• unlimited access with print and download
    $ 37 00
  • read full document, no print or download, expires after 72 hours
    $ 4 99
More info
Unlimited access including download and printing, plus availability for reading and annotating in your in your Udini library.
  • Access to this article in your Udini library for 72 hours from purchase.
  • The article will not be available for download or print.
  • Upgrade to the full version of this document at a reduced price.
  • Your trial access payment is credited when purchasing the full version.
Buy
Continue searching

Household saving behavior: Empirical evidence based on Mexican Households Surveys

Dissertation
Author: Jesus Sandoval-Hernandez
Abstract:
In this dissertation I study different aspects of households saving behavior in Mexico. In chapter one, I use micro data from ten rounds of the Households Surveys of Income and Expenditures (ENIGH), to offer a multivariable analysis and investigate the empirical determinants of household saving behavior in Mexico for 1984-2006. Using the Life Cycle Hypothesis (LCH) as theoretical background, I identify some stylized facts of household saving behavior and address the issue of to what extent this behavior is in line with the LCH's predictions. Through the implementation of cross-sectional analysis and synthetic panels, I carry out a comprehensive characterization of saving patterns that complements and extends previous studies in several dimensions. The results suggest that demographics characteristics and family composition capture some of the household's motivation for saving more at the early and last stages of the life cycle, resulting in a U-shaped age saving profile. The inclusion of variables capturing the relaxation of liquidity constraints exhibits the negative correlation between access to financial resources and lower saving rates. In the second chapter, I analyze the impact of financial liberalization and aggregate shocks on households saving behavior. In particular, I investigate to what extent households with access to financial resources modify their saving behavior when the economy experiences either credit expansions--a byproduct of financial liberalization and macroeconomic stability--or credit contractions resulting from financial and banking crises. I estimate a saving function and employ a difference in difference technique to identify the effects of these two types of "macroeconomic shocks" on households saving behavior. It is found that the effects of macro shocks and financial liberalization are stronger for households with better access to financial resources. In 1997, the Mexican government approved a comprehensive pension reform that affected the formal labor force in the private sector. The legislation did not affect workers employed in the public sector, however. In chapter three, I use this policy change as a natural experiment to evaluate the impact of pension reforms on household saving behavior. The results indicate that households adopting the new funded pension system increase their saving rates systematically.

Table of Contents List of tables vi List of figures vii Abstract viii Dedication x Acknowledgments xi Introduction 1 Determinants of households saving 11 1 Introduction 11 2 Literature Review 16 2.1 Incentives to save 16 2.2 Demographics and saving 18 2.3 Precautionary saving and liquidity constraints 20 3 Data and definitions 22 3.1 Data 22 3.2 Definitions of households' saving and variables 23 4 Characterization of saving patterns 25 4.1 Cross-sectional analysis 25 4.1.1 Descriptive analysis 25 4.1.2 Stylized facts 29 4.1.3 Regression equation 32 4.1.4 Results from cross-sectional regressions 37 iii

4.2 Synthetic panels 46 4.2.1 Synthetic panels results 50 5 Concluding remarks 55 Households' saving, financial liberalization and aggregate shocks 57 1 Introduction 57 2 Macroeconomic environment and financial system reforms 60 2.1 Antecedents 60 2.2 Re-privatization of the banks and the peso crisis 62 2.3 The 1990s post crisis reforms and the internationalization of banks 63 3 Empirical strategy 65 4 Results 70 4.1 1984-1989: Credit contraction, crisis and recession 71 4.2 1989-1992: Financial liberalization and credit expansion... .72 4.3 1994-1996: Credit contraction and crisis 74 4.4 1998-2000: Financial reforms, credit contraction, and positive growth 74 4.5 2000-2002: Mild recession 75 4.6 Period 2002-2006: Credit expansion 73 5 Concluding remarks 77 Households saving and pension reforms 79 1 Introduction 79 2 Pension Reforms 83 2.1 Background 83 2.2 Why the reforms were necessary? 85 2.3 The reforms 87 3 Methodology 89 3.1 Estimation of the ATT based on the propensity score 90 3.1.1 Selection bias problem 90 IV

3.1.2 The propensity score matching method 91 3.2 Characteristics of the treated and control groups 93 3.3 Propensity score estimations 96 3.4 Difference in difference estimation 97 4 Results 98 5 Conclusions 101 Tables 103 Figures 123 Appendix A Income and expenditures households surveys (ENIGH) 136 B Mexico's regions, INEGH's classification 139 C Matching methods 140 Bibliography 143 v

List of Tables Table 1: Mexico's National Saving 100 Table 2: Medians of income, consumption, saving, and saving rates 1984-2006.... 101 Table 3: Demographics and Family's Composition, ENIGHs 1984-2006 102 Table 4: Percentages of Households with Dummies=l 103 Table 5: Households Saving Rates (medians) at Different Percentiles 103 Table 6: Median Regressions of Saving Determinants 1984-2006 104 Table 7: Cohorts definitions and cell sizes by head of household 108 Table 8: Cohorts saving rates (Median values) 109 Table 9: Saving rate median regressions. Dependent Variable S/Y 110 Table 10: Accumulated Growth Rate of Credit During selected Periods I l l Table 11: Timing of credit and business cycles, and financial liberalization 1984- 2006 112 Table 12: Probit regressions of access to credit indicators 113 Table 13: Effects of credit cycles and macro shocks on household saving 114 Table 14: Contributions to the Pension System before and after the reforms 116 Table 15: Main characteristics of the sample 117 Table 16: Propensity score estimation, probit regressions 118 Table 17: S aving rates average effects of the treatment over the treated 119 vi

List of Figures Figure 1: Savings and Growth: Latin America and East Asia (WEO, 2007) 120 Figure 2: Medians of disposable income, consumption, and saving rates by age of HH of household 121 Figure 3: Medians of saving rates by educational attainment and age 123 Figure 4: Income and consumption expenditure profiles by cohorts 125 Figure 5: Income and consumption expenditures by cohorts smoothed (median values, selected cohorts) 125 Figure 6: (a) Saving by cohort; (b) Saving by cohort smoothed 126 Figure 7: (a) Saving rates by cohort; (b) Saving rates by cohort smoothed 126 Figure 8: Panel A Saving by cohorts; Panel B Saving by cohort smoothed 127 Figure 9: Total credit, credit to consumption, to credit cards, and to mortgages 128 Figure 10: Annual percentage change of GDP 129 Figure 11: (a) Share of assets (%0 Dec, 1997; (b) Share of assets (%) Dec, 2006 130 Figure 12: Estimated propensity scores 131 vn

An Abstract of Household Saving Behavior: Empirical Evidence based on Mexican Households Surveys Jesus Sandoval-Hernandez In this dissertation I study different aspects of households saving behavior in Mexico. In chapter one, I use micro data from ten rounds of the Households Surveys of Income and Expenditures (ENIGH), to offer a multivariable analysis and investigate the empirical determinants of household saving behavior in Mexico for 1984-2006. Using the Life Cycle Hypothesis (LCH) as theoretical background, I identify some stylized facts of household saving behavior and address the issue of to what extent this behavior is in line with the LCH's predictions. Through the implementation of cross-sectional analysis and synthetic panels, I carry out a comprehensive characterization of saving patterns that complements and extends previous studies in several dimensions. The results suggest that demographics characteristics and family composition capture some of the household's motivation for saving more at the early and last stages of the life cycle, resulting in a U-shaped age saving profile. The inclusion of variables capturing the relaxation of liquidity constraints exhibits the negative correlation between access to financial resources and lower saving rates.

In the second chapter, I analyze the impact of financial liberalization and aggregate shocks on households saving behavior. In particular, I investigate to what extent households with access to financial resources modify their saving behavior when the economy experiences either credit expansions —a byproduct of financial liberalization and macroeconomic stability— or credit contractions resulting from financial and banking crises. I estimate a saving function and employ a difference in difference technique to identify the effects of these two types of "macroeconomic shocks" on households saving behavior. It is found that the effects of macro shocks and financial liberalization are stronger for households with better access to financial resources. In 1997, the Mexican government approved a comprehensive pension reform that affected the formal labor force in the private sector. The legislation did not affect workers employed in the public sector, however. In chapter three, I use this policy change as a natural experiment to evaluate the impact of pension reforms on household saving behavior. The results indicate that households adopting the new funded pension system increase their saving rates systematically.

Dedication This dissertation is dedicated to the memory of my mother Margarita Hernandez Barbosa, from whom I learned the most essential values that guide my life, including the value of books. This is for you "Mama." The arduous journey through graduate school was alleviated by the support and love of my family. I want to dedicate this thesis to my father Jose Sandoval, my sisters Armida, Elvia, Rocio, and Alma, my brothers Damian, Guillermo, and Gello, my nieces and nephews, especially to Andres. To Rito in memoriam. x

Acknowledgment I thank, above all, my advisor Yin-Wong Cheung for his never-ending help and encouragement since I started my doctoral studies, and for his exceptional advising. I would also like to thank Professors Michael P. Dooley, Keneeth Kletzer, Joshua Aizenman, and Michael H. Hutchinson for valuable guidance and support. I am grateful for financial support from the Graduate Research Mentorship Program of the Graduate Division at UCSC, and from the UC MexUS Research grants. xi

Introduction In the neoclassical growth theory saving plays an essential role in shaping an economy's growth trajectory. High saving rates foster capital accumulation, generate higher levels of investment, and trigger a virtuous cycle of saving, investment and economic growth.1 Developed countries may have an almost unlimited access to international financial markets and possess highly sophisticated financial systems. These factors render less relevance of domestic saving in these economies. In contrast, developing countries face considerably more obstacles for external financing. Usually, capital inflows are more volatile and often generate balance of payment problems and economic crises.2 Domestic saving plays a critical role in the development process of poor countries. They function as a buffer stock during a crisis period and adjust to expectations about the future and to the current financial markets conditions.3 1 The relationship between saving and growth is crucial in the neoclassical growth models of Solow (1956), Koopmans (1965), and Ramsey (1928), and the AK models of Domar (1946), Frankel (1962) and Romer (1990). 2 The Mexican Peso crisis in the mid-90s, and the East Asia and Russian Crises of the late 1990's are the textbook examples. 3 Aghion et al (2006) argue that domestic saving affect growth in poor countries even with international capital mobility. In these countries saving provide with collateral, which induces local entrepreneurs into the effort needed to make foreign investment profitable. Their model shows how foreign direct investment is required to transfer frontier technological knowledge to local innovating sectors in these countries. The effect of saving on growth is reduced for rich countries because their small distance to technological frontier makes them less dependent on FDI to adopt new technologies. 1

National saving rates in developing countries differ a lot. For example, since 1980, average domestic saving rates in countries of the East Asia region surpassed 30 percent of their GDP. In contrast, The Latin America region's saving rates revolve around 20 percent. Figure 1 shows the staggering contrast between the high economic and saving growth rates in East Asia and the low rates in Latin America for most of the last three decades. This gap between the two regions could hardly be explained by differences in property rights protection, financial development or access to international financial markets. Some authors argue that East Asia benefited in the recent past from demographic shifts, in which the relative size of the age groups that produce and save the most has been increasing. Accordingly, this demographic change is one of the main reasons why East Asian countries' economic performance and their saving rates have been enhanced (Bloom and Williamson, 1999, Schultz, 2004). For households in developing countries, saving is a crucial determinant of welfare. Facing income risks, the ability of liquidity-constrained households to smooth consumption is limited. In such conditions, households may decide to accumulate more assets or engage in precautionary saving in "good" times to shelter consumption in "bad" times. Saving is one of the only means to accumulate assets in the absence of efficient financial intermediation or insurance markets; the capacity to save becomes one of the best vehicles of social mobility and of enhancing future income possibilities. 2

An accurate characterization of household saving patterns in an economy is important not only for the central role of saving discussed above, but because different saving decisions imply different patterns of saving among individuals and in the aggregate economy. Different saving decisions also imply different responses of both individual welfare and aggregate saving to policies such as interest rates, direct saving incentives, such as tax deductions, or the provision of social security. In the saving and consumption literature, a researcher faces the dilemma of using aggregate data to obtain high frequency information (weekly, monthly, quarterly, etc.), but with this type of data, problems of aggregation arise. Alternatively, she can use microeconomic data, collected directly from the individuals who ultimately make the decisions on the relevant variables. With this type of data aggregation problems are eliminated, but it constitutes low frequency data4. In this dissertation, I take the microeconomic approach to study one component of private saving, households saving. Specifically, I characterize the determinants of household saving decisions in Mexico during 1984-2006. I exploit the availability of a large number of household surveys from the Mexican Household Survey of Income and Expenditures (Encuesta Nacional de Ingresos y Gastos de los Hogares (ENIGH). Only a handful of previous works examined saving behavior at micro level in Mexico. However, all of these studies were conducted for the late 4 Usually households' surveys are collected at most every year. 3

1980s and early 1990s.5 With the additional most recent surveys, the data spans for a long time period (twenty-two years), covering three recessions (1986, 1994-95, and 2001), and three periods of recovering (the late 1980s-early 1990s, 1996-2000 and 2002-2006). The characterization of households saving's pattern for an important emerging market such as Mexico constitutes an interesting case study on its own. The data covers periods of major structural economic reforms and economic distress; this is particularly relevant because it offers an opportunity to gauge the effects of this macroeconomic environment on households' saving. In the macroeconomic sphere, Mexico made considerable progress with the elimination of the budget deficit, a sharp reduction of inflation, the liberalization of its trade regime—including the implementation of the North American Free Trade Agreement (NAFTA)—, the adoption of both current and capital-account convertibility in its international transactions, and the re-privatization of its banking sector. In the late 1990s Mexico adopted a second wave of structural reforms where the country sharply changed the legal and regulatory environment of the financial sector. These institutional changes included the opening of the banking sector to foreign ownership, the tightness of risk management practices and capitalization 5 Papers on households saving using micro data mushroomed in the mid to late 90's, mostly for developed countries. There are a few studies for Mexico, and they use at most five surveys in their analysis. The most recent survey used was the one from 1996 (see for example Szekely (1998), Calderon (1998), Villagomez and Zamudio (2000) Aportella (2001), Fuentes and Villagomez (2000), and Attanasio and Szekely (2000)). 4

requirement, and the creation of a bureau of credit to have a better information system about borrowers.6 The scarcity of studies analyzing the impact of these events on households saving at micro level is quite surprising. This work attempts to fill the gap. Another contribution of this thesis is related to the fact that Mexico went through the process of privatizing most of its social security system in 1997. In spite of the important policy issues that retirement saving raise, research focusing on the repercussions of different pension schemes on households' saving has been scant. This paper contributes to the literature on pension reforms and household saving behavior by examining empirically the impact of such reforms on households saving rates. Household saving rates increased from 7.1 percent in 1984 to 8.4 percent in 1989 followed by a decreasing trend with its lowest point in 1996. After the peso crisis, household started increasing their saving; however, by 2006 they saved only 6.8 percent of their income. Although relatively high, the later figure indicates a decline of 24 percent or a drop of 1.6 percentage points compared with saving in 1989 (see table 1). The steady decline of households' saving and the changing macroeconomic environment raise a variety of questions. Are demographical 6 These reforms have been considered to play an important role for the current healthy state of the Mexican financial sector. For instance, in a recent article The Economist reported that Banamex, the second largest bank in Mexico and a subsidiary of Citigroup, helped its "crisis stricken parent" with more than 40 percent of Citi net income originated from overseas operations between January and March of 2008, ("Mexico's fast-growing banks appear unusually unaffected by the financial crisis north of the border", The Economist, May 1st, 2008). 5

changes the reasons for this drop? What are the roles of financial liberalization, credit cycles, and economic crises in shaping household savings patterns? Does the access to financial resources have an impact on households' saving decisions given the changes in the macroeconomic environment? How the replacement of the public pay-as-you-go pension system by a private voluntary pension system had affected households' saving behavior? These are important questions with policy implications addressed in this dissertation. In chapter one, I use the framework of the Life-Cycle Hypothesis (LCH) of Modigliani and Brumberg (1954) to conduct an empirical investigation of households' saving patterns. The unifying framework of the LCH allows for a consistent analysis of the large amount of data included in the household surveys. The LCH in its simplest version implies that individuals' consumption and saving through their lives evolve according with their age. The analysis of the evolution of saving behavior by households that go through different stages of their life is also useful to characterize the response of saving to different shocks. Rather than only considering the initial formulation of the life-cycle model, I also incorporate family's composition, demographic factors, and liquidity constraints considerations into what could be characterized as a more general incarnation of the LCH theory. The analysis is conducted in various steps. First, I inquire whether, in line with previous findings, saving behavior is sturdily influenced by age and demographic factors. To address this issue, I constructed cross-sectional age saving 6

profiles. Then, for each survey, I implemented cross sectional regressions of household saving on a group of theoretical determinants of saving. The regressions include the Life-Cycle factors, a group of proxies for access to financial resources, and a group of controls. By incorporating proxies for access to financial resources, I seek to analyze the role of liquidity constraints and their linkages with the accessibility to financial resources on households' saving.7 Although cross-sectional analysis does not capture time and cohort effects, it may be still informative on the effects of these variables on saving for a specific point in time. Second, given the impossibility of analyzing an intrinsic dynamic phenomenon, as saving, using cross-sectional analysis, a further step in this exercise is to include the dynamic aspects of saving decisions. Since individual households are not followed across the ENIGH surveys, I cannot identify them in different surveys. To overcome these shortcomings, I make extensive use of the synthetic cohort techniques proposed by Browning, Deaton and Irish (1985) and widely used, since then, in a variety of situations . In principle, the method amounts to following the average behavior of groups whose membership is assumed to be fixed over time. This procedure allows me to study the dynamic behavior of the average of the variable of interest in different years. 7 In this work, liquidity constrained households have limited or null access to credit (formal and informal), have not access to financial instruments such as credit cards, receive not transfers in the form of remittances, government's and/or relatives' transfers, and have little or not real and financial wealth. 8 For studies using this technique on the ENIGH surveys see for example Szekely (1998), Attanasio and Szekely (1998), Calderon (1998), Attanasio and Szekely (2000), and Fuentes and Villagomez (2001) among others. 7

To anticipate the results, demographic characteristics and family composition capture some of the household motivation for saving more at the early and last stages of the life cycle, resulting in a U-shaped age saving profile for most of the period. The explanatory power of the model increases considerably when proxies for access to financial resources, net wealth, access to health services, and home ownership are incorporated into the cross-sectional regressions,. The inclusion of variables capturing the relaxation of liquidity constraints exhibits the negative correlation between access to financial resources and lower saving rates. The main findings in the cross-sectional analysis part are validated by the cohorts' analysis. In chapter two, I investigate the impact of financial liberalization and aggregate shocks on households saving behavior; the latter defined as pronounced economic downturns due to currency and/or financial crises. In particular, I investigate to what extent households with access to financial resources modify their saving behavior when the economy experiences either credit expansions —a byproduct of financial liberalization and macroeconomic stability— or credit contractions resulting from financial and banking crises. I estimate a saving function9 and employ a difference in difference (DD) technique to identify the effects of these two types of "macroeconomic shocks" on households saving behavior. If liquidity constraints affect the ability of households to smooth consumption, then in response to income risk households may decide to accumulate more assets or engage in precautionary saving in 'good' times in order to shelter consumption in 9 It incorporates the life-cycle and liquidity constraints components considered in the first chapter. 8

'bad' times. This implies, on one hand, that the effect of a negative shock, say for example the economic crisis of the mid-1990s, would have a greater impact on saving rates of less liquidity constrained households. On the other hand, although the effects of financial liberalization on households' saving are theoretically ambiguous, it is possible that in the short-run financial liberalization would loosen liquidity constraints by increasing the availability of credit and affecting the interest rates. If so, financial liberalization should be correlated with households' access to financial services; for example by an increase of the availability of credit for consumption. Consequently, financial reforms should be negatively related with households saving. The negative impact of these reforms on saving rates should be greater for less liquidity constrained households. The findings in this chapter suggest that during periods of economic crisis and credit contraction households with better access to finance increased their saving rates, and during periods of financial liberalization, credit expansion, and economic expansion, these households decrease their saving. In the third chapter, I evaluate households' saving behavior under two different pension regimes. The 1997's social security reform affected the formal labor force in the private sector. Under this reform, workers were forced to move from a state-run pay-as-you-go (PAYG) system to a private pension plan. Workers in the public sector were unaffected, however. This made possible to identify two segments of the labor force with similar characteristics but under different pension 9

regimes and evaluate the impact on savings decisions that a change in policy, in this case the introduction of a different pension system, may have. By extracting two subsets from the data set, I am implicitly assuming that they are random sub-samples of the population. However, it is possible that enrollment to a pension regime would be correlated with individuals personal attributes, implying a selection problem that would bias the results. To "correct" for the selection bias and be able to make valid inferences, I use a matching approach10. This means that I mach, using a propensity score approach, households enrolled in the "new" pension system with otherwise similar households but that still are under the PAYG system. Once I have matched households in this manner, I will then be able to compute the effect of the privatized pension system on households saving rates. This effect takes the form of 'an average treatment on the treated" effect, where the treatment is taken as whether a household is in the privatized pension system. In general, the results show strong evidence of the positive correlation between saving rates and the membership of a private pension regime. 10 Rosenbaum and Rubin (1983) proposed propensity score matching as a method to reduce the bias in the estimation of treatment effects with observational data sets. 10

Chapter One Determinants of Households Saving 1 Introduction In the last few decades, the Mexican economy alternated between periods of economic prosperity followed by abrupt downturns. Apparently, the remarkable economic reforms pursued by different administrations were not totally effective in containing sharp swings in the economy. It is conceivable that the relative slow economic growth experienced by the Mexican economy is a by-product of the wild economic fluctuations. Why is the Mexican economy performance so volatile? One possible reason is that Mexico does not have a level of national savings that cushion the economy from shocks. Without a proper savings buffer, consumption, and investment will respond quite vigorously to adverse shocks originated from either within and outside 11

the economy. Further, a low level of savings encourages neither investment nor growth. Although the average national saving rate in Mexico during the period 1980- 2006 was similar to those of Latin America and OECD countries (22.6, 21.0 and 21.3 percent respectively), the impressive high average saving rates of East Asia and China (36.4 and 39.6 percent respectively) exhibit the dearth of the Mexican national savings rates.11 One interesting observation linking volatility of economic performance and saving is that following a year of economic contraction or economic crisis, external saving (capital inflows from abroad) dropped, but its decline was largely compensated for by an increase in domestic saving, where the private component of domestic saving increased considerably. For instance, when external saving dropped from 7.1 to 0.5 percent between 1994 and 1995, domestic saving increased from 14.7 to 19.3 percent, and private domestic saving increased from 11.3 to 15.0 percent in the same period (Table 1). In the aftermath of a strong negative economic shock one would expect that at least the private sector would try to smooth out the shock by reducing saving rates to maintain consumption and investment levels. Although not inconsistent with the evidence from other countries, the increase of private domestic saving during the recession years seems puzzling. 11 IMF World Economic Outlook, 2007 and Banco de Mexico. 12 What it is known as the "Paradox of Thrift" has been observed recently in the United States in the aftermath of the 2008's financial crisis. For instance, citing predictions of Goldman Sachs, The Wall 12

Unfortunately, the Mexican system of National Accounts does not break down private saving into its household and corporate components13. Therefore, it is not possible to know what has been the role played by individuals or firms in the determination of aggregate saving. Aggregate household saving could move in the short run for a variety of reasons. Without a disaggregated analysis of household behavior and without identifying the structural reasons for changes in household saving, it is not possible to interpret movements in aggregate saving rates. The focus in this chapter is to shed light on the structural changes of household saving. For that purpose, I characterize household saving behavior in Mexico for the period 1984-2006, using data from the Mexican Household Survey of Income and Expenditures (ENIGH). The characterization of household saving patterns, not only constitutes an interesting case study on its own, but it offers an opportunity to gauge the effects of major structural economic reforms and crises on household saving behavior. The time span is of particular interest because during this period the Mexican economy not only was transformed by the adoption of remarkable economic reforms but it also went through a period of great volatility.14 Street Journal reported that, "... it is expected that personal saving in the US could increase from recent negative rates to 6 to 10 percent in 2009" ("Hard-Hit Families Finally Start Saving, Aggravating Nation's Economic Woes" The Wall Street Journal, January 6, 2009). 13 Instead, Private Saving must be calculated as "the residual of a residual," in two steps: first National Saving is obtained as the difference between Gross Investment and External Saving, and then Private Saving is calculated as the difference between National Saving and Public Saving. For a comprehensive review of Mexico's structural reforms see Lusting (1998). Three periods of economic stagnation (1984- 1986, 1994-1996, and 2001-2002) and three periods of economic recovering (1987-1994, 1996-2000, and 2002-2006) can be identified. 13

The use of micro data makes it possible to identify some factors not available from aggregate data, but that are relevant for policy considerations. In particular, micro evidence permits a decomposition of macroeconomics savings evolution, between individual behavioral changes and the effects of aggregating different types of individuals. Further, the analysis of microeconomic evidence allows for the possibility of studying the theoretical predictions on individual behavior at the household level. The analysis is conducted within the framework of the Life Cycle Hypothesis (LCH) of Modigliani and Brumberg (1954). The unifying framework of the LCH allows for a consistent analysis of the large amount of data included in the household surveys. The LCH in its simplest version implies that individuals' consumption and saving through their lives evolve in accord to their age. The analysis of the evolution of saving behavior by households that go through different stages of their life is useful in characterizing the response of saving to different shocks. Rather than considering only the initial formulation of the life-cycle model, I also incorporate family composition, demographic factors, and liquidity constraint considerations into what can be characterized as a more general incarnation of the LCH theory. Only a handful of studies had examined saving behavior at the micro level in Mexico. They are mostly of descriptive nature and use at most the first five available 14

surveys in their analysis. Unlike previous works, this paper examines household saving behavior in Mexico using both cross-sectional and synthetic panel approaches. I identify some stylized facts of household saving behavior and analyze the impact of changes in demographics and family composition on household saving. In addition, I present evidence on the negative correlation between access to financial resources and household saving rates. This work presents probably the most definitive picture of microeconomic households saving behavior in Mexico to date. The exercise has two components. First, I conduct cross-sectional analysis by constructing, for each survey, age saving, income, and consumption profiles. Then, I implement cross sectional regressions. The benchmark regression include the Life- Cycle factors, and it is augmented by adding family composition and demographic variables, a group of proxies for access to financial resources (liquidity constraints), and a group of controls. I6 Although cross-sectional analysis does not capture time and cohort effects, it may still be informative regarding the effects of these variables on saving for a specific point in time. The second component includes the dynamic aspects of saving decisions. Since individual households are not followed across the ENIGH surveys, it is not possible to identify them in different surveys. To overcome these shortcomings, I 15 The most recent survey used was the one from 1996. See for example Szekely (1998), Calderon (1998, 1999), Villagomez and Zamudio (2000) Aportella (2001), Fuentes and Villagomez (2000), Attanasio and Szekely (1998), Attanasio and Szekely (2000). 16 In this work, liquidity constrained households have limited or null access to credit (formal and informal), have not access to financial instruments such as credit cards, receive not transfers in the form of remittances, government's and/or relatives' transfers, and have little or not real and financial wealth. 15

Full document contains 162 pages
Abstract: In this dissertation I study different aspects of households saving behavior in Mexico. In chapter one, I use micro data from ten rounds of the Households Surveys of Income and Expenditures (ENIGH), to offer a multivariable analysis and investigate the empirical determinants of household saving behavior in Mexico for 1984-2006. Using the Life Cycle Hypothesis (LCH) as theoretical background, I identify some stylized facts of household saving behavior and address the issue of to what extent this behavior is in line with the LCH's predictions. Through the implementation of cross-sectional analysis and synthetic panels, I carry out a comprehensive characterization of saving patterns that complements and extends previous studies in several dimensions. The results suggest that demographics characteristics and family composition capture some of the household's motivation for saving more at the early and last stages of the life cycle, resulting in a U-shaped age saving profile. The inclusion of variables capturing the relaxation of liquidity constraints exhibits the negative correlation between access to financial resources and lower saving rates. In the second chapter, I analyze the impact of financial liberalization and aggregate shocks on households saving behavior. In particular, I investigate to what extent households with access to financial resources modify their saving behavior when the economy experiences either credit expansions--a byproduct of financial liberalization and macroeconomic stability--or credit contractions resulting from financial and banking crises. I estimate a saving function and employ a difference in difference technique to identify the effects of these two types of "macroeconomic shocks" on households saving behavior. It is found that the effects of macro shocks and financial liberalization are stronger for households with better access to financial resources. In 1997, the Mexican government approved a comprehensive pension reform that affected the formal labor force in the private sector. The legislation did not affect workers employed in the public sector, however. In chapter three, I use this policy change as a natural experiment to evaluate the impact of pension reforms on household saving behavior. The results indicate that households adopting the new funded pension system increase their saving rates systematically.