1Class 5Causality in Social ScienceClass Outline• Causality in Social Science• Causal Diagrams• Spurious Causation• Intervening and Confounding VariablesNomothetic Explanation• Idiographic - Seeks to fully understand the causes of what happened in a single instance. Deterministic.Prevalent in case studies and everyday life. • Nomothetic - Seeks to explain a class of situations or events rather than a single one. Probabilistic. Prevalent in social sciences.2Causality in Probability• Classical notion of causality in natural sciences (before quantum mechanics) is deterministic. – If X happens, then Y happens.• Causality in social science is probabilistic.– If X happens, then Y will happen with probability p. X raises the probability of Y.– Example: smoking and lung cancer. Necessary and Sufficient Causes• Necessary cause represents a condition that must be present for the effect to follow.• Sufficient cause represents a condition that if present, guarantees the effect in question.• Causes that are necessary and sufficient are the most satisfying outcome in research.Implications of Probabilistic Causality• Importance of multiple observations (sample size N)– The law of large numbers: as sample size increases, the sample mean approaches the population mean.• Importance of replications– Replications: Repetitions of a study using the same research methods to answer the same research question– Meta-analysis: Using statistical techniques to synthesize past study results.3Example of Meta-AnalysisSource: Hedges, Larry and Amy Nowell. 1995. “Sex Differences in Mental Test Scores, Variability, and Numbers of High-Scoring Individuals.” Science. Vol. 269(7), pp. 41-5.Variability in Response to the Same Stimuli• Variability is the essence of social science.– Population variability (i.e., variability across different individuals) – Contextual variability (i.e., variability across time and space)• The same stimuli often lead to different responses.• Compare this to natural scienceCriteria for Probabilistic Causality• A statistical association between the two variables.• The cause takes place before the effect.• There is no third variable that can explain away the observed correlation as spurious.4Positive Association Example:Education and IncomeEarnings05000100001500020000250003000035000400004500050000Less than HS High School Some College College Grads Master and Ph.D.Causal Diagram: Two VariablesDrinkingEarningsIndependent variableDependent variableExamples:Water pollutionCholera• Identify independent variables and type of association.• How to state a hypothesis.Sex Verbal abilitySpuriousnesswhen in fact the correlation is induced only by a common cause age:Shoe sizeMath knowledgeAgeShoe sizeMath knowledgeWe observe a correlation:Shoe size Math knowledgeand we claim causation:5Simpson’s Paradox30%183545%2691% admittedapplicants% admittedapplicantsFemaleMalegender admissionAre female students discriminated against in UC Berkeley’s admission process? ?Simpson’s Paradox30%183545%2691total7%3416%373F24%39328%191E35%37533%417D34%59337%325C68%2563%560B82%10862%825A% admittedapplicants% admittedapplicantsMajorFemaleMalegenderadmissionmajorAfter controlling for major, there is no effect of gender on admission.Simpson’s Paradox• Simpson’s Paradox refers to the reversal of the direction of a comparison or an association when data from several groups are combined to form a single group.• Lurking variable – major• Another example: the electoral college6Causal Diagram: InterveningSex admissionmajorExample:Independent variableDependent variableIntervening variableCausal Diagram: ConfoundingShoe size Math knowledgeAgeExample:Independent variableDependent variableConfounding variableNo Booze? You May Lose:Why Drinkers Earn More Money ThanNondrinkers• A number of theorists assume that drinking has harmful economic effects, but data show that drinking and earnings are positively correlated. We hypothesize that drinking leads to higher earnings by increasing social capital. If drinkers have larger social networks, their earnings should increase. Examining the General Social Survey, we find that self-reported drinkers earn 10-14 percent more than abstainers, which replicates results from other data sets. We then attempt to differentiate between social and nonsocial drinking by comparing the earnings of those who frequent bars at least once per month and those who do not. We find that males who frequent bars at least once per month earn an additional 7 percent on top of the 10 percent drinkers’ premium. These results suggest that social drinking leads to increased social
View Full Document
Unlocking...