cash transfer policy impact on education: the uruguayan ... de... · keywords: asignaciones...
TRANSCRIPT
Cash Transfer Policy impact on education: The Uruguayan Case “Asignaciones Familiares”.
Nicolás González Pampillón§
June, 2007
Abstract A Cash Transfer Policy not only should contribute to poverty reduction. It also should aim to
increase the educational level of benefit receivers and thus their work opportunities will rise
in the future. This paper assesses the impact of the Cash Transfer Policy, “Asignaciones
Familiares”, on education. The results reveal that this policy does not have an impact on
education of the benefit receivers during the year 2005. Therefore we think that the system
requires a redesign, emphasizing the role of education on the beneficiaries and their families
to attain its goal.
JEL classification: I21, I38, J18
Keywords: Asignaciones Familiares, educational level, work opportunities.
§ Nicolás González is a student of Economics, Universidad de Montevideo, Uruguay. He wishes to thank Alejandro Cid for his valuable comments and for his help in defining some of the variables explored. The author alone is responsible for the opinions expressed and for the errors in this paper. Contact information: [email protected]
Introduction Education is an essential tool for human development from an individual and social point of
view. Actually, the role of education in the socio-politic scheme raised for the postmodernist
world acquires a great relevance because its correct working depends, in part, on the
population education level. High education levels extends the opportunity scope, contributes
with a high social mobility, reduces social marginalization and poverty, and therefore diminish
the violence related to impotence and frustration generated by social exclusion.
People live in households with different socio-economic contexts. These differences
experiment a progressive increase, generating a considerable gap between people who live
in favorable context households and those who have to manage difficult situations, mainly
developing countries. This fact requires a government intervention, by means of social
policies attempting to increase opportunities for society sectors living in depressed contexts,
mainly for kids at early childhood.
Latin America and the Caribbean, have implemented conditional cash transfer (CCT)
programs to mitigate the negative effects of poverty and inequality in families with kids and
having low incomes; contributing to their human development and poverty alleviation1. These
programs link safety nets directly to human capital development, by making the transfer
conditional on school attendance and health care checkups. In most cases, the amount of
transfer is dependent to school attainment, creating positive incentives to invest in human
capital. CCT has generally been successful in terms of rising educational level and reducing
child work in short term. The effects are not clear in the long run.
Despite the fact that this kind of policies has been applied recently in Latin America and the
Caribbean, Uruguay has cash transfer regulations since 1943, called: “Asignaciones
Familiares”. The original system suffered a lot of legislative modifications according to
historical changes in the region and the world, associated with laws about cash transfer
policies. The actual system differs substantially with the original one in the main aim,
implementation and functioning. This paper investigates “Asignaciones Familiares”
implementation; mainly their effects on beneficiary schooling attainments.
The work is divided into four sections. Next section (Section 1) describes the legal aspects of
the system, mentioning the most important modifications that lead to the actual form.
Additionally it includes statistically information describing its evolution. Section 2 proceeds to
1 See Skoufias, E. and S. Parker (2001); and see Rawlings, L. and Rubio, G. (2003).
describe the sample which is used to formulate the model that measures the impact on
education. Section 3 presents the methodology and results. Finally, Section 4 contains
conclusions and policy recommendations. Appendix showed additional information used.
Section 1: Law regulations and statistical information. I) Legal and Institutional aspects
In Uruguay we have the law n° 10.449 of 12.XI.1943 (Consejo de Salarios y Asignaciones
Familiares). “Asignaciones Familiares” means the extra salary sum of money that a family
receives from the State because of the fact of having children. The law was approved
because the State wanted to help workers to pay their family expenses. The employer must
pay for each employee he/she has in his/her company. With this extra payment, the family
receives a sum of money depending on the number of children the family has. The money
comes from a fund named “Compensation Cash”2.
The previous law was replaced by the actual one (15.084 of 9th December 1980), where
“Asignaciones Familiares” is defined as a sum of money (benefit amount) received by an
employee who has children under his/her responsibility. Under the same regulations the
benefit will be given to: a) those who are on the dole during the first six months; b) workers;
c) maids and rural workers; d) small rural producers who pay their taxes and work in their
own plots; e) pensioners and retired people from private activity and pensioners from the
private activity as well as from private banks. The money is charged for their own children or
those who are under their care. The benefit amount implies an 8% of the minimum national
salary.
Since 1980, when the law n° 15.080 was applied, there have been subsequent laws that
modified it substantially:
1) Law n° 16.697 of 25.VI.1995
Focus and doubled. The benefit amount implies a 16% of the minimum national salary. The
people that will receive this benefit must have a salary which is less than six minimum
national salaries. While the benefit amount will be of 8% of the minimum national salary per
each person, when they receive salaries which are more than six but not less than ten
minimum national salaries and when the family has also two children. The upper limit
increases in one minimum national salary when the family has another child after the second
one. Households whose salaries are more than the upper limit do not receive the benefit. To
2 Pérez del Castillo, S. "Manual Práctico de Normas Laborales", Edición nº 10.
determine the income level of a family (household income) they also add the spouse’s salary
to the partner who lives at the same place of that who receives the benefit.
2) Law n° 17.139 of 16.VII.1999
Households with Low Incomes (HLI). The benefit includes children who belong to households
with low incomes. These kind of households are those whose monthly household income
whatever it is implies a sum which is no more than six “Bases de Prestaciones y
Contribuciones” (BPC)3. The benefit amount is 16% of the BPC and the administrators are:
• Women living alone and in charge of the household;
• Workers, men or women under the dole benefit and when the dole is over, as
well;
• Women who are pregnant will receive a sum of money from the beginning of
their pregnancy and during twelve months after their babies are born.
3) Law n° 17.758 of 4.V.2004
HLI Extension. The benefit also includes the household with any kind of income with an
upper limit of 3 BPC and if they were not included in other laws.
• The man who has not been on the dole;
• The woman who supports her home on her own.
The sum of money is a 16% of the BPC for each beneficiary; for those who have any kind of
inability the amount of money will be doubled. To be benefit administrator the person must
have a legal capacity and show a legal certificate stating that he is in charge of the
youngsters.
Summarizing, the legal system of today is ruled by the law 15.084 of 1980 and its following
modifications: Law nº 16.697 of 1995, Law nº 17.139 of 1999 and Law nº 17.758 of 2004 and
Decreto n° 227/981 and Decreto n° 596/985.
3 This substitutes the national minimum salary as a reference value. The amount of money which represents this is fixed by the President and his Ministers. In July of 2005 1 BPC was $ 1.397 equivalent to USD 58.
Considerations concerning the original law modifications
The changes in the system made by the above mentioned laws are related to changes which
took place in the different cash transfer programs in the world. However, in other countries
many of these changes have focused on the schooling children achievements while the
modifications made in Uruguay have highlighted the role of “Asignaciones Familiares” as
simple resource transferences. The fact is that the benefit has been spread to households
with low incomes, workers who are no part of the social security and other situations where
there’s no economical support, modified the original aim of the 1943 law.
The increase in poverty, homeless people and informal workers in Uruguay, justifies the
existence of a group of people who take advantage of its benefit. However, the aim is not
clearly expressed and the existence of so many laws on the same subject makes it more
difficult to clarify it. While in other parts of Latin America and the Caribbean where the cash
transference programs have very strict conditions, the Uruguayan system is defined as a
benefit depending on tested means.
Age limit benefit
The benefit administrators charge for their children until they are 14. This period can be
extended in the following cases:
i) Until 16 years old, when the beneficiary can’t finish the primary school at 14
because of justified reasons; or if it is a dead worker son; or if the beneficiary
suffers any kind of inability; or if it is in prison.
ii) Until 18 years old, if the beneficiary is attending a secondary school (either public
or private).
iii) A life benefit is received if the beneficiary has a physical or mental handicap of
such kind which prevents him from working normally.
Benefit amount for each youngster
i) 16% of one BPC, if the household income is not more than 6 BPS or if one HLI.
ii) 8% of the BPC if these incomes are between 6 and 10 BPC and the family has
two children. If the family has more than two children, the upper limit is of 10 BPC
increasing in one if there is one more child who exceeds of two. So, with three
beneficiaries the upper limit is of 11 BPC. With four beneficiaries is 12 BPC.
There is no limit for retired people.
iii) If it is a beneficiary who has an inability his benefit will be twice the sum detailed
before.
The benefit amount payment is made bimonthly.
Requirements
The ones who receive the benefit amount must fulfill the following requirements at Banco de
Previsión Social (BPS), which is the public organ in charge of paying the benefit:
A) The family income must be signed in a document by the benefit administrators having
the benefit application form enclosed as well.
B) The fact of an usual attendance to teaching institutes either public or private as well as
having medical assistance controls required by BPS.
C) If people have an inability of such kind which prevents them from working according to
BPS doctors. Every three years medical controls will be made in order to evaluate if the
person has the same level of inability which justifies the fact of receiving money.
BPS can do any kind of inspections which consider convenient in order to know the truth
about the income declaration as well as the attendance of young children to schools and
medical centers.
Inter-institutional coordination
The law claims a fluent communication between the Public Teaching Administration or the
Private ones and the BPS in order to check the information which administrators provide
about the beneficiary attendance to teaching institutions. The “Instituto Nacional del Menor”
(INAME)4 will communicate BPS about the circumstances which imply suspension,
interruption or cancellation of the benefit given. If the BPS realizes any kind of false
information, the benefit could be suspended as well as sanctioned because of breaking the
law.
4 Actually INAU, “Instituto del Niño y el Adolescente del Uruguay”.
In the original law the benefit was given to those who belonged to a formal working condition
either state or private, but after the modifications it is permitted the informal working
conditions, so people could falsify their income when they fill in the application form in order
to receive the benefit. This adds a supplementary problem to the legal working system which
is difficult to avoid.
Medical care
Children under the benefit can have medical assistance at medical centers provided by BPS,
in its “Centros Materno-Infantiles”; or by means of medical insurance organizations called
“Instituciones de Asistencia Médica Colectiva” (IAMC).
II) Statistical Information
Here we describe some indicators, which give information about how the system works. The
evolution of the number of beneficiaries in the system experimented great fluctuations owing
to two main factors: 1) legal regulation changes in the system, which explain the significant
fluctuations; 2) the macroeconomic context5. In the following figure the most important
fluctuations are observed in 1995, 2000, 2001 and 2004.
Figure 1
Number of Beneficiaries and Growth Rate
0
100.000
200.000
300.000
400.000
500.000
600.000
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Years
Num
ber
of B
enef
icia
ries
-15,0
-10,0
-5,0
0,0
5,0
10,0
15,0
20,0
%
Number ofBeneficiaries(left)
Growth Rate(right)
5 A prosper macroeconomic context reduces the households necessity of cash transfer. The opposite happen when the macroeconomic context is instable and so that shows indicators performing improperly.
Source: Personal Elaboration based on the Stadistical Report: “Prestaciones de Actividad”, BPS In 1995, law nº 16.697, which established limits for the family income, was applied. This law
is restrictive because its upper limit excludes households that were in the system. The
number of beneficiaries dropped dramatically and its growth rate reaches a peak. The awful
macroeconomic context6 could also be contributing to this downfall.
In the case of law n° 17.139 of 1999, which extends the concept of benefit administrator,
started to rule in 2000. This fact was reflected in the substantial growth of the number of
beneficiaries. Albeit in the next year, 2001, the number of beneficiaries declined sharply. The
reasons behind this fact is the increase that the control level of BPS had and besides, this
law includes INAME as another organ of control which inspects whether the beneficiaries
fulfill the requirements or not , particularly the one referred to school attendance.
During the following years we observe a rising evolution sustained by the unfavourable
economical context. In 2004, law nº 17.758, which is a further extension of the benefit for
HLI, started to rule and for this reason the beneficiary growth rate reach a peak. Finally, in
2005 the number of beneficiaries increased the most.
Some Statistical Remarks
According to the information taken from ECLAC7, in 1994 the 9.7% of the urban population
was living below poverty line while in 2004 it was the 20.9%. We take these percentages as
representative of the whole population to calculate the number of people below poverty line.
We obtain the underestimated values of 306,885 and 677,370 respectively. The people in
poverty situation raised strongly, 120.7%, meanwhile the number of beneficiaries grew
slightly, 12.3% for the same period. In 1994 the system had 413,426 beneficiaries while in
2004 it had 464,397. Regarding to these underestimate values we consider that there were
people in poverty situation who were out of the system and probably had to be taking part of
it. This situation improved in 2005 because the number of poor people diminished to 609,309
and the number of beneficiaries grew to 521,905. In conclusion, despite the fact that the
situation is improving, we consider that there are cases out of the target.
6 We present some indicators as an evidence of the economic situation in 1995: Gross Domestic Product (GDP) real growth rate = -1.4%; Inflation = 35.4%; Unemployment rate = 10.3%. 7 Economic Commission for Latin America and the Caribbean
It is also observed that in other countries which applied cash transfer policy have a bigger
number of beneficiaries in the rest of the country than in the capital of the Country8. In
Uruguay the number of beneficiaries who live in the rest of the country grows progressively.
In 1991 the 57% of the beneficiaries lived in the rest of the country, while in 2005 it was a
68%. Since 1996 until 2004 the population in the rest of the country has raised in nearly
100,000 inhabitants9 under condition of poverty.
Benefit Amount
The benefit amount does not play an important role in the benefit administrator budget.
Particularly it represents a little part of the Basic Goods Set (BGS)10, and it also represents
an insignificant proportion of the poverty line (see Table 1).
Table 1
BENEFIT AMOUNT AND SOCIAL INDICATORS 1998 1999 2000 2001 2002 2003 2004 2005 Benefit Amount (USD) 14.83 14.54 14.02 13.12 8.37 6.65 7.13 9.04 Benefit Amount / BGS Montevideo 24.8% 25.4% 24.7% 24.7% 22.1% 19.1% 18.7% 18.3%Benefit Amount / BGS Rest of Urban Zone 32.6% 33.4% 32.5% 32.5% 29.0% 25.2% 24.6% 24.0%Benefit Amount / PL Montevideo 7.2% 7.4% 7.2% 7.2% 6.4% 5.6% 5.4% 5.3% Benefit Amount / PL Rest of Urban Zone 11.5% 11.8% 11.5% 11.5% 10.3% 8.9% 8.7% 8.5% Source: Elaboration extracted from "Asignaciones familiares, distribución del ingreso y pobreza en Uruguay. Un análisis para el período 2001-2004". Vigorito, A., based on data from National Institute of Statistics of Uruguay
The above indicators do not take into account the inflation effects. Whether we consider the
inflation effect in the evolution of Benefit Amount, we appreciate lose in real terms in recent
years (see Figure 2). We also have to remember that the benefit administrators take the
benefit amount bimonthly. In line with these matters, we think that the effect of the policy is
insignificant in relation with the development of households.
8 Most of developing countries are centralized in their capitals. In Uruguay all the important services and institutions are centralized in Montevideo, its capital. Therefore we differentiate between Montevideo and rest of the Country. 9 Data extracted from National Institute of Statistics of Uruguay. 10 The BGS contain goods that people usually consume. The National Institute of Statistics considers one Basic Goods Set for Montevideo and other for the Rest of Urban Zone.
Figure 2
Amount Real Growth Rate
1.5%
-24.3%
-2.1%
2.0% 3.3%
-0.6%-4.8%
-30.0%
-25.0%
-20.0%
-15.0%
-10.0%
-5.0%
0.0%
5.0%
1999 2000 2001 2002 2003 2004 2005
%
Source: Author elaboration based on data provided by the National Institute of Statistics of Uruguay.
Finally, the expenditures of the system do not represent an important proportion of BPS
expenses, as well as of Public Social Expenditures and Total Public Expenditures (see Table
2). System expenditures are partly financed by General Government Revenues. The other
part that is referred to the expenditure generated for benefit administrators who came into the
system owing to the last reform, is financed by availabilities of the Government Treasury and
by income of Banking Recovery Fund (law n° 17.758). Table 2
SYSTEM FINANCIAL INFORMATION (thousand
USD) 2002 2003 2004 2005 TOTAL PUBLIC EXPENDITURES (TPE) 4,660,235 4,300,264 4,346,299 5,415,036SOCIAL PUBLIC EXPENDITURES (SPE) 3,010,820 2,414,486 2,554,927 3,284,046BPS EXPENDITURES 1,813,293 1,422,009 1,488,015 1,909,867 1. Social Security 1,738,807 1,363,844 1,415,513 1,812,167 2. Social Work 74,485 58,165 72,501 97,700 2.1. “Asignaciones Familares” Expenditures (AFE) 39,080 30,574 41,734 56,755 2.2. Others 35,405 27,591 30,767 40,945 AFE / BPS EXPENDITURES 2,2% 2,2% 2,8% 3,0% AFE / SPE 1,3% 1,3% 1,6% 1,7% AFE / TPE 0,8% 0,7% 1,0% 1,0% Source: Author elaboration based on data provided by the “Boletín Estadístico del BPS” and data from "Identificación y ánalisis del Gasto Público Social en Uruguay. 2002-2005", Bertoni, R, Azar, P., y Silveira, M”.
Section 2. Sample Description. The sample was done by using cross-sectional data of the year 2005 from the Continuous
Households Survey (Encuesta Continua de Hogares – ECH), which is conducted by the
National Institute of Statistics (Instituto Nacional de Estadística – INE) of Uruguay. The
survey is carried out on urban regions of the country and it has a total sample size of 54,330
observations. It also includes different economical information of households and their
memberships, which is presented in qualitative and quantitative variables.
The sample extracted contains household structure composed by a head of household
(HOH), a spouse or cohabitee (of HOH), and child or children between eight and sixteen
years old. Child (or children) can be son or daughter of the couple, or only of one member of
the couple. This selection was made considering the research target, which focuses on
evaluating the impact of the policy on children educational attainment. In Uruguay
educational system starts with primary school at the age of six years old, so the selected
household group might have schooling child (or children). We excluded children under eight
because it is difficult to observe if they really have a schooling gap11. We excluded children
over sixteen because factors that explain schooling gap, which are difficult to specify in a
model, start to influence. We mention some of these factors in the following lines: leave the
educational system as a consequence of getting into the work market, drug problems,
internal behavioural aspects of each person, and negative influences of their social
environment. Although these factors can be affecting the interval selected, they have more
practical consequences in subsequent ages.
Sample selected remarks
In 2005, a Government with a different socio-politic ideology from the previous one, assumed
the political power. The new authorities introduced social programs attempting to reduce the
extent of people in poverty situation. The group of programs or plans was denominated
Social Emergency Plan (“Plan de Emergencia Social”). Since April 1st of 2005 that social
policy has started to coexist with “Asignaciones Familiares” and therefore it could be a
problem taking this year into account for the analysis. However, the impact of this social
policy on children school performance affects subsequent years. According to this idea we
selected year 2005, which includes “Asignaciones Familiares” last reform of year 2004.
11 Schooling gap indicates the relative lag behind the age-appropriate schooling level.
Sub-sample differences
The sample has 5261 observations. The 63% of the sample receives the benefit (3372
elements), and the 37% does not receive (1934 elements), thus we have two subsamples.
We generate different quantitative and qualitative variables to describe these subsamples.
The following table shows differences of the quantitative variables.
Table 3
QUANTITATIVE VARIABLES DESCRIPTION
Educational
Gap Children AgeMother
Age Father Age
Father Education
Mother Education
Not receive the benefit (1) 0.09 12.13 40.96 44.38 10.36 10.75
Receive the benefit (2) 0.15 12.66 38.18 41.80 7.76 8.20
Difference -0.06** 0.47** 2.77** 2.58** 2.60** 2.55**
Number of Household Members
Family Income
Father Experience
Mother Experience Wealth Index
Household Income
Not receive the benefit (1) 4.83 18,658.75 29.03 25.20 0.55 26,579.81
Receive the benefit (2) 5.53 72,42.98 29.03 25.00 0.36 12,142.74
Difference -0.70** 11,415.77** 0 0.2 0.19** 14,436.07**Source: Personal Elaboration based on data from Continuous Households Survey 2005. ** means are statistically different at 5% * means are statistically different at 10%
We find differences in almost of quantitative variables except for “Father Experience” variable
and “Mother Experience” variable. Relating to “Household Income” and “Wealth Index” 12, we
find a great difference which seems to be reasonable because the group who receives
Assignment has lower income.
The generated qualitative variables are binary variables, which takes 1 or 0 values. The
following Table gives information about what percentage of elements takes 1 value in each
group. Regarding to each qualitative variable shown in the Table, taking 1 value respectively
means: children have married parents; children live with biological parents; children have
father or mother with a degree given by a university; children have mother employed;
children have father employed; children have father or mother with public job; children have
mother or father subemployed; children have father or mother with formal job; children have 12 The formulation as well as the explanation of the Wealth Index is showed in the Appendix.
mother or father who did not go to work because of the bad weather; children have mother or
father irresponsible because he or she are underemployed and did not do nothing to find a
better job.
Table 4
QUALITATIVE VARIABLES DESCRIPTION
Married Parents
Biological Parents
Father or Mother Professional
Mother Labour Situation
Father Labour Situation
Not receive the benefit (1)
82% 87% 27% 62% 93%
Receive the benefit (2)
68% 82% 4% 52% 92%
Difference 14%** 5%** 23%** 10%** 1%
Public
Job Subemployed Formal Job Reasons for not
going to work Irresponsibility
Not receive the benefit (1)
36% 40% 78% 1% 5%
Receive the benefit (2)
8% 60% 61% 3% 5%
Difference 28%** 20%** 17%** 2%** 5% Source: Personal Elaboration based on data of Continuous Households Survey 2005. ** means are statistically different at 5% * means are statistically different at 10%
We also find differences in almost of qualitative variables except for “Father Labour Situation”
and “Irresponsibility. Concluding, this way of dividing the sample generates two homogenous
subsamples of households, which have different characteristics that seem to be reasonable a
priori.
Dependent Variable or Outcome Variable
Educational gap is defined as:
( ) ( )[ ]
( )6AgeSchoolingofYears6Age
Y−
−−=
Let Y be the variable that measures educational gap, which is defined as quotient where
numerator represents the absolute lag behind the age-appropriate schooling level and
denominator represent the expected years of schooling. Therefore, educational gap is
expressed as a proportion of the expected years of education, which takes values between
one and zero. When it takes one means that this child has 100% of educational gap, when it
takes zero means that this child has 0%. There are infinite possibilities inside interval [0,1]
because Y is a continuous variable.
Section 3. Methodology, Endogeneity and Results. I) Methodology
The methodology applied to measure the cash transfer policy effectiveness in this paper is
called “Treatment Evaluation”. This methodology implies comparing the average outcomes of
the treated and nontreated groups, with those which actually took part in the program. That is
what we call Average Treatment Effect on the Treated (ATT). In this case the treated group
is the one receives benefits and the non treated group is the one which does not receive. We
define the outcome variable previously as Educational Gap.
The members of each group, treated and nontreated present different characteristics as we
observed in Section 2. Our interest is to compare group members with similar observable
characteristics. In the terminology of treatment evaluation we call Matching, the process that
identify a comparison group with similar observable characteristics shown by x (a vector of
covariates), than the treated group. We use a matching estimator based on the Propensity
Score13, which is a conditional probability measure of treatment participation given x.
Formally
[ ]xX1DPrp(x) ===
There are different matching estimators that are based on Propensity Score. We use the
Kernel Matching Method14. “With Kernel Matching all treated are matched with a weighted
average of all controls with weights that are inversely proportional to the distance between
the propensity scores of treated and controls”15. Formally, the kernel matching estimator is
given by
( )( )∑ ∑
∑∈ ∈
−∈
−
⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
−=Ti Ck h
ppCj h
ppCjT
iTK
n
ik
n
ij
G
GYY
N1τ
Where is a kernel function and is a bandwidth parameter. ( )⋅G nh
13 Propensity Score concept is used when there is not a random assignment of the treatment. In other words there is a self-selection into treatment, like in our case. Propensity Score could be calculated using a Probit or a Logit Model. We use the last one. The results are on the Appendix. 14 See Heckman, J.J., Ichimura, H., Todd, P. (1997). 15 Becker, O.S., Ichino, A.(2004) “Estimation of average treatment effects based on propensity scores”.
After obtaining the result, we made a matching test to see if there were people compared
who have similar characteristics.
II) Endogeneity
The methods for estimation of treatment effects rely on assumptions to let meaningful
comparisons between the outcome of the two groups. Under some of these assumptions
there will be no simultaneity bias or selection bias and omitted variables bias, of endogeneity.
In the following lines we list and explain these assumptions.
Conditional independence assumption16
,xD,yy 10 ⊥ States that conditional on x, the outcome are independent of treatment. Behavioral
implications of this assumption is that participation in the treatment program does not depend
on outcomes, after controlling for the variation in outcomes induced by differences in x.
Unconfoundedness or Ignorability assumption
D,y 0⊥
Which implies conditional independence of participation and yo. The assumption is
tantamount to treatment assignment that ignores outcome.
These assumptions seem to be valid in this work. Atributaries do not receive Family
assignment as a consequence of their children educational gap.
16 Colin, A., and Trivedi, P.K (2005): “Micro-econometrics: methods and applications”.
III) Results
Average Effect of Treatment on the Treated17 (ATT) - estimation with the Kernel Matching
method.
Variable Sample Treated Controls Difference S.E. T-stat
Unmatched .14887498 .091892335 .056982645 .00497961 11.44Educational Gap ATT .14887499 .140813841 .008061138 .006031434 1.34
Note: S.E. for ATT does not take into account that the propensity score is estimated.
Untreated 3,327 Treated 1,934 Total 5,261
Bootstrapped standard errors
Bootstrap Normal - based Observed Coef. Std. Err.
z P > ׀z׀ [ 95% Conf. Interval ]
.0080611 .00066313 1.22 0.224 -.0049359 .0210582 Matching Test
Mean %reduct t-test Variable Sample Treated Control %bias |bias| t P > |t|
Unmatched .48862 .46739 4.3 1.49 0.137
Sex Matched .48862 .48668 0.4 90.9 0.12 0.904
Unmatched 5.9586 6.7553 -31.1 -10.81 0.000
Education Matched 5.9586 6.0465 -3.4 89.0 -1.09 0.277
Unmatched .04292 .24737 -60.6 -19.61 0.000
Private Education Matched .04292 .04289 0.0 100.0 0.00 0.997
Unmatched .08118 .11181 -9.5 -3.28 0.001 Household
Members > 60 Matched .08118 .08051 0.2 97.8 0.07 0.944
Unmatched .00155 .00331 -3.6 -1.19 0.234 Domestic Service
Matched .00155 .00255 -2.0 42.9 -0.69 0.491 17 The ATT measure shows the average gain from treatment for the treated. Therefore, ATT measure allows assessing the effectiveness of the policy. We use “psmatch2” command developed by Edwin Leuven & Barbara Sianesi, in STATA software.
Unmatched .01965 .02615 -4.3 -1.49 0.136
Biological Father Matched .01965 .02059 -0.6 85.6 -0.21 0.836
Unmatched 7.7601 10.358 -71.6 -23.98 0.000
Father Education Matched 7.7601 7.8284 -1.9 97.4 -0.72 0.473
Unmatched 29.04 29.025 0.2 0.06 0.955
Father Experience Matched 29.04 29.247 -2.2 -1305.5 -0.69 0.489
Unmatched .00827 .15179 -54.8 -17.27 0.000 Father -
Professional Matched .00827 .00908 -0.3 99.4 -0.27 0.787
Unmatched .03257 .03787 -2.9 -1.00 0.320 Father - Inactive
Matched .03257 .03397 -0.8 73.7 -0.24 0.809
Unmatched 8.2019 10.751 -71.2 -23.91 0.000 Mother Education
Matched 8.2019 8.259 -1.6 97.8 -0.58 0.559
Unmatched 24.98 25.204 -2.9 -1.03 0.302 Mother Experience Matched 24.98 25.054 -1.0 66.8 -0.30 0.760
Unmatched .02947 .19687 -54.8 -17.61 0.000 Mother-
Professional Matched .02947 .03014 -0.2 99.6 -0.12 0.904
Unmatched .37901 .29606 17.6 6.21 0.000 Mother - Inactive
Matched .37901 .36776 2.4 86.4 0.72 0.470
Unmatched .08014 .36099 -72.0 -23.61 0.000 Public Job
Matched .08014 .08082 -0.2 99.8 -0.08 0.939
Unmatched .04705 .04629 0.4 0.13 0.899 Irresponsible Parents Matched .04705 .04539 0.8 -117.1 0.25 0.806
Unmatched .60445 .40487 40.7 14.24 0.000
Subemployee Matched .60445 .61421 -2.0 95.1 -0.62 0.534
Section 4. Conclusions and Policy Recommendations
• The actual system is based on four different laws, which are overlapped in some
aspects. As a result of that, the aim of the system is not clearly defined. We consider
important to emphasize the impact of the legal modifications on the significant
fluctuation experimented by the number of beneficiaries.
• The system which exists nowadays could be defined as a benefit depending on
tested means, which means that the benefit amount is given whether the
requirements are fulfilled without considering the educational achievement of the
beneficiary.
• The result obtained shows that “Asignaciones Familiares” does not have impact on
educational attainment of beneficiaries for the year 2005. In contrast with this result,
the cash transfer policy result was positive. This kind of policy, which attempts to
increase households’ investment in human capital, was applied by different Latin
American and Caribbean countries.
• The system has some functional problems, which are difficult to overcome. On one
hand, there are benefit administrators of the informal work market who probably
declare lower incomes than the ones they actually have. On the other hand, there are
people who are probably out of the system and should be in because of their socio-
economical features.
• The system is very complex as a result of the number of laws that rule it. The fact of
taking part of the new law means to fill in a new application form in order to continue
receiving the benefit.
• The group of the population who receive the benefit present different features
compared with the group which do not receive it. We observe the main difference
between each group on the wealth and the household income, which seems to be
reasonable because the ones who do not receive the benefit have lower economical
status.
• The benefit administrators receive bimonthly the benefit amount. This amount
represents a little proportion of the Basic Goods Set, as well as it also represents an
insignificant proportion of the Poverty Line. Whether we consider the inflation effect in
the evolution of the benefit amount, we appreciate loses in real terms in recent years.
• Finally, the expenditures of the system do not represent an important proportion of
the BPS expenses, as well as of the Public Social Expenditures and the Total Public
Expenditures.
Other authors, Bucheli (1997) and Vigorito (2004), conclude that “Asignaciones Familiares”
make a contribution to poverty alleviation. Bucheli (1997) also concludes that the school
attendance requirement limits the number of beneficiaries in low resources households as a
result of their high drop out rate of the educational system. Therefore there is a conflict
between the poverty alleviation target and the school attendance target. Moreover, Vigorito
(2004) considers that the system is concentrated on improving household incomes without
following the aim of keeping the children on the educative system. She also concludes that
“Asignaciones Familiares” does not contribute significantly to poverty reduction.
Nowadays, the Social Development Ministry (Ministerio de Desarrollo Social, Uruguay),
together with BPS, prepare a bill which will merge the different laws regulating this issue. We
think that a legal reform ought to redesign the system aiming to generate the necessary
incentives that make profitable for the families to invest in the education of their children. The
monetary incentive is the most important for these households because of their low
economical status. In these exists a trade off between sending them to school and sending
them to work, and this depends on the children’s age. After certain age that depends on
each particular economical situation, parents or tutors, consider more profitable to send their
children to work and not going to school.
The economic incentives that transfer policies give have to affect household’s trade off in the
sense of extending their human capital investment. The educational investment generates
positives externalities, specially the reduction of children work as a consequence of their
retention in the educational system. Therefore the benefit amount should increase in
accordance with school attainments to give the families an incentive to invest in education. It
also has to take into account the family income to avoid the not desirable effect of excluding
Low Resources Households. Future researches should analyze other factors, besides the
economic one, which affect the trade off decision work-education, and also the suitable
benefit amount.
References Andersen, L (2001): “Social Mobility in Latin America: Links with Adolescent Schooling”. Research Network Working Paper R-433, Inter-American Development Bank, July 2001. Banco de Previsión Social (2005): “Boletín Estadístico”. Asesoría Económica, BPS, Montevideo. Becker, O.S. and Ichino, A. (2004): “Estimation of average treatment effects based on propensity scores”. pp. 1-19.
Bertoni, R., Azar, P. and Silveira, M. (2006): “Identificación y análisis del Gasto Público Social en Uruguay”. October, 2006. Bucheli, M. (1997): “Equidad en las Asignanciones Familiares de Uruguay” LC/MV/R149, Montevideo, CEPAL. Cid, A., Presno, I. and Viana, L. (2004): “Institutions, Family and Economic Performance”. Revista de Ciencias Empresariales y Economía. Año III. 2004. Colin, A., and Trivedi, P.K (2005): “Microeconometrics: methods and applications”. Cambridge University Press. Handa, S. and Davis, B. (2006): “The Experience of Conditional Cash Transfers in Latin America and the Caribbean”. ESA Working Paper No. 06-07. May 2006. Heckman, J.J., Ichimura, H. and Todd, P. (1997): “Matching As An Econometric Evaluation Estimator”. Machin, S. (2004): “Education Systems and Intergenerational Mobility”. Draft Paper Prepared for CESifo/PEPG Conference, Munich, September 2004. Pérez del Castillo, S. (2004): "Manual Práctico de Normas Laborales” Edition nº 10. Rawlings, L.B. and Rubio, G.M. (2003): “Evaluating the Impact of Conditional Cash Transfer Programmes. Lessons from Latin America”. World Bank Policy Research Working Paper 3119, August 2003. Skoufias, E. and Parker, S.W. (2001): “Conditional cash transfers and their impact on child work and schooling: Evidence from the PROGRESA program in Mexico”. FCND Discussion Paper No. 123. Food Consumption and Nutrition Division. International Food Policy Research. October 2001. Vigorito, A. (2004): “Asignaciones familiares, distribución del ingreso y pobreza en el Uruguay. Un análisis para el período 2001-2004”. UNICEF, October 2005. Wooldridge, J.M. (2002): “Econometric Analysis of Cross Section and Panel Data” Massachusetts Institute of Technology.
APPENDIX
Wealth Index
∑ === 13i1i id
131
IndexWealth
Where di is a dummy variable. There are one binary variable for each comfort good which the ECH
consider in the survey. The survey consider thirteen comfort goods: hot water heater, electric tea
kettle, refrigerator, color television, cable TV service, VCR player/recorder, washing machine,
dishwasher, microwave, computer, internet connection, automobile for personal use, telephone
service. The dummy variable takes value 1 if the house has this good or service, and 0 otherwise.
Propensity Score Estimations
We use “pscore” command developed by Becker and Ichino (2002) in STATA software. Here is the
result of the estimation:
**************************************************** Algorithm to estimate the propensity score **************************************************** The treatment is Family Assignment Asign. | Familiares | Freq. Percent Cum. ------------+----------------------------------- 0 | 3,327 63.24 63.24 1 | 1,934 36.76 100.00 ------------+----------------------------------- Total | 5,261 100.00 Estimation of the propensity score Iteration 0: log likelihood = -3460.0114 Iteration 1: log likelihood = -2834.6528 Iteration 2: log likelihood = -2766.756 Iteration 3: log likelihood = -2758.4392 Iteration 4: log likelihood = -2757.9766 Iteration 5: log likelihood = -2757.973 Iteration 6: log likelihood = -2757.973 Logistic regression Number of obs = 5261 LR chi2(17) = 1404.08 Prob > chi2 = 0.0000 Log likelihood = -2757.973 Pseudo R2 = 0.2029 ---------------------------------------------------------------------------------------- Family Assignment | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+-------------------------------------------------------------------------- Sex | .0862989 .0654926 1.32 0.188 -.0420643 .2146621 Education | -.0600159 .0135564 -4.43 0.000 -.0865859 -.0334458 Private Education | -1.103291 .1365232 -8.08 0.000 -1.370872 -.8357106 Household Members >60 | -.1892374 .1076793 -1.76 0.079 -.4002849 .0218101
Domestic Service | 2.416161 .7446026 3.24 0.001 .9567671 3.875556 Biological Father | -.5297162 .2204675 -2.40 0.016 -.9618246 -.0976079 Father Education | -.0789878 .0143297 -5.51 0.000 -.1070735 -.050902 Father Experience | -.0058908 .0056468 -1.04 0.297 -.0169582 .0051767 Father-Professional | -1.34669 .2850693 -4.72 0.000 -1.905416 -.7879646 Father-Inactive | -.3447425 .1821063 -1.89 0.058 -.7016644 .0121793 Mother Education | -.083292 .0149226 -5.58 0.000 -.1125397 -.0540442 Mother Experience | -.0311728 .0066128 -4.71 0.000 -.0441335 -.018212 Mother-Professional | -.4388714 .1784506 -2.46 0.014 -.7886282 -.0891146 Mother-Inactive | -.0261581 .074186 -0.35 0.724 -.1715599 .1192438 Public Job | -1.656179 .0981924 -16.87 0.000 -1.848633 -1.463725 Irresponsible Parents | -.0550458 .1594606 -0.35 0.730 -.3675829 .2574912 Subemployed | .1281823 .0737382 1.74 0.082 -.016342 .2727065 Constant | 2.755889 .2422784 11.37 0.000 2.281032 3.230747 ---------------------------------------------------------------------------------------- Note: the common support option has been selected The region of common support is [.02316895, .81566826] Description of the estimated propensity score in region of common support Estimated propensity score ------------------------------------------------------------- Percentiles Smallest 1% .0276072 .0231689 5% .0510063 .0232448 10% .0838864 .0233693 Obs 4884 25% .1770681 .0234152 Sum of Wgt. 4884 50% .4408883 Mean .3948883 Largest Std. Dev. .218496 75% .5823016 .8054642 90% .6606876 .810742 Variance .0477405 95% .6983418 .8132967 Skewness -.2027905 99% .7547129 .8156683 Kurtosis 1.682612 ****************************************************** Step 1: Identification of the optimal number of blocks Use option detail if you want more detailed output ****************************************************** The final number of blocks is 9 This number of blocks ensures that the mean propensity score is not different for treated and controls in each blocks ********************************************************** Step 2: Test of balancing property of the propensity score Use option detail if you want more detailed output ********************************************************** The balancing property is satisfied This table shows the inferior bound, the number of treated and the number of controls for each block Inferior | of block | Family Assignment of pscore | 0 1 | Total -----------+----------------------+---------- .0231689 | 574 42 | 616 .1 | 658 103 | 761 .2 | 189 45 | 234 .25 | 112 58 | 170 .3 | 288 152 | 440 .4 | 366 294 | 660 .5 | 407 551 | 958 .6 | 354 686 | 1,040 .8 | 2 3 | 5 -----------+----------------------+---------- Total | 2,950 1,934 | 4,884
Note: the common support option has been selected ******************************************* End of the algorithm to estimate the pscore *******************************************
In this work propensity score was estimated by a logit model. To see the robustness of the Logit
estimation there are some test commonly used. Here we show its.
Correctly Classified Outcomes – Logit estimation of propensity scores for Family Assignment.
Area under ROC curve = 0.7858
Note: Sensitivity means the fraction of observed positive-outcome cases that are correctly classified;
specificity is the fraction of observed negative outcomes cases that are correctly classified. A model with no
predictive power has area 0.5, a perfect model has area 1.
lstat Measure – Logit estimation of propensity scores for Family Assignment. Logistic model for Family Assignment -------- True -------- Classified | D ~D | Total -----------+--------------------------+----------- + | 1240 764 | 2004 - | 694 2563 | 3257 -----------+--------------------------+----------- Total | 1934 3327 | 5261 Classified + if predicted Pr(D) >= .5 True D defined as asignacion_familiar != 0 -------------------------------------------------- Sensitivity Pr( +| D) 64.12% Specificity Pr( -|~D) 77.04% Positive predictive value Pr( D| +) 61.88% Negative predictive value Pr(~D| -) 78.69% -------------------------------------------------- False + rate for true ~D Pr( +|~D) 22.96% False - rate for true D Pr( -| D) 35.88% False + rate for classified + Pr(~D| +) 38.12% False - rate for classified - Pr( D| -) 21.31% -------------------------------------------------- Correctly classified 72.29% --------------------------------------------------
Hosmer-Lemeshow goodness-of-fit test – Logit estimation of propensity scores for Family Assignment. Number of observations = 5261
Number of groups = 20
Hosmer-Lemeshow (18) = 39.93 2χ
Prob > = 0.0021 2χ
After obtaining the “Correctly Classified Outcomes”, the “lstat Measure” and “Hosmer-Lemeshow”
goodness-of-fit test, we cannot reject our logit model for the propensity scores.