The Impact of Social Security Income on Cognitive Function at Older Ages Padmaja Ayyagari, University of Iowa David E. Frisvold, University of Iowa The research reported herein was performed pursuant to a grant from the U.S. Social Security Administration (SSA) funded as part of the Retirement Research Consortium. The opinions and conclusions expressed are solely those of the authors and do not represent the opinions or policy of the SSA or any agency of the Federal Government. The authors thank the SSA for funding support. The authors also thank Kristine Brown and participants of the Comparative International Research Based on the HRS Family of Data conference for helpful comments. Overview Are there modifiable factors or policies that reduce or delay cognitive decline and the onset of Alzheimer’s Disease? • Prior work: • Examines education, retirement age, childhood environment • Also finds that income is correlated with cognition (Cagney and Lauderdale, 2002; Lee et al., 2006, 2010) • Our paper: • Is there a causal impact of Social Security income on cognitive function? • Quasi-experimental approach using Social Security Notch • Results: Increase in SS income improves multiple dimensions of cognition 2 Why Might Income Affect Cognition? • Lower financial strain (Mendes De Leon et al., 1994) • Reduces stress and depressive symptoms, which are linked to cognition (Andel et al., 2012) • Better access to health care • Even among the elderly through Medigap (Fang et al., 2008) • Earlier detection (Valcour et al., 2000) • Earlier treatment (Petersen et al., 2005) and management of comorbidities (Daviglus et al., 2010) • Earlier retirement (Snyder and Evans, 2006) • Leads to faster declines in cognition (Rohwedder and Willis, 2010) • Overall effect of income is theoretically ambiguous 3 Background: Cognition, Age, and Social Security Income (HRS Data) 14 16 18 20 22 24 Total Cognition 65 68 71 74 77 High SS Income 80 Age 83 86 89 92 95 Low SS Income 4 Background: Cognition, Age, and Social Security Income (HRS Data) • Individuals with above average Social Security income have a total cognition score that is on average 0.9 units higher • This difference widens with age • But, does this represent a causal relationship? • Job complexity • Correlated with income • And with cognition (Finkel et al., 2009) • Childhood environment • Correlated with income (Almond and Currie, 2011) • And with cognition (Case and Paxson, 2009) 5 The Social Security Notch • Different cohorts with the same work history received different benefits • 1972: annual automatic COLA introduced • Benefits increased faster than inflation (double indexation) • 1977: COLA formula corrected for 1917 and later cohorts • Grandfathering provisions: benefit changes were permanent • 1915-1917 cohorts: largest benefits relative to trend • Prior research using the Social Security notch: • prescription drug use (Moran and Simon 2006), weight (Cawley, Moran, and Simon 2010), mental health (Golberstein 2015), mortality (Snyder and Evans 2006), labor supply (Krueger and Pischke 1992), living arrangements (Engelhardt, Gruber, and Perry 2005) and utilization of long term care services (Goda, Golberstein, and Grabowski 2011) 6 Methodology • Instrumental Variable Approach • Two stages: • First stage: Estimate effect of being born in 1915-1917 on SS income • Second stage: Estimate effect of predicted SS income (from the first stage) on cognition • Identifies the impact of changes in SS income due to the 1977 amendments on cognition • Based on the birth year of the primary Social Security beneficiary • Widowed or divorced females are excluded 7 Data: AHEAD • Data requirements: • SS income, cohorts born around 1915, cognition measures • 1993 wave of the Study of Assets and Health Dynamics among the Oldest Old (AHEAD) • Sample: 1901-1930 cohorts receiving at least $100/month of SS income • 4139 persons, 673 born 1915-1917 • Cognition measures • • • • • Serial 7: measures working memory, range 0-5 Word Recall: measures episodic memory, range 0-20 Mental Status: measures knowledge, language, and orientation, range 0-15 Total Cognition: overall measure, word recall + mental status, range 0-35 Normal, CIND, Demented (Crimmins et al., 2011; Hsu and Willis, 2013) 8 .1 Data: Histogram of 27-Point Cognition Score Normal CIND .06 .04 0 .02 Density .08 Demented 0 2 4 6 7 8 10 12 14 16 Cognition 18 20 22 24 26 27 Point scale constructed from the word recall, serial 7, and backwards counting tests. 9 Results • First stage: • Individuals born between 1915 and 1917 receive an additional $1,160 in annual Social Security benefits • Second stage: An additional $1000 of SS income results in • 2.2% improvement in the serial 7 (working memory) • • 1.1% improvement in mental status (knowledge, orientation, language) • 1.4% improvement in total cognition score (sig. at 10% level) 10 Results • Are these changes clinically meaningful? • An additional $1000 of SS income reduces the rate of “demented” individuals by 1.9 percentage points • Compared to the average rate (8.8%), this represents a 21% decline 11 Heterogeneity • Income may have heterogeneous effects at different parts of the distribution of cognition • At the 10th quantile of the total cognition score, a $1000 increase in SS income improves cognition by 0.2 units • At the 90th quantile, a $1000 increase in SS income improves cognition by 1.3 units • Individuals with better cognition benefit more from increases in SS income 12 Robustness/Validity • Flu pandemic in 1918 • Results are robust to excluding 1918 and 1919 cohorts • More comparable control groups • Results are robust to restricting the sample to the birth years 1905 to 1925 • Could estimates reflect cohort differences in cognition? • Height and educational attainment - well known to exhibit cohort differences and correlated with cognition • Should not be affected by the notch which only affected postretirement income • We find no effect on height or education 13 Conclusion • Our results suggest that increase in Social Security income increased cognition • A $1000 annual increase improved total cognition by 1.4 % • Results are based on 1901-1930 cohorts and 1977 SSA amendment, but still relevant for today • Proposed changes to SS • Chained CPI: would result in lower annual increases for recipients • Increase the minimum benefits for low-wage workers • Results also relevant for Medicare and Medicaid • Cognitive impairment is associated with significant health care spending 14