the relationship between food consumption and socio...
TRANSCRIPT
The relationship between food consumptionand socio-economic status:
evidence among the British population
Paola De Agostini∗†
Abstract
This paper investigates the relationship between nutrition andsocio-economic status among British youths. It describes the dynam-ics of consumption over age and time using data from the BritishNational Food Survey (NFS) covering the period 1975-2000. Dailyintakes-age and food-age relationships for men and women are esti-mated by solving a non-linear least square model with a roughnesspenalty function approach. Focusing on young age groups, trends ofconsumption over the 25-year period of study and the cohorts effecthave been explored across three classes of age. Finally, an explorationof specific trend variations in eating habits has been implemented con-trolling for income distribution, region of residence, household struc-ture and the presence of children.
∗Institute for Social and Economic Research (ISER), University of Essex - UK; email:[email protected]
†Department of Economics, University of Verona - Italy.This paper is an extract of my PhD thesis. It would not have been possible without
the outstanding supervision which I have been receiving from Marco Francesconi. Thanksare also due to Stephen P. Jenkins, Federico Perali, Andrew Chesher and the participantsat the EEA Summer School in Microeconometrics and at the 97th EAAE Seminar forhelpful suggestions. I am grateful to the data depositors of the NFS (Department forEnvironment, Food and Rural Affairs), and to the UK Data Archive, University of Essex,for providing access to the data. Finally, I would like to thank Chimera, Institute forSocio-Technical Innovation and Research at the University of Essex for funding. Theauthor alone is responsible for errors and opinions.
1
1 Introduction
In a growing number of industrialized Countries obesity has become a phe-
nomena of unprecedented proportions. In 2000, 300 million adults were obese
and 700 million were classified as overweight around the world (OECD, 2003).
There are however notable differences in obesity rates across countries. In
the United Kingdom, the obesity rate among adults has tripled over the last
twenty years to stand at 22% in 2001. This is higher than in nearly all other
OECD countries, but lower than in the United States (31% in 1999) and
Mexico (24% in 2000), and comparable to Australia (21% in 1999). Accord-
ing with the OECD data, if the average rate of increase in the prevalence of
obesity between 1980 and 1998 continues, over one fifth of men and about a
quarter of women in England will be obese by 2005, and over a quarter of
all adults by 2010. This would bring levels of obesity in England up to those
experienced now in the United States.
Obesity and overweight are not only adults issues. On the opposite, they
are increasing also among children and adolescents. In 1995, 18 million under
five and 155 million children between five and seventeen years old in 2000
were classified as overweight around the world. Over the past two decades the
number of overweight children and teens nearly double. The Surgeon General
Report for the United States reports that in 1999 13% of children aged 6 to
11 years and 14% of adolescents aged 12 to 19 years were overweight. This
prevalence has nearly tripled for adolescents in the past two decades. Even if
European and British indicators are not as high as in the US, the proportion
of children classed as overweight or obese increased between the mid-80s and
mid-90s. Studies looking at boys and girls in England and Scotland, aged
between 4 and 11, show that approximately 5% of British children in 1984
were overweight. A decade later, 9% were overweight.
The primary concern of overweight and obesity is one of health and not
appearance. Overweight and obese people are at risk for a number of health
problems including heart disease, diabetes, high blood pressure, and some
forms of cancer. Overweight adolescents have a 70% chance of becoming
overweight or obese adults then their non overweight counterparts. This
2
increases to 80% if one or more parent is overweight or obese (genetic predis-
position). Overweight children and adolescents compared to children with a
healthy weight are at higher risk of developing such diseases listed above. On
the other hand, the most immediate consequence of overweight as perceived
by the children themselves is social discrimination. This is associated with
poor self-esteem and depression.
The second concern of overweight and obesity is one of health service
costs. As reported by the Summary of Intelligence on Obesity (2004) the
cost of obesity in UK is estimated at 3.7 billion per year and 7.4 billion when
adding the cost of overweight. Moreover, considering the time lag between
the onset of obesity and related chronic diseases, researchers suggest that
the rise in obesity that has been occurring in the last 20 years, will have
substantial implications on future costs.
The main causes of obesity have been identified on excessive consump-
tion (unbalance diet), lack of physical activities, genetic predisposition and
disorders that affect the normal bodily functions as metabolism and growth.
Leaving genetic predisposition aside, it remains unclear the exact relative
responsibility of an unbalance diet and reduce exercise.
Food choices depend on many social factors such as history, culture, and
environment, as well as on energy and nutrient needs.
The modern sedentary life, the growing number of fast food and restau-
rant, technological changes and women participation in the labor market
are often popular justifications of the growing calories consumption and the
reduce physical activity.
Lakdawalla and Philipson (2002) analyze the energy equation from the
energy expenditure point of view. They concluded that a sedentary worker
will be heavier than someone in a highly active job. Further, they estimated
that about 60% of the total growth in weight in the United States may be due
to demand factors, such as a decrease in physical activity, and about 40% is
due to expansion in calories, potentially through increased food abundance
due to agricultural innovations.
Increasing availability of fast food and ready meals have changed the rela-
tive costs of meals preparation and consumption increasing the consumption
3
of some nutrient intakes as saturated fats, sugars and calories also because
of the bigger portion size.
Women labor participation increases the cost of time, reducing the time
spent cooking healthy home-meals and contributes at increasing number of
meals eaten out and consumption of ready meals.
Although few studies analyze the implication of economic and technolog-
ical changes as possible reasons of the change of food choices (Cutler et al.,
2003; Chou et al., 2001) finding the expected positive relationship between
obesity and the number of fast food per capita, there is little evidence on
the relation between the increasing number of women at work and the rising
in demand for eating out. Chou, Grossman and Suffer (2001) argue that
expanding labor market opportunity for women have resulted in significant
increases in families’ command of real resources and higher standards. Also
Cutler, Claser and Shapiro reject the theory for which the increasing number
of women at work have increased the demand for eating out, pointing out that
the main reason for increasing calorie, saturated fat and sugar consumption
is mainly consumption of snacks outside the main meals.
Another important question that has been addressed is whether differ-
ences in nutrition depend from differences in income. Are poorer eating
worst than wealthier? In fact, family choices associated with health and
the processes of biological programming are strongly mediated by their so-
cial context. Of course, the most powerful aspect of social context associated
with health is poverty because it is often associated with poor diet and conse-
quent poor likelihood of growth and development (Baeker 1998), with raised
risk of infection. But also low level of education are associated with poor
health behavior, for example in terms of diet and exercise, and in raised of
overweight and obesity.
Curry and Bhattacharya (2000) use data on Americans youth to deter-
mine the causes of poor nutritional outcomes. Their finding suggest that poor
nutrition is a problem for American youth neither entirely related to a lack
of household resources nor to family background. They measure information
about the relationship between nutrition and health through education and
age of head of household and the content of television programming. They
4
argue that information and technology matter. TV viewing has a negative
effect on nutrition outcome both because of the content of the programming
and because it reduces physical activities. Similar results are found also by
Chou, Grossman and Suffer (2001), who argue that wealthier and more edu-
cated people are less likely to become obese or overweight, whereas Hispanic
and black individuals are more likely to suffer from obesity.
Cutler, Glaeser and Shapiro (2003) find that obesity and income are neg-
atively related, mainly because incomes were not increasing greatly at the
bottom of the income distribution in United States in the period they con-
sidered.
Case, Lubotsky and Paxson (2002) point out the importance for future
research of explaining the mechanisms that underlie the relationship between
income, nutrition and children’s health outcomes. Their findings show that
the robust relationship between children’s health status and family income
may be due to differences in parent’s and child’s health-related behaviors at
different levels of income. Choices made concerning how often a child sees a
doctor or about his eating habits, may have both short-term and long-term
health implications. Many of these behaviors are correlated with socioeco-
nomic status, and so may potentially explain at least part of the association
between children’s health and household income. In their results, inclusion of
these health-related behaviors reduces the observed income gradient among
Americans, but only slightly. Healthful diets help children grow, develop, and
do well in school. Food choices also can help to reduce the risk of chronic
diseases, such as heart disease, certain cancers, diabetes, stroke, and osteo-
porosis, that are leading causes of death and disability among Americans.
Good diets can reduce major risk factors for chronic diseases factors such as
obesity, high blood pressure, and high blood cholesterol.
The UK Department of Health recognizes that promoting diet changes
and physical activities increase would help, but it is difficult to be done be-
cause it would imply changes in preferences and consumer behaviors. They
suggest that preventing obesity and overweight in childhood is perhaps a
more effective approach in the long term. Child health is of the greatest
importance for the future health of a nation, not only because today’s chil-
5
dren grow up to become the next generation of parents and workers, but also
because recent research in child health shows that early life health is, for
each child, the basis of health in adulthood and nutrition is one of its basic
determinants. As the UK Food Standard Agency Report points out:
A healthy balance of foods provides the energy and nourishment
everyone needs to survive and to enjoy life. Eating too little food
soon leads to illness, but eating too much or the wrong balance
of foods can lead to problems in the long term. So it is important
to get the balance right both in the amount and in the types of
foods eaten. A healthy and balance diet in childhood can reduce
the risk in anaemia and dental decay. In the longer term, it can
help to prevent ill health later in life. For example, it can reduce
the risk of heart diseases, obesity, stroke and some cancers.
National governments and international organizations, such as the World
Health Organization (WHO) and the Food and Agriculture Organization
(FAO), have been working on nutritional guidelines extension, in particular
recommending a reduction in total fat and sugar consumption.
But many questions regarding the reasons why people are becoming obese
remain unanswered. Have diet or physical activity changed over time? Do
we eat more? How has our diet changed across time? Who has been affected
more by this changes? (why them? and why was that?) Do we eat ”better”
today than in the past? This paper carries on a first exploration of eating
habit variation across age and time among the British population. Moreover,
it presents some evidence on the relationship between nutrition and socio-
economic status in Britain using cross sectional data from the National Food
Survey covering the period 1975-2000. In doing so, we use Chesher’s method-
ology for providing a through decomposition of the National Food Survey
data and identify original regularities for basic demographic subgroups.
In his paper from 1997, Chesher decomposes intakes household supply
into individuals consumption using a non parametric model applied to the
National Food Survey data pooled by three years from 1974 to 1994. He
estimates the age profile of nutrient intakes such as calories, fat, calcium
6
and vitamin C controlling also for the potential effect of eating out and the
presence of visitors. Effects of household characteristics such as region of
residence and family income, are also considered.
The aim of this paper is to extend Chesher’s work using data from 1975
to 2000 considering both intakes and food groups consumption. It describes
how eating habits have changed by gender and age, by gender and time for
all age groups and particularly for children aged 0-17, and by generations in
the United Kingdom. Moreover, it will try to shed a light on the importance
of social and economic environmental to nutrition changes considering the
effect of income separately. There might have been many forces that have
affected people’s (especially children’s) food habits. This paper will simply
describe patterns without testing one explanation against another. We will
consider some hypotheses in a later work.
This work is organized as follows. Section 2 describes the data. Section 3
introduces the consumption specification using nonparametric techniques for
the estimation of average daily nutrition intakes and food group consump-
tion within a household model that adopts a roughness penalty function
and controls for income distribution, eating out, presence of visitors, family
structure, region of residence and presence of children. Section 4 presents
the results. Conclusions and extensions for future research are summarized
in Section 5.
2 Data
The data used in this study come from the National Food Survey (NFS). This
is a cross-sectional survey started in 1940, and it has run continuously since
1942. Its initial aim was to monitor the diet of the urban ”working class”
during the war years. In 1950 it was extended to the whole population in
Britain to collect data on food consumption and expenditures. Since 1992 the
NFS collects information also about confectionary, alcohol and soft drinks;
and since 1996 it has been extended to Northern Ireland.
The NFS collects weekly data over one year on household food acquisition
for a large nationally representative sample of British adults and children.
7
It collects information from roughly 7,000 household in the UK every year
(corresponding to a response rate of 65 percent). It contains year and month
specific information about all food entering into the household. After a short
interview, the household’s member who does the most of the shopping is
asked to keep a diary where reporting expenditures in British pence as well
as physical quantities of food purchased among more than 200 food items
listed. Each of the other members, age 11 and over, are requested to col-
lect information on personal expenditure on snack, meals, sweets, and drinks
consumed outside the home. The data also record the number and type of
meals (breakfast, lunch or dinner) offered to guests. In addition, the survey
records some demographic characteristics, for example age and sex of each
member of the family, number of male and female working, household char-
acteristics, region of residence, and socio-economic variables, such as income
and occupation of head of household.
The time period considered in this paper covers 26 years from 1975 to
2000, in which 201,032 households and 521,000 individuals were observed.
2.1 Sample Characteristics
Descriptive statistics for the main sample are reported in Table 1. After
controlling for missing values, dropping households from North Ireland1 and
dropping people over 91 years because their number was not enough to pro-
duce significant figures, the final sample ends up containing 130,789 house-
holds and 353,989 individuals.
The individual average age in the sample is 35 years. Head of house-
hold on average are 49 years old and their wives are just one year younger.
Children are on average 8 years old and the sample seems roughly equal
distributed over age groups.
Information on eating out are summarized from the net balance variable.
This variable varies for each person from 0 and 100. It takes value zero
1Data on North Ireland have been collected only from 1996. In order to include thissample into the analysis, it is necessary to weight the data.
8
if the person eats always out; it takes value 100 if the person eats every
meal at home. When a person eats outside the household, his net balance is
diminished of a certain amount depending on which of the main three meals
he did not took from the household. In particular, if a person has breakfast
outside his net balance will diminish of 3 points, if he has lunch outside, his
net balance will decrease of 4 points, and if he eats dinner outside, the net
balance will decrease of 7 points.
Table 1: Descriptive Statistics (1975-2000) - Household obs. 130,789 - Indi-viduals obs. 353,989
Descriptive StatisticsObs. Mean Std. Dev. Min Max
Individual Characteristicsage 353989 34.87747 22.73049 0 91age of wife of hoh 130789 47.68864 17.7052 16 92age of children 101001 8.15483 5.043785 0 17
Head of Household Characteristicsage 130759 49.34549 17.52024 16 91age if male 97252 47.27158 16.27708 16 91age if female 33507 55.36488 19.49869 16 91
Family Characteristicsnumber of members 130789 2.610946 1.366265 1 13number adult male 130789 .8797376 .5462758 0 6number adult female 130789 .9941738 .4376216 0 7number children 130789 .7370345 1.081004 0 10number adult aged greater than 64 130789 .3526214 .6372804 0 4number person aged 0000 130789 .0383518 .1935124 0 3number person aged 0104 130789 .177232 .4594763 0 5number person aged 0507 130789 .131953 .3800701 0 3number person aged 0811 130789 .1681028 .4514887 0 5number person aged 1215 130789 .1563052 .4449344 0 6number person aged 1617 130789 .0650972 .2599484 0 3
Eating Outnet balance per person 353989 87.47551 14.93684 0 100total net balance per household 130789 238.7557 127.2163 0 1247
In the sample considered here, the average individual records a net bal-
ance per week of 87.47 points. In other words, in average in a week a person
9
has almost 13 percent of his net balance from outside the household. This
corresponds almost to one full day eating out (one breakfast, one lunch and
one dinner per week).
In average there are about 2 or 3 members per household, with a maxi-
mum of 13 members. Approximately 9 percent of the individuals lives alone,
59 percent lives in household without children, while 41 percent lives in
household with only adults (Table 2).
The sample is roughly 5 percent from Wales, 9 percent from Scotland, 7
percent from the Northern, 9 percent from York and Humberside, 11 percent
from the North West, 7 percent from East Midland, 10 percent from West
Midland, 8 percent from South West, 3 percent from East Anglia and 30
percent from South East.
The set of economic variables available from the data set includes net
family income, total expenditure on food, specific food expenditures on par-
ticular items and quantity of food purchased during the period of study.
The dependent variables used in this paper are quantities of major food
group and nutrient intakes consumed, whose allocations among different fam-
ily members have been compute according to the methodology described in
the following section.
Following the World Health Organization and Government Guidelines,
food items have been summarized into 14 key food groups: diary products,
meat, fish, eggs, oils and fats, sugar and preservatives, vegetable, fruit, cere-
als, beverage, miscellaneous, soft drinks, confectionary and alcoholic drinks.
Moreover, following standard grouping used by food analysts (e.g. Food
Standard Agency, US Department of Agriculture, etc.) meat and fish have
been clustered together into a single group2.
The number of observation on soft drinks, confectionaries, alcohol, mis-
cellaneous and beverage are not enough to produce significant estimations,
therefore the analysis will proceed distinguish the first six standard main
food categories used by nutritionists and others.
2Meat, fish and eggs belong to the food group providing proteins, therefore they areusually classified together. In this case, eggs are measured in number of eggs purchasedby the household, while meat and fish are measured in grams. Therefore in the rest of thepaper I will consider only the category meat and fish.
10
Table 2: Region of residence, Number of family members and OccupationCode - Household obs. 130,387 - Individuals obs. 338,060
Variables Households IndividualsFreq. Percent Freq. Percent
Hhld composition1 adult only 29,737 22.74 30,634 8.651 adult and 1 or more children 5,510 4.21 15,642 4.422 adult only 41,354 31.62 85,491 24.152 adults, 1 child 12,987 9.93 40,655 11.482 adults, 2 children 18,237 13.94 76,102 21.502 adults, 3 children 6,158 4.71 32,360 9.142 adults, 4 or more children 2,026 1.55 13,710 3.873 adults 6,693 5.12 20,450 5.784 or more adults 2,057 1.57 8,521 2.413 or more adults, 1 or 2 children 5,145 3.93 24,053 6.793 or more adults, 3 or more children 885 0.68 6,371 1.80
Hhld with children 50,948 38.95 208,893 59.01Hhld without children 79,841 61.05 145,096 40.99
Region of ResidenceWales 6,819 5.21 18,298 5.17Scotland 11,726 8.97 32,234 9.11Northern 8,932 6.83 23,837 6.73York and Humberside 12,279 9.39 33,066 9.34North West 14,938 11.42 41,224 11.65East Midland 9,145 6.99 25,314 7.15West Midland 12,577 9.62 35,523 10.04South West 11,082 8.47 29,032 8.20East Anglia 4,528 3.46 11,565 3.27South East 38,763 29.64 103,896 29.35Total 130,789 100.00 353,989 100.00
11
Table 3 reports descriptive statistics on net family income, total expen-
diture and quantities (in grams) of food purchased on average per household
in a week period on each food groups.
On average households spend 29 pounds each week on food, and earn
almost 150 pounds per week. In particular the average household spends 4
pounds in diary products, 8 pounds in meat, 2 pounds in fats and sugars,
6.50 pounds in vegetables and fruit, 4 pounds in cereals and pasta, 2 pounds
in soft drinks and 10 pounds in alcohol3.
Over the time period, the average household (of 2 or 3 people) buys in
one week almost 6 liters of diary products, 3 kg of meat, 12 eggs, almost 1
kg of fats and oils, 1.4 kg of sugars and preservatives, 6 kg of vegetables, 2.7
kg of fruit, 4 kg of cereals, pasta and rise, 4 liters of soft drinks and 3 liters
of alcohol.
2.2 Derived Variables
As many nutritionist point out, in order to have a healthy diet is important
to have the right balance of nutrients needed to be healthy. Therefore, it
is obviously important to study both consumption of nutrient intakes and
consumption of major food groups. A number of balance diets are based on
either combination of intakes or combination of food types or both.
Therefore, this paper considers also a second set of dependent variables:
nutrient intake quantities consumed. They are computed from the basic data
using the conversion factor tables from the Department for Environmental
Food and Rural Affairs (DEFRA, 1999). The full detail of reported food
purchased is used, with weights converted to intakes using the intake content
factors. DEFRA table reports 47 nutrient intakes on it. This paper considers
13 of them: calories, proteins, fat, carbohydrates, calcium, iron, vitamin C,
D, E, B6 and B12, potassium, magnesium. However, conversion factor for
potassium, magnesium, vitamin B6, vitamin B12 and vitamin E are available
only from 1992 and they do not produce significant estimation. Therefore the
analysis will proceed focusing on calories, fat intake, proteins, carbohydrates,
3Data on soft drinks, alcohol, miscellaneous and beverage are available from 1992.
12
Table 3: [Continue] Descriptive Statistics (1975-2000) - Household obs.130,789 - Individuals obs. 353,989
Descriptive StatisticsObs. Mean Std. Dev. Min Max
Family Income and Expendituretotal household food expenditure 130789 29.15207 23.38368 0 604.96net family income 130789 150.1823 156.1296 10 2978.00
Family Food Expenditure (GBP)diary products 128332 3.841969 3.04438 0 64.63meat 121361 8.152233 7.689984 0 280.26fish 79799 2.225025 2.567056 0 75.40eggs 77267 .8009915 .5522406 0 29.43fats and oils 94085 1.211639 1.019039 0 27.12sugar and preservatives 66282 .8645352 .7003412 0 17.21vegetable 125033 3.991886 3.706174 0 59.50fruit 106087 2.514873 2.841234 0 69.89cereals 127236 4.808307 4.435264 0 101.42soft drinks 28251 2.347129 2.250002 0 46.83confectionaries 19082 1.979311 2.235051 0 40.90alcohol 15179 9.815807 13.10797 0 512.06miscellaneous 91765 1.797628 2.079407 0 149.65beverage 72702 1.70735 1.53138 0 47.98
Quantity of Food purchased (g)diary products 128332 6635.383 4463.997 42.00071 66640.75meat 121361 2846.82 2757.956 25.00042 140219.1fish 79799 592.8705 567.0425 28.00047 20723.85eggs (N. of eggs) 77267 12.18848 8.105958 0 162fats and oils 94085 962.3533 877.0756 11.00019 40161.2sugar and preservatives 66282 1397.647 1078.614 14.00023 62511.75vegetable 125033 6159.51 5846.392 20.00034 117907.7fruit 106087 2783.287 2588.278 9.977799 81364.5cereals 127236 4159.729 3274.509 19.00032 303111.1soft drinks 28251 4034.825 4035.488 74.8335 85050.78confectionaries 19082 437.463 455.752 7.370999 8500.323alcohol 15179 3188.451 3958.576 10.47816 123138.2miscellaneous 91765 1250.973 1584.85 0 61696.84beverage 72702 344.5001 268.0583 9.00015 10000.17
13
calcium, iron and vitamin C. However, given that in the White Paper ”The
Health of the Nation” the Government has specifically set targets for the
proportion of energy from fats to be no more than 35% by 2005, I will report
the main results on fats also in terms of proportion of energy from fats
(PEF)4.
Table 4 shows household consumption of nutrient intakes in a week period.
In average a household purchases 37,414 calories per week and almost 1.5 kg
of fats, that means that almost half of the total calories purchased derives
from fats.
Table 4: [Continue] Descriptive Statistics (1975-2000) - Household obs.130,789 - Individuals obs. 353,989
Descriptive StatisticsObs. Mean Std. Dev. Min Max
Quantity of Intakes purchasedcalories (Kcal) 130789 37414.1 25811.94 0 1260780Proteins (g) 130789 769.6949 551.5648 0 22950.1Fat intake (g) 130789 1687.037 1278.57 0 39593.91Proportion of energy from fats (Kcal) 130789 15183.33 11507.13 0 356345.2Carbohydrates (g) 130789 4559.545 3475.143 0 263892.5Calcium (mg) 130789 16269.43 10108.62 0 228228.3Iron (mg) 130789 195.7957 134.4513 0 2731.77Vitamin C (mg) 130789 1030.547 921.3632 0 24667.22Vitamin D (mg) 130789 54.94992 55.28951 0 1568.759Potassium (mg) 52279 45858.38 31445.96 0 627095.6Magnesium (mg) 52279 4103.089 2821.844 0 75137.1Vitamin B6 (mg) 52279 35.06806 27.61754 0 1131.76Vitamin B12 (ug) 52279 93.21448 84.2328 0 3088.253Vitamin E (mg) 52279 4055.283 3122.384 0 57431.36
4Total amount of energy from fat is obtained multiplying total amount of fats by 9.00.The proportion of energy from fats is the ratio between total amount of calories from fatand total amount of calories.
14
3 Methods
Expected consumption of individual p is assumed to be function of individual
characteristics xp (e.g. sex and age) and household characteristics z. The
theoretical model follows Chesher (1997).:
E[cp|x, z] = f(xp, z)
Thus average household consumption is:
E[c|x, z] =∑P
p=1 f(xp, z)
where P denotes total household members. The National Food Survey
(NFS) collects information about total food acquisition and expenditure per
household. There is no information on individual consumption. In order to
take into account consumption variation with respect to age and sex of dif-
ferent household members this model relates food acquisition with household
composition.
If a household consumes qci , quantity of food i, and therefore the amount
of nutrient contained in each unit of food i, νi, the total quantity of nutri-
ent consumed by the household can be expressed as c =∑
νiqci . The total
quantity of nutrient entering the household is the total amount of nutrient
contained into total food purchased5, y =∑
νiqi. In the long term it is prob-
ably reasonable to assume that total amount of food entering the household
is equal to the total amount of food consumed by each family. Therefore the
expected value of total food acquisition is assumed to be equal to expected
value of total food consumed:
E[y|x, z] = E[c|x, z] =∑P
p=1 f(xp, z)
Where y represents the total quantity of food entering into the household
and c the total quantity of food consumed.
Assuming that f is a separable function with respect to xp and z allows
f to be written as product of two functions:
5The NFS records amount of food i entering the household h: qci 6= qi.
15
f(xp, z) = g(xp) · u(z)
The amount of food consumed by a person p still depends from his/her
own characteristics, such as age and sex, and from the characteristics of the
family. However, if we consider two persons of the same age and sex, living
in two different households, their ratio of consumptions will be the same. In
other words, the model proposed here estimate the average consumption of
food/nutrient for a person p of age a and sex s. Identical individuals in dif-
ferent households consume the same amount of food/nutrient independently
from their family structure uz. This assumption also implies that diet of chil-
dren and adults are not affected by family structure. In order to take that
into account, household composition and presence of children are introduced
in z.
It is known that consumption differs over age among male and female,
even in early ages. So, let gS(ap) be the function representing the relationship
between consumption and age for each sex S = M, F . The total distribution
of consumption over age can be express as follows:
g(xp) = g(age, sex) = spgM(ap) + (1− sp)gF (ap)
where gS(ap) are complex and non-linear functions and sp is a dummy
variable taking value 1 if the individual observed is male, and 0 otherwise.
3.1 Demand for food and nutrients
The demand for food and nutrient changes through lifetime with level of
activity and preferences. Following Chesher (1997), I use a non-parametric
approach in defining gS(ap) and add household characteristics in parametric
form, as follows:
u(z) = exp(z′γ).
The function gS(ap) represents the relationship between food consumption
and age for each gender S. For each individual p, let wp = [wp,0, wp,1, wp,2, . . . , wp,99]
16
be a vector of dummy variables allocating the value 1 to the dummy corre-
sponding to the class of age to which the individual belongs, that is:
wp,a =
1 if a ≤ ap ≤ a + 1
0 otherwise
So, for example, if family 1 holds 3 members, aged 50, 2 and 0 years old
respectively, the matrix of vectors wp where p = 1, 2, 3 identifies person p
(second column) in household h (first column), will look as follows:
X =
1 1 0 0 0 . . . 1 . . . 0
1 2 0 0 1 . . . 0 . . . 0
1 3 1 0 0 . . . 0 . . . 0...
......
......
......
......
Given the assumptions above, the relationship between age and intake
for males and females can be approximated by the discrete form:
gS(ap) = w′pβ
S =
wp,0
wp,1
wp,2
...
wp,99
(βS
0 βS1 βS
2 . . . βS99
)
Where βSa are the coefficients estimated at each age for S = M, F . They
represent the amount of nutrient consumed by a person p of age a and sex
S.
At this point we can formalize the expected value of household consump-
tion6 as follows:
E[y|x, z] =P∑
p=1
[sp · gM(ap) + (1− sp) · gF (ap)] · exp(z′γ)
6Therefore food acquisition in the long term - see first assumption above.
17
E[y|x, z] =P∑
p=1
[spw′pβ
M + (1− sp)w′pβ
F ] · exp(z′γ)
It should be noticed that∑
p spw′p represents the number of males of age
a living in the household and∑
p(1− sp)w′p represents the number of females
of age a living in the household. Thus, for each household, the expected
nutrient consumption is going to be:
E[y|x, z] =P∑
p=1
A∑
a=0
[nMpaβ
Ma + nF
paβFa ] · exp(z
′γ)
where A is the maximum value taken by the variable age and βSa represents
the amount of nutrient consumed by any individual of age a and gender S.
3.2 Penalized least square regression
In its simplest form the roughness penalty approach is a method for relaxing
the model assumptions in classical linear regression in a slightly different
way from polynomial regression (Green and Silverman, 1995). In order to
estimate βM , βF and γ, and given the discontinuity of age, I use non-linear
least squares with a roughness penalty function methodology and minimize
the following object:
minβMβF γ
[H∑
h=1
(yh −(β0 +
99∑
a=0
(nMhaβ
Ma + nF
haβFa )
)exp(z
′hγ))
]2
+
+λ2M
99∑
a=0
(βMa − 2βM
a+1 + βMa+2)
2 + λ2F
99∑
a=0
(βFa − 2βF
a+1 + βFa+2)
2
where β0 is included to capture flows of nutrients into households that are
unrelated to the number of household members (e.g. food for pets) and the
last term is the discrete version of the roughness penalty function capturing
the smoothness of the relationship between age and consumption. The same
18
object is representable in matrix form as follows:
min(Y −Xβ)′(Y −Xβ) + λ2β
′W
′Wβ
where λ > 0.
Using matrix the data structure can be summarized as follows. Let
D =
yh i NM NF Z
0 0 λ · A 0 0
0 0 0 λ · A 0
=
Y X Z
0 λ ·W 0
where:[i, NM , NF
]= X, and
0 λ · A 0
0 0 λ · A
= λ ·W
The final sum of squared model without considering Z is7:
minS = (Y −Xβ)′(Y −Xβ) + λ2β
′W
′Wβ
so the β estimator turns out to be biased and the bias depends on λ:
β(λ) = (X′X + λ2W
′W )−1X
′Y
with expected value and variance given by:
E[β(λ)
]= (X
′X + λ2W
′W )−1X
′E(Y |X)
= (X′X + λ2W
′W )−1X
′(Xβ + ε)
= (X′X + λ2W
′W )−1X
′Xβ + 0
=⇒ E[β(λ)
]= (X
′X + λ2W
′W )−1X
′Xβ
and
V[β(λ)
]= σ2(X
′X + λ2W
′W )−1X
′X(X
′X + λ2W
′W )−1
7The vector Z will be introduced later on in this chapter.
19
3.3 How to choose the degree of smoothness λ
As indicated by Green and Silverman, the most common method used to
identify λ is Cross Validation (CV). This methodology requires to omit ar-
bitrarily an observation i and estimate the curve from the remaining data.
This new object, denoted g−i(tj, λ), is the minimizer of:
∑
j 6=i
{yj − g (tj)}2 + λ∫ (
g′′)2
The value of λ derives from minimizing the sum of square differences be-
tween observed and estimated values, this time considering also observation
i omitted before:
minCV (λ) = n−1n∑
k=1
{yk − g−i(tk, λ)}2
Following Chesher (1997), this paper considers three possible values of
the degree of smoothness: no smoothness (λ = 0), λ = 57.3, that it is the
value that minimizes Wahba’s (1975) generalized cross-validation criterion,
and λ = 100.
Calories distribution over age for men and women using data from 1975
are shown in Figure 1 for each value of λ specified above. The same model
has been run for every year of the NFS considered.
For each year, estimate using λ = 0 show high variability across ages,
and all the models show that the trend of consumption of calories increases
during early ages and decreasing after 60, with two main local maximum at
age 15 and 50. Considering that the main aim of this work is to describe
variations of eating habits over age, and that the differences between using
λ = 57.3 or λ = 100 are not very big, all the estimation results that follow
will use λ = 100.
20
Figure 1: Estimated energy-age curves for male and female using data from1975 with roughness penalty λ=0, λ=57.3 and λ=100.
080
016
0024
0030
00K
cal
0 10 20 30 40 50 60 70 80 90age
(a) λ=0.
080
016
0024
0030
00K
cal
0 10 20 30 40 50 60 70 80 90age
(b) λ=57.3.
080
016
0024
0030
00K
cal
0 10 20 30 40 50 60 70 80 90age
(c) λ=100.
21
3.4 Introducing information on eating out and visitors
The National Food Survey provides some information about food eaten out
and visitors. Although it does not record the amount of food obtained from
no household supplies, for each person a measure of the number of meals
taken from the household during the survey week is available. This measure
is known as ”net balance” and it varies from 0 to 100. It is equal to 100
when the person obtain all his meals from household supply. It has value 0
when all the meals are eaten out. For each meal eaten out, the net balance is
reduced of 3, 4 or 7 depending from whether the missed meal was breakfast,
lunch or dinner respectively.
In the following estimation the model controls for eating out interpreting
the net balance as the proportion of food coming from household supplies for
each person (bp). If βMa and βF
a are interpreted as total food supplies from
the household and from outside the home, then the total amount of food
coming from the household is bpβSa and the initial model can be written as
follows:
E[y|x, z, b] = β0 +P∑
p=1
A∑
a=0
[nMpabpβ
Ma + nF
pabpβFa ] · exp(z
′γ)
= β0 +P∑
p=1
A∑
a=0
[b′MβM
a + b′F βF
a ] · exp(z′γ)
where for each household bS is a vector containing the net balance for
each individual at each year of age.
The net balance information is available also for each visitor. Using this
information, the model takes into account each visitor as an additional mem-
ber of the household, by age and sex, who takes from the household the
proportion of food indicated by his net balance.
22
4 Results
With the aim here of providing a through decomposition of the NFS data and
identify original regularities for basic demographic subgroups, this section de-
scribes the estimates of the Roughness Penalty Function Model obtained from
non-linear ordinary least squares (NL-OLS) method to account for function
smoothness. The paper investigates nutrition curves - using nutrient intakes
and major food groups - with the objective to see how they have changed by
gender and age over the recent time period and by gender and time for all
age groups and, particularly, for children aged 0-17. We will also consider
the effect of income and other household characteristics separately.
4.1 AGE
The relationship between nutrient intake (major-food-group) and age have
been estimated separately by each sample year in two stages: in a first stage,
the paper deals only with the results from the non parametric analysis (nu-
trient intake/food in relation to gender and age), while in a second stage,
we control for other household characteristics such as income, region of resi-
dence, household composition, presence of children.
4.1.1 Age curves estimation
At the first stage, we estimate nutrient and food consumption only in relation
to household members characteristics (i.e. age and gender). Coefficients
estimated separately for each year have been averaged up over the whole
sample period 1975-2000. The findings for each nutrient intake and food
group for males and females separately, by each completed year of age from
0 to 91, are reported graphically in Figures 2 and 3 respectively.
Both for male and female the distribution of consumption over the life
cycle show an inverse U shape, increasing rapidly until age 14 for girls and 16
for boys, then it declines until around age 25, and it increases again showing
a peak at the age 55 for females and 60 for males. After that there is a steady
decline.
23
The estimates show that on average males consume more then females at
any age. This picture is quite similar along all the period for all nutrients
and foods considered, with some exceptions such as calcium and vitamin
C. Figure 2 panels g) and h) show that on average females consume more
calcium than males after 40 years old, and more vitamin C along all the
life cycle. The reason of that is probably the higher consumption of food
estimated for females, like milk (after age 35) and fruits, shown from Figure
3 panels a) and f).
The peak at puberty is consistent with consolidation of body height and
weight during the adolescence period. The peak occurs 2 years earlier in
girls than in boys, as puberty itself does. Similarly, the fall in consump-
tion after middle age can be explained by the fact that elderly people lose
weight and spend less energy. It is also important to note the steady rise in
calories/nutrients consumption after 30. This usually coincides with a pe-
riod in life when people exercise less and increase weight, but these are not
necessarily the only explanations.
The age patterns of fats intake, carbohydrates and iron (Figure 2, panel
b), e) and f) respectively) are very similar to those of calories intake. For both
men and women, they increase during childhood, slightly decrease between
age 15 and 30, and then increase again, but more rapidly for women than for
men.
Proteins consumption distribution shows less differences among genders
((Figure 2, panel d)), showing the biggest difference between age 10 and 40.
Fat intakes converts to calories at the rate of 9 kcal per gram. Therefore,
multiplying estimate fat consumption by 9 and dividing it by total calories at
each age we obtain the distribution of proportion of energy from fat (PEF)
shown in Figure 2, panel c). Over the life cycle, in average PEF results
between 30 and 40 percent up to age 25, and over 40 afterwards. WHO
recommends proportion of energy from fat and saturated fats to be reduced
by 2005 to 35%. It will, therefore, be interesting to see in the following
sections how it has changed over the period of study and how its trend
moves over time.
Figure 3 shows the estimated food-age curves using a linear model not
24
Figure 2: Estimated intake-age curves using linear model with roughnesspenalty λ =100 - weighted average over 1975-2000.
30
08
00
13
00
18
00
23
00
28
00
Kca
l/day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Calories
(a) Calories.
02
04
06
08
01
00
12
01
40
gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Fat intakes
(b) Fat Intake.
20
25
30
35
40
45
%K
cal/d
ay/
capita
fro
m fat
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Proportion of energy from fat
(c) Proportion of energy from fat.
02
04
06
08
0gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Protein
(d) Proteins.
90
19
02
90
39
0gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Carbohydrate
(e) Carbohydrate.
05
10
15
gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Iron
(f) Iron.
20
04
00
60
08
00
10
00
gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Calcium
(g) Calcium.
02
04
06
08
01
00
gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Vitamin C
(h) Vitamin C.25
accounting for income, eating out, visitors, region of residence, etc.
Here we consider six main groups of food8, namely: 1) milk products, 2)
meat and fish products, 3) fat, oils, sugar and preservatives, 4) cereal, pasta,
rise and bread, 5) vegetable and 6) fruit.
Also in this case, the estimated curves show an inverse U shape, with
males consuming more than females, but milk products and fruit for which,
as we said above, women consume more milk products than men after 40,
and more fruit in average.
It is interesting to notice the stable increase in consumption of fat, oils,
sweets and preservatives from childhood up to age 70 for males and 55 for
females.
4.1.2 Estimate Non-linear model controlling for income, eating
out and visitors
This section presents the estimated intake/food-age curves obtained intro-
ducing other variables in the model previously estimated. Controlling for
other household characteristics such as income, region of residence, house-
hold composition, presence of children makes the model non-linear. Here we
compare the new nutritional-age curves with those obtained without taking
into account eating out and presence of visitors. Later in the paper we will
present estimates of coefficients on the other variables listed.
The model has been estimated separately for each year and for each nutri-
ent intake and food group considered. Figures 4 and 5 represent graphically
estimated coefficients using roughness penalty λ = 100 averaged up for the
period 1975-2000.
Introducing information about eating out, we would expect that the es-
timates produced would be higher for almost all ages. While controlling for
presence of visitors, we assume that some of the food bought from the house-
hold is consumed by visitors not by household members, causing a decrease
8Other foods have been analyzed using the same methodology, as for example, beverage,miscellaneous, soft drink, confectionaries and alcohol, but the NFS data are collected onlyfrom 1992 onwards. Consumption distributions over age for these food groups appear veryunstable and irregular with numerous peaks, so they have been left out from the rest ofthe analysis.
26
Figure 3: Estimated food groups-age curves using linear model with rough-ness penalty λ =100 - weighted average over 1975-2000.
01
00
20
03
00
40
0gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Milk
(a) Diary Products.0
50
10
01
50
20
02
50
30
0gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Meat and fish
(b) Meat and Fish.
04
08
01
20
16
02
00
gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Oils and fats, sugar and preservatives
(c) Fat, oils, sugar and preserva-tives.
05
01
00
15
02
00
25
03
00
35
0gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Cereals, rice and pasta
(d) Cereal, pasta, rise and bread.
50
15
02
50
35
04
50
55
0gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Vegetables
(e) Vegetable.
05
01
00
15
02
00
gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Fruit
(f) Fruit.
27
in the individual consumption.
At this point it is difficult to say which of the two effects is driving
the individual consumption estimation. Although showing a similar shape
to Figures 2 and 3 and confirming that men consume in general more than
women, the coefficients estimated from the non-linear model result lower than
the previous ones and the curves represented in Figures 4 and 5 are shifted
downward and look flatter. The biggest effect is that on fruit consumption
that becomes almost null. It is also to notice that part of the steep rise showed
before after age 30 disappears. This might be the effect of the presence of
visitors. If people after age 30 receive visitors in their home and invite them
to eat with them, then the age dependence relation estimated here will take it
into account assigning a lower amount of food and nutrients from household
supply to each individual.
The decrement in quantity consumed at home, might be also caused by
the use of net balance information to take into account eating out. If a
member of the family with a net balance of 86 eats out one day per week.
For that day what before attributed to his consumption given age and gender,
it will be now redistributed among the other household members with the
effect of increasing their consumption. The higher decrement from previous
estimates is for people age after 30. If people at this age range tend to eat
outside rather than take food from household supply, part of the change in
the shape of the curve might be due to the incidence of eating out.
Controlling for eating out and presence of visitors has also an effect on
proportion of calories from fats, that now varies between 25 and 35 percent
for both men and women.
28
Figure 4: Estimated intake-age curves using non-linear model with roughnesspenalty λ =100 - weighted average over 1975-2000.
30
08
00
13
00
18
00
23
00
28
00
Kca
l/day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Calories
(a) Calories.
02
04
06
08
01
00
gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Fat intakes
(b) Fat Intake.
20
25
30
35
40
%K
cal/d
ay/
capita
0 10 20 30 40 50 60 70 80 90age
male female
Proportion of energy from fat
(c) Proportion of energy from fat.
01
02
03
04
0gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Protein
(d) Proteins.
90
19
02
90
39
0gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Carbohydrate
(e) Carbohydrate.
05
10
15
mg/d
ay/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Iron
(f) Iron.
10
02
00
30
04
00
50
06
00
mg/d
ay/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Calcium
(g) Calcium.
03
69
12
mg/d
ay/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Vitamin C
(h) Vitamin C.29
Figure 5: Estimated food-age curves using non-linear model with roughnesspenalty λ =100 - weighted average over 1975-2000.
05
01
00
15
02
00
25
0gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Milk
(a) Diary Products.
03
06
09
01
20
15
0gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Meat and Fish
(b) Meat and Fish.
05
01
00
15
02
00
25
03
00
gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Oils and fats, sugar and preservatives
(c) Fats, oils, sugar and preserva-tives.
01
00
20
03
00
40
05
00
gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Cereals
(d) Cereals, pasta, rise and bread.
01
00
20
03
00
40
05
00
60
0gr/
day/
capita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Vegetables
(e) Vegetable.
−1
00
10
20
30
gr/
da
y/ca
pita
0 10 20 30 40 50 60 70 80 90age
male female 95% confidence interval
Fruit
(f) Fruit.
30
4.2 TIME
The focus of this section is on changes in food consumption and nutrient
intakes among younger people (age 0-17). Nutritional trends have been an-
alyzed for both the whole population and, in particular for all youth dis-
tinguishing three age classes (0-6, 7-12, 13-17) and gender. The aim of
this section is chart patterns over time. There might be many forces that
have affected people’s (especially children’s) food habits. Examples of such
drivers are, home technology improvements, changes of parental costs of time,
parental preferences and information about children: food and diet which
could affect intra-household resource allocations (Cutler et al., 2003).
In this section we simply describe patterns without testing one expla-
nation against another. We consider some of these hypotheses in a later
chapter.
4.2.1 Nutrient Intakes
Figures 6, 7, 8, 9, 10 and 11 aim to illustrate how eating habits have changed
over the last 26 years. In particular looking at age groups under 17, they
report trends of grams of nutrients consumed by the youngest three age
groups.
Figure 6 shows distribution of nutrient intakes over the time period 1975-
2000.
Calories and iron consumption (Figure 6 panel a) and f)) remain quite
stable over time for both men and women. Trend of calories as well as
carbohydrates show a steep rise from 1994 to 1995 that is probably due
to changes in the NFS data collection, as for example the introduction of
alcohol, soft drinks and confectionaries.
Amount of fats intakes consumed by people at the beginning of the time
period is instead very different from those consumed at the end. Panel b) in
Figure 6 shows a steady increase of consumption that almost double by the
end of the 90s. Similar finding are presented in panel d) for proteins.
The increasing incidence of fat on production of calories is also shown in
panel c). Proportion of energy from fat show a clear increasing trend up to
31
1990, that becomes flat and stable around 35 percent in the last decade.
As for the last two panels in Figure 6 they represent consumption of
calcium and vitamin C, respectively. Trend of calcium shows an inverse U
shape along the period. While the consumption of vitamin C results very
low. Given the steep rise shown from 1995 onwards, the model estimation
for this nutrient will require further study.
The trends of nutrient intakes consumption for younger people are shown
in Figure 7 and 8 separately for boys and girls by age groups. The first
column of graphs shows average daily intakes consumption for males, while
the second column shows estimates for females. In general younger children
consume less than older children. The variation of consumption of different
age groups results from the distance between the curves and respects the
different age distribution in consumption shown above.
Trends of consumptions do not show huge differences among the time
period under analysis. In particular calories and carbohydrates trends are
quite flat (Figure 7, panels a) and b), and Figure 8 panels a) and b)). Like
for the whole population, the trend of consumption of calories for older boys
show a jump upward at the beginning of the 90s when soft drinks, alcohol
and confectionaries were introduced in the survey. For girls the effect is
smaller. If this discontinuity in the trend captures the effect of soft drinks
and confectionaries among children, than it is interesting to notice that the
effect is bigger for boys than for girls (they show bigger differences between
the curves) and it affects their diet very early in life (there is no effect on the
youngest group 0-6, but from 7 onwards).
Proportion of energy from fat has increased along the time period (Figure
7, panels e) and f)). Even if it does not change a lot across age groups, it is
higher at the end of the period than in the 70s. It results slightly higher for
girls than for boys and it is quite flat around 35 percent from the 90s.
Fats intakes and proteins trends (Figure 7, panels c), d), g) and h)) for
both genders at all ages tend to increase slightly from 1978 up to 2000. The
difference between ages is higher for boys in fats and it tends to increase
more from the 90s onwards especially for proteins.
The others intakes show a trend very similar to the whole population.
32
Figure 6: Estimated intakes-year curves using non-linear model with rough-ness penalty λ =100 - weighted average over age (male and female).
05
00
10
00
15
00
20
00
25
00
30
00
Kca
l/day/
capita
1975 1980 1985 1990 1995 2000time
(a) Calories.
02
55
07
51
00
12
5gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(b) Fat Intake.
01
02
03
04
05
0%
Kca
l/day/
capita
fro
m fat
1975 1980 1985 1990 1995 2000time
(c) Proportion of energy from fat.
01
02
03
04
0gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(d) Proteins.
01
50
30
04
50
60
07
50
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(e) Carbohydrate.
02
46
81
0m
g/d
ay/
capita
1975 1980 1985 1990 1995 2000time
(f) Iron.
01
00
20
03
00
40
05
00
60
07
00
mg/d
ay/
capita
1975 1980 1985 1990 1995 2000time
(g) Calcium.
02
46
81
0m
g/d
ay/
capita
1975 1980 1985 1990 1995 2000time
(h) Vitamin C.33
Figure 7: Estimated intakes-year curves for children using non-linear modelwith roughness penalty λ =100 - weighted average by class of age (boys andgirls).
30
08
00
13
00
18
00
23
00
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(a) Calories - boys.3
00
80
01
30
01
80
02
30
0gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(b) Calories - girls.
02
04
06
08
01
00
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(c) Fat Intake - boys.
02
04
06
08
01
00
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(d) Fat Intake - girls.
10
20
30
40
50
%K
cal/d
ay/
cap
ita f
rom
fa
t
1975 1980 1985 1990 1995 2000time
male 0−6 male 7−12 male 13−18
Proportion of energy from fat
(e) Proportion of energy from fat -boys.
10
20
30
40
50
%K
cal/d
ay/
cap
ita f
rom
fa
t
1975 1980 1985 1990 1995 2000time
female 0−6 female 7−12 female 13−18
Proportion of energy from fat
(f) Proportion of energy from fat -girls.
01
02
03
04
0gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(g) Proteins - boys.
01
02
03
04
0gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(h) Proteins - girls.
34
Figure 8: [Cont.]Estimated intakes-year curves for children using non-linearmodel with roughness penalty λ =100 - weighted average by class of age(boys and girls).
01
00
20
03
00
40
05
00
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(a) Carbohydrates - boys.0
10
02
00
30
04
00
50
0gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(b) Carbohydrates - girls.
02
46
81
0m
g/d
ay/
capita
1975 1980 1985 1990 1995 2000time
(c) Iron - boys.
02
46
81
0m
g/d
ay/
capita
1975 1980 1985 1990 1995 2000time
(d) Iron - girls.
01
00
20
03
00
40
05
00
60
0m
g/d
ay/
capita
1975 1980 1985 1990 1995 2000time
(e) Calcium - boys.
01
00
20
03
00
40
05
00
60
0gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(f) Calcium - girls.
02
46
81
0m
g/d
ay/
capita
1975 1980 1985 1990 1995 2000time
(g) Vitamin C - boys.
02
46
81
0m
g/d
ay/
capita
1975 1980 1985 1990 1995 2000time
(h) Vitamin C - girls.
35
4.2.2 Major Food Groups
The same analysis has been conducted also on major food groups as milk
products, meat and fish products, fats from oils and sugars, cereals, vegetable
and fruit and results are presented in Figure 9.
In this case we notice that in general average daily consumption of food
tends to increase in the last decade considered. Consumption of meat and
fish increases during the 1990s as well as consumptions of fats and fruit for
both genders9 (Figure 9 panel b), c) and f)).
Bread, cereal, pasta and rice consumption trend shows a steep jump from
1992 to 1994. As for calories and carbohydrate it is probably link to the
introduction of new food items in the survey.
Also quantity of vegetables (Figure 9, panel e))consumed results mainly
constant over the whole period of study, with two local peaks between 1982
and 1987 and between 1991 and 1994 both for men and women, and declines
afterwards.
Figures 10 and 11 show trends for boys and girls by three age groups.
Trends of consumption follow in general the trend of the whole population.
9The very low estimates shown at the beginning of the time period have required furtherinvestigation of the data. However we did not find anything that can explain such lowvalues.
36
Figure 9: Estimated food-year curves using non-linear model with roughnesspenalty λ =100 - weighted average over age (male and female).
05
01
00
15
02
00
25
0gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(a) Diary Products.
05
01
00
15
02
00
25
03
00
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(b) Meat and Fish.
05
01
00
15
02
00
25
03
00
35
04
00
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(c) Fats, oils, sugar and preserva-tives.
01
50
30
04
50
60
07
50
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(d) Cereals, pasta, rise and bread.
03
00
60
09
00
12
00
15
00
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(e) Vegetable.
05
10
15
20
25
30
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(f) Fruit.
37
Figure 10: Estimated food groups-year curves for children using non-linearmodel with roughness penalty λ =100 - weighted average by class of age(boys and girls).
05
01
00
15
02
00
25
0gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(a) Diary Products - boys.
05
01
00
15
02
00
25
0gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(b) Diary Products - girls.
05
01
00
15
02
00
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(c) Meat and fish - boys.
05
01
00
15
02
00
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(d) Meat and fish - girls.
05
01
00
15
02
00
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(e) Fats, oils, sugar and preserva-tives - boys.
05
01
00
15
02
00
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(f) Fats, oils, sugar and preserva-tives - girls.
38
Figure 11: [Cont.]Estimated food groups-year curves for children using non-linear model with roughness penalty λ =100 - weighted average by class ofage (boys and girls).
01
00
20
03
00
40
05
00
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(a) Cereals, pasta, rise and bread -boys.
01
00
20
03
00
40
05
00
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(b) Cereals, pasta, rice and bread -girls.
02
50
50
07
50
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(c) Vegetable - boys.
02
50
50
07
50
gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(d) Vegetable - girls.
02
46
81
0gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(e) Fruit - boys.
02
46
81
0gr/
day/
capita
1975 1980 1985 1990 1995 2000time
(f) Fruit - girls.
39
4.3 COHORTS
The focus of this section is on differences in nutrient intakes and food con-
sumption among generations. The NFS is a series of household cross sectional
data and, thus, it does not follow the same individuals over time. In order
to see whether there exists some generation’s effect on differences between
people born at different times, we consider ten cohorts (1945, 1950, 1955,
1960, 1965, 1970, 1975, 1980, 1985 and 1990).
Cohorts are constructed by date of birth of each individual. For each
survey we average estimate nutrient intakes and food consumption by age
and then track the sample from the same cohort one year older in the next
survey. We do not distinguish here by gender. For example, people who
were born in 1945 are observed from age 30 (in 1975) to age 55 (in 2001),
while cohort 1975 is observed from age 0 to age 25. The last three cohorts
(those born in 1980, 1985 and 1990) are the youngest cohorts in the sample
who were born after the beginning of the survey, therefore they are observed
only for a short period of time: twenty, fifteen and ten years, respectively.
Results on nutrient intakes and food groups are shown in Figures 12 and 13
respectively.
Figure 12 shows the cohort intakes consumption curves beginning with
those born in 1990. In panel a) of Figure 12, the first line segment connects
the average consumption of calories of those who were zero years old in 1975
to the average consumption of calories of 1 year old in 1976, until the last
observation of the cohort in 2000, when they were 10 years old. The second
line segment repeats the exercise for those who were five years older until the
last cohort considered in this graph of those born in 1945.
There is a visible life-cycle pattern rising with age as we saw from the
previous sections. With few exceptions at older ages, the lines for the younger
cohorts are very often but not always above the lines for the older cohorts,
even when they are observed at the same age, that is when the cohorts
overlap.
Comparing nutritional habits of different generations at the same age,
calories consumption is slightly different for different cohorts at different
40
ages. Between age 0 and 10 younger generation consumed less than older
ones, while between age 10 and 18 they consume slightly more calories than
their older counterparts.
Figure 12 panel b) plots fats intake patterns. It is interesting to notice
that younger generations consume higher amount of fats intakes at all ages.
In particular looking at children between 0 and 10 years old we compare
cohorts from 1975, 1980, 1985 and 1990. Children born in 1990 eat more fats
than those born earlier since age 4. Consumption of fats intakes maintains
the same structure, with younger generation eating more fats than older
ones, at all ages. Similar patterns are shown for consumption of proteins
(Figure 12, panel d)). For all generations consumption of proteins sharply
increases with age, with younger generations consuming more proteins that
older generations at the same age.
Similar conclusion can be drawn from Figure 13 representing consumption
of food by cohort where again we notice that the younger cohorts tend to
have a higher average consumption than the older but this is not always the
case because of the within cohort movements. Each cohort seems to follow
the time effect studied in the previous section. A deeper analysis of these
figures would require us to distinguish the three effects: age, cohort and years
effects. This would require to estimate the decomposition of effects regressing
cohort averages of consumption against dummy variables for all three sets of
effects (Deaton and Paxson, 1994). We do not present this analysis here, but
it can be the object of future studies.
41
Figure 12: Estimated intakes-cohort curves using non-liner model with rough-ness penalty λ =100.
50
01
00
01
50
02
00
02
50
03
00
0K
cal/d
ay/
capita
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Calories
(a) Calories.0
25
50
75
10
01
25
gr/
day/
capita
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Fat Intakes
(b) Fat Intake.
10
20
30
40
50
%K
cal/d
ay/
capita
fro
m fat
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Proportion of energy from fat
(c) Proportion of energy from fats.
01
02
03
04
0gr/
day/
capita
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Proteins
(d) Proteins.
01
00
20
03
00
40
05
00
gr/
day/
capita
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Carbohydrates
(e) Carbohydrates.
05
10
15
mg/d
ay/
capita
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Iron
(f) Iron.
02
00
40
06
00
80
0m
g/d
ay/
capita
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Calcium
(g) Calcium.
03
69
12
15
mg/d
ay/
capita
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Vitamin C
(h) Vitamin C.
42
Figure 13: Estimated food-cohort curves using non-liner model with rough-ness penalty λ =100.
01
00
20
03
00
gr/
day/
capita
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Milk
(a) Diary products.
05
01
00
15
02
00
25
03
00
gr/
day/
capita
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Meat and fish
(b) Meat and fish.
01
00
20
03
00
40
0gr/
day/
capita
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Oil and fats, sugar and preservative
(c) Fats, oils, sugar and preserva-tives.
01
50
30
04
50
60
07
50
gr/
day/
capita
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Cereals, Pasta and Rice
(d) Cereals, pasta, rise and bread.
02
00
40
06
00
80
01
00
0gr/
day/
capita
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Vegetables
(e) Vegetable.
01
02
03
0gr/
day/
capita
0 10 20 30 40 50 60age
cohort 1945 cohort 1950 cohort 1955 cohort 1960 cohort 1965
cohort 1970 cohort 1975 cohort 1980 cohort 1985 cohort 1990
Fruits
(f) Fruit.
43
4.4 INCOME
The previous sections demonstrated that in general nutrition varies over age.
It has also been shown that children consumption have changed over time.
In particular, the findings show a common change among all the aspects an-
alyzed around 1993 and some other little and single changes over the last
26 years. We turn now to examine whether the accumulation of household
income has played a role in the way people eat. Poorer people may be more
likely to malnutrition that leads to poorer health status. In addition, their
families may be less able to provide the investment necessary to maintain
good diet in the presence of low income. In doing so I explore the relation-
ship between children’s consumption and per capita family income10, and I
analyze time trends of such a relationship.
To do this we use the estimates on the log of net family income per capita
from the non-linear least square model with Roughness Penalty Function of
nutrient intakes and food consumed by the whole household in one week
period. The estimated coefficients on log family income per capita which
represent elasticities of consumption with respect to income for each nutri-
ent intake and food group, are reported in Tables 5 to 6. This provides
alternative evidence on the health-income gradient discussed by a number of
analysts (e.g., Case et. al., 2002).
The income elasticity reported here measure the proportionate rate of
change in quantity of a nutrient or food consumed from household supply
due to a unit proportionate change in household income per capita, other
individual and household characteristics held constant.
Table 5 and 6 show the estimated income elasticities for each year ob-
tained using NFS data together with estimated standard errors. In all cases
the results indicate that nutrient intakes and food groups, with some excep-
tion in some years for fats from oils, sugar and preservatives, cereals and
vegetables, are ”normal” goods, quantity purchased increases as income rises
10Per capita family income derived from net family income divided by number of mem-bers of the household. The model also controls for family composition. Therefore, we donot use equivalence scales to compute income per capita.
44
at a slower rate (elasticity less than 1) than the rate at which income in-
creases.
For example, the first column in Table 5 reports income elasticity of
calories consumption for each year of the sample (γ) and panel a) in Figure
14 describes its trend over the whole period of study graphically. Elasticity
of calories with respect to income varies in a range between -0.029 and 0.10,
being negative only in 1996. Therefore, apart in 1996, an increase in family
income would augmented daily calories consumption.
Nutrient intakes show relatively low income elasticities. In fact, most of
the elasticities are close to zero (i.e. calories, carbohydrates, iron, calcium).
Elasticity of vitamin C results a bit higher than other intakes, but it does not
exceed 0.5. Fats intakes and proteins show to have been more sensitive to
income variation than other nutrients in the past. However their sensitiveness
to income changes becomes lower with time.
Looking at elasticity of food groups estimates show a general positive
relationship between quantity of food consumed and increment of per capita
family income (γ resulting positive). There are, however, some exceptions.
Elasticity of vegetable consumption (Table 6) being negative or zero from
1980 as well as income elasticity of cereal, pasta, rice and bread. Thus, in this
case, the effect of a rise in per capita family income is a decrease in quantity
of vegetables consumed. Cereals elasticity of consumption has floated around
zero along most of the period considered (panel d of Figures 15) revealing a
general insensitiveness of cereals consumption to income variations.
Different trends are observable for meat and fish products and for fruit.
Here income elasticity is positive, slightly higher than 0.5 at the beginning
of the time period. Both food groups show a trend downwards starting from
1985 (Figure 15, panel b) and f)).
Income elasticities for all milk products and fats from oils, sugars and
preservatives are in average of similar orders of magnitude, between 0 and
0.25. In particular elasticity trends of oils and sugars has been stable since
middle 70s until the beginning of 90s varying between 0.03 and 0.17. Between
1992 and 1998 income elasticities of those products show bigger variability
(range -0.07 to 0.42).
45
Tab
le5:
Ela
stic
ity
ofin
take
consu
mpti
onre
spec
tto
fam
ily
inco
me
per
capit
a[γ
from
NL-O
LS].
Inta
kes
γco
effici
ents
.C
alo
rie
s.e.
fat
inta
ke
s.e.
pro
tein
ss.
e.ca
rbohydra
tes
s.e.
Calc
ium
s.e.
Iron
s.e.
Vit
am
inC
s.e.
1975
.0521827
.013635
.1167741
.0167911
.1315306
.0192628
.0044775
.0152588
.0761469
.0106096
.0676739
.0144421
.433347
.0197529
1976
.0194216
.0146524
.0776474
.01728
.2034976
.0193733
-.0242167
.0170687
.0560549
.0116148
.033079
.0164354
.4520648
.019529
1977
.0223059
.0153302
.0928694
.0177008
.2389275
.0185802
-.0500879
.0178905
.0680019
.0127519
.0582841
.0168321
.4770511
.0201174
1978
.1005065
.0167508
.1643151
.019016
.2342044
.020263
.0533032
.0206522
.104681
.0132066
.1249448
.017267
.5038663
.020817
1979
.0652827
.0158257
.1333762
.0192426
.5644395
.014227
.0104416
.0176412
.0957686
.0132073
.083899
.0169489
.5598125
.0216536
1980
.0396437
.0170043
.1163824
.0201602
.222104
.021684
-.0279415
.0190227
.092391
.0125377
.0931727
.0173877
.5059993
.0184271
1981
.0534028
.0167465
.1280234
.0195514
.1698825
.0189892
-.0038197
.0188858
.0798878
.0138231
.103069
.0168843
.5054457
.0197932
1982
.0638088
.0165285
.1246046
.0192073
.1746992
.0201141
.0122623
.0184933
.1016478
.0126084
.098097
.0160846
.4910612
.0185739
1983
.0536164
.0159645
.094478
.0182227
.1601869
.0161979
.0208739
.019004
.1083945
.0138056
.1002752
.0163094
.528724
.0214616
1984
.0364573
.0173916
.0936876
.0196823
.1225217
.0188115
-.0173819
.0200418
.0893111
.0136746
.0984656
.0170675
.5084264
.0214614
1985
.0202297
.015328
.0773886
.0180477
.1507015
.0180142
-.0408154
.0170538
.1016032
.0132153
.0833701
.0159827
.5252997
.0204438
1986
.0534933
.0167879
.1236182
.0192564
.1654341
.0183505
-.0044793
.0187338
.1080226
.0139527
.1086249
.0165982
.515379
.0223539
1987
.0607489
.0156137
.1116535
.0182843
.2174186
.0173792
.0117148
.0173355
.1035324
.0133749
.1252697
.0160744
.5169719
.0214016
1988
.0559557
.0148751
.0686199
.0178326
.0973729
.0171512
.0484176
.0162578
.1085295
.0125052
.1115609
.0153175
.5027014
.0218625
1989
.0201637
.0157693
.0325466
.0190589
.074367
.0161255
.0156426
.0172759
.0899822
.0124514
.0749165
.0153058
.4983603
.0181812
1990
.024642
.015773
.0599366
.0186001
.11897
.0177573
-.0105793
.0173192
.0988949
.013101
.0895318
.0172017
.4626124
.0197981
1991
.0370192
.01399
.0637211
.0166486
.0979982
.014205
.0124217
.0163904
.0751826
.0124917
.0796878
.0146764
.4228986
.0176878
1992
.0467337
.0143042
.06801
.0169676
.0945982
.0142942
.0220627
.0161621
.0781438
.0125673
.0943081
.0150117
.4356766
.0154821
1993
.0655788
.0145824
.0944741
.0170285
.1084111
.0143728
.0316449
.0161369
.098801
.0121917
.1251528
.0146895
.4635633
.0155525
1994
.0316565
.0131103
.0221296
.0159683
.0841295
.0133948
.0153433
.0141781
.0731581
.0113561
.1031358
.0136138
.434798
.0150316
1995
.0101366
.0133681
.0148092
.0157804
.0876294
.0134178
-.027448
.0145453
.0612391
.0109298
.0927613
.0133341
.3949915
.0155369
1996
-.029737
.0148042
.0158789
.0152261
.043583
.0134069
-.1145085
.0182407
.0834132
.0113091
.0805311
.013016
.375404
.0147514
1997
.0286663
.0146283
.0334735
.0183954
.0729259
.0149547
.0095784
.0155846
.0710729
.0122255
.1218994
.0145097
.4197164
.0152957
1998
.0557527
.0143096
.0558842
.0182315
.0749743
.0155672
.0323474
.01523
.0480176
.0129171
.1215937
.0151048
.3726525
.0203701
1999
.028998
.016552
.0022434
.0190215
.0635345
.0154713
.0245117
.0190138
.0634992
.0124109
.1060496
.0151578
.3363529
.0190049
2000
.0121733
.0139528
-.0088689
.0165324
.054557
.014712
.0020739
.0156034
.033195
.0119333
.092326
.0152234
.3156308
.0192067
46
Finally, there is some evidence of changes through time in income elastic-
ities both for nutrient intakes and food groups. Effects of variation in income
are slightly stronger on food than on nutrients consumption. This may re-
flect the fact that consumers, at different income level, substitute between
food groups in a way that substitution within nutrients results very little
(Subramanian and Deaton, 1996). Possible drivers of such effects might be
sought in changes through time in the nature of food and in the way they are
presented to households, changes in the technology available for preparing
foods, changes in household circumstances including increased labor market
participation and cost of time, and so forth.
Although during the period of study increments of income have implied
little positive changes in quantity of intakes consumed, at this point it is not
possible to say whether consuming more nutrient intakes implies a better
diet and therefore a better health status.
47
Figure 14: Estimated nutrient intakes elasticity trend, λ =100.
−1
−.7
5−
.5−
.25
0.2
5.5
.75
1e
last
icity
1975 1980 1985 1990 1995 2000time
elasticity 95% confidence interval
Calories
(a) Calories.
−1
−.7
5−
.5−
.25
0.2
5.5
.75
1e
last
icity
1975 1980 1985 1990 1995 2000time
(b) Fat Intakes.
−1
−.7
5−
.5−
.25
0.2
5.5
.75
1e
last
icity
1975 1980 1985 1990 1995 2000time
(c) Proteins.
−1
−.7
5−
.5−
.25
0.2
5.5
.75
1e
last
icity
1975 1980 1985 1990 1995 2000time
(d) Carbohydrates.
−1
−.7
5−
.5−
.25
0.2
5.5
.75
1e
last
icity
1975 1980 1985 1990 1995 2000time
(e) Iron.
−1
−.7
5−
.5−
.25
0.2
5.5
.75
1e
last
icity
1975 1980 1985 1990 1995 2000time
(f) Calcium.
−1
−.7
5−
.5−
.25
0.2
5.5
.75
1e
last
icity
1975 1980 1985 1990 1995 2000time
(g) Vitamin C.
48
Tab
le6:
Ela
stic
ity
offo
od
grou
ps
consu
mpti
onre
spec
tto
fam
ily
inco
me
per
capit
a[γ
from
NL-O
LS].
Food
Gro
ups
γco
effici
ents
milk
s.e.
mea
tgr
oup
s.e.
fats
grou
ps.
e.ce
real
ss.
e.ve
geta
bles
s.e.
frui
ts.
e.
1975
.094
9065
.012
1076
.455
0134
.024
3509
.141
6234
.019
671
.000
9174
.016
4525
.065
4628
.026
8947
.529
3232
.024
9877
1976
.089
37.0
1296
65.5
4075
28.0
2094
89.1
1421
48.0
2311
96-.03
0950
3.0
1812
35.0
0362
69.0
2594
19.5
9246
39.0
2447
6519
77.1
2675
29.0
1303
25.6
5036
39.0
2612
43.1
2385
48.0
2274
65-.07
3628
1.0
1810
47.0
4293
42.0
2977
93.5
8628
6.0
2401
5119
78.1
1321
64.0
1377
06.5
1399
15.0
2389
06.1
7564
54.0
2841
9.0
5342
14.0
2031
03.1
7688
61.0
3160
38.5
7985
46.0
2737
1979
.110
7348
.013
9747
.529
6338
.023
1536
.077
1633
.023
7504
-.01
6206
5.0
1832
06.1
7917
45.0
2826
97.6
9127
37.0
2756
619
80.1
1266
02.0
1211
93.5
5441
05.0
2379
27.0
3368
21.0
2915
32-.02
0917
4.0
1926
12.1
5326
51.0
2692
6.5
6981
92.0
2405
0219
81.1
0298
12.0
1319
37.5
0428
61.0
2297
46.1
0578
4.0
2561
48.0
0993
73.0
1995
51-.02
2562
7.0
2850
73.5
7591
12.0
2335
919
82.1
1842
27.0
1218
26.4
8538
79.0
2087
36.1
4935
77.0
2236
09.0
1805
6.0
1970
89.0
2325
88.0
2546
21.5
9386
89.0
2526
6819
83.1
1804
85.0
1461
46.4
5380
04.0
1903
71.0
9883
58.0
2443
78.0
2595
78.0
1875
8-.01
5696
5.0
2823
15.5
5643
52.0
2582
6519
84.1
1216
57.0
1425
08.4
7777
24.0
2101
26.1
2703
7.0
2472
05.0
1490
42.0
1931
76-.02
1243
7.0
3019
91.6
1357
7.0
2806
1119
85.1
3200
23.0
1428
54.4
4364
15.0
1968
17.0
5356
89.0
2378
26-.02
0432
2.0
1714
49-.10
1822
9.0
2828
31.6
4237
65.0
2829
3119
86.1
3033
73.0
1480
44.1
9073
9.0
2705
22.1
4329
23.0
2383
13.0
4742
89.0
1843
64-.05
0159
9.0
2822
88.5
9510
9.0
2523
2519
87.1
0725
05.0
1493
22.5
0992
55.0
1884
6.0
8998
62.0
2141
47-.00
8804
7.0
1832
35.0
5162
96.0
2417
78.5
6892
.025
8693
1988
.124
4639
.013
6427
.137
8042
.021
8584
.111
4309
.022
8924
.026
037
.016
0657
.003
298
.024
8569
.524
5033
.023
5646
1989
.097
3338
.013
5209
.408
8185
.017
7708
.153
3527
.022
854
.015
3771
.017
3801
-.01
9797
9.0
2498
08.5
8483
76.0
2382
1119
90.1
3685
58.0
1447
18.4
4620
07.0
1889
27.1
4365
42.0
2507
98.0
1460
31.0
1684
89-.01
9562
1.0
2705
15.5
2351
9.0
2391
1519
91.0
7423
19.0
1426
69.1
2276
74.0
2113
98.0
7855
66.0
2248
57.0
3887
97.0
1590
96-.05
1530
8.0
2438
98.4
8939
45.0
2009
5419
92.0
8605
6.0
1428
91.1
3030
84.0
2126
88.0
5421
48.0
2581
32.0
1822
65.0
1675
32-.04
3096
5.0
2220
2.5
0515
59.0
1975
319
93.1
1177
82.0
1295
3.0
7181
59.0
2355
9.4
2392
69.0
1772
33.0
0703
8.0
1682
09-.03
0240
8.0
2353
16.4
9032
18.0
1983
5119
94.0
8540
39.0
1264
86.0
6962
73.0
1891
5-.02
7651
7.0
2677
65.0
1563
58.0
1482
17.0
0462
47.0
2075
15.4
8955
26.0
1848
7919
95.0
7059
17.0
1230
53.0
8611
06.0
1992
19-.14
1324
3.0
3096
51-.02
7134
9.0
1515
33-.03
4977
8.0
1999
98.4
4136
2.0
1814
8519
96.0
9216
48.0
1256
22-.06
5591
2.0
2058
96-.07
3020
9.0
2554
22-.13
4876
2.0
2104
73.0
2554
26.0
1833
25.4
0572
24.0
1619
1819
97.0
5080
53.0
1363
12.0
3679
01.0
2188
65.3
8450
1.0
2180
99.0
3075
9.0
1593
64.0
1638
36.0
2125
05.4
6248
88.0
1721
3719
98-.00
1461
2.0
1509
23.0
4434
7.0
2565
21-.03
8087
8.0
3114
21.0
4502
69.0
1627
45.0
4804
98.0
2137
87.4
2579
53.0
2399
5619
99.0
2650
9.0
1464
39.0
7562
69.0
2646
06-.03
0935
7.0
3391
91.0
4478
54.0
2135
93.0
2881
59.0
2260
49.4
1014
32.0
2156
1120
00.0
3484
87.0
1463
69.0
6979
74.0
2383
11-.02
8336
6.0
3040
83-.00
2115
2.0
1756
18.0
5608
45.0
1986
89.3
3971
32.0
2255
79
49
Figure 15: Estimated food groups income elasticity trend, λ =100.
−1
−.7
5−
.5−
.25
0.2
5.5
.75
1e
last
icity
1975 1980 1985 1990 1995 2000time
(a) Diary Products.
−1
−.7
5−
.5−
.25
0.2
5.5
.75
1e
last
icity
1975 1980 1985 1990 1995 2000time
(b) Meat and fish.
−1
−.7
5−
.5−
.25
0.2
5.5
.75
1e
last
icity
1975 1980 1985 1990 1995 2000time
(c) Fats, oils, sugar and preserva-tives.
−1
−.7
5−
.5−
.25
0.2
5.5
.75
1e
last
icity
1975 1980 1985 1990 1995 2000time
(d) Cereals, pasta, rise and bread.
−1
−.7
5−
.5−
.25
0.2
5.5
.75
1e
last
icity
1975 1980 1985 1990 1995 2000time
(e) Vegetable.
−1
−.7
5−
.5−
.25
0.2
5.5
.75
1e
last
icity
1975 1980 1985 1990 1995 2000time
(f) Fruit.
50
4.5 Household Characteristics: household composition,
region of residence, presence of children
The models employed consider also some household characteristics, such as
per capita family income, household composition, region of residence and
presence of children. Summary statistics for the explanatory variables were
given in Table 2.
As shown by Chesher (1997) the effects of alternative amounts of smooth-
ing on estimates of the coefficients on household characteristics are negligible.
Tables 7 and 8 show coefficients estimated for regions, household compo-
sition and presence of children in 2000 using λ = 100 employed earlier11 on
nutrient intakes and food consumption respectively.
Estimates coefficients on nutrient intakes consumption for Scotland are
uniformly negative and significantly different from zero for most of the mod-
els employed (calories, fat intakes, carbohydrates, iron and vitamin C) with
exception of calcium, meaning that Scotish household consumption of intakes
is lower than in London and South East of England. Estimates for Northern,
Central, South West of England and Wales result significantly different from
zero only on calcium and vitamin C models. Consumption of calcium results
between 3 and 4 percent higher in the rest of England than in London and
the South East, whilst coefficients estimated on vitamin C are uniformly neg-
ative for the three region considered, meaning that they consume in average
less than people in London and South East (respectively 20 percent less in
Scotland, 12 percent in Northern England and 5 percent less in the Central,
South West and Wales).
Differences in intakes consumption might be due to different diets across
regions. In fact, as shown in Table 8, there is a significant regional effect on
consumption of milk, fat from oils and sugar, vegetable and fruit. Although
most of the coefficients across region result significantly different from zero,
consumption of meat and fish, and cereal, pasta, bread and rice do not differ
by region.
11The coefficients estimated for each year of survey considered are reported in Appendixat the end of this chapter
51
In particular, Scotish households consume less fruit and vegetable than
londoners and south-eastern, respectively 21 and 17 percent less. Northern
England consume 7 percent more milk, but 9 percent less fruit as well as
Central, South West of England and Wales. The latter consumes also more
vegetable (6 percent) and less fat from oils and sugar.
The analysis carried on in this paper assumes that expected consumption
of people of a given age and sex is independent from the presence, demo-
graphic characteristics of other household members. Of course this is an
approximation, in the sense that the coefficients estimated for each age and
gender represent average consumption across household composition types.
Therefore, when introducing household characteristics, we control also for
household composition introducing a dummy variable by type of household12
and an indicator of presence of children.
The coefficients estimates on presence of children are always positive and
significant at different critical levels for all nutrient intakes and food groups.
This clearly means that the presence of children is a main explanation of dif-
ferences in household consumption. In particular the magnitude of nutrient
intakes variation is in average around 10 percent higher respect a household
without children. This is mainly due to variation arising in food consump-
tion, as for example the increase in consumption of fat from oils and sugar of
about 13 percent as well as a 10 percent higher consumption of cereals and
vegetable.
In general, the effect of household composition is not largely significant,
and the presence of children remains the main explanation of differences
in household consumption. However it is worthy to highlight some results.
Coefficients estimated on the effect of different household composition on
consumption of fat from oils and sugar result uniformly negative (with ex-
ception of household with a single adult) and significantly different from zero.
The base category here is a household with two adults. It is interesting to
notice that a person living alone consume in general 33 percent more fats
121) one adult; 2) one adult and 1 or more children; 3) 2 adults; 4) 2 adults and 1 child;5) 2 adults and 2 children; 6) 2 adults and 3 children; 7) 2 adults and 4 or more children;8) 3 adults; 9) 4 or more adults; 10) 3 or more adults and 1 or 2 children; 11) 3 or moreadults and 3 or more children
52
Tab
le7:
Model
sfo
rnutr
ient
inta
kes:
esti
mat
edco
effici
ents
for
hou
sehol
dch
arac
teri
stic
s(t
valu
esin
bra
cket
s)-
2000
.
Nut
rien
tIn
take
sC
alor
ies
Fat
Inta
kes
Pro
tein
sC
arbo
hydr
ate
Cal
cium
Iron
Vit
amin
CSc
otla
ndvs
Lon
don
and
SEE
ngla
nd-0
.079
-0.0
79-0
.026
-0.0
870.
000
-0.0
93-0
.195
(-2.
810)
(-2.
360)
(-0.
870)
(-2.
730)
(0.0
20)
(-2.
930)
(-4.
660)
Nor
ther
nE
ngla
nd-0
.022
-0.0
380.
028
-0.0
240.
034
-0.0
03-0
.118
(-1.
170)
(-1.
700)
(1.4
30)
(-1.
160)
(2.0
70)
(-0.
130)
(-4.
370)
Cen
tral
,So
uth-
Wes
tan
dW
ales
0.00
90.
006
0.00
70.
012
0.03
90.
014
-0.0
55(0
.520
)(0
.310
)(0
.370
)(0
.590
)(2
.560
)(0
.720
)(-
2.25
0)1
adul
ton
lyvs
2ad
ults
0.03
20.
022
-0.0
320.
065
0.02
6-0
.013
-0.1
31(0
.710
)(0
.420
)(-
0.65
0)(1
.260
)(0
.650
)(-
0.24
0)(-
1.94
0)1
adul
tan
d1
orm
ore
child
ren
0.08
80.
065
0.12
00.
099
0.16
30.
150
0.34
6(1
.410
)(0
.860
)(1
.750
)(1
.450
)(2
.950
)(2
.130
)(3
.700
)2
adul
ts,1
child
-0.0
12-0
.029
-0.0
080.
001
0.05
00.
050
0.14
2(-
0.29
0)(-
0.60
0)(-
0.20
0)(0
.020
)(1
.400
)(1
.110
)(2
.440
)2
adul
ts,2
child
ren
0.01
70.
034
0.02
80.
006
0.05
90.
056
0.30
8(0
.370
)(0
.620
)(0
.570
)(0
.110
)(1
.420
)(1
.040
)(4
.310
)2
adul
ts,3
child
ren
-0.0
69-0
.082
-0.0
49-0
.060
0.02
5-0
.007
0.36
8(-
1.17
0)(-
1.18
0)(-
0.76
0)(-
0.93
0)(0
.470
)(-
0.10
0)(3
.960
)2
adul
ts,4
orm
ore
child
ren
-0.0
33-0
.030
0.09
2-0
.048
0.06
50.
099
0.50
1(-
0.45
0)(-
0.34
0)(1
.140
)(-
0.60
0)(1
.020
)(1
.190
)(4
.160
)3
adul
ts-0
.021
-0.0
310.
027
-0.0
140.
010
-0.0
160.
054
(-0.
670)
(-0.
870)
(0.8
40)
(-0.
390)
(0.3
70)
(-0.
450)
(1.1
709
4or
mor
ead
ults
-0.0
94-0
.097
-0.0
05-0
.095
-0.0
73-0
.041
-0.0
45(-
2.10
0)(-
1.86
0)(-
0.12
0)(-
1.87
0)(-
1.79
0)(-
0.79
0)(-
0.61
0)3
orm
ore
adul
ts,1
or2
child
ren
-0.0
27-0
.040
-0.0
79-0
.007
0.00
00.
019
0.18
8(-
0.61
0)(-
0.77
0)(-
1.62
0)(-
0.15
0)(0
.010
)(0
.360
)(2
.710
)3
orm
ore
adul
ts,3
orm
ore
child
ren
-0.1
77-0
.214
-0.1
87-0
.143
-0.0
53-0
.125
0.11
8(-
2.47
0)(-
2.48
0)(-
2.27
0)(-
1.87
0)(-
0.85
0)(-
1.49
0)(0
.970
)P
rese
nce
ofch
ildre
n0.
117
0.11
50.
102
0.13
70.
076
0.08
20.
113
(4.5
60)
(3.8
30)
(3.8
80)
(4.6
70)
(3.3
20)
(2.8
50)
(3.0
60)
53
Tab
le8:
Model
sfo
rfo
od
grou
ps:
esti
mat
edco
effici
ents
for
hou
sehol
dch
arac
teri
stic
s(t
valu
esin
bra
cket
s)-
2000
.
Food
grou
psM
ilkM
eat
and
Fis
hFa
tsan
dsu
gars
Cer
eals
Veg
etab
leFr
uit
Scot
land
vsLon
don
and
SEE
ngla
nd0.
044
0.00
4-0
.021
-0.0
40-0
.172
-0.2
11(1
.490
)(0
.070
)(-
0.35
0)(-
1.12
0)(-
3.85
0)(-
3.99
0)N
orth
ern
Eng
land
0.07
40.
042
-0.0
450.
014
0.00
1-0
.091
(3.6
10)
(1.2
80)
(-1.
060)
(0.5
80)
(0.0
30)
(-2.
790)
Cen
tral
,So
uth-
Wes
tan
dW
ales
0.05
00.
023
-0.0
820.
022
0.06
3-0
.089
(2.5
70)
(0.7
00)
(-2.
060)
(0.9
50)
(2.5
30)
(-3.
010)
1ad
ult
only
vs2
adul
ts0.
026
-0.0
340.
333
0.11
5-0
.070
-0.1
19(0
.500
)(-
0.44
0)(4
.960
)(1
.970
)(-
1.04
0)(-
1.77
0)1
adul
tan
d1
orm
ore
child
ren
0.24
60.
076
-0.0
890.
079
0.10
90.
363
(3.4
20)
(0.6
70)
(-0.
600)
(1.0
20)
(1.2
00)
(3.3
70)
2ad
ults
,1
child
0.14
7-0
.065
-0.2
510.
000
0.02
20.
184
(3.1
10)
(-0.
900)
(-2.
800)
(0.0
10)
(0.3
90)
(2.6
30)
2ad
ults
,2
child
ren
0.14
9-0
.062
-0.2
06-0
.038
0.07
90.
428
(2.6
90)
(-0.
730)
(-2.
100)
(-0.
640)
(1.1
90)
(5.1
00)
2ad
ults
,3
child
ren
0.12
7-0
.033
-0.3
62-0
.116
0.15
30.
558
(1.8
80)
(-0.
300)
(-2.
850)
(-1.
590)
(1.7
80)
(4.8
30)
2ad
ults
,4
orm
ore
child
ren
0.26
80.
081
-0.4
85-0
.100
0.38
20.
711
(3.3
20)
(0.6
10)
(-2.
820)
(-1.
120)
(3.5
60)
(4.3
10)
3ad
ults
0.00
9-0
.014
-0.3
76-0
.029
0.04
70.
035
(0.2
40)
(-0.
260)
(-5.
650)
(-0.
710)
(1.1
20)
(0.6
40)
4or
mor
ead
ults
0.03
90.
026
-0.5
47-0
.042
-0.1
30-0
.054
(0.7
20)
(0.3
30)
(-4.
930)
(-0.
750)
(-1.
950)
(-0.
560)
3or
mor
ead
ults
,1
or2
child
ren
0.04
0-0
.090
-0.4
170.
000
-0.0
030.
311
(0.7
30)
(-1.
050)
(-4.
560)
(0.0
00)
(-0.
040)
(3.6
00)
3or
mor
ead
ults
,3
orm
ore
child
ren
0.13
9-0
.156
-0.4
67-0
.145
-0.0
750.
340
(1.7
50)
(-1.
060)
(-2.
960)
(-1.
710)
(-0.
690)
(2.0
90)
Pre
senc
eof
child
ren
0.03
80.
076
0.13
50.
105
0.10
80.
045
(1.2
50)
(1.7
40)
(2.7
50)
(3.1
10)
(3.0
70)
(1.0
30)
54
than two adults living together, while increasing the number of adults in the
household (3 or 4 people) consumption of fats decreases up to respectively
38 and 55 percent less. There might be many factors driving this results.
People living alone might prefer to buy more ready meals or eat outside the
home because they do not have time to cook or because they do not enjoy it.
On the other side, adults sharing accommodation with other adults might
be single looking for partner and therefore caring more about their physical
aspect and their diet, they might also find in the time spent cooking a mo-
ment for socializing, and sharing rent might allowed them to have a easier
access to healthy but more expensive food like vegetable and fruit. The last
point does not however arise from the estimates of vegetable and fruit where
the coefficients result negative and not significantly different from zero for
household of adults.
It is also interesting to see that when children arrive, the couple decreases
significantly the amount of fats and sugar consumed and increases amount
of fruit and vegetable13.
There is not way to know here whether the presence of children or living
with others improve the quality of diet. The negative signs associated with
household composition in the fats and sugar model and the positive signs
shown in the fruit consumption model give a first hint on this direction.
Therefore this might be a point to be developed in future studies.
5 Discussion and Outline for future work
This paper has started to explore how eating habits of people in Britain have
changed over the last twenty-five years of the twentieth century. Using data
from 1975-2000 from the National Food Survey this paper reports an exten-
sive descriptive analysis that investigates the relationship between average
nutrient intake and key food groups consumption across ages and over time.
In doing so we estimates a Roughness Penalty Function Model obtained from
ordinary least squares method to account for function smoothness (Chesher,
13Coefficients estimated on consumption of vegetable are positive but not significantlydifferent from zero.
55
1997). I investigate nutrition curves - using nutrient intakes and major food
groups - with the objective to see how they have changed by gender and age
over time and by gender and time for all age groups and, in particular, for
children 0-17.
We stress five main results. First findings demonstrated that in general
nutrition varies over age and by gender. In general males consume more food
and nutrient intakes than females. Nutrients consumption strongly increase
during childhood until puberty, decrease at the beginning of adulthood age
and increase later on, decreasing again when people get older. Age distribu-
tion of consumption by major food groups show a general increase up to age
50 and decline afterwards.
The second finding focuses on changes in food consumption and nutrient
intakes over time and in particular among British youth by three age groups.
The results show a change in trend for some nutrient intakes such as fat
intake and proteins that increase along all the time period of study and for
calcium from the 90s, and a tendency to decrease in consumption of iron.
In particular the proportion of energy from fat increased at the end of the
80s to 35 percent and it is stable from there since. The variations emerged
in nutrient intakes might be due to variation in food consumption and to
some variation of the data collection process. In particular the data show an
increases in milk products, meat and fish products, fat from oils and sugars
and cereals since the beginning of the 90s.
The third finding focuses on cohort analysis in order to see whether dif-
ferent generations eat differently. We compare ten birth-cohorts and present
results both for nutrients intakes and food groups. The most interest find-
ings regard calories and fat intakes. While calories consumed do not change
a lot across generations, younger generations consume higher quantity of fats
intakes and proteins. The consequences of this can be seen in the proportion
of energy from fats, that for younger generation results much higher than
older generation at the same age. Younger generation consume also more
milk and fat from oil and sugars and less vegetable than their parents at the
same age.
56
In the fourth part we consider the effect of income on eating habits.
Therefore, focusing on the relation between eating habits and income distri-
bution, trends of elasticity of intakes and food consumption with respect to
income have been computed. The findings highlight that changes among nu-
trient intakes and food consumption due to income variations are relatively
low (all less than 1) and in general positive, meaning that as income rises
consumption rices as well but less than proportionally. In general the sensi-
tiveness of consumption to income variation becomes smaller with most of the
trends tending to zero. Finally, there some evidence of changes through time
in income elasticities both for nutrient intakes and food groups. However,
the effect of family income variation is much higher on food groups than on
intake nutrients consumption. This means that as households become richer,
the substitution between foods is much quicker than the variation of diet
through substitution within nutrient intakes consumption. In other words,
for people is much easier to change food quantity consumed than quantity of
intakes. However, at this moment it is not possible to say whether a positive
variation of family income improves nutrition and therefore health status.
The fifth finding reports on the effect of region of residence, household
composition and presence of children. The effects on nutrient intakes are not
very large, on the contrary the major findings appear in the food estimates.
Region mainly differ in consumption of vegetable and fruit. The presence of
children results the main explanation of differences in household consumption
for all nutrients and foods. The presence of children increases the amount of
fruit and reduces the amount of fat from oils and sugar consumed.
In considering these results, some extension should be considered for fu-
ture research. Many might be the causes of eating habits changes resulting
from the analysis carried on in this paper. For example technical change, in-
come growth, lifestyle changes, mass media and advertising, and changes in
relative prices. In fact, technical changes have provided food supply system
with mechanisms that increase productivity and improve food conservation
and distribution system. Moreover, development of supermarkets has greatly
changed supply chains system. Today, supermarkets make many new prod-
ucts available wherever in the world - either from other countries and out of
57
season - while small and local shops are increasingly less present.
In the next chapter we will study a demand system for some food groups
(like for example fats from oils and sugar, and vegetable and fruit) in order
to explore the effect of prices on household food demand.
58
References
Case A., Lubotsky D. and Paxson C. (2002), Economic Status and Health in
Childhood: The Origins of the Gradient, The American Economic Review,
92, 5, pp. 1308-1334.
Chesher A. (1997), D iet Revealed?: Semiparametric Estimation of Nutrient
Intake-Age Relationships, Journal of Royal Statistical Society A, n.160. part
3, pp. 202-203
Chesher A. (1998), Individual demands form household aggregates: time and age
variation in the composition of diet, Journal of Applied Econometrics, 13,
pp. 505-524
Currie J. & Stabile M. 2002, Socioeconomic Status and Health: why is the rela-
tionship stronger for older children?, NBER Working Paper 9098, National
Bureau of Economic Research, Cambridge MA.
Cutler, Glaeser & Shapiro (2003). W hy have Americans become more obese?,
Journal of Economic Perspectives, American Economic Association, 17(3):
93-118.
Deaton A. (1985). Panel Data from Time Series of Cross-Sections, Journal of
Econometrics, 30: 109-126.
Deaton A. (1997). The Analysis of Household Survey, The John Hopkins Uni-
versity Press.
Deaton A., Paxson C. (1994). Saving, Growth, and Aging in Taiwan, in D.A.
Wise (ed.), S tudying in the Economics of Aging, Chicago University Press.
Department of Health (2004). The Summary of Intelligence on Obesity
Green P.J. & Silverman B.W. (1994). Nonparametric Regression and Generalized
Linear Models. A roughness penalty approach. London: Chapman & Hall
Marmot M. & Wilkinson R.G. (1999). Social Determinants of Health, Oxford
University Press, Oxford.
Maxwell S. & Slater R. (2003). Food Policy Old and New, Development Policy
Review, 21 (5-6):531-553.
59
Ministry of Agriculture, Fisheries and Food (1999). National Food Survey, Lon-
don: Her’s Majesty’s Stationery Office.
OECD Health Policy Unit (2003). H ealth at a glance - OECD indicators. Briefing
note.
Schmidhuber J. (2003). The outlook for long-term changes in food consumption
patterns: Concerns and policy options, presented at FAO Scientific Work-
shop 2003.
Subramanian S., Deaton A. (1996). The demand for food and calories, Journal
of Political Economy, 104 (1): 133-162.
Joint WHO/FAO expert consultation on Diet, Nutrition and the Prevention of
Chronic Diseases (2002: Geneva, Switzerland) - WHO technical report se-
ries: 916.
World Health Organization (2004). Obesity and Overweight., Geneva: WHO.
Available on the world wide web: http://www.who.int/dietphysicalactivity/publications/facts/obesity/en/.
60
6 Appendix
61
Tab
le9:
Effec
tof
pre
sence
ofch
ildre
non
nutr
ient
inta
kes
consu
mpti
on.
Coeffi
cien
tsass
oci
ate
dw
ith
pre
sence
ofch
ildre
nby
Inta
kes
.C
alo
rie
s.e.
fat
inta
ke
s.e.
pro
tein
ss.
e.ca
rbohydra
tes
s.e.
Calc
ium
s.e.
Iron
s.e.
Vit
am
inC
s.e.
1975
0.0
88
0.0
22
0.1
02
0.0
26
0.0
91
0.0
30
0.0
78
0.0
25
0.0
87
0.0
18
0.0
59
0.0
24
0.1
59
0.0
37
1976
0.0
86
0.0
22
0.1
14
0.0
26
0.0
48
0.0
34
0.0
73
0.0
26
0.0
70
0.0
18
0.0
38
0.0
25
0.1
11
0.0
38
1977
0.0
92
0.0
23
0.1
00
0.0
26
0.1
12
0.0
28
0.0
88
0.0
28
0.0
92
0.0
21
0.0
69
0.0
26
0.0
66
0.0
40
1978
0.0
94
0.0
27
0.0
74
0.0
31
0.0
43
0.0
34
0.1
22
0.0
33
0.0
74
0.0
23
0.0
74
0.0
29
0.1
71
0.0
41
1979
0.1
54
0.0
23
0.1
83
0.0
27
0.2
68
0.0
38
0.1
23
0.0
27
0.0
93
0.0
20
0.1
58
0.0
24
0.1
27
0.0
39
1980
0.0
76
0.0
28
0.0
93
0.0
33
0.0
44
0.0
37
0.0
66
0.0
32
0.0
72
0.0
23
0.0
36
0.0
30
0.1
58
0.0
42
1981
0.0
56
0.0
26
0.0
74
0.0
31
0.0
50
0.0
29
0.0
48
0.0
31
0.0
31
0.0
23
0.0
34
0.0
28
0.0
52
0.0
42
1982
0.1
04
0.0
26
0.1
13
0.0
30
0.0
42
0.0
32
0.1
08
0.0
29
0.0
95
0.0
21
0.1
09
0.0
25
0.1
46
0.0
38
1983
0.0
52
0.0
26
0.0
91
0.0
30
0.0
67
0.0
26
0.0
22
0.0
33
0.0
45
0.0
23
0.0
40
0.0
27
0.0
92
0.0
45
1984
0.0
93
0.0
29
0.0
97
0.0
32
0.1
04
0.0
30
0.0
83
0.0
34
0.0
93
0.0
24
0.0
96
0.0
29
0.0
72
0.0
47
1985
0.0
94
0.0
25
0.1
10
0.0
29
0.0
98
0.0
28
0.0
86
0.0
29
0.0
60
0.0
22
0.0
76
0.0
27
0.1
37
0.0
41
1986
0.0
79
0.0
26
0.0
64
0.0
31
0.0
52
0.0
29
0.1
04
0.0
30
0.0
42
0.0
22
0.0
45
0.0
27
0.0
80
0.0
44
1987
0.1
03
0.0
26
0.1
00
0.0
31
0.0
80
0.0
30
0.1
16
0.0
29
0.0
87
0.0
24
0.0
98
0.0
28
0.1
89
0.0
48
1988
0.0
79
0.0
25
0.0
57
0.0
30
0.0
64
0.0
30
0.1
03
0.0
28
0.1
18
0.0
22
0.0
69
0.0
27
0.1
41
0.0
46
1989
0.1
34
0.0
25
0.1
77
0.0
30
0.1
76
0.0
25
0.0
94
0.0
30
0.0
73
0.0
22
0.0
93
0.0
26
-0.0
05
0.0
41
1990
0.1
48
0.0
25
0.1
62
0.0
30
0.1
52
0.0
29
0.1
40
0.0
28
0.1
08
0.0
23
0.1
15
0.0
29
0.1
92
0.0
39
1991
0.1
52
0.0
26
0.1
56
0.0
31
0.1
43
0.0
25
0.1
52
0.0
31
0.1
25
0.0
25
0.1
08
0.0
28
0.0
91
0.0
43
1992
0.1
02
0.0
26
0.1
28
0.0
32
0.1
26
0.0
26
0.0
93
0.0
30
0.0
89
0.0
24
0.0
91
0.0
28
0.1
71
0.0
40
1993
0.0
44
0.0
28
0.0
61
0.0
33
0.0
87
0.0
25
0.0
18
0.0
32
0.0
75
0.0
23
0.0
46
0.0
28
0.1
08
0.0
42
1994
0.1
67
0.0
24
0.1
85
0.0
29
0.1
44
0.0
24
0.1
64
0.0
27
0.1
38
0.0
22
0.1
30
0.0
25
0.1
71
0.0
35
1995
0.1
03
0.0
24
0.1
32
0.0
28
0.0
96
0.0
23
0.0
88
0.0
28
0.0
68
0.0
21
0.0
61
0.0
25
0.1
75
0.0
35
1996
0.1
34
0.0
28
0.1
22
0.0
29
0.0
95
0.0
24
0.1
63
0.0
37
0.1
22
0.0
21
0.0
93
0.0
24
0.0
97
0.0
32
1997
0.0
98
0.0
27
0.1
24
0.0
34
0.0
99
0.0
27
0.0
96
0.0
31
0.0
74
0.0
23
0.0
46
0.0
28
0.1
49
0.0
35
1998
0.1
55
0.0
25
0.1
89
0.0
31
0.1
27
0.0
26
0.1
41
0.0
28
0.1
05
0.0
24
0.1
03
0.0
27
0.0
68
0.0
37
1999
0.1
18
0.0
31
0.1
50
0.0
35
0.1
67
0.0
27
0.0
88
0.0
37
0.0
97
0.0
24
0.0
72
0.0
29
0.1
05
0.0
38
2000
0.1
17
0.0
26
0.1
15
0.0
30
0.1
02
0.0
26
0.1
37
0.0
29
0.0
76
0.0
23
0.0
82
0.0
29
0.1
13
0.0
37
62
Tab
le10
:E
ffec
tof
pre
sence
ofch
ildre
non
food
grou
ps
consu
mpti
on[fro
mN
L-O
LS].
Coe
ffici
ents
asso
ciat
edw
ith
pres
ence
ofch
ildre
nby
Food
Gro
ups
milk
s.e.
mea
tgr
oup
s.e.
fats
grou
ps.
e.ce
real
ss.
e.ve
geta
bles
s.e.
frui
ts.
e.
1975
0.09
40.
024
0.10
80.
050
0.06
70.
035
0.07
20.
027
0.09
60.
046
0.10
80.
046
1976
0.07
90.
024
0.07
50.
048
0.04
40.
036
0.08
00.
028
0.11
30.
040
-0.0
370.
050
1977
0.09
10.
024
0.16
90.
046
0.05
70.
037
0.09
40.
029
0.09
60.
048
0.02
80.
048
1978
0.03
80.
027
0.07
00.
058
0.04
60.
050
0.08
80.
034
0.21
80.
055
0.15
10.
050
1979
0.04
90.
024
0.13
10.
050
0.09
50.
036
0.16
10.
028
0.19
00.
046
0.02
50.
053
1980
0.09
50.
025
0.12
70.
062
0.06
90.
048
0.06
70.
034
0.14
50.
052
0.18
10.
051
1981
0.03
30.
025
0.10
30.
050
-0.0
030.
043
0.07
10.
035
0.03
00.
044
-0.0
490.
050
1982
0.04
30.
024
0.10
90.
044
0.05
50.
040
0.14
00.
031
0.12
00.
042
0.09
60.
048
1983
0.05
00.
028
0.06
90.
042
0.07
30.
046
0.00
30.
034
0.07
10.
046
0.07
40.
051
1984
0.06
60.
029
0.12
60.
056
0.06
40.
049
0.10
80.
035
0.02
00.
051
0.06
90.
057
1985
0.07
50.
026
0.15
10.
045
0.02
90.
043
0.07
30.
030
0.15
00.
048
0.06
60.
051
1986
0.00
70.
027
0.06
00.
045
0.11
40.
041
0.09
80.
032
0.08
30.
047
-0.0
120.
052
1987
0.05
10.
030
-0.0
100.
048
0.04
90.
043
0.09
40.
033
0.21
90.
042
0.09
60.
057
1988
0.11
00.
027
0.03
50.
042
0.00
10.
042
0.10
80.
028
0.17
40.
043
0.18
80.
047
1989
0.05
20.
027
0.40
40.
042
0.06
50.
043
0.08
90.
031
0.12
40.
041
-0.0
470.
049
1990
0.05
60.
029
0.20
10.
048
0.07
30.
046
0.12
00.
030
0.22
90.
040
0.14
00.
046
1991
0.05
90.
033
0.13
90.
039
0.09
30.
048
0.17
20.
031
0.01
60.
047
-0.0
020.
050
1992
0.07
50.
030
0.17
10.
039
-0.0
250.
054
0.08
90.
032
0.14
60.
040
0.06
40.
047
1993
0.07
70.
027
0.07
80.
044
0.07
50.
054
-0.0
420.
034
0.06
10.
042
0.11
40.
051
1994
0.09
10.
027
0.14
60.
036
0.11
20.
047
0.13
40.
030
0.15
60.
037
0.11
20.
042
1995
0.05
50.
026
0.06
70.
035
0.06
10.
052
0.11
00.
030
0.06
10.
036
0.17
90.
041
1996
0.09
90.
027
0.10
60.
038
-0.0
700.
043
0.15
40.
047
0.14
80.
033
0.05
60.
038
1997
0.07
50.
029
0.11
00.
045
0.19
30.
074
0.08
50.
033
0.16
00.
036
0.02
90.
042
1998
0.05
40.
030
0.14
10.
043
0.11
30.
054
0.11
90.
031
0.09
80.
036
0.04
70.
045
1999
0.09
70.
030
0.23
00.
047
0.05
70.
060
0.07
80.
042
0.07
50.
040
0.03
90.
045
2000
0.03
80.
031
0.07
60.
044
0.13
50.
049
0.10
50.
034
0.10
80.
035
0.04
50.
044
63