Context and Electoral Studies

1
Context and Electoral Studies

• Michael Marsh
• Trinity College Dublin

2
Types of context effects 1

• Global effects, influencing all voters in a
  locality
• Eg different parties on ballot paper in different
  constituencies, or different rules
• Election itself as context in time series of
  election studies

3
Types of context effects 2

• Compositional effects (sometimes called
  contextual effects) where composition of local
  units varies in relation to key dependent or
  independent variables
• Eg importance of majority party in an area
  McLean and constant swing

4
Types of context effects 3

• Non equivalence of measures, giving apparent
  contextual effects
• Eg farmer might not mean the same thing in
  different areas
• Or a party itself may have a very different
  character in different places, or at different
  times

5
Why does it matter

• Missing variable bias
• False assumption of equivalence, and hence
  measurement error
• Context variables may help explain processes

6
How does it matter

• Compositional effects through
• Interpersonal communications, the classic
  neighbourhood effect
• Locality as a site of direct experience
• Selection biases, that appear as compositional
  effects but are not phantom contextual effects

7
Implications for research design

• Ask voters about their locality and local
  experience and contacts
• Supplement individual data with contextual data
• Allows explorations of regional effects in
  British elections are there regional differences
  if everything else is controlled
• Some explorations of neighbourhood effects and
  the processes which underlie them
• Tactical voting

8
Implications for research design 2

• Design study specifically to explore the
  mechanisms through which context matters
  (Huckfeldt and Sprague)
• Local site
• Geographical context data added
• Additional interviews with contacts of informants
• Context as communication networks
• Problematic to apply to full election study
• Parsimony - too many possible contacts
• Modest effects

9
Implications for research design 3

• Over-sample interesting contexts, as is typically
  done in cross national research e.g. European
  Election Studies, CSES
• Disproportionate stratified random sample with as
  many cases from Ireland as from Germany
• Add on extensive contextual data, and some
  questions tailored to specific context
• Helps show how far relationships conditional ( or
  not) on institutional and other differences
  between countries and helps explain differences
  tht are found

10
Conclusions

• Compositional effects interesting but small and
  generally should not be focus of election studies
• Global effects more promising, easier to deal
  with in national studies and more comparative
  data is becoming available
• But continuing problems of equivalence