Determinants of return migration intentions: Evidence from Zimbabwean migrants living in South Africa

Abstract

This paper uses a dataset of Zimbabwean migrants living in South Africa to examine the determinants of the probability of their returning to their country of origin. It analyses migrants' return migration intentions using a logistic regression that examines 10 demographic and socioeconomic factors. Six factors – reason for migrating, the number of dependants supported in the home country, the level of education, economic activity in the host country, the level of income and the duration of stay in the host country – are found to be statistically significant determinants of the return migration intentions. The main policy implication of these findings is that the chances of attracting back skills are high if political and economic stability can be achieved.

Keywords:

• migration theory
• return migration
• Zimbabwean migrants in South Africa

1. Introduction

In the literature the phenomenon of return migration is explained in the context of international migration theory. Three broad theoretical frameworks are employed to explain international migration: the neoclassical economic theory of migration, the new economics of labour migration theory (NELM) and the structuralist theory.

According to the neoclassical economic theory of migration as expounded by Todaro (1969), wage differences between countries are the major driving force of migration. People migrate for economic reasons, and it is assumed that before migrating they will have complete information about opportunities in the intended destination countries as compared with opportunities in the home countries. Return migration, according to the neoclassical theory, is a manifestation of a failed migration experience, meaning that the expected higher earnings have failed to materialise. In other words, lower income levels in a host country than expected could possibly prompt a return migration decision (Constant & Massy, 2002; Cassarino, 2004). This could be especially so if the lower income level is not commensurate with the education level of the migrant. Evidence in support of the neoclassical theory has been mixed. For instance, Cohen & Haberfeld (2001) found returning migrants to have higher schooling levels than those who did not return, while Reagan & Olsen (2000) could not find skill bias in their analysis of returnees from the US.

The second theoretical approach, the NELM theory as expounded by Stark (1991), views migration as a livelihood strategy employed by households and families to diversify income risks and overcome market constraints such as difficulty in obtaining credit and insurance in the country of origin. It is argued that households send out their best-equipped individuals to earn income abroad and the money they remit serves to spread income risks, increase income, improve living conditions and enable investment (see also Cassarino, 2004). According to this theory, return migration is a natural outcome of a successful migration experience, meaning that migrants would have met their goals of earning higher incomes and accumulating savings during their stay in the host country. Thus, according to this view, migrants stay in the host country for a limited period of time and any prolonged stay is a result of repeated postponement of return in order to achieve goals (Cassarino, 2004). Prolonged stays are associated with maintenance of strong transnational ties with origin societies. Some empirical evidence has offered support for the NELM theory. For instance, Ammassari (2004) has found some evidence of higher levels of entrepreneurship among African migrants that could be attributed to savings in the diaspora, while Alberts & Hazen (2005) have reported increased job prospects in the home country for Tanzanian students in the US.

A study by Constant & Massey (2002) simultaneously testing the neoclassical and NELM theories found mixed support for both, which suggests there is no unitary process of return migration. De Haas et al. (2009) suggest that the relation between integration processes and return migration is likely to depend on initial motives for migrating, opportunities in the origin and destination societies, and educational, cultural and other specific features of immigrant groups. In other words, they suggest there is no one-size-fits-all theory. Furthermore, it has been argued that both the neoclassical and the NELM theories have shortcomings in that the motivation for return is considered to be largely determined by financial or economic factors (Cassarino, 2004).

In a departure from the neoclassical economic theory and the NELM theory, the structuralist theory argues that the decision to return is a social and contextual issue affected by situational and structural factors (Cassarino, 2004). In particular, this author notes that the structuralist approach focuses on the extent of the impact that returning migrants may have on their origin societies and argues that the success or failure of returnees should be analysed by juxtaposing the ‘reality’ of the home economy and society with the expectations of the returnee. Writing more than 30 years ago, Gmelch (1980) argued that it could be difficult for migrants to obtain good information about the social, economic and political changes that would have occurred in their home countries during their absence. However, given the advances in communication technologies in today's age of information, this argument may no longer hold.

Cerase's (1974) typology of reasons for returning – ‘return of failure’, ‘return of conservatism’, ‘return of retirement’, and ‘return of innovation’ – can be seen as an attempt to model the structuralist approach to return migration. ‘Return of failure’ refers to migrants who return because they failed to integrate in the host countries. They would also have failed to get jobs that are necessary for survival and been unable to send remittances back home. ‘Return of conservatism’ refers to those who return after accumulating savings, as initially planned before migrating, ‘return of retirement’ to those who return to the country of origin in their old age to live on their pensions, and ‘return of innovation’ to those who return after acquiring new skills in the host countries.

While the theoretical framework of return migration has been subjected to extensive contextualisation, empirical studies on the determinants of the decision to return are sparse, owing to the paucity of data on returnees. Most empirical work on return migration has largely focused on what returning migrants contribute to economic development (Diatta & Mbow, 1999; Thomas-Hope, 1999; McCormick & Wahba, 2001). This paper contributes to the return migration debate by examining 10 factors hypothesised to be determinants of return migration: legal status in the host country, reason for migrating, gender, age, marital status, number of dependants in the home country, level of education, economic activity in the host country, level of income and length of stay in the host country. The choice of these factors was informed by empirical evidence which can be considered not yet conclusive. Further information relevant to return migration was drawn from studies by Stark & Galor (1990), who observe that migrants have a propensity to acquire additional skills and to save more money than their counterparts in the receiving countries, which facilitates their return; Dustmann et al. (1996), who show that return propensities increase with the age at entry, but decrease with the number of years of residence; Constant & Massey (2002), who observe that having a spouse in the country of origin increases the likelihood of return; and Zhao (2002), who observes that returning migrants are predominantly married male workers who are older than their fellow migrants and that spousal separation is a dominant factor in the decision to return.

Focusing on return migration intentions, the study used data obtained from a survey of 4654 Zimbabwean migrants living in South Africa (Makina, 2007). Among other things, respondent migrants were asked about their return intentions in the event of a stable economic and political environment being re-established in Zimbabwe. Two thirds of the respondents (66%) said they would return to the home country should there be stability. Two other survey studies conducted independently, by Bloch (2005), whose sample comprised Zimbabwean migrants in South Africa and the UK, and Chetsanga & Muchenje (2003) whose sample comprised Zimbabwean migrants worldwide, also showed that 66% of the migrants surveyed had intentions of returning. Although their sample sizes differed, these three studies used the same methodology of convenience sampling and hence their results can be directly compared.

Crucially, the decision to return should be viewed against the determinants of migration. Naude (2010), investigating determinants of migration from 45 sub-Saharan African countries over the period 1965 to 2005, found armed conflict and lack of job opportunities to be significant determinants. Similarly, Makina (2010), focusing on a sample of Zimbabwean migrants living in South Africa, found that the main determinants of migration from Zimbabwe were political reasons and lack of economic opportunities in the home country.

This paper focused on return migration intentions rather than actual return migration behaviour for two reasons. First, data on actual returnees are not available. Dustmann & Weiss (2007) observe that return migration is difficult to measure, pointing out that while countries may have registration procedures for assessing the number of incoming immigrants, there are no procedures that track those exiting. Second, as De Haas et al. (2009) observe, return might not be driven by an actual desire to return, so there could be discrepancies between intentions and actual migration behaviour. Research on return intentions should therefore provide indications of actual return migration.

The rest of the paper is structured as follows. Section 2 provides an overview of migration trends for Zimbabweans entering South Africa, Section 3 analyses the migrant data that were used to identify factors influencing return migration and presents some preliminary indications from descriptive statistics, Section 4 describes the estimation model used to determine the statistical significance of determinants of return migration, Section 5 presents the logistic regression results and discusses these with reference to similar empirical studies, and Section 6 concludes.

2. Contemporary migration trends for Zimbabweans

As is well-known, the controversial land programme and disputed general elections that have characterised Zimbabwe since 2000 caused an unprecedented political and economic meltdown in this once stable country. The IMF (2009) reports that by the end of 2008 GDP had cumulatively declined by 54% since 2000 and macroeconomic instability had given way to hyperinflation. Over the years, Zimbabweans have responded to the crisis by emigrating in large numbers, to destinations varying from nearby southern African countries to others as far away as New Zealand, Australia, Canada, the UK and the US (UNDP, 2010).

While a sizeable number of Zimbabweans are to be found in almost every country across the globe, the most favoured destination is neighbouring South Africa. Figure 1 shows that the Zimbabwean migrant population growth in South Africa had soared to an estimated 2.12 million by the end of 2009 (UNDP, 2010). Globally, the UNDP (2010) estimate the number of Zimbabweans living outside the country to be between three and four million (about a quarter of the country's population) distributed in different countries and regions. Those living in Africa are estimated to constitute at least 2.52 million – 83% of the total Zimbabwe diaspora. South Africa alone houses over two thirds of the total, thus making the country a unique laboratory for studying Zimbabwean migrants.

Figure 1: Trend of Zimbabwe population in South Africa

3. Data and analysis of determinant variables

The data for this study were obtained from a survey of Zimbabwean migrants living in South Africa that was conducted in 2007 and reported by Makina (2007, 2010). The survey, conducted in three suburbs of the inner city of Johannesburg (Hillbrow, Berea and Yeoville), provided some insights into the profile, activities and migrant behaviour of Zimbabwean migrants living in South Africa's largest city. While it is acknowledged that Zimbabwean migrants are dispersed throughout South Africa, the 2001 South African Census reported that 80% of the total recorded Zimbabwean migrant population lived in inner-city Johannesburg. A total of 4654 Zimbabwean migrants (excluding visitors and informal cross-border traders) living in South Africa were interviewed using a structured questionnaire. Makina (2010) explains that random sampling methods could not be used to identify respondents due to the lack of a reliable sampling frame and so convenience sampling was used instead. The researcher justified the representativeness of the sample on the basis of its large size and the fact that so many Zimbabweans are concentrated in inner-city Johannesburg (Makina, 2010).

Among many other findings, the survey reported that 66% of respondent migrants said they intended to return to Zimbabwe in the future (Makina, 2010). In the original 2007 study the return intentions were not broken down in terms of socioeconomic characteristics of the migrants. In this paper the dataset is further manipulated so that the decision to return or to stay is broken down as it pertains to legal status in the host country, reason for migrating, gender, age, marital status, number of dependants left at home, level of education, economic activity in the host country, level of income, and length of stay in the host country. The resulting factors are shown in Table 1. It is noteworthy that data of 4639 migrants out of the original sample of 4654 could be manipulated to enable analysis of the 10 variables hypothesised to be determinants of the return migration decision.

Table 1: Determinant variables and analysis of the decision to return

		Return or stay
		Stay in South Africa		Go back home		Total
Determinant variables		N	Percentage	N	Percentage	N	Percentage
Gender	Female	663	41.8%	1262	41.3%	1925	41.5%
	Male	925	58.2%	1790	58.7%	2715	58.5%
	Total	1588	100.0%	3052	100.0%	4640	100.0%
Legal status	Documented	579	36.5%	1427	46.8%	2006	43.2%
	Undocumented	1009	63.5%	1625	53.2%	2634	56.8%
	Total	1588	100.0%	3052	100.0%	4640	100.0%
Number of dependants in Zimbabwe	None	144	9.1%	154	5.1%	298	6.4%
	1–2	340	21.4%	633	20.8%	973	21.0%
	3–4	631	39.8%	1321	43.4%	1952	42.2%
	5 +	472	29.7%	935	30.7%	1407	30.4%
	Total	1587	100.0%	3043	100.0%	4630	100.0%
Marital status	Single	596	38.2%	1064	35.2%	1660	36.2%
	Married	789	50.6%	1738	57.5%	2527	55.1%
	Widowed	55	3.5%	91	3.0%	146	3.2%
	Divorced	120	7.7%	130	4.3%	250	5.5%
	Total	1560	100.0%	3023	100.0%	4583	100.0%
Age	<18 years	17	1.1%	15	.5%	32	.7%
	18–20 years	94	6.0%	129	4.2%	223	4.8%
	21–30 years	625	39.7%	1224	40.3%	1849	40.1%
	31–40	632	40.1%	1211	39.9%	1843	40.0%
	41–50	177	11.2%	422	13.9%	599	13.0%
	51+ years old	30	1.9%	37	1.2%	67	1.5%
	Total	1575	100.0%	3038	100.0%	4613	100.0%
Level of education	University degree	34	2.2%	150	4.9%	184	4.0%
	Professional qualifications	135	8.6%	693	22.8%	828	18.0%
	Post-secondary diploma/certificate	112	7.1%	354	11.7%	466	10.1%
	Secondary education	1175	74.7%	1685	55.5%	2860	62.0%
	Primary education and other lower forms of education	118	7.5%	155	5.1%	273	5.9%
	Total	1574	100.0%	3037	100.0%	4611	100.0%
Economic activity	Domestic worker/gardener	181	11.6%	331	11.0%	512	11.2%
	Security	245	15.7%	369	12.2%	614	13.4%
	Hairdressing	93	6.0%	119	3.9%	212	4.6%
	Shop assistant	105	6.7%	173	5.7%	278	6.1%
	Teacher	53	3.4%	270	8.9%	323	7.1%
	Driver	40	2.6%	75	2.5%	115	2.5%
	Health professional	26	1.7%	133	4.4%	159	3.5%
	Hawking	173	11.1%	269	8.9%	442	9.7%
	Miscellaneous professions	37	2.4%	98	3.2%	135	2.9%
	Artisan	117	7.5%	296	9.8%	413	9.0%
	Hospitality workers	247	15.8%	338	11.2%	585	12.8%
	Other (miscellaneous menial jobs)	242	15.5%	549	18.2%	791	17.3%
	Total	1559	100.0%	3020	100.0%	4579	100.0%
Monthly income	R1000 or less	308	20.1%	605	20.6%	913	20.4%
	R1001–2000	639	41.7%	1077	36.7%	1716	38.4%
	R2001–4000	415	27.1%	785	26.7%	1200	26.8%
	Over R4000	171	11.2%	471	16.0%	642	14.4%
	Total	1533	100.0%	2938	100.0%	4471	100.0%
Reason for migrating	Economic reasons	588	37.2%	1435	47.1%	2023	43.7%
	Economic and political reasons	173	10.9%	421	13.8%	594	12.8%
	Economic and other reasons	49	3.1%	74	2.4%	123	2.7%
	Political reasons	718	45.4%	1026	33.7%	1744	37.7%
	Political and other reasons	13	0.8%	21	0.7%	34	0.7%
	Other reasons	41	2.6%	70	2.3%	111	2.4%
	Total	1582	100.0%	3047	100.0%	4629	100.0%
Length of stay	Arrived before 2000	165	10.4%	190	6.2%	355	7.7%
	Arrived 2000 onwards	1423	89.6%	2861	93.8%	4284	92.3%
	Total	1588	100.0%	3051	100.0%	4639	100.0%
Source: Adapted from the dataset used by Makina (2007, 2010).

While not conclusive, the analyses in Table 1 give preliminary indications as to how the 10 variables might influence the return migration decision. Looking at the likelihood of returning to the home country, we find male migrants (58.7%) more likely to return than female migrants (41.3%), undocumented (illegal) migrants (53.2%) more likely than documented (legal) migrants (46.8%), migrants with one or more dependants at home more likely than those without dependants, married migrants more likely than single migrants, and migrants in the age range 21 to 40 years (81.2%) more likely than any other age ranges.

With regard to socioeconomic factors, preliminary indications are that migrants with university degrees, professional qualifications and post-secondary education are more likely to return than those with primary education. The majority of those in all professions said that they would return, the highest proportions being those employed as hospitality workers, artisans and teachers. However, migrants earning more than R4000 (US$500) per month are less likely to return than those earning a lower monthly wage, indicating that income level is a stronger determinant of the return migration decision than profession.

Another finding was that migrants who left for purely economic reasons (47.1%) or political reasons (33.7%) were more likely to return than those who left for other reasons or for a combination of political and economic reasons. Indications are that migrants who left Zimbabwe from the year 2000 onwards, when the country was plunged into an economic and political crisis, are more likely to return than those who left before the year 2000.

Since the above simple analyses do not provide the statistical significance of the hypothesised determinants of return migration intentions, the study proceeded to use a more rigorous logistic regression model which enables the identification of relationships while controlling for the other variables that may be related to return migration decisions.

4. Estimation model

The return migration decision has a dichotomous outcome: the migrant either returns to the country of origin or decides to settle in the host country. Logistic regression is employed to analyse such dichotomous outcomes (Cox & Snell, 1989; Cabrera, 1994). The main advantage of using logistic regression is that we do not have to make any assumption about the distribution of the independent variables as they need not be normally distributed. Furthermore they do not have to be linearly related to the dependent variable or of same variance within same category. Logistic regression has been recently used to investigate the phenomenon of return migration (see Zhao, 2002; Thomas, 2008).

In this study, the dependent variable is the return migration decision, which takes the value one, with the probability of returning being q. Alternatively, the return migration decision could take the value zero, indicating that not returning (staying) would have a probability of (1 – q). The predictor (independent) variables, i.e. the 10 determinant variables being tested, can take any form. The logistic regression equation that is estimated is given as follows:

where a = the constant of the equation and b = the coefficient of the predictor variables entered into the model, namely β₁ for legal status, β₂ for reason for migrating, β₃ for gender, β₄ for age, β₅ for marital status, β₆ for dependants in home country, β₇ for level of education, β₈ for economic activity in host country, β₉ for income level and β₁₀ for duration of stay.

The model computes probability of return using odds ratios. If the odds ratio of a predictor variable is greater than one, returning is more likely to happen than not. If the odds ratio of a predictor variable is less than one, then return migration is less likely to happen. The Wald test is used to test the statistical significance of each coefficient in the logistic regression model and, if statistically significant, the variable concerned predicts the return migration decision. The Hosmer and Lemeshow statistic evaluates the goodness-of-fit of the regression and ordinarily should not be significant at 5% significance level if there is a good fit. If the Hosmer and Lemeshow statistic is significant, this indicates that the regression does not have a good enough fit to enable meaningful analysis to be done. In other words, the variables that would have been included in the regression would not be suitable for analysis.

5. Logistic regression results and discussion

Using the SPSS statistical package, the estimation results of the logistic regression Equation (1) are as shown in Table 2.1

Table 2: Results of binary logistic regression

Variables		B	S.E.	Wald	df	Sig.	Exp (B)
Constant		–0.902	0.527	2.932	1	0.087	0.406
Legal status		–0.110	0.081	1.849	1	0.174	0.896
Reason for migrating				35.074	5	0.000***
	Economic	0.238	0.245	0.942	1	0.332	1.269
	Economic and political	0.344	0.261	1.737	1	0.188	1.411
	Economic and other	–0.393	0.312	1.591	1	0.207	0.675
	Political	–0.139	0.250	0.308	1	0.579	0.871
	Political and other	–0.164	0.442	0.138	1	0.710	0.848
Gender		0.060	0.079	0.587	1	0.710	1.062
Age				10.929	5	0.053*
	18–20	0.236	0.468	0.255	1	0.613	1.267
	21–30	0.416	0.449	0.860	1	0.354	1.516
	31–40	0.159	0.453	0.123	1	0.726	1.172
	41–50	0.283	0.463	0.374	1	0.541	1.327
	50 +	0.052	0.528	0.010	1	0.921	1.053
Marital status				6.338	3	0.096*
	Married	0.140	0.085	2.728	1	0.099*	1.151
	Widowed	0.087	0.203	0.183	1	0.668	1.091
	Divorced/separated	–0.180	0.156	1.340	1	0.247	0.835
Number of dependants in home country				31.003	3	0.000***
	1–2	0.783	0.163	23.145	1	0.000***	2.188
	3–4	0.845	0.157	29.031	1	0.000***	2.328
	5 +	0.850	0.158	28.752	1	0.000***	2.339
Level of education				102.685	4	0.000***
	University degree	0.958	0.262	13.395	1	0.000***	2.605
	Professional qualifications	1.155	0.183	39.876	1	0.000***	3.174
	Post-secondary diploma/certificate	0.620	0.186	11.105	1	0.001***	1.858
	Secondary	–0.071	0.142	0.254	1	0.614	0.931
Economic activity in host country				33.813	11	0.000***
	Security	–0.260	0.140	3.461	1	0.063*	0.771
	Hairdressing	–0.508	0.175	8.404	1	0.004**	0.602
	Shop assistant	–0.191	0.165	1.339	1	0.247	0.826
	Teacher	0.155	0.209	0.555	1	0.456	1.168
	Driver	0.082	0.242	0.115	1	0.734	1.086
	Health professionals	0.406	0.264	2.371	1	0.124	1.501
	Hawking	–0.289	0.145	3.954	1	0.047**	0.749
	Miscellaneous professions	–0.048	0.243	0.039	1	0.843	0.953
	Artisans	0.093	0.165	0.317	1	0.573	1.097
	Hospitality workers	–0.436	0.138	9.931	1	0.002	0.647
	Other	–0.080	0.144	0.311	1	0.577	0.923
Monthly income level				11.603	3	0.009***
	R1001–2000	–0.176	0.097	3.289	1	0.070*	0.839
	R2001–4000	–0.342	0.110	9.614	1	0.002***	0.711
	Over R4000	–0.392	0.144	7.453	1	0.006***	0.676
Duration of stay		0.594	0.130	20.798	1	0.000***	1.811
	–2 Log likelihood	Cox and Snell R-squared		Nagelkerke R-squared
	5135.187	0.087		0.121
Hosmer and Lemeshow test
	Chi-square	Degrees of freedom		Significance
	13.959	8		0.083
Note: *, ** and *** denote significance at the 10%, 5% and 1% confidence interval respectively.

According to the Nagelkerke R-squared, the predictor variables account for 12.1% of the amount of variance in the dependent variable (return migration decision). This means that 87.9% of the variance in the return migration decision is accounted for by other variables not included in the model. However, at a 95% confidence level the Hosmer and Lemeshow test is not statistically significant, meaning that the logistic regression model fits the data adequately, thus ruling out the need for further stepwise iteration of the logistic regression model. Interpretation of results is undertaken in terms of the Wald statistics and odd ratios for sub-groups of variables shown as Exp (B) in Table 2.

In terms of the Wald test, the coefficients of legal status, age, gender and marital status are not statistically significant at the 95% confidence level, meaning that they are not significant predictors of the return migration intentions. As in the study on return intentions of Moroccans by De Haas et al. (2009), in this study return intentions were not found to be significantly affected by the respondents' age and gender. Gibson & McKenzie (2009), however, found age to be significant and gender to be insignificant for return intentions among migrants in the Pacific.

The other six variables, level of education, reason for migrating, economic activity in host country, number of dependants left at home, duration of stay and level of income, are statistically significant at the 95% confidence level, showing that they are statistically significant predictor group variables for the return migration intentions.

The odds of returning for the sub-groups of the predictor variables are represented by Exp (B) in Table 2. An odds ratio that is greater than one indicates a greater probability of return migration. The odds ratios for the sub-groups of the predictor variables (shown as statistically significant by the Wald test) are discussed below.

Level of education: The reference for comparison here is a migrant with primary education and other lower forms of education. It can be observed that migrants with university degrees are 2.605 times more likely to return in terms of the odds ratio. Migrants with professional qualifications (such as teaching, nursing, artisans) are more than three times as likely to return (an odds ratio of 3.174). Those with post-secondary education (diplomas and certificates) have an odds ratio of 1.858. Generally, the higher the migrant's level of education, the higher the probability of return migration.

The level of education was also found to be a significant predictor of the return migration decision by Gibson & McKenzie (2009), who examined actual professional returnees in countries of the Pacific (New Zealand, Tonga and Papua New Guinea). They found more migrants with PhDs had returned than those without a PhD. Similarly, Zhao (2002) found returning migrants to have the highest educational levels. Using data from Uganda, Thomas (2008) observed that native-born returning migrants with university degrees had a higher chance of getting employment than both native-born non-migrants and foreign-born professionals. The policy implication of these empirical results for a country experiencing a brain drain is that the chances of attracting skills back to the country are high. Skilled migrants are more likely to return than unskilled migrants who might not have better employment prospects in the home country. De Haas et al. (2009) found the effect of educational attainment to be positively related to return intentions but non-linear among Moroccan migrants. They found that migrants who had completed pre-school or primary education had the highest likelihood of return intentions, followed by the highest educated migrants. In contrast, the unqualified migrants were found to have the lowest likelihood of return intentions. This behaviour seems to tally with what was observed in the present study, that teachers, health professionals and artisans are more likely to return than those without trades. However, this is not consistent with the liquidity constraints theory that predicts that poorer individuals are more likely to return because they would have migrated to alleviate liquidity constraints at home. Gibson & McKenzie (2009) also observed this inconsistency, finding wealth to be positively correlated with return migration.

Reason for migrating: The reference for comparison here is a migrant who migrated for other reasons, such as family reunification or further studies. It can be observed that migrants who left for purely economic reasons and for a combination of economic and political reasons are more likely to return than those who left for other reasons. The odds ratio of those who left for economic reasons is computed to be 1.269, while those who left for economic and political reasons have an odds ratio of 1.411. No other combination of reasons has an odds ratio of more than one. It should be noted that while the reason for migrating as a group is a significant predictor of the decision to return, the odds ratios for specific reasons such as political and economic reasons though greater than one (thus indicating high probability of return) are not significant.

That the reason for migrating is a significant predictor of return migration is an important finding. Given that the migrants said they intended to return should the political and economic situation in Zimbabwe stabilise, it is not surprising that either economic reasons or a combination of economic and political reasons for migration are strong predictors of return migration. Political refugees (migrants) usually return to the home country when a preferred political system is in place. The behaviour of migrants who migrated for purely economic reasons appears to be consistent with the NELM theory. De Haas (2009) also observed that improved economic conditions in the home country started a trend of return migration to Turkey.

Economic activity in host country: The reference for comparison here is a migrant employed as a domestic worker. It can be observed that migrants working as teachers (odds ratio of 1.168), health professionals (odds ratio of 1.501), artisans (odds ratio of 1.097) and drivers (odds ratio of 1.086) are the most likely to return. Among the four employee categories, health professionals have the highest odds ratio and hence the highest probability of returning to the home country. While the odds of returning are greater than one, they are not significant. However, economic activity in the host country as a group category is significant (refer to Table 2).

The role that economic activity in the host country plays in the decision to return has also been observed in other empirical studies. It has been reported that return migration is negatively related to the economic development in the host country (Lianos, 1980; Glytsos, 1991). Furthermore, Robolis & Xideas (1996) have observed that the variance of the return migration flow is determined by a combination of income and employment opportunities in both the host country and the country of origin. It can be argued from the results in this paper that economic conditions in South Africa and employment prospects in Zimbabwe (that would ordinarily favour skilled migrants) are likely to play a major role in influencing the return migration decision.

Number of dependants left at home: The reference for comparison is a migrant who has no dependants left in the home country. It is observed that migrants with one or more dependants left at home are more likely to return. The odds ratio is 2.188 for those who left one to two dependants and this increases fractionally as the number of dependants increases. The odds ratios are significant at 1% significance level (refer to Table 2). The positive relationship between migrants who left dependants at home and return migration is corroborated by evidence from Gibson & McKenzie (2009).

Duration of stay: Migrants who arrived in South Africa from the year 2000 onwards are more likely to return, having an odds ratio of 1.811, than those who relocated to the country before 2000. Therefore, the recent migrants have a higher probability of returning than older migrants who are likely to have integrated into the host country.

Consistent with expectations, the paper finds a negative relationship between duration of stay and return migration intention. The longer the migrant stays in the host country, the more likely it is that he or she will become integrated and unlikely to return to the home country. This is consistent with the neoclassical economic theory of migration. However, contrary to expectations, De Haas et al. (2009) found that among Moroccan migrants the duration of stay had a linear positive impact on the likelihood of intending to return: the longer they live abroad, the more likely it is that they will consider returning. The authors suggest that this could mean that many Moroccan migrants intend to return to the home country for retirement, that is, the ‘return of retirement’ in Cerase's (1974) typology. It is significant that the recent out-migration from Zimbabwe was triggered by a political and economic crisis and thus the return migration intentions observed here tend to support the structuralist theory of return migration.

Level of income: The reference for comparison here is a migrant earning a monthly wage of R1000 or less. It is found that migrants earning more than R1000 per month are unlikely to return. The wage ranges of R1001 to 2000, R2001 to 4000 and over R4000 all have odds ratios of less than one, indicating a low probability of return. The odds ratio decreases with higher income levels, showing that the higher the income level in the host country, the lower the probability of return.

The finding that the probability of return decreases as the level of income in the host country increases supports the neoclassical theory of migration, where migration is explained according to economic factors, specifically the wage difference between the sending country and the receiving country. However, other researchers have observed that the level of income alone as a determinant of return migration intention may not fully explain return migration. For example, Gibson & McKenzie (2009) found, among migrants in the Pacific, that the decision to return is most strongly associated with some social factors, with family and lifestyle reasons being stronger predictors of return than the extent of income gains from migration.

6. Conclusion

In summary, this study identified six significant determinants of return migration intentions: reason for migrating, age, the number of dependants supported in the home country, the level of education, economic activity in the host country, the level of income and the duration of stay in the host country. While these results may not be generalised to all countries, the policy implication for Zimbabwe in particular is that the chances of attracting back skills are high if political and economic stability is achieved. Strong ties seem to exist between the home country and migrants in South Africa. Skilled migrants are more likely to return than unskilled migrants who might not have better employment prospects in the home country. The promising observation is that those who left after the 2000 political and economic crisis are more likely to return than those who left before the crisis, which supports the structuralist view of return migration. However, militating against return migration is the widening wage difference between Zimbabwe and South Africa (see UNDP, 2010:36). While skilled migrants will have aspirations to return to contribute to the development of the country, they will be constrained by the low wage levels in Zimbabwe.

Notes

¹ This author first reported partial results of this work in a 2010 UNDP Working Paper 11, ‘The Potential Contribution of the Zimbabwe Diaspora to Economic Recovery’, which he co-authored (see UNDP, 2010).