Document not found! Please try again

Left to Right or Right to Left? - Survey Sampling International

4 downloads 179 Views 46KB Size Report
Feb 12, 2016 - Pete Cape. Director, Global Knowledge, SSI. In laying out an agree/disagree battery in a grid format the
 

Left  to  Right  or  Right  to  Left?:  How  to  Order  Scales  for  Best  Results   Pete  Cape   Director,  Global  Knowledge,  SSI     In  laying  out  an  agree/disagree  battery  in  a  grid  format  the  researcher  faces  a  choice  of  which  order   to  display  the  scale:  Agree  on  the  left  or  Agree  on  the  right?   Does  the  choice  matter?  Researchers  aware  of  a  primacy  bias  (a  bias  toward  choosing  the  first  item   presented  visually)  would  counter  this  bias  by  flipping  the  scale  such  that  half  the  sample  sees   ‘Agree’  on  the  left  and  half  sees  ‘Disagree’.  If  they  were  also  concerned  about  recency  bias  (a  bias   towards  choosing  the  last  item  when  presented  aurally)  then  the  same  solution  would  work.   But  might  this  action  actually  cause  more  problems  than  it  solves?   Research  done  in  2004  by  Roger  Tourangeau  demonstrated  that  most  participants  expect  the  most   positive  category  to  be  presented  first  -­‐  so  there  appears  a  natural  order  in  ordinal  scales.  (An   ordinal  scale  is  one  in  which  the  order  of  the  scale  points  is  important,  but  there  is  no  precise   measurement  of  the  gaps  between  each  point.)  In  order  to  prevent  confusion  therefore  the  scales   should  be  presented  in  their  natural  order.  Under  this  reasoning  the  scale  for  satisfaction  (for   example)  ought  to  start  with  “extremely  satisfied”  and  end  with  “not  at  all  satisfied”).  Scales   presented  the  “wrong  way  around”  may  be  biased  towards  the  negative  end.  Any  necessary  mental   correction  that  the  participant  has  to  make  to  the  way  the  scale  is  being  displayed  (now  “up  means   bad”  rather  than  “up  means  good”)  should  result  in  slightly  longer  being  taken  to  process  the  grid.   Thus  we  might  expect  both  time  and  data  differences  if  the  scale  is  presented  in  an  un-­‐natural  order.   This,  however,  is  not  what  is  found  in  the  case  of  online  surveys.  According  to  another  questionnaire   design  guru  Don  Dillman,  in  Internet,  mail  and  mixed-­‐mode  surveys  no  data  differences  are  found   when  the  scales  are  reversed  –although  the  “un-­‐natural”  version  does  take  longer  to  do.   An  agree-­‐disagree  scale  however  (unlike  an  “extremely”  to  “not  at  all”  scale)  has  no  natural  order.   The  “good”  end  of  the  scale  depends  firstly  on  the  item  being  rated  and  secondly  in  the  eye  of  the   beholder  (on  the  basis  that  you  would  think  “good”  that  which  you  are  congruent  on  –  whether   agreement  or  disagreement).   Our  small  experiment  also  shows  no  difference  in  data  on  an  agree/disagree  scale  resulting  from  the   order  the  scale  categories  are  presented,  and  showed  no  statistically  significant  differences  when   the  item  being  rated  tended  to  be  agreed  with  or  disagreed  with:  

February  12,  2016                                                                                                                                                                                                                                                                [email protected]  |  surveysampling.com  

 

     

%  Agreeing   Agree  on  left  

Disagree  on  left  

Base  

155  

154  

Sports  is  an  important  part  of  your  life  

52%  

47%  

A  sedentary  lifestyle  has  no  effect  on  your  long   term  health   You  would  like  to  be  more  active  and  fit  by   engaging  in  sports  or  physical  exercise  but  you   can’t  find  the  time   You  always  ensure  you  make  time  for  yourself  to   engage  in  sports  or  physical  activities   average  time  

18%  

21%  

52%  

45%  

51%  

44%  

40  seconds  

43  seconds  

median  time  

36  seconds  

40  seconds  

  However,  we  did  see  a  slight  increase  in  time  taken  when  the  scale  is  presented  Disagree-­‐Agree   rather  than  Agree-­‐Disagree.   SSI  therefore  recommends  that  ordinal  scales  be  ordered  left  to  right  (or  top  to  bottom)  in  terms  of   best  to  worst.  If  flipping  of  a  scale  is  called  for  the  order  of  the  scale  should  be  kept  the  same   throughout  an  individual’s  survey  in  order  not  to  cause  confusion.     About  SSI   SSI   is   the   premier   global   provider   of   data   solutions   and   technology   for   consumer   and   business-­‐to-­‐ business   survey   research,   reaching   respondents   in   100+   countries   via   Internet,   telephone,   mobile/wireless  and  mixed-­‐access  offerings.  SSI  staff  operates  from  30  offices  in  20  countries,  offering   sample,   data   collection,   CATI,   questionnaire   design   consultation,   programming   and   hosting,   online   custom  reporting  and  data  processing.  SSI’s  3,600  employees  serve  more  than  2,500  clients  worldwide.     Visit  SSI  at  www.surveysampling.com.    

February  12,  2016                                                                                                                                                                                                                                                                [email protected]  |  surveysampling.com