Chulalongkorn Journal T. of Chomtohsuwan Economics 18(1), : Differences April 2006: in 63-82 Household Demand 63 Differences in Household Demand for Water Supply in Thailand and Tax Policy Implication Thoedsak Chomtohsuwan* Faculty of Economics Chulalongkorn University Abstract This paper attempts to estimate the household demand for water supply 1 in each income class in Thailand, and investigate the tax policy that is effective to decrease both demand for water supply and differences in ability to pay. Linear Expenditure System (LES) is utilized to estimate the system of demand equations. Data used in this analysis come from two main sources, i.e., the household expenditure data from the 2002 Household Socio-Economic Survey (SES), and the regional commodity price data from the Department of Business Economics under the Ministry of Commerce. Households are classified into four income classes, i.e., poor, almost poor, moderate, and rich. Commodities are divided into three main groups, i.e., water supply, food, and non-food. The results assure us that, water is a necessary good. Moreover, they demonstrate that the price elasticities of demand for water supply in each household income class are significantly different. High-income households price elasticities of demand for water * Graduate Student, Author for correspondence- E-mail address: Thoedsak@msn.com 1 Water supply, in this study, means the water supplied by the Waterworks Authority.
@ 64 Chulalongkorn Journal of Economics 18(1), April 2006 supply are less than those of low-income households. The present pricing and tax policy for water supply in Thailand is not effective to control the demand for water supply adequately. Furthermore, it also causes an increase in inequity between different households. The Double Non-Uniform Tax (DNUT) policy, which is developed from the optimal taxation in a manyperson economy, however is effective in reducing the demand, improving vertical inequity, and also increasing government tax revenue, simultaneously. JEL classification: C81, H24, Q25 Keywords: Thai household demand estimation, Inequality, Income distribution, Double Non-Uniform Tax 1. Introduction It is a well-known problem that demands for water resources are increasing continuously all over the world because of the increase in the world population and the activities consuming water, while water resources remain almost constant. The authorities of many countries have attempted to increase the production efficiency, while at the same time many public policy planners have also attempted to use public policies to decrease demand by activating price mechanisms. Many research studies attempt to estimate the price elasticity of demand for water in various countries 2. However there is no explicit study which investigated the price elasticity of household demand for water in Thailand especially in each different household income group yet. In the view of tax policy, even though Ramsey (1927) inferred that the government can raise revenue by increasing taxes on goods which have a low price elasticity of demand 3, the inequity between the poor and the rich in ability to access public subsistence resources has not 2 Fernando Arbue@s Gracia, MarIa A@ngeles GarcI@a Valin)as and Roberto MartInez- @ Espin)eira (2001). 3 Known as Ramsey tax. Afterwards, base on Ramsey (1927), the inference about optimal taxation in a many-person economy, which applies different tax rates to different goods, was developed by Diamond and Mirrless (1971).
T. Chomtohsuwan : Differences in Household Demand 65 been of concern yet. Therefore, in the case of subsistence goods, Ramsey tax and optimal taxation in a many-person economy theory are not enough. The reasons mentioned above inspired the author to tackle the differences in price elasticity of demand for water supply in each income class and study policies that are able to reduce water demands and also improve the inequality in ability to pay between different households at the same time. The next section is theoretical framework. The third section in this paper is the data sets analysis. The data sets used in this analysis come from two main sources, i.e., the household expenditure data from the 2002 Household Socio-Economics Survey (SES), and the regional commodity price data from the Department of Business Economics under the Ministry of Commerce. The fourth section presents the estimation of household demand for water supply, food, and non-food in each income class by using Linear Expenditure System (LES) to estimate the system of demand equations. The fifth section discusses the tax policy implications. The last section offers concluding remarks. 2. Theoretical Framework Linear Expenditure System (LES), introduced by Stone R. (1954), is utilized in this study to estimate the parameters of the consumption demand for water supply function. The two important economic properties of LES are additivity and homogeneity. However, LES does not allow inferior goods in the system and separates saving from consumption. The LES model assumes that the utilities which consumers get from each good consumed can be combined, and all consumers want to maximize their own utility subject to a budget constraint. This is shown in equation [1] and [2], where U is the utility function, β i is the marginal expenditure on commodity i out of total expenditure, with β i > 0 and the sum of β i equals one, Q i is the total consumption level on commodity i, γ i is the consumer s committed consumption level of commodity i, with 0 < γ i < Q i, E is the consumer s total expenditure, and P i is the price of commodity i.
66 Chulalongkorn Journal of Economics 18(1), April 2006 Maximize...[1] Subject to...[2] The system of demand functions is found by solving the maximization problem, as shown in equations [3], [4], and [5]....[3]...[4]...[5] Furthermore, the parameters in the LES model can also be used to derive other important figures such as the expenditure elasticity, the uncompensated (Marshallian) price elasticities, and the compensated (Hicksian) price elasticities by using equations [6] to [12] 4, where I i is the expenditure demand elasticity, w i is the expenditure share for commodity i, Φ is the supernumerary ratio, ε ii is the Marshallian own price elasticity of demand, ε ij is the Marshallian cross price elasticity of demand, η ii is the Hicksian own price elasticity of demand, η ij is the Hicksian cross price elasticity of demand....[6]...[7]...[8] 4 These were introduced by Lluch and William (1975).
T. Chomtohsuwan : Differences in Household Demand 67...[9]...[10]...[11]...[12] Unfortunately, because there are differences in the unit of each good and differences in the unit price in each region, the relative price index is used to calculate the representative price for each commodity group in the five regions, as shown in equation [13], where P ir is the relative price index of good i in region r, P ir is the price of good i in region r, P io is the price of good i in the reference region, o, which is Bangkok in this study, ω ir is the expenditure weight share of good i in region r....[13] In policy simulations, the present price policies, which consist of increasing block rate and VAT, is used to be based policy, The uniform price increase (UPI) policy, which increases the water supply price before tax, and the uniform tax increase (UTI) policy, which increases the VAT for water supply, are used to investigate the effect on household demand for water supply and inequality. Besides the UPI and UTI policy, this study introduces the Double Non-Uniform Tax (DNUT) whichis developed from the optimal taxation in a many-person economy of Diamond and Mirrless (1971)....[14] DNUT used a difference in price elasticities in each household income class to define the different rate in each class. DNUT taxes different goods at a different rate and taxes some goods, especially
68 Chulalongkorn Journal of Economics 18(1), April 2006 subsistence goods, at a different rate for each household, as shown in h equation [14], where R τ is the partial government tax revenue, τ g is the VAT rate for goods g consumed by household h, P g is price before h tax of goods g and Q g is the consumpton of goods g of household h. 3. Data Sets 3.1 Household Expenditure Data The household expenditure data used in this study come from the 2002 Household Socio-Economics Survey (SES), which are secondary cross-section data. The advantages of SES are that the range of observations covers the whole country. There is an enormous amount of observations, including micro data about each household such as household characteristics and expenditures. This dataset has 34,785 household observations in five regions of Thailand, i.e., Bangkok, Central and East, North, Northeast, and South. There are 23,155 household observations, i.e. around 67% of households, with access to the water supply and 11,630 household observations, i.e. around 33% of households, without access to the water supplied by the Waterworks Authority. From the results of the data analysis, as shown in Table 1, it can be can seen that the proportion of households in urban areas that can access water supply is around 73.70% of the whole country. The highest proportion around 90.49%, is in Bangkok, The lowest proportion around 62.30%, is in the South. In the other areas, the proportions are almost the same at around 70%. The remaining households use water from other sources such as underground water, ground water, etc. The proportion of households in rural areas that cannot access water supply is around 45.18% of the whole country. The rates are almost the same at around 40% in the Central and East, the North, and the Northeast. The highest proportion of households that cannot access water supply, around 65.16%, is in rural areas of the South.
T. Chomtohsuwan : Differences in Household Demand 69 Table 1 Number of household observations which have access to water per region Number of household observations Region Urban area Rural area Can Cannot Total Can Cannot Total access access access access Bangkok 1,761 185 1,946 0 0 0 (90.49%) (9.51%) (0.00%) (0.00%) Central and East 4,706 1,130 6,036 2,547 1,784 4,331 (77.97%) (22.03%) (58.81%) (41.19%) North 3,283 1,374 4,657 1,929 1,385 3,314 (70.50%) (29.50%) (58.21%) (41.79%) Northeast 4,278 1,653 5,931 1,927 1,185 3,112 (72.13%) (27.87%) (61.92%) (38.18%) South 1,866 1,129 2,995 858 1,605 2,463 (62.30%) (37.70%) (34.84%) (65.16%) Total 15,894 5,671 21,565 7,261 5,959 13,220 (73.70%) (26.30%) (54.92%) (45.18%) Source: The 2002 Household Socio-Economics Survey, Arranged and calculated by author Base on the SES dataset, the 2002 Thai income distribution is shown in Figure 1. It can be noted that the average per capita income is around 4,446 bath/month. More than 66% of the Thai population have a monthly current income below the minimum monthly wage rate of around 4,000 baht/month, and around 9.8% earn less than the Thai poverty line of 922 baht/month. Just 5.9% earn more than 12,500 baht/month. In this study, households were classified
70 Chulalongkorn Journal of Economics 18(1), April 2006 into four classes of income 5, i.e. Very Poor (VP), Poor (P), Moderate (M), and Rich (R), by using criteria which reflect the household ability to pay and the quality of life, i.e. poverty line, minimum wage rate, and minimum income on which personal income tax has to be paid, as shown in Table 2. All goods in the SES dataset, i.e., 614 goods, were classified into three groups, namely water supply (ws), food (f), and non-food (nf), as shown in Table 3. Source: The 2002 Household Socio-Economics Survey, Arranged and calculated by author Figure 1 The 2002 Thai income distribution 5 In the estimation, households were classified by the amount of money that they used per month instead of current monthly income that they received because of two reasons. Firstly, expenditure can reflect wealth better than income. Secondly, one of the weaknesses of the SES data is that some people usually declare their income lower than their real income to avoid government tax. However, the disadvantage of this method is that personal expenditures of each household member were assumed equal to the average household expenditure.
T. Chomtohsuwan : Differences in Household Demand 71 Table 2 Household income classification Household Monthly % Number of Classifying Criteria income class current income observations (Baht/Month) Poor (P) 0 to 922 9.8 841 2002 Thai poverty line Almost Poor (AP) 923 to 4,000 56.9 15,231 2002 Thai minimum wage rate Moderate (M) 4,001 to 12,500 27.5 6,440 2002 Thai minimum income on which is payable personal income tax Rich (R) More than 12,500 5.9 643 The rest Source: The 2002 Household Socio-Economics Survey, Arranged and calculated by author Table 3 Commodity group classification Commodity group Number of Kinds of goods goods Water supply 1 Water supply Food 454 Food Other kinds of water except water supply Non-Food 159 Clothes Rent Energy Appliances Medicine Travel Communication Entertainment Education Beer, Wine, Cigarettes Gambling Service, etc. Source: SES Data dictionary, Arranged and calculated by author
72 Chulalongkorn Journal of Economics 18(1), April 2006 3.2 Commodity Price Data One of the weaknesses of the SES dataset is the lack of price information in the household consumption data. Therefore this study also uses data from the Department of Business Economics under the Ministry of Commerce and combines the price data with the SES data. As there are differences in the unit of each good and differences in the unit price in each region, the relative price index is used to calculate the representative price for each commodity group in the five regions. Table 4 Relative price index in each region Commodity Relative price index group Bangkok Central North Northeast South and East Water supply 1 0.942946058 0.855809129 1.06224066 0.967842324 Food 1 0.923514687 0.892714354 0.97539007 0.972724425 Non-food 1 0.912741240 0.873995302 0.91031402 0.943064740 Source: The Department of Business Economics under Ministry of the Commerce, Arranged and calculated by author The results in Table 4 show that the relative price index of water supply in the Northeast is the highest, followed by Bangkok, the South, the Central and the East, and the North, respectively. The figures in Table 4 are used to calculate the other relative price indices in each household income class, as shown in Table 5. As the results in Table 5 show, the poor face the highest relative price of water supply because most of the poor live in the Northeast. The second class which faces a high relative price of water supply comprises the rich because most of the rich live in Bangkok.
T. Chomtohsuwan : Differences in Household Demand 73 Table 5 Relative price index in each household income class Household Commodity Relative price Average Expenditure income class group index Each group Total Whole country Water supply 0.9625538 44.2166 Food 0.9420717 1246.4776 4464.9504 Non-food 0.9135491 3174.2562 Poor Water supply 0.9732865 9.4910 Food 0.9420299 400.5671 767.9290 Non-food 0.9005611 357.8709 Almost Poor Water supply 0.9607389 30.3944 Food 0.9391624 908.6017 2319.6811 Non-food 0.9084552 1380.6849 Moderate Water supply 0.9644005 65.6106 Food 0.9463350 1737.3900 6486.9703 Non-food 0.9223064 4683.9697 Rich Water supply 0.9663577 95.3529 Food 0.9499494 2716.1027 20315.0045 Non-food 0.9264807 17503.5489 Source: The Department of Business Economics under Ministry of the Commerce, Arranged and calculated by author
74 Chulalongkorn Journal of Economics 18(1), April 2006 4. Estimation Results After using LES to estimate parameters in the system of household demand equation, the results are as follow. According to the results in Table 6, the marginal expenditure (β) of the low-income classes on water supply and food is more than that of the high-income classes. On the other hand the marginal expenditure of the high-income class on non-food is more that than of the low-income class. The committed consumption (γ) of the low-income class on every commodity group is less than that of the high-income class. Table 6 Estimated parameters of demand equation systems of each household income class Household Commodity Marginal expenditure Committed consumption level income class group (β) (γ) Coeff. t - Stat Prob. Coeff. t - Stat Prob. Whole country Water supply 0.0039 78.42 0.00 37.5275 257.89 0.00 Food 0.1013 122.79 0.00 1119.5087 297.12 0.00 Non-food 0.8948 1065.27 0.00 1450.4634 87.74 0.00 Poor Water supply 0.0106 4.96 0.00 2.9588 4.12 0.00 Food 0.4639 21.08 0.00 149.7283 4.53 0.00 Non-food 0.5255 111.9597 3.46 0.00 Almost Poor Water supply 0.0129 54.23 0.00 22.4289 159.02 0.00 Food 0.2792 106.45 0.00 764.0077 368.82 0.00 Non-food 0.7080 267.44 0.00 982.2689 187.64 0.00 Moderate Water supply 0.0048 19.38 0.00 56.5323 152.89 0.00 Food 0.1182 36.57 0.00 1534.5302 257.43 0.00 Non-food 0.8770 267.76 0.00 2763.6364 96.23 0.00 Rich Water supply 0.0010 5.05 0.00 92.6053 76.57 0.00 Food 0.0270 7.57 0.00 2682.1326 114.74 0.00 Non-food 0.9720 268.34 0.00 12282.5353 53.69 0.00 Source: Estimated by author
T. Chomtohsuwan : Differences in Household Demand 75 The results in Table 7 show that the expenditure elasticities of water supply and food are less than one, which imply that these are necessary goods. Whereas expenditure elasticities of non-food are more than one implying that the latter are inferior goods. There are significant differences in the price elasticity of demand for goods in each household income class. Overall, the own price elasticity of demand for water supply is -0.1844, which is close to the figure estimated in studies of other countries 6. In the partial analysis, for each income class, they are between -0.0627 and -0.5779. The results assure us and reveal the fact that the price elasticity of demand for water supply of the low-income class is higher than that of the high-income class. The own price elasticities for food and non-food range from -0.2 to -0.8 and from -0.8 to -1, respectively 7. Therefore, these results highlight the importance of separating the dataset into many income classes. In this study, the forecasted demand without income classification exceeds the forecasted demand using four income classes by around 5.24%. This infers that policy studies that are concerned about differences in each individual household or vertical equity would get more reasonable outcomes if those policies were simulated using parameters that were estimated by the basis of classified datasets. 6 The average figure of other studies ranges from -0.067 to -0.96, for example Fernando Arbue@s Gracia, MarI@a A@ngeles GarcI@a Valin)as and Roberto MartI@nez- Espin)eira (2001). 7 In the study of Sarntisart (1995), goods in the food group vary from -0.0802 to -1.0261 and goods in the non-food group vary from -0.2161 to -1.1010.
76 Chulalongkorn Journal of Economics 18(1), April 2006 Table 7 Estimated Price Elasticity of demand of whole country and each income class Source: Calculated by author The errors in the estimations may result from three main causes. The first is the six limitations of the SES data collecting process 8, i.e., observations that were not taken from the entire population, non-sampling errors that were caused by many sources, the errors made in recording data, the mistakes made in coding and estimating missing data, effort that was made to minimize errors, and the under-reporting of income and expenditure data by the sample household. The second ferers to the effects of high aggregation of goods in some goods group and the large difference of price or type 8 The limitations of the SES data are well explained in the Report of the 2002 Households Socio-Economic Survey by the National Statistical Office (2002).
T. Chomtohsuwan : Differences in Household Demand 77 of each good in the group. The third cause is the limitations of the LES theory that express the demand and the utility function in the Linear Geary-Stone utility function form. 5. Tax Policy Implication As results in the study, the price elasticity of demand for water supply of the low-income class is higher than that of the highincome class. This means that an increase in price or tax will affect the demand for water supply of low income households more than that of high income households. Besides, water is a necessary good, it is a subsistence good. Therefore, the policy for water should be treated differently to acknowledge the difference in households ability to access water resources. The policy planner should realize the vertical inequity and human rights issue because there are significant differences in the forecasted results between using only one class and using many income classes, governments or policy planners should use demand equations for many household classes to accurately forecast the effect of a new policy. The two ways to decrease the demand for water supply and reduce the inequity are price and tax change. From the results, as shown in Table 8, price change or tax change have the same effect on the demand of consumers. However, the most important issue is that if the government changes the water supply price, much of the change in revenue will be transferred from the wealth loss of consumers to private groups who own the waterworks service. If the government changes the water supply tax, much of the change in revenue will accrue to government and could be returned to consumers in another form i.e. a public good or service. Consequently, tax change has greater advantages regarding both, controlling demand and improving the welfare of consumers, which may be decreased due to the increase in water expenditure.
78 Chulalongkorn Journal of Economics 18(1), April 2006 The simulation results shown in Table 8 show that the present policies, which consist of increasing block rate and 7%VAT, have caused an inequality of 0.1792. Households in the high-income class have a higher ability to consume water supply than the lowincome households. Although both, the uniform price increase (UPI) policy, which increases the water supply price before tax by 5%, and the uniform tax increase (UTI) policy, which increases the VAT for water supply from 7% to 12.5%, are able to decrease the total demand for water supply by the same rate of 1.07%, inequality is increased from 0.1792 to 0.1818. DNUT is not only able to decrease the total demand for water supply at the same rate of 1.07%, but also decrease the inequality from 0.1818 to 0.1785 which is also less than the 0.1792 of the present policy. From these results, it can be deduced that DNUT is better. Table 8 The effect of policies on each household income class Unit : Cubic meter per month Household Average demand for water supply per person income class Present Policy UPI Policy UTI Policy DNUT Policy Poor 8.6567 8.3838 8.3838 8.7658 (-3.1525 %) (-3.1525 %) (+1.2603 %) Almost Poor 31.6703 31.2164 31.2164 31.4103 (-1.4332 %) (-1.4332 %) (-0.8210 %) Moderate 68.6336 68.0444 68.0444 67.6130 (-0.8585 %) (-0.8585 %) (-1.4870 %) Rich 99.1227 98.8079 98.8079 98.3100 (-0.3176 %) (-0.3176 %) (-0.8199 %) Theil Index 0.1792 0.1818 0.1818 0.1785 (+1.4509 %) (+1.4509 %) (-0.3906 %) Source: Study results, Simulated by author
T. Chomtohsuwan : Differences in Household Demand 79 There are six issues that should be tackled in the DNUT policy implementation. The first is the imperfect household information problem such as inaccuracy and non up-to-date household income information. How will we know the accurate household data? The second is the definition of the ability to pay. Should it include other fix assets, such as land and bequest, besides the current income? The third is the criteria for the household classification. What are the proper number of household classes and the proper range of each income class? The fourth is the accuracy of the price elasticity of demand for water supply in each household income dass. How much should we trust in the results? The fifth is the households which have to pay a higher rate. How to pursuade them? And the sixth is firms or entities, which also consume water supply. What is the class in which firms or entities should be? Is it better to separate firms or entities from the pricing system? The policy planner or government in each country or area should consider these issues when deciding on suitable choices. 6. Concluding Remarks The results of this study show that, for the whole country analysis, the own price elasticities of demand for water supply, food, and non-food commodities are -0.18, -0.25, and -0.95, respectively. The household level analysis further shows that the own price elasticities of demand for water supply in each household income class are significantly different. The price elasticities of demand for water supply are -0.5885, -0.3009, -0.1818, and -0.0636 for Poor, Almost poor, Moderate, and Rich income classes, respectively. The expenditure elasticities for water supply in every income class assure us that water supply is a subsistence good and should be treated specially. The other expenditure elasticities confirm that the commodities in the food group are necessary goods and the commodities in the non-food group are luxury goods. The price elasticities of the demand for water supply of the high-income classes are less than those of the low-income classes. It demonstrates that each different income class has individual characteristics and the difference should not be ignored if the govern-
80 Chulalongkorn Journal of Economics 18(1), April 2006 ment would like to address vertical inequity. It is better to estimate price elasticity for each class of income separately than to aggregate all people into one group. This result is very useful for the government to conduct public policy efficiently and address vertical inequity. The water supply policies in Thailand, nowadays, are too weak to solve the upcoming water supply shortage and they also create inequity in ability to access the water supply between different household income classes. The government should implement a tax increase policy, not a price increase policy, to prevent the water supply firms from over-profiting and to be able to return the increased tax revenue to people through other ways in the future. Double Non-Uniform Tax (DNUT) is one of the good tools to reduce demand, redistribute income and increase the ability of the low-income households to access subsistence goods, which are water, food, medicine etc. Further studies may separate underground water from the non-food commodity group and drinking water from the food commodity group and relax the definition of the commodity group to a variety of commodity groups. Acknowledgments The author would like to thank Assist.Prof.Dr.Worawet Suwanrada and Assoc.Prof.Dr.Isra Sarntisart for helpful suggestions, and the Kasikorn Bank for the financial support of this study. References Decaluwe, B., Patry, A. and Savard, L. (1999). When Water is no longer heaven sent: Comparative pricing analysis in an AGE model. Laval University. Department of Business Economics. (2002). Price of Consumer Goods, by Community and Region Region 1998-2002. Department of Business Economics. Bangkok: Thailand. Diamond, P. A. and Mirrless, J. A. (1971). Optimal Taxation and Public Production I: Production Efficiency. American Economic
@ @ T. Chomtohsuwan : Differences in Household Demand 81 Review, Vol. 61: 8-27. Diamond, P. A. and Mirrless, J. A. (1971). Optimal Taxation and Public Production II: Tax Rules. American Economic Review, Vol. 61: 261-78. Division of Agricultural Economics, Ministry of Agriculture and cooperatives. (1979). Demand Elasticities for Food of Farmers 1977-1978. Thailand. (Thai Version) Fernando Arbue@s Gracia, MarIa A@ngeles GarcIa @ Valin)as and Roberto MartI@nez-Espin)eira. (2001). The Literature on the Estimation of Residential Water Demand. Department of Applied Economics. University of Zaragoza. Llunch, C. and Williams, R. (1975). Consumer Demand System and Aggregate Consumption in USA: An Application of the Extended Linear Expenditure System. The Canadian Journal of Economics, Vol. 8, No. 1: 49-66. Mark, W. Rosegrant, Ximing Cai and Sarah, A. Cline. (2002). Global Water Outlook to 2025: Averting and Impending Crisis. Washington D.C. USA: IFPRI. Meesook, O.A. (1976). Regional Consumer Price Indices for Thailand. Discussion Papers, Series 49, Thammasat University, Mimeographed. (Thai Version) National Statistical Office. (2002). Report of the 2002 Households Socio-Economic Survey. National Statistical Office (NSO). Bangkok: Thailand. Patamakitsakult, P. (1988). The Analysis of the Household s Food share Consumption. In Kraiyudht Dhiratayakinant, ed. The knowledge of Thai economists, The Economic Association of Thailand, p. 146-68. (Thai Version) Patamasiriwat, D. (1989). A Study on pattern of Consumption Expenditure from National Account Data. Chulalongkorn Journal of Economics Vol. 1 No. 2 September 1989. Bangkok: Chulalongkorn University, (Thai Version) Patamasiriwat, D. and A. Wetmanat. (1988). The Analysis of the Consumption pattern of Thai Households in 1986. In Kraiyudht
82 Chulalongkorn Journal of Economics 18(1), April 2006 Dhiratayakinant, ed. The knowledge of Thai economists, The Economic Association of Thailand, p. 117-45. (Thai Version) Ramsey, F. P. (1927). A Contribution to the Theory of Taxation. The Economic Journal, Vol. 37, No. 145: 47-61. Sarntisart, Isra. (1995). An Estimation of Consumer Demand. Chulalongkorn Journal of Economics Vol. 7, No. 3, Bangkok: Chulalongkorn University. Stone, R. (1954). Linear Expenditure System and Demand Analysis: An Application to the pattern of British Demand. The Economic Journal, Vol. 64, No. 255: 511-27. Techaposai, S. (1989). Household Consumption and Saving Pattern in Thailand. Master Thesis, Thammasat University. Trairatvorakul, P. (1983). Income and Price Changes and Food Consumption of the Poor in Thailand. Thammasat Economic Journal, Vol. 1, No. 4: 139-87. (Thai Version)