how do persons with disabilities earn a living?: findings from the field survey conducted in
Post on 11-Sep-2021
2 Views
Preview:
TRANSCRIPT
Very Preliminary
How Do Persons with Disabilities Earn a Living?: Findings from the Field
Survey Conducted in Metro Manila, the Philippines*
Jose Ramon Albert†, Soya Mori‡, Celia Reyes†, Aubrey Tabuga†, Tatsufumi Yamagata‡, and
Josef Yap†
April 2009
Abstract
How do persons with disabilities (PWDs) earn a living? This question is more critical in developing countries from the view point of poverty reduction of PWDs. This paper investigates economic activities of PWDs in the Philippines where disability‐related legislation is relatively progressive among developing countries. A field survey was conducted in cooperation with Disability People’s Organizations (DPOs) with a tailor‐made questionnaire in four representative cities in Metro Manila in August 2008. The level of income and education of PWDs and determinants of them are examined with the Mincer regression. Main conclusions of this study are: (1) both the incidence and depth of poverty are greater among sample PWDs than that of total population in Metro Manila; (2) strong association between education and income and solid heredity of low level of education to a PWD from her/his father; (3) female PWDs are likely to be educated less than male PWDs after taking the endogeneity between income and education into account and controlling individual, parents’ and environmental characteristics.
Keywords: Disability; Philippines; Poverty; Mincer regression; Rate of return on education; Gender. JEL classifications: D31, I12, I20, I31, J16, J31, O53
* The authors acknowledge Marites Racquel Corpuz, Abner Manlapaz, Lauro Purcil for overall consultation, and Ronina D. Asis, Junalyn T. Bayona, Ma. Blessila G. Datu, Kathrina G. Gonzales, Christian D. Mina, for tireless assistance. A comprehensive report of the field survey is published as Mori, Reyes and Yamagata eds. [2009]. † Philippine Institute for Development Studies. ‡ Institute of Developing Economies, Japan External Trade Organization. E‐mail: soya_mori@ide.go.jp; tatsufumi_yamagata@ide.go.jp.
1
1. Introduction
Disability has received due treatment in the context of international development only recently.
It was regarded as a part of diseases and injuries for a long time, so that data on disability was not
actively collected, and this inactiveness is applicable more to developing countries than developed
countries. In these days some international organizations which constitute the “Washington City Group”
put forth intensive efforts to determine how disability should be identified as a part of the questionnaire
for population census (Altman [2006] among others). However, even they pay attention mostly to the
identification of disability, and socio‐economic life of persons with disabilities (PWDs) was rarely
featured in field surveys on disability.1
This study focuses on livelihood of PWDs in the Philippines which is a country enacting
disability‐related laws widely among developing countries. The Institute of Developing Economies (IDE),
Japan, and the Philippine Institute for Development Studies (PIDS) conducted a field survey in four cities
in Metro Manila, in August 2008. Yap et al. [2009] covers basic facts derived from the data. In this paper,
how PWDs make a living is fully examined with the data.
Main conclusions are threefold: (1) both the incidence and depth of poverty is higher among
sample PWDs than that of total population in Metro Manila; (2) there are strong association between
education and income and solid heredity of low level of education to a PWD from her/his father; (3)
female PWDs are likely to earn and to be educated less than male PWDs after taking the endogeneity
between income and educational into account and controlling individual, parents’ and environmental
characteristics.
The rest of paper is organized as follows. The next section briefs the sampling strategy. Section
3 reviews education and income of sample PWDs. Features concerning education is scrutinized because
it is a prospective determinant of earnings. The incidence and depth of poverty among sample PWDs is
investigated in this section, too. Section 4 demonstrates regression exercises to explore deep causes and
mechanisms towards the poverty. Determinants of income and education of PWDs are sought. The final
section provides concluding remarks.
2. Sampling method
1 An exception is a set of field surveys conducted by Washington University in St. Luis and the National Institute on Disability and Rehabilitation Research in the United States, which are known by the names of the “Community Receptivity Project” and “Assistive Technology in the Community Assessment.”
2
The Philippines is known as a developing country where laws and institutions have been
better and more comprehensively formulated than others. The first law for PWDs dates back to Republic
Act 1179 entitled “An Act to Provide for the Promotion of Vocational Rehabilitation of the Blind and
Other Handicapped Persons and Their Return to Civil Employment,” enacted in 1954.2 Since then seven
Republic Acts have been promulgated. The culmination of them is Republic Act 7277 named “An Act
Providing for the Rehabilitation, Self‐Development and Self‐Reliance of Disabled Persons and their
Integration in to the Mainstream of Society and for Other Purposes,” which is more widely known as
“Philippine Magna Carta for Disabled Persons” enacted in 1992. The Magna Carta declares the rights of
persons with disabilities in terms of employment, education, health, social services, etc. It was amended
further in 2007 so that the Philippine society fully integrates PWDs.
Metro Manila is the capital of the Philippine, which consists of seventeen cities known as
“Local Government Units” (LGUs). Needless to say, Metro Manila is a megacity to and from which a
large number of people migrate both domestically and internationally. In this sense, Metro Manila is a
representing place of the Philippines even though it never be a typical place to live in the country. For the
first approach to this issue, the capital was selected for this intensive study.
Among the seventeen LGUs, Makati, Pasay, Quezon and Valenzuela were selected for this
survey (Figure 1). These four cities represent certain features of Metro Manila from various viewpoints.
Makati contains the most luxurious quarter in the Philippines, so that its fiscal affluence is outstanding
among the seventeen LGUs (see Figure 2). PWDs living in ordinary residential districts of Makati benefit
from the great amount of revenue. Quezon is a great city which makes up 20 percent in Metro Manila in
terms of both area and population. The city is known as an educational district where some famous
universities are located. Pasay and Valenzuela are as poor as other LGUs of Metro Manila. While Pasay is
like an adjacent downtown to Makati, Valenzuela is considered to be one of peripheries of Metro Manila.
It is supposed that an intensive study of these four cities sheds light on necessary spots of Metro Manila
to scrutinize the livelihood of PWDs living in Metro Manila.
There are three types of mother lists of PWDs living in Metro Manila. The seemingly most
legitimate list is the one formulated through the process of 2000 Census of Population and Housing (NSO
[2004]). According to this list which was provided by the National Statistics Office (NSO), the numbers of
PWDs by city and disability are those shown in Table 1. In fact, this mother list was not useful for our
2 For more in details, see National Council on Disability Affairs Website at www.ncda.gov.ph.
3
field study because of inaccuracy of information of PWDs in name, disability, age, sex and address. Thus,
we could not but give up relying on this source.3
The second source is member lists of Disabled People Organizations (DPOs) maintained by
each DPO. This list is considered to cover relatively upper class of PWDs more than average, because
highly motivated people are more likely to become members of DPOs.
The third source is PWDs lists maintained by LGUs. Each LGU has and updates the list to
comprehend and to identify who should be beneficiaries of social service for PWDs offered by the LGU.
By construction, this list covers lower segment of PWDs who need social services. Therefore, we used
these lists provided by the four LGUs to identify PWDs to be interviewed.4
There is a unique policy in this survey, which is that PWD enumerators were hired and they
interviewed PWDs with the same sort of disability. This policy was adopted on two grounds. The one is
that PWDs know the situation of peers with the same disability best, so that PWDs may be the most
qualified enumerators. The other is the empowerment of PWDs, both enumerators and interviewees, by
realizing their potential through their mutual interactions5. In order to stick to this policy, the sorts of
disabilities to be focused were limited to mobility6, visual and hearing disabilities. Persons with other
sorts of disabilities such as cognitive, mental, etc. were interviewed only if they have either mobility,
visual or hearing impairment.
The identification of qualified PWDs, whose age falls in the range of working ages, i.e. 15‐70
years old, at their correct addresses was immediately followed by interviews. The target sample size was
set as 360, so that each sort of disability secures 120 samples and that samples are roughly similarly
distributed across LGUs.
The distribution of samples is displayed in Table 2. Sample PWDs categorized under
“multiple” have more than two sorts of disabilities among the three. In total, 403 samples are collected
and there is no extreme concentration of samples by city and disability.
3. Education and economic activities of sample PWDs
3 See Yap et al. [2009], pp. 26-36 for more information. The inaccuracy is alluded by extraordinarily small figures for the number of persons with hearing impairments in Table 1. 4 Precisely speaking, 3.0, 9.9 and 7.7 percents of samples were drawn from the first, second and other sources. For more details, see Table 6 of Yap et al. [2009]. 5 A flip side is that “abled” people are also empowered by learning about PWDs. 6 More precisely, upper-limb impairments were also categorized under “mobility impairments” in this study.
4
The purpose of this study is to explore determinants of income of PWDs. Before proceeding to
empirical analyses, basic features of key variables are reviewed in this section.
Education
Generally speaking, the level of education is a key variable to determine the level of income. It
is meaningful to overview general tendencies of education of sample PWDs.
Table 3 shows distribution of samples by disability and highest educational attainment, while
Table 4 presents the share of samples with each education category to the total number of PWDs by
disability. As seen in Table 4, persons who have never been to school make up 7.9 percent of total
samples. In fact, almost a quarter (24.3 percent) of samples did not finish the primary school. On the
other hand, another quarter (exactly 25.0 percent) have been to a college, university or graduate school.
Thus, there is high disparity in education among PWDs. This disparity is accentuated for people with
visual impairments. That is, while 13.2 percent of them have never been to school, 27.8 percent have been
to a college or a higher educational institution.
Table 5 summarizes the average years of schooling7 by disability and sex. Note that the a
years of schooling is shorter for females irrespective of sorts of disability. The table shows that persons
with mobility impairments go to school the longest while those with hearing do the shortest.
verage
Jobs
Even if a PWD is willing to earn money by herself / himself, if an income‐generating
opportunity is not available, her / his willingness is not to be realized. Table 6 exhibits how much likely
PWDs get income‐generating jobs8. About a half sample PWDs were engaged in any economic activities.
In our data persons with visual impairments are more likely to get a job. Table 7 reveals that 71.5 percent
of them have jobs, which is far higher than the same ratios for those with mobility and hearing
impairments of 44.2 percent and 31.5 percent, respectively. According to a detailed review in this issue,
many of those with visual impairments are massagers (Yap et al. [2009]). Institutional conditions such as
training schools, collective shops, and long‐established reputation seem to let persons with visual
7 The interviewees were asked highest educational degree the interviewee attained. The years of schooling was constructed as follows; (1) Kindergarten/Prep: 1 year; (2) Grade I to V: 3 years; (3) Elementary graduate: 6 years; (4) 1st to 3rd Year High School: 8 years; (5) High School Graduate: 10 years; (6) Vocational school: 10 years; (7) Post-secondary (diploma courses/ certificate): 11 years; (8) College level: 12 years; (9) College or University graduate: 14 years; (10) Master or higher: 15 years. 8 The term of “income generating jobs” includes self-employment throughout this paper.
5
impairments straightforwardly consider the massager a promising occupation and relatively readily get
a job by following predecessors.
Income by Disability
The average annual income of PWDs by source and disability is shown in Table 8. The simple
average of annual income of sample PWDs is 60,173 peso which is roughly equivalent to 1,200 US dollars.
Again, disparities in the average income by disability are remarkable. The average income of persons
with visual impairments is 76,270 peso which is 1.67 and 1.37 times greater than those of persons with
hearing and mobility impairment, respectively. Another important issue is that the amount of private
income transfer is the greatest to persons with hearing impairments whose income is the lowest, while
persons with visual impairments, who are the richest, received private income transfer the least. These
observations imply that private income transfer is made in the direction to offset the income differences.
In other words, the private income transfer conceals real potential to earn money to a certain extent.
The income difference by disability is statistically examined and its results are displayed in
Table 9. The personal income is regressed to dummy variables of impairments. The reference disability is
the mobility impairment, so that a coefficient on a disability dummy incorporates a difference in income
between persons with the particular disability and those with mobility impairments9. The conspicuity of
persons with visual impairments is confirmed with this analysis, too. The estimated coefficient on the
visual impairment dummy is significantly positive of 95 percent significance level if the robust standard
error is invoked. Since the p‐value of the same coefficient with the ordinary standard error is as low as 5.3
percent, the significance of the visual impairment dummy is not affected very much by choice of
standard error.
Poverty Indices
For evaluation of welfare and economic achievements of PWDs, the comparison in poverty
indices between PWDs and non‐PWDs (or just total population) is meaningful.
Figure 3 depicts sample distribution in terms of personal income as well as the poverty line
applied to Metro Manila. The poverty line is 19,345 peso which was constructed by the National Statistics
Office (NSO) for the year 200710. It is apparent at a glance that there are a small number of persons who
have high earnings, which implies high skewness of the distribution.
9 There is only one sample who has a mental impairment. Since the mental impairment dummy sometimes destabilizes estimation results, the mental impairment dummy is suppressed.
6
The second column of Table 10 presents the estimates of standard poverty indices calculated
with our data. The head count ratio turns out to be as high as 40.8 percent11. Let’s take a reference to
evaluate the ratio worked out of our data. A natural reference is the same ratio of Metro Manila derived
from the latest Family Income and Expenditure Survey (FIES) which is the one conducted in 2006. The ratio
of 2006 Survey for Metro Manila was 10.4 percent, which is far lower than 40.8 percent. Even though the
sampling methods, survey periods, survey areas are not exactly the same, the difference between the two
is so great. Thus, the great difference between the two ratio indicates that the ratio of PWDs living below
the poverty line is higher than that of total population in Metro Manila.
In addition to the head count ratio, the poverty gap ratio and the squared poverty gap ratio
were computed with our data12. The poverty gap ratio reflects the depth of poverty, while the squared
poverty gap ratio takes inequality among the poor into account. As is the difference in the head count
ratio, the differences in the poverty gap ratio and the squared poverty gap ratio between our PWD
samples and FIES total population are substantial. These observations suggest that the depth of poverty
and inequality among the poor are greater among PWDs than non‐PWDs. Collateral evidence to the
inequality is that quite a few PWDs living below the poverty line earn nothing. Out of 162 sample PWDs
living below the poverty line, 74 PWDs earn no monetary income even without receiving income
transfer.
There are an aspect which makes the poverty indices to be overestimated and another which
makes the indices to be underestimated. The former is due to the sampling method. Since the
registration to the Local Government Units is for the purpose of receiving social services, rich PWDs may
be less keen to be registered. Thus, our data set may have downward bias in terms of the average income.
The latter aspect contributing to undervaluation of the poverty indices is the possible undervaluation of
poverty line to PWDs. The price and quantity of essential commodities and services to PWDs are
considered to be greater than those to non‐PWDs. If so, the amount of poverty line is not sufficient for
PWDs to entertain the same standard of living as non‐PWDs do. Then, the estimated poverty indices
based on the poverty line computed for non‐PWD have downward bias. In sum, it is indeterminate
which aspect dominates, and we cannot tell our estimates of poverty indices have any inclination to
deviate the population indices.
4. Empirical analyses of determinants of income and education
10 See the following site (http://www.nscb.gov.ph/poverty/2006-2007/pov_th_07.asp). 11 A more precise estimate with region-wise poverty lines is 36.5 percent. See Table 150 of Yap et al.[2009].
7
In order to explore how poor PWDs can get out of poverty, it is necessary to know what are
determinants of their income. A celebrated and tractable method is the Mincer regression (Mincer [1958]),
which regress the logged income13 on years of education and other control variables in order to estimate
the rate of return on education and impacts of the control variables. Since the Mincer regression has been
used in a large number of studies, it is easy to compare estimation results of this study with existing
literature.
In addition, determinants of the years of education is also investigated. While the rate of return
on education is a key interest, the years of education may have bi‐directional causalities with the level of
income. Thus, how to handle the endogeneity is an important concern to estimate the effect of education
on income accurately.
The rest of estimation exercises proceed as follows. First, a simplified benchmark Mincer
equation is estimated. A high rate of return on education and a significant negative coefficient of sex
dummy are tentative findings. Then, the endogeneity of years of education is explored by seeking
exogenous determinants of education, which may be used as instruments to estimate the rate of return
better. All variables used in those exercises are summarized in Table 11.
Benchmark Mincer regression
As a first experiment, the Mincer regression is conducted with fundamental explanatory
variables such as years of schooling, age, age squared, sex dummy (female = 1), marriage dummy14 (the
married =1), and the disability dummies (the reference is mobility disability). The regression equation is
specified as follows:
( ) ( ) ( ) ( ) SexD42
3210 ageageschooling of yearsincomeln βββββ ++++=
uDDDDD OthersalIntellectuHearingVisualMarriage +
+++++ 98765 βββββ . (1)
Since the dependent variables is left‐censored at zero, Tobit regression method is applied as well as OLS.
Table 12 displays estimation results of this benchmark regression. The results are qualitatively
12 For details of these poverty indices, see Deaton [1997] and Foster, Greers and Thorbecke [1984] among others. 13 As indicated in the previous section, 74 sample PWDs do not have any monetary income. Since logarithm is applicable only positive numbers, 1 peso is assigned to the income of the 74 sample PWDs in place of zero. One peso is as small as zero peso as annual income of the Philippines, so that there will not be any discernible effect. 14 The common-law marriage is also counted as a marriage.
8
similar between OLS and Tobit regressions. The coefficient on years of schooling is interpreted as the rate
of return on education. The point estimates are 24.7 percent with OLS and 30.1 percent with Tobit
regression, respectively. These estimates are high as a rate of return on education15. For example, the OLS
estimate implies that one year increase in education raises income by 24.7 percent.
Other significant explanatory variables are age, age squared, sex dummy and visual
impairment dummy. The point estimates of coefficient on age are higher than the estimated coefficients
on years of education, irrespective of estimation method. This coefficient is conventionally interpreted as
the “rate of return on experience in life”. Precisely speaking, effects of age squared must be taken into
account to compute the rate of return on experience. Since the estimates on age squared are significantly
negative, an increment in income according to age decreases as aging.
Another interesting finding is that the coefficient on sex dummy is significantly negative, and
that the magnitude is so great as the absolute values of the estimated coefficients exceed unity for both
OLS and Tobit regression. The significantly negative coefficient implies that the woman earns
significantly lower income than the man with the same years of schooling, age, marital status and
disability. Moreover, the coefficient exceeding unity in the absolute value means that the difference in
income between the woman and the man with the same education, age, marital status and disability is
more than twice16. Thus, female PWDs incur double disadvantages, the one is as a PWD and the other as
a woman.
Finally, the estimates of coefficient on visual impairment dummy are significantly high for both
OLS and Tobit regression. This is consistent with the result of simple comparisons of income by disability
shown in the previous section.
As a whole, while very impressive findings such as the high rate of return on education, high
rate of return to life experience, low income of the woman, and high income of persons with visual
impairments are indicated, the robustness of them is yet to be confirmed. Various empirical trials will be
made in the next subsection to check the robustness.
Experiments of some modifications
15 The 95 percent confidence interval is (14.3%, 35.1%) with OLS and (17.1%, 43.1%) with Tobit regression. The estimates of rate of return on education as high as 20-40 percent were not very unusual among studies for developing countries in the 1960s-70s (Willis [1986: 540-541]). However, the estimates commonly observed in the world in these years (Card [1999: 1834-1854]) as well as those found with data from the Philippines are likely to be around 10 percent (Maluccio [1998], Schady [2003], Yamauchi [2005: 965]). 16 When the coefficient on the sex dummy is one, the conditional male to female income ratio is e1 (=2.718...).
9
In order to elaborate the robustness of the above findings, three ways of experiments are made.
Firstly, income minus private income transfer is used as the proxy for earnings to capture only
autonomous earnings instead of total income. As found in Table 8, private income transfer is offsetting
differences in autonomous income so that it obscures real earning potential of PWDs. Secondly, possibly
omitted variables are introduced as additional explanatory variables to the benchmark set of explanatory
variables. Newly added explanatory variables are area dummies, parents’ attributes and timing of onset
of disability.17 The area dummies are based on which LGU a respondent lives in. Makati is the reference
city. The reason to be adopted as explanatory variables are that residential environments and LGUs’
policies critically affect income generating opportunities of PWDs. Variables used for parents’ attributes
are (1) whether respondent’s father / mother is alive; (2) father’s / mother’s years of schooling; and (3)
whether father / mother has any disabilities or not. Those variables are assumed to represent latent
family‐factors affecting income generating potentials of PWDs. Those born in a rich family probably
have advantages to attain income generating opportunities through reputation of the family, networks
constructed by parents, and learning by seeing parents’ deeds, unless parents died in respondent’s
younger age. The above three sorts of variables might incorporate such family effects.
Thirdly, the endogeneity in the years of schooling is taken into account. There could be a
reverse causality running from income level of a PWD towards the PWD’s level of education if there is a
latent factor such as family effects. For example, a rich family can afford to send its child with disability
to school longer, and at the same time the family offers some advantages in income generation to the
child. And, if the additional explanatory variables do not completely represent this family effects, the
estimated rate of return on education may have an upward bias due to positive correlation between the
family effect in error term and years of education. Therefore, instrumental variables which are exogenous
to current income of PWD and which are highly correlated with PWD’s years of education were sought.
The instruments used in this analysis are: (1) PWD’s years of schooling while father / mother was alive;
(2) its square; (3) difference in age with father / mother; (4) its square; and all explanatory variables used
for the Mincer regression but the years of schooling. The instrument (1) is used as an instrument because
parent’s early death must affect continuation of PWD’s schooling. The justification of the instrument (3) is
that young parents tend to have difficulties to finance schooling of their child. In the end, determinants of
“years of schooling” are explicitly sought by specifying an “education equation” and estimating
simultaneously with the Mincer equation.
17 Since the question on the year of onset of hearing disability was not successfully collected, dummy variables concerning whether a hearing disability was obtained by born; whether the disability occurs in the pre-lingual period
10
Results of the modified estimations
Tables 13 and 14 are results of the experiments to elaborate the determinants of income and
education.
First, logged income‐minus‐private income transfer is used as the dependent variable and
regressed to the same set of explanatory variables as used for the benchmark regression shown in Table
12. The result of this experiment is exhibited in the second column of Table 13. The most impressive
change is that a decrease in the point estimate of the rate of return on education from 20‐30 percent
shown in Table 12 to 14.2 percent, which is closer to those found in the related literature.
The second experiment is to augment explanatory variables. The third and fourth column for
OLS and Tobit estimation demonstrate the effects of the augmentation. First, the point estimates of rate of
return on education decrease from the figures attained by less explanatory variables exhibited in Table 12.
Therefore, the upward bias might be reduced by adding variables which were omitted in the benchmark
regression and positively correlated with years of schooling. An annoying new feature is higher
estimates of coefficients of age, which does not bear any reasonable explanation. Another interesting
issue is that the level of significance of the sex dummy dropped from 95 percent to 90 percent.
The third experiment is to use the instrumental variables for years of schooling. The 2SLS result
is exhibited in the above‐left quarter of Table 14, while that of IV Tobit is shown in the fifth column of
Table 13. A general inclination with IV exercises applied in this study is that the standard error increases
several times while the point estimates do not change that much. Thus, the significance declines
accordingly. The point estimates of rate of return on education are 0.152 with 2SLS (Table 14) and 0.206
with IV Tobit (Table 13). In the meantime, the robust standard errors are 0.362 with 2SLS and 0.851 with
IV Tobit, respectively. Then, with both methods, the rate of return on education turns insignificant. This
trade‐off between the treatment of endogeneity and a decrease in precision in the form of an increase in
standard error is a common concern to the instrumental variable method (Wooldridge [200?]). The same
difficulty appears with our study.
Similarly, the sex dummy becomes insignificant due to increases in standard errors with both
IV Tobit and 2SLS. Again, the point estimates are not affected by adoption of IV methods very much.
Results of tests of relevancy and validity of instruments are acceptable levels. Since the number
of instruments is great, first‐stage R2 and adjusted R2 are not low. In the meantime, Shea’s partial R2 is as
(0-2 years old); or whether it occurs in the post-lingual period (3 years old or above); were used for estimation.
11
low as 0.036.18 This relatively low relevancy might be a cause of high standard error with the IV methods.
The validity of the instruments are confirmed with the test of overidentifying restrictions in the sense that
there is no significant association between the errors and the instruments.
The final exercise is estimation of system of equations, i.e. the Mincer equation and the
education equation. The education equation is that with years of schooling as the dependent variable
and those listed in the second half of the first column of Table 14. The second and third columns display
results of 2SLS for the Mincer equation and OLS for the education equation for reference. Iterated 3SLS is
used for estimation of system of equations.
The result of iterated 3SLS of the Mincer equation is qualitatively similar to the counterpart
from 2SLS. The point estimate of rate of return on education is 15.9 percent, though its standard error is
so great. The sex dummy takes high point estimate which is close to unity in the absolute value.
However, its standard error is also high, so that the negative figure is statistically insignificant.
In the meantime, a significant gender difference is found in the education equation (Table 14).
The coefficient on the sex dummy is significantly negative of 90 percent significance level. Even though
the gender difference in income becomes insignificant probably because of weak instruments, that in
education still remains. Another intriguing finding is a significantly positive impact of father’s years of
schooling to PWD’s own years of schooling. According to the result of iterated 3SLS, the coefficient is
0.158 with the standard error of 0.071. Incidentally, mother’s education does not show a significant effect.
They imply that high level of education of father is likely to be inherited to his daughter / son with
disability. In other words, a highly educated father can afford and/or wants to give his child higher
education. And high level of education of a father might be a proxy for high level of income and/or
wealth of him. If that is the case, the significantly positive coefficient of father’s years of schooling implies
that the wealth of father enables his child with disability to go to school longer. A flip side is that children
with disability who have an uneducated (poor) father has less scope to be in school for a long time. In
other words, a vicious circle of poverty (precisely, poor education) runs between a father and his child.
6. Concluding Remarks
It is highly probable that persons with disability is poorer than those without disability on
average. This is the first study providing micro‐level evidence for this hypothesis in the Philippines,
18 Shea [1997] states that such low partial R2 is common to many well-known studies.
12
though the scope of research is limited to four cities in Metro Manila. As long as PWDs living in the four
cities are concerned, the incidence, depth and severity of poverty appear far greater than non‐PWDs.
What are causes of low income? A simple Mincer regression indicates that education
differentiate people with high education and else drastically. In addition, female PWDs seem to have
disadvantage in earning income.
The effects of education and gender are obscured when (1) alternative proxy for income; (2)
omitted variables; and (3) endogeneity of education; are taken into account. Then, other aspects emerge
as determinants of years of schooling of PWDs, which are sex and father’s education. What remain
important are: (1) female PWDs face disadvantage in education; and (2) rich human capital in terms of
length of schooling of father (not mother) is inherited to his child with disability. To sum up,
conventional aspects to explain poverty such as gender and inheritance of (human) wealth matter to
PWDs in Metro Manila, too. These observations are not surprising, but vital to PWDs.
References
Barbara Altman [2006] “The Washington Group: Origin and Purpose,” in Barbara Altman and Sharon N.
Barnartt eds., International Views on Disability Measures: Moving Toward Comparative Measurement ‐
Research in Social Science and Disability, Volume 4, Amsterdam: Elsevier Science, B.V., pp.1‐8.
Card, David [1999] “The Causal Effect of Education on Earnings,” in Orley Ashenfelter and David Card
ed., Handbook of Labor Economics, Vol. 3A, Amsterdam: Elsevier Science, B. V., pp. 1801‐1863.
Deaton, Angus [1997] The Analysis of Household Surveys: A Microeconometric Approach to Development Policy,
Baltimore and London: Johns Hopkins University Press.
Foster, James, Joel Greer and Eric Thorbecke [1984] “A Class of Decomposable Poverty Measures,”
Econometrica, Vol. 52, No. 3, May, pp. 761‐766.
Maluccio, John [1998] “Endogeneity of Schooling in the Wage Function: Evidence from the Rural
Philippines,” Discussion Paper No. 54, Washington, D.C.: International Food Policy Research
Institute.
Mincer, Jacob [1958] “Investment in Human Capital and Personal Income Distribution,” Journal of
Political Economy, Vol. 66, No. 4, August, pp. 281‐302.
Mori, Soya; Celia Reyes; and Tatsufumi Yamagata, eds. [2009] Poverty Reduction for the Disabled in the
Philippines: Livelihood Analysis from the Data of PWDs in Metro Manila, Joint Research Program
13
14
Series No. 151, Chiba, Japan: Institute of Developing Economies.
National Statistics Office (NSO) [2004] 2000 Census of Population and Housing, Manila: NSO.
Schady, Norbert [2003] “Convexity and Sheepskin Effects in the Human Capital Earnings Function:
Recent Evidence from Filipino Men,” Oxford Bulletin of Economics and Statistics, Vol. 65, Issue 2,
May, pp. 171–96.
Shea, John [1997] “Instrument Relevance in Multivariate Linear Models: A Simple Measure, Review of
Economics and Statistics, Vol. 79, No. 2, May, pp. 348‐352.
Willis, Robert J. [1986] “Wage Determinants: A Survey and Reinterpretation of Human Capital Earnings
Functions,” in Orley Ashenfelter and Richard Layard ed., Handbook of Labor Economics, Vol. 1,
Amsterdam: Elsevier Science Publishers, B.V., pp. 525‐602.
Wooldridge, Jeffrey M. [2002] Econometric Analysis of Cross Section and Panel Data, Cambridge, MA: MIT
Press.
Yamauchi, Futoshi [2005] “Why Do Schooling Returns Differ? Screening, Private Schools, and Labor
Markets in the Philippines and Thailand,” Economic Development and Cultural Change, Vol. 53, No.
4, July, pp. 959‐981.
Yap, Josef; Celia Reyes; Jose Ramon Albert; and Aubrey Tabuga [2009] “Preliminary Results of the Survey
on Persons with Disabilities (PWDs) Conducted in Selected Metro Manila Cities,” Soya Mori;
Celia Reyes; and Tatsufumi Yamagata, eds., Poverty Reduction for the Disabled in the Philippines:
Livelihood Analysis from the Data of PWDs in Metro Manila, Joint Research Program Series No.
151,Chiba, Japan: Institute of Developing Economies, pp. 17‐143.
Table 1. Total Number of PWDs by Sampled City: 2000
City Type of Impairment
Mobility Visual Hearing Others Total Makati 233 4,637 133 227 5,230 Pasay 161 1,189 66 126 1,542 Quezon 850 4,701 372 720 6,643 Valenzuela 203 1,990 63 193 2,449 Source: NSO [2004].
Table 2. Number of Respondents by Type of Impairments and by Area
City Type of Impairment
Mobility Visual Hearing Multiple Total
Makati 54 31 38 2 125 Pasay 29 27 23 5 84 Quezon 28 58 32 4 122 Valenzuela 27 28 15 2 72 Total 138 144 108 13 403 Percentage Makati 39 22 35 15 31 Pasay 21 19 21 38 21 Quezon 20 40 30 31 30 Valenzuela 20 19 14 15 18 Total 100 100 100 100 100 Source: Yap et al. [2009], Table 8.
15
Table 3. Respondents by Highest Educational Attainment and by Impairments
Highest Educational Attainment Type of Impairment
Mobility Visual Hearing Multiple TotalNever go to school 5 19 5 3 32 Kindergarten/preparatory school 0 0 2 0 2 Grade I to V 12 22 28 2 64 Elementary graduate 11 14 6 0 31 1st to 3rd year high school 25 12 20 0 57 High school graduate 26 24 24 5 79 Vocational school 20 13 1 1 35 Post‐secondary 2 0 0 0 2 College level 26 22 17 1 66 College or university graduate 11 15 5 1 32 Master or higher 0 3 0 0 3 Total 138 144 108 13 403 Source: Yap et al. [2009], Table 21a.
Table 4. Percentage of Respondents by Highest Educational Attainment and by Impairments
Highest Educational Attainment Type of Impairment
Mobility Visual Hearing Multiple Total Never go to school 3.6 13.2 4.6 23.1 7.9 Kindergarten/preparatory school 0.0 0.0 1.8 0.0 0.5 Grade I to V 8.7 15.3 25.9 15.4 15.9 Elementary graduate 8.0 9.7 5.6 0.0 7.7 1st to 3rd year high school 18.1 8.3 18.5 0.0 14.1 High school graduate 18.8 16.7 22.2 38.5 19.6 Vocational school 14.5 9.0 0.9 7.7 8.7 Post‐secondary 1.4 0.0 0.0 0.0 0.5 College level 18.8 15.3 15.7 7.7 16.4 College or university graduate 8.0 10.4 4.6 7.7 7.9 Master or higher 0.0 2.1 0.0 0.0 0.7 Total 100 100 100 100 100 Source: Yap et al. [2009], Table 21b.
16
Table 5. Average Years of Schooling by Type of Impairment and Sex
Impairment Female Male Total Mobility 8.5 9.3 9.1 Visual 7.6 8.1 7.9 Hearing 7.0 7.8 7.5 Multiple 5.5 8.4 7.1 Total 7.6 8.5 8.1 Source: Yap et al. [2009], Table 23i.
Table 6. Respondents with Income‐Generating Job by Sex
Status Sex
Total Female Male
With 61 142 203 Without 93 105 198 No answer 0 2 2 Total 154 249 403 Percentage With 39.6 57.0 50.4 Without 60.4 42.2 49.1 No answer 0.0 0.8 0.5 Total 100 100 100 Source: Yap et al. [2009], Table 106.
Table 7. Percentage of Respondents with Income‐generating Job to Total Respondents by Type of Impairment
Impairment Status
Without With No answer Total Mobility 76 61 1 138 Visual 41 103 0 144
Hearing 73 34 1 108 Multiple 8 5 0 13 Total 198 203 2 403 Percentage Mobility 55.1 44.2 0.7 100 Visual 28.5 71.5 0.0 100 Hearing 67.6 31.5 0.9 100 Multiple 61.5 38.5 0.0 100 Total 49.1 50.4 0.5 100 Source: Yap et al. [2009], Table 108.
17
Table 8. Mean Annual Income of Respondents from Various Source by Type of Impairment Source Mobility Visual Hearing Multiple All Wages and Salaries 10,460 58,315 13,053 6,111 28,127
Profits from business 15,320 4,745 1,870 16,622 8,004
Rent for buildings/rooms/lands 1,733 1,894 3,906 462 2,331Interests and dividends from bonds, savings and stocks
92 0 47 0 44
Pension 7,690 592 1,048 5,700 3,256Benefit/allowance from government
4,412 296 0 0 1,617
Receiving money from family members/friends
12,229 7,352 24,967 15,554 13,995
Others 1,759 3,076 10 0 1,706
Total Income 55,681 76,270 45,667 44,077 60,173
Source: Yap et al. [2009], Table 138.
Table 9. Difference in Personal Income by Disability
Variable Coefficient Ordinary Standard Error Robust Standard Error
Standard Error
t‐value p‐value Standard Error
t‐value p‐value
Constant 55,224.7 7,377.0 7.49 0.000 6,948.7 7.95 0.000Dummy: Visual 19,188.6 9,897.5 1.94 0.053 9,390.0 2.04 0.042Dummy: Hearing ‐9,641.2 10,596.2 ‐0.91 0.363 10,770.0 ‐0.90 0.371Dummy: Cognitive 2,287.8 43,539.0 0.05 0.958 24,895.0 0.09 0.927Dummy: Others ‐14,772.2 19,567.7 ‐0.75 0.451 12,044.6 ‐1.23 0.221Note: The dependent variable is personal income. The reference disability is the mobility impairment. The R‐squared and the adjusted R‐squared are 0.022 and 0.012, respectively. The number of observations is 397.
Table 10. Poverty Indices in Metro Manila (%) IDE‐PIDS Survey 2008 FIES 2006
Head Count Ratio (P0) 40.8 10.4
Poverty Gap Ratio (P1) 30.6 1.5
Squared Poverty Gap Ratio (P2) 27.0 0.5
Note: FIES is the abbreviation of the Family Income and Expenditure Survey which was conducted by the National Statistical Coordination Board in the whole country in 2006. The figures of the FIES 2006 are cited from Tables 2 and 11 of the following site: http://www.nscb.gov.ph/poverty/ 2006_05mar08/tables.asp. The squared poverty gap ratio is called the “severity of poverty” on the site.
18
Table 11. Summary of Variables for Estimations
Variable Obs. Mean Std. Dev. Min. Max. Personal income 397 58,996.42 84,943.26 1 660,000Personal income minus private income transfer 389 45,712.22 68,392.78 1 600,000Age 403 38.42 12.63 15 67Sex dummy (Female = 1) 403 0.38 0.49 0 1Marriage dummy (Married = 1) 402 0.47 0.50 0 1Years of schooling 403 8.15 4.08 0 15Disability dummy: Mobility (reference) 403 0.37 0.48 0 1Disability dummy: Visual 403 0.38 0.49 0 1Disability dummy: Hearing 403 0.29 0.45 0 1Disability dummy: Cognitive 403 0.01 0.11 0 1Disability dummy: Others 403 0.05 0.22 0 1Area dummy: Makati (reference) 403 0.31 0.46 0 1Area dummy: Quezon 403 0.30 0.46 0 1Area dummy: Pasay 403 0.21 0.41 0 1Area dummy: Valenzuela 403 0.18 0.38 0 1Year of onset of mobility impairment 138 1,985.09 16.87 1950 2008Year of onset of visual impairment 144 1,980.42 14.28 1950 2007By‐born hearing disability dummy (reference) 108 0.58 0.50 0 1Pre‐lingual hearing disability dummy 108 0.22 0.42 0 1Post‐lingual hearing disability dummy 108 0.19 0.40 0 1Father’s life dummy: Alive 390 0.49 0.50 0 1Mother’s life dummy: Alive 395 0.68 0.47 0 1Father’s years of schooling 341 8.27 3.86 0 15Mother’s years of schooling 359 7.46 3.80 0 15Years of schooling while father was alive 359 13.97 3.23 0 15Years of schooling while mother was alive 376 14.62 1.91 0 15Father’s disability dummy (Yes=1) 386 0.08 0.26 0 1Mother’s disability dummy (Yes=1) 389 0.08 0.28 0 1Difference in age with father 337 30.78 8.63 14 60Difference in age with mother 359 27.13 7.06 11 47
19
Table 12. Benchmark Mincer Regression Explanatory variables OLS Tobit
Intercept 0.371 (2.229)
‐1.826 (2.855)
Years of schooling 0.247*** (0.053)
0.301*** (0.066)
Age 0.292** (0.116)
0.357** (0.146)
Age squared ‐0.003** (0.001)
‐0.004** (0.002)
Sex dummy (female = 1) ‐1.115** (0.437)
‐1.360** (0.537)
Marriage dummy (the married = 1) ‐0.104 (0.450)
‐0.174 (0.537)
Disability dummy: Visual 1.825*** (0.514)
2.170*** (0.628)
Disability dummy: Hearing 0.900 (0.562)
1.152* (0.689)
Disability dummy: Cognitive ‐1.490 (2.375)
‐1.929 (2.970)
Disability dummy: Others ‐0.078 (1.171)
‐0.076 (1.428)
Number of observations 396 396 R2 0.133 0.025 Note: The dependent variable is the logarithm of income of PWD. For respondents who do not have any income, the income of them is assumed to be one peso instead of zero peso, because of the convenience with the log transformation. As annual income, one peso is taken to be as small as zero peso in Metro Manila. The figures in parentheses are heteroscedasticity‐robust standard error. The coefficients with *** , ** and * are statistically significant with the significance of 99%, 95%, and 90%, respectively. As for the Tobit estimation, 73 observations are left‐censored at the log income of zero. The log pseudo likelihood is ‐1,053.8. For information, the value of R2 is the pseudo R2.
20
Table 13. Estimations of Augmented Mincer Equation Dependent variable: ln (income minus income transfer) OLS OLS Tobit IV Tobit
Intercept ‐4.147 (2.526)
‐1.724 (3.316)
‐8.611* (5.076)
‐5.979 (5.409)
Years of schooling 0.142** (0.060)
0.134* (0.069)
0.201** (0.099)
0.206 (0.851)
Age 0.496*** (0.133)
0.340** (0.154)
0.558** (0.232)
0.466 (0.348)
Age squared ‐0.006*** (0.002)
‐0.004* (0.002)
‐0.006** (0.003)
‐0.005 (0.004)
Sex dummy (Female = 1) ‐1.156** (0.512)
‐1.029* (0.580)
‐1.412* (0.819)
‐1.361 (1.054)
Marriage dummy (Married = 1) 0.260 (0.554)
0.516 (0.612)
0.547 (0.824)
0.594 (1.261)
Disability dummy: Visual 2.646*** (0.589)
5.168*** (1.637)
7.586*** (2.598)
6.178** (2.810)
Disability dummy: Hearing ‐0.179 (0.708)
‐0.600 (1.613)
‐0.656 (2.301)
‐2.981 (2.977)
Disability dummy: Cognitive ‐2.971 (2.295)
‐5.089** (2.205)
‐9.071** (4.374)
‐5.955 (5.738)
Disability dummy: Others (except mobility) 0.320 (1.369)
‐1.759 (1.739)
‐2.377 (2.490)
‐3.096 (2.656)
Area dummy: Quezon 1.108
(0.701) 1.754* (1.013)
1.630 (1.797)
Area dummy: Pasay 2.801***
(0.783) 4.104*** (1.078)
4.078*** (1.141)
Area dummy: Valenzuela 0.519
(0.844) 0.961 (1.195)
0.651 (1.875)
Father’s life dummy: Alive ‐0.138
(0.663) ‐0.217 (0.880)
0.447 (1.015)
Mother’s life dummy: Alive 0.050
(0.701) ‐0.022 (0.915)
‐0.047 (1.220)
Father’s years of schooling 0.051 (0.091)
0.092 (0.126)
0.037 (0.212)
Mother’s years of schooling ‐0.134
(0.105) ‐0.199 (0.144)
‐0.131 (0.173)
Father’s disability dummy (Yes=1) ‐0.213
(1.065) ‐0.346 (1.503)
‐0.147 (1.762)
Mother’s disability dummy (Yes=1) ‐0.468
(1.159) ‐0.618 (1.611)
‐0.214 (1.818)
Year of onset of mobility impairment * Mobility disability dummy 3.3×10‐4
(8.3×10‐4) 6.3×10‐4 (1.2×10‐3)
‐2.9×10‐5 (0.001)
Year of onset of visual impairment * Visual disability dummy ‐1.4×10‐4*
(8.0×10‐4) ‐0.002* (0.001)
‐0.002* (0.001)
Pre‐lingual hearing disability dummy * Hearing disability dummy 1.252
(1.232) 1.999 (1.917)
2.781 (2.146)
Post‐lingual hearing disability dummy * Hearing disability dummy ‐0.662
(1.586) ‐1.361 (2.745)
‐1.358 (3.054)
Sample size 388 307 307 273
Goodness of fit R2=0.171 R2=0.231 Pseudo R2=0.049
Log pseudo likelihood = ‐1404.3
Left censored observations ‐ ‐ 96 85 Uncensored observations ‐ ‐ 211 188
Wald test of exogeneity ‐ ‐ ‐ χ2(1)=0.00 [0.946]
Note: Figures in parentheses are robust standard errors. The coefficients with *** , ** and * are statistically significant with the significance of 99%, 95%, and 90%, respectively. The instruments for the IV Tobit are the same as explanatory variables of years of schooling, which appear in the latter half of Table 14.
21
Table 14. Simultaneous Equations Model (Iterated 3SLS): Income and Education Coef. Std. Err. Coef. Std. Err. Mincer (Dependent variable: ln [income minus income transfer]) 2SLS Iterated 3SLS Intercept 0.193 (3.402) ‐0.182 (3.377) Years of schooling 0.152 (0.362) 0.159 (0.404) Age 0.266 (0.190) 0.280 (0.194) Age squared ‐0.003 (0.002) ‐0.003 (0.002) Sex dummy (Female = 1) ‐0.982 (0.655) ‐0.974 (0.639) Marriage dummy (Married = 1) 0.581 (0.732) 0.652 (0.784) Disability dummy: Visual 4.284** (1.714) 4.247* (2.571) Disability dummy: Hearing ‐2.204 (1.926) ‐2.184 (1.844) Disability dummy: Cognitive ‐3.377 (2.796) ‐3.405 (3.971) Disability dummy: Others (except mobility) ‐2.232 (1.703) ‐2.209 (1.445) Area dummy: Quezon 1.058 (0.953) 1.061 (0.982) Area dummy: Pasay 2.828*** (0.801) 2.823*** (0.849) Area dummy: Valenzuela 0.342 (1.054) 0.348 (1.044) Father’s life dummy: Alive 0.326 (0.706) 0.344 (0.712) Mother’s life dummy: Alive 0.006 (0.786) 0.044 (0.867) Father’s years of schooling 0.008 (0.115) 0.007 (0.117) Mother’s years of schooling ‐0.089 (0.112) ‐0.089 (0.105) Father’s disability dummy (Yes=1) ‐0.118 (1.218) ‐0.123 (1.232) Mother’s disability dummy (Yes=1) ‐0.166 (1.272) ‐0.146 (1.111) Year of onset of mobility impairment * Mobility disability dummy ‐1.5×10‐4 (1.0×10‐3) ‐1.8×10‐4 (9.1×10‐4) Year of onset of visual impairment * Visual disability dummy ‐15.3×10‐4* (8.0×10‐4) ‐0.002 (0.001) Pre‐lingual hearing disability dummy * Hearing disability dummy 1.721 (1.317) 1.697 (1.331) Post‐lingual hearing disability dummy * Hearing disability dummy ‐0.692 (1.647) ‐0.708 (1.534)
Sample size and overall significance n=273 χ2=135.9 [0.000]
n=273 χ2=90.4 [0.000]
Fist stage, Goodness of fit: R2, Adjusted R2, Partial R2 0.231, 0.139, 0.036
Test of overidentifying restrictions: Score χ2 (d.f.=7) [p‐value] 5.83 [0.559]
Education (Dependent variable: Years of schooling) OLS Iterated 3SLS Intercept 1.557 (4.642) ‐10.079 (10.841) Sex dummy (Female = 1) ‐0.873* (0.478) ‐0.822* (0.466) Area dummy: Quezon ‐1.439** (0.585) ‐1.624*** (0.565) Area dummy: Pasay ‐0.106 (0.722) ‐0.134 (0.680) Area dummy: Valenzuela ‐1.643** (0.671) ‐1.616** (0.658) Father’s years of schooling 0.176** (0.074) 0.158** (0.071) Mother’s years of schooling 0.088 (0.070) 0.111 (0.075) Father’s disability dummy (Yes=1) ‐0.579 (0.829) ‐0.411 (0.988) Mother’s disability dummy (Yes=1) ‐0.364 (0.845) ‐0.675 (0.874) Years of schooling while father was alive 0.280 (0.441) 0.336 (0.419) (Years of schooling while father was alive)2 ‐0.014 (0.024) ‐0.016 (0.023) Years of schooling while mother was alive ‐0.533 (0.526) 1.551 (2.228) (Years of schooling while mother was alive)2 0.034 (0.029) ‐0.057 (0.105) Difference in age with father 0.138 (0.185) 0.164 (0.186) (Difference in age with father)2 ‐0.002 (0.003) ‐0.003 (0.003) Difference in age with mother 0.154 (0.255) 0.180 (0.236) (Difference in age with mother)2 ‐0.002 (0.005) ‐0.002 (0.004) Disability dummy: Visual 1.283 (1.001) 0.660 (2.072) Disability dummy: Hearing ‐0.510 (1.235) ‐0.679 (1.488) Disability dummy: Cognitive 2.252 (1.436) 3.999 (3.189) Disability dummy: Others (except mobility) ‐0.987 (1.411) ‐1.081 (1.124) Year of onset of mobility impairment * Mobility disability dummy 8.110‐4 (6.110‐4) 5.9×10‐4 (7.3×10‐4) Year of onset of visual impairment * Visual disability dummy 1.610‐4 (6.810‐4) 3.5×10‐4 (0.001) Pre‐lingual hearing disability dummy * Hearing disability dummy 1.162 (1.120) 1.034 (1.040) Post‐lingual hearing disability dummy * Hearing disability dummy 0.219 (1.158) 0.606 (1.199)
Sample size and overall significance n=285 χ2=2.43 [0.000]
n=273 χ2=56.1 [0.000]
Note: The standard error of 2SLS and OLS is the robust S.E. The coefficients with *** , ** and * are statistically significant with the significance of 99%, 95%, and 90%, respectively.
22
Figure 1. Cities to Be Studied and Distribution of Respondents
Source: Yap et al. (2009), Figure 2.
23
Figure 2. Per Capita Revenue and Scale of Population by City
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
0 500 1,000 1,500 2,000 2,500
Per C
apita
Rev
enue
(pes
o)
Population (thousand persons, 2000)
Pasay
Valenzuela
Quezon
Makati
Figure 3. Histogram of Samples in Personal Income
020
4060
8010
0Fr
eque
ncy
(Per
sons
)
0 200,000 400,000 600,000 800,000Personal Income (Peso)
19,345
Note: The broken line indicates the poverty line in Metro Manila in 2007, which was released by the National Statistical
Coordination Board, the Philippines.
24
top related