Research Design and Methods
Causal Inference What is causal inference “…learning about CAUSAL effects from the data observed.” (KKV, 8) Different from descriptive, philosophical, normative, prescriptive approaches Causal inference is uncertain What evidence would increase confidence in causal argument
Selecting a Research Question International Environmental Policy question that interests you Question that lends itself to causal analysis Question that is of policy and/or theoretical significance Think whether and how you could measure your dependent and explanatory/independent variables
What Is a Variable? Variable is a measure of something that we are interested in. It varies in certain, measurable ways. Something which can have at least two values. Dependent variable = effect Independent variable = cause Not always clear which direction causality runs. Sometimes issues of simultaneity or endogeneity.
Hypotheses Probabilistic proposition about how variables relate, keeping everything else constant Sources of hypotheses Theory Existing literature, case studies Logic
Methods Experimental methods – not widely used in social science Statistical methods Comparative (case study) method Case study method
Research Design Identify a puzzle, research question Identify theories that are relevant to that question. Often achieved through literature review. Have some preliminary hypotheses about how the variables of interest could correlate Identify observable implications of your hypotheses Select research and analysis method
Improving Case Selection Increase number of cases or observations (disaggregate across regions, across time, etc.) Control for as many variable as possible. Focus on comparable cases Select cases that similar in a lot of aspects, but vary along the variable of interest Be explicit on why this particular set of cases were selected and make a good comparison. Focus analysis on key variables, avoid “laundry list” type of explanations
Increasing Confidence in Causal Argument Observable implications -generate as many as possible, check against evidence Counterfactuals Could the causal relationship be spurious (both the d.v. and i.v are caused by a third variable affecting each simultaneously and are therefore unrelated) Distinguish correlation from causation.
Use of Data Make sure you have a “measure” of your dependent and explanatory variables Data and methods should be public, transparent, well documented Make use or at least show awareness of the full range of data sources on a particular topic Ask whether the particular measure of the variable is driving the results. Show awareness of a possible problem Be cautious when using Internet sources, do not rely only on internet sources.