On the direct and indirect effects of the Great East Japan earthquake on self rated health through social connections: Mediation analysis
Introduction
The Great East Japan Earthquake that occurred in March 2011 caused enormous damage. A Japanese National Police Agency report (2017) confirmed 15,894 deaths, 6152 injured, and 2562 people missing across twenty prefectures. People who were forced to change their residence because their homes had been burned or otherwise destroyed lost their connections with other people (Hikichi et al., 2017). Therefore, it can be concluded that the earthquake had both direct adverse health effects on its victims, and indirect negative effects due to reduced social connections (Hikichi et al., 2016a; 2016b). In this paper, by analyzing self-rated health (SRH) outcomes, we empirically investigate the Great East Japan Earthquake’s direct and indirect impacts on houses.
We found that a high degree of damage to houses negatively affected victims’ SRH (direct effect), and decreased the level of their social connections, which precipitated harmful effects on their SRH (indirect effect). In this paper, we use the term social connections (hereafter SC) as a concept related to social capital. It has also been shown that although the direct impacts of earthquake damage disappear relatively quickly, the indirect effects of a decrease in social connections remain for some time. In other words, the direct health effects caused by earthquakes resolve in the short term, but the earthquake’s negative aftereffects, which include a reduction in social connections, may continue to be felt for a relatively long period of time. We conducted a sensitivity analysis of the robustness of the results noted above. The robustness check shows that the results of the mediation analysis are established only within a fairly limited range. This means that the correlation between the SC and SRH error terms must be less than 0.068, in order to allege these results. This suggests the need to conduct further verification analysis using methods that will prove a causal relationship. We then performed an Instrumental Variable (IV) estimation, to verify the robustness of the assumptions that are necessary to assure the validity of the estimated results of the mediation analysis.
The extent to which the houses were destroyed is a natural type of experiment, and we may be able to regard the damage as a random event that occurred irrespective of the individuals’ attributes. However, as for SC, it may be difficult to accept these suppositions, because the following counter-arguments must also be considered: Firstly, there may be the opposite causality that you can interact with people because you are fine (a reverse causality problem). Second, there may be a correlationship between the ability to actively interact with people and the ability to be healthy, so these variables are correlated (an omitted variables problem). If such unobserved factors exit, false observation results are obtained.
In this paper, we use geographical information describing the area in Iwanuma City where the respondents lived before the earthquake. That is, the distance from the coast ≒ the magnitude of the effects of the tsunami is used as the principal instrumental variable. In addition, we also use the type of ownership (such as owned or rented) and the type of building (such as single-family or apartment) as instrumental variables. The results of the IV estimation show that there are larger mediation effects, and that in the short-term, the indirect impacts of the earthquake made up about 55 % of all the impacts. We also used these instrument variables to test whether the extent of housing damage and the SC variables are endogenous variables. The results show that the extent of housing damage is exogenous, and SC is endogenous to SRH.
In recent years, a large number of papers have examined the influences of social connections on health (Choi et al., 2014; De Silva et al., 2005; Gargiulo and Benassi, 1999; Lund et al., 2010). In particular, after Kawachi et al. (1997) showed that social connections are related to overall mortality, the relationship between high levels of social connections and good health outcomes has become the subject of intensive research. Social connections can be captured at the individual or regional level. In this paper, we focused on the individual’s level of social connections, mainly because we assessed the earthquake’s impact in a small area, Iwanuma city, which has an area of 60.45 km2.
It is difficult to enumerate all the studies that have been conducted on the relationship between social connections and health. This paper will introduce only Kawachi et al. (2008), who surveyed these studies. Several studies have examined the relationship between social connections and health in disaster situations. These include Ali et al. (2012); Beaudoin (2007, 2011), Beiser et al. (2010); Wind et al. (2011); Wind and Komproe (2012); Fergusson et al. (2014); Frankenberg et al. (2008); Kumar et al. (2007), and Van Griensven et al. (2006), but, with the exception of Fergusson et al., 2014 few studies have dealt with panel data gathered before and after the occurrence of natural disasters. In addition, Hikichi et al. (2016a, b), and Tsuboya et al. (2016) used the data analyzed in this study in a previous study. This article adopts an original approach by considering the broad concept of SRH as outcomes, an analysis that has not been conducted elsewhere.
In this study, we use subjective variables based on recollections to ask questions about changes in health conditions before and after the disaster. This step is often considered a weak point in the research process. However, in a study examining the quality of life (QOL) this point can be advantageous, because subjective variables evaluate their own health based on some criteria present in their own minds. This also applies to subjective variables such as life satisfaction. After experiencing a huge influence such as the Great East Japan Earthquake, the health conditions of the people in the respondents’ circle have often deteriorated. Based on such circumstances, the respondents’ subjective health levels will have great value, even if their health levels have not changed. Also, if they become more aware of the value of health, they will be biased to report that their health levels have increased. In other words, by changing the evaluation criteria, it becomes difficult to evaluate changes in subjective variables even within the same individual. When asking respondents to recollect changes before and after the disaster, we may be able to obtain more accurate information, since the evaluation criteria have not been changed.
Several studies on life satisfaction and happiness have reported that changes in these evaluation criteria actually occurred as a result of the impacts of the Great East Japan Earthquake. Some papers report that this earthquake increased the level of happiness in Japan (Ishino et al., 2011; Yamamura et al., 2015), or at least no statistically significant nation-wide drop in happiness has been observed after the disaster (Tiefenbach and Kohlbacher, 2015). We think the reason these results were observed is that the logic described above is working, because there is little objective evidence that social and economic conditions improved after the earthquake. Unless the evaluation criteria are unified when subjective evaluations are made, an accurate analysis cannot be conducted. Research on anchoring vignettes is underway, as a means of dealing with these problems (Grol-Prokopczyk et al., 2011; Salomon et al., 2004).
Section snippets
Data
The Japan Gerontological Evaluation Study (JAGES) is a nationwide cohort study initiated in 2010, by a research group established for the purpose of prospectively investigating personal and regional predictors of healthy aging. We used two of the JAGES cohort’s investigations conducted in 2010 (baseline) and 2013 (following the aftermath of the Great East Japan Earthquake) for this study. This study’s research profile has been described in detail earlier by Hikichi et al. (2016a,b), and Tsuboya
Causal/Sensitivity mediation analysis
The Mediation Analysis Framework is shown in Fig. 1 below.
In our analyses, these acronyms correspond to the following variables:
Y: State of Health (Self Rated Health)
M: Social Connections
T: Degree of Overall Housing Damage
SRH is composed of SRH1, SRH2, and SRH3. These are the responses to the questions asked about changes in respondents’ health conditions before and after the disaster. More specifically, SRH 1 is the response given to the following question: How had your health condition
Conclusion
In this research, we found that a high degree of housing damage not only affected victims’ SRH (direct effect) negatively, but it also decreased the level of social connections, which subsequently had a harmful effect on SRH (indirect effect). We also found that although the direct impact of earthquake damage disappears relatively quickly, the indirect effects of a decrease in social connections remain for a while. To consider the possible problem of endogeneity, we conducted a first-difference
Funding
We appreciate the support and cooperation of the Iwanuma Mayor’s office, and the staff of the Department of Health and Welfare of Iwanuma city government. This work was supported by a grant from the National Institutes of Health (R01 AG042463); Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (KAKENHI 15H01972, KAKENHI 23243070, KAKENHI 22390400, KAKENHI 22592327 and KAKENHI 24390469); a Health Labour Sciences Research Grant from the Japanese Ministry of
References (33)
- et al.
Assessment of prevalence and determinants of posttraumatic stress disorder in survivors of earthquake in Pakistan using Davidson Trauma Scale
J. Affect. Disord.
(2012) - et al.
Human-initiated disaster, social disorganization and post-traumatic stress disorder above Nigeria’s oil basins
Soc. Sci. Med.
(2010) - et al.
Does social participation improve self-rated health in the older population? A quasi-experimental intervention study
Soc. Sci. Med.
(2013) - et al.
Poverty and common mental disorders in low and middle income countries: a systematic review
Soc. Sci. Med.
(2010) - et al.
Predictors of depressive symptoms following the Great East Japan Earthquake: a prospective study
Soc. Sci. Med.
(2016) - et al.
The mechanisms that associate community social connections with post-disaster mental health: a multilevel model
Soc. Sci. Med.
(2012) - et al.
The moderator–mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations
J. Pers. Soc. Psychol.
(1986) News, social connections and health in the context of Katrina
J. Health Care Poor Underserved
(2007)Hurricane Katrina: addictive behavior trends and predictors
Public Health Rep.
(2011)- et al.
Problems with instrumental variables estimation when the correlation between the instruments and the endogenous explanatory variable is weak
J. Am. Stat. Assoc.
(1995)