Goals Identify General Patterns and Relationships Testing Theories Making Predictions
Principles of Quantitative Research Generality (relationship holds across a wide range of different cases of a population) Parsimony (as few variables as possible should explain as much as possible) Direct contrast to qualitative but also comparative research
Example: study on income difference in US (how to become rich?) Quantitative approach: search for individual level attribute with strongest relation to income level (race, education, income of father etc.). Assumption: all cases are more less the same, different sides of the same coin Comparative approach: “how different causes combine in complex and sometimes contradictory ways to produce different outcomes” (Ragin 1994: 138). Identify diverse ways of succeeding and of failing to achieve material success. It’s not about which variables covary better, but which different pathways do exist!
Process Framing: clear definition of population of cases, dependent variable and independent variables (deduced from theory) Operationalization: transform concepts/variables into measurable indicators/indices Data Collection: conduct survey, use secondary sources Data Analysis: search for covariation Interpretation: do covariations found, support theory? do covariations found, mean causality (which direction?)? Ragin: construction of images, relate image to theory/idea
Terms Variable: general features that may differ (vary) from one case to the next Indicator: instrument to measure variable Index: a single number calculated from different indicators
2. Operationalization & measuring variables Example: study on job satisfaction (theory testing) Theory: people are happy, when they can do what they are good at Cases: employees (specify) Dependent variable: job satisfaction Independent variable: match between skills of employee and job characteristics
Job Satisfaction INDICATORS: Survey of employees (ratings) Absenteeism Ask supervisors (performance) Pro vs Con Level of measurement (yes/no, 1-5, 1-100) (nominal, ordinal, interval, ratio)
Validity & Reliability Validity: appropriateness of a measure (indicator) –Does it measure what it is supposed to measure? Reliability: how much randomness is in a measure? –Will repetition of measurement bring the same result?
Defining variables Operationalizing them Defining your levels of measurement Most important steps for quantitative research. They decide about the quality of the whole research. Once this is done, the researcher has not much influence on the study (Collecting & processing data is very technical)
3. Examining correlations Positive correlation: high values ind. V. and high values dep. V.; low values ind. V. and low values dep. V. Ex: unemployment – crime rate Negative correlation: low values ind. V. and high values dep. V; high values ind. V. and low values dep. V. Ex: bureaucracy – job satisfaction
Degree of Correlation Correlation not in all cases (of thousands) to be found. Does not matter Correlation Coefficient –To what degree do variables correlate –Pearson‘s r –varies between and perfect negative correlation perfect positive correlation 0.0 no correlation at all
Correlation ≠ Causation Correlation might be coincidence –If we find shoe size and income covary? Correlation might be effect of something else (3rd variable that effects both). –Positive correlation: sun protector / skin cancer (both high because strength of sun) Direction of causation might be unclear – education levels / GNP
Correlation ≠ Causation Correlation + Theory suggests causation
4. Using correlation coefficients Ex: how to explain life expectancy Correlation coefficients can help to compare competing causes (how much variation can be explained by the ind. variable a, how much by b?)
Dep VariableIndependent Variable Life expectancy Calorie consumption GNP / capita Doctors / capita Life expectancy Calorie consumption GNP per capita Doctors per capita
causes of higher life expectancy Nuitrition seems to be most important. Calorie consumption highest correlation (.802) GNP per capita seems also to play a role (.651). But not independent of calories (.848 correlation). GNP effects via calories Doctors/capita (.721) quite high + relatively independent from calories (.321)