REVISTA INGENIO

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the

Year 2030

Tendencias Demográcas: Pronóstico de la Población Infantil de 0 a 3 años para el Año 2030

Gandy Rene López Fuertes | Ministerio de Inclusión Económica y Social, MIES -Ecuador

Guillermo Alexis Albuja Proaño | Universidad Central del Ecuador, UCE - Ecuador

Iván Cristian Naula Reina | Universidad Central del Ecuador, UCE - Ecuador

Jeimy Sugey Hernández Martínez | Universidad de Pueblos y Nacionalidades del Ecuador Amawtay

Wasi, UAW -Ecuador

Gustavo Napoleón López Fuertes | Instituto Superior Universitario Central Técnico, ISUCT - Ecuador

https://doi.org/10.29166/ingenio.v8i2.7545 pISSN 2588-0829

2025 Universidad Central del Ecuador eISSN 2697-3243

CC BY-NC 4.0 —Licencia Creative Commons Reconocimiento-NoComercial 4.0 Internacional ng.revista.ingenio@uce.edu.ec

      

    ,  (),  - , . -



Este trabajo calcula la cantidad de niños y niñas de 0 a 3 años en Ecuador que viven en condiciones de

pobreza, utilizando información combinada de la Encuesta de Condiciones de Vida (ECV) 2014 y el

Censo de Población y Vivienda (CPV) 2010, ya que los datos actuales no ofrecen suciente detalle por

regiones. Para ello, se emplea el método de Estimación para Áreas Pequeñas (SAE), basado en el modelo

Fay-Herriot, que utiliza regresión lineal múltiple para integrar datos de encuestas y censos. Este enfoque

permite analizar subpoblaciones especícas, como cantones o distritos, donde las encuestas por sí solas

no brindan resultados conables debido a muestras pequeñas. Además, se simula la pobreza basada en

el consumo y se proyecta la población infantil hasta el año 2030, comparando estas proyecciones con

estimaciones de la Comisión Económica para América Latina y el Caribe (CEPAL) para vericar su

exactitud. El propósito principal de esta investigación es ofrecer información precisa y detallada sobre la

infancia en pobreza, apoyando así la creación de políticas públicas efectivas en Ecuador.



is paper estimates the number of children aged 0-3 living in poverty in Ecuador, using combined data

from the 2014 Living Conditions Survey (LCS) and the 2010 Population and Housing Census (CPV), as

the current data do not provide sucient detail by region. To do so, we employ the Small Area Estimation

(SAE) method, based on the Fay-Herriot model, which uses multiple linear regression to integrate survey

and census data. is approach allows the analysis of specic sub-populations, such as cantons or districts,

where surveys alone do not provide reliable results due to small samples. In addition, consumption-based

poverty is simulated and the child population is projected up to 2030, comparing these projections with

estimates from the Economic Commission for Latin America and the Caribbean (ECLAC) to verify their

accuracy. e main purpose of this research is to provide accurate and detailed information on children in

poverty, thus supporting the creation of eective public policies in Ecuador [1].

Recibido: 18/11/2024

Recibido tras revisión: 27/12/2024

Aceptado: 24/3/2025

Publicado: 10/07/2025

 

Child poverty, child welfare, multiple

regression, Fay-Herriot model, demogra-

phic projections

 

Pobreza infantil, bienestar infantil, regre-

sión múltiple, modelo Fay-Herriot, pro-

yecciones demográcas.

1. Introduction ders the ght against child poverty, but could also aggra-

vate existing social and economic gaps in the country.

e lack of geographically disaggregated informa-

tion for certain population groups, such as the child

population, is a critical problem that directly aects deci-

sion-making in economic and social policies. In Ecuador,

this limitation prevents the eective targeting of resour-

ces to areas of greatest need, which can exacerbate re-

gional inequalities [2]. At the international level, similar

e absence of specic data on child poverty at the lo-

cal level in Ecuador is a critical barrier to designing and

implementing social policies that actually work. Without

knowing how many children aged 0-3 face poverty in

each canton or district, decision-makers are unable to

distribute resources eectively or create programs that

respond to the realities of the most neglected commu-

nities [1], [2]. is lack of detailed information not only

hin-

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

countries such as Perú and Colombia face similar cha-

llenges, where the lack of disaggregated data has hindered

the implementation of child poverty reduction programs

[3]. is study is relevant because it oers an innovative

methodological solution to overcome these limitations,

allowing for better planning and resource allocation. e

central research question is: How can the child population

in Ecuador be estimated according to poverty levels using

the Small Area Estimation methodology? is research

is novel in combining survey and census data to generate

accurate district-level estimates, which is crucial for the

design of eective public policies in similar contexts [4].

e problem of the scarcity of disaggregated data is

not unique to Ecuador. In several Latin American coun-

tries, this limitation has slowed progress in reducing po-

verty and inequality. For example, in Colombia, the lack

of accurate information on rural areas has caused rural

support programs to lose impact [5]. In México, the lack

of detailed statistics at the municipal level has made it di-

cult for health policies to eectively reach indigenous

communities [6]. ese cases show how the absence of

reliable data can hinder policy decisions in diverse con-

texts, highlighting the need for solutions such as the one

this study proposes.

What makes this study special is its innovative

approach in employing the Small Subpopulation Estima-

tion (SAE) methodology, based on the Fay-Herriot mo-

del and multiple regression, to accurately estimate child

poverty in specic regions of Ecuador. Unlike conventio-

nal methods, which typically rely on small sample sur-

veys, this technique combines information from sources

such as the 2014 Living Conditions Survey (LCS) and the

2010 Population and Housing Census (CPV), achieving

robust results even in places where direct data are scarce

[1]. is advance not only raises the quality of estimates,

but also provides a practical tool for policy to reach whe-

re it is most needed.

e main objective of this work is to determine how

many children aged 0-3 will be living in poverty in Ecua-

dor by 2030, with a detailed breakdown by canton and

district. e question guiding this research is: How many

children aged 0-3 will be living in consumption poverty

in 2030 in Ecuador, and how will this reality be distribu-

ted among the dierent regions of the country? Answe-

ring this question is key to guiding child development

policies and ensuring that eorts are concentrated in the

most critical areas.

e study of poverty and social injustice represents

a complex challenge due to the interaction of multiple

factors, elements and intrinsic variables that converge in

the same objective. is problem has been the subject of

much research and debate, which has proposed various

concepts and realities that demonstrate that an individual

cannot disassociate himself from a just environment that

will keep him out of poverty and allow him to overcome

the inequalities of his environment. e analysis of pover-

ty encompasses various aspects. Some of these aspects are

under the control of the individual, but most are inuen-

ced by social forces and economic and political power.

e dierences between theories and conceptualizations

are based on how the problem of poverty and vulnerabili-

ty is addressed. However, they all share the common goal

of achieving a more equitable and just society to combat

social inequality [7]. Approximately 700 million people

live on less than $2.15 a day, which is considered the ex-

treme poverty line. is situation remains prevalent in

parts of sub-Saharan Africa, fragile and conict-aected

regions, and rural areas. Aer decades of progress, the

rate of global poverty reduction began to slow in 2015,

coinciding with moderate economic growth. e Sustai-

nable Development Goal of eradicating extreme poverty

by 2030 still seems unattainable.

e COVID-19 pandemic and a series of major shoc-

ks between 2020 and 2022 dealt a severe blow to eorts

to reduce global poverty, setting back three years of pro-

gress. Low-income countries were hardest hit and have yet

to recover. In 2022, a total of 712 million people were li-

ving in extreme poverty worldwide, an increase of 23 mi-

llion compared to 2019. Poverty and inequality cannot be

reduced without also addressing interrelated global cha-

llenges, such as slow economic growth, fragility and con-

ict, and climate change.

Climate change is hampering poverty reduction and

poses a major threat to the future. e lives and liveli-

hoods of the poor are most exposed to climate-related

risks. Every year, millions of households fall into pover-

ty or become trapped in poverty due to natural disas-

ters. Higher temperatures are already causing declining

productivity in Africa and Latin America, and will fur-

ther reduce economic growth, especially in the poorest

regions of the world. Eradicating poverty requires addres-

sing its multiple dimensions. Countries cannot adequate-

ly address this problem without also improving people’s

well-being in a comprehensive manner, including more

equitable access to health, education, and basic infrastruc-

ture and services, including digital services. Policymakers

must intensify their eorts to grow their countries eco-

nomies in ways that create high-quality jobs and employ-

ment and protect the most vulnerable. Employment is the

surest way to reduce poverty and inequality. Its impact is

further multiplied in communities and across generations

by empowering women, girls and youth [8].

To further contextualize the relevance of our study, it

is important to recognize that the small area estimation

(SAE) methodology has been successfully applied in a va-

riety of geographical and socio-economic contexts to ad-

dress similar challenges related to measuring poverty and

inequality at the local level [4]. For example, research in

Latin American countries has used SAE to generate ac-

curate poverty estimates in rural and marginalized areas

López G. et al.

where census data are limited or outdated [9]. ese stu-

dies have shown that the SAE can be a valuable tool for

evidence-based decision-making and targeting of public

policies aimed at the most vulnerable population. By del-

ving deeper into this previous research, we can better un-

derstand the strengths and limitations of the SAE and

adapt its application to the specic case of Ecuador, in or-

der to obtain more accurate and relevant results.

e Small Area Estimation (SAE) method is a power-

ful tool that has been successfully used in dierent coun-

tries to measure poverty and inequality at the local

level. For example, in Latin America, recent studies have

applied this approach to obtain accurate poverty data in

rural and marginalized areas where census information is

scarce or old [10]. Similarly, in the European Union, SAE

is used to produce detailed regional statistics on income

and living conditions, helping governments to target their

policies more eectively [11]. In Canada, this method has

been used to estimate the incidence of chronic diseases in

indigenous communities, improving the allocation of re-

sources for health programs [12]. ese cases demonstrate

how SAE can be adapted to dierent contexts, providing

a solid basis for our analysis in Ecuador [1].

e identication of individuals in situations of vul-

nerability and poverty is the rst essential step for the

implementation, execution and monitoring of inclusion

programs. is is fundamental to advance in the creation

of opportunities that facilitate equality and social cohe-

sion, also considering the gender approach as a cross-cu-

tting element in any social policy [13]. Data from the VII

population and VI housing census, conducted in 2010, are

a key source for obtaining population information down

to the lowest level of territorial disaggregation, i.e. at the

district level. e variables that measure poverty through

Unsatised Basic Needs (UBN) and consumption allow us

to get closer to the social reality of individuals. However,

the use of the UBN poverty indicator has certain limita-

tions. Firstly, this type of poverty is based on a multidi-

mensional logic that emphasizes services such as drinking

water and sanitation, as well as housing conditions. Given

the current national coverage of these services, which is

low by international standards, lack of access may aect

households that are not necessarily in vulnerable condi-

tions. is can lead to an overestimation of the popula-

tion in need of care [14].

According to the 2010 Population and Housing Cen-

sus, the incidence of poverty measured by UBN was

60.1%, which represents the percentage of people in hou-

seholds that do not satisfy one or more basic needs. In

contrast, poverty measured by consumption was 25.8%

in 2014, indicating a percentage of people whose per ca-

pita consumption is below the poverty line. is suggests

that regardless of the method used to estimate poverty,

the Living Conditions Survey provides a more accurate

estimate. According to [15], poverty measured through

consumption is more aligned with the notion of vulne-

rability, as it depends mainly on the labor market and

available resources [13]. e dependence of consumption

poverty on factors such as labor demand and available re-

sources implies that any macroeconomic shock can aect

overall consumption levels among inhabitants, which has

a signicant impact on those in extreme poverty. Unlike

poverty measured by UBN, the latter is more related to

structural issues linked to the country’s infrastructure and

basic services. In the specic case of Ecuador, the com-

ponents related to sanitation and overcrowding have the

greatest inuence on the indicator [14].

e current government is implementing a program-

med to improve access to drinking water and sanitation,

complementing housing initiatives. When a household

has these basic services, it is automatically no longer con-

sidered poor; furthermore, the chances of falling back

into this condition are minimal because UBN poverty is

closely related to structural aspects [13]. is situation li-

mits the design of projects aimed at creating conditions

that foster both economic and social inclusion, which is

crucial for achieving greater equity among members of

society. erefore, it is necessary to measure poverty from

a perspective more focused on socio-economic compo-

nents such as consumption poverty [14].

However, the initial sources to obtain these indicators

present a maximum territorial disaggregation at the pro-

vincial and regional levels, which hinders eective plan-

ning at the cantonal and district levels. In this context, it

is proposed to implement methodologies such as Small

Area Estimation (SAE), which allows estimating popula-

tion parameters based on census data applied to popula-

tion subsets obtained through surveys [13]. SAE facilitates

obtaining parameters measured in surveys and transla-

ting them to census data with high population disaggre-

gation. is makes it possible to identify populations in

small areas using variables not available in the 2010 Cen-

sus but present in surveys such as the 2014 Living Con-

ditions Survey [14].

is methodological proposal makes it possible to put

into practice the fundamental concepts of the estimation

of small areas, which responds to an estimation of popu-

lation parameters. Consequently, we know that at present

our country in particular, does not have geographically

disaggregated information of the entire population in

general, and specically of a certain population group,

which is used for planning and decision making for eco-

nomic and social policy, as indicated by [4]. e identi-

cation of people in vulnerability and poverty is the rst

step for the implementation, execution and monitoring

of inclusion programs, with the aim of making progress

in the generation of opportunities that allow equality and

social cohesion to be achieved, in addition to the intrinsic

consideration of gender and ethnicity as a cross-cutting

core of any social policy, as mentioned by [16].

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

e data from the VII population census and VI housing

census of 2010, as a source from which population data

is obtained up to the lowest territorial disaggregation le-

vel, i.e. district. According to [17], the variables poverty

by Unsatised Basic Needs and poverty by consumption

allow an approximation to be made of the social reality

of individuals. e Small Area Estimation (SAE) me-

thodology is a mathematical and statistical method that

models information from one or more data sources to

produce estimates of a variable of interest, for example

poverty at the local level, as indicated by [18]. SAE is

most useful when domain (sub-population) sample sizes

are too small to provide adequate precision for estima-

tors. Consequently, the methodology (SAE) will allow

estimating at a disaggregated level the population of chil-

dren aged 0-3 years from 2014 to 2030, based on the Li-

ving Conditions Survey of 2014 and the VII Population

Census and VI Housing Census of 2010.

To fully understand the applicability and relevance of

Small Area Estimation (SAE) in our context, it is essential

to recognize its use and evolution in other countries and

regions. Globally, SAE has been successfully implemented

in a number of nations to address similar challenges rela-

ted to measuring poverty and inequality at the local level.

For example, in the European Union, the EAS is used to

generate detailed regional statistics on income and living

conditions, enabling member states to design more eec-

tive and targeted policies [19]. In Latin America, coun-

tries such as Brazil and México have used SAE to improve

the accuracy of poverty estimates in rural and marginali-

zed areas, where census data are limited or outdated [20].

e adoption of SAE is not only limited to pover-

ty measurement, but also extends to other areas such as

public health, education and agricultural development.

In Canadá, for example, SAE is used to estimate the pre-

valence of chronic diseases in indigenous communities,

facilitating the allocation of resources and the implemen-

tation of prevention programs [12]. In the African con-

text, SAE has proven valuable for monitoring sustainable

development indicators at the local level, allowing gover-

nments and non-governmental organizations to assess the

impact of their interventions and adjust their strategies

accordingly [21].

ese examples illustrate the versatility and potential

of SAE as a tool for evidence-based decision-making at

the local level. By adapting and applying SAE in our spe-

cic context, we can benet from lessons learned and best

practices identied in other countries and regions, whi-

le addressing our own unique needs and challenges. e

originality of our research lies in the combination of cen-

sus data and household surveys to estimate child pover-

ty at the small area level, which will allow us to generate

valuable information for planning and policy design for

the most vulnerable population.

Poverty and social inequality are complex issues that

involve a variety of factors and require comprehensive

approaches to their analysis. is study seeks to estimate

how many children aged 0-3 live in poverty in Ecuador,

using the Small Area Estimation (SAE) method to inte-

grate data from the 2014 Living Conditions Survey (LCS)

and the 2010 Population and Housing Census (CPV).

is is intended to generate detailed estimates at the can-

ton and district levels, providing valuable information for

designing public policies that address the country’s most

vulnerable groups [8].

In order to enrich the theoretical and methodological

framework of this study, a comprehensive review of re-

cent scientic literature on the application of Small Area

Estimation (SAE) in similar contexts has been conducted.

is review has identied relevant studies that address

specic methodological challenges, such as the selection

of appropriate regression models, the validation of as-

sumptions and the assessment of the sensitivity of esti-

mates to dierent scenarios [22]. Research exploring the

use of machine learning and big data techniques to im-

prove the precision of small area estimates and overcome

the limitations of traditional methods has also been inclu-

ded [23]. e incorporation of these recent studies helps

to strengthen the evidence base of our research and pla-

ce it in the context of more current debates on the mea-

surement and analysis of child poverty.

Recent studies support the value of the SAE methodo-

logy in similar situations. For example, an analysis by [8],

applied this technique to measure poverty in rural areas of

Latin America, conrming its accuracy when compared

to census data. At the national level, the National Institu-

te of Statistics and Census of Ecuador [23] has started to

use SAE to rene its ocial statistics, showing promising

results. ese experiences, both local and global, rein-

force the strength and importance of the approach this

study adopts.

is research has been based on an extensive review

of the scientic literature, both in the eld of Small Area

Estimation (SAE) methodology and in the analysis of

child poverty and demographic projections. roughout

the study, references are presented to support both the

methodology and the data used in this study.

For studies that have used SAE in other geographical

contexts, we recommend consulting the work of [24] in

the eld of public health, as well as the studies of [25] in

the context of agriculture. ese works provide concre-

te examples of how SEM can be applied in dierent areas

and with dierent data sources.

In terms of updated data sources, the most recent

World Bank and ECLAC reports on poverty and inequa-

lity in Latin America have been reviewed, as well as the

most recent population and housing censuses of the coun-

tries in the region. ese reports provide updated data on

López G. et al.

the socio-economic situation of the population and can be

used to validate and complement the results of this study.

Finally, for recent studies on population projections

and poverty, we have consulted the work of [26] on demo-

graphic trends at the global level, as well as the studies of

[27] on poverty and inequality in the world. ese wor-

ks provide a global perspective on demographic trends

and poverty, which can help contextualize the results of

this study.

2. Methodology

2.1 SMALL AREA ESTIMATION

Small Area Estimation (SAE) techniques are a set of sta-

tistical methods for estimating parameters for small sub-

populations. ese techniques are especially useful when

the subpopulation of interest is contained within a larger

study. In this context, the term “small area” generally re-

fers to a small geographical boundary, such as a provin-

ce, canton, parish or district. It may also encompass a

“small domain”, which refers to a specic demographic

group within a larger region [28].

When a survey is conducted at the national or state

level, the sample size in small areas may be insucient to

generate accurate estimates. To address this challenge, ad-

ditional data, such as census records, can be used, which

are available for these small areas and help to obtain more

reliable estimates [29]. e SAE methodology allows in-

formation to be obtained at levels of territorial disaggre-

gation that cannot be achieved by standard methods. is

is valuable as it provides reliable estimates on one or seve-

ral relevant variables in areas where existing information

is insucient to provide valid estimates [30]. All SAE me-

thods rely on auxiliary data available for sub-populations.

However, in many cases, the information may be limited

and only allow estimates to be made for larger geographic

areas or certain population subgroups [31].

Since not all poverty-related variables are collected in

a census that covers the entire population, it is essential to

resort to tools that provide more detailed information on

specic localities, such as surveys or administrative records

[32]. e availability of accurate data on living conditions

is essential for targeting policies and programs to reduce

poverty. Having adequate poverty estimates is crucial for

managing programs and allocating resources to local juris-

dictions [28]. Identifying vulnerable and poor populations

is essential for the design and implementation of programs

that promote economic and social inclusion. is aims to

generate opportunities that facilitate equality and social co-

hesion [30]. e SAE methodology is based on modelling

information from various sources to produce estimates on

variables of interest at the local level. It is particularly useful

when sample sizes in the domains are too small to provide

acceptable precision in the estimators [29].

According to [4], the Small Area Estimation (SAE)

methodology is a valuable technique for obtaining reliable

estimates for small areas, which are subpopulations with

little or no sample information. is methodology uses

mixed models, which combine information from the sam-

ple and from auxiliary sources, such as censuses or admi-

nistrative records. is makes it possible to obtain more

accurate and ecient estimators than direct estimators.

In addition, the SAE methodology oers several methods

to calculate the mean square error and condence inter-

vals, which are fundamental to measure the uncertainty

of small area estimators. In recent years, the SAE metho-

dology has incorporated advanced techniques, such as

machine learning, to improve the performance and e

xibility of small area estimators. ese techniques allow

adapting to changing conditions, and provide more accu-

rate estimates for small areas. However, it is important to

remember that these are only estimates and may be sub-

ject to errors and variations due to the uncertainty inhe-

rent in any estimation process.

e SAE methodology has a number of advantages

over traditional estimation methods. Firstly, SAE allows

more precise estimates to be obtained in small areas. Se-

condly, SAE allows information from dierent sources

to be incorporated, which can improve the precision of

the estimates. ird, SAE is a exible technique that can

be adapted to dierent types of data. All small area esti-

mation (SAE) methods rely on auxiliary data available at

the sub-population level, such as administrative records,

specic surveys or data from the latest census. is infor-

mation is essential to construct predictor variables that

are integrated into a statistical model, which is then used

to predict the values of the variable of interest in various

small areas [29].

In order to generate estimates using SAE, it is essen-

tial to take into account several factors. It is necessary to

assess whether there is a demand for such statistics, whe-

ther there is the commitment and willingness of an entity

to provide methodological support and expert sta, whe-

ther the available information is sucient, and whether the

auxiliary data and variables of interest are correlated [32].

e relevance of these aspects lies in the fact that, over

time, policy makers have increasingly started to use quan-

titative and qualitative information to design, implement

and monitor projects and programs of public interest.

National or regional information is no longer sucient,

especially in a context where decentralized public mana-

gement is sought. ere is a growing demand for data that

provide more detailed and specic territorial disaggrega-

tion; for example, it is not uncommon to request informa-

tion on specic population groups within certain parishes

or districts [33]. Currently, various sources of informa-

tion are used to adequately dene target populations in

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

order to direct concrete actions for the benet of these

communities. More detailed data therefore facilitates de-

cision-making and is essential for the design of targeted

public policies and realistic nancial planning [34].

e main source of information used remains the

census, as it provides a wide range of household data and

covers the entire population, allowing for well-disaggrega-

ted information. However, since the census is conducted

every ten years, many variables may lose relevance over

time [31]. On the other hand, there are surveys that co-

llect specic information more frequently; however, the-

se only allow estimates to be made for large geographic

areas or larger population groups.

Mechanisms need to be sought to generate more spe-

cic and disaggregated statistics. One option would be to

modify the design of surveys to meet information needs

or to increase the sample size of existing surveys, althou-

gh this may involve high costs. Another alternative would

be to collect information in smaller localities; however,

this would require each local government to nance the

process, which is not always feasible [10]. In this context,

estimation techniques such as SAE are gaining populari-

ty by providing statistical information at smaller territo-

rial levels. SAE produces estimates on variables of interest

using census data and other sources that provide more de-

tailed information on specic localities [29].

e eectiveness of SAE methods depends to a large

extent on the availability of suitable predictor variables

that are measured uniformly across the population. It is

crucial to analyses each variable used in the estimation

and select the most appropriate predictive model. In addi-

tion, the eective use of SAE methods requires a thorou-

gh assessment of model quality. Finally, when estimating

in smaller domains, it is important to validate their accu-

racy. Fortunately, there have been signicant advances in

the development of SAE models that minimize the mean

square error and provide more accurate estimates [29].

To ensure the transparency and reproducibility of our

study, it is important to provide additional details on the

statistical models used and the data sources selected. In

particular, the model for estimating consumption poverty

is based on multiple linear regression [35]. e selection

of the independent variables was made based on previous

scientic literature and the availability of data in the se-

lected sources, such as the 2010 Population and Housing

Census and the 2014 Living Conditions Survey [14]. e-

se data were carefully processed and validated to ensure

their quality and reliability.

e Small Area Estimation (SAE) methodology was

chosen because of its ability to overcome the limitations

of traditional estimation methods, which oen require

large samples to generate accurate estimates [4]. In con-

trast, SAE allows combining survey data with census in-

formation to obtain reliable estimates at the local level,

even in areas with small samples [35]. is methodology

is particularly useful when seeking to analyze variables

such as consumption poverty, which are not available in

censuses but are available in surveys such as the Living

Conditions Survey (LCS 2014) [14].

Econometric models, specically multiple linear re-

gression and generalized linear regression (GLS), were se-

lected to integrate data from the 2014 LCS and the 2010

Census of Population and Housing (CPV). ese models

allow for predicting variables such as per capita consump-

tion in small areas by relating them to demographic and

socioeconomic factors [9]. e choice of these techniques

is justied by their ability to handle multiple variables and

correct for possible correlations between them, which is

essential for obtaining accurate estimates of child pover-

ty [36].

e sample used was based on a probability sampling

design, ensuring representativeness of the population stu-

died. Small areas were selected based on geographic rele-

vance and data availability, allowing for adequate coverage

of the most vulnerable regions [84]. For data analysis, Sta-

ta soware was used, which oers advanced tools for im-

plementing complex statistical models [37].

Independent variables were selected based on their re-

levance in explaining consumption poverty. Demographic

variables such as age and gender were included, as well as

socio-economic variables such as household income and

parental education level [38]. Dummy variables, such as

membership in a specic ethnic group, were included to

capture signicant dierences in poverty among dierent

population groups [39].

2.2 METHODOLOGICAL ASPECTS OF SAE

Domains can be dened on the basis of characteristics

that segment the population into a set of sub-populations

that are mutually exclusive. Generally, these domains are

established according to geographical areas, such as sta-

tes, provinces, municipalities, regions, districts or metro-

politan areas, as well as by socio-demographic groups,

which may include categories such as age, gender or

ethnicity within a broader region [40]. Small area esti-

mation (SAE) is classied into two types of estimators:

direct and indirect. A direct estimator is based exclusi-

vely on sample information from a specic domain. is

type of estimator can incorporate known auxiliary infor-

mation, such as the values of an auxiliary variable X, X

that is related to the variable of interest Y, Y. However, a

disadvantage of direct estimators is that they can have

large standard errors, especially if the sample size in the

domain is small or non-existent [19], [41].

On the other hand, an indirect estimator uses values

of the variable of interest Y, Y from a related domain or

from a specic time period. is approach increases the

“eective” sample size and reduces the standard error. e

information used for these estimators is obtained from

López G. et al.

recent censuses or current administrative records. ere

are three main types of indirect estimators: the indirect

domain estimator, which uses values from another do-

main; the indirect time estimator, which uses values from

another period; and the estimator that combines both as-

pects, using values from another domain and another pe-

riod simultaneously [28].

2.3 GENERAL MATHEMATICAL MODEL FOR INDI

RECT ESTIMATION AT THE AREA LEVEL

In Small Area Estimation (SAE) [42], an area-level model

is used when the data from a survey, such as the Living

Conditions Survey (LCS 2014), is available for specic

areas but lacks sucient sample size to make reliable

estimates at a smaller geographic level (e.g., cantons,

districts). e area-level model combines these direct

survey estimates with auxiliary information available for

the entire population from sources like the Census of Po-

pulation and Housing (CPV 2010).

In [42], indicates the general mathematical model for

area-level SAE is based on the Fay-Herriot model, which

assumes the following structure:

(1)

Where:

is the direct estimate of the parameter of interest (e.g.,

poverty rate or consumption poverty) for area obtained

from the survey (LCS 2014).

is a vector of auxiliary variables for area (e.g., so-

cio-economic variables from CPV 2010).

is a vector of unknown regression coecients that re-

late the auxiliary variables to the area-level estimates.

is the random area eect, capturing the variation spe-

cic to area that is not explained by the auxiliary varia-

bles. It follows a normal distribution .

is the sampling error associated with the direct es-

timate , assumed to follow a normal distribution

2.3.1 Steps in the Model

Model specication: e Fay-Herriot model combines

direct estimates from a survey with auxiliary data, as-

suming a linear relationship between the two.

2. Error structure:

: e survey sampling error, which depends on the

sample size in each area.

: e area-level random eect, which accounts for

unexplained variability between areas.

e goal is to estimate the population parameter for each

area using both the survey data and the auxiliary data.

e empirical best linear unbiased predictor (EBLUP) is

oen used for this purpose.

2.4. ESTIMATION PROCESS FOR THE FAYHERRIOT

MODEL

Step 1: Linear model setup

e model assumes that for each area , the estimate

from the survey can be modeled as (1). e vector

contains the auxiliary variables (e.g., census data from

CPV 2010), and the regression coecients describe

the relationship between these variables and the survey

estimate .

Step 2: Estimation of using Generalized Least Squares

(GLS)

e regression coecients are estimated using Gene-

ralized Least Squares (GLS), accounting for the heteros-

cedastic errors across areas:

(2)

Where is the variance-covariance matrix for area ,

dened as:

(3)

The term is the variance of the area-specic eect

, and is the sampling error variance of the direct

survey estimate .

Step 3: Empirical Best Linear Unbiased Prediction

(EBLUP)

Once the regression coecients are estimated, the

EBLUP for the parameter of interest (e.g., poverty rate)

in area is given by:

(4)

Where:

(5)

is formula adjusts the direct estimate by incorpora-

ting the auxiliary data from , eectively combining the

direct survey data with census information to produce a

more reliable estimate at the area level.

2.4.1 Particular Model for LCS 2014 and CPV 2010

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

In your case, you have two data sources: Living Condi-

tions Survey (LCS 2014): is provides direct estimates

of consumption poverty but lacks sucient sample size

to give reliable estimates for smaller geographic areas

(e.g., cantons or districts).

Census of Population and Housing (CPV 2010): is

provides detailed auxiliary information at a highly disag-

gregated geographic level but does not contain direct me-

asures of consumption poverty.

Given this, the indirect method should be applied at

the area level because: LCS 2014 provides area-level es-

timates of poverty for larger areas (e.g., provinces), but

these estimates are subject to high variability due to sma-

ll sample sizes in some areas. CPV 2010 provides auxi-

liary data at smaller geographic levels (e.g., cantons or

districts), which can be used to improve the estimates

from LCS 2014. e particular model for your case would

look like this (1). Where:

is the direct estimate of consumption poverty for area

from LCS 2014.

is the vector of auxiliary variables (e.g., income le-

vels, housing characteristics) from CPV 2010.

is the vector of regression coecients to be estimated.

is the random area eect specic to area .

is the sampling error from the LCS 2014 estimate.

2.4.2 Model interpretation: Auxiliary data from CPV

2010 helps predict the consumption poverty rate for

smaller areas. Random eects () account for area-specic

deviations not explained by the census data. e EBLUP

estimator combines direct survey estimates from LCS

2014 with predictions based on census data from CPV

2010 to produce more accurate small-area estimates.

Decision: For your case, you should work at the area

level because the data from LCS 2014 provides estima-

tes for areas, while CPV 2010 provides auxiliary varia-

bles at a more detailed geographic level. By applying the

Fay-Herriot model, you can improve the reliability of the

consumption poverty estimates for small areas by com-

bining the survey and census data.

2.5 LOGARITHMIC FAYHERRIOT MODEL FOR INDI

RECT ESTIMATION

In [42], it is indicated that, in order to include the logari-

thmic consumption poverty model in the indirect small

area estimation (SAE) model, the formulation is slightly

changed. e (ln) transformation helps deal with skewed

distributions of the dependent variable (poverty by con-

sumption), making the model more robust and suitable

for handling extreme values. In this case, we model the

logarithm of the poverty by consumption variable from

the Living Conditions Survey (LCS 2014). e indirect

model at the area level is now written as follows:

(6)

is the log-transformed direct estimate of

consumption poverty for area from LCS 2014.

is the vector of auxiliary variables (e.g., socio-

economic indicators, housing conditions) from the

Census of Population and Housing (CPV 2010).

is the vector of regression coecients to be

estimated.

is the random area eect, which accounts for

unexplained variation specic to area . It follows a

normal distribution .

is the sampling error associated with the direct

estimate , assumed to follow a normal distribution

2.5.1 Estimation Steps with Logarithmic Model

Step 1: Linear Model in Log Space

e direct estimates from LCS 2014 are transformed by

taking the logarithm of the poverty by consumption esti-

mates (6). is model allows the log-transformed direct

estimate to depend linearly on the auxiliary data from

CPV 2010, with adjustments for random eects and

sampling error.

Step 2: Estimation of using GLS (Generalized Least Squares)

As in the standard Fay-Herriot model, the regression co-

ecients are estimated using Generalized Least Squares

(GLS). However, now we are working with the log-trans-

formed estimates:

(7)

Where is the variance-covariance matrix for area i, de-

ned as:

(8)

is accounts for both the variability between areas

(through ) and the sampling error (through ).

Step 3: Empirical Best Linear Unbiased Prediction

(EBLUP) in Log Space

e EBLUP for the log-transformed poverty by con-

sumption in area is given by:

(9)

Where:

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

(10)

is predictor combines the direct estimate of log-trans-

formed poverty by consumption with predic-

tions based on auxiliary data from CPV 2010, adjusted

for area-specic random eects.

Step 4: Back-Transformation to Original Scale

To obtain the estimate of poverty by consumption on the

original scale (non-logarithmic), the EBLUP estimates

are back-transformed:

(11)

is back-transformation gives the nal estimate of po-

verty by consumption for area .

Particular Model for Log-transformed Consumption Po-

verty in Ecuador

For your case with LCS 2014 and CPV 2010, the loga-

rithmic Fay-Herriot model applied to poverty by con-

sumption would look like this:

is the log-transformed estimate of consumption

poverty for area from LCS 2014.

represents a set of auxiliary variables from CPV 2010,

such as income, education, employment, housing condi-

tions, and other socio-economic factors.

is the vector of coecients to be estimated.

captures random area eects for area i, reecting varia-

tions not explained by the auxiliary data.

is the sampling error of the direct estimate from LCS 2014.

By taking the logarithm of the poverty by consump-

tion variable, you ensure the model handles skewness and

extreme values more eectively. e auxiliary data from

CPV 2010 allows the model to make reliable predictions

at smaller geographic levels (cantons, districts) where LCS

2014 alone would be unreliable due to small sample sizes.

e back-transformation ensures that the nal estimate

is returned to the original scale of poverty by consump-

tion aer tting the model.

is model improves the estimates of consumption

poverty at smaller geographic areas by leveraging the de-

tailed auxiliary information available from the census

(CPV 2010), while correcting for small sample sizes in

the survey (LCS 2014) through the log transformation

and random eects adjustments.

2.6 APPLYING THE SAE METHODOLOGY IN ECUADOR

In the study by [43], he used the advanced statisti-

cal technique known as Small Area Estimation (SAE),

summarized in Figure 1, to analyze consumption and

poverty in specic areas of Ecuador. is technique is es-

pecially useful when trying to make precise estimates in

small geographic areas where sample sizes are too small

to provide reliable estimates.

For the application of the SAE in [43], we used the Indi-

rect method called Multiple Regression Estimator at the

geographic area level of the natural logarithm of pover-

ty by consumption, whose mathematical formulation is

described in the previous section. In this initial process

we used the Guide for the use of databases of the Natio-

nal Institute of Statistics and Census [44], this involved

several important steps to analyses and eectively use

data from two main sources: the Living Conditions Sur-

vey (LCS) 2013 - 2014, and the VII Population Census

and VI Housing Census (CPV) 2010.

Step 1: Data collection was conducted using the Li-

ving Conditions Survey (LCS) for the years 2013-2014.

is survey provided a wide range of information on

the living conditions of the inhabitants of Ecuador, co-

vering aspects such as income, expenditure, education

and health, among others. In parallel, data from the VII

Figure 1.

SAE scheme

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

Population Census and VI Housing Census (CPV) con-

ducted in 2010 were used. In addition, dummy variables

were generated for each category. Dummy variables are

binary variables that represent categories or groups in a

statistical analysis. For example, a dummy variable could

be used to code gender, assigning a value of 1 for “male”

and 0 for “female”. ese variables allow categorical in-

formation to be included in the analysis, which generally

only accepts numerical data [45].

Dummy variables are fundamental in statistical analy-

sis, as they allow the inclusion of qualitative variables in

models that require numerical data. ese variables take

binary values (0 or 1) and are used to represent specic

attributes or categories within the data set. For example,

when analyzing the impact of gender in a regression mo-

del, a dummy variable can be created to indicate whether

an individual is female (1) or not (0), in [22]. is makes

it easier to interpret how categorical dierences aect the

dependent variable in the model [46]. In summary, the

creation of dummy variables is a crucial step in statisti-

cal analysis that allows for the integration of qualitative

information into quantitative models. is approach not

only improves the analytical power of the model, but also

allows for more precise conclusions about how dierent

categories inuence the variables of interest [47].

Step 2: e homogenization of the databases was ca-

rried out in this process. In [24], he explains that reconci-

liation involves aligning or correlating data from dierent

sources so that they are comparable with each other. In

this case, the aim was to identify common variables be-

tween the Living Conditions Survey (LCS) and the Po-

pulation and Housing Census (CPV). is means that

variables that were present in both databases were selec-

ted for further analysis.

Standardization is an essential step in data analysis, as

it ensures that information extracted from dierent sour-

ces is consistent and comparable. By identifying common

variables between the LCS and the LCP, joint analysis of

the data is facilitated, leading to more robust and infor-

med conclusions about living conditions in Ecuador.

Once homogenization is complete, aggregation at the

household level follows, followed by a critique of the data.

is step is crucial in any analysis, as it involves checking

the data for errors, inconsistencies or outliers that may

distort the results. In this case, outliers were removed and

the database was cleaned, preparing it for the next phase

of the process [48]. Table 1 presents the homologous va-

riables obtained. In addition, in [43], the code necessary

to carry out the homologation and aggregation of varia-

bles at the household level can be found.

or explanation. In this analysis, the dependent variable

was consumption poverty, while the independent varia-

bles were demographic and socio-economic in nature. It

should be noted that only variables included in both da-

tabases were considered, as presented in Table 1 [24]. Si-

multaneously, econometric models were developed with

data from both sources to predict per capita consumption.

Econometric models are fundamental statistical tools in

economics that allow for the examination of relationships

between dierent variables. In this case, the variables se-

lected for the models were similar to those used in the VII

Population and VI Housing Census (2010), which ensu-

red consistency in the analysis across data sources [49].

Once the models were developed, they were applied

to census data, allowing for predictions of per capita con-

sumption in specic geographic areas. ese predictions

were then used to calculate the consumption poverty rate

in the districts of interest. In this case, poverty was deter-

mined by considering as poor those individuals whose per

capita consumption was below a pre-established threshold

[43] study provided a clear picture of consumption po

verty in specic areas of Ecuador. By employing the Small

Area Estimation (SAE) technique, accurate estimates were

achieved even in areas with small sample sizes, highligh-

ting the importance of advanced statistical techniques in

economic and social studies [48].

In summary, the process allowed us to extract valua-

ble information from the Living Conditions Survey (LCS)

and the Population and Housing Census (CPV), facilita-

ting a better understanding of consumption poverty at

the household level in Ecuador. Subsequently, econome-

tric models were estimated using Ordinary Least Squa-

res (OLS) and Generalized Least Squares (GLS), following

the SAE methodology. In addition, heteroskedasticity is-

sues were analyzed using non-constant variance models,

as recommended in the literature [40]. Finally, these para-

meters were applied to the census data, allowing us to pre-

dict consumption and construct the consumption poverty

rate in small areas, robust to the level of the 365 Ecuado-

rian districts [49].

Step 4: Coding and application of the SAE metho-

dology was carried out. is process began with the use

of the Living Conditions Survey (LCS) standardized at

household level. From this database, an explanatory mo-

del was created for each province, modelling the natu-

ral logarithm of per capita consumption. is approach

is common in econometric analysis, as it allows transfor-

ming the data to make it more manageable and reduce

the inuence of extreme values on the results [49]. Sub

sequently, we integrated the analysis of heteroskedastici-

ty, a common feature in economic data, which refers to

the non-constant variation of errors in a regression mo-

del. is step was essential to t the models appropriate-

ly, as heteroscedastic errors can inuence the precision of

Step 3: Aer standardizing and cleaning the two da-

tabases at the household level, the dependent and inde

pendent variables were identied. Dependent variables

are those variables that we seek to predict or explain, whi-

le independent variables are used to make that prediction

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

N° Variable Name De s crip t ion Azu a y Bolívar Cañar Carchi Cotopaxi Chimborazo El Oro Esmeraldas Guayas Imbabura Loja Los Ríos Manabí

Morona

Santiago

Nap o Pastaza Pichincha Tungurahua

Zamora

Chinchipe

Galápagos Sucumbíos Orella na

Santo Domingo

de los

Tsáchilas

Santa Elena

Constant - 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

2Analfabeto

Head of household is illiterate 110111011110001101101000

3analfabeto_cl

Percentage of heads of household who cannot read and write at cluster level 1 0 1 1 1 0 1 1 1 0 0 1 0 1 1 0 0 1 1 0 11 0 0

4cocina_ex

The household has a kitchen exclusively for the use of the household. 100000111001000010000010

5cocina_ex_cl

Percentage of households that have a kitchen exclusively for household use at cluster level 0 0 1 1 0 11000001110101011 0 1

6edad_p

Average household age 1 1 1 0 0 1 1 0 1 1 0 0 1 0 1 0111010 0 0

7edad_p_cl

Average household age at cluster level 1 1 1 0 1 1 0 1 1 0 0 1 1 0 1 1 0 1 0 0 1 1 1 1

8edad2_

Age squared of the household 0 1 0 0 0 0 1 1 1 0 1 0 0000000010 0 0

9edad2 cl

Age squared average household age at cluster level 101001100011111001110100

10 edad2_p

Average age of the household squared 1 0 1 1 1 1 1 1 1 1 0 0 1 0 1 0 1 1 1 0 1 1 1 1

11 edad2_p_cl

Mean squared household age at cluster level 111011001011011011111110

12 h01

Number of bedrooms in the dwelling 1 1 1 1 1 1 1 0 1 1 1 1 1 1 0 1 1 1 0 10 1 1 1

13 h01_cl

Average number of bedrooms in the dwelling at the cluster level 1 1 1 1 0 1 1 0 1 0 0 0 0 0 1 1 100000 0 1

14 h04_2

The shower is shared with other households 0 1 0 0 0 0 0 0 0 1 0 0 0 1 0 0 1 1 0 0 0 0 0 0

15 h04_3

No shower 1 1 1 1 1 0 1 1 1 0 1 0 1 001111000 1 0

16 h15_3

You own the house and you are paying for it 0000101110001010110000 1 0

17 h15_4

Housing is owned and fully paid for 1 0 1 1 0 0 1 0 0 0 0 1 0 0 0 0 1 1 0 1 0 0 1 0

18 h15_5

The hous e is ceded 000100100000001000100000

19 h15_6

Housing is received by services 0 0 0 1 0 1 1 0 0 0 0 0 1 0 0 0 0 1 0 10 0 1 0

20 internet

The household has internet service 111111101111100111110111

21 internet_cl

Percentage of households having internet service at cluster level 1 1 1 1 0 1 0 0 1 0 1 0 1 1 0 0 1 0 0 0 1 0 0 0

22 migra_d

Household received money from friends or relatives abroad in the last 12 months 0 1 0 0 0 0 1 0 1 0 1 1 0 0 00000011 0 1

23 migra_d_cl

Percentage of households that in the last 12 months received money from friends or relatives abroad at cluster level 0 0 0 0 1 1 0 0 0 0 0 0 0 1 0 0 0 0 1 01 0 0 0

24 mu je r

If the head of household is female 0 0 1 0 0 1 0 0 0 0 0 10010010001 1 0

25 mujer_cl

Percentage of female heads of household by cluster 010000010000010001101010

26 nivel_2

Level of education of the head of household: GBS 0 0 1 0 0 1 1 0 0 0 0 1 0 0 0 1 1 0 0 10 0 1 1

27 nivel_3

Level of education of the head of household: incomplete secondary education 0010011010111001110100 1 1

28 nivel_4

Level of education of the head of household: medium 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 1 1

29 nivel_5

Level of education of head of household: non-university tertiary education 1 0 1 0 0 1 1 0 1 1 0 1 0 1 1 0 111110 1 0

30 nivel_6

Level of education of the head of household: university education 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 011 1 1

31 no_empleo

Head of household is unemployed 0 0 0 1 1 1 0 0 0 0 1 0 1000000001 1 0

32 no_empleo_cl

Percentage of unemployed household heads 111001100110000101101000

33 p03

Age of head of household 0 1 0 1 0 0 1 0 1 0 1 0 0 0 1 0 0 0 0 010 1 0

34 p03_cl

Average age of heads of households at cluster level 1 0 1 0 1 1 1 0 0 1 1 1 1 1 1 1 0 1 1 001 0 1

35 p16_2

Ethnicity of head of household: Afro-descendant 1 0 1 1 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 00 0 0 0

36 p16_3

Ethnicity of head of household: Montubio 1 0 1 1 0 0 1 0 0 0 0 0 0 1 0 1 0 00100 0 0

37 p16_4

Ethnicity of the head of household: mixed race 0 0 1 0 1 0 0 1 0 1 0 0 0 1 1 0 0 1 1 000 0 0

38 p16_5

Ethnicity of head of household: white or other 1 0 0 0 0 1 0 1 0 1 1 0 0 1 11010011 0 0

39 p34_2

Marital or conjugal status of the head of household: common-law union 0010011000110001100011 0 0

40 p34_3

Marital or conjugal status of the head of household: separated 0 0 0 0 0 1 0 0 0 1 0 0 1 0 1 0 0 0 0 000 0 1

41 p34_4

Marital or conjugal status of the head of household: divorced 0 0 0 0 0 0 0 0 1 1 0 1 1 0 0 0 0 1 0 000 0 0

42 p34_5

Marital or conjugal status of the head of household: widowed 0 1 0 0 0 1 0 110000000110000 0 0

43 p34_6

Marital or conjugal status of the head of household: single 0 0 0 1 0 0 0 1 0 0 0 0 0 0 1 1 0 0 0001 1 1

44 pa_ad

Su itab le ho u s in g wall materials 1 0 1 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 1 000 1 1

45 pa_ad_cl

Percentage of dwellings with adequate wall materials at cluster level 1 0 1 1 1 1 0 0 0 0 0 0 0 0 1 0 0 0 0 000 0 0

46 pi_ad

Su itab le flooring materia ls 1001110001100000010000 0 0

47 pi_ad_cl

Percentage of dwellings with adequate flooring materials at cluster level 0 0 1 1 0 1 0 1 1 0 0 01000101000 0 0

48 rat_dp

Male/Female Ratio at household level 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 111 1 1

49 rat_dp_cl

Average Male/Female Ratio at cluster level 1 0 0 0 0 1 0 0 1 0101100010000 0 0

50 rural

Rural areas 0 0 0 1 0 0 1 0 1 0 1 1 1 0 0 0 1 0 1 010 0 1

51 T_am

Number of persons over 65 per household 0 0 1 1 0 1 1 0 1 0 0 1 1 1 1 1 0 0 1 110 0 0

52 T_am_cl

Percentage of people aged 65+ at cluster level 1 0 1 1 0 1 1 0 1 1 0 0 1 1 0 0 010010 0 0

53 T _ me

Number of persons aged 5-17 per household 1111111111111101111111 1 1

54 T_me_cl

Percentage of households with 5–17-year-olds at cluster level 1 0 1 1 1 0 1 1 1 0 1 0 1 110101011 0 0

55 T_nn

Number of persons under 5 per household 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 111 1 1

56 T_nn_cl

Percentage of households with persons older than 5 years old at cluster level 0 1 1 1 1 0 0 0 1 0 0 0 0 0 1 1 0 1 0 011 0 0

57 T_pcd

Number of persons with disabilities per household 11 1 0 1 0 1 0 1 1 1 0 1 0 0 0 1 1 0 011 1 0

58 T_pcd_cl

Percentage of households with persons with disabilities at cluster level 0 0 1 0 0 1 0 0 1 0 0 0 0 1 0 0 0 1 1 000 0 0

59 T_per

Number of persons per household 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 11111 1 1

60 T_per_cl

Average number of people at cluster level 10 0 1 011010000011101011 0 1

61 te_ad

Adequate roofing materials 0 0 0 0 0 0 0 0 0 0 0 0 1 1 01000001 0 1

62 te_ad_cl

Percentage of dwellings with adequate roofing materials at cluster level 0 1 0 0 1 1 0 0 1 0 0 0 0 1 0 0 0 0 0 011 0 0

63 telef_f

Household has conventional telephone service 1 1 1 1 1 1 1 1 1 1 1 11010111111 1 1

64 telef_f_cl

Percentage of households with conventional telephone service at cluster level 00 0 0 0 1 1 1 0 0 0 1 1 0 0 0 1 1 1 001 0 0

65 tvcable

The household has cable television service 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 111 1 1

66 tvcable_cl

Percentage of hous eholds with cable TV service at cluster level 0 0 0 1 0 0 1 1 0 0 0 0 1 1 0 0 0 0 1000 0 0

67 v08_2

The water supply is piped outside the house but on the lot. 0 1 1 1 1 1 1 0 0 1 0 0 1 0 1 1 1 1 0 0 11 1 1

68 v08_3

Water supply is piped outside the dwelling, lot or land 0 0 0 0 1 0 0 0 0 0 0 0 1 0 1 0110010 0 0

69 v08_4

Water supply is not piped but received by other means. 1 0 1 0 1 1 0 0 0 0 0 1 1 1 1 1 1 1 0 111 1 1

70 v09_2

Ty p e of t oilet fa cilitie s : toilet an d s ept ic ta nk 0 0 1 0 1 0 0 0 0 0 0 0 0101101000 0 0

71 v09_3

Type of toilet facilities: toilet and cesspool 00 0 1 1 0 0 1 0 0 0 0 0 1 0 0 1 1 1 00 1 0 0

72 v09_4

Ty p e of t oilet: latrine 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 1 0 0 010 0 0

73 v09_5

Type of toilet: none 0 0 0 1 0 1 0 0 0 1 0 1 0 1 0 0 0 0 1 011 0 1

74 v12a

Number of energy-saving light bulbs used by the household 1111111 1 1 1 1 1 1 1 1 1 1 1 1 111 1 1

75 v12a_cl

Average number of saved light bulbs used by households at the cluster level 1 0 1 1 0 1 1 1 0 1 0 1 0 1 1 0 1 0 1 0 1 0 1 1

76 vap_2

Main access to the house: cobblestones 0 0 0 0 0 0 0 1 0 1 1 1 1 1 0 0 1 1 1 0 1 1 1 0

77 vap_3

Main access to the dwelling: ballasted/dirt road 111000 1 0 1 1 1 0 0 0 1 0 1 0 0 00110

78 vap_4

Main access to the dwelling: pathway/walkway/other 1 1 1 0 0 0 0 0 1 1 1 0 0 0 0 0 1 0 0 1 0 1 0 0

79 vap_5

Main access to the dwelling: river/sea/lake 1 0 0 0 0 00 0 1 0 0 1 1 0 1 0 00 0 0 1 1 0 0

80 vtv_2

Type of housing: flat in house or building 0 0 0 0 1 0 1 1 1 0 1 1 0 0 0 1 1 0 0 0 1 0 1 1

81 vtv_3

Type of housing: room in a tenement house 101000 0 0 0 0 0 0 0 0000 0 0 1 1 0 0 0

82 vtv_4

Type of dwelling: half-hut, hut or hut 00 0 0 0 1 0 0 0 0 0 0 1 0 0100 0 0 0 0 00

83 vtv_5

Type of dwelling: ranch 0 0 0 10 0 0 0 1 0 1 1 1 0 0 0 0 0 0 0 0 0 1 1

Source: [15]

Table 1.

Indicators and/or standardized variables included in the methodology

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

the estimated coecients and thus the reliability of per ca-

pita consumption predictions [48].

e output of these routines was a matrix of estima-

ted coecients, which was subsequently used to estimate

consumption poverty in Ecuador’s Census of Population

and Housing CPV database. is step allowed us to apply

the models tted in the LCS to the CPV, thus generating

more accurate predictions about the distribution of con-

sumption poverty at the provincial level, which is key for

planning eective public policies in the country. e co-

ding used in [43] is shown below:

* --------------------------------------------------------------- *

* Small Sample Estimation (SAE)

* Consumption Poverty Estimation from LCS 2014 in CPV

2010 * *

* General Estimation Syntax *

* *-------------------------------------------------------------- *

clear all

set more o

set matsize 11000

set maxvar 30000

set seed 1234

/* -------------------------------------------------------------- *

Declare Programme in Mata for GLS estimation

* -------------------------------------------------------------- */

mata

void sigma(real scalar N, real scalar Ncl){

//*** 1. Construct the Omega matrix for GLS with cluste-

red errors.

Si=J(N, N, 0)

M=st_matrix(“max”)

m=st_matrix(“min”)

sigma2_h=st_matrix(“sigma2_h”)

sigma2_cl=st_matrix(“sigma2_cl”)

for (cl=1; cl<=Ncl; cl++) {

m_=m[cl,1]

M_=M[cl,1]

for (i=m_; i<=M_; i++) {

for (j=m_; j<=M_; j++) {

if (j==i){

Si[j,i]=sigma2_h[i,1];

}

if (j!=i){

Si[j,i]=sigma2_cl[i,1];

}

//*** 2. Estimation v’a GLS

y=st_matrix(“y”)

X=st_matrix(“X”)

// Estimation of coecients

inv_Si=invsym(Si)

b_gls=invsym(X’*inv_Si*X)*X’*inv_Si*y

// Variance-Covariance Coecient Variance Matrix Esti-

mation

V_gls=invsym(X’*inv_Si*X)

// Adjustment Estimation (R^2)

e=y-X*b_gls

sse=e’e

iota=J(N,1,1)

y_=y-iota*mean(y)

sst=y_’y_

R2=1-sse/sst

// Matrix of Coecients and Standard Errors

se = sqrt(diagonal(V_gls))

report=(b_gls, se, b_gls:/se)

// Extract matrices back to Stata

st_matrix(“report”, report)

st_matrix(“R2”, R2)

st_matrix(“b_gls”, b_gls)

st_matrix(“V_gls”, V_gls)

}

end

/* ---------------------------------------------------------------*

Call database

* ---------------------------------------------------------------*/

use “C:/SAE_2/1. Databases /base_hogares.dta”, clear

forv d=1/24{ /* Start of loop for provinces */

preserve

*local d=1

keep if dominio==`d’

/* ---------------------------------------------------------------*

Denitions

* ---------------------------------------------------------------*/

*gen ln_con=ln(cpcf)

* Declare dependent variable (dep) and independent va-

riable (ind)

global dep ln_con

global ind T_per T_nn T_me T_am T_pcd rat_dp

p03 edad2_ ///

edad_p edad2_p rural

p16_2-p16_5 mujer nivel_2-nivel_6 p34_2-p34_6 analfa-

beto no_empleo ///

migra_d vap_2-vap_5

v09_2-v09_5 h01 v08_2-v08_4 h04_2-h04_3 telef_f inter-

net tvcable h15_3-h15_6 v12a ///

te_ad pa_ad pi_ad vtv_2-vtv_5

cocina_ex T_per_cl-cocina_ex_cl

* Declare cluster level

gen cluster=ciudad

/* ---------------------------------------------------------------*

Estimation of the Consumption Model for each domain

* -------------------------------------------------------------- */

*** 1. Initial adjustment by OLS

reg $dep $ind, vce(r)

* Eliminate omitted variables in OLS (Must run more than

once for some provinces)

global ind: colnames(e(b))

global ind= subinstr(“$ind”, “_cons”, “”, .)

forv r=1/100 {

foreach x of global ind {

local a = substr(“`x’”,1,2)

if “`a’” == “o.” {

global ind = regexr(“$ind”, “`x’”, “”)

}

* Selecting independent variables with 10% signicance or

less using a backward stepwise

sw, pr(.10): reg $dep $ind, vce(r)

keep if e(sample)

* Extract variables used eectively

global ind: colnames(e(b))

global ind= subinstr(“$ind”, “_cons”, “”, .)

* Create matrix X with actually used variables including

a column of 1’s

gen c=1

global ind2 “${ind} c”

*** 2. Decompose errors between parts of each household

and cluster.

scalar K=colsof(r(table))

predict res, residual

egen res_cl=mean(res), by(cluster)

gen sigma2_cl=res_cl^2

gen res_h=res-res_cl

gen e2=res_h^2

mkmat $dep, mat(y)

mkmat $ind2, mat(X)

/* ---------------------------------------------------------------*

Estimation of the Heterocedasticity Model for each domain

* ---------------------------------------------------------------*/

qui sum e2

scalar A=1.05*r(max)

gen het=ln(e2/(A-e2))

reg het $ind, vce(r)

scalar K_a=colsof(r(table))

scalar sigma_r=e(rss)/(e(N)-K_a)

predict xb if e(sample), xb

gen B=exp(xb)

gen sigma2_ch=(A*B/(1+B))+0.5*sigma_r*(A*B*(1-B)/

((1+B)^3))

gen sigma2_h=sigma2_ch+sigma2_cl

* Output matrices

mat a_het`d’=e(b)

mat a_V`d’=e(V)

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

/* ---------------------------------------------------------------*

Estimation of the Model v’a GLS

* ---------------------------------------------------------------*/

*** 1. Construction of estimated sigma matrix of variance

covariance of errors **Diagonal elements equal to varian-

ce of each household + variance of the cluster component

**O-diagonal elements equal to variance of the cluster

component within the cluster

** 1.1 Initial denitions *Order the clusters

egen ncl=group(cluster)

sort ncl

*Identify the number of clusters

qui sum ncl

local Ncl=r(max)

*Identify the size of each cluster

bysort cluster: gen s_cl=_N

*Identify the starting position of each cluster

gen m_cl = 1 if ncl == 1

forval po = 2 / `Ncl’ {

local i = `po’ - 1

qui sum s_cl if ncl == `i’

local max1 = r(max)

qui sum m_cl if ncl == `i’

local max = `max1’+r(max)

replace m_cl = `max’ if ncl == `po’

}

*Set upper limit for each cluster

gen M_cl=m_cl+s_cl-1

*Minimum and Maximum Matrices to ll clusters in the

variance-covariance matrix

qui mean m_cl, over(ncl)

mat min=e(b)’

qui mean M_cl, over(ncl)

mat max=e(b)’

** 1.2 Filling in the estimated sigma matrix

mkmat sigma2_cl, mat(sigma2_cl)

mkmat sigma2_h, mat(sigma2_h)

local N=_N

** 1.3 Estimation of the declared function in Mata

mata sigma(`N’, `Ncl’)

** 1.4 Output matrices

mat b`d’=b_gls’

mat V`d’=V_gls’

mat dom=J(K,1,`d’)

mat mod`d’=(dom, report)

mat rownames mod`d’=$ind2

mat R2_=(nullmat(R2_)\(`d’,R2))

mat N=(nullmat(N)\(`d’,_N))

mat msigma_r=(nullmat(msigma_r)\(`d’,sigma_r))

mat mA=(nullmat(mA)\(`d’,A))

**1.5 Extract cluster variance matrices

collapse (mean)sigma2_cl, by(cluster)

cd “C:/SAE_2/3. Estimacion/Resultados”

save “eta_`d’.dta”, replace

/* ---------------------------------------------------------------*

Simulations of coecients

* ---------------------------------------------------------------*/

/* Create 100 random vectors of coecients for each pro-

vince averaging the vector estimated by GLS (or OLS in

case of the alpha model) and their respective variance-co-

variance matrices. */

** Alpha Model (Heteroscedasticity Model)

local d1= colsof(a_het`d’)

drawnorm a_het1-a_het`d1’, n(100) mean(a_het`d’) co-

v(a_V`d’) clear

local i=0

macro drop inda

foreach x of global ind2 {

local i = `i’+1

rename a_het`i’ a_`x’

global inda “$inda a_`x’”

}

gen dominio = `d’

gen sim = _n

reshape wide $inda, i(dominio) j(sim)

temple alpha_`d’

save `alpha_`d’’

**Beta Model (Model of the mean)

drawnorm b1-b`d1’, n(100) mean(b`d’) cov(V`d’) clear

local i=0

macro drop indb

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

foreach x of global ind2 {

local i = `i’+1

rename b`i’ b_`x’

global indb “$indb b_`x’”

}

gen dominio = `d’

gen sim = _n

reshape wide $indb, i(dominio) j(sim)

temple beta_`d’

save `beta_`d’’

restore

} /* Loop closure for provinces */

* Create a single coecient le for all 24 provinces aer

running the loop.

global var “alpha beta”

foreach x of global var {

use ``x’_1’, clear

forval p = 2 / 24 {

append using ``x’_`p’’

}

cd “C:/SAE_2/3. Estimation/Results “

save `x’, replace

}

cd “C:/SAE_2/3. Estimation/Results “

use beta, clear

merge 1:1 dominio using alpha

drop _m

order dominio, rst

cd “C:/SAE_2/3. Estimation/Results “

save “sim_b.dta”, replace

*Crear matriz de etas por cluster

use eta_1, clear

forval p = 2 / 24 {

append using eta_`p’

}

cd “C:/SAE_2/3. Estimation/Results “

save “eta_T.dta”, replace

* Create matrix of variance parameters per household

clear

svmat msigma_r

svmat mA

drop mA1

rename msigma_r1 dominio

rename msigma_r2 sigma_r

rename mA2 A

save “A_sigma.dta”, replace

**Extract model matrices, t levels and sample sizes

mat2txt2 R2_ N using Ajuste.txt, replace

mat2txt2 mod1 ///

mod2 ///

mod3 ///

mod4 ///

mod5 ///

mod6 ///

mod7 ///

mod8 ///

mod9 ///

mod10 ///

mod11 ///

mod12 ///

mod13 ///

mod14 ///

mod15 ///

mod16 ///

mod17 ///

mod18 ///

mod19 ///

mod20 ///

mod21 ///

mod22 ///

mod23 ///

mod24 using modelos.txt, replace

Step 5: Corresponding simulations were carried out

using the standardized Census of Population and Hou-

sing (CPV) base together with the matrices and coe-

cients obtained in Step 4. ese simulations allow for the

inherent variability in the models, and are a common

approach in econometrics to assess the robustness of es-

timates. In this case, 100 simulations were run to estima-

te per capita consumption poverty at the household level,

which is essential to obtain a more accurate and detailed

picture of the distribution of poverty [50].

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

Each of these 100 simulations generated a dierent es-

timate of consumption poverty, capturing possible uc-

tuations in the data and ensuring that the predictions are

more representative of reality. ese multiple simulations

approach is useful when working with estimates in econo-

metrics, as it allows for more reliable results to be obtai-

ned by averaging the dierent simulations, thus reducing

uncertainty in the models [51]. From these estimates, the

population of children aged 0-3 living in consumption po-

verty at the household level was selected. is step was

key to identifying the most vulnerable groups and targe-

ting more eective public policies. e precise estimates

obtained through this process allow the design of targe-

ted interventions that have a direct impact on improving

the living conditions of the child population [43]. e co-

ding is presented below:

* ----------------------------------------------------------- *

* Small Sample Estimation (SAE) *

*Estimation of Consumption Poverty from LCS 2014 in

CPV 2010 *

*Simulations in the CPV

* ----------------------------------------------------------------*

clear all

macro drop _all

set more o

set matsize 11000

set maxvar 30000

set seed 1234

/* -------------------------------------------------------------- *

Create simulation bases

* ---------------------------------------------------------------*/

* For speed of calculation, each simulation of coecients

needs to be separated (100 bases are created - 1 for each

simulation).

cd “C:/SAE_2/3. Estimation/Results “

use sim_b, clear

forval y = 1 / 100 {

foreach x of varlist b_T_per1 - a_te_ad100 {

local v : var label `x’

tokenize “`v’”

if `y’ == `1’ {

global var “$var `2’`1’”

}

preserve

keep dominio $var

cd “C:/SAE_2/3. Estimacion/Temp”

recode b_* a_* (.=0)

save sim_`y’.dta, replace

restore

macro drop var

}

/* ---------------------------------------------------------------*

Call CPV database

* ---------------------------------------------------------------*/

use “C:\SAE_2\1. Databases \base_trabajo_cpv.dta”, clear

*sample 10

gen cluster=ciudad

*Demographic composition of the household

egen T_per=sum(numpers), by(id_hogar)

gen menor=(p03>=5 & p03<=17)

egen T_nn=sum(p_5), by(id_hogar)

egen T_am=sum(p_65), by(id_hogar)

egen T_me=sum(menor), by(id_hogar)

egen T_h=sum(num_hom), by(id_hogar)

egen T_m=sum(num_muj), by(id_hogar)

gen rat_hm=T_h/T_m

gen rat_dp=(T_nn+T_me+T_nn)/T_per

egen edad_p=mean(p03), by(id_hogar)

gen edad2_p=edad_p^2

gen edad2_=p03^2

gen pcd=(p08==1)

egen T_pcd=sum(pcd), by(id_hogar)

* Ethnic recoding

recode p16 (2/4=2)

* Recoding level of education

gen nivel=0 if p23==1

replace nivel=1 if p23>=2 & p23<=4

replace nivel=1 if p23==5 & p24<=3

replace nivel=1 if p23==6

replace nivel=2 if p23==5 & p24>3 & p24<6

replace nivel=2 if p23==7 & p24<3

replace nivel=3 if p23==5 & p24==6

replace nivel=3 if p23==7 & p24==3

replace nivel=4 if p23==8

replace nivel=5 if p23==9 | p23==10

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

label dene nivel 0”Ninguna” 1”EGB” 2”Media Incomple-

ta” 3”Media” 4”Superior no Universitaria” 5”Universita-

ria”

label val nivel nivel

* Identication of suitable housing materials

gen te_ad=(v01>=1 & v01<=4)

gen pa_ad=(v03>=1 & v03<=4)

gen pi_ad=(v05>=1 & v05<=5)

* Keeping only the head of household and household in-

formation

keep if p02==1

*** Create dummies for each variable to create the X ma-

trix of the GLS model.

foreach v in p16 nivel p34 vap v09 v08 h04 h15 vtv {

qui tab `v’, gen(`v’_)

}

* Correct ethnicity others for few remarks. Joins whites

replace p16_5=1 if p16_6==1

* Correct housing access others for few remarks. Joins *a

sendero

replace vap_4=1 if vap_6==1

* Correct antichresis housing tenure for few observations.

Joins *a arriendo

replace h15_1=1 if h15_2==1

* Correct dwelling type shack and hut for few observations.

Joins *a mediagua

replace vtv_4=1 if vtv_6==1 | vtv_7==1

gen rural=(urp==2)

gen mujer=(p01==2)

gen analfabeto=(p19==2)

gen no_empleo=(p27==2)

gen migra_d=(m1==2)

gen telef_f=(h07==1)

gen internet=(h09==1)

gen tvcable=(h11==1)

gen cocina_ex=(h02==1)

* Cluster level average variables

foreach v in T_per T_nn T_me T_am T_pcd rat_dp eda-

d_p edad2_p p03 edad2_ mujer ///

analfabeto no_empleo migra_d h01 tele-

f_f internet tvcable v12a ///

te_ad pa_ad pi_ad cocina_ex {

egen `v’_cl=mean(`v’), by(ciudad)

}

clonevar dominio=provincia

*Keep only variables to be used

keep cluster id_hogar dominio T_per T_nn T_me T_am

T_pcd rat_dp p03 edad2_ ///

edad_p edad2_p rural p16_2-p16_5 mujer ni-

vel_2-nivel_6 p34_2-p34_6 analfabeto no_empleo ///

migra_d vap_2-vap_5 v09_2-v09_5 h01

v08_2-v08_4 h04_2-h04_3 telef_f internet tvcable

h15_3-h15_6 v12a ///

te_ad pa_ad pi_ad vtv_2-vtv_5 cocina_ex T_

per_cl-cocina_ex_cl

*Identifying the independent (explanatory) variables

global ind T_per T_nn T_me T_am T_pcd rat_dp p03

edad2_ ///

edad_p edad2_p rural p16_2-p16_5 mu-

jer nivel_2-nivel_6 p34_2-p34_6 analfabeto no_empleo ///

migra_d vap_2-vap_5 v09_2-v09_5

h01 v08_2-v08_4 h04_2-h04_3 telef_f internet tvcable

h15_3-h15_6 v12a ///

te_ad pa_ad pi_ad vtv_2-vtv_5 cocina_

ex T_per_cl-cocina_ex_cl

*Keep only remarks with complete information.

foreach v in $ind{

drop if `v’==.

}

order dominio cluster id_hogar, rst

/* ---------------------------------------------------------------*

Simulations

* ---------------------------------------------------------------*/

* Run each of the 100 simulations

forv s=1/100{

disp `s’

preserve

cd “C:/SAE_2/3. Estimacion/Temp”

destring dominio, replace

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

* Call the normalisation parameters of heteroscedasticity

per domain

merge m:1 dominio using “C:/SAE_2/3. Estimacion/Re-

sultados/A_sigma.dta”

drop _m

* Call eta estimates (cluster-level heteroscedasticity)

merge m:1 cluster using “C:/SAE_2/3. Estimacion/Resul-

tados/eta_T.dta”

drop if _m==2

drop _m

egen sigma2_cl_=mean(sigma2_cl), by(dominio)

replace sigma2_cl=sigma2_cl_ if sigma2_cl==.

* Call the model coecients Alpha and Beta for each si-

mulation.

merge m:1 dominio using sim_`s’

keep if _m==3

drop _m

* Estimate the logarithm of per capita consumption accor-

ding to the beta model (ln_con_d)

*Estimate the heteroscedasticity according to the alpha

model (B_)

gen ln_con_d=b_c`s’

gen B_=a_c`s’

foreach v of varlist $ind {

di “`v’”

replace ln_con_d=ln_con_d + `v’*b_`v’`s’

replace B_=B_ + `v’*a_`v’`s’

}

gen B=exp(B_)

gen sigma2_ch=(A*B/(1+B))+(0.5*sigma_r*(A*B*(1-B)/

((1+B)^3)))

* Randomly estimate assuming normality the error com-

ponents (at household level u_h and cluster level u_c).

gen u_h=rnormal(0, sqrt(sigma2_ch))

gen u_c=rnormal(0, sqrt(sigma2_cl))

* Bringing together the three components of estimated per

capita consumption

egen ln_con=rsum(ln_con_d u_h u_c)

gen con_pc`s’=exp(ln_con)

keep id_hogar con_pc`s’

* Store a baseline of the results for each household of the

100 simulations of the estimated per capita consumption.

cd “C:/SAE_2/3. Estimation/Results “

if `s’==1{

save cpv_con_.dta, replace

}

else{

merge 1:1 id_hogar using cpv_con_.dta, nogen

save cpv_con_.dta, replace

}

restore

}

/* ---------------------------------------------------------------*

Poverty Estimation

* ---------------------------------------------------------------*/

cd “C:\SAE_2\1. Databases “

use “base_trabajo_cpv.dta”, clear

gen id_sector=substr(id_hogar,1,12)

merge m:1 id_sector using “id_hogar-circuitos.dta”

drop if _m==2

drop _m

keep id_hogar izodec idist ciudad urp p01 p03 p08

cd “C:/SAE_2/3. Estimacion/Resultados”

merge m:1 id_hogar using “cpv_con_.dta”

keep if _m==3

drop _m

gen canton=substr(ciudad, 1, 4)

gen prov_=substr(ciudad, 1, 2)

*sample 2

*Dene the poverty line of the LCS 2014

scalar lp=83.22367

* Estimate for each household 100 poverties according to

its estimated per capita income in each of the 100 simu-

lations.

forv s=1/100{

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

gen pobre`s’=(con_pc`s’<lp)

}

* Creating care populations

gen cnh=(p03>=0 & p03<=3)

gen cibv=(p03>=1 & p03<=3)

gen joven=(p03>=15 & p03<=29)

gen ad_m=(p03>=65 & p03!=.)

gen pcd=(p08==1)

gen nac=1

*Create planning area with prex Z

tostring izodec, replace

gen z=”Z”

egen zonap=concat(z izodec)

* Create province with prex P

cap tostring prov_, replace

gen p=”P”

egen prov=concat(p prov_)

keep pobre* cnh cibv joven ad_m pcd nac zonap idist can-

ton prov

* Create poverty for each population for each simulation

foreach p in cnh cibv joven ad_m pcd nac {

foreach d in nac zonap idist canton prov {

foreach stat in mean sum {

preserve

collapse (`stat’) pobre* if `p’==1,

by(`d’)

cd “C:/SAE_2/3. Estimation/Re-

sults “

temple `p’_`d’_`stat’

rename `d’ id

capture tostring id, replace

forv i=1/100{

rename pobre`i’ po-

bre_`p’_`i’

}

save “``p’_`d’_`stat’’”, replace

restore

}

foreach stat in mean sum {

foreach p in cnh cibv joven ad_m pcd nac {

use “``p’_nac_`stat’’”, clear

foreach d in zonap idist canton prov {

append using “``p’_`d’_`stat’’”

}

cd “C:/SAE_2/3. Estimacion/Temp”

save `p’_`stat’.dta, replace

}

foreach stat in mean sum {

cd “C:/SAE_2/3. Estimacion/Temp”

use cnh_`stat’, clear

foreach p in cibv joven ad_m pcd nac {

append using `p’_`stat’

}

cd “C:/SAE_2/3. Estimation/Results “

collapse (sum) pobre*, by(id)

save “pobreza_res_`stat’.dta”, replace

}

/* ---------------------------------------------------------------*

Output les

* ---------------------------------------------------------------*/

foreach stat in mean sum {

cd “C:/SAE_2/3. Estimation/Results “

use “pobreza_res_`stat’.dta”, clear

reshape long pobre_cnh_ pobre_cibv_ pobre_jo-

ven_ pobre_ad_m_ pobre_pcd_ pobre_nac_, i(id) j(s)

collapse (mean) cnh=pobre_cnh_ cibv=pobre_cibv_ jo-

ven=pobre_joven_ ad_m=pobre_ad_m_ pcd=pobre_pcd_

nac=pobre_nac_ ///

(sd) cnh_sd=pobre_cnh_ cibv_sd=pobre_cibv_ joven_

sd=pobre_joven_ ad_m_sd=pobre_ad_m_ pcd_sd=po-

bre_pcd_ nac_sd=pobre_nac_, by(id)

gen l=length(id)

sort l id

drop l

foreach u in cnh cibv joven ad_m pcd nac{

gen `u’_cv=`u’_sd/`u’*100

}

export excel using “C:/SAE_2/3. Estimation/Re-

sults/Base_nal_`stat’.xls”, rstrow(variables) replace

}

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

2.7 LOGARITHMIC FAYHERRIOT MODEL FOR IN

DIRECT ESTIMATION AND ITS RELATION TO THE

MULTIPLE LINEAR REGRESSION MODEL FOR THE

LOGARITHM OF CONSUMPTION POVERTY

e Fay-Herriot model [42], is a methodology used for

indirect estimation in small areas, where sample size is

limited. It is generally applied in situations such as cen-

suses or national surveys, where data are scarce in cer-

tain geographic areas or sub-populations. e model

assumes that direct estimates of a target variable for a

small area can be improved by using auxiliary variable

information and empirical models. From the above syn-

tax, the Fay-Herriot model is identied mainly in the

estimation structure using the generalized least squares

(GLS) method and in the way variance components are

handled. e model is presented in the following form:

2.7.1 Fay-Herriot mathematical model (logarithmic).

In [31], the Fay-Herriot model for small area estimation

is given in (6).

2.7.2 Parameter estimation

In the syntax provided, generalized least squares (GLS)

methods are used to estimate the coecients and the va-

riances of the model components (heteroscedasticity and

cluster variance). is is observed in the following code

fragment the application of the formula for the estimated

coecients by GLS is: , donde is the variance-covariance

matrix of the errors, which in this case has been estima-

ted by separating the variance of each household and the

errors at the cluster level.

// Coecient estimation using GLS

inv_Si = invsym(Si)

b_gls = invsym(X’*inv_Si*X)*X’*inv_Si*y

Here, the inversion of the variance-covariance matrix is

performed that this represented as ( ) and it is estimated

, the tted coecients of the model.

2.7.3. Decomposition of errors

e Fay-Herriot model takes into account the decompo-

sition of errors into two components:

Within-household variance : individual variance for

each unit within an area or cluster.

Inter-cluster variance : variance between the dierent

areas or clusters. is treatment is evident in the syn-

tax with the creation of the variance matrices:

gen sigma2_cl = res_cl^2

gen res_h = res - res_cl

gen e2 = res_h^2

2.7.4 Logarithmic model

e dependent variable modelled is the natural logari-

thm of per capita consumption. is is stated with:

gen ln_con = ln(cpcf)

is suggests that the Fay-Herriot model is being tted

to a log-transformed dependent variable. Model sum-

mary, the model implemented is a GLS-adjusted Fay-He-

rriot model, considering heteroskedasticity and variance

eects at the cluster level, and applying a log transfor-

mation to the dependent variable to model per capita

consumption in small areas. In mathematical terms, the

model can be expressed as:

(12)

Where the coecients are estimated and variances

(cluster) and (household) using generalized least squa-

res, and an adjustment for heteroskedasticity is included

using , the diagonal variance-covariance matrix.

2.8 RESULTS OBTAINED FROM THE APPLICATION

OF THE FAYHERRIET MODEL IN THE PARAMETER

ESTIMATION AND THE ADJUSTED MODEL

When applying the Fay-Herriot model in parameter es-

timation, it is essential to verify that the assumptions of

the model are met to ensure the validity and reliability

of the results. e Analysis of Variance - ANOVA, in-

dicates that the Fay-Herriot model has proven to be a

valuable tool in the analysis of complex data, particularly

in contexts where estimation for small areas is crucial.

As shown in table 2, the variability decomposition shows

that the sum of squares of the regression reaches a value

of 9,634.1391, reecting the portion of variability explai-

ned by the model. With 82 degrees of freedom, the mean

square of the regression stands at 117.490, indicating a

considerable explanation of the variability in the data

[42]. On the other hand, the residuals present a sum of

squares of 4,698.2966, with 27.696 degrees of freedom,

suggesting a considerable sample size and a mean squa-

re of residuals of 0.16963809, a value that evidences the

accuracy of the model in minimizing estimation errors

[24]. e R² of approximately 0.67 (67.2%) underlines the

relevance of the included auxiliary variables, as they ex-

plain more than two-thirds of the variability in the data,

which is remarkable in small-area estimation contexts

where data are oen sparse and of higher variability [27].

is model not only provides reliable estimates, but also

oers a solid basis for evidence-based decision-making,

especially in situations where precision in geographically

dened areas is essential for social and economic inter-

ventions [52]. Table 3 (annexes) presents the estimation

of the consumption poverty parameters obtained by the

Fay-Herriot model.

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

e following graph presents the kernel density es-

timate of the model residuals together with a theoretical

normal distribution for comparison. e solid curve re-

presents the estimated distribution of the observed resi-

duals, while the dotted line shows a normal distribution

with the same mean and standard deviation.

Regarding the shape of the curve, it is observed that

the kernel estimate closely resembles a normal distri-

bution, which is a good sign for the Fay-Herriot model.

However, there are some interesting details:

e higher part of the kernel curve looks a bit atter

compared to the theoretical normal distribution.

e tails of the kernel curve seem to spread out a bit

more, suggesting that there might be more extreme va-

lues than we would expect in a standard normal dis-

tribution.

e bandwidth used in the kernel estimation is 0.0457,

which determines how much the data is smoothed. A

smaller bandwidth would make the curve follow the raw

data more closely, while a larger one would produce a

smoother curve.

In summary, the residuals of the Fay-Herriot model

are approximately normally distributed, which is essen-

tial for the model to be valid. e similarity between the

kernel curve and the theoretical normal distribution in-

dicates that the residuals comply quite well with the as-

sumption of normality. However, the small dierences in

the shape of the curve suggest that the residuals might

have a slightly leptokurtic distribution, i.e. with heavier

tails than those of a standard normal distribution. Al-

though this slight deviation from normality is noticea-

ble, it does not invalidate the model completely [40], [53].

Source: [15]

In general, most of the points align quite well with the refe-

rence line, suggesting that the residuals follow an approxi-

mately normal distribution. is alignment is especially

noticeable in the central range of the data, where the resi-

duals t well to the expected theoretical distribution.

However, at the extremes of the distribution, particu-

larly at the bottom le of the graph, a slight deviation of

the points from the straight line is observed. is devia-

tion can be interpreted as an indication that the residuals

have slightly heavier tails than those of a standard normal

distribution, which is consistent with the previous obser-

vation of a leptokurtic distribution in the kernel density

curve analysis.

In conclusion, the residuals of the Fay-Herriot mo-

del reasonably comply with the assumption of normali-

ty, which is fundamental for the validity of the statistical

inferences of the model. Although there is a slight trend

towards heavier tails, this deviation is not sucient to in-

validate the model. However, it may be benecial to per-

form a more detailed analysis of the residuals to identify

possible outliers or patterns not captured by the current

model. Overall, the model proves adequate for the data

analyzed, providing a solid basis for estimates and predic-

tions in small area contexts [40], [54].

Source: [15]

Source of vari-

ation

SS Df. MS

Regression 9,634.1391 82 117,490

Residues 4,698.2966 27696 0.16963809

Total 14,332.4357 27778 0.51596356

Number of obs.

27.779

F(82, 27696)

692.59

Prob > F

0,0000

0.6722

R2 – Adjusted

0.6712

Root MSE

0.41187

Table 2.

ANOVA - Analysis of Variance

Source: [15] Figure 3 presents a visual comparison between the quan-

tiles of the residuals of the Fay-Herriot model and the

corresponding quantiles of a theoretical normal distri-

bution. In this graph, each point represents an observed

residual, while the straight line shows the normal distri-

bution expected under the assumption of normality.

Figure 2.

Kernel density estimation

Figure 3.

Distribution of Residuals for the Fay-Herriot Model

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

Figure 4, which presents the residuals versus tted va-

lues, provides a detailed insight into the behavior of the

Fay-Herriot model. e residuals, which represent the

dierences between the observed values and those pre-

dicted by the model, are randomly distributed around

the reference line (zero), which is a positive sign. is

dispersion around zero suggests that the model is not

systematically biased, i.e. it does not tend to consistently

underestimate or overestimate the true values.

Homoscedasticity, i.e. the constancy of the variance

of the residuals over the tted values, is clearly seen in the

graph. is feature is fundamental in regression models,

as violations of this assumption could indicate problems

in the specication of the model or the need for transfor-

mations in the variables.

Furthermore, the absence of discernible patterns in

the distribution of residuals (such as systematic curves or

trends) indicates that the model has adequately captured

the relationship between the explanatory variables and the

dependent variable. is suggests that the auxiliary varia-

bles included in the model are relevant and that the func-

tional form of the model is appropriate.

Finally, no extreme residuals are identied that could

distort the results, which is important for the robustness

of the estimates. Taken together, these features of the re-

sidual plot versus tted values support the validity of the

Fay-Herriot model for these data, fullling the assump-

tions necessary for statistical inference to be reliable in

small-area estimation contexts [40], [54].

Source: [15]

Table 3 (Annexes) presents a GLS-adjusted Fay-Herriot

model obtained in [43], with the logarithm of consump-

tion as the dependent variable. Household characte-

ristics and their impact on consumption are analyzed,

showing estimated coecients, standard errors, t-values,

and condence intervals for each independent variable

and the intercept. e table shows that consumption is

inuenced by various socio-demographic and economic

characteristics, providing a broad picture of the deter-

minants of welfare in terms of household consumption.

From the results obtained, a thorough and detailed

analysis was carried out which considers the following:

Household size and composition: Household size and

composition, in particular the number of children aged

0-3 years, have signicant eects on consumption po-

verty. As the number of dependents increases, be they

children or older adults, per capita consumption tends

to decrease, suggesting greater economic pressure on

larger households or households with more depen-

dents.

·Education: Education emerges as a key driver of con-

sumption, with higher levels of education associated

with higher levels of consumption, highlighting the im-

portance of education in the ght against poverty.

Technological and housing conditions: Ownership of

technological goods such as xed telephone, internet,

and cable TV is positively related to consumption, re-

ecting a higher standard of living in households with

access to these technologies.

Gender: Households where the reference person is fe-

male tend to have slightly lower consumption, which

could point to dierences in economic opportunities

between men and women.

Regional dierences: Households in rural areas have

slightly higher consumption compared to urban areas,

although this eect is small, suggesting that geographi-

cal dierences are not as pronounced in this sample.

Additionally, Table 4 and Map 1, presents the goodness of

t R² of the econometric models used through the SAE

methodology for each of the domains analyzed. e R² of

the models, as a whole, are above the 50% recommended

in the literature [42], [55]. Next, we include the values co-

rresponding to the coecient of determination - goodness

of t, which is obtained in the estimation of small areas at

the National level where we observe that: e R² measures

what percentage of the variability in the dependent varia-

ble (e.g. consumption) is explained by the independent

variables included in the Fay-Herriet Model. is is in the

multiple regression model [42]. An R² value close to 1 in-

dicates that the model explains the variability of the data

well. A value close to 0 indicates that the model does not

explain the variability well. In the table, the R² values vary

from 0.60 (Carchi) to 0.82 (Napo), indicating that the mo-

del has dierent levels of explanatory power depending on

the region. A higher R² suggests that the model is more

appropriate for that region.

In addition to this the number of observations (n) re-

fers to how many data were used to estimate the model

in each domain. Regions with more observations, such as

Guayas (2973) and Pichincha (2880), tend to have more

robust models because the sample is larger. Regions with

fewer observations, such as Morona Santiago (611) or

Galápagos (556), might have less precise estimates due

to the smaller sample size. Consequently, we look at the

Figure 4.

Residuals versus Fitted Values in the Fay-Herriot Model

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

quality of the model by region: Where Regions with good

t: Napo (R² = 0.82), Morona Santiago (R² = 0.79), Pasta-

za (R² = 0.75), Orellana (R² = 0.76). e model explains

more than 75% of the variability in these regions, indi-

cating that the model is highly predictive. Regions with

moderate t: Azuay (R² = 0.70), Chimborazo (R² = 0.69),

Loja (R² = 0.67), among others. e model performs well

in these regions, but there could still be factors not captu-

red by the model. Regions with the lowest t: Carchi (R²

= 0.60), El Oro (R² = 0.62), Guayas (R² = 0.61), Galápagos

(R² = 0.62). Although the model still explains a signicant

proportion of the variability, there is room for improve-

ment in these regions, perhaps by considering additional

variables or adjusting the model.

Adicionalmente, en el anexo se incluyen resultados

obtenidos

Dominio R2Number of observations

AZUAY 0.70 1929

BOLÍVAR 0.63 855

CAÑAR 0.66 923

CARCHI 0.60 858

COTOPAXI 0.63 1089

CHIMBORAZO 0.69 1117

EL ORO 0.62 1950

ESMERALDAS 0.66 1037

GUAYAS 0.61 2973

IMBABURA 0.72 1051

LOJA 0.67 1139

LOS RÍOS 0.62 1219

MANABÍ 0.67 1231

MORONA SANTIAGO 0.79 611

NAPO 0.82 681

PASTAZA 0.75 570

PICHINCHA 0.68 2880

TUNGURAHUA 0.67 1151

ZAMORA CHINCHIPE 0.67 682

GALÁPAGOS 0.62 556

SUCUMBÍOS 0.67 735

ORELLANA 0.76 652

SANTO DOMINGO DE LOS TSÁCHILAS 0.65 976

SANTA ELENA 0.66 914

Source: [15]

It is crucial to recognize that child poverty is not a

one-dimensional phenomenon, but is inuenced by a

complex interaction of socio-economic and demogra-

phic factors. While our analysis has focused on variables

such as consumption and unmet basic needs, it is impor-

tant to consider the possible inuence of other factors

not considered that could aect children’s vulnerability.

For example, parental education, especially that of the

mother, has been shown to be a key factor in reducing

child poverty, as it inuences employment opportunities

and household income [21].

In addition, access to basic services such as quali-

ty health, nutrition and education plays a critical role in

child development and long-term poverty prevention

[56]. e availability and quality of these services can vary

signicantly across regions and communities, which may

explain some of the observed dierences in child pover-

ty levels. In addition, factors such as domestic violence,

child abuse and lack of access to a safe and stimulating

environment can have a negative impact on children’s we-

ll-being and increase their risk of falling into poverty [57].

For future research, it would be advisable to explo-

re the inclusion of these additional variables in the small

area estimation model in order to gain a more complete

and accurate understanding of the factors contributing to

child poverty. is may require additional data collection

through targeted surveys or the use of administrative data

sources such as health and education records. By broade-

ning the scope of our analysis, we will be able to identify

more eective and targeted interventions to break the cy-

cle of poverty and ensure a better future for all children.

3. Results and discussion

As mentioned in the methodological section, once the

multiple linear regression models were estimated for

each of the domains, simulations were carried out for

each domain. It should be noted that methodologically,

the estimation of consumption poverty using the SAE

Small Area Estimation, particularly in the Fay-Herriet

model involved most of the time and work. Finally, once

the consumption poverty levels were calculated down

to the smallest geographical disaggregation, i.e. the dis-

tricts, we proceeded to make projections for the popula-

tion of children aged 0-3 years, up to the year 2030 [43].

3.1 ANALYSIS OF THE INCIDENCE OF CONSUMP

TION POVERTY

Consumption poverty is analyzed at dierent geographical

levels, showing signicant variations between provinces,

cantons and districts. At the national level, the consump-

tion poverty rate is 25.8%, but varies considerably in regions

such as the highlands and the coast, where rates can be up

to 30% higher in rural areas [14]. is geographical hete-

rogeneity is crucial for the implementation of public poli-

cies, as it allows targeting resources to areas of greatest need

and designing interventions adapted to local conditions.

For example, conditional cash transfer programs that have

proven eective in reducing child poverty can be applied in

areas with high poverty rates [3]. Heterogeneity of poverty

helps in policy implementation by allowing better alloca-

tion of resources to the most vulnerable populations, which

can improve the eectiveness of social interventions [2].

Table 4.

Goodness-of-t R2 of econometric models

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

Province Incidence Children from 0

to 3 years old

Standard

deviation

Coecient of

variation

Napo 81,3% 7.121 2,0% 2,5%

Chimborazo 79,1% 24.929 3,1% 3,9%

Santa Elena 67,4% 17.726 6,4% 9,5%

Bolívar 66,1% 8.078 4,4% 6,7%

Zamora Chinchipe 60,9% 4.532 6,6% 10,9%

Esmeraldas 59,8% 25.358 2,3% 3,9%

Imbabura 55,4% 16.440 2,0% 3,6%

Morona Santiago 54,4% 6.142 2,5% 4,6%

Manabí 53,1% 51.156 3,5% 6,5%

Los Ríos 48,2% 28.339 2,0% 4,1%

Sucumbíos 45,6% 6.081 6,2% 13,5%

Cotopaxi 45,4% 13.494 2,9% 6,3%

Loja 42,4% 13.794 3,7% 8,6%

Tungurahua 40,6% 13.843 2,3% 5,8%

Azuay 39,4% 20.734 3,3% 8,4%

Guayas 37,6% 99.474 3,0% 8,0%

Carchi 37,4% 4.301 2,4% 6,4%

Pastaza 36,9% 2.124 4,0% 10,9%

Orellana 36,3% 4.035 2,7% 7,5%

Santo Domingo de los Tsáchilas 34,1% 10.371 2,8% 8,2%

Pichincha 25,1% 46.329 1,9% 7,5%

El Oro 18,5% 7.867 2,5% 13,4%

Cañar 15,2% 2.635 3,1% 20,2%

Galápagos 1,2% 20 0,7% 61,0%

National total 41,1% 434.926 0,9% 2,3%

Source: [15]

It is important to mention that the incidence of con-

sumption poverty shows signicant variations across age

groups. In this case, consumption poverty in children

aged 0-3 years was 41.1% respectively at the national le-

vel. e provincial disaggregation is shown in Table 5 on

the incidence of poverty in children aged 0-3 years in di-

erent provinces of Ecuador, together with the standard

deviation and the coecient of variation (CV), which in-

dicate the precision and variability of the estimates.

Poverty incidence: Represents the percentage of children

aged 0-3 living in poverty in each province. e provinc-

es with the highest incidence of child poverty are Napo

(81.3%), Chimborazo (79.1%) and Santa Elena (67.4%),

while Galápagos (1.2%) has the lowest incidence.

Number of children: Shows the total number of chil-

dren aged 0-3 aected by poverty in each province. e

provinces with the highest number of children in pov-

erty are Guayas (99,474), Manabí (51,156) and Pichin-

cha (46,329), due to their larger child population. On

the other hand, Galapagos has the lowest number of

children aged 0 to 3 (20), which coincides with its low

incidence of poverty.

Standard deviation: A measure of dispersion that in-

dicates how much estimates of poverty incidence vary.

Provinces with a higher standard deviation, such as

Santa Elena (6.4%) and Zamora Chinchipe (6.6%),

have greater uncertainty in their estimates, while prov-

inces such as Pichincha (1.9%) and Esmeraldas (2.3%)

show less variability in estimates.

Coecient of variation (CV): e CV is a measure that

reects the ratio of the standard deviation to the mean,

indicating the relative precision of the estimates. A low-

er CV suggests greater precision. e table shows that

Gálapagos (61%), Cañar (20.2%) and El Oro (13.4%)

have high coecients of variation, suggesting that pov-

erty estimates in these provinces have greater uncer-

tainty. In contrast, Napo (2.5%) and Esmeraldas (3.9%)

have lower Coecients of Variation, indicating greater

reliability in their estimates.

Below, in Map 2, the distribution of consumption pover-

ty for children aged 0-3 years is presented at the disag-

gregation level as Planning Zones [43], Planning Zone 1

(53.9%) and Planning Zone 3 (53.1%) show the highest

consumption poverty scores for children aged 0-3 years.

On the other hand, planning zone 9 (22.3%) reects the

best situation at the level of this entire population and for

the country as a whole. Planning zones 2, 4 and 8 show

rates below the national average (see Annex Table 6).

Table 5.

Consumption poverty for children aged 0-3 years, by Province

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

Consumption poverty for children aged 0-3 years, by Planning

Area

Source: [15] Source: [15]

Map 1.

Consumption poverty for children aged 0-3 years, by Province

Map 2.

Map 3.

Consumption poverty for children aged 0-3 years, by Canton Map 4.

Consumption poverty for children aged 0-3 years, by Distrito

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

In addition, Map 3 shows the distribution of poverty

by consumption for children aged 0 to 3 years, at the can-

tonal level, where it was observed that the cantons with

the highest incidence of consumption poverty are: Gua-

mote (97.8%), Colta (95.5%), Carlos Julio Arosemena

Tola (92.7%). Meanwhile, the cantons with the lowest in-

cidence of consumption poverty are: Santa Cruz (1.4%),

Isabela (0.6%), San Cristóbal (0.3%).

Finally, Map 4 shows the distribution of consumption

poverty for children aged 0 to 3 years old at the district

level, where the districts with the highest consumption

poverty rates are: 06D04 (96.9%), 14D05 (87.5%), 06D05

(83.8%) and the districts with the lowest rates are: 03D03

(10.2%), 03D01 (6.2%), 20D01 (1%). In summary, the re-

sults show that there is signicant heterogeneity in con-

sumption poverty within territorial units in Ecuador. is

allows for better targeting of public policies at the disag-

gregated level [43], (see Annex Estimation of consump-

tion poverty for children 0-3 years old, according to the

SAE methodology).

3.2 ESTIMATED POPULATION OF CHILDREN AGED

0 TO 3 YEARS OLD UP TO 2030, BY CONSUMPTION

POVERTY.

e projection of the child population up to 2030 was

made using the SAE methodology, which allows estima-

ting consumption poverty in small areas. However, the

lack of updated Living Conditions Survey (LCS) data

introduces uncertainty into the projections, as recent

economic changes can signicantly aect consumption

levels [2]. To mitigate this problem, econometric models

were used that adjust the projections according to cu-

rrent economic trends. In addition, a homogeneous dis-

tribution of poverty in the areas studied was assumed,

which could aect accuracy if not met. It is important to

consider methodological adjustments to reect possible

variations in the distribution of poverty and improve the

accuracy of the projections [58].

e methodology used to project the population of

children aged 0-3 until 2030 is based on the use of ad-

vanced statistical techniques. ese techniques allow for

more precise estimates in small geographic areas whe-

re sample sizes are insucient to provide reliable data.

is methodological approach is particularly relevant in

contexts where traditional data sources cannot accurately

capture child population dynamics at the local level [59].

It is important to note that there is no update of the Li-

ving Conditions Survey (LCS), only the one conducted

in the years 2013-2014. is survey is crucial as it pro-

vides detailed information on the living conditions and

demographic structure of the Ecuadorian population.

e lack of updating of this data source limits the ability

to make more accurate projections on the child popula-

tion. Population estimates depend on current data, and

the absence of such data can compromise the accuracy

of the results [60].

In addition, the demographic projections are based on

the latest available information, including socio-economic

and demographic data. If these sources are not up to date,

projections may not adequately reect recent changes in

population structure, which could lead to less eective or

inecient public policy decisions [61].

A relevant aspect to highlight is that, despite the up-

date of the VII Population and VI Housing Census of

2010 to the VIII Population, VII Housing and I Com-

munities Census of 2022, the use of the latter census is

not recommended for the application of the Small Area

Estimation (SAE). is is because dierences in the me-

thodologies used in the 2010 and 2022 censuses may sig-

nicantly aect the accuracy and validity of the SAE. e

SAE relies on consistent and representative data to ge-

nerate reliable estimates at the local level [4]. One of the

main considerations is the change in data collection me-

thodology. While the 2010 Census was conducted on a

single day with student participation, the 2022 Census

adopted a mixed approach, including both virtual and

face-to-face collection. is change may have an impact

on the coverage and representativeness of the data at the

local level, which is critical for the proper functioning

of SAE models, which require homogeneous and relia-

ble data [62].

In addition, the 2022 Census incorporated new varia-

bles in its questionnaire, such as questions on sexual iden-

tity, pet ownership and female fertility. ese variables

open up new possibilities for analysis and estimation at

the small area level. However, they also imply the need for

adjustments to previously employed SAE models, as these

must be adapted to the new types of data collected [63].

Finally, comparability of data between the 2010 and 2022

censuses will be a signicant challenge. Dierences in me-

thodologies and variables collected could hinder longitu-

dinal analyses and, consequently, the ecient application

of the SAE. is challenge requires detailed evaluation to

ensure that estimates remain valid and accurate [64]. Me-

thodological changes, especially in the questionnaire for

the 2010 and 2022 Censuses in Ecuador, could have im-

portant implications for the implementation of the Small

Area Estimation (SAE). ese modications aect both

the coverage and representativeness of the data, as well as

the variables available for analysis at the local level. As a

result, the accuracy of population projections may be li-

mited. In particular, when working with models that rely

on consistent data, any variation in data sources can com-

promise the reliability of estimates [40].

Going forward, it will be essential to closely monitor

the quality and comparability of census data to ensure

accurate application of the SAE. Methodological chan-

ges must be carefully managed to avoid distortions in the

results obtained from the new data. Consistency in data

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

collection and comparability with previous censuses are

critical to ensure that estimates remain reliable and useful

for decision-making [63]. Despite these limitations, the

present study successfully used the methodology of [43]

to make annual projections of the population of children

aged 0-3 years until 2030. ese projections provide a va-

luable tool for planning in key sectors such as education,

health and social services. In these areas, accurate popu-

lation estimates are crucial for allocating resources and

designing services eciently.

From a methodological perspective, the projection of

the population of children aged 0-3 towards 2030 requi-

red the use of alternative data sources and was develo-

ped in two dierent scenarios. In both cases, data derived

from income poverty were used as the fundamental ba-

sis. Income poverty is an economic indicator that measu-

res the ability of an individual or household to meet their

basic needs from their income. is measure is crucial

because it provides an accurate picture of the economic

constraints faced by families, which is critical for accu-

rate demographic projections [65]. e information on

income poverty used for these projections comes from

the National Survey of Employment, Unemployment

and Underemployment (National Survey on Employ-

ment, Unemployment and Underemployment, ENEM-

DU). is survey, which is administered periodically by

the National Institute of Statistics and Census (INEC) of

Ecuador, collects data on the employment situation of the

population, including employment, unemployment and

underemployment. Given that the economic situation of

households signicantly inuences demographic dyna-

mics, the ENEMDU becomes a key source for the analy-

sis and construction of long-term population estimation

models [66].

We decided to use the income poverty measure ins-

tead of consumption poverty because of limitations in the

data available to calculate the latter. Consumption pover-

ty is calculated from the Living Conditions Survey (LCS),

which is conducted on a non-periodic basis. ere was

an 8-year gap between the last LCS rounds, which im-

plies that the data may be outdated and not adequate for

reliable annual forecasting. is represents a signicant

challenge for long-term projections [67]. In contrast, the

National Survey of Employment, Unemployment and Un

deremployment (ENEMDU) is conducted periodically,

which provides more recent and relevant information for

annual population projections. is is crucial when con-

sidering the evolution of income poverty, which can di-

rectly inuence population dynamics. erefore, we have

chosen to use ENEMDU data to obtain more accurate

estimates of the population of children aged 0-3 years

around 2030 [66].

e use of income poverty as an indicator of the eco-

nomic situation of an individual or household is a com-

mon practice in population projection studies. This

measure allows for a more accurate capture of the eco-

nomic uctuations that directly aect households, which

in turn impacts on demographic trends. In this sense, in-

come poverty becomes a suitable proxy for assessing the

current socio-economic situation and projecting future

scenarios [65].

In economic terms, an individual income can be

broken down into two main components: consump-

tion and savings. is can be expressed by the equation

Y=C+A, where Y represents total income, C consump-

tion and A savings. A person’s consumption is strongly

inuenced by his or her level of income, which can be

modelled by the equation C=kY, where k is the margi-

nal propensity to consume. is propensity indicates the

fraction of an increase in income that is spent on con-

sumption rather than saving. ere is a direct and positive

relationship between income and consumption: as inco-

me increases, so does consumption. Generally, changes in

income lead to changes in consumption in similar pro-

portions. However, in some contexts, such as in Ecuador,

savings can be very limited. In such cases, income beco-

mes a good proxy for consumption, which means that it

can be used to represent the level of economic well-being

of an individual or household [68].

e relationship between income, savings and con-

sumption is fundamental to understanding the economic

situation of a household. Income poverty is a measure ba-

sed on this relationship and is used to make accurate po-

verty estimates for small geographic areas and specic

population groups. By considering how income aects

both consumption and savings, it is possible to obtain

a clearer picture of the economic well-being of dierent

segments of the population [69]. is methodology is par-

ticularly relevant for designing public policies to address

the economic needs of the most vulnerable communi-

ties. erefore, by analyzing consumption behavior and

its relationship to income, it is possible to deduce how va-

riations in income impact people standard of living. Ac-

curate estimates of poverty and well-being are crucial for

implementing eective strategies to promote social and

economic development [70].

3.2.1.First scenario

e rst scenario (see Annex Table 4), which is based on

the following assumptions to estimate the population of

children aged 0-3 years according to consumption po-

verty:

e variation in the growth rate of income poverty is

constant at -8.87%, according to the latest data from

the National Survey of Employment, Unemployment

and Underemployment [71].

Poverty is evenly distributed across the country. - e

incidence of poverty in the ve self-reported cities

does not show statistically signicant variations.

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

is scenario uses the ENEMDU data, and the SAE

methodology to predict consumption and construct

the consumption poverty rate for the dierent popu-

lation subgroups dened in the districts.

3.2.2.Second scenario

An approach based on growth rates derived from Ecua-

dor Gross Domestic Product (GDP) is used to estima-

te the population of children aged 0-3 until 2030. is

analysis assumes that variations in the level of produc-

tion, and hence economic income, inuence poverty le-

vels in dierent regions of the country. is is considered

to be a conservative scenario (see Annex Table 5), given

that the country has relatively low annual growth rates.

e methodology is based on several assumptions:

Income elasticity of poverty reduction: Dened as

the percentage change in poverty resulting from a 1%

increase in per capita income. is elasticity is held

constant and is calculated using the most recent data

on income poverty variation provided by the National

Survey of Employment, Unemployment and Underem-

ployment (ENEMDU), as well as ECLAC growth pro-

jections for the same year.

Homogeneous distribution of poverty: Poverty is as-

sumed to be evenly distributed across the territory.

Minimum thresholds in the projections: No projec-

tion is lower than an established minimum threshold.

In this context, rates of change of GDP per capita are

calculated using growth estimates provided by ECLAC

and population projections published by INEC. e

paper includes poverty elasticity calculations and pro-

jections for the child population aged 0-3 years, consi-

dering consumption poverty at the district level [72].

is approach allows for a deeper understanding of how

economic changes aect vulnerable populations. Income

elasticity is crucial for assessing the impact of economic

policies on poverty, as a higher elasticity indicates that

an increase in income has a signicant eect on pover-

ty reduction. Moreover, by considering a homogeneous

distribution, it aims to simplify the analysis, although

in practice there may be signicant variations between

dierent regions [73]. Finally, by setting lower limits on

the projections, it ensures that the estimates do not un-

derestimate the real needs of the child population. is

is essential for designing eective policies to address de-

privation and promote social well-being. e resulting

projections provide valuable information for planning

resources and services for this critical age group [74].

3.2.3.Additional clarications:

Adjustments to the SAE methodology: To improve ac

curacy, additional variables reecting recent economic

changes could be incorporated or the model could be

adjusted to account for possible biases in the distribu-

tion of poverty [4].

Homogeneous distribution of poverty: A homogeneous

distribution was assumed to simplify the model, but if

poverty is not evenly distributed, this could aect the

precision of the estimates. It would be useful to explo-

re models that incorporate geographical variations in

the distribution of poverty [75].

Projection scenarios: In the first scenario, histori-

cal data were used without significant adjustment

to reect current trends. In the second scenario, ad-

justments were included to reect recent economic

changes, allowing for a more realistic projection of the

child population [9].

3.3 ANALYSIS OF THE RESULTS OBTAINED IN THE

SCENARIOS.

CELADE estimates of the child population in Latin

America are relevant to this analysis, as they provide a

comparative framework for assessing regional demogra-

phic trends [76]. Figure 5 compares Ecuador child po-

pulation projection scenarios with CELADE estimates,

showing signicant dierences in the population growth

rate. Factors such as education and birth control are cru-

cial to explain these dierences, as they directly inuen-

ce population dynamics and poverty reduction [77]. e

two projections dier in their approach: the rst is based

on historical trends, while the second incorporates ad-

justments to reect recent economic changes, resulting

in a more conservative projection of the child population

in high-poverty areas.

In Figure 5 illustrates that a decline in the population

of children aged 0-4 years is anticipated in all projected

scenarios between 2022 and 2030. is trend can be at

tributed to a number of factors, such as declining birth

rates, changes in population policies, as well as improve-

ments in education and increased knowledge about bir-

th control. CELADE estimates, which cover children up

to the age of 4, show a more gradual decline. is might

suggest that ECLAC foresees a slow change in the current

conditions aecting the child population. In contrast, esti-

mates from other scenarios, obtained using the SAE me-

thodology, show more rapid reductions for children aged

0-3 in consumption poverty. is suggests that these sce-

narios anticipate more signicant or rapid changes in the

conditions impacting this age group [78].

It is crucial to note that these projections are mere-

ly hypothetical scenarios and that future reality may be

subject to multiple unforeseen factors. While downward

trends may cause concern, they may also reect positive

outcomes, such as increased access to education and bir-

th control. erefore, while the graph provides valuable

insight into child population projections, it is critical to

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

contextualize these data with other metrics and socio-eco-

nomic factors to gain a full understanding of the situation

[79]. Finally, in assessing these projections, it is essen-

tial to consider how they relate to public policies and so-

cial programs aimed at supporting families and fostering

a favorable environment for child growth. e interac-

tion between demographic changes and the policies im-

plemented will be critical in addressing future challenges

related to population and social well-being [80].

To facilitate understanding of the results obtained, it

is useful to make clearer comparisons between the die-

rent scenarios considered in the study. In particular, the

simulation to estimate consumption poverty and project

the child population up to 2030 (see Annex Table 6) re-

veals signicant dierences between the baseline, pes-

simistic and optimistic scenarios [35]. In the baseline

scenario, child poverty is projected to increase modera-

tely compared to current levels, while the pessimistic sce-

nario shows a considerable increase due to factors such

as the economic slowdown and rising unemployment. In

contrast, in the optimistic scenario, child poverty is pro-

jected to decline due to the implementation of eecti-

ve public policies and inclusive economic growth. ese

comparisons provide a better understanding of the poten-

tial impact of dierent factors and policies on the evolu-

tion of child poverty in Ecuador.

3.4 DISCUSSION

e application of the Small Area Estimation (SAE) me-

thodology in this study has overcome the limitations of

traditional estimation methods, providing accurate re-

sults at the local level. However, it is important to recog-

nize that the utility of SAE is not without its challenges.

One of the main advantages of SAE is its ability to inte-

grate survey and census data, which facilitates obtaining

reliable estimates in small areas [13]. However, the tran-

sition between positive results and limitations needs to

be smoothed to maintain a clear thread.

Variations in census questions and collection metho-

dology can introduce biases or errors in the data used for

EDC, aecting the precision of estimates [58]. To mitigate

these problems, strategies such as cross-validation of data

and adjustment of models to reect changes in local con-

ditions can be implemented [75]. In addition, the lack of

updating of the Living Conditions Survey (LCS) hinders

the application of the methodology, as recent economic

changes may not be reected in the data, aecting the ac-

curacy of the projections [2].

Undetected variations at the community level or

within ethnic groups can have signicant implications

for the accuracy of estimates. For example, if commu-

nity-specic cultural or socio-economic dierences are

not considered, models may not adequately capture lo

cal needs [39]. In terms of model performance, a good t

was observed in urban areas, but limitations aected ru-

ral areas more, where the lack of updated data was more

pronounced [23].

A fundamental aspect to consider in the interpreta-

tion of the results obtained through SAE lies in the inhe-

rent limitations of this methodology. While SAE oers a

valuable tool for generating estimates at the disaggrega-

ted level, it is crucial to recognize that its accuracy and

validity depend to a large extent on the selection of auxi-

liary variables and their representativeness in the dierent

geographical contexts analyzed [4]. e choice of varia-

bles that are not strongly correlated with the variable of

interest, or that have measurement biases, can compro-

mise the reliability of the estimates and lead to erroneous

conclusions about the distribution of child poverty. It is

also important to bear in mind that the SAE is based on

statistical models that simplify the complexity of social

reality, which can lead to the omission of relevant factors

that inuence children’s vulnerability [75].

Estimation of girls and boys from 0 to 3 years old according to the SAE methodology, vs. Estimation of the population of girls and boys

from 0 to 4 years old according to ECLAC until 2030.

Source: [15]

Figure 5.

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

e discussion of the results of this study is enriched

by linking them closely to their practical and theoreti-

cal implications, as well as by contrasting them with n-

dings from another relevant research in the eld [9]. For

example, our results on the impact of maternal education

on child poverty are consistent with previous studies that

have shown a strong correlation between these two factors

[81]. However, we also found some signicant dieren-

ces compared to research conducted in other geographi-

cal contexts, suggesting the importance of considering

the specic characteristics of the Ecuadorian population

when designing public policies aimed at reducing child

poverty [82]. By analyzing these similarities and dieren-

ces, we can gain a deeper understanding of the factors that

inuence child poverty and design more eective inter-

ventions tailored to local needs.

e results obtained by applying the SAE methodolo-

gy make it possible to generate precise estimates of the in-

cidence of consumption poverty for dierent age groups,

especially for children aged 0-3 years, at provincial, can-

tonal, parish and district levels (see Annex Table 6). is

level of disaggregation is not possible with traditional sta-

tistical methods, which makes the SAE methodology a

crucial tool for planning and decision-making in public

policy. e ability to estimate poverty rates more accura-

tely at the district level provides government institutions

with a solid basis for designing more eective and targe-

ted strategies [4].

e usefulness of the SAE methodology is particu

larly evident when considering the limitations of current

data sources, which do not allow reliable statistical infe-

rences to be generated in small areas using conventional

methods. SAE overcomes these limitations by exploiting

both census and survey data and allows for the creation

of tight and robust predictions in small population do-

mains. e results of this approach show signicant va-

riations in consumption poverty levels between dierent

planning areas, provinces, cantons and districts, highligh-

ting the geographical heterogeneity of the poverty pheno-

menon. is underlines the need for specic and targeted

policies to address these regional disparities [80].

At the district level, the results obtained through SAE

are a valuable tool to visualize and understand consump-

tion poverty gaps in the country. However, the unavaila-

bility of certain household-level socio-economic variables

in the Census restricts the use of some predictors, which

may limit the precision of the estimates. Moreover, me-

thodological dierences between the 2010 and 2022 cen-

suses present additional challenges in the implementation

of the SAE. ese changes may aect the consistency and

comparability of the data, underscoring the importance

of adjusting methodologies to maintain the reliability of

the analyses [83].

Comparison of the 2010 and 2022 Ecuadorian censu-

ses reveals important methodological discrepancies that

compromise the accuracy of small area estimates (SAE).

Changes in questionnaire design, data collection techni-

ques, population size and composition, and the amount of

information collected aect the comparability of data and

the accuracy of SAE estimates. Alterations in the ques-

tionnaires make direct comparison with the 2022 census

dicult, introducing potential biases. Demographic chan-

ges also inuence the accuracy of estimates if they are not

adequately considered. To ensure the validity and relia-

bility of the SAE estimates, it is essential to take these di-

erences into account and adjust the methodology [84].

e lack of an update of the National Living Condi-

tions Survey (LCS) hinders the application of the SAE

methodology, limiting the possibility of integrating au-

xiliary information to improve the precision of the esti-

mates. While acknowledging the limitations of the data

sources used in this study, the models employed showed

a good t to the available data [40]. Finally, the limited

spatial resolution of the data sources may prevent the de-

tection of minor variations at the community or ethnic

group level. Despite this limitation, the methodological

approach used is useful for small area-level analysis [52].

While our study has revealed signicant variations in

consumption poverty levels between dierent regions, it

is crucial to delve deeper into the underlying reasons for

these disparities. Factors such as the economic structu-

re of each region, the availability of employment, levels

of education and the public policies implemented can in-

uence the distribution of wealth and the vulnerability of

households. For example, regions with a higher concen-

tration of informal agricultural activities may be more

susceptible to poverty due to commodity price volatility

and lack of access to nancial services and training [84].

Likewise, regions with a higher proportion of indige-

nous or Afro-descendant populations may face additional

barriers to accessing employment and education oppor-

tunities due to discrimination and social exclusion [85].

Lack of investment in infrastructure and basic services in

rural and marginalized areas can also contribute to perpe-

tuating poverty and limiting economic development. To

better understand these dynamics, it would be advisable

to conduct detailed case studies in some of the regions

with the highest levels of poverty, in order to identify the

specic factors that are contributing to the situation and

design more eective interventions tailored to local needs.

It is important to recognize the inherent limitations

of any study based on statistical data, especially in areas

where the available information presents a higher level of

uncertainty. In our case, the representativeness of the LCS

data may be limited in some regions, which could aect

the validity of our projections. In addition, there may be

measurement errors or biases in the data, which could in-

uence the results. It is therefore essential to interpret our

ndings with caution and to recognize that there are mar-

gins of error associated with our estimates.

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

To improve the accuracy and reliability of future re-

search, it would be advisable to explore the possibility of

combining dierent data sources, such as administrative

records, household surveys and satellite data, in order to

obtain a completer and more accurate picture of the child

poverty situation. It is also important to invest in training

local sta in data collection and analysis to ensure the

quality and relevance of the information. By addressing

these limitations, we will be able to generate more solid

and useful evidence for decision-making and the imple-

mentation of public policies aimed at the most vulnera-

ble population.

It is essential to recognize that the SAE methodology

has certain limitations that must be considered when in-

terpreting the results and designing public policies based

on them. One of the main challenges of SAE lies in the

potential bias introduced by the selection of auxiliary va-

riables, which may not be representative of the socio-eco-

nomic reality of all the geographical areas analyzed [25].

To mitigate this risk, it is crucial to conduct a thorough

analysis of the quality and relevance of the variables used,

as well as to assess their ability to capture the specic dy-

namics of each local context.

It is also important to bear in mind that the SAE re-

lies on survey data that may not be updated frequently

enough to reect the socio-economic changes occurring

in society [86]. In this sense, it is advisable to complement

survey data with information from other sources, such as

administrative records and geographic information sys-

tems, in order to obtain a completer and more accurate

picture of the child poverty situation [86].

Furthermore, it is crucial to recognize that poverty

dynamics can vary signicantly over time and in die-

rent geographical contexts, which can inuence the eec-

tiveness of policy programs designed on the basis of these

data [87]. erefore, it is necessary to conduct regular

evaluations of implemented programs and adjust strate-

gies based on the results obtained, in order to ensure that

interventions are relevant and eective in reducing child

poverty.

Living Conditions Surveys (LCS) play a crucial role

in policy formulation by providing detailed information

on the socio-economic characteristics of households and

their access to basic services [88]. LCS data can inuen-

ce the design of policy interventions for social inclusion

and poverty reduction by identifying the most vulnera-

ble groups and targeting resources where they are most

needed. However, it is important to bear in mind that

LCA are only one tool among many, and that their eec-

tiveness depends on the quality of the data collected, the

thoroughness of the analysis carried out and the politi-

cal will to implement the recommendations derived from

the results.

e ndings of this study, which estimate the popula-

tion of children in poverty in Ecuador using SAE, provide

a notable advance in knowledge about this problem at

the local level. Compared to previous research in Latin

America, such as that of [10], which used SAE to mea-

sure rural poverty, our work stands out for focusing spe-

cically on children aged 0-3 years in Ecuador, a group

little studied at this scale. e results show great diver-

sity in poverty levels across regions, provinces, cantons

and districts, highlighting the need for detailed data to

target resources eciently. Furthermore, when compa-

ring our projections with those of ECLAC, we nd a rea-

sonable alignment, which reinforces the reliability of our

estimates in the face of the scarcity of updated territo-

rial data in the country [2]. is innovative approach not

only improves the precision of estimates in small areas,

but also establishes a starting point for future studies on

child poverty and vulnerability, providing policy makers

with a key tool to prioritize interventions where they are

most needed [10].

4. Conclusions

is study has demonstrated the usefulness of the SAE

methodology for estimating consumption poverty in sma-

ll areas, specically contributing to this end by providing

accurate estimates at the district level. e results obtai-

ned allow for a more specic identication of the target

population required for decision-making regarding the

implementation of specic public policies. By using re-

cent data from 2014 to 2025 and projections up to 2030, a

clear picture of demographic trends and child poverty in

Ecuador is provided. is approach is particularly useful

for the country economic and social planning, as it allows

targeting resources in areas of greatest need and designing

interventions tailored to local conditions.

One of the main contributions of this study is the abi-

lity to optimize the use of available information, com-

bining survey and census data to generate accurate

estimates. e numerical values of the estimates and in-

dicators provide evidence of the credibility and precision

of the SAE methodology, which allows for better public

policy decision-making [53]. Furthermore, this study sets

an important precedent for future research by demonstra-

ting the feasibility of the SAE in the Ecuadorian context.

Limitations of the study, such as the lack of updated

Living Conditions Survey (LCS) data, suggest the need

for additional research to validate the results and explore

new methodologies. It is recommended that future stu-

dies consider incorporating updated administrative data

and conducting sensitivity analyses to assess the impact

of the assumptions used in the projections [2]. e iden-

tication of vulnerable groups has been consolidated in

this study, allowing for a better understanding of social

and economic inequalities in Ecuador.

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

Emphasize the importance of replication of the me-

thod and its relevance in the eld of social change. In a

context of constant social and economic change, the SAE

methodology oers a valuable tool for monitoring and

evaluating the impact of public policies on child pover-

ty reduction. Its replication in dierent geographical and

social contexts can contribute to a better understanding

of the dynamics of poverty and the identication of eec-

tive strategies to overcome it [8].

e SAE methodology provides accurate and reliable

estimates of population variables at detailed geographi-

cal levels, which is essential for national socio-economic

planning [89]. In the Ecuadorian context, the application

of the SAE has made it possible to estimate consumption

poverty at the district level between 2014 and 2030, fa-

cilitating the precise identication of populations requi-

ring priority attention for the design of targeted public

policies [86]. is analysis reveals considerable variabi-

lity in consumption poverty levels between dierent re-

gions, provinces, cantons and districts, highlighting the

need for disaggregated data for ecient resource alloca-

tion [90]. Heterogeneity at the district level is particularly

signicant, and the availability of reliable data at this sca

le allows for better management of social inclusion pro-

grams [91]. In summary, the SAE methodology proves to

be an eective tool for generating reliable estimates, even

for small geographical areas, thus fullling a crucial need

in public planning and management.

is study identies, with great geographical preci-

sion, the most vulnerable population groups, which faci-

litates the ecient allocation of public resources for social

assistance programs [92]. e integration of gender and

ethnic perspectives is crucial for designing interventions

that promote equity and social inclusion [93]. By applying

econometric models to data from the 2014 Living Con-

ditions Survey (LCS) and the 2010 Population Census,

vulnerability estimates have been obtained at the district

level, maximizing the use of available information [94].

e SAE methodology has proven to be a valuable tool

for identifying unmet needs at the local level, enabling

more informed and eective public decision-making [89].

e possibility of replicating this study periodically would

allow for monitoring socio-economic evolution and adap-

ting strategies to changing social realities [95].

e regression model employed shows a good t in

most geographical regions, with R² values generally abo-

ve 0.60, indicating that the model explains more than 60%

of the variability in consumption [96]. However, a less sa-

tisfactory t is observed in some specic regions, such

as Carchi and Guayas, suggesting the need for improve-

ments or the incorporation of additional variables to op-

timize the model in these areas [97]. To improve the t of

the model in regions with R² values close to 0.60, it is re-

commended to explore the inclusion of additional expla-

natory variables or to re-evaluate the model specication.

is could involve the consideration of region-specic so-

cio-economic or geographical factors [98].

In regions with a limited number of observations,

even if the R² is high, the uncertainty in the results can

be considerable. In these cases, further data collection or

the application of additional statistical techniques is re-

commended to strengthen the conclusions [99]. e mo-

del performs better in regions with higher R² values and a

larger number of observations, such as Napo and Orella-

na, which increases the reliability of the results obtained

in these areas. is contrast highlights the importance of

data quality and quantity for the accuracy of econome-

tric models [100]. e provinces of Napo and Chimbora-

zo have a high prevalence of child poverty, which justies

the prioritization of resources and public policies in the-

se regions. Targeted intervention is required to improve

socio-economic and infrastructure conditions, with the

aim of signicantly reducing child poverty levels in the-

se areas [101].

e high variability in poverty estimates in provin-

ces such as Galápagos, Cañar and Zamora Chinchipe de-

mands special attention. e lower precision of estimates

in these areas suggests the need to improve the quality of

data collection and the design of more targeted interven-

tions to ensure a more reliable assessment of the situa-

tion and the eectiveness of implemented policies [102].

Despite having relatively moderate poverty rates (Gua-

yas: 37.6%, Manabí: 53.1%, Pichincha: 25.1%), the pro-

vinces of Guayas, Manabí and Pichincha are home to a

high number of children in poverty. is indicates that

the implementation of programs aimed at this vulnera-

ble population would have a signicant impact on pover-

ty reduction at the national level. It is crucial to develop

and implement specic programs that address the parti-

cular needs of this population [103].

Provinces such as El Oro and Cañar, despite exhibi

ting low child poverty rates (18.5% and 15.2%, respecti-

vely), show high variability in their estimates, indicating

considerable uncertainty in the data. It is essential to con-

tinue monitoring these provinces to conrm whether the

low incidence of poverty is a real trend or an artefact of

measurement imprecision. Rigorous monitoring will help

determine the need for targeted interventions [58]. e

formu10lation of eective child poverty reduction poli-

cies must consider not only the magnitude of the problem

in each province, but also the quality and reliability of the

available data. Prioritizing interventions based on accura-

te and reliable data is crucial to ensure that resources are

allocated eciently and that the strategies implemented

are truly eective [104].

In order to enrich the comparative analysis and con

textualize our ndings, we have expanded the references

section to include relevant international studies that have

applied the SAE methodology in other contexts [105]. e-

se studies address a variety of issues related to child poverty,

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

such as the measurement of food insecurity, access to heal-

th and education services, and the impact of public policies

on poverty reduction [106]. By examining this research,

we can gain a broader perspective on the challenges and

opportunities associated with implementing SAE in die-

rent geographical and socio-economic contexts, and adapt

best practices to the Ecuadorian reality [107].

References

[1] A. Gutiérrez, X. Mancero, and S. Guerrero, “Poverty ma-

pping in Latin America: ECLAC experiences on small

area estimation,” Statistical Journal of the IAOS, vol. 38,

no. 3, pp. 1021–1033, 2022.

[2] Comisión Económica para América Latina y el Caribe (CE-

PAL), Panorama social de América Latina y el Caribe

2024: Desafíos de la protección social, [Online]. Available:

https://www.cepal.org/es/publicaciones/80858-panora-

ma-social-america-latina-caribe-2024-desaos-la-pro-

teccion-social .

[3] Banco Mundial, Panorama de la pobreza y la desigualdad

en América Latina y el Caribe , Washington, D.C.: Ban-

co Mundial, 2024.

[4] J. N. Rao and I. Molina, Small Area Estimation, John Wi-

ley & Sons, 2015.

[5] DANE, Encuesta Nacional de Calidad de Vida 2023 , De-

partamento Administrativo Nacional de Estadística de

Colombia, 2023. [Online]. Available: https://microda-

tos.dane.gov.co/index.php/catalog/827 .

[6] INEGI, Censo de Población y Vivienda 2024, Instituto Nacio-

nal de Estadística y Geografía de México, 2024. [Onli-

ne]. Available: https://www.inegi.org.mx/app/buscador/

default.html?q=SAE .

[7] A. Sen, Development as Freedom , Oxford University Press,

1999.

[8] Banco Mundial, Pobreza: Panorama general , 2024. [Onli-

ne]. Available: https://www.bancomundial.org/es/topic/

poverty/overview .

[9] W. H. Greene, Econometric Analysis (8th ed.), Pearson Edu-

cation, 2018.

[10] A. Arias-Salazar, A. Gutiérrez, S. Guerrero-Gómez, X.

Mancero, N. Rojas-Perilla, and H. Zhang, “Small area

estimation for composite indicators: the case of multi-

dimensional poverty incidence,” Journal of Ocial Sta-

tistics , 0282423X241300751, 2023.

[11] Eurostat, City Statistics - Small Area Estimation , European

Commission, 2024. [Online]. Available: https://ec.euro-

pa.eu/eurostat/statistics-explained/index.php?title=Ci-

ty_statistics_-_small_area_estimation

[12] Statistics Canada, Canadian Community Health Survey -

Annual Component (CCHS), [Online]. Available: https://

www23.statcan.gc.ca/imdb/p2SV.pl?Function=getSur-

vey&SDDS=3226 .

[13] A. Molina, J. Rosero, M. León, R. Castillo, F. Jácome, D.

Rojas, et al., Reporte de pobreza por consumo Ecuador

2006-2014, 2016.

[14] INEC-BM, Reporte de Pobreza por Consumo Ecuador 2006-

2014, Quito: Instituto Nacional de Estadística y Cen-

sos, 2014.

[15] S. Dercon, Assessing Vulnerability , World Bank Publica-

tions, 2001.

[16] Z. Villa Juan-Albacea, Small Area Estimation of Poverty Sta-

tistics , No. 2009-16, PIDS Discussion Paper Series, 2009.

[17] D. M. Stukel and J. N. K. Rao, “Estimation of regression

models with nested error structure and unequal error

variances under two and three stage cluster sampling,”

Statistics & Probability Letters , vol. 35, no. 4, pp. 401–

407, 1997.

[18] Y. You and J. Rao, “A pseudo-empirical best linear unbiased

prediction approach to small area estimation using sur-

vey weights,” Canadian Journal of Statistics, vol. 30, no.

3, pp. 431–439, 2002.

[19] J. N. K. Rao and I. Molina, Small Area Estimation: A Review

of Methods Based on the Application of Mixed Models,

2015. [Online]. Available: https://www.researchgate.net/

publication/252669243_Small_Area_Estimation_A_

Review_of_Methods_Based_on_the_Application_of_

Mixed_Models .

[20] C. Elbers, J. O. Lanjouw, and P. Lanjouw, Small Area Es-

timation of Poverty Indicators, ResearchGate, 2003.

[Online]. Available: https://www.researchgate.net/pu-

blication/230806613_Small_Area_Estimation_of_Po-

verty_Indicators .

[21] UNICEF, e State of the World’s Children 2023 , [On-

line]. Available: https://www.unicef.org/reports/sta-

te-worlds-children-2023 .

[22] Studocu, Variables dummy de estadística , [Online].

Available: https://www.studocu.com/ec/document/

instituto-tecnologico-bolivariano-de-tecnologia/admi-

nistracion/variables-dummy-de-estadistica/15406065 .

[23]J. Jiang and P. Lahiri, “Mixed model prediction and

small area estimation,” Test , vol. 15, pp. 1–96, 2006.

[24] I. Molina, J. N. K. Rao, and G. S. Datta, “Small area estima-

tion under a Fay-Herriot model with preliminary testing

for the presence of random area eects,” Survey Metho-

dology , vol. 41, no. 1, pp. 1–20, 2015.

[25] M. Ghosh and J. N. Rao, “Small area estimation: an apprai-

s a l ,” Statistical Science , vol. 9, no. 1, pp. 55–76, 1994.

[26] W. Lutz, W. P. Butz, and K. E. Samir (Eds.), World Popu-

lation and Human Capital in the Twenty-First Century ,

OUP Oxford, 2014.

[27] A. Van Delden, B. J. Du Chatinier, and S. Scholtus, “Accu-

racy in the application of statistical matching methods

for continuous variables using auxiliary data,” Journal

of Survey Statistics and Methodology , vol. 8, no. 5, pp.

990–1017, 2020.

[28] M. Ghosh and J. N. Rao, “Small area estimation: an apprai-

s a l ,” Statistical Science , vol. 9, no. 1, pp. 55–76, 1994.

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

[29] I. Molina and Y. Marhuenda, Small Area Estimation for

Mapping Local Indicators , 2019. [Online]. Available: ht-

tps://www.makswell.eu/attached_documents/3_sma-

ll-area-estimation-for-mapping-local-indicators.pdf .

[30] P. Corral, I. Molina, A. Cojocaru, and S. Segovia, Guideli-

nes to Small Area Estimation for Poverty Mapping , Was-

hington: World Bank, 2022.

[31] United Nations Population Fund, Small Area Estimation for

Mapping Local Indicators , 2019. [Online]. Available: ht-

tps://www.unfpa.org/sites/default/les/pub-pdf/19-310_

SAE_Brochure_A4-SINGLE-PROOF6.pdf .

[32] [14] Asian Development Bank, Introduction to Small Area

Estimation Techniques: A Practical Guide for National

Statistics Oces , 2020. [Online]. Available: https://www.

adb.org/publications/small-area-estimation-guide-na-

tional-statistics-oces .

[33] World Bank, Guidelines to Small Area Estimation for Po-

verty Mapping , 2023. [Online]. Available: https://

www.worldbank.org/en/events/2023/02/07/guideli-

nes-to-small-area-estimation-for-poverty-mapping .

[34] RTI International, Small Area Estimation , [Online]. Avai-

lable: https://www.rti.org/brochures/small-area-estima-

tion .

[35] A. de Mesa, Los sistemas de pensiones en América Latina:

institucionalidad, gasto público y sostenibilidad nancie-

ra en tiempos del COVID-19 , 2020.

[36] J. M. Wooldridge, Introductory Econometrics: A Modern

Approach (6th ed.), Cengage Learning, 2016.

[37] M. C. Nguyen, P. A. C. Rodas, J. P. W. De Azevedo, and Q.

Zhao, “Small area estimation and Poverty map in Stata,”

in 2017 Stata Conference , No. 20, 2017.

[38] M. H. Boyle et al., “e inuence of economic develop-

ment level, household wealth and maternal education

on child health in the developing world,” Social science

& medicine , vol. 63, no. 8, pp. 2242–2254, 2006.

[39] F. F. J. Bogliaccini, Metas del milenio y desarrollo: las con-

guraciones del desarrollo social en América Latina , Do-

cumento de Trabajo del IPES, Montevideo, 2008.

[40] J. N. K. Rao and I. Molina, Small Area Estimation: A Review

of Methods Based on the Application of Mixed Models

, 2015. [Online]. Available: https://www.researchgate.

net/publication/252669243_Small_Area_Estimation_A_

Review_of_Methods_Based_on_the_Application_of_

Mixed_Models .

[41] E. A. Herrera Herrera, Comparación de estimadores directos

e indirectos en estimación de áreas pequeñas , Universidad

Santo Tomás, 2018. [Online]. Available: https://reposi-

tory.usta.edu.co/bitstream/handle/11634/12530/2018e-

dgarherrera.pdf?sequence=1

[42] R. E. Fay and R. A. Herriot, “Estimates of income for small

places: An application of James-Stein procedures to cen-

sus data,” Journal of the American Statistical Association

, vol. 74, no. 366, pp. 269–277, 1979.

[43] G. N. López Fuertes, Estimación de la población de ni-

ñas y niños de 0 a 3 años del ecuador, según Pobreza por

Consumo, mediante la metodología Small Area Estima-

tion (SAE) hasta el año 2021 , Master’s thesis, Quito:

UCE, 2019.

[44] Instituto Nacional de Estadística y Censos (INEC), Guía de

Uso de Base de Datos , 2021. [Online]. Available: https://

www.ecuadorencifras.gob.ec/documentos/web-inec/

EMPLEO/2021/Febrero-2021/202102_%20Guia%20

de%20usuario%20BDD_ENEMDU.pdf

[45] J. Parra, Introducción a las variables cticias o variables

dummy , [Online]. Available: https://www.javierparra.

net/ecoknowmic/introduccion-a-las-variables-cti-

cias-o-variables-dummy/ .

[46] Studocu, Variables dummy de estadística , [Online].

Available: https://www.studocu.com/ec/document/

instituto-tecnologico-bolivariano-de-tecnologia/admi-

nistracion/variables-dummy-de-estadistica/15406065 .

[47] M. C. Howard, Regresión con Variables Dummy en Jamovi

, [Online]. Available: https://mattchoward.com/regre-

sion-con-variables-dummy-en-jamovi/ .

[48] M. Carvajal, Análisis de datos y detección de valores atípi-

cos en econometría , Quito: Ediciones Económicas, 2004.

[49] T. Hastie, R. Tibshirani, J. H. Friedman, and J. H. Fried-

man, e Elements of Statistical Learning: Data Mining,

Inference, and Prediction , vol. 2, New York: Springer,

2009.

[50] A. C. King, M. P. Rao, and C. D. Tregillis, “Econometric

Analysis,” in Litigation Services Handbook: e Role Of

e Financial Expert , 2017, pp. 1–62.

[51] R. McElreath, Statistical Rethinking: A Bayesian Course

with Examples in R and Stan , Chapman and Hall/CRC,

2018.

[52] D. Pfeermann, “New important developments in small

area estimation,” Quality control and applied statistics ,

vol. 59, no. 1, pp. 103–106, 2014.

[53] B. W. Silverman, Density Estimation for Sta-

tistics and Data Analysis , Routledge, 2018.

[54] J. Fox, Applied Regression Analysis and Generalized

Linear Models , Sage Publications, 2015.

[55] N. Tzavidis et al., “A framework for producing small area

estimates based on the Fay-Herriot model,” Statistical

Modelling , vol. 19, no. 4, pp. 309–328, 2019.

[56] World Health Organization (WHO), Global Strategy

Data: Maternal, Newborn, Child, Adolescent and Age-

ing , [Online]. Available: https://platform.who.int/data/

maternal-newborn-child-adolescent-ageing/global-stra

tegy-data .

[57] Save the Children, World more dangerous than ever for chil-

dren , [Online]. Available: https://www.savethechildren.

org.uk/news/media-centre/press-releases/world-mo-

re-dangerous-than-ever-for-children#:~:text=Save%20

the%20Children%20report%20reveals,day%20were%20

killed%20or%20maimed .

[58] A. F. Correa, On the measurement of multidimensional po-

verty as a policy tool: empirical applications to Chile, Co-

lombia, Equador and Peru , 2017.

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

[59] S. K. Smith, J. Tayman, and D. A. Swanson, State and Lo-

cal Population Projections: Methodology and Analysis ,

Springer, 2019.

[60] United Nations, Principles and Recommendations for Po-

pulation and Housing Censuses , United Nations Statis-

tics Division, 2020.

[61] S. H. Preston, P. Heuveline, and M. Guillot, Demography:

Measuring and Modeling Population Processes , Blackwe-

ll Publishing, 2022.

[62] National Research Council, Reengineering the 2020 Census:

Risks and Challenges , National Academies Press, 2020.

[63] D. Pfeermann, “Small Area Estimation: A Review and

Some Directions,” Statistica Sinica , vol. 29, no. 1, pp.

289–326, 2019.

[64] OECD, Challenges of the 2022 Census and the Implications

for Data Comparability , Organisation for Economic

Co-operation and Development, 2022.

[65] J. Foster, J. Greer, and E. orbecke, “A class of decompo-

sable poverty measures,” Econometrica: journal of the

econometric society , pp. 761–766, 1984.

[66] Instituto Nacional de Estadística y Censos (INEC), Encues-

ta Nacional de Empleo, Desempleo y Subempleo (ENEM-

DU) , Quito, Ecuador: INEC, 2023.

[67] P. Samaniego, “Evolución de la pobreza y la desigualdad en

Quito,” Questiones Urbano Regionales. Revista del Insti-

tuto de la Ciudad , vol. 1, no. 2, pp. 77–94, 2013.

[68] N. G. Mankiw and M. P. Taylor, Economía , Cengage Lear-

ning, 2021.

[69] J. De Gregorio, Macroeconomía , McGraw-Hill, 2007.

[70] R. Prebisch, El desarrollo económico de América Latina y

sus principales problemas , Naciones Unidas, 2012.

[71] Instituto Nacional de Estadística y Censos (INEC), Bole-

tín de pobreza: Encuesta Nacional de Empleo, Desempleo

y Subempleo (ENEMDU), Junio 2024 , Quito, Ecuador:

INEC, 2024. [Online]. Available: https://www.ecua-

dorencifras.gob.ec/documentos/web-inec/POBRE-

ZA/2024/Junio/202406_Boletin_pobreza_ENEMDU.

pdf .

[72] J. Andrade, E. Marinho, and G. Lima, “Crecimiento eco-

nómico y concentración del ingreso: sus efectos en

la pobreza del Brasil,” CEPAL , 2017. [Online]. Avai-

lable: https://repositorio.cepal.org/bitstream/hand-

le/11362/42693/1/RVE123_Araujo.pdf .

[73] del Ecuador, O. S., Situación de la niñez y adolescencia en

Ecuador. Una mirada a través de los ODS , 2019.

[74] INEC, Proyecciones Demográcas 2022-2030 , 2024. [Onli-

ne]. Available: https://www.ecuadorencifras.gob.ec/pro-

yecciones-poblacionales/ .

[75] R. L. Chambers and R. Clark, An Introduction to Model-Ba-

sed Survey Sampling with Applications , Oxford Univer-

sity Press, 2012.

[76] CELADE-División de Población de la CEPAL y División

de Población de las Naciones Unidas, Estimaciones y pro-

yecciones de población a nivel nacional, de los 20 países

de América Latina , CEPAL, 2023. [Online]. Available:

https://www.cepal.org/es/temas/estimaciones-proyec-

ciones-poblacion/estimaciones-proyecciones-pobla-

cion-nivel-nacional-los-20-paises-america-latina .

[77] UNICEF, Estado Mundial de la Infancia 2024 , Fondo de

las Naciones Unidas para la Infancia, 2024.

[78] Institute for Health Metrics and Evaluation, TL Capstones

global fertility , 2024. [Online]. Available: https://www.

healthdata.org/sites/default/les/2024-03/TL%20Caps-

tones%20global%20fertility_ES.pdf .

[79] Dialoguemos, ¿Cuáles son los países de América Latina

con la natalidad más baja y más alta? , 2024. [Online].

Available: https://dialoguemos.ec/2024/10/cuales-son-

los-paises-de-america-latina-con-la-natalidad-mas-ba-

ja-y-mas-alta/ .

[80] P. Corral, I. Molina, A. Cojocaru, and S. Segovia, Guideli-

nes to Small Area Estimation for Poverty Mapping , Was-

hington: World Bank, 2022.

[81] T. Lynch et al., “Denitions, theories, and measurement of

stress in children,” Journal of Pediatric Nursing , vol. 66,

pp. 202–212, 2022.

[82] G. D. E. Loaiza, “Impacto de la inversión social en la reduc-

ción de la pobreza y la desigualdad en Ecuador: Análisis

de políticas públicas y resultados 2010-2023,” Ciencia y

Educación , vol. 5, no. 12, pp. 122–138, 2024

[83] N. Tzavidis, L. C. Zhang, A. Luna, T. Schmid, and N. Ro-

jas-Perilla, “From start to nish: a framework for the

production of small area ocial statistics,” Journal of the

Royal Statistical Society: Series A (Statistics in Society) ,

vol. 181, no. 4, pp. 927–979, 2018.

[84] Food and Agriculture Organization of the United Nations

(FAO), [Título del recurso] , [Online]. Available: https://

openknowledge.fao.org/items/65139780-d06c-4b7c-

a2cd-3ed4256eaa1c .

[85] Economic Commission for Latin America and the Ca-

ribbean (ECLAC), Social Panorama of Latin Ame-

rica and the Caribbean 2024: Accessible Version ,

[Online]. Available: https://www.cepal.org/en/publica-

tions/81294-social-panorama-latin-america-and-carib-

bean-2024-accessible-version .

[86] S. Parandekar, R. Vos, and D. Winkler, Ecuador: Crisis, po-

verty, and social protection , 2002.

[87] A. V. Banerjee and E. Duo, Poor Economics: A Radical

Rethinking of the Way to Fight Global Poverty , Public

Aairs, 2011.

[88] A. Deaton, e Analysis of Household Surveys: A Microeco-

nometric Approach to Development Policy , World Bank

Publications, 1997.

[89] C. Elbers, J. O. Lanjouw, and P. Lanjouw, “Micro–Level Es-

timation of Poverty and Inequality,” Econometrica , vol.

71, no. 1, pp. 355–364, 2003.

[90] M. F. M. Macías and M. M. B. Salinas, “Programas de ser-

vicios sociales para erradicar la pobreza en el Ecuador,”

ECA Sinergia , vol. 12, no. 1, pp. 59–69, 2021.

[91] G. Wan, “Regional income inequality in rural China,

1985–2002: Trends, causes and policy implications,”

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

in Understanding Inequality and Poverty in China: Me

thods and Applications , London: Palgrave Macmillan

UK, 2008, pp. 89–116.

[92] I. A. Eyzaguirre, A. Y. Iwama, and M. E. Fernandes, “In-

tegrating a conceptual framework for the sustainable

development goals in the mangrove ecosystem: a syste-

matic review,” Environmental Development , p. 100895,

2023.

[93] W. Naudé, M. McGillivray, and S. Rossouw, Measuring the

vulnerability of subnational regions , No. 2008/54, WI-

DER Research Paper, 2008.

[94] S. Alkire and M. E. Santos, “Measuring acute poverty in

the developing world: Robustness and scope of the mul-

tidimensional poverty index,” World Development , vol.

59, pp. 251–274, 2014.

[95] A. Ahrens, C. Aitken, and M. E. Schaer, “Using machine

learning methods to support causal inference in econo-

metrics,” in Behavioral Predictive Modeling in Economics,

2021, pp. 23–52.

[96] P. R. Voss, D. D. Long, R. B. Hammer, and S. Friedman,

“County child poverty rates in the US: a spatial regres-

sion approach,” Population Research and Policy Review ,

vol. 25, pp. 369–391, 2006.

[97] N. Shrestha, “Factor analysis as a tool for survey analysis,”

American journal of Applied Mathematics and statistics

, vol. 9, no. 1, pp. 4–11, 2021.

[98] J. Cohen, Statistical Power Analysis for the Behavioral Scien-

ces , Routledge, 2013.

[99] A. Field, Discovering Statistics Using IBM SPSS Statistics ,

Sage Publications Limited, 2024.

[100] J. Hentschel, Combining census and survey data to study

spatial dimensions of poverty: A case study of Ecuador ,

vol. 1928, World Bank Publications, 1998.

[101] C. Elbers, J. O. Lanjouw, and P. Lanjouw, Micro-level esti-

mation of welfare , vol. 2911, World Bank Publications,

2002.

[102] UNICEF, e state of the world’s children 2006: excluded

and invisible , UNICEF, 2005.

[103] D. Pfeermann, “New important developments in sma-

ll area estimation,” Quality control and applied statistics

, vol. 59, no. 1, pp. 103–106, 2013.

[104] Economic Commission for Latin America and the Cari-

bbean (ECLAC), Small area estimation (SAE): Resources

, [Online]. Available: https://statistics.cepal.org/portal/

sae/resources.html?lang=en .

[105] F. O. Rico, C. F. T. Piñerez, and N. Ramirez-Vargas, “A

Review of the Use of Small Area Estimation in Colom-

bia,” Revista Colombiana de Estadística , vol. 47, no. 2,

p. 407, 2024.

[106] A. Zgodic et al., “Estimates of childhood overweight and

obesity at the region, state, and county levels: a multile-

vel small-area estimation approach,” American journal

of epidemiology , vol. 190, no. 12, pp. 2618–2629, 2021.

[107] T. F. Crossley and L. J. Curtis, “Child pover-

ty in Canada,” Review of Income and Weal-

th , vol. 52, no. 2, pp. 237–260, 2006.

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

Annexes

ln_con Coef. Std. Err. t P>|t| [95% Conf. Interval]

_cons 5,054579 0,1943018 26,01 0,0000 4,673738 5,435420

T_per -0,1849809 0,004099 -45,13 0,0000 -0,1930151 -0,1769466

T_nn 0,1968746 0,0107344 18,34 0,0000 0,1758347 0,2179145

T_me 0,1011308 0,0057637 17,55 0,0000 0,0898337 0,1124279

T_am -0,0602178 0,0076539 -7,87 0,0000 -0,0752198 -0,0452159

T_pcd -0,064163 0,0067434 -9,51 0,0000 -0,0773804 -0,0509455

rat_dp -0,5575776 0,0270586 -20,61 0,0000 -0,6106139 -0,5045413

p03 -0,0042259 0,0015088 -2,8 0,0050 -0,0071832 -0,0012685

edad2_ 0,0000331 0,0000139 2,39 0,0170 5,94E-06 0,0000603

edad_p 0,0125164 0,0015418 8,12 0,0000 0,0094944 0,0155383

edad2_p -0,0001247 0,0000134 -9,32 0,0000 -0,0001509 -0,0000985

rural 0,0212506 0,0074766 2,84 0,0040 0,0065961 0,0359052

p16_2 0,0929644 0,0144991 6,41 0,0000 0,0645455 0,1213832

p16_3 0,1053171 0,0151142 6,97 0,0000 0,0756924 0,1349418

p16_4 0,0987475 0,0092817 10,64 0,0000 0,0805549 0,11694

p16_5 0,1496131 0,0168492 8,88 0,0000 0,116588 0,1826383

mujer -0,0398968 0,0086264 -4,62 0,0000 -0,056805 -0,0229886

nivel_2 0,0463438 0,0150709 3,08 0,0020 0,016804 0,0758836

nivel_3 0,1248864 0,0194194 6,43 0,0000 0,0868235 0,1629493

nivel_4 0,1692531 0,0170471 9,93 0,0000 0,1358399 0,2026662

nivel_5 0,2381168 0,0273787 8,7 0,0000 0,1844531 0,2917805

nivel_6 0,3887886 0,0177058 21,96 0,0000 0,3540843 0,423493

p34_2 -0,031441 0,0070084 -4,49 0,0000 -0,0451778 -0,0177042

p34_3 0,0290773 0,0112747 2,58 0,0100 0,0069783 0,0511762

p34_4 0,0906552 0,0154005 5,89 0,0000 0,0604695 0,1208409

p34_5 0,0514007 0,0122753 4,19 0,0000 0,0273404 0,075461

p34_6 -0,0121323 0,0115162 -1,05 0,2920 -0,0347047 0,01044

analfabeto -0,063958 0,0126419 -5,06 0,0000 -0,0887368 -0,0391793

no_empleo -0,0158142 0,008339 -1,9 0,0580 -0,0321591 0,0005306

migra_d 0,0422481 0,0081524 5,18 0,0000 0,0262689 0,0582272

vap_2 -0,0722079 0,0111353 -6,48 0,0000 -0,0940337 -0,050382

vap_3 -0,0586908 0,0064927 -9,04 0,0000 -0,0714168 -0,0459648

vap_4 -0,0938256 0,0105939 -8,86 0,0000 -0,1145902 -0,073061

vap_5 -0,0520068 0,0411179 -1,26 0,2060 -0,1325999 0,0285863

v09_2 0,0242875 0,0074351 3,27 0,0010 0,0097144 0,0388606

v09_3 -0,0393046 0,011603 -3,39 0,0010 -0,0620472 -0,0165621

v09_4 -0,0308275 0,0185361 -1,66 0,0960 -0,0671593 0,0055042

v09_5 -0,0369643 0,0129003 -2,87 0,0040 -0,0622495 -0,0116791

h01 0,0552463 0,0032684 16,9 0,0000 0,0488401 0,0616525

v08_2 -0,0914424 0,0073258 -12,48 0,0000 -0,1058013 -0,0770835

v08_3 -0,0396555 0,0264744 -1,5 0,1340 -0,0915467 0,0122356

v08_4 -0,1080225 0,0104776 -10,31 0,0000 -0,128559 -0,0874859

h04_2 -0,0503693 0,0129574 -3,89 0,0000 -0,0757663 -0,0249722

h04_3 -0,0933007 0,0085004 -10,98 0,0000 -0,1099619 -0,0766394

telef_f 0,125342 0,007444 16,84 0,0000 0,1107514 0,1399326

internet 0,1345116 0,0081625 16,48 0,0000 0,1185126 0,1505105

tvcable 0,177595 0,00684 25,96 0,0000 0,1641884 0,1910017

h15_3 0,0781627 0,0166848 4,68 0,0000 0,0454597 0,1108656

h15_4 0,0516225 0,0087074 5,93 0,0000 0,0345555 0,0686895

h15_5 0,0231404 0,0096073 2,41 0,0160 0,0043095 0,0419712

h15_6 0,0555352 0,0195722 2,84 0,0050 0,0171727 0,0938978

v12a 0,0252396 0,0008545 29,54 0,0000 0,0235647 0,0269146

te_ad 0,0254308 0,0211323 1,2 0,2290 -0,0159896 0,0668512

pa_ad -0,0574557 0,0254689 -2,26 0,0240 -0,107376 -0,0075354

pi_ad 0,0507765 0,0120756 4,2 0,0000 0,0271077 0,0744453

Table 3.

Estimation of consumption poverty parameters

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

vtv_2 0,0669437 0,0085981 7,79 0,0000 0,0500909 0,0837965

vtv_3 -0,0113254 0,0155061 -0,73 0,4650 -0,0417182 0,0190673

vtv_4 -0,0328144 0,0115216 -2,85 0,0040 -0,0553972 -0,0102316

vtv_5 -0,0803702 0,0277981 -2,89 0,0040 -0,1348559 -0,0258845

cocina_ex 0,0300794 0,0059608 5,05 0,0000 0,0183959 0,0417629

T_per_cl -0,0773363 0,0203935 -3,79 0,0000 -0,1173087 -0,037364

T_nn_cl 0,1625616 0,0536592 3,03 0,0020 0,0573869 0,2677364

T_me_cl -0,0034573 0,0282182 -0,12 0,9020 -0,0587665 0,0518518

T_am_cl 0,0593213 0,0449806 1,32 0,1870 -0,0288428 0,1474854

T_pcd_cl -0,1433981 0,0350462 -4,09 0,0000 -0,2120905 -0,0747058

rat_dp_cl -0,5178252 0,1589009 -3,26 0,0010 -0,8292788 -0,2063715

edad_p_cl 0,0075774 0,0097266 0,78 0,4360 -0,0114873 0,0266421

edad2_p_cl -0,0002427 0,0000813 -2,99 0,0030 -0,000402 -0,0000834

p03_cl -0,00767 0,0095427 -0,8 0,4220 -0,0263743 0,0110342

edad2__cl 0,0000665 0,0000851 0,78 0,4340 -0,0001003 0,0002334

mujer_cl -0,111091 0,0342351 -3,24 0,0010 -0,1781936 -0,0439884

analfabeto_cl 0,1662323 0,0369349 4,5 0,0000 0,093838 0,2386265

no_empleo_cl -0,0140697 0,0450944 -0,31 0,7550 -0,1024571 0,0743176

migra_d_cl 0,1869325 0,041368 4,52 0,0000 0,1058491 0,268016

h01_cl -0,1104284 0,01241 -8,9 0,0000 -0,1347527 -0,0861041

telef_f_cl 0,1285071 0,0255726 5,03 0,0000 0,0783834 0,1786307

internet_cl 0,1339712 0,0407815 3,29 0,0010 0,0540374 0,213905

tvcable_cl -0,0778569 0,0216295 -3,6 0,0000 -0,1202519 -0,035462

v12a_cl 0,0341016 0,0034544 9,87 0,0000 0,0273309 0,0408724

te_ad_cl 0,7739045 0,058824 13,16 0,0000 0,6586066 0,8892025

pa_ad_cl -0,1189062 0,0333034 -3,57 0,0000 -0,1841826 -0,0536298

pi_ad_cl -0,0504654 0,0288163 -1,75 0,0800 -0,1069468 0,0060159

cocina_ex_cl -0,019565 0,017731 -1,1 0,2700 -0,0543185 0,0151886

Source: [15]

Income elasticity of poverty reduction -0,0887 -0,0887 -0,0887 -0,0887 -0,0887 -0,0887 -0,0887 -0,0887 -0,0887 -0,0887 -0,0887 -0,0887 -0,0887 -0,0887 -0,0887 -0,0887

Geographical Identier id 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030

National 1 434.926 396.349 361.194 329.157 299.962 273.356 249.110 227.015 206.879 188.529 171.807 156.568 142.681 130.026 118.493 107.983 98.405

Esmeraldas, Imbabura, Carchi, Sucumbíos. Z1 52.180 47.551 43.333 39.489 35.986 32.794 29.885 27.234 24.818 22.617 20.611 18.783 17.117 15.599 14.215 12.954 11.805

Pichincha (excepto el cantón Quito), Napo, Orellana Z2 22.378 20.393 18.584 16.936 15.434 14.065 12.817 11.680 10.644 9.700 8.840 8.056 7.341 6.690 6.097 5.556 5.063

Cotopaxi, Tungurahua, Chimborazo, Pastaza Z3 54.391 49.566 45.170 41.164 37.513 34.186 31.154 28.391 25.873 23.578 21.487 19.581 17.844 16.261 14.819 13.505 12.307

Manabí, Santo Domingo de los Tsáchilas Z4 61.528 56.070 51.097 46.565 42.435 38.671 35.241 32.115 29.266 26.670 24.304 22.148 20.184 18.394 16.762 15.275 13.920

Santa Elena, Guayas (excepto los cantones de Gua-

yaquil, Samborondón y Durán), Bolívar, Los Ríos y

Galápagos

Z5 89.586 81.639 74.398 67.799 61.785 56.305 51.311 46.760 42.612 38.832 35.388 32.249 29.389 26.782 24.406 22.241 20.268

Cañar, Azuay, Morona Santiago. Z6 29.512 26.894 24.509 22.335 20.354 18.549 16.904 15.405 14.039 12.794 11.659 10.625 9.683 8.824 8.041 7.328 6.678

El Oro, Loja, Zamora Chinchipe Z7 26.193 23.869 21.752 19.823 18.065 16.463 15.003 13.672 12.459 11.354 10.347 9.429 8.593 7.831 7.136 6.503 5.926

Guayaquil, Samborondón y Durán Z8 64.053 58.372 53.195 48.477 44.177 40.259 36.688 33.434 30.468 27.766 25.303 23.059 21.014 19.150 17.451 15.903 14.492

Distrito Metropolitano de Quito. Z9 35.107 31.993 29.155 26.569 24.212 22.064 20.107 18.324 16.699 15.218 13.868 12.638 11.517 10.495 9.564 8.716 7.943

AZUAY P01 20.734 18.895 17.219 15.692 14.300 13.032 11.876 10.823 9.863 8.988 8.191 7.464 6.802 6.199 5.649 5.148 4.691

BOLÍVAR P02 8.078 7.362 6.709 6.114 5.572 5.078 4.628 4.218 3.844 3.503 3.192 2.909 2.651 2.416 2.202 2.007 1.829

CAÑAR P03 2.635 2.401 2.188 1.994 1.817 1.656 1.509 1.375 1.253 1.142 1.041 949 865 788 718 654 596

CARCHI P04 4.301 3.919 3.571 3.254 2.965 2.702 2.462 2.244 2.045 1.864 1.699 1.548 1.411 1.286 1.172 1.068 973

COTOPAXI P05 13.494 12.297 11.206 10.212 9.306 8.481 7.729 7.043 6.418 5.849 5.330 4.857 4.426 4.033 3.675 3.349 3.052

CHIMBORAZO P06 24.929 22.718 20.703 18.867 17.194 15.669 14.279 13.012 11.858 10.806 9.848 8.975 8.179 7.454 6.793 6.190 5.641

EL ORO P07 7.867 7.169 6.533 5.954 5.426 4.945 4.506 4.106 3.742 3.410 3.108 2.832 2.581 2.352 2.143 1.953 1.780

ESMERALDAS P08 25.358 23.108 21.058 19.190 17.488 15.937 14.523 13.235 12.061 10.991 10.016 9.128 8.318 7.580 6.908 6.295 5.737

GUAYAS P09 99.474 90.651 82.610 75.283 68.606 62.521 56.976 51.922 47.317 43.120 39.295 35.810 32.634 29.739 27.101 24.697 22.506

IMBABURA P10 16.440 14.982 13.653 12.442 11.338 10.332 9.416 8.581 7.820 7.126 6.494 5.918 5.393 4.915 4.479 4.082 3.720

LOJA P11 13.794 12.570 11.455 10.439 9.513 8.669 7.900 7.199 6.560 5.978 5.448 4.965 4.525 4.124 3.758 3.425 3.121

LOS RÍOS P12 28.339 25.826 23.535 21.447 19.545 17.811 16.231 14.791 13.479 12.283 11.194 10.201 9.296 8.471 7.720 7.035 6.411

MANABÍ P13 51.156 46.619 42.484 38.716 35.282 32.153 29.301 26.702 24.334 22.176 20.209 18.417 16.783 15.294 13.937 12.701 11.574

MORONA SANTIAGO P14 6.142 5.598 5.101 4.649 4.237 3.861 3.519 3.207 2.923 2.664 2.428 2.213 2.017 1.838 1.675 1.526 1.391

NAPO P15 7.121 6.490 5.914 5.389 4.911 4.475 4.078 3.716 3.386 3.086 2.812 2.563 2.336 2.129 1.940 1.768 1.611

PASTAZA P16 2.124 1.936 1.764 1.608 1.465 1.335 1.217 1.109 1.011 921 839 765 697 635 579 528 481

PICHINCHA P17 46.329 42.220 38.475 35.062 31.952 29.118 26.535 24.181 22.036 20.081 18.300 16.677 15.198 13.850 12.622 11.502 10.482

TUNGURAHUA P18 13.843 12.615 11.496 10.476 9.547 8.700 7.928 7.225 6.584 6.000 5.468 4.983 4.541 4.138 3.771 3.437 3.132

Table 4.

SCENARIO 1: Estimated consumption poverty of children aged 0-3 years, according to the methodology (SAE) up to 2030

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

ZAMORA CHINCHIPE P19 4.532 4.130 3.764 3.430 3.126 2.849 2.596 2.366 2.156 1.965 1.791 1.632 1.487 1.355 1.235 1.125 1.025

GALÁPAGOS P20 20 18 16 15 14 13 12 11 10 9 8 7 6 5 5 5 5

SUCUMBÍOS P21 6.081 5.542 5.050 4.602 4.194 3.822 3.483 3.174 2.892 2.635 2.401 2.188 1.994 1.817 1.656 1.509 1.375

ORELLANA P22 4.035 3.677 3.351 3.054 2.783 2.536 2.311 2.106 1.919 1.749 1.594 1.453 1.324 1.207 1.100 1.002 913

SANTO DOMINGO DE LOS TSÁCHILAS P23 10.371 9.451 8.613 7.849 7.153 6.519 5.941 5.414 4.934 4.496 4.097 3.734 3.403 3.101 2.826 2.575 2.347

SANTA ELENA P24 17.726 16.154 14.721 13.415 12.225 11.141 10.153 9.252 8.431 7.683 7.002 6.381 5.815 5.299 4.829 4.401 4.011

Income elasticity of poverty reduction -0,0637 -0,0187 0,0359 -0,0828 -0,0318 -0,0318 -0,0318 -0,0630 -0,0329 -0,0164 -0,0164 -0,0164 -0,0164 -0,0164 -0,0164 -0,0164

Geographical Identier id 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030

National 1 434.926 407.221 399.616 413.965 379.685 367.592 355.884 344.549 322.833 312.228 307.099 302.055 297.094 292.214 287.414 282.693 278.050

Esmeraldas, Imbabura, Carchi, Sucumbíos. Z1 52.180 48.856 47.944 49.666 45.553 44.102 42.697 41.337 38.732 37.460 36.845 36.240 35.645 35.060 34.484 33.918 33.361

Pichincha (excepto el cantón Quito), Napo, Orellana Z2 22.378 20.953 20.562 21.300 19.536 18.914 18.312 17.729 16.612 16.066 15.802 15.542 15.287 15.036 14.789 14.546 14.307

Cotopaxi, Tungurahua, Chimborazo, Pastaza Z3 54.391 50.926 49.975 51.769 47.482 45.970 44.506 43.088 40.372 39.046 38.405 37.774 37.154 36.544 35.944 35.354 34.773

Manabí, Santo Domingo de los Tsáchilas Z4 61.528 57.608 56.532 58.562 53.712 52.001 50.345 48.742 45.670 44.170 43.444 42.730 42.028 41.338 40.659 39.991 39.334

Santa Elena, Guayas (excepto los cantones de Guayaquil,

Samborondón y Durán), Bolívar, Los Ríos y Galápagos Z5 89.586 83.879 82.313 85.269 78.208 75.717 73.305 70.970 66.497 64.313 63.257 62.218 61.196 60.191 59.202 58.230 57.274

Cañar, Azuay, Morona Santiago. Z6 29.512 27.632 27.116 28.090 25.764 24.943 24.149 23.380 21.906 21.186 20.838 20.496 20.159 19.828 19.502 19.182 18.867

El Oro, Loja, Zamora Chinchipe Z7 26.193 24.524 24.066 24.930 22.866 22.138 21.433 20.750 19.442 18.803 18.494 18.190 17.891 17.597 17.308 17.024 16.744

Guayaquil, Samborondón y Durán Z8 64.053 59.973 58.853 60.966 55.917 54.136 52.412 50.743 47.545 45.983 45.228 44.485 43.754 43.035 42.328 41.633 40.949

Distrito Metropolitano de Quito. Z9 35.107 32.871 32.257 33.415 30.648 29.672 28.727 27.812 26.059 25.203 24.789 24.382 23.982 23.588 23.201 22.820 22.445

AZUAY P01 20.734 19.413 19.050 19.734 18.100 17.524 16.966 16.426 15.391 14.885 14.641 14.401 14.164 13.931 13.702 13.477 13.256

BOLÍVAR P02 8.078 7.564 7.423 7.690 7.053 6.828 6.611 6.400 5.997 5.800 5.705 5.611 5.519 5.428 5.339 5.251 5.165

CAÑAR P03 2.635 2.467 2.421 2.508 2.300 2.227 2.156 2.087 1.955 1.891 1.860 1.829 1.799 1.769 1.740 1.711 1.683

CARCHI P04 4.301 4.027 3.952 4.094 3.755 3.635 3.519 3.407 3.192 3.087 3.036 2.986 2.937 2.889 2.842 2.795 2.749

COTOPAXI P05 13.494 12.634 12.398 12.843 11.779 11.404 11.041 10.689 10.015 9.686 9.527 9.371 9.217 9.066 8.917 8.771 8.627

CHIMBORAZO P06 24.929 23.341 22.905 23.727 21.762 21.069 20.398 19.748 18.503 17.895 17.601 17.312 17.028 16.748 16.473 16.202 15.936

EL ORO P07 7.867 7.366 7.228 7.488 6.868 6.649 6.437 6.232 5.839 5.647 5.554 5.463 5.373 5.285 5.198 5.113 5.029

ESMERALDAS P08 25.358 23.742 23.299 24.136 22.137 21.432 20.749 20.088 18.822 18.204 17.905 17.611 17.322 17.037 16.757 16.482 16.211

GUAYAS P09 99.474 93.138 91.399 94.681 86.840 84.074 81.396 78.804 73.837 71.411 70.238 69.084 67.949 66.833 65.735 64.655 63.593

IMBABURA P10 16.440 15.393 15.106 15.648 14.352 13.895 13.452 13.024 12.203 11.802 11.608 11.417 11.229 11.045 10.864 10.686 10.510

LOJA P11 13.794 12.915 12.674 13.129 12.042 11.658 11.287 10.928 10.239 9.903 9.740 9.580 9.423 9.268 9.116 8.966 8.819

LOS RÍOS P12 28.339 26.534 26.038 26.973 24.739 23.951 23.188 22.449 21.034 20.343 20.009 19.680 19.357 19.039 18.726 18.418 18.115

MANABÍ P13 51.156 47.898 47.004 48.692 44.660 43.238 41.861 40.528 37.974 36.727 36.124 35.531 34.947 34.373 33.808 33.253 32.707

MORONA SANTIAGO P14 6.142 5.751 5.644 5.847 5.363 5.192 5.027 4.867 4.560 4.410 4.338 4.267 4.197 4.128 4.060 3.993 3.927

NAPO P15 7.121 6.668 6.543 6.778 6.217 6.019 5.827 5.641 5.285 5.111 5.027 4.944 4.863 4.783 4.704 4.627 4.551

PASTAZA P16 2.124 1.989 1.952 2.022 1.855 1.796 1.739 1.684 1.578 1.526 1.501 1.476 1.452 1.428 1.405 1.382 1.359

PICHINCHA P17 46.329 43.378 42.568 44.096 40.444 39.156 37.909 36.702 34.389 33.259 32.713 32.176 31.647 31.127 30.616 30.113 29.618

TUNGURAHUA P18 13.843 12.961 12.719 13.176 12.085 11.700 11.327 10.966 10.275 9.937 9.774 9.613 9.455 9.300 9.147 8.997 8.849

ZAMORA CHINCHIPE P19 4.532 4.243 4.164 4.314 3.957 3.831 3.709 3.591 3.365 3.254 3.201 3.148 3.096 3.045 2.995 2.946 2.898

GALÁPAGOS P20 20 19 19 20 18 17 16 15 14 14 14 14 14 14 14 14 14

SUCUMBÍOS P21 6.081 5.694 5.588 5.789 5.310 5.141 4.977 4.818 4.514 4.366 4.294 4.223 4.154 4.086 4.019 3.953 3.888

ORELLANA P22 4.035 3.778 3.707 3.840 3.522 3.410 3.301 3.196 2.995 2.897 2.849 2.802 2.756 2.711 2.666 2.622 2.579

SANTO DOMINGO DE LOS TSÁCHILAS P23 10.371 9.711 9.530 9.872 9.055 8.767 8.488 8.218 7.700 7.447 7.325 7.205 7.087 6.971 6.856 6.743 6.632

SANTA ELENA P24 17.726 16.597 16.287 16.872 15.475 14.982 14.505 14.043 13.158 12.726 12.517 12.311 12.109 11.910 11.714 11.522 11.333

Source: [15]

Table 5.

SCENARIO 2: Estimated consumption poverty of children aged 0-3 years, according to the methodology (SAE) up to 2030

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

Geographical Identier Incidence (%) Standard deviation

(%)

Coecient of variation

(%)

1National 41,1 1,03 2,54

1 Planning zone 1 53,9 1,20 2,23

2 Planning zone 2 47,7 1,70 3,58

3 Planning zone 3 53,1 1,53 2,89

4 Planning zone 4 48,0 2,54 5,28

5 Planning zone 5 50,3 1,71 3,39

6 Planning zone 6 35,2 2,70 7,69

7 Planning zone 7 31,6 2,59 8,21

8 Planning zone 8 33,6 2,72 8,10

9 Planning zone 9 22,3 1,95 8,75

1Napo 81,3 4,23 11,15

6 Chimborazo 79,1 4,98 7,47

24 Santa Elena 67,4 3,18 22,72

2Bolívar 66,1 3,54 9,42

5 Zamora Chinchipe 60,9 3,07 6,93

8 Esmeraldas 59,8 3,62 4,61

10 Imbabura 55,4 3,67 20,34

8Morona Santiago 54,4 1,89 3,18

9Manabí 53,1 2,67 7,16

10 Los Ríos 48,2 1,96 3,52

11 Sucumbíos 45,6 4,05 9,57

5Cotopaxi 45,4 2,26 4,87

11 Loja 42,4 3,21 6,10

18 Tungurahua 40,6 2,56 4,67

1 Azuay 39,4 2,31 2,86

9Guayas 37,6 4,34 12,27

4Carchi 37,4 1,98 7,86

16 Pastaza 36,9 2,79 6,98

19 Orellana 36,3 6,06 9,66

23 Santo Domingo de los Tsáchilas 34,1 0,46 47,73

17 Pichincha 25,1 4,37 9,63

7 El Oro 18,5 4,09 11,55

3Cañar 15,2 2,71 8,02

20 Galápagos 1,2 6,57 9,98

0101 Cuenca 32,6 3,72 11,43

0102 Girón 50,7 6,93 13,67

0103 Gualaceo 46,5 6,01 12,92

0104 Nabón 63,8 6,31 9,89

0105 Paute 46,3 5,89 12,73

0106 Pucara 62,4 8,13 13,02

0107 San Fernando 44,5 6,91 15,52

0108 Santa Isabel 46,4 5,57 12,01

0109 Sigsig 53,6 6,14 11,45

Table 6.

Estimation of consumption poverty of children aged 0-3 years, according to the methodology (SAE).

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

0110 Oña 69,4 7,18 10,35

0111 Chordeleg 42,7 5,95 13,93

0112 El Pan 48,1 9,82 20,39

0113 Sevilla de Oro 49,5 7,29 14,72

0114 Guachapala 38,8 10,65 27,49

0115 Camilo Ponce Enríquez 52,5 5,32 10,14

0201 Guaranda 67,1 4,40 6,56

0202 Chillanes 77,5 5,80 7,48

0203 Chimbo 65,5 5,99 9,16

0204 Echeandía 63,7 6,16 9,68

0205 San Miguel 64,2 5,79 9,01

0206 Caluma 57,3 6,84 11,94

0207 Las Naves 70,5 6,75 9,58

0301 Azogues 6,7 2,79 41,73

0302 Biblián 4,8 2,93 60,87

0303 Cañar 28,7 3,77 13,14

0304 La Troncal 10,2 4,40 42,96

0305 El Tambo 21,1 4,46 21,12

0306 Deleg 6,0 3,23 53,37

0307 Suscal 26,7 7,18 26,91

0401 Tulcán 30,2 3,79 12,54

0402 Bolívar 52,6 4,99 9,50

0403 Espejo 40,9 4,98 12,16

0404 Mira 54,3 4,44 8,18

0405 Montufar 45,9 4,18 9,10

0406 San Pedro de Huaca 28,2 5,71 20,26

0501 Latacunga 37,3 2,66 7,13

0502 La Maná 45,4 3,51 7,73

0503 Pangua 57,1 4,23 7,41

0504 Pujilí 51,7 4,99 9,66

0505 Salcedo 40,8 3,11 7,63

0506 Saquisilí 49,9 3,92 7,86

0507 Sigchos 61,6 5,05 8,20

0601 Riobamba 68,3 5,03 7,37

0602 Alausí 86,5 2,66 3,07

0603 Colta 95,5 1,80 1,88

0604 Chambo 78,8 4,23 5,37

0605 Chunchi 75,1 6,34 8,44

0606 Guamote 97,8 1,41 1,44

0607 Guano 84,5 3,99 4,72

0608 Pallatanga 84,4 3,90 4,62

0609 Penipe 79,1 6,19 7,82

0610 Cumanda 76,9 6,32 8,22

0701 Machala 14,5 3,35 23,15

0702 Arenillas 22,6 4,23 18,74

0703 Atahualpa 14,5 5,07 35,07

0704 Balsas 20,1 4,99 24,78

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

0705 Chilla 31,4 8,78 27,94

0706 El Guabo 27,1 4,28 15,76

0707 Huaquillas 16,8 3,90 23,26

0708 Marcabelí 26,6 5,57 20,91

0709 Pasaje 21,7 3,91 18,05

0710 Piñas 12,0 4,32 36,00

0711 Portovelo 20,1 4,69 23,35

0712 Santa Rosa 18,1 4,01 22,16

0713 Zaruma 19,9 5,28 26,48

0714 Las Lajas 22,6 6,05 26,81

0801 Esmeraldas 47,9 2,15 4,48

0802 Eloy Alfaro 73,3 2,77 3,78

0803 Muisne 75,1 2,95 3,92

0804 Quinindé 65,6 2,17 3,31

0805 San Lorenzo 62,5 3,22 5,15

0806 Atacames 59,1 2,70 4,58

0807 Rioverde 78,5 2,60 3,31

2302 La Concordia (*) 57,5 2,96 5,15

0901 Guayaquil 33,6 2,80 8,32

0902 Alfredo Baquerizo Moreno 57,9 3,92 6,77

0903 Balao 47,5 4,63 9,74

0904 Balzar 50,4 3,50 6,94

0905 Colimes 55,5 4,10 7,39

0906 Daule 37,0 2,29 6,19

0907 Duran 35,1 2,69 7,67

0908 Empalme 56,8 3,41 6,00

0909 El Triunfo 48,2 3,11 6,47

0910 Milagro 41,7 2,98 7,14

0911 Naranjal 48,2 3,28 6,80

0912 Naranjito 47,7 3,35 7,02

0913 Palestina 48,7 3,48 7,15

0914 Pedro Carbo 42,4 4,00 9,44

0916 Samborondón 26,3 1,91 7,26

0918 Santa Lucía 47,6 4,24 8,92

0919 Salitre 55,4 3,96 7,15

0920 San Jacinto de Yaguachi 47,8 3,21 6,72

0921 Playas 37,1 4,68 12,62

0922 Simón Bolívar 52,2 3,82 7,32

0923 Coronel Marcelino Maridueña 42,9 3,17 7,38

0924 Lomas de Sargentillo 39,2 4,32 11,03

0925 Nobol 44,9 4,01 8,92

0927 General Antonio Elizalde 45,7 2,99 6,54

0928 Isidro Ayora 42,9 4,12 9,60

1001 Ibarra 39,3 1,89 4,80

1002 Antonio Ante 52,7 2,06 3,91

1003 Cotacachi 74,6 2,49 3,34

1004 Otavalo 69,6 2,51 3,60

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

1005 Pimampiro 70,3 2,66 3,78

1006 San Miguel de Urcuquí 79,0 3,09 3,91

1101 Loja 29,3 3,88 13,26

1102 Calvas 49,3 4,50 9,14

1103 Catamayo 43,3 5,45 12,58

1104 Celica 51,7 4,73 9,15

1105 Chaguarpamba 59,5 5,07 8,52

1106 Espíndola 68,7 4,51 6,56

1107 Gonzanama 62,1 4,99 8,04

1108 Macara 42,6 5,02 11,78

1109 Paltas 58,3 4,05 6,96

1110 Puyango 52,7 5,00 9,48

1111 Saraguro 61,9 4,83 7,79

1112 Sozoranga 67,0 5,53 8,26

1113 Zapotillo 61,6 5,44 8,83

1114 Pindal 56,0 4,83 8,62

1115 Quilanga 59,4 5,75 9,68

1116 Olmedo 67,6 5,77 8,54

1201 Babahoyo 40,5 2,52 6,21

1202 Baba 55,0 2,69 4,89

1203 Montalvo 40,0 2,77 6,92

1204 Puebloviejo 47,5 3,20 6,75

1205 Quevedo 40,4 2,64 6,53

1206 Urdaneta 45,7 2,76 6,03

1207 Ventanas 43,7 2,85 6,53

1208 Vinces 48,7 2,44 5,01

1209 Palenque 61,7 2,74 4,44

1210 Buena Fe 52,6 2,96 5,62

1211 Valencia 56,1 2,90 5,18

1212 Mocache 57,6 2,50 4,34

1213 Quinsaloma 53,6 2,53 4,71

1301 Portoviejo 43,3 3,65 8,43

1302 Bolívar 59,3 2,90 4,89

1303 Chone 55,7 2,82 5,06

1304 El Carmen 47,8 3,56 7,46

1305 Flavio Alfaro 58,8 3,85 6,54

1306 Jipijapa 65,7 4,86 7,41

1307 Junín 63,5 4,02 6,33

1308 Manta 38,8 3,30 8,50

1309 Montecristi 50,0 5,85 11,69

1310 Paján 71,4 3,56 4,99

1311 Pichincha 64,8 3,39 5,23

1312 Rocafuerte 48,9 4,43 9,07

1313 Santa Ana 68,0 3,01 4,42

1314 Sucre 57,2 4,10 7,16

1315 Tosagua 56,9 3,96 6,97

1316 24 de Mayo 75,6 2,98 3,94

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

1317 Pedernales 63,1 3,25 5,15

1318 Olmedo 73,8 3,00 4,06

1319 Puerto López 63,0 5,45 8,66

1320 Jama 75,9 4,12 5,42

1321 Jaramijó 57,3 6,86 11,97

1322 San Vicente 55,5 4,70 8,46

1401 Morona 47,1 3,38 7,17

1402 Gualaquiza 50,5 2,47 4,89

1403 Limón Indanza 53,3 2,87 5,39

1404 Palora 56,0 3,25 5,81

1405 Santiago 59,5 3,64 6,12

1406 Sucúa 34,9 3,94 11,29

1407 Huamboya 86,7 2,91 3,36

1408 San Juan Bosco 51,7 4,11 7,96

1409 Taisha 87,5 2,38 2,72

1410 Logroño 68,9 3,57 5,18

1411 Pablo Sexto 59,7 5,48 9,19

1412 Tiwintza 77,0 2,72 3,53

1501 Tena 77,8 2,18 2,80

1503 Archidona 89,7 2,10 2,34

1504 El Chaco 77,9 5,17 6,64

1507 Quijos 68,0 5,37 7,89

1509 Carlos Julio Arosemena Tola 92,7 3,43 3,70

1601 Pastaza 35,2 4,23 12,03

1602 Mera 30,4 4,34 14,29

1603 Santa Clara 44,1 6,41 14,55

1604 Arajuno 47,3 8,31 17,56

1701 Quito 22,3 1,95 8,75

1702 Cayambe 53,3 2,73 5,13

1703 Mejía 35,5 3,07 8,65

1704 Pedro Moncayo 51,1 3,35 6,56

1705 Rumiñahui 21,8 1,96 9,00

1707 San Miguel de los Bancos 53,0 4,41 8,31

1708 Pedro Vicente Maldonado 49,9 3,92 7,85

1709 Puerto Quito 55,0 3,74 6,81

1801 Ambato 38,2 2,56 6,70

1802 Baños de Agua Santa 26,7 2,98 11,16

1803 Cevallos 31,0 4,00 12,89

1804 Mocha 39,2 5,16 13,16

1805 Patate 48,1 3,65 7,58

1806 Quero 52,3 4,79 9,15

1807 San Pedro de Pelileo 45,1 3,86 8,55

1808 Santiago de Pillaro 45,4 3,81 8,40

1809 Tisaleo 44,5 4,25 9,56

1901 Zamora 53,4 6,62 12,39

1902 Chinchipe 61,2 9,14 14,92

1903 Nangaritza 81,2 5,83 7,18

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

1904 Yacuambi 71,8 6,39 8,90

1905 Yantzaza 65,1 5,91 9,08

1906 El Pangui 60,4 4,84 8,02

1907 Centinela del Cóndor 59,9 5,98 9,99

1908 Palanda 65,1 9,07 13,93

1909 Paquisha 74,4 9,19 12,35

2001 San Cristóbal 0,3 0,33 131,97

2002 Isabela 0,6 0,85 145,36

2003 Santa Cruz 1,4 0,68 48,17

2101 Lago Agrio 40,7 4,63 11,39

2102 Gonzalo Pizarro 43,1 5,03 11,66

2103 Putumayo 63,0 4,84 7,69

2104 Shushundi 47,8 4,48 9,36

2105 Sucumbíos 35,7 5,97 16,72

2106 Cascales 56,5 5,30 9,37

2107 Cuyabeno 59,6 5,40 9,05

2201 Orellana 27,5 4,38 15,95

2202 Aguarico 69,2 4,59 6,64

2203 La Joya de los Sachas 32,6 4,81 14,74

2204 Loreto 63,3 3,66 5,78

2301 Santo Domingo 33,8 2,71 8,02

2401 Santa Elena 66,7 6,63 9,94

2402 La Libertad 64,2 7,55 11,76

2403 Salinas 66,6 6,05 9,08

9901 Las Golondrinas (**) 65,6 2,17 3,31

9903 Manga del Cura (**) 47,8 3,56 7,46

9904 El Piedrero (**) 48,2 3,11 6,47

01D01 01D01 31,5 3,86 12,25

01D02 01D02 33,5 3,63 10,83

01D03 01D03 51,1 6,28 12,30

01D04 01D04 45,7 5,91 12,94

01D05 01D05 64,9 6,30 9,71

01D06 01D06 46,3 6,47 13,97

01D07 01D07 52,5 5,32 10,14

01D08 01D08 53,6 6,14 11,45

02D01 02D01 67,1 4,40 6,56

02D02 02D02 77,5 5,80 7,48

02D03 02D03 64,7 5,79 8,94

02D04 02D04 62,1 6,32 10,19

03D01 03D01 6,2 2,72 43,68

03D02 03D02 27,6 3,85 13,95

03D03 03D03 10,2 4,40 42,96

04D01 04D01 30,1 3,86 12,84

04D02 04D02 48,0 3,93 8,19

04D03 04D03 46,9 4,45 9,49

05D01 05D01 37,3 2,66 7,13

05D02 05D02 45,4 3,51 7,73

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

05D03 05D03 57,1 4,23 7,41

05D04 05D04 51,2 4,61 9,01

05D05 05D05 61,6 5,05 8,20

05D06 05D06 40,8 3,11 7,63

06D01 06D01 68,8 4,95 7,19

06D02 06D02 83,9 3,45 4,11

06D03 06D03 80,3 5,08 6,32

06D04 06D04 96,9 1,44 1,49

06D05 06D05 83,8 4,11 4,91

07D01 07D01 24,3 4,00 16,49

07D02 07D02 14,5 3,35 23,15

07D03 07D03 19,3 4,74 24,58

07D04 07D04 15,9 4,26 26,87

07D05 07D05 19,0 3,87 20,37

07D06 07D06 18,1 4,01 22,16

08D01 08D01 47,9 2,15 4,48

08D02 08D02 73,3 2,77 3,78

08D03 08D03 65,0 2,46 3,79

08D04 08D04 65,6 2,17 3,31

08D05 08D05 62,5 3,22 5,15

08D06 08D06 78,5 2,60 3,31

08D07 08D07 57,5 2,96 5,15

09D01 09D01 32,9 2,70 8,20

09D02 09D02 35,2 2,87 8,15

09D03 09D03 24,6 2,54 10,31

09D04 09D04 32,7 2,87 8,79

09D05 09D05 11,6 1,47 12,65

09D06 09D06 27,4 2,68 9,77

09D07 09D07 32,8 2,84 8,65

09D08 09D08 51,4 3,80 7,39

09D09 09D09 18,3 1,64 8,99

09D10 09D10 47,3 4,53 9,56

09D11 09D11 55,0 3,64 6,62

09D12 09D12 48,0 3,54 7,37

09D13 09D13 51,3 3,39 6,61

09D14 09D14 41,7 3,91 9,37

09D15 09D15 56,8 3,41 6,00

09D16 09D16 47,7 2,98 6,26

09D17 09D17 41,7 2,98 7,14

09D18 09D18 46,6 3,14 6,73

09D19 09D19 40,1 2,76 6,89

09D20 09D20 55,4 3,96 7,15

09D21 09D21 47,8 3,21 6,72

09D22 09D22 37,1 4,68 12,62

09D23 09D23 26,3 1,91 7,26

09D24 09D24 35,1 2,69 7,67

10D01 10D01 44,4 1,85 4,17

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

10D02 10D02 64,8 2,27 3,51

10D03 10D03 74,6 2,49 3,34

11D01 11D01 29,3 3,88 13,26

11D02 11D02 48,2 5,15 10,69

11D03 11D03 58,3 4,05 6,96

11D04 11D04 53,0 4,59 8,65

11D05 11D05 68,7 4,51 6,56

11D06 11D06 53,3 4,51 8,46

11D07 11D07 50,0 4,92 9,84

11D08 11D08 61,9 4,83 7,79

11D09 11D09 61,6 5,44 8,83

12D01 12D01 43,2 2,37 5,48

12D02 12D02 46,8 2,81 6,02

12D03 12D03 43,5 2,41 5,55

12D04 12D04 45,7 2,64 5,78

12D05 12D05 51,8 2,36 4,55

12D06 12D06 54,0 2,83 5,24

13D01 13D01 43,4 3,66 8,43

13D02 13D02 42,7 3,95 9,25

13D03 13D03 64,9 4,90 7,55

13D04 13D04 71,1 2,80 3,94

13D05 13D05 47,8 3,56 7,46

13D06 13D06 60,7 3,09 5,10

13D07 13D07 56,2 2,88 5,13

13D08 13D08 64,8 3,39 5,23

13D09 13D09 71,4 3,56 4,99

13D10 13D10 66,9 3,27 4,88

13D11 13D11 56,8 4,18 7,36

13D12 13D12 52,9 4,03 7,61

14D01 14D01 47,1 3,38 7,17

14D02 14D02 71,7 2,59 3,61

14D03 14D03 41,7 3,57 8,56

14D04 14D04 50,7 2,57 5,07

14D05 14D05 87,5 2,38 2,72

14D06 14D06 60,0 2,65 4,41

15D01 15D01 81,8 2,05 2,51

15D02 15D02 73,8 5,03 6,81

16D01 16D01 34,7 4,19 12,08

16D02 16D02 47,3 8,31 17,56

17D01 17D01 47,0 3,73 7,95

17D02 17D02 22,2 2,75 12,37

17D03 17D03 23,1 1,92 8,32

17D04 17D04 24,2 2,09 8,65

17D05 17D05 16,4 1,51 9,24

17D06 17D06 18,6 1,88 10,13

17D07 17D07 26,2 2,27 8,67

17D08 17D08 24,1 2,15 8,91

Demographic Trends: Forecast of the Child Population From 0 to 3 years Old for the Year 2030

17D09 17D09 26,8 2,40 8,98

17D10 17D10 52,6 2,83 5,38

17D11 17D11 29,0 2,44 8,40

17D12 17D12 52,9 3,65 6,89

18D01 18D01 40,9 2,42 5,92

18D02 18D02 36,2 2,72 7,52

18D03 18D03 26,7 2,98 11,16

18D04 18D04 45,7 3,72 8,14

18D05 18D05 45,4 3,81 8,40

18D06 18D06 44,6 4,15 9,30

19D01 19D01 57,0 6,11 10,71

19D02 19D02 70,6 6,26 8,87

19D03 19D03 63,3 8,86 14,01

19D04 19D04 63,7 5,44 8,54

20D01 20D01 1,0 0,46 47,73

21D01 21D01 48,0 4,58 9,53

21D02 21D02 40,7 4,63 11,39

21D03 21D03 61,5 4,86 7,91

21D04 21D04 47,8 4,48 9,36

22D01 22D01 32,6 4,81 14,74

22D02 22D02 35,2 4,02 11,43

22D03 22D03 69,2 4,59 6,64

23D01 23D01 31,9 2,70 8,46

23D02 23D02 35,9 2,78 7,76

23D03 23D03 57,5 2,96 5,15

24D01 24D01 66,7 6,63 9,94

24D02 24D02 65,2 6,67 10,23

(*) Updated code

(**) Aggregates, value of nearby locality is assumed.

Source: [15]