國立臺灣大學政治科學論叢

Abstract / Chinese PDF Download

In  the  last  few  years,  empirical  researches  in  political  science  have  oftenneglected  to  systematically  examine  their  measurement  indices  and  methods.Concerning such a general neglect, this paper uses the large-scale surveys of TEDSconducted  in  2003  and  2004  as  the  bases  for  analyzing  a  number  of  possibleproblems  related  to  the  reliability  of  the  measurements  of  political  attitudes  andbehavioral scales. The degree of reliability can be measured by performing a test-retest  and  then  examining  the  result  for  the  extent  of  internal  consistency.  Test-retest is carried out with the Kappa coefficient and the weighted Kappa coefficientwith  quadratic  weighting.  Internal  consistency  is  tested  with  the  widely  usedCronbach’s  α  coefficient  and  the  Guttman  formula  of  split-half  reliability.  Usingthese four measures to examine the two TEDS surveys in 2003 and 2004, we findthat, on the whole, the two surveys pass our tests for reliability. It is found that anumber  of  previously  considered  very  sensitive  political  topics,  such  as  votersupport,   ethnic   identity,   Taiwan   independence,   and   party   identification,   allrelatively reveal stable conditions of results. We also find that the most importantmeasures  of  political  attitude  display  a  high  degree  of  internal  consistency.  Inaddition  to  the  stability  found  in  the  acquired  results,  our  tests  also  reveal  someunexpected results on a small number of items, which may require further revisionin our future research.