Table 4

Baseline OLS and IRT Analysis of Age, Gender and Education on Cognition

Independent VariablesMental Intactness (0–11)Episodic Memory (0–10)
OLSOLSIRTOLSOLS
Demographics(1)(2)(3)(4)(5)
    Age (+45)0.033
(0.056)		0.126***
(0.047)		0.058***
(0.019)		−0.002
(0.052)		0.112***
(0.049)
    Age square/100−0.067
(0.045)		−0.125***
(0.038)		−0.055***
(0.015)		−0.043
(0.043)		−0.121***
(0.040)
    Female−0.955***
(0.095)		−0.286***
(0.096)		−0.115***
(0.038)		−0.323***
(0.085)		0.095
(0.095)
    Urban0.438***
(0.085)		0.166***
(0.034)		0.377***
(0.083)
    Zhejiang0.958***
(0.086)		0.398***
(0.034)		0.179**
(0.085)
    Height0.112**
(0.005)		0.005**
(0.002)		0.007
(0.005)
SES
Education (illiterates omitted)
    Sishu/home school and below1.391***
(0.114)		0.584***
(0.045)		0.598***
(0.114)
    Elementary school1.582***
(0.118)		0.656***
(0.046)		0.804***
(0.116)
    Middle school1.941***
(0.133)		0.766***
(0.052)		1.464***
(0.129)
    High school and above2.399***
(0.152)		1.000***
(0.060)		1.671***
(0.147)
Log household PCE0.183***
(0.038)		0.097***
(0.017)		0.136***
(0.038)
Observations1,8731,8731,8731,6631,663
R–squared0.0920.3750.4030.0810.231

  • Note: Standard errors in parentheses.

  • ↵*** p < 0.01,

  • ↵** p < 0.05, * p <0.1. Columns 1–3 are models with Mental Intactness (on a scale of 0–11) as the dependent variable, while Columns 4 and 5 model Episodic Memory (the average of the ten word memory recalls (immediate and delayed)). Models 1, 2, 4, and 5 are based on Ordinary Least Squares (OLS) while Model 3 uses Item Response Theory (IRT). IRT models the response of a respondent of a given ability to each item in the test. The estimated coefficients from the Bayesian expected posterior model are listed.