The inequality of nutrition intake among adults in China

ABSTRACT

This paper constructs a multidimensional Theil Index to estimate and decompose the inequality of nutrition intake among the adult population in China. Using 1991–2009 China Health and Nutrition Surveys (CHNS), this paper features two major findings. First, we show that the nutrition inequality has remained small in contrast to the large and increasing inequality in population income. Second, using Theil decompositions and Oaxaca–Blinder (O-B) decompositions, we find that, unlike income inequality, nature factors (such as age and gender) and regional factors play a more important role than socioeconomic factors in nutrition inequality. This finding provides a plausible explanation to the different time trends of nutrition inequality and income inequality. Moreover, it suggests that policies that aim to reduce the socioeconomic disparities may not automatically transfer to closing the gap in nutrition intake, which in turn is a potentially important determinant of population health and the long-term economic development.

KEYWORDS:

• Nutrition intake
• inequality
• Theil decomposition
• China

Research highlights

1. We construct a multidimensional measure to estimate and decompose nutrition inequality.

2. Nutrition inequality in China has remained small over the period 1991–2009.

3. Nature and regional factors contribute to the inequality more than socioeconomic factors. This may explain the different trends of nutrition inequality and income inequality in China.

4. Policies that reduce socioeconomic disparities may not be as effective in reducing nutrition inequality.

Acknowledgments

This study was supported by National Natural Science Foundation of China (Grant No. 71573003), Beijing Social Science Foundation Research Base Grant (No. 16JDLJB001) and Peking University School of Economics Research Seed Grant. The authors are responsible for all remaining errors.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. The 10 properties are: (1) monotonicity (MON), (2) symmetry (SYM), (3) normalization (NORM), (4) separability (SEP), (5) rank-dependent separability (RSEP), (6) weak ratio-scale invariance (WSI), (7) strong ratio-scale invariance (SSI), (8) weak translation invariance (WTI), (9) replication invariance (REP) and (10) restricted aggregation (RA).

2. The Pigou–Dalton Transfers Principle states that inequality should fall as income is transferred from a richer to a poorer person without modifying their relative ranks.

3. According to Cowell (1988), it is possible to have a change in distribution that does not change the mean of any group but increases the inequality within every group and reduces the overall Gini coefficient. For example, considering the change from [(Group A: 3, 3, 13) (Group B: 6, 7, 8)] to [(Group A: 1, 6, 12) (Group B: 6, 6, 9)]. It does not change the mean of Group A (6.333) or that of Group B (7), while the inequality in each group goes up (Group A from 0.3509 to 0.3860 and Group B from 0.0635 to 0.9520) and the overall inequality goes down (from 0.2750 to 0.2667).

4. For example, Akita (2003) decomposes China’s income inequality by regions (Western, Central, Eastern and Northeastern China) and finds that the between-group inequality is 26% of the total inequality; he then decomposes the within-region inequality according to provinces (using the two-stage Theil decomposition) and finds the introduction of more groups raises the between-group inequality to 36% of the total inequality.

5. For example, if being male is associated with 0.1 standard deviation increase in the consumption of nutrition, gender should create a gap of 0.1 standard deviation in nutrition intake between men and women. However, since men generally earn higher income than women, and higher income may lead to better nutrition, the observed nutrition disparity may be greater than 0.1 standard deviation.

6. According to Yun (2005), choosing different base groups in a categorical variable may reveal different results in a decomposition.

7. Since our paper is concerned with the relationship between nutrition intake and demographic and regional variables, it is not necessary to control for these variables in the calculation of the z-scores.

8. The Theil index can only be calculated on positive values. As a result, a total of 1,203 observations with zero or negative net income are dropped in the calculation of Theil indices of individual income. Specifically, 169, 198, 184, 127, 251, 181 and 93 observations are dropped in the calculation of the Theil index in years 1991, 1993, 1997, 2000, 2004, 2006 and 2009, respectively.

9. We carry out paired t-tests on the following null hypothesis:

$H_0:Theilindexoffatintake<Theilindexofnutrienti$

The results reject the null hypothesis with p-values less than 0.0001 for all other nutrients (carbohydrate, protein, CES nutrition intake indicator with zero beta and with unit beta).

10. Using data from Table 2, the correlations between the Theil index of fat intake and that of carbohydrate and protein intake are −0.0728 and −0.2702, respectively.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China [71573003] and Beijing Social Science Foundation Research Base Grant [16JDLJB001].

Notes on contributors

Shaofei Jiang

Shaofei Jiang is a Ph.D. student in economics at the University of Texas at Austin.

Xuezheng Qin

Xuezheng Qin is a professor and deputy dean in the School of Economics at Peking University, where he started his academic career after earning his Ph.D. in Economics from the State University of New York at Buffalo. His primary research interests include health economics, economics of human capital, and applied econometrics. He was a guest editor of China Economic Review and Emerging Markets Finance and Trade, and he also serves on the editorial board of China Health Review.