Table 1.

Linear Regressions Predicting Memory

Variables	Cross-sectional associations (βs)								Longitudinal associations (βs)
	Memory performance				Self-rated memory				Memory performance				Self-rated memory
	M1	M2	M3	d	M1	M2	M3	d	M1	M2	M3	d	M1	M2	M3	d
Demographics
Age	−.33*	−.33*	−.32*		−.12*	−.12*	−.12*		−.24*	−.24*	−.24*		−.01	−.02	−.02
Age squared	−.09*	−.08*	−.09*		.05*	.07*	.06*		−06*	−.06*	−.07*		.00	.00	.00
Sex	.16*	.16*	.16*		.02	.02*	.03*		.08*	.09*	.09*		.02	.03*	.03*
Ethnicity (black)	−.14*	−.14*	−14*		−.06*	−.08*	−.08*		−.05*	−.05*	−.05*		−.03*	−.04*	−.04*
Ethnicity (other)	−.06*	−.06*	−.06*		−.02*	−.03*	−.03*		−.01	−.01	−.01		−.01	−.01	−.01
Education	.29*	.26*	.26*		.22*	.16*	.15*		.15*	.14*	.13*		.09*	.06*	.06*
Baseline performance	—	—	—		—	—	—		.44*	.43*	.43*		.53*	.49*	.49*
Personality
Neuroticism	—	−.06*	−05*	.23	—	−09*	−.07*	.43	—	−.03*	−.02*	.08	—	−.06*	−.05*	.22
Extraversion	—	−.02	−03*	.12	—	.10*	.09*	.45	—	−.02	−.03*	.01	—	.04*	.03*	.18
Openness	—	.05*	.05*	.21	—	.10*	.11*	.51	—	.02*	.03*	.07	—	.06*	.06*	.19
Agreeableness	—	.01	.01	.13	—	−07*	−.06*	.27	—	−.02	−.02	.03	—	−.04*	−.04*	.09
Conscientiousness	—	.06*	.06*	.27	—	.13*	.12*	.60	—	.03*	.03*	.10	—	.05*	.04*	.22
Health-risk factors
Hypertension	—	—	−.02*	.06	—	—	−.02	.10	—	—	−.00	.02	—	—	−.01	.05
Diabetes	—	—	−.03*	.09	—	—	−.00	.08	—	—	−.04*	.09	—	—	−.01	.05
Treatment for distress	—	—	−.03*	.15	—	—	−.08*	.32	—	—	−.02*	.08	—	—	−.02*	.12
Obesity	—	—	.02*	.03	—	—	.01	.03	—	—	.01	.00	—	—	.00	.02
Physical inactivity	—	—	−.03*	.12	—	—	−.03*	.18	—	—	−.01	.04	—	—	−.02*	.09
Ever smoker	—	—	.00	.02	—	—	−.03*	.10	—	—	.00	.01	—	—	−.02	.04
R 2 adjusted	.292	.305	.309		.077	.146	.153		.429	.431	.433		.316	.329	.331
R 2 change	.292*	.014*	.004*		.078*	.069*	.008*		.429*	.003*	.002*		.316*	.014*	.002*
Sample size	N = 13,140				N = 13,140				N = 10,565				N = 10,592

Notes: βs = standardized regression coefficients; d = Cohen’s d (absolute values), computed by comparing top and bottom quartiles of each personality trait and the absence/presence of each risk factor, controlling for the demographic covariates. A total of 13,987 participants had at least one valid measure of personality and memory at baseline. Given that not all participants completed both measures at follow-up, N ranged from 10,565 to 10,592 for longitudinal analysis. The effects were similar when the HRS sampling weights were used in the analyses. M1 = Model 1, which included the demographic factors as predictors on cognition. M2 = Model 2, which additionally included personality. M3 = Model 3, which additionally included the health-risk factors. Health-risk factors were coded 0/1 (i.e., absence/presence, respectively).

*p < .01.