Human trafficking among Ethiopian returnees: its magnitude and risk factors

The current study was conducted at three cross border towns along the three major human trafficking (exit) corridors of Ethiopia, namely Mettema Yohannes, Moyale, and Galafi situated bordering Sudan, Kenya, and Djibouti, respectively. Mettema Yohannes and Moyale are just on points where the Cairo-Cape town High-way enters and leaves Ethiopia, respectively. Galafi is situated on the Ethio-Djibouti border through which the high-way passes. Therefore, all the three towns are the most important gates of human trafficking from all over the country. Individuals who returned to Ethiopia either willfully or by deportation via the three gates were contacted in person and included in the study from May 2016 to October 2016. Returnees might have left Ethiopia through other gates, including Bole International Airport, Bosaso, Humera, and Gambella in addition to the three gates; however, they had to return through the three gates to be included in the study.

A quantitative cross-sectional study was conducted to determine the magnitude of human trafficking among Ethiopian returnees and factors associated with it. A pilot study was conducted on 196 returnees (of whom 103 were trafficked persons) at Metemma Yohannes to check the acceptability and applicability of procedures and tools, and to get inputs for sample size determination for the main work.

The study population consisted of migrants who were abroad seeking better opportunities and returned or would be returned either willfully or by deportation during times closer to the study period. From the pilot study we learned that migrants could recall experiences, situations, and events which happened to them during the past two years. Therefore, Ethiopian migrants whose time of departures from home were during the last 3 to 24 months and returned via the three major trafficking (exit) corridors were included in the study.

Thus, the sample size was determined using the Epi-Info software version 7 for the cross-sectional study with a 1: 1 ratio of exposed to unexposed groups. After comparing the sample sizes determined using different factors in the pilot study, finally, a proportion of human trafficking of 53.48% for junior and lower educational level gave us the maximum sample size. In all cases, a confidence level of 95%, a power of 80%, an odds ratio of 1.5, a contingency of 15% for non-responses and a design effect of 1.5 were assumed. The adequacy of the sample to address the magnitude of human trafficking was checked, assuming an expected proportion of 50%, a margin of error of 5%, and a 95% confidence level. With these assumptions, the sample size was determined to be 1432 returnees. All emigrants coming back home either by deportation or by their own will are usually expected to report to the Ethiopian emigration offices located at the three border towns. Therefore, all returnees who were eligible for the study were interviewed at each check point or in the hotel they booked during the study period until the required sample size was secured.

The outcome variable of the study was trafficking status and was leveled as “trafficked” and “non-trafficked”. The trafficking status of each participant was ascertained by interviewing about their age and conditions during their recruitment, travel, and at destination. If a returnee was a child during migration (age less than 18), then by the UN 2000 definition of human trafficking he/she would be considered as trafficked irrespective of consent. Moreover, if a returnee was initially recruited by deception, fraud, or force by brokers or anyone else, and if there was any subsequent exploitation, may be either labor or sexual exploitation or child soldering, then he/she would be considered as a victim of human trafficking [1].

Sometimes, there were situations where traffickers bribe border guards and officials to smuggle migrants. Three or more folds of that money would be the means to control migrants and would be a debit bondage. There were also situations where multiple folds of the money paid by brokers for food, accommodation, transport, etc. was accounted to migrants’ credit. Or brokers could sell migrants to other brokers or traffickers; thus, they could be trapped by the network of continual exploitation unless they escape or give their hands to foreign governments to be deported. In all these and other similar conditions, even though they returned before reaching their destinations, we considered them as victims of human trafficking.

To answer questions about the underlying factors that drive migrants into the trafficking process, data were examined quantitatively using different potential risk factors. These include socio demographic factors (age, sex, marital status, ethnicity, religion, residence, and education), economic factors (household head occupation, household wealth index, family credit pressure before departure), smuggling status when crossing neighboring countries, exit corridor, responsibility to family livelihood, social support, pulling factor (included a questionnaire with 10 items about how people were attracted by foreign conditions), and belief and trust related pushing factor (based on a questionnaire consisting of 11 items that measures the degree of belief and trust people have on their country in changing their life by working at home); trust implies here the confidence they have in home country opportunities, resources, and governance relative to that of their potential destination countries. After conducting factor analyses on the two factors named as ‘pulling factor’ and ‘belief and trust related pushing factor’ (Additional file 1), two underlying factors (components or constructs) were emerged from each.

In consultation with sociologists, after examining the respective items of the puling factor that loaded on each underlying component, the emerged constructs were named as ‘exposure to seductive information about oversea life’ and ‘desire for successful oversea life’. Similarly, considering the belief and trust related pushing factor, the two underlying emerged constructs were named as ‘feeling hopelessness at success in home country’ and ‘risk-opportunity imbalance’. Risk-opportunity imbalance represents the extent returnees under-rate risks that they could face during trafficking and over-estimate opportunities at destination, or it is the extent of outweighing opportunities at destination to risks during trafficking. Thus, the score of constructs named as ‘feeling hopelessness at success in home country’ and ‘desire for successful oversea life’ were each scaled as weak, moderate, and strong by grouping the respective scores. On the other hand, ‘exposure to seductive information about oversea life’ and, ‘risk-opportunity imbalance’ were scaled as low, medium and high.

Human smuggling always involves illegal entry into any other country; thus, individuals involved in this process would be illegal immigrants or criminals [18]. Unlike trafficked persons who would be victims of deception or coercion and continual exploitation [1], smuggled persons would be free from their smugglers control when they reach destinations because their contractual agreement would be to transfer them through international borders up to a certain destination [18]. However, while they are traveling illegally through borders, they could fall in the hands of traffickers, or smugglers could transfer them to traffickers. Therefore, data regarding the smuggling status of each returnee immediately after departure, from Ethiopia to neighboring countries, were collected. However, to minimize the dilemma of its temporal relationship with trafficking, smuggling started at non-neighboring countries or at later times were not considered as far as they were not initially smuggled from Ethiopia to the neighboring countries.

A structured questionnaire was prepared in English (Additional file 2) and Amharic languages to collect data using face to face interview technique. At Metemma Yohannes where most of the returnees were encountered, two male and other two female data collectors, and a field supervisor were assigned. At Moyale and Galafi towns, two data collectors, one from each sex were deployed. Data collectors took training, including field exercise for two days to help them familiarized with the tool. Study participants were interviewed separately by interviewers of similar sex in the waiting rooms of each immigration office or in the hotels they booked. Interpreters were employed when data collectors and respondents did not speak a common language. The principal investigator and field supervisors made on site supervisions during the whole data collection period.

Each copy of the questionnaire filled with data was checked for completeness before it was fed into the computer. The study variables were coded and data were entered into Epi-Info software version 7 and transferred to Stata 14 for analysis (Additional file 3). Descriptive findings including the prevalence of human trafficking by different characteristics of returnees were presented using both texts and tables.

Before making regression analysis, two different groups of questions that were supposed to affect trafficking status [14] were examined using factor analysis. Thus, the first group of questions was about the pulling factor of migration that included 10 items, and the second group constituted the factor related to the degree of belief, attitude, or confidence that individuals had in improving their life by working in their home country. The latter factor which was named as belief and trust related pushing factor consists of 11 questions.

Both eigenvalue above one and scree test criteria were used to identify and retain the meaningful components of each of the two factors that explained much of the variability after varimax (orthogonal) rotation. After rotating the components, decision about factor patterns was based on whether each variable loads 0.40 or more on a component and less than 0.40 on the other components. Thus, variables that load on more than one component were dropped from the analysis. When two constructs emerged after using the principal component analysis, we checked whether both of them were divergently and discriminately valid by observing the correlation matrix, and their reliability was examined with Cronbach’s alpha.

Accordingly, two components each with more than 1 eigenvalue emerged from the factor named as the puling factor. One of the two constructs emerged was named as ‘exposure to seductive information about oversea life’ which represents the extent to which opportunities abroad were advertised to them before departure, and the other construct was named as ‘desire for successful oversea life’ (the level of attraction by materials and quality life that could be maintained or achieved abroad). The reliability test gave a Cronbach’s alpha of 0.84 to the construct named as exposure to seductive information about oversea life and 0.81 to desire for successful oversea life.

In the second factor named as belief and trust related pushing factor, two other components each with eigenvalue greater than one emerged. The construct that emerged first was named to be the level of feeling hopelessness at success in home-country (having more trust and confidence on host-country opportunities than that of home country), and it had a reliability test result of 0.70. The other emerged construct was named as ‘risk-opportunity imbalance’ (the extent of outweighing opportunities at destination to risks during trafficking), and it had a Cronbach’s alpha of 0.80.

Moreover, a composite wealth index was determined by weighting urban and rural specific indexes using the principal component analysis. Some of the questions included in the tool to measure wealth quintile were about the number of people in the household, type of toilet facility, construction material of house, etc. which were adapted from Ethiopian demographic and health survey [19].

To handle the correlation due to clustering of human trafficking by region/zone using a working correlation structure, the Generalized Estimating Equation (GEE) technique was employed so that the marginal effects of variables on human trafficking were determined. Independent variables with p-values of 0.20 or less were taken into the multi-variable analysis [20]. The effect of each covariate on the dependent variable was measured by the Adjusted Odds Ratio (AOR), and a p-value of less than 0.05 was considered as statistically significant.

A total of 1342 study participants with a response rate of 93.72% were included in the study. The mean age of the study participants was estimated at 23.95 + 4.55 standard deviations. The youngest and oldest migrants were 14 and 50 years old, respectively. Over half (59.69%) of the participants were males. Of the total participants, about 826 (61.55%) were Muslim, 274 (20.42%) Orthodox, and 223 (16.62%) Protestant by religion. Five hundred sixty-five (42.10%) of the participants were from Oromyia, 332 (24.74%) southern nations, nationalities, and peoples regional state (SNNPRS), and 287 (21.39%) Amhara regions (Table 1).

Of the total study participants, 683 were found to be victims of human trafficking, yielding a prevalence of 50.89% (95% CI: 48.22–53.57). Three hundred sixty-three (45.32%) of the males and 320 (59.15%) of the females were victims of human trafficking. The prevalence of human trafficking by exit corridor was 387 (47.54%) for Metemma Yohannes, 92 (51.11%) for Moyale, 181 (71.83%) for Galafi, and 23 (23.96%) for other gates (Table 2).

After the bi-variable analysis, ethnicity, religion, household head’s occupation, family credit pressure before departure, responsibility for family livelihood, and social support were excluded from the multi-variable analysis as their effects on human trafficking were not significant at 0.20 level of significance. Therefore, after adjusting for other variables, the odds of being trafficked for returnees aged 18–20 was about five times more than that of older migrants aged 30–50 (AOR = 4.65, 95% CI: 2.39–9.05). For females, the odds of being trafficked was 1.55 times (AOR = 1.55, 95% CI: 1.10–2.17) higher than that of males. The odds of being victims of human trafficking for migrants from rural areas was 1.49 times higher than that of migrants from urban areas (AOR = 1.49, 95% CI: 1.03–2.16). The odds of trafficking was more than four-fold for smuggled persons at departure when compared with non-smuggled ones (AOR = 4.48, 95% CI: 3.19–6.29).

Migrants who were moderately exposed to seductive information about oversea life had more than two-fold odds of being trafficked when compared with those who were exposed to such information at a lower level (AOR = 2.32, 95% CI: 1.55–3.48). On the other hand, the odds of being trafficked for migrants who had strong desire for successful oversea life were almost four-fold (AOR = 3.98, 95% CI: 2.63–6.02) times more than that of migrants who had a low level of exposure.

Returnees with a high level of risk-opportunity imbalance before their departures had more than six-fold odds of being trafficked when compared with those who had low level of risk-opportunity imbalance. On the other hand, the odds of being trafficked for returnees who felt strongly hopeless at success in home-country were about 9-fold (AOR = 8.64, 95% CI: 5.62–13.30) more than that of returnees who felt it at a low level (Table 3).

First, the study represents only returnees or those who would return closer in time to the study period. Because it is unclear that the reason why individuals who stay abroad and would not return is related to human trafficking and its consequences, our findings may not be generalized to such segment of the population. Second, migrants who returned by air were not examined quantitatively in this study. Moreover, a returnee might cross the border of a neighboring country without being smuggled, legally; even if he/she was smuggled to cross non-neighboring countries latter, we didn’t consider such returnees as smuggled. This was of course an attempt made to minimize the dilemma of temporality between smuggling and trafficking. Despite these limitations, we identified and quantitatively examined the different risk factors for human trafficking that were contextually relevant for developing countries, like Ethiopia.

Evidences showed that human trafficking is a severe form of violence affecting basic human rights and health [17]. Thus, our findings particularly the high prevalence of trafficking among returnees may imply the need to conduct other studies examining the magnitude of subsequent traumatic experiences such as physical, sexual, and psychological violence by recruiting only victims of trafficking. It is also justifiable to conduct similar assessments regarding common health consequences of trafficking such as physical, sexual, and mental health problems; a considerable number of returnees from trafficking in the country might have been suffering from such traumatic events and health outcomes that may need the attention of concerned institutions.

The current study examined socio-demographic and economic factors, smuggling status, exit corridors, and pushing (e.g. feeling hopelessness at success in come-country and risk-opportunity imbalance) and pulling factors (e.g. exposure to seductive information about oversea life and desire for successful oversea life) against trafficking status. However, still there could be a need to understand how these factors are associated to each other concomitantly to affect the trafficking status of individuals. For instance, it is unclear how exposure to seductive information about oversea life is related to a desire for successful oversea life; it is possible that a person who is exposed to an information advertising about oversea life may develop a desire for leading successful life abroad. Conversely, if a person has a desire to lead successful oversea life, then he/she may strive to get information that advertises about opportunities abroad, and vice versa. Thus, to examine this and other similar interwoven relationships among various factors as well as trafficking status, future researchers could employ appropriate statistical methods such as structural equation modeling that examines various variables simultaneously. Because human trafficking is a phenomenon influenced by various factors at individual, relationship/interpersonal, community, and societal levels, it is also worth considering the social ecological model theoretical framework [16] in analyzing its risk factors quantitatively in all these levels using multilevel analysis.

The higher risk of being trafficked for younger group of returnees than the oldest ones could imply the need for paying more attention to minors and younger groups. Because children or younger groups are immature both physiologically and psychologically, the consequences of exploitive and adverse trafficking conditions may be severe among them than older ones. The health seeking behavior of this segment of the trafficked persons may also be poor which could further worsen their health. On the other hand, the high vulnerability of female returnees to be trafficked than male counterparts could suggest that because of disparities in socio-cultural factors, males are relatively in a better condition with regard to the risk of being trafficked not only at home but also in other stages of trafficking after departure. In turn, the disproportionately trafficked women and girls may also be exposed to health problems such as unwanted pregnancy, abortion, and related complications which are unique only to this vulnerable group.