S + M, Chapter 6

Research Design

Attributing Causation Through Control

  • Understanding causation is crucial for addressing social and political issues.

  • This chapter will cover:

    • The nature of causation in social science.

    • Importance of logical warrants in scientific research.

    • Key characteristics and benefits of experimental research.

    • Challenges and limitations of experimental research.

    • Common experimental methods used in political science.

The Nature of Causation in Social Science

  • Causation is a complex concept that requires careful examination:

    • Everyday examples of causation seem straightforward (e.g., a car not starting).

    • Scientific causation often needs inferential reasoning based on theory and observation.

    • Direct observation of causation is rare; inference based on evidence is essential.

    • Example: Proving a chemical causes cancer involves statistical comparisons and theoretical mechanisms of action.

Types of Causation

Probabilistic Causation

  • Unlike natural sciences, social sciences often observe probabilistic rather than absolute causation.

  • Example: Higher education correlates with political activity, but does not guarantee it on an individual level.

Multiple Causation

  • Most social phenomena are caused by multiple factors:

    • Political party identification can be influenced by background, peer pressure, and socio-economic status.

    • Recognition of multiple influences is vital for understanding complex social events.

Indirect Causation

  • Causation can occur through intervening variables:

    • Example: Racial prejudice may indirectly cause school segregation by creating economic divides, leading to residential segregation.

Establishing Causation

  • Ethical considerations limit claims of causation in social science:

    • Researchers must provide clear justification for causal claims.

  • Correlations do not imply causation:

    • Without ruling out other explanations, causation cannot be inferred.

Experimental Research Design

The Classic Experiment

  • The classic experimental model encompasses key components:

    • A hypothesis predicting a relationship between variables.

    • An experimental group exposed to an independent variable and a control group not exposed.

    • Random assignment of participants to minimize bias.

  • Pretests and posttests are critical for measuring changes.

Variations on Classic Experimental Design

Solomon Designs

  • Address reactivity and control variations:

    • Solomon two-control-group design removes the test effect from results.

    • Solomon three-control-group design includes a group with no pretest, allowing assessment of extraneous influences.

Randomization and Assigning Cases

  • Randomization is essential for creating comparable groups:

    • Ensures similar characteristics among participants, reducing bias.

  • If randomization is not possible:

    • Precision matching attempts to pair subjects with similar characteristics.

    • Frequency distribution control ensures groups share average characteristics.

Field Experiments

  • Conducted in natural settings with less control over extraneous factors:

    • Aim to assess real-world implications and effects of interventions.

    • Example: Assessment of voter turnout through different campaigning methods.

Quasi-Experimental Designs

  • Common in political science when random control is infeasible:

    • Ex post facto designs observe existing conditions for causal inference.

    • Time-series designs track metrics before and after an event to determine effects.

Challenges in Causal Research

  • External validity is a critical challenge; labs offer control but lack realism:

    • Ensure findings are applicable to real-world contexts.

    • Field experiments provide a compromise between control and realism, with potential ethical implications.

Conclusion

  • Experimental research is beneficial but complex, requiring a balance between control and real-world applicability.

Key Terms

  • causation

  • logical warrant

  • control

  • experimental design

  • experimental group

  • control group

  • pretest

  • posttest

  • test effect

  • Solomon designs

  • randomization

  • precision matching

  • field experiment

  • quasi-experimental designs

  • ex post facto experiment

  • time-series designs

  • regression toward the mean

  • controlled time-series designs

References

  • Blass, Thomas. 2000. Obedience to Authority.

  • Braman, Eileen. 2009. Law, Politics, and Perception.

  • Cook, Thomas D., and Donald T. Campbell. 1979. Quasi-Experimentation.

  • Green, Donald P., and Alan S. Gerber. 2004. Get Out the Vote!.