• No results found

Methods employed

In document Trojans in Wireless Sensor Networks (Page 79-88)

H7 Electoral system

3.3 Methods employed

This  dissertation  employs  a  mix  of  quantitative  (multivariate  linear  regression  and  factor  analy-­‐ sis)  and  qualitative  (case  study)  methods,  in  addition  to  data  collection  techniques  such  as  question-­‐ naires  and  interviews.  

Multivariate  linear  regression  analysis.    Bivariate  linear  regression  analysis  offers  an  estimate  of   the  change  in  the  dependent  variable  when  the  independent  variable  increases  by  one  unit  (Pollock   2006,  139).    For  example,  when  the  impact  of  country-­‐level  political  tolerance43  (tolerance  toward  the  

least-­‐liked  group)  on  democracy  is  analyzed,  the  b  coefficient  (i.e.,  the  slope  of  the  regression,  which                                                                                                                            

 

indicates  the  magnitude  of  the  aforementioned  effect)  is  1.7.    That  is,  for  every  one  unit  (standard  devi-­‐ ation)  increase  in  the  level  of  tolerance  towards  least  liked  groups,  the  model  predicts,  on  average,  a  1.7   units    increase  in  the  level  of  democracy  (operationalized  using  Voice  and  Accountability  scores).    So  if   the  average  political  tolerance  of  the  citizens  of  country  X  is  one  unit  higher  than  the  political  tolerance   of  the  citizens  of  country  Y,  then  country  X  will  be  more  democratic  with  1.7  units  on  the  Voice  and  Ac-­‐ countability  scale.  

It  is  seldom  the  case  in  social  science  to  be  able  to  offer  a  satisfactory  and  comprehensive  ac-­‐ count  for  variability  in  a  dependent  variable  in  terms  of  a  single  independent  variable,  and  the  above   example  is  no  exception.    What  is  also  known  is  that  tolerance  is  usually  positively  correlated  with  de-­‐ velopment  and,  in  turn,  development  is  positively  correlated  with  democracy.    That  being  the  case,  in   order  to  obtain  an  unbiased  estimate  of  the  impact  of  tolerance  on  democracy,  development  must  be   included  as  a  control  variable  in  a  multivariate  regression  model,  even  if  the  impact  of  development  is   not  an  object  of  interest.  

Indeed,  data  analysis  confirms  these  contentions.    The  two  independent  variables,  degree  of   tolerance  towards  least  liked  group  and  Human  Development  Index,  are  highly  correlated  (0.33),  and   both  are  also  highly  and  positively  correlated  with  the  dependent  variable  (0.42  and  0.72,  respectively).     When  both  independent  variables  are  included  as  predictors  for  democracy,  the  estimated  impact  of   tolerance  (partial  regression  coefficient),  which  describes  the  contribution  of  this  independent  variable   on  the  dependent  variable,  controlling  for  the  impact  of  development,  is  only  0.81.    This  indicates  that   half  of  the  (apparent)  impact  of  tolerance  in  the  original  bivariate  model  was  actually  due  to  develop-­‐ ment,  rather  than  tolerance.    Table  3.3  shows  the  relationship  among  the  three  variables  and  depicts   the  importance  of  introducing  control  variables.      

Table  3.3.  Bivariate  Correlations  

  Democracy   Tolerance  

Tolerance   .421*    

Development   .725**   .330*  

**Significant  at  p=0.01  level     *Significant  at  p=0.05  level   N=36  

 

The  above  example  has  important  substantive  implications  for  this  study.    For  instance,  Chapter   4  presents  a  multivariate  analysis  of  cross-­‐national  levels  of  democracy  as  a  function  of  the  average  lev-­‐ els  of  political  tolerance  and  social  tolerance  in  those  countries,  controlling  for  the  level  of  socioeconom-­‐ ic  development.  

Factor  analysis.    The  goal  of  factor  analysis  is  the  identification  of  underlying  dimensions  among   a  number  of  variables.    These  dimensions,  called  factors,  can  be  seen  as  “averages”  of  closely  related   variables  (Lijphart  1999,  245).    Often,  the  factors  obtained  using  the  initial  factor  extraction  may  be  diffi-­‐ cult  to  interpret.    Consequently,  most  researchers  use  rotation,  a  technique  that  helps  in  obtaining  fac-­‐ tors  that  are  easier  to  interpret.    In  cases  such  as  the  study  when  there  are  prior  expectations  about  the   number  of  factors,  the  most  appropriate  approach  is  using  confirmatory  factor  analysis.      

The  comparative  method  is  the  “method  of  testing  hypothesized  empirical  relationships  among   variables  on  the  basis  of  the  same  logic  that  guides  the  statistical  method,  but  in  which  the  cases  are   selected  in  such  a  way  as  to  maximize  the  variance  of  the  independent  variables  and  to  minimize  the   variance  of  the  control  variables”  (Lijphart,  1975,  164).    This  study  benefits  from  the  comparative  meth-­‐ od  first  because  “the  comparative  method  does  not  select  its  cases  in  random  ways  (as  do  experimental   and  statistical  studies).    Rather  comparative  studies  unabashedly  select  their  cases  on  the  dependent   variable”  (Moses  and  Knutsen,  2007,  95).    The  ability  to  select  the  cases  included  in  the  analysis  is  con-­‐ sidered  one  of  the  main  strengths  of  the  comparative  method.    It  also  allows  the  researcher  to  compare  

 

es  or  most  dissimilar  cases.    In  general  the  shortcomings  of  this  methodological  approach  are  avoided  by   the  researcher  employing  both  deductive  and  inductive  directions  of  determining  the  relationship  be-­‐ tween  the  variables.  

Case  study.    Two  important  ways  in  which  a  variable  like  tolerance  can  be  analyzed  are  multi-­‐ case  comparisons  and  single-­‐case  studies.    Scholars  such  as  Inglehart  (1997;  2005)  and  Huntington   (1991)  looked  at  dozens  of  cases  to  explain  cross-­‐national  differences  in  political  culture  or  patterns  of   democratization.    One  major  advantage  of  this  approach  is  the  possibility  of  testing  hypotheses  on  a   large  number  of  cases,  thus  opening  up  the  prospect  of  generalizing  the  results.    However,  while  broad   in  scope,  such  studies  tend  to  be  short  on  details,  lacking  an  in-­‐depth  perspective  on  processes  and  indi-­‐ vidual  cases.    This  is  exactly  one  of  the  strengths  of  case  studies,  which  offer  the  researcher  the  oppor-­‐ tunity  to  conduct  a  fine  grained  analysis  of  a  single-­‐case  and  subsequently  to  offer  a  comprehensive  ac-­‐ count  of  the  findings.    Robert  Putnam’s  Making  Democracy  Work  is  one  of  the  finest  examples  of  using  a   multi-­‐method  approach  to  explain  why  the  same  institutions  work  well  in  the  North  of  Italy  but  poorly  in   the  South.    Arend  Lijphart  is  one  of  the  most  influential  authors  in  comparative  politics;  however,  the   starting  point  for  the  development  of  his  typology  of  democratic  regimes  (1984;  1999)  was  Politics  of   Accommodation  (1968),  a  book  on  his  native  Netherlands.  

Both  multi-­‐case  comparisons  and  single-­‐case  studies  are  valuable  tools  for  research.    The  two   methods  are  complementary,  rather  than  mutually  exclusive.    Often,  a  case  study  is  only  the  first  step  in   a  long  term  research  project,  as  it  was,  for  instance,  in  Lijphart’s  case.    This  is  also  the  direction  of  this   research.    The  main  focus  of  this  dissertation  is  the  study  of  political  tolerance  and  its  micro-­‐  and  macro-­‐ level  determinants  in  general,  not  just  in  Romania.    This  study  only  uses  Romania,  a  country  whose  citi-­‐ zens  are  comparatively  intolerant  as  previous  research  demonstrates  (e.g.  Viman-­‐Miller  and  Fesnic  

2010),  yet  it  has  not  been  studied  extensively  so  far,  as  an  exemplary  case  to  develop  and  test  hypothe-­‐ ses  which  can  be  tested  subsequently  in  other  countries.      

Quasi-­‐experiment.  Previous  studies  on  immigration  in  Western  Europe  show  that  the  newcom-­‐ ers  adopt  a  more  democratic  profile  of  citizenship.    The  final  part  of  this  dissertation  seeks  to  add  a  new   dimension  to  study  of  political  tolerance  –  i.e.,  the  impact  of  migration.    Does  temporary  migration  influ-­‐ ence  the  level  of  political  tolerance  of  migrants?    In  order  to  answer  this  question,  this  study  will  analyze   both  qualitative  and  quantitative  data,  comparing  the  tolerance  of  Romanians  who  have  migrated  to   that  of  Romanians  who  have  never  left  the  country,  seeking  to  isolate  the  independent  effects  of  migra-­‐ tion  on  attitudes.    Based  on  the  overview  of  the  literature  on  socio-­‐economic,  demographic,  and  attitu-­‐ dinal  determinants  of  political  tolerance,  Figure  3.2  presents  a  recursive  model  of  political  tolerance  de-­‐ rived  from  my  theoretical  argument.    The  social  and  demographic  characteristics  of  respondents  are   control  variables,  while  exposure  to  the  ideas  and  institutions  characteristics  for  a  more  democratic  so-­‐ ciety  is  the  intervening  variable  which  increases  the  political  tolerance  of  migrants.  

 

Figure  3.2.  Migration  and  political  tolerance:  a  recursive  model  

 

This  model  is  tested  in  Chapter  7  using  data  provided  by  an  original  survey  of  Romanian  college   students.    This  is  a  quasi-­‐experimental  design,  and  the  survey  was  conducted  in  Cluj,  Romania  using  stu-­‐ dents  from  “Babes-­‐Bolyai”  University,  a  state  university44  which  is  the  largest  institution  of  higher  educa-­‐

tion  in  the  region  of  Transylvania,  with  over  50,000  students  enrolled  in  2008.    Two  groups  were  used,  a   “treatment”  group  and  a  “control”  group.    The  first  group  included  students  who  have  traveled  to  the  

                                                                                                                         

44  To  see  how  representative  this  sample  is  for  the  population  of  state  university  students  in  Romania,  or  for  the  broader  stu-­‐ dent  population  in  the  country,  there  are  surveys  conducted  on  nationally  representative  samples  of  students  at  our  disposal.     One  such  example  is  a  survey  conducted  on  two  samples,  one  that  is  representative  for  students  enrolled  in  Romanian  state   universities,  while  the  other  sample  is  representative  for  students  enrolled  in  Romanian  private  universities.  (Direcţia  pentru  

US  with  the  Work  and  Travel  program45  and  the  second  group  included  students  who  have  never  trav-­‐

eled  to  the  US.      

The  survey.  The  practical  means  of  implementing  the  survey  were  self-­‐administered  question-­‐ naires,  and  the  setting  was  group  administration  in  classes  with  large  enrollment.    The  cost  of  this  ap-­‐ proach  was  very  low  and  the  completion  rate  was  near  100%,  and  these  were  major  advantages  under   conditions  of  limited  time  and  resources  (Johnson  and  Reynolds  2008,  303).    Moreover,  unlike  face-­‐to-­‐ face  interviews,  self-­‐administered  questionnaires  facilitated  asking  sensitive  questions  (Johnson  and   Reynolds  2008,  318)  and  alleviated  the  pressure  to  give  “socially  desirable”  answers  (Traugott  and  Price   1992,  246).  

To  address  the  problem  of  an  absence  of  a  pre-­‐test,  multiple  items  in  the  questionnaire  asked   the  respondents  to  self-­‐assess  their  position  at  the  moment  of  the  questionnaire  administration  com-­‐ pared  to  what  it  was  before  their  American  experience.    The  survey  used  a  combination  of  questions  to   assess  students’  political  tolerance  before  and  after  their  involvement  in  the  Work  and  Travel  program.     This  approach  is  not  novel;  social  psychology  and  political  psychology  research  often  rely  on  information   obtained  from  surveys  using  recall.    Ansolabehere  and  Iyengar  (1995)  employed  the  recall  method  and   researched  the  level  of  influence  that  negative  advertisement  had  on  voter  turn  out.    They  found  that   negative  advertisement  tends  to  depress  the  voter  turnout.  Later,  Wattenberg  and  Brians  (1999)  found   that  negative  campaign  advertisement  stimulates  voter  turnout.    They  based  their  research  on  the  same   methodological  approach  as  their  predecessors.    Druckman  et.  al  (2011)  debate  thoroughly  the  pros  and   cons  of  subjects’  memory  recall  in  applied  social  sciences.    They  find  that  in  experimental  political  sci-­‐ ence  recall  cannot  be  avoided  and  that  researchers  must  pay  more  attention  to  data  interpretation.                                                                                                                              

 

They  make  a  distinction  between  recall  and  recognition  (100).    A  similar  approach  is  also  used  in  political   sociology,  especially  in  studies  of  voting  behavior,  where  survey  respondents  are  often  asked  how  they   voted  in  the  last  election.    Lizotte,  Lodge  and  Taber  (2005)  argue  that  results  from  emotional  recall  can-­‐ not  be  trusted  if  they  are  obtained  from  a  format  that  asks  direct  questions.    However,  their  attempt  to   demonstrate  this  experimentally  was  unsuccessful.    Table  3.4.  presents  a  summary  of  the  research  de-­‐ sign,  comparing  current  political  tolerance  of  the  treatment  group  with  that  of  the  control  group.  

 

Table  3.4.  Assessing  the  impact  of  Work  and  Travel  experience:  a  quasi-­‐experiment   Time:  

Group:  

t  –  1    

(before  Work  and  Travel):   t:  Work  and  Travel  (“Treatment”?)   (present):  t  +1    

“Control”   Tolerance  C,  t-­‐1   No     Tolerance  C,  t  +1  

  ≈     ≠  

“Treatment”   Tolerance  T,  t-­‐1   Yes  (democratic  expo-­‐

sure  &  learning)   Tolerance  T,  t  +1  

 

A  major  advantage  of  surveys  is  the  fact  that  they  provide  a  large  number  of  cases  which  enable   multivariate  analyses  –  here,  testing  the  impact  of  migration  on  tolerance  using  background  and  attitu-­‐ dinal  variables  as  controls.    Survey  data  are  helpful  to  answer  the  “if”  question,  but  is  less  helpful  to  an-­‐ swer  the  “why”  question.    If  statistical  analysis  indicates  that  migration  does  have  an  impact  on  toler-­‐ ance,  it  is  still  to  be  determined  why  that  is  the  case,  and  the  processes  through  which  greater  exposure   to  a  democratic  culture  via  migration  leads  to  a  change  of  the  migrant’s  level  of  tolerance.    Thus,  a  mul-­‐ ti-­‐method  approach  was  necessary,  one  that  included  both  quantitative  and  qualitative  methods.  

Interviews.  To  complement  the  data  provided  by  the  surveys,  twenty  systematic  semi-­‐structured   interviews  were  conducted,  trying  to  get  a  more  in-­‐depth  understanding  of  the  processes  that  may  lead   to  an  increase  in  tolerance,  and  also  to  assess  whether  such  an  increase  has  indeed  occurred  in  the  first  

place.    The  richness  of  interview  data  collection  allowed  a  better  understanding  of  the  impact  of  toler-­‐ ance  at  the  individual-­‐level.    These  were  “nonscheduled  standardized  interviews,”  defined  by  Gray  et  al.   (2007,  161)  as  a  method  by  which  “all  questions  are  asked  of  each  respondent,  but  they  may  be  asked  in   different  ways  and  in  different  sequences.”    According  to  Manheim  and  Rich  (1995,  162),  the  central   goal  of  this  kind  of  interviewing  is  not  so  much  “the  collection  of  prespecified  data,  but  the  gathering  of   information  to  assist  in  reconstructing  some  event  or  discerning  a  pattern  of  specific  behaviors”  –  in  this   case,  how  migration  affects  tolerance.    Among  other  advantages  of  qualitative  interviews  is  the  fact  that   it  allows  the  interviewer  to  become  more  personal  and  gain  the  trust  of  the  interviewees.    This  allows   and  encourages  introspection  from  the  respondents  (Gray  et.  al  2007).    For  this  study,  respondents  who   were  previously  enrolled  in  Work  and  Travel  programs  were  selected,  discussing  with  them  how  this   experience  has  changed  their  political  and  social  tolerance.  

To  minimize  any  bias  in  the  selection  of  respondents  for  these  interviews,  every  student  previ-­‐ ously  enrolled  in  the  Work  and  Travel  program  who  filled  the  questionnaire  was  asked  if  he  or  she  was   willing  to  be  interviewed.    Given  the  time  and  financial  constraints,  on  the  one  hand,  and  the  relatively   low  number  of  people  in  this  sample  with  previous  Work  and  Travel  experience,  on  the  other  hand,  it   was  unrealistic  to  try  to  interview  more  than  thirty  people.    The  expectation  was  that,  if  at  least  half  of   those  from  whom  an  interview  were  requested  would  answer  affirmatively,  taking  into  account  the  goal   of  having  at  least  one  hundred  students  with  Work  and  Travel  experience  filling  the  questionnaire,  this   should  have  also  ensured  about  twenty  interviews.    In  actuality,  the  total  number  of  students  from  the   total  sample  (N  =  1,514)  with  at  least  one  Work  and  Travel  experience  was  129  (8.5%),  and  the  number   of  interviews  completed  was  20.    The  main  point  of  these  interviews  was  to  get  additional  insights  on   the  respondents’  experience  in  the  US  (qualitative  information),  in  addition  to  the  quantitative  infor-­‐ mation  provided  by  survey  data.    The  interviews  were  not  meant  to  be  representative,  but  they  were  

 

designed  to  provide  important  additional  contextual  information  that  could  help  illuminate  patterns  in   the  survey  data.  

In document Trojans in Wireless Sensor Networks (Page 79-88)