Chinese Income Growth and Animal Product Consumption. A Demand Analysis

Abstract

In this study, National Bureau of Statistics of China (NBS) data from 1995 to 2012 was used to investigate the link between growing income and rising animal foods demand in China. Due to China’s population size, the latter has implications for mankind on a global scale, ranging from the acceleration of climate change to the exacerbation of health risks and threats to food security. It is therefore vital to understand to what extent the increasing prosperity of the world’s most populous country connects to higher demand for animal products. Two approaches were chosen to compute income elasticities: an OLS estimator and an Almost Ideal Demand System (AIDS), which lead to significantly different results. The AIDS model produced more reliable estimates, indicating a rise of poultry consumption parallel to increasing income, a rising share of income being spent on aquatic products, and decreasing demand for eggs for urban households. Due to the lack of out-of-home consumption data, these results likely understate the real demand.

1 Introduction

Significant challenges for mankind are exacerbated due to the consump­tion of foods from animal sources, i.e. pork, beef, mutton, poultry, fish, eggs and milk. First, animal agriculture propels climate change, as it causes the emission of a major share of global greenhouse gas (GHG) emissions. While beef and milk cattle largely contribute through their methane emissions, other land animals emit less, but still considerable amounts of CO2 equivalents. Global livestock accounts for 14.5% of human-caused GHG (FAO, 2013). A diet including aquatic products in­stead of meat is not as detrimental to the climate as is livestock, but more so than a vegetarian or vegan food composition (Scarborough et al., 2014).

Apart from their impact on the global climate, animal products affect

the availability of natural resources, and therefore food security. While 1 kg of paddy rice requires 2300 L of freshwater (Hoekstra and Chapagain, 2008) and 3 m2 of land (Gerbens-Leenes, 2006), the same amount of pork - providing less nutritional energy (Voedingscentrum, 1998) - causes 4850 L of freshwater (Hoekstra and Chapagain, 2008) and 9 m2 of land (Gerbens-Leenes and Nonhebel, 2005) to be used. With the prospect of continuing world population growth, an efficient use of agricultural resources is of interest for food security.

Third, multiple health risks for humans increase due to animal food consumption. About 60% of identified human diseases are transmit­ted through animals (Steinfeld et al., 2006) and many recently observed infections, such as bovine spongiform encephalopathy (BSE) or swine influenza, are transmitted by livestock (WHO, 2016a). To prevent the emergence of such pathogens, the regular use of antibiotics is common in animal farming. The latter has led to increasing occurrence of mul- tiresistant pathogens (Gilchrist et al., 2007), which the WHO classifies as ”one of the biggest threats to global health today” (WHO, 2016b).

Abbildung in dieser Leseprobe nicht enthalten

Due to its strong economic growth and large population, China is significantly involved in exacerbating the aforementioned challenges. As figure 1.1 and 1.2 show, both Chinese real income and animal product
consumption have increased over the past years. Taking a share of around 28% of all meat consumed worldwide in 2015 (OECD, 2016), the Chinese population already largely contributes to the impacts of animal agricul­ture. Concerning the effects on human health, it is noteworthy that the Health Ministry of China recently published guidelines recommending a reduction of per capita meat consumption by 50% (Health Ministry of China, 2016). Thus, China is both responsible for the threats invoked by demand for animal sourced foods, and itself interested in diminishing them.

Products

— Pork Beef& Mutton
— Poultry
— - Eggs

Aquatic

The goal of the present paper is to show the link between income growth and animal food consumption in China in the past years, i.e. to what extent an increase in individual net income caused larger amounts of animal products to be demanded. To this end, the questions to be answered are: (i) What are the recent income elasticities of animal foods consumed in China? (ii) How do they differ between rural and urban households? (iii) How did they evolve over time?

These research questions point to three important aspects of the re­search topic of this study. The first question posed allows to understand the relationship between income and animal foods at several places and at several points in time. The second question aims at considering the differences between these places. Further, the third question investigates differences over time to enable reasonable predictions of future demand. The results of this study will allow for a prediction of animal foods con­sumption on the basis of income data. Thus, they will enable the scale of environmental challenges and health risks imposed by animal prod­uct consumption to be defined clearer and reacting policies to be more precise. The rest of this study is structured as follows. In section 2, pre­vious research is reviewed, before section 3 introduces the data obtained. Then, the estimation methods are explained, followed by the description and discussion of the results. Section 6 concludes.

2 Review of Relevant Literature

There are three kinds of relevant literature for this study: research in­vestigating animal product consumption in China, research observing animal foods consumption worldwide, and research focusing on China's statistics. The worldwide perspective is important for this project as it shows the general relationship between economic growth and animal product consumption.

2.1 Literature on China

The analysis of Zhou et al. (2012) observes food consumption in China with NBS data from 1978 to 2010 and provides estimates for demand in 2020. Importantly, the authors do not differentiate between rural and urban consumers in their income elasticity estimates. The study shows rising consumption of animal products with rising income, as well as declining income elasticities. Their descriptive analyses are detailed, the econometric methods used to estimate income elasticities, however, are not sophisticated, as the authors state themselves.

Using Household and Income Expenditure Survey (HIES) data from 1991-2001, Ma et al. (2004b) found income growth was associated with an increase in poultry, fish and milk consumption of Chinese households. Crucially, the authors include food consumed away from home in their analysis, which they found to make a significant difference. Also, dif­ferences between Chinese regions are stressed. The investigation of the 1990s by Fuller et al. (2000) reports rising animal product consumption as well.

Hsu et al. (2001) engage in a hypothetical calculation utilizing NBS data from the year 2001 to predict China's rising consumption of animal sourced products. They argue that increasing urbanization can be ex­pected in accordance with the target of the Chinese government to lift it from 36% in 2000 to 50% in 2020 (Hsu et al., 2001). The authors apply the consumption pattern of animal products in 2000 to an urbanization rate of 50%, finding an increase of 5-16% in consumption of meat, poul­try, eggs and fish. Hsu et al. note that their urban consumption data may be incomplete, as the data used (China Statistical Yearbook 2011) possibly underestimates food consumption away from home. Further, the downside of this analysis is its use of a single cross-sectional data set. It is unclear, whether the year 2000 is representative for Chinese food consumption. The analysis conducted by Liu et al. (2009), which shows a dependency of meat consumption on income and the meat price, also features a data set based on consumer surveys conducted in a single year (2005).

Guo et al. (2000) investigated the importance of other factors than income on animal product consumption. In their observation of China Health and Nutritional Survey data of the years 1989-1993, they point to changes in diet composition which cannot be explained by higher incomes alone, as they state. Instead, the authors mention the possible influence of Western culture.

2.2 Literature on the World

Gerbens-Leenes et al. (2010) examine food consumption in 57 countries in 2001, as well as large-scale time series for parts of Europe. Their study confirms the trend observed by the aforementioned authors cov­ering China: higher incomes are associated with higher quantities of animal foods. York and Gossard (2004) find this in their analysis of cross-national variation in 132 countries too, although they also show that economic growth affects meat and fish consumption differently, de­pending on the region.

2.3 Literature on China’s Statistics

Both GDP and animal product consumption data from the NBS, which have been used by multiple authors, show inconsistencies. For instance, the production of meat, poultry and eggs in China rose by 209% between 1985 and 1996, while the feed grain consumption increased by only 88% in the same period . For 1996, NBS reported pork and poultry production may be 39% and 70% too high respectively (Fuller et al., 2000). Further, Ma et al. (2004a) explain that NBS data from 1988 to 1999 shows the supply to demand ratio of pork increasing from 1.15 to 1.77 and use several statistical tests to demonstrate this rise is implausible. Including away-from-home consumption and other sources in their data set, the authors manage to correct for the inconsistencies.

Regarding GDP measures, Rawski (2001) argues that growth from 1997 to 2001 was only one third as high as the NBS claims. The impos­sibility of reconstructing NBS data for Chinese household consumption in these years (Holz, 2004) underlines Rawski’s point. However, a recent examination of macroeconomic proxies for GDP does not show significant discrepancies with NBS data (Mehrotra and Paäkkonen, 2011).

3 Data

For this project, the issues of the ’’China Statistical Yearbook” from 1995 to 2012 were used, which is published by the NBS. It includes mul­tiple annual social and macroeconomic indicators, which are aggregated on a national and provincial level. The data of interest can be found in the Yearbook's section on living conditions, which the NBS collects through yearly surveys of urban and rural households, and in the section on prices. An urban household is part of the sample for three years, its rural counterpart for five years (China Statistical Yearbook, 2015). Its survey methods have become more and more sophisticated (Gale, 2002), leading to the use of stratified sampling of more than two million house­holds in the most recent issue (China Statistical Yearbook, 2015).

Several variables were analyzed for both rural and urban households. These are per capita animal food1 consumption, per capita income and consumer price indices (CPIs) for animal foods as well as a general CPI. The animal product CPIs were used to calculate 1995 prices of these com­modities, while the general CPI served to convert income to real income at 1995 prices. Unfortunately, only one CPI for meat and its products was available from the yearbooks, which lead to the use of one compound price for beef, pork, mutton and poultry. Another downside of the data concerns exclusivity of certain variables for urban households. The per capita consumption expenditure on individual animal foods could only obtained for the latter. It was utilized to compute expenditure shares.

4 Estimation Method

Due to the data restrictions mentioned above, the central findings of this study, income elasticities, are estimated with two separate approaches. First, urban and rural consumption data is used to calculate the elasticities by the means of an ordinary least squares (OLS) estimator. Second, the available urban expenditure data serves the estimation of a linear ap­proximate AIDS (LA/AIDS), which aligns better with classical demand theory. Thus, the procedure Ma et al. (2004b) used to analyze consump­tion in the 1990s in China will be applied to a more recent data set of urban households. Even though this two-step approach does not deliver sophisticated results for the whole data set, i.e. both urban and rural households, it does indicate a trend for all households, and, by not rely­ing on a single estimation method, allows to evaluate the effect of using a certain econometric method.

York, beef and mutton, poultry, aquatic product and egg consumption were obtained for urban residents, while the data available for rural households features the difference of not distinguishing pork, beef and mutton.

4.1 OLS-Estimation

The OLS estimator used by Zhou et al. (2012) was adopted to large parts, leading to the following form:

Abbildung in dieser Leseprobe nicht enthalten

where Qit is the quantity of an animal product i consumed in year t, ine is the real income per capita in year t, and pjt is the price of a substitute or complement j of animal food i in year t. This equation is applied to both urban and rural households. To include ^ as a covariate allows for an income elasticity function which converges to a1i with rising income:

Abbildung in dieser Leseprobe nicht enthalten

If a2 is negative, eit decreases with rising per capita income, which reflects microeconomic theory[1]. Equation 4.2 is derived from equation

4.1 through application of the following standard definition of income elasticity:

Abbildung in dieser Leseprobe nicht enthalten

The OLS model described so far allows to infer income elasticities from the consumption data available for this study. However, certain issues occur. As equation 4.1 is estimated for several years, the risk of serial correlation arises. Also, contemporaneous correlation between the different estimates of consumption is likely, as households probably do, for instance, take into consideration how much poultry they just consumed when deciding upon how much pork they are inclined to buy.

Moreover, the possibility of reverse causality between quantity of a product demanded and price of product arises. A rising price might cause decreasing demand, but a rising price might as well be due to an increase in demand. This possible endogeneity problem could cause biased OLS estimates. However, the Chinese food market is competitive (EU SME Centre, 2015, Kreab, 2013), which indicates that prices are predominantly influenced by production costs instead of a firm’s mark­up. Thus, a causality of prices for demand seems more likely than the other way around. The following model is less prone to problems such as the just mentioned.

4.2 LA/AIDS-Estimation

The AIDS procedure was established by Deaton and Muellbauer (1980), and is popular among multiple authors investigating household consump­tion in China (Han and Wahl, 1998, Huang and Bouis, 2001, Ma et al. 2004b). The following AIDS equation will be used for urban households only (as the necessary expenditure data is only available for the latter):

Abbildung in dieser Leseprobe nicht enthalten

where w denotes the share of expenditures spent on animal product i, pj indicates the price of animal product j, X is the total expenditure in all observed years, and P is a price index for all of these years. The Stone price index will be used to estimate P:

Abbildung in dieser Leseprobe nicht enthalten

which leads to the LA/AIDS. The latter constitutes a regression equa­tion system, which will be estimated using the seemingly unrelated re­gression (SUR) procedure, avoiding contemporaneous correlation. Then, the following equation is suitable to estimate expenditure (i.e. income) elasticities (Green and Alston, 1990):

Abbildung in dieser Leseprobe nicht enthalten

Other than the SUR estimation, the use of an AIDS has theoretical advantages. For instance, it allows for non-linear Engel curves (Moschini, 1998), thus allowing to estimate a relationship between income and food expenditure in which the latter increases less with increasing income. Also, the following standard theoretical restrictions can be tested with the AIDS (Deaton and Muellbauer, 1980):

Abbildung in dieser Leseprobe nicht enthalten

[...]