Summary of results from testing of experimental Subjective Well

Summary of results from testing of
experimental Subjective Well-being
questions - December 2012
Measuring Subjective Well-being,
Measuring National Well-being Programme, ONS
www.ons.gov.uk/well-being
Background
From April 2011, ONS included four subjective well-being (SWB) questions
covering evaluative, eudemonic and affect measures of well-being on the
Annual Population Survey (APS) and the Opinions & Lifestyle Survey (OPN)1.
The four questions are as follows:
• Overall, how satisfied are you with your life nowadays? (evaluative)
• Overall, to what extent do you feel the things you do in your life are
worthwhile? (eudemonic)
• Overall, how happy did you feel yesterday? (affect)
• Overall, how anxious did you feel yesterday? (affect)
All answered on a scale of 0 to 10 where 0 is 'not at all' and 10 is 'completely‘.
For more information on the four questions and the different ways of measuring
subjective well-being please see Measuring Subjective Well-being in the
UK on the ONS website: http://www.ons.gov.uk/ons/index.html
1. The questions included on the APS have not changed since their introduction in 2011; All testing work involves the OPN with exception to the month on month which was APS.
2
Experimental Statistics
•
As with all survey based questions, responses to subjective well-being questions
can be affected by various methodological issues such as mode of interview,
question order and context effect, question wording, and response scales.
•
ONS is undertaking an iterative process of both quantitative and qualitative testing
in order to investigate how well the subjective well-being questions capture the
required information.
•
In taking this work forward it should be recognised that measuring subjective wellbeing is still an ‘emerging’ science. Our aim at ONS is to develop a suite of
questions which are both statistically robust and practical for wider roll-out across
policy relevant surveys.
•
ONS statistics have been labelled ‘experimental’ to signal their on-going
refinement and development.
•
This presentation reports the results of the second phase of quantitative testing
carried out on the OPN survey. The first results can be found in section 8 of ‘Initial
investigation into Subjective Well-being from the Opinions Survey’ (ONS,
December 2011).
For a full copy of the March 2012 cognitive testing report, see link http://www.ons.gov.uk/ons/guide-method/user-guidance/well-being/advisory-groups/well-being-technicaladvisory-group/index.html
3
OPN Survey Testing
•
The OPN Survey offers a flexible testing vehicle (1,000 respondents
monthly; new module each month).
•
Split trials carried out on OPN in this phase of testing include:
 question order,
 alternative prompts/preambles for the ‘yesterday’ questions,
 question wording,
 show-cards,
 response variation for days of the week.
•
Data from the APS survey was used to provide information on the extent
to which responses varied across months of the year.
•
We have previously looked at interviewer led versus self- completion
interviews and face to face interviews versus self-completion interviews,
for further information please see Annexes A, B and C.
•
Additionally Dolan et al (2012) investigated the differences between
interviewer administered and telephone administered responses to the
APS. http://eprints.lse.ac.uk/45273/
4
Question Order Analysis
•
Responses to SWB questions can be affected both by the respondent’s
current mood and the survey context.
•
It is important to find out if the order in which the different subjective
well-being questions are asked affects the responses to the questions.
•
For this reason, split trials were conducted in which the four subjective
well-being questions were asked in different orders.
5
Question Order – 4 Samples
6
Testing Question Order
•
Sample 1 replicates the order in which the SWB questions are asked in ONS surveys.
Key reasons for the existing question order include:
 ONS cognitive testing found that respondents preferred positive questions to be
placed first as they were seen as being easier to answer.
 the ‘life satisfaction’ question appears first because it is considered easiest to
answer; the ‘satisfaction’ and ‘worthwhile’ questions both ask about respondent’s
lives overall, and are therefore placed next to each other.
 the ‘happy’ and ‘anxious’ questions ask about the respondent’s emotions the
previous day and are therefore asked next to each other.
 the ‘anxious’ question is considered to be the most negative and is therefore
placed last. The scale is also the reverse to the others so putting it last avoids
confusion.
•
As the context in which the questions are asked can affect responses it was
important to test the questions in different orders. However, it should be noted that
the ‘satisfaction’ question was asked first 3 out of 4 times and the ‘anxiety’ question
was never asked first for the reasons described above.
7
Question Order – Results
Subjective well-being mean scores by question order sample1.
Sample 1
7.9
7.8
7.7
7.6
7.5
7.4
7.3
7.2
7.1
7.0
6.9
Sample 2
Sample 3
Sample 4
The differences between the order of the
questions were small but the differences we
do see are as we might have expected.
•
When ‘life satisfaction’ followed the affect
questions (sample 3) its mean rating was
lower (7.3) compared with the other orders
(7.5).
•
For the ‘worthwhile’ question the lowest
mean of 7.6 was observed when asked
after the anxious question.
•
The ‘happy yesterday’ question had the
lowest mean (7.1) when asked after the
‘anxious yesterday’ question (Sample 4).
•
However, none of these differences were
found to be statistically significant. This
suggests that the 4 different question
orders tested here did not result in any
systematic difference in how people
answered the questions.
•
For further information see annex D.
7.7
7.6
7.5
7.6
7.6
7.5
7.5
7.3
7.2 7.2
7.2
7.1
Overall, how satisfied are you Overall, to what extent do you Overall, how happy did you
feel yesterday?
with your life nowadays?
feel that the things you do in
your life are worthwhile?
Anxious mean scores by question order sample1.
3.5
•
Sample 1
Sample 2
Sample 3
Sample 4
3.4
3.3
3.2
3.2
3.1
3.1
3.2
3.1
3.0
2.9
2.8
2.7
2.6
2.5
Overall, how anxious did you feel yesterday?
1. Data compiled using five months data, September, October, December 11, January and
April 12.
8
Testing Show Cards
•
We tested show cards in the OPN testing although these are not normally used in ONS surveys
generally or for the well-being questions in the APS.
The key reasons for testing show cards were because:

ONS cognitive testing found that the scale for the ‘anxiety’ question was misunderstood by
some respondents e.g. these respondents would give a high rating (indicating high anxiety)
to this question when they had meant to give a low rating (indicating low anxiety).
Why was the scale for the ‘anxiety’ question misunderstood?
•
•
•
•
For the ‘life satisfaction’, ‘worthwhile’, and ‘happy yesterday’ questions a 10 out of 10 is the best
possible result and would indicate ‘complete satisfaction’ etc, however, for the ‘anxiety’ question a
rating of 10 out of 10 would indicate the most negative result, e.g. that the respondent was
‘completely anxious’.
Cognitive testing indicated that when show cards were introduced they helped respondents to
reverse the scale and therefore interpret the anxiety question correctly.
The split trial testing was used to see what difference show cards would make to survey
responses on a larger scale.
Show cards were presented with the scale horizontally across the show card.
9
Show Cards Versus No Show Cards
Life Satisfaction, Worthwhile and Happy Yesterday: Show cards and No
Show cards, February/ March 20121
Showcard
No Showcard
7.9
7.8
7.7
7.6
7.5
7.4
7.3
7.2
7.1
7.0
6.9
7.7
•
Comparing responses when show cards were used
versus when they were not, the means were significantly
higher for ‘life satisfaction’ (at the 5% level) and happy
yesterday (at the 10% level) when show cards were
used.
•
There were no significant differences found between
show cards versus no show cards for either the
‘worthwhile’ or ‘anxious yesterday’ questions.
•
The ‘worthwhile’ question seems to be the most stable
with or without show cards (with a mean of 7.7 in both
cases).
•
These are surprising results given the previous ONS
cognitive testing work (see previous slide). However, the
sample used here was small, therefore further testing
may be helpful.
•
We have also examined the distribution of responses to
each question in cases where statistically significant
differences were found between the means when show
cars were used or not. The patterns of response
distributions did not differ in any interesting way
regardless of whether show cards were used.
•
For further information see annex E.
7.7
7.6
7.4
7.4
7.2
Overall, how satisfied are you Overall, to what extent do you Overall, how happy did you
feel yesterday?
with your life nowadays?
feel that the things you do in
your life are worthwhile?
Anxious Yesterday: Show cards and No Show cards, February/ March
20121
3.5
3.4
3.3
3.2
3.1
3.0
3.0
2.9
2.9
2.8
2.7
2.6
2.5
Showcard
No Showcard
1. Only two months data available and therefore based on small sample sizes.
10
Testing Different Versions Of The
‘Yesterday’ Questions
•
Previous ONS cognitive testing found that some respondents did not like the
reference to ‘yesterday’ when answering the ‘happy & anxious yesterday’ questions
as it was not seen as being representative of their general state.
•
This led to the trial of an additional instruction from the interviewer; 'please think
about yesterday even if it was not a typical day'.
This instruction was tested in two alternative ways:
•
 Sample 1 tested the instruction immediately after the question: 'Overall, how happy
did you feel yesterday?
Please think about 'yesterday' even if it was not a typical day?’
 Sample 2 tested the instruction as preamble to the two affect questions: 'The next two
questions ask about how you felt yesterday. Please think about 'yesterday' even if it
was not a typical day.
Overall, how happy did you feel yesterday?‘
11
Happy Yesterday Variation Results
Happy yesterday question variation, July/August/September
20121
•
There was no significant difference
between the mean scores for the ‘happy
yesterday’ question regardless of which
version of the preamble to the question
was used.
•
This indicates that the alternative ways of
administering the preamble do not affect
the way that respondents understand or
answer the questions.
•
We have not yet tested the ‘anxious
yesterday’ question with the alternative
preambles, or conducted a split trial of
the ‘happy’ and ‘anxious’ yesterday
questions with the addition of ‘please
think about ‘yesterday’ even if it was not
a typical day’ versus the original wording.
This is planned for ongoing split trial
testing work.
•
For further information see annex F.
7.9
7.8
7.7
7.6
7.5
7.4
7.3
7.3
7.3
7.2
7.1
7.0
6.9
Overall, how happy did you feel
yesterday? Please think about
‘yesterday’ even if it was not a
typical day.
The next two questions ask about
how you felt yesterday. Please
think about ‘yesterday’ even if it
was not a typical day. Overall, how
happy did you feel yesterday?
1. Data compiled using three month’s data.
12
Testing Alternatives To The ‘Life
Satisfaction’ Question
Overall, it is important to note that no specific time frame for the ‘life
satisfaction’ question is intended, rather the aim of the question is to capture
information from ‘more recent times’ in people’s lives.
For the ‘life satisfaction’ question three different endings of the question were
tested:
• Overall, how satisfied are you with your life nowadays?
• Overall, how satisfied are you with your life these days?
• Overall, how satisfied are you with your life?
Why were these alternatives tested?
 ONS cognitive testing had found that some younger respondents felt the
term ‘nowadays’ used in the original question is old fashioned.
 Additionally it is important to test alternative versions of the ‘life
satisfaction’ question as it is possible that an international standard
harmonised version of the ‘life satisfaction’ question may not necessarily
have exactly the same wording as the ONS question. Therefore it is helpful
to know whether other versions of the question will elicit similar responses
or not.
13
Life Satisfaction ‘These Days’ Versus
‘Nowadays’ Results
Life satisfaction question wording - November 20111
•
‘These days’ had a higher mean score
of 7.5 compared to ‘nowadays’ at 7.4.
This difference was not statistically
significant.
•
This finding would indicate that the
change from ‘nowadays’ to ‘these
days’ does not alter the way in which
respondents answer this question, but
the sample sizes were small so
findings should be cautiously
interpreted.
•
This would suggest that, even if
respondents understand these terms
in slightly different ways as indicated
by cognitive testing, this does not
mean on average that we get very
different results.
•
For further information see annex G.
7.9
7.8
7.7
7.6
7.5
7.5
7.4
7.4
7.3
7.2
7.1
7.0
6.9
Overall, how satisfied are you with your Overall, how satisfied are you with your
life nowadays?
life these days?
1. Only one month’s data available and therefore based on small sample sizes.
14
Life Satisfaction ‘These Days’ Versus ‘No
Ending’ Results
•
Asking the question with no time reference (i.e
no ending) resulted in a significantly higher
mean score of 7.7 compared to 7.5 when a time
reference of ‘these days’ was used.
•
This is a significant difference at the 5% level,
but is based on a small sample size and
therefore should be interpreted cautiously.
•
This finding suggests that the two versions of
the question may collect slightly different
information from respondents.
•
However the distributions of both variations
follow a similar pattern.
•
This difference could be due to the fact that the
use of ‘these days’ focuses respondents more
on recent times in their lives, whereas in the
version with no time reference respondents
may be considering their whole lives.
•
Further investigations maybe required in the
future to augment the sample sizes.
•
For further information see annex G.
Life satisfaction question wording - July/August 20121
7.9
7.8
7.7
7.7
7.6
7.5
7.5
7.4
7.3
7.2
7.1
7.0
6.9
Overall, how satisfied are you with
your life these days?
Overall, how satisfied are you with
your life?
1. Only two month’s data available and therefore based on small sample sizes.
15
Life Satisfaction ‘Nowadays’ Versus ‘No
Ending’ Results
•
There was no difference found in the
mean scores when the ‘life satisfaction’
question was asked with no time
reference versus the ‘nowadays’ time
reference. Therefore both versions of the
question had a mean score of 7.7.
•
Although the sample sizes were small
here, this appears to suggest that there is
no important difference between the way
that respondents answer these two
versions of this question.
•
This is surprising given the significant
difference in response to the ‘these days’
and ‘no ending’ versions of the question
(see previous slide). Again, this suggests
there would be value in replicating this
test to increase the sample sizes and
enable question confidence in the
findings.
•
For further information see annex G.
Life Satisfaction question wording - September 20121
7.9
7.8
7.7
7.7
7.7
7.6
7.5
7.4
7.3
7.2
7.1
7.0
6.9
Overall, how satisfied are you with
your life nowadays?
Overall, how satisfied are you with
your life?
1. Only one month’s data available and therefore based on small sample sizes.
16
Testing alternatives to the ‘Worthwhile’
Question
•
The ‘eudemonic’ approach draws on self-determination theory and tends to measure
such things as people’s sense of meaning and purpose in life. The original ONS
eudemonic question is as follows: ‘Overall, to what extent do you feel that the things
you do in life are worthwhile’
•
Previous ONS cognitive testing suggested that answers to this question may be
affected by social desirability bias which could lead to people giving inflated scores.
•
Additionally some respondents with low levels of education did not understand the
term ‘worthwhile’.
•
For this reason the words ‘meaningful’ and ‘purpose’ were tested as alternatives to
‘worthwhile’.
The question variations trialled were:
 Overall, to what extent do you feel that the things you do in your life are meaningful?
and
 Overall, to what extent do you feel that the things you do in your life have purpose?
17
‘Meaningful’ Versus ‘Purpose’ Question
Results
Testing alternatives to ‘worthwhile’ - July/August/Sept
20121
•
‘Purpose’ had higher mean scores than
‘meaningful‘ (7.7 and 7.5 respectively), this
difference is significant at the 5% level. The
distribution of responses to the question
followed a similar pattern for both versions of
the questions.
•
These findings suggest that people interpret
‘purpose’ and ‘meaningful’ differently and are
more likely to give a higher rating out of 10
when asked about the things they do in their life
having ‘purpose’, compared with when they are
asked about the things in their life having
‘meaning’.
•
No testing has been carried out yet with
comparing ‘meaningful’ versus ‘worthwhile’.
•
The terms 'purpose‘ and 'worthwhile' have
previously been tested on the OPN Survey2.
•
For further information see annex G.
7.9
7.8
7.7
7.7
7.6
7.5
7.5
7.4
7.3
7.2
7.1
7.0
6.9
Overall, to what extent do you feel that the Overall, to what extent do you feel that the
things you do in your life are meaningful?
things you do in your life have purpose?
1. Data compiled using three month’s data.
2. http://www.ons.gov.uk/ons/guide-method/user-guidance/well-being/publications/previouspublications/index.html
18
Alternatives To The ‘Anxious Yesterday’
Question
•
Previous ONS cognitive testing found differences in the way that
respondents interpreted the terms ‘anxious’ and ‘worried’.
•
Qualitative evidence suggested that ‘worry’ was considered by respondents
to be less serious than ‘anxious’.
•
The cognitive testing also found that some respondents felt that the term
‘anxious’ had a level of stigma attached to it which could potentially lead to
lower scores (indicating lower anxiety) being given for this question (due to
social acceptability).
Two versions of the questions were therefore trialled during OPN testing:

Overall, how anxious did you feel yesterday?
and

Overall, how worried did you feel yesterday?
19
Alternatives To ‘Anxious Yesterday’
Results
•
‘Worried yesterday’ had a higher mean score
than ‘anxious yesterday’ (3.3 and 3.0
respectively), however, this difference was
not statistically significant.
•
This suggests that there is no systematic
difference in how people respond to these
questions in a survey context.
•
‘Anxious’ and ‘worry’ was previously
compared in the 'Initial investigation into
subjective well-being from the Opinions
Survey' December 20112. However, the
responses to the 'anxiety' question had a
higher mean score than the mean score for
the 'worry' question (In Dec 2011 anxious
was 3.6 and worried 3.1)3.
•
This difference could be due to the fact that
only one month’s worth of data was used in
the current analysis, or that previously the
questions were administered as part of a
suite of negative affect questions rather
than a split trial.
•
For further information see annex G.
Anxious yesterday question wording - November 20111
3.5
3.4
3.3
3.3
3.2
3.1
3.0
3.0
2.9
2.8
2.7
2.6
2.5
Overall, how anxious did you feel
yesterday?
Overall, how worried did you feel
yesterday?
1. Only one month’s data available and therefore based on small sample sizes.
2. http://www.ons.gov.uk/ons/rel/wellbeing/measuring-subjective-wellbeing-in-theuk/investigation-of-subjective-well-being-data-from-the-ons-opinions-survey/initial-investigationinto-subjective-well-being-from-the-opinions-survey.html
3. UK average for anxious is 3.1, on the testing used above it is 3.0
20
Month On Month Analysis (APS)
•
It is interesting to compare month by month subjective well-being
data in order to investigate whether there are any significant
fluctuations, caused by seasons, weather or other national
events which may have impacted on the whole population.
•
For monthly analysis, the APS data was used rather than OPN
data due to the larger sample size of the APS, and consistency in
the administration of the four subjective well-being questions over
time.
•
Our expectation was that the ‘yesterday’ questions would show
greater fluctuations month on month than the longer term
assessments entailed in the ‘life satisfaction’ and ‘worthwhile’
questions.
21
Month On Month Results For Life Satisfaction,
Worthwhile And Happy Yesterday
Mean satisfaction with life, worthwhile, happy yesterday by month1
1. Data complied using one year’s data from April 2011 to March 2012.
•
A significant difference at the 5% level was
found between months of the year for ‘life
satisfaction’, although for all months the mean
scores for ‘life satisfaction’ ranged very little
(between 7.4 and 7.5).*
•
No significant difference was found between
months of the year for the ‘worthwhile’ question.
For all months the mean scores for ‘worthwhile’
varies from 7.6 to 7.7.*
•
A significant difference at the 1% level was
found between months for the ‘happy
yesterday’ question.
•
For all months the mean scores for ‘happy
yesterday’ fluctuated between 7.2 and 7.4 over
the period.
•
For further information see annexes H and I.
*Please note, whole (unrounded) numbers were used in significance
tests
22
Month On Month Results For Anxious
Yesterday
Mean anxious yesterday by month1
3.5
3.4
3.1
Responses to the ‘anxious’ yesterday
question are considered separately
to show differences in responses
more clearly than would be possible
if the full 11 point scale was used in
the chart.
•
A significant difference at the 1%
level was found between months for
the ‘anxious yesterday’ question.
•
For all months the mean scores for
‘anxious yesterday’ ranged from 3.0
to 3.3 over the year.
•
For further information see annexes
H and I.
3.3
3.3
3.2
•
3.2
3.1
3.2
3.2
3.1
3.2
3.1
3.1
3.0
3.0
3.1
3.0
2.9
2.8
2.7
2.6
2.5
1. Data complied using one year’s data from April 2011 to March 2012.
23
Events/Weather Conditions Which May
Have Impacted Month On Month Results
There were significant differences for the ‘life satisfaction’, ‘happy yesterday’ and ‘anxious yesterday’ questions.
•
Though there was a significant difference for ‘life satisfaction’, there was no national or global events found to aid
explanation for the peaks or troughs over the 12 month period.
•
There was a 1% significant difference for the ‘happy yesterday’ question with April 2011 and March 2012 being the ‘happiest’
months with a score of 7.4. Possible reasons for this could include:
 A heat wave in April 2011 sees warmest UK April for more than 100 years.
 The royal wedding resulted in an extra bank holiday for many.
 March 2012 was the 3rd warmest, 3rd sunniest and 5th driest since records began.
•
The least happiest month was November 2011 with a score of 7.2, possible reasons for this are:
 Tuition fees demonstration in Westminster
 Pension strikes across UK
•
There was a 1% significant difference for ‘anxious yesterday’ question with May 2011 being the most ‘anxious’ month with a
score of 3.3. Possible reasons for this could include:

May 2011 Osama Bin Laden was killed.

May 2011 Iceland’s volcano erupted resulting in ash cloud and cancelled airline flights in the UK.
•
Whilst December 2011 and March 2012 were the least ‘anxious’ months with a score of 3.0 , a possible reasons for this
could be:
 In December 2011 it was announced that the UK economy grew slightly faster than expected in the third
quarter .
•
There were no significant differences found for the ‘worthwhile’ question.
In order to investigate monthly variation in more detail several years worth of APS data would be needed. This
is something ONS will re-visit when more data is available.
24
Days Of The Week Results
Interviews are not normally conducted on a Sunday, therefore respondents answering on a Sunday were excluded from this
analysis due to a very small sample size.
•
Overall, how satisfied are you with your life
7.9
7.8
7.7
7.6
7.5
7.4
7.3
7.2
7.1
7.0
6.9
7.5
7.5
7.5
nowadays?1
7.5
7.5
7.4
•
•
•
Monday
Tuesday
Wednesday
Thursday
Friday
Saturday
•
Overall, to what extent do you feel that the things you do in your life
are worthwhile?1
7.9
7.8
7.7
7.6
7.5
7.4
7.3
7.2
7.1
7.0
6.9
7.8
7.7
7.7
7.6
7.7
7.6
•
•
Monday
Tuesday
Wednesday
Thursday
Friday
Other studies have found significant differences
in responses depending on which day of the
week the questions were asked. Therefore we
looked at whether that also arises in the OPN
data.
There were no significant differences found
between days of the week for life satisfaction
ratings.
Little fluctuations were present over the week.
There was significant difference at the 5% level
for days of the week for worthwhile ratings.
The Tuesday mean score was significantly
lower than Mondays, Wednesdays and Fridays
and the Wednesday mean score also
significantly higher than Thursday’s.
For more information on the significant
differences between days of the week see in
annex – Days of the week analysis – day on
day significant differences.
For further information see annexes J and K.
Saturday
1. Data complied using one year’s data from September 2011 to June 2012.
25
Days Of The Week Results
The ‘Happy’ and ‘Anxious’ questions ask 'how happy/ anxious did you feel yesterday'. Therefore when answering respondents are
thinking of the previous day. These charts show the days of the week that their answer referred to and not the day of the interview.
Overall, how happy did you feel yesterday?1
7.9
7.8
7.7
7.6
7.5
7.4
7.3
7.2
7.1
7.0
6.9
•
A significant difference at the 1% level was
found between days of the week for the
‘happy yesterday’ question.
7.3
•
People were significantly happier on a
Sunday compared to other days of the week,
however, this difference was not as strong
between Sundays and Fridays.
Friday
•
A significant difference at the 1% level was
also found between days of the week for the
‘anxious yesterday’ question.
3.3
•
The findings were similar to the ‘happy
yesterday‘ question: mean scores were
significantly lower on a Sunday (meaning
people were less anxious) compared to
other days of the week. The difference was
significant at the 1% level.
•
Respondents were the most anxious on a
Friday (3.3).
•
For further information on significant
differences see annexes J and K.
7.5
Sunday
7.2
7.2
7.2
7.2
Monday
Tuesday
Wednesday
Thursday
Overall, how anxious did you feel yesterday?1
3.5
3.4
3.3
3.2
3.1
3.0
2.9
2.8
2.7
2.6
2.5
3.2
3.2
3.2
3.1
2.7
Sunday
Monday
Tuesday
Wednesday
Thursday
1. Data complied using one year’s data from September 2011 to June 2012.
Friday
26
Days Of The Week – Overview of Results
•
The findings showed that the affect questions are more volatile than the
evaluative (life satisfaction) or eudemonic (worthwhile) questions.
•
There is evidence of a weekend/weekday effect suggesting:
o
that people are happier and less anxious at the weekend than during
the week. However, there are fewer respondents on weekends (due to
interviewing practices) which may have affected results.
•
Interviews are not normally carried out on a Sunday with a few exceptions.
Respondents answering on a Sunday were therefore excluded from days of
the week analysis due to a very small sample size. This means that no data
are available for respondents answering affect questions in relation to
Saturdays.
•
Our results replicate those of others showing a significantly different pattern
of affect responses at the weekend than on weekdays, see links below:
http://www.tandfonline.com/doi/abs/10.1080/17439760.2012.691980?tab=pe
rmissions & and http://www.bbc.co.uk/news/health-19316104
27
Annexes
28
Alternative Question Overview – Annex A
29
Interviewer Led Versus Self-Completion Interviews
Means and Response Rates – Annex B
For further information and a full copy of the Initial investigation into Subjective Well-being from the Opinions Survey see
http://www.ons.gov.uk/ons/rel/wellbeing/measuring-subjective-wellbeing-in-the-uk/investigation-of-subjective-well-being-data-from-the-onsopinions-survey/index.html
30
Face To Face Versus Telephone
Interviews By Average Means – Annex C
For further information and a full copy of the First ONS Annual Experimental Subjective Well-being
Results see http://www.ons.gov.uk/ons/rel/wellbeing/measuring-subjective-wellbeing-in-the-uk/firstannual-ons-experimental-subjective-well-being-results/index.html
•
Dolan, Paul and Kavetsos, Georgios (2012) analysed ONS data from the APS and found that individuals
consistently reported higher subjective well-being over the phone than compared with face-to-face
interviews.
'Happy talk: mode of administration effects on subjective well-being‘
http://eprints.lse.ac.uk/45273/
31
Question Order Analysis Response Rates
and Sample Sizes – Annex D
32
Show Cards Analysis Response Rates
and Sample Sizes – Annex E
33
‘Yesterday’ Question Analysis Response
Rates and Sample Sizes – Annex F
34
Question Wording Analysis Response
Rates and Sample Sizes Annex G
35
Month On Month Analysis Response
Rates and Sample Sizes - Annex H
36
Significant Differences Highlighted in the
Month on Month Data – Annex I
37
Days Of The Week Analysis – Day On Day
Significance Testing – Annex J
38
Days Of The Week Analysis Response
Rates and Sample Sizes – Annex K
39
Notes on Significance Testing
Significance Tests used for Well-being Split Trial Analysis
Parametric statistical tests have been used to test the significance of the
differences between different conditions. Parametric tests rely on assumptions
about the shape of the distribution (i.e. assume a normal distribution) in the
underlying population and about the form or parameters (i.e. means and
standard deviations) of the assumed distribution. Nonparametric statistical
procedures have the advantage of no or few assumptions about the shape or
parameters of the population distribution from which the sample was drawn;
however it is difficult to take account of the survey design when using these
procedures. Therefore, parametric tests were used.
Studentised t-test
The t-test assesses whether the means of two groups are statistically different
from each other.
40
Notes on Significance Testing continued
F Test - ANOVA
The F-test in one-way analysis of variance (ANOVA) is used to assess whether the
means of three or more groups are statistically different from each other. The advantage
of the ANOVA F-test is that we do not need to pre-specify which treatments are to be
compared, and we do not need to adjust for making multiple comparisons. The
disadvantage of the ANOVA F-test is that if we reject the null hypothesis, we do not
know which treatments can be said to be significantly different from the others. In this
situation it is necessary carry out post hoc analysis.
The formula for the one-way ANOVA F-test statistic is
or
Note that when there are only two groups for the one-way ANOVA F-test, F = t2 where t
is the Studentised t statistic.
41
Notes on Significance Testing continued
Two way, between subjects ANOVA
This is an extension of the F test ANOVA which deals with situations when there is more
than one condition. This was used to analysis the June data in which there was a
wording condition and a showcard condition. In this situation, there may also be an
interaction effect. The analysis of variance can be presented in terms of a linear model,
Xijk = µ + αi + βj + αiβj + εijk
Where denotes the overall mean, denotes the main effect of condition A (i.e. wording),
denotes the main effect of condition B (i.e. showcards), denotes the interaction effect
and captures an extra difference due to the combination of condition A and condition B
that is not accounted for in the main effects.
42
Notes on Significance Testing continued
Assumptions
All of these tests have the same assumptions:
• Independence of observations – the samples for the different conditions
are independent of each other
• Normally distributed data
• Homoscedasticity - variances of the different populations are assumed to
be equal.
The independence assumption has been met. Although the data is not
normally distributed, the tests are robust with respect to non-normality if the
skew in the data is not too extreme (as with this data). The variances of the
different conditions are similar for most of the analysis. In situations where
they were not, for example the days of the week analysis (where the variance
of Sunday is different from the others days) that condition was removed for
some of the analysis.
43
Notes on Significance Testing continued
Sample sizes
Sample sizes vary across the different split trials. Some split trials were only
tested for one month, meaning sample sizes were small and significance test
results should therefore be interpreted cautiously.
The larger the sample used in a significance test the more robust the result. A
larger sample will also allow testing to detect very small statistically significant
differences.
Whole numbers
Whole (unrounded) numbers were used in significance testing in this analysis.
44
Notes on the Opinions & Lifestyle Survey (OPN)
•
The OPN uses a random probability sample stratified by: region, the proportion of households
with no car, the proportion in National Statistics Socio-economic categories one to three and the
proportion of people aged over 65 years. In common with other ONS social surveys, it uses the
Royal Mail’s small user postcode address file to draw the sample from across Great Britain.
•
An initial sample of 2,100 addresses is drawn each month and advance letters are sent to all
addresses giving a brief account of the survey. Participation is purely voluntary and interviewers
only call at addresses where no refusal has been made to the advance letter. The interviewer will
make up to 20 calls at an address at different times in the day and the week to try to make
contact, after which the address is marked as a non-contact.
•
The interviewing period starts in the first week of the calendar month and continues for the
duration of the month in question. The interviewer uses a Kish grid to randomly select one of the
adults (aged 16 and over) living within the household for interview. All interviews are carried out
face-to-face (except for a very small number of telephone reissues) by ONS interviewers trained
to carry out National Statistics surveys.
•
The final achieved sample is around 1,100 adults (aged 16 and over) per month with an
approximate overall survey response rate of around 60 per cent.
•
The allocation of the split trails in the OPN was based on address number, each address number
is allocated to a sample. All interviewers have all versions of the trial questions, the version used
depends on the addresses that they have been allocated.
45
Notes on the Opinions & Lifestyle Survey (OPN)
continued
•
All estimates in this report are weighted. By weighting the estimates, we
ensure that they are more representative of the population but with the
assumption that those people who did not respond to the survey would
provide on average the same ratings of subjective well-being as those that do.
•
There are two weights in the Opinions Survey, firstly a weight that adjusts for
the differences in the probability of an individual being selected due to
different household sizes and sample design and secondly a weight that
calibrates the sample so that it is representative of the overall population
levels in Great Britain by age, sex and region. We used the second of these
two weights when carrying out this analysis.
•
For more information on the methodology of the Opinions and Lifestyle
Survey please find link to the survey user guide:
http://www.ons.gov.uk/ons/about-ons/who-we-are/services/opinions-andlifestyle-survey/opinions-and-lifestyle-survey--opn-.html
46
Feedback and Further Information
www.ons.gov.uk/well-being
[email protected]
Discuss National Well-being
http://www.statsusernet.org.uk
47