view presentation - MS Powerpoint

advertisement
An analysis of the relationship between
time spent on active leisure and
educational qualifications
Shu-li Cheng
Centre for Research on Innovation and Competition (CRIC)
University of Manchester
Presentation to UPTAP Workshop, Leeds, 21st March 2007
Combining the strengths of UMIST and
The Victoria University of Manchester
Objectives
•
Applying zero-inflated modelling approach to estimate the amount of time
spent on infrequent activities in the UK 2000 Time Use Survey
– The typical distribution of time spent on an infrequent activity during
survey period is highly positively skewed and has a large proportion of
zeros
•
To examine the relationship between educational qualifications and active
leisure allowing for potential socio-economic and demographic
characteristics using the survey
– The effect of education on participation in active leisure
– The effect of education on the amount of time spent on active leisure
Combining the strengths of UMIST and
The Victoria University of Manchester
The UK 2000 Time Use Survey
•
The main aim of the survey is to measure the amount of time spent by
the UK population on various activities.
•
Stratified random sampling was used to collect the data
– People aged 8 years and over from the randomly selected household
were interviewed and were asked to record their activities in the
given diaries
» Note: only respondents aged 16 and over are included
in the following analysis
•
Two diaries per respondent
– One weekday and one weekend day
•
The diary was pre-divided into 10-minute intervals
– Primary and secondary activities were recorded
– Information on with whom the primary activity was carried out
Combining the strengths of UMIST and
The Victoria University of Manchester
Zero-inflated modelling (mixture model)
Pr(Y = 0) = p + 1  p 1   /  
Pr(Y = y) = 1  p 
-τ
 y   
1 

y! 

y=0
 1    
-τ
-y
y = 1, 2 …
Zeros are from two sources:
1. structural zeros:
p proportion of the respondents had never participated in the activity
2. sampling zeros:
(1 – p) proportion of the respondents participated (or had an intention of
participating) in the activity; however, occurring by chance, they spent no
time on the activity
Combining the strengths of UMIST and
The Victoria University of Manchester
Response variables (1)
Walking
Actively participating in sports
•
Walks, rambles
•
Outdoor team games
•
Other outdoor hobbies (e.g. painting)
•
Non-team ball hitting sports
•
Running, jogging, cross-country, track and field
•
Golf
•
Fishing
•
Bowls
•
Martial arts
•
Swimming and other water sports
•
Keep fit, yoga, aerobics, dance practice
•
Cycling
•
Other outdoor sports
•
Other indoor sports
•
Horse rides
•
Hunting, shooting, fishing, etc.
•
Other participation in sport and active leisure activities
Include general “outdoors” variables
Include communication for the purposes of active leisure
Combining the strengths of UMIST and
The Victoria University of Manchester
Include general variables such as “other” active leisure or
“other” sport
Response variables (2)
Variable
Walking
Actively participating in sports
Observation
16576
16576
Note:
‘Walking’ contains 74% of zero observations
‘Active sports’ contains 89% of zero observations
Total number of households (cluster): 4881
Combining the strengths of UMIST and
The Victoria University of Manchester
Mean
13.3
11.8
Std. Dev.
36.3
46.4
Min
0
0
Max
1270
850
Explanatory variables
•
Educational qualifications
–
–
–
•
Sex
–
–
•
•
Combining the strengths of UMIST and
The Victoria University of Manchester
No (reference category)
Yes
Urban or rural household
–
–
•
Lowest 25%
Middle 25% (reference category)
Highest 25%
Whether respondent has disability
–
–
•
Married/cohabiting (reference category)
Not living with a spouse or partner
Number of children under age 18 in the household
Hours of paid work last week
Income
–
–
–
•
Man (reference category)
woman
Age
Marital status
–
–
•
•
•
Incomplete secondary or less (reference category)
Complete secondary
Above secondary
Urban/suburban (reference category)
Rural/semi-rural
Weekday or weekend
–
–
Weekday (reference category)
Weekend
Results: time spent walking
Secondary
Above secondary
Male
Age-46
Not married
No. child aged<18
Hours paid work
Income lowest 25%
Income highest 25%
Disability
Rural household
Weekend
Constant
Combining the strengths of UMIST and
The Victoria University of Manchester
Logit
Coefficient
-0.007
-0.159
-0.196
0.009
-0.359
0.093
0.012
-0.321
0.087
0.228
0.293
-0.397
1.178
z-value
-0.10
-2.18
-4.03
4.87
-5.81
3.28
8.33
-4.26
1.16
3.29
2.83
-9.48
14.60
Negative binomial
IRR
z-value
0.940
-1.14
0.891
-1.89
1.088
2.25
1.000
-0.33
1.003
0.06
0.972
-1.28
0.998
-1.34
1.149
2.32
1.085
1.14
0.921
-1.60
0.976
-0.33
1.164
3.92
Whether spent time on walking (logit model)
•
All explanatory variables are statistically significant.
•
Educational qualifications are important in predicting the likelihood of spending
time on walking.
– Those gaining qualifications above secondary level are significantly more
likely to spend time walking compared with those gaining no qualifications.
But there is no significant difference between secondary level and below
secondary.
•
Men are more likely to walk than women as well as not married than married.
•
The least income group is more likely to spend time walking than the middle
income group. But no significant difference between middle and highest groups.
•
The likelihood of spending time walking decreases with age, number of
dependent children, hours of work, as well as having disability and being in a
rural household.
•
The likelihood of walking is significantly higher at the weekend than on a
weekday.
Combining the strengths of UMIST and
The Victoria University of Manchester
Time spent walking (negative binomial model)
•
The amount of time spent walking is 1.1 times longer for a man, compared
with time spent by a woman.
•
The amount of time the lowest income group spent on walking is 1.1 times
longer than the middle income group.
•
Time spent walking at the weekend is 1.2 times longer than on a weekday.
Combining the strengths of UMIST and
The Victoria University of Manchester
Results: time spent actively participating in sports
Secondary
Above secondary
Male
Age-46
Not married
No. child aged<18
Hours paid work
Income lowest 25%
Income highest 25%
Disability
Rural household
Weekend
Constant
Combining the strengths of UMIST and
The Victoria University of Manchester
Logit
Coefficient
-0.230
-0.521
-0.436
0.007
-0.289
0.045
0.010
0.493
-0.218
0.345
-0.021
-0.084
2.326
z-value
-2.49
-5.27
-6.13
2.52
-3.31
1.13
4.99
4.30
-2.30
3.08
-0.16
-1.49
20.16
Negative binomial
IRR
z-value
0.912
-1.36
0.908
-1.30
1.492
7.76
0.998
-0.84
1.021
0.30
0.986
-0.51
0.997
-2.01
0.844
-2.10
0.935
-0.97
0.854
-1.74
1.002
0.02
1.318
5.74
Whether spent time on active sports (logit model)
•
There is a gradient among educational groups on the likelihood of spending
time on active sports.
– The most educated are the most likely group to spend time on active
sports, while the least educated are the least likely to participate.
•
Men are more likely to participate than women.
•
The likelihood of participating decreases with age as well as hours doing
paid work.
•
Not married people are more likely to spend time on active sports than
those married.
•
Similar to the impact of education, the effect of income shows an upward
gradient on the likelihood of participation.
•
Disability has a negative impact on participating in active sports.
Combining the strengths of UMIST and
The Victoria University of Manchester
Time spent on active sports (negative binomial model)
•
Education is not statistically significant in predicting time spent on active
sports after controlling for other explanatory variables.
•
Gender, hours of paid work, income and weekend are the significant
variables.
– The time a man spent on active sports is 1.5 times longer than a
woman.
– An additional hour of paid work reduces about 0.3% of time on active
sports.
– The amount of time spent by the lower income group is 84% of the
middle income group.
– The amount of time spent on active sports at the weekend is about 1.3
times longer than on a weekday.
Combining the strengths of UMIST and
The Victoria University of Manchester
Conclusion
•
•
•
•
•
The benefits of zero-inflated modelling approach
– Allowing for structural zeros and sampling zeros in the data makes the
model more general
– The influence of explanatory variables can be examined in both logit part
and negative binomial (Poisson) part of the model
The effect of educational qualifications is significant for participating in leisure
walking and in active sports after controlling for other explanatory variables;
however, no difference among educational levels on the amount of time spent.
Gender differences are found in active leisure. Men are more likely to
participate than women. They also spend longer time doing it.
Income influences both participation and the amount of time spent on active
leisure. The effect seems more obvious for the lower income group.
Day of the week is highly significant in predicting time spent on active leisure.
– More time spent over the weekend on active leisure than during the week
after taking into account of hours of paid work and other explanatory
variables.
Combining the strengths of UMIST and
The Victoria University of Manchester
Download