Thriving Amidst Adversity: Longitudinal Subjective Well-Being and Economic Mobility in the Puerto Rican Climate Diaspora

Introduction

“The experience of adversity is virtually inevitable, but suffering in its wake need not be”

(Park, 2012, pg. 121)

As extreme weather events become more frequent in the Atlantic, the United States mainland is poised to witness a sustained exodus of climate survivors from U.S. territories such as Puerto Rico and the U.S. Virgin Islands, as well as nations states such as Cuba, Haiti, and the Dominican Republic that are geographically vulnerable to recurrent climate-related events, including severe storms and floods (World Bank, 2024). This large exodus of people seeking refuge from devastation can exert considerable pressure on migrant-receiving states, particularly neighboring states such as Florida, which are already grappling with the consequences of climate change (Gerrard, 2019). Rising sea levels, hurricanes, and floods continue to stretch mainland states’ resources thin and to challenge their capacity to provide adequate services to increasing numbers of climate migrants. This situation underscores the urgent need for comprehensive strategies for effective resource allocation vis-à-vis disaster preparedness and response.

Given the urgent need to alleviate suffering among climate migrants, there is a tendency to focus solely on mental health and economic shortcomings among climate migrants, as well as on the fiscal burden that their demands for social programs and services may represent. However, such a deficit-focused approach often overlooks the positive intrapersonal and community-level factors present among climate migrants. Positive factors can help them not only recover from the material and emotional losses associated with the disaster and the challenges of reconstructing their lives in new and sometimes unwelcoming contexts, but also to thrive in their new environments. Said another way, we should also focus on climate migrants’ ability to “get back on their feet” and to become economically independent and contributing members of their communities despite facing considerable adversity.

This shift in approach can help reframe the mainstream belief that climate migrants are primarily broken and fiscal resource-consuming individuals. Instead, it positions climate migrants as agentic individuals capable of capitalizing on their positive intrapersonal and community resources, and of leveraging the initial momentum provided by programs and services available during resettlement, as they recover and regain stability. A positive-oriented approach can also provide key evidence-based insights to inform the strategic allocation of aid in accordance with typologies of thriving within subgroups of climate migrants during and after relocation. Such an approach to policy recognizes both the humanistic and fiscal value of leveraging the positive intrapersonal and community-level resources inherent in climate migrants to expedite their recovery and facilitate the rebuilding of a thriving life.

With this framework in mind, the present study was designed to examine the emergence of unique latent thriving trajectory classes across three time points among a sample of 319 Puerto Ricans who survived Hurricane Maria and subsequently relocated to the U.S. mainland. Our theoretical approach is rooted in the positive psychology literature and is informed by the two-continua model of mental health (Westerhof & Keyes, 2010), which posits that mental health and mental illness are related but distinct dimensions. Consistent with this conceptual approach, we examined heterogeneity of subjective well-being and economic mobility, and controlled for distinct sources of stress and distress (i.e., hurricane trauma, cultural stress, and mental distress).

The Puerto Rican Triad of Stress: Hurricane Maria, Crisis Migration, and Cultural Stress

In September 2017, Hurricane Maria claimed thousands of lives and inflicted extensive damage on homes, businesses, and infrastructure in Puerto Rico. The disaster, compounded by decades of pre-disaster corruption and fiscal/governance failures, transformed the storm’s devastation into a humanitarian crisis, leaving most residents without essentials such as fresh food, potable water, electricity, and healthcare for months (Clark-Ginsberg et al., 2023). In addition to exposure to potentially traumatic hurricane-related experiences, forced displacement added an additional layer of exposure to post-migration stressors for Puerto Rican survivors in their mainland receiving communities (Schwartz et al., 2022).

Crisis migrants are individuals compelled to leave their homelands due to large-scale catastrophic events such as climate disasters, armed conflicts, wars, or political and economic collapse. According to the Crisis-Informed Theory of Cultural Stress (Salas-Wright et al., in press), crisis migration represents a two-fold dimensional stressor. First, crisis migration often takes place in a state of desperation, without a clear resettlement plan, and within significant economic constraints. Second, the lives of crisis migrants are simultaneously influenced by pre-migration stressors related to the disaster that prompted displacement as well as by post-migration stressors in their receiving communities. Among the most common cultural post-migration stressors are discrimination (exclusion, mockery, attacks, or suspicion due to race or ethnicity), a negative context of reception (perceived limited opportunities and unwelcomeness), and language stress (feeling embarrassed about one’s accent or difficulty understanding spoken English). Cultural stress has been positively and significantly associated with increased internalizing symptoms (i.e., anxiety and depression) and post-traumatic stress across distinct groups of crisis migrants (Salas-Wright et al., 2024; Schwartz et al., 2022; Vos et al., 2021),

Because Puerto Ricans are U.S. citizens at birth, the post-Maria Puerto Rican diaspora illustrates the concept of “citizen crisis migration” whereby a large influx of citizens migrate en masse to other areas within their country of origin to escape the aftermaths of a major disaster (Schwartz et al., 2022). Compared to international crisis migrants in the United States, Puerto Ricans are U.S. citizens and are not subject to the scrutiny and intense demands of the U.S. immigration system, nor do they face access barriers to jobs and services tied to legal status in the country. However, cultural disparities, such as their deeply rooted Hispanic traditions and predominantly Spanish language, set Puerto Rican migrants displaced by Hurricane Maria apart from other groups of mainland climate survivors, such as African American Hurricane Katrina survivors displaced from New Orleans to Houston (Cepeda et al., 2010). These cultural disparities place Puerto Rican migrants from Maria closer culturally to immigrants from Spanish-speaking countries in the Americas than to other groups of U.S. citizen crisis migrants, thereby exposing them to multiple sources of cultural stress when they relocate to mainland communities (Schwartz et al., 2015).

Previous cross-sectional studies conducted by our research group, using baseline data from the current sample of Maria migrants, found that cultural stress is a significant predictor of mental distress, even more so than exposure to hurricane-related experiences such as loss, disruption and life-threatening events during the storm. Such mental distress includes higher rates of anxiety and depressive symptoms (Montero-Zamora et al., 2023). Among cultural stressors, discrimination has been identified as the primary intervening variable in the association between hurricane exposure and problematic alcohol consumption in the present sample of Maria migrants (Schwartz et al., 2022).

In terms of post-traumatic stress, we found that exposure to hurricane-related traumatic experiences is significantly and positively associated with cultural stress at baseline. Both hurricane-related experiences and cultural stress then predict posttraumatic stress symptoms and conservative criteria for PTSD in 20.5% of Maria migrants in our sample, 3–4 years after Hurricane Maria (Hodges et al., 2023).

Conceptualization of Thriving

Thriving has been examined across various populations by pioneering scholars in developmental psychology (Lerner, 2004) and positive psychology Seligman & Csikszentmihalyi, 2000). More recently, it has also been explored through the lens of evolutionary psychology (King et al., 2017). Although most scholars define thriving as a construct that extends beyond mere survival and the absence of pathology to encompass growth and the realization of one’s potential, the conceptualization of thriving is not yet consensual.

Positive psychology focuses on counterbalancing traditional pathology-centered approaches by empirically examining positive states and optimal functioning (Seligman, 2019; Su et al., 2014). The two continua model of mental health proposed by Westerhof and Keyes (2010) further distinguishes mental health from mental illness, examines their interactions, and defines optimal functioning in terms of emotional, psychological, and social well-being, even, although not exclusively, in the face of adversity. Positive states such as satisfaction with life, positive emotions, and optimism (Carver, 1998; Su et al., 2014) have been identified as key elements present within thriving individuals. However, over the decades, thriving has often been used interchangeably with other constructs within positive psychology, such as post-traumatic growth, flourishing, and psychological well-being (Cohen et al., 1998; Seligman, 2019; Su et al., 2014), contributing to significant conceptual ambiguity (King et al., 2017).

Given the substantial material hardships experienced by most crisis migrants (Salas-Wright et al., 2021; 2023) in their receiving contexts, the link between subjective well-being (SBW) defined as satisfaction with life and positive emotions (Diener & Chan, 2011), and post-migration economic security is of critical significance in this population. Among Maria migrants who relocated to the mainland, the relationship between subjective well-being (SWB) and economic mobility during and after resettlement is still not well understood.

Findings by Clark-Ginsberg et al. (2023) indicate that Maria migrants generally arrive with very limited economic resources and face considerable material hardship during resettlement. Indeed, anecdotal experiences relayed by our community partners suggest that the Maria diaspora was largely composed of individuals from lower socioeconomic backgrounds and was primarily driven by pre- and post-disaster economic hardship. We propose an operational definition of thriving tailored to crisis migrants, which builds on the positive psychology literature and the premise that “the overarching objective for all individuals, including those who have endured trauma, is to achieve a fulfilling and prosperous life (Park, 2012, p.121). We define thriving individuals as those exhibiting trajectories of positive changes in (a) SBW (i.e., satisfaction with life and relative happiness on the mainland) and (b) economic prosperity (i.e., increase in income during and after resettlement compared to Puerto Rico). Especially within Latin American populations, where cultural collectivism often predominates over individualism and self-focus (Vignoles et al., 2016), thriving would likely also include relational components such as close friendships and satisfying family relationships. However, these relational variables were not available in the dataset used for the present study.

Thriving is also often connected with protective factors that reduce the likelihood of negative emotional, social, or behavioral outcomes (Cobb et al., 2025). Protective factors can operate at a variety of levels, including individual, family, and community. With respect to community level protective factors, prosocial behavior (a measure of how much individuals engage in actions that benefit others) and collective efficacy (the social cohesion among neighbors and the willingness and ability of the group to work toward a common good) have been associated with decreased levels of mental distress among Maria migrants (Cobb et al., in press; Piñeros-Leaño et al., 2022). Similarly, findings by Clark-Ginsberg et al. (2023) suggest that neighbors, community, and educational institutions constitute important sources of resilience and well-being in Maria survivors upon relocation to Florida.

Association Between Subjective Well-Being and Economic Mobility

Prior research indicates that individuals’ level of disaster preparedness is significantly and positively correlated with happiness and satisfaction with life (SWL) among survivors of severe disasters (Chen et al., 2021; Di Bucci et al., 2023). Among Maria migrants, (Schwartz et al., 2022) found that during resettlement, cultural stress emerged as a significant predictor of decreased levels of SWL. The association between SWL and economic mobility has been a subject of interest among scholars from distinct disciplines. A study conducted by Hadjar and Samuel (2015), using longitudinal multilevel analysis with a European panel dataset, found that SWL was associated with economic mobility. In the United States, Kahneman and Deaton (2010) found that, although SWL increases with upward economic mobility, this correlation seems to plateau at the $75,000 annual income mark. Incomes above $75,000 per year do not directly improve happiness. Rather, contrary to mainstream beliefs that more money can make people happy, upward economic mobility does not increase happiness but instead can reduce the unhappiness associated with poverty and financial hardship.

In the present study, we sought to identify unique latent trajectory classes of thriving among Maria migrants on the U.S. mainland. In addition, we examined the associations among sociodemographic characteristics, hurricane exposure experiences, mental distress (i.e., symptoms of depression, post-traumatic stress, and anxiety), cultural stress, and collective wellness (i.e., prosocial behavior and collective efficacy) across thriving trajectory classes. Prior research with the current sample of Maria migrants indicates that subgroups of Maria migrants vary both in their degree and type of positive states (i.e., intrapersonal and community factors) (Cobb et al., in press). The present study was guided by two research questions:

RQ1. What distinct latent thriving trajectory classes would emerge from our study sample?

RQ2. How do these emerging latent thriving trajectory classes differ in terms of sociodemographic characteristics, hurricane trauma exposure, mental health, cultural stress, and collective wellness?

Although it is difficult to advance a priori predictions regarding the number of classes that will emerge from the data, we hypothesized that multiple classes would emerge, each with a distinct thriving profile. We also hypothesized that classes with higher subjective well-being (SWB) would report higher levels of upward economic mobility.

Method

Participants.

Maria Puerto Rican adult survivors residing on the U.S. mainland at the time of data collection were recruited to participate in the present study. Eligible participants included participants aged 18 and older, who had lived in Puerto Rico during the storm, moved to the U.S. mainland between 2017–2020, and expressed no intention of returning to Puerto Rico at baseline. The average age was 38.7 years old (SD = 12.1, range 18–77 years), and most study participants (75%) resided in Central Florida, specifically Orlando and Kissimmee. Various other locations across the U.S. accounted for the remaining of participants: Texas (6%), the New England area (4%), Illinois (3%), Delaware (2%), South Carolina (2%), and other 8%.

Approximately one-third (29.8%) of participants rated their spoken English as very poor, another third (31.7%) as poor, around one-quarter (23.2%) as good, and only a small fraction (15.4%) described their English oral fluency as very good. The majority (71%) of our sample consisted of women. This limitation is attributed to the population accessible by our community partner organization, the Episcopal Church of Jesus of Nazareth in Orlando. Importantly, most study participants recruited through the church’s food pantry and other outreach services, are Catholic and not members of the Episcopal congregation. Almost half (43.9%) of study participants reported an annual household income less than $10,000 (29.8%), or between $10,000 and $14,999 (14.1%). The additional 56.1% of the study sample reported a household income between $15,000–$24,999 (22.3%), $25,000–$34,999 (14.3%), $35,000–$49,999 (10.7%), or $50,000 or more per year (8.8%).

Procedures.

Participants were recruited between August 2020 and October 2021 via a referral system where 111 initial seed participants were referred through our community partner organization, and seed participants referred and additional 208 participants to the study in exchange for additional $30 compensation for each successful referral. Study participants who provided informed consent completed a 60-minute survey online and received a $100 gift card as an incentive. Data collection for Time 2 started in October 2021 and Time 3 in May 2022. All data collection activities were conducted virtually using Qualtrics software. All participants completed the survey in Spanish, although an English version was available. Retention rates were 90% at T2 and 87% at T3. Ethical approval for all procedures was obtained from the institutional review board at the university that was responsible for data collection.

Measures

Sociodemographic variables.

Sociodemographic variables included age, sex (male = 0, female = 1), year of arrival on the mainland (2017 = 1, 2018 = 2, 2019 = 3, 2020 = 4), educational attainment (less than high school = 1, high school/GED = 2, some college or associated degree = 3, and college degree or more = 4), spoken English ability (very low = 1, low = 2, good = 3, very good = 4), and employment status (full time = 1, part-time = 2, unemployed = 3, retired = 4, and other = 5).

Subjective Well-Being (SWB).

We assessed SWB in terms of relative happiness and SWL. Relative happiness was assessed using a single question: “How do you compare your level of comfort and happiness living in the United States now compared to when you arrived?” Response answers ranged from 1= much worse to 5= much better. SWL was assessed using the Satisfaction with Life Scale (Diener et al., 1984) was used to assess global cognitive judgments of respondents’ satisfaction with life. Sample items include “If I could live my life over, I would change almost nothing” and “In most ways my life is close to my ideal.” Participants indicated how much they agreed or disagreed with each of the 5 items using a 7-point scale that ranged from 1 = Strongly Disagree to 7 = Strongly Agree. Cronbach’s alpha coefficients across waves ranged from .88 to .90 in the present sample.

Economic mobility.

Economic mobility was assessed using a single item question: “Is your economic situation better in the United States than it was in Puerto Rico?” Given that many people moved to the mainland for economic reasons, attaining more favorable economic standing following migration would indeed reflect a form of economic thriving. Response options were “No” and “Yes.”

Collective wellness.

We assessed collective wellness in terms of prosocial behavior and collective efficacy. Prosocial behavior was assessed using an adapted version of the Prosocial Tendencies Measure-Revised (Carlo et al., 2002). We used a total of 8 items, that is the Dire (4 items) and Emotional (4 items) subscales from the Prosocial Tendencies Measure (Carlo, 2002). We did not assess altruistic, compliant, public, and anonymous prosocial behaviors to prevent respondent burden. Sample items include “I tend to help people who are in a real crisis or need” and “when people ask me to help them, I don’t hesitate.” Response options ranged from 1 = Never to 5 = Always. In the present sample, Cronbach’s alpha coefficients across waves ranged from .90 to .91. We assessed collective efficacy using the 10-item Collective Efficacy Scale (Sampson et al., 1997). Sample items include “My neighbors would get involved if children were showing disrespect to an adult” and “My neighbors would get involved if a fight broke out in front of their house” with response options ranging from 1 = Strongly agree to 5 = Strongly disagree. Cronbach’s alpha coefficients across waves ranged from .92 to 94.

Mental distress.

We assessed mental health in terms of symptoms of PTSD and internalizing symptoms (i.e., depression and anxiety). PTSD symptoms were assessed using the PTSD Checklist–Civilian Version (PCL-C; Weathers et al., 1993), consisting of 17 items that measured the extent to which participants experienced PTSD symptoms attributable to Hurricane Maria in the month preceding the assessment. Sample items include “Having difficulty concentrating” and “Being “super alert” or watchful on guard.” Participants rated each item on a 5-point scale, with response options ranging from 1= Not at all to 5= Extremely. Cronbach’s alpha coefficients across waves ranged from .93 to .96 in the present sample.

Depressive symptoms were assessed using the 10-item Center for Epidemiologic Studies Depression Scale Boston Form (CES-D-10; Grzywacz et al.). The CES-D-10 scale measures both negative (e.g., “I felt depressed” “My sleep was restless) and positive (e.g., “I was happy”, “I enjoyed life”) symptoms in the week prior to the administering survey with response options from 1 = rarely or never to 4 = almost always. Higher values on the negative symptoms indicate greater depressive symptoms. We used 8 of the 10 items measuring negative symptoms and did not include the 2 items assessing positive symptoms. This was done because the positive items provided low factor loadings and they decreased the internal consistency (even after reverse-coding these responses). Cronbach’s alpha coefficients across waves ranged from .89 to .90.

Internalizing Symptoms was assessed in terms of anxiety and depressive symptoms. Anxiety symptoms were assessed using the 7-item Generalized Anxiety Disorder (GAD) scale by Spitzer et al. (2006). This instrument gauges various anxiety symptoms experienced in the two weeks leading up to the assessment. Sample items include “Worrying too much about different things” and “Trouble relaxing.” Cronbach’s alpha coefficients ranged from .94 to .95.

Cultural stress.

We examined three types of perceived cultural stressors in the present study: discrimination, negative context of reception, and language-related stress. Discrimination was assessed using Phinney et al.’s (1998) seven-item assessment tool, which inquired about the frequency of experiences such as being treated unfairly. Participants were required to indicate the frequency of each event described by the items, with responses ranging from 1 = Not at all to 5 = Almost every day. Cronbach’s alpha across waves ranged from .95 to .96 in the present sample.

Negative context of reception was assessed using the 6-item Negative Context of Reception Scale (Schwartz et al., 2015). The items referred to disparities in opportunities available to Puerto Ricans in comparison to other migrant groups and the perception that the members of host community did not welcome or accept Puerto Ricans. Sample items included “Puerto Ricans are not welcome here.” Participants indicated agreement with each item on a 5-point scale, ranging from 1 = Strongly Disagree to 5 = Strongly Agree. Cronbach’s alpha ranged from .84 to .90.

Language-related stress was assessed using the corresponding subscale from the Hispanic Stress Inventory (Cervantes et al., 1990). The seven items in this subscale to feelings and experiences such as being embarrassed about one’s accent when speaking English. Respondents rated their agreement with each item on a 5-point scale, ranging from 1 = Strongly Disagree to 5 = Strongly Agree. Cronbach’s alpha coefficients across waves ranged from .88 to .91 in the present sample.

Hurricane Exposure Experiences.

We assessed hurricane exposure using a modified version of the Hurricane-Related Traumatic Experiences Questionnaire (HURTE; Vernberg et al., 1996). We adapted the 17-item HURTE originally designed for adolescents. We omitted five items and re-worded items that seemed redundant or irrelevant for an adult demographic. Consequently, participants were presented with 12 items, five of which measured life-threatening situations such as “Did you get hurt during the hurricane?” The remaining items explored the losses and damages participants encountered during and following Hurricane Maria, such as “Were your possessions ruined by the hurricane?” Response options were “No” and “Yes”. The Cronbach’s alpha coefficient for this measure ranged between .89 and .92.

Data Analysis

We employed Class Growth Mixture Modeling (GMM) using Mplus 8.3 software to identify unique classes of thriving and their trajectories in terms of sociodemographic variables, hurricane trauma exposure, mental health, cultural stress, and collective wellness. We used robust maximum likelihood (MLR) to estimate parameters with standard errors and a chi-square test statistic that are robust to nonnormality and nonindependence. Proportions of missing data ranged from 4.7% to 20.2% for domains of thriving. Given that data missing at random is an assumption for use of full information likelihood estimation (FIML; Dong & Peng, 2013), we employed Little’s Missing Completely at Random Test (Little, 1988).

Results indicated that data across time were missing at random, χ²(60) = 52.83, p = .59. As a result, we addressed missing data using FIML. Based on the methods set out by Grimm et al. (2016) and Jung and Wickrama (2008), we conducted the analyses in three steps. First, we estimated an unconditional latent growth model to assess model fit for a single class. Suggested values for good model fit are CFI ≥ 0.95, RMSEA < 0.08, and SRMR < 0.08 (Kline, 2015). We include the chi-square statistic, where a nonsignificant p-value indicates good fit. Second, we followed a multistage process to estimate a series of GMMs and identify the optimal number of classes, and their corresponding trajectories, that provided the best fit to the data. To accomplish this, we first assessed whether variance estimates for the growth factors were significant. If they were not, we would employ a GMM without the ALGORITHM = INTEGRATION command in Mplus, which sets the variances and covariances of the growth factors to equality.

This is essentially equivalent to a latent class growth analysis where variances for the growth factors do not vary across classes. We assessed four key criteria to determine the optimal class solution: (1) We considered the Vuong–Lo–Mendell–Rubin likelihood ratio test (VLMR-LRT) that indicates the extent to which the 2-log likelihood value for a model with k classes is significantly smaller than the likelihood value for a model with k-1 classes; (2) We assessed Akaike information criterion (AIC) and Bayesian information criterion (BIC) as a basis for comparing models, where smaller values suggest better model fit; (3) Each class had to represent at least 5–10% or more of the sample to provide additional confidence in the stability of the class solution (Nylund et al., 2007); and (4) Entropy values and posterior classification probabilities should be >.75.

Third, we employed the Bolck–Croon–Hagenaars (BCH) method to assess mean differences between classes across study variables. The BCH method accounts for uncertainty in latent class classification by assigning observation weights that reflect measurement error of the latent class variables (Asparouhov & Muthen, 2014). The BCH is the preferred method for assessing class differences when using continuous and categorical distal outcomes. The BCH provides a chi-square test of mean differences for the distributions of covariates across classes and permits the evaluation of class differences without influencing the initial individual profile classifications.

Results

Descriptive Statistics.

Table 1 contains demographic and outcome variables by latent thriving trajectory class.

Table 1.

Demographic and Study Variables by Latent Thriving Trajectory Class

Variable	Class 1 Elevated SWB/Low UEM (n=81, 25%.)	Class 2 Moderate SWB/Low UEM (n=23, 7%)	Class 3 High SWB/High UEM (n=217, 68%)	Overall c2
Sociodemographic
Sex %				2.76
Male	26%	43%	28%
Female	74%	57%	72%
Age (M, SD)	41.49 (1.58)	37.94 (2.23)	37.59 (0.86)	4.18
Education %				.678
Less than high school	2.20%	4.40%	0.00%
High school/GED	30.90%	29.40%	32.30%
Some college/associate degree	36.20%	39.70%	42.40%
College or more	30.80%	26.50%	25.30%
Employment % T1				18.39*
Full time	30.50%	13.20%	42.70%
Part-time	26.90%	20.40%	23.30%
Unemployed	18.10%	48.70%	16.30%
Retired	7.30%	4.40%	3.00%
Other	17.30%	13.30%	14.70%
Employment % T3				21.28**
Full time	35.40%	50.00%	48.40%
Part-time	20.20%	10.00%	30.00%
Unemployed	23.80%	30.10%	10%
Retired	3.0%	5.0%	3.0%
Other	17.80%	5.0%	11.80%
Spoken English (M, SD)	2.12 (0.06)	2.35 (0.20)	2.28 (0.07)	1.20
Year of arrival				28.88***
2017	58.4%a	30.8%b	64.2%a
2018	30.4%	34.2%	28.8%
2019	6.7%	35.1%	3.0%
2020	4.5%	0.0%	4.0%
Hurricane Stress
Did windows or doors break in the place you stayed during the hurricane?	67.0%a	91.8%b	68.9%a	12.47**
Did you get hurt during the hurricane?	11.70%	53.90%	7.80%	18.03***
Were your possessions ruined by the hurricane?	67.70%	91.40%	62.40%	17.16***
Did you or your family have difficulties accessing food and water?	69.70%	60.10%	83.90%	5.38*
Community Wellness
Prosocial behavior T1 (M, SD)	34.04	30.53	31.87	8.39*
Prosocial behavior T3 (M, SD)	32.44 (0.83)	35.16 (1.39)	31.33 (0.46)	7.23*
Collective efficacy (M, SD)	25.38 (1.17)	19.10 (2.73)	25.90 (0.50)	5.80
Mental Distress
PTSD (M, SD)	32.34 (2.03)	34.66 (5.15)	30.91 (1.08)	.77
Depression (M, SD)	13.90 (0.80)	14.10 (1.71)	13.06 (0.38)	1.05
Anxiety (M, SD)	12.06 (0.72)	12.23 (1.70)	11.48 (0.42)	.53
Cultural Stressors
Discrimination (M, SD)	15.21 (0.93)	15.10 (2.06)	14.32 (0.50)	1.21
Context reception (M, SD)	13.77 (0.81)	14.00 (1.80)	13.62 (0.40)	.06
Language stress (M, SD)	16.39 (0.83)	14.31 (1.75)	16.24 (0.41)	1.22

Note: SWB: positive psychological well-being; UEM: upward economic mobility; GED: general educational diploma; PTSD: post-traumatic stress disorder. Demographic variables are reported at T1, and outcome variables are reported at T3. If patterns of significance across classes are different between T1 and T3, then that variable will be reported at both T1 and T3.

Unconditional latent growth model.

Results for the single class unconditional latent growth curve model of thriving indicated appropriate fit: χ²(36) = 744.511, p < .001, CFI = .93, TLI = .90, RMSEA = .08, 95% CI [0.057, 0.097], SRMR = .07. Results indicated an elevated and significant intercept growth factor of 21.82 (SE = 0.37, p < .001), a significant positive slope factor of .05 (SE = 0.20, p = .02) for SWL, a significant intercept growth factor of 3.85 (SE = 0.10, p < .001), a non-statistically significant negative slope factor of −.01 (SE = 0.06, p = .90) for relative happiness, and a significant positive slope factor of .18 (SE = .07, p < .001) for improved income in the mainland compared to income on Puerto Rico. In addition, the variance of the intercept for SWL was significant at 20.10 (SE = 2.38, p < .001), at .48 (SE = .13, p < .001) for relative happiness, and at .77 (SE = .07, p < .001) for improved income. Results suggest that Maria migrants in our sample varied in degree of their levels of thriving at baseline, and longitudinal trajectories appeared to also vary for SWL and improved income.

Unconditional GMMs and number of class solution.

Following a sequential process of class selection, we found that a 3-class solution was optimal based on a lower AIC, BIC, and ABIC values, higher entropy, and posterior probabilities, as well as a significant VLMR-LRT p-value criterion (Table 2). Such set of criteria suggest that the 3-class solution fit the data better than a 2-class solution for the latent growth curve analysis. In contrast to the 3-class model, although the fit indices were lower for the 4- class model, the VLMR-LRT was nonsignificant, suggesting that the model with 4 classes did not fit the data better than the model with 3 classes.

Table 2.

Model Indices for Class Solutions

# of Classes	AIC	BIC	ABIC	VLMR-LRT (k 1)	Entropy	Posterior probabilities	Class proportion ranges
1	8648.834	8731.806	8662.025
2	8662.318	8745.289	8675.509	<.001	0.849	0.943–0.965	25–75%
3	8556.877	8670.020	8574.865	<.001	0.865	0.924–0.982	.07–66%
4	8500.559	8640.102	8522.744	.052	0.866	0.905–0.994	.04–65%

Moreover, one of the class proportions in the 4-class solutions fell below 5% (i.e., 4%), which is not recommended as it would have provided a solution with a very small number of cases in one or more classes (Jung & Wickrama, 2007). To provide maximum confidence regarding our preferred class solution, we performed both the adjusted likelihood ratio test and the bootstrapped likelihood test. Both tests favored the 3-class over the 4-class solution (adjusted LRT p = .06, where a nonsignificant p-value favors the solution with fewer classes). In turn, we opted for the more parsimonious 3-class solution (Figure 1), which yielded three classes.

Figure 1.

Latent thriving trajectory classes at baseline

Characteristics of the latent classes.

The three emerging classes were named primarily according to their baseline levels (intercepts). Class # 1 Elevated Subjective Well-Being/Low Upward Economic Mobility (Elevated SWB/Low UEM) (n = 81, 25%.): This class comprised a quarter of the sample and was characterized by moderate intercept and slope factors for SWL and relative happiness, as well as the lowest slope means for improved income compared to the other two classes across time points. All intercept and slope variances were significant. Class # 2 Moderate Subjective Well-Being/Low Upward Economic Mobility (Moderate SWB/Low UEM) (n =23, 7%): This class comprised a small fraction of the sample and was characterized by the lowest mean intercept but the highest slope factors for all three constructs of thriving across time points (i.e., SWL, relative happiness, and improved income). The intercept variances for all three constructs were statistically significant. However, the slope variances were all highly significant.

Class # 3 High Subjective Well-Being/High Upward Economic Mobility (High SWB/High UEM) (n =217, 68%). This class comprised over three quarters of the sample and was characterized by the highest intercept for SWL and relative happiness compared to the other two classes, whereas the slope factors for SWL and improved income were both moderate and the slope factor for relative happiness was negative. Similar to Class 2, the intercept variances for all three constructs were significant, whereas the slope variances were nonsignificant except for relative happiness. Intercept and slope mean and standard deviations for classifying variables across latent profiles are provided in Table 3.

Table 3.

Latent Thriving Trajectory Classes by Indicator Intercepts and Slopes

Variable	Class 1 Elevated SWB/Low UEM (n=81, 25%.)	Class 2 Moderate SWB/Low UEM (n=23, 7%)	Class 3 High SWB/High UEM (n=217, 68%)
Satisfaction With Life
Intercept (SD)	19.90*** (4.15)	15.72*** (4.15)	23.24*** (4.15)
Slope (SD)	0.48 (0.27)	1.77*** (0.27)	0.40 (0.27)
Relative Happiness
Intercept (SD)	3.53*** (0.30)	0.55*** (0.30)	4.38*** (0.30)
Slope (SD)	−0.22*** (0.24)	1.53*** (0.24)	−0.11** (0.24)
Economic Mobility
Intercept	−4.36*	−5.25***	0.00a
Slope	0.29	1.09***	0.68

^aThe mean intercept for class 3 was constrained to zero to facilitate model identification.

Class differences for main variables.

Using the BCH method in Mplus, where auxiliary variables are compared across classes but are not used to create the classes, several class differences emerged across study variables. Post-migration mental distress (i.e., symptoms of PTSD, depression, anxiety) and cultural stress did not differ significantly among classes. Additionally, no statistically significant differences emerged for age, sex, educational attainment, or spoken English ability. However, the overrepresentation of women (71%) may have biased the results and could skew the findings towards women’s experiences. We identified significant class differences in retrospectively reported hurricane exposure experiences, year of arrival on the mainland, employment status, and prosocial behavior, as follows:

Hurricane exposure experiences:

Although a significant majority of individuals across all classes reported hurricane distressful experiences, results suggest that Class 2 members experienced the most material damage and injury during the hurricane. Class 2 individuals reported the highest percentage of broken windows or doors (91.8%), personal injuries (53.9%), and ruined possessions (91.4%). In contrast, participants in Class 1 and Class 3 reported lower rates of these experiences. These sets of results indicate that, whereas individuals in Class 1 and Class 3 may not have been as severely affected as those participants in Class 2, they still experienced other types of hurricane-related stressors. For instance, people in Class 3 reported the most significant challenges in securing food or water after the hurricane (83.9%).

Year of arrival on the mainland:

In terms of migration patterns to the U.S. mainland between 2017 to 2020, Class 1 and Class 3 largely outnumbered Class 2 in terms of arrivals during 2017. By 2018, all classes evidenced more balanced migration patterns, with Class 2 slightly surpassing the other two classes. In 2019, Class 2 saw a significant increase in arrivals, whereas Class 1 and Class 3 dropped drastically. By 2020, arrivals across all classes decreased.

Employment:

Class 2 reported the highest percentage of full-time workers (50.0%), the highest percentage of unemployed individuals (30.1%), as well as the lowest part-time workers (10%). The percentage of retired individuals were relatively evenly distributed across classes. Collective wellness: Prosocial behavior differed significantly across classes (p = .027), with Class 2 displaying the highest mean value (35.16). Conversely, no significant variation across classes (p= .055) was found for collective efficacy.

Indeed, members of Class 2 stand out from those in the other two classes in several significant ways, as follows: the vast majority (>90%) of participants in Class 2 reported extensive property damage and loss of possessions due to the storm, and over half of them reported personal injuries. The high likelihood that these individuals were among the most socioeconomically disadvantaged in Puerto Rico prior to the disaster may explain their extensive exposure to hurricane-related hardships and injuries, likely due to residing in structures ill-equipped to withstand the hurricane’s force. The compounded effects of material devastation and physical harm may account for the scattered migration patterns observed in Class 2. Access barriers to healthcare in Puerto Rico following Hurricane Maria (Clark-Ginsberg et al., 2023), coupled with economic constraints and the financial burden of air travel, may help explain delayed migration arrangements for individuals in Class 2, who mostly arrived on the mainland in 2019. Pre-migration hardships, such as poverty, material loss, and disability, may also help explain the substantial rate of unemployment (30.1%) among members of Class 2.

Discussion

Drawing from a sample of Puerto Rican adults displaced to the U.S. mainland following Hurricane Maria, findings from the present study shed new light on thriving (defined as subjective well-being and upward economic mobility) in the context of climate crisis migration, as migrant individuals recover and build a thriving life in their new environments. Our first key finding relates to identifying three distinct trajectory classes of thriving. These classes are Class 1 (Elevated SWB/Low UEM), Class 2 (Moderate SWB/Low UEM) and Class 3 (High SWB/High UEM). In addition to heterogeneity in thriving indicators, significant differences in hurricane exposure experiences, employment status, year of arrival, and collective wellness (i.e., prosocial behavior but not collective efficacy) emerged across classes.

The most noteworthy of these differences concerns the highest increases in thriving observed among the subset of individuals (Class 2) who were the most severely affected by Hurricane Maria. Despite comprising a minority (7%) within our study sample, Class 2 results offer valuable insights into the potential underlying mechanisms through which Maria migrants may transform significant adversity into opportunities for intrapersonal and material thriving. A potential explanation for the unexpected levels of upward economic mobility among Class 2 Maria migrants could be that relocating to new communities on the mainland provided them with opportunities to interact with wealthier individuals – opportunities that may not have been as available in Puerto Rico. Cross-class interaction, or relationships with higher-income peers, can increase an individual’s chances of escaping poverty in the future (Chetty et al., 2022). Even if Maria migrants settled in low socioeconomic areas on the mainland, these communities likely experienced lower poverty levels than the staggering 44.5% rate in Puerto Rico following Hurricane Maria (U.S. Census Bureau, 2019. This rate is nearly double that of the most impoverished mainland states (Buckley & Burnette, 2023).

Prior work suggests that upward economic mobility can help decrease psychological distress and increase satisfaction with life (SWL) among internally displaced migrants (Chiang et al., 2021). It has been proposed that improved levels of SWL among low-income individuals in the U.S. associated with upward economic mobility (UEM) can be explained in terms of the interplay between UEM and decreased levels of unhappiness associated with poverty (Kahneman & Deaton, 2010). The aforementioned studies suggest that the interplay between UEM and subjective well-being among Class 2 individuals in our sample may be twofold. First, UEM appears to contribute to overcoming baseline economic hardship associated with the significant material loss in Puerto Rico due to the storm and to subsequent economic constraints during resettlement on the mainland. Second, such economic recovery may help explain the increases in SWL associated with higher comfort and happiness on the mainland compared to Puerto Rico.

Indeed, we conducted post hoc analyses to assess the trajectory of employment status among all three thriving classes emerging from our sample. We found that Class 2 individuals reported the steepest growth in full-time employment, increasing from 13.2% at baseline to 50.0% at T3, and a decrease in part-time employment from 20.4% to 10.0% during that same time frame. This finding suggests that not only were more individuals able to secure full-time jobs over time, but also that half those people with part-time employment at baseline potentially transitioned into full-time positions. Similarly, the baseline unemployment rate of 48.7% among Class 2 individuals—the highest among the three classes—was notably reduced to 30.1% by T3. Although unemployment remained significantly elevated, this trend in economic mobility indicates that, despite enduring the most severe impact from Maria and arriving on the mainland under dire material conditions, Class 2 individuals were progressively achieving economic security.

Although we cannot ascertain the directionality of the relationship between subjective well-being and UEM, our findings offer new insights into this association. For instance, it is plausible that unemployment among our study participants is linked with increased levels of unhappiness due to difficulties in affording basic necessities such as food, clothing, and housing, during resettlement. Consequently, it is unsurprising that SWL increases as external factors potentially contributing to unhappiness associated with material hardship are mitigated. Conversely, it is also possible that, over time, SWL increases as Maria migrants find themselves in new environments, away from the long-term dire conditions they experienced in Puerto Rico. Despite the challenges occurring during resettlement, such as material hardship, cultural stress, and mental distress, Maria migrants become more comfortable and happier as they learn to navigate their new realities. This enhanced subjective well-being may facilitate the establishment of new relationships, strengthen their sense of connectedness to their communities, and increase their openness to new job opportunities.

Indeed, our results suggest that the elevated levels of prosocial behaviors reported across all three classes in our study may have contributed to cultivating a sense of belonging and mutual support among Maria migrants within their new communities. Furthermore, findings from our post hoc analysis indicate that only Class 2 members reported a significant increase in rates of prosocial behavior over time compared to the other two classes in our sample. Such a growth trajectory in prosocial behavior among Class 2 members suggests that prosocial behavior may represent a significant community-level asset among thriving climate migrants.

This finding is supported by prior work conducted among our sample of Maria migrants using baseline data, which found that prosocial behavior serves as a protective factor in this population (Piñeros-Leaño et al., 2022). Prior research among our sample of Maria migrants also found that collective efficacy can be a protective factor, providing Maria migrants with a valuable network of support and resources to navigate post-migration challenges and uncertainties, thereby helping these individuals feel safer in their communities (Cobb et al., in press). To our surprise, we did not find a significant difference in collective efficacy across classes in our study. We should note, however, that the means in Table 1 suggest low-to-moderate levels of collective efficacy (item means ranging from 2.1 and 2.9 on a 1–5 scale). It appears that all three classes viewed their neighborhood social cohesion as at least somewhat lacking.

Our second key finding concerned moderate levels of thriving among the 68% of our sample placed into Class 3. Individuals in Class 3 exhibited growth in SWL and upward economic mobility, alongside a slight decline in relative happiness over time. It is important to underscore that these individuals reported high levels in all three domains of thriving at baseline, leaving limited room for further growth. Furthermore, Class 3 included the greatest number of Maria migrants who were employed either full-time or part-time at baseline (66%) and at T3 (78.4%), and reported the lowest unemployment rate (16.3%) during resettlement, as well as the steepest decline (7%) in unemployment.

Although Class 3 individuals also faced significant post-disaster hardship while in Puerto Rico, early migration to the mainland right after Hurricane Maria may have offered these individuals an advantage, potentially protecting them from long-term challenges such as shortages in basic services (i.e., drinking water, electricity) that those who remained on the island after the storm had to endure. Regarding Class 1, which comprised 25% of our sample, and unlike the other two classes, Class 1 members reported significantly elevated levels of all three domains of thriving at baseline but showed no significant growth over time. Instead, there was a slight but significant decline in relative happiness in T3. Whereas the rate of individuals employed full-time increased from 30.50% at baseline to 35.40% at T3 in this class, the unemployment rate increased from 18.10% to 23.80% over this same time span.

Our third key finding is that post-migration mental distress (i.e., symptoms of PTSD, depression, anxiety) and cultural stress did not predict class membership. This finding is supported by the two continua model of mental health (Westerhof & Keyes, 2010), which postulates that mental health and mental illness are related but distinct dimensions. Therefore, individuals can endure high levels of psychological distress yet still exhibit positive mental health (i.e., feeling good, functioning well, and being of social value). We found that mean values of both cultural stress and indicators of mental distress at baseline were slightly above the scale midpoint across the three classes. Our post hoc analysis further indicated that longitudinal mean levels of cultural stress and symptoms of depression and anxiety remained relatively stable across time.

Conversely, mean values of post-traumatic stress symptoms evidenced a slight decrease over time within all classes, ranging from 35.91 to 38.72 at baseline to 30.90 to 34.73 at T3. A previous study using a different sample of Puerto Rican Hurricane Maria migrants conducted by our research group in 2018, shortly after the storm, found that a majority of Maria migrants in Florida (65.7%) self-reported clinically elevated levels of PTSD symptoms (Scaramutti et al., 2019). In contrast, subsequent research from our group, utilizing baseline data from the present study sample collected three to four years after the storm (Hodges et al., 2023), indicated that nearly half of participants (44.7%) reported clinically significant decreases in PTSD screening diagnoses. Although we cannot assess the comparability of the Scaramutti et al. (2019) sample with the baseline data used by Hodges et al. (2023), the decrease in PTSD rates reported by the latter suggests a downward trend over time.

Extensive research on resilience conducted over the past few decades (Bonnano et al., 2010) supports our findings, indicating that “resilience is the most common response” to stressors and potential trauma (Galatzer-Levy et al., 2018). This body of literature also posits that disasters produce multiple patterns of outcomes, including a confluence of typologies of mental distress and psychological resilience (Bonanno et al., 2010). Furthermore, positive psychology rejects the notion that negative and positive states constitute opposite sides of a dichotomous continuum and questions the idea that well-being is solely achievable through the absence of negative states (Seligman, 2011), or that human growth is disentangled from adversity (Linley & Joseph, 2004; Tedeschi & Calhoun, 2004). Instead, positive psychology asserts that negative and positive emotions (Fredrickson, 2001) and states are not only separate psychological phenomena, each consisting of unique constructs, but that they can coexist (Larsen & McGraw, 2011). In the words of Seligman, (2018), “arriving at the good is a lot more than just eliminating the bad” (p. 20).

Implications for policy and practice

Government, private institutions, and mental health care providers concerned with assisting climate survivors displaced to the U.S. mainland can prepare for the next natural disaster in the Atlantic in at least two ways. First, it is important to triage newcomers not only in terms of mental distress but also in terms of economic insecurity. Research shows that when mental health care is matched with economic aid (e.g., expedited access to personal relief loans) to assist climate survivors in alleviating economic insecurity after a natural disaster, it can lead to a two- to six-fold decrease in PTSD cases one year following the disaster (Cohen et al., 2019).

Second, the service needs (e.g., mental care and housing) of Maria migrants in their receiving contexts are not monolithic. These needs may vary based on (1) pre-migration exposure (i.e., exposure to hurricane-related experiences) and (2) post-migration economic insecurity. Sustained economic aid during resettlement, coupled with social services such as English literacy and assistance in navigating the employment market, as well as having individuals who can engage migrants in Spanish, can help alleviate psychological distress associated with significant financial hardship. This, in turn, can expedite recovery and facilitate migrants’ inclusion into the labor force.

Additionally, social engagement in environments where community building takes place (e.g., workplaces, faith-based organizations) can facilitate social inclusion and adaptation, and promote a sense of belonging among newcomers.

Strengths and Limitations

Findings from the present study underline the importance of longitudinal designs in identifying the likelihood of change in thriving among climate migrants, which is otherwise unattainable using cross-sectional data. Our results also highlight the need to empirically examine the interplay between thriving and distinct post-migration stressors, as well as the bidirectional nature of the relationship between positive psychological states and desirable external outcomes, such as economic mobility. Despite these contributions, our findings should be interpreted in light of several limitations. First, data collection took place three years after Hurricane Maria. Consequently, participants’ recollections of their experiences during the storm may introduce recall bias inherent in retrospective self-reported data. Second, the use of a convenience sample represents a limitation of the present study. However, although we obtained seed participants from an Episcopal church, the use of participant referrals helped to diversify the sample beyond individuals affiliated with the church. Indeed, the fact that we were able to recruit participants in Texas, New York, and other states suggests that our sample may be more generalizable than one consisting solely of individuals directly connected to a church. Within our sample, only 11 seeds were recruited directly from the church, and the remaining 308 participants were recruited through referrals. Further, because there is likely no universal registry of Puerto Rican Hurricane Maria survivors who relocated to the U.S. mainland, population-based sampling likely would not have been possible. Third, the generalizability of our findings may be limited to Maria survivors who migrated to the U.S. mainland after the storm; therefore, their experiences may not be representative of the larger population of Puerto Rican Hurricane Maria survivors who remained on (or returned to) the island. Fourth, we were not able to include relational variables as components of thriving. Although the availability of longitudinal data with a unique crisis migrant sample may offset this limitation to some extent, it is important for future studies to include close friendships, satisfying family relationships, and other relational variables as indicators and components of thriving.

Conclusion

Overall, our findings illuminate the significant positive association between subjective well-being and upward economic mobility among a sample of climate migrants, emphasizing the necessity for further investigation of this relationship in future research. Notably, individuals who were most severely impacted by the hurricane and were in the poorest financial condition at baseline demonstrated the most pronounced positive and increasing trends in both domains of thriving over time. Our results contribute to the expanding body of research demonstrating that mental health and thriving represent not merely the absence of mental illness; rather, thriving operates at least somewhat separately from mental distress. An exclusive focus on deficits, while highly relevant, may have inadvertently led to the oversight that many climate migrants possess substantial intrapersonal and community-level resources that enable them to thrive and contribute to their new environments despite adversity. We hope that the present findings inspire further research in this direction.