230 likes | 244 Views
This course provides an overview of research questions, their construction, and the relevance of causality in social science research. Topics covered include different types of research questions, the construction of hypotheses, and the centrality of causality in providing a better understanding of social phenomena.
E N D
Research question, Causality and hypotheses PhD Course in Research Design Roskilde University Spring 2018
Overview • The introduction – to yourPhD • The research question • What it is • Whyyouneed it • How youconstruct it • Causality • Why it is so relevant for social science • Twotypes => • Different types of Research Questions • Hypotheses • Whyyoumay (or may not) wantthem • Types • How youconstructthem
The Introduction I • Explaining the dissertation’s research problem / question • Reviewing the literatureabout the research problem • Identifyingdeficiences in the literature
The Introduction II • Explaining the relevance of the research problem: • For whom, for whatis this problem relevant? • Academic, administration, society.. • Account for the purpose, e.g: • How do youintend to answer the problem? AND/OR • What is your argument / what do youintend to show? • Account for youconclusion: • Whatwereyourresults? The answer to yourquestion?
Whyyouneed a research question • To guide youentirePhD dissertation • The choice of theory • The choice of methods • The design of your (emprical) analysis • How youanalyzeyour data • Whatyoucanpossiblyconclude • EVERYTHING IS STRUCTURED BY YOUR RQ (and to someextentyour IQ..)
What is a research question? • A puzzling and answerablequestionaboutyour research problem = YOUR CONTRIBUTION! • It must constitute a puzzle • Arousingyourcuriosity and interest • Is must be non-trivial – to you and manyotheracademics • It must beanswearable • What is possiblewithin the scope of yourPhDproject? • Be realisticand listen to your supervisor • Yourcontribution is the sum of 1 and 2 • MAKE SURE YOUR PASS THE CONTRIBUTION TEST: • Can youexplain in threesentenceswhatyourcontribution is?
Exercise • In groups of two students, revisit and discuss whether the PhD Research Design text you submitted for the course clearly defines and describes a good research problem • Write around half a page each on any revisions that might be needed to clarify the research problem in your project
How do youconstruct the RQ? I • Useyourexistingknowledgeabout society • Economic, politics, social issues, etc.=> • The societalrelevance argument • Useyourliteraturereview • The academicgap argument • Contradict the literature! • Resonance with yourtheory • Most RQsonly go with a limited set of theories => • The theoryinforms, but does not determineyour RQ • The theoryinformsyouranalysis and possibleconclusions • BUT the theory must NEVER determineyouranswer
How do youconstruct the RQ? II- Contradicting the literature • Developmental / historicalcontradictions • You can claim that others have mistaken the origin, development, or history of your object of study, e.g.: • Some argue that EU immigration is rising for the first time, this has happenedbefore • Cause-effectcontradictions • You can claim that assumed causal relationships do not exist (or vice versa), e.g.: • A new way to stop juveniles from becoming criminals is the “boot camp” concept. But evidence suggests that it does little good. • Contradicitions of perspective • These contradictions run deeper by suggesting to look at things in a new way, e.g.: • It has been assumed that advertising is best understood as a purely economic function, but it has (also) served as a laboratory for new art forms and styles.
How do youconstruct the RQ? III • Resonance with yourmethod • Again, most RQsonly go with certainmethods • Start with the RQ, thenchoosemethod(s) • Think of the Quant / Qualdivide, but allow for mixed methods • An on-goingendeavour • Expect to re(de)fine it throughout the firsthalf of yourPhDproject • But beware of the dangers of strayingtoo far – in particular if youaretrying to answer a quant type of question • Be ambitious AND humble! • Youwant to beable to answeryourquestion in depth • Don’texpect to beable to answerunanswerablequestions (leavethat to journalists and priests)
Causality: Its centrality to social science • No social science research only seeks to describe an event for the sake of description • Allsocial science research seeks to provide a better understanding of a certain phenomenon • E.g. by asking HOW did this event take place? (by what arguments, practices, circumstances, etc.) • Even this descriptive task assumes that some circumstances are more important than others because they somehow interact with the phenomenon • Muchsocial science research seeks to explain • E.g. by asking WHY did this event happen or what was the consequences of this event? • But NO agreement within the social science: • What Explanation is / requires • If explanation requires identifying causal relations • What causality is – or what it takes to identify it
Causality: Two types • Qualititative vs. Quantitative • Of course this is just one typology! But it covers key differences between non-positivist and positivist approaches
Causality: Quantitative approaches • Effects-of-causes approach • E.g what is the effect of outsourcing the unemployment services? • Case-selection on independent variables (causes) – ideally, • e.g. a “natural experiment”, such as the Danish municipal reform • Correlational causes (probabilistic understanding) • E.g. there is (no) statistically significant correlation between outsourcing of unemployment services and people getting into jobs • Additive causation (all factors weighed and added), • E.g. 1/0 outsourcing + X*costs for services + Y*number of employees + Z*level of employee experience + … • Non-fit/systematic treated as errors or outlines (=ignored!) => substantively important cases often ignored
Causality: Qualitative approaches • Causes-of-effect (often individual cases = one effect), • e.g. the Russian revolution • Case selection often on dependent variable (effect) • Necessary and sufficient causes (deterministic conception) • E.g. Marx’ theory of capitalism necessary for Russian revolution • Clusters of factors amount to a cause (set theory logic) • E.g. Marx theory AND feudal regime AND poverty AND WW1 = sufficient causes of Russian Revolution • Lack of fit: • Non-conforming (untypical) cases are often more interesting than the typical ones, e.g. Foucault’s analysis of crime and madness
Exercise • In groups of two: • Discuss if/how you intend to address causality in you PhD project • If you do not intend to do causal analysis, then discuss what you do instead to provide a better explanation / understanding of your RQ
Types of research questions I • Overall purpose of question • Descriptive (what and how) vs. Explanatory (why) • Normative (good/bad) vs. Explanatory • In much of social science today, youoftenneed to justify the absence of whyquestion(s) • Types of explanatoryquestions • IntersubjectiveUnderstanding vs. ObjectivistExplanation • Causal processes and relations vs. Statistical correlations • No methodmonopoly on explanatoryRQs: • Explanatoryquestionscanbeansweredboth by qualitative and by quantitativemethods
Types of research questions II- Qualitative • Typicalquestons: • Whathappened and/or how? • Explanation by way of: • Describingactors’ actions, explanations, justification, etc. • Describingwider (collective) ideas, logics or rationalitiesthatseveralactorsrefer to whenjustifyingtheir actions • Followingconcrete actions, events and processes • Document relationshipsbetween statements, ways of thinking, actions, decisions, routines • Explanation of actions is the result of suchmappings / descriptions
Types of research questions III- Quantitative • Typicalquestion: • What is the relationshipbetween variable X and variable Y? • Explanation by way of: • Randomizedcontrolledtrials OR • Statistical correlation • NOT by mapping the social practicesconstituting the causal relations – unless mixed method => • Beware of conflatingcorrelation with causality • Explanation by experimental or statistical proxy: • The calculated and controlledcorrelationsbetween abstract objectification of two or more sets of social practices
Whyyoumaywant to formulatehypotheses • Hypothesis: • A falsifiablepostulateabout a social practiceor causalrelationship, e.g.: • High personalincomelevelsareconducive to voluntarywork • May beuseful to guide your research, i.e. a furtheroperationalization of your RQ • Onlymakehypothesis if: • Youintend to examine / test them • Theyactuallyhelpyouoperationalizeyour RQ
How do youconstructyourhypotheses? • Deductive: • Based on theory • Something sounds wrong, letssee if it holds.. • Something sounds right, but does it alsoapplyhere? • This theory is veryvague, mycontribution is to turn (parts of) it intotestable propositions • Inductive (or Ab-/Retroductive): • Based on eithertheory, yourpreconceptions or the (lay) understandingsfound in the fieldyouexamine => • Hypothesesformulatedbefore OR during data collection => • Not usedveryoften in the strict test version, but more as a workinghypothesisguidingyour research
Types of hypotheses • Test hypotheses vs. Workinghypotheses • Test hypothesis • An empiricallyfalsifiableclaimabout reality usually but not alwaysabout a causalrelationship, e.g. • Test Hypothesis= CV writingcontributes to getting more people in job • Falsification/confirmation is fundamental (= RQ answer) • Usuallylinked to quantitativemethodsand the use of regression analysis • Workinghypothesis or Assumption(hypothesis a loaded term): • Assumptionthat guides you research, e.g.: • Workinghypothesis: Soft forms of power play an influentialrole in contemporaryDanish employmentpolicy => • I examinehowsoft power is exercised in DK employment policy (+ why + with whatconsequences..) • Falsification/Confirmationmaybepossible, but is not the keyobjective
Types of (test) hypotheses • Nullhypothesis • Assumesnorelationshipbetweentwo or more variables • Directionalhypothesis • Assumes a directionalrelationshipbetweentwo or more variables, e.g. X => Y • Precondition: It is possible to distinguishbetween independent (X) and dependent variable (Y) – if not, thentautology • Non-directionalhypothesis • Assumesthat the presence of X has an influence on Y, but not howPrecondition: • Possible to distinguishbetween variable X and Y, but • Not the direction of theirinterdependency
Summing up • Research Question • Mandatory • Your basic tool for guiding your entire project • Needs to be revised on a regular basis – though not too much as you approach project deadline! • Causality • Part and parcel of most social science, but there are very different ways of approach causality => • Be explicit about how you do it • Hypotheses • Optional • May help you sharpen you RQ and your design