ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

What Affects Changes in Cereal Consumption?

This article diagnoses the factors that contribute to changes in cereal consumption at the household level across four expenditure groups, both for rural and urban regions. These factors can be decomposed into own price effect, substitution price effect, income effect and tastes and preferences.


What Affects Changes in Cereal Consumption?

This article diagnoses the factors that contribute to changes in cereal consumption at the household level across four expenditure groups, both for rural and urban regions. These factors can be decomposed into own price effect, substitution price effect, income effect and tastes and preferences.


s per Engel’s law, as the income of poor people rises consumption of staple food, which is a cheaper source of calories, is likely to decline. Decline in cereal consumption is substituted with the increased consumption of high value commodities. The shift away from cereal consumption is prominent across all households in different expenditure groups1 over time. The decline in cereal consumption can be attributed towards the diversification of food production, easy access of high value commodities, changed tastes and preferences and reduction in prices of cereals relative to other food commodities [Radhakrishna 2005; Kumar 1998; Murthy 2000].

Past studies [Radhakrishna and Ravi 1992; Kumar 1997; Rao 2000] have shown that economic growth, rise in per capita income, urbanisation, changing tastes and preferences, market integration, etc, are the dominating factors for the change in the per capita consumption of cereals. Mechanisation of agriculture, improvement in infrastructure and medical facilities also contribute to the reduction in energy requirement and thus less cereal consumption [Rao 2000]. This goes with the arguments by Kumar and Mathur (1996) that the demand for food is not only influenced by income changes, but also by differences in urban and rural lifestyles, the development of more advanced marketing systems and occupational changes that are closely linked with increasing per capita income. Tastes and preferences are also moving towards high value commodities. With the availability of cash money and high value commodities in the food basket, the decline in calories by reduced cereal consumption is partly compensated by intake of highly nutritive and vitamin rich food products. Rao (2000) observes that a reduction in the intake of foodgrains on this account should not be taken as deterioration in human welfare.

In this context, the present paper attempts to develop a demand model to quantify the effect of change in cereal consumption. The paper investigates the variables like price, income or choice of the consumers that lead to change in the consumption of cereals. The present paper is a methodological contribution to the existing empirical literature in this field. It also examines the implications of change in cereal consumption.

Data and Methodology

The study uses the consumer expenditure household survey data of National Sample Survey (NSS) rounds number 38, 43, 50 and 55 pertaining to the periods 1983, 1987-88, 1993-94 and 1999-2000 to compute the demand parameters needed for decomposition. The data refers to the average per capita consumption over the 30 days recall in all the rounds. The paper decomposes the factors effecting cereal consumption in rural and urban regions and across different expenditure groups.

A multi-stage (two-stage) budgeting framework is used to model the consumption behaviour of households. Earlier this model has been applied by Dey (2000), Kumar and Dey (2004) and Kumar et al (2005) for fisheries. The model is an extension of the “Almost Idle Demand System”. The model gives away the assumption of linearity in the expenditure function and assumes there is a non-linear relationship between income and expenditure. A quadratic equation is used as a specific case rather than a non-linear function. The model is quadratic in per capita expenditure and is thus named as the QUAIDS model.

In the first stage, the household makes decisions on how much of its total income (expenditure) is to be allocated for food consumption, conditional on the consumption of non-food goods and household and demographic characteristics. In the second stage, the household allocates the total food expenditure among different items/ groups (rice, wheat, coarse cereals, pulses, milk, edible oils, vegetables, fruits, meat, fish, eggs, sugar, other food, and nonfood). Following Blundell et al (1993), Dey (2000) and Kumar and Dey (2004), the specific functional form used in the two stages are as follows:

Stage 1: Food Expenditure Function

Ln(M) =α + γ1Ln(Pf)+ γ2Ln(Pnf)+ β0Ln(Y)

+ β1(lnY)2 + ΣθjZ …(1)

where M is the per capita food expenditure; Y is the per capita total expenditure (income); Pf is the household specific price index for food; Pnf is price index of nonfood. Socio-demographic and conditioning variables (vector Z) include the ratio of adults in the household, family size, and urban dummy. Equation 1 was estimated by the OLS method, and homogeneity of degree zero in prices and income was imposed by restricting γ1 + γ2+ β0 +2β1 Ln(Y) = 0 at the sample mean of Ln(Y).

Stage 2: Quadratic-AIDS (QUAIDS) Model

In stage 2 of the analysis, the quadratic extension to Deaton and Muellbauer’s (1980) almost ideal model (QUAIDS) for the food demand system will be used. This model is quite popular and was adopted recently by Meenakshi and Ray (1999) for the India food model, Dey (2000) for the fish demand model of Bangladesh and Kumar and Dey (2004) for the fish

Economic and Political Weekly February 3, 2007

demand model of India. The specific functional form for ith items/groups is as follows:

Si =ai+ ΣbijLn(FPi)+ coiLn(M/I) + cli(LnM/I)2


+ diUrban + ΣeikIMRk …(2)k

where FPi is the price of ith items/groups; I is the Stone geometric price index; Urban is a binary dummy variable for urban areas.2 The parameters of the model (ai, bij, ci, di and eik) were estimated by imposing the homogeneity (degree zero in prices), symmetry (cross price effects are same across the good), and adding up (all the budget shares add up to one) restrictions. The following restrictions were econometrically imposed.

n Homogeneity: Σ bij = 0; j=1

Symmetry: bij = bji, c11/c10 = c21/c20 =...= cn1/cn0;

The homogeneity and symmetry restrictions are imposed at the sample mean. Given the quadratic specification of the demand system (Equations 1 and 2) a test of symmetry additionally requires that the ratio of coefficients on food expenditure and the square terms in food expenditure be the same for all items/ groups [Blundell et al 1993]. The predicted value of food expenditure obtained from stage 1 will be used as the explanatory variable. The model is estimated separately for each expenditure class across rural and urban regions. Separate coefficients are generated for each expenditure group for the purpose of decomposition across each group. Coefficients are presented in Table 1. The price variables are significant at 1 per cent significance level, but in some cases, income variables are not significant even at 10 per cent significance level.

The estimated model is decomposed in two stages. In stage 1 (Equation 1) the coefficient of Pf explains the magnitude of change in the prices of food items on total food expenditure. Thus the first term explains the per cent change in budget expenditure on food with a 1 per cent change in prices of food; this defines the own price effect. The coefficient of Pnf explains the impact of the change in the price of non-food item on the total budget allocated to food. Thus the second term explains the per cent change in budget expenditure on food with a 1 per cent change in the prices of non-food items – substitution price effect, the coefficients of third and fourth term in equation 1 gives the magnitude of the income effect. The impact of taste and preferences is explained by the time variable.3

In stage 2 the estimated share equations are used to decompose the factors responsible for the change in cereal consumption. The first term in equation 2 explains the price effect. Own price is defined as the effect of the change in the price of cereals on its share in the total food expenditure budget. Substitution price effects are obtained by adding up the impact of the change in the prices of other food groups on the share of cereal in food expenditure. The linear and quadratic income terms in Equation 2 explains the income effect in stage 2 of the model and time trend is assumed to be reflecting the change in taste and preferences of the households in different groups and regions.


The concern about very poor sections of society in terms of the decline in cereal consumption can be countered by the argument that there is change in their consumption basket towards non-cereals and high value commodities. The increase in per capita income and decline in relative food prices play a key role in this diversification, even for very poor income groups. The factors that affect the consumption behaviour of a household can be decomposed into own price effect, substitution price effect, income effect and tastes and preferences. The decomposition is done

Table 1: Coefficients of Estimated Two Stage QUAIDS Model

Stage 1: Estimated Food Expenditure Function, India

Variable Very Poor Poor Non-Poor Rich

Intercept -144.166 -147.493 -148.590 -125.81 (-28.29) (-47.39) (-63.45) (-26.50) Ln(price index for food) 0.163 0.182 0.151 0.143

(3.74) (3.95) (3.89) (4.24) Ln(price index for non-food) -0.585 -0.165 -0.388 -0.587

(-8.73) (-6.64) (-9.54) (-10.26) Ln(per capita total expenditure) 0.348 0.236 -0.274 -0.646

(1.56) (0.91) (-1.10) (-3.35) Ln(per capita total expenditure)2 0.007 -0.023 0.047 0.101

(0.32) (-0.99) (2.11) (6.58) Family size -0.007 -0.001 0.011 0.013

(-0.64) (-0.10) (1.04) (1.01) Urban dummy 0.098 0.193 0.133 0.030 Time trend 0.074 0.076 0.077 0.066 Adjusted R2 0.97 0.96 0.97 0.97 Number of observations 212 229 239 240

Stage 2: Estimated Parameters of QUAIDS Demand System for Cereals

Food Groups Very Poor Poor Non-Poor Rich

Intercept 3.978 5.629 5.032 5.626

(1.97) (2.46) (2.39) (3.46) Cereal 0.187 0.179 0.097 0.081

(11.17) (9.53) (6.01) (5.95) Pulses -0.050 -0.060 -0.028 -0.019

(-10.13) (-12.47) (-7.01) (-5.33) Vegetables and Fruits -0.230 0.031 0.019 0.006

(4.41) (6.05) (3.56) (1.02) Milk -0.115 -0.083 -0.091 -0.100 (-11.47) (-9.94) (-9.04) (-10.88) Edible Oil -0.060 -0.626 -0.027 -0.017 (-14.33) (-14.72) (-7.00) (-4.12) Sugar -0.062 -0.062 -0.033 -0.027

(-16.53) (-17.43) (-11.16) (-10.00) Meat, fish and eggs 0.078 0.056 0.062 0.077

(6.99) (4.69) (5.55) (7.02) Ln(per capita food exp) 4.159 5.155 2.107 0.227

(4.80) (5.45) (2.61) (0.52) Ln(per capita food exp)2 -0.657 -0.863 -0.327 -0.054

(-4.76) (-5.37) (-2.71) (-0.90) Urban dummy -0.174 -0.177 -0.145 -0.111 Time trend -0.005 -0.006 -0.004 -0.003 DW statistics 4.8127 8.3494 3.0574 2.9406

Note: Figures in the parenthesis are the t-value.

Economic and Political Weekly February 3, 2007 in two stages and the results are presented in Table 2.

It is observed by different studies that the share of the budget allocated to food in total household expenditure has declined over time. The factors accounting for this change are explained through stage 1 of the decomposition model. The factors are decomposed for changes in per capita budget expenditure on food. The own price effect has a positive sign because as food becomes relatively expensive, the expenditure on food items increases. On the other hand, the substitution price effect sign is negative. This implies that as nonfood items become relatively cheaper than food items, consumers tend to shift some part of their expenditure on food to nonfood products. Table 2 shows that for very poor rural households, with a 1 per cent increase in the price of food, the budget expenditure on food increases by 0.21 per cent while with a decline in the price of non- food items, the budget allocated to food increases by 0.75 per cent. Thus as food becomes relatively expensive, its share in the budget expenditure increases. The income effect is positive for poor and very poor groups but negative for the upper two income groups. This is true for both rural and urban regions. The extra real income is net of the own price effect and substitution price effect. With an increase in real income, the poor and very poor groups increase their budget allocated to food while the non-poor and rich shift away budget expenditure from food as their income levels increase. As observed in Table 2, taste and preferences play an important role in favour of expenditure on food items for all the groups but its magnitude is much higher in poor groups than non-poor groups.

In stage 2 of the decomposition model, the factors affecting share of per capita cereal in total food expenditure is also decomposed into price effects, income effect and taste and preferences. Own price and substitution price have expected signs, thus as cereals become relatively expensive to their substitute, there is a shift away from cereal consumption.4 The food groups included as substitutes to cereals are pulses, fruits and vegetables, milk, edible oil, sugar and meat, fish and egg. The substitution effect is the net effect of all these food items. For the very poor group, the magnitude of own price effect and substitution price effect is same, thus the net change in share of cereals in total food consumption remains same for this group in spite of price change. Income has a positive effect on the share of cereal consumption in lower expenditure groups. In the poorest expenditure group extra income will lead to purchase of more cereals. The income effect is negative for non-poor and rich groups. This implies that as income rises, the poor continue to consume more cereals but after a certain income level they shift away from cereals to their substitutes. With time, the taste and preferences of households change, which leads to the decline in the share of cereal consumption in total food expenditure by all expenditure groups. The change in the share of cereals may tend to be zero with time but total cereal consumption would not be zero. The magnitude of this decline is higher for the poorest group, which implies that with time, preferences are shifting away from cereal consumption. This is due to the increased availability of cereal substitutes in nearby markets, urbanisation and demonstration effect. In the last two decades, with the development of rural markets, rural regions as well as the poor sections of society have found that various cereal substitutes are available at relatively cheaper prices. There are factors like urbanisation, development of market infrastructure, demonstration effect, eating out, etc, which cannot be quantified in the total estimated change. The total estimated change might not be equal to the observed change in consumption. Thus the difference in the estimated changes and observed changes gives the magnitude of these unexplained factors.


There is an ongoing debate that the decline in cereal consumption is an indicator of the lack of purchasing power and increasing poverty. The analysis presented above contradicts this argument. Studies have shown that over time, the decline in cereal consumption in the consumption bundle is accounted for by an increase in the consumption of high value commodities like fruits and vegetables, meat products and milk [Rao 2000; Kumar and Mathur 1996]. Thus, when a complete food consumption picture is taken into view, it is very well evident that a decline in cereal consumption is due to the diversification of the consumption bundle. This diversification has made even the poorest

Table 2: Stage Decomposition of Change in Cereal Consumption, 1983-99

(In per cent)

Ist Stage Decomposition
Expenditure Decomposition of Factors Affecting Budget Expenditure on Food
Groups Own Substitution Income Taste and Total Change in Change
Price Effect Price Effect Effect Preferences Per Capita Food Not
Expenditure Explained
Estimated Observed by Model
Very poor 0.21 -0.75 0.49 1.18 1.13 1.19 0.07
Poor 0.23 -0.21 0.27 1.22 1.51 1.20 -0.31
Non-poor Rich 0.19 0.18 -0.50 -0.75 -0.28 -0.70 1.24 1.06 0.65 -0.20 1.24 1.31 0.59 1.51
Very poor Poor 0.23 0.26 -0.73 -0.21 0.52 0.28 1.18 1.22 1.20 1.54 1.31 1.24 0.11 -0.30
Non-poor 0.21 -0.49 -0.28 1.23 0.67 1.24 0.57
Rich 0.20 -0.77 -0.72 1.06 -0.22 1.31 1.54
2nd Stage Decomposition
Expenditure Decomposition of Factors Affecting Share of Per Capita Cereal Consumption
Groups Own Substitution Income Taste and Total Change in Change
Price Effect Price Effect Effect Preferences Share Cereal Not
Consumption Explained
Estimated Observed by Model
Very poor 0.21 -0.21 0.47 -0.10 0.36 -0.14 -0.51
Poor 0.23 -1.67 1.90 -0.08 0.38 -0.08 -0.46
Non-poor 0.12 -0.12 -0.80 -0.06 -0.87 -0.10 0.77
Rich 0.10 -0.66 -0.36 -0.04 -0.97 -0.10 0.87
Very poor 0.22 -0.22 1.44 -0.10 1.34 -0.10 -1.44
Poor 0.23 -1.62 2.26 -0.08 0.79 -0.06 -0.85
Non-poor 0.12 -0.12 -0.59 -0.06 -0.65 -0.03 0.62
Rich 0.11 -0.63 -0.35 -0.04 -0.92 -0.04 0.88

Economic and Political Weekly February 3, 2007

household better off, due to the addition of high value food in the basket. This diversification can also be taken as an indicator of improvement in the quality of life. At the household level, food security refers to the ability of the household to secure adequate food to meet the dietary needs of all the members of the household. Thus, a lack of purchasing power would not have enabled very poor households to consume non-cereal high value food items. It is the taste and preferences as well as the non-price unexplained factors that are really responsible for the decline in cereal consumption for poor households.

If the decline in cereal consumption is taken as a measure of increasing poverty and malnutrition, then this effect is more due to the increase in relative prices of cereals vis-a-vis other food commodities, rather than worsening income levels. Ravallion (1990) states that the positive correlation between food prices and poverty is not an income distribution effect. Rather it appears to be due to covariant fluctuations between average consumption and food prices due to other variables including food supply, bad agriculture year, lower rural living standards and increase in food prices. In other words, rural poverty increases when foods prices rise and not because rural income distribution worsens, but rather, because rural average income declines. Thus, as income increased and non-cereal food items became relatively cheap in the last few years, the consumption of high value commodities increased. This can also be viewed as an increase in the welfare of the very poor group, due to the diversification of their consumption bundle with the addition of food items that were considered almost luxury commodities for them earlier.


The change in cereal consumption is the combined effect of various factors. As cereals are becoming relatively expensive, the positive own price and negative substitution price effect are leading to a change in consumption away from cereals. The positive income effect is illustrated by the results for the poor group while it is negative for the higher income groups. As income increases, the lower income groups tend to increase their utility levels by increasing the consumption of cereals, whereas the upper income groups move towards other food items. Taste and preferences play a very important role for all the income groups and even the lowest income groups have a tendency to shift away from cereal consumption. This shift away from cereals and towards high value agricultural commodities can be interpreted as a rise in the welfare of households due to a diversification in their consumption bundle.




[I am grateful to Praduman Kumar and Arvind Virmani for their valuable comments and suggestions. This paper is drawn from ICRIER working paper No 184 referred to as Mittal (2006). The comments of an anonymous referee are acknowledged. Views expressed in the paper are those of the author and not of ICRIER.]

1 Four expenditure groups are formed for both rural and urban households on the basis of the poverty lines adopted by the Planning Commission [Radhakrishna and Ravi 1990; Kumar 1998]. Based on the expenditure groups of the NSS, persons with expenditure below 75 per cent of the poverty line are defined as very poor; those between 75 per cent and the poverty line as poor; those at 150 per cent of the poverty line are termed as non-poor and those above 150 per cent of the poverty line are considered rich.

2 The urban dummy is used to pool data across all the four groups for both rural and urban regions. This is done to increase the number of observations. The dummy is used as a scale effect at different mean levels of rural and urban observations.

3 Since taste and preferences change with time and cannot be quantified in number, the decomposition model assumes that the time trend will be capturing the effect of changing tastes and preferences on cereal consumption.

4 The impact of non-food prices becoming relatively cheaper also affects cereal consumption, which is captured indirectly in stage 2 through this multistage framework model.


Blundell, R, P Pashardes and G Weber (1993): ‘What Do We Learn about Consumer Demand Pattern from Micro Data?’ American Economic Review, Vol 83, pp 570-97.

Dey, Madan Mohan (2000): ‘Analysis of Demand for Fish in Bangladesh’, Aquacultures, Economics and Management, Vol 4, pp 63-81.

Deaton, A S and J Muellbauer (1980): ‘An Almost Ideal Demand System’, American Economic Review, Vol 70, pp 359-68.

Kumar, Praduman and V C Mathur (1996): ‘Structural Changes in Demand for Food in India’, Indian Journal of Agricultural Economics, 51(4), pp 664-73.

Kumar, Praduman (1997): ‘Food Security: Supply and Demand Perspective’, Indian Farming, December, pp 4-9.

– (1998): ‘Food Demand and Supply Projections for India’, Agricultural Economics Policy Paper 98-01, Indian Agricultural Research Institute, New Delhi, India.

Kumar, Praduman and M M Dey (2004): ‘A Study on Modelling of Household Demand for Fish in India’, Indian Journal of Agricultural Economics, 59(3), pp 465-75.

Kumar, Praduman, M M Dey and F J Paraguas (2005): ‘Demand for Fish by Species in India: Three Stage Budgeting Framework’, Agricultural Economics Research Review, 18(2), pp 167-86.

Meenakshi, J V and R Ray (1999): ‘Regional Differences in India’s Food Expenditure Pattern: A Completed Demand Systems Approach’, Journal of International Development, Vol 11, pp 47-74.

Murthy, K N (2000): ‘Changes in Taste and Demand Pattern for Cereals’, Agricultural Economics Research Review, 13(1), pp 26-53.

Mittal, Surabhi (2006): ‘Structural Shift in Demand for Food: India’s Prospects in 2020’, ICRIER Working Paper 184, August.

Radhakrishna, R and C Ravi (1990): ‘Food Demand Projections for India’, Centre for Economics and Social Studies, Hyderabad.

– (1992): ‘Effects of Growth, Relative Price and Preferences of Food and Nutrition’, Indian Economic Review, Vol 27, pp 303-23.

Radhakrishna, R (2005): ‘Food and Nutrition Security of the Poor’, Economic and Political Weekly, Vol XL, No 18, April 30-May 6, pp 1817-21.

Rao, C H Hanumantha (2000): ‘Declining Demand for Foodgrains in Rural India: Causes and Implications’, Economic and Political Weekly, January 22, pp 201-06.

Ravallion, Martin (1990): ‘Income Effects on Under-Nutrition’, Economic Development and Cultural Change, Vol 38(3), University of Chicago Press, pp 489-515.

Back Volumes

Back Volumes of Economic and Political Weekly from 1976 to 2006 are

available in unbound form. Write to: Circulation Department,

Economic and Political Weekly

Hitkari House, 284 Shahid Bhagat Singh Road, Mumbai 400 001.

Economic and Political Weekly February 3, 2007

Dear Reader,

To continue reading, become a subscriber.

Explore our attractive subscription offers.

Click here

Back to Top