Cognition and SES Relationships Among the Mid-Aged and Elderly: A Comparison of China and Indonesia

John Strauss, Firman Witoelar, Qinqin Meng, Xinxin Chen, Yaohui Zhao, Bondan Sikoki, Yafeng Wang

NBER Working Paper No. 24583
Issued in May 2018
NBER Program(s):Development Economics

In this paper, we use a measure of fluid intelligence, an adaptive number series test, to measure that part of cognition for respondents in two developing countries: China and Indonesia, both with very low educated elderly populations. This test was specially adapted by us and our collaborators from measures used in the United States to better fit such populations. We also use a measure of episodic memory and one measuring mental state intactness and examine their distributions and then the socio-economic gradients associated with each, concentrating on gender differences and how those change as SES and variables measuring community development are added.

We find large variation in our cognition measures in both countries, even among those 60 and over with no schooling. We explore the bivariate socio-economic gradients for these measures, separately for different age groups: 45-59 and 60 and above. We find strong gender, education and rural-urban gradients. Of these, the education gradient is the strongest, followed by the rural-urban gradient. China has a stronger rural-urban gradient than Indonesia, which is associated with the hukou residential permit system in China.

We find a significant, negative multivariate differential for women, that is significantly larger in China than Indonesia. The gender differential in both countries is smaller for the mid-aged, 45-59, for whom the gender schooling differentials are smaller. The gender differential declines substantially, and the China-Indonesia differential disappears once we control for SES characteristics. Adding community measures related to mean schooling and asset levels does not affect the gender differential.

Schooling levels are monotonically and significantly related to higher levels of cognition for all three of the variables we use. The magnitudes of the schooling coefficients are relatively large. Higher log of household per capita expenditure (pce) is positively associated with cognition, more so in China. Other SES characteristics such as height, are also positively related to the cognition measures, again more strongly so in China. Rural respondents have substantially lower levels of cognition measures, with a significantly stronger gradient in China. Mean community level schooling and log pce are also positively related to cognition outcomes, especially for elderly women.

download in pdf format
   (5787 K)

email paper

Machine-readable bibliographic record - MARC, RIS, BibTeX

Document Object Identifier (DOI): 10.3386/w24583

NBER Videos

National Bureau of Economic Research, 1050 Massachusetts Ave., Cambridge, MA 02138; 617-868-3900; email:

Contact Us