Social Science & Medicine 153 (2016) 62e70
Contents lists available at ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
Poverty dynamics in Germany: Evidence on the relationship between persistent poverty and health behavior Katja Aue a, Jutta Roosen a, *, Helen H. Jensen b a b
Technical University of Munich, TUM School of Management, Chair of Marketing and Consumer Research, Alte Akademie 16, 85354, Freising, Germany Iowa State University, Department of Economics, Heady Hall, Ames, IA, 50011-1070, USA
a r t i c l e i n f o
a b s t r a c t
Article history: Received 18 October 2015 Received in revised form 24 December 2015 Accepted 24 January 2016 Available online 29 January 2016
Previous studies have found poverty to be related to lower levels of health due to poor health behavior such as unhealthy eating, smoking or less physical activity. Longer periods of poverty seem to be especially harmful for individual health behavior. Studies have shown that poverty has a dynamic character. Moreover, poverty is increasingly regarded as being a multidimensional construct and one that considers more aspects than income alone. Against this background this paper analyzes the relationship between health behavior and persistent spells of income poverty as well as a combined poverty indicator using data of the German Socio-Economic Panel (2000e2010). Next to cross-sectional logistic regression models we estimate fixed-effects models to analyze the effect of persistent poverty on dietary behavior, tobacco consumption, and physical activity. Cross-sectional results suggest that persistent poverty is related to poor health behavior, particularly regarding tobacco consumption and physical activity. Results also show that multidimensional and dynamic aspects of poverty matter. Complementary panel analyses reveal negative effects for the combined poverty indicator only for dietary behavior in the total sample. However, by analyzing the sample by gender we identify further effects of persistent poverty on health behavior. The analyses show that not only do individuals in poverty but also those in precarious situations show health-damaging behavior more often. © 2016 Elsevier Ltd. All rights reserved.
Keywords: GSOEP Health behavior Multidimensional poverty Panel analysis Persistent poverty
1. Introduction Studies in the area of health economics and public health have shown that poverty and low socio-economic status (SES) are related to lower levels of health (Benzeval and Judge, 2001; Cohen et al., 2003; Helmert, 2003; Mackenbach et al., 2008; Mielck, 2000). Inequality in healthy life expectancy can be observed. For example, rates of premature mortality are higher among those with lower levels of education, occupational status or income. Rates of morbidity are also higher (Lampert and Kroll, 2009; Mackenbach, 2006). Attempts to explain these differences have often made reference to the observation that poor health behavior such as unhealthy dietary behavior, smoking, or physical inactivity clusters in poverty groups or for those with a low SES (Contoyannis and Jones, 2004; Lynch et al., 1997a). For instance, McGinnis and Foege (1993) have
* Corresponding author. E-mail address: [email protected] (J. Roosen). http://dx.doi.org/10.1016/j.socscimed.2016.01.040 0277-9536/© 2016 Elsevier Ltd. All rights reserved.
shown that approximately 38% of all deaths in the US were caused by behavior-related factors. Also Mokdad (2004) confirmed this relationship for the United States in 2004. Likewise in Europe, the World Health Organization (WHO) (2002) reported that the total burden of disease in Europe is considerably influenced by health behavior and by poverty and income inequalities. Concerning poverty dynamics and health there are only a small number of longitudinal studies available. Furthermore, the available research shows that persistent poverty is more important than current income (Benzeval and Judge, 2001). Health behavior (in comparison to health outcomes), however, has been only studied in a few cases (c.f. Lynch et al., 1997b; Smith and Middleton, 2007; Smith and Zick, 1994) which underlines the need for studies in this area. Furthermore, to the best of the authors' knowledge, there is no study looking at health behavior and the dynamics of multidimensional poverty. Against this background the objective of this paper is to empirically examine the relationship between dichotomous and multidimensional persistent poverty measurements (at-riskpoverty-rate vs. the combined poverty indicator by Groh-Samberg,
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
2008) and health behavior. It is the first time that such a multidimensional indicator is used for analyses in the area of health behavior research. Our analyses are based on the conceptual framework by Mackenbach (2006) outlined in Section 2.1. We use five years of observations of health behavior variables from the period 2004e2010 in the German population using data of the German Socio-Economic Panel (GSOEP) and combine these data with information on persistent poverty status. Health behavior dimensions include dietary behavior, smoking and physical activity. Poverty is measured by two measures: relative income poverty and a combined poverty indicator by Groh-Samberg (2009). The data are analyzed using a cross-sectional approach as well as a panel data estimation. The results reveal that considering multidimensional and dynamic aspects of poverty matters. Both poverty indicators show that persistent poverty is negatively associated with health behavior. The paper is organized as follows. The next section will give some background on explaining health behavior and measures of poverty. Section 3 then presents the data and methods of analysis. Results are discussed in Section 4 before the paper concludes. 2. Background 2.1. Explaining health behavior Health behavior can be explained by several theories including economic and social sciences. An important model explaining health inequality was developed by Mackenbach (2006) and is depicted in Fig. 1. Mackenbach considers health behavior as one of three mediators between SES and health. Traditional components of SES are income, education and occupation (Adler and Ostrove, 1999). Hence health behavior is influenced by socio-economic parameters directly as well as indirectly via material and psychosocial factors. Health in turn is determined by health behavior, material factors and psychosocial factors. Material factors influence the health status directly as well as via psychosocial factors and health behavior. Because these different determinants are heterogeneously distributed across socio-economic groups, they are seen as the main explanation of health inequality. In this model, specific health determinants comprise the main explanation of health inequality. The group of material factors, including poverty, considers financial aspects, especially the income situation which influences psychosocial aspects such as stress, subsequent risk-taking
63
health behavior as well as the access to health-promoting facilities and products. These types of psychosocial stress may lead to illhealth either through biological or behavioral pathways. Further material factors are health risks related to occupation and housing. In this context, it should be noted that health behavior may respond immediately to deprivation whereas the health status develops over time and may be the result of long-term effects of health behavior. This study therefore focuses on individual health behavior as a key factor in Mackenbach's model (Kroll, 2010; Knoops et al., 2004; Olshansky and Ault, 1986; Osler, 2006). Against this background we hypothesize that persistent poverty is associated with detrimental health behavior (cross-sectional analysis) and that it increases the likelihood of such behavior (panel analysis). 2.2. Defining poverty Poverty is still present in many developed countries like Germany. While most poor individuals are not affected by physical deprivation or hunger, relative poverty, mostly defined by low income status, still concerns many to this day (e.g., Duncan et al., 1993; Eurostat, 2011). There exist several approaches to the definition of poverty in developed countries, however, there is no universally valid definition. Definitions are based on absolute and relative concepts as well as subjective approaches (Wagle, 2002). Despite the fact that the definition of relative poverty is difficult and normative (O'Boyle, 1999), our study focuses on approaches of relative poverty. In that we follow the European Commission that defines people as poor “[…]if their income and resources are so inadequate as to preclude them from having a standard of living considered acceptable in the society in which they live […]. They are often excluded and marginalised from participating in activities (economic, social and cultural) that are the norm for other people and their access to fundamental rights may be restricted” (European Union, 2010). Altogether, income-based measurements are widely used to describe income poverty or the at-risk-of-poverty rate (Nolan and Whelan, 2007). For example, in the European Union the at-risk-poverty-rate is one of the so-called Laeken indicators and defined to be at 60% of median net-equivalence income (Dennis and Guio, 2003). Nevertheless, information on income may be insufficient to determine the degree to which a person is at risk of deprivation. Some households are able to maintain an acceptable standard of living although they are on a low level of income, either because income poverty is only temporary or because of other
Fig. 1. Explanation of health inequality. Solid lines describe the causation hypothesis. The dashed line represents the explanation of natural or social selection that was included in Mackenbach et al. (1994) first model. This aspect is no longer included in the updated model from 2006. Source: modified according to Mackenbach et al. (1994) and Mackenbach (2006).
64
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
resources such as savings or gifts. To capture the multifaceted nature of poverty, the British sociologist Peter Townsend provided a definition based on concepts of deprivation and social exclusion and introduced a deprivation index that describes poverty as a multidimensional phenomenon rez-Mayo, 2004). The index surveys whether (Townsend, 1979; Pe essential items, namely goods or practices of everyday life, are missing due to financial restrictions. A household is accordingly defined as deprived if a given amount of items is missing, where the necessary amount is oriented on the common standard of living in society. There have been several attempts to describe multidimensional poverty since Townsend developed the deprivation in€d, 1994, 1995; Mack and Lansley, dex (Berthoud et al., 2004; Hallero rez-Mayo, 2004; Whelan and Maître, 2005). 1985; Muffels, 1993; Pe Since many researchers propose to supplement the measurement of income poverty with direct measurements of standard of living and of life domains (Benzeval and Judge, 2001; GrohSamberg, 2009; Moll, 2006; Nolan and Whelan, 2007), this study focuses on an existing multidimensional measurement that considers not only income but also additional aspects of deprivation €d, 1995; Pe rez-Mayo, 2004; Whelan et al., 2001). (cf. Hallero Therefore the index combines income poverty with four life domains: housing, consumption, financial reserves and unemployment (Groh-Samberg, 2008). As a result, a household may be deprived in the income and material dimensions or poverty may be inconsistent, e.g. a household is affected by multiple deprivations €d, 1995). but has adequate income or vice versa (cf. Hallero Since poverty is not a static phenomenon, dynamic aspects should also be considered (Ashworth et al., 1994; Duncan et al., rez-Mayo, 2004). In the 1990s the research on poverty 1993; Pe dynamics gained in importance in Europe following the research ideas being generated in the United States (Leisering, 2008). In Germany, Groh-Samberg (2008) stated that poverty becomes more persistent over time. In addition, other statuses such as a precarious living condition or temporary poverty are also observable over time. 3. Data and estimation 3.1. Data description We use data from the German Socio-Economic Panel (GSOEP) to examine the relationship between multidimensional aspects of poverty and deprivation and health behavior. The GSOEP is an ongoing panel survey of households and individuals conducted annually since 1984, and is representative of the resident population of Germany. Further details on design and methodology of GSOEP have been described elsewhere (Wagner et al., 2007). Individuals report repeatedly on an annual basis. The following analyses use the 2000e2010 waves. The panel data design is unbalanced to consider as many individuals as possible. 3.2. Measurement of health behavior We analyze three types of health behavior: dietary behavior, physical activity and smoking. Dietary behavior is measured on a four-point scale by the question: “To what extent do you follow a health-conscious diet?” Since there is evidence that there is a direct association between self-reported healthfulness of diet and actual dietary quality, this question is used to represent the dietary behavior within the German population (Basiotis et al., 1995). Answers are aggregated to a binary variable: healthy diet (¼1, includes answers 1 and 2) and no healthy diet (¼0; includes answers 3 and 4). Smoking is associated with higher risk for cardiovascular diseases and several
forms of cancer (Krueger and Chang, 2008; Moore and Hughes, 2001). It is measured in the data set by counting the number of cigarettes/day (pipes and cigars are counted as two cigarettes) (Mueller and Heinzel-Gutenbrunner, 2001). We summarize the responses in a dichotomous variable indicating whether the respondent is non-smoker (Y ¼ 1) or smoker (Y ¼ 0). Physical activity is defined as sufficient if the respondent does sport at least once per week or more often (Y ¼ 1), which is near the recommendations of the Robert Koch-Institute (2005) for Germany and the Physical Activity Guidelines for Americans (Centers for Disease Control and Prevention, 2008). 3.3. Independent variables 3.3.1. Income poverty In the European Union, income poverty is often defined as the at-risk-of-poverty rate using a threshold of 60% of median netequivalence income based on the OECD modified equivalence scale (Atkinson et al., 2002; Dennis and Guio, 2003). Since persistent income can be represented by the average of income over time, the mean of the equivalized net household income (ENIt) is computed over five years (Goebel et al., 2008; Hagenaars et al., 1994), so that for year t:
ENI EU;t ¼
4 X ENIEU;ts s¼0
5
(1)
The period over five years was chosen in accordance with previous panel analyses: on the one hand it allows considering as many survey years as possible and on the other hand it avoids a high rate of panel attrition that may be expected, especially for individuals affected by poverty (Moll, 2006). A household is considered to be affected by persistent income poverty if ENI EU;t is smaller than 60% of the median equivalized net household income. 3.3.2. Combined poverty index by Groh-Samberg Different from the European definition of income poverty, GrohSamberg uses the “old” OECD equivalence scale (OECD, 1982). Comparable to equation (1) the persistent equivalent net household income in year t following Groh-Samberg (2009) (ENIGS,t) is computed as follows:
ENI GS;t ¼
4 X ENIGS;ts s¼0
5
:
(2)
Groh-Samberg defines three income situations with regard to the mean of the ENI: “income poor” (75% of mean) assuming that a mean is more suitable to consider changes for higher incomes than the median (Groh-Samberg, 2009). In addition to income poverty, the four deprivation dimensions considered are housing, consumption, financial reserves and unemployment. They also refer to the household level. Housing deprivation includes insufficient room and a lack of basic equipment. Consumption is aggregated on a scale of commodities. This scale includes a large number of items such as owning a car or TV. A deprivation threshold of one standard deviation below the index mean is applied. To determine deprivation in the area of financial reserves households are regarded as deprived if they have no assets and no significant savings at all. Finally unemployment is a state of deprivation too, because it can be seen as one of the most important non-monetary dimensions of social exclusion and lowers life satisfaction substantially (Groh-Samberg, 2008).
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
65
Combining income poverty with deprivation measurements results in nine combinations. Altogether six groups, namely extreme, moderate and one-sided poverty as well as vulnerability, fragile prosperity and secure prosperity are specified (Table 1). To derive a combined index of persistent poverty, multiple deprivations over five years can be observed if the sum of deprivations is greater than 8 (z2 deprivations/year), a single deprivation exists if the sum of deprivations is less than 8 and greater than 2 (z1 deprivation/year), and no deprivation if the sum of deprivations does not exceed 2 deprivations over five years (GrohSamberg, 2008).
4. Results
3.4. Multivariate analyses
4.1.2. Income poverty For all relevant five-year periods, Table 3 shows the respective number of observations, the percentage of the population at risk of poverty and the poverty line. In addition, we report the data for individual years at the top of the table. Although at-risk-of-poverty is defined at the household level, the data are reported for observations of individuals to match the analyses reported below. Table 3 shows that the share of individuals whose equivalized net-household income is below the at-risk-of-poverty threshold of 60% of median net equivalence income increases gradually from 11.85% in 2000e16.17% in 2010. At the same time, the threshold increases over the years from 10,222 Euros to 12,185 Euros. Looking at measures of persistent poverty, the table reveals that more than 10% of the GSOEP population is affected by persistent income poverty.
Firstly, dietary behavior, smoking and physical activity of individual i are analyzed as dependent binary variables (Yi 2 {0, 1}) using information on persistent income poverty and the combined poverty indicator in the respective year as explanatory variables in logistic regression models. Since dietary behavior and smoking are measured in even years and physical activity in uneven years we use GSOEP waves of 2009 and 2010. To interpret the results we use odds ratios (OR). Secondly, panel data analyses are conducted for the three types of health behavior using information on persistent poverty as explanatory variables in logistic regression panel analyses. According to the data availability in GSOEP, panel analyses regarding dietary behavior and smoking are conducted for the period 2004 to 2010, whereas analyses on physical activity are conducted for 2005 to 2009. As evident in Table 2, each health behavior is surveyed four times in these time spans. We estimate fixed effects logistic models. Fixed-effect models of binary dependent variables have two specific characteristics. First, the logistic regression analyses use only within-person variation, so that individuals without variation in the dependent variable are excluded from the sample. Second, the fixed-effects models do not produce any parameter estimates for the time-invariant explanatory variables (e.g. gender and migration background) that are confounded in the constant term.
3.5. Confounders Models are adjusted for subjective health status, education, (un) employment, occupational status, gender, age, marital status, migration background, region of residence (former East/West Germany) and number of children (0e14 years living in a household). Following Mackenbach (2006) economic worries are also considered as measurement for psychosocial stress regarding financial aspects. Furthermore we adjust our models for risk preferences (cf. Anderson and Mellor, 2008). Since health-related behavior varies by gender, analyses are conducted not only for the total population but also separated by gender.
4.1. Descriptive statistics 4.1.1. Health behavior Table 2 shows the distribution of health behaviors for the total sample and by gender during the period 2004e2010. Information on healthy diet and non-smoking is collected in 2004, 2006, 2008, and 2010; information on physical activity in 2005, 2007, 2008, and 2009. Generally, more women behave in a health-promoting way than do men, so that analyzing health behavior by gender is necessary (cf. Wardle et al., 2004).
4.1.3. Combined poverty index by Groh-Samberg Results for the concurrent and persistent combined poverty index are given in Table 4. As in the case of the at-risk-of poverty rate, poverty in single years is more common than persistent poverty. Compared to the persistent at-risk-of poverty rate that shows an average of 10.2% to 13.6% of the German population at risk of poverty, the persistent combined poverty indicator makes different aspects of poverty visible. Extreme persistent poverty (income poverty and multiple deprivations) is observed for less than 5% of the sample for most periods. Fragile prosperity and secure prosperity is achieved by about 80% of the sample. Low income and single deprivations (vulnerability) go together more often (about 10% of the population) than being opposed to each other and resulting in one-sided poverty (around 2% of the population). 4.2. Logistic regression models of diet and smoking (2010) and physical activity (2009) We first present results for the cross section analyses based on the persistent at-risk-of-poverty rate and then for the combined poverty indicator. All tables first show the results for the entire sample and then by gender. Additional results for concurrent poverty are available in the online appendix.
Table 1 Characteristics of the combined poverty index by Groh-Samberg. Income
Deprivation Multiple deprivations (2)
Single deprivation (¼1)
No deprivation (¼0)
Income poverty (75% of mean)
Extreme poverty Moderate poverty One-sided poverty
Moderate poverty Vulnerability Fragile prosperity
One-sided poverty Fragile prosperity Secure prosperity
Source: Groh-Samberg (2008).
66
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
Table 2 Distribution of health behavior in the GSOEP population (%)*. Health behavior
Sample
Healthy diet
Total Male Female
Non-smoking
Total Male Female
Sufficient physical activity
Total Male Female
*Weighted results,
#
Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No
2004
2005
2006
2007
2008
2009
2010
45.44 54.56 36.07 63.93 54.58 45.42 64.60 35.40 60.45 39.55 68.64 31.64
#
44.95 55.05 35.12 64.88 54.35 45.65 64.30 35.70 61.17 38.83 67.29 32.71
#
#
43.68 56.32 32.54 67.46 54.23 45.77 66.83 33.17 62.87 37.13 70.56 29.44
#
33.72 66.28 33.36 66.64 34.07 65.93
#
35.49 64.51 34.25 65.75 36.67 63.33
44.98 55.02 35.01 64.99 54.51 45.49 66.42 33.58 62.65 37.35 70.02 29.98 40.67 59.33 39.67 60.33 41.63 58.37
# # # # #
# # # # # # # # # # #
# # # # # # # # # #
# # # # #
# # # # # # # # # # #
37.64 62.36 38.77 61.23 36.45 63.55
# # # # # #
Data not available.
4.2.1. Persistent at-risk-of-poverty rate Results presented in Table 5 show the logistic regression models using the persistent at-risk-of-poverty rate for 2006e2010 in the regressions regarding dietary behavior and smoking and for 2005e2009 for physical activity. For the categories diet and smoking, odds ratios are not significantly different from one for individuals who are persistently at risk of poverty compared to the reference group of prosperity. In contrast, persistent risk of poverty leads to a lower likelihood of being sufficiently physically active (OR ¼ 0.795). However, these results are only significant for females (OR ¼ 0.702), but not for men (OR ¼ 0.987) in the subsample analyses. Results underline that persistent risk-of-poverty affects health behavior in different ways, potentially because of gender-specific coping mechanisms with being at the risk of poverty.
Table 3 Concurrent and persistent at-risk-of-poverty rate based on ENIEU (%). Study period
#
Sample size (N)
Measures for individual years 2000 19,498 2001 17,984 2002 19,530 2003 18,495 2004 18,081 2005 17,384 2006 18,393 2007 17,228 2008 16,092 2009 16,878 2010 15,119 Measures for five-year periods 00e04 14,298 01e05 13,806 02e06 13,165 03e07 13,864 04e08 13,128 05e09 12,246 06e10 12,284
At-risk-of-poverty rate (%)
Threshold (Euro)
11.85 12.26 12.76 13.43 14.50 15.27 16.15 14.95 15.88 15.70 16.17
10,222.33 10,522.00 10,566.60 10,878.57 10,971.00 10,965.71 11,045.43 11,101.20 11,594.29 11,827.30 12,185.10
10.17 11.12 11.49 12.64 12.79 13.06 13.59
10,326.24 10,555.87 10,907.66 10,986.06 11,056.52 11,238.41 11,579.46
4.2.2. Persistent combined poverty indicator Results are presented in Table 6. Compared to Table 5, the sample size is reduced due to missing data on deprivations. First, dietary behavior differs significantly in all groups from that of the
Table 4 Concurrent and persistent poverty based on combined poverty indicator (%). Study period
N
Measures for individual years 2000 19,498 2001 17,984 2002 19,530 2003 18,495 2004 18,081 2005 17,384 2006 18,393 2007 17,228 2008 16,092 2009 16,878 2010 15,119 Measures for five-year periods 00e04 12,832 01e05 12,464 02e06 13,332 03e07 12,722 04e08 11,175 05e09 11,337 06e10 11,466
Poverty
Precarity
Prosperity
Extreme poverty
Moderate poverty
One-sided poverty
Vulnerability
Fragile prosperity
Secure prosperity
4.23 4.84 5.28 6.65 6.56 7.71 7.97 8.55 7.68 7.76 5.70
8.24 8.71 9.63 9.16 9.21 11.02 10.78 10.56 10.56 10.03 10.27
8.67 7.94 7.87 8.39 7.77 8.05 8.99 8.38 8.57 10.76 8.09
5.94 6.03 7.45 6.98 8.21 6.75 6.36 6.61 6.10 6.05 6.19
24.06 21.25 23.40 20.53 22.97 20.80 22.00 20.47 22.67 21.86 22.63
48.87 51.24 46.36 48.29 45.28 45.67 43.90 45.44 44.42 43.55 47.12
2.58 3.05 4.62 4.89 4.63 5.28 4.70
5.23 5.82 5.90 7.10 6.21 6.82 7.30
2.15 2.43 2.14 2.08 2.17 2.08 2.02
9.16 9.35 9.93 9.38 9.89 9.42 8.34
23.98 22.09 23.70 22.55 23.98 24.03 23.61
56.89 57.26 53.70 54.01 53.11 52.37 54.02
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
67
Table 5 Resultsa,b,c of the logistic regression models using the persistent at-risk-of-poverty rate of 2006e2010d and 2005e2009.e Dependent variables
Diet (healthy diet ¼ 1) N ¼ 12,221 Smoking (non-smoking ¼ 1) N ¼ 12,221 Physical activity (sufficient physical activity ¼ 1) N ¼ 12,221 a b c d e
Independent variables
Total sample
Male sample
Female sample
Odds ratio
95% CI
Odds ratio
95% CI
Odds ratio
95% CI
At-risk-of-poverty
0.933
[0.751, 1.160]
1.098
[0.793, 1.522]
0.868
[0.655, 1.151]
At-risk-of-poverty
0.890
[0.723, 1.095]
0.838
[0.618, 1.137]
0.888
[0.666, 1.184]
At-risk-of-poverty
0.795*
[0.620, 1.019]
0.987
[0.677, 1.439]
0.702**
[0.506, 0.975]
Reference group ¼ prosperity. Results based on estimation including confounders. Full estimation results available upon request. Significance level:***,**,* ¼ 1%, 5%, 10% and indicated in bold. Regarding dietary behavior and smoking. Regarding physical activity.
Table 6 Resultsa,b,c of the logistic regression models using the persistent combined poverty indicator of period of 2006e2010d and 2005e2009.e Dependent variables
Diet (healthy diet ¼ 1)
N ¼ 10,464 Smoking (non-smoking ¼ 1)
N ¼ 10,464 Physical activity (sufficient physical activity ¼ 1)
Independent variables
Total sample
Male sample
Odds ratio
95% CI
Extreme poverty Moderate poverty One-sided poverty Vulnerability Fragile prosperity
0.666** 0.618*** 0.580** 0.712*** 0.835**
[0.453, [0.460, [0.370, [0.560, [0.714,
Extreme poverty Moderate poverty One-sided poverty Vulnerability Fragile prosperity
0.369*** 0.612*** 0.695* 0.621*** 0.875
Extreme poverty Moderate poverty One-sided poverty Vulnerability Fragile prosperity
0.409*** 0.408*** 0.652 0.621*** 0.771***
Female sample
Odds ratio
95% CI
0.979] 0.831] 0.910] 0.906] 0.976]
1.027 0.635* 0.744 0.755 0.824
[0.583, [0.389, [0.391, [0.520, [0.652,
Odds ratio
95% CI
0.510 *** 0.607*** 0.473** 0.677** 0.841
[0.315, [0.417, [0.264, [0.496, [0.680,
0.824] 0.883] 0.845] 0.925] 1.040]
[0.256, [0.460, [0.461, [0.495, [0.738,
0.530] 0.814] 1.046] 0.778] 1.036]
0.326*** 0.472*** 0.515** 0.518*** 0.771**
[0.19, 0.559] [0.312, 0.716] [0.295, 0.900] [0.377, 0.711] [0.608, 0.977]
0.398*** 0.753 0.970 0.753* 1.012
[0.241, [0.504, [0.521, [0.543, [0.795,
0.659] 1.127] 1.807] 1.045] 1.288]
[0.250, [0.283, [0.380, [0.471, [0.646,
0.668] 0.589] 1.119] 0.817] 0.920]
0.581 0.456*** 1.181 0.641** 0.873
[0.271, [0.270, [0.527, [0.425, [0.670,
0.321*** 0.365*** 0.383*** 0.592*** 0.690***
[0.179, [0.219, [0.200, [0.407, [0.545,
0.578] 0.607] 0.732] 0.861] 0.873]
1.808] 1.037] 1.417] 1.096] 1.041]
1.243] 0.770] 2.646] 0.967] 1.137]
N ¼ 10,464 a b c d e
Reference group ¼ prosperity. Results based on estimation including confounders. Significance level:***,**,* ¼ 1%, 5%, 10% and indicated in bold. Regarding dietary behavior and smoking. Regarding physical activity.
reference group of prosperity. Extremely poor, moderately poor and one-sided poor individuals (OR ¼ 0.580 to 0.666) are least likely to have a healthy diet, followed by respondents of the groups “vulnerability” and “fragile prosperity” (OR ¼ 0.712 and 0.835). Looking at the male and female samples, significant odds ratios result for moderately poor men (OR ¼ 0.635) and for all women (OR ¼ 0.473e0.677) except women in fragile prosperity. There are also significant differences regarding non-smoking for all groups of poverty and precarity, except “fragile prosperity”. The lowest likelihood is among persistently extremely poor individuals (OR ¼ 0.369). In comparison to the latter result, differences are smaller for persistently moderately poor (OR ¼ 0.612), one-sidedly poor (OR ¼ 0.695) as well as vulnerable individuals (OR ¼ 0.621), who are also less likely to be non-smokers. Significant results on smoking differ if analyses of males and females are separated, and poor men are more likely to be smokers than women. Prosperous individuals are more likely to be sufficiently physically active than their poor and precarious counterparts. The lowest odds ratios are observable for the persistently extremely and moderately poor (OR ¼ 0.409 and 0.408). In addition, individuals who belong either to the vulnerability group (OR ¼ 0.621) or to the fragile prosperity group (OR ¼ 0.771) are less likely to practice
sufficient sports. No significant effects are observable for persistently one-sided poor individuals. The described effects are stronger for female than for male individuals. Persistently extremely poor women are least likely to be sufficiently physically active (OR ¼ 0.321). ORs less than 0.4 are also observable for women who are affected by persistent moderate and one-sided poverty. Persistently precarious women (“vulnerability” and “fragile prosperity”) are also less likely to be sufficiently physically active. In contrast, the significant ORs for men amount to 0.456 for the group of moderate poverty and 0.641 for vulnerability. 4.3. Panel analysis We next turn to panel analysis using the persistent poverty indicators and multiple year observations. Additional results for concurrent poverty are available in the online appendix. 4.3.1. Persistent at-risk-of-poverty rate Results of the fixed-effects logistic regression models using the persistent at-risk-of-poverty rate are presented in Table 7. The dependent variable has been measured every other year between 2004 and 2010 for healthy diet and non-smoking, and in the years 2005 and 2007e2009 for physical activity. The data are
68
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
Table 7 Resultsa,b,c of the fixed-effects logistic regression models using the persistent at-risk-of-poverty rate of 2000e2010d and 2001e2009.e Dependent variables
Independent variables
Diet (healthy diet ¼ 1) At-risk-of-poverty Number of observations: 16,781; number of individuals: 4,486 Smoking (non-smoking ¼ 1) At-risk-of-poverty Number of observations: 5,975; number of individuals: 1,591 Physical activity (sufficient physical activity ¼ 1) At-risk-of-poverty Number of observations: 14,243; number of individuals: 3,853 a b c d e
Total sample
Male sample
Female sample
Odds ratio
95% CI
Odds ratio
95% CI
Odds ratio
95% CI
0.975
[0.822, 1.155]
0.659***
[0.492, 0.883]
1.226*
[0.991, 1.517]
0.839
[0.645, 1.092]
1.267
[0.807, 1.990]
0.685**
[0.492, 0.955]
0.856
[0.698, 1.050]
0.924
[0.664, 1.287]
0.810
[0.622, 1.057]
Reference group ¼ prosperity. Results based on estimation including confounders. Significance level:***,**,* ¼ 1%, 5%, 10% and indicated in bold. Regarding dietary behavior and smoking. Regarding physical activity.
combined with persistent poverty indicators calculated based on data for five consecutive years for 2000e2010 in the first case, and from 2001 to 2009 in the second. The basis for these analyses are five-year periods during which an individual stays in a certain poverty group. As discussed in Section 3.4, only those observations with a change in the dependent variable are included in the fixed effect logistic regression. With regard to the total sample there are no effects of being persistently at risk of poverty and the three categories of health behavior. However, gender-specific effects can be found. Compared to the reference group that is not persistently at-risk-of-poverty, persistently poor men are less likely to follow a healthy diet (OR ¼ 0.659) whereas women are more likely to do so (OR ¼ 1.226). Additionally, women who are persistently at risk of poverty are less likely to be a non-smoker. In conclusion, the results underline the importance of separate analysis for men and women. Effects for being persistently at risk of poverty are only present for men regarding diet and for women concerning diet and smoking. 4.3.2. Persistent combined poverty indicator Panel analyses on health behavior and the combined poverty indicator produce a number of significant differences with regard
to the reference group of secure prosperity (Table 8). Individuals who belong to any of the five groups of persistent poverty or precarity decrease their probability of having a healthy diet. The highest risk of eating unhealthily is indicated for the groups of persistent extreme poverty (OR ¼ 0.485) and persistent moderate poverty (OR ¼ 0.623). Altogether an inverse gradient is observed, so that the odds ratio for a fragile prosperous individual is only 0.839. If the sample is separated by gender, males and females vary regarding dietary behavior. The lowest likelihood of having a healthy diet is among moderate poor males (OR ¼ 0.501) and extreme poor females (OR ¼ 0.469). Furthermore, persistent extreme poor or vulnerable men as well as women who are in the group of persistent moderate poverty, vulnerability and fragile prosperity are less likely to have a healthy diet. Odds ratios range between 0.547 and 0.803. With regard to smoking, results are less clear. Only individuals who are persistently fragile prosperous are more likely to be nonsmokers (OR ¼ 1.253). Interestingly, only vulnerable men are less likely to be non-smokers, whereas ORs for women in the groups of persistent extreme poverty, vulnerability and fragile prosperity have ORs greater than 1 (OR ¼ 1.400 to 1.788). Persistent extreme poverty is associated with the greatest chances of behaving in a
Table 8 Resultsa,b,c of the fixed-effects logistic regression models using the persistent combined poverty indicator of 2000e2010d and 2001e2009.e Dependent variables
Independent variables
Diet (healthy diet ¼ 1)
Extreme poverty Moderate poverty One-sided poverty Vulnerability Fragile prosperity Number of observations: 15,065; number of individuals: 4,057 Smoking (non-smoking ¼ 1) Extreme poverty Moderate poverty One-sided poverty Vulnerability Fragile prosperity Number of observations: 5,319; number of individuals: 1,429 Physical activity (sufficient physical activity ¼ 1) Extreme poverty Moderate poverty One-sided poverty Vulnerability Fragile prosperity Number of observations: 13,221; number of individuals: 3,583
a b c d e
Reference group ¼ prosperity. Results based on estimation including confounders. Significance level:***,**,* ¼ 1%, 5%, 10% and indicated in bold. Regarding dietary behavior and smoking. Regarding physical activity.
Total sample
Male sample
Odds ratio
95% CI
0.485*** 0.623*** 0.768* 0.724*** 0.839***
[0.357, [0.502, [0.578, [0.615, [0.758,
1.411 1.011 0.993 1.038 1.253*** 1.018 1.103 0.837 0.974 0.994
Female sample
Odds ratio
95% CI
Odds ratio
95% CI
0.658] 0.773] 1.021] 0.852] 0.928]
0.547** 0.501*** 0.727 0.683*** 0.916
[0.324, [0.348, [0.438, [0.522, [0.778,
0.923] 0.723] 1.206] 0.894] 1.078]
0.469*** 0.708** 0.793 0.750*** 0.803***
[0.320, [0.539, [0.560, [0.610, [0.704,
0.688] 0.928] 1.121] 0.923] 0.915]
[0.847, [0.710, [0.635, [0.795, [1.058,
2.350] 1.439] 1.553] 1.356] 1.485]
1.177 1.325 1.378 0.696* 1.185
[0.517, [0.756, [0.671, [0.467, [0.933,
2.682] 2.322] 2.831] 1.036] 1.506]
1.788* 0.930 0.908 1.459** 1.400***
[0.914, [0.581, [0.503, [1.004, [1.093,
3.500] 1.487] 1.639] 2.121] 1.794]
[0.637, [0.814, [0.590, [0.788, [0.869,
1.628] 1.495] 1.187] 1.204] 1.136]
2.311** 2.545*** 3.192*** 1.885*** 1.582***
[1.153, [1.537, [1.765, [1.313, [1.266,
4.635] 4.213] 5.774] 2.706] 1.977]
0.734 0.790 0.403*** 0.761** 0.797***
[0.379, [0.532, [0.256, [0.582, [0.671,
1.421] 1.173] 0.634] 0.996] 0.946]
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
health-promoting way in this category. Analyses of physical activity and the persistent combined poverty indicator show that there are no significant differences for the total sample. Nevertheless, significant odds ratios are present for men and women. Men who belong to one of the persistent poverty or precarity groups are more likely to pursue sufficient sports compared to their prosperous counterparts. The highest odds ratios are observable for one-sided poverty (OR ¼ 3.192), moderate poverty (OR ¼ 2.545) and extreme poverty (OR ¼ 2.311). In contrast, women behave differently. Women who are part of the persistent one-sided poverty group are least likely to be physically active (OR ¼ 0.403). Additionally, being affected by persistent vulnerability or fragile prosperity is associated with lower odds ratios (OR ¼ 0.761 and OR ¼ 0.797). No effects are observed for persistent extreme or moderate poverty and pursuing sufficient sports. In conclusion, significant results are mainly observed for dietary behavior. Particularly regarding non-smoking and physical activity, it is important to consider the results by gender. Being affected by persistent precarity and poverty is associated with a higher likelihood of being a non-smoker for women, whereas men are more likely to be physical active. Results of confounding variables are available in the online appendix. For space consideration they are not discussed here. 5. Discussion Our analysis provides new evidence that individuals who are persistently either at risk of poverty, poor, or precarious are less likely to behave in a health-promoting way than their prosperous counterparts. Hence, the importance of persistent poverty on health behavior can be confirmed. The results are based on the analysis of three different types of health behavior using an income-based poverty measure in addition to one that is based on income and four dimensions material deprivation. Using the two different poverty measures reveals the multifaceted aspects of the relation between poverty and health behavior. With regard to diet, we observe a health behavior gradient and a steady decline in following a healthy diet as combined poverty becomes more severe. For other types of health behavior (non-smoking and physical activity), the results are more heterogeneous. Comparing both persistent poverty indicators, the smallest differences between persistently poor and prosperous individuals are reported for the at-risk-of-poverty-rate. It can be concluded that the two determinants e income poverty and the combined poverty indicator e cover different mechanisms explaining health behavior. For this reason they cannot be used interchangeably. This is in line with the literature presented above (cf. Benzeval and Judge, 2001; Groh-Samberg, 2009; Moll, 2006; Nolan and Whelan, 2007). These findings also confirm Mackenbach's “explanation of health inequality” depicted in 2.1. Material factors, including income, influence health behavior directly through resource constraints and via psychosocial factors such as negative life events, daily hassles, “effort-reward imbalance”, or a combination of high demands and low control (Mackenbach, 2006). As a result, a critical material situation can be reinforced by psychosocial stress. Thus, the results show, for the first time in the context of health behavior, the relevance of considering multidimensional aspects of poverty that have been identified by various poverty researchers (cf. Bradshaw and Finch, 2003; Groh-Samberg, 2009; €d, 1995; Mack and Lansley, 1985; Nolan and Whelan, 2007). Hallero Finally, findings of the analyses show that poverty affects several dimensions of health behavior. The greatest differences are observed for tobacco consumption and physical activity, but they are also important for dietary behavior. The multidimensional
69
measurement shows that not only do the most persistent nonprosperous groups, namely those in poverty, but also individuals living in precarious situations report a higher likelihood for healthdamaging behavior than their prosperous counterparts. Panel analyses (Tables 7 and 8) show that health behavior is influenced by changes in poverty status over time. Also here, the weakest impact on health behavior is reported for the at-risk-ofpoverty rate. These findings underline the relevance of persistent poverty (cf. Duncan et al., 1993; Groh-Samberg, 2008). These results are consistent with Benzeval and Judge's (2001) observation that persistent poverty implies greater health risks than occasional episodes. Furthermore, variations between men and women occur and gender specific effects differ from those of the general sample. Additionally, Lynch et al. (1997b) observe lower rates of physical activity for persistently poor individuals. The present analyses show, however, that this is only valid for women using the combined poverty indicator but not for men. Thus, data of GSOEP confirm previous empirical results as well as the “explanation of health inequality” by Mackenbach (2006), which describes an inverse relationship between socio-economic status and dietary behavior, smoking and physical activity. These results emphasize the necessity of supporting individuals in both situations of poverty as well as precarity. This study has several limitations. Data regarding health behavior is part of GSOEP but not a focus. In comparison to pure health (behavior) surveys that consider only limited information on socio-economic aspects, information on health behavior is not very detailed. Nevertheless the GSOEP offers a relevant data base for the questions addressed here, as epidemiological surveys offer fewer details regarding socio-economic aspects which this study can consider in the analyses. In addition, social selectivity must be assumed. Like many empirical surveys, GSOEP is not able to cover homeless persons, illegal immigrants, addicts, or persons who are highly deprived (cf. Groh-Samberg, 2009). In addition, future research could consider the potential correlations among the different types of health behavior in the analysis. Despite these limitations, the results show that not only being at risk of persistent poverty, but also being persistently poor and in precarious situations, is associated with detrimental health behavior. Findings underscore the relevance of cumulative poverty as well as dynamic aspects. We could also show that the relationship between poverty and health behavior is gender-specific. This result has to be taken into consideration for approaches to support health-promoting behavior. Acknowledgments We acknowledge DIW for providing the GSOEP data to us. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.socscimed.2016.01.040. References Adler, N., Ostrove, J., 1999. Socioeconomic status and health: what we know and what we don't. Ann. N. Y. Acad. Sci. 3e15. Anderson, L.R., Mellor, J.M., 2008. Predicting health behaviors with an experimental measure of risk preference. J. Health Econ. 27 (5), 1260e1274. http://dx.doi.org/ 10.1016/j.jhealeco.2008.05.011. Ashworth, K., Hill, M., Walker, R., 1994. Patterns of childhood poverty: new challenges for policy. J. Policy Anal. Manag. 13 (4), 658e680. Atkinson, A.B., Cantillon, B., Marlier, E., Nolan, B., 2002. Social Indicators: the EU and Social Inclusion. Oxford University Press, Oxford. Basiotis, P., Guthrie, J., Bowman, S., Welsh, S., 1995. Construction and evaluation of a diet status index. Fam. Econ. Nutr. Rev. 8 (2), 2e13.
70
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
Benzeval, M., Judge, K., 2001. Income and health: the time dimension. Soc. Sci. Med. 52, 1371e1390. Berthoud, R., Bryan, M., Bardasi, E., 2004. The Dynamics of Deprivation: the Relationship between Income and Material Deprivation over Time. Retrieved from. http://campaigns.dwp.gov.uk/asd/asd5/rports2003-2004/rrep219.pdf. Bradshaw, J., Finch, N., 2003. Overlaps in dimensions of poverty. J. Soc. Policy 32 (04), 513e525. http://dx.doi.org/10.1017/S004727940300713X. Centers for Disease Control and Prevention, 2008. Fact Sheet for Health Professionals on Physical Activity Guidelines for Adults. Retrieved from. http:// www.cdc.gov/nccdphp/dnpa/physical/pdf/PA_Fact_Sheet_Adults.pdf. Cohen, D.A., Farley, T.A., Mason, K., 2003. Why is poverty unhealthy? Social and physical mediators. Soc. Sci. Med. 57 (9), 1631e1641. Contoyannis, P., Jones, A., 2004. Socio-economic status, health and lifestyle. J. Health Econ. 23, 965e985. Dennis, I., Guio, A., 2003. Poverty and Social Exclusion in the EU after Laeken e Part 1: Population and Social Conditions. Retrieved from. http://ec.europa.eu/ eurostat/documents/3433488/5551812/KS-NK-03-008-EN.PDF/9b0cdf8b-943b4d63-894f-3a0669fe2d93. Duncan, G., Gustafsson, B., Hauser, R., Schmauss, G., Messinger, H., Muffels, R., …, Ray, J.-C., 1993. Poverty dynamics in eight countries. J. Popul. Econ. 6 (3), 215e234. European Union, 2010. Combating Poverty and Social Exclusion: A Statistical Portrait of the European Union 2010: A Statistical Portrait of the European Union 2010. Publications Office of the European Union, Luxembourg. Retrieved from. http://epp.eurostat.ec.europa.eu/cache/ITY_OFFPUB/KS-EP-09-001/EN/ KS-EP-09-001-EN.PDF. Eurostat, 2011. Glossary: At-risk-of-poverty Rate. Retrieved from. http://epp. eurostat.ec.europa.eu/statistics_explained/index.php/Glossary:At-risk-ofpoverty_rate. Goebel, J., Grabka, M., Krause, P., Kroh, M., Pischner, R., Sieber, I., Spieß, M., 2008. €ngsschnittstudie SozioMikrodaten, Gewichtung und Datenstruktur der La oekonomisches panel (SOEP). Vierteljahrsh. Wirtsch. 77 (3), 77e109. Retrieved from. http://ejournals.duncker-humblot.de/doi/pdf/10.3790/vjh.77.3. 77. Groh-Samberg, O., 2008. Persistent poverty is increasing in Germany. In: Heady, B., Holst, E. (Eds.), A Quarter Century of Change: Results from the German Socioeconomic Panel (SOEP), pp. 41e48. Berlin. Retrieved from. http://www.diw. de/documents/publikationen/73/diw_01.c.91272.de/soep_wave_report_2008. pdf. Groh-Samberg, O., 2009. Armut, soziale Ausgrenzung und Klassenstruktur. VS Verlag für Sozialwissenschaften, Wiesbaden. Hagenaars, A., Vos, K., Zaidi, M., 1994. Poverty Statistics in the Late 1980s: Research Based on Micro-data: Research Based on Micro-data. Office for Official Publications of the European Communities, Luxembourg. €d, B., 1994. A New Approach to the Direct Consensual Measurement of Hallero Poverty, vol. 50. Social Policy Research Centre (SPRC). Discussion Paper. €d, B., 1995. The truly poor: direct and indirect consensual measurement of Hallero poverty in Sweden. J. Eur. Soc. Policy 5 (2), 111e129. Helmert, U., 2003. Soziale Ungleichheit und Krankheitsrisiken. Maro, Augsburg. Knoops, K.T.B., Groot, L.C.P.G.M., de, Kromhout, D., Perrin, A., Moreiras-Varela, O., Menotti, A., van Staveren, W.A., 2004. Mediterranean diet, lifestyle factors, and 10-year mortality in elderly European men and women. J. Am. Med. Assoc. 292 (12), 1433e1439. Kroll, L.E., 2010. Sozialer Wandel, soziale Ungleichheit und Gesundheit: Die Entwicklung sozialer und gesundheitlicher Ungleichheiten in Deutschland zwischen 1984 und 2006. VS Verlag für Sozialwissenschaften, Wiesbaden. Krueger, P., Chang, V., 2008. Being poor and coping with stress: health behaviors and the risk of death. Am. J. Public Health 98, 889e896. € konomischen Status in soziaLampert, T., Kroll, L., 2009. Die Messung des sozioo lepidemiologischen Studien. In: Richter, M., Hurrelmann, K. (Eds.), Gesundheitliche Ungleichheit. Grundlagen, Probleme, Perspektiven, second ed. VS Verlag für Sozialwissenschaften, Wiesbaden, pp. 309e334. Leisering, L., 2008. Dynamik von Armut. In: Huster, E.-U., Boeckh, J., MoggeGrotjahn, H. (Eds.), Handbuch Armut und Soziale Ausgrenzung. VS Verlag für Sozialwissenschaften, Wiesbaden, pp. 118e132. Lynch, J., Kaplan, G., Salonen, J., 1997a. Why do poor people behave poorly? Variation in adult health behaviours and psychosocial characteristics by stages of the socioeconomic lifecourse. Soc. Sci. Med. 44 (6), 809e819. Lynch, J., Kaplan, G., Shema, S., 1997b. Cumulative impact of sustained economic hardship on physical, cognitive, psychological, and social functioning. N. Engl. J. Med. 337 (26), 1889e1895. Mack, J., Lansley, S., 1985. Poor Britain. Allen & Unwin, London. Mackenbach, J., 2006. Health Inequalities: Europe in Profile. Retrieved from. http://
ec.europa.eu/health/ph_determinants/socio_economics/documents/ev_ 060302_rd06_en.pdf. Mackenbach, J.P., van de Mheen, H., Stronks, K., 1994. A prospective cohort study investigating the explanation of socio-economic inequalities in health in the Netherlands. Soc. Sci. Med. 38 (2), 299e308. http://dx.doi.org/10.1016/02779536(94)90399-9. Mackenbach, J.P., Stirbu, I., Roskam, A.-J.R., Schaap, M.M., Menvielle, G., Leinsalu, M., Kunst, A.E., 2008. Socioeconomic inequalities in health in 22 European countries. N. Engl. J. Med. 358 (23), 2468e2481. http://dx.doi.org/10.1056/ NEJMsa0707519. McGinnis, J., Foege, W., 1993. Actual causes of death in the United States. JAMA 270 (19), 2207e2212. http://dx.doi.org/10.1001/jama.1993.03510180077038. Mielck, A., 2000. Soziale Ungleichheit und Gesundheit: Empirische Ergebnisse, €rungsansa €tze, Interventionsmo €glichkeiten, first ed. Huber, Bern. Erkla Mokdad, A.H., 2004. Actual causes of death in the United States, 2000. JAMA 291 (10), 1238e1245. http://dx.doi.org/10.1001/jama.291.10.1238. Moll, S., 2006. Dynamik von Armut in Deutschland: Ergebnisse mikrooekonometrischer Analysen. Dissertation. University of Hohenheim. Retrieved from. http://opus.ub.uni-hohenheim.de/volltexte/2006/154/pdf/dissertation. pdf. Moore, M., Hughes, J., 2001. The health care consequences and its regulation. Forum Health Econ. Policy 4, 31e70. Mueller, U., Heinzel-Gutenbrunner, M., 2001. Krankheiten und Beschwerden (subjektive Gesundheit) unter Bewertung der eigenen Gesundheit. Mater. Bev€lkerungswiss. 102c. Retrieved from. http://www.bib-demografie.de/nn_ o 750460/SharedDocs/Publikationen/DE/Download/Materialienbaende/ 102c,templateId¼raw,property¼publicationFile.pdf/102c.pdf. Muffels, R., 1993. Welfare Economic Effects of Social Security. Essays on Poverty. Social Security and Labour Market: Evidence from Panel Data. TISSER, Tilburg Inst. for Social Security Research, Dep. of Social Security Studies, Tilburg. Nolan, B., Whelan, C., 2007. On the multidimensionality of poverty and social exclusion. In: Jenkins, S.P., Micklewright, J. (Eds.), Inequality and Poverty Reexamined. Oxford University Press, Oxford, pp. 146e165. O'Boyle, E.J., 1999. Toward an improved definition of poverty. Rev. Soc. Econ. 57 (3), 281e300. Retrieved from. http://search.ebscohost.com/login.aspx? direct¼true&db¼sih&AN¼2452162&site¼ehost-live. OECD, 1982. The OECD List of Social Indicators. Paris. Organisation for Economic Cooperation and Development, Washington, D.C. Sold by OECD Publications and Information Center. Olshansky, S.J., Ault, A.B., 1986. The fourth stage of the epidemiologic transition: the age of delayed degenerative diseases. Milbank Q. 64 (3), 355e391. Osler, M., 2006. The life course perspective: a challenge for public health research and prevention. Eur. J. Public Health 16 (3), 230. rez-Mayo, J., 2004. Consistent Poverty Dynamics in Spain. In: IRISS Working Pe Paper Series. Retrieved from. http://iriss.ceps.lu/documents/irisswp50.pdf. €rperliche Aktivita €t. Gesundh. des Bund. 26, 13e14. Robert Koch-Institut, 2005. Ko Smith, N., Middleton, S., 2007. A Review of Poverty Dynamics Research in the UK. Joseph Rowntree Foundation, York. Smith, K.R., Zick, C.D., 1994. Linked lives, dependent demise? Survival analysis of husbands and wives. Demography 31 (1), 81e93. Retrieved from. http://search. ebscohost.com/login.aspx?direct¼true&db¼sih&AN¼9410253102&site¼ehostlive. Townsend, P., 1979. Poverty in the UK: A Survey of Household Resources and Standards of Living. Penguin Books, Harmondsworth. Wagle, U., 2002. Rethinking poverty: definition and measurement. Int. Soc. Sci. J. 54 (171), 155e165. http://dx.doi.org/10.1111/1468-2451.00366. Wagner, G., Frick, J., Schupp, J., 2007. The German socio-economic panel study (SOEP): scope, evolution and enhancements. Schmollers Jahrb. 1 (127), 139e170. Retrieved from. http://www.diw.de/documents/publikationen/73/ diw_01.c.60184.de/diw_sp0001.pdf. Wardle, J., Haase, A.M., Steptoe, A., Nillapun, M., Jonwutives, K., Bellisie, F., 2004. Gender differences in food choice: the contribution of health beliefs and dieting. Ann. Behav. Med. 27 (2), 107e116. http://dx.doi.org/10.1207/ s15324796abm2702_5. Whelan, C., Maître, B., 2005. Comparing poverty and deprivation dynamics: issues of reliability and validity. J. Econ. Inequal. 4 (3), 303e323. http://dx.doi.org/ 10.1007/s10888-005-9017-1. Whelan, C., Layte, R., Maître, B., 2001. What is the Scale of Multiple Deprivation in the European Union? Retrieved from. http://www.esri.ie/UserFiles/ publications/20070208160751/OPEA014.pdf. WHO, 2002. The World Health Report 2002-Reducing Risks, Promoting Healthy Life. Retrieved from. http://www.who.int/whr/2002/en/whr02_en.pdf.
Contents lists available at ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
Poverty dynamics in Germany: Evidence on the relationship between persistent poverty and health behavior Katja Aue a, Jutta Roosen a, *, Helen H. Jensen b a b
Technical University of Munich, TUM School of Management, Chair of Marketing and Consumer Research, Alte Akademie 16, 85354, Freising, Germany Iowa State University, Department of Economics, Heady Hall, Ames, IA, 50011-1070, USA
a r t i c l e i n f o
a b s t r a c t
Article history: Received 18 October 2015 Received in revised form 24 December 2015 Accepted 24 January 2016 Available online 29 January 2016
Previous studies have found poverty to be related to lower levels of health due to poor health behavior such as unhealthy eating, smoking or less physical activity. Longer periods of poverty seem to be especially harmful for individual health behavior. Studies have shown that poverty has a dynamic character. Moreover, poverty is increasingly regarded as being a multidimensional construct and one that considers more aspects than income alone. Against this background this paper analyzes the relationship between health behavior and persistent spells of income poverty as well as a combined poverty indicator using data of the German Socio-Economic Panel (2000e2010). Next to cross-sectional logistic regression models we estimate fixed-effects models to analyze the effect of persistent poverty on dietary behavior, tobacco consumption, and physical activity. Cross-sectional results suggest that persistent poverty is related to poor health behavior, particularly regarding tobacco consumption and physical activity. Results also show that multidimensional and dynamic aspects of poverty matter. Complementary panel analyses reveal negative effects for the combined poverty indicator only for dietary behavior in the total sample. However, by analyzing the sample by gender we identify further effects of persistent poverty on health behavior. The analyses show that not only do individuals in poverty but also those in precarious situations show health-damaging behavior more often. © 2016 Elsevier Ltd. All rights reserved.
Keywords: GSOEP Health behavior Multidimensional poverty Panel analysis Persistent poverty
1. Introduction Studies in the area of health economics and public health have shown that poverty and low socio-economic status (SES) are related to lower levels of health (Benzeval and Judge, 2001; Cohen et al., 2003; Helmert, 2003; Mackenbach et al., 2008; Mielck, 2000). Inequality in healthy life expectancy can be observed. For example, rates of premature mortality are higher among those with lower levels of education, occupational status or income. Rates of morbidity are also higher (Lampert and Kroll, 2009; Mackenbach, 2006). Attempts to explain these differences have often made reference to the observation that poor health behavior such as unhealthy dietary behavior, smoking, or physical inactivity clusters in poverty groups or for those with a low SES (Contoyannis and Jones, 2004; Lynch et al., 1997a). For instance, McGinnis and Foege (1993) have
* Corresponding author. E-mail address: [email protected] (J. Roosen). http://dx.doi.org/10.1016/j.socscimed.2016.01.040 0277-9536/© 2016 Elsevier Ltd. All rights reserved.
shown that approximately 38% of all deaths in the US were caused by behavior-related factors. Also Mokdad (2004) confirmed this relationship for the United States in 2004. Likewise in Europe, the World Health Organization (WHO) (2002) reported that the total burden of disease in Europe is considerably influenced by health behavior and by poverty and income inequalities. Concerning poverty dynamics and health there are only a small number of longitudinal studies available. Furthermore, the available research shows that persistent poverty is more important than current income (Benzeval and Judge, 2001). Health behavior (in comparison to health outcomes), however, has been only studied in a few cases (c.f. Lynch et al., 1997b; Smith and Middleton, 2007; Smith and Zick, 1994) which underlines the need for studies in this area. Furthermore, to the best of the authors' knowledge, there is no study looking at health behavior and the dynamics of multidimensional poverty. Against this background the objective of this paper is to empirically examine the relationship between dichotomous and multidimensional persistent poverty measurements (at-riskpoverty-rate vs. the combined poverty indicator by Groh-Samberg,
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
2008) and health behavior. It is the first time that such a multidimensional indicator is used for analyses in the area of health behavior research. Our analyses are based on the conceptual framework by Mackenbach (2006) outlined in Section 2.1. We use five years of observations of health behavior variables from the period 2004e2010 in the German population using data of the German Socio-Economic Panel (GSOEP) and combine these data with information on persistent poverty status. Health behavior dimensions include dietary behavior, smoking and physical activity. Poverty is measured by two measures: relative income poverty and a combined poverty indicator by Groh-Samberg (2009). The data are analyzed using a cross-sectional approach as well as a panel data estimation. The results reveal that considering multidimensional and dynamic aspects of poverty matters. Both poverty indicators show that persistent poverty is negatively associated with health behavior. The paper is organized as follows. The next section will give some background on explaining health behavior and measures of poverty. Section 3 then presents the data and methods of analysis. Results are discussed in Section 4 before the paper concludes. 2. Background 2.1. Explaining health behavior Health behavior can be explained by several theories including economic and social sciences. An important model explaining health inequality was developed by Mackenbach (2006) and is depicted in Fig. 1. Mackenbach considers health behavior as one of three mediators between SES and health. Traditional components of SES are income, education and occupation (Adler and Ostrove, 1999). Hence health behavior is influenced by socio-economic parameters directly as well as indirectly via material and psychosocial factors. Health in turn is determined by health behavior, material factors and psychosocial factors. Material factors influence the health status directly as well as via psychosocial factors and health behavior. Because these different determinants are heterogeneously distributed across socio-economic groups, they are seen as the main explanation of health inequality. In this model, specific health determinants comprise the main explanation of health inequality. The group of material factors, including poverty, considers financial aspects, especially the income situation which influences psychosocial aspects such as stress, subsequent risk-taking
63
health behavior as well as the access to health-promoting facilities and products. These types of psychosocial stress may lead to illhealth either through biological or behavioral pathways. Further material factors are health risks related to occupation and housing. In this context, it should be noted that health behavior may respond immediately to deprivation whereas the health status develops over time and may be the result of long-term effects of health behavior. This study therefore focuses on individual health behavior as a key factor in Mackenbach's model (Kroll, 2010; Knoops et al., 2004; Olshansky and Ault, 1986; Osler, 2006). Against this background we hypothesize that persistent poverty is associated with detrimental health behavior (cross-sectional analysis) and that it increases the likelihood of such behavior (panel analysis). 2.2. Defining poverty Poverty is still present in many developed countries like Germany. While most poor individuals are not affected by physical deprivation or hunger, relative poverty, mostly defined by low income status, still concerns many to this day (e.g., Duncan et al., 1993; Eurostat, 2011). There exist several approaches to the definition of poverty in developed countries, however, there is no universally valid definition. Definitions are based on absolute and relative concepts as well as subjective approaches (Wagle, 2002). Despite the fact that the definition of relative poverty is difficult and normative (O'Boyle, 1999), our study focuses on approaches of relative poverty. In that we follow the European Commission that defines people as poor “[…]if their income and resources are so inadequate as to preclude them from having a standard of living considered acceptable in the society in which they live […]. They are often excluded and marginalised from participating in activities (economic, social and cultural) that are the norm for other people and their access to fundamental rights may be restricted” (European Union, 2010). Altogether, income-based measurements are widely used to describe income poverty or the at-risk-of-poverty rate (Nolan and Whelan, 2007). For example, in the European Union the at-risk-poverty-rate is one of the so-called Laeken indicators and defined to be at 60% of median net-equivalence income (Dennis and Guio, 2003). Nevertheless, information on income may be insufficient to determine the degree to which a person is at risk of deprivation. Some households are able to maintain an acceptable standard of living although they are on a low level of income, either because income poverty is only temporary or because of other
Fig. 1. Explanation of health inequality. Solid lines describe the causation hypothesis. The dashed line represents the explanation of natural or social selection that was included in Mackenbach et al. (1994) first model. This aspect is no longer included in the updated model from 2006. Source: modified according to Mackenbach et al. (1994) and Mackenbach (2006).
64
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
resources such as savings or gifts. To capture the multifaceted nature of poverty, the British sociologist Peter Townsend provided a definition based on concepts of deprivation and social exclusion and introduced a deprivation index that describes poverty as a multidimensional phenomenon rez-Mayo, 2004). The index surveys whether (Townsend, 1979; Pe essential items, namely goods or practices of everyday life, are missing due to financial restrictions. A household is accordingly defined as deprived if a given amount of items is missing, where the necessary amount is oriented on the common standard of living in society. There have been several attempts to describe multidimensional poverty since Townsend developed the deprivation in€d, 1994, 1995; Mack and Lansley, dex (Berthoud et al., 2004; Hallero rez-Mayo, 2004; Whelan and Maître, 2005). 1985; Muffels, 1993; Pe Since many researchers propose to supplement the measurement of income poverty with direct measurements of standard of living and of life domains (Benzeval and Judge, 2001; GrohSamberg, 2009; Moll, 2006; Nolan and Whelan, 2007), this study focuses on an existing multidimensional measurement that considers not only income but also additional aspects of deprivation €d, 1995; Pe rez-Mayo, 2004; Whelan et al., 2001). (cf. Hallero Therefore the index combines income poverty with four life domains: housing, consumption, financial reserves and unemployment (Groh-Samberg, 2008). As a result, a household may be deprived in the income and material dimensions or poverty may be inconsistent, e.g. a household is affected by multiple deprivations €d, 1995). but has adequate income or vice versa (cf. Hallero Since poverty is not a static phenomenon, dynamic aspects should also be considered (Ashworth et al., 1994; Duncan et al., rez-Mayo, 2004). In the 1990s the research on poverty 1993; Pe dynamics gained in importance in Europe following the research ideas being generated in the United States (Leisering, 2008). In Germany, Groh-Samberg (2008) stated that poverty becomes more persistent over time. In addition, other statuses such as a precarious living condition or temporary poverty are also observable over time. 3. Data and estimation 3.1. Data description We use data from the German Socio-Economic Panel (GSOEP) to examine the relationship between multidimensional aspects of poverty and deprivation and health behavior. The GSOEP is an ongoing panel survey of households and individuals conducted annually since 1984, and is representative of the resident population of Germany. Further details on design and methodology of GSOEP have been described elsewhere (Wagner et al., 2007). Individuals report repeatedly on an annual basis. The following analyses use the 2000e2010 waves. The panel data design is unbalanced to consider as many individuals as possible. 3.2. Measurement of health behavior We analyze three types of health behavior: dietary behavior, physical activity and smoking. Dietary behavior is measured on a four-point scale by the question: “To what extent do you follow a health-conscious diet?” Since there is evidence that there is a direct association between self-reported healthfulness of diet and actual dietary quality, this question is used to represent the dietary behavior within the German population (Basiotis et al., 1995). Answers are aggregated to a binary variable: healthy diet (¼1, includes answers 1 and 2) and no healthy diet (¼0; includes answers 3 and 4). Smoking is associated with higher risk for cardiovascular diseases and several
forms of cancer (Krueger and Chang, 2008; Moore and Hughes, 2001). It is measured in the data set by counting the number of cigarettes/day (pipes and cigars are counted as two cigarettes) (Mueller and Heinzel-Gutenbrunner, 2001). We summarize the responses in a dichotomous variable indicating whether the respondent is non-smoker (Y ¼ 1) or smoker (Y ¼ 0). Physical activity is defined as sufficient if the respondent does sport at least once per week or more often (Y ¼ 1), which is near the recommendations of the Robert Koch-Institute (2005) for Germany and the Physical Activity Guidelines for Americans (Centers for Disease Control and Prevention, 2008). 3.3. Independent variables 3.3.1. Income poverty In the European Union, income poverty is often defined as the at-risk-of-poverty rate using a threshold of 60% of median netequivalence income based on the OECD modified equivalence scale (Atkinson et al., 2002; Dennis and Guio, 2003). Since persistent income can be represented by the average of income over time, the mean of the equivalized net household income (ENIt) is computed over five years (Goebel et al., 2008; Hagenaars et al., 1994), so that for year t:
ENI EU;t ¼
4 X ENIEU;ts s¼0
5
(1)
The period over five years was chosen in accordance with previous panel analyses: on the one hand it allows considering as many survey years as possible and on the other hand it avoids a high rate of panel attrition that may be expected, especially for individuals affected by poverty (Moll, 2006). A household is considered to be affected by persistent income poverty if ENI EU;t is smaller than 60% of the median equivalized net household income. 3.3.2. Combined poverty index by Groh-Samberg Different from the European definition of income poverty, GrohSamberg uses the “old” OECD equivalence scale (OECD, 1982). Comparable to equation (1) the persistent equivalent net household income in year t following Groh-Samberg (2009) (ENIGS,t) is computed as follows:
ENI GS;t ¼
4 X ENIGS;ts s¼0
5
:
(2)
Groh-Samberg defines three income situations with regard to the mean of the ENI: “income poor” (75% of mean) assuming that a mean is more suitable to consider changes for higher incomes than the median (Groh-Samberg, 2009). In addition to income poverty, the four deprivation dimensions considered are housing, consumption, financial reserves and unemployment. They also refer to the household level. Housing deprivation includes insufficient room and a lack of basic equipment. Consumption is aggregated on a scale of commodities. This scale includes a large number of items such as owning a car or TV. A deprivation threshold of one standard deviation below the index mean is applied. To determine deprivation in the area of financial reserves households are regarded as deprived if they have no assets and no significant savings at all. Finally unemployment is a state of deprivation too, because it can be seen as one of the most important non-monetary dimensions of social exclusion and lowers life satisfaction substantially (Groh-Samberg, 2008).
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
65
Combining income poverty with deprivation measurements results in nine combinations. Altogether six groups, namely extreme, moderate and one-sided poverty as well as vulnerability, fragile prosperity and secure prosperity are specified (Table 1). To derive a combined index of persistent poverty, multiple deprivations over five years can be observed if the sum of deprivations is greater than 8 (z2 deprivations/year), a single deprivation exists if the sum of deprivations is less than 8 and greater than 2 (z1 deprivation/year), and no deprivation if the sum of deprivations does not exceed 2 deprivations over five years (GrohSamberg, 2008).
4. Results
3.4. Multivariate analyses
4.1.2. Income poverty For all relevant five-year periods, Table 3 shows the respective number of observations, the percentage of the population at risk of poverty and the poverty line. In addition, we report the data for individual years at the top of the table. Although at-risk-of-poverty is defined at the household level, the data are reported for observations of individuals to match the analyses reported below. Table 3 shows that the share of individuals whose equivalized net-household income is below the at-risk-of-poverty threshold of 60% of median net equivalence income increases gradually from 11.85% in 2000e16.17% in 2010. At the same time, the threshold increases over the years from 10,222 Euros to 12,185 Euros. Looking at measures of persistent poverty, the table reveals that more than 10% of the GSOEP population is affected by persistent income poverty.
Firstly, dietary behavior, smoking and physical activity of individual i are analyzed as dependent binary variables (Yi 2 {0, 1}) using information on persistent income poverty and the combined poverty indicator in the respective year as explanatory variables in logistic regression models. Since dietary behavior and smoking are measured in even years and physical activity in uneven years we use GSOEP waves of 2009 and 2010. To interpret the results we use odds ratios (OR). Secondly, panel data analyses are conducted for the three types of health behavior using information on persistent poverty as explanatory variables in logistic regression panel analyses. According to the data availability in GSOEP, panel analyses regarding dietary behavior and smoking are conducted for the period 2004 to 2010, whereas analyses on physical activity are conducted for 2005 to 2009. As evident in Table 2, each health behavior is surveyed four times in these time spans. We estimate fixed effects logistic models. Fixed-effect models of binary dependent variables have two specific characteristics. First, the logistic regression analyses use only within-person variation, so that individuals without variation in the dependent variable are excluded from the sample. Second, the fixed-effects models do not produce any parameter estimates for the time-invariant explanatory variables (e.g. gender and migration background) that are confounded in the constant term.
3.5. Confounders Models are adjusted for subjective health status, education, (un) employment, occupational status, gender, age, marital status, migration background, region of residence (former East/West Germany) and number of children (0e14 years living in a household). Following Mackenbach (2006) economic worries are also considered as measurement for psychosocial stress regarding financial aspects. Furthermore we adjust our models for risk preferences (cf. Anderson and Mellor, 2008). Since health-related behavior varies by gender, analyses are conducted not only for the total population but also separated by gender.
4.1. Descriptive statistics 4.1.1. Health behavior Table 2 shows the distribution of health behaviors for the total sample and by gender during the period 2004e2010. Information on healthy diet and non-smoking is collected in 2004, 2006, 2008, and 2010; information on physical activity in 2005, 2007, 2008, and 2009. Generally, more women behave in a health-promoting way than do men, so that analyzing health behavior by gender is necessary (cf. Wardle et al., 2004).
4.1.3. Combined poverty index by Groh-Samberg Results for the concurrent and persistent combined poverty index are given in Table 4. As in the case of the at-risk-of poverty rate, poverty in single years is more common than persistent poverty. Compared to the persistent at-risk-of poverty rate that shows an average of 10.2% to 13.6% of the German population at risk of poverty, the persistent combined poverty indicator makes different aspects of poverty visible. Extreme persistent poverty (income poverty and multiple deprivations) is observed for less than 5% of the sample for most periods. Fragile prosperity and secure prosperity is achieved by about 80% of the sample. Low income and single deprivations (vulnerability) go together more often (about 10% of the population) than being opposed to each other and resulting in one-sided poverty (around 2% of the population). 4.2. Logistic regression models of diet and smoking (2010) and physical activity (2009) We first present results for the cross section analyses based on the persistent at-risk-of-poverty rate and then for the combined poverty indicator. All tables first show the results for the entire sample and then by gender. Additional results for concurrent poverty are available in the online appendix.
Table 1 Characteristics of the combined poverty index by Groh-Samberg. Income
Deprivation Multiple deprivations (2)
Single deprivation (¼1)
No deprivation (¼0)
Income poverty (75% of mean)
Extreme poverty Moderate poverty One-sided poverty
Moderate poverty Vulnerability Fragile prosperity
One-sided poverty Fragile prosperity Secure prosperity
Source: Groh-Samberg (2008).
66
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
Table 2 Distribution of health behavior in the GSOEP population (%)*. Health behavior
Sample
Healthy diet
Total Male Female
Non-smoking
Total Male Female
Sufficient physical activity
Total Male Female
*Weighted results,
#
Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No
2004
2005
2006
2007
2008
2009
2010
45.44 54.56 36.07 63.93 54.58 45.42 64.60 35.40 60.45 39.55 68.64 31.64
#
44.95 55.05 35.12 64.88 54.35 45.65 64.30 35.70 61.17 38.83 67.29 32.71
#
#
43.68 56.32 32.54 67.46 54.23 45.77 66.83 33.17 62.87 37.13 70.56 29.44
#
33.72 66.28 33.36 66.64 34.07 65.93
#
35.49 64.51 34.25 65.75 36.67 63.33
44.98 55.02 35.01 64.99 54.51 45.49 66.42 33.58 62.65 37.35 70.02 29.98 40.67 59.33 39.67 60.33 41.63 58.37
# # # # #
# # # # # # # # # # #
# # # # # # # # # #
# # # # #
# # # # # # # # # # #
37.64 62.36 38.77 61.23 36.45 63.55
# # # # # #
Data not available.
4.2.1. Persistent at-risk-of-poverty rate Results presented in Table 5 show the logistic regression models using the persistent at-risk-of-poverty rate for 2006e2010 in the regressions regarding dietary behavior and smoking and for 2005e2009 for physical activity. For the categories diet and smoking, odds ratios are not significantly different from one for individuals who are persistently at risk of poverty compared to the reference group of prosperity. In contrast, persistent risk of poverty leads to a lower likelihood of being sufficiently physically active (OR ¼ 0.795). However, these results are only significant for females (OR ¼ 0.702), but not for men (OR ¼ 0.987) in the subsample analyses. Results underline that persistent risk-of-poverty affects health behavior in different ways, potentially because of gender-specific coping mechanisms with being at the risk of poverty.
Table 3 Concurrent and persistent at-risk-of-poverty rate based on ENIEU (%). Study period
#
Sample size (N)
Measures for individual years 2000 19,498 2001 17,984 2002 19,530 2003 18,495 2004 18,081 2005 17,384 2006 18,393 2007 17,228 2008 16,092 2009 16,878 2010 15,119 Measures for five-year periods 00e04 14,298 01e05 13,806 02e06 13,165 03e07 13,864 04e08 13,128 05e09 12,246 06e10 12,284
At-risk-of-poverty rate (%)
Threshold (Euro)
11.85 12.26 12.76 13.43 14.50 15.27 16.15 14.95 15.88 15.70 16.17
10,222.33 10,522.00 10,566.60 10,878.57 10,971.00 10,965.71 11,045.43 11,101.20 11,594.29 11,827.30 12,185.10
10.17 11.12 11.49 12.64 12.79 13.06 13.59
10,326.24 10,555.87 10,907.66 10,986.06 11,056.52 11,238.41 11,579.46
4.2.2. Persistent combined poverty indicator Results are presented in Table 6. Compared to Table 5, the sample size is reduced due to missing data on deprivations. First, dietary behavior differs significantly in all groups from that of the
Table 4 Concurrent and persistent poverty based on combined poverty indicator (%). Study period
N
Measures for individual years 2000 19,498 2001 17,984 2002 19,530 2003 18,495 2004 18,081 2005 17,384 2006 18,393 2007 17,228 2008 16,092 2009 16,878 2010 15,119 Measures for five-year periods 00e04 12,832 01e05 12,464 02e06 13,332 03e07 12,722 04e08 11,175 05e09 11,337 06e10 11,466
Poverty
Precarity
Prosperity
Extreme poverty
Moderate poverty
One-sided poverty
Vulnerability
Fragile prosperity
Secure prosperity
4.23 4.84 5.28 6.65 6.56 7.71 7.97 8.55 7.68 7.76 5.70
8.24 8.71 9.63 9.16 9.21 11.02 10.78 10.56 10.56 10.03 10.27
8.67 7.94 7.87 8.39 7.77 8.05 8.99 8.38 8.57 10.76 8.09
5.94 6.03 7.45 6.98 8.21 6.75 6.36 6.61 6.10 6.05 6.19
24.06 21.25 23.40 20.53 22.97 20.80 22.00 20.47 22.67 21.86 22.63
48.87 51.24 46.36 48.29 45.28 45.67 43.90 45.44 44.42 43.55 47.12
2.58 3.05 4.62 4.89 4.63 5.28 4.70
5.23 5.82 5.90 7.10 6.21 6.82 7.30
2.15 2.43 2.14 2.08 2.17 2.08 2.02
9.16 9.35 9.93 9.38 9.89 9.42 8.34
23.98 22.09 23.70 22.55 23.98 24.03 23.61
56.89 57.26 53.70 54.01 53.11 52.37 54.02
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
67
Table 5 Resultsa,b,c of the logistic regression models using the persistent at-risk-of-poverty rate of 2006e2010d and 2005e2009.e Dependent variables
Diet (healthy diet ¼ 1) N ¼ 12,221 Smoking (non-smoking ¼ 1) N ¼ 12,221 Physical activity (sufficient physical activity ¼ 1) N ¼ 12,221 a b c d e
Independent variables
Total sample
Male sample
Female sample
Odds ratio
95% CI
Odds ratio
95% CI
Odds ratio
95% CI
At-risk-of-poverty
0.933
[0.751, 1.160]
1.098
[0.793, 1.522]
0.868
[0.655, 1.151]
At-risk-of-poverty
0.890
[0.723, 1.095]
0.838
[0.618, 1.137]
0.888
[0.666, 1.184]
At-risk-of-poverty
0.795*
[0.620, 1.019]
0.987
[0.677, 1.439]
0.702**
[0.506, 0.975]
Reference group ¼ prosperity. Results based on estimation including confounders. Full estimation results available upon request. Significance level:***,**,* ¼ 1%, 5%, 10% and indicated in bold. Regarding dietary behavior and smoking. Regarding physical activity.
Table 6 Resultsa,b,c of the logistic regression models using the persistent combined poverty indicator of period of 2006e2010d and 2005e2009.e Dependent variables
Diet (healthy diet ¼ 1)
N ¼ 10,464 Smoking (non-smoking ¼ 1)
N ¼ 10,464 Physical activity (sufficient physical activity ¼ 1)
Independent variables
Total sample
Male sample
Odds ratio
95% CI
Extreme poverty Moderate poverty One-sided poverty Vulnerability Fragile prosperity
0.666** 0.618*** 0.580** 0.712*** 0.835**
[0.453, [0.460, [0.370, [0.560, [0.714,
Extreme poverty Moderate poverty One-sided poverty Vulnerability Fragile prosperity
0.369*** 0.612*** 0.695* 0.621*** 0.875
Extreme poverty Moderate poverty One-sided poverty Vulnerability Fragile prosperity
0.409*** 0.408*** 0.652 0.621*** 0.771***
Female sample
Odds ratio
95% CI
0.979] 0.831] 0.910] 0.906] 0.976]
1.027 0.635* 0.744 0.755 0.824
[0.583, [0.389, [0.391, [0.520, [0.652,
Odds ratio
95% CI
0.510 *** 0.607*** 0.473** 0.677** 0.841
[0.315, [0.417, [0.264, [0.496, [0.680,
0.824] 0.883] 0.845] 0.925] 1.040]
[0.256, [0.460, [0.461, [0.495, [0.738,
0.530] 0.814] 1.046] 0.778] 1.036]
0.326*** 0.472*** 0.515** 0.518*** 0.771**
[0.19, 0.559] [0.312, 0.716] [0.295, 0.900] [0.377, 0.711] [0.608, 0.977]
0.398*** 0.753 0.970 0.753* 1.012
[0.241, [0.504, [0.521, [0.543, [0.795,
0.659] 1.127] 1.807] 1.045] 1.288]
[0.250, [0.283, [0.380, [0.471, [0.646,
0.668] 0.589] 1.119] 0.817] 0.920]
0.581 0.456*** 1.181 0.641** 0.873
[0.271, [0.270, [0.527, [0.425, [0.670,
0.321*** 0.365*** 0.383*** 0.592*** 0.690***
[0.179, [0.219, [0.200, [0.407, [0.545,
0.578] 0.607] 0.732] 0.861] 0.873]
1.808] 1.037] 1.417] 1.096] 1.041]
1.243] 0.770] 2.646] 0.967] 1.137]
N ¼ 10,464 a b c d e
Reference group ¼ prosperity. Results based on estimation including confounders. Significance level:***,**,* ¼ 1%, 5%, 10% and indicated in bold. Regarding dietary behavior and smoking. Regarding physical activity.
reference group of prosperity. Extremely poor, moderately poor and one-sided poor individuals (OR ¼ 0.580 to 0.666) are least likely to have a healthy diet, followed by respondents of the groups “vulnerability” and “fragile prosperity” (OR ¼ 0.712 and 0.835). Looking at the male and female samples, significant odds ratios result for moderately poor men (OR ¼ 0.635) and for all women (OR ¼ 0.473e0.677) except women in fragile prosperity. There are also significant differences regarding non-smoking for all groups of poverty and precarity, except “fragile prosperity”. The lowest likelihood is among persistently extremely poor individuals (OR ¼ 0.369). In comparison to the latter result, differences are smaller for persistently moderately poor (OR ¼ 0.612), one-sidedly poor (OR ¼ 0.695) as well as vulnerable individuals (OR ¼ 0.621), who are also less likely to be non-smokers. Significant results on smoking differ if analyses of males and females are separated, and poor men are more likely to be smokers than women. Prosperous individuals are more likely to be sufficiently physically active than their poor and precarious counterparts. The lowest odds ratios are observable for the persistently extremely and moderately poor (OR ¼ 0.409 and 0.408). In addition, individuals who belong either to the vulnerability group (OR ¼ 0.621) or to the fragile prosperity group (OR ¼ 0.771) are less likely to practice
sufficient sports. No significant effects are observable for persistently one-sided poor individuals. The described effects are stronger for female than for male individuals. Persistently extremely poor women are least likely to be sufficiently physically active (OR ¼ 0.321). ORs less than 0.4 are also observable for women who are affected by persistent moderate and one-sided poverty. Persistently precarious women (“vulnerability” and “fragile prosperity”) are also less likely to be sufficiently physically active. In contrast, the significant ORs for men amount to 0.456 for the group of moderate poverty and 0.641 for vulnerability. 4.3. Panel analysis We next turn to panel analysis using the persistent poverty indicators and multiple year observations. Additional results for concurrent poverty are available in the online appendix. 4.3.1. Persistent at-risk-of-poverty rate Results of the fixed-effects logistic regression models using the persistent at-risk-of-poverty rate are presented in Table 7. The dependent variable has been measured every other year between 2004 and 2010 for healthy diet and non-smoking, and in the years 2005 and 2007e2009 for physical activity. The data are
68
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
Table 7 Resultsa,b,c of the fixed-effects logistic regression models using the persistent at-risk-of-poverty rate of 2000e2010d and 2001e2009.e Dependent variables
Independent variables
Diet (healthy diet ¼ 1) At-risk-of-poverty Number of observations: 16,781; number of individuals: 4,486 Smoking (non-smoking ¼ 1) At-risk-of-poverty Number of observations: 5,975; number of individuals: 1,591 Physical activity (sufficient physical activity ¼ 1) At-risk-of-poverty Number of observations: 14,243; number of individuals: 3,853 a b c d e
Total sample
Male sample
Female sample
Odds ratio
95% CI
Odds ratio
95% CI
Odds ratio
95% CI
0.975
[0.822, 1.155]
0.659***
[0.492, 0.883]
1.226*
[0.991, 1.517]
0.839
[0.645, 1.092]
1.267
[0.807, 1.990]
0.685**
[0.492, 0.955]
0.856
[0.698, 1.050]
0.924
[0.664, 1.287]
0.810
[0.622, 1.057]
Reference group ¼ prosperity. Results based on estimation including confounders. Significance level:***,**,* ¼ 1%, 5%, 10% and indicated in bold. Regarding dietary behavior and smoking. Regarding physical activity.
combined with persistent poverty indicators calculated based on data for five consecutive years for 2000e2010 in the first case, and from 2001 to 2009 in the second. The basis for these analyses are five-year periods during which an individual stays in a certain poverty group. As discussed in Section 3.4, only those observations with a change in the dependent variable are included in the fixed effect logistic regression. With regard to the total sample there are no effects of being persistently at risk of poverty and the three categories of health behavior. However, gender-specific effects can be found. Compared to the reference group that is not persistently at-risk-of-poverty, persistently poor men are less likely to follow a healthy diet (OR ¼ 0.659) whereas women are more likely to do so (OR ¼ 1.226). Additionally, women who are persistently at risk of poverty are less likely to be a non-smoker. In conclusion, the results underline the importance of separate analysis for men and women. Effects for being persistently at risk of poverty are only present for men regarding diet and for women concerning diet and smoking. 4.3.2. Persistent combined poverty indicator Panel analyses on health behavior and the combined poverty indicator produce a number of significant differences with regard
to the reference group of secure prosperity (Table 8). Individuals who belong to any of the five groups of persistent poverty or precarity decrease their probability of having a healthy diet. The highest risk of eating unhealthily is indicated for the groups of persistent extreme poverty (OR ¼ 0.485) and persistent moderate poverty (OR ¼ 0.623). Altogether an inverse gradient is observed, so that the odds ratio for a fragile prosperous individual is only 0.839. If the sample is separated by gender, males and females vary regarding dietary behavior. The lowest likelihood of having a healthy diet is among moderate poor males (OR ¼ 0.501) and extreme poor females (OR ¼ 0.469). Furthermore, persistent extreme poor or vulnerable men as well as women who are in the group of persistent moderate poverty, vulnerability and fragile prosperity are less likely to have a healthy diet. Odds ratios range between 0.547 and 0.803. With regard to smoking, results are less clear. Only individuals who are persistently fragile prosperous are more likely to be nonsmokers (OR ¼ 1.253). Interestingly, only vulnerable men are less likely to be non-smokers, whereas ORs for women in the groups of persistent extreme poverty, vulnerability and fragile prosperity have ORs greater than 1 (OR ¼ 1.400 to 1.788). Persistent extreme poverty is associated with the greatest chances of behaving in a
Table 8 Resultsa,b,c of the fixed-effects logistic regression models using the persistent combined poverty indicator of 2000e2010d and 2001e2009.e Dependent variables
Independent variables
Diet (healthy diet ¼ 1)
Extreme poverty Moderate poverty One-sided poverty Vulnerability Fragile prosperity Number of observations: 15,065; number of individuals: 4,057 Smoking (non-smoking ¼ 1) Extreme poverty Moderate poverty One-sided poverty Vulnerability Fragile prosperity Number of observations: 5,319; number of individuals: 1,429 Physical activity (sufficient physical activity ¼ 1) Extreme poverty Moderate poverty One-sided poverty Vulnerability Fragile prosperity Number of observations: 13,221; number of individuals: 3,583
a b c d e
Reference group ¼ prosperity. Results based on estimation including confounders. Significance level:***,**,* ¼ 1%, 5%, 10% and indicated in bold. Regarding dietary behavior and smoking. Regarding physical activity.
Total sample
Male sample
Odds ratio
95% CI
0.485*** 0.623*** 0.768* 0.724*** 0.839***
[0.357, [0.502, [0.578, [0.615, [0.758,
1.411 1.011 0.993 1.038 1.253*** 1.018 1.103 0.837 0.974 0.994
Female sample
Odds ratio
95% CI
Odds ratio
95% CI
0.658] 0.773] 1.021] 0.852] 0.928]
0.547** 0.501*** 0.727 0.683*** 0.916
[0.324, [0.348, [0.438, [0.522, [0.778,
0.923] 0.723] 1.206] 0.894] 1.078]
0.469*** 0.708** 0.793 0.750*** 0.803***
[0.320, [0.539, [0.560, [0.610, [0.704,
0.688] 0.928] 1.121] 0.923] 0.915]
[0.847, [0.710, [0.635, [0.795, [1.058,
2.350] 1.439] 1.553] 1.356] 1.485]
1.177 1.325 1.378 0.696* 1.185
[0.517, [0.756, [0.671, [0.467, [0.933,
2.682] 2.322] 2.831] 1.036] 1.506]
1.788* 0.930 0.908 1.459** 1.400***
[0.914, [0.581, [0.503, [1.004, [1.093,
3.500] 1.487] 1.639] 2.121] 1.794]
[0.637, [0.814, [0.590, [0.788, [0.869,
1.628] 1.495] 1.187] 1.204] 1.136]
2.311** 2.545*** 3.192*** 1.885*** 1.582***
[1.153, [1.537, [1.765, [1.313, [1.266,
4.635] 4.213] 5.774] 2.706] 1.977]
0.734 0.790 0.403*** 0.761** 0.797***
[0.379, [0.532, [0.256, [0.582, [0.671,
1.421] 1.173] 0.634] 0.996] 0.946]
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
health-promoting way in this category. Analyses of physical activity and the persistent combined poverty indicator show that there are no significant differences for the total sample. Nevertheless, significant odds ratios are present for men and women. Men who belong to one of the persistent poverty or precarity groups are more likely to pursue sufficient sports compared to their prosperous counterparts. The highest odds ratios are observable for one-sided poverty (OR ¼ 3.192), moderate poverty (OR ¼ 2.545) and extreme poverty (OR ¼ 2.311). In contrast, women behave differently. Women who are part of the persistent one-sided poverty group are least likely to be physically active (OR ¼ 0.403). Additionally, being affected by persistent vulnerability or fragile prosperity is associated with lower odds ratios (OR ¼ 0.761 and OR ¼ 0.797). No effects are observed for persistent extreme or moderate poverty and pursuing sufficient sports. In conclusion, significant results are mainly observed for dietary behavior. Particularly regarding non-smoking and physical activity, it is important to consider the results by gender. Being affected by persistent precarity and poverty is associated with a higher likelihood of being a non-smoker for women, whereas men are more likely to be physical active. Results of confounding variables are available in the online appendix. For space consideration they are not discussed here. 5. Discussion Our analysis provides new evidence that individuals who are persistently either at risk of poverty, poor, or precarious are less likely to behave in a health-promoting way than their prosperous counterparts. Hence, the importance of persistent poverty on health behavior can be confirmed. The results are based on the analysis of three different types of health behavior using an income-based poverty measure in addition to one that is based on income and four dimensions material deprivation. Using the two different poverty measures reveals the multifaceted aspects of the relation between poverty and health behavior. With regard to diet, we observe a health behavior gradient and a steady decline in following a healthy diet as combined poverty becomes more severe. For other types of health behavior (non-smoking and physical activity), the results are more heterogeneous. Comparing both persistent poverty indicators, the smallest differences between persistently poor and prosperous individuals are reported for the at-risk-of-poverty-rate. It can be concluded that the two determinants e income poverty and the combined poverty indicator e cover different mechanisms explaining health behavior. For this reason they cannot be used interchangeably. This is in line with the literature presented above (cf. Benzeval and Judge, 2001; Groh-Samberg, 2009; Moll, 2006; Nolan and Whelan, 2007). These findings also confirm Mackenbach's “explanation of health inequality” depicted in 2.1. Material factors, including income, influence health behavior directly through resource constraints and via psychosocial factors such as negative life events, daily hassles, “effort-reward imbalance”, or a combination of high demands and low control (Mackenbach, 2006). As a result, a critical material situation can be reinforced by psychosocial stress. Thus, the results show, for the first time in the context of health behavior, the relevance of considering multidimensional aspects of poverty that have been identified by various poverty researchers (cf. Bradshaw and Finch, 2003; Groh-Samberg, 2009; €d, 1995; Mack and Lansley, 1985; Nolan and Whelan, 2007). Hallero Finally, findings of the analyses show that poverty affects several dimensions of health behavior. The greatest differences are observed for tobacco consumption and physical activity, but they are also important for dietary behavior. The multidimensional
69
measurement shows that not only do the most persistent nonprosperous groups, namely those in poverty, but also individuals living in precarious situations report a higher likelihood for healthdamaging behavior than their prosperous counterparts. Panel analyses (Tables 7 and 8) show that health behavior is influenced by changes in poverty status over time. Also here, the weakest impact on health behavior is reported for the at-risk-ofpoverty rate. These findings underline the relevance of persistent poverty (cf. Duncan et al., 1993; Groh-Samberg, 2008). These results are consistent with Benzeval and Judge's (2001) observation that persistent poverty implies greater health risks than occasional episodes. Furthermore, variations between men and women occur and gender specific effects differ from those of the general sample. Additionally, Lynch et al. (1997b) observe lower rates of physical activity for persistently poor individuals. The present analyses show, however, that this is only valid for women using the combined poverty indicator but not for men. Thus, data of GSOEP confirm previous empirical results as well as the “explanation of health inequality” by Mackenbach (2006), which describes an inverse relationship between socio-economic status and dietary behavior, smoking and physical activity. These results emphasize the necessity of supporting individuals in both situations of poverty as well as precarity. This study has several limitations. Data regarding health behavior is part of GSOEP but not a focus. In comparison to pure health (behavior) surveys that consider only limited information on socio-economic aspects, information on health behavior is not very detailed. Nevertheless the GSOEP offers a relevant data base for the questions addressed here, as epidemiological surveys offer fewer details regarding socio-economic aspects which this study can consider in the analyses. In addition, social selectivity must be assumed. Like many empirical surveys, GSOEP is not able to cover homeless persons, illegal immigrants, addicts, or persons who are highly deprived (cf. Groh-Samberg, 2009). In addition, future research could consider the potential correlations among the different types of health behavior in the analysis. Despite these limitations, the results show that not only being at risk of persistent poverty, but also being persistently poor and in precarious situations, is associated with detrimental health behavior. Findings underscore the relevance of cumulative poverty as well as dynamic aspects. We could also show that the relationship between poverty and health behavior is gender-specific. This result has to be taken into consideration for approaches to support health-promoting behavior. Acknowledgments We acknowledge DIW for providing the GSOEP data to us. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.socscimed.2016.01.040. References Adler, N., Ostrove, J., 1999. Socioeconomic status and health: what we know and what we don't. Ann. N. Y. Acad. Sci. 3e15. Anderson, L.R., Mellor, J.M., 2008. Predicting health behaviors with an experimental measure of risk preference. J. Health Econ. 27 (5), 1260e1274. http://dx.doi.org/ 10.1016/j.jhealeco.2008.05.011. Ashworth, K., Hill, M., Walker, R., 1994. Patterns of childhood poverty: new challenges for policy. J. Policy Anal. Manag. 13 (4), 658e680. Atkinson, A.B., Cantillon, B., Marlier, E., Nolan, B., 2002. Social Indicators: the EU and Social Inclusion. Oxford University Press, Oxford. Basiotis, P., Guthrie, J., Bowman, S., Welsh, S., 1995. Construction and evaluation of a diet status index. Fam. Econ. Nutr. Rev. 8 (2), 2e13.
70
K. Aue et al. / Social Science & Medicine 153 (2016) 62e70
Benzeval, M., Judge, K., 2001. Income and health: the time dimension. Soc. Sci. Med. 52, 1371e1390. Berthoud, R., Bryan, M., Bardasi, E., 2004. The Dynamics of Deprivation: the Relationship between Income and Material Deprivation over Time. Retrieved from. http://campaigns.dwp.gov.uk/asd/asd5/rports2003-2004/rrep219.pdf. Bradshaw, J., Finch, N., 2003. Overlaps in dimensions of poverty. J. Soc. Policy 32 (04), 513e525. http://dx.doi.org/10.1017/S004727940300713X. Centers for Disease Control and Prevention, 2008. Fact Sheet for Health Professionals on Physical Activity Guidelines for Adults. Retrieved from. http:// www.cdc.gov/nccdphp/dnpa/physical/pdf/PA_Fact_Sheet_Adults.pdf. Cohen, D.A., Farley, T.A., Mason, K., 2003. Why is poverty unhealthy? Social and physical mediators. Soc. Sci. Med. 57 (9), 1631e1641. Contoyannis, P., Jones, A., 2004. Socio-economic status, health and lifestyle. J. Health Econ. 23, 965e985. Dennis, I., Guio, A., 2003. Poverty and Social Exclusion in the EU after Laeken e Part 1: Population and Social Conditions. Retrieved from. http://ec.europa.eu/ eurostat/documents/3433488/5551812/KS-NK-03-008-EN.PDF/9b0cdf8b-943b4d63-894f-3a0669fe2d93. Duncan, G., Gustafsson, B., Hauser, R., Schmauss, G., Messinger, H., Muffels, R., …, Ray, J.-C., 1993. Poverty dynamics in eight countries. J. Popul. Econ. 6 (3), 215e234. European Union, 2010. Combating Poverty and Social Exclusion: A Statistical Portrait of the European Union 2010: A Statistical Portrait of the European Union 2010. Publications Office of the European Union, Luxembourg. Retrieved from. http://epp.eurostat.ec.europa.eu/cache/ITY_OFFPUB/KS-EP-09-001/EN/ KS-EP-09-001-EN.PDF. Eurostat, 2011. Glossary: At-risk-of-poverty Rate. Retrieved from. http://epp. eurostat.ec.europa.eu/statistics_explained/index.php/Glossary:At-risk-ofpoverty_rate. Goebel, J., Grabka, M., Krause, P., Kroh, M., Pischner, R., Sieber, I., Spieß, M., 2008. €ngsschnittstudie SozioMikrodaten, Gewichtung und Datenstruktur der La oekonomisches panel (SOEP). Vierteljahrsh. Wirtsch. 77 (3), 77e109. Retrieved from. http://ejournals.duncker-humblot.de/doi/pdf/10.3790/vjh.77.3. 77. Groh-Samberg, O., 2008. Persistent poverty is increasing in Germany. In: Heady, B., Holst, E. (Eds.), A Quarter Century of Change: Results from the German Socioeconomic Panel (SOEP), pp. 41e48. Berlin. Retrieved from. http://www.diw. de/documents/publikationen/73/diw_01.c.91272.de/soep_wave_report_2008. pdf. Groh-Samberg, O., 2009. Armut, soziale Ausgrenzung und Klassenstruktur. VS Verlag für Sozialwissenschaften, Wiesbaden. Hagenaars, A., Vos, K., Zaidi, M., 1994. Poverty Statistics in the Late 1980s: Research Based on Micro-data: Research Based on Micro-data. Office for Official Publications of the European Communities, Luxembourg. €d, B., 1994. A New Approach to the Direct Consensual Measurement of Hallero Poverty, vol. 50. Social Policy Research Centre (SPRC). Discussion Paper. €d, B., 1995. The truly poor: direct and indirect consensual measurement of Hallero poverty in Sweden. J. Eur. Soc. Policy 5 (2), 111e129. Helmert, U., 2003. Soziale Ungleichheit und Krankheitsrisiken. Maro, Augsburg. Knoops, K.T.B., Groot, L.C.P.G.M., de, Kromhout, D., Perrin, A., Moreiras-Varela, O., Menotti, A., van Staveren, W.A., 2004. Mediterranean diet, lifestyle factors, and 10-year mortality in elderly European men and women. J. Am. Med. Assoc. 292 (12), 1433e1439. Kroll, L.E., 2010. Sozialer Wandel, soziale Ungleichheit und Gesundheit: Die Entwicklung sozialer und gesundheitlicher Ungleichheiten in Deutschland zwischen 1984 und 2006. VS Verlag für Sozialwissenschaften, Wiesbaden. Krueger, P., Chang, V., 2008. Being poor and coping with stress: health behaviors and the risk of death. Am. J. Public Health 98, 889e896. € konomischen Status in soziaLampert, T., Kroll, L., 2009. Die Messung des sozioo lepidemiologischen Studien. In: Richter, M., Hurrelmann, K. (Eds.), Gesundheitliche Ungleichheit. Grundlagen, Probleme, Perspektiven, second ed. VS Verlag für Sozialwissenschaften, Wiesbaden, pp. 309e334. Leisering, L., 2008. Dynamik von Armut. In: Huster, E.-U., Boeckh, J., MoggeGrotjahn, H. (Eds.), Handbuch Armut und Soziale Ausgrenzung. VS Verlag für Sozialwissenschaften, Wiesbaden, pp. 118e132. Lynch, J., Kaplan, G., Salonen, J., 1997a. Why do poor people behave poorly? Variation in adult health behaviours and psychosocial characteristics by stages of the socioeconomic lifecourse. Soc. Sci. Med. 44 (6), 809e819. Lynch, J., Kaplan, G., Shema, S., 1997b. Cumulative impact of sustained economic hardship on physical, cognitive, psychological, and social functioning. N. Engl. J. Med. 337 (26), 1889e1895. Mack, J., Lansley, S., 1985. Poor Britain. Allen & Unwin, London. Mackenbach, J., 2006. Health Inequalities: Europe in Profile. Retrieved from. http://
ec.europa.eu/health/ph_determinants/socio_economics/documents/ev_ 060302_rd06_en.pdf. Mackenbach, J.P., van de Mheen, H., Stronks, K., 1994. A prospective cohort study investigating the explanation of socio-economic inequalities in health in the Netherlands. Soc. Sci. Med. 38 (2), 299e308. http://dx.doi.org/10.1016/02779536(94)90399-9. Mackenbach, J.P., Stirbu, I., Roskam, A.-J.R., Schaap, M.M., Menvielle, G., Leinsalu, M., Kunst, A.E., 2008. Socioeconomic inequalities in health in 22 European countries. N. Engl. J. Med. 358 (23), 2468e2481. http://dx.doi.org/10.1056/ NEJMsa0707519. McGinnis, J., Foege, W., 1993. Actual causes of death in the United States. JAMA 270 (19), 2207e2212. http://dx.doi.org/10.1001/jama.1993.03510180077038. Mielck, A., 2000. Soziale Ungleichheit und Gesundheit: Empirische Ergebnisse, €rungsansa €tze, Interventionsmo €glichkeiten, first ed. Huber, Bern. Erkla Mokdad, A.H., 2004. Actual causes of death in the United States, 2000. JAMA 291 (10), 1238e1245. http://dx.doi.org/10.1001/jama.291.10.1238. Moll, S., 2006. Dynamik von Armut in Deutschland: Ergebnisse mikrooekonometrischer Analysen. Dissertation. University of Hohenheim. Retrieved from. http://opus.ub.uni-hohenheim.de/volltexte/2006/154/pdf/dissertation. pdf. Moore, M., Hughes, J., 2001. The health care consequences and its regulation. Forum Health Econ. Policy 4, 31e70. Mueller, U., Heinzel-Gutenbrunner, M., 2001. Krankheiten und Beschwerden (subjektive Gesundheit) unter Bewertung der eigenen Gesundheit. Mater. Bev€lkerungswiss. 102c. Retrieved from. http://www.bib-demografie.de/nn_ o 750460/SharedDocs/Publikationen/DE/Download/Materialienbaende/ 102c,templateId¼raw,property¼publicationFile.pdf/102c.pdf. Muffels, R., 1993. Welfare Economic Effects of Social Security. Essays on Poverty. Social Security and Labour Market: Evidence from Panel Data. TISSER, Tilburg Inst. for Social Security Research, Dep. of Social Security Studies, Tilburg. Nolan, B., Whelan, C., 2007. On the multidimensionality of poverty and social exclusion. In: Jenkins, S.P., Micklewright, J. (Eds.), Inequality and Poverty Reexamined. Oxford University Press, Oxford, pp. 146e165. O'Boyle, E.J., 1999. Toward an improved definition of poverty. Rev. Soc. Econ. 57 (3), 281e300. Retrieved from. http://search.ebscohost.com/login.aspx? direct¼true&db¼sih&AN¼2452162&site¼ehost-live. OECD, 1982. The OECD List of Social Indicators. Paris. Organisation for Economic Cooperation and Development, Washington, D.C. Sold by OECD Publications and Information Center. Olshansky, S.J., Ault, A.B., 1986. The fourth stage of the epidemiologic transition: the age of delayed degenerative diseases. Milbank Q. 64 (3), 355e391. Osler, M., 2006. The life course perspective: a challenge for public health research and prevention. Eur. J. Public Health 16 (3), 230. rez-Mayo, J., 2004. Consistent Poverty Dynamics in Spain. In: IRISS Working Pe Paper Series. Retrieved from. http://iriss.ceps.lu/documents/irisswp50.pdf. €rperliche Aktivita €t. Gesundh. des Bund. 26, 13e14. Robert Koch-Institut, 2005. Ko Smith, N., Middleton, S., 2007. A Review of Poverty Dynamics Research in the UK. Joseph Rowntree Foundation, York. Smith, K.R., Zick, C.D., 1994. Linked lives, dependent demise? Survival analysis of husbands and wives. Demography 31 (1), 81e93. Retrieved from. http://search. ebscohost.com/login.aspx?direct¼true&db¼sih&AN¼9410253102&site¼ehostlive. Townsend, P., 1979. Poverty in the UK: A Survey of Household Resources and Standards of Living. Penguin Books, Harmondsworth. Wagle, U., 2002. Rethinking poverty: definition and measurement. Int. Soc. Sci. J. 54 (171), 155e165. http://dx.doi.org/10.1111/1468-2451.00366. Wagner, G., Frick, J., Schupp, J., 2007. The German socio-economic panel study (SOEP): scope, evolution and enhancements. Schmollers Jahrb. 1 (127), 139e170. Retrieved from. http://www.diw.de/documents/publikationen/73/ diw_01.c.60184.de/diw_sp0001.pdf. Wardle, J., Haase, A.M., Steptoe, A., Nillapun, M., Jonwutives, K., Bellisie, F., 2004. Gender differences in food choice: the contribution of health beliefs and dieting. Ann. Behav. Med. 27 (2), 107e116. http://dx.doi.org/10.1207/ s15324796abm2702_5. Whelan, C., Maître, B., 2005. Comparing poverty and deprivation dynamics: issues of reliability and validity. J. Econ. Inequal. 4 (3), 303e323. http://dx.doi.org/ 10.1007/s10888-005-9017-1. Whelan, C., Layte, R., Maître, B., 2001. What is the Scale of Multiple Deprivation in the European Union? Retrieved from. http://www.esri.ie/UserFiles/ publications/20070208160751/OPEA014.pdf. WHO, 2002. The World Health Report 2002-Reducing Risks, Promoting Healthy Life. Retrieved from. http://www.who.int/whr/2002/en/whr02_en.pdf.
