Abstract: The objective of this study was to assess the factors that influence young migrant’s duration of stay in a high risk environment. Studies on return migration have predominantly focused on former migrants who have return to their home countries, in the case of international migration, or former migrants who have returned to their villages or hometowns, in the case of internal migration. Differing from the predominant cases in this literature, this study, examined the determinants of the estimation of return migration propensities among young migrants in a slum in the city of Accra (place of destination). Drawing on the concepts of urban environmental vulnerability or risks and human encumberments, this study explained the justification for analyzing the determinants of estimated return migration propensities in a slum setting. Using ordinal regression, this study undertake multivariate analysis of the determinants of return migration among young migrants in the face of risks and stressors at the place of destination. It was found that sex, marital intentions of young migrants, the current income being earned, whether an individual is learning a trade or not and whether young migrants know where to go to get information concerning how to deal with stressors are statistically significant and in fact, are the proximate determinants of the estimation of return migration propensities.
INTRODUCTION
Return migration determinants have prevalently been estimated among migrants who have returned to their countries of origin in the case of emigration (Kirdar, 2009; Rooth and Saarela, 2007) or their towns/villages of origin in terms of out-migration in the context of internal migration (Ni Laoire, 2007; Zhao, 2002). However, return migration propensities have not been explored. In this study, it was explained that the concepts of human encumberments and urban environmental vulnerability or risks thereby highlighting the vulnerability of young migrants. This study subsequently justified the analysis of the determinants of the estimation of migration propensities.
The model of development in Ghana has been argued to be associated with the patterns of internal migration experienced in the country. The clear differences in the levels of poverty (Table 1) between the South and the North and the unequal capacities of both geographical locations to respond to economic opportunities have affected the patterns of internal migration. This development pattern created a spatial dichotomy between the Northern and Southern parts of the country, (Kwankye et al., 2007) thereby, triggering the migration of people in the economically active age from the North to work in the agricultural and mining sectors in the South.
| Table 1: | Incidence of poverty by selected administrative regions from 1991-2006 |
| Source: Ghana Statistical Services, 2007 | |
Therefore, the Northern, Upper East and Upper West Regions of Ghana became a manual labor store for the relatively economic buoyant South (Kwankye et al., 2007; Awumbila, 2007; Anarfi et al., 2003; Gyimah, 2001).
More recently, young people have been migrating independent of their parents and guardians. These young people find themselves in different social and cultural areas and hence, become more vulnerable and encounter several risks related to inadequate shelter, reproductive health and injustices (Yeboah, 2008; Kwankye et al., 2007). Many end up living under very poor circumstances and are vulnerable to both physical and reproductive health risks, particularly reproductive tract infections through rape, thereby impacting adversely their reproductive health rights (Awumbila and Ardayfio-Schandorf, 2008; Awumbila 2007). These risks, among others to be discussed, are grounds for conceptualizing return migration propensities.
| Fig. 1: | A part of Old Fadama, an urban slum in the city of Accra, Ghana is showing |
Next, this study explained the conceptual underpinning of the study and situate the need to understand the determinants of return migration propensities. These concepts are urban environmental vulnerability or risk and human encumberments.
Urban environmental vulnerability or risk is defined as the product of hazard, assets and fragility (Davis, 2007) where, hazard is the frequency and magnitude of natural events, asset is population and shelter exposed to hazard. Fragility is the physical features of the built environment which could also be explained as being synonymous with methodical government neglect of environmental safety. Old Fadama, popularly called Sodom and Gomorrah, the study site, is an epitome of urban environmental risk. The place has inherent natural hazards of an urban environment (Fig. 1). Typical of most third world slums, Old Fadama lacks adequate latrines, potable water, drainage systems. Form of accommodation is mostly wooden structures with rubber roofs or rusted sheets (Tschakert and Tutu, 2008). Old Fadama is a squatter settlement and hence, highly fragile since it receives no support from government for infrastructure or for environmental safety programs. Therefore, the product of a high magnitude of natural events, high exposure to hazards and high fragility is high environmental vulnerability or risk. Old Fadama is a high risk place by all standards. Further, residents of Old Fadama suffer social exclusion. Social exclusion is conceptualized here in the sense, that the poor are excluded from higher valued places in the cities and left behind in the insecure areas such as refuse dump sites and places with inadequate sanitary facilities (Davis, 2007) which, without doubt, is the case for Old Fadama and it residents.
Human encumberments is a situation in which slum dwellers are perceived as societal burden and hence, regularly suffer the wrath of governments in the name of city beautification through remaking spatial boundaries to the benefit of property owners, investors and the elite (Davis, 2007; Stren, 1990). Poor residents in poor neighborhoods are forcibly evicted with most of the people becoming perpetual nomads or remaining in a state of permanent relocation (Agbola, 1997). For example, in Zimbabwe, the shantytowns of Harare where, the poor residents were perceived to be sympathizers of the opposition were treated as human encumberments. The police, mandated by the President embarked on operation drive out trash (Davis, 2007). In other words, the slum dwellers are conceptualized as trash or people with no dignity who ought to be treated as the powers-that-be wish.
Residents of Old Fadama, this study site, have been at the receiving end of this perception. The nick name of the slum, that is, Sodom and Gomorrah, is an embodiment of how the slum is conceptualized and thought of more recently, there have been several calls from the civil society, media and from the corridors of power about the need to evict the residents of Old Fadama. Most of the reasons cited do not suggest the need to resettle residents and enhance their welfare but to get rid of the nuisance in the city (Bentil, 2009; Koomson, 2009). The slum has been labeled as a national security threat, an environmental nightmare to residents of the city of Accra and city authorities have vowed not to allow illegalities to become legal (Bentil, 2009).
Young people who have moved from the Northern parts of Ghana to the city of Accra, in search of greener pastures, suddenly find themselves living in a slum. They encounter risks that are associated with where they live and they are treated as trash by other residents in the city and government officials (Bentil, 2009; Koomson, 2009). They are constantly threatened with ejection without being offered alternative means of accommodation (Bentil, 2009). As young people, they might be contemplating leaving the slum or abandoning their earlier dream of a happy city life.
The impact of these concepts, that is, urban environmental vulnerability or risk and human encumberment on return migration have not been explored and hence absent from the literature. Also, the insights these concepts bring to examining return migration propensities have not been assessed and hence leaves gaps in the literature. Consequently, based on respondents’ estimation of their likelihood of return to their places of origin, this study estimated the determinants of the estimated return migration propensities.
MATERIALS AND METHODS
Instruments: The measurement instrument for this study was a survey questionnaire which consisted of four broad sections. The questions were based on literature as well as the conceptual underpinnings discussed. The first part of the instrument collected information regarding the demographics of the respondents. The demographic information included age (completed years), age at entry into the slum, sex, ethnicity, education, marital status, household size, fertility intentions, as well as marital intentions. Part two elicited information regarding other social and economic variables such as employment status, occupation/type of employment, current income/earnings, social capital and networks and whether or not migrants were making arrangement for a prospective migrant to join them. In the next section, after a discussion of the risks encountered by these young migrants and current duration of stay (covariate), they were asked to rate their likelihood of returning home within the next 5 years: that is, very likely, possible, or not likely at all.
Further, the last part of the questionnaire elicited information regarding activities and actions that enhance well-being of the young migrants. These included learning a trade (e.g., seamstress, tailoring, hairdressing) and getting information about how to deal with stressors.
Collection of data: Field work was carried out during the summer of 2009 in Old Fadama. The population of interest is the young people aged between 10 to 29 years old who have migrated to the slum. In conformity with conventional definition of a migrant, the population of interest was to have lived in the slum for at least 6 months. The sampling method was random sampling and snow balling. Sampling was done through the aid of opinion leaders of the community. A total of 104 young people responded to the survey questionnaire consisting of 70% females and 30% males.
Analysis of data: To determine whether, there was an association between the response variable (likelihood of return) and the independent variables discussed, this study used an ordinal regression model. An ordinal regression model is used to model data when the response variable is ordinal. To arrive at the most important independent variables in the study, this study used a screening process where, only those independent variables that had an association with the response variable were included in the final model. Specifically, A One-Way Analysis of Variance (ANOVA) was used to ascertain the association between each independent variable and the dependent variable. Subsequently, χ2-test is used to assess the possible problem of multi-colinearity. The independent variables used in the ordinal regression equation after screening were: Sex (Male/Female), Marital Intention in Accra (No/Yes), Current Income (per week) in Ghana Cedis (None, 10-19, 20-29, 30-39, 40-49 and 50-59), Learning a trade (No/Yes) and Know where, to get information (No/Yes). An enumeration of the counts and percentages of each category for each variable is shown in Table 2. Though, the current age of respondents and the social capital (resources accruing to young migrants from social relations and institutions that enable them to withstand stressors) of respondents were not significantly associated with the dependent variable (likelihood of return), a bivariate analysis is undertaken to understand how they relate.
Limitations of the study: The findings, with regard to the determinants of return migration propensities, can not be generalized for all return migration propensity estimations.
| Table 2: | Description of variables used in the model |
This is because, the study is place-specific and hence, Old Fadama has peculiar characteristics with regard to urban environmental vulnerability that impact residents. Second, as with all surveys, this study is subject to sampling errors and bias that are inherent in random and snowball sampling. The reduction in the occurrence of this bias and representativeness were, however, minimized through ensuring that no two or more participants were selected from the same household. Also, respondents were selected across the length and breadth of the study area.
RESULTS
Multivariate: First of all, the goodness of fit tests which are the Pearson test (p-value = 0.439 and Deviance test (p-value = 0.512) indicated that the data had fit the model well. In other words, there is insufficient evidence to claim that the model does not fit the data adequately. If the p-values had been small (say<0.05) then, it would have concluded that data did not fit the model well.
In ordinal regression, one of the categories of the categorical variables is chosen as the reference category. In this study, for the response (dependent) variable, the Not likely at all category was chosen as the reference category. The categories labeled Very likely and Possible are estimated intercepts for the logits of the cumulative probabilities of the likelihood of returning for very likely and for possibility of returning, respectively. Because, the cumulative probability for the last response category Not likely at all is 1, there is no need to estimate an intercept for non likelihood of returning. The coefficient estimates (intercept estimates) here are, therefore, not of importance to the analysis. These intercept values do not depend on the values of the independent variable for a particular case. This is similar to the intercept in linear regression where the significance is usually not important. However, in this case, each logit has its own intercept. I now discuss the main results from Table 3.
The coefficient of 0.7710 for sex is the estimated change in the logit of the cumulative probability of the likelihood of returning when the males are compared to females, with all other independent variables being held constant. Because the p-value for estimated coefficient is 0.151, there is sufficient evidence to conclude that sex has an effect upon likelihood of return.
With regards to the variable Marital Intention in Accra, it can be seen that this is significant predictor of likelihood of return (p-value<0.0001). Since, the estimate of the parameter for this variable is negative, (estimate = -2.122) this indicates that people who said No (that is, not having the intentions of marrying in Accra) were less likely than those who answered Yes (to having the intentions of marrying in Accra) to have higher ratings for the response variable (likelihood of return). In other words, people who said Yes were more unlikely to return than those who said No.
Using the same reasoning, it was noticed that the two variables-Learning a trade and Know where, to get information were also significant predictors of the response variable with p-values of 0.033 and 0.029, respectively. Both also had negative values for estimates for the coefficients. These estimates are -1.163 and -1.142, respectively. In both cases, Yes was used as the reference category and No was used as the estimable category. Therefore, it was concluded that people who responded No are less likely than those who responded Yes (to both learning a trade and knowing where to get information) to have higher ratings for the response variable for both of these variables. In other words, for the variable Learning a Trade, those who answered No were more unlikely to return than those who answered Yes. Likewise, for the variable Know where to go for Information, those who answered No were more unlikely to return than those who responded Yes.
The only other independent variable in the ordinal regression model was income. This variable had 7 categories. In this case, the over 59 Ghana Cedis per week category was the reference group. Income was also a significant predicator of likelihood of return with p-values<0.001 for each of the estimable categories. Also, all the coefficients are negative; this means that people who have smaller incomes are less likely than those with higher incomes (over 59 Ghana Cedis per week) to return to their former home or place of origin.
| Table 3: | Multivariate analysis of return migration propensities: ordinal regression |
| aThis parameter is set to zero because it is redundant | |
Bivariate: As mentioned briefly above, age is an important demographic variable relevant for return migration in the literature. Also, social capital is a crucial element so far as migration is concerned. Though each of these variables were not statistically significant in relation to the response variable, a bivariate analysis reveals an interesting relationship between them.
First, the age of respondents are grouped into 5 year age groups. These are early adolescence (10-14) late adolescence (15-19) early adulthood (20-24) and adulthood (25-29). From Fig. 2, it was found that 50% of early and late adolescence indicated that it is very likely they will return to their places of origin within the next 5 years. On the other hand, only 45 and 15% of early adult and adults, respectively indicated they would return within the next five years. With respect to not likely at all category, while 50 and 20% of early adolescents and late adolescents, respectively indicated it was not at all likely that they will return home in the next 5 years, 28 and 58% of early and late adults indicated same. Therefore, adolescent migrants do have higher propensities of return migration and adults have relatively lower propensities of return migration. That is, migrants within the ages of 10-19 have a shorter expected duration of stay in the slum compared to migrants aged 20-29.
Second, this study examined the relationship between social capital and expected duration of stay in the face of risk. Social capital index is computed for each individual migrant based on the following social capital constructs: interpersonal networks (consisting of family, relative and friend), relationships of trust (consisting of boyfriends/girlfriends and sugar dads/mums), membership in groups (ethnic associations, fraternity groups and others) and access to social institutions (non-governmental organizations and governmental institutions).
| Fig. 2: | Likelihood of return by age of respondents |
Social capital index is a composite measure of the cumulative connectedness of the individual migrants based on the social capital constructs mentioned. Migrants, consequently, fell into the following social capital quintile categories: lowest, low, medium, high and highest. From Fig. 3, respondents who had the lowest social capital had the highest return migration propensities. Forty percent indicated they will very likely leave for their places of origin within the next 5 years and 60% indicated it was possible they will leave within the same period of time.
| Fig. 3: | Likelihood of return by social capital index |
Young migrants in the highest social capital index had the lowest return migration propensities. Hundred percent of the respondents in the highest social capital index indicated they will not likely at all return within the next 5 years.
DISCUSSION
Due to the innovative nature of this study, the findings are unique since, this study have examined the factors that influence return migration propensities at the place of destination. This assertion is backed by the fact that return migration studies have focused on migrants who have returned to their places of origin. From the findings, it can be seen that return migration is not simply an inactive reaction to only economic factors but mediated by factors other than economic. These factors include sex, an individual’s marital intentions, as well as access to information.
Young people who have an intention of marrying at their destination irrespective of the risks have a longer expected duration of stay in the slum than their counterparts who do not have the intention of marrying at the destination. It must be noted that intentions are important aspect of migration decision-making framework (De Jong, 2000) with its roots from the theory of planned behavior (Ajzen, 1991). According to this theory, intentions are the principal determinant of people’s behavior in as mush as the implementation of these intentions is perceived to lead to the fulfillment of objectives. It was hence, concluded that marital intention is a proximate determinant of estimated return migration propensities.
| Table 4: | Respondents educational background by sex |
It was found that young males are associated with higher code value on likelihood of return. In other words, young males are more likely to have a shorter duration of stay in the face of risks. Young males are expected to be back home to take care of the elderly and the property, mostly land, of the family. Further, from Table 4, young male respondents have higher education than young female respondents. It is clear that over 55% of the women had no formal education, while none of them had a high school education. On the contrary, only 26% of their male counterparts had not had a formal education and 42% have had a high school education. Young males, therefore, have a higher motivation to return to school than their female counterparts.
Research participants who said they were learning a trade (for example, tailoring, seamstress, auto mechanics, hairdressing and wielding) have higher return migration propensities. That is, young migrates who are learning on a job have shorter expected duration of stay in the face of risks and adversity. Being treated as trash coupled with the persistence in the call for dwellers in Old Fadama to be ejected, we argue that, skilled young people will have a shorter expected duration of stay than their unskilled counterparts as evident in the results of this study. Further, we argue that, the findings is expected because after the training, they are likely to go back and expected to establish their own shops to do business thereby earning enough income, which will help improve their lives. These young migrants learning a trade do have a hope of a job upon return. This is another illustration of Ajzen (1991) planned behavior theory. It is usual to find in the city of Accra beauty shops, popularly known as Saloons. These Saloons are managed by professional hairdressers (mistresses) who either had a formal or informal training in hair management. These mistresses take on people who are also trained on the job in these shops. The same is true for tailoring/seamstress; auto mechanics; and wielding. These trainees are called apprentices. Therefore, the participants who said they are learning trade are apprentices. It is expected that after training, they also establish their shops and employ apprentices. Further, the duration of training for apprentices in the above professions usually takes 3 to 4 years. As a result, if these apprentices should go by the social expectation of starting a shop with apprentices when they finish their training, it is likely they will return home soon after graduation to start businesses in their places of origin, which have less number of people with such professions. These young people see enormous opportunity in their places of origin after graduation and do prefer those to the poor conditions of the slum such as poor drainage system, inadequate latrine facilities and inadequate potable water. More so, Zhao (2002), in a study among returned migrants in rural China, found that return migrants are an important source of investment activities. Return migrants invest in both farming and non-farming activities (cloth making). The findings of Zhao’s study indicate that a migrant who gains working skills will invest in the activity of which he or she attained skills. It is, therefore, not surprising those young migrants who are learning trades have shorter expected duration of stay in the slum. Consequently, it was concluded that, learning a trade is a proximate determinant of estimated return migration propensities.
Access to information is very crucial in every respect. Studies have shown the importance of social networks in information flow that result in positive outcomes (Devillanova, 2008). Devillanova (2008) assessed role of information networks in influencing immigrants’ access to primary health care with regard to the reliance on strong and weak ties. It was found that relying on friends and kin (strong ties) in order to get information considerably accelerated health care utilization than reliance on just acquaintances (weak ties). Most importantly, in the case of my respondents in a high risk place (that is, for example, living in a mosquito infested, muddy and fire prone environment) is to know exactly where to get information about how to deal with the stressors they encounter as young individuals. It was found that young migrants in a high risk area who do know where to get information to deal with the stressors they encounter have a shorter expected duration of stay compared with their counterparts who do not know where to go for information regarding how to deal with stressors. This makes sense since, they are likely to access the needed information to deal with stressors as well as information that will enable them get out of the slum.
As can be seen from the results, income is a major determinant of estimated return migration propensities among young people in the face of risks. It was found that lower incomes are associated with longer expected duration of stay and higher incomes are associated with shorter expected duration of stay in the face of risks and stressors. When young people reside in high risk environments such as Old Fadama, in terms of hazards and assets, for example, residents fall prey to flooding of rooms during moderate to torrential rains. This implies high discomfort during rains and worse of all rainy seasons. Such discomforting situation will encourage higher savings towards early movement from that environment. Since, income opportunities in places of origin have been found not to be strongly associated with return migration in the context of emigration (Gibson and McKenzie, 2009) it makes sense that higher incomes at the destination, especially in a high risk environment, is associated with shorter expected duration of stay. Further, we use the example that follows to illustrate the finding. Young people who earned 60 Ghana Cedis or more per week (about US $53 per week) would be able to save a substantial amount towards a venture in their home of origin. Ideally, living in the slum, the young individual may not spend half of their income as daily expenditure; hence, not less than half of that amount could be saved. Assuming half of this is saved; the young persons may have about US $1,192.50 per year in savings all other things being equal (26.5x45 weeks). Such an amount is a good enough capital for ventures like continuing education, starting fitting shop, opening a barbering/Saloon shop etc., in the Northern parts of Ghana where, cost of living is relatively low. If we estimate on the low side, for instance, a quarter of $1,192.50 will be US $298.13 per year in savings. Over a period of three years, with financial discipline, any young person who wishes to continue his education or start a small scale venture in the North will be able to do that with ($298.13x3 years) that is, US $894.40.
CONCLUSION
In sum, in the face of risks and stressors at a place of destination, young migrants do make decisions about their expected duration of stay. The decision to either stay shorter or longer is a function of a number of factors both remote and proximate. It was found that sex, marital intentions, learning a trade, knowledge of accessing information and income are the proximate determinants of estimated duration of stay of young migrants in a high risk area.
ACKNOWLEDGMENTS
I am grateful to Dr. Petra Tschakert, Dr. Ikubolajeh Logan, Dr. Lakshman Yapa and Dr. Kevin Thomas for their valuable insights and guidance towards this study. Also, I express my utmost gratitude to Isaac Dialsingh, a Ph.D Candidate at the Department of Statistics, The Pennsylvania State University, for his priceless comments on previous drafts of this study. The Africana Research Center and the Department of Geography at The Pennsylvania State University are acknowledged for funding this study. I also, thank Naa Dodua Dodoo, a student at the Regional Institute for Population Studies, at the University of Ghana for her help with the collection of data.