Why do asylum seekers choose Hungary as an entry point to the European Union? An econometric analysis 1

Transcription

1 Why do asylum seekers choose Hungary as an entry point to the European Union? An econometric analysis 1 Tamás Barczikay 2 Balázs Szent-Iványi 3 - András TÉTÉNYI 4 Abstract: The number of applications for asylum to the European Union (EU) has been steadily increasing since 2010, having peaked at 1,321,600 asylum applications during the calendar year of In parallel with the growing number of asylum seekers, the entry points to the European Union have also undergone considerable change: whereas in 2010 less than 1% of asylum seekers applied for refugee status in Hungary, by the end of 2015 this had increased to 16.5% (Eurostat 2015). Empirical research on the determinants of asylum seeker behaviour has so far neglected why a growing number of asylum seekers have chosen East Central Europe, and mainly Hungary, as their point of entry to the European Union. The article utilises fixed and random effect regression analysis to discover the elements of the relative attractiveness of Hungary as an entry point. The results from a dyadic panel over the time period from 2002 to 2015 demonstrates, that the recognition rate and relative income difference were the two main variables which have led to the rise of asylum applications in Hungary, despite the low acceptance rates. The article contributes to our understanding of asylum seeker behaviour, particularly in relation those seeking entry in East Central Europe, and therefore assists in providing appropriate policy responses both on the local and on the European level. 1) Introduction Asylum applications in the countries of the European Union (EU) in 2015 have surpassed all previous data: the member states (MS) of the EU have registered 1,321,600 applications (Eurostat 2016). While in the early 2000 s asylum seeker flows were relatively constant (see Table 1), after 2010 a sizeable increase could be observed: the number of registered asylum seekers increased by 5.09 between 2010 and An aspect of this increase is, that it is not only the traditional Western European destination countries which are affected by the increase in asylum applications, but Eastern European countries (notably Hungary where 82.7% of Eastern European MS applications were lodged in 2015) as well (Eurostat 2016). The accession states of 2004 and 2007 are relative newcomers in refugee protection as many had no tradition of it before As with many policy areas (such as the creation of 1 To be presented at the ECPR General Conference in Prague, September Attendance at the conference has been supported by the Hungarian National Bank and Corvinus University of Budapest. 2 Barczikay.tamas@gmail.com; Institute of World Economy, Corvinus University of Budapest, Budapest, Hungary 3 b.szent-ivanyi@aston.ac.uk; School of Languages and Social Sciences, Aston University, Birmigham, United Kingdom 4 andras.tetenyi@uni-corvinus.hu; (corresponding author), Institute of World Economy, Corvinus University of Budapest, Budapest, Hungary 1

2 international development policy) the creation of asylum policy in accession countries as well as the adoption of the acquis communautaire in the field of refugee protection (joining the Common European Asylum System) was a precondition towards the EU accession of 2004 and The details on the establishment of the current Common European Asylum System (CEAS) has been discussed at length elsewhere (for instance Guild 2006, Lavenex 2001a, 2001b, Toshkov and de Haan 2013) therefore this article will only reflect on the major milestones of the process. The Single European Act of 1986 sought to abolish the controls on the movement of goods, services, people and capital by 1992, however the movement of refugees and asylum seekers was not mentioned in the document, it was left to the member states. Interestingly enough the concept of the determination of refugee status by the member state where the asylum applicant was first lodged can be traced back to an internal market logic, similar to that of goods arriving to the EU. The Schengen Implementing Convention of 1990 and the Dublin Convention of 1990 provided the right to the Member States to, on the one hand to pool their responsibility towards asylum seekers as regards to rejection, and on the other hand to determine where the asylum application will be decided. The Maastricht Treaty of 1992 introduced the concept of unfounded application: for instance, when the applicant has passed through a safe third country on their way to lodge an asylum application in the Member State (Guild 2006). This logic has been increasingly used in 2015 and 2016 by Hungarian authorities to reject asylum applications on the grounds that the asylum applicant has passed through Serbia (which Hungary considers a safe third country). The Amsterdam Treaty of 1997 inserted asylum into EU law and specified in Article 73(k) that asylum should be in accordance with the Geneva Convention of 28 July 1951 and the Protocol of 31 January 1967 relating to the status of refugees and other relevant treaties (European Commission 1997). Two years later (1999) the Tampere Conclusions centered around the principles of agreeing on common minimum standards (which this article will be investigating later on in relation to Hungary) and the principle of mutual recognition. Thus in 2003 the Dublin II Regulation introduced the rules for evaluating where an asylum application should be assessed (essentially where the asylum seeker was registered). The Reception Conditions Directive (2003) introduced minimum standards in terms of housing, healthcare and general terms of living in the member state, while the Qualification Directive (2004) collected those cases where an asylum applicant is in need of protection. Lastly, the Asylum Procedures Directive (2005) regulates what minimum standards in terms of procedure should MS pursue (Toshkov and de Haan 2013). All of the afore mentioned directives were implemented by the accession countries, which could have meant that asylum seeker flows would be distributed relatively evenly among the East Central European (ECE) member states, as all of them have the same minimum standards in qualifications, reception conditions and asylum procedures. However according to Table 1, out of 11 ECE Member States Hungary, Poland and Bulgaria countries accounted for 78.6% of registered asylum seekers between 2002 and Furthermore, the data shows that out of 555,075 asylum registrations in ECE between 2002 and 2015, 48.7% were lodged in Hungary. The objective of this article is to investigate what push and pull factors contributed to asylum seekers choosing Hungary as an entry point to the European Union. The article is structured as follows. The following section will review the quantitative literature analyzing asylum seeker destination choices, after which the article will introduce the research design and introduce the results of the estimation. The final section discusses findings and tentative implications on asylum seeker destination choices. 2

3 Table 1: Total asylum applications to East Central European Member States ( ) GEO/TIME European Union (27 421,47 344,80 276,67 234,67 197,41 222,63 225,15 263,83 259,40 309,04 335,29 431,09 626,96 1,321,60 countries) Bulgaria 2,890 1, , ,385 7,145 11,080 20,365 Czech Republic 8,485 11,400 5,300 3,590 2,730 1,585 1,645 1, ,145 1,515 Estonia Croatia 1, Latvia Lithuania Hungary 6,410 2,400 1,600 1,610 2,115 3,420 3,175 4,665 2,095 1,690 2,155 18,895 42, ,135 Poland 5,170 6,810 7,925 5,240 4,225 7,205 8,515 10,590 6,540 6,885 10,750 15,240 8,020 12,190 Romania 1, , ,720 2,510 1,495 1,545 1,260 Slovenia 650 1,050 1,090 1, Slovakia 9,745 10,300 11,395 3,550 2,850 2, Source: Eurostat

4 2) The review of the literature Studying the destination choices of asylum seekers has been relatively well researched in the last twenty years. Böcker and Havinga (1998) have analysed asylum seeker movements between in the then 12 countries of the EU, looking at 44 countries of origin. The authors investigate former colonial ties, differences in the relative wealth of the receiving country, the distribution of different regions of origin of asylum seekers, and acceptance rates within receiving countries. The article does not mention t-statistics, therefore it is not known which variables and effects are significant. Nevertheless, based on the descriptive statistics used by the authors they argue, that asylum seekers favoured the richer and larger countries of the EU, while asylum seekers from former colonies also tended to go to the former mother countries. Thielemann (2003) inspects 20 OECD countries as destination countries between using relative asylum application numbers to make the destination countries comparable over time. He finds that the lagged number of unemployed people and the deterrence rate in the recipient country impacts negatively on asylum seeker numbers, whereas the lagged stock of foreign nationals and relative ODA payments impacts positively on asylum applications. Also interestingly enough the distance from the origin country and GDP growth data in the recipient country does not influence the dependent variable. Neumayer (2004) shows that between the time period of and observing 125 origin countries, economic factors in the Western European destination countries such as unemployment rate and social welfare are not significant and do not impact on asylum destination choices. However richer countries do receive marginally more asylum seekers, even though GDP growth is negatively associated with asylum applications. It is also interesting to note, that countries with right-wing populist parties attract lower number of asylum seekers as opposed to countries with more dominant left-wing parties. A higher positive recognition rate in previous years, also leads to higher asylum share in the present. Also countries which have been former colonies, have a common language and are closer receive more asylum seekers. Building on his previous article Neumayer (2005) analyses 15 Western European destination countries between the years of and 127 countries of origin from the Global South. These countries of origin account for approximately 85 percent of all asylum seekers in Western Europe. The dependent variable is the annual number of asylum seekers from a country of origin, applying for asylum in a Western European destination country. The results show that if GDP per capita increases in the destination country, asylum seeker numbers will decrease. Lower economic growth however increases the number of asylum seekers to that particular destination country. The higher the distance is from the country of origin, it will also negatively influence the application numbers in the destination country. Other significant push factors which increase asylum applications are autocracy in the country of origin, human rights violations in the country of origin, state failure and economic discrimination. Colonial ties, aid, trade and number of Christians in the country of origin fleeing from religious persecution, were all insignificant. Moore and Shellman (2007) have created the probably most comprehensive study up until now about whether refugees are pushed by violence or drawn by economic opportunity. The authors assess the years between for all countries they have been able to obtain data on. The results show that average wages in non-bordering destination countries negatively influence refugee flows, which according to the authors is a result of higher average wage countries are typically located far away from countries of origin making relocation expensive. Interesting to note that democratic institutions in the host country do not influence asylum seekers in their choice. Asylum seekers also avoid destination countries which experience 4

5 genocide and international conflicts. It is also quite unexpected that asylum seekers are negatively influenced by shared language in a non-bordering destination country. Summarizing: asylum seekers are not drawn in large numbers to countries of the Global North despite colonial connections and diasporas. Refugees do relocate first to neighbouring countries, from which they choose the wealthier ones, however they do not necessarily move on the non-neighbouring countries. Rotte and Vogler (1999) analysed migration and asylum seeker flows to Germany the largest recipient country within the European Union. The article concludes that an increase in the income differential between origin and destination country leads to the increase in asylum seekers; the increase in political terror in the sending country also increases asylum seeker numbers while simultaneously political rights and civil liberties have no impact on it. 3) Research methodology The article answers the question why asylum seekers have chosen Hungary as an entry point to the European Union and what significant factors contribute to this choice. The reason why the article focuses on Hungary is, that it has accounted for almost half of registered asylum seekers in the ECE region between 2002 and Due to the fact that the ECE Member States of the European Union all use the same minimum standards of the Common European Asylum System, a rationally thinking asylum seeker should be choosing the country with the most net benefits (Neumayer 2004) and according to the data this country is Hungary. The article conceptualizes asylum seeker flows as aggregate observable flows of individuals (Neumayer 2004, 2005; Moore and Shellman 2004, 2007; Rotte and Vogler 1999 and Thielemann 2003). The dependent and independent variables of the article are collected in Table 2. The dependent variable The dependent variable (REGISTRATIONS) used in the article is the stock of the raw number of registered asylum seekers (Schmeidl 1997:291, Neumayer 2005:396, Moore and Shellman 2007:820) from country of origin to the destination country (in this case Hungary). Even though we have annual data on the registered number of asylum seekers from Eurostat (2016) between 1998 and 2015, the data available on Eurostat (2016) provides a by-origin-country asylum acceptance and rejection rate only from 2002 onwards. Hence the article uses the registered number of asylum seekers by country of origin between 2002 and The article has selected 24 countries of origin out of the approximately available 130 Global South countries, by checking the annual asylum registrations by country of origin. In order to make the findings of the article more robust the authors have included in their analysis only those countries or origin from which Hungary has received at least 300 asylum registrations over the period of 14 years. These 24 countries however account for 95.05% of registered asylum claims. 5

6 Table 2: Variables used All data between Source Units Literature Dependant variable To Hungary from Country in a defined Year Eurostat No. of registrations Schmeidl 1997:291; Moore and Shellman 2007:820; Neumayer 2005: 396 Independent variables Freedom rating Civil liberties rating Political rights rating Freedom House Freedom House Freedom House PF, NF, F Moore and Shellman 2007:821 1 to 7 Neumayer (2005:397); Rotte and Vogler (1999:32) 1 to 7 Neumayer (2005:397) Purdue political terror scale 5 Purdue school 1 to 5 Moore and Shellman 2007:821); Neumayer (2005: 53), Neumayer (2005:397); Rotte and Vogler (1999:32) Decisions made Eurostat No. of decisions Recognition rate Eurostat No. of positive decisions Böcker and Havinga (1998:250), Neumayer (2004:168; ) 5 PTS eventually was not used due to lack of data for 2015 at the time of writing 6

7 Aid 6 OECD ODA in million USD (current) Trade Intracen Volume of trade=export+import; USD (current) Neumayer (2005:397); Rotte and Vogler (1999:32); Rotte and Vogler (1999:32); Berthélemy and Beuran and Maurel (2009:1593) Neumayer (2005:397); Rotte and Vogler (1999:32); Rotte and Vogler (1999:32); Berthélemy and Beuran and Maurel (2009:1593) Income differential between sending and recipient country World Bank, UN ln(gnp per capita Hungary/GNP per sending country) (current USD) Rotte and Vogler (1999:32) Unemployment in destination country KSH % Thielemann (2003:19); Neumayer (2004:167; 2005: 52); (Moore and Shellman 2007:821) Average social expenditure in destination country 7 KSH In Hungarian Forint Neumayer (2004:169) Distance In Kms Thielemann (2003:20); Neumayer (2004:168); Neumayer (ISQ2005:397); Rotte and Vogler (1999:32); Berthélemy (2009:1593) 6 Aid eventually was not used due to lack of data for 2015 at the time of writing 7 Social expenditure eventually was not used due to lack of data for 2015 at the time of writing 7

8 The independent variables We grouped the independent variables into two larger groups: origin specific (which push asylum seekers from their home country) and destination specific (which pulls asylum seekers to Hungary). In order to capture what origin specific political conditions have led people to flee their home countries, we took the freedom (PARTFREE and FREE) (Moore and Shellman 2007:821), civil liberties (CIVLIB) (Neumayer 2005:397, Rotte and Vogler 1999:32) and the political rights rating (POLRIGHT) (Neumayer 2005:397) from Freedom House in addition to the Purdue Political Terror Scale (PTS) (Moore and Shellman 2007:821; Neumayer 2005: 53; Neumayer 2005:397; Rotte and Vogler 1999:32) index. In initial models the PTS index was used as an independent variable, however the effect turned out to be insignificant in the majority of cases, in addition to data not being available for was the year in which the European Union and Hungary recorded the highest numbers of asylum registrations, therefore we decided in order to capture that effect, not to use the PTS index in the final models. We also checked whether the location of the destination country played a role in the decision to enter the EU through Hungary by measuring the distance between capitals of the destination country and the country of origin (DISTANCE) (Thielemann 2003:20; Neumayer 2004:168; Neumayer 2005:397; Rotte-Vogler 1999:32; Berthélemy 2009:1593). For the destination specific independent variables and in order to measure personal links and contacts between the origin and destination countries, which might increase the flow of migration, we used aid and trade data based on (Neumayer 2005:397; Rotte and Vogler 1999:32; Berthélemy and Beuran and Maurel 2009:1593). For the AID data we used the absolute numbers of Official Development Assistance (ODA) Hungary was providing in a given year to the country of origin as provided by the OECD (Organisation for Economic Cooperation and Development 2016). Unfortunately, similar to the PTS index, there was no available data for 2015 at the time of writing of the article, therefore in order to capture the 2015 asylum registration effects we once again left out this variable from later models. In initial models however AID was insignificant. For the TRADE variable the article used total value of trade in a given year (that is the sum of exports and imports) between the country of origin and Hungary. The data was measured in absolute figures and used the data set of the International Trade Centre (2016). To capture the relative differences in terms of economic development between origin and destination country, the article uses the logarithmic version of the income differential (L_INCOMEDIFF) which makes the GNP of the origin and destination countries comparable (Rotter and Vogler 1999:32). For calculating the income differential, the data was taken from the World Bank (2016), apart from the GNP of Syria which was taken from the UN (2016) for reasons of availability. The general idea behind using the income differential as a pull indicator is, that the higher the income difference is between the country of origin and destination, the more likely an asylum seeker might decide to go there if they are not only seeking safe haven, but are thinking about long term economic prospects. To check whether asylum seekers were motivated in their choices by economic factors in the destination country in addition to the income differential, the article checked the unemployment rate UNEMP in the destination country (Thielemann 2003:19; Neumayer 2004:167, 2005: 52; Moore-Shellman 2007:821) and the annual average social welfare expenditure WELFEXP (Neumayer 2004:169) in Hungary. Both datasets were taken from the Hungarian Central Statistical Office (KSH 2016), however for welfare expenditure there was no data for 2015 and was insignificant in early models, therefore was not used in later models similar to PTS and AID. 8

9 In order to investigate how the decisions of the Hungarian authorities on whether to grant or not to grant protection have impacted on asylum seeker flows, the article introduced the variable: recognition rates (RECOGNITION) (Böcker and Havinga 1998:250, Neumayer 2004:168). We would expect asylum seeker numbers to increase if asylum recognition is high and decrease if rejection is high. Hungary has also been quoted to be a country of transit, or point of entry to the European Union for asylum seekers, which means that they enter Hungary, but go on soon enough to Austria, Germany or other Western European countries (Guardian 2015). According to the Dublin 3 legislation, those asylum seekers who have lodged their application in Hungary and have gone on to other EU countries, can be returned anytime to Hungary for their application process to be heard and a decision to be made on their asylum claim recognition, therefore we would expect asylum seekers to choose relatively wisely where they register. To test, whether Hungary is really a country of transit we included the variable DECISIONS, which is the annual number of final decisions on asylum applications made by Hungarian officials taken from Eurostat (2016). Those cases where a decision is made are typically cases where the asylum seeker went through the whole process of qualifying for asylum and did not leave at one point for Western Europe. If that would have been the case the process would have been discontinued. We theorize that if DECISIONS is positively associated with REGISTRATIONS than at least some asylum seekers do choose to come to Hungary and not only as a point of entry to the EU. However, if the association is negative, that would mean that number of completed application decisions, have been falling while asylum registrations increasing. This could signal to Hungarian and EU authorities that Hungary is only a country of transit. Depending on the model used, we have between 250 to 284 observations. As we have explained earlier, we have dropped the PTS, AID and WELFEXP variables due to lack of data for 2015 the year we wanted to include in our analysis. The number of entries should be enough as according to Szent-Iványi (2012) most regression textbooks agree that there should be at least ten times as many observations as variables, which the article fulfills. The descriptive statistics for the final model are in Table 3. Table 3: Descriptive statistics Variable Obs Mean Std. Dev. Min Max REGISTRATI~S PARTFREE FREE CIVLIB POLRIGHT DECISIONS RECOGNITION TRADE e+07 L_INCOMEDIFF UNEMP DISTANCE ) Research Design The article includes a level level and a log level model. The first three columns contain the level level, whereas the second three columns the log level models (Table 4 and 5). The reason for adopting a log level model is that this way REGISTRATIONS will follow a more normal distribution as opposed to the level level model. We also used the White test to have heteroscedasticity-robust standard errors. 9

10 Table 4, Model 1: Estimates for REGISTRATIONS as the dependent variable and not including DECISIONS ( ) Pooled OLS REGIST~S RE REGIST~S FE REGIST~S Pooled OLS L_ REGIST~S RE L_ REGIST~S FE L_ REGIST~S PARTFREE (726.5) (751.4) (2043.6) (0.449) (0.801) (1.071) FREE (956.5) (876.0) (1334.3) (0.729) (1.101) (1.041) CIVLIB (255.8) (272.4) (319.7) (0.175) (0.309) (0.367) POLRIGHT (158.5) (79.87) (451.9) (0.150) (0.125) (0.194) TRADE e-08 *** *** -7.87e-08 ( ) ( ) ( ) (3.58e-08) (4.25e-08) ( ) L_INCOME~F ** (2040.2) (2701.7) (3445.9) (0.904) (1.006) (1.337) L.L_INCOMED IFF * ** (2050.4) (2747.6) (2980.0) (0.915) (1.064) (1.201) UNEMP * * *** *** ( ) ( ) (107.7) (3.737) (3.182) (0.112) L.UNEMP * *** *** *** (7530.4) (7700.3) (173.5) (1.583) (1.301) (0.121) DISTANCE (0.0468) (0.0301) (.) ( ) ( ) (.) RECOG~N * *** ** *** *** *** (12.95) (5.212) (8.416) ( ) ( ) ( ) L.RECOG~N *** *** *** *** *** (7.727) (1.002) (0.858) ( ) ( ) ( ) Syria (367.0) (320.8) (.) (0.448) (0.814) (.) Afghanistan *** * (3667.3) (647.0) (.) (0.589) (0.750) (.) Iraq *** (486.4) (150.7) (.) (0.308) (0.347) (.) Pakistan *** *** *** 0 (840.6) (147.4) (.) (0.421) (0.339) (.) Constant * * *** *** * ( ) ( ) (3480.7) (12.88) (10.94) (4.165) Observation s R Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01 Model 1 in Table 4 contains our first estimates for analyzing the reasons why asylum seekers register and apply for asylum status in Hungary. We have used time dummies to control for the effects of individual years on registration numbers and for the pooled and random effect log - level models they were significant at the 1% level. We also analysed the effects of individual countries of origin with a large number of registered asylum seekers and found that only asylum seekers from Pakistan (19,630 registered asylum seekers between ) influenced registrations significantly. That is, if on average there is one additional asylum seeker coming from Pakistan, it will increase the registered number of asylum seekers by 127%. (In the pooled effect model Afghanistan was marginally significant at 10%). We have included a one-year lag for three independent variables (L_INCOMEDIFF, UNEMP and RECOGNITION) in order to test if the change in economic variables and positive asylum recognitions in the past will influence registrations in the present. The conclusions from the log level models are the 10

11 following: recognition numbers are significant at 1% in all three models (pooled, random effect (RE) and fixed effect (FE)) and with roughly the same results. If one additional asylum seeker receives asylum, the number of registrations will increase by about 2.48% (FE). Interestingly enough the results also show a negative, but significant connection between the one-year lag of RECOGNITION and registrations. A tentative conclusion here might be, that those asylum seekers who have received refugee status previously, when talking to their friends and family in the country of origin, might dissuade them to come to, or through Hungary, because of difficulties in the registration process (Tétényi 2016). These results are visible in all three loglevel models. We have received an R , which means that the model has sufficient explanatory power. Additional results vary between the pooled, random effect and fixed effect models, but nevertheless we have decided to include all of them. Using the pooled model depends on the effects of the dependent and independent variables being constant in time and the pooled groups being relatively similar or homogenous. However, people may decide to leave their country of origin for different reasons: these may include different intensities of war or internal conflict breaking out at different times, different culture, links to family abroad, knowledge of languages and different local religions, making the pooled groups heterogenous. Therefore, we will be also using the fixed effects (FE) model in order to handle the unobserved culture specific fixed effects of the countries of origin, which were not measured in the independent variables. The usage of the random effects (RE) model depends on the fact that the unobserved effects of the country of origin are uncorrelated with all of the independent variables. Since we measure at least 9 independent variables with an R 2 of it seems unlikely that the unobserved effect is uncorrelated with the explanatory variables. We have also conducted the Hausman test to determine the usage of the RE or FE models and for both the first and second models it rejected using the RE model. In addition to the RECOGNITION being significant in the FE model, the lagged log INCOMEDIFF is also significant at 5%, meaning that if the income differential between countries of destination and origin increases by 1%, registrations in Hungary will be increasing by 2.538%. The result for the lagged unemployment is somewhat perplexing as it would suggest that if unemployment has increased in the destination country, more asylum registrations will occur. It may be the case, that the global financial crisis was influencing (as seen in the INCOMEDIFF variable) people to migrate, thus the increase in unemployment in the destination country was lower than in the country of origin. In the second model (Table 5) we included DECISIONS as an additional independent variable, which we believe indicates that the asylum seeker decided to come to Hungary not only as a point of entry to the EU but also as a destination country. The explanatory power (R 2 ) of the second model grew quite significantly to RECOGNITION which was significant in the pooled, FE and RE models is now significant at only the 10% level for the FE model, however L_RECOGNITION stayed significant at the 1% level and stayed negative as in Model 1. The result for DECISIONS shows that if there is one additional decision made (that is one more person decides to stay in Hungary until the asylum procedure is completed) the registrations into the country will increase by 0.579%. This seems quite logical and shows that Hungary is a transit country. If the decisions would be made in the same year as the asylum application, and Hungary would not be a country of transit the relationship between DECISIONS and REGISTRATIONS would probably be more equal, that is a one person increase in DECISIONS would result closer to a one person increase in REGISTRATIONS. In addition, L.L_INCOMEDIFF is still significant, although only on the 10% level for the FE model. 11

12 Table 5: Estimates for REGISTRATIONS as the dependent variable with DECISIONS included ( ) Pooled OLS REGIST~S RE REGIST~S FE REGIST~S Pooled OLS L_ REGIST~S RE L_ REGIST~S FE L_ REGIST~S PARTFREE (833.7) (810.1) (1897.3) (0.381) (0.709) (0.692) FREE (1186.2) (1112.2) (1415.6) (0.613) (1.049) (0.816) CIVLIB (233.4) (198.3) (300.3) (0.154) (0.224) (0.297) POLRIGHT (157.6) (119.6) (330.0) (0.133) (0.191) (0.145) TRADE e-08 *** -8.95e-08 *** -6.16e-08 ( ) ( ) ( ) (2.92e-08) (2.93e-08) ( ) L_INCOMED~F ** * (2046.5) (2723.2) (3092.3) (0.965) (1.256) (1.174) L.L_INCOMED IFF * * (2038.2) (2732.1) (3010.2) (0.971) (1.247) (1.142) UNEMP * * *** *** ( ) ( ) (153.4) (3.214) (2.774) (0.110) L.UNEMP * * * *** *** *** (7205.1) (7319.2) (223.0) (1.365) (1.152) (0.105) DISTANCE (0.0497) (0.0432) (.) ( ) ( ) (.) RECOG~N ** * (13.63) (17.28) (21.75) ( ) ( ) ( ) L.RECOG~N *** *** *** ** ** (10.23) (3.342) (3.801) ( ) ( ) ( ) DECISIONS * * *** *** *** (3.935) (4.161) (4.511) ( ) ( ) ( ) L.DECISIONS (1.214) (1.076) (1.323) ( ) ( ) ( ) Syria (455.0) (593.0) (.) (0.447) (0.718) (.) Afghanistan *** ** * 0 (3594.6) (782.8) (.) (0.508) (0.632) (.) Iraq ** (537.4) (288.8) (.) (0.287) (0.245) (.) Pakistan ** ** *** 0 (775.6) (334.9) (.) (0.314) (0.270) (.) Constant * * *** *** * ( ) ( ) (2767.2) (11.09) (9.571) (3.114) Observation s R Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < ) Results, limitations and implications Our analysis in Model 1 has confirmed that even though the acceptance rate of refugees in Hungary was relatively low, it was still one of the main factors which influenced asylum seekers to register there. The explanatory power of the model can be considered to be quite high at 0.489, but there are still quite a large number of effects which our model was not able to capture. The positive relationship between lagged unemployment in the destination country and the number of registrations is a good example, but also the negative relation between lagged recognition and registrations. Therefore, some additional push factors for the country of origin such as unemployment, or the number of terrorist attacks could be included at a later stage to 12

13 make the findings of the model more accurate. One of the problems we have been facing in the data collection stage of the research, that there was a lack of available data for 2015, which effect we wanted to capture. Therefore, the model did not include the political terror scale of a country of origin, which coupled with the number of terrorist attacks might have provided a more robust result. Model 2 increased the explanatory power of our estimation to by including the number of decisions made by Hungarian authorities. The positive connection between decisions and registration shows that Hungary is not only a country of transit, but a country of destination as well. Recognition remained significant in Model 2, but the lagged recognition also shows that maybe with hindsight, asylum seekers may want to choose a different country of destination. The results of the model are consistent with that of the literature. Rotte and Vogler (1999), who used an estimation technique for a single destination country also received the result that log income differential is significant, whereas the Freedom House indices are insignificant. The importance of recognition rates was also significant in Neumayer (2004), same as the per capita GDP in the destination country. We are also aware that 14 years and one destination country (even though it is the largest recipient of asylum seekers in East Central Europe) may not truly capture the phenomena of an increasing activity from the side of asylum seekers in entering the European Union through its eastern borders. However, as our conclusions are in line with the limited evidence available on the choices of asylum seekers, we believe that the shortcomings of the models do not invalidate the results. 6) References Berthélemy, J.-C. and Beuran, M. and Maurel, M. (2009), Aid and Migration: Substitutes or Complements?, World Development, 37(10): Böcker, A. and Havinga, T. (1998), Asylum Applications in the European Union: Patterns and Trends and the Effects of Policy Measures, Journal of Refugee Studies, Vol. 11, No. 3: Eurostat (2016), Asylum and managed migration database, date of download: 23 rd of May Freedom House (2016), Report freedom in the world. date of download: 5 th of June Guardian (2015), First refugees arrive from Hungary after Austria and Germany open borders, date of download: 29 th of August Guild, E. (2006), The Europeanisation of Europe's Asylum Policy, International Journal of Refugee Law, 18 (3-4): International Trade Centre (2016), International Trade Statistics, Date of download: 1 st of June KSH (2016), Hungarian Central Statistical Office, Date of download: 13 th of June Lavenex, S. (2001a): The Europeanization of Refugee Policies: Normative Challenges and Institutional Legacies. Journal of Common Market Studies, 39(5): Lavenex, S. (2001b): Migration and the EU's new eastern border:between realism and liberalism. Journal of European Public Policy, 8(1): Moore, W. and Shellman, S. (2004), Fear of Persecution: Forced Migration, , The Journal of Conflict Resolution, 48(5):

14 Moore, W. and Shellman, S. (2007), Whither Will They Go? A Global Study of Refugees Destinations, , International Studies Quarterly, 2007(51): Neumayer, E. (2004), Asylum Destination Choice - What Makes Some West European Countries More Attractive Than Others?, European Union Politics, 5 (2): Neumayer, E. (2005), Bogus Refugees? The Determinants of Asylum Migration to Western Europe, International Studies Quarterly, 2005(49): OECD (2016), Aid (ODA) disbursements to countries and regions (DAC2a), Date of download: 8 th of June Purdue Political Terror Scale (2016), The Political Terror Scale, Date of download: 9 th of June Rotte, R. and Vogler, M. (1999), The Effects of Development on Migration: Theoretical Issues and New Empirical Evidence, IZA Dp No. 46. Schmeidl, S. (1997), Exploring the Causes of Forced Migration: A Pooled Time-Series Analysis, , Social Sciences Quarterly, 78(2): Szent-Iványi, B. (2012), Aid allocation of the emerging Central and Eastern European donors, Journal of International Relations and Development, 15(1): Tétényi, A. (2016), The Europeanization of Asylum Policies in Hungary, Presentation at the ISA-CEEISA Conference in Ljubljana, 23 rd to 25 th of June Thielemann, E. (2003), Does Policy Matter? On Governments Attempts to Control Unwanted Migration, EI Working paper Toshkov, D. de Haan L. (2013), The Europeanization of asylum policy: an assessment of the EU impact on asylum applications and recognitions rates, Journal of European Public Policy, 20(5): UN (2016), UNDATA, Date of download: 10 th of June World Bank (2016), World Bank Development Data, Date of download: 10 th of June