Background
Low socioeconomic status (SES) in childhood has a long-lasting adverse impact on numerous physical and mental health outcomes in adulthood [
1‐
3]. Studies have shown that the experience of having low SES in childhood is associated with poor health in adulthood, largely due to harmful habitual behaviors, such as smoking, maintaining a sedentary lifestyle, poor dietary habits, and excessive drinking [
4]. Childhood poverty is likely to reflect a various aspects of low SES in childhood and affect later-developing health-risk behaviors. Indeed, childhood poverty is often accompanied by parental absence or less parental structure (lack of rules or routines, such as regular bedtimes), poor quality housing, poor diet, and family conflicts. Poverty experienced in childhood is also correlated with an increased risk of academic underachievement and lower income in the future [
5]. With the accumulation of these adversities, children may have reduced resources and opportunities to engage in healthy behaviors throughout their life course [
6,
7].
Adult SES is one possible mediator of the impact of childhood poverty on adult health-risk behaviors. The experience of childhood poverty might reduce an individual’s likelihood of good educational attainment, occupational achievement, and future earning potential [
5‐
7]. These factors could, in turn, lead to an increased likelihood of smoking, excessive alcohol consumption, poor dietary habits, and a sedentary lifestyle [
8,
9]. A number of studies have examined the association of childhood SES, mostly measured by either the father’s occupation or education, with these health-risk behaviors in adulthood [
10‐
13]. These studies generally found significant associations between lower childhood SES and health-risk behaviors, and discovered that these associations are largely explained by an individual’s own educational attainment and occupations. However, very few studies have specifically examined the impact of the experience of childhood poverty [
14,
15], despite the fact that financial deprivation may have a distinctive effect on later health behaviors [
16]. Furthermore, the mediating effect of adult SES on these associations has not been examined.
While social support is often considered to moderate the influence of childhood poverty on later health-risk behaviors, it could mediate this relationship [
17,
18]. Childhood poverty may introduce individuals to certain social networks that could either promote or discourage positive health behaviors [
19]. However, no previous study has examined the mediating effect of adult social support on the impact of the experience of childhood poverty on adult health-risk behaviors.
In the current study, we expand on previous research by examining a broader type of health-risk behavior using cross-sectional data collected from Japanese community residents. We particularly focused on smoking, lack of exercise, poor dietary habits, and excessive drinking, as these health-risk behaviors are considered key behavioral risk factors for a number of diseases and mortality [
20]. To reduce recall bias and the subjectivity of retrospective self-assessment of childhood poverty, we constructed a binary variable of childhood poverty based on an ordered probit model. In this model, a categorical variable of self-reported childhood standard of living was estimated by the results obtained from multiple correspondence analysis of a set of parental SES measures. To further assess the relative importance of adult SES and social support as mediating factors, we used a mediation analysis wherein we examined the differences in the magnitude of the mediating effects among different sets of health-risk behaviors.
Childhood poverty is currently a major policy concern in Japanese society. The relative poverty rate (the ratio of individuals whose household-size-adjusted income is below the poverty line, i.e., 50 % of the median of the household-size-adjusted income of the population) of children in Japan was ranked 9th out of those of 35 developed countries, with 14.9 to 15.7 % of all children living in the country being under this poverty line [
21,
22]. Single-parent families are at greater risk of experiencing poverty in Japanese society; in 2012, the relative poverty rate in single-parent households was 54.6 %, while that of the general Japanese population was 16.1 % [
21]. A recent study also showed that Japanese children have relatively lower levels of material well-being (i.e., monetary and material deprivation) compared with children of other advanced countries, while other dimensions of child well-being, such as education and health-risk behaviors, are relatively favorable [
22]. As such, empirical evidence is needed on the impact of childhood poverty on later life, as this will help support the development of mitigating policy measures.
Results
In brief, the results indicated that adult SES and social support together mediated a substantial portion of the impacts of the experience of childhood poverty on adult health-risk factors. Specifically, we found that the experience of childhood poverty was a reliable predictor of smoking, lack of exercise, and poor dietary habits (Table
3). Furthermore, educational attainment, household income, occupational status, and social support were associated with experience of childhood poverty (Table
4). After controlling for adult SES and social support, the strength of the association between the experience of childhood poverty and adult health-risk behaviors was reduced (Table
5). Adult SES and social support had the largest mediating effect on the impact of the experience of childhood poverty on smoking (Table
6). Among the mediators investigated, educational attainment had the largest mediating effect for smoking (Table
6).
We analyzed the data of 3836 respondents (men:
n = 1777; women:
n = 2059), excluding 281 responses that were missing data on key variables for the analysis. Table
1 displays the sociodemographic/economic characteristics of respondents and the proportions of respondents who reported health-risk behaviors in the total sample and in the subsamples according to experience of childhood poverty.
Table 1
Socio-demographic characteristics and health-risk behaviors of the total sample and subsample of childhood poverty
Sex | Men | 46.3 | 43.9 |
Women | 53.7 | 56.1 |
Age (years) |
M (SD) | 37.6 (7.22) | 38.7 (6.9) |
25–29 years | 18.4 | 14.1 |
30–39 | 21.8 | 21.3 |
40–50 | 20.9 | 24.1 |
Marital status | Married | 70.8 | 73.6 |
Unmarried/separated | 29.2 | 26.4 |
Education | College or higher | 43.7 | 28.4 |
Junior college | 12.3 | 11.3 |
High school or lower | 44.0 | 60.3 |
Occupation | Regular employment | 51.6 | 48.8 |
Non-regular employment | 24.0 | 26.0 |
Self-employed | 5.7 | 6.3 |
Unemployed | 1.9 | 2.0 |
Homemaker/other | 16.8 | 16.9 |
Social support | High | 31.4 | 32.1 |
Middle | 32.3 | 31.5 |
Low | 35.1 | 36.8 |
Smoking | | 23.7 | 27.2 |
Lack of exercise | Seldom | 41.7 | 47.2 |
Less than once per month | 61.2 | 64.3 |
Less than 1–2 days per week | 81.2 | 81.6 |
Poor dietary habits | | 23.5 | 24.7 |
Excessive drinking | | 8.8 | 9.1 |
Number of observations | | 3836 | 1279 |
Table
2 shows the results of the multiple correspondence analysis. As shown in the table, the first dimension explained 41.8 % of the total inertia, while the explanatory power declined to 17.7 and 11.7 % for the second and third dimensions, respectively. These three dimensions captured 71.2 % of the variance of the set of variables denoting parental SES. We utilized the row coordinates corresponding to these three dimensions to explain participants’ recall of their standard of living at age 15, based on the ordered probit model. The rightmost two columns of Table
2 report the estimated coefficients—all of which were highly significant—on the row coordinates corresponding to the three dimensions and their standard errors. Based on these results, we constructed a binary variable of childhood poverty, as explained in the Methods section.
Table 2
Results of the multiple correspondence analysis for parental SES variables and ordered probit models to explain childhood poverty (N = 3836)
1 | 0.111 | 41.8 | 41.8 | 0.100***
| (0.018) |
2 | 0.047 | 17.7 | 59.5 | 0.248***
| (0.018) |
3 | 0.031 | 11.7 | 71.2 | −0.218***
| (0.018) |
Table
3 present the estimation results for Models 1, in which adult health-risk behavior was predicted by the experience of childhood poverty without controlling for the potential mediators (i.e., adult SES and social support). As the first step of the mediation analysis, we confirmed that the experience of childhood poverty was positively associated with smoking, lack of exercise, and poor dietary habits, but not with excessive drinking.
Table 3
Estimated associations between the experience of childhood poverty and adult health-risk behaviors (Models 1)a (N = 3836)
Childhood poverty | Yes | 1.53*** (1.30–1.80) | 1.55*** (1.29–1.85) | 1.48*** (1.25–1.74) | 1.12 (0.87–1.43) |
No | 1 | 1 | 1 | 1 |
At the second step of the mediation analysis, we estimated Models 2 to predict adult SES and social support by the experience of childhood poverty. For illustrative purposes, we re-categorized the adult SES and social support variables as binary variables and present their estimated associations with the experience of childhood poverty in Table
4. Specifically, we re-categorized adult SES and social support as follows: “high school or lower educational institution” and higher (educational attainment); “the lowest quintile of household income” and higher (household income); “non-regular employment or unemployment” and others (occupational status); and “the lowest tertile of social support total score” and higher (social support). We found that the experience of childhood poverty had positive bivariate associations with low levels of SES and social support in adulthood. Those who experienced childhood poverty were more likely to have a high school degree or less, earn within the lowest quintile of household income, stay in unstable employment (non-regularly employed or unemployed), and receive the lowest tertile of social support.
Table 4
Estimated associations between the experience of childhood poverty and adult SES and social support (Models 2)a (N = 3836
Childhood poverty | Yes | 1.47*** (1.17–1.83) | 1.58*** (1.33–1.88) | 1.16 (0.99–1.37) | 1.28** (1.09–1.51) |
No | 1 | 1 | 1 | 1 |
After controlling for adult SES and social support in the third step of the mediation analysis, the strength of the association between the experience of childhood poverty and adult health-risk behaviors was reduced. By comparing the results in Tables
3 and
5, we found the odds ratios in response to the experience of childhood poverty declined from 1.53 to 1.21, 1.55 to 1.40, and 1.48 to 1.36 for smoking, lack of exercise, and poor dietary habits, respectively, after controlling for a set of mediators. We did not conduct this third-step analysis for excessive drinking, as it had no association with childhood poverty in the first step.
Table 5
Estimated associations between the experience of childhood poverty and adult health-risk behaviors controlling for adult SES and social support (Models 3)a (N = 3841)
Childhood poverty | 1.21* (1.01–1.43) | 1.40*** (1.22–1.62) | 1.36*** (1.14–1.61) |
Adult SES | | | |
Educational attainment |
College or higher | 1 | 1 | 1 |
Junior college | 1.16 (0.82–1.64) | 1.23 (0.98–1.54) | 0.96 (0.69–1.32) |
High school or lower | 2.44*** (2.03–2.94) | 1.14 (0.983–1.33) | 1.25* (1.04–1.49) |
Household income |
1st quintile (highest) | 1 | 1 | 1 |
2nd quintile | 0.91 (0.70–1.18) | 1.34** (1.08–1.66) | 1.11 (0.86–1.44) |
3rd quintile | 1.04 (0.79–1.36) | 1.66*** (1.33–2.08) | 1.15 (0.88–1.50) |
4th quintile | 1.21 (0.93–1.57) | 1.59*** (1.27–1.98) | 1.24 (0.95–1.61) |
5th quintile (lowest) | 1.13 (0.86–1.49) | 1.88*** (1.49–2.37) | 1.09 (0.83–1.42) |
Occupational status |
Regularly employed | 1 | 1 | 1 |
Non-regularly employed | 1.10 (0.87–1.39) | 0.98 (0.81–1.18) | 0.97 (0.77–1.21) |
Self-employed | 0.74 (0.52–1.06) | 0.83 (0.61–1.12) | 1.05 (0.74–1.48) |
Unemployed | 0.58 (0.31–1.10) | 0.84 (0.51–1.38) | 1.02 (0.59–1.75) |
Homemaker/other | 0.60** (0.44–0.83) | 1.14 (0.92–1.42) | 0.74* (0.55–1.00) |
Adult social support |
High | 1 | 1 | 1 |
Middle | 0.89 (0.72–1.10) | 1.20* (1.02–1.42) | 1.28* (1.04–1.58) |
Low | 1.12 (0.90–1.40) | 1.54*** (1.29–1.85) | 1.63*** (1.30–2.04) |
Finally, as shown in Table
6, the proportions of the impact mediated by adult SES and social support varied substantially across the health-risk behaviors: 64.0 % for smoking, 29.0 % for lack of exercise, and 30.6 % for poor dietary habits. Educational attainment mediated the largest proportion of the impact of the experience of childhood poverty on smoking (58.2 %) and poor dietary habits to a lesser extent (18.4 %), while household income mediated the largest proportion of the impact on lack of exercise (19.1 %). We found no significant mediating effect of occupational status for any of the health-risk behaviors, probably because of its close relationship with educational attainment and household income. The mediating effect of social support was modest, but statistically significant, for lack of exercise (7.1 %) and poor dietary habits (5.7 %), but was non-significant for smoking.
Table 6
Estimated propositions (%) of the direct and mediated effects of the experience of childhood poverty on adult health-risk behaviorsa
Direct (unmediated) effect | 36.0c (4.0–54.7) | 70.6c (53.9–83.3) | 69.4c (44.3–85.3) |
Mediated effect via: |
Adult SES | 62.6c (44.2–94.1) | 23.7c (11.7–39.0) | 23.5c (8.5–45.2) |
Educational attainment | 58.2c (40.7–90.2) | 6.4 (−4.1–17.9) | 18.4c (4.8–37.5) |
Household income | 5.2 (−2.4–14.2) | 19.1c (11.6–31.0) | 4.4 (−4.3–15.0) |
Occupational status | −0.8 (−10.4–6.2) | 1.8 (−8.2–1.8) | 0.8 (−4.8–7.3) |
Adult social support | 1.4 (−1.6–5.3) | 5.7c (2.5–11.2) | 7.1c (2.9–14.9) |
Total | 64.0c (45.3–96.0) | 29.4c (16.7–46.1) | 30.6c (14.7–55.7) |
Total | 100 | 100 | 100 |
Discussion
The current study highlighted the importance of childhood poverty in developing health-risk behaviors over time in a Japanese context. Our findings have provided empirical evidence that the experience of childhood poverty increases the likelihood of engaging in unhealthy behaviors and having an unhealthy lifestyle in adulthood [
14‐
16]. The results were consistent with those of previous studies focusing on specific aspects of childhood SES, such as parental education and occupation, social class, and household income [
4,
10‐
13]. Our findings were also supportive of the notion that unfavorable socioeconomic and psychosocial circumstances carried throughout the life course [
5,
6] play an important role in the association between childhood SES and adult health behaviors [
8,
9].
A key contribution of this study was the quantitative evaluation of each mediator of the impact of childhood poverty on adult health behaviors to determine its relative importance. This will help in prioritizing intervention targets. In our sample, adult SES and social support together mediated 29.4–64.0 % of the impact of childhood poverty on adult health-risk behaviors. We found that educational attainment was a key mediator, especially for smoking, which corroborates with existing literature [
30]. This finding may be because children in poverty tend to be less successful in school, which could constrain the informational, financial, and social resources needed to acquire and maintain favorable health behaviors later in life [
5]. We also found that social support had a modest, but significant, mediating effect on the impact of childhood poverty on poor dietary habits and lack of exercise. Social exclusion and adverse interpersonal experiences accompanying childhood poverty may hinder an individual’s access to social networks that provide positive support for healthy habits [
17‐
19].
Our mediation analysis also showed that direct (unmediated) effects accounted for 36.0–70.6 % of the impact of the experience of childhood poverty of adult health-risk behaviors. These effects might be at least partially accounted for by mediators other than adult SES and social support. For instance, the psychological vulnerability and behavioral dysfunction caused by family turmoil, parental psychopathology, and dysfunctional coping styles often associated with poverty might also contribute to the increase in future health-risk behaviors [
31].
In terms of analytic methodology, one novelty in the current study was the measurement of childhood poverty using limited longitudinal information in a cross-sectional study. Well-designed longitudinal studies have been conducted on the health impacts of childhood SES in New Zealand, the UK, the US, and Scandinavian countries [
32,
33]. However, such research has not been conducted in most other countries, including Japan, because of limited resources for conducting long-term follow-up studies. The assessment of childhood financial condition in cross-sectional studies is often based on retrospective self-reported data, which is highly subjective and is vulnerable to recall bias [
34]. In the current study, we constructed a variable of childhood poverty based on the results of a model wherein respondents’ retrospective and subjective assessment of the living standards in childhood were explained by parental SES variables that were more objective and thus likely less susceptible to recall bias [
27].
It should be noted, however, that our measurement of childhood poverty had some limitations. First, it was not fully free from recall bias, because the parental SES variables were based on the respondents’ retrospective answers as well. Second, a binary variable of childhood poverty seems too limited to represent a concept that is inherently ordinal (i.e., related to distinct social classes), even though it had sufficient statistical power. Third, we measured standard of living only at age 15. As such, we may have failed to capture the potential diversity of changes in living conditions from infancy across respondents. Varied duration and timing of exposure to poverty may yield different consequences for health behaviors [
35]. Thus, longitudinal data are needed to assess childhood poverty more rigorously and to overcome these limitations.
In addition to these limitations on the measurement of childhood poverty, the current study has several features that require us to be cautious in interpreting estimation results. First, our analysis was based on a cross-sectional dataset; thus, causality between adult health-risk behaviors and mediators is not assured. In other words, poor health status caused by health-risk behaviors might have affected adult SES and social support [
36]. Second, the J-SHINE survey had a low response rate, and our respondents reported higher SES relative to the survey respondents of other studies. In addition, the reported health conditions tended to be poorer among the respondents excluded from our analysis due to missing data. These traits of the J-SHINE data may have resulted in an underestimation of the impact of low childhood and adult SES. Third, we defined “lack of exercise” as seldom exercising for 10 min or more, which is less frequent than the nationally recommended Japanese standard [
24]. The results of this study indicate, however, that those who have experienced childhood poverty would be at a greater risk for exercising less than the national standard.
Acknowledgements
We used the data from the Japanese Study of Stratification, Health, Income, and Neighborhood (J-SHINE), for which data collection was supported by a Grant-in-Aid for Scientific Research on Innovative Areas (No. 20240061) from the Ministry of Education, Culture, Sports, Science, and Technology, Japan. The study was financially supported by Grants-in-Aid for Scientific Research (No. 26245039 and 15H03339) from the Ministry of Education, Culture, Sports, Science, and Technology, Japan.
Competing interests
The authors certify that there is no competing interest with any financial organization regarding the material discussed in the manuscript.
Authors’ contributions
All authors were involved in the conception and design of this study. TO was in charge of data analysis, MU drafted the initial manuscript, and MF completed the manuscript with a literature review. All of them read and approved the final version of the manuscript.