Intragenerational Social Mobility and Preventive Behaviors regarding COVID-19: A Case Study of the Slum and Non-Slum People in Dhaka City


 Background: The policy design for eliminating health inequality from social mobility perspective has been prioritized in the sociological studies. The World Economic Forum has also developed Global Social Mobility Index taking into account the fixing of social inequality including health inequality. Unfortunately, the policy design and explanation for preventive behavior gap regarding COVID-19 from social mobility perspective has hitherto been an unexplained phenomenon in Bangladeshi context. This study, hence, evaluates the effectiveness of Health Belief Model (HBM) in explaining the preventive behavior gap index (PBGI) regarding COVID-19 in relation to Blau and Duncan’s socioeconomic status (SES) index and explores other extraneous factors. Methods: The study follows explanatory sequential mixed-method approach. In the first phase, the study has conducted pilot test for confirmatory factor analysis. The second phase is also quantitative in nature which follows the perfectly experimental research design. In this regard, the study first used SPSS software (Windows version 25) and then converting into CSV format linked with SmartPLS 3 for developing structural equation model. In the third phase, qualitative data were collected using key informant interview to explore other extraneous factors that can explain the preventive behaviors of slum and non-slum peoples of Bangladesh regarding COVID-19.Results: The high SES index score of non-slum people have positively significant effect on their PBGI regarding COVID-19 mediated through perception gap index (PGI) while the low SES index score of slum people have negatively non-significant effect on the same constructs. Conclusions: The HBM as dysfunctional should be revisited in line with the free provision of protective equipment for understanding the health behavior of those people who live from hands to mouth.

health beliefs are potential contributors to COVID-19 related PBGI. The empirical studies also support the HBM constructs including susceptibility, severity, bene ts, barriers and self-e cacy [10]. The previous study ndings related to the determinants of PBGI regarding COVID-19 are not identical. This is the shortcoming of those studies to understand why the PBGI regarding COVID-19 signi cantly varies from Western developed world to developing countries like Bangladesh. Even, in the context of Bangladesh, some studies reveal that knowledge, attitude and practice (KAP) model is effective while other studies nd the opposite result. In fact, the related existing studies could not understand the developed nations are directed toward the upward social mobility, universalism and achievement while Bangladesh is in the stage of particularism, ascription and many communities in the country are still now directed toward downward mobility. At the same time, a uent families or communities of Bangladesh are directed toward upward mobility. Thus, merely knowledge and perceived health beliefs fail to predict or explain PBGI regarding COVID-19. Rather, the social class or strata may matter in the societies. Hence, the study has an endeavor to evaluate the usefulness of HBM in explaining the PBGI regarding COVID-19 among the slum and non-slum people of Bangladesh depending on the constructs of HBM in relation to socioeconomic status (SES) index of [11].
Theoretical Framework Health belief model (HBM) The HBM was rst developed in the 1950s-by social psychologists of the U.S. Public Health Service to explain the failure of disease preventive programs [12]. This model has 5 core constructs: perceived severity (beliefs about severe and consequences of the condition), perceived susceptibility (beliefs about magnitude of the possibility to be exposed/suffering from the condition), cues to action (contributing to perceived threat), perceived bene ts (belief about the usefulness of taking a particular action), perceived barriers (negative belief of taking a particular action) [13] The perceived severity, susceptibility and cues to action taken together refer to perceived threat [14]. Recently, the perceived self-e cacy is added with HBM as another construct [10]. The degree of effectiveness and magnitude of such constructs are not same. For instance, perceived susceptibility explained health behavior better for prevention than treatment while the opposite was true for perceived bene ts and perceived severity [15]. In this viewpoint, the HMB is not affected by the perceived barriers and the cues to action is rarely considered as a determinant of HBM [16]. Observational literature suggests that inclusion of few modifying factors such as socio-demographics, socio-psychological and structural variables in the HBM are essential for assessing the beliefs, perception and the likelihood of preventive behavior of a certain disease [12,14].
However, there are numerous criticisms and inconsistencies of effectiveness of HBM when the sociostructural variables are not properly addressed. Inclusion of such variables in the HBM are essential for assessing personal behavior like immunization behavior and risk practices [17]. Based on the guideline of SES index of [11], we considered few socio-economic variables in our proposed model that are explained in the following.

Model speci cation
We speci ed the hypothesized model of PBGI regarding COVID-19 (see Fig.1

Research design and measurement of variables
The study follows explanatory sequential mixed method approach. The rst step of the study is pre-test pilot survey through which the study conducts con rmatory factor analyses and checks the internal consistency, reliability and validity of data.
The second step is also quantitative in nature that follows the perfectly experimental research design. In this viewpoint, we used structured questionnaire to collect data on the new respondents. The respondents interviewed for pre-test were not included in the main survey. To test the proposed hypothesized models, the study considered four constructs SES, MOBIN, PGI and PBGI regarding COVID-19. The indicators of exogenous construct SES (education, occupation and income) were primarily used for constructing MOBIN and then for measuring PGI and PBGI regarding COVID-19 among the slum and non-slum people of the study area. Though the data on demographic variables were collected for understanding the characteristics of the respondents, the hypothesized model did not include them for not being signi cant in the pre-test survey. The SES has been de ned as the socio-economic status for the slum or non-slum people based on the aggregate index score formed from their education, occupation and income. The levels of educational status were coded as Below JSC (including illiterate) =1, SSC= 2, HSC= 3, graduate= 4, postgraduate= 5, MPhil= 6 and PhD and above= 7. On the other hand, occupational status has been operationalized with the occupational levels adopted from the Quarterly Labor Force Survey 2017 with modi cation of codes considering the rank scores and the category others. In this viewpoint, In the third step, the study follows qualitative technique of data collection to explore other extraneous factors that can explain the preventive behaviors of slum and non-slum people of Bangladesh regarding COVID-19. In this regard, 10 Key Informant Interviews (KIIs) from slum and 10 KIIs from non-slum people were conducted.
Sampling technique, sample size determination and data collection Purposive sampling techniqe has been used to conduct this study for considering the very purpose of the study and selecting the relevant respondents who can provide the required information for conducting the study. Accordingly, the data has been collected from 100 slum people and 100 non-slum people of Mirpur area in Dhaka city. The reason of selecting Mirpur area as the sample respondents considering the availability of both types of respondents (i.e., slum and non-slum people coming of a uent families). To collect the data from a uent non-slum people, the study has selected DOHS area. On the other hand, for collecting the data from slum people, the study selected 5 slums of Mirpur including Jhilpar Slum of The previous studies following Smart PLS to develop structural equation model suggested that a sample size ranging from 100 to 200 was usually a standard starting point in carrying out path modeling [18]. The present study has determined a sample size of 200 (100 from slum people and 100 from non-slum a uent families). The formula for determining sample size was not used since the study adopted purposive sampling technique.
For the proper empirical investigation, a pre-test survey was conducted by us. 10 respondents from slum areas and 10 from non-slum a uent families were requested to participate in this pre-test survey. After successfully completing that pre-test, we moved forward to carry out nal survey. Respondents of pre-test were not included in the nal survey. The survey was conducted in Mirpur area from 25 June to 20 July 2021. While collecting data, appropriate sampling procedure was followed in order to avoid the sampling error. In this regard, all survey interviews were conducted by a group of trained data collectors. All respondents were explained the signi cance and implication of the present study. The interview of each respondent was taken care of for a long time. The data collectors did not indulge in any personal and irrelevant gossiping to avoid anchoring or in uencing the answers of the respondents. The survey strictly followed data collection protocols subject to maintaining compliance with human research. In addition, 20 KIIs (10 from slum and 10 from non-slum people) were also conducted since the study follows the explanatory sequential mixed method approach.

Results
The study found the whole sample as to be skewed towards greater representation of individuals with diversi ed education, occupation and income including demographic variables. Table 1 shows the distribution of socio-demographic and socio-economic status characteristics of the respondents.  Finally, the discriminant validity has been evaluated through Heterotrait-Monotrait (HTMT) ratio which should be less than 0.90 [21]. On the other hand, [22] argue that it should be less than 0.85. Our estimated HTMT ratios of SES NON SLUM, SES SLUM, MOBIN, PGI and PBGI are lower than 0.85 that make conformity of discriminant validity of data.

Structural Model
The structural model signi es the inner models where there are relationships between exogenous and endogenous constructs [23,24]. To evaluate the structural model, the study follows few steps. First of all, multicollinearity has been checked. The estimated values of the variance in ation factor (VIF) ensure that the constructs of the model are fairly free from multicollinearity problem since most of the VIF of lesser than 3 and three factors are beyond 5 [19].
Secondly, in terms of coe cient effect, the exogenous construct SES of slum people has negative effects on all the constructs such MOBIN (-0.206), PGI (-0.011) and PBGI (-.022) regarding COVID-19. On the other hand, the exogenous construct SES of non-slum people has positive effects on MOBIN (0.902) and PGI (0.545) but negative effect on PBGI (-0.143). The negative effects of SES among slum people imply that the slum people experience downward mobility while the positive effects of SES among non-slum people imply that the slum people experience upward mobility. However, the high SES of non-slum people have no direct signi cant effect on their PBGI regarding COVID-19. Rather, their SES has signi cant effect on PBGI mediated through MOBIN and PGI. These effects mean that the gap of SES between slum and nonslum people plays key role in increasing further perception gap and preventive behavior gap among them. The existing studies also predicts such effect of SES index on health beliefs and the likelihood of action or preventive behavior [11]. The study also shows that all the structural path coe cients between SES among slum people have been found to be non-signi cant at 5 percent level with the T-statistics lesser than 1.96 and P values greater than 0.05. By contrast, though the structural path relationship between SES of non-slum people and PBGI regarding COVID-19 is not directly signi cant, their SES has a mediated effect through MOBIN and PGI because of their greater T-statistics than 1.96 and lesser P-value than 0.05. However, PGI is found to be the strongest predictor of PBGI where the main contributor is SES of non-slum people ( Table 2 and Fig. 3).
Thirdly, the coe cient of determination (R 2 ) value (0.933) which explains the explanatory power of model indicates that 93.3% of the variation of the endogenous latent variable PBGI is explained by the associated exogenous SES of slum and non-slum people mediated through PGI which is directly in uenced by SES of non-slum and slum people and also mediated through PGI. Since the PBGI value exceeds 0.67, the model is strongly explained by both its positive and negative predictors. Thus, the hypothesized model is acceptable.
Fourthly, the effect size (f 2 ) is another instrument where it is possible to assess the effect size of exogenous constructs on the endogenous constructs. According to [25], the effect sizes of 0.02, 0.15 and 0.35 represent respectively small effect, medium effect and large effect of exogenous on endogenous constructs. The present study reveals that the construct SES NON-SLUM has directly large effect on MOBIN (2.14) and the largest mediated effect (SES NON-SUM*PGI) on PBGI (8.415) regarding COVID-19.
It should be noted that the construct SES NON-SLUM has small effect on PBGI (0.131). By contrast, the construct SES SLUM has very negligible effect on PBGI (0.006) and no effect on PGI while the construct has small effect on MOBIN (0.112). But, due to the negative structural coe cient effect, such SES effect of slum people indicates the negative effect on MOBIN.
Finally, based on the Blindfolding method, estimated predictive power or relevance of endogenous construct (Q 2 ) ensures that PBGI regarding COVID-19 has a strong predictive power and relevance since the Q 2 value is 0.921 that is greater than the threshold value of 0. This estimated value is consistent with the suggested Q 2 value of the Stone-Geisser model [26,27]. This Q 2 value indicates the model has strong predictive capacity and relevance in the study context.

Discussion
Since the study follows the explanatory sequential mixed method approach, in the nal step the study explores the open views of the slum people regarding their health perception and preventive behavior regarding COVID-19. The reason is that through this qualitative insight from the disadvantaged groups can help the social researchers to revisit the HBM and to nd out the in uential factors that can explain the poverty stricken people of our country. In this regard, 10 key informants from slum people and 10 key informants from non-slum people were interviewed.
The a uent non-slum people who are highly educated, occupationally prestigious and belonged to upper class in terms of earning can perceive that they are susceptible to be infected with COVID-19. They can also perceive severity of COVID-19, bene ts of preventive behaviors and self-e cacy regarding preventive action. They do not have negativity (barriers) of maintaining preventive measures. By and large, they try to follow the instructions of medical practitioners as evident from the quantitative results of the study. The strongest factor of a uent families is their education. For example, when they were asked about the reason of their perception of the severity of COVID-19 and preventive behavior regarding it, one of the key informants put it: What we see in the society is that the more you educated, the more you earn through occupations. The individuals who have higher education, they can get the high level job. And job facilities and their earnings lead them to aware of the health risks and thus they try to prevent themselves from risk behavior.
Our study nds the non-slum a uent families as to be upward in social mobility. However, the slum people were found to be downward in all respects.
Most of the slum people were downward in terms of education, occupation and income as evident from the quantitative ndings. The qualitative results were also consistent with the quantitative one in line with their perception and preventive behaviors regarding COVID-19.
The slum people are neither afraid of the severity of COVID-19 nor have perception regarding the possibility of their infection. As one of the key informants put it: Corona virus disease is not for us. I think, I am not at risk of coronavirus disease. I heard and see in television that coronavirus disease leads to death, but I don't see. I am reluctant about COVID-19.

Another informant says:
There is no bene t of wearing masks, using hand sanitizers or following other instructions given by doctors or health organizations. Rather, we have heard that due to receiving vaccine against COVID-19, many people have died.
One Hujur (religious leader) who is the slum dweller puts interesting observation: Corona virus is the Ajab (punishment) from Allah. Because, we have forgotten our creator. We are much more involved in crimes and offences than the past. So, Allah is teaching us and warning us to be corrected. But we are not correcting ourselves. So, corona virus is being continued. I think, if we cannot understand our offences and are not serious to prevent from bad deeds, this punishment from Allah will be continued.
Among the preventive measures regarding COVID-19, wearing masks is the common trend among most of the slum dwellers. However, they do not use surgical masks, but masks made of different fabrics. It should be noted that they usually do not wear mask. They wear mask in restricted places like cantonment or check post. One of the respondents said "I earn very little money by driving a rickshaw. We all can't afford surgical mask. We can wash and use the cloth-masks again and again. Hand sanitizer and safety kit is a luxury goods in slum area. This is their reality that they must live from hand to mouth. In terms of using hand sanitizer, most of them put it "We didn't use hand sanitizer. Even, we had no knowledge about it. But when World Vision provided Hexisol Hand Rub for us, then we began to use it." Slum houses were like box having no windows. When they were asked if they maintain household ventilation, one of the key informants, described the situation saying that: "The ventilation process takes place on its own. They do not have that luxury to afford window or any ventilation system in their houses. Only the door is the way of the wind ow in their houses. Even, we do not have adequate access to light and air." Social distancing is another barrier of the slum people. They could not maintain at least one-meter distance apart from each other. Every house is very small and one family lives in only one room. As a result, there is no scope for quarantine in a separate room if someone is infected with COVID-19. Due to the nancial crisis and extreme poverty, they cannot disinfect their houses and surroundings. Similarly, they cannot use tissues or handkerchiefs while coughing or sneezing. One of key informants put it: "Most of the time in a day, my family is starving for living from hand to mouth. Then, how can I buy facial masks, hand-sanitizers or other costly items for preventing from COVID-19 infection. You can tell our government to provide these items free and rehabilitate us with modern facility-based housing. Then we can follow all the instructions regarding COVID-19. For example, we are taking vaccine since vaccines are given free." Interestingly, when the Hujur was asked that will you take vaccine? He replied yes and argued that it is free. He also added that both Dua (prayers) and Dawah (medicine) are important for Shifa (cure from disease).
Despite rigorously following the scienti c methods, the present study has some limitations. The rst limitation is to conduct the study based on the respondents of only one area of Bangladesh, using purposive sampling considering the very purpose of the study and limited budget. However, if the number of sample areas were expanded, the study ndings would have been more nationally representative. In this regard, the future researchers are recommended for expanding the larger sample size including the different areas of Bangladesh. To do so, the researchers are recommended to conduct such study having necessary fund. The second limitation is to explain PBGI from social mobility theory and HBM though there are many other health seeking behavior models such as theory of reasoned action, ecological model etc. The third limitation of the study is not to predict the COVID-19 related preventive behavior among religious and non-religious people in Bangladesh since religion matters the pandemic that is empirically grounded. The nal limitation is to include only SES mobility developed by [11]

Conclusions
The general objective of the study is to understand the effect of the Intragenerational social mobility among slum and non-slum people on PBGI regarding COVID-19. This is the rst study to explain and predict the PBGI in the social mobility context of Bangladesh. The quantitative study ndings reveal that upward SES of non-slum a uent families in Bangladesh has strong mediating effect (SES NON-SLUM*PGI) on PBGI regarding COVID-19. Thus, the upward SES mobility of non-slum people increases the unequal health perception and preventive behavior regarding COVID-19 in the society. On the other hand, the downward SES mobility of slum people leads them to lie in the false perception and the risk behavior regarding COVID-19. This result has been consistent with that of [11] in their study on "Occupational Structure in America". But, when they were asked through open questions using qualitative approach, the facts became clearer suggesting that only SES based on education, occupation and income differs the perception index and preventive behavior index of the slum and non-slum people regarding COVID-19.
However, based on the estimated results and the qualitative insights, this study has some implications for the policy makers of our government as well as the medical sociologists and psychologists. First of all, the policy makers and government should take proper steps to boost up the mobility and increasing universalism without compromising the slum or destitute people who live from hand to mouth. Secondly, administrators should prioritize the educational and occupational aspects of all groups and communities of Bangladesh since these three indicators are the key role maker of unequal health perception and preventive behavior regarding COVID-19 in Bangladesh. Thirdly, the government should provide free equipment for the disadvantaged people like slum dwellers who live from hand to mouth. Fourthly, medical sociologists should revisit HBM adding with another socioeconomic factor 'free provision of protective equipment'. Finally, social scientists should conduct further studies in this eld to    Measurement model