MIGRATION

Edited by Elzbieta Gozdziak, Georgetown University

doi:10.1111/j.1468-2435.2011.00701.x

Methods, Approaches and Data Sources for Estimating Stocks of Irregular Migrants
Michael Jandl*

ABSTRACT
This paper presents a comprehensive review of available methods for sizing irregular migrant populations as a particular group in the study of hidden populations. Based on the existing body of literature on the subject, a generic classication scheme is developed that divides existing estimation procedures into subcategories like approaches, methods and estimation techniques. For each of these categories, basic principles, methodical strengths and weaknesses, as well as practical problems, are identied and discussed with the use of existing examples. Special emphasis is placed on data requirements, data shortcomings and possible estimation biases. In addition, based on the empirical classication and quality assessment of country-specic estimates developed in the CLANDESTINO research project, the potential and requirements for replicating best practice models in other countries are explored. Finally, a number of conclusions on the appropriate design of estimation projects are offered.

INTRODUCTION
In social science research we frequently encounter phenomena about which we have little or only incomplete information. So-called hidden populations are particularly difcult to observe and quantify because their members may wish not to be identied out of fear or shame (e.g. illiterate, HIV-positive or homeless persons) or because their activities are illicit or illegal (e.g. drug dealers, contraband smugglers or moonlighters). However, for various reasons, knowing more about the size, characteristics and social behaviour of such hidden populations
* United Nations Ofce on Drugs and Crime (UNODC), Vienna, Austria. Michael Jandl is a Research Expert with the UNODC. This article was written in a personal capacity of the author and the views expressed are those of the author only and not those of UNODC or of its Member States. 2011 The Author International Migration 2011 IOM International Migration Vol. 49 (5) 2011 ISSN 0020-7985

Published by Blackwell Publishing Ltd., 9600 Garsington Road, Oxford OX4 2DQ, UK, and 350 Main Street, Malden, MA 02148, USA.

54

Jandl

is of considerable interest to policy-makers as well as social scientists. Such knowledge is of relevance for policy-making in a number of elds such as regulation and enforcement, provision of basic services and health care, the protection of fundamental human rights and the implementation of assistance programmes. Researching and sizing any of these hidden populations is complicated by many of the same methodical problems. A full count of the target groups is, by denition, impossible. It is also not possible to draw a representative sample from such a population, as the structure of the underlying total is unknown. This is also true for researching and sizing irregular migrants, who have incentives to deliberately hide from public authorities. Thus, the number of irregular migrants that are documented at any one time in ofcial statistics is inevitably only a subset of the total population of irregular migrants. To know more about the total population, we need to obtain an estimate of the dark gure, that is, that part of the irregular migrant population that is not documented in the data but is likely to constitute the major part of it. To do so, social scientists draw conclusions from observed statistical indicators on the dimensions and characteristics of the unobserved phenomena (e.g., illegal residence or illegal employment) by using various estimation methods and techniques. While all such estimates are necessarily imprecise and prone to large margins of error, the use of scientically sound methods is preferable in that they specify the assumptions, data and methods used and can thus be discussed and criticized by external evaluators. This paper provides a systematic overview and categorisation of available estimation methods and techniques, discussing their basic logic, assumptions and limitations1.

A CLASSIFICATION OF METHODS FOR ESTIMATING STOCKS OF IRREGULAR MIGRANTS

Before diving into the methodological discussion a note on terminology and denitions is in order. In this paper, the terms irregular, undocumented, unauthorized or illegal migration are used synonymously, while in other contexts a more nuanced use of these terms may denote different facets of the wider phenomenon of irregular migration. More specically, when referring to irregular or illegal (foreign) residents, the term denotes residents without a legal residence status in the country in which they are residing. The term irregular or illegal entrants refers to persons who cross an international border without the required valid documents, either un-inspected over land or sea, or over ports of entry.2 As with the data on legal migration, the fundamental distinction between all estimates on irregular migration is that they refer to either stocks or ows. In the following, the discussion is limited to available methods for sizing irregular migrant stocks. Methods for the estimation of stocks of irregular migrants can be divided into direct and indirect approaches and into further sub-categories labelled estimation techniques.3 Direct approaches are based on data that capture the subject of

2011 The Author. International Migration 2011 IOM

Estimating stocks of irregular immigrants

55

research (e.g., illegally resident foreigners or irregular migrant workers) directly, while indirect approaches do not rely on such data. Data sets that contain (a subset of) the target population directly are immigration enforcement data (e.g., apprehended illegal residents), administrative records (e.g., data on regularizations of irregular migrant workers) and survey data (e.g., irregular migrants identied through snowball sampling techniques). On the other hand, residual methods based on the difference between, for example, the total population represented in census gures and some estimate of the legally resident population (see sections below) are classied indirect, rather than direct, approaches, because at no point do they identify illegal residents directly in their data. Table 1 provides an overview of this classication scheme and the required data sources.4 In addition to such methodologically pure methods for the estimation of stocks of irregular migrants as presented in Table 1, many approaches make use of a combination of estimation techniques for sizing stocks of irregular migrants (see last section).

DIRECT APPROACHES
Multiplier methods The multiplier principle in estimation methods begins from the proposition that the size of the unknown total can be directly deduced from the size of a known subtotal by use of an appropriately derived multiplier. Once this proposition is accepted (and the size of the subtotal has been established with an acceptable degree of accuracy), then the estimation problem is redened as one of nding the correct multiplier (Vogel, 2002). The use of multipliers is probably the most common method employed for estimating the dark gure of hidden populations in migration, as in other elds5. The multiplier principle is also the starting point for more sophisticated statistical estimation techniques (see below). Simple multiplier techniques Today, simple multiplier techniques are frequently used for estimating irregular migration phenomena on the basis of apprehension data, with the multiplier often resulting from expert opinions or other subjective methods. However, this is unsatisfactory for a statistical technique, as the basis for a subjective estimate cannot be scrutinized and evaluated. Even simple multipliers should therefore have a statistical basis. Furthermore, it has been pointed out that commonly available (aggregate) apprehension data tend to mix stocks and ows and are therefore unsuitable for the application of the multiplier technique (cf. Jandl and Kraler, 2006). Moreover, apprehension data are strongly dependant on the intensity of law enforcement efforts. This means that any (estimated or derived) multiplier has to take into account both the time and the risk dimension of the indicator.

2011 The Author. International Migration 2011 IOM

TABLE 1 METHODS FOR STOCK ESTIMATES Estimation technique Simple Multiplier Capture-recapture Repeated capture Matching of registers Random effect mixed modelling approach Evidence based on regularisation data Main Idea of Calculation

56

Approach

Data sources

Method

Direct approaches

Multiplier methods

Methods of self-identication

Jandl

Immigration enforcement data; registers of residents or migrants including records of irregular residents Administrative data of regularizations; residence permits data Sample surveys of migrants Using data on status adjustments Center sampling technique Surveys using the randomized response technique 3 cards method Snowball sampling techniques Respondent driven sampling

2011 The Author. International Migration 2011 IOM

Survey methods

Estimation of total with a simple multiplier based on derived or estimated ratio of dark gure vs. clear gure Estimation based on probabilistic function derived from multiple recaptures of individuals in sample Estimation based on implied non-registrations in two or more individually matched registers Estimation using statistical regression model assuming comparable apprehension rates of legal illegal residents with statistical adjustment for random effects Inferences on the size and composition of irregular migrant stocks prior to regularization from data on applications for and grants of regularizations Inferences derived from data on changes in residence status after a period of irregular residence Reconstruction of a random sample of regular and irregular migrants through a re-weighting of the probability of contacts Eliciting information about the share of respondents employing irregular migrants through a non-threatening survey design Inferences about share of irregular migrants in sample survey using non-threatening survey design combined with residual estimation results Estimation using chain referral methods to obtain a sample of persons not registered vs. persons registered Recruitment of interviewees through peers and an incentive system leads to equilibrium sample of respondents after several recruitment waves which is independent from original sample and can be analysed statistically

TABLE 1 (CONTINUED) Estimation technique Main Idea of Calculation Indirect estimation based on the calculated difference between census data and data on legal immigrants

Approach

Data sources

Method

Indirect approaches

Census data; administrative registers or sample surveys of immigrants

Residual methods

Census data; admin. registers; demographic data

Demographic methods Expected population methods

Indirect estimation based on a comparison of two or more registers with data on the same target population Inferences on demographic subgroups based on the comparison of real and expected birth or death rates Indirect estimation of illegal resident population from comparison of emigration estimates with data on legally resident immigrants at destination

Estimating stocks of irregular immigrants

2011 The Author. International Migration 2011 IOM Differences between census results and legal immigration data Simple comparison of various registers Use of birth death rates Comparison of census emigration data and immigration statistics Calculating the stock through ow gures Using information on correlated phenomena as basis of calculation Expert surveys Delphi surveys

Administrative data Various complementary data sources (economic data, health data, etc.)

Flow-stock methods Indirect inferences

Surveys of key informants (experts, employers, etc.)

Subjective Estimations Indicator Methods

Using estimated inow- and duration of stay indicators to estimate a steady-state stock of illegal residents Making inferences on subgroups of irregular migrants on the basis of indirectly related phenomena and estimates such as irregular work, sector-specic demand for irregular services, school attendance or health services (e.g. inference of share and size of irregular foreign workers from econometric estimates on the shadow economy) Survey of key informants on their assessments of sizes, ratios and characteristics of target population Anonymous multiple round survey of key informants mediated by researcher to attain a convergence of opinions

57

58

Jandl

A good example of deriving a multiplier statistically from a sample population is provided by Burgers (1996, cited in Pinkerton, 2004: 14f) who uses a record and survey-based method to determine the number of illegal foreign residents in Rotterdam. First, Burgers uses police data on apprehended foreigners over a six year period to gain a measure of how many illegally resident criminal offenders were captured. Then he uses in-depth interviews with a (non-random) sample of 145 illegal residents to determine the proportion of migrants involved in criminal activities over this period. This (inverse) proportion is then applied to the apprehension number as a multiplier.6 One example of a multiplier method recently applied to data on crime suspects that acknowledged the uncertainties of the database by putting forward a range of (minimum-maximum) estimates. This multi-year (20012008) estimate of illegal residents in Austria (Jandl 2009) is based on crime suspect data by residence status, including illegal residents and excluding non-residents such as tourists. It argues that the proportion of crime suspects among illegal residents falls between those of natives and other foreigners.7 Repeated capture methods Capture-Recapture methods have a long tradition in population biology and were originally developed to estimate animal populations in the wild (cf. Jandl, 2004). The capture-recapture method is basically a multiplier method in which the multiplier is developed through a repeated sampling of the same population. In an advanced form, the capture-recapture method has also been applied to sizing the number of illegal residents in the Netherlands. Van der Leun et al. (1998) rst used the repeated capture method for estimating the size of the illegally resident population in the four largest Dutch cities.8 The estimate was based on one single data set (7,000 les of all apprehensions of illegal residents in the four largest Dutch cities in 1995). A crucial data requirement for the method to work is that the case les can be analysed to determine who has been captured once, twice, three times and so on, with no or few linking errors (which occurs through the use of ngerprinting). The same method was later applied by Engbersen et al. (2002) and by van der Heijden et al. (2006) for producing updated estimates of the number of illegal residents in the Netherlands. Once the numbers of single and repeated captures has been established, the method postulates that these counts follow a probabilistic distribution, known as the Poisson distribution. On the basis of the available data, the parameter determining the Poisson distribution can be estimated, which is then used to calculate the probability that an individual is never (zero times) caught by the police. Adding to this estimated number the number of illegal residents actually captured various times in the police les produces an estimate of the total illegally resident population (cf. van der Heijden, Cruijff et al., 2003). Despite its statistical elegance, the repeated capture method depends on two critical assumptions: rst, a constant probability of apprehension that does not change after a previous capture, and second, a constant population over the

2011 The Author. International Migration 2011 IOM

Estimating stocks of irregular immigrants

59

period studied. The second assumption can be relaxed through a separate estimation for the group of effectively expelled persons (van der Heijden, Bustami et al., 2003). Moreover, van der Heijden (2006) demonstrates that group-specic differences in the risk of being caught (heterogeneity) can be included in the model. Matching of registers The matching of registers technique uses the fact that illegal residents sometimes appear in a certain register (e.g., police apprehension les, aliens register) and can be identied and re-identied when appearing in another register as well. If the probability of appearing in the rst register is independent of the probability of appearing in the second, the identication of the same individual in both registers constitutes, metaphorically speaking, a recapture of the same individual and the total can be derived from the sizes and capture rates of the two samples. The method can be illustrated with the use of a so-called contingency table (cf. Sheldon, 2002) that crosses two registers N1 and N2. The resulting table has four elds, where each eld lists the number of persons who are to be found either in both registers (n11), in one but not the other the two registers (n12 and n21), and in neither register (n22). If the probabilities of being included in the rst or the second register are independent from each other, then the number of persons not included in either register can be calculated as n22 = (n12 n21) n11. Like the repeated capture method, the simple application of this estimation technique is based on some crucial assumptionsmainly a closed population over the period and no linking errors (i.e. a person can be correctly identied as appearing in only one or both registers). Heterogeneity (differences in the probability of being captured) can be accommodated in the model if a sufcient set of covariates (nationality, socio-economic variables) is available (cf. Sikkel et al., 2006). A random effects mixed modelling approach This approach tries to estimate the size of the illegally resident population (M) through statistical modelling from data on registered legal residents (N) and data on apprehended legal and illegal residents by nationality (ni and mi respectively, where the subscript i denotes nationality) (Zhang, 2008). While the proportion of apprehended foreigners is known among the legal residents (i.e., n N), it is unknown for the illegal residents (i.e., m M is not known). The model realistically assumes that there are a number of reasons why these rates differ between nationalities, and without knowing the difference one cannot estimate M directly. The model thus introduces a random effect that is able to accommodate the perceived heterogeneity in relative apprehension rates. The basic features of the model are, rst, that it assumes that the ratios of the illegally resident to the legally resident subpopulations (Mi Ni) are a random deviation from a global proportionality coefcient. Next, it assumes that the catch-rate (mi Mi) of the illegally resident subpopulations can be represented

2011 The Author. International Migration 2011 IOM

60

Jandl

as a function of the catch-rate of the legally resident subpopulations (ni Ni), including a random effect. Given the random effect, both assumptions are by denition true. The question is whether the model can be sensibly estimated and whether the data allow for the desired interpretation. In applying the model to data on registered and apprehended foreigners in Norway, Zhang (2008) nds that the regression of the model to the data results in a good t and central estimates of the illegally resident population (by former residence status) fall within reasonable condence intervals. Methods of self-identication Methods that rely on the self-identication of members of hidden populations are often mistaken to provide counts of the total. This also applies to data from regularization programmes of illegally resident (or illegally working) migrants, whereas it is far from clear that most irregular migrants would be able and willing to reveal their irregular status to the authorities. Evidence based on regularization data Large-scale regularizations (or amnesties) are often carried out to solve the problems related to the presence of irregular migrants or irregular migrant workers. Almost as a by-product they also provide data on the size and structure of their target groups (cf. Pinkerton, 2004: 39f). When interpreting such data, it should thus be clear that participation in such programmes is already restricted by the terms of the amnesty on offer, such as restrictions pertaining to certain subgroups (e.g., former asylum seekers, family members of regular migrants, minors, etc.), a certain cut-off date, a required length of residence or proof of employment. Moreover, even among those eligible to apply, not everyone will be willing to reveal her or his identity to the authorities. There are a number of other reasons for which regularization data cannot serve as direct estimates. First, the number of applications is sometimes inated because the same persons apply more than once in order to increase their chances (ibid: 40). Second, some migrants who possess a legal residence permit that does not grant them the right to work may apply to obtain a different kind of residence permit that does grant this right (Cangiano, 2008: 103). Third, previously regularized migrants may later fall back into irregularity (thus appearing repeatedly in successive amnesty programs).9 Finally, an unknown number of nonresidents (e.g., irregular migrants from other countries) may also le an application. Using data on status adjustments In lieu of publicly announced (mass) regularization programmes, many countries make use of more discretionary programmes of ongoing regularizations on a case-by-case basis that can also shed some light on previous illegal residence. For example, Thierry and Rogers (2004) use data on rst contacts with the

2011 The Author. International Migration 2011 IOM

Estimating stocks of irregular immigrants

61

prefectures in France to determine the number of persons who obtained a regular status after a period of illegal residence. Though they do not provide an overall estimation of irregular residence, the data do offer a minimum estimate of new illegal residencies after the special regularizations in 1997 and 1998.10 Most of the individuals regularized had been in France for many years. Similarly, in the UK Worker Registration Scheme after EU-enlargement in May 2004, a considerable number of applicants from the 10 new EU Member States may already have been in the UK before May 2004 (Home Ofce, 2005: 1, 8). Survey methods Survey methods for the study of hidden populations are dominated by two types of techniques: Location sampling techniques are suitable when the target population is large and geographically concentrated (overall or at certain times), while chain-referral sampling techniques are suitable when members of the target population are fewer and interconnected through traceable contact patterns (Heckathorn, 2002). Center sampling technique When the number and share of irregular migrants among the general migrant population is large, and when migrants do not systematically misrepresent their status in interviews, a direct large-scale survey of both regular and irregular immigrants can be constructed for the measurement of both groups. Such surveys have been carried out by the Foundation ISMU in Italy since the early 1990s with samples of up to 30,000 migrant interviews.11 The survey technique applied by ISMU is the so-called centre sampling technique, which assumes that each foreigner in the sampled area keeps a set of contacts with one or more centres or gathering places located in that area. Once a sufciently wide set of centres is identied, any foreigner can be described by her or his set of contacts with various centres). Based on the information derived from the migrant interviews, a matrix of all centres visited by each interviewee is drawn up and used to adjust the non-random sample to obtain a representative sample of interviewees. To do so, rst the centres are selected randomly (with replacement), and then within each centre one interviewee is drawn. The number of drawn interviewees in a certain centre will depend on its size, and the probability of inclusion of each migrant depends on the number of centres frequented by him or her and the number of people frequenting these centres. Thus, the re-weighted sample becomes representative and the share, characteristics and size of the irregular resident population can be calculated (Baio et al., 2008). Direct surveys with randomized response techniques Randomized response is a method designed to facilitate surveys of sensitive topics (like drug use, sexuality or illegal residence). The survey is designed such that

2011 The Author. International Migration 2011 IOM

62

Jandl

answers to sensitive questions are partly determined by chance and partly by a pre-determined rule (forced response), while the aggregate results can be used for statistical analysis (cf. Sikkel et al., 2006: 153f). A randomized response method for surveying employers of irregularly employed migrants has been applied in the Netherlands by Regioplan Policy Research (cf. van der Leun and Ilies, 2008: 19f). Direct surveys with the 3 cards method An alternative to randomized response surveys is the so-called 3 cards method, which was developed by the US Government Accountability Ofce (GAO) for the United States (cf. Larson and Droitcour, 2003). Rather than posing the sensitive question on a persons immigration status directly in a survey, respondents are asked to choose one of three slightly different answer cards, each of which arranges several mutually exclusive answer categories on their immigration status in three boxes. Information on the sensitive answer category (undocumented) can then be estimated indirectly by statistical inference. Direct surveys using network characteristics The snowball sampling technique is a non-random sampling method that relies on a number of initial respondents for locating additional waves of respondents of hidden populations. Zhang (2008: 37ff) describes the so-called singlestage link-tracing sampling method in the classic study on cocaine users by Frank and Snijders (1994), in which initial respondents are also asked to nominate other members of the target population but where the sample is restricted to one wave only. Both methods are based on the (problematic) assumption that the resulting samples would be representative of the target population. Respondent-driven sampling (RDS) is another form of the chain referral technique that explicitly tries to reduce the bias introduced through the initial (nonrandom) sample selection. In RDS respondents are asked to recruit their peers into participation in the study by offering them an award for participating in the interview as well as a reward for recruiting others into the study. After ve to six recruitment waves, the resulting sample should be independent from the original sample (Heckathorn, 2002).

INDIRECT APPROACHES
Methods subsumed under indirect approaches do not use actual counts of irregular migrants (such as apprehensions data, etc.) as the basis for their estimations, but derive such estimates from data related to the phenomenon to be measured.

2011 The Author. International Migration 2011 IOM

Estimating stocks of irregular immigrants

63

Residual estimation methods The use of residual estimation methods is not common in Europe, mostly because the censuses and other registries are likely to seriously undercount irregular migrants in European states (cf. Lederer, 2004: 197). The major exceptions are the UK and Spain. Comparison of census results with legal immigration gures Based on the proposition that the Census provides a valid count of the total immigrant population, an estimate of the number of illegal residents can be obtained by subtracting the number of legally resident immigrants from this total. This method is most commonly applied in the United States (Passel, 2006). The data are adjusted for a potential undercounting of immigrants in the census and regularly updated (cf. Hoefer et al., 2008). As there is no population registry in the US, the major difculty in this model is in calculating the total legally resident foreign-born population (adjusted for emigration, mortality and temporary migrants). Following the US example, the residual method has also been applied on the 2001 census data in the United Kingdom. The calculations are similar but another category of migrants who are in the UK on a quasi-legal basis is subtracted from the total (e.g., asylum-seekers or refused applicants with leave to remain; cf. Woodbridge, 2005). In the absence of a registry, the stock of legal permanent immigrants is obtained by adding the number of migrants granted settlement each year since 1970 and adjusting the number for likely emigration and death.12 No explicit undercount estimates of the foreign-born population in the census are available, so a range of undercounts (lower-middle-upper) is applied, resulting in a range of estimates of the residual (illegally resident) population. Simple comparison of registers In Spain the law allows foreigners, irrespective of their legal status, to be included in the municipal register (Padron) if they can provide evidence of residence in that municipality. Irregular migrants face positive incentives (such as free health benets) for registering, and consequently most are thought to be recorded. Hence, the number of irregular foreigners in the country can be estimated by subtracting the number of registered foreigners with residence permits from the gures in the Padron. Despite the apparent simplicity of this estimation technique, there are sev lez Enriquez, 2008: 27ff). eral problems related to the data used (cf. Gonza First, the Padron is likely to contain many foreigners who have already left the country but did not de-register13 and some persons who have never lived in Spain but have been registered by their relatives. Second, the residence
2011 The Author. International Migration 2011 IOM

64

Jandl

permits registry may also over-represent the number of legally resident foreigners in the country (due to emigration, death, naturalization, etc.). And third, there are quite likely a number of irregular migrants who are not recorded in either register. Demographic methods These are based on the idea that over the course of ones life each individual faces a certain likelihood (or demographic risk) of experiencing a demographic event like giving birth, dying or being hospitalised, and that also this holds true for illegal residents. Hence, a comparison of recorded demographic events by age, sex and group cohorts can theoretically be used for sizing the illegally resident population (cf. Wanner, 2002: 11ff). In practice, there are few studies applying demographic methods. Delaunay and Tapinos (1998: 65) present an older Italian study (Natale, 1990) that analyzes the growth in recorded birth rates of foreigners and compares it with the growth of legal foreign residents. Lederer (2004: 195) cites a study by Robinson (1980) on specic death rates in 10 US States but critically notes the wide range of the resulting estimate.14 A particular problem with demographic methods is that most demographic risks for illegally resident foreigners are likely to differ from those of legal residents as migration is highly (self-) selective as in cases where vulnerable migrants choose to return as a result of impending demographic risks (giving birth, chronic illness or death). On the other hand, those demographic events (such as fatal road accidents) that suffer from less bias through self-selectivity, return migration or non-registration (cf. Vogel, 1999: 172) are likely to be too rare for estimation purposes in most (smaller) European countries. Comparison of immigration and emigration statistics In principle, comparing the number of all emigrants to the number of legally resident immigrants from one country to another can yield an estimate of the number of illegal residents from that country of origin. Lederer (2004: 192ff) describes two studies of this type on undocumented migrants from Mexico in the US. One study uses Mexican census results (1960 and 1970) while the other study compares observed and expected sex-ratios in Mexico (missing males) to estimate total emigration to the US. In the European context, Delaunay and Tapinos (1998: 4256) make their own calculations on data from Morocco and Tunisia to estimate the number of illegal residents from these countries in the main European destination countries. The authors use both data from Moroccan censuses in 1982 and 1994 and Tunisian censuses in 1984 and 1994, and also test the sex-ratio method on the Moroccan case. However, the authors admit that their results for Morocco15 suffer from serious data deciencies in the census and that the estimation for Tunisia was not possible due to a lack of reliable data.

2011 The Author. International Migration 2011 IOM

Estimating stocks of irregular immigrants

65

Flow-stock methods In physics, the formula for calculating the (steady-state) equilibrium of a (continuously replenishing) stock is: stock equals ow per period multiplied by the average duration of stay within the system (cf. Sheldon 2002: 22ff). This principle can also be applied to the stock of illegal residents in a country (cf. Tapinos, 1999: 232). While data on ows and average duration of stay are as difcult to obtain as other indicators of illegality, in certain cases utilizable data may be readily available. Taking up a suggestion by Wanner (2002: 14f), Sheldon (2002: 23f) gives the example of the (unknown) stock of disappeared asylum seekers: if 1,000 rejected asylum seekers per month disappear into illegality and remain hidden for an average of 6 months until their apprehension, it follows that the stock of disappeared asylum seekers is 6,000 on average. It is important to point out that the steady-state formula applied in this example requires knowledge of the average completed duration of stay in the system, as opposed to the average duration of stay of persons still within the system. Methods based on indirect inferences Methods that make only indirect, rather than direct, inferences on irregular migration make use of data on correlated phenomena, rather than data on (regular or irregular) migration as such. A widely used estimation technique in this group makes inferences on the involvement of irregular (foreign) workers from calculations of the likely labour demand in a certain sector or branch. For example, Hess (2006, as cited in Cyrus, 2008: 44) estimates the overall number of seasonal workers required in German agriculture based on data of oor space, yields and output, as well as the average (estimated) working time and contract duration for seasonal workers. Subtracting the number of ofcially placed seasonal workers from this estimated labour demand yields an estimate of the number of irregularly employed foreign workers in agriculture. Another group of estimation techniques makes inferences on illegal foreign employment based on the estimated participation of foreigners in the informal economy. Estimates on the output (and relative size) of the shadow economy16 can be derived from econometric models that use the amount of cash in circulation or the amount of electricity consumed as indicators (cf. Schneider and Enste, 1999). Applying the method to the Austrian and German economies and using data on per-capita productivity, the estimates of the shadow economy output are then transformed into estimates of informal labour inputs required (in full-time equivalents). Finally, the estimated labour inputs are split into those by native and foreign workers, though the precise procedure for this crucial step remains unclear (cf. Schneider, 2006; for a critique, see Kraler et al., 2008: 21ff). Indirect inferences can also be based on non-economic indicators, such as administrative data on health care. For example, the French government bases

2011 The Author. International Migration 2011 IOM

66

Jandl

its ofcial estimate of the illegally resident population on the number of people who have received free medical aid from the state. However, in this ofcial estimate, no explicit adjustment is made for the share of illegal residents not claiming any medical aid during their residence in France because they had no need to or because they did not full the three eligibility criteria for beneting from medical aid (cf. Courau, 2008).17 One problem with methods based on indirect inferences is that they add one or several levels of uncertainty to the calculations. Thus, estimates on the shadow economy do not automatically translate into estimates on irregular foreign workers, which is yet again a different category from irregular foreign residents. The same applies to indirect inferences on the basis of other indicators. Subjective estimations Indicators methods Estimation methods that rely on the collection and aggregation of subjective opinions and estimations of experts will necessarily result in subjective indicators, no matter how large the sample or sophisticated the information processing18. Regarding the latter, we can distinguish between (one-round) expert surveys and (multi-round) Delphi surveys. Expert surveys Experts (also called insiders or key informants) are individuals who are assumed to have privileged knowledge of a subject (e.g., the target population). The validity of estimates derived from this method is thus dependent on whether the experts selected really have such privileged insights, for example through direct interaction with the target population. It follows that the accuracy of any individual subjective estimate is potentially higher the smaller the (geographical) scope of the estimate. To produce estimates with a wider scope (e.g. total economy or nation-wide), both a suitable sample of experts as well as an appropriate aggregation method is needed. Ideally, the sample covers experts with insights into the whole target population, while the aggregation method leads to a representative estimate of the expert sample. Both are difcult to achieve in practice. Lederer (2004: 237ff) relates the method and results of a survey on irregular migrants in Germany carried out by Caritas Germany in 1995. A written questionnaire was disseminated to 1,000 key informants from NGOs, law rms, churches, etc. working with (irregular) migrants. However, it was not possible to collect conclusive information on the overall size of the target population. Cyrus (2008: 24f) concludes that the selective and non-representative nature of the clients served by various charities makes such surveys unsuitable for general inferences on irregular migrant populations. Targeting a different group of key informants to estimate the extent of illegally employed foreigners, Piguet and Losa (2001) have carried out an anonymous survey of 821 employers in Switzerland, using indirect questions

2011 The Author. International Migration 2011 IOM

Estimating stocks of irregular immigrants

67

pertaining to the estimated percentage of illegally employed foreigners in their particular branches of business. It is noteworthy, however, that the results depend heavily on the choice of statistical aggregation.19 Similar surveys of employers hiring irregular migrant workers were also carried out in the Netherlands (cf. van der Leun and Ilies, 2008: 17f) and in the agricultural sector in northern Greece (cf. Pinkerton et al., 2004:16f). Small-scale expert surveys, involving interviews with both key informants and irregular migrants have been carried out in the German cities of Leipzig and Munich (Alt 2003). Alt concluded that the share of the illegally resident population ranges between 1 per cent and 3 per cent in large German cities. Delphi surveys The Delphi method is a variant of expert surveys that involves a certain number of independent experts in an interactive process of exchange through the use of written questionnaires, designed to foster convergence and consensus (cf. Jandl et al., 2007: 14f). Experts are given the opportunity to state their views on a given topic and react to the (anonymized and consolidated) views and assessments of other experts in the subsequent rounds (minimum of two rounds of questionnaires). While Delphi surveys are frequently used to collect qualitative inputs on hard-to-research populations from a diverse range of experts, there have been few attempts to apply the method to the quantication of irregular migration. In a recent collaborative research effort, two-round Delphi surveys have been employed to produce estimations on the extent of irregular migrant work in three central European countries (see Jandl et al. 2007 on Austria; Drbohlav , 2008: 20, on the Czech Republic; and Futo, 2008: 31f, on the and Lachmanova study by Juhasz on Hungary). While the resulting estimations are clearly based on subjective evaluations, some experts see the method as a valuable addition to a system of multiple indicators on irregular migration (cf. Lederer, 2004: 207f).

THE USE OF REGRESSION ANALYSIS FOR EXTRAPOLATION IN SPACE

As already mentioned, several approaches use regression analysis as part of their estimation technique (cf. 3.1.4. and 4.5). In addition, regression analysis is also used for extrapolating estimates in space from the local level to the country level, using the correlation of the small-scale estimates to a set of known variables in multiple regression analysis. Of course, any results obtained through the extrapolation of small-scale to large-scale estimates depend crucially on the quality of the former. The window method, also called postal code area method rst seeks to establish the size of a hidden population in well-dened and restricted areas in order to then extrapolate to a larger area. The advantage of using well specied postal

2011 The Author. International Migration 2011 IOM

68

Jandl

code areas is that detailed statistics on socio-economic characteristics (e.g., purchasing power, employment, industries) within such statistical units are usually readily available for extrapolation. When suitable characteristics that are statistically correlated to the hidden population are employed, the use of multiple indicators should improve the estimation against a simple extrapolation with population totals (cf. Sikkel et al., 2006: 153f). The application of the window method on estimating the size of the illegally resident population is currently tested in an ongoing research project in the , 2008: 25f). In the project, a full Czech Republic (Drbohlav and Lachmanova survey of actual residents (regardless of their legal status) in several areas of Prague is to be implemented (via observation and interview method). The observed numbers will then be compared with ofcial numbers of legal foreign residents to derive the number of illegal residents as a residual for each locality. Using available socio-economic indicators, the micro-level results will then be extrapolated to other areas of Prague. Similarly, a Swiss study uses multiple regression analysis to extrapolate results from a Delphi-survey in six cantons to the whole country using their implied correlation with various indicators such as share of work force in agriculture, share of registered foreigners and income level (GFS, 2005). Methods and data sources used in europe Of the various approaches and methods for estimating stocks of irregular migrants, some have found broad application in European countries, while others have not. In this section, a brief evaluation of methods and data sources which have recently been applied in the European context is provided. The analysis is based on the CLANDESTINO dataset and classication scheme (cf. Vogel and Kovacheva, 2010 in this volume).20 Applying this classication, our subsequent discussion is limited to high quality and medium quality estimates in the CLANDESTINO country tables. Table 2 arranges these estimates by the methods employed to produce them. We can immediately see that researchers were able to exploit a variety of data sources to achieve either high or medium quality estimates. Both case-specic enforcement les and special survey data were the basis for high quality estimates and a number of other data sources (administrative data, census results, crime statistics, registers and expert surveys) were used in medium quality estimations. Regarding methods, it is striking that only two methods (the centre sampling technique and the repeated capture method) have resulted in estimates deemed to be of high quality satisfying usual academic standards. Apart from these, there are a number of estimates produced by an array of methods that are considered medium quality. The great majority of estimates reviewed by the CLANDESTINO project, however, are not included in Table 2, as they have been classied as low quality. There are several reasons for this lack of solid estimates:

2011 The Author. International Migration 2011 IOM

TABLE 2

METHODS AND ESTIMATION TECHNIQUES USED TO PRODUCE MEDIUM- OR HIGH-QUALITY STOCK ESTIMATES OF IRREGULAR FOREIGN RESIDENTS OR IRREGULAR FOREIGN WORKERS IN EUROPE: COUNTRIES AND YEARS OF ESTIMATES CLANDESTINO classication of the quality of the estimates High quality Medium quality CZ (2000): IFR; G (2006): IFR-working max1; G (2006): IFW-companies max; A (2009): IFR; G (2009): IFR NL (2001): IFR

Method

Estimation technique

Multiplier methods NL (2000): IFR; NL (2001 2002 2003): IFR; NL (2005): IFR I (2005 2006 2007): IFR 15+

Simple Multiplier

Capture-recapture Repeated capture

Methods of self-identication Survey methods

Evidence based on regularisation data Direct survey methods

I (2001): IFR 15+ min.2 GR (2004a): IFR SP (2007): IFR min NL (2004): IFW; NL (2006): IFW UK (2001): IFR SP (2001 to 2008): IFR max; GR (2004b): IFR 18+; GR (2007a): IFR min; GR (2007b): IFR-children; Italy (2000 to 2005): IFW fte3

Estimating stocks of irregular immigrants

2011 The Author. International Migration 2011 IOM

Residual methods

Employer surveys using randomized response Differences census results legal immigration data Simple comparison of various registers

69

70

TABLE 2 (CONTINUED) CLANDESTINO classication of the quality of the estimates High quality Medium quality A (2000): IFW, A (2003): IFW; G (2001) IFW fte4 Jandl G (2003): IFR-Munich5; G (2007): IFR-in sex work max.6

Method

Estimation technique

Indirect inferences

Subjective Estimations

Using information on correlated phenomena as basis of calculation Expert surveys

2011 The Author. International Migration 2011 IOM

Source: Own compilation on the basis of the CLANDESTINO classication of estimates. 1 Max. = maximum estimate of irregular foreign residents working in companies. 2 Minimum estimate on illegally resident persons over 15 years. 3 Estimation of irregular foreign workers in full-time equivalents (fte). 4 Estimation of irregular foreign workers in full-time equivalents (fte) in seasonal agriculture only. 5 Estimate based on expert and migrant interviews for the city of Munich only. 6 Maximum estimate of irregular foreign residents in sex work.

Estimating stocks of irregular immigrants

71

First, not all of the methods included in Table 1 are fully appropriate for the production of good estimates. Subjective estimation methods, for example, are not (directly) based on hard data and have consequently never been classied as high quality estimates. Second, for most countries there is simply a lack of the required statistical data sources to put available methods to work. For example, residual methods could be applied only in Spain and in the UK, because there are data sources that are thought to cover illegal residents reasonably well. Finally, despite the high level of media attention devoted to any new gures and estimates on irregular migration, there is apparently less interest in distinguishing reasonable estimates from purely speculative ones. This lack of interest in a solid information base for policy making is exemplied by the lack of funding to implement high-quality research designs for the production of better estimates.

CONCLUSIONS
This paper has provided an overview of theoretically available methods and techniques for the estimation of irregular migration phenomena and their practical application in Europe. In conclusion, several methodological and practical considerations shall be presented. Far from being inaccessible to serious study, there are in fact a considerable number of methods and estimation techniques available for the quantication of irregular migration phenomena. Most of these have originally been developed for the study of other hidden populations but are also applicable to irregular migration research. Thus, a lot can and should be learned from methods and approaches applied in related social science elds (such as studies on homelessness, drug abuse or criminal deviance), From the many methods presented in this paper, not all are suitable in all circumstances and migration contexts. Some methods presuppose the existence of certain data sets (such as census results or certain registers), while others require the production of primary data through surveys. Data availability thus limits the applicability of certain methods to certain administrative contexts or time periods. In addition to valid and relevant data, the construction of any solid estimate requires the mobilization of substantial resources and expertise. This requires not only the technical prociency for statistical analysis, but also a substantive appreciation of irregular migration processes and the data sources used. As such contextual knowledge is often hard to obtain, quantitative research on irregular migration should ideally be accompanied by qualitative research into the subject. Even if suitable data are available and an appropriate methodological approach is followed, the inherent uncertainties of the subject matter together with the inevitable practical data shortcomings imply that no one estimate, based on a single methodological approach, will be perfect. Therefore, the use of multiple, complementary methodical approaches is recommendable. As the external validation of the accuracy of estimates is generally not possible,

2011 The Author. International Migration 2011 IOM

72

Jandl

a triangulation of several independent estimates can help to improve condence in the outcome of specic estimates. Finally, while the present article has systematized a multitude of available methods, the methodological toolkit for the quantication of irregular migration is far from exhausted. New data sources and methods can and should be developed to further our understanding of irregular migration phenomena. Only a decade ago, Delaunay and Tapinos (1998: 72) in their comprehensive review of applied estimation methods in Europe had identied practically nothing that is well-founded. As this paper demonstrates, since then a number of new studies, methods and estimation techniques could be added to our methodical toolbox. Technological and administrative innovations (e.g., biometric visa or new databases) will give rise to new data sources and new research avenues. In an uncertain and highly dynamic eld such as this one, research should also remain open to progressive methodological innovation.

NOTES
1. The current overview is necessarily limited in both length and depth. For a comprehensive overview of methods for estimating stocks and ows of irregular migrants, see Jandl (2008). 2. The terms illegal residents or illegal entrants should not be mistaken as in indication of criminalizing a group of persons. In fact, legal regulations towards the right to entry and the right to residence vary widely, making the breach of such regulations in most states administrative infractions rather than criminal offences. At the same time, a precise use of the terms should denote whether the entry into or the residence in the country are considered to be outside existing laws and regulations. For the same reason, the term illegal migrant should not be used as it does not indicate which activity is considered to be outside the law. 3. It should be noted that the classication proposed here has some similarities but also signicant differences to the typology proposed by Delaunay and Tapinos (1998: 36ff), which is also taken up in the excellent overview provided by Pinkterton et al. (2004: 33ff). Differences are due to different criteria used for classication. 4. It should be mentioned that the link between methods and data sources as presented in Table 1 is not exclusive but illustrative of their main applications. In addition, many methods could be applied also to other data sources than those directly mentioned. 5. For an illustrative application of the multiplier principle on problem drug users, cf. UNODC 2007, p. 266. 6. The same multiplier method is applied both to crime data and health records on the Belgium case (van Meeteren et. al., 2007). Similarly, Cyrus (2008: 41f) puts forward maximum estimates of irregular foreign workers and illegal residents in Germany on the basis of derived multipliers applied to labour inspection data. 7. For the application of this minimum maximum logic to the city of Hamburg, combined with qualitative validation efforts, see Vogel and Aner (2009). 8. The study was part of a wider study known as the Unknown City Project. The estimation result is surprisingly close to the estimate provided by Burgers, 1996. 9. Depending on the terms of the regularizations on offer, it is likely that the extent of this experience varies from being only a marginal phenomenon to a situation where it could comprise the majority of beneciaries of recent regularization programmes.

2011 The Author. International Migration 2011 IOM

Estimating stocks of irregular immigrants

73

10. From 1999 to 2001, an average of 11% of all third-country-nationals admitted for residence in France had previously been in an illegal situation (Thierry and Rogers, 2004, p.654ff). 11. For more information, see: http://www.ismu.org/ORIM/. 12. The number of legal immigrants granted settlement in the UK before 1970 was estimated with an alternative method. 13. Since 2003, third country citizens without permanent residence permits must renew their registration in the Padron every two years or are automatically deregistered lez Enriquez 2008:26). (Gonza 14. The stock of illegal foreign residents in the 10 US states was estimated at between 570,000 and 4.7 million persons. 15. The clandestine population from Morocco in Europe was estimated by the expected population method to be around 370,000 and by the sex-ratio method to be between 600,000 and 700,000 persons. 16. The shadow economy is dened as ,,all unregistered economic activities that contribute to the ofcially calculated GDP, see Schneider and Klinglmaier, 2004: 4 17. The three eligibility criteria are: being able to produce an identity card; having an income of less than 606,- per month; and showing proof of residence in France for at least three months. 18. The use of the term subjective estimations in this context should not imply that other estimations are necessarily objective. Rather, the connotation implies that a set of subjective opinions of experts, which are not further scrutinized and justied by the use of rationalized methods and data sources, are taken as the basis of a constructed estimate. 19. Applying the (adjusted) average of the survey results, the calculated number of illegally employed foreigners is 182,556. However, applying the median of the survey results, the calculated number is only 73,100. 20. The CLANDESTINO project covers 12 European countries, namely: Austria, the Czech Republic, France, Germany, Greece, Hungary, Italy, the Netherlands, Poland, Slovakia, Spain and the United Kingdom. A full database with quantitative estimates at the country-level, classied by their quality, is available at http://irregularmigration.hwwi.net (last accessed 22 July 2010). In addition, the database provides aggregate EU-level estimates for the years 2002, 2005 and 2008. Estimates of the number of illegal residents are provided in ranges and for various geographical areas and years: for the EU-15, for example, the estimates were 3.15.3 million in 2002, 2.24.8 million in 2005 and 1.83.3 million in 2008. The estimate for the EU-27 in 2008 was put only slightly higher at 1.93.8 million persons.

REFERENCES
Alt, J. 2003

Leben in der Schattenwelt. Problemkomplex illegale Migration, Karlsruhe, von Loeper. Baio, G., G.C. Blangiardo, and M. Blangiardo 2008 Center sampling technique in foreign migration surveys (forthcoming). Burgers, J. 1996 Natte vingers en vuile handen. Over het schatten van het aantal illegale vreemdelingen, Migrantenstudies, 12(1): 1426 and 3233.

2011 The Author. International Migration 2011 IOM

74

Jandl

Cangiano, A. 2008 Foreign migrants in Southern European countries: evaluation of recent data, in J. Raymer and F. Willekens (Eds), International Migration in Europe: Data, Models and Estimates, John Wiley & Sons, Ltd, New York: 89114. Courau, H. 2008 French Country Report for CLANDESTINO, November 2008, http://irregularmigration.hwwi.net/ (accessed on 9 December 2008). Cyrus, N. 2008 German Country Report for CLANDESTINO, revised version, August 2008, http://irregular-migration.hwwi.net/ (accessed on 7 October 008). Delaunay, D., and G. Tapinos 1998 La mesure de la migration clandestine en Europe, Volume 1: Rapport de synthese, Eurostat Working Papers, Series Population et conditions sociales, Eurostat, Luxemburg. Drbohlav, D., and L. Lachmanova 2008 Czech Country Report for CLANDESTINO, revised version, August 2008, http://irregular-migration.hwwi.net/ (accessed on16 January 2008). Engbersen, G., R. Staring, J. van der Leun, et al. 2002 Illegale vreemdelingen in Nederland. Omvang, overkomst, verblijf en uitzetting, Erasmus Universiteit, Rotterdam. Frank and Snijders 1994 Estimating the size of hidden populations using snowball sampling, Journal of Ofcial Statistics, 10: 5367. Futo, P. 2008 Hungarian Country Report for CLANDESTINO, revised version, September 2008, http://irregular-migration.hwwi.net/ (accessed on 10 October 2008). Forschung fu r Politik, Kommunikation und Gesellschaft (Gfs.bern) 2005 Sans Papier in der Schweiz: Arbeitsmarkt, nicht Asylpolitik ist entscheidend, Schlussbericht im Auftrag des Bundesamtes fu r Migration, Bern. lez Enriquez, C. Gonza 2008 Spanish Country Report for CLANDESTINO, revised version, September 2008, http://irregular-migration.hwwi.net/ (accessed on 6 October 2008). Heckathorn, D. 2002 Respondent-driven sampling II: deriving valid population estimates from chain-referral samples of hidden populations, Social Problems, 49(1): 1134. Hoefer, M., N. Rytina, and B. Baker 2008 Estimates of the Unauthorized Immigrant Population Residing in the United Sates: January 2007, Ofce of Immigration Statistics, Policy Directorate, US Department of Homeland Security, http://www.dhs.gov/xlibrary/assets/statistics/publications/ois_ill_pe_2007.pdf. Home Ofce 2005 Accession Monitoring Report May 2004 June 2005, Home Ofce, London, August. Jandl, M. 2004 The estimation of illegal migration in Europe, Studi Emigrazione Migration Studies, XLI, (153): 141155. 2008 Methods for estimating stocks and ows of irregular migrants, in M. Jandl, D. Vogel and K. Iglicka (Eds), CLANDESTINO Report on Methodological Issues: 1957, http://irregular-migration.hwwi.net/ (accessed on 2 April 2009).

2011 The Author. International Migration 2011 IOM

Estimating stocks of irregular immigrants

75

A multiplier estimate of the illegally resident TCN population in Austria based on crime suspect data, http://irregular-migration.hwwi.net/ (accessed on 2 July 2009). Jandl, M., and A. Kraler 2006 Statistics on refusals, apprehensions and removals: an analysis of the CIREFI data, in M. Poulain, N. Perrin and A. Singleton (Eds), THESIM Towards Harmonized European Statistics on International Migration, UCL Presses, Louvain-La-Neuve: 271285. Jandl, M., C. Hollomey, and A. Stepien 2007 Migration and irregular work in Austria: results of a delphi-study, International Migration Papers Nr. 90, International Labour Ofce, Geneva, http:// www.ilo.org. Kraler, A., D. Reichel, and C. Hollomey 2008 Austrian Country Report for CLANDESTINO, revised version, August 2008, http://irregular-migration.hwwi.net/ (accessed on 19 December 2008). Larson, E.M., and J.A. Droitcour 2003 Estimating the Illegal Alien Population in the United States: Some Methodological Considerations, United States General Accounting Ofce, Washington, DC, October. Lederer, H.W. 2004 Indikatoren der Migration. Zur Messung des Umfangs und der Arten von Migration in Deutschland unter besonderer Berucksichtigung des Ehegatten- und Familiennachzugs sowie der illegalen Migration, efms, Bamberg. Natale, M. 1990 Limmigrazione straniera in Italia: consistenza, caratteristiche, prospettive, Polis, IV (1): 540. Passel, J. 2006 The Size and Characteristics of the Unauthorized Migrant Population in the U.S.: Estimates based on the March 2005 Current Population Survey, Pew Hispanic Center, Washington, DC, http://pewhispanic.org/les/reports/61.pdf. Piguet, E., and S. Losa 2001 Travailleurs de lombre ? Demande de main-duvre du domaine de lasile et trangers non de clare s en Suisse, Rapport de recherche , ampleur de lemploi de Swiss Forum for Migration Studies, 19 2001 Pinkerton, C.G., McLaughlan, and J. Salt 2004 Sizing the Illegally Resident Population in the UK, Home Ofce Online Report 58 04, http://www.homeofce.gov.uk/rds/pdfs04/rdsolr5804.pdf. Robinson, J.G. 1980 Estimating the approximate size of the illegal alien population in the United States by the comparative trend analysis of age-specic death rates, Demography, 17(2): March. Schneider, F. 2006 Nur noch leicht sinkende Schattenwirtschaft in Deutschland im Gegensatz zu anderen OECD-Landern im Jahr 2006: Fluch oder Segen?, Universita t Linz, Linz. Schneider, F., and D. Enste 1999 Shadow economies around the world: size, causes, and consequences, IMF Working Paper WP 00 26,http://www.imf.org/external/pubs/ft/wp/2000/ wp0026.pdf.

2009

2011 The Author. International Migration 2011 IOM

76

Jandl

Sheldon, G. 2002 Forschungsdesign zur Bestimmung des Ausmasses und der Struktur der Schwarzarbeit in der Schweiz, Forschungstelle fur Arbeitsmarkt- und Industrieokonomik, Universita t Basel, Basel. Sikkel, D., P. van der Heijden, and G. van Gils 2006 Summary, Methoden voor omvangschattingen van verborgen populaties, met name illegalen, WODC rapport 243, Boom Juridische Uitgevers, Meppel. Tapinos, G. 1999 Clandestine immigration: economic and political issues, in OECD Sopemi, Trends in International Migration. OECD, Paris: 229251. Thierry, X., and G.I. Rogers 2004 Recent immigration trends in France and elements for a comparison with the United Kingdom, Population, 59(5): 635672. United Nations Ofce on Drugs and Crime 2007 Methodology, World Drug Report 2007, United Nations Ofce on Drugs and Crime, Vienna: 257274. Van der Heijden, P. 2006 Estimating the size of the illegal population in the Netherlands, Graduate School Lecture, University of Utrecht, Utrecht, 9 October. Van der Heijden, P., R. Bustami, Cruijff, et al. 2003 Point and interval estimation of the truncated Poisson regression model, Statistical Modelling, 3: 305322. Van der Heijden, P., M. Cruijff, and H. van Houwelingen 2003 Estimating the size of a criminal population from police registrations using the truncated Poisson regression model, Statistica Neerlandica, 57: 289304. Van der Heijden, P., G. van Gils, M. Cruijff, et al. 2006 Summary, An Estimate of the Number of Illegal Aliens Residing in the Netherlands in 2005, University of Utrecht, Utrecht. Van der Leun, J., and M. Ilies 2008 Dutch Country Report for CLANDESTINO, revised version, August, http:// irregular-migration.hwwi.net/ (accessed on 15 October 2008). Van der Leun, J., G. Engbersen, and P. van der Heijden 1998 Illegaliteit en criminaliteit: schattingen, aanhoudingen en uitzettingen. Onderzoeksrapport, Faculteit der Sociale Wetenschappen Sociologie, Erasmus Universiteit, Rotterdam Van Meeteren, M., M. van San, and G. Engbersen 2007 Irreguliere immigranten in Belgie. Inbedding, uitsluiting en criminaliteit, Erasmus Universiteit, Rotterdam. Vogel, D. berlegungen zur Date1999 Illegaler Aufenthalt in Deutschland methodische U nnutzung und Datenerhebung. Zeitschrift fu lkerungswissenschaft, Jg, r Bevo 24: 165185. 2002 Ausla nder ohne Aufenthaltsstatus in Deutschland. Methoden zur Scha tzung ihrer Zahl in E. Minthe (Ed.), Illegale Migration und Schleusungskriminalitat. Eigenverlag der Kriminologischen Zentralstelle, Wiesbaden: 6578. Vogel, D., and Kovacheva 2010 in this volume Wanner, P. thodes destimation de la population sans autori2002 Compter les clandestins. Me ` partir des donne es sur la population, SFM Discussion jour a sation de se Paper 13 2002, Swiss Forum for Migration Studies, Neuchatel.

2011 The Author. International Migration 2011 IOM

Estimating stocks of irregular immigrants

77

Woodbridge, J. 2005 Sizing the Unauthorised (Illegal) Migrant Population in the United Kingdom in 2001, Home Ofce Online Report 29 05, Home Ofce, London. Zhang, L. 2008 Report on Developing Methods for Determining the Number of Unauthorized Foreigners in Norway, Statistics Norway, Document 2008 11, Oslo, June, http://www.ssb.no/emner/00/90/doc_200811/doc_200811.pdf.

2011 The Author. International Migration 2011 IOM