1. Introduction
Since its reform and opening in 1978, China has experienced improvements in food consumption and nutrition intake. According to the China Nutrition and Chronic Diseases Status Report (2015) (Abbreviated CNCDSR in the following text), the average calorie intake of the Chinese population in 2012 was 2172 kcal (consisting of 301 g of carbohydrate, 80 g of fat, and 65 g of protein). Nutrient intake is vital for health promotion, social equity, and long-term economic development. Direct economic losses caused by malnutrition are estimated to be in the range of 3 to 5% of GDP in developing countries [
1]. Therefore, it is imperative to conduct research on nutrition intake and transition.
Debate regarding the relationship between calorie and income has been widely discussed in various studies [
2,
3]. Bouis and Haddad [
4] measured how household calorie intakes have changed with income and estimated elasticities, which range from 0.08 to 0.14. Tian and Yu [
3] reported that calorie intake increases with income growth, but with decreasing marginal returns. Aromolaran [
5] revealed that increasing women’s household income share decreases household per capita calorie intake in low-income households in rural southwestern Nigeria using instrumental variable (IV) methods. This finding implies that food calorie intake responds negatively to a reallocation of household income from men to women.
Some studies have documented the impact of economic openness (globalization) on nutrition transition. Rayner et al. [
6] found that trade liberalization can affect the food supply chain via factors such as food imports, exports, and FDI in food processing. They used FDI, supermarketization, and cultural change to illustrate complex linkages between trade liberalization and diet transition. Thow [
7] studied the complex relationship between trade policy and nutrition transition. On the one hand, by increasing the availability and affordability of processed food and animal products, trade liberalization policies could facilitate nutrition transition in developing countries [
7,
8]. On the other hand, the dietary patterns resulting from nutrition transition are associated with diet-related chronic diseases [
7]. This means that, although globalization has the potential to improve nutrition, some aspects of the globalization process may deteriorate human nutrition and health [
9]. Globalization has played an important role in changing energy consumption patterns, dietary intake, and resulting diseases in the world [
10]. However, most of these studies lack empirical analysis based on microdata. A number of related studies have explored the impact of economic openness on health with mixed findings. For example, Vogli et al. [
11] and Burns et al. [
12] found that FDI is positively associated with BMI and other health indices. However, Kawachi [
13] identified a negative relationship between economic globalization and labor health.
Moreover, the impact of economic openness on different individuals’ nutrition intake is often neglected. In particular, there is no study focusing on intrahousehold individual nutrition intake from the impact of economic openness perspective, as far as we know. Mussa [
14] found that intrahousehold nutrition inequalities are more pervasive than interhousehold ones. The evidence of intrahousehold bargaining power may explain the inequality to some extent [
15,
16]. There is a reallocation effect between family members within each household [
17,
18,
19,
20]. Typical examples include situations in which surplus laborers in rural households migrate to seek more off-farm income [
21] and regularly send their salaries home to support their families, which affects their nutrition intake, and mothers’ economic independence possibly benefiting their children [
18]. The intrahousehold reallocation effect may protect vulnerable people, such as the elderly, children, unhealthy, and disabled people. Parents distribute household resources to their children because of altruism [
22,
23]. Baeten et al. [
24] noted that seniors over 70 years old were still supported by family-based self-insurance mechanisms in rural China. However, different family roles may have different outcomes in nutrition reallocation. For example, Shimokawa [
23] found a strong gender bias against girls in cities while children and the elderly were both affected in rural areas when allocating intrahousehold calories. In previous studies, researchers analyzed this intrahousehold reallocation with a negative shock. For example, Carson et al. [
25] explained that intrahousehold reallocation of working hours between family members are used to reduce the potential risk induced by arsenic exposure. In contrast, our paper will test the effect with a positive shock (i.e., the impact of economic openness on intrahousehold nutrient intake).
We will utilize the CHNS data to analyze the effect of FDI on nutrient intake across various family roles to identify the different family roles’ heterogeneous nutrition intake responses to economic openness. Following previous studies, this paper will use key factors such as gender, marital status, family responsibilities, social connections, and social status [
26,
27] to study this topic. With this thorough analysis, more targeted food policies and economic development policies can be formulated. The paper proceeds as follows. The next section introduces empirical models and variables.
Section 3 presents data and descriptive statistics of the sample.
Section 4 provides econometric results with discussion, and the final section draws conclusions and policy implications.
2. Methods and Variables
It is believed that economic openness changes nutrition intake through both income and non-income effects. For the nutrient intake of the income effect, Braunstein and Brenner [
28] found that FDI could have a positive impact on individual income. Researchers have shown that the impact of economic openness on income varies across genders and regions. For example, Chen et al. [
29] noted that globalization could encourage female employment and reduce gender discrimination, which is beneficial for female income improvement. Rising earnings resulting from FDI openness may be beneficial to food consumption and nutrition. The non-income effect on nutrient intake mainly refers to the change in eating habits. Influenced by the Western diet with more fat, more animal products, and high-energy density foods, the Chinese traditional diet, with a focus on grain and plant products, has been changing [
30]. Urbanization and globalization may increase consumption of non-traditional foods, such as processed foods [
10,
31]. Dietary patterns are affected by price changes, production practices, and the presence of trade and markets in the developing world [
31]. FDI may also make more highly processed foods available to more people by lowering prices, establishing new purchasing channels, optimizing the effectiveness of marketing and advertising, and increasing sales [
9]. Therefore, we speculate that FDI-nutrient intake coefficients are significantly positive both in urban and rural areas.
Hypothesis 1:
DI openness would increase intrahousehold nutrient intake in both urban and rural areas.
Those with labor force roles could obtain higher income levels than other members, which means a higher bargaining power in the household. The intrahousehold reallocation effect on food consumption and nutrient intake may protect non-labor force members who are more vulnerable (e.g., children and the elderly). Therefore, labor force roles may consume more nutrients for both higher intrahousehold bargaining power and higher labor supply intensity under the impact of FDI openness. Another Hypothesis 2 is assumed as follows.
Hypothesis 2:
The positive nutrient intake coefficients of labor force roles are higher than other family roles in an open economy.
To study the regional effect, we will adopt subsample regression disaggregated by rural and urban areas. Roemling and Qaim [
32] noted that food choices, job types, and personal hobbies in urban areas are quite different from those in rural areas. Burggraf et al. [
33] separated rural and urban samples when analyzing nutrition transition in China and found considerable elasticity differences between them. This paper will not only split the sample by region, into urban and rural, but will further split the sample by family role into the husband sample, wife sample, son sample, and so on. The purpose of splitting samples by family role is to capture the different nutrition intake changes of each family role responding to economic openness.
(1) The family role regression
This study focuses on ten family roles based on the traditional Chinese family structure: husband, wife, father, mother, son, daughter, son-in-law, daughter-in-law, grandson, and granddaughter. Each family role’s sample regression is based on the equation below:
In Equation (1), subscript
i refers to a specific family role group. The term
is a time-invariant and group-specific unobserved term, and
is a random error term. Variable
N is the three-day average calorie intake of each family role group. For measuring nutrition, calorie intake quantity changes have been widely utilized [
34,
35]. Hence, calorie intake (calorie) is adopted as the core dependent variable here because it measures the energy provided from all nutrients and contains more information about the nutritional status [
36].
FDI openness is used as a proxy of economic openness. The accumulated FDI stock divided by GDP of each province and each sample year measures FDI openness. The variable has been converted by the official exchange rate, accumulated consumption price index (CPI) based on the year 2011, and annual depreciation rate (9.6%). The FDI stock variables and initial values are constructed following previous studies [
37,
38]. Exchange rate conversion is necessary since the unit of FDI is dollars while GDP is measured in CNY. The accumulated consumption price index (CPI) could eliminate price factors since FDI contains price fluctuations. According to the estimation of Zhang et al. [
39] of China’s provincial capital stock depreciation, FDI stock needs to get rid of 9.6% depreciation.
Figure 1 provides the trends in
FDI variable statistics across sample regions and years.
Age, education, and BMI are used to capture individual demographic characteristics [
14,
32,
40,
41]. In addition, gender and education level are the main variables that reflect the bargaining power of each family role [
15,
32]. Therefore, variables controlling individual and family characteristic include age (
Age), education (
Edu) level, body mass index (
BMI), family size (
Fsize), and total household income (
Hinc). The definition of family roles in our model automatically implies the gender characteristics, so, to avoid redundancy, it is not included here. The family size variable should be controlled since it measures intrahousehold resource availability and allocation issues about nutrition [
41]. We speculate that the family size variable could affect intrahousehold food consumption and nutrient intake.
The regional price factor involves a series of accumulated food price indices (
Pgrain,
Poil,
Pmeat,
Pegg,
Paquatic, and
Pvegetable). Food price reduction increases poor people’s access to food [
33] and has a positive impact on people’s nutritional well-being [
42]. Bhargava [
43] also documented that high food price led to an energy intake reduction. Thus, this study calculated the accumulated food price indices of grains, oil, meat, eggs, aquatic products, and vegetables to capture the price effect on nutrition. The accumulated food price is calculated by consumption price indices of each province in the base year 2011. To show the calculation process, this study takes the 1991 Beijing grain price index as an example. With the price index of 2011 set to 1, we first multiply the Beijing grain CPI across 1991 to 2010. Then, dividing the 1991 Beijing grain CPI by the multiplied results, we obtain the converted Beijing 1991 grain CPI at the base year 2011. The province-specific effect and time-fixed effect are controlled using dummies of twelve regions and eight waves of survey data in the sample period from 1991 to 2011 in China.
(2) The joint consumption regression
Intrahousehold decisions on food may affect each family member’s calorie intake [
44]. To examine the responses of intrafamily members’ nutrient intake under the impact of FDI, it is critical to consider the intrahousehold joint nutrient consumption decision. Within a household, people are related by blood and kinship. They live together, pool all or most of their income for living, and generally share the same food supply [
22]. Therefore, intrahousehold food is the main nutrition source for each individual member of a family. The FDI is used as the external shock in this paper to study nutrient intake responses. When we control the household total income, the response of each family member’s nutrient intake to FDI might be affected by two factors: (a) the family utility maximization function [
20] and (b) family members’ bargaining power reflected by factors such as individual income or education level [
16]. Senauer and Jacinto [
22] introduced two household economic models to incorporate both factors. One model is to maximize the household joint utility function under the budget constraint. The other is based on the bargaining effect to reconcile the differences across family members. In our model, the household income is used as the budget constraint condition. Education level and individual income are used to reflect bargaining power.
A multivariate regression model is employed to explore the nutrient intake variations of subgroups since residuals of each family member are correlated. The model is the same as the above equation, but the joint consumption decision is considered, instead of the individual’s decision [
45]. For each family member, the left side of the equation is the nutrient intake of all members of the family. Their personal characteristics (age, education level, and BMI index) are added to the control vector (
Z) on the right side of the equation. The remaining variables remain the same as in the family role’s regression. Correlation coefficients of residuals are also reported to capture the interrelation of different family members’ nutrient intake decisions. To further provide more detailed nutrient intake responses, we also use the consumption amount of protein (
Protein), fat (
Fat), and carbohydrate (
Carbo), instead of the calorie intake amount, in the regression equation as independent variables. The amount of calorie, fat, protein, and carbohydrates provided by the CHNS dataset is calculated based on the Chinese Food Composition Table [
46].
(3) The two-stage least squares (2SLS) regression for individual income effect
Here this study will analyze the individual income transmission effect of FDI on nutrient intake. FDI openness stimulates individuals’ earnings [
28], which are directly linked to their food consumption and nutrient intake. The intrahousehold joint consumption effect was another factor in analyzing the income effect, since one family member’s nutrient intake may be indirectly affected by another’s. A typical example is the couple’s interaction. For example, the remittances between husband and wife may generate an invisible income source [
47] that impacts the nutrient intake of both.
This study only focuses on prime-age adults who are between 18 and 60 years old when estimating the income effect. Based on the efficiency wage theory [
48], we use instrument variables (IVs) for individual income to perform regressions by 2SLS. For IVs of income, we use each household’s asset value, adding the wage rate for different local jobs of the working family members in the household. The household assets here include agriculture appliances (
Ktrans), professional appliances (
Kagr), and transportation carriers (
Kprof) in detail. The wage rate here includes the average daily wage of a male factory worker (
Wage_male), female factory worker (
Wage_female), domestic helper (
Wage_helper), construction worker (
Wage_const), and driver (
Wage_driver) monthly income in the 2SLS analysis. This wage information of the most common job types in China is obtained from the CHNS community level survey. The variables
Ktrans,
Kagr, and
Kprof are obtained at the household level, and the rest are at the community level. Those wage rates that relate to individual income depend closely on regional economic prosperity and are less relevant to personal nutrient intake decisions. The household business asset is adopted by Aromolaran [
5] to instrument women’s income when studying its impact on women’s nutrition intake. You et al. [
49] used the five-year average provincial annual growth rate of average wage (per worker) to deal with the endogeneity problem between income and nutrient intake. They illustrated that there is a strong link between personal wages and the provincial wage growth rate, and the provincial wage growth rate may not be related to personal nutrient intake decisions.
All variables are used in the logarithm form to level out the skewed distribution, except for dummies and food price indices, which are normalized.
Table 1 provides the detailed definitions of each variable.