Capturing causation in political science : The perspective of research design