Human rights violations and the gender gap in asylum recognition rates

ABSTRACT

During the recent European refugee crisis, more than two million people applied for asylum in Germany. Female applicants stand a higher chance to gain protection than male applicants. Whilst small on average, this gender gap in asylum recognition rates varies strongly across countries of origin which has remained little noticed. We analyse the gender gap in asylum recognition rates for the 56 major countries of origin of refugees whose asylum claims are decided on over the period 2012–2018. We show that both general or gender-unspecific human rights abuses and gender-specific human rights violations are associated with cross-country variation in the gender gap in recognition rates – but in opposite ways. Specifically, we find that the gender gap is lower for refugees coming from countries with worse general human rights abuses in the form of political terror perpetrated by state agents but is higher for refugees from countries with a higher prevalence of female genital mutilation and a higher prevalence of child marriage.

KEYWORDS:

• Asylum
• gender
• human rights
• refugees

Introduction

In absolute numbers, Germany was by far the main target country for asylum-seekers during the recent though still ongoing ‘European refugee crisis’, which Scipioni (2018, p. 1358) submits ‘was brought about by a combination of weak monitoring, lack of policy harmonization, low solidarity, and absence of central institutions’ within the European Union (EU).¹ Between 2012 and 2018, the last year for which data are available, of the 5.3 million refugees applying for asylum in one of the countries of the European Union, more than two million did so in Germany with more than 1.2 million of them arriving in 2015 and 2016 alone.

It has remained little noticed that recognition rates for the protection of refugees in Germany differ by gender. Over the period from 2012 to 2018, women’s overall rate for receiving a recognized protection status was 56.6 per cent, whereas that of men was 53.3 per cent. Accordingly, the overall gender gap stood at a rather unremarkable 3.3 percentage points. However, this overall figure hides that the gender gap in recognition rates varies strongly across asylum-seekers’ countries of origin. While women benefit from a higher recognition rate than men in all countries bar Myanmar and Zimbabwe, the gender gap in recognition rates varies from 16.9 percentage points in favour of men in the case of Myanmar to a staggering 57.5 percentage points in favour of women for refugees from Guinea.² Five more countries have a gender gap larger than 30 percentage points: Burkina Faso, Gambia, Jordan, Mali and Sierra Leone. At the other extreme, the gender gap is smaller than 5 percentage points for 23 out of the 56 countries in our sample. Syria, by far the largest single source of refugees, Eritrea, Russia, Ukraine, Vietnam and all former Yugoslavian states fall into this category. Syria is a special case since virtually all Syrian refugees applying for protection received a positive decision. As a consequence of this exceptionally high recognition rate, the gender gap cannot be large for Syrian refugees.

This article explores whether gender-specific and gender-unspecific, general human rights violations in the countries of origin can explain a substantial part of the variation in the gender gap in asylum recognition rates at the aggregate level across countries of origin. In many countries, women do not enjoy equal rights compared to men and face more discrimination and gender-specific human rights violations. German law explicitly accepts certain forms of oppression and persecution that predominantly affect women as valid grounds for protection, e.g., genital mutilation, forced and child marriage (BAMF, 2010; Ellinger, 2001; Liebner, 2017, p. 68f.). While these gender-specific human rights violations affect women more adversely than men, the opposite can be expected for gender-unspecific human rights violations. The typical presumption is that men on average are more actively involved in public political opposition to a regime that abuses human rights and persecutes opponents (Coffé & Dilli, 2015; Crawley, 2000; Desposato & Norrander, 2009; Isaksson et al., 2014). If this holds, then men are more likely to be persecuted by these regimes. The prediction from human rights violations on the gender gap in asylum recognition rates therefore needs to differentiate between gender-specific and gender-unspecific or general human rights violations: on the one hand, the gender gap in recognition rates is positively associated with the degree to which women’s rights in the refugees’ country of origin are restricted or violated with respect to the specific gender-related potential grounds for protection. On the other hand, the gender gap in recognition rates is negatively associated with general human rights violations in the country of origin.

We find evidence consistent with these predictions in our analysis of the gender gap in recognition rates across the 56 major countries of origin over the period 2012–2018 in Germany. Our study at the aggregate cross-country level provides an important and novel empirical insight that cannot be gained from the few existing studies at the individual level reporting a higher chance for women to gain protection (Ecker et al., 2020; Emeriau, 2019; Holzer et al., 2000; Mascini & Van Bochove, 2009). Firstly, without exception any recognized asylum status is due to action or inaction by governments and their agents in the refugees’ countries of origin who either actively violate human rights or fail to protect individuals from such violations. Thus, even if, for example, women flee a country because of the threat of forced marriage or of female genital mutilation, the grounds for protection are that her government does not protect her from this violation of her gender-specific rights. Secondly, whilst by law individual asylum claims are assessed based on their individual merit, individual cases in micro-level analysis should not be considered as being independent from each other where asylum-seekers come from the same country of origin. The plausibility of any stated claim for individual protection is not independent from what deciders hear from other claimants or from what they know of and are briefed about government action or inaction with regards to human rights violations in countries of origin. These two features of the asylum recognition process are best studied in a macro-level analysis that captures the extent to which governments actively or passively violate human rights, which means there is value in such macro-level analysis complementing micro-level analyses with additional insights.

Human rights violations and their impact on the gender gap in asylum recognition

In this section, we look at the impact of human rights violations on grounds for granting asylum and the gender gap in asylum recognition rates. We distinguish between general or gender-unspecific human rights violations on the one hand and gender-specific human rights violations on the other. Gender-specific human rights violations, which attracted increasing attention by researchers from the 1990s onwards (Kelly, 1993), typically affect women much more than men and should therefore make it easier for women from countries that violate these rights to make a persuasive claim for protection (Crawley, 2000). At first glance, gender-specific grounds for protection provide the most straightforward explanation for a gender gap in recognition rates. Women share all of the potential grounds that normally result in being granted protection as a result of general human rights violations and then there are additional gender-specific grounds that affect women exclusively or predominantly.

Yet, the mere fact that the number of ways in which women can be persecuted exceeds the number of ways in which men can be persecuted does not logically imply that a larger share of women than men are in fact persecuted or that women are more intensely persecuted than men and therefore stand a higher chance of receiving recognized protection status. One also needs to take into account whether general or gender-unspecific human rights violations affect men more adversely than women, rendering it easier for men from countries that violate these rights to make a persuasive claim for protection. In this section, we will argue that this is indeed the case and as a consequence we expect human rights violations to have two effects that run in opposite directions. We predict that a larger gender gap in recognition rates is associated with more gender-specific human rights violations but a smaller gender gap is associated with more gender-unspecific human rights violations.

Human rights violations provide the classical ground for granting asylum (Neumayer, 2005). Article 14 of the Universal Declaration of Human Rights states that ‘everyone has the right to seek and to enjoy in other countries asylum from persecution.’ Persecution of citizens by their own government, thus, has been the prime internationally codified ground for asylum. In Germany, the asylum recognition practice has been shaped not only by international law and conventions like the Geneva Refugee Convention, but also by the interpretation of article 16a of the German constitution on the Right to Asylum by the powerful Federal Constitutional Court. According to the Court’s rulings, a person can experience political persecution in the sense of article 16a not only if he or she experiences violations of his or her human rights by the state itself but also by third persons if these violations can be indirectly attributed to actions or inactions of the government and agents of the state. Yet further grounds for protection include, among other things, the unwillingness or inability of the government to keep its citizens safe from arbitrary force within an international or domestic armed conflict.

Accordingly, like many other countries, Germany provides a tripartite level of protection for those people seeking protection from persecution whose claims are assessed as having merit (BAMF, 2019). The strongest form of protection is based on asylum protection, which requires that the person granted asylum was persecuted by state actors on the basis of their race, nationality, political orientation, religious conviction or belonging to a particular social group (including groups based on sexual orientation) and continues to be threatened with violations of their human rights if they were to return to their country of origin. If asylum is denied, protection can still be granted in the form of refugee protection under the Geneva Refugee Convention. The grounds for granting protection are the same as under asylum, except persecution can come from non-state actors, too. Failing that, subsidiary protection can still be granted for a person who can persuasively demonstrate that returning to their country of origin would result in significant personal harm to them in the form of the death penalty, torture, inhuman or degrading treatment or punishment or fear of life or a serious individual threat to the life or integrity of the person, including rape, as a result of arbitrary force within an international or domestic armed conflict. Practically all Syrian refugees coming to Germany have been granted subsidiary protection status. Deciders in the Federal Office for Migration and Refugees are specially trained and acquire knowledge not just of the German asylum, refugee and immigration laws but also of the human rights situation in the countries of origin of refugees to help them in their decision-making. These deciders assess the merit of the asylum claim and the plausibility of the accompanying narrative relative to what the law stipulates as valid grounds for protection.

The typical presumption is that men on average are more actively involved in public political opposition to a regime that violates human rights and persecutes opponents (Crawley, 2000). Whilst women equally share with men race, nationality, political orientation, religious conviction or belonging to a particular social group on which claims of persecution resulting in protection can be based, men arguably face more political repression and persecution than women simply because they are more likely to openly and publicly oppose repressive and human rights violating regimes due to their higher political participation rate (Coffé & Dilli, 2015; Desposato & Norrander, 2009; Isaksson et al., 2014). Even feminist critics who point out that women oppose human rights abusing regimes in ways that are less publicly visible admit that asylum law typically ‘privileges male-dominated “public” activities over the activities of women, which take place largely in the “private” sphere’ (Crawley, 2000, p. 17). If this holds, men are relatively more likely than women to make a valid protection claim if they come from countries of origin abusing general, as opposed to gender-specific, human rights. This prediction is based on the assumption that men are more likely to actively and openly oppose human rights-violating regimes. If this assumption is valid, men will be disproportionately affected by persecution. Note, however, that if this logic is correct or at least plausibe, it becomes relatively easier for men to tell a persuasive narrative of persecution even if they themselves were not persecuted. It is in the nature of the asylum process that many of the claims cannot be proven or documented such that the validity of any claim is subjective and will not be independent of what deciders know about the general state of human rights violations in the applicant’s country of origin. All other things equal, this results in the empirically testable prediction that the gender gap in recognition rates is smaller for countries that have a worse general human rights record.

While important, general human rights violations are not the only reason to flee one’s country of origin. There are also gender-specific human rights violations and, as mentioned in the introduction, Germany’s refugee protection system explicitly allows for a large number of gender-specific grounds for providing asylum that predominantly affect women (BAMF, 2010; Ellinger, 2001; Liebner, 2017, p. 68f.). These are the existence of forced marriage and the absence of legal protection from forced marriage; child marriage; lack of legal protection from domestic violence; threat of ‘honour’ killing or dowry killing; rape or sexual abuse if resulting in expulsion from the household or threat to life or imprisonment; behavioural, clothing and other social or cultural norms if violating these norms results in persecution; forced prostitution; enslavement and human trafficking; genital mutilation; and persecution because of a person’s sexual orientation, which may however potentially affect men more than women if men on average show their non-heterosexual orientation more openly.

Women from countries that oppress and persecute women in these gender-specific dimensions of human rights will find it comparatively easier than men to produce a persuasive narrative of gender-specific rights violations and therefore make a valid protection claim even if they themselves are not in truth affected by these abuses. This is not to belittle the fact that those whose rights have in fact been violated will not always find it easy to substantiate gender-specific oppression and the risk of persecution at home. Ellinger (2001, p. 17) asserts that it is rare that female asylum-seekers invoke gender-specific grounds for protection. Yet, one of the specially-commissioned case office workers at the Federal Office strongly denied in a public speech that these gender-specific grounds for protection are considered only rarely by deciders (Liebner, 2017, p. 69). These specially-commissioned case officers receive additional training to deal with unaccompanied minors, victims of torture, victims of trauma and persons persecuted because of their gender, as well as victims of trafficking in human beings (BAMF, 2019). All female asylum-seekers can request a female decider if they have gender-specific grounds for seeking protection and if they feel more comfortable for their claim to be heard by a woman. Female asylum-seekers are made aware of this provision (Liebner, 2017, p. 61). More generally, Germany’s asylum policy has been characterized as ‘moderately women-friendly’ (Emmenegger & Stigwall, 2019, p. 1304).

In sum, we argue that general or gender-unspecific human rights violations have a different and in fact opposite effect on the gender gap in asylum recognition rates to gender-specific human rights violations. The empirically testable predictions are that while general human rights abuses and repression result in a lower gender gap in asylum recognition rates, gender-specific oppression increases the gender gap in asylum recognition rates. Note that even if we find evidence that the gender gap in recognition rates varies systematically with gender-unspecific and gender-specific human rights violations, this may still represent what is known as ‘statistical discrimination’ (Arrow, 1973; Phelps, 1972). Deciders may use observable characteristics such as the gender of refugees as a proxy for otherwise unobservable characteristics that are relevant to their decision-making. Deciders who know or believe that women suffer on average more from gender-specific human rights violations and that men suffer on average more from gender-unspecific human rights violations may let this belief about aggregate group characteristics influence their evaluation of individual asylum claims by men and women without behaving irrationally or without being personally prejudiced against either men or women.

Alternative explanations for the gender gap in asylum recognition rates

Human rights violations are of course not the only factor potentially impacting the gender gap in asylum recognition rates across countries of origin. We discuss two alternative or additional explanations to motivate the selection of control variables in the empirical model specification.

A general preference for women

The first additional explanation of the gender gap in asylum recognition rates can be motivated by recent experimental survey research. Bansak et al. (2016) have conducted a conjoint experiment using vignettes of hypothetical cases of asylum-seekers to assess attitudes of European citizens. Whilst not the main focus of their analysis, the results in Bansak et al. (2016) demonstrate that survey participants prefer female to male asylum-seekers.

Asylum decisions in Germany are not determined by public vote or citizen committees, however. While there is little reason to believe that deciders at the Federal Office have gender preferences that systematically differ from the broader German population, the question is whether these preferences actually influence their decision-making.

Since each asylum request is assessed individually and via personal face-to-face interview, typically involving translators, deciders have very significant leeway in their decision-making such that any general preference for women over men could in fact influence their decision-making. The Federal Office for Migration and Refugees aspires to assure a uniform and purely merit-based asylum process via procedure management, namely in the form of official instructions, internal orientation aids and work guidelines (BAMF, 2019). However, social science research has demonstrated that recognition rates vary between German states (Riedel & Schneider, 2017), which casts doubt on uniform and purely merit-based asylum decision making.³

There is micro-level evidence demonstrating that the personal attributes of decision-makers and the gender of the individuals whose cases they decide on affect their decision-making. Keith et al. (2013) find that liberal immigration judges in the United States respond to asylum-seekers’ characteristics differently than conservative judges. Mascini (2008) in his analysis of decisions taken by 98 Dutch asylum caseworkers demonstrates how decisions by these caseworkers are influenced by, amongst other factors, their political orientation, professional background and role definition. There is an established literature demonstrating that female offenders are treated more lightly than male offenders by judges of both sexes (Ecker et al., 2020). Interviews with sixteen Dutch immigration officials suggest that these officials are more likely to regard male refugees as economic refugees and female refugees as seeking protection on humanitarian grounds (Van Wetten et al., 1998). Similarly, Mascini and Van Bochove (2009) speculate that men are more often suspected of being calculating economic refugees than women, whereas women are more often regarded as ‘defenceless victims’. Thus, it is possible that the gender gap in recognition rates is partly driven by a general preference for female as opposed to male asylum-seekers. This would probably provide evidence for taste-based discrimination by deciders, because if statistical discrimination exists, preferential treatment for women will vary across countries of origin since the degrees to which men and women are persecuted vary across countries of origin. Yet, the stark differences in the gender gap across countries of origin which suggests a potential general taste-based preference of asylum deciders for women over men cannot represent anywhere near the full story.

Demographic influences

Another influence stems from demographic factors for which individual-level studies provide evidence. Holzer et al. (2000) analyse approximately 180,000 asylum decisions at the decentralized cantonal level in Switzerland. Controlling for a range of cantonal attributes plus the age and marital status of applicants together with the duration and date of decision-making as well as some selected country of origin dummies, they find that marital status has a positive impact upon recognition chances but that the impact is larger for men than for women. Mascini and Van Bochove (2009) examine almost 162,000 asylum claims of individuals in the Netherlands and find that the lower success rate of men is influenced by two demographic factors: men are less likely to be married or accompanied by children and are less likely to follow their spouse for family reunification. Ecker et al. (2020) analyse nearly 41,000 asylum adjudications in Austria, finding that controlling for regions of origin, female applicants are statistically significantly more likely to be successful. They suggest that the gender gap is larger if a decider’s caseload is dominated by men.

Female asylum-seekers coming to Germany are more likely to be married than male asylum-seekers. If we take the marriage rates of refugees already resident in Germany as a proxy variable, on average across countries of origin at 38 per cent the share of women who are married is 10 percentage points higher than the share of married men. This imbalance has important consequences. To start with, deciders may give preference to asylum-seekers who are married (Mascini & Van Bochove, 2009). But even if deciders are not more likely to grant asylum to married rather than to unmarried men and women as such, the fact that a higher share of female asylum-seekers are married implies that they disproportionately benefit from what is known as ‘family asylum’ in German law. The instrument of ‘family asylum’ stipulates that ‘if a principal person (…) has been recognized as entitled to asylum, his or her family members who are in Germany are also granted asylum on application’ (BAMF, 2019, p. 26). For the purpose of the decision process, two asylum-seekers count as married if they already were married in their home country. This provision increases the probability of gaining protection for married individuals because only one ‘principal’ family member has to be entitled for the entire family to be granted protection in Germany. All other things equal the gender gap in recognition rates will be larger the higher the marriage rate of female asylum-seekers and will be the lower the higher the marriage rate of male asylum-seekers from a country of origin.

The share of minors or children among refugees varies dramatically across countries of origin, from a low of just above 6 per cent in the case of Gambia and Morocco to slightly above 52 per cent in the case of Russia and Montenegro. Since we analyse aggregate data, we have no information on the share of adult female versus adult male asylum-seekers who are accompanied by minors. However, if Mascini and Van Bochove’s (2009) finding that women are more likely than men to be accompanied by minors extends from asylum-seekers in the Netherlands to those coming to Germany then, all other things equal, a higher share of minors amongst all refugees increases the female recognition rate more than the male recognition rate. This is because refugee children enjoy a higher protection rate which is likely to extend to their accompanying adult.

Yet, all other things are not equal in this particular respect. The recognition rate of refugee children is not only almost double that of adults on average but also the gender gap in recognition rates among refugee children is much smaller than that of adults with the latter almost four times larger than the former on average. This implies that a higher share of minors amongst refugees also has the opposite effect, namely a levelling effect on the overall gender gap in recognition rates: The essential lack of a sizeable gender gap in recognition rates amongst refugee children means that, all other things equal, a higher share of refugees coming from a country of origin who are below the age of 18 reduces the overall gender gap in recognition rates. Which of the two opposing effects of a higher share of minors among refugees dominates is therefore an empirical question. Alternatively, the two effects could also cancel each other out.

Lastly, women might receive preferential treatment because they are outnumbered by male asylum-seekers. Almost two thirds of refugees coming to Germany were men and the number of men exceeds that of women among asylum-seekers in all but five countries of origin and by a factor of more than 10 for refugees coming from Tunisia, Gambia, Mali, Algeria, Bangladesh, Senegal and Morocco. Only Moldova, Zimbabwe, Vietnam, Mongolia and Kenya have more female than male refugees coming to Germany. Ecker et al. (2020) explain their finding of a preferential treatment of women by individual deciders whose caseload is highly skewed toward male applicants with psychological factors. Most decisions by deciders are negative in their study and most cases decided are with regards to men. When women come up as rare cases, they are seen as atypical, allowing deciders more easily to come to an atypical, i.e., positive, decision on the case, a variant of statistical discrimination. In our study, most decisions are not negative since the overall protection rate is above 50 per cent but we test for the relevance of this potential alternative explanation in our empirical analysis. If this explanation holds, the gender gap in recognition rates is the higher the larger the ratio of male to female applicants from a country of origin.

Research design

Our dependent variable is the gender gap in recognition rates in first instance decisions taken in a particular year for a particular country of origin. Asylum-seekers have the right to appeal against the first instance decision made by the Federal Office to the administrative court (Verwaltungsgericht) and, under certain conditions, to higher level courts up until the German Federal Constitutional Court. However, due to lack of reliable data, we exclusively focus on first-instance decisions. The gender gap is defined as the female rate of having received recognized protection status minus the male recognition rate. The recognition rate of women and men is the share of, respectively, women or men who have received a recognized protection status amongst all men or women seeking protection. Recognized protection status refers to protection granted on whatever grounds and independently of whether it has been granted on an open-ended or time-limited basis. Whilst we use the terms asylum-seekers, refugees and people seeking protection interchangeably, formally asylum-seekers and refugees form subsets of all individuals seeking protection defined as all foreigners present in Germany invoking humanitarian grounds for their stay (BAMF, 2019). Data are taken from Eurostat’s Asylum and Managed Migration database, which is also the source for data for explanatory variables unless otherwise noted.⁴

We exclude all countries of origin from which fewer than 200 first instance decisions were made in any one year so as to avoid the possibility that the protection status of a very small number of individuals can make a large difference on the aggregate recognition rates and therefore the gender gap in recognition rates. This results in a sample of 56 countries of origin, which are listed in Section 1 of the Online Appendix together with the female and male recognition rates and the resulting gender gap in recognition rates, averaged over the sample period 2012–2018.

To test our two predictions with regards to gender-specific and general or gender-unspecific human rights violations, we include measures of the prevalence of female genital mutilation in a country of origin, specifically the percentage of girls and women aged 15–49 who have been the victim of female genital mutilation, and of the prevalence of child marriage, specifically the percentage of children aged 15 or below who are married. These two variables are time-invariant with data referring to the most recent estimate that the data source provides. We further include a variable that captures the extent to which laws provide women with rights in relation to their husband in five domains, namely whether laws prohibit domestic violence, whether women are legally required to obey their husband and similar rights relating to the standing of women vis-à-vis men in the domestic household.⁵ We reverse this variable so that higher values on a scale from 0 to 5 indicate lower women’s legal rights. Lack of data prevent us from capturing other gender-specific human rights violations. General human rights violations in a country of origin are measured by the Political Terror Scales variable, where in order to maximize sample size we take the measure that is based on information contained in the U.S. State Department reports.⁶

To control for alternative explanations, we firstly include the marriage rate of, respectively, female and male refugees already resident in Germany in the year of decision.⁷ Note that, unfortunately, we have no data on the marital status of those whose asylum claims are decided upon in any particular year. However, whilst this introduces measurement error, the marital status of the already existing stock of refugees in Germany should be highly correlated with the marital status of those whose claims are newly decided. We include as a second demographic composition variable the share of minors or children, i.e., those below the age of 18, amongst all refugees from a country of origin whose claims are decided on in a specific year. Finally, we include the ratio of male to female asylum-seekers whose claims are decided on in a specific year. Section 2 of the Online Appendix provides summary descriptive variable statistics. Further potentially relevant control variables are added in a robustness test.

The existence of general pro-female preferences of those making protection decisions would show up in the intercept, which captures any gender gap in recognition rates unaccounted for by other explanatory variables. The intercept represents the average preferential treatment women receive across countries of origin conditional on controlling for other explanatory factors. Of course, any omitted variables will bias its coefficient.

Our sample covers the period 2012–2018. Clearly, observations from the same country of origin in different years are not necessarily statistically independent and there may be auto-correlation in the data. In the main estimations, we employ a random effects linear estimator with an assumed autoregressive error term of order 1 (AR1), which is based on a Prais-Winsten transformation to estimate the following equation:$y_{it}=\alpha +x_{it}\text{β}+{\epsilon }_{it}$, with ${\epsilon }_{it}=\rho {\epsilon }_{it-1}+{\eta }_{it}$where $\left|\rho \right|<1$ to avoid asymptotic estimates approaching infinity and the ${\eta }_{it}$ part of the error term is presumed to be independently and identically distributed. In a robustness test, we show that our results are very similar if we use a standard random effects linear estimator with a lagged dependent variable, year fixed effects and standard errors clustered on countries of origin. We do not include country fixed effects since some of our core explanatory variables are time-invariant and others vary only little over time. We employ Jackknife standard errors with replications clustered on countries of origin throughout.

Results

Model 1, for which results are reported in Table 1, only includes the intercept and therefore tells us that the average recognition rate across countries of origin in our sample is 12.1 percentage points higher for women than for men.⁸ Model 2 includes the human rights explanatory variables without control variables. We find that all three of the gender-specific human rights violation variables have the expected positive sign, whereas the general human rights violation variable has the expected negative sign. A higher prevalence of child marriage, a higher prevalence of female genital mutilation and lower women’s legal rights in the household are therefore all associated with a higher gender gap in protection status. By contrast, the gender gap is lower for refugees from countries with a worse record on general human rights abuses.

Table 1. Estimation results.

	M1	M2	M3
Lack of women’s legal rights in household		2.263**	1.096*
		(0.568)	(0.555)
Female genital mutilation prevalence		0.187**	0.198**
		(0.034)	(0.032)
Child marriage prevalence		0.740**	0.595**
		(0.140)	(0.169)
Political terror		−2.569**	−2.361*
		(0.972)	(0.951)
Share female refugees married			0.427**
			(0.119)
Share male refugees married			−0.441**
			(0.121)
Share of children in cases decided			−0.093
			(0.074)
Ratio men to women in cases decided			−0.213
			(0.116)
Constant	12.108**	10.033**	11.808**
	(0.995)	(2.788)	(4.186)
R-squared	n.a.	0.415	0.457
Chi squared	n.a.	74.15	100.2
Rho (AR1)	0.0976	0.107	0.101

Note: N=259. Number of countries=56. Jackknife standard errors with replications clustered on countries of origin in parentheses. Level of statistical significance: *p < 0.05; **p < 0.01.

Model 3 additionally includes control variables capturing the demographic composition of refugee populations. The coefficient of the variable capturing lack of women’s legal rights in the household becomes much smaller. By contrast, the results from model 2 on the other human rights variables are hardly affected. In substantive terms, an additional point on the lack of legal rights index which runs from 0 to 5, raises the gender gap in recognition rate by 1.1 percentage point. An additional percentage point higher prevalence in female genital mutilation and in child marriage raise the gender gap in recognition rates by 0.2 and 0.6 percentage points, respectively. This may appear small but standard deviation increases in these variables raise the gender gap by 5.9 and 3.6 percentage points, respectively. The predicted gender gap is 2.4 percentage points lower for every point increase on the 5-point political terror scale, which is also approximately its standard deviation. With respect to the control variables, we find that the coefficients of the marriage rates of, respectively, male and female refugees already resident in Germany have the expected positive and negative sign, respectively, and are almost identical in absolute size: increasing by one percentage point the share of the female or male refugee population who are married raises the gender gap by 0.35 percentage points and lowers it by 0.29 percentage points, respectively. This contradicts a finding reported in Holzer et al. (2000) which suggested that being married has a larger influence on recognition rates for male than for female applicants. We observe no statistically significant coefficients for the share of minors or for the ratio of men to women in refugee cases being decided. Interestingly, even in model 3 with all explanatory variables included, the coefficient of the intercept remains virtually the same as in model 1. The uncertainty of the estimate increases but the results do not allow us to reject the possibility of a fairly stable general ‘preference’ for female asylum applicants even after structural factors are included in the estimation model that together can account for almost half of the variation in the dependent variable as indicated by the R-squared.

Robustness tests

We conduct two sets of robustness tests. In the first set, we include further control variables to test for potential omitted variable bias and we alter the dynamic model specification. In the second set, we exclude the top quartile and the bottom quartile of asylum flows to check whether our results are driven by the inclusion of relatively large or relatively small asylum flows.⁹

In model 4, we include five additional control variables, namely a dummy variable for country years in which a country had been declared a so-called safe country of origin by German parliament, the total recognition rate, a variable capturing the share of the population of Muslim faith in the country of origin and two variables capturing, respectively, the share of male and female asylum-seekers who are between the age of 18 and 34.¹⁰ Article 16a.3 of the German basic constitution allows the German parliament with the approval by the ‘Bundesrat’, the Federal Council of all sixteen German states, to specify states

in which, on the basis of their laws, enforcement practices, and general political conditions, it can be safely concluded that neither political persecution nor inhuman or degrading punishment or treatment exists. It shall be presumed that a foreigner from such a state is not persecuted, unless he presents evidence justifying the conclusion that, contrary to this presumption, he is persecuted on political grounds.

A declaration of safe country should lower the total recognition rate for refugees from a country of origin since individuals will find it more difficult to demonstrate that valid grounds of protection apply to them despite the general presumption of safety in their country of origin. All other things equal, a low total recognition rate can be expected to result in a lower gender gap in recognition rates. Likewise, however, a very high total recognition rate also makes it more difficult for the gender gap in recognition rates to be large. We therefore control not only for the total recognition rate but also add its squared and cubic term to allow for great flexibility in functional form. Emeriau (2019) reports a significantly lower chance of gaining recognition for Muslim refugees in France and we want to ensure that our results are not simply driven by an aversion against men from predominantly Muslim countries of origin given the Islamophobia among a minority of German society, which finds its vocal political representation in the Alternative for Germany party. Lastly, Holzer et al. (2000) report that the age group between 18 and 34 is the relatively least likely to gain recognition in Switzerland and that the effect appears stronger for men than for women. As model 5 shows, apart from the total recognition rate, none of the coefficients of these control variables are statistically significant nor does their inclusion have any major impact on the results for the variables included in our baseline model except for the lack of women’s legal rights which becomes negative but statistically insignificant. The coefficient of the intercept becomes much smaller in size and statistically insignificant.

The estimated autocorrelation coefficients are around 0.1 in the estimations we have reported above and thus very low, suggesting that temporal dependence does not pose a major inferential threat. This issue is explored further in robustness test model 5, in which we replace the baseline model’s random effects linear estimator with an assumed autoregressive error term of order 1 (AR1) with a standard random effects linear estimator that includes the lagged dependent variable and year-specific fixed effects. Results are robust except for the marriage rate of male refugees. Note that with year fixed effects included in the model, the coefficient and standard error of the intercept are no longer directly comparable to the coefficient and standard error of the intercept of other models. The small estimated coefficient of the lagged dependent variable of only 0.22 corroborates the finding from the AR1 random effects estimator that temporal dependence poses no major threat to the interpretation of our estimation results (Table 2).

Table 2. Robustness tests 1.

	Baseline	M4	M5
Gender gap in recognition rate (t-1)			0.221
			(0.138)
Lack of women’s legal rights in household	1.096*	−0.086	1.260**
	(0.555)	(0.565)	(0.474)
Female genital mutilation prevalence	0.198**	0.157**	0.141*
	(0.032)	(0.034)	(0.056)
Child marriage prevalence	0.595**	0.319*	0.494**
	(0.169)	(0.148)	(0.143)
Political terror	−2.361*	−3.098**	−2.164**
	(0.951)	(0.927)	(0.789)
Share female refugees married	0.427**	0.359*	0.244
	(0.119)	(0.147)	(0.170)
Share male refugees married	−0.441**	−0.378**	−0.254
	(0.121)	(0.134)	(0.197)
Share of children in cases decided	−0.093	−0.046	−0.103
	(0.074)	(0.113)	(0.053)
Ratio men to women in cases decided	−0.213	−0.044	−0.027
	(0.116)	(0.310)	(0.254)
Total protection rate		1.294**
		(0.214)
(Total protection rate)^2		−0.026**
		(0.006)
(Total protection rate)^3		0.000**
		(0.000)
Safe country of origin		0.007
		(0.950)
Share of population who are Muslim		0.036
		(0.023)
Share female refugees aged 18–34		−3.122
		(13.590)
Share male refugees aged 18–34		0.104
		(0.106)
Constant	11.808**	1.230	7.451*
	(4.186)	(10.054)	(3.772)
R-squared	0.457	0.628	0.613
Rho (AR1)	0.101	0.106	n.a.

Note: N=259 (257 in model 5). Number of countries=56. Model 5 contains year fixed effects. Jackknife standard errors with replications clustered on countries of origin in parentheses. Level of statistical significance: *p < 0.05; **p < 0.01.

In the next set of robustness tests, we explore whether our results are driven by relatively large or relatively small numbers of refugees from certain countries of origin in the estimation model. In model 6, we exclude the bottom quartile of asylum flows whereas in model 7 we exclude the top quartile of asylum flows. Results are reported in Table 3.

Table 3. Robustness tests 2.

	Baseline	M6	M7
Lack of women’s legal rights in household	1.096*	1.673**	1.101
	(0.555)	(0.574)	(0.587)
Female genital mutilation prevalence	0.198**	0.171**	0.213**
	(0.032)	(0.033)	(0.035)
Child marriage prevalence	0.595**	0.502**	0.557**
	(0.169)	(0.155)	(0.172)
Political terror	−2.361*	−2.047*	−2.668*
	(0.951)	(0.949)	(1.079)
Share female refugees married	0.427**	0.074	0.391**
	(0.119)	(0.120)	(0.113)
Share male refugees married	−0.441**	−0.247	−0.363**
	(0.121)	(0.138)	(0.117)
Share of children in cases decided	−0.093	−0.141*	−0.148*
	(0.074)	(0.068)	(0.074)
Ratio men to women in cases decided	−0.213	0.043	−0.266*
	(0.116)	(0.388)	(0.126)
Constant	11.808**	18.892**	13.473**
	(4.186)	(4.416)	(4.521)
Number of observations	259	197	195
R-squared	0.457	0.519	0.474
Rho (AR1)	0.101	0.0872	0.130

Note: Jackknife standard errors with replications clustered on countries of origin in parentheses. Level of statistical significance: *p < 0.05; **p < 0.01.

In a further unreported robustness test, we account for the fact that the Federal Office underwent some administrative reforms after the peak of the refugee crisis in 2016 and had to hire many more deciders. We split the sample into the period 2012–2016 and 2017–2018. We find no statistically significant differences in estimated coefficients, except for the ratio of men to women in cases decided, which becomes positive and statistically significant.

Conclusion

We have explored whether human rights violations in the asylum applicants’ countries of origin statistically explain a substantial share of the cross-country variation in the gender gap in recognition rates over the period 2012–2018 in Germany. We have found evidence that they do. The gender gap in recognition rates varies systematically and in a substantively relevant way with gender-specific human rights violations and with general or gender-unspecific human rights violations in ways that theory predicts. The gender gap is larger for countries of origin with a higher prevalence of female genital mutilation and of child marriage but is smaller for countries with higher levels of political terror. It is important to note that even if the aggregate gender gap in recognition rates varies systematically in ways theory predicts, this nevertheless suggests that both men and women suffer from statistical discrimination as knowledge, or informed beliefs, of deciders about average persecution of men and women will influence their decision-making on individual asylum claims beyond the merit of individual cases. Such statistical discrimination is practically inevitable given asymmetric information with deciders often unable to verify the truth of asylum claims made. Even carefully executed individual-level studies like that of Emeriau (2019) cannot reliably code the actual merit of individuals’ asylum claims, which would otherwise allow testing the extent to which asylum decision-making is purely based on individual merit.

Of course, with respect to gender-specific human rights violations, we would not interpret these associations as being exclusively determined by the prevalence of female genital mutilation and of child marriage. Rather we suspect that these variables pick up some of the effects of omitted other gender-specific human rights violations with which they are correlated but for which no data exist. For the same reason, the results on the coefficient of the intercept need to be interpreted with great caution. Our estimations do not rule out the possibility of a general taste-based preference for women even after structural factors have been accounted for but that is all one should infer.

That the recognition rate of female asylum-seekers exceeds that of male applicants is not just a German phenomenon. We have taken Germany as our destination country of choice given it was the main target for refugees in the ongoing European refugee crisis. Future research should address whether the gender gap in recognition rates varies systematically across German states, similar to what Riedel and Schneider (2017) found for total recognition rates, and whether our findings generalize to other European countries. An early study found substantial differences across European countries and lack of convergence in recognition rates over time (Neumayer, 2005) and whilst a more recent study by Toshkov and de Haan (2013) reports some convergence since the Common European Asylum System (CEAS) was introduced in the early 2000s, the authors warn that ‘important national differences in the recognition of applicants from the same country of origin persist’ (Toshkov & de Haan, 2013, p. 661).¹¹ It is unclear to what extent these findings also apply to the gender gap in recognition rates.

Across all EU-28 countries plus Switzerland, Norway and Iceland, the total aggregate gender gap in recognition rates is largest in Romania, Switzerland, and Portugal, and smallest in Belgium, Latvia, Poland and the UK. Naturally, such total aggregate differences will in part be driven by heterogeneity in the country of origins of those seeking asylum in various destination countries but future studies should analyse whether there is convergence or divergence in the gender gap in recognition rates across destination countries for the same countries of origin and whether both gender-unspecific and gender-specific human rights violations explain the gender gap in recognition rates similarly well as they do for Germany.

Statistical replication materials and data

Supporting data and materials for this article can be accessed at https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/YHIPQ1

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Thomas Plümper

Thomas Plümper is Professor of Quantitative Social Research, Vienna University of Economics.

Eric Neumayer

Eric Neumayer is Pro-Vice Chancellor Faculty Development (from 1 September 2020: P-VC Planning and Resources) and Professor of Environment and Development, London School of Economics and Political Science.

Notes

1 We employ the terms ‘asylum-seekers’, ‘refugees’ and ‘persons seeking protection’ on humanitarian grounds interchangeably. In Section “Human rights violations and their impact on the gender gap in asylum recognition”, we explain how the former two formally form a subset of the latter, which is the official term now used by the German Office for Migration and Refugees.

2 Here and elsewhere, data refer to averages over the period 2012–2018 and exclude country years with fewer than 200 asylum recognition decisions.

3 The same holds for Swiss cantons (Holzer et al., 2000). This variation within federally organized states has led to the charge that the asylum process resembles a ‘lottery’ (Schneider & Riedel, 2017) or ‘roulette’ (Ramji-Nogales et al., 2011).

4 https://ec.europa.eu/eurostat/data/database.

5 Data on female genital mutilation come from https://data.unicef.org/topic/child-protection/female-genital-mutilation/, data on under-15 marriage rates from https://data.unicef.org/topic/child-protection/child-marriage/, and data on domestic violence legislation and rights relating to the standing of women in the family household from https://wbl.worldbank.org/.

6 Data taken from http://www.politicalterrorscale.org/.

7 Data taken from www-genesis.destatis.de.

8 Note that this is only seemingly at odds with the size of the gender gap in recognition rate for all refugees coming to Germany, which as mentioned in the Introduction stands at 3.3 percentage points. That overall figure is heavily influenced by refugees from Syria representing the single largest refugee population with a very small gender gap due to an extra-ordinarily high recognition rate for both men and women.

9 See Neumayer and Plümper (2017) for a broader discussion of the logic of robustness tests.

10 Data on the Muslim population share variable are taken from Graham and Tucker (2019), the list of so-called safe countries of origin can be found in BAMF (2019).

11 With regards to more general asylum policy harmonization at the EU level, scholars have noted that a ‘race to the bottom’, feared by some, did not occur and have analysed the reasons behind this development (Kaunert & Léonard, 2012; Zaun, 2016).