WHY SOCIAL SCIENCES ARE NOT SCIENTIFIC ENOUGH
Taagepera, 2008
Chiara Francavilla 27th September 2010
AIMS
Highlight problems in social sciences studies Provide suggestions
PROBLEMS
Overuse and misuse of statistical analysis
In particular linear regression
PROBLEMS (CONT.)
Prevalence of Descriptive models rather than Predictive
Reversed role of scientists and statisticians
SUGGESTIONS
A better balance of methods, more model building
Directional vs quantitative prediction
dx\dt ‹0 x=f(x)
SUGGESTION (CONT.)
Quantitatively predictive logical methods More than directional Based on logical considerations Knowledge of conditions under which the model can/cannot
apply Introducible without numerical data
SUGGESTIONS (CONT.)
How to construct predictive logical models? In general: consider broad conceptual constrains and specific
problems
Guidelines: Parsimony Absurdity avoidance:
Contradiction avoidance
SUGGESTIONS (CONT.)
The more the merrier or the Occam’s razor? Statisticians’ vs scientists’ goals
Interdependent variables
REFERENCES
Taagepera, 2008. ‘Why social sciences are not scientific enough’ and ‘How to construct predictive models: Simplicity and Nonabsurdity’ in Making Social Sciences More Scientific, Oxford: Oxford UP
Fearon and Laitin, 2003. ‘Ethnicity, Insurgency and Civil War’. American Political Science Review, 97(1)
QUESTIONS
1. Can we carry out researches in the social science field with the same instruments and methods used in the natural science one?
2. In referring to the Occam’s razor principle, how can we determine in advance which considerations are not essential?
3. The construction of a logical model before the collection of the data might influence the latter and lead to biased results. Discuss
LINEAR REGRESSION
DESCRIPTIVE VS PREDICTIVE MODEL
Main question
Nature Core method Mathematical format
Direct output Indirect output
How? Descriptive Statistical data analysis
Generic Statistical
Nonfalsifiable postdiction
Limited scope postdiction-based prediction
Why? Explanatory Logical consideration
Subject-specific conceptualization
Prediction falsifiable upon testing
Broader substantiated prediction
Taagepera, 2008:7
EXAMPLE: DESCRIPTIVE MODEL
Number and Percentage of Countries with Ongoing Civil Wars by Year from 1945 to 1999 (Fearon and Laitin, 2003)
EXAMPLE: PREDICTIVE MODEL
Probability of Civil War Onset per Five-Year Period (Fearon and Laitin, 2003)
DIRECTIONAL VS QUANTITATIVE PREDICTIONTaagepera, 2008:26-27