Measuring impact of education and socio-economic factors on health for Pakistan.
Asghar, Zahid ; Attique, Nazia ; Urooj, Amena 等
Education affects health not because of the knowledge and practices
one can learn at school, but rather it shapes individuals life and can
alter the characteristics of an individual to be healthier. Measurement
of health is an abstract concept and health itself is affected by a
number of factors. This study aims at exploring whether there is any
relation in education, gender, and health for Pakistan. Exploratory data
analysis and ordinal logistic regression are used here to assess
relationship between health, education and other socio-economic factors.
It is evident that individuals with higher education level tend to have
better health status than a person with lower levels of education. There
is also evidence of gender being an important determinant of health in
Pakistan. The variation in health status of individuals is studied with
reference to gender, culture, occupations, and social economic status.
This study provides a useful piece of information for the policy-makers
in health and education sectors. The data used in this study was
collected by Pakistan and Medical Research Council under National Health
Survey of Pakistan [NHSP (1990-1994)].
JEL classification: L00, L20, C21
Keywords: Education, Self-reported Health, Ordered Logit, Health
Conditions, Pseudo [R.sup.2]
1. INTRODUCTION
It is a common understanding that people with higher level of
education lead a more healthy life due to their enhanced level of
awareness compared to the less educated individuals. Two important
prerequisites for an effective health policy are; monitoring and
forecasting the population's health and its health determinants.
Health of any individual of that of a society of community is not
dependent on a particular single factor. In fact it is the product of
the interaction of our environments, socio-economic status,
psycho-social conditions and cultural norms and beliefs with our genetic
inheritance. "The social conditions, in which people live,
powerfully influence their chances to be healthy. Indeed factors such as
poverty, social exclusion and discrimination, poor housing, unhealthy
early childhood conditions and low occupational status are important
determinants of most diseases, deaths and health inequalities between
and within countries" [WHO (2004)].
The research on the subject reveals that people belonging to
different socioeconomic groups experience different levels of health,
whereas the factors that lead to different health conditions need to be
identified. [Wilkinson and Marmot (2003)].
The 'social determinants' are the socio-economic
conditions of the people which determine their health. WHO and other
health organisations have identified these determinants; which are
illustrated in Figure 1.
[FIGURE 1 OMITTED]
Studies reveal that schooling is associated with several non-market
outcomes.
Among these non-market returns to schooling, there has recently
been a growing interest in the health returns and is believed that
besides human capital, health capital also emerges from the education.
So it is important to analyse whether education policies help to improve
health. The subject study is aimed at estimating the effect of education
on health in particular and exploring the relation between health and
some other social factors in general in Pakistan. The study has two main
objectives; first to elucidate and analyse the effect of education,
gender, occupation etc. on health; second to understand the mechanism,
by which education, gender and other socio-economic factors can
profoundly affect the health status of an individual.
Keeping the objectives of the study we focus on the relation from
socio-economic factors to health by applying general linear model in
multivariate framework. This is of tremendous importance for our
understanding of determinants of health as well as for our understanding
of how schooling affects and shape individual lives.
The study is outlined as: In Section 2 brief review of the issue is
discussed. Section 3 describes the data and the methodological frame
work applied in the study. Section 4 consists of the exploratory data
analysis of the variables. Section 5 is about the use of ordinal
logistic models, along with the empirical analysis of the estimation
techniques used, and finally we conclude the study.
2. REVIEW OF LITERATURE
Since long, it has been observed and documented that the
educational differences have great impact on health status. Grossman
(1972, 1975) has deeply explored the correlation between education and
health. Over the decades, a number of important mechanisms are proposed
through which director indirect influence of schooling on health can be
studied. The impact of past health on current health and years of formal
schooling is studied by Grossman (1975) where he uses a recursive model
to identify the causal relationship between education and health. In his
model, health capital is measured in terms of Self-reported Health (SRH)
and it is shown that with past health, keeping other variables constant,
schooling has a positive and significant effect on present health.
Health disparities between better and less well educated people
often increase when a new health technology is introduced. Health
disparities between better and less well educated people often increase
when a new health technology is introduced [Case (2001)].
Treating schooling as endogenous to health suggests that most of
the correlation between schooling and health is attributable to
unobserved heterogeneity except possibly at low levels of schooling for
individuals with low cognitive ability [Christopher and Sidhu (2005)].
They also identify the role of cognitive ability in the health education
relation and shows that both schooling and ability are strongly
associated with health at low levels but less related or unrelated at
high levels. Arendt (2001) analysed the extent to which heterogeneity in
health and endogeneity of education explained the gradient in health. By
making use Self-reported Health (SRH), Body Mass Index (BMI) and
indicators for high blood pressure and never been smoking, he shows that
education is related to SRH when controlling for the three other health
measures, which can be interpreted as inputs in health production. The
diverse demographic and socio- economic conditions and the availability
of educational facilities affect the extent of heterogeneity in health
and health related quality of life. Shumueli (2003) decomposed the
demographic and socioeconomic factors to study this heterogeneity.
The complex multidimensional structure of health has been of keen
interest for researcher since years. Using the demographic and
socio-economic factors, attempts of identifying the complex relation of
health and education have been made. Fuchs (2004) observes that there
are considerable uncertainties concerning the socio-economic correlates
of health, the extent to which they reflect causal chains and their
implications for policy and studies the possible reasons for this
uncertainty. The inequality in health from the perspective of
socio-economic factors is analysed in a study by Syed, et al. (2006), in
which he has considered two ethnic groups and has found a large
diversity of SRH and prevalence of diabetes and distress among the
ethnic groups.
The possibility of a causal relationship between education and
health is explored by [Arendt (2005)]. Along with SRH, the study
includes BMI and an indicator of never been smoking as supplemental
outcomes. The study shows that education is associated with better SRH
for both men and women. In an attempt to investigate the direct
relationship between education and health, Cutler and Lleras (2007) find
that better educated individuals have more positive health outcomes even
after controlling for job characteristics, income and family background.
Ardent (2008) articulates this causal relationship in terms of
hospitalisation and finds the significant effect of increase in
education on decrease in hospitalisation especially for females.
Evidence for a causal relationship running from better schooling to
better health can be found in an investigation conducted by Silles
(2009). In which by relying on changes in educational participation
caused by raising the school the minimum school-leaving age, also
provides evidence of the causal effect of schooling on health.
Cutler, et al. (2005) described that the link between social status
and health as complex, perhaps too complex for a single explanation.
Discussing the direct causal mechanisms running from income to health,
they have pointed that the link between income and health is a result of
the latter causing the former rather than the reverse. There is most
likely a direct positive effect of education on health but there are no
well stated causal mechanisms.
Hartong and Osterbeek (1998) have studied the returns to education
in terms of health status, financial wealth and happiness, and have
concluded that IQ independently affects health status, even after
controlling for schooling. Returns to education have also been
calculated by relating the value of health gain to the average income
per capita [Groot and Brink (2007)]. The effect of education on health
is analysed by giving some tests for causality, and control for
unobserved heterogeneity; it is found that of gender, the education and
the number of years of education have a positive effect on the quality
of education.
Cutler and Richardson (1998) measured the change in health capital
by age, gender, race and income and concluded that measuring changes in
health by income or education is more difficult than by measuring it by
race and gender. More insight in the size of the quality of health
effect can be obtained by relating the value of the health gain to the
average income per capita.
Costa and Uch6a (2004) determined the factors associated with
self-rated health among adults, considering five dimensions of socio
demographic variables. And it has been observed that self-rated health
among older adults is multidimensional in structure, being influenced by
socioeconomic conditions, social support, health status (with emphasis
on mental health), and access/use of healthcare services but not by the
life style.
3. DATA DESCRIPTION
The data used in this study is collected under National Health
Survey of Pakistan (NHSP 1990-1994). NHSP round II by PMRC is under
progress and its data is not available till date. It is a cross
sectional survey which comprise of sample of size 19862 collected
randomly all over Pakistan. The survey uses three separate
questionnaires for children, adult male and adults female
simultaneously. It gives detailed Information on several health profile
of individual and provides a base for the analysis of determinations of
health especially education.
In the NHSP (1990-94), the respondent's health is measured in
term of self- reported health (SRH) which is measured on ordinal scale
having five categories as excellent, very good, good, fair, and poor.
The other variables involved in this study are education level,
occupation, social status, age, gender, marital status, residence and
province.
SRH is subjective in the medical sense of being a state perceptible
to the individual and not to those who observe or examine the
individual. It is not subjective in the psychological sense of being
moodily introspective or illusory. It reports something real, but
directly observable only by the individual reporting [Mirowsky (2003)].
In the last three decades, self-reported health has been used
increasingly as a measure in the psychological and gerontological areas,
as well as in epidemiological surveys [Lima-Costa and Fernanda Uchao
(2004)]. Self-reported health shows good reliability and its validity is
equivalent to that of other more complex measures of health status
[Idler and Benyamini (1997)]. BMI is a reliable indicator of total body
fat, which is related to the risk of diseases and death. It's a
useful, indirect measure of body composition because it correlates
highly with body fat in most people. Weight in kilograms is divided by
height in meters squared (kg/m2).
Table 1 gives the description of variables used in this study. The
across the sample distribution of the variables used is given in Table
la which is supported by a detailed descriptive analysis of these
variables in Section 4.
It seems that now-a-days, people are better aware off about their
health and due to several possible factors majority groups and
segmentations of the society are believed to be in weaker conditions. As
evident from the whole sample being surveyed, as shown in Figure 2, that
most of the people report their health as 'fair' of
'poor'. About 33 percent people report their health as
'good' and only 6 percent (approx) have very good or
'excellent' health. This overall distribution highlights the
poor condition of health in the society and emphasises the need of
attention in this regard. Hence, it is desired to identify the factors
which lead to low health. Therefore, now discussing it in detail, we
come across some interesting results related to the relation between
education and health.
4. EXPLORATORY ANALYSIS
Studying the relationship between the health and education by
visually describing the data reveals several interesting aspects. It
seems from the graphical display that education, gender play significant
role in determining health. There is large variation in the categories
of SRH.
According to Figure 3, there is quite large variation among the
five categories of SRH over all age groups. However, the category of
'good' and 'fair' remain stable and almost similar
for all age groups. Most of the people from the sample report their
health as 'good' and 'fair'. The share of reporting
health as 'poor' only rises significantly for the people above
60 years of age. From Figure 4, the response of men and women to the SRH
reveals significant the gender differentials. The high bars of
'fair' and 'poor' categories in SRH of females
indicate that women report themselves to be in a worse health condition
as compared to the men. This may be due to several possible factors/
reasons need to be identified with strong evidence.
There is a lot of economic and social disparity in urban and rural
population but here we do not find any significant pattern in
urban-rural area. However, the response pattern is entirely different
across the provinces as evident from Figures 5 and 6. It might be due to
the lack of awareness about health that a large proportion of sample
from Sindh declare them to be in good health. While, major proportion of
people from Punjab identifies them to be in the low categories of
health. However, evidence is needed to justify this argument.
The cross tabulation of SRH with respect to education and gender
reveals some interesting behaviour patterns. It is evident from the
Table 2 that majority of men reported good to excellent health. However,
the majority of females lie in the category of fair and poor.
Now we consider the association between education and some more
objective measures of health; the self-reported number of chronic
conditions. The percentages are shown in Table 3.
It seems prevalence of the disease or conditions is not same for
different levels of education. From the table it can be seen that the
individuals having education up to degree level or above have reported
lesser diseases as compared to the less educated.
On the other hand, we observe that the ratio of reporting different
diseases by individuals having primary or less education is
comparatively lower than the other education levels till matric. This
may be due to the reason that due to less education and lack of
awareness they fail to understand or realise the health problems they
face. Some other reasons e.g. financial factors may also play their
underlying role here. In general educated persons report the chronic
conditions less frequently than less educated persons.
4.1. Comparison of Different Patterns of Self-reported Health by
Education and Social Status
Now in order to study the effect of education on health status, we
make comparison by analysing the pattern of SRH across educational
levels. But as noted earlier the response behaviour of men and women are
entirely different suggesting that the gender differentials impart a
significant effect on health. On comparing the patterns of SRH at
different health status, we observe that with the increase in education
women are getting more awareness about health as indicated from the
decreasing percentage of females under poor social status reporting to
be in very good health. However, the pattern is not similar in case of
men. The pattern is clear from the Figures 8 and 9 respectively.
In Figures 10 and 11 we find that as the level of education
increases especially above middle there is a very sharp decline in
reporting poor health, this decline is steeper in females than in men.
However, the increase in reporting very good health is quite gradual for
women but is very sharp for men having matric or higher education.
Even in case of declaring their health as good improves with the
increase in education. In Figures 10 and 11, on comparing the categories
of good vs. poor in SRH, we find very significant role of education in
women indicated by sharp increase in good health supported by a sharp
decline in reporting poor health among females having middle or higher
education. However, the behaviour in men here remains quite stable and
the role of education generating awareness etc. is not evident here.
Similarly, reporting health as fair increases with the increase in
education among female as evident from Figure 12 and Figure 13. But
remarkably, it declines among men. Hence, we may conclude that education
plays a significant role in generating awareness about health.
4.2. Self-reported Health by Body Mass Index
Now the percentage distribution of the respondents for SRH by body
mass index is given as under:
Figure 14 shows the percentage distribution of SRH for each level
of BMI. The first thing we observe that though response to an
"excellent" health status is at minimal level i.e. 0.3
percent, 0.7 percent and 1.1 percent, but it exist from under weight to
over weight respectively. At the same time among obese nobody reports
the state of excellent health. Further it seems that there is a gradual
increase in the percentage response in favour of "very good"
health status. But for the obese it again declines. The overweight
category shows the maximum response of the individuals (39.1 percent),
of having "good" health. Again we see that this response
increases from underweight up to overweight but again for obese it
decrease. So, without the loss of generality, we can conclude here that
on the basis of BMI vs. SRH, we can say that reporting different
categories of health is close to reality and the general awareness about
health has improved. The interesting point here is that people do not
consider obesity a disease as evident by the highest bar of
'fair' health at the sate of obese under BMI.
4.3. Body Mass Index by Education
Now we see the distribution of BMI for only two groups, one is
highly educated in our sample and other is poorly educated.
From the Figure 15 it is clear that usually less educated appears
to report high level of underweight conditions as compared to more
educated people. Moreover they are also more obese than the educated
persons. About 60 percent of the highly educated persons ate healthy.
Ignoring some fluctuations these patterns suggest almost negative
relationship exist between SRH and BMI, and in BMI and education i.e.
with the increase in education people become more aware of their health
and hence, we can observe significant change in reporting BMI and health
status.
Hence, not only has the general awareness about health had improved
enabling them report much better about their health but it also has
improved the health conditions as indicated by BMI, that educated people
are more concerned about their health and physic.
4.4. Health Care Utilisation and Education
We also investigate whether there is any link between health care
behaviours and education or not. So for this purpose we make comparison
between literates and illiterates. We also illustrate in Figure 16 the
behaviour of individuals with in the literato category only. This is
based on the response of the persons when they are asked at the time of
interview that whether they seek any medical care in the last two weeks
before the survey.
From Figure 17, we see that with the increase in level of
education, the frequency of seeking any medical care decreases. 29
percent people with education level less than primary seek the medical
care, the percentage reduces gradually with an increase in education
level and 20 percent with education degree and above seek the medical
care.
5. EMPIRICAL MODEL AND ESTIMATION
After having descriptive analysis we use ordinal logistic
regression models in order to assess the impact of several variables on
SRH. We assume that the latent health variable is measured by education,
(1) objective health (2) and by some other individual characteristics in
the following way:
[H.sup.*] = [[beta].sub.0] + [H.sup.0] [[beta].sub.1] +
X[[beta].sub.2] + [epsilon]
Where [beta]'s are the vectors of the coefficients and C is
random term capturing unmeasured and immeasurable effects on the true
health status. And X contains the variables i.e. education, occupation
or personal characteristic. The subjective health status is taken as
dependent variable in the model. As the response variable is measured on
ordinal scale, so it is going to be treated as ordinal under the
assumption that the levels of SRH have natural ordering ranges from
'excellent' to 'poor', but the distance between
adjacent levels are unknown. So for this purpose we have again recoded
the data in the following way; "5 excellent, 4 good, 3 fair, 2 poor
and 1 very poor" for response variable.
5.1. Estimation Results
Specifically, our objective is to estimate the effect of education
on health, holding other factors constant. Therefore, taking a closer
look at the possible factors and identifying whether they play any
significant role in determining the health status. Table 4 presents the
results from the estimation of model I to III, model I contains
schooling variables, age and gender. Beside education at all levels,
gender and age also play a significant role. However, it is interesting
to note that the only age group significant here is of 20-29 years old
respondents.
Model II contains education, employment level, and social status
indicating that the employment level and the socio-economic status play
a significant role in determining the subjective health status. Model
III is an extension of the model II by including additional individual
characteristics i.e. age, gender, marital status, residence type and
province. Although controlling for these variables does not alter the
significance of education in determining the self-reported health (SRH)
but nevertheless the significance of employment status and
socio-economic status has changed. Purpose of adding more socio-
economic variables in the model II and model III is to make sure that
our model should not be misperceived. The marital status and residence
type are statistically significant.
Table 5 presents the results from the estimations of model IV where
education level and past health are included as regressors. Model IV
excludes the insignificant variables and include the objective health
measures i.e. disease and conditions, to study their role in determining
the self-reported health status. All diseases included as the objective
measure of health are significant except the knee pain. It is
interesting to note that including the objective health status measure
(all diseases) has not hampered the significance of the levels of
employment. In model IV the various forms of employment status except
the disability are significantly affecting the self-reported health
measure. However, the Pseudo [R.sup.2] is now reduced. In each of the
above considered models we see that by adding more variables the effect
of education on SRH remains significant.
Model V contains all socio-economic, past health and education
variables. We note that Pseudo [R.sup.2] improves in model V as compared
to other four possible models. Log likelihood and chi square also
support this model over other models. Table 6 presents the results of
model V from the estimations of an extensive model containing a large
set of potential variables. It can be observed that significance of the
category having less than primary education is almost on border line but
rests of the categories are highly significant. As the subjective health
variable runs from excellent to poor, a negative sign of the estimate of
the coefficients of the explanatory variables indicates that a decrease
in the level of the variable is associated with a decrease in the
quality of health.
6. CONCLUSION
The existing literature documents extensively on the existence of
the educational differences in health. The need for investigating
whether any causal component lies in the observed relationship between
health and education has been emphasised since long. Often reported
literature largely reflects a causal effect of schooling and education
on health. By analysing the responses to self-reported health, we get a
body of empirical evidence that a variety of socioeconomic and
socio-demographic characteristics lead to perception of health, among
individuals, in a varied manner. Among several socioeconomic variables,
schooling, gender, occupation, economic status and provinces are the
significant determinants of self-reported health. However, schooling
(education) seems to have the most significant impact on health status.
Moreover, the association both between health and education is not very
sensitive to either including or excluding the other variables. As the
level of education increases the heath of the individuals seems to be
affected in a positive way. It is likely that these health differences
are the result of the differences in behaviour across education groups.
An exogenous increase in education causes better health among
individuals. Hence, those with more years of education can take care of
the risks factors leading to health. Thus, health policy researchers and
analysts should emphasis that health and education represent a mutual
approach in improving population health.
REFERENCES
Arendt, J. N. (2001) Education Effects on Health: Causal or from
Unobserved Components? A Panel Data Analysis with Endogenous Education.
Submitted for publication at Institute of Economics, University of
Copenhagen. Available on
www.econ.ku.dk/okojn/Public/Research/jarendt.pdf
Arendt, J. N. (2005) Does Education Cause Better Health? A Panel
Data Analysis using School Reforms for Identification. Economics of
Education Review 24, 149-160.
Arendt, J. N. (2008) In Sickness and In Health-Till Education Do Us
Part: Education Effects on Hospitalisation. Economics of Education
Review 27, 61-172.
Case, A. (2001) The Primacy of Education. Research Program in
Development Studies, Princeton University. (Working Paper No. 203.)
Christopher, M. and N. Sidhu (2005) Schooling, Cognitive Ability
and the Health. Health Economics. John Wiley & Sons, Ltd. 14:10,
1019-1034.
Cutler, D. and E. Richardson (1998) The Value of Health: 1970-90.
American Economic Review 88, 97-100.
Cutler, D., A. Deaton, and A. M. Lleras (2005) The Determinants of
Mortality. Journal of Economic Perspective 20, 1-50.
Cutter, D. and A. M. Lleras (2007) Education and Health: Theories
and Evidences. (NBER Working Paper No. 12352). available on
http\\:www.npc.umich.edu
Fuchs, V. R. (2004) Reflections on Socio-economic Correlates of
Health. Journal of Health Economics 23, 653-661.
Groot, W. and H. Brink (2007) The Health Effects of Education.
Economics of Education Review 26,186-200.
Grossman, M. (1972) On the Concept of Health Capital and the Demand
for Health. Journal of Political Economy 80, 223-255.
Grossman, M. (1975) The Correlation between Health and Education.
(NBER Working Paper No. 22).
Hartong, J. and H. Osterbeek (1998) Health, Wealth and Happiness:
Why Pursue a High Education. Economics of Education Review 17,245-256
Idler EL, Benyamini Y. (1999) Self-Rated Health and Mortality: A
Review of Twenty Seven Community Studies. Journal of Health and Social
Behaviour 38, 21-37.
Lima-coasta, M. F. and Fernanda Uchao (2004) The Structure of Self
Rated Health among Older Adults: The Bambui Health and Aging Study
(BHAS). Rey. Saude Publica 38:6, Sao Paulo.
Mirowsky, J. and C. E. Ross (2003) Education, Social Status, and
Health. Published by Aldine de Gruyter. 242 pp.
Pakistan, Government of (1998) National Health Survey of Pakistan
1990-1994. Islamabad: Pakistan Medical Research Council.
Shumueli, A. (2003) Socio Economic and Demographic Variation in
Health and Its Measures, The Issue of Reporting Heterogeneity. Social
Science and Medicine 57, 125-134.
Silles, M. A. (2009) The Causal Effect of Education on Health:
Evidence from United Kingdom. Economics of Education Review 28, 122-128.
Syed, H. R., et al. (2006) Inequalities in Health: A Comparative
Study between Ethnic Norwegians and Pakistanis in Oslo, Norway.
International Journal of Equity Health. 5:7.
The World Health Report (2004)
http://www.who.int/whr/2004/en/report04_en.pdf
Wilkinson, R. and M. Marmot (eds.) (2003) Social Determinants of
Health: The Solid Facts. 2nd edition/edited for World Health
Organisation 2003 published by International Centre for Health and
Society.
Comments
The paper analyses the socio-economic determinants of health with
special focus on education. There is ample evidence to support the
notion that educational level and health status are positively
correlated and the findings of the study endorse and reinforce this
viewpoint. To begin with, the authors give a brief account of the
literature explaining that educational differences have great impact on
health status and describe that education has a direct effect on health
but there are no well stated causal mechanisms to explain the
relationship.
The authors have used self reported health (SRH) as an indicator of
health status measured on ordinal scale with five categories reflecting
the respondents' perceptions of health status and not of those who
observe or examine the individual. Using data from National Health
Survey of Pakistan (NHSP, 1990-94), exploratory data analysis and
ordinal logistic regression models are used to assess relationship
between health, education and other socio-economic variables.
I would like to appreciate the authors' effort to undertake a
comprehensive analysis of data on pattern of SRH and relate them to
educational differences. The results, however are questionable as
individuals with less than primary education have reported lower
prevalence of diseases than those with middle, matric and degree levels
of education--a finding contrary to our expectation which needs further
exploration.
My comments pertain to three major areas--the methodology; the data
set used, and the results and analysis--which all have limitations and
interpretation biases. First the methodology used specifies
self-reported health (SRH) as dependent variable ordered in five
response categories as excellent, very good, good, fair, and poor. One
major limitation to use ordinal logistic regression is the difficulty to
quantify and interpret the distance between categories of responses
which in turn are subjective perceptions of respondents about their
health status. The data set used is the National Health Survey of
1990-94 which is about 15-20 years old. Since that time, tremendous
changes in health related perceptions and disease patterns have
occurred, and the results may not be reflective of the current situation
of the relationship studied.
Looking at the characteristics of respondents (as shown in Table
la), it appears that the sampled population has a young age structure
(70 percent between ages 20-39 years), of which 70 percent are males,
has education mostly between primary and matric levels, and the majority
has reported health status between good and fair categories, thereby
reflecting the biases of the self reported health status.
As for the results and its analysis, the education categories used
are not precise but indicate a range such as primary <middle; middle
< matric; matric < degree--etc. It would be more appropriate to
make more precise categories of education since education as predictor
of health is the main focus of the study.
Looking at the ordered logit estimates in Table 4, there are some
contradictory results that need further explanation. For example, among
age categories, the coefficient is significant only for age group of
20-29 years and has a negative sign, implying that younger people have
lower self-reported health status than the older age groups- -a finding
contrary to one's expectations. Similar is the situation with male
and urban residence indicating lower self-reported health than the
corresponding category in comparison which needs further investigation.
Overall, the results of the study reinforce the fact that increase
in level of education is positively related to health status of
individuals, but the analysis raises questions about the usefulness and
efficacy of using self-reported health perceptions as best indicators of
health status. The question needs to be addressed for further research
with a more recent and precise data set on health related matters.
Naushin Mahmood
Pakistan Institute of Development Economics,
Islamabad.
(1) Generally the education is included in the model by years of
schooling but for this study we use it as dummy variable.
(2) Objective health is also considered as past health.
Zahid Asghar <z_asgar@yahoo.com> is Assistant Professor at
Statistics Department, Quaid-i-Azam University, Islamabad. Nazia Attique
<naziattique@gmail.com> is Research Executive, OASIS Insights
Islamabad. Amena Urooj <dayashka@gmail.com> is Assistant Professor
at the Pakistan Institute of Development Economics, Islamabad.
Table 1
The Profile of the Study Population (Based oft NHSP 1990-94)
Variables Description
Health Variables Subjective Health We take SRH graded as
(1:excellent, 2:very good, 3:good,
4:fair, 5:poor)
Objective Health Arthritis, asthma,
diabetes, hemorrhoids, tuberculosis,
heart disease, pain in back, pain in
knees, vision problem and dental problem
Schooling Variables For schooling five different levels of
education are taken.
i. Less Than Primary
ii. Primary But Less Than Middle
iii. Middle But Less Than Matric
iv. Matric But Less Than Degree
v. Degree and Above
Social Status It is divided into only two categories
high and low
Occupation It includes seven categories employed,
self-employed unemployed, work in home,
student, disabled and other
Other Background Variables
Age Age is defined from 20 up to 90 years
Gender Gender is classified as males and
females
Marital Status There are three categories, single,
married and other
Residence Type It consists of two categories urban and
rural
Provinces Punjab Sindh Khyber Paktunkhwa and
Balochistan
Table 1a
Socio-demographic Characteristics of the Sample Population with
Descriptive Statistics, NHSP (1990-94)
Variables Percentage
Self-reported Health
Excellent 0.5
Very Good 9.2
Good 37.6
Fair 35.2
Poor 17.5
Social Status
High 57.2
Low 42.8
Occupation
Employed 24.7
Self-employed 28.5
Unemployed 1.7
Student 4.5
Work in Home/ Sick 22.7
Disability 3
Other 14.9
Education Level
< primary 7.6
Primary but < middle 31.4
Middle but < matric 20.5
Matric but < degree 31.1
Degree and above 9.4
Gender
Male 70.6
Female 29.4
Residence
Urban 56.5
Rural 43.5
Provinces
Punjab 50.9
Sindh 22.8
Khyber Pakhtunkhwa 19.9
Balochistan 6.5
Age Groups
20-29 years 42.7
35-39 years 27.2
40-49 years 15.4
50-59 years 9.4
60 years and above 5.3
Marital Status
Single 25
Married 72
Others 3
Table 2
Subjective Quality of Health by Education and Gender
Self-reported Health
Education Excellent (%) Very Good (%) Good (%)
Women
< Primary -- -- 27.7
Primary but < Middle -- 1.9 14.6
Middle but< Matric 0.9 3.4 21.4
Matric but <degree -- 9.8 29.9
Degree and Above 4.9 12.2 34.1
Total 0.5 5.1 22.8
Men
<Primary -- 2.8 52.8
Primary but <Middle -- 8.5 38
Middle but <Matric 0.3 5.3 43.9
Matric but <degree 1.6 12.8 41.7
Degree and Above 1.2 27 49.1
Total 0.7 10.9 42.8
Education Fair (%) Poor (%)
Women
< Primary 34 38.3
Primary but < Middle 36.8 46.7
Middle but< Matric 42.7 31.6
Matric but <degree 37.4 23
Degree and Above 43.9 4.9
Total 38.4 33.2
Men
<Primary 31.1 13.2
Primary but <Middle 39.2 14.3
Middle but <Matric 38.6 11.9
Matric but <degree 34.3 9.6
Degree and Above 17.2 5.5
Total 34.3 11.3
Note: the estimates are based on using study sample (1).
Table 3
Objective of Health Status by Education
Education (Values are in %)
Objective < Primary> Primary but Middle but
Health < Middle < Matric
Arthritis 9.9 45.6 19.8
Asthma 9.3 33.3 16.0
Diabetes 8.2 26.5 20.4
Hemorrhoids 5.5 38.4 23.2
Tuberculosis 18.8 28.1 15.6
Heart Disease 7.3 29.3 19.5
Pain in Back 10.2 41.9 18.6
Pain in Knees 9.8 42.9 20.9
Vision Problem 6.0 35.8 22.0
Dental Problem 9.0 33.6 20.1
Education (Values are in %)
Objective Matric but Degree and
Health < degree Above
Arthritis 19.8 4.8
Asthma 32.0 9.3
Diabetes 36.7 8.2
Hemorrhoids 24.4 8.5
Tuberculosis 37.5 0.0
Heart Disease 29.3 14.6
Pain in Back 23.8 5.4
Pain in Knees 21.2 5.3
Vision Problem 26.8 9.4
Dental Problem 29.4 7.8
Table 4
Ordered Logit Estimates for Health Status
Explanatory Variables Model I
Coefficients p-value
(Std. error)
< primary -1.275 .000
(.213)
Primary but < middle -1.487
.000
(.166)
Middle but < matric -1.330
.000
(.174)
Matric but < degree -.933
(.164) .000
Employed
Self-employed
Unemployed
Student
Work in Home/ Sick
Disability
Social-economic Status
Higher
Urban
Punjab
Sindh
Khyber Pakhtunkhwa
Male -1.301 0.000
(0.1)
Single
Married
Age 20--29 Years -.557
(.200) .005
Age 35--39 Years -.247
(.205) .227
Age 40--49 Years -.014
(.216) .947
Age 50--59 Years .163
(.233) .483
-2 Log Likelihood 580.095
Chi Square 287.654
Sig. 0.000
Pseudo R-square 0.154
Explanatory Variables Model II
Coefficients p-value
(Std. error)
< primary -1.142 .000
(.219)
Primary but < middle -1.331
.000
(.173)
Middle but < matric -1.207
.000
(.178)
Matric but < degree
(.1865) .000
Employed -.534
.000
(.143)
Self-employed -.525
(.139) .000
Unemployed -.920
(.346) .008
Student -.581
.013
(.234)
Work in Home/ Sick -1.729
(.150) .000
Disability -.770 .004
(.270) .019
Social-economic Status .211
Higher (.090)
Urban
Punjab
Sindh
Khyber Pakhtunkhwa
Male
Single
Married
Age 20--29 Years
Age 35--39 Years
Age 40--49 Years
Age 50--59 Years
-2 Log Likelihood 706.581
Chi Square 268.913
Sig. 0.000
Pseudo R-square 0.144
Explanatory Variables Model III
Coefficients p-value
(Std. error)
< primary -.868 .000
(.231)
Primary but < middle .925
.000
(.183)
Middle but < matric .671
.000
(.187)
Matric but < degree -.483
(.172) .005
Employed .035
.832
(.164)
Self-employed .016
(.164) .924
Unemployed -.790
(.366) .031
Student .209
.404
(.250)
Work in Home/ Sick .005
(.232) .983
Disability -.118 .725
(.335) .036
Social-economic Status .209
Higher (.100)
Urban -.038 .702
(.100) .000
Punjab -1.062
(.189)
Sindh 1.227
(201) .000
Khyber Pakhtunkhwa 1.526
.000
(.209)
Male -1.501 .000
(186)
Single .231
(.305) .450
Married .285
.301
(.276)
Age 20--29 Years .589
(.256) .021
Age 35--39 Years .234
(.249) .348
Age 40--49 Years -.146
(.258) .571
Age 50--59 Years -.178
(.264) .499
-2 Log Likelihood 2775.392
Chi Square 866.303
Sig. 0.000
Pseudo R-square 0.4
Table 5
Ordered Logit Estimates for Health Status
Model-IV
Explanatory Variables Coefficients S.E p-value
< primary -0.952 0.224 0
Primary but <middle -1.141 0.176 0
Middle but < matric -1.142 0.18 0
Matric but < degree -0.826 0.167 0
Employed -0.296 0.146 0.043
Self-employed -0.336 0.142 0.018
Unemployed -0.958 0.352 0.007
Student -0.524 0.238 0.028
Economically Inactive -1.476 0.158 0
Disability -0.233 0.281 0.406
SES Higher 0.206 0.092 0.026
Urban
Punjab
Sindh
Khyber Pakhtunkhwa
Male
Single
Married
Age 20--29 Years
Age 35--39 Years
Age 40--49 Years
Age 50--59 Years
Disease and Conditions
Arthritis -0.895 0.149 0
Asthma -1.013 0.24 0
Diabetes -0.82 0.286 0
Hemorrhoids -0.543 0.17 0.001
Tuberculosis -0.946 0.366 0.01
Heart Disease -0.562 0.312 0.072
Pain in back -0.508 0.11 0
Pain in knees 0.165 0.148 0.267
Vision Problem -0.836 0.111 0
Dental Problem -0.554 0.096 0
-2Log Likelihood 2624.119
Chi Square 504.825
Sig. 0
Pseudo R-sq 0.271
Table 6
Ordered Logit Estimates for Health Status
Explanatory Variables Coefficients p-value
(Std. error)
Urban -0.044 0.670
(0.103)
Punjab -1.249 0.000
(0.196)
Sindh 976 0.000
(0.208)
Khyber Pakhtunkhwa 1.114 0.000
(0.217)
Male 1.387 0.000
(0.195)
Single 256 0.416
(0.315)
Married 0.422 0.139
(0.285)
Age 20-29 years 305 0.250
(0.265)
Age 35-39 years 0.032 0.902
(0.258)
Age 40-49 years -0.215 0.416
(0.265)
Age 50-59 years -0.155 0.565
(0.270)
SES higher -0.200 0.048
(0.101)
Less than primary 751 0.001
(0.235)
Primary but less than 0.805 0.000
middle (0.185)
Middle but less than matric 640 0.001
(0.189)
Matric but less than degree 452 0.009
(0.173)
-2Log Likelihood 3501.102
P value 0.000
Employed +.035 0.835
(0.166)
Self Employed -0.025 0.883
(0.166)
Unemployed -0.862 0.020
(0.372)
Student 168 0.507
(0.253)
work in home/ Sick -0.009 0.968
(0.237)
Disabled 70 0.838
(0.343)
Health Conditions
Arthritis -0.775 0.000
(0.155)
Asthma -1.154 0.000
(0.251)
Diabetes -0.725 0.017
(0.304)
Hemorrhoids -0.358 0.045
(0.178)
Tuberculosis -1.187 0.002
(0.388)
Heart disease -0.674 0.039
(0.326)
Pain back -0.438 0.000
(0.116)
Pain knee +.067 0.665
(0.155)
Vision problem -0.554 0.000
(0.118)
dental problem -0.314 0.002
(0.101)
Chi square 1030.974
Pseudo R sq 0.457
Fig. 2. Percent of Persons Reporting SRH
Excellent 0.46%
Very Good 5.52%
Good 33.42%
Fair 38.94%
Poor 21.66%
Note: Table made from pie chart.
Fig. 3 Percent of Persons Reporting SRH by Gender
Age Groups Excellent Very Good Good Fair Poor
20-29 years 0.8% 11.6% 37.5% 33.9% 16.2%
30-39 years 0.0% 7.9% 37.3% 38.0% 16.8%
40-49 years 1.0% 6.4% 36.4% 37.4% 18.8%
50-59 years 1.1% 8.0% 35.6% 36.2% 19.1%
> 60 years 0.8% 6.7% 35.0% 31.7% 25.8%
Note: Table made from bar graph.
Fig. 4 Percent of Persons Reporting SRH by Age
Gender Excellent Very Good Good Fair Poor
Male 0.7% 10.8% 42.7% 34.5% 11.3%
Female 0.5% 5.1% 22.8% 38.4% 33.2%
Note: Table made from bar graph.
Fig. 5. Percent of Persons Reporting SRH by Residence
Area of Residence Excellent Very Good Good Fair Poor
Urban 0.8% 10.4% 36.3% 35.0% 17.6%
Rural 0.4% 7.6% 37.8% 36.5% 17.6%
Note: Table made from bar graph.
Fig. 6. Percent of Persons Reporting SRH by Province
Province Excellent Very Good Good Fair Poor
Punjab 0.2% 2.0% 20.5% 49.2% 28.2%
Sindh 1.1% 14.3% 61.1% 17.9% 5.6%
NWFP 1.6% 23.7% 50.0% 19.5% 5.2%
Balochistan 5.9% 44.9% 36.8% 2.5%
Note: Table made from bar graph.
Fig. 8. Pattern of SRH Very Good vs. Poor for Women
Level of Education Very Good Poor
Less Than Primary 0 38.3
Primary But Less Than Middle 1.9 46.7
Middle But Less Than Matric 3.4 31.6
Matric But Less Than degree 9.8 23
Degree and Above 12.2 4.9
Note: Table made from line graph.
Fig. 9. Pattern of SRH Very Good vs. Poor for Men
Level of Education Very Good Poor
Less Than Primary 2.8 13.2
Primary But Less Than Middle 8.5 14.3
Middle But Less Than Matric 5.3 11.9
Matric But Less Than degree 12.8 9.6
Degree and Above 27 5.5
Note: Table made from line graph.
Fig. 10. Pattern of SRH Good vs. Poor for Women
Level of Education Good Poor
Less Than Primary 27.7 38.3
Primary But Less Than Middle 14.6 46.7
Middle But Less Than Matric 21.4 31.6
Matric But Less Than degree 29.9 23
Degree and Above 34.1 4.9
Note: Table made from line graph.
Fig. 11. Pattern of SRH Good vs. Poor for Men
Level of Education Good Poor
Less Than Primary 52.8 13.2
Primary But Less Than Middle 39 14.3
Middle But Less Than Matric 43.9 11.9
Matric But Less Than degree 41.7 9.6
Degree and Above 49.1 5.5
Note: Table made from line graph.
Fig. 12. Pattern of SRH Fair vs. Poor for Women
Level of Education Fair Poor
Less Than Primary 34 38.3
Primary But Less Than Middle 36.8 46.7
Middle But Less Than Matric 42.7 31.6
Matric But Less Than degree 37.4 23
Degree and Above 43.9 4.9
Note: Table made from line graph.
Fig. 13. Pattern of SRH Fair vs. Poor for Men
Level of Education Fair Poor
Less Than Primary 31.1 19.2
Primary But Less Than Middle 39.2 14.3
Middle But Less Than Matric 38.6 11.9
Matric But Less Than degree 34.3 9.6
Degree and Above 17.2 5.5
Note: Table made from line graph.
Fig. 14. SRH Reporting of Persons Belonging to Four Categories of BMI
BMI Excellent Very Good Good Fair Poor
Underweight 0.3 5.5 33.1 37.5 23.7
Normal 0.7 9.7 38.4 34.9 16.2
Overweight 1.1 12.1 39.1 32.0 15.7
Obese 9.6 34.9 42.2 13.3
Note: Table made from bar graph.
Fig. 15. Percent of Persons Reporting on BMI Scale by Education
BMI Less Than Primary Degree and Above
Underweight 21.60 12.1
Normal 54.9 59.3
Overweight 17.6 25.6
Obese 5.9 3
Note: Table made from bar graph.
Fig. 16. Percent of Persons Seeking Medical Care by Levels
of Education
Percent
Illiterate 23.30%
Literate 24.80%
Note: Table made from bar graph.
Fig. 17. Percent of Persons Seeking Medical Care by Literacy
Levels of Education Percent
Degree and Above 20.10
Matric But Less Than degree 23.20
Middle But Less Than Matric 26.90
Primary But Less Than Middle 26.20
Less Than Primary 29.10
Note: Table made from bar graph.
