Patterns of household consumption in Vietnam

ing me the opportunity to participate in this course, in which I have received many benefits in terms of knowledge, approach and methodology of doing research. I wish to express my since thanks to my supervisor, Dr. Ngo Huy Duc for his patient attention, guidance and precious advice. His assistance helped me overcome difficulties and encourage me during the period of writing this thesis. I am indebted to Dr. Adam Mc.Carty and Prof., Dr.Sc. Vu Thieu for their efforts in running the Project and their strict disciplines toward writing a thesis. Their advice would still accompany with my life and my career in the future. My thanks are also sent to all my teachers in the Vietnam-Netherlands Project for Master Program in Economics of Development for their enthusiastic supports. I also wish to express my appreciation for the assistance of the project staffs; all my friends and classmates who helped me fulfill this thesis Finally, I would like to thank my family for their understanding and support during the thesis preparation process. TABLE OF CONTENTS CHAPTER 1: INTRODUCTION 1 The relevance of the thesis 1 Focus and scope of the thesis 2 Methodology 3 Data sources 3 Structure of the thesis 3 CHAPTER 2: THEORETICAL FRAMEWORK 5 Concepts and definitions 5 Household 5 Consumption 6 Consumption theories 6 Keynes’ consumption function 6 Fisher’s intertemporal choice model 8 Modigliani’s life-cycle hypothesis 10 Friedman’s permanent-income hypothesis 13 Hall’s random-walk hypothesis 15 Empirical study review 16 Chapter remarks 18 CHAPTER 3: AN OVERVIEW OF HOUSEHOLD CONSUMPTION IN VIETNAM 20 Characteristics of household in Vietnam 20 Household consumption in Vietnam in 2002 21 Household consumption by household income 21 Household consumption by region and quintiles 22 Household consumption between urban and rural area 24 Household consumption by household size and number of children 25 Household consumption by household head’s education 26 Household consumption and head’s age and gender 27 Household consumption by occupation of household head 29 Chapter remarks 29 CHAPTER 4: MODEL SPECIFICATION AND EMPIRICAL RESULTS 31 Model specification 31 Econometric formulation 31 Variables 32 Dependent variable 32 Explanatory variables 32 Data and estimation procedure 36 Data description 36 Estimation procedure 39 Estimation results 39 Chapter remarks 44 CHAPTER 5: CONCLUSIONS AND RECOMMENDATIONS 45 Conclusions 45 Recommendations 46 Limitations and suggestion for further studies 47 Bibliography 48 Appendix 51 LIST OF TABLES Table 3.1: Household consumption and HH income by quintiles (’000 VND/year) 22 Table 3.2: Household consumption by regions and quintiles (’000 VND/year) 23 Table 3.3: Household consumption by rural/urban and consumption quintiles (’000 VND/year) 24 Table 3.4: Household head’s number of schooling year by consumption quintiles 27 Table 3.5: Age of household head by consumption quintiles 27 Table 3.6: Household consumption by head’s gender and consumption quintiles (’000 VND/year) 28 Table 3.7: Household consumption by occupation of household head and consumption quintiles (’000 VND/year) 29 Table 4.1: Variables description 38 Table 4.2: Estimation of household consumption in Vietnam in 2002 40 LIST OF FIGURES Figure 1.1: Keynes’ consumption function 7 Figure 1.2: The consumer’s budget constraint 10 Figure 1.3: The life-cycle hypothesis of consumption 11 Figure 1.4: The Life-cycle consumption functions 12 Figure 1.5: Household consumption and permanent income 15 LIST OF APPENDIXES Appendix 3.1: Household consumption by number of schooling year 51 Appendix 3.2: Household consumption by household size 52 Appendix 3.3: Household consumption by number of children 53 Appendix 4.1: Regions classification in the data set 54 Appendix 4.2: Urban and rural classification 55 Appendix 4.3: Variable definition 56 Appendix 4.4: Estimation of full model 57 Appendix 4.5: Estimation of reduced model 58 Appendix 4.6: Analysis of correlation among variable 59 Appendix 4.7: Test normality of residual 60 Appendix 4.8: Other usual tests 61 ABBREVIATIONS ADB Asian Development Bank APC Average propensity to consume CPI Consumer Price Index GDP Gross Domestic Product GNP Gross National Product GSO General Statistic Office LCH Life-cycle hypothesis MPC Marginal propensity to consume MSE Residual Mean Square OLS Ordinary Least Square PIH Permanent income hypothesis P-|t| Associated p-value SOE State-owned enterprise Std.Dev. Standard Deviation UNDP United Nations Development Program VHLSS Vietnam Household Living Standard Survey VLSS Vietnam Living Standard Survey VND Vietnamese Dong WB World Bank CHAPTER 1 INTRODUCTION “Consumption is the sole end and purpose of all production” - Adam Smith - THE RELEVANCE OF THE THESIS In any society, one of the ultimate objectives of the economic system is to deliver goods and services to its members. The success of an economy can be measured by its ability to provide for its people, to feed them, to clothe and shelter them, and to offer them access to good health, to education and to a wide range of consumer goods (Deaton and Case, 1987). Based on such things, to evaluate the prosperity and social welfare of an economy, consumption is the most appropriate measurement. At macro level, consumption, or more precisely aggregate consumption, is a major component of aggregate demand or so-called gross national product (GNP). It runs about 65 percent of GNP in the United States (Branson, 1989) and even higher in many developing counties, accounted for about three fourth of annual GNP (Deaton, 1997). Because of its significance, consumption becomes direct or indirect target of many economic polices such as demand stimulating policy, other fiscal policies and national programs. With respect to consumer behavior at micro level, household consumption is also of interest. Patterns of consumption within the budget constrain are important in examining the structure of household activity, while the amounts consumed of various goods, particularly foodstuffs, are usually used in determining nutritional status, health and life-expectancy, which in turn results in the quality and quantity of human resource in the economy. Considered to be one of the most successful economies in transition, Vietnam has experienced high growth rate of about 8 percent annually (GSO, 1998). Consequently, a large number of poor household have been escaped from poverty and living standard has much improved. A decade ago, 58 percent of the population had an expenditure level that was insufficient to support a healthy life but this figure had fallen to 29 percent by 2002 (Rama, 2004). To measure of these changes in Vietnamese living standards, consumption is used extensively and taken an important component of human welfare. In much of the World Bank’s operational work as well as in applied research, consumption constructed from survey data have been used to measure poverty, to analyze changes in living standards over time, and to assess the distributional impacts of various programs and policies in Vietnam. Although there were a great number of researches attempting to analyze and evaluate different aspects of poverty in Vietnam by using VLSS 1992 and VLSS 1998 data sets, less work has been done on household consumption. Moreover, the limited accuracy of data on consumption from VLSS 1992 and VLSS 1998 has hindered researchers’ endeavors in studying on household consumption composition, as well as its determinants. For those reasons, basing on an accurate data set, VHLSS 2002, to find out determinants of household consumption is worth of researching. Beyond the analysis of household consumption, a foundation for proposing policies on consumption stimulation established. It is particularly relevant to developing countries like Vietnam of which consumption occupies a large share of aggregate demand. FOCUS AND SCOPE OF THE THESIS This thesis focuses on the demand side of the household consumption, of which household expenditure is an official proxy. Main purpose of this paper is investigating determinants of household consumption. Other factors in supply side including price, inflation, and supply quantity will be neglected. The thesis also tries to examine fixed effects of household characteristics on consumption level rather than random effects. Applied model is an expansion of fundamental macroeconomic consumption model introduced by Keynes. Major objective of the thesis is finding out the answer to the central question: “What are determinants of household consumption in Vietnam?” Sub questions: To what extent and in what dimension does the household characteristics affect to consumption? Is there any geographic effect on patterns of household consumption? What are implications for consumption stimulating and development policy? METHODOLOGY The study is developed in analytical framework from demand side of consumption. This thesis will firstly review fundamental theoretical issues relating to household consumption. Then, it will concentrate on descriptive analysis of household consumption relying on data from VHLSS 2002. A quantitative analysis using linear econometric model will be applied to demonstrate main characteristics of household affecting to the household consumption level. Finally, it will show key findings and suggest some policy implications for enhancing household consumption in Vietnam. DATA SOURCES For primary analysis, relying data is Vietnam Household Living Standard Survey (VHLSS) carried out by the General Statistic Office (GSO) in 2002. The nation-wide household survey was part of the Living Standards Measurement Study (LSMS) household surveys with technical assistance from the World Bank. It is cross sectional in design and administered at the individual level with nearly 140,000 observations. However, the number of observation at household level is only nearly 30,000. The VHLSS 2002 used questionnaires contained 9 sections each of which covered a separate aspect of household activities. Those are: (i) Household member; (ii) Education; (iii) Employment; (iv) Health; (v) Income and Source of income; (vi) Expenditure; (vii) Fixed assets and durable things; (viii) Housing; (iv) Participant in poverty alleviation program. The whole data set is recorded in STATA data files. Those files include data about household size, age, sex, education, occupation, expenditure, income, and other household compositions and geographic features that are available for studying. STRUCTURE OF THE THESIS The thesis includes five chapters. After the introduction, chapter 2 provides key concepts, and general theoretical framework on household consumption. This chapter firstly reviews fundamental consumption theories, which show numerous approaches explaining consumption behavior of household. Then, it goes over key findings of some relevant empirical studies on the aspect. An overview of household consumption in Vietnam relying on data set VHLSS 2002 is presented in chapter 3. This chapter points out intuitive patterns of household consumption in Vietnam with descriptive tendency and expectations before coming to quantitative analysis. The core part of the thesis is chapter 4, which focus on quantitative analysis of household consumption including highlighting major features of household consumption, rationing and formulating econometric model, describing data, presenting and interpreting empirical results. Based on chapter 3 and chapter 4, Chapter 5 is dedicated to key findings and proposing some policy implications. CHAPTER 2 THEORETICAL FRAMEWORK This chapter is dedicated to discussing theoretical framework and empirical studies on household consumption. Some basic concepts and definitions will be presented in the first section of this chapter. The second section will concentrate on the evolution of theories on consumption. The third section is devoted to reviewing some outstanding empirical studies on household consumption. CONCEPTS AND DEFINITIONS Household The concept of household is defined and discussed by economists, feminists, and anthropologists. Chen and Dunn (1996) defined household as a domestic residential group whose members live together in intimate contact, rear children, share the proceeds of labor and other resources held in common, and in general cooperate on a day-to-day basis. However, there is much controversy on the definition. Generally, economists consider household as an essential unit of analysis and developing theories based on modeling household behavior. Chayanov model is the earliest one that integrates production and consumption activities in analyzing the peasant households. By combining utility maximization from consumption theory with the production from production theory, Chayanov model provides a foundation for integrated model of household behavior (Nguyen, 2003). As a conceptual concept, household is reckoned as a portfolio of economic resources, economic activities, and flows between them (Tran, 2000). Household resources are set of human, physical, and financial resources available for use by household in a given period. Household activities include consumption, production and investment activities that household member undertake in a given period. Household flows are sets of inputs and expenditures coming from household resources to support household activities. Besides, there are some definitions relating to the household. First, the household characteristics are defined as things like household composition or household typical features such as number of adults, children, occupation, age, and marital status. Second, household endowments are circumscribed as assets, which cover various factors of household and natural talent quality or ability such as land, labour, capital, inherit…(Nguyen, 2003). Consumption Consumption is the use of resources, goods, or services to satisfy want and need. It is exactly not the same thing as the household expenditure on the consumer goods and services. Households want to maintain a stable flow of services consumption those are the sources of utility to the households from a given kind of commodity (Sach and Larrain, 1993). Consumer goods comprise durables and nondurables. The distinction between them is the time they provide utility to user. Commodities considered to be non-durables if they give utility only in the course of using them up - a meal, a newspaper, and a weekend vacation. By contrast, other commodities providing satisfaction to the household over a long period as motorcycle, television sets are durables. According to Sach and Larrain (1993), consumption is properly measured as the sum of expenditures on non-durables and the flow of services rendered by existing stock of consumer durables. That is, most of current expenditures on consumer durables are actually investment spending rather than consumer spending. To some extend, definition of consumption is somewhat similar to expenditure. However, consumption is still primarily analyzed in the context of utility, demand and other importance to market exchanges. CONSUMPTION THEORIES Keynes’ Consumption Function Consumption is one of the key concepts in economics. As an initiative, in the General Theory published in 1936, John Keynes made the consumption function central to his theory of economic fluctuations and it has played a crucial role in macroeconomic analysis ever since. Keynes started his theory with three conjectures about consumption function, which rest with his causal observation. Those conjectures are: marginal propensity to consume (MPC) - the amount consumed out of an additional dollar of income - is between zero and one, average propensity to consume (APC) – defined as the ratio of consumption to income – falls as income rises, and income is the primary determinants of consumption and interest rate does not have important role (Mankiw, 1997). On this basis, a short run consumption model is often written as following: C = C0 + MPC.Y C0 > 0, 0<c<1 (1.1) Where C : Household consumption C0 : Autonomous consumption, the constant consumption which is not based on income MPC : Marginal propensity to consume Y: Current income C = C0 + MPC.Y C0 C Y Figure 1.1 Keynes’ consumption function Source: Branson, 1989 This function reflects the tendency that as income increase people tend to spend a decreasing percentage of income (Figure 1.1). The slope of linear consumption function is MPC and the slope from the origin is the APC at that point. Apparently from the graph, the MPC is less than APC. Keynes reasoned that as income falls relative to recent levels, people will protect consumption standards by not cutting consumption proportionally to the drop in income, and conversely as income rises, consumption will not rise proportionally (Branson, 1989). Based on this standard function, another alternative consumption function with additional calculus’s (adapted from Bui, 2002) could be: Ln (C) = c0 + c1Ln (Y) (1.2) The later one exposes clearer relationship between consumption and income. This result from the most distinguished features of log-log consumption model is that the slope coefficient c1 represents the elasticity of consumption (C) with respect to income (Y). It can be interpreted as the percentage change in consumption for a given percentage change in income, ceteris paribus. If c1 equal to one, any percentage change in income is fully absorbed by consumption regardless of long run or short run change, permanent income or transitory income change. Immediately after John Keynes introduced his consumption function, economists tried to test his conjectures. The earliest studies indicated that the Keynes’ consumption function is a good approximation of how household behave (Mankiw, 1997). However, incongruity arose when economists examine long time-series household data. Typically, Kuznets discovered that APC is fairly stable from decade to decade, despite large increase in income. Although the consumption function that Keynes proposed has some limitations, it is considered to be a milestone in the development of knowledge in this area. The function is still an excellent illustration of how household behave in the short run. Yet, due to its simplicity, it is not commonly used in consumption analysis. Fisher’s Intertemporal Choice Model Differently from current income based consumption model of Keynes, Irving Fisher developed the model in which intertemporal choices of forward-looking household are taken into account. This model illuminates the budget constraint that households face, the time preferences they have, and how these factors influence to their consumption. This paper focuses on the intertemporal budget constraint, which is the most relevant concept to scope of the study. For simplicity, Fisher explored decision making of a household who lives in two periods with his income constraint. He earns Y1 in the first period and Y2 in the second one. His consumption is C1 and C2 respectively in the two periods. In the first period, without any initial endowment, saving (S) equals income minus consumption. S = Y1 – C1 (1.3) In the second period, since no bequest left, consumption equals second period income (Y2) plus accumulated saving included interest earning. C2 = (1 + r)S + Y2 (1.4) where r is interest rate. Substitute the equation (1.3) into the equation (1.4) we get: C2 = (1 + r)(Y1 – C1) + Y2 (1.5) Divide both side of the equation (1.5) by (1 + r) and rearrange, we have C1 + C2 = Y1 + Y2 (1.6) (1+r) (1+r) This equation can be understood as the present value of consumption and income. It is the standard way of expressing the household’s intertemporal budget constraint. The implicitly in Fisher’s model is that a household can consume more or less than its income in a specific period but over its lifetime, it cannot consume more that its resources. The budget constraint is interpreted as the combination of first-period and second-period consumption that the household can choose. Figure 1.2 presents household’s budget constrained by a budget line. Any point on this budget line represents possible intertemporal choice of the household. At point A, the household consumes exactly what it earns in each period. If it chooses points between A and B, the household consumes less than its income in the first period and save the rest for the second period. In this case, the household shifts its present consumption to future consumption and vice versa if the household chooses a point between A and C. C2 B (1 + r)Y1 + Y2 Y2 A Y1 C C1 Y1 + Y2/ (1 + r) Figure 1.2 The household’s budget constraint Source: Mankiw, 1997 The simple two-period budget constraint can be expanded to many periods case with or without initial endowment and left bequest. That is: (1.7) in which (1+r)E is the present value of initial endowment including interest payment; t is number of periods (t = 1,2,…,n); Bt is the amount of bequest left at the end of period t (adapted from Bui, 2002). In conclusion, Irving Fisher developed intertemporal choice model showing how rational, forward-looking household make its choices. The central idea is that household consumption does not simply base on its current income but on the present value of current and future income. Relying on this fundamental model, Franco Modigliani built up his life-cycle hypothesis (LCH) whereas Milton Friedman introduced permanent income hypothesis (PIH) of consumption. These expanded models are discussed in the next sections. Modigliani’s Life-Cycle Hypothesis To explain the three conjectures in Keynes’s consumption function discussed earlier, Franco Modigliani postulated a life-cycle hypothesis of consumption. He emphasized that income varies systematically over people’s lifetime and that saving allows consumer to move income from those time in life when income is high to those time in life when it is low (Branson, 1989). As his argument, a typical individual has an income stream that is relatively low at the beginning and end of his life. This income stream is shown as Y curve in Figure 1.3 where Y, C, S, and T are expected income, consumption, saving and expected lifetime respectively. C Y T Y,C S Figure 1.3 The life-cycle hypothesis of consumption Source: Branson, 1989 This model suggests that in the early years of people’s life, he is a net borrower, the first shaded area in Figure 1.3. In the middle years, he saves to cover his debt and reserve income for smoothing consumption in retirement period. He dissaves in the later years in his lifetime, the second shaded area. Modigliani relied on two main assumptions. Those are: opportunity: smooth individual income until the retirement and after that his income turns to zero, and preference: consumption remains constant over individual lifetime and no bequest left to next generation. He differentiated his LCH on consumption by individual’s lifetime resources composition. He stated that consumption is not solely based on expected income but also on wealth. Consider a consumer who expects to live another T years has wealth of W, and expects to earn annual income Y until he retires R years from now. Thus, consumer’s lifetime resources include both initial wealth W and lifetime earning of R*Y. Assuming that consumer wishes to smooth his consumption, he will divide his lifetime resources of W + R*Y evenly among T remaining years and in each year he consumes: C = W + R*Y (1.8) T The function can be rearranged as: C = 1 W + R Y (1.9) T T This equation shows that an extra dollar increase in annual income raises consumption by (R/T) dollar per year, and an extra dollar increase in wealth raise consumption by (1/T) dollar. In other words, (R/T) is the MPC out of income and (1/T) is the MPC out of wealth. This consumption function can be depicted in Figure 1.4 below. C1 = f(W1, Y) C2 = f(W2, Y) C (1/T)W2 Y (1/T)W1 Figure 1.4 The Life-cycle consumption function The consumption function predicted by Modigliani is similar to the conventional consumption function introduced by Keynes. Nonetheless, the intercept term (1/T)W is not a fixed value. It, instead, depends on initial wealth. If the wealth increases, the consumption curve is shifted upward as presented in Figure 1.4. Modigliani also solved Keynes puzzle. Derive from equation (1.9), APC can be calculated by dividing both side this equation by Y and we get: C = 1 . W + R (1.10) Y T Y T The economist argued that because wealth does not vary proportionally with income in the short run, high income induces low APC. However over a long lifetime, wealth and income grow together, resulting in a constant ratio W/Y and thus a constant APC (Mankiw, 1997). In brief, Franco Modigliani successfully developed intertemporal choices in his LCH on consumption, which emphasized that income changes systematically over person’s life and that saving allows peoples to smooth their consumption. Friedman’s Permanent-Income Hypothesis Another consumption theory expanded from intertemporal approach is permanent income hypothesis, which is proposed by Milton Friedman in 1957. Unlike the LCH, which emphasized that income follows a regular pattern over a person’s lifetime, the PIH highlighted that people experience random and temporary changes in their income from year to year (Friedman, 1957). He suggested that income should be viewed as a sum of permanent income (YP) and transitory income (YT). Y = YP + YT (1.11) In which permanent income is defined as the average income that the household should expect over a long time horizon while transitory income is the random deviation from that average. Similarly, total consumption in any period is permanent consumption (CP) and random transitory consumption (CT), which represents positive, negative, or zero deviation from the usual permanent consumption level. C = CP + CT (1.12) As Branson realized, PIH rests with three main assumptions. There is no correlation between permanent and transitory income. There is no relationship between permanent and transitory consumption. There is no relation ship between transitory consumption and transitory income. Therefore, transitory consumption is not correlated with ether permanent consumption or transitory income. It means that changes in permanent or transitory income does not affect to transitory consumption. The transitory component has an expected value of zero (E[YT] = 0) reflecting the notion that over time transitory gains are offset by future transitory losses and vice-versa. Thus in the long run observed levels of income 'Y' are equal to permanent income 'YP'. Consequently, Friedman concluded that consumption depends primary on permanent income because household uses saving and borrowing to smooth consumption in response to transitory changes in income. That is C = a * YP (1.13) in which, a is a constant that measures the fraction of consumption out of permanent income. Applying intertemporal approach in the PIH, Friedman formulated consumption function with budget constraint in two periods as following: C1 + C2 = YP + YP (1.14) (1+r) (1+r) It is clear that the PIH must satisfy the equation: YP + YP = Y1 + Y2 (1.15) (1+r) (1+r) Solve this equation to find YP: (1.16) A graphical representation of permanent income consumption is provided in Figure 1.5. Because YP is the average expected income over the two periods, the intersection of 450 line and the budget constraint line is expected YP. Point A is an extreme case when consumption is exactly equal to permanent income in each period. In this special case of utility maximization, people tries to maintain fairly stable consumption path so that he consumes the same every period. Saving and borrowing is used in response to temporary changes in transitory income. Period 2 A C2 = YP Utility curve Y2 E 450 Period 1 C1 = YP Y1 Figure 1.5 Household consumption and permanent income Source: Sach and Larrain, 1993 In short, PIH on consumption postulated by Milton Friedman is an application of intertemporal approach. It suggested that income should be viewed as a combination of permanent income and transitory income. Accordingly, consumption should depend primarily on permanent income, and saving is considered to be responsive factor to temporary income changes. Hall’s Random-Walk Hypothesis Rested with Irving Fisher’s intertemporal choice model, Franco Modigliani concentrated on the structural relationship of expected income, wealth, and current consumption whereas Milton Friedman developed his PIH showing that current consumption essentially relies on permanent income. Developing this analogy, more recently Robert Hall has reformulated consumption theory by adding the assumption of rational expectation to the intertemporal choice approach. The rational expectation assumption states that people use all available information to make optimal forecasts about he future income (Branson, 1989). Moving directly to reduced-form forecasting equation, let us discuss Hall’s insight somewhat more carefully in terms of the intertemporal consumption model. (1.17) where t is the current point of time and t+1 is the next period. The equation gives the expected value of the next period consumption Ct+1, conditional on this period consumption, Ct, which is already known. When people receive news or information, his anticipation must be revised and expected consumption is changed consequently. Hence, the expected value for consumption is a total of permanent consumption and transitory consumption, which is the random term. If the present v._.alue of actual consumption Ct incorporates all the relevant information in an unbiased estimation of permanent income, then the expectations are rational, and the transitory component in the equation (1.17) will indeed be random over time (Branson, 1989). Thus, the above equation can be seen as a forecasting equation of anticipated consumption. Adding rational expectation assumption to the basic Fisher’s intertemporal model, Robert Hall gives us a reduced-form forecasting equation for consumption. It is apparently consistent with the basic relationship of the whole consumption theory: current consumption depends on the present value of the entire future income stream. This is the general starting point we have discussed. EMPIRICAL STUDY REVIEW Taking these theories of consumption as benchmark, various types of consumption function have been examined by economists. Typical researcher who relied on life-cycle hypothesis of consumption is Deaton (1997, 2002). He explored age profiles of consumption and saving in the C«te d’Ivoire, Thailand and Taiwan. He stated that consumption profile is very close to the income profiles. Implicitly, peoples smooth their consumption over the lifetime. More interestingly, he concluded “…while it is possible that the age profiles of consumption are simply those that people like, and just happen to match the shape of income profiles, it stretches belief that the coincidence should happen for every educational and occupational groups as well as for a wide range of countries…” Another direct test of permanent income hypothesis presented by De Juan, Seater and Wirjanto (2003) uses time-series data of 48 contiguous US states. The research sets out to implement a direct test of the implication of the PIH that the size of consumption revision due to an income innovation is equal to the size of permanent income revision due to the same income innovation. For each state, the authors jointly estimated the univariate income generating process and the equation describing the relation between consumption revision and income innovation. They then obtained an estimate of the magnitude of the revision in permanent income associated with an income innovation using the estimated parameters of the stochastic process generating income. The result implies that innovations in current income contain new information about the expected future path of income that leads consumers in each state to revise both their consumption and permanent income. This support for the PIH stands in contrast to the widely held views that credit constraints, myopia, or behavior toward risk. In some other literatures, economists attempt to examine the interrelationship of household characteristics and household consumption. This had been learnt in detail by Wondom (1999). The central issue of his research is finding the micro determinants of growth in per capita consumption over time in Bangladesh. Separated regressions were used for rural and urban area because, as author’s explanation, the returns to household characteristics may differ between these two areas. The independent variables are geographic location, household size, age and sex of the household head, marital status, education variables, and occupation variables. He found that households with female head have lower consumption and heads without spouse fare better than households with a spouse do. Moreover, a remarkable statement was a household with both the head and the spouse having completed secondary school has an expected per capita consumption almost double that of a similar household with illiterate head and spouse. In addition, return of education is larger in urban area, and non-farm household consume more than farm household in Bangladesh’s rural area does. In Vietnamese context, household expenditure is more greatly concerned than household consumption. “What determinants explain for differences in living standard of households” is the central research question of Vu T. M. (1999). The author firstly gave a descriptive analysis of per capita expenditure based on quintile classification of VLSS 1992-1993. Then he examined conventional household characteristics. He concluded that household whose head attain higher education level and those who work in state-owned sectors has higher per capita expenditure. The larger household size, the higher probability of being the poor. In addition, his essential finding that characterizes Vietnamese households is gender of head does not affect prosperity of household if others being equal. This finding gave delight to the author himself because in descriptive analysis, statistically, household head by female has higher probability of being rich than others are. The author also investigated the significance of the geographic effect. Households live in South East and Mekong delta seems spend more, partly because they are wealthier. These findings are valuable for ongoing studies. However, if the author had combined the data set VLSS 1998, it would have been an exhaustive research on both fixed effect and random effects on expenditure. Relying on recent data set VHLSS 2002, Le Thuc Duc et.al (2004) have evaluated impact of openness on the poor. A part of their study is identifying the poor based on expenditure necessary to secure an intake of 2100 calories per adult equivalent per day. They also figured out poverty rate and poverty gap for various domains in Vietnam. A logistic model was applied with emphasis on the variables like ethnicity, education, government and SOEs job, region, electrification, and landless. After all, they conclude that ethnicity explains most of the regional difference in poverty rate, especially in Northern Uplands. Return on education moderately differs as well. The government or SOEs job have greater antipoverty effect in the North than in the South. CHAPTER REMARKS This chapter provides a review of the outstanding theoretical and empirical studies on household consumption. The first section of this chapter presents a widely accepted definition of household as well as definition and measurement of consumption. The second part, the center of this chapter, illustrates the evolution of economic theories on consumption. Keynes initially argued that current income is the primary determinants of consumption. Despite some limitations in the Keynes’ consumption function, it still plays a vital role in the development of ideas in this area. Intertemporal choices model developed by Irving Fisher shows how rational, forward-looking individual make his choice. This model implies that consumption depends on the present value of current and future income. Based on intertemporal choice model, Franco Modigliani built on his LCH on consumption. The hypothesis emphasized that income varies systematically over people’s life and saving allows consumer to move income from those time in life when income is high to those time in life when it is low. Another expansion of intertemporal choice model is PIH postulated by Milton Friedman as a complement to LCH. Friedman claimed that people follow random and temporary changes in income from year to year. Current income of a person comprises permanent income and transitory income. Thus current consumption primarily based on permanent income rather than transitory income. Also rest with intertemporal approach, Robert Hall provided an estimation equation for expected consumption by adding the rational expectation assumption. He found that the expected value for consumption is a total of permanent consumption and transitory consumption. The final section is devoted to reviewing relevant key findings and conclusions of outstanding empirical studies on consumption of foreign economists as well as Vietnamese researchers. These findings appear to be consistent with theoretical framework on consumption. CHAPTER 3 AN OVERVIEW OF HOUSEHOLD CONSUMPTION IN VIETNAM In this chapter, we will analyze the relationship of household consumption and household characteristics as well as geographic features. First, the chapter aims to study characteristics of household in Vietnam. Then it will focus on examining household consumption by tool of descriptive and comparative analysis in order to highlight main patterns of household consumption in Vietnam. CHARACTERISTICS OF HOUSEHOLD IN VIETNAM During the past decade, Vietnam’s economy has been growing steadily at level of 7 percent (GSO, 2002). This achievement has contributed to remarkable increase in per capita income and living standard enhancement. As a result, characteristics of Vietnamese households have been changed gradually. Since the reunification in 1976, the percentage of females has been always higher than that of males. Fortunately, the gap has narrowed. The proportion of males and females were 48.78 and 51.22 percent respectively in 1990 and has reduced to 49.16 and 50.84 percent in 2002 (GSO, 2002). These figures were consistent with those calculated from VHLSS 2002. According to VHLSS 2002 survey, males accounted for 48.43 percent and female accounted for 50.57 percent. The differences in proportion of male and female may be explained by the differences in the lifetime of males and females, and a great numbers of man have died during the long-decades of war. Although the renovation process has brought about urbanization, the number of households living in the rural areas made up a large percentage. In 1998, the urban population accounted for 22.4 percent and the rural population accounted for 77.57 percent (GSO, 1998). These figures are consistent with those calculated from VHLSS 2002. The most densely populated areas are the Mekong Rive Delta and Red Delta, which made up more than 35 percent of the whole population VHLSS 2002. In the recent year, thanks to the propaganda of family planning and the use of model technology in birth control, the population growth has fallen. The average household size also changes from region to region with the greatest size of 5.2 people in rural South and the smallest size of 4.0 people in urban North. On average, Vietnamese household size was 4.49. One interesting finding here is that the number of children dropped gradually along with income quintiles, which means that the more chance the household can earn money the less baby they have. The area with the highest number of children in a household was the Central Highlands (4.58) and the area with the lowest number was the Red River Delta (3.08). Household heads play important roles in household management. In Vietnam, due to the history and agricultural culture most of the household heads were men. The VHLSS 2002 shows that male household heads made up 76.54 percent and the remaining of 23.4 percent were women. The data also reveals that the higher the education of household heads the higher the income of household, Unfortunately, most household heads have low level of education (81.02 percent have not finished general education of which 7.99 percent never attending school). HOUSEHOLD CONSUMPTION IN VIETNAM IN 2002 In this part, the thesis will analyze household consumption by factors including income, household size, number of children, region, age, gender, education and occupation. Household consumption defined as all expenditure on non-durable goods and adjusted to price differences among regions and to inflation. In analysis, household consumption is calculated from VHLSS 2002 data set and grouped in 5 quintiles that represented for 5 groups of household in different economic situation. Household consumption is measured in thousand VND per year. Household consumption by household income As throughout analyzed in theoretical chapter, household consumption always depends on household income. In other words, household consumption is positively related to household income. Once again, it is true when examining household consumption in Vietnam. Statistical results from VHLSS 2002 data set are shown in table 3.1. In the first quintile, household consumption accounted for 4,975 thousand VND per year where as household income was 6,671 thousand VND per year or household consumption took a share of 0.74 of household income. Household consumption increased as household income rise across quintiles. Interestingly, the share of household consumption on household income did not increase as income rise. This share fallen from 0.746 in the first quintile to 0.75 in the forth quintile. It is slightly higher (0.737) in the fifth quintile. This finding is consistent with Keynesian consumption model, which stated that as income rise, household would save a greater and greater fraction of their income, it would, in turn, reduce fraction of their consumption. To examine magnitude of consumption fraction to income in the whole data set, quantitative model in the next chapter will bring a full-sided analysis. Table 3.1: Household consumption and HH income by quintiles (’000 VND/year) Consumption quintile Consumption Income Consumption/ Income 1st quintile 4,975.41 6,671.02 0.746 2nd quintile 8,246.56 11,215.19 0.753 3rd quintile 10,921.68 14,995.82 0.728 4th quintile 14,895.84 20,823.29 0.715 5th quintile 28,558.61 38,742.77 0.737 Source: Author’s calculation from VHLSS 2002 Household consumption by region and quintiles To examine the pattern of household consumption in Vietnam across region, some descriptive analysis was done basing on VHLSS 2002 data set with 22,462 observations. The result shows that consumption level of household living in the South was highest (16,764.90 mill VND per year), that of household living in the North was lowest (11,800.80 mil VND per year) and consumption of household living in the Central was in the middle (11,968.5 mill VND per year). This thesis also applies cluster methodology by using quintile of consumption level for each region. The obtained results partly answer to the question about geographic impact on consumption level. Table 3.2 reveals differences in consumption level across regions, the North, the Central, and the South of Vietnam. The lowest and the highest consumption quintile show clearly this difference. In the first quintile, consumption level of household in the North accounted for about 4,984 thousand VND per year, that of household located in the South reached 5,113 thousand VND per year, whereas consumption level of household in the Central was only 4,878 thousand VND per year. There is a similar picture for consumption level of the highest quintile. Household consumption level in the North, the Central, and the South respectively was 26,965 thousand VND per year, 26,613 thousand VND per year and 30,151 thousand VND per year. Table 3.2: Household consumption by regions and quintiles (’000 VND/year) Consumption quintile North Central South 1st quintile 4,984.385 4,877.992 5,113.045 2nd quintile 8,238.136 8,217.052 8,293.896 3rd quintile 10,893.83 10,914.62 10,963.39 4th quintile 14,832.50 14,832.92 14,995.8 5th quintile 26,964.92 26,613.46 30,151.03 Source: Author’s calculation based on VHLSS 2002 data set In overall, five consumption quintiles display a apparent distance of consumption level of household living in the South with that of household living in two other regions. It can be explained by following reasons: In the North, especially in the Northeast and the Northwest regions, a large number of household is living in poor infrastructure such as having bad accommodation, using unpurified water, lighting by petrol in stead of electricity. Thus consumption level of household in these regions is certainly lower than that of household in region with good infrastructure like in the South where gather large industrial zones and processing zone. Household in the Central also consume less than household in the South because in the Central, household often faces with natural disaster (flood, storm…). These events force Central household have to make forward-looking decision. Maintaining a certain saving by consume less in a given budget constrain is the best way. This is accompanied by PIH hypothesis shown in the previous chapter. People in the South have a more generous perception and habit in consumption partly because services sector in this region is more active to satisfy wants and needs of household. Moreover, life in the South is apparently more modern than in other regions resulting from historical inheritance. Thus, it is reasonable to expect that household living in the South has higher level in both autonomous consumption and MPC than household living in the North and in the Central. However, the comparison of consumption level of household in the North and in the Central does not show clear relation. This will be further studied in the next chapter. Household consumption between urban and rural area In many developing countries, income, expenditure and consumption of household located in urban area is often higher than that of household in rural area. Vietnam is not exclusion. According to VHLSS 2002, consumption level of household located in urban area (20,661 thousand VND per year) is nearly double that of rural household (11,348 thousand VND per year). It can be explained by following reasons: in rural area, making businesses and finding jobs is more difficult due to unfavorable location, lower quality of human capital (e.g. knowledge, skills and chance to access to information…) As a result, household income could be lower. This is similar to findings of Vu P.H.D. (2000). most of rural household’s income comes from agriculture, aquaculture, and sylviculture which faces with vulnerable conditions and general living standard in rural area is lower than that in urban area. This argument is further supported as shown in table 3.3 where difference in consumption is more substantial across consumption quintiles. Table 3.3: Household consumption by rural/urban and consumption quintiles (’000 VND/year) Consumption quintile Urban Rural Discrepancy 1st quintile 4,988.045 4,974.353 14 2nd quintile 8,305.059 8,239.13 66 3rd quintile 11,058.68 10,893.74 165 4th quintile 15,167.5 14,789.15 378 5th quintile 31,677.41 25,119.95 6558 Source: Author’s calculation from VHLSS 2002 In overall, figures in above table are consistent with the idea of household in urban area consume more than that in rural area. It can be also observed that the discrepancy of household consumption level in each quintile increases gradually. Surprisingly, the difference in the 5th quintile reaches 6,558 thousand VND per year. Base on these stylized facts, it can be expected that household’s consumption level in the urban area is greater than that of household in rural area. Linear model regression is used in testing this hypothesis in chapter 4. Household Consumption by household size and number of children A fact that cannot be denied is increase in household member induces household consumption expansion. Event the member is adult or children the household must use a larger share of income to satisfy, at least, his necessity need. Appendix 4.2 strongly supports to this idea. If a family consisted of only one member, its consumption level was 4,481 thousand VND per year. Household consumption gradually in creased long with household size. It reached a peak of 38,034 thousand VND per year for households of 14 members. Household included 15,17 and 18 persons occurred only one time for each in the data set. Thus theirs consumption level consider to be unidentical result. In brief, higher household consumption level is always associated with larger household size. However, to that extent does the household size affect to consumption level, it need to be further analyzed in the next chapter where quantitative model is used. There is similar picture when examining relationship between number of children and household consumption. The higher number of children, the larger household consumption level. Evidence illustrated in Appendix 4.3 where household consumption in the families without children was about 12,767 thousand VND per year, with one child it accounted for 14,593 thousand VND per year and so on. Especially the household with 9 children, consumption level was 22,829 thousand VND per year. This number was much higher than the previous consumption level of household with 8 children (15,075 thousand VND per year). All above figures resulted from VHLSS 2002 and the term number of children is measured in counting number of children under fifteen year old. Because below 15 year old, children cannot self-rely. At 15, all most children pass secondary school and they can work to feed themselves since then. Household consumption by household head’s education According to many previous studies, education impact not only directly on consumption level but also indirectly through income level. Appendix 3.1 shows relationship between education of household head and household consumption. Once again, result from VHLSS 2002 confirms positive relationship between education and consumption. As higher educated household head, household consumption level will be greater. Here, education is measured in number of schooling year of household head. There are some clear breakdowns of consumption level at number of schooling year of 9, 12, 15, 16, 18, and 21. It’s can be understood that there is a significant change in consumption level if household head educated secondary school, upper secondary school, college, university, master degree, and doctor degree. For those who have never attended school (represented by number of schooling year is 0), his household consumption level was about 9,931 thousand VND per year. The consumption level increased gradually as the household head gets higher education. If the household head passed secondary school, his family consumption level reached nearly 13,000 thousand VND per year. However, there was a remarkable rise in consumption level of household whose head participates in upper secondary school. The rise was approximately 3,000 thousand VND per year. Changes in consumption level also occurred in circumstance of household head getting higher education. If household head has studied college, consumption level was 19,259 thousand VND per year. For graduated university head, his household consumption level accounted for 27,358 thousand VND per year. Especially, consumption level pushed up to 46,607 thousand VND per year, if household head has received master degree and reached a pick of 63,591 thousand VND per year for household which head getting doctor degree. Examining household head’s number of schooling year by consumption quintiles further supports the positive relationship between education and consumption. Table 3.4 reflects high consumption level is always associated with higher education. Table 3.4: Household head’s number of schooling year by consumption quintiles Consumption quintile Number of schooling year 1st quintile 5.15424 2nd quintile 6.493989 3rd quintile 6.827732 4th quintile 7.33363 5th quintile 8.557111 Source: Author’s calculation from VHLSS 2002 Explanation for the positive relationship is that high-educated household head often have good planning for the future and over-handed attitude about consumption. That is direct impact of education on consumption. Moreover, the indirect impact through income is that: income is higher if head better educated, as income rise, want and need further expanded, necessity goods reduced and luxury goods increased in the household consumption bundle of goods of household owned higher income. Household consumption and head’s age and gender. Basing on VHLSS 2002 data set, household consumption was analyzed by using cluster method. Table 3.5 presents average age of household head by consumption quintiles. Table 3.5: Age of household head by consumption quintiles Consumption quintile Age of household head 1st quintile 49.41776 2nd quintile 44.97841 3rd quintile 46.08079 4th quintile 47.61028 5th quintile 49.7328 Source: Author’s calculation from VHLSS 2002 It’s easily to recognize that (i) in all consumption quintiles, household heads were above 40 years old and (ii) there was no clear linear relationship between household consumption and household head’s age. In the lowest and the highest consumption quintiles, age of household head are both more than 49. The age was lower in other consumption quintiles and the lowest age of household head accounted for nearly 45 in the second consumption quintiles. In most studies about relationship between household expenditure, heath care expenditure or household saving and household head’s age, there was a linear correlation between those and household head’s age. However in analyzing household consumption, with above figures, it seems to be irrelevant. This finding further strengthens lifecycle hypothesis of consumption stated in the previous chapter. Therefore, using “square function” of household head’s age proved to be more suitable in investigating impact of household head’s age on household consumption. In general, statistical result from VHLSS 2002 shows that household, which headed by female consumes more than those headed by male. Average consumption level of household headed by female accounted for 13,873 thousand VND per year whereas that of household headed by male was only 13,410 thousand VND per year. However if household consumption level is analyzed in quintiles, the nature with not stay the same. Table 3.6: Household consumption by head’s gender and consumption quintiles (’000 VND/year) Consumption quintile Female head Male head 1st quintile 4,345.217 5,292.823 2nd quintile 8,217.488 8,253.204 3rd quintile 10,902.85 10,926.01 4th quintile 14,930.97 14,887.23 5th quintile 31,221.25 27,599.83 Source: Author’s calculation from VHLSS 2002 The above table reveals that in the three first quintiles, consumption level of household headed by male was slightly higher than that of household headed by female. Nonetheless in the two remained quintiles, surprisingly, household headed by female seemed to consume more than those headed by male. So, there is no apparent relationship between household head’s gender and household consumption level. It’s consistent with many empirical studies in Vietnam especially that of Le (2004) and Vu (1999). Household consumption by occupation of household head Occupation in this thesis is narrowed in state sector (including public sector) and other sector. Reason for this classifying is to examine the difference in household consumption between household with rather stable income and other. Relying on VHLSS 2002 data set, if household head works in state sector, the household consumption level was 20,284 thousand VND per year while consumption level of household which head works in other sector was only 12,836 thousand VND per year. The dissimilar in consumption level can be observed more clearly by analyzing consumption by quintiles as shown in below table. Table 3.7: Household consumption by occupation of household head and consumption quintiles (’000 VND/year) Consumption quintile State sector jobs Others Discrepancy 1st quintile 4,964.565 5,558.647 594 2nd quintile 8,239.183 8,412.598 173 3rd quintile 10,913.27 11,024.57 111 4th quintile 14,864.06 15,125.15 261 5th quintile 27,950.92 31,027.71 3077 Source: Author’s calculation from VHLSS 2002 In table 3.5, household consumption varies across occupation classification discrepancy is quite large, especially in the 5th quintile. This fact raises a question of whether household which head works in state sector always associated with higher household consumption. This question will be answered in the next chapter where quantitative model will be used. This model will examine where occupation variable is significant in determining household consumption CHAPTER REMARKS The descriptive analysis on pattern of household consumption based on VHLSS 2002 shown that household consumption is affected by demographic characteristics (gender, age, household size, number of children), by geographic characteristics, and by other factors such as household income, education, and occupation. However, the relationship between household head’s age and household consumption is not clear. It is also not apparent when examining impact of household head’s gender and household consumption. Some important conclusions in this chapter are found. First, there is a difference in consumption level of household located in the North, the Central and the South of which family living in the South consume more than others. Second, consumption level of household in urban area is nearly double that of household in rural area. Third, higher consumption level is always associated with larger household size and increase in number of children. Forth, the higher educated household head, the greater household consumption level. Fifth, household which head works in state sector is always get along with higher household consumption. These conclusions imply that household consumption depends on a lot of factors. All above conclusions, however, are only based on descriptive analysis. Therefore, in order to confirm these findings and point out what main factors determine household consumption, a quantitative analysis of determinants of household consumption will be carried out in the next chapter. The combination of two methods will provide useful recommendations in the final chapter. CHAPTER 4 MODEL SPECIFICATION AND EMPIRICAL RESULTS So far, we have highlighted the outstanding theoretical and empirical studies on the household consumption. Theoretical frameworks and empirical results introduced in Chapter 1 postulates the hypothesis of intertemporal choice o._.