Childhood Obesity in the UK: Is Fast Food a Factor? Peter Dolton

Childhood Obesity in the UK:
Is Fast Food a Factor?
Peter Dolton
(Royal Holloway College, University of London
and
Centre for Economic Performance, London School of Economics)
MOTIVATION
Source: Brunello et al Economic Policy (2008)
•Annual
Reviews
Motivation
• Obesity is 2nd biggest cause of death in the UK
(after smoking).
• 23% of men obese and 24% of women [2008]
• Wanless report obesity costs - £3.6 billion a
year.[2007]
• Obesity rates have risen by 65% for boys and
51% for girls in the last 10 years. [up to 2005]
• 10% of kids obese by the time of starting primary
school (17% in deprived areas) [2008]
• Fat kids make fat adults.
KEY QUESTIONS
Main Research Questions:
• Can we explain what causes childhood obesity
(and why it is rising)?
Main Research Questions:
• Can we explain what causes childhood obesity
(and why it is rising)?
• Does Fast Food have a causal impact on
childhood obesity (which can be identified)?
Main Research Questions:
• Can we explain what causes childhood obesity
(and why it is rising)?
• Does Fast Food have a causal impact on
childhood obesity (which can be identified)?
• Does Fast Food Outlet Proximity have a
causal impact on Childhood Obesity?
ANSWER
• The answer is NO!!!!! – There is no causal
link between fast food proximity and
obesity
• Note – this does not mean that I am
saying FF if you eat enough of it does not
make you fat.
• It merely says that having a FF outlet near
does not cause you to be fat.
How Do We Answer the Question?
• To answer this question properly we would
need a controlled experiment – with some
children exposed randomly to FF and
others not.
• We would also need to know exactly what
else the children ate to control for
variations in eating patterns and calorie
intake.
• Such data does not exist.
Currie, Della Vigna, Moretti and
Pathania AEJ(P),NBER 2009
• ‘Taken together, the weight of the
evidence is consistent with a causal effect
of fast-food restaurants on obesity rates
among 9th graders…’
• “We find that among 9th grade children, a
fast food restaurant within one tenth of a
mile from a school is associated with at
least a 5.2% increase in obesity rates”.
WRONG!!
• US is ‘full saturated’ with FF and so no
control group.
• They only find their effects at .1 of a mile –
in the school playground!!!
• Hence I will look at the nearest thing we
have to an experiment.
UK Cohort Data
• 1946 NSHD Cohort
On food rationing up to age of 8.
• 1958 NCSD Cohort
“Control cohort”
• 1970 BCS Cohort
“Fast Food Cohort”
0
.05
.1
.15
.2
Male BMI at age 16 by Cohort
10
15
20
NSHD
25
Male BMI at 16
NCDS
30
35
BCS
0
.05
.1
.15
Female BMI at age 16 by Cohort
10
15
20
25
Female BMI at 16
NSHD
NCDS
30
35
BCS
Fraction Obese by Cohort at Age
15/16.
Girls
Boys
NSHD - 1961
0.06
0.04
NCDS- 1974
0.06
0.05
BCS- 1986
0.09
0.09
WHAT MY RESEARCH DOES
• Takes a cohort of children born in 1970
(BCS 1970) and follows them through to
adulthood.
• Merges all their data with information on
WHERE and WHEN ALL fast food outlets
opened in Britain.
• Estimates the effect of fast food outlet
proximity on children’s BMI.
BCS70 DATA
BCS70 Follow-ups
BCS70 Follow-ups and sources of information 1970-2000
BBS
(1970)
Birth
Mother
CHES
(1975)
5
CHES
(1980)
10
Youthscan
(1986)
16
BCS70
(1996)
26
BCS70
(2000)
30
— Subject
— Subject
— Parents — Parents — Parents
School
Tests
—
Tests
— School
—
Tests
Medical — Medical — Medical — Medical
16,135a
13,135
Subject
— Subject
14,875
11,628
9,003
11,261
Note: a Achieved Sample – when at least one survey instrument partially completed.
Fast Food Data
• Collected data on all the fast food outlets
over 1968-1986 and their year of opening
and their exact location.
• Includes:
– McDonalds
– Wimpy
– Burger King
– Kentucky Fried Chicken
What counts as Fast Food??
• Use Counter service
• Kids can buy it – served at source to eat
immediately, at any time of the day, and take
away.
• Fish and Chips was everywhere – so is assume
to be a constant background effect – (as is
changing supermarket food – like microwave
meals)
• Fish and Chips only available at ‘frying times’.
When did they open?
0
.1
.2
.3
Opening of UK Fast Food Outlets 1968-1986
1968
1970
1972
1974
1976
1978
1980
1982
Year of Opening of Fast Food Outlet
Source: Author's Calculations
1984
1986
How can we measure Fast Food
Treatment?
• Distance to nearest outlet.
• Number of outlets
• Currie et al (2009) use dummy if there is a
FF outlet with .1 of a mile. (Also measure
obesity as fraction in class x year who are
obese.)
• I use Distance x Years of opening – to
give intensity of treatment.
Key Explanatory Variables
•
•
•
•
•
•
•
Psychological Problems – ‘Any evidence of any emotional or behavioural
problem since 10’ as reported by doctor.
Physical Handicap – ‘Any evidence of significant illness, developmental
problem, defect or handicap’ as reported by doctor.
Exercise – Participation in any sport 3 times a week or more – self
reported.
High Calorie Diet – Consumes any of: hamburger/beef burger, pudding,
butter, chocolate, sweets, ice cream, cake or buns, biscuits, chips,
takeaway, white bread more than three times a week – reported by mother.
Eating Problems – Parental reported eating or appetite problems.
Screen Hours – Total of TV time, DVD time and computer time in a day.
Teen Eats Takeaway – Teen eats Takeaway meal as reported by the
mother.
Table 4 Determinants of Age 16 BMI for Girls in BCS70.
Fast Food Treatment
Intensity
BMI at 10
(1)
0.003
(0.005)
Mother's BMI
(2)
-0.005
(0.003)
0.818
(0.028)***
0.062
(0.016)***
Asian
European
Psychological
Problems
Physical Handicap
Eating Problems
(3)
-0.004
(0.003)
0.818
(0.028)***
0.063
(0.016)***
0.352
(0.736)
0.375
(0.636)
-0.251
(0.335)
-0.168
(0.192)
0.005
(0.132)
High Calorie Diet
Exercise
Smoker
Mother Works
Inner London
(4)
-0.001
(0.004)
0.824
(0.030)***
0.051
(0.017)***
0.315
(0.841)
0.169
(0.727)
-0.238
(0.384)
-0.177
(0.209)
0.071
(0.142)
0.168
(0.185)
0.341
(0.323)
0.842
(0.928)
-0.118
(0.116)
-1.173
(0.788)
Screen Hours
(5)
-0.005
(0.007)
0.673
(0.050)***
0.112
(0.032)***
1.258
(1.625)
0.861
(1.518)
-0.052
(0.661)
-0.192
(0.355)
0.113
(0.247)
0.010
(0.320)
0.027
(0.590)
-0.154
(1.493)
-0.241
(0.198)
Constant
21.539
(0.071)***
5.777
(0.538)***
5.400
(0.831)***
5.770
(0.932)***
0.093
(0.046)**
5.919
(1.772)***
Observations
R-squared
2668
0.00
2111
0.32
2070
0.32
1798
0.31
694
0.25
Table 5
Determinants of Age 16 BMI for Boys BCS70
Fast Food Treatment
Intensity
BMI at 10
(1)
0.003
(0.005)
Mother's BMI
(2)
-0.003
(0.004)
0.812
(0.026)***
0.106
(0.015)***
Asian
European
Psychological
Problems
Physical Handicap
Eating Problems
(3)
-0.002
(0.004)
0.804
(0.027)***
0.100
(0.015)***
-0.441
(0.720)
-0.395
(0.586)
0.897
(0.384)**
-0.014
(0.216)
0.016
(0.136)
High Calorie Diet
Exercise
Smoker
Mother Works
Inner London
(4)
-0.004
(0.004)
0.794
(0.027)***
0.108
(0.015)***
-0.750
(0.821)
-0.860
(0.697)
1.330
(0.406)***
0.072
(0.222)
-0.009
(0.139)
-0.115
(0.201)
0.636
(0.501)
0.240
(1.753)
0.237
(0.114)**
3.307
(1.447)**
Screen Hours
Constant
Observations
R-squared
21.539
(0.071)***
2668
0.00
5.339
(0.510)***
2273
0.34
5.946
(0.789)***
2221
0.33
6.227
(0.881)***
1907
0.36
(5)
-0.002
(0.006)
0.754
(0.041)***
0.091
(0.023)***
0.636
(1.417)
0.616
(1.282)
1.698
(0.760)**
-0.050
(0.316)
-0.242
(0.207)
-0.441
(0.306)
1.173
(0.738)
-1.906
(2.502)
0.176
(0.166)
0.082
(0.046)*
5.572
(1.550)***
914
0.32
Summary of BCS70 Findings @ 16
• NO FAST FOOD EFFECT
• Think of average kid with BMI of 20 then 1 unit of
BMI = 5% extra body mass.
• Both
– Own BMI at age 10 (1.25 units of BMI @ 10= 1 unit of BMI)
– Mothers BMI (10 units of Mothers BMI = 1 unit of BMI)
– Screen Time (approx 10 hrs per day= 1 unit BMI)
• For Boys only:
– Psychological Problems (extra .8-1.6 units of BMI)
– Inner London (up to 3 units of BMI)
– Mother Works (2.5 unit of BMI)
Extra Findings
• Number of Takeaways per week does not
correlate with BMI.
• Fizzy drinks alone is significant for boys –
5 drinks a day = 1 BMI
• Urban/Rural location is not significant.
• No Fast food effect at 26.
MAIN Conclusions
• In the UK Fast Food Proximity does not
have an impact on childhood BMI/Obesity.
• In so far as it is identifiable – the effect of
eating FF is not clear on BMI – we just
don’t observe EXACTLY what an
individual eats.
• Proximity of FF outlets does have an effect
on FF consumption.
Other Findings
• Main factors in BMI at 16 are own BMI at 10, Mothers
BMI.
• Exercise has no impact on BMI.
• Fizzy drinks do impact on BMI.
• Screen Time does impact on BMI.
• For boys emotional /behavioural problems are a factor.
• Increased FF consumption reduces consumption of Vit C
and fibre and increases consumption of crisps, fizzy
drinks and fatty food.
• FF outlet location is not related to where fat people live.
Tentative Implications
• Blaming Fast food outlets may be wrong.
• A huge fraction of overweight risk is from your
genes and your family environment.
• Onus must be on mothers to control kids diet
(and their own diet) when young.
• Fast food outlets don’t have information to locate
where fat people are – but they do cite where
poor people in work and ethnic minorities are.
EXTRA SLIDES
Outline of Talk
•
•
•
•
•
Motivation
Literature
Identification and Causality?
Measuring Childhood Obesity
BCS Cohort Data and Fast Food Data &
Fast Food Treatment.
• BCS 1970 Results
• Conclusions and Policy Implications
David Cameron 2008
“We talk about people being at risk of
obesity instead of talking about people
who eat too much and take too little
exercise – its as if obesity is a purely
external event like a plague or bad
weather”
Source: Anderson et al NBER paper (2007)
The Literature - Effect
Chou et al (2004) (J of H Econ)– state wide US
data find an effect.
Dunn (2008) (mimeo) IV of interstate exits –
10%increase in FF density – gives .33 of BMI
Davis and Carpenter (2009) (Am J of Pub Health)
Uses US data from CHKS over the 20022005 period.
Currie et al (2009) (AEJ(P), NBER) Use US data
from 1999-2007 for Californian kids.
The Literature – No effect
Burdette et al (2004) (Preventive Med) 7,000 Ohio kids.
Simmons et al (2005) (Int J of Obesity) find no
relationship in Australia.
Powell (2009) (J of H Econ)– uses US data from NLSY
data over teenagers from 1997-2003
Anderson and Matsa (2009) (mimeo) Use adjacent
highways as IV and find no effect.
What this paper does that is new:
• (To my knowledge) First study in the UK
• Set at a time when fast food outlets were growing
exponentially.
• Use distance from home not school.
• Know all - not just the closest restaurants.
• Know the timing of the opening of the fast food
restaurants - hence can work out a measure of ‘duration
intensity of fast food treatment'.
• Uses medically measured height and weight (not self
reported)
• Other detailed data – Mothers BMI, Screen time,
mothers work etc.
• New identification clarity.
Identification and
Causality?
Base Model
Currie et al (2009), Chou et al (2004), Burdette et al (2004), Powell
(2009)
BMI   2 R  X  2  X  2   2
*
•
•
•
•
BMI – is a measure of obesity
R is a measure of (EXOG) FF treatment
X is observable controls
X* unobservables
(1)
How are R & BMI measured?
• Chou et al (2004), Powell (2009) measure
R by number of restaurants at the state
level.
• Currie et al (2009) measure R as the
proximity of the FF outlets from school and
BMI as the % of kids in a class who are
‘fit’.
IV Model
Anderson and Mata (2009) & Dunn(2008)
R  X 1  X  1  Z1  1
(2.1)
BMI   2 R  X 2  X  2   2
(2.2)
*
*
• Assumption here is that R is endogenous and
we need to find IVs, Z for it. Anderson & Matsa
use interstate highways and Dunn uses exits.
Full Model
R  X 1  X  1  Z1  1
(3.1)
T  X 3  R1   3
(3.2)
*
BMI  2T  X  2  X  2   2
*
(3.3)
Here we make a distinction between
consumption of FF and Proximity of FF outlets.
T is a measure of Take-away consumption. May
be endog – in which case could use FF outlet
proximity, R as an IV.
Measuring Childhood
Obesity
How do we measure BMI?
• BMI – Quetelet Index
BMI 
Weight (k )
 Height (m)
2
BMI Classification for Adults
BMI Score
Category
< 18.5
Underweight
18.5 - 24.9
Normal Weight
25 - 29.9
Overweight
30 - 39.9
Obese
40
Morbidly Obese
BMI for kids
• Use the LMS tables generated by
Pan and Cole for the MRC
• Gives Obese BMIs as:
Age
Boys
Girls
10
20.27
21.35
16
24.9
25.85
Fraction Obese by Cohort at Age
15/16.
Girls
Boys
NSHD - 1961
0.067
0.044
NCDS- 1974
0.064
0.054
BCS- 1986
0.087
0.086
Local Examples
• Norwich
– 1978 Wimpy opened (Ipswich had one in 1977!!!)
– 1981 McDonalds opened
– 8 Fast food outlets by 1981
• Lancaster
• 1977 Wimpy opened
• 1980 McDonalds opened
• 4 Fast food outlets by 1980
Duration Intensity Treatment of
Fast Food
• Where Y is years since outlet k opened,
and d is distance from individual i. This
measures DITFF for less than 5 miles.
Yk
DITFF5  
k:d 5 d k
Example: Duration Intensity of
Fast Food Treatment
Additive Measure
8 year @ 1 mile ==2(1 year @ 4 miles)
0
.02
.04
.06
.08
Kernal Density of Total Fast Food Treatment Intensity
0
20
40
60
Fast Food Intensity
80
100
Reasons why it is interesting to look at
location of kids home and fast food
outlets:
• 1. At least 50% of the days in a year kids dont go to
school if we count weekends and holidays and absence.
They are only there for 6 hours and all but 1 are in
lessons. So only around 2-3% of time can get FF at
school.
• 2. Only 2.2% of primary kids age 10 in 1980 ate their
midday meal outside school but not in the family home.
• 3. Only 15% of secondary kids age 16 in 1986 ate their
midday meal outside school but not in the family home.
Table 6
Determinants of Age 16 BMI for Girls BCS70: Currie et al Specification.
FF distance <.1 mile
FF distance <.25 mile
FF distance <.5 mile
FF distance <1 mile
(1)
_16BMI
-4.630
(1.683)***
1.551
(1.002)
0.271
(0.529)
-0.109
(0.259)
(2)
_16BMI
-0.508
(1.591)
-0.901
(0.860)
-0.003
(0.449)
-0.011
(0.216)
0.808
(0.027)***
0.060
(0.015)***
(3)
_16BMI
-0.427
(1.617)
-1.033
(0.903)
0.072
(0.455)
-0.049
(0.220)
0.807
(0.027)***
0.062
(0.015)***
0.603
(0.703)
0.509
(0.598)
-0.100
(0.323)
-0.103
(0.189)
0.010
(0.130)
(4)
_16BMI
-0.231
(2.000)
-1.463
(1.084)
0.175
(0.483)
-0.009
(0.235)
0.810
(0.029)***
0.051
(0.017)***
0.808
(0.834)
0.447
(0.720)
-0.101
(0.374)
-0.111
(0.206)
0.063
(0.140)
0.217
(0.181)
0.344
(0.321)
0.847
(0.924)
-0.061
(0.114)
20.945
(0.066)***
2598
0.00
5.962
(0.523)***
2172
0.31
5.446
(0.792)***
2130
0.32
5.675
(0.913)***
1852
0.31
BMI at 10
Mother's BMI
Asian
European
Psychological
Problems
Physical Handicap
Eating Problems
High Calorie Diet
Exercise
Smoker
Mother Works
Screen Hours
Constant
R-squared
Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
(5)
_16BMI
-0.927
(3.009)
-1.637
(1.616)
-0.104
(0.729)
-0.464
(0.390)
0.661
(0.049)***
0.119
(0.032)***
1.620
(1.625)
0.738
(1.508)
0.137
(0.637)
-0.075
(0.350)
0.130
(0.243)
0.050
(0.318)
0.009
(0.572)
-0.193
(1.482)
-0.127
(0.195)
0.077
(0.045)*
6.074
(1.758)***
710
0.26
Table 7
Determinants of Age 16 BMI for Boys BCS70: Currie et al Specification.
FF distance <.1 mile
FF distance <.25 mile
FF distance <.5 mile
FF distance <1 mile
(1)
_16BMI
2.087
(1.809)
-0.522
(0.874)
-0.426
(0.532)
0.394
(0.263)
(2)
_16BMI
0.496
(1.643)
-0.414
(0.731)
-0.550
(0.462)
0.183
(0.225)
0.804
(0.026)***
0.108
(0.015)***
(3)
_16BMI
0.508
(1.629)
-0.407
(0.725)
-0.610
(0.460)
0.282
(0.228)
0.794
(0.026)***
0.104
(0.015)***
-0.444
(0.717)
-0.353
(0.577)
0.696
(0.368)*
0.007
(0.209)
0.020
(0.133)
(4)
_16BMI
-0.452
(1.847)
-0.296
(0.710)
-0.116
(0.458)
-0.112
(0.227)
0.787
(0.027)***
0.111
(0.015)***
-0.716
(0.821)
-0.774
(0.691)
1.013
(0.386)***
0.089
(0.216)
-0.020
(0.137)
-0.098
(0.197)
0.656
(0.510)
0.204
(1.751)
0.233
(0.112)**
21.549
(0.066)***
2752
0.00
5.423
(0.499)***
2353
0.34
5.977
(0.772)***
2299
0.33
6.192
(0.868)***
1978
0.36
BMI at 10
Mother's BMI
Asian
European
Psychological
Problems
Physical Handicap
Eating Problems
High Calorie Diet
Exercise
Smoker
Mother Works
Screen Hours
Constant
R-squared
Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
(5)
_16BMI
0.000
(0.000)
-0.199
(0.991)
-0.549
(0.669)
0.277
(0.333)
0.769
(0.040)***
0.093
(0.023)***
0.569
(1.423)
0.609
(1.274)
1.501
(0.723)**
-0.085
(0.312)
-0.275
(0.202)
-0.561
(0.298)*
1.200
(0.734)
-1.986
(2.484)
0.145
(0.163)
0.082
(0.045)*
5.308
(1.531)***
936
0.34
Full Model
R  X 1  X  1  Z1  1
(3.1)
T  X 3  R1   3
(3.2)
*
BMI  2T  X  2  X  2   2
*
(3.3)
Here we make a distinction between
consumption of FF and Proximity of FF outlets.
T is a measure of Take-away consumption. May
be endog – in which case could use FF outlet
proximity, R as an IV.
Eqn (3.2) First Stage IV
FF distance <.25 mile
Fast Food Treatment
Constant
Observations
R-squared
Total
0.070
(0.034)**
0.000
(0.000)**
0.067
(0.003)***
10324
0.02
Girls
0.074
(0.046)
0.000
(0.000)*
0.067
(0.004)***
5145
0.02
Boys
0.063
(0.053)
0.000
(0.000)*
0.068
(0.004)***
5179
0.02
Table 11 Marginal Effects Estimates for Different Fast Food Treatment Intensity
Measures. Boys in the BCS70 at 16.
Fast Food Treatment
Measure
Intensity
Nearest
School Area
Teen Eats Takeaways IV
(1)
(2)
(3)
(4)
(5)
0.003
(0.005)
0.004
(0.009)
0.002
(0.006)
0.991
(4.332)
-0.003
(0.004)
-0.009
(0.008)
-0.005
(0.006)
-3.359
(3.348)
-0.002
(0.004)
-0.007
(0.008)
-0.005
(0.006)
-2.506
(3.156)
-0.004
(0.004)
-0.011
(0.008)
-0.009
(0.006)
-2.734
(2.917)
-0.002
(0.006)
-0.019
(0.008)
-0.019
(0.010)*
-0.104
(9.203)
Table 12 Marginal Effects Estimates for Different Fast Food Treatment Intensity
Measures. Girls in the BCS70 at 16.
Fast Food Treatment
Measure
Intensity
Nearest
School Area
Teen Eats Takeaways IV
(1)
(2)
(3)
(4)
(5)
0.003
(0.005)
0.004
(0.009)
-0.003
(0.006)
-2.193
(3.026)
-0.005
(0.003)
-0.008
(0.006)
-0.006
(0.005)
-3.540
(2.536)
-0.004
(0.003)
-0.008
(0.006)
-0.006
(0.005)
-3.744
(2.724)
-0.001
(0.004)
-0.008
(0.009)
-0.001
(0.006)
-11.471
(18.914)
-0.005
(0.007)
-0.028
(0.015)*
0.000
(0.011)
19.038
(31.354)
Identification Issue
Does living near Fast Food outlet ‘cause’
you to be fat?
OR
Do Fast Food outlets cite near people who
are fat and/or do people who have a
propensity to be fat choose to live near FF
outlets?
Full Model
R  X 1  X  1  Z1  1
(3.1)
T  X 3  R1   3
(3.2)
*
BMI  2T  X  2  X  2   2
*
(3.3)
Here we make a distinction between
consumption of FF and Proximity of FF outlets.
T is a measure of Take-away consumption. May
be endog – in which case could use FF outlet
proximity, R as an IV.
Trying to establish Causality?!
• Took all the Youth Cohort Surveys 1-7
over the whole period of the 1983-1992 to
characteristics of the LEAs:
Table 13 Explanation of Fast Food Duration Intensity by LEA Characteristics.
Duration Intensity Fast
Food
Mean Overweight
% Non Manual Fathers
% Owner Occupiers
Average Education Score
% Youth Unemployment
% Truancy
% Dads in Work
% One Parent Families
Duration Intensity Fast
Food
16.667
(22.489)
-52.185
(36.267)
23.958
(25.440)
2.553
(0.955)***
-565.945
(82.830)***
161.684
(51.942)***
-100.999
(36.099)***
112.402
(59.109)*
% Non-White
Constant
Observations
R-squared
65.339
(39.660)
96
0.55
Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
21.596
(16.853)
-21.448
(27.412)
-16.182
(19.655)
1.859
(0.720)**
-282.110
(70.809)***
40.614
(41.538)
-16.345
(28.890)
55.262
(44.799)
96.066
(11.556)***
16.240
(30.284)
96
0.75
Source from Cummins et al (2005) Int J of Behavioral Nutrition and Physical Activity.
Possible explanations of why I dont find
anything of a fast food effect:
1. Using home rather than school.
2. Just too early - the data which finds an effect is for the US and in the 2000s.
Powell uses US data from NLSY data over teenagers from 1997-2003
Davis (2009) Uses US data from CHKS over the 2002-2005 period.
Currie et al (2009) Use US data from 1999-2007 for Californian kids.
3. Its just a US effect - At present there is no evidence from any other country that there
is an effect.
Simmons et al (2005) find no relationship in Australia.
4. Methodology or claimed causality on other papers is dubious. How can we be sure
that fast food restaurants do not locate where fat people (or people who have a
propensity to consume more highly) live. i.e. fast food outlets locate where there
are already fat people rather than the location of the fast food outlet causes people
to become fat. All the papers purporting to find this effect do not satisfactorily
tackle this question.
5. My measure of intensity of treatment is wrong.
6.
There simply is no causal effect for proximity of fast to children's whereabouts.
7.
Takeaways do not induce BMI gain as kids substitute calories when they take fast
food – we do not have good enough data to tell us about how FF substitutes for
other food.
Table 14. Teen Version of What they Ate: Poisson Regression of Number of Times ate
Takeaways this Week.
Regressors: Ate Yesterday
Total portions of chips
Total bags of crisps
Total carbohydrate/starch score
Total vitamin c score
Total fibre score
Total carbohydrate/sugar score
Total fizzy drinks
Constant
0.260
(0.028)***
0.173
(0.017)***
-0.016
(0.011)
-0.014
(0.009)
-0.051
(0.009)***
0.069
(0.009)***
0.052
(0.015)***
-0.113
(0.046)**
4652
Observations
Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
Issues in Obesity Literature.
• Does maternal working play a role in childhood obesity?
(Anderson et al 19 )
• Is SES related to obesity? (Baum and Ruhm )
• Does mothers BMI impact on childs obseity (Parsons et
al )
• Height is related to Earnings. (Case and Paxson)
• BMI is negatively related to earnings for women.
Blanchflower and Sargent (1980) Morris
• Childhood obesity is not linked to adult outcomes (Viner
and Cole)
• Relation between screen viewing and obesity.
Other Geographical Studies
• Cummins & MacIntyre (2005) (Int J of Epidemiolgy)
• Cummins et al (2005) Int J of Behavioral Nutrition and
Physical Activity.
• Morland & Evenson (2009) (Health and Place)
The Data
• 1946 NSHD Cohort
On food rationing since up to age of 8.
• 1958 NCSD Cohort
“Control cohort”
• 1970 BCS Cohort
“Fast Food Cohort”
NHSD Follow-ups
• Socially structured sample of 5362
singletons born in one week in March
1946.
• Followed at Birth
• 2,4
response 95%
• 6,7,8,9,10,11,13,15
response 85%
NCDS Follow-ups
NCDS Follow-ups and sources of information 1958-2000
PMS
(1958)
Birth
NCDS1
(1965)
7
NCDS2
(1969)
11
NCDS3
(1974)
16
EXAMS
(1978)
20
NCDS4
(1981)
23
NCDS5
(1991)
33
NCDS6
(2000)
42
17,733a
16,883
16,835
16,915
16,906
16,457
15,600
15,145
Mother
— Parents — Parents — Parents
School
— School
— School
Tests
— Tests
— Tests
— School
Medical — Medical — Medical — Medical
Subject — Subject — Subject ————— — Subject — Subject — Subject
Census ————— — Census
Spouse/
Partner
Motherc
Children
17,414b
15,568
15,503
14,761
14,370
12,537
11,407
Notes:
a: Target Sample - immigrants with appropriate date of birth included for NCDS1-3.
b: Achieved Sample - at least one survey instrument partially completed.
c: This could be the Cohort Member, their Spouse, or Partner.
11,419
0
.05
.1
.15
.2
Male BMI at age 16 by Cohort
10
15
20
NSHD
25
Male BMI at 16
NCDS
30
35
BCS
0
.05
.1
.15
Female BMI at age 16 by Cohort
10
15
20
25
Female BMI at 16
NSHD
NCDS
30
35
BCS
0
.0002
.0004
.0006
.0008
Female Birth Weight by Cohort
1000
2000
3000
Birth Weight
NSHD
BCS
4000
NCDS
5000
0
.0002
.0004
.0006
.0008
Male Birth Weight by Cohort
1000
2000
3000
Birth Weight
NSHD
BCS
4000
NCDS
5000
Additional features of the data:
• Have a 'rationed' and 'control' generation to
examine the context of weight gain of teenagers
over the period 1962-1986.
• Have repeated measure of BMI - i.e. longitudinal
data and can model changes and identify
changes in intensity of fast food treatment on
weight gain.
• Know an extremely rich set of covariates
including mothers and father own BMI - no other
studies have access to such controlling data to
provide a context of genetic and family
environment factors.
Table 2 Determinants of Age 10/11 BMI by Cohort.
Male
Medical Problems
Birth Weight in gms
Mother Works
Social Class I
Social Class II
Social Class III
Social Class IV
NSHD
NCDS
BCS
Combined
0.313
(0.076)***
0.396
(0.079)***
0.001
(0.000)***
0.391
(0.082)***
0.084
(0.192)
-0.053
(0.136)
-0.189
(0.117)
-0.086
(0.137)
0.460
(0.049)***
0.157
(0.088)*
0.001
(0.000)***
0.120
(0.050)**
-0.050
(0.131)
0.091
(0.098)
0.125
(0.086)
0.117
(0.099)
0.358
(0.039)***
0.077
(0.056)
0.0004
(0.000)***
0.156
(0.039)***
-0.289
(0.095)***
-0.123
(0.068)*
-0.075
(0.060)
-0.080
(0.080)
14.838
(0.284)***
3920
0.03
14.995
(0.181)***
11567
0.02
15.431
(0.139)***
11287
0.02
0.394
(0.029)***
0.188
(0.042)***
0.001
(0.000)***
0.169
(0.030)***
-0.165
(0.074)**
-0.048
(0.054)
-0.029
(0.047)
-0.018
(0.057)
-0.774
(0.032)***
-0.339
(0.046)***
15.566
(0.107)***
26774
0.04
BCS Cohort
NSHD Cohort
Constant
Observations
R-squared
Table 3
Determinants of Age 15/16 BMI by Cohort
Male
Medical problems
BMI at 10/11
Mother Works
Social Class I
Social Class II
Social Class III
Social Class IV
NSHD
NCDS
BCS
Combined
0.824
(0.061)***
0.019
(0.063)
0.894
(0.013)***
0.030
(0.062)
-0.185
(0.147)
0.008
(0.102)
-0.024
(0.089)
0.091
(0.106)
0.483
(0.044)***
0.035
(0.080)
0.787
(0.009)***
0.018
(0.047)
-0.297
(0.111)***
-0.147
(0.071)**
0.031
(0.059)
0.132
(0.080)*
0.617
(0.117)***
0.046
(0.126)
0.848
(0.028)***
-0.022
(0.117)
-0.100
(0.284)
0.184
(0.219)
0.133
(0.208)
0.572
(0.276)**
4.167
(0.237)***
3319
0.61
6.531
(0.157)***
8917
0.50
6.473
(0.521)***
2067
0.31
0.580
(0.035)***
0.047
(0.048)
0.816
(0.007)***
0.015
(0.036)
-0.248
(0.086)***
-0.073
(0.057)
0.013
(0.049)
0.157
(0.065)**
1.219
(0.053)***
-0.334
(0.045)***
5.956
(0.133)***
14303
0.50
BCS Cohort
NSHD Cohort
Constant
Observations
R-squared
Main Results of Cross Cohort
Comparison
•
•
•
•
•
•
At 10/11 mother works increases BMI
At 10/11 social class effects
At 10/11 Birth weight effects.
At 15/16 still social class effects
At 15/16 no mother works effect
At 15/16 BMI @ 10/11 is main driver.