Review of Agricultural Economics—Volume 31, Number 4—Pages 873–893 DOI: 10.1111/j.1467-9353.2009.01471.x

An Analysis of Food Demand in China: A Case Study of Urban Households in Jiangsu Province

Zhihao Zheng and Shida Rastegari Henneberry

C

hina has had one of the world’s most rapidly developing economies for at least the past two decades. Population growth, accompanied by recent economic growth and rapid urbanization, has led to an increase in food demand and a considerable change in the composition of foods consumed in China. Rural households (roughly 60% of China’s consumers) decreased their per capita at-home consumption of food grains from 262 kg per person in 1990 to 219 kg per person in 2004, a decrease of over 16%. At the same time, they raised their per capita at-home consumption of foods of animal origin (meats, poultry, eggs, aquatic products, and dairy products), from 28 kg per person in 1990 to 42 kg per person in 2004, an increase of 50%. Urban at-home consumption of foods has changed even more drastically. The per capita consumption of food grains declined by 40%, from 131 kg in 1990 to 78 kg in 2004; whereas per capita consumption of foods of animal origin increased by 78%, from 41 kg in 1990 to 73 kg in 2004 [China’s National Bureau of Statistics (NBS), 1991–2005]. Considering that China has over one-fifth of the world’s consumers and an economy that has grown at an average rate of 9–10% annually since 1978, this country’s changing food consumption patterns have the potential to significantly impact the global magnitude and pattern of food demand. Research is needed to provide a better understanding of China’s food buyer preferences and the potential for marketing food in China. Several studies have been conducted on China’s household demand for food. However, these studies have not taken into account the more recent changes in economic structure in China, including the rapidly rising incomes during the past decade. These past studies have used a variety of data, including aggregate time-series data (Lewis and Andrews), aggregate city-level cross-sectional data (Wu, Li, and Samuel), pooled time-series and cross-sectional data at the provincial


Zhihao Zheng is an associate professor in the College of Economics and Management at China Agricultural University and a former postdoctoral research associate in the Department of Agricultural Economics at Oklahoma State University.
Shida R. Henneberry is a professor in the Department of Agricultural Economics, Oklahoma State University.

874

Review of Agricultural Economics

level (Fan, Wailes, and Cramer), and pooled time-series and cross-sectional data at the county level (Zhang, Mount, and Boisvert). Some of the more recent studies have used household survey data collected by NBS. The NBS household survey data are considered superior compared to the available time-series data for research because they include detailed demographic characteristics that allow for the assumption of heterogeneity in preferences across households. Additionally, the large sample size included in the NBS household survey data allows estimating a relatively large demand system. Among the published studies based on the household survey data, Halbrendt et al. and Gao, Wailes, and Cramer focus on rural households in Guangdong and Jiangsu provinces, respectively. Zhang and Wang; and Yen, Fang, and Su concentrate on urban households in China in 1998 and 2000, respectively. Liu and Chern; and Gould and Villarreal analyze the food demand of urban families using the household survey data for Shandong, Jiangsu, Heilongjiang, Henan, and Guangdong provinces in 1997 and 2001, respectively. Some studies have taken advantage of the availability of household survey data over the years and have analyzed food demand using the available pooled time-series and cross-sectional data at the household level. Gould uses three consecutive years of NBS urban household survey data (1995–97) for Jiangsu, Shandong, and Guangdong provinces to estimate a system of demands for food commodities; and Guo et al. use data for 1989, 1991, and 1993 from the China Health and Nutrition Survey to examine food consumption behaviors of both urban and rural households across income levels. This study goes beyond the previous studies in data use by utilizing a more recent data set—the 2004 NBS urban household survey data. Furthermore, this study uses a generalized almost ideal demand system (GAIDS) that “allows the demand shifters to be included in a fashion that is flexible, parsimonious, and maintains the model’s invariance to changes in units of measurement” (Alston, Chalfant, and Piggott, p. 77). The objective of this study is to estimate the impacts of economic factors (prices and expenditures) and noneconomic factors (demographic variables) on urban household demand for ten broad food categories in Jiangsu province. Jiangsu is one of China’s major provinces with its gross domestic product (GDP) accounting for more than 9% of China’s national GDP. Jiangsu’s urban per capita disposable income was ranked seventh among thirtyone provinces in the nation in 2004 (NBS 2005). Thus, a study of urban household food consumption patterns in this province may help in understanding China’s national demand and the factors that affect it. China has undergone a massive urbanization during the twenty-first century, which has had a dramatic effect on its food demand (Hsu, Chern, and Gale). According to China’s official statistics, only 42% of China’s population lived in cities and towns in 2004 (NBS 2005). This urban population share is expected to grow to 50% by 2020 (Hsu, Chern, and Gale). Given that urban residents in China have much higher per capita incomes compared to rural residents, urban households have been the driving force behind the growth in food demand and the emerging demand for better quality foods in China.1 This changing food demand has led to a significant increase in the number of supermarkets, convenience stores, and food-away-from-home (FAFH) outlets that offer greater convenience, variety, and quality to consumers (Gale and Huang). By shedding light on China’s contemporary consumer preferences, the results of this study are expected to

An Analysis of Food Demand in China

875

be useful to policy makers and exporters in the United States and other food exporting countries in developing effective trade policies and marketing strategies for trade with China. More specifically, the food demand elasticities provided by this study may be used to analyze the impacts of trade policies on China’s economy and the world food markets. The remainder of this study is organized as follows. A model of urban household food demand in China is presented in the following section. The data and estimation procedures are then described. Results are presented next, followed by the summary and conclusions.

Model Specification The almost ideal demand system (AIDS), developed by Deaton and Muellbauer, has been widely used in empirical demand studies due to its theoretical advantages (Henneberry, Piewthongngam, and Qiang). The AIDS specification, however, does not maintain the desired property of estimated economic effects (elasticities) being invariant to disproportionate changes in units of measurement when incorporating demand shifters in a traditional way (Alston, Chalfant, and Piggott). One solution suggested by Alston, Chalfant, and Piggott is to adopt a generalized version of the AIDS, the GAIDS, first derived by Bollino. The GAIDS model includes precommitted quantities as parameters. These precommitted quantities are independent of prices and expenditure. By augmenting these precommitted quantities to be a function of demand shifters, the lack of invariance problem that may be present in the AIDS specification can be avoided in the GAIDS model. Following Bollino, as well as Piggott and Marsh, the GAIDS model can be expressed in share form as

(1)

  ∗  N  pi c i M M∗ wi = i j ln( p j ) + i ln + i + , M M P j=1

i, j = 1, . . . , N,

where wi represents the budget share associated with the ith good; pi is the price of the ith good; c i denotes the precommitted quantity of good i; M is the total exN penditure exhausted in the system; M∗ = M − i=1 pi c i denotes supernumerary N expenditure, where i=1 pi c i is the mathematical representation of the total preN N N committed expenditure; ln(P) = 0 + i=1 i + 0.5 i=1 j=1 i j ln( pi ) ln( p j ) is a nonlinear price index; and i , i j , and i are parameters to be estimated. Demographic variables that affect consumption behavior are incorporated into the model by allowing the precommitted quantities in equation (1), the c i s, to be a function of these variables. That is (2)

c i = c i0 +

K 

c ik dk ,

k=1

where k = 1, . . . , K is used to identify the demographic variables considered in the  study, c i0 and c ik s are parameters to be estimated, and dk is the kth demographic

876

Review of Agricultural Economics

variable. Thus, the resulting share equations become (3)

   ∗  K N   M∗ pi M wi = c ik dk + i j ln( p j ) + i ln c i0 + i + M M P k=1 j=1

where  ∗

M = M−

N 

 c i0 +

K 

i=1

c ik dk

 pi .

k=1

The properties from neoclassical demand theory can be imposed on model equation (3) by restricting its  parameters. The adding-up restriction ( wi = 1) is given by (4a)

N 

N 

i = 1,

i=1

i j = 0, and

i=1

N 

i = 0.

i=1

Homogeneity (wi unchanged by a proportional change of all prices and income) is imposed as (4b)

N 

i j = 0 for any j.

j=1

Slutsky symmetry is given by (4c)

i j =  ji for any i and j.

A major problem when estimating a complete demand system is the endogeneity of expenditure term in the system (LaFrance). To control for the expenditure endogeneity in the GAIDS model, a nonlinear full information maximum likelihood (FIML) estimation procedure is used in this study. As pointed out by Greene and Dhar, Chavas, and Gould, this procedure can generate consistent and asymptotically efficient estimates under the assumption that the error terms are normally distributed. Similar to Blundell, Pashardes, and Weber, this study specifies a reduced form expenditure equation where household food expenditure is a function of disposable household income, prices of the studied goods, and demographic variables that are the same as those used in equation (2). The reduced form expenditure equation is specified as (5)

M = a0 +

K  l=1

a l dl +

N 

br pr + c y ln(y)

r =1

where dl is the lth demographic variable defined above; pr is the price of the rth good; y denotes disposable household income, which is used as an identifying instrument; and a 0 , a l s, br s, and c y are parameters to be estimated.

An Analysis of Food Demand in China

877

The uncompensated (Marshallian) price elasticities of demand are calculated as (6) e iuj

    

 N  p j c j pjc j M∗ ci  × × +  = i j − 1 − −  +  ln( p ) + ij i j kj k qi M∗ M∗ pi q i k=1 

where q i denotes the quantity of good i, and i j = −1 + c i /q i if i = j and zero otherwise. The expenditure elasticity of demand is  e i = (1 − c i /q i ) × M M∗ + i /wi .

(7)

The compensated (Hicksian) price elasticities of demand are e icj = e iuj + w j e i .

(8)

The demographic elasticity of demand is  (9)

d e ik

= 1 − i −

N 

 i j ln( p j ) − i ln

j=1

M∗ P



 − i ×



c ik dk qi

 ,

where dk = 1 if dk is a binary (0/1) variable and the mean of the variable otherwise.

Data and Estimation Procedures The data set used for this study was collected by NBS for Jiangsu province in 2004. The NBS conducts a nationwide urban household survey annually. As an official statistical activity, the urban household survey collects extensive socioeconomic information on income, consumption, employment, housing, demographics, education, and asset ownership. Unlike most income and expenditure surveys that cover only a short period, the urban household survey in China captures expenditures and consumptions through a diary kept by the chosen households over the course of an entire year. These households are selected by NBS and represent the households belonging to various income classes in urban Jiangsu. Thus, the data set used for this study reflects actual consumption patterns of a set of households during an entire year (Gale and Huang). The sample of households selected for the survey in Jiangsu province is chosen from twenty-eight cities and towns and has 4,600 households, which accounts for about 0.05% of total urban households in the province in 2004. However, the data set available for this study has only 922 households, which are drawn systematically from the 4,600 sample households. After deleting the households that contain missing observations for two or more food categories, only usable data for 902 households remain that are used for this study.2 The food products that are analyzed in this study consist of ten broad food categories: grains, oils and fats, meats (encompassing pork, beef, and mutton), poultry, eggs, aquatic products, dairy products, vegetables, fruits, and other foods (including starches

878

Review of Agricultural Economics

and tubers, alcoholic beverages, beverages, and cakes). On a per capita basis, the average total expenditure on the studied food products accounts for 82.6% of total expenditures for food consumed at home and 69.4% of total food expenditures in the data set for the 902 surveyed households. Households report their food expenditures and the physical quantities that pertain to their food consumption in the survey diary. The prices paid by the households are calculated by dividing the consumer expenditures on a food product by its corresponding quantity. Hence, the price calculated in this manner (unit value) is household specific, representing household purchase decisions. In most instances, consumers choose both the quantity and the quality of consumption simultaneously. Therefore, the calculated price should be adjusted for quality differences among households before it can be used to estimate commodity demand functions from cross-sectional data. The quality and price adjustments follow the procedure proposed by Cox and Wohlgnant. Not all households purchased all food products during the survey period. If no expenditure or quantity occurs, the quality-adjusted price is equal to the regional (city or town) average price for the consuming households in that region.3 In this study, the food products examined are treated to be weakly separable from other food and nonfood items in the consumer’s budget. Consequently, the demand for an individual food product depends only on the expenditure on the studied foods, the prices of the foods within the system, and the included demand shifters representing demographic changes. Demographic variables used in this study include region (south vs. north), city size (towns vs. cities), household size, ratio of the number of seniors (aged sixty-one and above) to total household members, ratio of the number of children (aged seventeen and below) to total household members, educational levels of household heads (college and above level vs. others), and ratio of expenditures for FAFH to total food expenditures. Table 1 contains summary statistics and a description of the variables used in the estimation. The average household size consists of three persons. The average per capita disposable income is 10,551 Yuan (equivalently US$1,286) per year. The average household in urban Jiangsu allocates 12% of its expenditures on the studied foods to grains, with 22%, 12%, and 13% allocated to meats, aquatic products, and vegetables, respectively. Thus, grains, meats, aquatic products, and vegetables are the main components in a consumer’s diet in urban Jiangsu. Additionally, as shown in the table, the low percentage of the number of children and the high percentage of the number of adults aged sixty-one and above indicate that population in urban Jiangsu has been aging, a result of government policy advocating later marriages, fewer births, and one birth per couple in urban areas. This changing age structure of the population is expected to affect the composition and the quantity of the food products consumed. Data are not available from NBS household survey data for oils and fats and dairy products for 5% and 11% of households, respectively. Thus, the consistent two-step (CTS) estimation procedure for a system of equations with limited variables, proposed by Shonkwiler and Yen, is used to account for zero expenditure shares resulting from missing values for these two dependent variables. In the first step of CTS, a probit model is estimated separately for oils and fats and dairy products, using the maximum likelihood estimation method to obtain the univariate standard normal probability density functions (pdf) and the standard normal

An Analysis of Food Demand in China

879

Table 1. Summary statistics, urban Jiangsu province, China, 2004a Variables Budget Share Grains Oils and fats Meats Poultry Eggs Aquatic products Dairy products Vegetables Fruits Other foods Quality-Adjusted Price (Yuan/Kg) Grains Oils and fats Meats Poultry Eggs Aquatic products Dairy products Vegetables Fruits Other foods Demographic Variables Per capita income Household size Ratio of seniorsb Ratio of childrenb Ratio of FAFH spendingb South (yes = 1; 0 otherwise) Town (yes = 1; 0 otherwise) College (yes = 1; 0 otherwise)

Mean

Standard Deviation

0.124 0.047 0.219 0.081 0.041 0.120 0.060 0.130 0.087 0.091

0.067 0.029 0.068 0.046 0.023 0.063 0.056 0.046 0.053 0.059

3.036 8.718 16.206 13.849 6.126 12.528 6.131 2.159 2.981 6.380

0.831 2.833 2.086 2.716 0.951 4.908 4.957 0.638 1.237 4.786

10551 3.004 0.206 0.143 0.135 0.459 0.282 0.074

8221 0.989 0.352 0.162 0.134 0.499 0.450 0.262

Source: Calculated based on the National Bureau of Statistics (NBS) data regarding 902 households in urban Jiangsu, China, 2004. a Statistics refer to food-consumed-at-home data. b Ratio of seniors, ratio of children, and ratio of food-away-from-home (FAFH) spending refer to the ratio of seniors (aged sixty-one and above) to total household members, the ratio of children (aged seventeen and below) to total household members, and the ratio of FAFH spending to total food expenditures, respectively.

cumulative distribution functions (cdf). The explanatory variables included in the estimation are logarithm of household disposable income, logarithms of prices of the ten studied food products, and the demographic variables that are the same as those used in equation (2). In the second step of CTS, the ten-good GAIDS system for nine equations of food categories encompassing grains, oils and fats, meats, poultry, eggs, aquatic products, dairy products, vegetables, and fruits are then estimated simultaneously with the reduced form expenditure equation (5)

880

Review of Agricultural Economics

using the FIML estimation method, with homogeneity and symmetry imposed. For equations for oils and fats and dairy products, the estimated cdf and pdf from the first-step probit estimations are applied (see Yen, Kan, and Su for a detailed explanation). The “other foods” category here is treated as a residual category with no specific demand of its own. Consequently the price and expenditure elasticities for the “other foods” category are derived using the adding-up restrictions N N  specified as i=1 wi e i = 1, i=1 wi e iuj = −w j , and Nj=1 e iuj + e i = 0 (Yen, Lin, and Smallwood);4 while the demographic elasticities for “other foods” category are calculated directly using the parameters of the adding-up restrictions.5

Results The estimated parameters and the adjusted R2 s of the GAIDS model are presented in table 2. Of the 190 parameters estimated in the system, more than onethird (88) are significantly different from zero at the 10% level. Moreover, most of the constant components (c i0 ) of the estimated precommitted quantities are negative, which is in contradiction with the results reported by Park et al. and  Piggott and Marsh. However, regularity conditions do not require the c i0 s to be  nonnegative and, consequently the signs of c i0 s should be regarded as empirical questions (Pollak and Wales). Additionally, parameter estimates for i (i.e., i ) are statistically significant in the oils and fats and dairy products equations, providing evidence that it is important to account for zero observations in these goods.6 Price and Expenditure Elasticities All price and expenditure elasticities are evaluated based on parameter estimates and sample means of explanatory variables. Standard errors of these elasticities are approximated using the delta method. The full matrix of the uncompensated (Marshallian) price elasticities for the ten studied food products is reported in table 3. Consistent with economic theory, all own-price elasticities are negative. With the exception of own-price elasticity for aquatic products, which is not statistically significant, own-price elasticities for the other studied food products are significant at the 10% level. Own-price elasticities for grains, oils and fats, dairy products, and other foods (including starch and tubers, alcohol beverage, beverages, and cakes) are greater than unity in absolute terms. The elasticities for the other studied food products (including meats, poultry, eggs, aquatic products, vegetables, and fruits) are less than unity in absolute values. The other foods category has the highest own-price elasticity in absolute value (−1.61), whereas the aquatic products category has the lowest own-price elasticity in absolute value (−0.10) among all the food products considered. Table 3 also presents the expenditure elasticities, which are all positive and significantly different from zero at the 5% level. Yet, only some of these elasticities (including grains, oils and fats, eggs, and vegetables) are found to be significantly different from 1.0 at the 10% level.7 Because the studied food products are treated as being weakly separable from other food and nonfood items in the consumer’s budget, the expenditure elasticities are conditional. The unconditional expenditure elasticity for any of the studied food categories may be calculated as the product of the expenditure elasticity for that food category and the elasticity

0.184∗∗

0.255∗∗ −0.008 −0.002 0.013 −0.006 0.000 −0.008 0.021∗∗ −0.012 0.004 −0.001 −0.024

Constant () ln(p-grains) ln(p-oils and fats) ln(p-meats) ln(p-poultry) ln(p-eggs) ln(p-aquatic products) ln(p-dairy products) ln(p-vegetables) ln(p-fruits) ln(p-other foods) ln(m∗ /P) b Adjusted R2 0.096

0.042

0.194

0.010 −0.005 0.002 −0.015∗∗ −0.003 −0.001 −0.009

0.077

−9.594 4.730∗∗ 2.074 3.960∗∗ 4.622∗ −20.451∗∗ 3.212 −24.027∗∗

Eggs

the first-step probit regressions.

0.059∗∗ −0.003 −0.023∗∗ 0.001 0.001 −0.007 0.002 −0.014

0.060∗∗ −0.021∗ 0.013∗∗ −0.041∗∗ −0.006 −0.019∗ 0.000 −0.004 −0.011

0.184∗

−21.443∗∗ 15.445∗∗ 2.377 2.231∗ 1.489 −12.592∗ 0.657 −21.733∗∗

−28.969∗∗ 13.350∗∗ 3.468 6.271∗∗ 5.081 −35.259∗∗ 0.695 −38.774∗∗ 0.287∗

Poultry

Meats

a Estimated using 2004 China’s NBS household survey data. b  indicates the univariate standard normal pdf estimated in ∗ , ∗∗ Significant at the 10% and 5% levels, respectively.

0.170

−6.915 2.733∗ −3.374∗∗ 3.047∗∗ 0.841 −4.921 1.417 −18.080∗∗

−40.656 25.831∗∗ −16.421∗ 30.930∗∗ 36.746∗∗ −154.873∗∗ −28.523 −170.159∗∗

Constant (c0 ) South Town Household size Ratio of seniors Ratio of children College Ratio of FAFH spending

−0.005 0.004 −0.003 0.002 −0.006 0.007∗∗ −0.004 0.007∗∗ 0.001 −0.019∗∗ −0.041∗∗ 0.119

Oils and Fats

Grains

Explanatory Variables

0.244

0.093∗∗ −0.002 −0.020∗∗ 0.016∗∗ −0.006∗∗ 0.028∗∗

−0.080

−17.202∗ 12.774∗∗ 7.567∗∗ 2.277 5.009 −34.624∗∗ −1.946 −19.577∗∗

Aquatic Products

Coefficients of

−0.024∗∗ 0.012∗∗ −0.012∗∗ 0.001 0.042∗∗ 0.041∗∗ 0.136

−0.209∗∗

−6.118 7.932∗∗ 1.761 −3.426 1.124 39.527∗∗ −2.479 22.370∗

Dairy Products

Table 2. Estimated GAIDS parameters, urban Jiangsu province, China, 2004a

0.197

0.057∗∗ −0.001 0.000 −0.034∗∗

0.456∗∗

−92.215∗∗ 77.619∗∗ −16.927 27.448∗∗ 48.325∗∗ −202.464∗∗ 4.612 −263.522∗∗

Vegetables

0.149

−0.006 0.002 0.029∗∗

−0.134

6.043 43.077∗∗ −2.037 −4.412 −8.990 −5.384 −3.464 −3.213

Fruits

–

– –

–

−5.364 7.143∗∗ 3.007 7.114∗∗ 6.594 −21.620∗∗ 7.188 −25.627∗∗

Other Foods

An Analysis of Food Demand in China 881

Grains

0.032 (0.057) −1.312∗∗ (0.116) 0.053 (0.040) 0.004 (0.073) 0.143 (0.099) −0.103∗∗ (0.041) 0.006 (0.048) 0.027 (0.040) 0.042 (0.049) −0.033 (0.035)

Oils and Fats

(0.099) 0.400∗∗ (0.135) −0.853∗∗ (0.122) −0.225 (0.155) 0.648∗∗ (0.197) −0.473∗∗ (0.084) −0.316∗∗ (0.114) −0.049 (0.089) −0.078 (0.105) −0.052 (0.078)

0.285∗∗

Meats

a Estimated using 2004 China’s NBS household survey b Standard errors are given in the parentheses. ∗ , ∗∗ Significant at the 10% and 5% levels, respectively.

Other foods

Fruits

Vegetables

Dairy pro.

Aquatic pro.

Eggs

Poultry

Meats

(0.171) −0.417∗∗ (0.109) 0.121∗∗ (0.052) −0.022 (0.114) 0.100 (0.133) −0.144∗∗ (0.068) 0.276∗∗ (0.098) −0.022 (0.062) −0.024 (0.078) 0.249∗∗ (0.071)

−1.221∗∗

Oils and fats

Grains

Commodity

data.

0.007 (0.083) 0.341∗∗ (0.096) −0.090 (0.058) −0.347∗ (0.193) −0.029 (0.151) −0.279∗∗ (0.064) −0.133∗ (0.072) 0.109∗ (0.065) −0.175∗∗ (0.082) −0.063 (0.053)

Poultry 0.032 (0.043) −0.218∗∗ (0.069) 0.101∗∗ (0.034) −0.024 (0.069) −0.849∗∗ (0.260) −0.079∗∗ (0.034) −0.035 (0.037) −0.125∗∗ (0.039) −0.079∗∗ (0.039) 0.154∗∗ (0.034)

Eggs −0.091 (0.076) −0.169∗∗ (0.082) −0.250∗∗ (0.049) −0.378∗∗ (0.098) −0.202∗ (0.117) −0.101 (0.107) 0.012 (0.089) −0.216∗∗ (0.058) 0.299∗∗ (0.082) 0.068 (0.054)

Aquatic Products

Price of

(0.048) 0.200∗∗ (0.044) −0.064∗∗ (0.030) −0.049 (0.046) −0.014 (0.054) 0.038 (0.036) −1.209∗∗ (0.163) 0.048∗ (0.029) −0.105∗∗ (0.051) −0.028 (0.039)

0.186∗∗

Dairy Products −0.012 (0.068) 0.182∗∗ (0.082) −0.061 (0.049) 0.137 (0.098) −0.409∗∗ (0.130) −0.294∗∗ (0.058) −0.024 (0.071) −0.500∗∗ (0.100) −0.162∗∗ (0.068) −0.024 (0.045)

Vegetables 0.011 (0.059) −0.153∗∗ (0.071) −0.023 (0.041) −0.155∗ (0.083) −0.151∗ (0.087) 0.195 (0.053) −0.190∗∗ (0.074) −0.067 (0.045) −0.865∗∗ (0.141) 0.158∗∗ (0.049)

Fruits −0.025 (0.050) 0.429∗∗ (0.079) 0.024 (0.034) 0.058 (0.051) −0.060 (0.069) 0.041∗∗ (0.038) 0.240∗∗ (0.068) −0.019 (0.036) 0.170∗∗ (0.057) −1.609∗∗ (0.224)

Other Foods

0.795∗∗ (0.092) 0.717∗∗ (0.085) 1.040∗∗ (0.062) 1.001∗∗ (0.109) 0.824∗∗ (0.097) 1.198∗∗ (0.085) 1.372∗∗ (0.147) 0.814∗∗ (0.065) 0.978∗∗ (0.094) 1.180∗∗ (0.112)

Expenditure

Table 3. Uncompensated (Marshallian) price and expenditure elasticities, urban Jiangsu province, China, 2004a

882 Review of Agricultural Economics

An Analysis of Food Demand in China

883

of total expenditure on the studied food categories with respect to income. Considering that income elasticity of food expenditure on the studied food products is positive, the signs of the resulting income elasticities indicate that all the studied food products are normal goods. Expenditure elasticities for meats, poultry, aquatic products, dairy products, and other foods are greater than unity, with dairy products elasticity having the largest value at 1.37. Expenditure elasticities for grains, oils and fats, eggs, vegetables, and fruits are less than one, with the oils and fats elasticity having the smallest value at 0.72. These findings suggest that as consumers’ food expenditures increase, consumers in urban Jiangsu spend proportionately more on dairy products, aquatic products, other foods, poultry, and meats; and less on fruits, eggs, vegetables, grains, and oils and fats following in order. The meat category (including pork, beef, and mutton) has a relatively lower expenditure elasticity compared to aquatic products, dairy products, and other foods (including starch and tubers, alcohol beverage, beverages, and cakes). Since pork accounts for more than 70% of the studied total meat expenditures in this study, the relatively low expenditure elasticity for meats is consistent with the fact that pork is the most widely consumed and affordable meat in China. According to the findings of this study, a consumer’s expenditure for the meat category in urban Jiangsu is expected to increase by a larger amount than that of poultry, fruits, eggs, vegetables, grains, and oils and fats as household incomes rise. The aquatic food category in this study has a relatively higher expenditure elasticity and lower own-price elasticity in absolute term compared to the other food categories examined in this study, which may suggest that aquatic food consumption is driven more by a change in expenditures than in price. Thus, the demand for aquatic products is expected to increase as the result of the rising per capita incomes. Moreover, these results for urban Jiangsu support the findings of Shono, Suzuki, and Kaiser, and Yen, Fang, and Su that indicate China’s dietary pattern is moving toward the diets of consumers in Japan, South Korea, Taiwan, and Hong Kong. These developed Asian countries and regions depend more on seafood as their source of protein than the western countries. Similar to the aquatic product category, poultry meat has a relatively larger expenditure elasticity and lower own-price elasticity in absolute value compared to other studied food products. Per capita at-home consumption of poultry meat in urban China almost tripled between 1990 and 2003, before leveling off during recent years (2004–6). The slow growth in poultry meat consumption in urban China since 2004 has been mainly due to frequent outbreaks of diseases in poultry production, which has caused consumers to switch to other meat products such as pork, beef, and mutton (personal communication with an NBS official). According to the estimated expenditure elasticity, poultry meat consumption in urban China is expected to substantially increase as household incomes continue to grow in the future and if poultry meat regains the reputation of being a safe product. The dairy category is more responsive to changes in own price and expenditure than most food products examined in this study. The high expenditure and ownprice elasticities (in absolute value) for dairy products illustrate that both income and own price play important roles in dairy food consumption. If the current price structure remains unchanged, the demand for dairy products is expected to increase as household incomes and food expenditures rise. Additionally, the

884

Review of Agricultural Economics

per capita consumption of dairy products in 2004 in urban Jiangsu is reported as 22 kg. The figure is much lower than those reported by Yen, Fang, and Su for Japan (65.8 kg), South Korea (28.6 kg), Taiwan (43.0 kg) and the United States (256.6 kg) in 2002. According to China’s official statistics, dairy product consumption has significantly increased in both urban Jiangsu and urban China over the past decade. The per capita consumption of fresh milk and yogurt in urban Jiangsu has increased from 4 kg in 1995 to 12 kg in 2000 and to 22 kg in 2004, an increase of more than 400% over the past ten years (NBS 1996–2005). Dairy demand in China is anticipated to increase in the future with growing urbanization and westernization associated with the growth in per capita incomes (Fuller, Beghin, and Rozelle; Yen, Fang, and Su). The nondiagonal elements in table 3 are the estimated uncompensated crossprice elasticities. These cross-price elasticities indicate a mixture of gross complements and substitutes. About half of the cross-price elasticities are significantly different from zero at the 10% or lower levels. Results indicate that the meats category is a gross substitute for grains and eggs; but a gross complement to aquatic products and dairy products. Moreover, the vegetables category is a gross substitute for poultry and dairy products; while it is a gross complement to eggs and aquatic products. Similar patterns also exist for other food products. Relative to own-price and expenditure elasticities, the cross-price effects are smaller in magnitude, which may indicate that China’s consumers are more responsive to changes in products’ own prices and food expenditures. Compensated (Hicksian) price elasticities are reported in table 4. Similar to the uncompensated elasticities, all compensated own-price elasticities are significant and negative, except for the poultry category and aquatic products category. The former is negative but statistically insignificant, while the latter is positive although very small in magnitude and statistically nonsignificant. Unlike their uncompensated counterparts that indicate a mix of gross substitutes and complements, the compensated cross-price elasticities suggest that net substitution is a dominant pattern. More specifically, among the ninety compensated crossprice elasticities, more than one-third (thirty-four) are positive and significant indicating net substitution; while about 10% (ten) are negative and significant indicating net complementarity. The findings are consistent with those reported by Yen, Fang, and Su for urban China.

Effects of Demographic Variables on Household Food Demand The effects of demographic variables on household food demand are measured by the estimated demographic elasticities in table 5. According to the results, households to the south of the Yangtze River (south, table 5) spend more of their annual income on nine of the ten studied food products (except for oils and fats category) compared to those living in the north of Yangtze River. Geographically, Jiangsu province consists of two parts, one to the north of the Yangtze River—a region that is less economically developed relative to the south, and another to the south of the Yangtze River—a region that is more economically developed. Thus, the estimated regional demographic elasticities in this study may mirror the difference in economic development levels between the south and the north regions.

Oils and Fats

0.068 (0.057) −1.280∗∗ (0.115) 0.100∗∗ (0.040) 0.050 (0.073) 0.180∗ (0.098) −0.049 (0.041) 0.068 (0.047) 0.063 (0.041) 0.086∗ (0.049) 0.020 (0.035)

Grains

−1.122∗∗ (0.169) −0.328∗∗ (0.109) 0.250∗∗ (0.052) 0.102 (0.114) 0.201 (0.133) 0.005 (0.068) 0.446∗∗ (0.098) 0.079 (0.063) 0.097 (0.080) 0.395∗∗ (0.072) data.

0.460∗∗ (0.096) 0.557∗∗ (0.133) −0.624∗∗ (0.117) −0.005 (0.155) 0.829∗∗ (0.195) −0.209∗∗ (0.082) −0.015 (0.114) 0.130 (0.085) 0.136 (0.104) 0.207∗∗ (0.083)

Meats

a Estimated using 2004 China’s NBS household survey b Standard errors are given in the parentheses. ∗ , ∗∗ Significant at the 10% and 5% levels, respectively.

Other foods

Fruits

Vegetables

Dairy pro.

Aquatic pro.

Eggs

Poultry

Meats

Oils and fats

Grains

Commodity 0.072 (0.082) 0.399∗∗ (0.096) −0.006 (0.059) −0.266 (0.190) 0.038 (0.151) −0.181∗∗ (0.063) −0.021 (0.070) 0.175∗∗ (0.064) −0.095 (0.080) 0.033 (0.056)

Poultry 0.065 (0.043) −0.188∗∗ (0.069) 0.144∗∗ (0.034) 0.017 (0.068) −0.816∗∗ (0.258) −0.030 (0.034) 0.021 (0.036) −0.091∗∗ (0.039) −0.040 (0.039) 0.202∗∗ (0.034)

Eggs 0.004 (0.075) −0.083 (0.082) −0.125∗∗ (0.048) −0.258∗∗ (0.097) −0.103 (0.117) 0.043 (0.102) 0.176∗∗ (0.088) −0.118∗∗ (0.058) 0.416∗∗ (0.082) 0.210∗∗ (0.058)

Aquatic Products

Price of

0.234∗∗ (0.048) 0.243∗∗ (0.044) −0.001 (0.030) 0.011 (0.046) 0.035 (0.054) 0.110∗∗ (0.037) −1.126∗∗ (0.160) 0.097∗∗ (0.029) −0.046 (0.050) 0.043 (0.038)

Dairy Products 0.092 (0.068) 0.276∗∗ (0.082) 0.075 (0.050) 0.268∗∗ (0.098) −0.301∗∗ (0.131) −0.138∗∗ (0.059) 0.155∗∗ (0.068) −0.394∗∗ (0.097) −0.034 (0.068) 0.130∗∗ (0.048)

Vegetables

Table 4. Compensated (Hicksian) price elasticities, urban Jiangsu province, China, 2004a

0.081 (0.057) −0.091 (0.070) 0.068∗ (0.040) −0.068 (0.082) −0.079 (0.088) 0.300∗∗ (0.053) −0.070 (0.072) 0.004 (0.045) −0.780∗∗ (0.137) 0.261∗∗ (0.049)

Fruits

0.047 (0.054) 0.494∗∗ (0.082) 0.119∗∗ (0.036) 0.150∗∗ (0.056) 0.015 (0.074) 0.151∗∗ (0.041) 0.365∗∗ (0.074) 0.055 (0.039) 0.259∗∗ (0.062) −1.501∗∗ (0.219)

Other Foods

An Analysis of Food Demand in China 885

0.052 (0.035) −0.362∗∗ (0.049) 0.932∗∗ (0.106) 0.058∗∗ (0.012) −0.193∗∗ (0.025) −0.335∗∗ (0.075) −0.208∗∗ (0.025)

0.107∗∗ (0.043) −0.068∗ (0.039) 0.383∗∗ (0.082) 0.031∗∗ (0.012) −0.091∗∗ (0.017) −0.118 (0.074) −0.095∗∗ (0.017)

South

data.

0.134∗∗ (0.041) 0.035 (0.040) 0.189∗∗ (0.076) 0.011 (0.013) −0.051∗∗ (0.017) 0.007 (0.076) −0.053∗∗ (0.014)

Meats

a Estimated using 2004 China’s NBS household survey b Standard errors are given in the parentheses. ∗ , ∗∗ Significant at the 10% and 5% levels, respectively.

Ratio of FAFH spending

College

Ratio of children

Ratio of seniors

Household size

Town

Oils and Fats

Grains

Variable 0.407∗∗ (0.055) 0.063 (0.059) 0.177∗ (0.104) 0.008 (0.018) −0.047∗ (0.025) 0.017 (0.106) −0.077∗∗ (0.023)

Poultry 0.128∗∗ (0.052) 0.056 (0.048) 0.322∗∗ (0.084) 0.026∗ (0.015) −0.079∗∗ (0.021) 0.087 (0.089) −0.088∗∗ (0.017)

Eggs 0.174∗∗ (0.051) 0.103∗∗ (0.047) 0.093 (0.094) 0.014 (0.016) −0.067∗∗ (0.021) −0.026 (0.088) −0.036∗∗ (0.017)

Aquatic Products

Table 5. Demographic elasticities, urban Jiangsu province, China, 2004a

0.113∗∗ (0.058) 0.025 (0.058) −0.147 (0.111) 0.003 (0.021) 0.081∗∗ (0.026) −0.035 (0.156) 0.043∗ (0.023)

Dairy Products 0.203∗∗ (0.040) −0.044 (0.038) 0.216∗∗ (0.072) 0.026∗∗ (0.013) −0.076∗∗ (0.017) 0.012 (0.073) −0.093∗∗ (0.013)

Vegetables 0.231∗∗ (0.052) −0.011 (0.049) −0.071 (0.095) −0.010 (0.015) −0.004 (0.023) −0.019 (0.083) −0.002 (0.018)

Fruits

0.067∗∗ (0.032) 0.028 (0.028) 0.201∗∗ (0.066) 0.013 (0.010) −0.029∗∗ (0.013) 0.068 (0.058) −0.033∗∗ (0.008)

Other Foods

886 Review of Agricultural Economics

An Analysis of Food Demand in China

887

City size exhibits a significant effect on the demand for several food products considered. Relative to households located in cities, households located in towns (town, table 5) tend to consume more aquatic products at home; but less grains and oils and fats. This finding is reflective of the fact that consumers in towns have a more limited access to the varieties and qualities of foods that are offered to consumers in cities through FAFH channels. Given their small living quarters, it is very common for urban Chinese households to eat some of their main meals away from home in eating establishments. Therefore, the results of this study support the urban food culture in China that consumers in bigger cities tend to consume less meats and aquatic products at home than those in smaller towns, with all else being the same. Household size has a positive relationship on the demand for grains, oils and fats, meats, poultry, eggs, vegetables, and other foods. Because of the government’s policy related to the one child per couple rule in urban China, household size in urban China has decreased significantly over the past two decades, from 3.50 persons per household in 1990 to 2.98 persons per family in 2004. As household size continues to decrease, the consumption of grains, oils and fats, meats, poultry, eggs, vegetables, and other foods is expected to decrease, with all else being equal. As expected, households with more seniors (aged sixty-one and above, ratio of seniors in table 5) tend to consume more grains, oils and fats, eggs, and vegetables. Households with more children (aged seventeen and below, ratio of children in table 5) tend to consume more dairy products; but less grains, oils and fats, meats, poultry, eggs, aquatic products, vegetables, and other foods. Hence, the older generation has generally different dietary habits compared to younger people. This finding is also consistent with the findings related to Japanese seniors in a study by Hsu, Chern, and Gale. Variables associated with educational levels of household heads have a statistically significant effect only on the consumption of grains (at the 11% significance level) and oils and fats. Results show that households headed by a better educated person (college and above, college in table 5) tend to consume less grains and oils and fats. This result might be because higher educated households usually have higher living standards than other families in contemporary China. Consequently, those higher income households can afford to have more higher valued products in their diets. The ratio of expenditures for FAFH to total food expenditures (ratio of FAFH spending, table 5) is one of the variables that significantly influence demand for most of the food products considered. This variable has a statistically significant positive effect on the at-home consumption of dairy products; but a negative effect on the at-home consumption of grains, oils and fats, meats, poultry, eggs, aquatic products, vegetables, and other foods. According to a survey conducted by the Chinese Academy of Agricultural Sciences in 1999, the urban Chinese consumer consumes about 27% of pork, 37% of beef and mutton, 51% of poultry, 13% of eggs, 43% of aquatic products, and 4% of dairy products outside his/her home during the course of a year (Wang and Zhou, pp. 96–97).8 Given that consumers spend a higher proportion of their FAFH expenditures on foods of animal origin; as the proportion of expenditures on FAFH increases, the at-home consumption of grains, oils and fats, meats, poultry, eggs, aquatic products, vegetables, and other

888

Review of Agricultural Economics

foods is expected to decrease; while the at-home consumption of dairy products is expected to increase, with all else being equal. Elasticity Comparisons with Other Studies Table 6 presents a comparison of own-price and food-expenditure elasticities from this study with those that have used the NBS urban household survey data after 1995. The studies that have used post-1995 data are more relevant to this study, compared to those that have utilized data prior to 1995. While expenditure elasticities for oils and fats and poultry from this study, particularly oils and fats, differ from those reported by past studies (Gould and Villarreal; Liu and Chern; Yen, Fang, and Su; Zhang and Wang); expenditure elasticities for the remaining studied food categories from this study fall within the range of estimates as reported in other studies (listed above). The differences between the magnitude of elasticities reported in this study and those given by others might be mainly attributable to the differences in model specification, make-up of food categories in the demand system, and data used. In particular, this study employs the nonlinear generalized version of the AIDS model and corrects for endogeneity of the expenditure term. This study is therefore different from most past studies that have used the linear or nonlinear versions of the AIDS model, but have not taken into consideration the possibility of the existence of expenditure endogeneity. Own-price elasticities for meats, eggs, dairy products, vegetables, and fruits from this study are very similar to the estimates reported by the past studies (listed above). However, own-price elasticities for grains and oils and fats from this study are higher (in absolute terms) while own-price elasticities for poultry and aquatic products from this study is smaller in absolute value, compared to those reported by these past studies. While own-price elasticity for aggregate oils and fats from this study is higher (in absolute terms) than those reported by Gould and Villarreal; Liu and Chern; Yen, Fang, and Su; and Zhang and Wang, it lies within the range of estimates reported by Fang and Beghin for disaggregate oils and fats (−0.22 to −1.32). The relatively high own-price elasticity for grains in absolute value might be attributed to the unexpected high grain prices in 2004. According to China’s official statistics, triggered by a sharp rise in rice prices in the late 2003, the grain price index in 2004 was 26.4% higher than those in 2003; while the consumer price index for food in 2004 was only 9% higher than the previous year (NBS 2005). Therefore, the relatively high own-price elasticity (in absolute value) for grains may be a reflection of household reaction to the unusual grain market situation in urban Jiangsu in that particular year (2004). The consumption of poultry meats accounts for a notable percentage of household meat consumption in urban Jiangsu. About 35% of at-home meat consumption in urban Jiangsu during the period 2002–2004 is poultry meats, considerably higher than those for the national average (28%). Moreover, the three-year average (2002–2004) per capita at-home consumption of poultry meat is 12.5 kg per person in urban Jiangsu, while it is 8.9 kg per person for the national average. Thus, the low own-price elasticity for poultry category may be a reflection of the food preferences of consumers in Jiangsu. Similar situation has occurred in the consumption of aquatic products in urban Jiangsu. Given that the consumers in Jiangsu already allocate a higher percentage of their food expenditures to poultry and aquatic products compared to the national average (NBS 1991–2005), the

a Estimates

1.03 1.09 1.17 1.16 0.89 1.24 1.00 0.87 0.92

1.17 1.18 1.18 0.90 1.27 1.00 0.96 0.79

1.14 1.09

0.50 1.37

1.13

Liu and Chern

0.94 1.41 1.26 0.77 1.41 1.40 0.83 0.60

0.98

0.82

Yen, Fang, and Su

0.97 1.14 1.24 1.04 1.05 1.19 1.11 0.96

0.99

1.18

Zhang and Wang

from Gould and Villarreal are those for urban Jiangsu province in 2001.

Grains Rice Wheat Other grains Oils and fats Meats Pork Beef Poultry Eggs Aquatic products Dairy products Vegetables Fruits

Commodity

Gould and Villarreala

Expenditure Elasticities

1.00 0.82 1.20 1.37 0.81 0.98

0.72 1.04

0.79

This Study

−0.79 −0.92 −1.00 −0.91 −0.91 −0.83 −1.07 −0.83 −0.90

−0.67 −0.97 −0.90 −0.68 −0.71 −0.44 −0.64 −0.60

−0.86 −0.95

Liu and Chern

−0.21 −0.96 −0.75 −0.70 −0.37 −1.40 −0.72 −0.76

−0.85 −0.38 −1.07 −0.73 −0.85

−0.72 −0.28

−0.53

−0.75

−0.90

−0.55

Zhang and Wang

Yen, Fang, and Su

Own-Price Elasticities

−1.00 −0.72

−0.72

Gould and Villarreal

Table 6. Elasticity comparisons between this study and other studies

−0.35 −0.85 −0.10 −1.21 −0.50 −0.86

−1.31 −0.85

−1.22

This Study

An Analysis of Food Demand in China 889

890

Review of Agricultural Economics

prices of poultry and aquatic products might not play such an important role in consumer purchasing decisions.

Summary and Conclusions This study examines the impacts of economic factors (price and expenditures) and noneconomic factors (demographic characteristics) on food consumption patterns in China using the 2004 NBS urban household survey data for Jiangsu province. A complete demand system of households is estimated for ten major food products (grains, oils and fats, meats, poultry, eggs, aquatic products, dairy products, vegetables, fruits, and other foods) using a nonlinear generalized version of the AIDS model. Moreover, the endogeneity of expenditure term in the GAIDS model is corrected using the full information maximum estimation procedure. Finally, the CTS estimation procedure proposed by Shonkwiler and Yen is used to account for zero budget shares resulting from missing values. Two major findings of this study are summarized as follows. First, the results of this study clearly indicate that the changing demographic profile of urban consumers in Jiangsu has had a significant impact on food demand. The most significant demographic effects come from region (south vs. north), city size (towns vs cities), household size, the ratio of seniors to household size, the ratio of children to total household members, and the ratio of expenditures for FAFH to total food expenditures. Variables related to the educational levels of household heads have a significant impact only on the demand for food grains and oils and fats. Second, the relatively large positive and statistically significant expenditure elasticities for the ten food categories analyzed in this study imply that income has been a notable driving force behind the changing food consumption patterns in the urban Jiangsu province of China. If the current price structure remains constant, expenditures on each of the studied food products are expected to grow as household incomes rise. However, the amount of the increase in demand in response to an increase in food expenditures varies across products. The demand for foods of animal origin (such as meats, poultry, aquatic products, and dairy products) is expected to grow by a larger magnitude than the other food categories included in this study. The findings of this study have important implications for both China and U.S. agriculture. This study indicates that with increases in per capita food expenditures, the demand for foods of animal origin is expected to increase by a larger magnitude than the demand for food grains in urban China. As a result, as incomes rise, the demand for feed grains to be used in livestock production is expected to increase by a larger percentage than the demand for food grains. China’s cropping patterns have changed considerably in the past decade in response to the growth in demand for foods of animal origin and the consequent increase in demand for feed grains. Between 1996 and 2006, planted hectares of corn and soybeans have increased by 16.2% and 25.9%, respectively; while planted hectares of rice and wheat have decreased by 6.6% and 8.5% correspondingly (NBS 2008). However, China’s ability to increase production of feed grains, particularly corn, might be hindered by its limited land and water resources. Given that China’s economy is expected to continue to grow in the future, China will be expected to continue to demand more foods of animal origin and, consequently, China will likely have to

An Analysis of Food Demand in China

891

resort to imported feed grains in order to increase its livestock inventory in the future (Crook and Colby; Fan, Wailes, and Cramer; Hayes). As a major feed grain exporting country and a prominent trading partner, the United States is expected to play an important role in feed grain markets in China.

Acknowledgments This study was partially funded by the Hatch Project No. 02702 of the Oklahoma State University Agricultural Experiment Station and the Cooperative Agreement No. 58-8000-8-0086 from the ERS, USDA on Study of China Livestock Industry Structure. This research benefited from the constructive input of Fred Gale, senior economist with China team at ERS, USDA regarding food consumption trends in China. The opinions expressed in this article are those of the authors and not necessarily those of the U.S. Department of Agriculture.

Endnotes 1 Average per capita income for urban Jiangsu province in 2004 was 10,482 Yuan, compared to 4,754 Yuan for rural Jiangsu (NBS 2005). 2 In order to examine how comparable the expenditures of the selected households (902) are with the expenditures of the NBS sample households for Jiangsu (4,600) published in China’s statistical yearbook, we compared the means of expenditures on each of the ten studied food products calculated from the two data sets. The results show that per capita expenditures on each of the studied food products based on the 902 households are consistent with the 2004 NBS sample household expenditures for Jiangsu. Moreover, only food consumed at home is considered in this study. Food away from home (FAFH) is not included because of the unavailability of data. China’s NBS only publishes data on total FAFH expenditures. 3 The approach developed by Cox and Wohlgnant may lead to bias because of sample selectivity and simultaneity problems. However, for a ten-good demand system using a data set with 902 observations, it would be computationally difficult to perform quality adjustment simultaneously with the estimation of the underlying model parameters. Consequently, this study employs the traditional procedure to generate quality-adjusted prices. 4 The elasticity estimates are not invariant to the residual good selected when using this approach to accommodate the adding-up restrictions (Dong, Gould, and Kaiser). However, “if there is a natural choice for this residual good, invariance is not of primary interest.” (Yen, Lin, and Smallwood, p. 460). For this study, the last food category (other foods) fits into the residual good category. 5 The unconditional mean of dependent variables in the second step of CTS procedure is specified as E(si ) = (zi i )w(xi i ) + i (zi i ), where si represents the observed budget share, wi denotes the deterministic budget share, zi and xi are vectors of exogenous variables used in the first-step probit and the second-step share estimations, and i represents parameters in the second-step estimation. Note that the common variables (i.e., price and demographic variables) that are used in both the probit and share equations (first- and second-step estimations), are expected to affect the dependent variables through xi i , standard normal probability i (zi i ), and density i (zi i ). Therefore, the formulae for price and demographic elasticities for oils and fats and dairy products should measure the full effects of the changes in the common variables. Based on Su and Yen (p. 733), the marginal effect of a common variable zi in xi and zi is specified as:

(10)

∂w(xi i ) ∂ E(si ) = (zi i ) × + w(xi i )(zi i )i − i (zi i )i . ∂zi ∂zi

The price and demographic elasticities for oils and fats and dairy products (equations 11 and 13) are subsequently revised as the following: (11)

     

   N  pjcj p c  ci  M∗ ˆ i i j 1 − i , ˆ i + ˆ i × j j + i j − i  j + e iuj = i j − 1 −   ln( p ) +  × k j k qi M∗ M∗ pi q i wi k=1 (12)

ˆ i /q i }) × ({M/M∗ }) + i  ˆ i /wi , e i = ({1 − c i 

892

Review of Agricultural Economics

and  (13)

d e ik = 1 − i −

N  j=1

 i j ln( p j ) − i ln

M∗ P



 − i ×



c ik dk qi



   ˆi + ˆ i ik dk 1 − i ,  wi

ˆ i /q i if i = j and 0 otherwise;  ˆ i and  ˆ i denote univariate standard normal cdf where i j = −1 + c i  and pdf; i j s and ik s are the parameters of logarithm of prices and demographic variables on food ˆ i in the demand estimated in the first-step probit functions, respectively; and i is the parameter of  second-step estimation, which also represents the error covariance between error terms of the censored system of equations (the second-step estimation) and error terms in a binary indicator function (the first-step estimation) (for a more detailed discussion, see Shonkwiler and Yen). 6 To save space, results for the reduced-form regression and the probit regressions are not reported here; however, they are available upon request. 7 The hypothesis test on expenditure elasticities being different from 1.0 was conducted using a Krinsky-Robb evaluation method. Specifically, a distribution of 1,000 values of each elasticity estimate was generated using a bootstrapping procedure proposed by Krinsky and Robb. The proportion of observations in this distribution with values greater than 1.0 is the p-value associated with the onesided hypothesis test that each elasticity estimate is greater than 1.0 (see Tonsor and Marsh for a more detailed explanation). 8 The Chinese Academy of Agricultural Sciences survey was conducted in six provinces and includes 359 urban households. Although the survey may have suffered from typical problems related to surveys conducted in China (not necessarily being a solid representation of the entire urban population, and being unofficial and ad hoc) it did support the fact that FAFH spending is an important component in urban Chinese consumer’s budget.

References Alston, J.M., J.A. Chalfant, and N.E. Piggott. “Incorporating Demand Shifters in the Almost Ideal Demand System.” Econ. Letters 70(2001):73–78. Blundell, R., P. Pashardes, and G. Weber. “What Do We Learn about Consumer Demand Patterns from Micro Data?” Amer. Econ. Rev. 83(June 1993):570–97. Bollino, C.A. “GAIDS: A Generalized Version of the Almost Ideal Demand System.” Econ. Letters 23(1987):199–202. China’s National Bureau of Statistics (NBS). China Statistical Yearbook (in Chinese). Beijing: China Statistical Press, 1991–2005. ——. Abstract of Results for the Second Agricultural Census in China (in Chinese). Beijing: China Statistical Press, 2008. Cox, T.L., and M.K. Wohlgnant. “Prices and Quality Effects in Cross-Section Demand Analysis.” Amer. J. Agr. Econ. 68(November 1986):908–19. Crook, F.W., and W.H. Colby. “China’s Livestock Sector Growing Rapidly.” In Agricultural Outlook, pp. 15–18. Washington DC: U.S. Department of Agriculture, Economic Research Service, November 1998. Deaton, A., and J. Muellbauer. “An Almost Ideal Demand System.” Amer. Econ. Rev. 70(June 1980):312– 26. Dhar, T., J. Chavas, and B. Gould. “An Empirical Assessment of Endogeneity Issues in Demand Analysis for Differentiated Products.” Amer. J. Agr. Econ. 85(August 2003):1605–17. Dong, D., B.W. Gould, and H. Kaiser. “Food Demand in Mexico: An Application of the AmemiyaTobin Approach to the Estimation of A Censored Food System.” Amer. J. Agr. Econ. 86(November 2004):1094–1107. Fan, S., E.J. Wailes, and G.L. Cramer. “Household Demand in Rural China: A Two-Stage LES-AIDS Model.” Amer. J. Agr. Econ. 77(February 1995):54–62. Fang, C., and J.C. Beghin. “Urban Demand for Edible Oils and Fats in China: Evidence from Household Survey Data.” J. Compar. Econ. 30(December 2002):732–53. Fuller, F., J. Beghin, and S. Rozelle. “Consumption of Dairy Products in Urban China: Results from Beijing, Shanghai and Guangzhou.” Australian J. Agr. Resource Econ. 51(December 2007):459–74. Gale, F., and K. Huang. Demand for Food Quantity and Quality in China. Washington, DC: U.S. Department of Agriculture, Economic Research Report No. 32, January 2007. Gao, X., E.J. Wailes, and G.L. Cramer. “A Two-Stage Rural Household Demand Analysis: Microdata Evidence from Jiangsu Province, China.” Amer. J. Agr. Econ. 78(August 1996):604–13. Gould, B.W. “Household Composition and Food Expenditures in China.” Agribusiness 18(Summer 2002):387–407.

An Analysis of Food Demand in China

893

Gould, B.W., and H.J. Villarreal. “An Assessment of the Current Structure of Food Demand in Urban China.” Agr. Econ. 34(2006):1–16. Greene, W.H. Econometric Analysis. Delhi (India): Pearson Education, Inc., 2003. Guo, X., T.A. Mroz, B.M. Popkin, and F. Zhai. “Structure Change in the Impact of Income on Food Consumption in China, 1989–93.” Econ. Devel. Cult. Change 48(July 2000):737–60. Halbrendt, C., F. Tuan, C. Gempesaw, and D. Dolk-Etz. “Rural Chinese Food Consumption: The Case of Guangdong.” Amer. J. Agr. Econ. 76(November 1994):794–99. Hayes, D. “China’s Role in World Livestock and Feed Grain Markets.” Choices 1(First Quarter 1999):25– 28. Henneberry, S.R., K. Piewthongngam, and H. Qiang. “Consumer Food Safety Concerns and Fresh Produce Consumption.” J. Agr. Resource Econ. 24(July 1999):98–113. Hsu, H., W.S. Chern, and F. Gale. “How Will Rising Income Affect the Structure of Food Demand?” In China’s Food and Agriculture: Issues for the 21st Century,F. Gale, ed., pp. 10–13. Washington, DC: U.S. Department of Agriculture, ERS-AIB-775, 2002. Krinsky, J., and A.L. Robb. “On Approximating the Statistical Properties of Elasticities.” Rev. Econ. Stat. 68(November 1986):715–19. LaFrance, J.T. “When Is Expenditure “Exogenous” in Separable Demand Models?” Western J. Agr. Econ. 16(1991):49–62. Lewis, P., and N. Andrews. “Household Demand in China.” Appl. Econ. 21(June 1989):793–807. Liu, K.E., and W.S. Chern. “Food Demand in Urban China: An Application of A Multi-Stage Censored Demand System.” Paper presented at AAEA annual meeting, Montreal, Canada, pp. 27–30 July 2003. Park, J.P., R.B. Holcomb, K.C. Raper, and O. Capps, Jr. “A Demand Systems Analysis of Food Commodities by U.S. Households Segmented by Income.” Amer. J. Agr. Econ. 78(May 1996):290–300. Piggott, N.E., and T.L. Marsh. “Does Food Safety Information Impact U.S. Meat Demand?” Amer. J. Agr. Econ. 86(February 2004):154–74. Pollak, R.A., and T.J. Wales. “Estimation of Complete Demand Systems from Household Budget Data: The Linear and Quadratic Expenditure Systems.” Amer. Econ. Rev. 68(June 1978):348–59. Shonkwiler, J.S., and S.T. Yen. “Two-Step Estimation of a Censored System of Equations.” Amer. J. Agr. Econ. 81(November 1999):972–82. Shono, C., N. Suzuki, and H.M. Kaiser. “Will China’s Diet Follow Western Diets?” Agribusiness 16(Summer 2000):271–79. Su, S., and S.T. Yen. “A Censored System of Cigarette and Alcohol Consumption.” Appl. Econ. 32(May 2000):729–37. Tonsor, G.T., and T.L. Marsh. “Comparing Heterogeneous Consumption in U.S. and Japanese Meat and Fish Demand.” Agr. Econ. 37(2007):81–91. Wang, J., and Z. Zhou. “Animal Product Consumption.” In Grains in China, Z. Zhou and W. Tian, eds., pp. 87–107. Burlington, MA: Ashgate Publishing Company, 2005. Wu, Y., E. Li, and S.N. Samuel. “Food Consumption in Urban China: An Empirical Analysis.” Appl. Econ. 27(June 1995):509–15. Yen, S.T., C. Fang, and S. Su. “Household Food Demand in Urban China: A Censored System Approach.” J. Compar. Econ. 32(September 2004):564–85. Yen, S.T., K. Kan, and S. Su. “Household Demand for Fats and Oils: Two-Step Estimation of a Censored Demand System.” Appl. Econ. 34(September 2002):1799–1806. Yen, S.T., B. Lin, and D.M. Smallwood. “Quasi- and Simulated-Likelihood Approaches to Censored Demand Systems: Food Consumption by Food Stamp Recipients in the United States.” Amer. J. Agr. Econ. 85(May 2003):458–78. Zhang, X., T.D. Mount, and R.N. Boisvert. “The Demand for Food Grain in China: New Insights into a Controversy.” Agr. Resource Econ. Rev. 30(April 2001):1–9. Zhang, W., and Q. Wang. “Changes in China’s Urban Food Consumption and Implication for Trade.” Paper presented at AAEA annual meeting, Montreal, Canada, pp. 27–30 July 2003.