Effect of Conditional Cash Transfers on Credit Market Decisions

17608 words (70 pages) Full Dissertation in Full Dissertations

06/06/19 Full Dissertations Reference this

Disclaimer: This work has been submitted by a student. This is not an example of the work produced by our Dissertation Writing Service. You can view samples of our professional work here.

Any opinions, findings, conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of UK Essays.

DO CONDITIONAL CASH TRANSFERS FACILITATE OR HINDER CREDIT MARKET PARTICIPATION?

Analysing the effect of CCTs on credit market decisions in rural Nicaragua.

Abstract

This dissertation researches the effect Conditional Cash Transfers (CCT) may have in credit application. Changes in credit markets due to the reception of a transfer (public or private) are ambiguous theoretically speaking. On one side, credit application could be motivated by the risk-protection the transfer brings and its change in the household production mix. In addition, it can improve the creditworthiness perceived by lenders. On the other side, the decreasing expected marginal returns of a loan after a transfer could disincentive credit application. An empirical evaluation is carried to discard the ambiguity. Using a panel data survey from poor households in Nicaragua, a Probit regression is run. The results show the CCT decreased the likelihood of requesting a loan in rural Nicaragua in a significant way. It is possible that getting a loan to have additional liquidity may have had a higher relative cost compared to the transfer’s actual cost, which is the conditionality. This may have occurred because its beneficiaries correctly followed the CCTs objectives. Nevertheless, the lack of consistency with other empirical results of the same case demand further accuracy with econometric methodologies. Additionally, the understanding of the different behaviours between formal and informal credit application are subject for further research, as this dissertation will show.

TABLE OF CONTENTS

Abstract

Acknowledgements.

Declaration.

Intellectual Property Statement.

Chapter 1: Introduction.

  • Conditional Cash Transfers, the unintended effects and its credit market implications.
  • The specific case of Nicaragua’s CCT and this dissertation research objective.
  • Dissertation Structure.

Chapter 2: An introduction to CCTs, RPS and its impact evaluations.

  • Conditional Cash Transfers, its impacts and progress.
  • Red de Protección Social and its impact.
  • The effect of CCTs on Credit Market Outcomes.
  • Chapter 3: A theoretical framework.
  • Conditional Cash Transfers as a means of growth.
  • Household decision to participate in the credit markets.
  • The credit supply sector.
  • The effect of transfers on the household decision to apply for a loan.
  • Research hypothesis.

Chapter 4: Data Analysis & Econometric Model.

  • Data collection.
  • Sample Characteristics.
  • Econometric Model.
  • Attrition.

Chapter 5: Estimation Results, Further Research & Limitations.

  • Probit Regression Results.
  • Remarks and discussion.
  • Further research opportunities.
  • Research limitations.
  • Policy recommendations.

Chapter 6: Conclusion.

Bibliography

Appendices

LIST OF TABLES

Table 1: Nicaragua RPS eligibility and benefits in Phase 1

Table 2: RPS average effect on per capita annual total household expenditures

Table 3: RPS average effect on enrollment, children age 7-13 in first to fourth grades

Table 4: Households participating in RPS surveys

Table 5: Sample characteristics on different variables (a) for the 3 surveys

Table 6: Number of households that applied at least once for credit (formal and/or informal), formal credit (regardless informal) and informal credit (regardless formal) in the 12 months previous to the survey

Table 7: Loan request determinants using a Probit model as shown in the equiation.

Table 8: Estimation of the probability of a household to request a loan by marginal effects of Probit Model 3

Table 9: Formal and informal loan request determinants using a Probit model as shown in the equiation

LIST OF FIGURES

Figure 1: CCTs in the world, 1997 and 2008

Figure 2: Household expected mg. Returns & costs of requesting a loan

Figure 3: Household expected mg. Returns & costs of requesting a loan after a transfer.

Figure 4: Transfer’s likely effect on the expected mg. returns curve

Figure 5: Transfer’s likely effect on the expected mg. costs curve

 LIST OF BOXES

Box 1: Nicaragua Credit Markets

Box 2: Migration and Remittances

 LIST OF APPENDICES

Appendix 1: Impact of CCTs on Poverty Measures, Various Years

Appendix 2: Impact of CCTs on School Enrollment and Attendance, Various Years, Latin America and Caribbean

Appendix 3: Statistical classification and Goodness-of-fit test for Probit Model 2

Appendix 4: F-statistic for the regression of observables on attrition

Chapter 1: Introduction.

The objective of this research is to understand what is the effect of Conditional Cash Transfers (CCTs) on credit market decisions in rural Nicaragua. An ambiguous theoretical framework will be used. It will theoretically explain the transfers’ effect on the household decision-making process to request a loan. An empirical Probit estimation will be done to quantitatively assess those effects. This study is aimed at bringing light to the matter fostering robustness to previous research.

Conditional Cash Transfers, the unintended effects and its credit market implications.

CCTs have experienced rapid growth in Latin America and the Caribbean since the 1990’s. They are considered poverty reduction initiatives when poverty is recognized as multidimensional and persistent. CCTs are programmes in which money to poor families is provided upon completion of certain conditions. Requirements are generally human capital investments such as school enrolment or regular health checks. This form of social protection is usually financed by taxes. Subsidizing consumption with cash provision reduces poverty in the short term, while facilitating and investing in human capital does it in the long run (Barrientos 2013a).

The promising evaluation results of the first generation programmes in Mexico, Brazil and Nicaragua have incentivise its popularity and rapid expansion among developing countries. These countries have applied experimental and quasi-experimental methodologies to show that the accumulation of human capital objectives has been successful. Empirical evidence showed an increase in school attendance, improvement in health care indicators and the raise of household consumption. Nevertheless, there are still many unanswered questions in terms of different countries’ implementation, sustainability and unintended effects (Rawlings & Rubio 2003). The latter is one of the most likely for further research because of its unknown nature and diversity in topics.

The impact evaluations have confirmed different desired and undesired effects from CCTs. Glewwe & Kassouf (2012) make an extensive literature review of studies that prove positive impacts in different cases. Their explanation is not subject to this dissertation, but benefits are seen in the cases of Nicaragua, Honduras, Colombia, Ecuador, Bangladesh and Cambodia. Studies have also focused in positive and negative undesired effects or impacts created by such policies. For example, in Argentina’s Asiganción Universal por Hijo[1] child labour was not considered in the policy objectives. However Salvia et al. (2015) found that the CCT reduced child labour by 2,6%. On a negative note, Garganta & Gasparini (2015) suggested the programme created a disincentive to labour formalization. In terms of unintended effects, probably the most studied cases are the ones of Mexico’s Progresa and Nicaragua’s Red de Protección Social.  This is because their randomized implementation reduced the difficulties in estimating programme effects greatly by allowing the use of difference-in-difference methodologies. Studies are so diverse that they even include the spill-over effects of those transfers on pregnancy and contraceptives (Darney et al. 2013).  Logically speaking, due to the nature of CCTs and its short-term safety net objective, one of the unintended effects could be manifested in the participation of household beneficiaries in credit markets. This topic has not been deeply studied yet.

Changes in credit markets due to the reception of a transfer (public or private) are ambiguous theoretically speaking. Amid a number of possibilities, incentives not to participate in credit markets may be created by CCTs for households. This is the case that because of the transfer, households do not need to pay in the future to have higher current liquidity. Meaning that getting a loan to have additional liquidity may have a higher relative cost compared to the transfer’s actual cost, which is the conditionality. In this case, the access to credit markets to fund other activities would be avoided by the household. On an opposite case, households that are beneficiaries of a CCT could have more probabilities to access the credit markets compared to those non-beneficiaries. Because the amount transferred is external and assured if the condition is fulfilled, the additional liquidity in this case is risk-free. Because of this, the household perceives an additional non-covariant funding source. This extra fund could be used for repayment reasons or shocks’ response. In this sense, the household creditworthiness perceived by lenders is improved by the change in the risk profile. Moreover, on the credit-demand side, the protection brought by the additional income leads to a change in the productive choices of the household. This change could lead to improved marginal returns from requesting a loan. In the end, the theory results ambiguous. In the latter case, the receipt of a CCT favours loan application for one reason, the other or the combination of both. While in the first case, being beneficiary of a CCT generates a disincentive to participate in credit markets (Morduch 1995; González Vega 2003; Hernandez et al. 2009).

The specific case of Nicaragua’s CCT and this dissertation research objective.

The Nicaraguan Red de Protección Social (Social Protection Network in English, from now on: RPS) was designed to influence the national poverty reduction strategy indicators. It was modelled after the pioneer case PROGRESA (now known as Oportunidades) in Mexico. RPS is a CCT that combines human capital investments with a social safety net access. It started in 2000 and with the support of the International Food Policy Research Institute (IFPRI) a three-year impact evaluation was conducted by Maluccio & Flores (2005). With a double difference methodology, the evaluation concluded positive and significant estimators in a wide range of outcomes associated to the objectives of the programme. Among many, the study found that enrolment was increased by 17.7%, attendance was raised by 11% and retention rate for children in grades from 1 to 4 was increased by 6.5% (Glewwe & Kassouf 2012). In terms of unintended effects, Maluccio (2007) assesses different productive investment behaviours. The study finds that there is a small increase in productive activities’ investment and a negative effect on beneficiaries’ labour supply. These findings echo the results of increased current consumption because of the specific CCT objectives in Maluccio & Flores (2005).

Hernandez et al. (2009) studied the effect RPS has had on the credit market outcomes. It was found that the credit request was not affected by RPS, while when controlling for potential endogeneity private remittances did increase it. The author argues that the reinforcement of long-term investments in human capital was well achieved, not incentivising the participation in credit markets. In spite of the well-founded calculations, the study holds one critique. Hernandez et al. (2009) use a linear probability model to assess a discrete variable. When conducting the research, the variable of interest responds to the question “Did you request for a loan in the last 12 months?” This question has a dichotomous variable as answer, meaning yes or no. Several econometric authors have agreed that when dealing with these kind of variables, more accurate results could be thrown by the Probit methodology instead of a linear probability model (Wooldridge 2016). Therefore, the effects of RPS on credit market outcomes could be subject for further research.

The present study will try to understand what is the probability of requesting a loan after receiving the CCT compared to non-beneficiaries. Following Maluccio & Flores (2005) a randomized community level evaluation will be used. Interviews to households were held before and after the intervention, both in treatment and control selected areas. The study will test the hypothesis that credit market participation decisions can be affected by the three reasons exposed in the contradictory theoretical model expressed above. Using a Probit methodology, the study will aim to bring more accuracy to the question raised by Hernandez et al. (2009). Because of the change in methodology new conclusions are arrived to, facilitating updated policy recommendations and further research. This study brings robustness to the field. Social sciences are generally criticized for the lack of robustness brought by repeating studies, compared to medical sciences. The above-mentioned reasons highlight the importance of conducting this dissertation.

Nevertheless, two research limitations must be pointed out. First, the results of this study at best will provide an insight into the direction of the relation between transfers and credit market participation, but it will not foster causality. Secondly, because of the nature of the information the model will address a probability, not being able to explain why that probability occurs. Understanding for what exact mix of the three theoretical reasons exposed above the probability is produced is out of scope of this study. Finally, further research could be conducted too address these limitations, external validity and formal and informal credit market outcomes.

Dissertation Structure.

This research focuses on the effects of a CCT on credit market outcomes. Overall there is evidence that RPS decreased the likelihood of requesting a loan in rural Nicaragua in a significant way. It is possible that getting a loan to have additional liquidity may have had a higher relative cost compared to the transfer’s conditionality. This may have occurred because its beneficiaries correctly followed RPS recommendations. The remaining sections provide firstly a brief literature review about CCTs in general and RPS specifically in Chapter 2. Secondly, the theoretical framework will be explained in Chapter 3. It will be done by understanding the household decision making process for loan requests, the credit-supply sector and how do transfers affect the household decision making process on the subject. Chapter 4 will focus on the explanation of the data available and the methodology used to understand if CTTs facilitate or hinder credit market participation in rural Nicaragua. Chapter 5 will conduct the empirical analysis highlighting the remarks and explaining the limitations. Lastly, the research will conclude summarizing the findings and providing policy recommendations and suggestions for further research.

Chapter 2: An introduction to CCTs, RPS and its impact evaluations.

The following sections provide a brief literature review about CCTs in general and RPS specifically. It is structured in three sub-subsections. Firstly, it reviews what are CCTs and what have been their impacts and progresses. Secondly, the specific case of RPS will be revised describing the program itself and the evaluations it has had. Lastly, the cases that have assessed the effects of CCTs on credit markets will be exposed.

 

Conditional Cash Transfers, its impacts and progress.

Conditional Cash Transfers are a form of social protection designed to reduce poverty across generations. Poverty is defined as the lack of certain goods & services or income level, and the difficulty of that poor person to revert it. This condition affects the last people of the well-being distribution (Barrientos 2013b; Hulme et al. 2012). CCTs aim to revert this unfortunate economic state of an individual or household.

Rawlings & Rubio (2005) define CCTs as a social contract. Money (cash) is provided to a poor family upon the completion of certain conditions. These are generally related to investment in human capital of children, like schooling and health care. In this way, CCTs facilitate and invest in human capital in the long term while enlarging consumption in the short term. Both objectives stimulate economic growth and social development. Income transfers with asset (human capital) accumulation are introduced with certain principles. According to Hulme et al. (2012), they are right based, non-contributory, long-term guaranteed and an important part of the population is covered. These social contracts generally form part of a larger development strategy.

CCTs have rapidly expanded among developing countries, strongly influencing the social assistance strategy (Handa & Davis 2006). Figure 1 describes the different CCTs found around the world across time. Starting in 1997, Progresa (now known as Oportunidades) was the pioneer case. Even though at the pilot state it had a small rural scope, since 2002 it started enlarging to urban communities. Today it is one of the largest programs in addition to Bolsa Familia in Brazil. The strategy has been dominating Latin American initiatives and expanded to diverse settings in developing countries in Africa and Asia (Fiszbein & Schady 2009; Glewwe & Kassouf 2012).  There are even few examples of shifting implementation to developed nations like the case of Opportunity NYC (Morais de Sá e Silva 2008).  Overall, these growing novel programmes are shifting social protection. In a constantly changing and informal economy, the strategy is slowly moving from contributory programmes (e.g. pensions) to social assistance (e.g. CCTs).

Figure 1: CCTs in the world, 1997 and 2008

Source: Fiszbein & Schady 2009

There has been accumulated evidence of the positive impacts and effects of CCTs. Vast studies and reviews have been done in combination with governments and international organizations’ efforts to produce quality data. Authors agree that the growth in CCTs is related to these successful evaluations. Its impacts make reference to the accomplished objectives of the programmes in poverty reduction and increased human capital. Appendix 1 and 2 provide a summary of the positive impacts in poverty reduction and school attendance in relevant cases. Moreover, additional research has been made around the unintended negative and positive effects that CCTs can have. Among many, child labour and teenage pregnancy reduction are found positive for example; disincentives on adult work in some cases create ambiguous conclusions. Nevertheless, positive impacts and effects seem to outbreak the negative ones (Fiszbein & Schady 2009).

In spite of the widely agreed positive outcomes of CCTs there are still further critiques to expose. To start with, the successful poverty reduction is only evaluated in the short term. The lack of several years’ evaluations hinders long-term outcomes analysis (Barrientos 2013b). Moreover, concerns have been raised about CCTs sustainability and cost-efficiency. Do specific CCTs pay for themselves after the years?[2]  Lastly, the relatively unique focus on building human capital leaves no room for broader development of poor settings (Handa & Davis 2006). The latter critique reinforces the importance of this dissertation.

Red de Protección Social and its impact.

Nicaragua is considered a developing country in Latin America with high poverty indexes. After a decade of political and economical instability, the 1990s started to be more prosperous. By following the Washington Consensus, the stop of hyperinflation and the transition to a market-led economy resulted in steady economic growth. In that decade and the early 2000s the national poverty headcount decreased from 51% to 46%. Nevertheless, rural poverty was still considered high. The 27% of rural population was extreme poor and 67% poor. These challenging figures demanded alternative solutions to the targets set in the Poverty Reduction Strategy Paper (PRSP) (Lacayo 2005).

RPS was a programme part of the Nicaraguan National Poverty Reduction Strategy (NNPRS). It was inspired by the Mexican case Progresa/Oportunidades and started a pilot phase in 2000. It was aimed at reducing present and future poverty by giving cash to targeted poor households in rural Nicaragua. The transfer was upon completion of certain conditions such as school attendance and health-care visits, guaranteeing human capital investment. RPS had several objectives, highlighting increased food expenditures, increased school enrolment and attendance among children aged 7-13, increased health care and nutrition of children aged under 9, and better pre and post-natal care for women. Benefits that fostered those objectives are described in Table 3 (Maluccio 2005; Maluccio 2007).

Table 1: Nicaragua RPS eligibility and benefits in Phase 1
Source: (Maluccio 2005)
  Programme Components
Food security, health, and nutrition Education
Eligibility: Geographic Targeting All households All households with children ages 7–13 who have not yet completed fourth grade of primary school
Demand-side benefits: Monetary transfers Bono alimentario (food security transfer) – C$2,880 per household per year (US$224) Bono escolar (school attendance transfer) – C$1,440 per household per year (US$112)
Mochila escolar (school supplies transfer) – C$275 per child beginning of school year (US$21)
Supply-side benefits: Provided and Monetary transfers *Health education workshops: Every 2 months
*Child growth and monitoring: Monthly: Newborn to 2-year-olds. Every 2 months: 2- to 5-year-olds
*Provision of antiparasite medicine, vitamins, and iron supplements
*Vaccinations (newborn to 5-year-olds) Bono
Bono a la oferta (teacher transfer) – C$80 per child per year given to teacher/school (US$6)

Three different categories of grants were provided by RPS as seen in Table 1. The benefits were on food security, education and health. All of them were distributed to the women head of household. While food security and education endowments were provided for a period maximum of 3 years, health ones were given for 5 years top (Maluccio 2005; Maluccio 2007).

RPS had two implementation phases. The first one was a pilot phase and lasted 3 years. The budget spent represented approximately 0,2% of GDP and was of $11 million. The Inter-American Development Bank (IADB) financed it under the condition of concluding an impact assessment. After the promising findings of the pilot stage an agreement between the Nicaraguan Government and IADB was held to prolong the programme. Phase II lasted four more years with a budget of $22 million. In this second phase the original beneficiaries were dismissed and new beneficiaries were included. The programme lasted 6 years in total (Lacayo 2005).

The pilot programme selected six municipalities as geographical coverage. These municipalities were chosen because of their poverty level and institutional capacity. All of them were located in the poor departments of Madriz and Matagalpa in the north of Nicaragua. A combination of geographical and household level targeting was followed for programme eligibility. Comarcas (localities) were grouped according to poverty level. The Comarcas with more than 55% of extreme poor households were geographically targeted as potential beneficiaries. In the rest less-poor Comarcas, proxy-mean test was used for household targeting. In the end 42 villages were eligible for the pilot phase. During the six years of implementation more than 30.000 households benefited from the programme (Maluccio & Flores 2005; Maluccio 2007).

In terms of impact[3] there is evidence that RPS was successful in achieving its core objectives. Rigorous empirical assessments were formulated thanks to the high-standards evaluation of Phase I conducted by Maluccio & Flores (2005). A randomized community-based intervention was used for the evaluation. It was measured before and after in treatment and control groups. The difference-in-difference methodology was implemented in the study. Overall, larger effects were seen for extreme poor households in the majority of indicators. Several indicators included in the NNPRS were ameliorated.

Studies confirmed improvements concerning food security, education and health. To start with, the increase in total annual per capita household expenditures was by 18% (Table 2). The majority of it was spent in food followed by education. Due to the economic crisis lived by Nicaraguans at that time, RPS worked as a traditional safety net according to the authors. Fiszbein & Schady (2009) confirm consumption impacts are generally found higher in cases where the transfer amount is greater, like the case of RPS. The 2002 national data confirms poverty fell by 5–9 points. Secondly, the programme exceptionally benefited school enrolment and attendance. The latter was increased by 20% while enrolment performed 13% better (Table 3). Furthermore, health improvements included: i) the rise of health-care programme participation of children under 3 years old by 16%; ii) an increase of 30% in vaccination rates; iii) and the decrease by 5.5% in the number of unconscious children. Moreover, promising undesired effects were found including the ones by the research of Maluccio (2003). The study suggested that child labour was reduced by 9% in children aged 10-13 after RPS implementation. The positive evidence in such indicators positively contributed to the decision to expand the programme in Phase II (Maluccio & Flores 2005).

Table 2: RPS average effect on per capita annual total household expenditures
Source: (Maluccio 2005)
Survey Round Intervention Control Difference
Follow-up 2002 4346 3378 968**
(385)
Follow-up 2001 4462 3194 1268***
(285)
Baseline 2000 4021 3738 283
(290)
Difference 2001-2000 441***
(146)
-544***
(172)
986***
(223)
Difference 2002-2000 325**
(153)
-360**
(168)
686***
(224)
** Significant at the 5% level.
*** Significant at the 1% level.
Standard error (in parentheses)
Table 3: RPS average effect on enrolment, children age 7-13 in first to fourth grades
Source: (Maluccio 2005)
Survey Round Intervention Control Difference
Follow-up 2002 92,7 79 13.5***
(2.6)
Follow-up 2001 93 74 19.2***
(3.1)
Baseline 2000 72 72 0.7
(4.1)
Difference 2001-2000 20.9***
(3.1)
2.4
(2.1)
18.5***
(3.7)
Difference 2002-2000 20.4***
(3.5)
7.6***
(2.6)
12.8***
(4.3)
** Significant at the 5% level.
*** Significant at the 1% level.
Standard error (in parentheses)

Furthermore, a recent study conducted by Barham et al. (2013) suggests schooling benefits go beyond the short-term positive effects seen in the above-explained evaluations. By focusing on a group of boys’ age 9-12 years, the assessment is done 10 years after the programme start. A sustained effect of half grade increase in schooling for boys is seen. Additionally, improved performances in maths and language achievement scores are statistically significant too. Hence, RPS is one of the few CCTs that has showed empirical evidence in long-term sustainability accomplishments.

On a supporting note to the findings expressed above, research on productive investments was carried out by Maluccio (2009). The results suggest not-promising conclusions. Even though expenditures were clearly increased by RPS, there is no evidence that investments in agricultural and non-agricultural activities were produced. This finding echoes the results on increased current consumption in food and education. In addition, the possible rising of long-term consumption due to increased investments is found narrow too.

Last but not least, other complementary analyses were done throwing ambiguous outcomes. Caldés et al. (2006) founds that the cost-benefit ratio of RPS was more expensive compared to other CCTs as Progresa. This raises questions about the efficiency of the programme, especially due to its scale. It is important to understand that the beneficial spill-over effects are generally contrasted by other investigations, in spite of the agreed positive impacts RPS has had.

The effect of CCTs on Credit Market Outcomes.

When understanding what has been the effect of CCTs on credit market outcomes, it is believed the number of studies have not been enough. The empirical literature addresses two specific cases. One occurs in Nicaragua and the other one in Mexico (Hernandez et al. 2009; Svarch 2011). Both of them arrive to different conclusions, increasing the importance of this specific research. Because of external validity or methodological accuracy, more cases or further research may be needed to bring robustness.

To start with, Hernandez et al. (2009) explore how public and private transfers may affect the decision of a household to request credit. In this study public transfers are in the form of RPS and private ones in the form of remittances. Using the household data that was utilized in previous impact evaluations, a linear probability regression with instrumented variables is conducted. They study concludes that the credit household decisions are not affected by CCTs on average. Nevertheless, remittances do increase the probability of requesting a loan when endogeneity is corrected.  The authors explain this may be the case because of the successful enforcement of RPS objectives. These findings are in line with lack of productive investments effects previously mentioned (Maluccio 2009).

In the case of Mexico, the Progresa/Oportunidades household data is empirically explored. A Probit regression and a linear regression with instrumented variables are run. In the end, both calculations suggest that households that received the CCT were more likely to request a loan than the ones that did not. In terms of the size of the loan demanded, a weak substitution effect is suggested since the effect on the size appears to be negative (Svarch 2011). These results are opposite to the ones found by Hernandez et al. (2009).

Differences in the cases are not subject to this dissertation, yet are important to highlight to understand the importance of this research. Firstly, the data used in the Nicaragua case was rural while the Mexican one was urban. Secondly, the scale was greatly different in both cases. Lastly, methodologies used in both cases differ. Hernandez et al. (2009) use a linear probability model to assess a discrete variable, while Svarch (2011) uses a Probit methodology. When conducting the RPS research, the variable of interest responds to the question “Did you request for a loan in the last 12 months?” This question has a dichotomous variable as answer, meaning yes or no. Several econometric authors have agreed that when dealing with these kind of variables, more accurate results could be thrown by the Probit methodology instead of a linear probability model (Wooldridge 2016). Therefore, the highlights exposed above conclude that the effects of RPS on credit market outcomes demand further research, subject to this dissertation.

Chapter 3: A theoretical framework.

This chapter will expose the theoretical framework used for this dissertation. Firstly, the general theory around the decision of a household to apply for a loan will be reviewed. The explanation of how the credit-supply sector works will continue. Lastly, what is the theoretical effect of transfers on the household decision to participate in the credit markets will be exposed. This section will conclude with the hypothesis to research in this dissertation. The theoretical framework explained below will be based on the work of: (Morduch 1995; González Vega 2003; Gonzalez-Vega 1982). All these authors are cited in the specific research papers of (Svarch 2011; Hernandez et al. 2009).

 

Conditional Cash Transfers as a means of growth.

In order to assure a permanent exit of poverty, sustained income growth is required. This is why Barrientos (2012) assessed the relation between social transfers and growth under a basic framework. The framework assumes that social transfers can lift credit constraints, provide greater security and/or enable household re-allocation of resources. All these consequences are thought to improve the productive capacity of a family, contributing to the poverty exit. With this framework an assessment of several cases was carried out. The study found that social transfers do have a positive effect on the productive capacity of a poor household. In the end, the chain of effects does stimulate growth at the micro-level. However, the framework seems simplistic for the complex analysis. The author concludes that more research is needed to offer a more comprehensive and systematic assessment of the growth effects of transfers (Barrientos 2012).

Chapter 2 showed clear evidence that CCTs in general and RPS specifically provide poor households a way to protect themselves from shocks. These risk-coping strategies enable long-term human development; for example allowing more schooling and lowering child labour.  The achievement of greater protection should also influence other household strategies. The case of requesting a loan could be one of those strategies (Hernandez et al. 2009).

This chapter aims to explore this advanced link by explaining a more complex theoretical framework than the exposed above. The complexity will consist in referring to the CCT’s effect on the household specific strategy to request a loan, as in comparison to a generalized means of growth like Barrientos (2012) does. The main interest to evaluate this specific effect grounds in two reasons. Firstly, money transfers are fungible. They may affect family choices about investment and consumption. These could have consequences in the credit market and indirectly in income growth in the long term. Secondly, transfers may affect the behaviour of other economic agents in the supply side like the case of lenders. Both reasons are in agreement with the basic transfers’ growth framework of Barrientos (2012) in a more specific way.

Household decision to participate in the credit markets.

The following paragraphs will describe the approach a credit-constraint household repeats every time they decide (or not) to apply for a loan. The explanation consists in the household maximization problem. The aim of it is to maximize the household expected profits by making use of its own resource endowments and a loan.

There is one economical condition required for the household to decide to participate in the credit markets. The theory claims a loan will be requested as long as the expected marginal returns from the use of the loan are greater than the marginal costs of borrowing. The latter vary depending the interest rates. Those rates are based on the lenders own maximization problem. Therefore, the household strategy will depend on two things. On one side it will depend on the household credit demand that reflects the expected marginal returns. On the supply side, the lenders profit maximization processes will determine it too. The credit supply factors emerge from the marginal costs of lending from alternative sources (González Vega 2003; Hernandez et al. 2009).

In the end, if the requested loan was given two possible outcomes could follow. One path could be that with the amount given the household demand could be 100% fulfilled.  A more common way could be that the amount needed by the household is actually bigger. This could lead to more credit applications that would turn the strategy into a credit-rationed one. The several borrowing options form the household loan portfolio (González Vega 2003; Hernandez et al. 2009). The latter option seems to represent more accurately the credit markets in rural Nicaragua, explained in Box 1.

Box 1: Nicaragua Credit Markets
The credit market in Nicaragua is composed by formal and informal sources. Generally diverse forms of credit reach the poorest households in rural Nicaragua. A study conducted in 1995 found that in the 90% of the credits provided to rural families, the time of re-payment was of less than 12 months. The report claims the extra cash was used mainly for consumption smooth and investment in agricultural inputs, such as fertilizers and seeds. For further references please see (Dauner 1998).

The credit supply sector.

Formal and informal loan products integrate the supply side. These are composed by the loan itself and the corresponding additional interest rate. The latter will depend on certain characteristics of the demanding household perceived by the lender. The loan given must be equal or less than the amount demanded by the borrower. The aim of the lender is to maximize the expected profits of giving a loan.

Two factors add the total expected cost of lending: the expected loss in case of default in addition to the administrative screening and contract costs. In this way, the costs increase in function to the increase of the loan amount and the borrower’s risk level. The level of risk is the result of the lender’s evaluation given a set of characteristics of the household that demands the credit. One example of characteristic could be the availability of collateral or not (Gonzalez-Vega 1982; Hernandez et al. 2009).

The riskier the perception is, the higher the interest rate will be. In a monopolistic credit-supply market the charges for interest will be higher than the marginal costs. However, the situation shifts when the borrowers are credit-constraint and the lenders are not. In this case the interest rate will fully represent the lenders marginal costs of lending to a demander with certain characteristics.  Therefore, the marginal costs function will change given these characteristics. The riskier the borrower is perceived, the higher the marginal costs will be, thus the higher interest rates that will result. In the end, if the household demanding extra cash is credit constraint the lender will equalize the interest rate to the expected marginal costs (Gonzalez-Vega 1982; Hernandez et al. 2009).

The effect of transfers on the household decision to apply for a loan.

The household marginal returns of using their own endowments will depend on three factors. Firstly, dependence will be on the production technology used. Secondly, the market opportunities available will affect them too. Nevertheless, the process of generating income is unpredictable due to the uncertainty of nature. The existence of shocks, like floods or droughts, could drastically impact the final outcome of the income-generating process in positive or negative ways (Morduch 1995). The ability of the household to cope with the uncertainty affects their marginal returns‘ capacity. The model assumes the riskier the technology is, the higher the potential marginal returns. Therefore, the household may choose between two production technologies with different safety degrees.  This choice will be depending on their risk-behaviour. A household without enough equipment to cope with risk may choose a safer mix. This will result in a lower productivity (Morduch 1995; González Vega 2003; Hernandez et al. 2009).

Therefore, the decision process that a household goes through when choosing to request credit is simple. The household does a comparison between the expected marginal returns from the use of their resources (from now on referred as: EMg(R)) and the expected marginal costs of requesting credit (from now on referred as: EMg(C)). The loan will be requested if the EMg(R) are higher than the EMg(C).

The process will be described by the following sequence. Once all the household resources have been allocated, the schedule of the EMg(R) of a given product mix will generate a demand curve for credit. This curve is downward sloping. It is assumed then that each point along this schedule will be associated with a probability of potential outcomes. In addition, the household may consider the option of default and its associated penalties (that is the probability that the returns may be below the level required for repayment). Overall, the size of the loan to be demanded at each level of EMg(C) of borrowing is determined by the maximum probability of default the household is willing to accept (Svarch 2011; Hernandez et al. 2009).

In Figure 2 R0 represents the initial household resource endowment. Because of the assumption of decreasing marginal returns, the curve EMg(R) is downward sloping. The area under EMg(R) represents the total expected gross returns; and the position of EMg(R) shows the household level of profitability. As explained earlier, this position will depend on the household product-mix choice, the market opportunities and the household risk toleration. In summary, the more productive mix, the better market opportunities and the higher risk toleration, the more above the curve will be located (compared to other households’ combinations). Following the model, EMg(C) will depend on two elements. Firstly, it will depend on the marginal costs perceived by the lenders participating in the market, meaning the interest. Secondly, the transaction costs must be considered. This is why EMg(C) is a curve sloped upwards and the area below it represents the cost of requesting a loan (L). A loan size B will be requested when the EMg(R) are equal to the EMg(C), marked as point B in the graph. Point B represents the complement between the expected gross returns of the use of their own endowments (which is the area below the curve EMg(R) to the left of R0 in the graph) and the maximum expected net returns from the use of the credit (which is the ABC triangle in the graph). The household can request several loans available until it reaches the point B (Svarch 2011; Hernandez et al. 2009).

Figure 2: Expected mg. returns & costs of requesting a loan

Source: Hernandez et al. 2009

GRAPH 1

Now, the model assumes the household endowments are increased because the household receives a transfer in the form of a CCT. How will the EMg(R) of the use of these resources vary?  Figure 3 explains it graphically at different levels of endowments. The model now assumes the receipt of the transfer is the only change in the situation. Now the initial household resource endowment will be represented by R0+T.  In this case, the chance of having increasing gross expected marginal returns reduces due to the decreasing marginal returns of the curve EMg(R). Even if the credit is big enough this chance will vanish. Graphically the model permits noticing the ABC area is smaller when the credit is available as endowment (Svarch 2011; Hernandez et al. 2009).

Figure 3: Household expected mg. Returns & costs of requesting a loan after a transfer.

Source: Hernandez et al. 2009

GRAPH 2

Nevertheless the transfer may produce other changes. If the CCT is correlated positively with the nature shocks, then the beneficiary will reduce income volatility.  This decrease can increase the household risk-toleration to a more productive-mix choice. Additionally it can make the household less reluctant to the default possibility. Thus, it can reduce the risk adjustments made to the EMg(R). Then the EMg(R) would shift upwards, creating a bigger ABC area as show in Figure 4. Overall, in this case requesting a loan is more attractive (Svarch 2011; Hernandez et al. 2009).

Figure 4: Transfer’s likely effect on the expected mg. returns curve

Source: Hernandez et al. 2009

GRAPH 3

In addition, the creditworthiness of the household beneficiary of the CCT could be improved. The improvement in the lender’s perception would shift the curve EMg(C) downwards. This case is represented by Figure 5, which further increases the ABC area. It is important to notice that this may or may not strictly occur given the asymmetries in information accessibility in the credit markets (Svarch 2011; Hernandez et al. 2009).

Figure 5: Transfer’s likely effect on the expected mg. costs curve

Source: Hernandez et al. 2009

GRAPH 4

In conclusion, the theoretical framework throws three different possibilities that may occur to the decision to request a loan after receiving a transfer like a CCT. Firstly, the decreasing marginal returns characteristic reduce the chance of the loan to add to the net expected marginal returns of the use of resources. If the transfer is big enough this chance may even disappear. Secondly, if the transfer acts as risk-protection from shocks for example, the expected marginal returns may become higher. This could increase the chance of the loan to become profitable. Finally yet importantly, the creditworthiness because of the transfer’s extra cash available could be possibly improved. This could result in more loan supply available for the borrower, increasing the chance to request a loan too. The decision will depend on which effect is the most dominating given the circumstances (Morduch 1995; González Vega 2003; Gonzalez-Vega 1982; Svarch 2011; Hernandez et al. 2009).

Research hypothesis.

The above-exposed effects are contradictory, making the theoretical framework ambiguous. Even though CCTs have been proved successful as risk-protection, the theoretical framework demands an empirical test that can throw more light. In addition there is a need to use a more accurate methodology to solve this complex model empirically. This dissertation will test the hypothesis that RPS has a negative effect on the credit market participation. The study will hypothesize the positive effects are offset by the disincentive the perception of the loan’s decreasing marginal effects brings to the household. Nevertheless, some research limitations must be pointed out. First, the results of this study at best will provide an insight into the direction of the relation between transfers and credit market participation, but it will not foster causality. Secondly, because of the nature of the information the model will address a probability, not being able to explain why that probability occurs. Understanding for what mix of the three theoretical reasons exposed above the probability is produced is out of scope of this study.

Chapter 4: Data Analysis & Econometric Model.

The following chapter will firstly focus on the explanation of the data collection and its descriptive statistics. It will later describe the econometric model to follow and it theories. Lastly it will address the issue of attrition.

Data collection.

The data used in this empirical study is collected for the RPS impact evaluation done by The International Food Policy Research Institute (IFPRI). The survey measured before and after the programme in control and treatment areas. As stated in the second chapter, 42 comarcas were eligible for the programme in a random way. For the purpose of the evaluation, a counterfactual measure was needed. Out of a pool of equally eligible candidates, 21 comarcas were selected as treatment group and 21 as control one. The control stated what would have happened in the case of programme inexistence. Since it was unknown if there were enough resources to implement the intervention in all households at once, random selection was considered fair. Thus ethical issues were not a concern in the programme evaluation.

The data was collected from a household panel data survey. The questionnaire was based on the Nicaraguan Living Standards Measurement Survey (NLSMS) from 1998. For evaluation purposed the questions were expanded in certain sections such as health and education to have better information of programme indicators. Consequently, the survey was contracted in other matters to lessen the answering burden and assure high quality in the collection. The survey was done in treatment and control areas before programme implementation in 2000 and after in 2001 and 2002 (Maluccio & Flores 2005). Due to the quality of the survey itself and the collection process this dataset has been used for many different studies and publications. Many of them are mentioned in this dissertation. This study has chosen this data because of its reliability, high standards and simplicity.

The household survey was asked in a stratified random sample of the 42 comarcas[4]. The initial target sample was of 1764 households. In each of the 42 comarcas, 42 households were selected. 10,4% households did not answer the base line survey, resulting in a 1581 household sample. The levels of attrition for treatment and control group for the follow-up surveys are displayed in Table 4. Nevertheless, randomization of the panel is considered accurate as long as attrition itself is random. This issue will be explained in further paragraphs (Maluccio & Flores 2005).

Table 4: Households participating in RPS surveys
Source: Authors based on Nicaraguan RPS evaluation data, StataCorp 2011
  Baseline 2000 Follow-up 2001 Follow-up 2002
Completed Interview 1581
(10.4)
1453
(8.1)
1397
(11.6)
Treatment 810
(8.2)
766
(5.4)
722
(10.9)
Control 771
(12.6)
687
(10.9)
675
(12.4)
Completed interviews in the three rounds 1359
(23.0)
1359
(14.0)
1359
(14.0)
Treatment 706
(20.0)
706
(12.8)
706
(12.8)
Control 653
(26.0)
653
(10.5)
653
(10.5)
The existence of households that participated in 2002 but not in 2001 and vice-versa creates a lean panel without attritors of 1359 households (and not 1397)
Percentage of attrition (in parentheses)

Sample Characteristics.

For the purpose of enlarging knowledge, the descriptive statistics will be associated to the three waves of surveys. Descriptions are showed in Table 5. However, for the econometric estimations only the second and third wave of surveys will be taken into account. This is because intervention implementation is needed to carry out the specific analysis. This descriptive analysis provides an initial explanation of the household behaviour during the pilot phase.

Table 5: Sample characteristics on different variables (a) for the 3 surveys
Source: Authors based on Nicaraguan RPS evaluation data, StataCorp 2011
(a) Description of variables can be found in the “Econometric Model” sub-section of this chapter
  2000   2000   2000
  All sample   Treatment   Control
Variable Obs Mean Std.   Obs Mean Std.   Obs Mean Std.
grupo 1359 0.519 0.500 706 1.000 0.000 653 0.000 0.000
byear 1359 2000 0 706 2000 0 653 2000 0
no_childr 1359 1.146 1.072 706 1.099 1.099 653 1.196 1.041
no_women 1359 2.377 1.468 706 2.326 1.452 653 2.432 1.484
no_men 1359 2.472 1.559 706 2.476 1.575 653 2.467 1.544
av_hhage 1359 23.035 11.515 706 23.808 12.650 653 22.199 10.091
credit 1359 0.179 0.383 706 0.160 0.367 653 0.199 0.400
credform 1359 0.082 0.275 706 0.076 0.266 653 0.089 0.285
credinform 1359 0.113 0.317 706 0.105 0.307 653 0.123 0.328
property 1359 0.814 0.389 706 0.816 0.388 653 0.812 0.391
social 1359 0.428 0.495 706 0.442 0.497 653 0.413 0.493
otherprog 1359 0.492 0.500 706 0.469 0.499 653 0.518 0.500
remitt 1359 0.038 0.192 706 0.031 0.174 653 0.046 0.210
shock 0 0 0
prev_remitt 0 0 0
assetind~100 1359 6.779 5.377 706 6.765 4.952 653 6.793 5.806
  2001   2001   2001
  All sample   Treatment   Control
Variable Obs Mean Std.   Obs Mean Std.   Obs Mean Std.
grupo 1359 0.519 0.500 706 1.000 0.000 653 0.000 0.000
byear 1359 2001 0 706 2001 0 653 2001 0
no_childr 1359 1.129 1.071 706 1.067 1.099 653 1.196 1.035
no_women 1359 2.577 1.564 706 2.514 1.563 653 2.645 1.564
no_men 1359 2.645 1.623 706 2.636 1.636 653 2.654 1.610
av_hhage 1359 23.265 11.222 706 24.104 12.439 653 22.357 9.666
credit 1359 0.140 0.347 706 0.129 0.335 653 0.152 0.359
credform 1359 0.063 0.244 706 0.042 0.202 653 0.086 0.280
credinform 1359 0.082 0.275 706 0.091 0.287 653 0.074 0.261
property 1359 0.834 0.372 706 0.834 0.372 653 0.833 0.373
social 1359 0.284 0.451 706 0.312 0.463 653 0.254 0.436
otherprog 1359 0.781 0.413 706 0.959 0.199 653 0.590 0.492
remitt 1359 0.046 0.210 706 0.033 0.178 653 0.061 0.240
shock 1359 0.869 0.338 706 0.853 0.355 653 0.887 0.317
prev_remitt 1359 0.038 0.192 706 0.031 0.174 653 0.046 0.210
assetind~100 1359 6.779 5.377 706 6.765 4.952 653 6.793 5.806
  2002   2002   2002
  All sample   Treatment   Control
Variable Obs Mean Std.   Obs Mean Std.   Obs Mean Std.
grupo 1359 0.519 0.500 706 1.000 0.000 653 0.000 0.000
byear 1359 2002 0 706 2002 0 653 2002 0
no_childr 1359 1.095 1.066 706 1.020 1.076 653 1.176 1.050
no_women 1359 2.735 1.622 706 2.646 1.620 653 2.832 1.620
no_men 1359 2.780 1.691 706 2.769 1.709 653 2.792 1.672
av_hhage 1359 23.759 11.182 706 24.632 12.425 653 22.814 9.579
credit 1359 0.110 0.313 706 0.105 0.307 653 0.116 0.321
credform 1359 0.037 0.188 706 0.028 0.166 653 0.046 0.210
credinform 1359 0.077 0.266 706 0.079 0.270 653 0.074 0.261
property 1359 0.853 0.354 706 0.867 0.340 653 0.838 0.369
social 1359 0.288 0.453 706 0.312 0.463 653 0.263 0.441
otherprog 1359 0.743 0.437 706 0.826 0.380 653 0.654 0.476
remitt 1359 0.065 0.246 706 0.069 0.254 653 0.060 0.237
shock 1359 0.834 0.372 706 0.847 0.360 653 0.819 0.385
prev_remitt 1359 0.046 0.210 706 0.033 0.178 653 0.061 0.240
assetind~100 1359 6.779 5.377 706 6.765 4.952 653 6.793 5.806

Although households in the control group seem to be slightly younger the whole sample tends to be young with an average age of 23 years old. On average each household has two women, two men and one child (under 6 years old) in the baseline. By 2002 both females and males are three per household, though children remain stable at one on average.

These families have an infrequent loan request. The 17% of the households asked for credit during the year before the baseline survey. Even though credit has different varieties as explained in Chapter 3, formal ways of requesting are lower than the informal ones. While 8% of the sample in 2000 asked for credit from well-known institutions, 11% did asking from a friend, relative, neighbour or informal organization. Both analyses are not exclusive, since a family could have requested credit from both types in the same period. When analysing the credit evolution it seems to have decreased over the year, reaching an 11% of credit request in the last 12 months in the 2002 wave. The trend applies for all types of credit. When focusing in the different groups, it seems the beneficiaries of RPS followed that downwards direction. Nevertheless, the non-beneficiaries kept the informal credit application steady in the treatment years (2002-2002) with around 7,4%. More specific details among beneficiaries and non-beneficiaries can be seen in Table 6. It is interesting to point out that formal credit request of the RPS households was reduced by more than half after 2 years of intervention.

Table 6: Number of households that applied at least once for credit (formal and/or informal), formal credit (regardless informal) and informal credit (regardless formal) in the 12 months previous to the survey
Source: Authors based on Nicaraguan RPS evaluation data, StataCorp 2011
  Credit Credit Formal Credit Informal
  2000 (lean panel: 1359 HH)
Sample 243.00 112.00 154.00
Treatment (RPS) 113.00 54.00 74.00
Control 130.00 58.00 80.00
  2001 (lean panel: 1359 HH)
Sample 190.00 86.00 112.00
Treatment (RPS) 91.00 30.00 64.00
Control 99.00 56.00 48.00
  2002 (lean panel: 1359 HH)
Sample 150.00 50.00 104.00
Treatment (RPS) 74.00 20.00 56.00
Control 76.00 30.00 48.00

In 2000 almost 4% of the sample had the receipt of a remittance. The definition of remittance from the survey is those household that received some sort of help (either in cash or in kind) from friends or family abroad or emigrated within the country. This concept differs to the one of access to remittances, which stands for having access to a migrant that could send the aid. The latter was around 7% (not shown) in the household’s beneficiaries of RPS in 2000. The receipt of remittances kept increasing over the years. Hernandez et al. (2009) explains this could be the case when families were mistaken in believing that the CCT would be taken away if they confirmed receiving extra help from a migrant. As years passed by, RPS beneficiaries might have realized it was not the case and started self-reporting more the reception of help.

In terms of other characteristics, more than 80% of the households report owning the property their live in (with or without an official title). The rest is assumed to be renting or others. The self-reported house ownership increases over the years. On another note, the average level of assets before the programme started is low. It is calculated in a simple index that shows how many assets a household reports to have out of the 25 items listed in the survey. This list of goods of the survey is composed of furniture, electronics, tools, sewing machines and vehicles such as bicycles, cars and motorbikes. The index is then is re-normalized to 100 (Moser & Felton 2009). Overall the index averages in 6,8 points out of 100, does not differ between control and treatment groups and has a maximum of around 50 points. The level of wealth seems to be homogeneous across households but low.

Finally, Hernandez et al. (2009) finds more interesting attributes to understand the sample. Families that participate in RPS predominantly work in agricultural schemes. Nevertheless around 12% of them are involved in other entrepreneurial activities.  The annual per capita consumption of a household is on average 3,900 Córdobas[5]. This is then considered in the less than a dollar per day category. Consumption of basic needs like food, housing, transport and education are captured in this figure.

To summarize, households of the sample have low access to remittances and do not request much credit. The average age is low, concluding families are composed of young members. In addition, they are categorized as very poor considering international standards and seem to have a low level of wealth. Overall these characteristics match with the livelihood in the poor rural setting around the world (Hulme & Lawson 2010).

Econometric Model.

This dissertation applies econometric exercises to understand the effects RPS might have on credit market outcomes in rural Nicaragua. Using a specific econometric model allows to identify household attributes that can influence credit market outcomes while controlling for all other effects. Moreover, this technique allows knowing specifically if being a beneficiary of RPS is a significant characteristic that can affect the likelihood of requesting a loan (positively or negatively) or not.

A Probit model will be used at household level to generate the empirical explanation. It will be initially presented in this chapter and then evolved along the dissertation due to different challenges. The Probit methodology is chosen because it is best fit when the dependant variable is a dichotomous one. In this case the dependent variable equals to 1 if the household requested a loan in the last 12 months and 0 otherwise.  On an opposite note an OLS model could be used to estimate coefficients when the variable of interest is continuous, which is not the case. To allow for a conclusion, probabilities will be interpreted from the marginal effects generated by the Probit model (Wooldridge 2016).

The econometric equation goes as follows:

Υ̂it=Pr⁡Υit=1=β0+β1 RPSi+β2t1+β3t2+ β4Cit+εit

Υ̂it

represents the expected probability that a household i in time t requested at least one loan in the 12 months prior to the interview.

β1 RPS

has a value of 1 if the household was assigned to the treatment group of RPS and zero otherwise.

β2t1+β3t2

represent year dummies 2001 & 2002 to capture conceptual variables common to all households.

εit

represents the random error of the estimated equation.

β4Cit

represents different controls to correct for other influences that may affect the decision to request a loan. These were selected after vast literature review and the contribution of previous studies such as the one of (Hernandez et al. 2012). The controls involve:

  • Labour endowment: represented by the variables of number of males (No_men), females (No_women) and children under 6 years old (No_childr) in the household. It is suggested by the literature that a greater labour endowment allows for strategy diversification for the household production capacity. This may influence the decision to apply for credit. Differentiation is done to consider different role assignments given gender and age. In addition, adding more control variables balances the fact that unobserved households characteristics cannot be corrected by Fixed Effects in a Probit model.
  • Experience & knowledge: represented by the variable of average age in the household (Av_hhage). The literature suggests it may determine the participation in the credit market by favouring the household’s ability to request a loan.
  • Wealth: represented by a dummy variable stating if the home plot is owned or not (Property) and value of durable goods resumed in an asset index (Assetindex100). The property ownership is a self-reported proxy that could give benefits to loan requests by acting as collateral or reduce administrative costs of credit. In the model it is treated as an exogenous variable considering the thin land markets rural Nicaragua has (Deininger et al. 2003). The asset index is a widely used as a measure of wealth in developing countries. The simple index calculation was explained in the Sample Characteristics sub-section. The asset index is considered only for those goods that were owned before RPS took place to avoid endogeneity problems. In this way, only goods that were acquired before the observed decision to request for credit are being taken into account. Both variables are self-reported.
  • Access to social networks: represented by a dummy variable that informs if any member of the household is a participant of a community organization related to sports, religion, cooperative, social development and others (Social). A greater access to social connections is believed to form more positive images to lenders and enable greater access to credit (Karlan 2007).
  • Access to other programs other than RPS: represented by the dummy variable stating if any member of the household has been benefited by other government socio-economic programmes or not (Otherprog). These may facilitate more transfers that could enable better access to credit.
  • Access to remittances: represented by a dummy variable stating if the household received cash or in-kind transfers from migrant friends or family in the 12 month previous to the interview (Remitt). It is believed that having extra transfers could affect the decision to apply for a loan. Nevertheless as Box 2 describes, remittances are the manifestation of a migration strategy to overcome income risks. This could result in an endogeneity problem since both variables could be influenced by each other. Additionally both are affected by the same explanatory variables[6]. In order to correct for this issue, a lagged value from the previous year will be used in the model (Prev_remitt).
Box 2: Migration and Remittances
Migration is an inter-temporal household strategy to reduce income risks. Migration involves remittances that are transfers sent in critical times. These are negatively correlated to household income and are seen as an informal method. Migration diversifies the household income-generating approaches spatially. It is chosen in markets where credit and insurance are missing or restricted. In addition, it is commonly used in rural settings. The New Economics of Labour Migration (NELM) claim this strategy can develop changes in investment and consumption choices compared to the decisions taken without the migration factor having occurred.

Vast literature review has empirically agreed remittances do increase consumption and investments. Nevertheless, the trade-off will always be associated to less availability of human resources. The self-selection characteristic in the process of migration is key to classify access to remittances as an endogenous variable when choosing to demand for credit (Hernandez et al. 2009; Stark 1991).

 

  • Suffering a shock as a motivation to insure: represented in a dummy variable stating the suffering of a self-reported shock or not (Shock). The survey asks if the household has suffered any of the following in the last 12 months: theft, lack of work, drought, flood or unprofitable coffee prices. Credit has been considered as a developing country solution to cope with these idiosyncratic shocks.

The model is assumed to be correctly specified. In order to arrive to this conclusion two different tests are run. Firstly, the classification statistics and table report is executed after the Probit estimations of Model 2 in Appendix 3. Although the Classification Table shows an 87% as correctly classified model, the model was only able to correctly classify the true negatives. This may be because 0.5 predicted probability of credit might not be the optimal cut-off. For further understanding and robustness, the goodness-of-fit test is run. The latter throws a non-significant chi-square, suggesting the model fit is rich (StataCorp 2011). The correct specification allows continuing to assess other challenges as the one of attrition. The objective is to arrive to the best estimation possible.

Attrition.

Attrition in panel data is referred to the non-answer of the following waves of the survey after the baseline. This problem has two negative consequences. Firstly, by having less number of observations the quality of the different estimators is diminished. Secondly, the characteristics of the attritors (households that are not again surveyed) can systematically differ to the ones of the non-attritors. In this case attrition is not random causing a bias. Because of the latter, attrition demands an analysis (Bendezu et al. 2007).

In this study, attrition is defined as a non-response in at least one of the two subsequent waves to the baseline. The characteristics of the overall initial sample (1581 households) will be compared to the characteristics (of the same 2000 survey) of the households that later became attritors (Bendezu et al. 2007). The aim is to determine if attrition was random or not.  A linear probability model is used with a dichotomous dependent variable. The latter is equal to zero if the household is an attritor and one otherwise. The independent variables are key observables available in the data that can vary between the households that failed to answer the follow-up surveys and the ones that did not. The null hypothesis tested in an F-statistic is that all parameter estimates are jointly zero. Attrition is random if we can fail to reject the null.

The F-statistics showed in Appendix 4 suggest that attrition was likely to be non-random. Households that had the support of other social programs than RPS were more likely to answer the follow-up surveys. Indeed, families that were assumed to have higher levels of wealth because they owned a property seem to have had less difficulty in completing the waves. On the other hand, families without the extra-support or advantages appear to have skipped one or more of the surveys due to their vulnerability. These findings relate to the ones of Hernandez et al. (2009). When comparing the control and treatment group of the lean panel without attritors, the authors find small but significant differences. The control group was formed of households with younger average age that had less per-capita consumption and had more credit market participation. Hernandez et al. (2009) suggest that the less advantaged again had less possibilities to overcome the costs demanded by the conditions of a cash transfer.

Because attrition has been tested not to be random it must be corrected. The methodology used for this procedure is the “Inverse Probability Weights”. To do this, the predicted probabilities from the unrestricted attrition Probit in Appendix 4 are calculated. The variables correlated with attrition (Property and Otherprog in this case) are excluded for the re-estimation. After calculating the predicted probabilities from the restricted attrition Probit, the inverse probability weights are calculated by taking the ratio of the restricted to unrestricted probabilities (Baulch & Quisumbin 2011). This process allows the model to use the weighted sample excluding the attritors, represented by Model 3 in Table 7.

Chapter 5: Estimation Results, Further Research & Limitations.

The following chapter will focus on the Probit regression results with its main remarks. In addition it will expose the topics that require further research. Lastly several research limitations will be shown. Limitations that are general, to the model and to the variables will be addressed.

Probit Regression Results.

Table 7 presents the different Probit models run in order to understand the effects of CCTs on credit market decisions in rural Nicaragua. At first sight it seems CCTs have a negative effect on credit applications. The coefficients intend to show the direction of the relation by its sign and its significance. Table 11 does not intend to present specific probabilities. Once the model is analysed, probabilities will be presented in further sub-sections by the marginal effects’ coefficients.

Model 1 presents the results taking into consideration the lean panel (without attritors of any wave). As it was previously mentioned remittances present an issue. Because of the nature of the information, it is not known if some households may participate in the credit markets because they have availability of remittances or vice versa. This endogeneity problem may possibly bias the model. Model 2 intends to correct such endogeneity using a lagged variable of the receipt of remittances. The one period lagged independent variable acts as a corrector for the potential reverse causality the model could be suffering (Wooldridge, 2016). This adjustment is the introduction of a variable of remittances that has the certainty of having been received in the previous year of the reception of the CCT. When looking at the results it is observed that in Model 2 the receipt of remittances looses its significance while our coefficient of interest becomes more significant. The correction of the simultaneity problem brings further accuracy to the model.  Last but not least, Model 3 evolves de analysis by using a weighted panel. As mentioned in Chapter 4 the intention is to correct for possible biases the non-random attrition could be causing. Model 3 will be the main analysis in this dissertation.

Table 7: Loan request determinants using a Probit model as shown in the equation.

Model 1 (Credit) Model 2 (Credit) Model 3 (Credit)
RPS -0.151*

(0.068)

-0.159**

0.068

-0.156**

(0.068)

No_childr -0.049

(0.036)

-0.052

(0.035)

-0.051

(0.035)

No_women -0.045*

(0.023)

-0.042*

(0.022)

-0.043*

(0.022)

No_men -0.004

(0.020)

-0.005

(0.020)

-0.005

(0.020)

Av_hhage -0.007*

(0.004)

-0.006

(0.004)

-0.006

(0.004)

Property 0.153*

(0.093)

0.144

(0.093)

0.150

(0.093)

Assetindex100 0.020***

(0.005)

0.020***

(0.005)

0.020***

(0.005)

Social 0.289***

(0.067)

0.287***

(0.067)

0.286***

(0.067)

Otherprog 0.227**

(0.087)

0.232***

(0.087)

0.231***

(0.087)

Remitt 0.379***

(0.125)

Prev_remitt -0.014

(0.158)

-0.024

(0.158)

Shock 0.109

(0.095)

0.106

(0.096)

0.102

(0.096)

T2 0.142**

(0.062)

0.134**

(0.062)

0.134**

(0.062)

_cons -1.459***

(0.185)

-1.446***

(0.184)

-1.144***

(0.184)

*** Significant at the 1% level, ** Significant at the 5% level, *Significant at 10% level.

Standard errors: (in parenthesis)

Number of observations: 2718 (1359 households)

Source: Authors based on Nicaraguan RPS evaluation data, StataCorp 2011

Remarks and discussion.

Our main model (3) shows that CCTs may hinder credit market participation in rural Nicaragua. The case seems to show that CCTs as subsidies of health and education diminish the likelihood of requesting a loan with its interests. Overall it can be understood that beneficiaries correctly followed the RPS investment recommendations. This may have led to increased current consumption in food and education and not in agricultural and non-agricultural activities. It is probable that the amounts of the RPS were correctly put to pursue its objectives. This was supported by a supply side intervention on education and health increasing the chances of consumption on these fields. The case shows that the transfer may not possibly be enough to give higher expectations of increased marginal returns of a loan to a household. Overall it means that the household perceives that the additional liquidity a loan could bring has a higher relative cost compared to the transfer’s cost. It is assumed that the CCT cost are the obligations due to its conditionality.

On a supporting note, the results could mean that in spite of having a transfer there was a lack of change in the production mix of the household. This may have not incentive investment in productive activities, which may have not incentive credit market participation. This could be expected especially in the more poor households, which were targeted for treatment and control areas. The poorer the household, the more expected current consumption. Moreover, it could be that before the CCTs investments in human capital were supported by credits. After the introduction of RPS they were not necessary.

These findings are in accordance with the different studies mentioned in Chapter 2. Firstly the achievement of improved education, nutrition and health (Maluccio & Flores 2005) shows policy recommendations were followed by beneficiaries. In addition, current consumption was found greater too, specially in education and food. Again, this could be related to the conditionality itself but also to the amount given to beneficiaries. The latter was considered high compared to other Latin-American CCTs. A greater fund tends to enable higher consumption impacts (Fiszbein & Schady 2009). Lastly, the lack of change in the production mix of the households supported by the fact that RPS did not foster expenditures in agricultural and non-agricultural activities (Maluccio 2009).

Further analysis about other household characteristics can be observed in Model 3. Firstly the number of women in the household seems to have a significant negative effect on the decision of requesting a loan. The negative parameter suggests household expenses administration and decision-making by female gender seems common in rural settings. Women are the household members that receive the cash and decide where to spend it. This is confirmed by studies that prove women have better transfer administration according to its objectives (Barrientos 2013a). Therefore it is not surprising that having a greater amount of women in the household may influence the decision to prioritize current consumption over credit application for investment in productive activities. It is possible that women have less risk toleration, thus decreasing the chances of credit application. The average household age as a household knowledge variable, does not show a significant influence. This may be the case because households are found to be on average young. Firstly the lack of older members does not add more knowledge that may influence positively the request of a loan. Secondly the lack of notable differences among households’ average age does not allow a significant effect analysis.

In terms of wealth and better opportunities indicators, the signs are as expected. Characteristics as owning a property (self-reported) and having experienced a shock are not significant but positive as the literature suggests. Higher wealth characteristics as the value of assets, access to social networks and having access to other programs other than RPS are found positively significant when influencing the decision to request for a loan. The latter may be a response of the government to unexpected shocks. This may be then a temporary effect as in comparison to the other two. It is possible that RPS has not been considered as a risk-protection from shocks since other programmes had that role. By not being considered as a shock protection, the expected marginal returns of getting a loan remained low. This has made the chances of applying to credit actually decrease as the econometric analysis shows. The negative effects have offset the positive.

Lastly, when analysing the influence of time effects it is positive and significant. It seems that the probability of requesting a loan in the second period after the introduction of RPS was higher than in the first period. However, the results show the probability of this effect in the case of the treatment group may have been lower than in the control group. Again this allows us to understand that the effect of RPS in increasing current expenditures may be greater than the time effect.

To summarize, Table 8 shows the marginal effects generated by the Model 3. According to the literature, these can be interpreted as probabilities (Wooldridge, 2016). The interpretation shows that RPS beneficiaries have a negative probability of 3.1% of requesting a loan. Overall the positive effects of reduced income risk can improve the creditworthiness of the household or increase their expected marginal returns to a loan because of the possibility of engaging in riskier productive activities. These positive effects are showed by the positive parameters of the model earlier exposed. However these seem to be offset by a larger effect that discourages the request of a loan. Because of the rise in liquidity the household faces decreasing marginal returns, which prove not to be enough over the marginal costs of requesting a loan. In this case RPS households have prioritized current consumption over credit request.

Table 8: Estimation of the probability of a household to request a loan by marginal effects of Probit Model 3.

Credit
RPS -0.031**

(0.013)

No_childr -0.010

(0.001)

No_women -0.008*

(0.004)

No_men -0.001

(0.004)

Av_hhage -0.001

(0.001)

Property 0.030

(0.001)

Assetindex100 0.004***

(0.001)

Social 0.056***

(0.013)

Otherprog 0.045***

(0.017)

Prev_remitt -0.005

(0.031)

Shock 0.020

(0.019)

T2 0.026**

(0.012)

*** Significant at the 1% level, ** Significant at the 5% level, *Significant at 10% level.

Standard errors: (in parenthesis)

Number of observations: 2718

Source: Authors based on Nicaraguan RPS evaluation data, StataCorp 2011

Further research opportunities.

Even though it is not the main purpose of this dissertation, it is important to mention that this study has had different results to the ones found by (Svarch 2011) and (Hernandez et al. 2009). As mentioned in Chapter 2 Svarch’s research is not under comparison because it involves the analysis of another case study. It could be possible that the opposite results of positive effects can be associated to the scale and urban setting the Mexican case has. In regards to (Hernandez et al. 2009) finding, it is possible different technicalities may allow different results. Firstly it is believed a Probit model may bring more accurate results than an OLS regression when the dependent variable is a dichotomous one (Wooldridge 2016). Moreover, in their model (Hernandez et al. 2009) do not control for attrition bias. The authors suggest the coefficient of interest may be upwards biased. This bias could be the one causing the contrast in the estimators between studies. While this dissertation suggests a negative significant effect of CCTs on credit market participation, (Hernandez et al. 2009) claim that CCTs do not modify in a significant way the probability of requesting a loan. Because the sample has low credit market participation on average, the differences on methodologies could be substantial. However, the understanding of the distinction between results and what is causing them is subject for further research, as it is not the main objective of this dissertation.

Another point subject to further research is the discrimination between types of credit. When analysing the raw data in Chapter 4 it was interesting to point out the evolution of formal credit request. This type of credit request was drastically reduced in the two years of intervention in the RPS households. However in control groups it was reduced by less than half only in the second year. Driven by these behaviours, two models are run to try to initially understand what are the effects of RPS on formal credit (Model 4) and informal credit (Model 5) requests.

As Table 9 shows, participation in RPS may discourage formal credit market participation and not informal one. It may be the case that formal credit request is associated to larger amounts used for productive investments while informal is to lower amounts for consumption. The formal credit higher costs (with expensive interest rates) may be causing the decreasing expected marginal returns of a household when requesting a formal loan while having more liquidity because of RPS. However, another reason for these results could be related to self-report issues. Households could be reporting less formal credit request after the reception of RPS. They may falsely fear the transfer may be taken away if the have access to other forms of private formal cash[7]. In addition, other reasons that this dissertation is not aware of could be playing a role in this analysis. Further qualitative and quantitative research should be conducted to understand this behaviour more clearly. The classification statistics and table of Model 4 show a 95% of correct classification (not shown). However, when running the goodness-of-fit test in Model 4 a significant chi-square is given (not shown). This suggests the Model 4 might be poorly specified and that there is still plenty room for improvement in these calculations. Because of the lack of consistent literature review and specification strategies supporting the results of Model 4 further research should be conducted.

Table 9: Formal and informal loan request determinants using a Probit model as shown in the equation.

Model 4 (Creditform) Model 5 (Creditinform)
RPS -0.454***

(0.097)

0.067

(0.077)

No_childr -0.083*

(0.049)

-0.014

(0.040)

No_women -0.004

(0.028)

-0.056**

(0.026)

No_men 0.046*

(0.026)

-0.027

(0.023)

Av_hhage 0.003

(0.005)

-0.010**

(0.005)

Property 0.283*

(0.153)

0.067

(0.099)

Assetindex100 0.024***

(0.006)

0.010

(0.007)

Social 0.605***

(0.088)

-0.002

(0.080)

Otherprog 0.463***

(0.139)

0.055

(0.094)

Remitt
Prev_remitt -0.026

(0.244)

0.091

(0.171)

Shock -0.140

(0.132)

0.213*

(0.109)

T2 0.303***

(0.087)

0.012

(0.070)

_cons -2.633***

(0.257)

-1.355***

(0.209)

*** Significant at the 1% level, ** Significant at the 5% level, *Significant at 10% level.

Standard errors: (in parenthesis)

Number of observations: 2718 (1359 households)

Source: Authors based on Nicaraguan RPS evaluation data, StataCorp 2011

Last but not least, the unexplored field of the effect of different CCT amounts on credit market participation could be investigated. Does the specific amount received influence the probability of requesting a loan?

Research limitations.

Besides the unanswered questions the use of different methodologies and types of credits bring for further research, other limitations of this study must be mentioned. Firstly, the results of this study at best provide an insight into the direction of the relation between transfers and credit market participation, but do not foster causality. Secondly, because of the nature of the information the model addresses a probability. This dissertation explains why that probability occurs by only associating theoretical links. Understanding the exact mix of the three theoretical reasons of why that probability is produced is out of scope of this study. Thirdly, at the expense of having a better technique for a dichotomous variable, the before and after treatment information is only used for sample characteristics descriptions. The popular Difference in Difference methodology used for other successful RPS studies is not taken advantage from in this dissertation.

In addition to the overall limitations, more specific variable related critiques could be mentioned. The reader must be aware that most of the variables were extracted from self-reported data. Additionally some variables as Property, Shock or Asset Index could be subject to critique. The latter could be qualified as simplistic because it does not differentiate the weight to a radio and a vehicle (Moser & Felton 2009). When analysing the property ownership the sign of the effect is hard to predict. Because of the lack of the available information, it is hard to know if it represents a form of collateral or if the household is in debt to have the asset ownership. In terms of the shock variable, it is difficult to discriminate between motivations to insure. The generalized coffee crisis experienced in Nicaragua at that time affected the high levels of self-reported shocks reports. These do not allow to contrast with other types of shock that could have happened (Maluccio 2005). Nevertheless it must be understood the outcome of this dissertation was done after the careful review of options with the best data available possible.

Last but not least, two model related limitations exist. Firstly, because of the use of the Probit methodology the model does not allow to control by fixed effects. An unobserved pattern like risk-propensity could exist. Nevertheless there is not a specific answer in the survey that can give information of those characteristics. This is why more variables such as number of men, women and children are added to the model to bring robustness under the absence of the fixed effects correction (Wooldridge 2016). Secondly, it is known that lagged variables are in some cases not the best solution of endogeneity (Bellemare et al. 2016). However Probit methodologies do not allow identification strategies such as Instrumented Variables. This is why a lagged variable of remittances is used, becoming the best strategy possible.

Policy recommendations.

In spite of the limitations mentioned above, this research can be use of several policy implications in design and implementation. On a specific note two findings must be highlighted. Firstly, after the results it is clear how relevant the amount decided for a transfer is to the programme objectives. In this case, larger amounts (compared to other CCTs in Latin America) foster current consumption in specific themes as food and education, over credit application. Secondly, the results support the importance of household composition for the pursuit of different objectives. In this study women are found to contribute to better CCT objectives of current consumption over credit. Therefore, women should continue to be the ones to receive the transfers when conditions want to be met.

On a more generalist note a last recommendation can be given. This dissertation reinforces the importance of clear policy objectives and strategy support. If correctly designed, policy objectives are met clearly without undesired effects as a consequence. However, this approach may hinder innovation for poverty reduction. After reviewing the case of RPS it is recommended to work coordinated with other governmental programmes while pursuing innovation in CCTs. In this case it seems that other public programmes that appear in specific risk situations fulfilled the risk protection role. If development was done coordinated like this case, it could leave room to broaden the CCTs objectives. What if a CCT that has a human capital conditionality (like school attendance), could aim to give cash for investment in productive activities rather than current consumption? It would be interesting to understand what could be that amount that allows a change in the household production mix, therefore allowing credit application and other growth strategies.

Chapter 6: Conclusion.

CCTs are a form of social protection that have had massive expansion in developing countries in the last two decades. With the impact evaluations of the pioneer cases in Latin America CCTs have been catalogued as successful poverty reduction strategies. Poverty reduction is done though its short-term safety net objective, in addition to the long-term human capital investment due to its conditions. Moreover, positive undesired effects like decreased child labour bring robustness to the strategy (Barrientos 2013a; Fiszbein & Schady 2009). Their impacts shape society in a long-lasting way. There is evidence that successful indicators in human capital themes, such as education, are still achieved even ten years after the intervention (Barham et al. 2013). However, CCTs have been subject to critique mainly in three aspects. Firstly, the lack of several years’ evaluations hinders long-term outcome analysis. Secondly, their sustainability and understanding if the policy pays for itself is a concern in the field (Glewwe & Kassouf 2012; Caldés et al. 2006). Thirdly, the solely focus on building human capital limits CCTs as a growth strategy. The specific objectives could be the key to its success. Nevertheless they leave no room for broader innovations in the development of the poor (Handa & Davis 2006). Because of the latter, this dissertation tried to analyse what are the effects of CCTs on rural credit market participation.

Access to credit is considered a profitable growth strategy at national, local and individual scale. However credit market participation at household level after the receipt of a transfer is theoretically ambiguous. On one side the additional cash could hinder credit market participation. Getting a loan to have even more cash may have higher relative cost compared to completing the CCT’s conditions. This may occur because of the decreasing marginal returns characteristic credit has. On the other hand, having additional liquidity due to the CCT may change the household production function. This could bring incentives to apply for credit to expand their production activities. Moreover, in the supply-side additional cash could be perceived as additional creditworthiness by the lenders (Hernandez et al. 2012; Svarch 2011; Morduch 1995; González Vega 2003). The contradictory theoretical framework demanded an empirical analysis to understand the specific case of the Nicaraguan RPS.

A Probit regression analysis with a dichotomous variable was run to bring light to the matter. The results show RPS significantly decreased the likelihood of requesting a loan in rural Nicaragua. Following our theoretical framework it is possible that RPS could have lowered the expected marginal returns of a loan compared to its expected costs. This may have occurred firstly because its beneficiaries correctly followed RPS recommendations. Support from right cash amounts and supply side interventions where necessary. Secondly it is possible the amount of the CCT was not enough to change the household’s production mix. In this sense the family may have perceived the costs of requesting a loan as higher than the obligations needed to perceive the CCT. Thirdly a larger amount of women could have contributed to this outcome. Women are considered better guardians of the transfer and its conditions. It is not surprising this possibly less risk-adverse gender pushed for the pursuit of the RPS objectives over credit participation. Lastly, the positive effects brought by household characteristics as wealth seem to be offset by a larger effect of RPS. It is possible other governmental programs other than RPS may have acted as risk-protection. Again the latter may have reinforced the use of the CCT for human capital consumption over credit. All the above-mentioned findings are supported by other studies. Other authors have already proved increased health and education indicators, higher consumption in food and education and no change in the expenses in agricultural and non-agricultural activities thanks to RPS.

After analysing the results, three main unanswered questions are left for further research. Firstly this dissertation does not analyse how the different amounts perceived from RPS may have an effect of credit market participation. Secondly, the fact that other authors have arrived to different results using a different methodology demands further exploration. Especially in the case where credit application in the sample was low and changes to methodology could be substantial. Thirdly, the relation between formal and informal credit application is unexplored. Some initial conclusions suggest RPS could have reduced the chance of formal loan applications but not informal ones. However the model itself that proves such theory is not robust enough. In addition, this effect could be attributed to an economic phenomenon or to a self-report issue. Further investigations are then required. The above-mentioned questions mean the field has not been studied enough and that more robustness is demanded to seek understanding.

To conclude, this dissertation analysis brings light to certain policy implications in design and implementation. To start with, the role the amount of the CCT plays on household decision-making is found crucial. If the amount would have been lower or higher the household production function may have changed the results. Secondly, the results support the importance of household composition for the pursuit of different objectives. In this case more women are found to contribute to better CCT objectives. Thirdly, when an evaluation is designed it is best to avoid self-reported data that could bias the impacts. This reduction could allow better and further conclusions. Lastly, this dissertation reinforces the importance of clear policy objectives and strategy support. The latter motivates concrete policy indicators achievements and discourages undesired effects. However, on the other side, it hinders innovation for poverty reduction. Creating a CCT with less restrictive objectives could enlarge growth opportunities. What if a CCT could enforce long-term human capital development through its obligations like school attendance (with the support of the supply side interventions); but on the short-term could foster the use of cash for productive activities and a change in the household production mix. What could the combination of human capital and consumption in productive riskier activities bring for short and long-term growth?


[1] Asiganción Universal por Hijo (Universal Child Allowance in english) is Argentina’s CCT. It started in 2009 and has had one of the fastest growth in the region ever since.

[2] Glewwe and Kassouf (2012) find that the long-term benefits to the economy of the additional years of education produced by Bolsa Scola are greater than the initial cost.

[3] ”Impact” is referred specifically to the evaluation of the programme objectives, while “effect” is referred to other consequences due to programme implementation.

[4] For further understanding of the sample size calculations and randomization methods used please refer to IFPRI (2005).

[5] Which is around US$304 at September 2000.

[6] Please refer to Model 1 in Table 11 to review the results of this variable.

[7] Other programs other than RPS could not have this issue because they are public funded.

Cite This Work

To export a reference to this article please select a referencing stye below:

Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.

Related Services

View all

DMCA / Removal Request

If you are the original writer of this essay and no longer wish to have the essay published on the UK Essays website then please:

McAfee SECURE sites help keep you safe from identity theft, credit card fraud, spyware, spam, viruses and online scams Prices from
£29

Undergraduate 2:2 • 250 words • 7 day delivery

Order now

Delivered on-time or your money back

Rated 4.0 out of 5 by
Reviews.co.uk Logo (23 Reviews)

Get help with your dissertation