Historically, quality of life is an interest of scholars in the field of health science, especially patients with disabilities ( Hoffman et al., 2006; Hu et al., 2011). Its importance has inspired many studies not to focus merely on individuals, but also on family ( e.g., FQOL measure). In 1998, WHO initiated a research attempted to devise a universal instrument of quality of life, which can cover both ‘ill’ and ‘well’ samples ( Division of Mental Health and Prevention of Substance Abuse, 1998). Its exploratory study consisted of patients with ill disease, ‘healthy’ patients, and health care providers. The final study involved 13 countries and 8,294 participants. The validity analyses suggested the measure to include six domains and 24 specific facets, including negative and positive feelings that usually refer to well-being measure. The measure has a two-fold indication. One, quality of life measure is a concern of every individual, regardless of their physical background. Two, the notion greatly overlaps with the concept of well-being; measuring one inevitably includes the other and vice versa. In times of the pandemic, people and their families are likely to be affected by ever-changing and pervasive health policies. Therefore, it is vital to investigate FQOL to predict the impact the pandemic has on a family and protect them against the risk factors resulted from the pandemic.

The conceptualization of FQOL relies on the following theoretical frameworks, (1) systemic concepts, (2) performance concepts, (3) individual-member concepts, and (4) family-unit concepts ( Zuna et al., 2009b). In a nutshell, the organizing frameworks suggest that the macro-level systems outside the family, e.g., healthcare, education, along with their policies and programs are associated with individuals’ demographic and characteristics, which directly impact individuals ( Hoffman et al., 2006). As each family member is associated to each other and to their habitat, any change experienced by a family member directly (and indirectly) impacts the whole family system ( Samuel et al., 2012). Therefore, a substantial change to normality brought by the Covid-19 pandemic is considered to highly impact family quality of life.

Besides health effects, the most direct impact of the Covid-19 pandemic comes from the economic aspect. Due to Covid-19, a lot of business are forced due to a wide-spread lockdown measure. Especially Indonesia, this has led to millions of job loss ( Idhom, 2020). Although the government has provided subsidies for those who lost jobs during the pandemic, the support is not sufficient to cover monthly bills and necessities for the whole family ( Saptoyo, 2021). Based on the family systems theory, the significant economic change will likely to disrupt the family balance ( Samuel et al., 2012). The relationship between parents’ adversity and child’s well-being work in a mutually reinforcing system; stress and disruption experienced by one party will affect the other. This significant economic change is expected to impact all dimensions of FQOL ( Hall & Clare, 2003).

In detail, FQOL is generally thought to include dimensions such as closeness, family interaction, family’s financial situation and a room for personal growth ( Hu et al., 2011). Based on its conceptualization, most of the FQOL scales emerge as an alternative and practical measure to the use of multiple measures. In this study, we specifically use the Beach Center FQOL scale ( Beach Center on Disability, 2012). The measure consists of five domains, namely family interaction, parenting, emotional well-being, material well-being, and disability-related support. For the purpose of the study that focuses on general population, we retain four dimensions and leave out the last dimension. This use can be equally compared to FQOL scale proposed by Zuna et al. (2009a).

Studies show that through aerosol droplets, a single-stranded ribonucleic acid (RNA) of the Covid-19 can infect up to 2.5 noninfected individuals ( Kaye et al., 2021). This infection has been shown to cost a human’s life or, at a less severe level, hospitalization. Due to the severity of the disease, the Covid-19 called for a special set of social and health protocols. As a result, the protocols inevitably started to disrupt supply chain of many trades, caused sudden unemployment, and forced closure of a great number of business ( Akbulaev et al., 2020; International Labour Organization, 2020). Not only does it pose economic cost on individuals, it has also precipitated many psychosocial impacts such as depression, substance abuse, etc ( Bu et al., 2020; Maital & Barzani, 2020). The latter impact, of course, can also be caused by the former due to a sudden change.

Specifically in Indonesia, the arrival of the Covid-19 has brought many unprecedented challenges to the families. According to a joint survey conducted by UNDP, UNICEF, Prospera and the SMERU Research Instittute (2021), 74.3% of all the households they interviewed during October to November, 2020 admit that their family was earning less than they were in January of the same year. This, in turn, heightened the proportion of low-income households across the country. In specific, this economic impact was felt greater among families in urban areas. The drop in income was also reported in all variety of income groups. On the other hand, the economic impact was exacerbated by increased cost of daily life, such as groceries. Covid-related expenditures, e.g., face mask, mobile communication also played a part in increasing living cost. Therefore, it is safe to say that the Covid-19 pandemic has brought economic hardships in most families across Indonesia.

In relation to FQOL, the economic impact seems to have a direct relation with the way families have to adapt. Hsiao (2018) shows that parental income plays a significant role on the FQOL. Many people are less satisfied with their financial circumstances during the pandemic ( SMERU et al., 2021). For others, the situation is even worse due to sudden unemployment or the closing of a business, and hence, forcing them to temporarily rely on government subsidy and aid ( Gusman, 2020; Idhom, 2020). Based on this, we expect that the Covid-19 economic impact, marked by less income and increased expenditure, is negatively related to FQOL (H1).

In general, parental stress is inherent in the parental role on a daily basis ( Cronin et al., 2015). The parental role includes providing care and developing intimate relationships, both of which can be exhausting and rewarding in itself. As such, it is common for parents to report some degree of parental stress in performing the role. Specifically, parental stress is defined as a set of physiological and psychological negative reactions towards the process of adaptation and demands in carrying out parental roles ( Pontoppidan et al., 2018). Without any specific family hardship, to some degree the stress is assumed to have no significant implication on the family. On the contrary, in times of difficulty, parental stress is likely to increase and will negatively affect parent–child relationship ( Respler-Herman et al., 2012). In addition, the degree of parental stress is strongly dependent on the stage of child development and the demands faced by parents ( Louie et al., 2017; Pontoppidan et al., 2018).

In looking further at parental stress, one should also consider it from the perspective of how parents respond to their stressful situations ( Cronin et al., 2015). This becomes a key for parents in managing the stress they experience and finding effective coping strategies. Specifically, when parents perceive the role of parents as rewarding, for example through happiness, enjoyment, optimism, intimacy and satisfaction with the child, then they are likely to reap satisfaction from performing the parental roles ( Berry & Jones, 1995). Those with high parental satisfaction are expected to enjoy performing their role as parents, enjoy being close to children and can show positive emotions when they are with the children. On the other hand, parents who perceive parenting as burdensome and demanding are expected to have a high degree of stress to the extent that they have difficulty in recognizing their children’s basic needs ( Respler-Herman et al., 2012).

In line with this, parental stress scale (PSS) was developed as a way to determine the extent of parental stress in carrying out parenting role ( Berry & Jones, 1995; Pontoppidan et al., 2018). This is measured through the dichotomy of the parental stress subscale and the parental satisfaction subscale. PSS has been adapted in various countries such as China, Spain, Portugal, Denmark, Malaysia, and many others ( Pontoppidan et al., 2018). Given the fact that the Covid-19 pandemic has brought many difficulties to a large number of families in Indonesia, it is likely that the level of parental stress is increased while the parental satisfaction is decreased. Events, such as supervising child (ren) while working at home, being restricted to indoor activity, are likely to cause disturbances among parents ( Griffith, 2020). Therefore, in times of the Covid-19 pandemic, we expect that parental stress is negatively related to FQOL (H2).

Resilience has been widely studied both at the individual as well as at the family levels ( Masten, 2019). While the former level refers to individual’s capacity to ‘bounce back’ from difficulty ( Smith et al., 2008), the latter refers to the capacity of a family system to survive and emerge from unfavourable circumstances, and to get stronger and more empowered ( Walsh, 2003). This definition is in line with the conception of family as an adaptive system and a context for human development.

Family resilience has extended the development of theory in the field of family stress, coping, and adaptation ( Patterson, 2002). During difficult times, family resilience will manifest itself in the extent to which a family is able to fulfil family functions, i.e., family membership, economic support, nurturance, and protection ( Li et al., 2019). We can claim that family resilience is a buffering factor in which families can rely on in times of hardship. However, it is important to note that there are times when families are able to swiftly bounce back from one adversity but take longer to recover from another type of adversity. This suggests that family resilience is also a process of continual growth and change across the life-span ( Walsh, 2003).

Furthermore, Walsh (2003, 2016) identify three key processes within domains of family functioning to define family resilience. The dimensions are (1) family belief system which consists of sub-components of making meaning of adversity, positive outlook, transcendence and spirituality; (2) organizational processes that consists of flexibility, connectedness, and mobilizing sources and economic resources sub-components; and (3) communication/problem-solving processes consisting of clarity, open emotional sharing, collaborative problem-solving. The three key processes enable families to work together in times of great stress in order to fulfil family functions mentioned earlier ( Prime et al., 2020).

Specifically, it is argued that family belief systems reflect by and large how families view happiness as well as adversity ( Walsh, 2003). The latter, indeed, shapes family members in their search for a meaning in times of crisis. During the Covid-19 pandemic, those with well-functioning belief system tend to look for positive motivation and encourage each other to get past the adversity. As Walsh (2016) argues, this type of family is likely to see adversity as a challenge which they can overcome with their available resources. At the same time, this way of thinking drives family members for a closer connection among each other. Through the times of the pandemic, well-functioning families show flexibility in adapting to new challenges. They are likely to adapt to a new situation by turning to each other for emotional and material resources and thus, strengthening their connectedness ( Prime et al., 2020). Furthermore, families with well-functioning belief system and positive organizational pattern are able to show positive collaboration in times of crisis. Through connectedness, they are able to share their opinions and feelings openly ( Walsh, 2003). This open communication allows family members to create effective decision-making process and hence, they are likely to operate effectively in their efforts to solve problems. On the whole, the combination of three key process of family resilience is a buffer to protect family functioning in times of crisis. Therefore, we hypothesize that family resilience positively moderates the relation between Covid-19 economic impact and FQOL (H3) . Similarly, parental stress is likely to be heightened due to unprecedented social policies implemented during the Covid-19 pandemic. However, taking into account the potential buffering effect of family resilience, we expect that it positively moderates the relation between parental stress and FQOL (H4) .

We employ demographic information to further explore the FQOL. Typical information such as gender, age, educational level, and number of child(ren) are included as individual’s characteristics. We also take demographic information as control variables to help ensure that there are no spurious relationships when considering the relevant predictors.

This study was conducted online from August 2021 until October 2021 using Qualtrics covering a wide coverage of Indonesia. This was done by involving local enumerators from several big cities, i.e., Bandung, Jakarta & its vicinities, Makassar, and Ambon, and instructed them to purposively distribute the Qualtrics link to their networks in local universities. Employing networks of local universities enable us to reach alumni and their extended networks. Although we did not aim for a nationwide generalizability, we believe that the selection of these cities represent a relevant case of Covid-19 impact on mostly urban areas (see Gusman, 2020; Idhom, 2020; Radhitya et al., 2020). Our selection criteria of participants were (1) a parent who lives together with their child(ren) and with or without their spouse or partner, and (2) has lived in their city of residence for at least two years. We acknowledge that there are limitations to our purposive sampling strategy, such as sampling representativeness. However, due to the Covid-19 measures, this was the best option to pursue in gathering participants from various locations. We also acknowledge that our participants would be biased towards middle income family due to the requirement of internet connection, accessible device and an intermediate skill of operating it. For this, sociodemographic information will be treated as control variables in the main analysis to ensure that there are no spurious relationships.

In a span of 3-month period, we distributed the survey link and successfully recruited 212 participants. The survey took about fifteen to twenty minutes to complete. In the end, we did not compute the time average taken by participants, because we set the online survey to allow participants to take a break during their participation thus, allowing multiple sessions or even days to finish the full questionnaire. The participants did not receive reward in any form. From 212 participants, we were only able to involve 169 participants due to either a substantial number of missing values or incomplete survey after 2 months. One from Takengon, two from Medan, 66 from Bandung and its vicinities ( e.g., Cimahi, Garut, Sumedang, Tasikmalaya), 41 from Jakarta and its vicinities ( i.e., Bekasi, Bogor, Serang, and Tangerang), 18 from Semarang and its vicinities (Cilacap, Yogyakarta, and Kebumen), four from Palangkaraya and its vicinities ( i.e., Tarakan and Pontianak), seven from Surabaya and its vicinities ( i.e., Nganjuk, Rembang, and Sampit), three from Denpasar and its vicinity ( i.e., Labuan Bajo), nine from Makassar and its vicinities ( i.e., Sengkang and Manado), 17 from Ambon, and finally one from Timika Papua. Table 1 provides the overview of our participants.

Furthermore, we looked more closely at the demographic information. Table 2 provides a detailed account of the information. First, in line with the mean value of the number of children ( M=2.27), most participants (44%) from our sample are shown to have 2 children. Second, we show that, on average, most participants (48.5%) experienced a decreased income during the Covid-19 pandemic. Although we notice that there is also a considerable number of participants (44.4%) who reported no change in their income. Meanwhile, a large number of participants (56.2%) reported an increased expense during the pandemic. Lastly, we also gathered a work status of participants and classified them into three broad categories, namely employee, entrepreneur and housewife/husband. It is shown that most of our participants (49.7%) are employees and the remaining are distributed almost equally across the other two categories. As a follow-up question, we asked participants whether there was a change due to the Covid-19 pandemic. The answer categories are 1 “no change”; 2 “instructed to stay home with a pay cut” for employees, “operational/economic capacity reduced to 50%” for entrepreneur and housewife/husband; 3 “instructed to stay home without pay” for employees, “operational/economic capacity reduced to more than 50%”; and 4 “laid off” for employees, “forced to close down” for entrepreneur and “economic capacity becomes impoverished” for housewife/husband. Here, a big majority of our employee and housewife/husband samples reported no change in their work and family. On the other hand, in the entrepreneur sample group, most participants (65.1%) reported that they had to reduce their operation by 50% or even reduced the operation to more than. While a small number of employee (1.2%) and entrepreneur (11.6%) reported to have been laid off or forced to close down their business.

Prior to filling in the questionnaire, the participants were given brief information of the study. Next to that, they were required to give an informed consent that they have been given sufficient time to carefully read the study information and there was no undue influence on their decision to participate. If agreed, participants were allowed to move on to the questionnaire. Otherwise, they would be directed to the end of the survey. No ‘consent for publication’ was necessary as the survey did not require personal information, such as name, initial, and other traceable information. We also disabled the recording of the internet protocol (IP) address on the Qualtrics survey to ensure that no unnecessary personal information was recorded. In terms of data management, only the research team had access to the dataset excluding the local enumerators.

We employed measures that have been tested in previous studies. To serve the purpose of this study, we translated the measures into Indonesian language. Specifically for parental stress and family resilience measures, we checked our translation with the Indonesian studies that have involved the measures previously. Our translation is similar to the previous studies. As for the FQOL measure, each of us translated the measure and compared the results. Subsequently, we had to reach a consensus on the selected translation for each statement.

Next, to ensure their applicability in our research context, we conducted confirmatory factor analysis (CFA) using lavaan package in R version 4.2.1 (RRID:SCR_001905; https://lavaan.ugent.be/). In addition, we calculated Alpha Cronbach to assess the internal consistency of all the measures and average variance extracted (AVE) and composite reliability (CR) to measure the amount of variance that is accounted for by the latent constructs ( Fornell & Larcker, 1981). The conventional benchmark for reliability level is larger than 0.70 or 0.80 ( Brunner & Süß, 2005). As for AVE, the value should be larger than any correlation found between any pair of latent constructs or minimum at 0.50 ( Fornell & Larcker, 1981). However, when AVE is found to be lower than 0.50 but CR is higher than 0.60, Fornell and Larcker (1981) argue that due to a more conservative estimate of convergent validity a construct with AVE lower than 0.50 can still be considered valid. Table 3 provides a bivariate correlation of the variables along with the value of AVE.

As mentioned earlier, we employed a measure of FQOL by Beach Center on Disability ( Hoffman et al., 2006). The scale assesses family member’s perception towards the different aspects of family life. The original scale consists of five domains, namely family interaction, parenting, emotional well-being, physical/material well-being, and disability-related support. There are 25 items in the original scale, but with the exclusion of disability-related dimension we employed 19 items spread across four dimensions. Family interaction is assessed through items such as, “My family enjoys spending time together”. Parenting is measured by items such as, “Family members help the children learn to be independent”. Emotional well-being is assessed through items such as, “My family has the support we need to relieve stress”. Finally, material well-being poses items such as, “My family has a way to take care of our expenses”. All items are rated on a five-point Likert scale, with 1 being very dissatisfied and 5 being very satisfied.

Based on the CFA results, we observe that the measurement model is indeed composed of four dimensions. However, the first model did not show a good fit with the data. We noticed that one item from each dimension of material well-being, family interaction, and parenting shared a substantial variance with other dimensions. After the removal the CFA shows a better fit model, Chi-squared = 191.94, p < 0.001, confirmatory fit index (CFI) = 0.94, root mean square error of approximation (RMSEA) = 0.08, and the standardized root mean squared error (SRMR) = 0.04 (see Appendix 1 in the extended data for a final list of items). According to Hooper et al. (2008), such fit measures are considered as a good fit model. The factor loadings of the remaining items range from 0.47 to 0.72, indicating a good level of accounted variance ( Peterson, 2000). In addition, the dimensions of the scale have a high level of reliability, with α = 0.89 for family interaction, α = 0.87 for parenting, α = 0.81 for emotional well-being and finally α = 0.74 for material well-being.

According to a report by SMERU et al. (2021), Covid-19 has brought economic challenges to many families across Indonesia. They based their conclusion on two main economic indicators, namely reduced income and increased expense. Therefore, to assess the economic impact brought by the Covid-19 pandemic, we asked two items regarding respondents’ change of income and expense. Respondents were asked to rate the items on a three-point Likert scale. The option categories range from 1 being income/expense is reduced to 3 income/expense is increased. Later in the analysis, we reversed the answer on a change of income item and calculated the total score for both items. Due to a two-item scale, we calculated its reliability by running a Pearson correlation and the result shows a significant correlation ( r = 0.17, p = 0.02).

Parental stress scale (PSS) was employed to determine the extent of parental stress in carrying out parenting role during the Covid-19 pandemic ( Berry & Jones, 1995; Pontoppidan et al., 2018). The scale is composed of two subscales, that is the level of parental stress and the parental satisfaction. Parental stress subscale asks respondents to rate their perception towards their parental role through items, such as “Having children has been a financial burden”. While parental satisfaction asks respondents to rate their satisfaction towards their parental role through items, such “I feel close to my child (ren)”. All items combined make up a total of 18 items and are rated on a five-point Likert scale, with 1 being strongly disagree and 5 being strongly agree. To compute the scale, we reversed the answer on all items in parental satisfaction subscale and then calculated the total score based on all items.

Through CFA, the first model shows a poor fit model. We observed that four items of parental stress subscale and one item of parental satisfaction subscale had a substantial shared variance with another subscale. After the removal of those items, the CFA shows a good fit model, Chi-squared = 116.89, p < 0.001, CFI = 0.95, RMSEA = 0.07, and the SRMR = 0.05. The factor loadings range from 0.51 to 0.71, indicating an acceptable level. The two subscales show a high level of reliability, α = 0.88 for parental stress and α = 0.90 for lack of parental satisfaction.

We assessed family resilience by employing family resilience framework by Walsh (2003). The framework proposes a multidimensional scale of family resilience, consisting of belief systems, organizational pattern and communication/problem solving ( Walsh, 2016). Belief systems asks respondents to rate themselves on statements, such as “We try to understand the stress situation and focus on our choice”. Organizational pattern consists of statements on family’s flexibility and the way they organize themselves in times of crisis, such as “We are flexible in adapting to a new challenge”. Finally, communication/problem solving assesses respondents’ way of communication in a family through statements, such as “We can express our opinion and be honest to one another”. Similar to previous scales, all items are rated on a five-point Likert scale, with 1 being strongly disagree and 5 being strongly agree. In total, the original scale consists of 32 items.

The first model in our CFA shows a poor fit model. Four items from belief systems dimension, and one item each from organizational pattern and communication/problem solving dimensions should be reconsidered due to cross loadings and low factor loading. After the removal of those items, the CFA demonstrates a good fit model, Chi-squared = 443.34, p < 0.001, CFI = 0.94, RMSEA = 0.06, and the SRMR = 0.05. The factor loadings of the remaining items range from 0.46 to 0.73, showing an acceptable level. Finally, all the dimensions of the scale show a high level of reliability, α = 0.91 for belief system, α = 0.93 for problem solving and α = 0.82 for organizational pattern.

We employed straightforward demographic questions concerning age, gender, educational level, and the number of child(ren). These items help us in the main analysis to ensure that there are no spurious relationships of interest when individual characteristics are factored in the statistical models.

Prior to running the main analyses, we conducted preliminary tests to ensure that our data fit the statistical assumptions, i.e., linearity, normal distribution, and multicollinearity. The ANOVA tests show that most of our predictors have a linear relationship with the FQOL. Moreover, the values of skewness and kurtosis of all our variables are in the range of −0.40 to 0.58 to −0.71 to 1.46, respectively, which suggest an acceptable range of normal distribution ( Kim, 2013). Finally, the multicollinearity tests show that the values of variance inflation factor (VIF) and tolerance statistics are within acceptable range, that is from 1.01 to 1.08 and 0.93 to 0.99, respectively ( Field, 2009; O’Brien, 2007). Based on all this, we can safely continue to our moderation analysis. In doing so, we mean-centred the predictors to enable easier interpretation on the results, i.e., high level of family resilience refers to above the mean and low level refers to below the mean.