سازگاری روانشناختی با زمین لرزه بزرگ هانشین آواجی در سال 1995: 16 سال بعد قربانیان هنوز سطوح پایین تر از بهزیستن ذهنی را گزارش می دهند

Abstract We investigated psychological adaptation to the Great Hanshin-Awaji Earthquake of 1995, using surveys conducted in 2001, 2003, 2005, and 2011. Respondents whose houses were damaged reported lower life satisfaction, more negative affect, and more health problems than those who did not suffer any damage in all surveys, including in 2011, or 16 years after the earthquake. Likewise, residents with at least one immediate family member who died in the earthquake reported lower life satisfaction, more negative affect, and more health problems than those who did not have any immediate family members killed in all surveys, including in 2011. Surprisingly, the effect of housing damage on subjective well-being remained significant, above and beyond human loss. Equally important, the 2011 survey data showed that pre-existing differences in socioeconomic status between the victims of housing damage and human loss did not change our main findings.

. Introduction Humankind has had to deal with all kinds of natural disasters since its earliest days on earth. From the time of ancient Greek epic poems and tragedies, stories around the world have told of our struggles with such events (Nussbaum, 1986/2000). How do people adapt to a major natural disaster such as an earthquake? We examined this question using surveys sampling victims of the 1995 Hanshin-Awaji Earthquake. Psychologists started to investigate hedonic reactions to major life events such as becoming disabled or winning the lottery in the late 1970s (e.g., Brickman, Coates, & Janoff-Bulman, 1978). Initial findings suggested that humans are capable of adapting to a diverse array of life events. Over the last decade, there have been large longitudinal studies aimed at studying these processes (Lucas, 2007; Lucas, Clark, Georgellis, & Diener, 2004). The main findings from these newer studies were that people tend to adapt to a positive life event, such as marriage, quickly (within a year or so), whereas people tend not to fully adapt to a major negative life event such as unemployment, widowhood, or divorce for an extended period of time (e.g., 7 years, see Diener et al., 2006 and Lucas, 2007 for review). Despite the active research on hedonic adaptation over the last decade, however, we know relatively little about human adaptation to major natural disasters. What research does exist suggests that many victims suffer from post-traumatic stress disorder and depressive symptoms, though many of them also report personal growth over time (e.g., Lowe, Manove, & Rhodes, 2013). Sri Lankans who were victims of the 2004 tsunami in that country reported higher levels of depressive symptoms one year later than those who did not experience the tsunami (Wickrama & Ketring, 2011). However, the victims’ depressive symptoms and physical health conditions did not differ from those of Sri Lankans who did not experience the tsunami 3 years later. Another study, of particular relevance to ours (Merdjanoff, 2013), analyzed Hurricane Katrina victims’ emotional distress roughly one year after Katrina. The victims’ level of distress remained quite high one year out, almost equivalent to that associated with mild mental illness. Furthermore, victims whose houses suffered serious damage were substantially more distressed than those whose houses were not damaged. Similarly, among the victims of two natural disasters in Greece, an earthquake and a wildfire, damage to property was a significant predictor of psychological distress one month after the earthquake and 11 months after the wildfire (Papanikolaou, Adamis, Mellon, Prodromitis, & Kyriopoulos, 2011). To our knowledge, there is only one study that examined psychological adaptation to a major disaster over a more substantial period of time (McFarlane & Van Hooff, 2009). This study found that Australians who experienced a major wildfire in childhood had a higher probability of lifetime anxiety disorder than non-victims roughly 20 years later; though they were no more likely to suffer from a variety of other disorders measured (i.e., lifetime depressive disorder, any DMV-IV disorder, PTSD, or an eating disorder). None of the aforementioned studies examining adaptation to a major disaster, however, assessed positive aspects of well-being such as life satisfaction. The current research investigates psychological adaptation of the victims of the 1995 Hanshin-Awaji Earthquake (Magnitude 7.3). The Hanshin-Awaji Earthquake took place on January 17, 1995. Its epicenter was close to the city center of Kobe, resulting in the death toll of over 6400, and the total destruction of over 100,000 homes in the city and its surrounding areas. To put the magnitude of the destruction in Kobe and its surrounding areas in perspective, the largest casualty count of any major earthquake in the U.S. over the last 100 years was 165 deaths in the 1946 Aleutian Islands earthquake and tsunami. Within the last 50 years, the largest casualty count of any earthquake in the U.S. was 65 deaths in the 1971 San Fernando earthquake (Magnitude of 6.6). The 1989 San Francisco-Oakland earthquake (also known as the Loma Prieta earthquake) and the 1994 Northridge earthquake, the two earthquakes most similar to the Hanshin-Awaji in a sense that they were close to a large city, had a magnitude of 6.9 and 6.7 and resulted in 63 and 33 deaths, respectively. Such comparisons (e.g., over 6400 deaths versus less than 70 deaths in the most major U.S. earthquakes over the last 50 years) make it clear that the 1995 earthquake in Kobe and its surrounding areas was a truly major natural disaster that likely affected the Kobe area’s 3.3 million residents for an extended period of time. Based on Merdjanoff (2013) and Papanikolaou et al. (2011), we predicted that housing damage would be an important predictor of the psychological adaptation of victims. In addition to housing damage, we also examined human loss, as previous research has shown relatively prolonged adaptation to the loss of a spouse or a child (Lehman et al., 1987 and Lucas et al., 2003). Because social relationships are often one of the strongest predictors of subjective well-being (e.g., Diener & Seligman, 2002), we expected that the effect of human loss would be stronger and more enduring than that of housing damage. Our survey data are ideally suited to examine the long-term effect of housing damage and human loss, as subjective well-being was assessed 6 years (2001), 8 years (2003), 10 years (2005) and 16 years (2011) after the earthquake.

3. Results The descriptive and test statistics are shown by year and by the degree of housing damage for each well-being scale in Table 1. First, the 2001 respondents whose houses/apartments were completely destroyed or burned down by the 1995 earthquake reported lower levels of life satisfaction (r[1193] = −.07, p = .015), and higher levels of negative affect (r[1125] = .14, p < .001) and health problems (r[1130] = .15, p < .001). Likewise, the 2003 respondents whose houses/apartments were completely destroyed or burned down by the 1995 earthquake also reported lower levels of life satisfaction (r[1, 470] = −.13, p < .001), and higher levels of negative affect (r[1, 435] = .14, p < .001) and health problems (r[1, 432] = .13, p < .001). The results were nearly identical among the 2005 participants: again, such respondents reported lower levels of life satisfaction (r[966] = −.15, p < .001), and higher levels of negative affect (r[946] = .21, p < .001) and health problems (r[944] = .19, p < .001). Finally, the 2011 participants whose houses/apartments were totally destroyed or burned down by the 1995 earthquake also reported lower levels of life satisfaction (r[1297] = −.11, p < .001), and higher levels of negative affect (r[1281] = .10, p < .001) and health problems (r[1278] = .08, p < .001). These residents who lost their houses had still not adapted to their misfortune even 16 years after the event. Table 1. Housing damage in the 1995 earthquake and well-being 6–16 years later (mean, SD). Housing damage in 1995 Year α None Partial Half Complete r 2001 Life satisfaction .78 3.17 (.84) 3.07 (.81) 3.04 (.78) 2.97 (.87) −.07 Negative affect .95 2.02 (.92) 2.18 (.97) 2.35 (1.11) 2.45 (1.17) .14 Health problems .95 1.49 (.69) 1.61 (.75) 1.70 (.85) 1.89 (.93) .15 N 212–221 521–551 218–231 176–192 2003 Life satisfaction .87 3.22 (.83) 3.02 (.84) 2.96 (.87) 2.85 (.83) −.12 Negative affect .94 2.02 (.97) 2.32 (1.04) 2.45 (1.12) 2.50 (1.06) .14 Health problems .90 1.52 (.72) 1.74 (.82) 1.80 (.89) 1.90 (.91) .13 N 250–255 638–655 297–304 247–258 2005 Life satisfaction .87 3.22 (.85) 3.14 (.82) 3.00 (.85) 2.83 (.90) −.15 Negative affect .94 2.06 (1.01) 2.21 (.97) 2.36 (1.00) 2.78 (1.17) .21 Health problems .89 1.50 (.69) 1.62 (.76) 1.66 (.78) 2.05 (1.01) .19 N 181–188 436–443 189–193 140–144 2011 Life satisfaction .88 3.20 (.83) 3.16 (.86) 2.96 (.83) 2.96 (.85) −.12 Negative affect .92 2.19 (.88) 2.32 (.94) 2.43 (.96) 2.48 (1.06) .11 Health problems .88 1.57 (.77) 1.58 (.75) 1.66 (.71) 1.73 (.84) .08 N 228–230 519–528 273–279 259–262 Meta-Analysis Across 4 Surveys r (95% C.I) Life satisfaction −.114 −.107;−.121 Negative affect .146 .107; .184 Health problems .133 .083; .184 Note.∗p < .05. ∗∗p < .01. The effect size r = −.114, .146, and .133 are equivalent to Cohen’s d = −.230, .295, and .268, respectively. Table options 3.1. What about human loss? Previous research on psychological adaptation to subjective well-being examined human loss (e.g., Lehman et al., 1987 and Lucas et al., 2003). Thus, we next examined whether respondents who lost a family member would report lower levels of well-being years later. As seen in Table 2, in every survey those who lost a family member in the earthquake reported lower levels of life satisfaction (Cohen’s ds ranged from −.18 to −.36 with a meta-analytically derived weighted mean effect size of −.30), higher levels of negative affect (Cohen’s ds ranged from .21 to .75 with a meta-analytically derived weighted mean effect size of .30), and more health problems (Cohen’s ds ranged from .56 to .67 with a meta-analytically derived weighted mean effect size of .62) than those who did not lose any family members in 1995. In other words, even 16 years after the earthquake, respondents who lost a family member reported substantially higher levels of health problems such as headaches and chest pain, slightly lower levels of life satisfaction and slightly higher levels of negative affect. Table 2. Human loss in the 1995 earthquake and well-being 6–16 years later (mean, SD). Year No loss Loss t-value df Cohen’s d 2001 Life satisfaction 3.10 (.813) 2.89 (.821) −3.10 1186 −.18 Negative affect 2.18 (1.01) 2.49 (1.11) 3.57 1120 .21 Health problems 1.60 (.767) 1.93 (.920) 4.28 194.16 .61 N 2003 Life satisfaction 3.11 (.833) 2.74 (.797) −6.69 1375 −.36 Negative affect 2.22 (1.03) 2.61 (1.06) 5.48 1342 .30 Health problems 1.66 (.793) 1.99 (.928) 5.43 379.28 .56 N 272–280 1070–1097 2005 Life satisfaction 3.15 (.823) 2.76 (.842) −5.85 909 −.39 Negative affect 2.17 (.964) 2.71 (1.13) 6.00 258 .75 Health problems 1.58 (.714) 1.96 (.999) 4.89 228.67 .65 N 181–186 710–725 2011 Life satisfaction 3.15 (.846) 2.86 (.811) −4.90 1254 −.28 Negative affect 2.28 (.940) 2.61 (1.01) 4.74 1238 .27 Health problems 1.55 (.721) 1.92 (.875) 5.97 314.14 .67 N 236–244 1003–1012 Meta-Analysis Across 4 Surveys d (95% C.I) Life satisfaction −.299 −.153; −.446 Negative affect .295 .051; .538 Health problems .618 .564; .672 Note. Degrees of freedom were smaller for health problems in all surveys and for negative affect for the 2005 survey because the homogeneity of variance assumption was violated (i.e., variance differed significantly). When calculating Cohen’s d, the following formula was used: 2t/SQRT(df). When the original t-values with the original degrees of freedom were used for health problems in all 4 surveys and negative affect for the 2005 survey (i.e., when the violation of the homogeneity of variance assumption was ignored), the meta-analytic weighted d changed to .346 for health problems (95% C.I = .282; .410), and .345 (95% C.I = .171; .522) for negative affect. Table options 3.2. Are housing damage and human loss associated with socioeconomic status (SES)? The observed associations between housing damage and human loss in 1995 and subsequent well-being could be due to a variety of third variables. One such candidate is the socioeconomic status of participants. For instance, it is reasonable to speculate that low SES individuals lived in homes that were shabbily constructed, and that their housing damage was therefore larger than those high in SES. The 2011 survey provided information regarding occupational status and homeownership in 1995. Thus, we were able to check to see if housing damage was associated with occupational status. There were 205 respondents (out of 1352 respondents) who had been CEOs, lawyers, doctors, professors, or held managerial positions at the time of the earthquake. As expected, high SES participants had slightly less housing damage (M = 1.30, SD = .947) than others (M = 1.47, SD = 1.02), t(296.15) = 2.32, p = .021, d = .27. If the association between housing damage in 1995 and well-being in 2011 was due to socioeconomic status in 1995, the inclusion of 1995 SES should reduce the effect of housing damage on well-being in 2011. Multiple regression analyses showed, however, that controlling for occupational status, housing damage in 1995 still predicted life satisfaction in 2011, b = −.069 (SE = .016), β = −.118, t(1296) = −4.27, p < .001, negative affect (b = .066, SE = .018, β = .100, t[1280] = 3.59, p < .001), and health problems (b = .040, SE = .015, β = .077, t[1277] = 2.76, p = .006). The standardized regression coefficients were nearly identical to the original simple correlation coefficients reported above (r = −.11 vs. β = −.12 for life satisfaction; r = .10 vs. β = .10 for negative affect; r = .08 vs. β = .08 for health problems), suggesting the effect of SES in 1995 is small. Another indicator of SES is homeownership, as homeowners tend to be wealthier than renters. Thus, we next examined whether homeowners suffered less than renters. In the 2011 survey, there were 905 respondents that had owned a house or condominium in 1995, whereas 431 had rented. Homeowners were not different from renters in terms of housing damage (Mowner = 1.45, SD = .965 vs. Mrenter = 1.42, SD = 1.10), t(756.94) = −.537, p = .591, d = −.04. Not surprisingly, multiple regression analyses showed that controlling for homeownership in 1995 did not change the association between housing damage in 1995 and well-being in 2011: life satisfaction (b = −.071, SE = .016, β = −.121, t[1296] = −4.44, p < .001), negative affect (b = .067, SE = .018, β = .102, t[1280] = 3.66, p < .001), and health problems (b = .042, SE = .015, β = .080, t[1277] = 2.89, p = .004). Thus, the 2011 survey data suggested that housing damage was not strongly covaried with the SES of respondents in 1995; rather housing damage was similar for homeowners and renters, though respondents with high occupational status in 1995 had slightly less housing damage than others. Equally important, however, statistically controlling for the effect of occupational status in 1995, housing damage in 1995 still predicted well-being in 2011. Thus, our results cannot be explained by slight SES differences between those who had greater housing damage and those who did not. Next, we examined whether people with higher SES suffered less human loss than others using the 2011 survey. There was no evidence that human loss was more severe among lower SES groups. Specifically, among the 198 respondents who enjoyed high occupational status in 1995, 37 respondents (or 18.7%) reported losing a family member in 1995. Among the 1088 respondents who did not have high occupational status in 1995, 217 respondents (or 19.9%) reported losing a family member in 1995, χ2 (1, 1286) = .167, p = .683. We also examined another indicator of SES: homeownership in 1995. Here we found the difference. Among 408 renters in 1995, 105 (25.7%) lost a family member, whereas 149 of the 878 home owners (17%) lost a family member, χ2 (1, 1286) = 13.50, p < .001. Thus, we next tested whether the addition of homeownership to the model would reduce the effect of human loss on later well-being. That was not the case. Multiple regression analyses showed that human loss remained a significant predictor of life satisfaction (b = −.135, SE = .030, β = −.126, t[1253] = −4.51, p < .001), negative affect (b = .164, SE = .035, β = .134, t[1237] = 4.73, p < .001), and health problems in 2011 (b = .178, SE = .027, β = .182, t[1237] = 6.50, p < .001), above and beyond homeownership in 1995. Thus, our findings on the association between human loss in 1995 and later well-being cannot be due to a slight SES difference between those who lost a family member and those who did not. 3.3. Controlling for socioeconomic status, gender, and age In addition to SES differences between victims of housing damage and non-victims, the victims of housing damage might have been, by the 2000s, older than other residents of the same area. To take into account these and other such factors, we conducted multiple regression analyses controlling for respondents’ age, gender, and SES. Because the 2001 survey lacked a question assessing household income, however, we simply conducted multiple regression analyses regressing well-being on housing damage, as well as age, gender, home ownership in 1995, and occupational status in 2001. The effect of housing damage was essentially the same as for the correlational analyses: it was negatively associated with life satisfaction (b = -.062, SE = .024, β = −.073, t(1194) = −2.56, p = .001), while positively associated with negative affect (b = .146, SE = .032, β = .134, t(1126) = 4.54, p < .001) and health problems (b = .117, SE = .024, β = .139, t(1131) = 4.76, p < .001). For the 2003 survey, we regressed well-being on housing damage, as well as age, gender, and three indicators of SES: household income in 2003, occupational status in 2003, and home ownership in 1995. The results were again essentially the same as the correlational analyses: life satisfaction dipped (b = −.092, SE = .023, β = −.106, t(1335) = −4.06, p < .001), whereas negative affect (b = .136, SE = .030, β = .124, t(1307) = 4.54, p < .001) and health problems (b = .099, SE = .023, β = .115, t(1304) = 4.23, p < .001) rose with housing damage. For the 2005 survey, we regressed well-being on housing damage, as well as age, gender, and three indicators of SES: household income in 2005, occupational status in 2005, and home ownership in 1995. The results were again essentially the same as the correlational analyses: lower life satisfaction (b = −.122, SE = .029, β = −.135, t(930) = −4.21, p < .001), higher negative affect (b = .206, SE = .036, β = .188, t(913) = 5.77, p < .001) and greater health problems (b = .131, SE = .028, β = .155, t(911) = 4.77, p < .001). Finally, for the 2011 survey, we again regressed well-being on housing damage, as well as age, gender, and three SES variables: household income in 2011, occupational status in 1995, and homeownership in 1995. Those whose houses were totally damaged in 1995 had lower levels of life satisfaction, b = −.089, SE = .023, β = −.105, t(1280) = −3.92, p < .001, and higher levels of negative affect, b = .095, SE = .027, β = .099, t(1262) = 3.55, p < .001, and health problems than those whose houses were not damaged, b = .052, SE = .021, β = .069, t(1260) = 2.48, p = .013. Thus, age, gender, and SES notwithstanding, having one’s house destroyed 16 years earlier was associated with lower levels of subjective well-being and more health problems. 3.4. Is housing damage confounded with human loss? It seemed plausible that individuals whose houses were totally damaged by the earthquake would be among those most likely to have their immediate family members killed in the earthquake. Indeed, across all the surveys, this was found to be the case, r(4, 918) = .27, p < .001. Because of this, it was important to test whether the effect of housing damage could be explained by losing a family member, or whether such damage causes harm over and above bereavement. Controlling for human loss, as well as age, gender, and SES variables, the effect of housing damage remained (marginally) significant for all outcome measures in 2001, 2003, and 2005 (see Table 3). In the 2011 data, housing damage still had a significant effect on life satisfaction and negative affect, but not on health problems. Finally, we tested whether the negative effect of housing damage was particularly strong when the housing damage also entailed human loss, and not as strong when the housing damage was not accompanied by the loss of a household member. To do this, we added an interaction term between housing damage and bereavement in the multiple regression analyses described above. None of the analyses yielded a significant interaction 1; thus, the effect of housing damage on well-being appears to be orthogonal to the effect of losing a family member. Table 3. Multiple regression analyses predicting well-being from housing damage and human loss, age, gender, and SES variables for each survey (unstandardized regression coefficients; standardized regression coefficients). Predictors R2 Housing damage Human loss 2001 Life satisfaction −.049; −.057+ −.080; −.069∗ .047 Negative affect .125; .115∗∗ .103; .069∗ .031 Health problems .093; .111∗∗ .118; .102∗∗ .056 2003 Life satisfaction −.058; −.068∗ −.135; −.130∗∗ .124 Negative affect .105; .097∗∗ .142; .110∗∗ .055 Health problems .065; .077∗∗ 132; .130∗∗ .068 2005 Life satisfaction −.069; −.077∗ −.161; −.153∗∗ .099 Negative affect .139; .128∗∗ .214; .170∗∗ .082 Health problems .081; .097∗∗ .156; .159∗∗ .083 2011 Life satisfaction −.066; −.078∗∗ −.114; −.106∗∗ .100 Negative affect .075; .078∗∗ .134; .109∗∗ .039 Health problems .030; .040 .167; .173∗∗ .065 Note.+p < .10. ∗p < .05. ∗∗p < .01. Housing Damage: 0 = no damage, 1 = partial damage, 2 = half damage, half burned; 3 = total damage, totally burned. Human Loss: 0 = no immediate family members killed; 1 = at least one immediate family member killed.