Labour Market Outcomes and Skill Acquisition in the Host Country: North African Migrants Returning Home from the European Union

D I S C U S S I O N P A P E R S E R I E S IZA DP No. 544 Labour Market Outcomes and Skill Acquisition in the Host Country: North Arican Migrants Returning Home rom the European Union Stephane Mahuteau Massimiliano Tani January 20 Forschungsinstitut zur Zukunt der Arbeit Institute or the Study o Labor

Labour Market Outcomes and Skill Acquisition in the Host Country: North Arican Migrants Returning Home rom the European Union Stephane Mahuteau NILS, Flinders University and Macquarie University Massimiliano Tani Macquarie University and IZA Discussion Paper No. 544 January 20 IZA P.O. Box 7240 53072 Bonn Germany Phone: +49-228-3894-0 Fax: +49-228-3894-80 E-mail: iza@iza.org Any opinions expressed here are those o the author(s) and not those o IZA. Research published in this series may include views on policy, but the institute itsel takes no institutional policy positions. The Institute or the Study o Labor (IZA) in Bonn is a local and virtual international research center and a place o communication between science, politics and business. IZA is an independent nonproit organization supported by Deutsche Post Foundation. The center is associated with the University o Bonn and oers a stimulating research environment through its international network, workshops and conerences, data service, project support, research visits and doctoral program. IZA engages in (i) original and internationally competitive research in all ields o labor economics, (ii) development o policy concepts, and (iii) dissemination o research results and concepts to the interested public. IZA Discussion Papers oten represent preliminary work and are circulated to encourage discussion. Citation o such a paper should account or its provisional character. A revised version may be available directly rom the author.

IZA Discussion Paper No. 544 January 20 ABSTRACT Labour Market Outcomes and Skill Acquisition in the Host Country: North Arican Migrants Returning Home rom the European Union * This paper studies the educational investment decisions o returning migrants while abroad in the context o their decisions about the choice o activity upon returning and the duration o migration. The theoretical model builds on Dustmann (999), Dustmann and Kirchkamp (992) and Mesnard (2004). Using data rom the MIREM database we explore whether the type o skills acquired by migrants while abroad is related to the activity chosen upon return and the duration o migration. The results suggest that the type o education plays a signiicant role in the migration decisions o those returning as wage earners or selemployed. In particular, there is a clear positive relationship between being sel-employed and having previously invested in vocational education in the host country. There is also a strong positive relationship between investing in university education abroad and becoming a wage earner. As international migration acilitates skill transers between developed and developing countries, the economic development o the latter will increasingly depend on migrants ability to access educational and vocational training in the developed world aside rom university education. Returning migrants with vocational and proessional training tend to be sel-employed ater returning home, and by so doing they contribute to reducing poverty in the host country. JEL Classiication: F2, J6 Keywords: return migration, human capital, education, duration o migration Corresponding author: Massimiliano Tani Department o Economics Macquarie University NSW 209 Australia E-mail: max.tani@mq.edu.au * We thank Jean Pierre Cassarino and the team at MIREM or providing us with the data.

Introduction Temporary migration, or a work spell abroad or a period o time that may span over several years, is an increasing eature o international labour lows, helped, among others, by globalisation and improved communication and transport technologies. Many temporary migrants remain economically active upon returning to the country o origin. This has prompted a number o authors to interpret migration as part o a work lie cycle. Temporarily moving rom a low to a high wage country (an observed phenomenon) is a strategy to ast track the accumulation o inancial resources to either start up a business activity or to increase uture consumption in the home country. Under this unctional interpretation o migration, its duration depends not only on the income earned in the country o origin and destination, as traditionally highlighted by the migration literature, but also on the migrants intended activity upon returning. For example, would-be entrepreneurs are predicted to have a shorter duration o migration than salaried workers and retirees. As they enjoy higher income in their home than their host country, they have no urther incentive to stay abroad once they accumulate enough savings to start their activity (e.g. Dustmann and Kirchkamp, 2002). However, while in the destination country, migrants do not enjoy a predetermined probability o economic success. This in turn depends on their individual characteristics, including the amount o human capital that they possess at the time o migration and, most importantly, the one which they subsequently acquire in the host country. Investing in human capital abroad enhances one s ability to integrate economically and socially there (Chiswick, 976). It also helps to maintain a degree o connectivity with the host country that may be strategic in promoting growth at home by acilitating international exchanges o goods and transers o knowledge and technology (e.g. Dosi et al., 988; Cohen and Levinthal, 989; Park, 2004). The choice o activity upon return and the duration o migration may thereore aect migrants incentives to invest in education while abroad: do they, and, i so, do they matter or the type o education in which returning migrants invest? This paper investigates these questions by analysing the educational choices, the activity chosen, and the duration o migration o Moroccan, Algerian, and Tunisian

migrants returning home rom a spell in the European Union. Among the studies ocusing on migration, little is known about the type o human capital that returning migrants acquire in the host country. Yet, this knowledge is relevant to understand the potential beneits and costs o temporary migration or both sending and receiving countries, and the potential need or policy action. This paper contributes to the existing literature by developing a theoretical model and by analysing data on returning migrants using the MIREM database. It organised as ollows. Section 2 reviews the recent literature on returning migrants. Section 3 presents a theoretical model o return migration in which migrants choose simultaneously the type o education investment in the host country, the activity upon returning, and the duration o migration. The model builds on (combining them) Dustmann (999), Dustmann and Kirchkamp (2002), and Mesnard (2004). Section 4 summarises the characteristics o the data. Section 5 presents the estimation strategy. Section 6 discusses the empirical results. Section 7 concludes. 2 Literature The existence o return migration to countries where the average wage is lower than the host country is at odds with the traditional approach o viewing migrants as income-maximising individuals. In such circumstances, migrants should intuitively remain abroad until retirement age. To explain the act that retirees are not the main group o returning migrants, existing studies typically introduce location-speciic preerences or the country o origin (e.g. Hill, 987; Djajic and Milbourne, 988; Galor and Start, 99; Dustmann and Kirchkamp, 2002; Mesnard, 2004). This approach relects both positive externalities o living at home (e.g. weather, riends, language, ood, culture) and negative externalities o living abroad (e.g. discrimination, racism). However, this is only the starting point or the explanations oered to interpret the observed variety o choice o activity and durations o migration. By ar, inancial considerations are the most commonly explored determinants o return migration and occupational choice. Migrants save while working abroad, and their savings can be used to acquire durable consumption goods and to invest in new business activities in the home country. Migrants oten become entrepreneurs upon 2

returning, and invest in new business ventures across several sectors o the economy, employing personnel, and positively contributing to poverty reduction and economic development. The preerence or sel-employment o many returning migrants is hence viewed as a consequence o the lack o adequately developed credit and inancial markets in the home country. Under such interpretation, would-be entrepreneurs acing local credit rationing meet the ixed setup costs o setting a new business with the savings made while temporarily working abroad (e.g. Mesnard, 2004). In such circumstances, their staying abroad depends on the degree o ineiciency o the local credit market: the more diicult it is to obtain credit locally, the longer is the duration o migration and the need to build up the savings needed or the desired investment (including a precautionary motive to ace unoreseen conditions). As access to a perect credit market implies that the decision to start a project is independent o the migrants personal inancial position, research has generally studied the relationship between sel-employment and the savings brought back. For example, Mesnard (2004) inds that the probability o choosing selemployment upon returning among Tunisian migrants is positively aected by the amount o savings accumulated abroad (albeit at a decreasing rate): quadrupling the average amount o savings raises the probability o sel-employment by almost 20%. Other empirical work based on survey data inds that savings accumulated abroad commonly inance investments in production in the country o origin and raise the probability o sel-employment. Examples include migrants returning to Pakistan (Ilahi, 999), Egypt (McCormick and Wahba, 200), and Ghana (Black, King, and Tiemoko, 2003). Aside rom the amount o investment required, inancial considerations also aect the duration o migration through income and substitution eects whenever income dierential between the home and the host country change. The sign o these eects are generally opposite. For example, an increase in the income earned abroad raises the incentive to stay in the host country or a wage earner, but lowers the marginal utility o wealth, reducing the utility o remaining abroad, and the duration o migration. The sign o the eects also depend on the choice o activity upon return. An increase in the income earned abroad generally increases the duration o migration or a wage earner, but not or a would-be entrepreneur, who can accumulate the target savings in a shorter period o time when oreign income rises. Extensive summaries o 3

the predicted eects o changes in income abroad and at home, and other (mostly inancial) variables are in Mesnard (2004, Table 3) and Dustmann and Kirchkamp (2002, Table 6). One valuable contribution o this literature is that the theoretical models developed are a useul ramework to analyse the duration o migration when there exist alternative occupations upon returning. Unortunately, these studies do not take into account the possible human capital investments o migrants while abroad, which motivates this paper. The role o human capital in determining the choice o activity and the duration o migration is seldom the main ocus o the existing literature. However, the educational level o returning migrants is commonly included as a control variable in the empirical analyses carried out. In those studies, the sign o the educational level o migrants in empirical applications is mixed. This possibly relects some peculiarity o the local labour market, and not only individual choices. For example, Ilahi (999) inds that Pakistan s returning migrants with a higher level o education are more likely to be wage earners, and suggests that the type o skills used by sel-employed, who oten operate in retail and wholesale trade, do not require a high degree o ormal education. A similar conclusion arises in Mesnard (2004), who inds that having no education raises the probability o becoming sel employed. In contrast, Dustmann and Kirchkamp (2002) ind that education has a positive eect on being selemployed. McCormick and Wahba (200) present a mixed case whereby the probability o becoming sel-employed upon returning rises with education or literate Egyptians, but not or those who are illiterate. This result may relect dierent types o entrepreneurial activity set up by returning migrants o higher and lower educational background (e.g. high value added services or the ormer and more basic activities or the latter). The little research on the human capital investment choices o returning migrants while abroad contrasts with the relatively developed literature highlighting that education and training completed in the host country improve migrants economic and social integration there, and the literature emphasising the role o human capital accumulation in promoting growth (e.g. Dosi, 988; Cohen and Levinthal, 989). I education raises productivity and success in the labour market, migrants would have an incentive in investing in human capital while abroad, as that would either raise 4

their returns and shorten their duration o migration. This possibility is explored by Dustmann (999), though with a ocus on oreign-speciic capital. He studies the relationship between acquiring language capital in the host country and the duration o migration or Turkish, Italian, Greek, Spanish, and Yugoslavian migrants working in Germany between the 950s and the 970s. The theoretical model that he develops suggests that the intended duration in the host country s labour market aects the acquisition o its language, and that both choices are made simultaneously. As a result, those intending to stay a longer period in the host country invest more in the oreign-speciic human capital, and vice-versa. An additional incentive in learning well the language o the host country is the value placed by the home country on the human capital gained abroad. I this is higher than the value placed by the host country, then there is a strong incentive to undertake the educational investment and return home (as it may be the case or international students). The empirical analysis supports this hypothesis: a longer intended duration has a signiicant positive eect on the probability o being luent in the host country s language. Furthermore, this eect is larger i the temporary migrant has a higher level o schooling, supporting the hypothesis that human capital is sel-producing (Ben-Porath, 967). One limitation o this theoretical approach however is the lack o distinction between dierent types o educational investment that are available to migrants. For example, migrants can either enrol in vocational and proessional training or university education (or do not invest in education). The model also does not allow migrants to have a choice o activity upon returning. This paper nests the theoretical ramework o educational investment into the broader theoretical model o duration o migration with multiple choices o activity upon returning. In particular, it contributes a theoretical model where the decision o investing in human capital in the host country can be studied vis-à-vis the choice o activity upon returning and the duration o migration. This theoretical approach enables one to study a more complex decision landscape aced by migrants, covering educational and labour market choices, as well as to analyse policy implications related to migration and education or both the home and the host country. 5

3 A theoretical model o education and activity choice or returning migrants The theoretical model discussed in this section combines the work o Dustmann (999), Dustmann and Kirchkamp (2002), and Mesnard (2004). The model describes the choices o a migrant arriving in the host country at time zero. The migrant has a human capital H, and his working lie ends at time T. The migrant makes two decisions during his working lie, which are illustrated in Figure. FIGURE THE MIGRANT S WORKING LIFE 0 s t T The irst decision is related to education, and occurs in the irst period o his stay, which is normalised at length. The migrant chooses between working using his existing human capital, or which he receives the host country wage w. Alternatively, he can orgo the wage w to invest a portion s o his unit o time to invest in human capital (0 s ). Although s is represented as a continuous variable, throughout the paper a small s (close to 0) is interpreted as representing short courses, such as proessional training and short vocational courses, and a large s (close to ) is viewed as a proxy or university education. The beneits o investing in education arise only ater the initial unit o time in the host country, and take the orm o a productivity improvement. By assumption, this does not depreciate or the rest o the migrant s working lie. The wage premium accruing to the additional human capital gained abroad is (s), with (s) = i s = 0, > 0, and < 0. The migrant makes his second decision at a time t, which is to return to the home country. There, he will enjoy a productivity premium ( s), with i = activity chosen i by the migrant upon returning and i to suggest that education gained abroad may have less home-speciic applications, and hence be valued less. In the home country the returned migrant becomes either sel-employed, in which case he earns w (or ( s) w i he has invested in education while abroad) and aces set up costs o BH SE (this cost is assumed to relate to the migrant s initial human capital to capture variation in the type o business activity undertaken, but it can also be a ixed cost B), 6

or a wage earner, in which case he earns w h (or ( s) w ). It is assumed that h W w > w > w h to highlight that a sel employed enjoys a higher real return at home than abroad, and that SE > W, to suggest that knowledge gained abroad can be applied only to a irm i sel-employed (and hence be possibly less transerable) rather than to a number o potential employers as is the case or salaried employment. The migrant s utility unction relects stronger consumption preerences or the home country, so: U = tuu ( c ) + ( T t) uh ( ch ) () where c indicates consumption, and the subscripts and h indicate host and home country, respectively, and u c ) < u ( c ). It is assumed that the utility unction u is concave so that u > 0 and u < 0. u ( h h tc The migrant s budget constraint is: ( s) + ( T t) ch Hw ( s) H ( t ) w ( s) h H ( T t) w + hbh h2h ( T t) wh ( s) = SE W 0 (2) where h = i the migrant chooses to be a sel-employed (h = 0 otherwise), and h 2 = i the migrant chooses to be a salary earner (h 2 = 0 otherwise). The migrant maximises utility U subject to the budget constraint BC with respect to the optimal consumption at home and abroad (c h, c ), the amount o time to invest in human capital in the host country s, and the duration o migration t. The maximisation generates a system where the education investment is determined simultaneously with the choice o activity and the duration o migration. As in the existing literature, the optimal choice o activity upon returning is obtained by comparing the indirect utility unctions or each activity. The properties o the model when duration is exogenous are investigated irst, as a number o countries impose a ixed term or temporary migration. Exogenous duration o migration 7

When t is exogenous, the irst order condition o the maximisation with respect to the educational investment s determines the optimal amount o investment in human capital while abroad. The condition is: / s = 0 = y '[( t ) w + h ( T t) w + h2 ( T t) wh SE W (3) From which one obtains: ds = "[( t ) w ' Axdx + h ( T t) w SE + h ( T t) w 2 h w (4) where A is the total dierential o the irst order condition with respect to x and x = w, w, wh, T, t, SE, W. The relationship between the investment in human capital abroad and the other variables is summarised in Table. TABLE THE MARGINAL EFFECTS ON EDUCATIONAL CHOICE WHEN DURATION IS EXOGENOUS Variable Marginal eect (ds/dx) w '( ) Dt D = w w h t Eect on s Uncertain: it is > 0 i s > ( ) and vice-versa t h SE ( T t) > 0 ' D h ( T t) > 0 ' 2 W D '( w hw SE h2whw ) < 0 i sel-employed (h =) D > 0 i wage earner (h 2 =) '( T t) h w < 0 SE 2 D SE W '( T t) h2 2 DW T w < 0 ( hw h + h w ' 2 SE h "( s)[( t ) w + h ( T t) w + h2 ( T SE W ) W > 0 D t) wh < 0. ' From Table it emerges that, or a given duration o migration, higher income rom sel-employment and wage in the home country increase the incentive to invest in 8

education abroad. The returning migrant in act enjoys a productivity premium rom the extra education obtained. However, an increase in the exogenous duration (e.g. a longer visa permit) reduces incentives to invest in human capital or would-be selemployed. Since they now have less time available to recoup the cost o setting up a new activity, they respond by lowering their investment in education abroad. This outcome does not occur to wage earners, as higher wages in the host vis-à-vis the home country coupled with the inability o choosing an earlier return, leads them to make a higher investment in education and enjoy the wage premium over a longer spell abroad. As expected, the lower transerability o skills reduces the incentives to invest in education abroad (ds/d SE < 0 and ds/d W < 0), and so does a shorter working lie (ds/dt > 0). Endogenous duration o migration When t is endogenously determined, the solution to the maximisation problem is obtained by totally dierentiating the derivatives with respect to t and s, and the budget constraint (2). This leads to: bt ds = b s b dt b x s dx (5) where b t, b s, and b t are the derivatives o y t) wh with respect to t, s, and x, '[( t ) w + h ( T t) w + h2 ( T SE respectively, and x = w w, w, γ, T,,., h SE W W It also yields dt = A A ( C λ t s / C λ ds + ) where y = x, H, B. Ay ( C y / Cλ ) dy A ( C / C ) λ t λ (6) This system highlights that the decision o investing in human capital in the host country depends on the choice o activity (through comparison o the corresponding indirect utility unctions) and the duration o migration. 9

The conditions or the existence o an internal solution consist o comparing the indirect utilities o the activity chosen upon return (sel-employed, wage earner, unemployed, not in the labour orce only the irst two activities are discussed in the paper), and ensuring that individuals emigrate and return. As in Dustmann and Kirchkamp (2002) this is done by calculating the value o the total dierential o the indirect utility unction or each activity and checking that the value obtained is positive as t 0 (so that emigration takes place) and that it is negative as t T (so that return occurs). 4 The sample The data on which the empirical analysis is perormed are extracted rom the sample o returning migrants interviewed through the MIREM project. The sample covers about,000 returning migrants rom Algeria, Morocco and Tunisia, who had a spell in the European Union. The survey was carried out in 2005-6, and the results can be accessed through the MIREM s website www.mirem.eu. A description o the data is also in Tani and Mahuteau (2008). Table 2 summarises the main variables used in the empirical analysis. TABLE 2 SUMMARY STATISTICS Code Mean Std Dev Description Current status AR_WFT 0.240 0.427 Currently ull time wage earner AR_SE 0.339 0.473 Currently sel-employed AGE_R 4.0 3.92 Age at return in home country R_PLACE 0.34 0.474 Returned to other place than beore emigrating B_STAY2 0.538 0.499 Intend to stay permanently ater returning R_K 0.288 0.453 Has kids ater returning Host country experience ANS_PPI 5.2 3.08 Nr years spent in main immigration country A_MARRY 0.39 0.466 Got married in host country VISA_WK 0.26 0.332 Working visa VISATOUR 0.283 0.45 Tourist visa B_REM 0.349 0.477 Sent money at least quarterly in last year abroad A_JOBT3 0.230 0.42 Got job within 3 months o emigrating Pre-emigration status B_WFT 0.466 0.499 Full-time wage earner beore returning B_WPT 0.37 0.344 Part-time wage earner beore returning B_SE 0.54 0.36 Sel-employed beore returning BD_SEC 0.369 0.483 Completed secondary education beore departure BD_TER 0.260 0439 Completed tertiary education beore departure BD_MP 0.273 0.446 Wanted to migrate permanently beore leaving BD_LAND 0.362 0.48 Possess land beore departure N 982 0

The respondents are rom the three countries o origin in broadly equal proportions, and are mostly males (87.4%). As shown in Table 2, 34% o the respondents are selemployed upon returning while 24% are ull time wage earners. Migrants appear both well educated and relatively well o: 26% o those surveyed have a university degree (37% have completed secondary school), and 36% owned land beore departing. This relatively high composition o skilled labour may contribute to explain why 23% o respondents obtained a job within three months o emigrating. About 27% o migrants wanted to migrate permanently beore leaving and 34% returned to a dierent town in the home country than where they were residing beore migrating. Less than 3% emigrated with a working visa, while over 28% moves abroad with a tourist visa, suggesting that a number o respondents may have illegally worked in the host country. The statistics give some indications about the work lie cycle experienced by the respondents: on average they remained 5 years in the host country, though there is signiicant variation, as the duration o migration ranges rom less than ive to over thirty one years. About 40% o the individuals have let less than 0 years beore the interview and we have no indication about how common this is among Moroccans, Algerians, and Tunisians who emigrated to Europe and subsequently returned to their respective countries o origin. It is possible that recent migrants are a sel-selected group relative to earlier migrant cohorts. Those who let ater 996 mostly stayed abroad or less than 5 years while those who were given the opportunity to stay longer (those who departed earlier) experienced longer durations in the host country. The overwhelming majority o those who departed beore 975 have stayed abroad more than 26 years, while those who departed between 976 and 985 stay also predominantly more than 26 years (out o a maximum o 29 years). Migrants were relatively young when they arrived in the host country, as the average age at the point o return is 4 years. It is thereore not surprising that a large portion o respondents got married while abroad, though 29% had children upon returning home. The labour market experience o migrants indicates that they were relatively successul in the host country. Most respondents were wage earners (47% ull time and 4% part time) while 5% were sel-employed (this contrasts with the 34% o

respondents in sel-employment ater returning). During the stay in the host country, there is relatively little movement between labour orce status categories. Persistence is highest among wage earners and sel-employed emigrants (86% and 82%, respectively This is not reported in Table 2). Respondents also sent regularly remittances to amily members in the home country, especially in the twelve months prior to returning. The majority o respondents do not appear to seek additional opportunities to re-emigrate, and intend to remain in the home country. Only 20% o those surveyed report to be prepared to re-emigrate. In addition to the data extracted rom the MIREM database, we created two new dependent variables. The irst is the educational investment variable. It is equal to zero i the migrant has no urther education in the host country; i the migrant undertakes vocational training or proessional courses in the host country; and 2 i the migrant undertakes university or university and vocational training (we collapsed these observations in the same category, as those migrants have undertaken university education). The second dependent variable is an activity choice upon returning. It takes the value o zero i the migrant retires or exists the labour orce upon returning; i the migrant is unemployed; 2 i the migrant is a salary earner; and 3 i the migrant is sel-employed. 5 Estimation strategy and empirical analysis Since the investment in human capital in the host country is simultaneously determined with the activity chosen upon returning and the migration o duration, our strategy involves estimating three equations: bt bx ds = dt dx b b s s The choice o activity upon returning, which consists o comparing the indirect utility unctions associated with each regime; and A Ay ( C y / C s λ ) dt = ds + dy A C / C ) A ( C / C ) λ ( t λ λ t λ In the estimations we also need to take into account the act that migrants may sel select into dierent patterns o migration based on their expected activity upon return. In particular, we expect determinants o the duration o migration or individuals who return as sel employed or wage earners to carry dierent weights, especially with 2

respect to those o retirees and unemployed. Unobservables such as individuals ability or motivation are likely to aect both choice o human capital investment and occupation upon return. The latter, in turns, determines the duration o migration ollowing Dustmann and Kirhckamp s (2002) hypothesis. The irst equation estimating the determinants o individuals education choice is estimated by multinomial logit whose results are kept to correct or selection in the choice o activity decision equation. This second equation then serves as a selection equation or the estimation o migration duration. We use a technique analogous to Lee (983) in order to estimate this second stage o the model. More precisely, the selection mechanism is based on a multinomial logit, where the probability that individual i chooses activity j ater returning (the selection variable) is given by: ( j) P LSFi = = e α ' jvi J ' α jvi + e j= where i stands or the individual observation and j or the choice category (0 = out o the labour orce; = unemployed; 2 = salaried work; 3 = sel-employed). The implied regression on the duration o stay in the main country o immigration is given by (see Lee, 983 and Greene, 2003): y y ( ) ' ( H j( α jvi) ) ' H j( α jvi) φ = β x + ( σ ) + η Φ ' j i j j j = β x + σ λ + η ' i i j j j j ( ) where H(.) is the inverse o the standard normal cd evaluated at P ( LFS j) =, φ (.) and Φ (.) are respectively the standard normal pd and cd. This is a two-steps procedure analogous to the Heckman selection model except that the selection mechanism is such that we irst estimate the multinomial logit on the choice o activity ater return. Then we estimate the durations given each choice o activity. This approach yields our sets o estimated parameters or the durations corresponding to each alternative choice (sel-employment, salaried work, unemployed, not in the labour orce), rom which it is possible to identiy the relevant determinants. We only present the results obtained or sel-employment and salaried work. 3

6 Results Table 3 presents the coeicients attached to each variable in determining the type o educational investment in the host country. The reerence group or the regression is migrants who did not pursue additional education in the host country, and who migrated in the period 996-2005. The irst column reports the coeicient obtained rom the regression, while the second and third columns show the corresponding standard deviation and statistical signiicance (t-statistic). The ourth column indicates whether the estimate obtained is statistically signiicantly dierent rom zero. TABLE 3 REGRESSION RESULTS: EDUCATIONAL CHOICE Variable Marg e St. err. b/st.err. P[ Z >z Probability to invest in vocational/ proessional training: Constant -3.679.445-8.480.000 M975.080.437 2.484.03 M985.757.47.607.08 M995.468.372.257.208 AR_WFT.447.429.040.298 AR_SE.629.283 2.22.026 BD_SEC.6.349 3.326.00 BD_TER.774.443 4.003.000 A_MARRY.00.267 3.750.000 Probability to invest in university or university and voc/pro training Constant -5.944.672-8.846.000 M975.345.480.720.47 M985.76.369 2.97.003 M995.48.255.638.0 AR_WFT.733.297 5.837.000 AR_SE.556.273 2.030.042 BD_SEC 4.460.636 7.02.000 BD_TER 5.59.664 8.308.000 A_MARRY.003.27 3.702.000 Number o observations: 796 Pseudo R2:.36 Inormation criterion:.399 Chi sq.: 459.40 logl: -496.92 df: 6 restricted LogL: -726.63 P(Chisq.>val): 0.00000 Percentage correct predictions: 73.86 The results displayed in Table 3 suggest that the choice o activity upon returning inluences the type o educational investment undertaken while abroad. In particular, the probability o undertaking vocational/proessional training in the host country is higher i one becomes sel employed ater return (AR_SE - coeicient: +.629, t = 2.22), but it is not i one becomes a ull-time wage earner (AR_WFT coeicient: 4

+.447, t =.040). In contrast, the probability o undertaking university and university and vocational education is signiicantly higher i one becomes a wage earner upon returning (coeicient:.733, t = 5.837) than is one becomes sel-employed (coeicient: +.556, t = 2.030). Prior education signiicantly aects the probability o undertaking both types o education in the host country, as the coeicients o the variables BD_SEC and BD_TER are positive and statistically signiicantly dierent rom zero, especially in the case o university and university and vocational education. Having migrated prior to 975 raises the probability o undertaking vocational studies but not o undertaking university education. This is instead inluenced by having migrated in the later decade (976-985). These outcomes may relect changes in sel-selection among migrants or changes in migration policy in Europe ater the irst oil shock, in 974, when tighter policies were introduced. Getting married in the host country, a sign o inancial commitment and increased responsibility, substantially raises the probability o undertaking urther education abroad. The marginal eects associated with these results are reported in Table 4. The irst column shows the marginal eect calculated at the sample mean o the relevant variables, while the remaining columns show the corresponding standard deviation and statistical signiicance (t-statistic). The last column represents the marginal eects averaged over individuals in the sample. TABLE 4 REGRESSION RESULTS: EDUCATIONAL CHOICE MARGINAL EFFECTS Variable Marg. E St. Error b/st.err P[ Z ]>z Marg e averaged over individuals Probability to get no urther education in host country Constant 0.903 0.070 2.950 0.000 0.759 M975-0.42 0.070-2.044 0.04-0.0976 M985-0.73 0.069-2.508 0.02-0.43 M995-0.085 0.05 -.674 0.094-0.0663 AR_WFT -0.98 0.064-3.8 0.002-0.828 AR_SE -0.4 0.042-2.708 0.007-0.0886 BD_SEC -0.5 0.06-8.39 0.000-0.4709 BD_TER -0.668 0.080-8.384 0.000-0.6034 A_MARRY -0.92 0.043-4.50 0.000-0.55 Probability to invest in vocational training Constant -0.344 0.05-6.780 0.000-0.63 5

M975 0.24 0.05 2.452 0.04 0.0879 M985 0.073 0.055.328 0.84 0.0375 M995 0.049 0.044.33 0.257 0.0303 AR_WFT 0.025 0.05 0.49 0.623-0.0098 AR_SE 0.066 0.033.999 0.046 0.0409 BD_SEC 0.066 0.042.588 0.2-0.024 BD_TER 0.22 0.054 2.270 0.023 0.0009 A_MARRY 0.04 0.03 3.323 0.00 0.067 Probability to invest in university education (or both) Constant -0.558 0.064-8.677 0.000-0.5977 M975 0.08 0.049 0.372 0.70 0.0097 M985 0.00 0.04 2.44 0.05 0.056 M995 0.036 0.026.365 0.72 0.036 AR_WFT 0.73 0.044 3.94 0.000 0.925 AR_SE 0.048 0.029.647 0.00 0.0477 BD_SEC 0.445 0.05 8.80 0.000 0.4949 BD_TER 0.545 0.069 7.97 0.000 0.6025 A_MARRY 0.088 0.03 2.846 0.004 0.0898 Being sel-employed ater returning home raises the probability o having undertaken vocational or proessional courses in the host country by about 6.6% (t =.999). As noted in the results displayed in Table 3, being a wage earner ater returning home does not appear to inluence the probability o undertaking vocational/proessional training (t =.49). Having undertaken secondary or tertiary education in the home country prior to migration raises substantially the probability o undertaking urther investments in human capital in the host country. This is not surprising as the sel-producing characteristic o education is well known. Having migrated prior to 975 raises the probability o undertaking vocational and proessional training education by 2.4% vis-à-vis not undertaking any urther investment in human capital abroad. With reerence to university education, being a wage earner upon returning raises the probability o undertaking university studies in the host country by a considerable amount (7%, t = 3.94) relative to not undertaking any education. Interestingly, becoming a sel-employed does not have a signiicant eect on such probability (t =.647) Being married raises the probability o undertaking urther education (either vocational training or university) by about 0%. These results suggest a clear division in the type o educational investment undertaken by migrants who then become selemployed and wage earners relative to those not undertaking any additional education 6

abroad. Future sel-employed privilege educational investments in vocational and proessional training while uture wage earners preer to invest in university education. Table 5 presents the estimation results on individuals choice o occupation ater return in their origin country. These results are corrected or the selection induced by the choice o education investment in the host country. Only the marginal eects are displayed with respect to either becoming a wage earner or sel-employed in the host country. The two variables PHKBPR0 and PHKBPR02 are the predicted probabilities o respectively investing in vocational training and undertaking a university degree. TABLE 5 REGRESSION RESULTS: CHOICE OF ACTIVITY MARGINAL EFFECTS Variable Marg e St. err. b/st.err. P[ Z >z Elasticity Probability o being wage earner Constant.449.3 3.406.000 PHKBPR0-8.835.895-9.863.000-3.879 PHKBPR02.885.45 6.080.000.46 AGE_R -.00.003 -.287.773 -.4 ANS_PPI.0.004 2.235.025.477 VISAWK.9.087.368.7.095 VISATOUR.007.065.4.909.009 R_PLACE.3.068.903.057.3 B_STAY2 -.083.06 -.353.76 -.9 R_K.043.062.68.496.06 B_WFT -.006.076 -.084.932 -.02 B_WPT -.042.088 -.48.63 -.026 B_SE -.222.099-2.240.025 -.60 Probability o being sel-employed Constant -.779.65-4.72.000 PHKBPR0 9.96.94 0.582.000.868 PHKBPR02 -.457.56-2.926.003 -.322 AGE_R.004.004.99.32.24 ANS_PPI -.04.005-2.785.005 -.28 VISAWK -.032.098 -.325.745 -.00 VISATOUR.03.075.78.859.007 R_PLACE -.065.08 -.808.49 -.027 B_STAY2.72.070 2.450.04.69 R_K.067.07.945.344.04 B_WFT.083.092.903.366.068 B_WPT.4.07.06.288.304 B_SE.432. 3.877.000 33 7

With reerence to the wage earners, gaining a university degree or a university degree and a vocational/proessional training in the host country substantially raises the probability o being a salary earner in the home country upon returning. Interestingly, having completed a vocational or proessional qualiication abroad decreases the probability o becoming a wage earner on return. It appears that those who undertake such education in the host country come back as sel employed. This result gives some support to the hypothesis that individuals engaging in vocational training in the host country learn a trade that is valued upon their return. The investment in education in the host country appears to enable them to set up a business or pursue the activity they have beore leaving the host country. As regards the marginal eects associated to university degrees, the results suggest that university degrees obtained in the host country are valued to some extent in the home country, as indicated by the higher probabilities to become a wage earner upon return. The marginal eects associated to individuals undertaking any orm o urther education in the host country suggest that individuals do not invest in the host country just or the sake o staying until retirement. Other statistically signiicant determinants that raise the probability o being a salary earner upon returning include going back to the same place rom which the migrant let (3%), while having kinds or having emigrated with a working visa do not appear to inluence the choice o activity upon returning. It is worth noting that while obtaining a university degree in the host country contributes to largely increase the probability that the individual will return as a wage earner, it reduces the probability o becoming sel-employed. According to the results, those who were wage earners last beore return are not signiicantly dierent rom individuals who were out o the labour orce as regards the probability to become a wage earner upon return. I anything, ormer wage earners are less likely to be out o the labour orce upon return than those who already were not participating beore settling back to the origin country. Being sel-employed beore leaving dramatically improves the probability that the individual will keep being sel-employed upon return (by as much as 43%). Besides, as noticed beore, individuals completing vocational training in the host country are more likely to become sel-employed upon return. Taken together, These results 8

suggest that access to vocational and proessional training in the host country can help returning migrants to become sel-employed and entrepreneurs, which in turn is likely to contribute positively to the economic development o the home country. Another actor that positively inluences the probability o becoming sel-employed is the intention to settle back in the origin country or good (7%). Those who, on the contrary, state that they wish to migrate again are more likely to be unemployed upon return. This result is not surprising since sel employment is associated with some sunk costs o establishing a business or building up a customer base. Tables 6a-b report the selection corrected regression results explaining duration (in years) or the two main labour orce statuses: wage earners and sel-employed, respectively. The regressions are obtained taking into account the truncation implied by the act that we do not observe individuals who stayed in the host country. This eect is captured by the lambda in the regression results. TABLE 6A. REGRESSION RESULTS WAGE EARNERS Variable Coe St. err. b/st.err. P[ Z >z Constant 5.592 3.488.603 AGE_M -.370.26-2.93 23.85 A_K 3.003 2.62.50.67 BD_LAND 6.686.944 3.438.303 A_JOBT3.622.993.84.38 MC_FR 3.002.789.678.55 B_WFT 6.34 2.22 2.866.393 B_WPT.847 2.67.692.8 B_SE -.674 3.022 -.22.2 B_REM -.977 2.8 -.448.227 BD_MP -.232 2.035.606.333 Lambda 5.257 2.80 2.42.93 Number o observations: 87 R squared.45 Mean o dep. Var. 9.74 Adj. R squared.295 St dev o dep. Var. 8.80 F(,54) 3.48 Age is the most important actor explaining the duration o migration or wage earners. For each year o age (second row), the duration o emigration drops by 0.37 years (t: 2.93). Owning land in the home country beore leaving raises the length o stay in the host country as does having worked as a wage earner beore leaving. There appears to be positive selection in the sample, as shown by the positive coeicient o 9

the selection regressor (lambda). This suggests that among those who return, wage earners tend to have stayed a longer period o time in the host country. TABLE 6B. REGRESSION RESULTS SELF-EMPLOYED Variable Coe St. err. b/st.err. P[ Z >z Constant.887.659.50 AGE_M.050 -.26.793 28.4 A_K.94 3.855.000.3 BD_LAND.979.824.068.2 A_JOBT3.236 -.62.87.3 MC_FR.85 2.067.038.494 B_WFT.98 3.874.000.544 B_WPT.46.04.298.27 B_SE.846.595.55.06 B_REM.947 3.609.000.255 BD_MP.034 -.60.07.200 Lambda.549 5.75.000.439 Number o observations: 222 R squared.399 Mean o dep. Var. 7.40 Adj. R squared.36 St dev o dep. Var. 6.8 F(,68) 0.8 The results reported in Table 6b suggest that the duration o migration or those who are sel-employed upon returning decreases with age but raises with having kids in the host country, having been a wage earner beore migrating, and having sent remittances regularly home in the last 2 months prior to retuning. Curiously, owning land in the home country beore leaving does not aect the duration o migration (t =.068). 7 Conclusions This paper analyses the determinants o the educational investment, choice o activity upon returning, and duration o emigration or a group o emigrants rom Algeria, Morocco and Tunisia who moved to Europe but returned to their home countries. The results suggest that the type o education plays a signiicant role in the migration decisions o those returning as wage earners or sel-employed. In particular, there is a clear positive relationship between being sel-employed and having previously invested in vocational education in the host country (and a negative relationship i one invested in university education while abroad). There is also a strong positive relationship between investing in university education abroad and becoming a wage earner upon returning. This result suggests that economics development in sending countries may occur through additional access to vocational and proessional 20

education in the host countries. As international migration acilitates skill transers between developed and developing countries, the economic development o the latter will increasingly depend on migrants ability to access educational and vocational training in the developed world in addition to university education. Successul development policies in receiving countries may thereore try to encourage migrants access to vocational and proessional training. Returning migrants with vocational and proessional training tend to be sel-employed ater returning home, and by so doing they contribute to reducing poverty in the host country. 2

Reerences: Black, R., King, R. and Tiemoko, R. (2003). Migration, return and small enterprise development in Ghana: a route out o poverty?, Sussex Centre or Migration Research, WP 9 (March). Dustmann, C. (2003). Return migration, wage dierentials, and the optimal Migration duration, European Economic Review, 47 (2), 353-369. Dustmann, C. and O. Kirchkamp (2002). The optimal migration duration and activity choice ater re-migration, Journal o Development Economics, 67 (2), 35-372. Greene, W. (983). Sample selection bias as a speciication error: comment. Econometrica, 49(3), 795 798. Lee, L.F., 983. Generalized econometric models with selectivity. Econometrica, 5(2), 507 53. McCormick, B. and Wahba, J. (200). Overseas work experience, savings and entrepreneurship amongst return migrants to LDCs, Scottish Journal o Political Economy, 48(2), Special issue, 64-78. 22