible. The skeptical social sclentist asks about the accuracy ofthe data (how do we know about eating habits? what sam- ple was used? are hoart attacks classified similarly in Japan and the United States so that we are comparing similar phenomena?) Assum- {ng thatthe data are accurate, what else might explain the effects: Ae there other variables (ther ditary differences, genetic features, life ‘Themes of This Volume + 30 steep te ot ea eine woe te cs then who i ut hve es tack tere eaten ‘tig nor our purpose tovcall such medical studies Into question. Rather we wish merely t illustrate how social scientists approach the [neue of cual inference: with skepticism and concer for alternative fxplanations that may have been overlooked, Causal inference thug Paomes a press where each conclusion becomes the occasion fof farther research to refine and testi, Through successive approximy tions we ly to come closer and loser to accurate causa inference. cHarTER2 Descriptive Inference SociaL science naseancit, whether quantitative or quali volves the dual goals of describing and explaining, Some scholars set ‘ut to describe the world; others to explain. Eich is essential We can- ‘ot construct meaningful causal explanations without good descrip: tion; description, in turn, loses mast of its interest unless linked fo some causa relationships. Description often comes first is hand to develop explanations before we know something about the world and ‘what needs to be explained on the basis of what characterstis, But the lationship between description and explanation i interactive. Somme- |}imes our explanations lead us to look for descriptions of different |Ipunts of the world; conversely, our descriptions may kad to new ‘usa explanations. Description and explanation both depend upon rules of scientific Jnferenc, In this chapter we focus on description and descriptive it erence. Description is far from mechanical or unproblematic since i fnvolves selection from the infinite numberof facts that could be re: corded, There are several fundamental aspects of scientific description, ne shat it involves inference: part of the descriptive tak isto infer Information about unobserved facts from the facts we have observed, Another aspect involves distinguishing between that which is systems ‘atic about the observed fais and that which fe noneystemati, ‘As should be clear, we disagree with those who denigrate “mere description. Even if explanation—connecting causes and effec—is {the utiate goa, description has a central roe in al explanation, and {tis fundamentally important in and of iselft fs not description ver: sus explanation that distinguishes scientific research from other re- search Iti whether systematic Inference is conducted according to valid proceduts. Inference, whether descriptive oF causa, quanti tive or qualitative isthe ullimate goal ofall good social science. Sy tematicaly collecting facts is a very important endeavor without ‘hich cience would not be possible But which doesnot by itself con slitut science. Good archival work or well-done summaries of histor cal facls may make good descriptive history, but neither are sufficient te constitute social science In this chapter, we distinguish description—the collection of facts— from descriptive inference. In secion 2.1 we discuss the relalionshi ive, in General Knowledge and Poul Facts = 35, between the scemingly contradictory goals of scholarship: discovering, {general knowledge and learning about particular facts. We are then Able to explain in more detail the concept of inference in section 2.2. (Our approsch in the remainder of the book is fo present ideas both verbally and through very simple algebraic models of research. In section 23 we consider the nature of these models. We then discuss ‘models for dala collection, for summarizing historical deal, and for descriptive inference in sections 24 25, anu 26, respectively. Finally, ‘we provide some specific erteria for judging descriptive inferences in section 27, 21 Gewenat. KNOWLEDGE AND PARTICULAR Facts ‘The world that social scientists study is made up of particulars: indi- Vidal voters, particular government agencies, specifi cites, tribes, ‘groups slates, provinces, and nations. Good social acne attempts to Bo beyond these particulars to more general knowledge. Generliza: tion, however, does not eliminate the importance ofthe paticulay, In fact the very purpose of moving from the particular tothe general i 'o improve our Understanding of both, The specific enlltes of the social world—or, more precisely, specific facts about these entities — provide the basis on which generalzations must rest. In addition, we, Almost always lea more about a specific cae by studying more gene «eral conclusions, If we wish to know why the foreign minister of Brazil resigned, it will help to lean why other ministers resigned in Brazil, ny foreign ministers in other countries have resigned, or why people in general resign from government or even nongovernmental jobs, Each of these will help us understand differen types of general facts and principles of human behavior, but they are very important even if ‘our one and only gol isto understand why the most recent Beailian foreign minister resigned. For example, by studying other ministers, ‘we might leash that all the ministers in Brazil resigned to protest the actions ofthe president, something we might not have realized by ex- mining only the actions ofthe foreign minster ‘Some socal science research tres to say something about a class of ‘events oF units without saying anything n particular about a speciic ‘event or unit. Studies of voting behavior using mass surveys explain the voting decisions of people in general, not the vote of any particular Individual. Studies of congressional finance explain the elle of ‘money on electoral outcomes across all congressional districts. Most ‘such studies would not mention the Seventh Congressional District in Pennsylvania or any other district except, perhaps, in passing oF 38 exceptions to a general rule. These studies fllow the injunction of 36+ Desrptivenfrence Praeworski and Teune (1982) eliminate proper names. However, though these studios may not seek to understand any particular di {cic they should not ignore—as sometimes is unfortunately done in this traction —the requirement thatthe facts about the various dis- tres tht go into the general analysis must be accurale, Other research ties 10 tell us something about a particular in- stance It focuses on the French Revolution or some other “important fevent and attempts to provide an explanation of how or why that ‘event came about. Research in this tradition woul be unthinkable fxrtainly uninteresting to most ofthe usa render of sich recon — Without proper names. A pois scientist may write effectively aunt tems .of relationships across the set of congressional campaigns Without looking at specific distil or specifi enndidates but imagine Robert Car's discussion (1983) ofthe 1948 Senate race in Texas with ‘ut Lyndon Johnson and Coke Stevenson. Particular events such as the French Revolution or the Democratic Senate primary in Texas In 1948 may indeed be of intensie interest they pique our curiosity and if they were preconditions for subsequent events (uch as the Napole- ‘onic Wars or Johnsons presidency), we may need to now about them to understand those later events: Moreover, knowledge about revel tion, rebellion or civil war in general wil provide invaluable informa: tion for any more focused study of the causes of the French Revolution in part We will consider these issues by discussing “interpretation,” @ ‘aimed alternative to scientific inference (ection 2.11, the concepts “of uniqueness and complexity of the subject of study (section 2.12); and the general area of comparative casestudies (section 21), 24 “Interpretation” and Inference In the human sciences, some historical and anthropological resazch- es claim to seck only specific knowledge through what they ell “i terpretation” lnterpeetivists seck accurate summaries of historical de- tal They also seek to place the events they describe in an intelligible ‘context within which the meaning of actions becomes explicable, As Fergjohn (in Goldstein and Keohane 1993:228) has written, “We want Nr cam we dss Caras someone in another sine a oul /togaper whos ol irs roma fh sc act a wor nner me I ht polit meta woul: Wht la ost fanaa ‘Simple Wat thao many aml amp aon carl oat Ws ‘tao cimpalgn cotton The canes aro parr edn ts fete te sutra sos oa wep wh snd fatten General Knowlalge ad Particular Facts = 37 social science theories to provide causal explanations of events {anal to give an account of the easons for or meanings of social action. ‘We want to know not only what caused the agent to perform some act but also the agent's reasons for taking the ation.” also writes that “it is notin our interest (o bleach human behavior of ihe very properties that interest us before we begin to examine i ‘Scholars who emphasize “interpretation” seck to illuminate the fy tentional aspects of human behavior by employing Verte Cem. phathy: understanding the meaning of actions and interactions frm the members own points of view" [Eckstein 197581. Inerpretivss, sock to explain the reasons fori ‘whole et of eoncepts and practices in which it isembedded, They also ‘employ standards of evaluation: “The must obvious tandads are o> hherenee and scope: an intrpecative acount shoul provide maximal coherence or intelligibility 10a st of social practices, and an interpre- itive account of a partic set of practices shouldbe eonaatnt with ther practices or traditions ofthe society” (Moon 1975: 173). Perhaps the single most important operational recommendation of {he interpretvists that researchers should Jen a great deal about culture prior to formulating research questions. For only with a deep, ‘cultural immersion and understanding of a subject can a researcher ask the right questions and formate useful hypotheses, For example, Duncier (1993) studied the collective fe of workingrclass black and White men at one integrated eafeteria in Chicago. By immersing him- ein this local eulture fo four years, he notices! several puzzles that had not previously occurred to him. For example, he observed that although these men were highly antagonistic i the Republican party, they articulated socially conservative positions on many issues. ‘Some scholars push the ole of interpretation even further, going 0 far as to suggest that i isa wholly diferent paradigm of inquiry for the social sctenees, "not an experimental science In search of law but an Interprative one tn search of meaning” (Certs 19733). In our view, however science (s we have defined it in section 1.1.2) and interpreta. lion are not fundamentally diferent endeavors aimed at divergent _gonle Both rely on preparing earful descriptions, galning deep under Standings ofthe world, asking good questions, formulating fasiable hypotheses on the basis of more general theotes, and collecting the evidence needed to evaluate those hypotheses, The dstnctive contei= bation of science i to presenta st of procedures for discovering the ‘sine to appropriately framed descriptive and eausal questions. ‘Our emphasis on the methodology of inference isnot Intend to denigrate the significance of the proces by which feutial questions fare formulated. On the conleary, we agrve with the ierpeaivss that 38 + Desiptve Inference itis crucial to understand a culture deeply before formulating hypoth- ‘sts or designing a systematic research projet to find an answer, We ‘only wish toad that evaluating the veracy of claims based on meth- ‘ds suchas participant observation can only be accomplished through the logic of scientific inference, which we describe. Finding the right, Answers tothe wrong questions is futile activity. Interpretation based fon Version Is often a rich source of insghtfal hypothests, Por in- Stance, Richard Fenno's close observations of Congress (Fenno 1978), ‘made through what he calls “soaking and poking,” have made major contributions to the study ofthat institution, particularly by helping to frame better questions for research. “Soaking. and poking.” says Putnam ina study of Valin regions (19:12), "requires the researcher to marinate herself inthe minutiae of an instituion—to experience ts ‘usloms and practices, is successes and its failings as thse who live itevery day do. Ths immersion sharpens our intuitions and provides Jnnumerable clues about how the institation fits together and how i adapts to its environment.” Any definition of sfence that dacs not i- ‘ude room for ideas regarding the generation of hypotheses as fol. ‘sh as an interpretive account that does not eate about discovering truth. ‘Yet once hypotheses have been formulated, demonstrating their cor rectness (with an estimate of uncertainty) roquites valid scientific i ferences, The procedures fr inference followed by interpetivist soca scientists, furthermore, must incorporate the same standards as those followed by other qualitative and quantitative researchers. That is, while agroeing that good socal science requires insighful interpreta tion or other methods of generating good hypotheses, we also insist that scence i essential for accurate interpretation, If we could under, stand human behavior only through Verstcen, we would never be abe o falsity our descriptive hypotheses or provide evidence for them beyond our experience. Our conclusions would never go beyond Ihe satus of untested hypotheses, and our interpretations would remain personal rather than scientific ‘One ofthe best and most famous examples inthe interpretative tr- dition is Ciilford Gecrte’s analysis of Gilbert Ryle’ discussion ofthe dliference between a twitch and a wink. Geert (1973:6) writes Consider ..two boys rapidly contracting the eyelid ftir right ees In ‘one tis isan involuntary tie; in the other, 9 canepiraral sigal toa {end The two movers are 35 movements, Sen from an am ‘amen, *phenomenalsic® observation af them alone, one could not tll Wich was witch and which was wink, or indeed whether bath or ether was titch oF wink. Yo the difference, however unphotographable, bee General Knowledge and Partcuae Facts + 99 ‘ween a trlch and a wink i as: a3 anyone unfortunate enough to have ‘nd the ist taken forthe second knows The winker bs commueating ed Indeed communicating in a presse and special way) deliberately.) 10 someone in partic, (to imprt a particular message, (0) acconfng toa socially established cod, and () wthot cognizance of the tf Ue eon ny. As ye points ut the wink las dane two ting, contracted is ‘yeti and snk, wil the ticher haw dane only one contd is ‘eyelids. Contracting your eyelids om purpose when thre crits» pubic ‘ede in which doing so counts a 9 conopitaorial signal i winking Geertz is making an important conceptual point. Without the con- ‘opt of “winking” given meaning, by a theory of communication, the ‘mast precise quantitative study of “eyelid contracting by human be- ings” Would be meaningless for stuxlens of social relations In this ex ample, the theory, which emerged from months of “soaking and pok; Ing” andl detailed cultural study, is exzential tothe proper question of ‘whether eyelid contraction even could be “witches” or “Winks” The ‘magnificent importance of interpretation suggested by this example is cleat it provides new ways of looking at the warld—new concepts to be considered and hypotheses to be evaluated. Without deep immer- sion ina situation, we might not even think of the right theories 16. evaluate n the present example if we did not think of the difference between Twiches and winks, everything would be lost, If interpreta: mor anything else—helps us arrive at new concepts oF hypothe ses, then it unquestionably useful, and interpretation, and simila forms of detailed cultural understanding, have been proven again and again “laving made relevant theoretical distinction, such as that between ‘wink and a twitch, the researcher then needs lo esluae the hypothe. sis that winking is taking place Iti in such evaluation that the logic of Selenite inference is unaurpaseed, Tati, the best Way of delete ing the meaning of eyelid contractions is through the ystematie meth ‘ds described in this book, If distinguishing a twitch from wink were Pivotal, we could easly design a esearch procedure to do 80, I for Instance, we believe that particular eyelid contractions are wink int: bcd with political meaning, then other similar instances must also be observable, since a sophisticated signaling device such as this (a “public code", nce developed, Is ikely wo be used again Given this Uikeihood, we might record every instance in which thi actor's eyelid ‘contracts, observe whether the other key actor is looking a the right time, and whether he responds. We could even design a series of ex- periments to see if individuals in Us culture are accustomed to com> ‘municatng in this fashion. Understanding the culture, carefully de- 40 ~ Desciptive Inference scribing the event, and having a deep familiarity with similar situa ions will all help us ask the right questions and even give us ad tional confidence in ur conclusions. But only with the methods of si ‘enlfc Inference will we be able to evaluate the hypothesis and see whether ts corec, Geer’ wink interpretation is best expressed asa causal hypothesis (ovhich we define precisely in section 3.1: the hypothetical causal ef- fect ofthe wink on the oer political actor i the other actor's response given the eyelid contraction minus his response If there were no move ‘ment and no other changes). I the eyelid contraction were a wink, the ‘causa effect would be positive; If were only twitch, the causal ef ect would be zero. If we decided to estimate this causal effect (and thus find out whether it was a wink ora twit, all the problems of Inference discussed at lng in the rest ofthis book would need to be understood if we were to arrive atthe best inference with respect to the interpretation ofthe observed behavior what we interpret as winks were actully involuntary witches, ‘our atlempts to derive causal inferences about eyelid contraction on ine basis of a theory of voluntary socal interaction would be rou. tinely unsuccessful we would not be able to generalize and we woulkd know it? Designing research to distinguish winks and twitches isnot likely to be a major part of most political scence research, but the same ‘methodologilisue arises in much of the subject area in which polit ial scientists work. We ae often called on to imerpret the mic an at Foreign policy decision makers send messages to eachother Is 8 particular message a threat, a negotiating point, a statement aimed 2 appealing o a domestic audience? Knovidge of cultural norms, of ‘conventions in international communications, and of the history of Particular actors, as well as close abservation of ancillary features of {the communication, wil all help us make such an interpretation, Or consider the folowing puzzle in qualitative research: Voters in the United States seem to be sending, a message by not turning out at the Polls. But what does the low turnout mean? Does it reflec alienation With the poltical system? calculation of the costs and benefits of vo ing with the costs being greater? Disappointment with recent eandi- dates or ecent campaigns? Could it be 9 consequence ofa change in ‘the minimum age of voting? Or sigh that nothing is sulicienly up- Fart she camps wath noting ht we cold nage a og eat ory wich see contin vi sk at il ea ‘ton ther aco Fer eam, etch sul be en msl oe ‘ws nes wheter he poo with th cy tation a ok, ‘nk ede fr tr ere comp th se. General Knowledge and Particular Facts + 41 salting to get them tothe polls? The decision of a citizen not to vot like a wink or a diplomatic message, can mean many things. The 90. Disticatd researcher should always work hard to ask the right ques ions and then carefully design scientific research to find out what the ambiguous act did in fact mene, ‘We would also like to briefly addres the exireme claims of afew Proponents of interpretation who argue that the goal of some research ‘ugh to be feelings an wo observable consequences, Th ‘but a small minority of re we claims are made suliciently force- that they seem worth addressing explicitly Like the overenthu- siastic claims of early pasiivists, who took the Untenable postion that "unobservable concepts had no place in scientific esare these ang ‘ments turn out to be inappropriate for empirical nacarch. For exam: ple, Psathas (1968510) angus that . ‘any behavior by focusing only on that prt whic sovet and niet i ‘concrete, dr observable acts is maiv, to sty the let The challange 12 {he socal cei who scks to understand socal wait, then isto under. anal the meaning tht the actors at hs hi. Psathas may be correct that social scientists who focuson only overt, olserenbe, behaviors are missing a lt, but how are we to know if we ‘cannot see? For example if two theories of sell-conception have dent «al observable manifestations, then no observer wil have sufficient in formation to distinguish the two. This is true no malter how clever oF culturally sensitive the observer i, how skilled she is at interpretation, how well she “brackets” her own presuppositions, or how hand se tuk. Interpretation, feeling, thick desertion, participant observation, nonpartcipant observation, depth interviewing, empathy, quantifies. tion and statistieal analysis, and all other procedures ant methods ro Inadequate tothe task of distinguishing two theories without differing ‘observable consequences. On the other han, i the two theories have some observable manifestations that differ, then the methods we de acribe in this book provide ways to distinguish between them. In practice, ethnographers (and all other good socal scientist) do look for observable behavior in orde to distinguish among thei theo Fie. They may immerse themselves inthe culture, but they all rely on various forms of obseration. Any further “understanding” ofthe cule {ural context comes directly from these or ether comparable observa- tions. Identifying relevant observations isnot always easy. On the cn teary, fining the appropriate observations is perhaps the most dif Part of a research project, especially and necessarily) for those, Breas of inquiry traditionally dominated by qualitative research, 42 + Descriptive Inference 212 “Unitas” Coup. Sinpliftin ‘Some qualitatively oriented researchers would reject the position that, general knowledge is either ncessary or useful (perhaps even posi ble) as the basis fr understanding a particular event, Their that the events or units they study are “unique.” In one sense, Fight. There was only one French Revolution and there is only one ‘Thailand, And no one who has read the biographical accounts or who lived through the 1960s ean doubt the fact that there was only one Lyndon B. Johnson, Dut they go further. Explanation, according to {heir position, is limited to that unique event or unit: nat why revolu- tions happen, but why the French Revolution happened; not why de- ‘mocratization sometimes sets to lg, bul why It lags in Thalland; not in 1948 oF 1964. Researchers in specifi they attempted to deal withthe general oil evolutions or

    You might also like