The Society for Social Work and Research

2013 Annual Conference

January 16-20, 2013 I Sheraton San Diego Hotel and Marina I San Diego, CA

Re-Examining the Relationship Between Work and Depression Among Older People: Using Propensity Score Analysis

Schedule:
Friday, January 18, 2013: 3:00 PM
Nautilus 2 (Sheraton San Diego Hotel & Marina)
* noted as presenting author
Eunhee Choi, MSW, Ph.D. Candidate, University of Pittsburgh, Pittsburgh, PA
Background & Purpose: Older people often become depressed as they retire but those who keep working tend to show lower depression levels. Work seems to buffer depression symptoms among older individuals, but existing studies show inconsistent findings. This inconsistency in results mostly stems from a methodological limitation in sampling, called selection bias, a failure to have balanced samples between a treatment group (workers) and a control group (retirees). Random assignment is crucial in examining a causal relationship between work and depression among older individuals because these two groups significantly differ in various aspects, particularly health and financial status. Although random assignment is almost impossible for studies of social science, propensity score analysis, an alternative statistical procedure, can balance data when a true experimental design is unattainable (Guo & Fraser, 2010). Thus, this study aims to re-examine the causal relationship between work and depression among older people by using propensity score analysis. Activity theory and role theory provide this study’s theoretical framework explaining the beneficial impact of work on older individuals.

Methods: This study employs the Health and Retirement Study 2008 dataset. This study only analyzes respondents who are at or older than the conventional retirement age of 65 (total sample size=1,578) and divides them into two groups: 1) those who currently work either part/full-time or are self-employed and 2) those who have retired. The outcome variable, depression, is measured by eight items of CESD. Control variables include demographic variables (age, gender, race, marital status, and education), health, household wealth and income. The data is analyzed by three steps of propensity score analysis: 1) propensity score estimation (developing propensity scores by using conditioning variables), 2) nearest neighbor matching within caliper .25*SD (a matching process creating matched samples), and 3) a regression analysis (the final analysis based on the matched data).

Results: The analysis results on the matched data show significant differences in depression levels by gender, marital status, education, and health. Older people who are female (b=.221, p=.007), not married (b=-.546, p=000), less educated (b=-.041, p=007), and with poorer health (b=.530, p=000) are more likely to be depressed. Interestingly, depression levels do not statistically differ by household wealth and income. When the selection bias and the influences of control variables are ruled out, no statistical difference is found in depression levels between older workers and retirees.

Conclusion & Implications: After selecting bias is removed from the data, this study finds no difference in depression levels between older individuals who stay in the labor market after retirement age and those who have retired. This result implies that work does not significantly buffer depression symptoms as individuals retire from their life time career. Future research is required to further investigate whether the effect of work on depression may vary by different work circumstances, such as occupational differences, reasons for working, and having any caregiving duties.