��������������������������
Similarly strong patterns in the NLMS are shown for poor health in relationship to education and life expectancy , with a five year difference in life expectancy for those with a college education compared to those who have not completed high school ( Braveman , Egerter , & Mockenhaupt , 2011 ). While the associations between income and education and health outcomes are often stylistically characterized as consistent “ social gradients ” ( Adler et al ., 1994 ), and it is true that there is a generally consistent and monotonic pattern , the form of this pattern and its strength varies across time and place ( Lynch 2003 ; van Raalte et al ., 2012 ). Why would this be true ? One possibility is that viewing the apparent simplicity of the association leads one to ignore the complexity of the social , behavioral , and biological processes that underlie the patterns that are observed , all of which may change over time . Figure 2 presents some , but not all , of the possible candidate pathways , each of which is supported by empirical findings . For example , while single measures of income are often used in epidemiologic studies , clearly a single measure does not represent the ebb and flow of income associated with life stage and the rises and falls of the economy ( Duncan , 1996 ), and income trajectories may be important as well ( Johnson-Lawrence , Kaplan , & Galea , 2013 ). Further , disposable income is produced by income from work , investments , government transfers , etc ., all of which are adjusted by the tax system . The job that generates income comes along with associated resources and risks ( Benach , Montaner , & Santana , 2007 ), and access to that job reflects both educational attainment and degrees ( Ewert , 2012 ), moderated by macroeconomic factors effecting the economic returns on education ( Hout , 2012 ) and current and historical patterns of discrimination and steering along racial / ethnic and gender lines ( Pager & Shepherd ,
2008 ). All of this occurs within a community context with its associated patterns of job availability and creation ( Owens , 2012 ), spatial mismatch ( Gobillon , Selod , & Zenou , 2007 ), social capital ( Sampson , 2009 ), schools , food supply , recreational resources ( Braveman et al ., 2010 ), healthcare , and residential mobility ( Sampson & Sharkey , 2008 ). Intergenerational effects are also critical ( McGarry & Schoeni , 1995 ), as is the effect of early health on future SEP ( Maslow , Haydon , McRee , Ford , & Halpern , 2011 ).
������������������������
While epidemiology has long focused on environmental exposures by “ place ,” the last 30 years or so have seen an expanding focus on other characteristics of places and their effects on a variety of health outcomes ( Diez Roux & Mair , 2010 ; Kaplan , 1996 ; Kawachi & Berkman , 2003 ). Much of this work has focused on spatial proxies for “ neighborhoods ,” often measured at scales ( census tract , zip code , etc .) driven more by the availability of data than by theory , but generally tapping variation along socioeconomic , occupational , and social dimensions ( Auchincloss , Gebreab , Mair , & Diez Roux , 2012 ). An early example is the work of Haan , Kaplan , and Camacho who examined the association between poverty area residence and risk of death for Oakland , CA respondents in the Alameda County Study ( Haan et al , 1987 ). Study participants who lived in Federally-designated poverty areas had an approximately 50 % higher risk of death over the next nine years compared to those who lived in nonpoverty areas , even when a wide variety of socio-demographic , behavioral , and psychological factors were taken into account . Poverty area residence was also associated with changes over time in physical activity and perceived health , as well
91